«

2002
Frascati Manual
PROPOSED STANDARD PRACTICE FOR SURVEYS
ON RESEARCH AND EXPERIMENTAL DEVELOPMENT Frascati Manual
The Measurement of Scientific and Technological Activities

In an era of rapid technological change, research and development (R&D) is an important PROPOSED STANDARD
element of economic growth. Monitoring the R&D efforts of industry, governments and
PRACTICE FOR SURVEYS ON
universities is key to successful policy making and analysis.
RESEARCH AND EXPERIMENTAL
The Frascati Manual has become the internationally recognised methodology for collecting DEVELOPMENT
and using R&D statistics, and is an indispensable tool for statistical offices around the
world. It includes definitions of basic concepts, guidelines for collecting data and the
classifications to be used in compiling statistics.

This updated sixth edition includes improved guidelines that address changes in OECD
economies, including the measurement of R&D in the service sector, globalisation of R&D
and human resources in R&D.

The Manual will also be published as a Web book to enable updates to be made as they
become available.

Frascati Manual
OECD's books, periodicals and statistical databases are now available via www.SourceOECD.org,
our online library.
This book is available to subscribers to the following SourceOECD themes:
Science and Information Technology
Statistics Sources and Methods
Ask your librarian for more details of how to access OECD books online, or write to us at
[email protected]

www.oecd.org

ISBN 92-64-19903-9
92 2002 08 1 P

2002 -:HSTCQE=V^^UXX: 2002

© OECD, 2002.
© Software: 1987-1996, Acrobat is a trademark of ADOBE.
All rights reserved. OECD grants you the right to use one copy of this Program for your personal use only.
Unauthorised reproduction, lending, hiring, transmission or distribution of any data or software is
prohibited. You must treat the Program and associated materials and any elements thereof like any other
copyrighted material.
All requests should be made to:
Head of Publications Service,
OECD Publications Service,
2, rue André-Pascal,
75775 Paris Cedex 16, France.
 THE MEASUREMENT OF SCIENTIFIC
AND TECHNOLOGICAL ACTIVITIES

Proposed Standard Practice for Surveys

on Research and Experimental Development

Frascati Manual

2002

ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT

FRONTMATTER1 Page 2 Friday, December 6, 2002 10:12 AM

ORGANISATION FOR ECONOMIC CO-OPERATION

AND DEVELOPMENT

Pursuant to Article 1 of the Convention signed in Paris on 14th December 1960,

and which came into force on 30th September 1961, the Organisation for Economic
Co-operation and Development (OECD) shall promote policies designed:
– to achieve the highest sustainable economic growth and employment and a
rising standard of living in Member countries, while maintaining financial
stability, and thus to contribute to the development of the world economy;
– to contribute to sound economic expansion in Member as well as non-
member countries in the process of economic development; and
– to contribute to the expansion of world trade on a multilateral, non-
discriminatory basis in accordance with international obligations.
The original Member countries of the OECD are Austria, Belgium, Canada,
Denmark, France, Germany, Greece, Iceland, Ireland, Italy, Luxembourg, the
Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, the United
Kingdom and the United States. The following countries became Members
subsequently through accession at the dates indicated hereafter: Japan
(28th April 1964), Finland (28th January 1969), Australia (7th June 1971),
New Zealand (29th May 1973), Mexico (18th May 1994), the Czech Republic
(21st December 1995), Hungary (7th May 1996), Poland (22nd November 1996),
Korea (12th December 1996) and the Slovak Republic (14th December 2000). The
Commission of the European Communities takes part in the work of the OECD
(Article 13 of the OECD Convention).

Publié en français sous le titre :

Manuel de Frascati 2002
Méthode type proposée pour les enquêtes sur la recherche et le développement expérimental

© OECD 2002

Permission to reproduce a portion of this work for non-commercial purposes or classroom use should be obtained through the
Centre français d’exploitation du droit de copie (CFC), 20, rue des Grands-Augustins, 75006 Paris, France, tel. (33-1) 44 07 47 70,
fax (33-1) 46 34 67 19, for every country except the United States. In the United States permission should be obtained through
the Copyright Clearance Center, Customer Service, (508)750-8400, 222 Rosewood Drive, Danvers, MA 01923 USA, or
CCC Online: www.copyright.com. All other applications for permission to reproduce or translate all or part of this book should
be made to OECD Publications, 2, rue André-Pascal, 75775 Paris Cedex 16, France.
 FOREWORD

Foreword
I n June 1963, the OECD met with national experts on research and development
(R&D) statistics at the Villa Falcioneri in Frascati, Italy. The result was the first official
version of the Proposed Standard Practice for Surveys of Research and
Development, better known as the Frascati Manual. This publication is the sixth
edition.
Since the fifth edition was issued in 1994, attention has increasingly been paid to
R&D and innovation as key elements in the knowledge-based economy. Reliable and
comparable statistics and indicators to monitor this area are of crucial importance.
This edition therefore makes an effort to strengthen various methodological
recommendations and guidelines, in particular for improving R&D statistics in the
services sector and collecting more detailed data on human resources for R&D. Because
globalisation is a challenge for R&D surveys, the Manual recommends some changes
in classifications in an attempt to address this issue.
Today’s R&D statistics are the result of the systematic development of surveys
based on the Frascati Manual and are now part of the statistical system of the OECD
member countries. Although the Manual is basically a technical document, it is a
cornerstone of OECD efforts to increase the understanding of the role played by science
and technology by analysing national systems of innovation. Furthermore, by
providing internationally accepted definitions of R&D and classifications of its
component activities, the Manual contributes to intergovernmental discussions on
“best practices” for science and technology policies.
The Frascati Manual is not only a standard for R&D surveys in OECD member
countries. As a result of initiatives by the OECD, UNESCO, the European Union and
various regional organisations, it has become a standard for R&D surveys worldwide.
The Frascati Manual is based on experience gained from collecting R&D
statistics in OECD member countries. It is a result of the collective work of national
experts in NESTI (the Working Party of National Experts on Science and Technology
Indicators). The Group, supported by an effective Secretariat, first led by the late Yvan
Fabian and subsequently by Alison Young, John Dryden, Daniel Malkin and Andrew
Wyckoff, has elaborated over the last 40 years on the concept of science and technology
indicators and developed a series of methodological manuals known as the “Frascati
Family”, which includes manuals on: R&D (Frascati Manual), innovation (Oslo

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 3

FOREWORD

Manual), human resources (Canberra Manual), technological balance of payments

and patents as science and technology indicators.
The Frascati Manual is also published in electronic format on the OECD Web
site. The idea is to update the electronic version more frequently, as newer material
becomes available. The electronic version is complemented by further material related
to R&D surveys.
The sixth edition of the Manual was prepared by teams of experts drawn from the
NESTI Group. The OECD Secretariat (especially Dominique Guellec, Laudeline Auriol,
Mosahid Khan, Geneviève Muzart and Sharon Standish) played an active role in
co-ordinating the process and drafting certain sections. Bill Pattinson (a former
Australian NESTI delegate) was responsible for preliminary revisions while working in
the OECD. Mikael Åkerblom (Statistics Finland and a Finnish NESTI delegate) worked
in the final stage for one year at the OECD to draft the Manual, incorporating various
comments and suggestions from NESTI members.
Thanks to a generous voluntary contribution to the OECD from the Japanese
government, this revision benefited from substantive contributions by experts and
proceeded in a timely fashion. Japan’s contribution is gratefully acknowledged. The
Manual is published on the responsibility of the Secretary-General of the OECD.

Takayuki Matsuo Giorgio Sirilli Fred Gault

Director for Science, Technology Chair of NESTI Current Chair of NESTI
and Industry, OECD initiating the fifth revision

4 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 TABLE OF CONTENTS

Table of Contents
Chapter 1. Aim and Scope of the Manual . . . . . . . . . . . . . . . . . . . . . . . 13
1.1. A preliminary word to the user of R&D data . . . . . . . . . . . . . 14
1.2. Coverage of the Manual and the uses of R&D statistics . . . . 14
1.3. The relationship between the Frascati Manual
and other international standards . . . . . . . . . . . . . . . . . . . . . 15
1.4. R&D input and output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.5. R&D and related activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.5.1. Research and experimental development (R&D) . . . . 17
1.5.2. Scientific and technological activities (STA) . . . . . . . . 18
1.5.3. R&D and technological innovation. . . . . . . . . . . . . . . . 18
1.5.4. The identification of R&D in software, social
sciences and service activities . . . . . . . . . . . . . . . . . . . 19
1.5.5. R&D administration and other supporting
activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.6. R&D in all fields of science and technology is covered. . . . . 20
1.7. Measures of R&D inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.7.1. R&D personnel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.7.2. R&D expenditures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.7.3. R&D facilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.7.4. National R&D efforts . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.8. Globalisation of R&D and R&D co-operation . . . . . . . . . . . . . 22
1.9. Classification systems for R&D . . . . . . . . . . . . . . . . . . . . . . . . 23
1.9.1. Institutional classifications . . . . . . . . . . . . . . . . . . . . . . 23
1.9.2. Functional distribution . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.10. R&D surveys, reliability of data and international
comparability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.11. Government budget appropriations or outlays for R&D
(GBAORD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.12. Topics of special interest. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.13. A final word to the user of R&D data. . . . . . . . . . . . . . . . . . . . 27

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 5

TABLE OF CONTENTS

Chapter 2. Basic Definitions and Conventions . . . . . . . . . . . . . . . . . . . 29

2.1. Research and experimental development (R&D). . . . . . . . . . 30
2.2. Activities to be excluded from R&D. . . . . . . . . . . . . . . . . . . . . 30
2.2.1. Education and training. . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2.2. Other related scientific and technological
activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2.3. Other industrial activities . . . . . . . . . . . . . . . . . . . . . . . 33
2.2.4. Administration and other supporting activities. . . . . 33
2.3. The boundaries of R&D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.3.1. Criteria for distinguishing R&D from related
activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.3.2. Problems at the borderline between R&D
and education and training . . . . . . . . . . . . . . . . . . . . . . 35
2.3.3. Problems at the borderline between R&D
and related scientific and technological activities . . 38
2.3.4. Problems at the borderline between R&D and other
industrial activities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.3.5. Problems at the borderline between R&D
administration and indirect supporting activities . . . 45
2.4. Identifying R&D in software development, in the social
sciences and humanities and in service activities
and industries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.4.1. Identifying R&D in software development . . . . . . . . . 46
2.4.2. Identifying R&D in the social sciences
and humanities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.4.3. Special problems for identifying R&D in service
activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

Chapter 3. Institutional Classification . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3.1. The approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.2. The reporting unit and the statistical unit . . . . . . . . . . . . . . . 52
3.2.1. The reporting unit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.2.2. The statistical unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3. Sectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3.1. Reasons for sectoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3.2. Choice of sectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3.3. Problems of sectoring . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.4. Business enterprise sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.4.1. Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

6 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 TABLE OF CONTENTS

3.4.2. The principal sector sub-classification . . . . . . . . . . . . 56

3.4.3. Other institutional sub-classifications . . . . . . . . . . . . 61
3.5. Government sector. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.5.1. Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.5.2. The principal sector sub-classification . . . . . . . . . . . . 63
3.5.3. Other institutional sub-classifications . . . . . . . . . . . . 64
3.6. Private non-profit sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.6.1. Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.6.2. The principal sector sub-classification . . . . . . . . . . . . 66
3.6.3. Other institutional sub-classifications . . . . . . . . . . . . 66
3.7. Higher education sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.7.1. Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.7.2. The principal sector sub-classification . . . . . . . . . . . . 71
3.7.3. Other institutional sub-classifications . . . . . . . . . . . . 72
3.8. Abroad . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.8.1. Coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.8.2. The principal sector sub-classification . . . . . . . . . . . . 72
3.8.3. Other institutional sub-classifications . . . . . . . . . . . . 73
3.8.4. Geographic area of origin or destination of funds . . . 73

Chapter 4. Functional Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

4.1. The approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.2. Type of R&D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.2.1. Use of distribution by type of R&D . . . . . . . . . . . . . . . . 77
4.2.2. The distribution list . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.2.3. Criteria for distinguishing between types of R&D . . . 79
4.3. Product fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.3.1. Use of distribution by product fields . . . . . . . . . . . . . . 82
4.3.2. The distribution list . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.3.3. Criteria for distribution . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.4. Fields of science and technology . . . . . . . . . . . . . . . . . . . . . . . 85
4.4.1. Use of distribution by field of science and technology 85
4.4.2. The distribution list . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.4.3. The criteria for distribution. . . . . . . . . . . . . . . . . . . . . . 86
4.5. Socio-economic objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.5.1. Use of distribution by socio-economic objectives . . . 86
4.5.2. Minimum recommended breakdown . . . . . . . . . . . . . 86
4.5.3. The distribution list . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.5.4. The criteria for distribution. . . . . . . . . . . . . . . . . . . . . . 88

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 7

TABLE OF CONTENTS

Chapter 5. Measurement of R&D Personnel. . . . . . . . . . . . . . . . . . . . . . 89

5.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.2. Coverage and definition of R&D personnel . . . . . . . . . . . . . . 92
5.2.1. Initial coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.2.2. Categories of R&D personnel . . . . . . . . . . . . . . . . . . . . 92
5.2.3. Classification by occupation . . . . . . . . . . . . . . . . . . . . . 93
5.2.4. Classification by level of formal qualification . . . . . . 95
5.2.5. Treatment of postgraduate students . . . . . . . . . . . . . . 96
5.3. Measurement and data collection . . . . . . . . . . . . . . . . . . . . . . 98
5.3.1. Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.3.2. Headcount data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.3.3. Full-time equivalence (FTE) data . . . . . . . . . . . . . . . . . 99
5.3.4. Recommended national aggregates and variables . . 102
5.3.5. Cross-classified data by occupation
and qualification. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.3.6. Regional data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

Chapter 6. Measurement of Expenditures Devoted to R&D . . . . . . . . 107

6.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.2. Intramural expenditures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.2.1. Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.2.2. Current costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.2.3. Capital expenditures . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.3. Sources of funds. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.3.1. Methods of measurement . . . . . . . . . . . . . . . . . . . . . . . 114
6.3.2. Criteria for identifying flows of R&D funds. . . . . . . . . 114
6.3.3. Identifying the sources of flows of R&D funds . . . . . . 116
6.4. Extramural expenditures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
6.5. Reconciling differences in performer-based
and source-based reporting . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.6. Regional distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
6.7. National totals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.7.1. Gross domestic expenditure on R&D (GERD) . . . . . . . 121
6.7.2. Gross national expenditure on R&D (GNERD). . . . . . . 121

Chapter 7. Survey Methodology and Procedures . . . . . . . . . . . . . . . . . 125

7.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
7.2. Scope of R&D surveys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
7.3. Identifying target population and survey respondents . . . . 127
7.3.1. Business enterprise sector. . . . . . . . . . . . . . . . . . . . . . . 127

8 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 TABLE OF CONTENTS

7.3.2. Government sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129

7.3.3. Private non-profit sector . . . . . . . . . . . . . . . . . . . . . . . . 130
7.3.4. Higher education sector. . . . . . . . . . . . . . . . . . . . . . . . . 130
7.3.5. Hospitals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.4. Working with respondents . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
7.4.1. Encouraging co-operation . . . . . . . . . . . . . . . . . . . . . . . 131
7.4.2. Operational criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
7.5. Estimation procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.5.1. Unit and item non-response . . . . . . . . . . . . . . . . . . . . . 134
7.5.2. Estimation procedures in the higher education
sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
7.6. Reporting to the OECD or to other international
organisations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

Chapter 8. Government Budget Appropriations or Outlays for R&D

by Socio-economic Objectives (GBAORD) . . . . . . . . . . . . . . 137
8.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
8.2. Relationship with other international standards . . . . . . . . . 138
8.3. Sources of budgetary data for GBAORD . . . . . . . . . . . . . . . . . 138
8.4. Coverage of R&D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
8.4.1. Basic definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
8.4.2. Fields of science and technology . . . . . . . . . . . . . . . . . 139
8.4.3. Identifying R&D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
8.5. Definition of government . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
8.6. Coverage of government budget appropriations
and outlays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
8.6.1. Intramural and extramural expenditures . . . . . . . . . . 140
8.6.2. Funding and performer-based reporting . . . . . . . . . . . 140
8.6.3. Budgetary funds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
8.6.4. Direct and indirect funding . . . . . . . . . . . . . . . . . . . . . . 141
8.6.5. Types of expenditure . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
8.6.6. GBAORD going to R&D abroad. . . . . . . . . . . . . . . . . . . . 142
8.7. Distribution by socio-economic objectives. . . . . . . . . . . . . . . 143
8.7.1. Criteria for distribution . . . . . . . . . . . . . . . . . . . . . . . . . 143
8.7.2. Distribution of budgetary items . . . . . . . . . . . . . . . . . . 144
8.7.3. The distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.7.4. Socio-economic objectives – SEO . . . . . . . . . . . . . . . . . 144
8.7.5. Principal areas of difficulty . . . . . . . . . . . . . . . . . . . . . . 148

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 9

TABLE OF CONTENTS

8.8. Main differences between GBAORD and GERD data . . . . . . . 149

8.8.1. General differences. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.8.2. GBAORD and government-financed GERD . . . . . . . . . 150
8.8.3. GBAORD and GERD by socio-economic objectives . . . 150

Annexes
1. Brief History and Origins of the Present Manual . . . . . . . . . . . . . . . 151
2. Obtaining Data on R&D in the Higher Education Sector. . . . . . . . . 158
3. The Treatment of R&D in the United Nations System
of National Accounts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
4. R&D Related to Health, Information and Communication
Technology (ICT) and Biotechnology . . . . . . . . . . . . . . . . . . . . . . . . . 181
5. Methods of Deriving Regional R&D Data . . . . . . . . . . . . . . . . . . . . . 193
6. Work on S&T Indicators in Other International Organisations . . . 195
7. Other Science and Technology Indicators . . . . . . . . . . . . . . . . . . . . 200
8. Practical Methods of Providing Up-to-date Estimates
and Projections of Resources Devoted to R&D . . . . . . . . . . . . . . . . . 212
9. R&D Deflators and Currency Converters. . . . . . . . . . . . . . . . . . . . . . 217
10. Supplementary Guidance on the Classification of Large R&D
Projects with Special Reference to the Defence and Aerospace
Industries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
11. Correspondence between the Categories of R&D Personnel
by Occupation in the Frascati Manual and ISCO–88 Classes . . . . . . 238

Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

Index by Paragraph Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246

List of Tables
1.1. OECD methodological manuals . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1. Supplementary criteria for separating R&D from related
scientific, technological and industrial activities . . . . . . . . . . . . 34
2.2. Borderline between R&D and education and training
at ISCED level 6. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3. Some cases at the borderline between R&D and other
industrial activities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.1. International Standard Industrial Classification arranged
for the purposes of R&D statistics. . . . . . . . . . . . . . . . . . . . . . . . . 57
3.2. Fields of science and technology . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.1. Utility of functional distributions . . . . . . . . . . . . . . . . . . . . . . . . . 76

10 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 TABLE OF CONTENTS

4.2. The three types of research in the social sciences

and humanities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.1. R&D and indirect support activities . . . . . . . . . . . . . . . . . . . . . . . 91
5.2. Standard key for ISCED levels and classes
of the Frascati Manual for R&D personnel by formal
qualifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.3a. Total national R&D personnel by sector and by occupation . . . 102
5.3b. Total national R&D personnel by sector and by level
of qualification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.4. R&D personnel classified by occupation and by formal
qualification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
6.1. Gross domestic expenditure on R&D (GERD) . . . . . . . . . . . . . . . . 122
6.2. Gross national expenditure on R&D (GNERD) . . . . . . . . . . . . . . . 123
8.1. Standard key between NABS 1992 and previous OECD
GBAORD objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.2. Standard key between NABS 1992 and Nordforsk GBAORD
objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

Annexes
1. Summary of sectors in the SNA and in the Frascati Manual. . . . 173
2. Sectors and producers in the SNA. . . . . . . . . . . . . . . . . . . . . . . . . 173
3. The SNA sectoring of units definitely and possibly included
in the Frascati Manual higher education sector . . . . . . . . . . . . . . 174
4. SNA classifications of government outlays and final
consumption expenditure of NPI serving households . . . . . . . . 175
5. Gross output and total intramural R&D . . . . . . . . . . . . . . . . . . . . 177
1. Identifying health-related R&D in GBAORD . . . . . . . . . . . . . . . . . 183
2. Health-related R&D from performer-reported data:
business enterprise sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
3. Identifying health-related R&D by field of science
and socio-economic objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
1. Terminology in common use in the defence and aerospace
industries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
2. Current classification of French, UK and US terminology
in the Frascati Manual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
3. Development of an armoured tank . . . . . . . . . . . . . . . . . . . . . . . 235
1. Correspondence between Frascati Manual categories of R&D
personnel by occupation and ISCO–88 classes . . . . . . . . . . . . . . 239

Figure
3.1. Decision tree for sectoring R&D units . . . . . . . . . . . . . . . . . . . . . . . . 55

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 11

 ISBN 92-64-19903-9
Frascati Manual 2002
Proposed Standard Practice for Surveys on Research
and Experimental Development
© OECD 2002

Chapter 1

Aim and Scope of the Manual

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 13

1 AIM AND SCOPE OF THE MANUAL

1.1. A preliminary word to the user of R&D data

1. This Manual was written by and for the national experts in member
countries who collect and issue national R&D data and submit responses to
OECD R&D surveys. Although many examples are given, the Manual remains
a technical document which is mainly intended as a reference work.
2. Chapter 1 is addressed principally to users of R&D data. It provides a
summary of the coverage and contents of the Manual in order to help them to
use it. It also indicates why certain types of data are, or are not, collected, the
problems of comparability they pose and what can be said about their
interpretation.

1.2. Coverage of the Manual and the uses of R&D statistics

3. The Manual was first issued nearly 40 years ago and deals exclusively
with the measurement of human and financial resources devoted to research
and experimental development (R&D), often referred to as R&D “input” data.
4. Over the years, input statistics have proved to be valuable indicators
and have been used in various national and international reports. The OECD
reports on science and technology indicators (OECD, 1984; OECD, 1986; OECD,
1989a); the Science and Technology Policy Review and Outlook series and the
Science, Technology and Industry Scoreboard (OECD, every second year) all provide
useful measures of the scale and direction of R&D in various countries,
sectors, industries, scientific fields and other categories of classification.
Administrations concerned with economic growth and productivity rely on
R&D statistics as one type of indicator of technological change. Advisors
concerned with science policy, but also with industrial policy and even general
economic and social policies, use them extensively. R&D statistics are now an
essential background element for many government programmes and provide
an important tool for evaluating them. In many countries, R&D statistics are
regarded as a part of general economic statistics.
5. However, R&D statistics are not enough. In the context of the
knowledge-based economy, it has become increasingly clear that such data
need to be examined within a conceptual framework that relates them both to
other types of resources and to the desired outcomes of given R&D activities.
This link may be made, for example, via the innovation process (see
Section 1.5.3) or within the broader framework of “intangible investment”,

14 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 1 AIM AND SCOPE OF THE MANUAL

which covers not only R&D and related S&T activities but also expenditures on
software, training, organisation, etc. Similarly, R&D personnel data need to be
viewed as part of a model for the training and use of scientific and technical
personnel. It is also of interest to analyse R&D data in conjunction with other
economic variables, such as value added and investment data. The Manual is
not based on a single model of the S&T system; its aim is to make it possible
to produce statistics that can be used to calculate indicators for use in various
models.
6. The Manual has two parts. The first consists of seven chapters in
addition to this introductory chapter. They present recommendations and
guidelines on the collection and interpretation of established R&D data. While
all member countries may not be able to comply with the recommendations
as stated, there is consensus that these are the standard to which all should
aspire.
7. The second part consists of eleven annexes, which interpret and
expand upon the basic principles outlined in the preceding chapters in order
to provide additional guidelines for R&D surveys or deal with topics relevant to
R&D surveys. These annexes can be used for information purposes but are not
necessarily an up-to-date interpretation of the subject.
8. The Manual is published both as a paper version and an electronic
version available on the Internet. The electronic version will be more
frequently updated with new material.

1.3. The relationship between the Frascati Manual and other

international standards
9. R&D is carried out throughout the economy but has certain
characteristics that distinguish it from the larger family of scientific activities
and from the economic activities of which it is a part. From the outset, it was
intended that the OECD should establish a set of guidelines on the
measurement of scientific and technological activities. For many years the
Frascati Manual was the only such manual; more recently, four others have
been added. In addition, other OECD methodological frameworks are available
for science and technology and related activities, such as education (see
Table 1.1).
10. The OECD did not set out to establish international norms for S&T
where these already existed. Thus, the Manual is consistent with UNESCO
recommendations for all scientific and technological activities (UNESCO,
1978), but is specific to R&D and to the needs of OECD member countries,
whose rather similar economic and scientific systems distinguish them from
non-OECD countries.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 15

1 AIM AND SCOPE OF THE MANUAL

Table 1.1. OECD methodological manuals

Type of data Title

A. The “Frascati family” The Measurement of Scientific and Technological Activities Series
R&D Frascati Manual: Proposed Standard Practice for Surveys of Research
and Experimental Development
R&D Statistics and Output Measurement in the Higher Education Sector.
“Frascati Manual Supplement” (OECD, 1989b)
Technology balance of payments “Manual for the Measurement and Interpretation of Technology Balance
of Payments Data – TBP Manual” (OECD, 1990)1
Innovation OECD Proposed Guidelines for Collecting and Interpreting Technological
Innovation Data – Oslo Manual (OECD, 1997a)
Patents “Using Patent Data as Science and Technology Indicators – Patent
Manual 1994” (OECD, OCDE/GD(94)114, 1994b)1
S&T personnel “The Measurement of Human Resources Devoted to Science
and Technology – Canberra Manual” (OECD, 1995)

B. Other methodological frameworks for S&T

High-technology “Revision of High-technology Sector and Product Classification”
(OECD, STI Working Paper 1997/2)
Bibliometrics “Bibliometric Indicators and Analysis of Research Systems, Methods
and Examples”, by Yoshiko Okubo (OECD, STI Working Paper 1997/1)
Globalisation Manual of Economic Globalisation Indicators (provisional title,
forthcoming)

C. Other relevant OECD statistical frameworks

Education statistics OECD Manual for Comparative Education Statistics (forthcoming)
Education classification Classifying Educational Programmes, Manual for ISCED-97
Implementation in OECD countries (OECD, 1999)
Training statistics Manual for Better Training Statistics – Conceptual, Measurement
and Survey Issues (OECD, 1997b)

1. Deals mainly with problems of classifying and interpreting existing information.

Source: OECD.

11. Because of the need to place R&D in a wider context, both conceptually
and in terms of databases, United Nations (UN) classifications are used as far
as possible, e.g. System of National Accounts – SNA (UN, 1968); Commission of
the European Communities – CEC (CEC et al., 1994); International Standard
In dus t ria l Cla s si fica t ion – ISIC (UN, 1 9 9 0) ; Int e rn ation al Sta n dard
Classification of Occupations – ISCO (International Labour Organization, 1990);
and International Standard Classification of Education – ISCED (UNESCO,
1997). Furthermore, wherever possible, the Manual draws on the experience of
regional organisations within the OECD area, notably the European Union (EU)
and the Nordic Industrial Fund.
12. The references to R&D in such classifications are relatively recent and
are generally based on the Frascati Manual as the established international
statistical framework.

16 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 1 AIM AND SCOPE OF THE MANUAL

13. As in the previous editions of the Manual, an attempt has been made
to make R&D surveys consistent with the principles laid down in the System
of National Accounts (SNA). As far as possible, complementary data should be
collected that allow for bridging Frascati Manual-type data with SNA-type data.
For this reason, recommendations for the breakdown of sources of funds and
of extramural R&D expenditures are quite detailed and a recommendation has
been introduced to collect data on software investments related to R&D.
Annex 3 discusses the relation between R&D surveys and national accounts in
more detail.

1.4. R&D input and output

14. This Manual is devoted to measuring R&D inputs. R&D covers both
formal R&D in R&D units and informal or occasional R&D in other units.
However, interest in R&D depends more on the new knowledg e and
innovations and the economic and social effects that result than on the
activity itself. Unfortunately, while indicators of R&D output are clearly
needed to complement input statistics, they are far more difficult to define
and produce.
15. The output of R&D or science and technology (S&T) in general can be
measured in several ways. Innovation surveys are an attempt to measure
outputs and the effects of the innovation process in which R&D plays an
important role. A manual on innovation surveys has been issued and revised
once (OECD, 1997a).
16. Another option is to use existing data sources. A substantial amount of
methodological work was required before recommending international
standard practice for using existing sources to derive S&T indicators. Manuals
on the technology balance of payments and on the use of patents as S&T
indicators have been published (OECD 1990, 1994b). Guidelines are also
available on bibliometrics and on the analysis of trade data in terms of the
“technology intensity” of the products or industries concerned (see Table 1.1).
These manuals and guidelines differ from this Manual in that they focus more
on problems of interpretation; the data used are not collected for the purpose
of S&T analysis but are extracted from existing sources and rearranged for this
purpose (for further details, see Annex 7).

1.5. R&D and related activities

1.5.1. Research and experimental development (R&D)
17. This Manual deals only with the measurement of research and
experimental development (comprising basic research, applied research and
experimental development). A full definition can be found in Chapter 2.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 17

1 AIM AND SCOPE OF THE MANUAL

18. R&D is an activity related to a number of others with a scientific and

technological basis. Although these other activities are often very closely
linked to R&D, through flows of information and in terms of operations,
institutions and personnel, they must be excluded when measuring R&D. R&D
and these related activities may be considered under two headings: the family
of scientific and technological activities (STA) and the process of scientific and
technological innovation.

1.5.2. Scientific and technological activities (STA)

19. UNESCO developed the broader concept of STA and included it in its
“Recommendation concerning the International Standardisation of Statistics
on Science and Technology” (UNESCO, 1978). In addition to R&D, scientific and
technological activities comprise scientific and technical education and
training (STET) and scientific and technological services (STS). The latter
services include, for example, S&T activities of libraries and museums,
translation and editing of S&T literature, surveying and prospecting, data
collection on socio-economic phenomena, testing, standardisation and
quality control, client counselling and advisory services, patent and licensing
activities by public bodies.
20. R&D (defined similarly by UNESCO and the OECD) is thus to be
distinguished from both STET and STS.

1.5.3. R&D and technological innovation

21. Technological innovation activities are all of the scientific,
technological, organisational, financial and commercial steps, including
investments in new knowledge, which actually, or are intended to, lead to the
implementation of technologically new or improved products and processes.
R&D is only one of these activities and may be carried out at different phases
of the innovation process. It may act not only as the original source of
inventive ideas but also as a means of problem solving which can be called
upon at any point up to implementation.
22. Besides R&D, other forms of innovative activities may be distinguished
in the innovation process. According to the Oslo Manual (OECD, 1997a), these
are “acquisition of disembodied technology and know-how, acquisition of
embodied technology, tooling up and industrial engineering, industrial design
n.e.c., other capital acquisition, production start-up and marketing for new or
improved products”.
23. Furthermore, in the case of innovations based on government R&D
programmes, the process may include a significant demonstration stage. “A
demonstration is a project involving an innovation operated at or near full
scale in a realistic environment for the purpose of: i) formulating national

18 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 1 AIM AND SCOPE OF THE MANUAL

policy; or ii) promoting the use of innovation” (Glennan et al., 1978). It should
be noted that the data collected and published by the International Energy
Agency at the OECD cover research, development and demonstration
(“R, D and D”).
24. Possibly the greatest source of error in measuring R&D is the difficulty
of locating the cut-off point between experimental development and the
related activities required to realise an innovation. Errors in this respect are
particularly significant: while many innovations may require costly R&D, the
costs of preparing for production are often higher still. Sections 2.3.4 and 2.4.1
of Chapter 2 are devoted to guidelines and conventions for dealing with these
problems and give examples. They provide guidelines on this borderline as it
relates to the development of computer software and large-scale projects,
notably for defence. Supplementary guidance on treating large-scale projects
is given in Annex 10, with examples distinguishing between R&D and pre-
production development.

1.5.4. The identification of R&D in software, social sciences and service

activities
25. In recent years, the desire for better information on R&D in service
activities has been expressed. The basic definitions in this Manual were
originally developed for manufacturing industry and research in the natural
sciences and engineering. Specific problems therefore arise for applying them
to service activities, which often involve software applications and research in
the social sciences. In Chapter 2, a new section (2.4) is devoted to a discussion
of these problems.

1.5.5. R&D administration and other supporting activities

26. To carry out the R&D activities described above, funds must be
provided and the project and its financing must be managed. The R&D
funding activities of policy agencies, such as ministries of science and
technology or research councils, do not constitute R&D. In the case of in-house
management of R&D projects and their financing, a distinction is made
between direct support for R&D by persons such as R&D managers closely
associated with individual projects, who are included in both the personnel
and expenditure series, and persons such as financial directors, whose
support is indirect or auxiliary and who are included in the expenditure series
only as an element of overheads. Auxiliary support by catering or transport
services is also included in overheads. These distinctions are discussed
further in Chapters 2, 5 and 6.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 19

1 AIM AND SCOPE OF THE MANUAL

1.6. R&D in all fields of science and technology is covered

27. The first two versions of the Manual covered only the natural sciences
and engineering. The social sciences and humanities were incorporated in
the third edition (OECD, 1976), adopted in 1974. Although the Manual
recommends standard practice, it is understood that, for various reasons,
some deviations may have to be accepted for the social sciences and
humanities (SSH). Experience in member countries differs: some find that
surveys can cover all sciences in all sectors, others find that common
procedures are not always appropriate.
28. The special problems of measuring SSH R&D are raised as they occur
in the various chapters of the Manual.

1.7. Measures of R&D inputs

29. For statistical purposes, two inputs are measured: R&D expenditures
and R&D personnel. Both inputs are normally measured on an annual basis: so
much spent during a year, so many person-years used during a year. Both
series have their strengths and weaknesses, and, in consequence, both are
necessary to secure an adequate representation of the effort devoted to R&D.

1.7.1. R&D personnel

30. Data on the utilisation of scientific and technical personnel provide
concrete measurements for international comparisons of resources devoted
to R&D. It is recognised, however, that R&D inputs are only one part of the
input of a nation’s human resources to the public welfare; scientific and
technical personnel contribute much more to industrial, agricultural and
medical progress through their involvement in production, operations, quality
control, management, education and other functions. The measurement of
these stocks of scientific and technical manpower is the subject of the
Canberra Manual (OECD, 1995); the focus in this Manual is instead the
measurement and classification of R&D resources.
31. For R&D personnel data, the problem arises of reducing such data to
full-time equivalent (FTE) or person-years spent on R&D (see Chapter 5,
Section 5.3). It is recommended therefore that data should also be collected in
terms of physical persons (“headcount”) so that they can be used in overall
models and databases on S&T personnel.
32. The national R&D effort requires a wide variety of personnel, from the
Nobel prize winner to the winner’s secretary, from the designer of space
experiments to the breeder of laboratory animals. Because of the range of
skills and education required, it is essential to classify R&D personnel into
categories.

20 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 1 AIM AND SCOPE OF THE MANUAL

33. Two systems are now used by OECD member countries to classify
persons engaged in R&D. Chapter 5, Section 5.2, contains definitions both for
a classification by occupation, linked as far as possible to the International
Standard Class ific ation of Occupation – ISCO (ILO, 19 90 ), and for a
classification by level of formal qualification based entire ly on the
International Standard Classification of Education – ISCED (UNESCO, 1997).
While it would be desirable to have data based on both classifications, most
member countries use only one. As data are available by occupation for most
OECD countries, the fact that a few still collect only qualification data for some
or all sectors means that serious problems of international comparability
remain. It may be argued that, in an efficient system, there should be no major
difference between the two – that all those employed as researchers, for
example, would have university degrees and that all university graduates
working on R&D would be employed as researchers. In practice, however, this
is not the case. For example, a number of mature researchers do not have
university-level qualifications, although they have other post-secondary
qualifications or equivalent experience. Conversely, an increasing number of
young university graduates are employed not as researchers but as high-level
technicians or as support staff.

1.7.2. R&D expenditures

34. The basic measure is “intramural expenditures”; i.e. all expenditures for
R&D performed within a statistical unit or sector of the economy. Another
measure, “extramural expenditures”, covers payments for R&D performed
outside the statistical unit or sector of the economy. For R&D purposes, both
current costs and capital expenditures are measured. In the case of the
government sector, expenditures refer to direct rather than indirect
expenditures. Depreciation costs are excluded. Further details on the coverage
and content of R&D expenditures are given in Chapter 6, Section 6.2, of the
Manual.
35. R&D is an activity involving significant transfers of resources among
units, organisations and sectors and especially between government and other
performers. It is important for science policy advisors and analysts to know who
finances R&D and who performs it. Chapter 6 deals with ways of tracing the
flow of R&D funds. It is stressed that such flows should be based on replies from
performers of R&D and not on replies from the source of funds (see Chapter 6,
Section 6.3). Guidelines are suggested for the treatment of public general
university funds (GUF), also called general university funds, i.e. that part of
university research which is financed from the general grant from ministries of
education, which is destined for both education and research. Such flows may
represent up to over half of all support for university research and are an
important share of all public support for R&D.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 21

1 AIM AND SCOPE OF THE MANUAL

36. The main disadvantage of expressing R&D input series in monetary

terms is that they are affected by differences in price levels between countries
and over time. It can be shown that current exchange rates often do not reflect
the balance of R&D prices between countries and that in times of high inflation
general price indices do not accurately reflect trends in the cost of performing
R&D. The Manual recommends using purchasing power parities (PPP) and the
implicit gross domestic price (GDP) price index for R&D statistics, although it is
recognised that they reflect the opportunity cost of the resources devoted to
R&D rather than the “real” amounts involved. Methods of developing special
R&D deflators and R&D exchange rates are discussed in Annex 9.

1.7.3. R&D facilities

37. Indicators of facilities available for R&D may be envisaged but are
seldom collected and are not discussed in the Manual. Standardised
equipment, library facilities, laboratory space, journal subscriptions and
standardised computer time would all be possible measures.

1.7.4. National R&D efforts

38. Although R&D activities take place throughout the economy, they are
often perceived as a whole for science policy purposes, i.e. as the “national
R&D effort”. One aim of the Manual is therefore to establish specifications for
R&D input data that can both be collected from a wide range of performers
and also be aggregated to meaningful national totals. The main expenditure
aggregate used for international comparison is gross domestic expenditure on
R&D (GERD), which covers all expenditures for R&D performed on national
territory in a given year. It thus includes domestically performed R&D which is
financed from abroad but excludes R&D funds paid abroad, notably to
international agencies. The corresponding personnel measure does not have a
special name. It covers total personnel working on R&D (in FTE) on national
territory during a given year. International comparisons are sometimes
restricted to researchers (or university graduates) because it is considered that
they are the true core of the R&D system.

1.8. Globalisation of R&D and R&D co-operation

39. Various studies have shown that R&D activities are more and more a
worldwide activity and that a bigger share of R&D is performed in co-operation
with individual researchers, research teams and research units. Multinational
enterprises play an increasing role as does R&D co-operation between
university and other research units and enterprises, both formally, via
organisations such as the European Union (EU) or the European Organization for

22 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 1 AIM AND SCOPE OF THE MANUAL

Nuclear Research (CERN), or informally, via multilateral and bilateral

agreements. There is a clear need for more information on these trends.
40. The present edition of the Frascati Manual takes the globalisation
process into account by suggesting more detailed breakdowns of sources of
funds for R&D and extramural R&D for transactions with units abroad. Further
information on the need for indicators of technological globalisation will be
found in a substantial review of different aspects of measuring globalisation
(Manual of Economic Globalisation Indicators, provisional title, forthcoming). As
the R&D activities of multinational groups of enterprises are usually
organised, managed and financed at group level or group division level, it is
sometimes very difficult, if not impossible, to identify R&D performed in units
of the group in different countries and to obtain information on R&D flows
between these units.
41. R&D co-operation is an area that is not traditionally covered in R&D
surveys. More information on R&D co-operation would be highly desirable for
policy makers. However, owing to lack of sufficient experience in member
countries, it has not been possible to include recommendations for data
collection on R&D co-operation in this edition of the Manual. There is some
relevant information on R&D flows between different kinds of institutions.
Experience with innovation surveys has shown that it is possible to ask
different types of units in different geographical regions a simple question on
co-operation. This could be tried in R&D surveys as well, so that, in future, it
may be possible to make explicit recommendations.

1.9. Classification systems for R&D

42. To understand R&D activity and its role, one must examine it in terms
of the organisations performing and funding R&D (institutional classification)
and in terms of the nature of the R&D programmes themselves (functional
distribution).
43. It is usual to use basic institutional classifications in national (and
international) R&D surveys, as they facilitate the survey process, and combine
them with functional distributions to obtain a fuller understanding of the
situation described by the statistics.

1.9.1. Institutional classifications

44. In the institutional approach, attention focuses on the characteristic
properties of the performing or funding institutions. All units are classified
according to their principal (economic) activity. In this approach, all of the
R&D resources of the statistical unit are allocated to one class or sub-class.
The advantage is that R&D data are generally collected within the same
framework as regular economic statistics; this simplifies surveying and

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 23

1 AIM AND SCOPE OF THE MANUAL

facilitates comparisons between R&D and other economic data. The main
disadvantage is that it does not exactly describe the R&D activities of the unit,
which may not always be directly related to its “official” activity.
45. Chapter 3 of the Manual deals with the institutional classifications
used. In order to ensure maximum comparability with regular economic or
s ocial statistics, thes e are, as far as possible, based on existing UN
classifications. The main institutional classification of national R&D efforts is
by sector. Five sectors are identified: business enterprise, government, private
non-profit (PNP), higher education and abroad. Sub-classifications are given
for three of the four national sectors (business enterprise, PNP and higher
education) and additional institutional classifications, designed to reveal
national differences in sectoring, are suggested.

1.9.2. Functional distribution

46. In the functional approach, discussed in Chapter 4, attention focuses
on the character of the R&D itself. The nature of the R&D activities performed
by the unit is examined, and these are broken down in various ways to show
their distribution by type of R&D, product field, objective, field of science, etc.
Thus, the functional approach provides data that are more detailed and, since
international differences in institutional patterns have less influence, they are
theoretically more internationally comparable than those resulting from
institutional classification. This approach is, however, sometimes difficult to
apply in practice. This is particularly true for analysis by type of R&D (basic
research, applied research, experimental development) which is, on the one
hand, of undoubted science policy interest but, on the other, is based on an
oversimplified model of the workings of the scientific and technological
system. It also involves an important element of subjective assessment by the
respondent. This question is discussed in Chapter 4, Section 4.2.3.
47. The distinction between military and civil R&D is considered as an
important functional breakdown of the national R&D effort. In most OECD
countries, defence R&D plays a relatively minor role. However, in a few
countries that perform a high level of R&D, defence R&D expenditure
approaches or exceeds half of total government R&D expenditure. As a result,
international comparisons differ, depending on whether defence R&D is or is
not included. The demand for defence R&D fluctuates with changing political
situations, and therefore its long-term trend varies differently from that of
civil R&D. This means that, within the overall picture of national R&D effort, it
will always be necessary to separate the two categories of R&D expenditure.
Defence R&D is further discussed in Annex 10.
48. While these functional distributions are more detailed than the
institutional classifications, they are still not detailed enough to be of use to one

24 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 1 AIM AND SCOPE OF THE MANUAL

significant class of potential users of R&D data, i.e. the person interested in one
very specific sub-item, such as a sub-field of science or a product field
(holography or computer controls for machine tools). As already pointed out,
the Manual is essentially designed to measure national R&D efforts and to
categorise them in various ways. Except for special inventories of specific fields,
few member countries have pushed sub-categorisation to such a detailed level,
and it is unlikely that such detail would be obtainable at the OECD level.
49. Furthermore, it is difficult to establish norms for categories of interest
to national governments when reviewing the types of research funded from
public monies, when such research may have various policy connotations.
Strategic research is one area that has received considerable attention. It is
generally taken to mean research which a nation sees as a priority for
developing its research base and ultimately its economy. What is and is not
strategic varies among member countries. Nevertheless, in recognition of the
policy importance of strategic research in certain countries, Chapter 4 of the
Manual gives some attention to its identification.

1.10. R&D surveys, reliability of data and international

comparability
50. While a certain amount of R&D data can be derived from published
sources, there is no substitute for a special R&D survey. Most of the Manual
has been drafted on the assumption that surveys of at least all the major
national performers of R&D will be made. Nevertheless, it may be necessary
for both respondents and surveying agencies to produce estimates; this
question is discussed at length in Chapter 7.
51. It is hard to generalise about how far such estimates are necessary or
how far they affect the reliability of the data, as the situation will vary from
country to country. Nevertheless, it is generally the case that “subjective”
estimation by respondents is probably greatest for the breakdown between
basic research, applied research and experimental development, while the use
of “rule of thumb” estimation by survey agencies is probably greatest for R&D
in the higher education sector. As a consequence, these data should be treated
with circumspection. Annex 2 and a special supplement to the 1980 edition of
the Manual give further guidance on this topic (OECD, 1989b).
52. National surveys which provide R&D data that are reasonably accurate
and relevant to national users’ needs may not be internationally comparable.
This may simply be because national definitions or classifications deviate
from international norms. Such cases are generally documented in footnotes.
The situation is more complex when the national situation does not
correspond to the international norms. This is often true for sector analysis;
for administrative reasons, apparently similar institutions may be placed in

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 25

1 AIM AND SCOPE OF THE MANUAL

different sectors in different countries. Moreover, national perceptions of

these norms may be different, notably for type of research analysis and for the
analysis of R&D personnel by occupation. Such differences are impossible to
quantify.

1.11. Government budget appropriations or outlays for R&D

(GBAORD)
53. GBAORD data are often available much earlier than the results of
retrospective R&D surveys and are framed in categories of particular interest
to policy makers.
54. This topic is discussed separately in Chapter 8. Although the general
definitions in Chapter 2 apply to GBAORD, specifications in the following
chapters, which are essentially designed for performer-based reporting, often
do not.
55. This type of analysis essentially seeks to ascertain government
intentions or objectives when committing money to R&D. R&D funding is thus
defined by the funder (including public GUF) and may be both forecast (budget
proposals or initial budget appropriations) or retrospective (final budget or
outlay). Whereas R&D statistics proper are collected by means of especially
designed surveys, government R&D funding data generally have to be derived,
at some stage or another, from national budgets, which are based on their own
standard methods and terminology. Although the links between survey and
GBAORD data have improved in recent years, the analysis will always be a
balance between what is desirable from the R&D point of view and what is
available from the budget or related sources.
56. The aim of classifying GBAORD by socio-economic objective is to help
governments to formulate science and technology policy. Consequently, the
categories have to be broad, and the series are intended to reflect the amount
of res ources devoted to each prim ary purpos e (de fen ce, in dustrial
development, etc.). Nevertheless, the fit is never perfect and always reflects
the policy intentions of a given programme rather than its precise content.
Because of this and because of methodological constraints on the way data are
compiled, the strict level of international comparability is probably lower for
GBAORD data than for most of the other series discussed in the Manual.

1.12. Topics of special interest

57. There is often a demand for R&D data for a specific priority area, which
cuts across the standard institutional and functional classifications. Data to
meet this demand often have to be built up from special extractions or
tabulations. Annexes 4 and 5 deal with currently popular priority areas.

26 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 1 AIM AND SCOPE OF THE MANUAL

58. Health R&D has become a policy concern in recent years, and various
international studies have been made. Health R&D data are not directly
available from any of the standard classifications described in the Manual. A
pragmatic method of deriving estimates of health-related R&D from existing
data sources is described in Annex 4. It is an aid to data compilation and
interpretation and should not be regarded as an international recommendation.
59. The OECD is developing statistics and indicators on the information
economy and information society. It is possible to calculate an aggregate for
R&D in selected information and communication technology (ICT) sectors on
the basis of the agreed list of industries belonging to the ICT sector, as
described in Annex 4.
60. Following information technology, biotechnology is expected to be the
next pervasive technology of great significance for future economic
development. The OECD has started work to develop a statistical framework
for biotechnology. Some ideas for questions on biotechnology in R&D surveys
and the concept of a special survey on biotechnology are presented in
Annex 4.
61. The regional distribution of R&D activities is of great policy interest not
only within the EU but also in other OECD countries, especially those with
federal constitutions. A recommendation to distribute some variables
by region is included in Chapters 5 and 6, and Annex 5 explains some
methodological aspects.

1.13. A final word to the user of R&D data

62. To conclude, four general points may be made about the use of both
R&D statistics and R&D funding data:
– Such series are only a summary quantitative reflection of very complex
patterns of activities and institutions. For this reason, it may be dangerous
to use them “neat”. They should, as far as possible, be analysed in the light
of relevant qualitative information. Particularly in the case of international
comparisons, the size, aspirations, economic structure and institutional
a rran gem en ts of th e c oun tr ie s c on c er n e d sh o uld be taken in to
consideration.
– Users generally refer to R&D data with a question in mind: “Is our national
university research effort declining?” “Does my firm spend a higher
proportion of its funds on basic research than the average for my industry?”,
etc. To answer such questions, it is necessary to identify the relevant basic
data and then use them to construct an R&D indicator. Some basic data may
be accurate enough to answer one question but not another. For example,
GBAORD data are useful for answering general questions about trends in
easily defined objectives: “Is there any sign that defence R&D is picking up

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 27

1 AIM AND SCOPE OF THE MANUAL

again in the OECD area?” They are not suitable for specific questions about
less easily defined objectives: “Does my country spend more or less in
absolute terms on R&D for environmental protection than country X?”
– One particularly useful way of constructing such indicators for international
comparisons is to compare R&D inputs with a corresponding economic
series, for example, by taking GERD as a percentage of GDP. Such broad
indicators are fairly accurate but may be biased if there are major differences
in the economic structure of the countries compared. For example, the
activities of big R&D-intensive multinationals may influence the GERD/GDP
ratio in a particular country quite significantly. The classifications and norms
used to collect R&D statistics are, as far as possible, compatible with those for
general statistics, and, while it is much more difficult to make detailed
comparisons between R&D and non-R&D series, establishing such
“structural” R&D indicators can be particularly revealing.
– The problems of data quality and comparability noted above are
characteristic of the whole range of data on dynamic socio-economic
activities – such as employment or international trade – which are
important to policy makers, managers, analysts and others. The philosophy
underlying the evolution of R&D statistical standards in the Manual has
been to identify and gradually resolve these problems by exploring various
approaches and learning from member countries’ experience.

28 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ISBN 92-64-19903-9
Frascati Manual 2002
Proposed Standard Practice for Surveys on Research
and Experimental Development
© OECD 2002

Chapter 2

Basic Definitions and Conventions

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 29

2 BASIC DEFINITIONS AND CONVENTIONS

2.1. Research and experimental development (R&D)

63.
Research and experimental development (R&D) comprise
creative work undertaken on a systematic basis in order to
increase the stock of knowledge, including knowledge of man,
culture and society, and the use of this stock of knowledge to
devise new applications.

64. The term R&D covers three activities: basic research, applied research and
experimental development; these are described in detail in Chapter 4. Basic
research is experimental or theoretical work undertaken primarily to acquire new
knowledge of the underlying foundation of phenomena and observable facts,
without any particular application or use in view. Applied research is also original
investigation undertaken in order to acquire new knowledge. It is, however,
directed primarily towards a specific practical aim or objective. Experimental
development is systematic work, drawing on existing knowledge gained from
research and/or practical experience, which is directed to producing new
materials, products or devices, to installing new processes, systems and services,
or to improving substantially those already produced or installed. R&D covers
both formal R&D in R&D units and informal or occasional R&D in other units.

2.2. Activities to be excluded from R&D

65. For survey purposes, R&D must be distinguished from a wide range of
related activities with a scientific and technological basis. These other
activities are very closely linked to R&D both through flows of information and
in terms of operations, institutions and personnel, but as far as possible, they
should be excluded when measuring R&D.
66. These activities will be discussed here under four headings:
– Education and training (Section 2.2.1).
– Other related scientific and technological activities (Section 2.2.2).
– Other industrial activities (Section 2.2.3).
– Administration and other supporting activities (Section 2.2.4).
67. The practical definitions given here are intended solely to exclude
these activities from R&D.

30 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 2 BASIC DEFINITIONS AND CONVENTIONS

2.2.1. Education and training

68. All education and training of personnel in the natural sciences,
engineering, medicine, agriculture, the social sciences and the humanities in
universities and special institutions of higher and post-secondary education
should be excluded. However, research by students at the PhD level carried out
at universities should be counted, whenever possible, as a part of R&D (see
Section 2.3.2).

2.2.2. Other related scientific and technological activities

69. The following activities should be excluded from R&D except when
carried out solely or primarily for the purposes of an R&D project (see
Section 2.3.1 for examples).

Scientific and technical information services

70. The specialised activities of:

}
– Collecting – Scientific and technical personnel
– Coding – Bibliographic services
– Recording – Patent services
– Classifying – Scientific and technical information,
by extension and advisory services
– Disseminating – Scientific conferences
– Translating
– Analysing
– Evaluating

are to be excluded, except when conducted solely or primarily for the purpose
of R&D support (e.g. the preparation of the original report of R&D findings
should be included in R&D).

General purpose data collection

71. General purpose data collection is undertaken generally by
government agencies to record natural, biological or social phenomena that
are of general public interest or that only the government has the resources to
record. Examples are routine topographical mapping; routine geological,
hydrological, oceanographic and meteorological surveying; astronomical
observations. Data collected solely or primarily as part of the R&D process are
included in R&D (e.g. data on the paths and characteristics of particles in a
nuclear reactor). The same re asoning applies to the processing and
interpretation of the data. The social sciences, in particular, are very
dependent on an accurate record of facts relating to society in the form of
censuses, sample surveys, etc. When these are specially collected or
processed for the purpose of scientific research, the cost should be attributed

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 31

2 BASIC DEFINITIONS AND CONVENTIONS

to research and should cover the planning, systematising, etc., of the data.
However, data collected for other or general purposes, such as quarterly
sampling of unemployment, should be excluded from R&D even if exploited
for research. Market surveys should also be excluded.

Testing and standardisation

72. This concerns the maintenance of national standards, the calibration
of secondary standards and routine testing and analysis of materials,
components, products, processes, soils, atmosphere, etc.

Feasibility studies
73. Investigation of proposed engineering projects, using existing
t e ch n iq ue s to p rov id e a d d iti on a l in f or ma ti on be f ore de c i di n g o n
implementation, is not R&D. In the social sciences, feasibility studies are
investigations of the socio-economic characteristics and implications of specific
situations (e.g. a study of the viability of a petrochemical complex in a certain
region). However, feasibility studies on research projects are part of R&D.

Specialised health care

74. This concerns routine investigation and normal application of
specialised medical knowledge. There may, however, be an element of R&D in
what is usually called “specialised health care”, when it is carried out, for
example, in university hospitals (see Section 2.3.2).

Patent and licence work

75. This includes all administrative and legal work connected with
patents and licences. However, patent work connected directly with R&D
projects is R&D.

Policy-related studies
76. In this context, “policy” refers not only to national policy but also to
policy at regional and local levels, as well as that of business enterprises in the
pursuit of their economic activity. Policy-related studies cover a range of
activities, such as the analysis and assessment of the existing programmes,
policies and operations of government departments and other institutions;
the work of units concerned with the continuing analysis and monitoring of
external phenomena (e.g. defence and security analysis); and the work of
legislative commissions of inquiry concerned with general government or
departmental policy or operations.

32 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 2 BASIC DEFINITIONS AND CONVENTIONS

Routine software development

77. Software-related activities of a routine nature are not considered to be
R&D. Such activities include work on system-specific or programme-specific
advances which were publicly available prior to the commencement of the
work. Technical problems that have been overcome in previous projects on the
same operating systems and computer architecture are also excluded. Routine
computer maintenance is not included in R&D (see Section 2.4.1 for a more
detailed discussion of borderline problems between software development
and R&D).

2.2.3. Other industrial activities

78. These can be considered under two, to some extent overlapping,
headings.

Other innovation activities

79. In the Oslo Manual (OECD, 1997a), these are defined as all those
scientific, technical, commercial and financial steps, other than R&D,
necessary for the implementation of new or improved products or services
and the commercial use of new or improved processes. These include
acquisition of technology (embodied and disembodied), tooling up and
industrial engineering, industrial design n.e.c., other capital acquisition,
production start-up and marketing for new and improved products.

Production and related technical activities

80. This covers industrial preproduction and production and distribution
of goods and services and the various allied technical services in the business
enterprise sector and in the economy at large, together with allied activities
using social science disciplines, such as market research.

2.2.4. Administration and other supporting activities

81. This category has two components.

Purely R&D-financing activities

82. The raising, management and distribution of R&D funds to performers
by ministries, research agencies, foundations or charities is not R&D. This is in
line with the instructions in the latest version of ISIC (UN, 1990).

Indirect supporting activities

83. This covers a number of activities which are not themselves R&D but
which provide support for R&D. By convention, R&D personnel data cover R&D
proper but exclude indirect supporting activities, whereas an allowance for

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 33

2 BASIC DEFINITIONS AND CONVENTIONS

these is included under overheads in R&D expenditure of performers. Typical

examples are transportation, storage, cleaning, repair, maintenance and
security activities. Administration and clerical activities undertaken not
exclusively for R&D, such as the activities of central finance and personnel
departments, also come under this heading.

2.3. The boundaries of R&D

2.3.1. Criteria for distinguishing R&D from related activities
84. The basic criterion for distinguishing R&D from related activities is the
presence in R&D of an appreciable element of novelty and the resolution of
scientific and/or technological uncertainty, i.e. when the solution to a problem
is not readily apparent to someone familiar with the basic stock of common
knowledge and techniques for the area concerned. Table 2.1 identifies some
supplementary criteria for distinguishing R&D.

Table 2.1. Supplementary criteria for separating R&D from related

scientific, technological and industrial activities

A. What are the objectives of the project?

B. What is new or innovative about this project?

Is it seeking previously undiscovered phenomena, structures or relationships?
Does it apply knowledge or techniques in a new way?
Is there a significant chance that it will result in new (extended or deeper) understanding of phenomena,
relationships or manipulative principles of interest to more than one organisation?
Are the results expected to be patentable?

C. What staff is working on the project?

D. What methods are being used?

E. Under what programme is the project funded?

F. How general are the findings or results of the project likely to be?

G. Does the project fall more naturally into another scientific, technological or industrial activity?

Source: OECD.

85. One aspect of these criteria is that a particular project may be R&D if
undertaken for one reason, but not if carried out for another, as shown in the
following examples:
– In the field of medicine, routine autopsy on the causes of death is the
practice of medical care and is not R&D; special investigation of a particular
mortality to establish the side effects of certain cancer treatments is R&D.
Similarly, routine tests such as blood and bacteriological tests carried out
for doctors are not R&D, whereas a special programme of blood tests in
connection with the introduction of a new drug is R&D.

34 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 2 BASIC DEFINITIONS AND CONVENTIONS

– The keeping of daily records of temperatures or of atmospheric pressure is

not R&D but the operation of a weather forecasting service or general data
collection. The investigation of new methods of measuring temperature is
R&D, as are the study and development of new systems and techniques for
interpreting the data.
– R&D activities in the mechanical engineering industry often have a close
connection with design and drawing work. In small and medium-size
enterprises (SMEs) in this industry, there is usually no special R&D department,
and R&D problems are mostly dealt with under the general heading “design
and drawing”. If calculations, designs, working drawings and operating
instructions are made for the setting up and operating of pilot plants and
prototypes, they should be included in R&D. If they are carried out for the
preparation, execution and maintenance of production standardisation
(e.g. jigs, machine tools) or to promote the sale of products (e.g. offers, leaflets,
catalogues of spare parts), they should be excluded from R&D.

2.3.2. Problems at the borderline between R&D and education

and training
General approach
86. In institutions of higher education, research and teaching are always
very closely linked, as most academic staff do both, and many buildings, as
well as much equipment, serve both purposes.
87. Because the results of research feed into teaching, and because
information and experience gained in teaching can often result in an input to
research, it is difficult to define where the education and training activities of
higher education staff and their students end and R&D activities begin, and
vice versa. Its elements of novelty distinguish R&D from routine teaching and
other work-related activities. Deciding whether or not to consider as R&D
those scientific activities that are the by-products of educational or training
activities does present a problem.
88. It exists for a number of the following cases:
– Postgraduate students at the PhD level and their activities.
– Supervision of students by university staff.
– Specialised health care.
– Personal education of academic staff (own reading).

Postgraduate students at the PhD level

89. In some OECD countries, the “postgraduate student” is not a standard
national category. In such cases, the R&D activity of such persons is probably
included with that of other part-time teaching staff.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 35

2 BASIC DEFINITIONS AND CONVENTIONS

90. However, in countries where such students constitute a recognised

category, the borderline between their R&D and their education and training is
particularly hard to establish. The activities of both the postgraduate students
themselves and of their teachers need to be taken into consideration.
91. Parts of the curricula for studies at ISCED level 6 are highly structured,
involving, for instance, study schemes, set courses, compulsory laboratory
work, etc. Here, the teacher transmits knowledge and trains in research
methods. Students who fall under this heading typically attend compulsory
courses, study the literature on the subject, learn research methodology, etc.
These activities do not fulfil the criterion of novelty specified in the definition
of R&D.
92. In addition, in order to obtain a final qualification at ISCED level 6,
students are also expected to prove their competence by undertaking
relatively independent study usually containing the elements of novelty
required for R&D projects and presenting their results. These activities should,
therefore, be attributed to R&D, and any supervision by the teacher should be
as well. In addition to R&D performed within the framework of courses of
postgraduate education, it is possible for both teachers and students to be
engaged in other R&D projects.
93. In addition, students at this level are often attached to or directly
employed by the establishment in which they study and have contracts or
similar engagements which oblige them to teach at lower levels or to perform
other activities, such as specialised medical care, while allowing them to
continue their studies and to do research.
94. Borderlines between R&D and education at ISCED level 6 are illustrated
in Table 2.2 which, together with much of the above text, is based on the
relevant Nordic Manual, R&D Statistics in the Higher Education Sector: Work on
Improved Guidelines (Nordforsk, 1986). The more practical problems of applying
these concepts are dealt with in Chapter 5 (see Section 5.2.5).

Supervision of students
95. Closely allied to the problem of identifying the R&D element of
postgraduate students’ work is that of extracting the R&D component of the
time spent by academic supervisors on supervising these students and their
research projects.
96. Such supervision activities should be included in R&D only if they are
equivalent to the direction and management of a specific R&D project
containing a sufficient element of novelty and having as its object to produce
new knowledge. In such cases, both the academic staff member’s supervision
and the student’s work should be included as R&D. If the supervision merely
deals with the teaching of R&D methods and the reading and correction of

36 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 2 BASIC DEFINITIONS AND CONVENTIONS

Table 2.2. Borderline between R&D and education and training

at ISCED level 6
Education and training at level 6 R&D Other activities

Teachers 1. Teaching students at level 6.

2. Training students at level 6
in R&D methodology,
laboratory work, etc.
3. Supervision of R&D
projects required for student
qualification at level 6.
4. Supervision of other R&D
projects and performance
of own R&D projects.
5. Teaching at levels lower
than level 6.
6. Other activities.

Postgraduate students 1. Course work for formal

qualification.
2. Performing and writing up
independent studies (R&D
projects) required for formal
qualification.
3. Any other R&D activities.
4. Teaching at levels lower
than level 6.
5. Other activities.

Source: OECD.

theses and dissertations or the work of undergraduate students, it should be

excluded from R&D.

Specialised health care

97. In university hospitals where the training of medical students is an
important activity in addition to the primary activity of health care, the
activities of teaching, R&D and advanced as well as routine medical care are
frequently closely linked. “Specialised health care” is an activity that is
normally to be excluded from R&D (see Section 2.2.2). However, there may be
an element of R&D in what is usually called specialised health care, when
carried out, for example, in university hospitals. It is difficult for university
doctors and their assistants to evaluate the part of their overall activities that
is exclusively R&D. If, however, time and money spent on routine medical care
are included in the R&D statistics, R&D resources in the medical sciences will
be overestimated.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 37

2 BASIC DEFINITIONS AND CONVENTIONS

98. Usually, such specialised health care is not considered R&D, and all
medical care not directly linked to a specific R&D project should be excluded
from the R&D statistics.

Personal education of academic staff

99. This activity covers time spent on activities such as continued
professional learning (“own reading”), attendance at conferences and seminars,
etc.
100. In distinguishing R&D from related activities, the question of whether
“own reading” should be included as part of R&D activities is often raised. It is
certainly part of the general professional development of research staff and, in
the long term, the knowledge and experience gained are incorporated into the
researcher’s thinking about, if not into the actual implementation of, R&D.
Own reading, in fact, constitutes a cumulative process, and when the
information gained from this activity is translated into research activity, it will
be measured as R&D.
101. Only personal education carried out specifically for a research project
should be considered as an R&D activity.

2.3.3. Problems at the borderline between R&D and related scientific

and technological activities
General approach
102. Difficulties for separating R&D from other scientific and technological
activities arise when several activities are performed at the same institution.
In survey practice, using rules of thumb to make distinctions facilitates the
identification of the R&D portion. For example:
– Institutions or units of institutions and firms whose principal activity is
R&D often have secondary, non-R&D activities (e.g. scientific and technical
information, testing, quality control, analysis). Insofar as a secondary
activity is undertaken primarily in the interests of R&D, it should be
included in R&D activities; if the secondary activity is designed essentially
to meet needs other than R&D, it should be excluded.
– Institutions whose main purpose is an R&D-related scientific activity often
undertake some research in connection with this activity. Such research
should be isolated and included when measuring R&D.
103. The following examples illustrate the use of such rules of thumb:
– The activities of a scientific and technical information service or of a
research laboratory library, maintained predominantly for the benefit of the
research workers in the laboratory, should be included in R&D. The
activities of a firm’s documentation centre open to all the firm’s staff should

38 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 2 BASIC DEFINITIONS AND CONVENTIONS

be excluded from R&D even if it shares the same premises as the company
research unit. Similarly, the activities of central university libraries should
be excluded from R&D. These criteria only apply when it is necessary to
exclude the activities of an institution or a department in their entirety.
Where more detailed accounting methods are used, it may be possible to
impute part of the costs of the excluded activities to R&D overheads.
Whereas the preparation of scientific and technical publications is,
generally speaking, excluded, the preparation of the original report of
research findings should be included in R&D.
– Public bodies and consumer organisations often operate laboratories whose
main purpose is testing and standardisation. The staff of these laboratories
may also spend time devising new or substantially improved methods of
testing. Such activities should be included in R&D.
– General-purpose data collection is particularly important to social science
research, since without it many aspects of this research would not be feasible.
However, unless it is collected primarily for research purposes, it should not
be classified as a research activity. On the other hand, the larger statistical
institutes may carry out some R&D (e.g. conceptual and methodological work
in relation to the development of completely new or substantially modified
surveys and statistical systems, work on sampling methodologies, small area
statistical estimates). Whenever possible, such R&D should be included.

Specific cases
104. In certain cases, the theoretical criteria for distinguishing between
R&D and related scientific and technological activities are particularly difficult
to apply. Space exploration, mining and prospecting and the development of
social systems are three areas involving large amounts of resources, and any
variations in their treatment will have important effects on the international
comparability of the resulting R&D data. Large-scale projects also pose
problems for the definition of their R&D; these are discussed in Section 2.3.4.
The following conventions apply in the four areas mentioned.

• Space exploration

105. The difficulty with space exploration is that, in some respects, much
space activity may now be considered routine; certainly, most costs are incurred
for the purchase of goods and services which are not R&D. However, the object
of all space exploration is still to increase the stock of knowledge, so that it
should all be included in R&D. It may be necessary to separate the activities
associated with space exploration, including the development of vehicles,
equipment and techniques, from those involved in the routine placing of
orbiting satellites or the establishment of tracking and communication stations.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 39

2 BASIC DEFINITIONS AND CONVENTIONS

• Mining and prospecting

106. Mining and prospecting sometimes cause problems owing to a

linguistic confusion between research for new or substantially improved
resources (food, energy, etc.) and the search for existing reserves of natural
resources, a confusion which blurs the distinction between R&D and
surveying and prospecting. In theory, in order to establish accurate R&D data,
the following activities should be identified, measured and summed:
– The development of new surveying methods and techniques.
– Surveying undertaken as an integral part of a research project on geological
phenomena.
– Research on geological phenomena per se, undertaken as a subsidiary part
of surveying and prospecting programmes.
107. In practice, the last of these presents a number of problems. It is
difficult to frame a precis e defin ition that would be me an in gful to
respondents to national surveys. For this reason, only the following activities
should be included in R&D:
– The development of new or substantially improved methods and
equipment for data acquisition and for the processing and study of the data
collected and for the interpretation of these data.
– Surveying undertaken as an integral part of an R&D project on geological
phenomena per se, including data acquisition, processing and interpretation
undertaken for primarily scientific purposes.
108. It follows that the surveying and prospecting activities of commercial
companies will be almost entirely excluded from R&D. For example, the
sinking of exploratory wells to evaluate the resources of a deposit should be
considered as scientific and technological services.

• The development of social systems

109. In general, but more particularly in the field of the social sciences, the
purpose of a study is to prepare the way for decisions by policy makers at the
level of government (central, regional, local) or in industrial and trading
enterprises. Usually, such studies employ established methodologies, but it is
sometimes necessary to modify existing methodologies or to develop new
ones. This requires an appreciable amount of research. In theory, such
modification or development should be included in R&D, but one must be
aware of the difficulties involved in evaluating the appropriate share of R&D in
a given study. In practice, despite the technical and conceptual problems, it
may be feasible either to assign studies which include an appreciable element
of novelty entirely to research or to attempt to estimate the proportion of
research in those studies and attribute this to R&D (see also Section 2.4.2). For

40 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 2 BASIC DEFINITIONS AND CONVENTIONS

determining whether a particular activity should be regarded as R&D or be

attributed to R&D, the fact that the activity is called a study or that the report
resulting from the activity performed is called a study is irrelevant. If a
particular activity falls within the definition of R&D, then it is regarded as or
attributed to R&D; if not, it is excluded.

2.3.4. Problems at the borderline between R&D and other industrial

activities
General approach
110. Care must be taken to exclude activities which, although undoubtedly
a part of the innovation process, rarely involve any R&D, e.g. patent filing and
licensing, market research, manufacturing start-up, tooling up and redesign
for the manufacturing process. Some activities, such as tooling up, process
development, design and prototype construction, may contain an appreciable
element of R&D, thus making it difficult to identify precisely what should or
should not be defined as R&D. This is particularly true for defence and large-
scale civil industries such as aerospace. Similar difficulties may arise in
distinguishing public technology-based services such as inspection and
control from related R&D, as for example in the area of food and drugs
(Table 2.3).

Table 2.3. Some cases at the borderline between R&D

and other industrial activities
Item Treatment Remarks

Prototypes Include in R&D As long as the primary objective is to make further

improvements.
Pilot plant Include in R&D As long as the primary purpose is R&D.
Industrial design and drawing Divide Include design required during R&D. Exclude design
for production process.
Industrial engineering and tooling up Divide Include “feedback” R&D and tooling up industrial engineering
associated with development of new products and new
processes. Exclude for production processes.
Trial production Divide Include if production implies full-scale testing and subsequent
further design and engineering. Exclude all other associated
activities.
After-sales service and trouble-shooting Exclude Except “feedback” R&D.
Patent and licence work Exclude All administrative and legal work connected with patents
and licences (except patent work directly connected with R&D
projects).
Routine tests Exclude Even if undertaken by R&D staff.
Data collection Exclude Except when an integral part of R&D.
Public inspection control, enforcement Exclude
of standards, regulations

Source: OECD.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 41

2 BASIC DEFINITIONS AND CONVENTIONS

111. Chapter 4 defines experimental development as “systematic work,

drawing on knowledge gained from research and practical experience, which is
directed to producing new materials, products or devices; to installing new
processes, systems and services; or to improving substantially those already
produced or installed”. It is difficult to define precisely the cut-off point between
experimental development and pre-production development, such as
producing user demonstration models and testing, and production that is
applicable to all industrial situations. It would be necessary to establish a series
of conventions or criteria by type of industry. The basic rule originally laid down
by the US National Science Foundation (NSF) provides a practical basis for the
exercise of judgement in difficult cases. Slightly expanded, it states:

“If the primary objective is to make further tech nical

improvements on the product or process, then the work comes
within the definition of R&D. If, on the other hand, the product,
process or approach is substantially set and the primary
objective is to develop markets, to do pre-production planning or
to get a production or control system working smoothly, the work
is no longer R&D.”

112. Despite this elaboration, it can be difficult to apply in individual

industries. It may not be clear when there is an appreciable element of novelty,
or when a product/process is substantially set.

Specific cases
113. Some common problem areas are described below.

• Prototypes

114. A prototype is an original model constructed to include all the

technical characteristics and performances of the new product. For example,
if a pump for corrosive liquids is being developed, several prototypes are
needed for accelerated life tests with different chemicals. A feedback loop
exists so that if the prototype tests are not successful, the results can be used
for further development of the pump.
115. Applying the NSF criterion, the design, construction and testing of
prototypes normally falls within the scope of R&D. This applies whether only
one or several prototypes are made and whether they are made consecutively or
simultaneously. However, when any necessary modifications to the prototype(s)
have been made and testing has been satisfactorily completed, the end-point of
R&D has been reached. The construction of several copies of a prototype to

42 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 2 BASIC DEFINITIONS AND CONVENTIONS

meet a temporary commercial, military or medical need after successful testing

of the original, even if undertaken by R&D staff, is not part of R&D.

• Pilot plants

116. The construction and operation of a pilot plant is a part of R&D as long
as the principal purposes are to obtain experience and to compile engineering
and other data to be used in:
– Evaluating hypotheses.
– Writing new product formulae.
– Establishing new finished product specifications.
– Designing special equipment and structures required by a new process.
– Preparing operating instructions or manuals on the process.
117. If, as soon as this experimental phase is over, a pilot plant switches to
operating as a normal commercial production unit, the activity can no longer
be considered R&D even though it may still be described as a pilot plant. As
long as the primary purpose in operating a pilot plant is non-commercial, it
makes no difference in principle if part or all of the output is sold. Such
receipts should not be deducted from the cost of R&D activity.

• Large-scale projects and costly pilot plants

118. Large-scale projects, of which defence and aerospace are the most
significant types, usually cover a spectrum of activity from experimental to
pre-production development. Under these circumstances, the funding and/or
performing organisation often cannot distinguish between the R&D and other
elements of expenditure. The distinction between R&D and non-R&D
expenditures is particularly important in countries where a large proportion of
government R&D expenditure is directed to defence. Annex 10 provides
supplementary guidelines on this question.
119. It is very important to look closely at the nature of very costly pilot
plants or prototypes, such as the first of a new line of nuclear power stations
or of icebreakers. They may be constructed almost entirely from existing
materials and using existing technology, and they are often built for
simultaneous use for R&D and for providing the primary service concerned
(power generation, ice breaking). The construction of such plants and
prototypes should not be wholly credited to R&D. Only the additional costs
due to the prototype nature of these products should be attributed to R&D.

• Trial production

120. After a prototype has been satisfactorily tested and any necessary
modifications made, the manufacturing start-up phase may begin. It is related
to full-scale production; it may consist of product or process modification or

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 43

2 BASIC DEFINITIONS AND CONVENTIONS

retraining personnel in the new techniques or in the use of new machinery.

Unless the manufacturing start-up phase implies further design and
engineering, it should not be counted as R&D, since the primary objective is no
longer to make further improvements to the products but to start the
production process. The first units of a trial production run for a mass
production series should not be considered as R&D prototypes even if they are
loosely described as such.
121. For example, if a new product is to be assembled by automatic welding,
the process of optimising the settings on the welding equipment in order to
achieve maximum production speed and efficiency would not count as R&D
(even if joint-strength requirements have to be met).

• Trouble-shooting

122. Trouble-shooting occasionally shows the need for further R&D, but
more frequently it involves the detection of faults in equipment or processes
and results in minor modifications of standard equipment and processes. It
should not, therefore, be included in R&D.

• “Feedback” R&D

123. After a new product or process has been turned over to production
units, there will still be technical problems to be solved, some of which may
demand further R&D. Such “feedback” R&D should be included.

• Industrial design

124. The vast bulk of design work in an industrial area is geared towards
production processes and as such is not classified as R&D. There are, however,
some elements of design work which should be considered as R&D. These
include plans and drawings aimed at defining procedures, technical
specifications and operational features necessary to the conception,
development and manufacturing of new products and processes.
125. For example, if an engineering product which incorporates machined,
heat-treated and/or electroplated components has been developed, the
drawing up and documenting of the requirements for surface smoothness,
heat treatment procedures or electroplating process requirements, whether
incorporated in the drawings or as separate specification sheets, are
considered R&D.

• Tooling up and industrial engineering

126. In most cases, the tooling-up and industrial engineering phases of any
project are considered to be part of the production process.

44 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 2 BASIC DEFINITIONS AND CONVENTIONS

127. Three phases of tooling up can be identified:

– The first-time use of components (including the use of components
resulting from R&D efforts).
– The initial tooling of equipment for mass production.
– Installing equipment linked with the start of mass production.
128. However, if the tooling-up process results in further R&D work, such as
developments in the production machinery and tools, changes to the
production and quality control procedures or the development of new
methods and standards, these activities are classified as R&D.
129. “Feedback” R&D resulting from the tooling-up phase should be defined
as R&D.

• Clinical trials

130. Before new drugs, vaccines or treatments can be introduced on the

market, they must be tested systematically on human volunteers to ensure
that they are both safe and effective. These clinical trials are divided into four
standard phases, three of which take place before permission to manufacture
is accorded. For the purposes of international comparison, by convention,
clinical trial phases 1, 2 and 3 can be treated as R&D. Phase 4 clinical trials,
which continue testing the drug or treatment after approval and manufacture,
should only be treated as R&D if they bring about a further scientific or
technological advance. Moreover, not all activities undertaken prior to
permission to manufacture are considered to be R&D, especially when there is
a significant wait after the completion of phase 3 trials, during which
marketing and process development activities may be started.

2.3.5. Problems at the borderline between R&D administration and

indirect supporting activities
131. The R&D activities described above are supported by a number of other
activities. In R&D statistics, the practice is that personnel data should cover
only R&D proper, whereas expenditure data should cover the full cost of R&D,
including the indirect supporting activities which are treated as overheads
(see Section 2.2.4).
132. Some activities, such as the provision of library or computer services,
are R&D proper if they are intended exclusively for R&D, but indirect
supporting activities if they are provided by central departments for both R&D
and non-R&D uses (see Section 2.3.3). The same argument applies for
management, administration and clerical activities. When these contribute
directly to R&D projects and are undertaken exclusively for R&D, then they are
part of R&D proper and included in R&D personnel. Typical examples are the

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 45

2 BASIC DEFINITIONS AND CONVENTIONS

R&D manager who plans and supervises the scientific and technical aspects of
the project or the person who produces the interim and final reports
containing the results of the project. It remains a moot point whether the
bookkeeping associated with a specific R&D project is direct (R&D proper) or
indirect (ancillary) activity. By convention, it is R&D proper rather than an
indirect supporting activity if it is carried out in close proximity to the R&D
(see Chapter 5, Table 5.1 and Section 5.1).

2.4. Identifying R&D in software development, in the social

sciences and humanities and in service activities and industries
133. The model on which the Manual was originally based was that of
institutionally structured R&D in the natural sciences and engineering leading
to tangible technological innovations in primary and secondary industries.
Software development has since become a major intangible innovation
activity with a high R&D content. In addition, an increasing share of relevant
activities draws on the social sciences and humanities, and, together with
advances in computing, leads to intangible innovations in service activities
and products, with a growing contributions from service industries in the
business enterprise sector.
134. The tools developed for identifying R&D in traditional fields and
industries are not always easy to apply to these new areas. This section deals
with the problems of identifying R&D in software development, in the social
sciences and humanities and in service activities.

2.4.1. Identifying R&D in software development

135. For a software development project to be classified as R&D, its
completion must be dependent on a scientific and/or technological advance,
and the aim of the project must be the systematic resolution of a scientific
and/or technological uncertainty.
136. In addition to the software that is part of an overall R&D project, the R&D
associated with software as an end product should also be classified as R&D.
137. The nature of software development is such as to make identifying its
R&D component, if any, difficult. Software development is an integral part of
many projects which in themselves have no element of R&D. The software
development component of such projects, however, may be classified as R&D
if it leads to an advance in the area of computer software. Such advances are
generally incremental rather than revolutionary. Therefore, an upgrade,
addition or change to an existing programme or system may be classified as
R&D if it embodies scientific and/or technological advances that result in an
increase in the stock of knowledge. Use of software for a new application or
purpose, however, does not by itself constitute an advance.

46 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 2 BASIC DEFINITIONS AND CONVENTIONS

138. A scientific and/or technological advance in software may be achieved

even if a project is not completed, because a failure can increase knowledge of
the technology of computer software by showing, for example, that a
particular approach will not succeed.
139. Advances in other fields resulting from a software project do not
determine whether an advance in computer software has occurred.
140. The following examples illustrate the concept of R&D in software.
Should be included in R&D:
– R&D producing new theorems and algorithms in the field of theoretical
computer science.
– Development of information technology at the level of operating systems,
programming languages, data management, communications software and
software development tools.
– Development of Internet technology.
– Research into methods of designing, developing, deploying or maintaining
software.
– Software development that produces advances in generic approaches for
capturing, transmitting, storing, retrieving, manipulating or displaying
information.
– Experimental development aimed at filling technology knowledge gaps as
necessary to develop a software programme or system.
– R&D on software tools or technologies in specialised areas of computing
(image processing, geographic data presentation, character recognition,
artificial intelligence and other areas).
141. Software-related activities of a routine nature which do not involve
scientific and/or technological advances or resolution of technological
uncertainties are not to be included in R&D. Examples are:
– Business application software and information system development using
known methods and existing software tools.
– Support for existing systems.
– Converting and/or translating computer languages.
– Adding user functionality to application programmes.
– Debugging of systems.
– Adaptation of existing software.
– Preparation of user documentation.
142. In the systems software area, individual projects may not be
considered as R&D but their aggregation into a larger project may qualify for
inclusion. For example, changes in file structure and user interfaces in a
fourth-generation language processor may be made necessary by the

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 47

2 BASIC DEFINITIONS AND CONVENTIONS

introduction of relational technology. The individual changes may not be

considered R&D if viewed in their own right, but the entire modification
project may result in the resolution of scientific and/or technological
uncertainty and thus be classified as R&D.

2.4.2. Identifying R&D in the social sciences and humanities

143. The social sciences and humanities are covered in the Manual by
including in the definition of R&D “knowledge of man, culture and society”
(see Chapter 2, Section 2.1). For the social sciences and humanities, an
appreciable element of novelty or a resolution of scientific/technological
uncertainty is again a useful criterion for defining the boundary between R&D
and related (routine) scientific activities. This element may be related to the
conceptual, methodological or empirical part of the project concerned.
Related activities of a routine nature can only be included in R&D if they are
undertaken as an integral part of a specific research project or undertaken for
the benefit of a specific research project. Therefore, projects of a routine
nature, in which social scientists bring established methodologies, principles
and models of the social sciences to bear on a particular problem, cannot be
classified as research.
144. The following are examples of work which might fall into this routine
category are generally not R&D: commentary on the probable economic effects
of a change in the tax structure, using existing economic data; use of standard
techniques in applied psychology to select and classify industrial and military
personnel, students, etc., and to test children with reading or other disabilities.

2.4.3. Special problems for identifying R&D in service activities

145. Defining the boundaries of R&D in service activities is difficult, for two
main reasons: first, it is difficult to identify projects involving R&D; and,
second, the line between R&D and other innovative activities which are not
R&D is a tenuous one.
146. Among the many innovative projects in services, those that constitute
R&D result in new knowledge or use of knowledge to devise new applications,
in keeping with the definition in the first paragraph of this chapter.
147. Identifying R&D is more difficult in service activities than in
manufacturing because it is not necessarily “specialised”. It covers several
areas: technology-related R&D, R&D in the social sciences and humanities,
including R&D relating to the knowledge of behaviour and organisations. This
last notion is already included in the criterion “knowledge of man, culture and
society”, but it is particularly important in the case of service activities.
Because these types of R&D may be combined in a given project, it is
important to circumscribe clearly the various forms of R&D involved. If the

48 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 2 BASIC DEFINITIONS AND CONVENTIONS

analysis is confined to technology-related R&D, for example, R&D may be

understated. In many cases, R&D findings in service industries are embodied
in software which is not necessarily innovative from the technical point of
view but innovates by virtue of the functions that it performs (see Section 2.4.1).
148. Also, in service companies, R&D is not always organised as formally as
in manufacturing companies (i.e. with a dedicated R&D department,
researchers or research engineers identified as such in the establishment’s
personnel list, etc.). The concept of R&D in services is still less specific and
s ometimes goes unrecognised by the ente rprises involved. As more
experience becomes available on surveying R&D in services, the criteria for
identifying R&D and examples of service-related R&D may require further
development.

Criteria for identifying R&D in services

149. The following are among the criteria that can help to identify the
presence of R&D in service activities:
– Links with public research laboratories.
– The involvement of staff with PhDs, or PhD students.
– The publication of research findings in scientific journals, organisation of
scientific conferences or involvement in scientific reviews.
– The construction of prototypes or pilot plants (subject to the reservations
noted in Section 2.3.4).

Examples of R&D in selected service activities

150. The R&D activities listed below may serve as examples of R&D in
service activities. The general and supplementary criteria for distinguishing
R&D presented in Section 2.3.1 have also to be taken into account.
151. The general boundaries of R&D as defined above, especially in
Sections 2.2, 2.3.3 and 2.3.4, also largely apply to service activities. The
element of novelty is a basic criterion for distinguishing R&D from related
activities.

Examples of R&D in banking and insurance

– Mathematical research relating to financial risk analysis.
– Development of risk models for credit policy.
– Experimental development of new software for home banking.
– Development of techniques for investigating consumer behaviour for the
purpose of creating new types of accounts and banking services.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 49

2 BASIC DEFINITIONS AND CONVENTIONS

– Research to identify new risks or new characteristics of risk that need to be

taken into consideration in insurance contracts.
– Research on social phenomena with an impact on new types of insurance
(health, retirement, etc.), such as on insurance cover for non-smokers.
– R&D related to electronic banking and insurance, Internet-related services
and e-commerce applications.
– R&D related to new or significantly improved financial services (new
concepts for accounts, loans, insurance and saving instruments).

Examples of R&D in some other service activities

– Analysis of the effects of economic and social change on consumption and
leisure activities.
– Development of new methods for measuring consumer expectations and
preferences.
– Development of new survey methods and instruments.
– Development of tracking and tracing procedures (logistics).
– Research into new travel and holiday concepts.
– Launch of prototype and pilot stores.

50 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ISBN 92-64-19903-9
Frascati Manual 2002
Proposed Standard Practice for Surveys on Research
and Experimental Development
© OECD 2002

Chapter 3

Institutional Classification

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 51

3 INSTITUTIONAL CLASSIFICATION

3.1. The approach

152. The institutional approach focuses on the characteristic properties of
the performing or funding institutions. All of the unit’s R&D resources are
classified to one class or sub-class, according to the unit’s principal activity.

3.2. The reporting unit and the statistical unit

3.2.1. The reporting unit
153. The reporting unit is the entity from which the recommended data
items are collected. It will vary from sector to sector and from country to
country, depending on institutional structures, the legal framework for data
collection, traditions, national priorities and survey resources. In some
countries, data may be collected from R&D units; in others, it may be gathered
at a more aggregate level. The Manual can make no recommendation to
member countries concerning the reporting unit.

3.2.2. The statistical unit

154. The statistical unit is the entity for which the required statistics are
compiled. It may be an observation unit in which information is received and
statistics are compiled or an analytical unit, which statisticians create by
splitting or combining observation units with the help of estimations or
imputations in order to supply more detailed and/or homogeneous data than
would otherwise be possible.
155. In principle, the statistical unit should be uniform, within sectors, for
all countries. In practice, however, this goal is never fully achieved. One reason
is that structures are different and names are different (or misleadingly
similar). Another is interaction with the reporting unit. If the reporting unit is
larger than the statistical unit, problems may arise for distributing the
data among the appropriate classification units. Various units will be
recommended in the following sections. Where necessary, reference is made
to the definitions of international standard classifications. However,
whenever member countries provide statistics for international comparisons,
the statistical units should be specified.

52 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 3 INSTITUTIONAL CLASSIFICATION

3.3. Sectors
3.3.1. Reasons for sectoring
156. To facilitate the collection of data, the description of institutional flows
of R&D funds and the analysis and interpretation of R&D data, the statistical
unit(s) classified should be grouped by sectors of the economy, following as
closely as possible standard classifications of economic activities. This offers
a number of important practical advantages:
– Different questionnaires and survey methods can be used for each sector
to take into account the different “mixes” of activities, the different
accounting s yste ms or the different re spon se pos sibilitie s of the
organisations.
– When measuring expenditure, the sectoral approach offers the most
reliable way of building up national aggregates.
– Sectoring offers a framework for analysing the flows of funds between R&D-
funding and R&D-performing entities.
– Since each sector has its own characteristics and its own kinds of R&D, this
classification throws some light on differences in the level and direction of
R&D.
– Insofar as the sectors are defined on the basis of a standard classification, it
may be possible to relate R&D to other statistical series. This may facilitate
an understanding of the role of R&D in economic development and the
formulation of science policy.
– The institutions of the various sectors are sensitive to different government
policy initiatives.

3.3.2. Choice of sectors

157. The System of National Accounts (SNA) (UN, 1968) stated that “in any
national accounting system transactors are necessarily grouped... but they
need not be grouped in the same way in all parts of the system and, indeed, it
is not desirable that they should be”. The SNA distinguishes between the
following sectors: non-financial corporations; financial corporations; general
government; non-profit institutions serving households; and households.
158. The following definitions of sectors for R&D surveys are largely based
on the SNA 93 (CEC et al., 1994), with the difference that higher education has
been established as a separate sector and households have, by convention,
been merged with the private non-profit (PNP) sector. Here, as in the SNA,
non-profit institutions (NPIs) have been distributed among sectors. For a more
detailed discussion of the relationships between SNA sectors and the sectors
proposed below for R&D surveys, see Annex 3.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 53

3 INSTITUTIONAL CLASSIFICATION

159. Five sectors are identified and discussed below:

– Business enterprise (see Section 3.4).
– Government (see Section 3.5).
– Private non-profit (see Section 3.6).
– Higher education (see Section 3.7).
– Abroad (see Section 3.8).
These, in turn, are divided into sub-sectors appropriate to each sector.

3.3.3. Problems of sectoring

160. In view of the diverse ways in which most contemporary institutions
have developed, sector definitions cannot be precise because, like the SNA
from which they are partly drawn, they are based on a combination of
sometimes conflicting criteria such as function, aim, economic behaviour,
sources of funds and legal status.
161. It is therefore not always clear in which sector a given institute should
be classified, and an arbitrary decision may have to be made. Institutions may
straddle two sectors; or, even if the conceptual distinction is clear, established
legal and administrative affiliations or political considerations may prevent
the application of this conceptual distinction in practice.
162. When two countries classify institutions with the same or similar
functions in different sectors, national survey results will not be completely
internationally comparable. Such divergences are unavoidable, as R&D
surveys are primarily undertaken to serve national purposes. For international
surveys, however, data should be collected and submitted in as much detail as
possible to allow for rearrangement for international comparisons. This is the
reason for the inclusion of “other institutional sub-classifications” for each
sector. Figure 3.1 presents a decision tree as a guide for classifying R&D units
by institutional sector.

3.4. Business enterprise sector

3.4.1. Coverage
163. The business enterprise sector includes:

● All firms, organisations and institutions whose primary

activity is the market production of goods or services (other
than higher education) for sale to the general public at an
economically significant price.
● The private non-profit institutions mainly serving them.

54 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002
Figure 3.1. Decision tree for sectoring R&D units

Institution

Does it purvey higher education?

Yes No

Higher Does it sell its output

education at an economical price?

Yes No

Business Is it controlled and mainly financed

enterprise by business enterprise units at home or abroad?

Yes No

Business enterprise Is it controlled and mainly financed government?

Yes No

Is it controlled and mainly financed by private

Is it administered by the higher education sector?
non-profit units serving households?

Yes No Yes No

Higher Is it administered Problem units where finance

Government
education by higher education? and control do not match

3 INSTITUTIONAL CLASSIFICATION
Yes No Is the unit mainly financed by:

Higher Private
education Business enterprise? Higher education? Government?
non-profit
Yes Yes Yes

Business enterprise Higher education Government?

Source: OECD.
55
3 INSTITUTIONAL CLASSIFICATION

164. The core of the sector is made up of private enterprises (corporations or

quasi-corporations), whether or not they distribute profits. Among these
enterprises may be found some firms for which R&D is the main activity
(commercial R&D institutes and laboratories). Any private enterprises
producing higher education services should be included in the higher
education sector.
165. In addition, this sector includes public enterprises (public corporations
and quasi-corporations owned by government units) mainly engaged in
market production and sale of the kind of goods and services which are often
produced by private enterprises, although, as a matter of policy, the price set
for these may be less than the full cost of production. To qualify as market
production in this context, the charges should be related to the value (quality
and quantity) of the goods and services furnished, the decision to purchase
them should be voluntary, and the price charged should significantly affect
supply and demand. Any public enterprises producing higher education
services should be included in the higher education sector.
166. This sector also includes non-profit institutions that are market
producers of goods and services other than higher education. These are of two
kinds.
167. The first are NPIs engaged in market production whose main activity
is the production of goods and services for sale at prices designed to recover
most or all of their costs. Research institutes, clinics, hospitals, medical
practitioners in private, fee-paying practices, etc., may be able to raise
additional funds in the form of donations or own assets generating property
income which allow them to charge below average prices.
168. The second are NPIs serving business. These are typically created and
managed by associations of businesses whose activities they are designed to
promote, such as chambers of commerce and agricultural, manufacturing or
trade associations. These NPIs are usually financed by contributions or
subscriptions from the businesses concerned which provide “institutional”
support for their R&D. However, NPIs that carry out similar functions but are
controlled or mainly financed by government – for example if they depend for
their existence on a block grant from government – should be included in the
government sector.

3.4.2. The principal sector sub-classification

The classification list
169. For international comparisons of R&D statistics, units in the business
enterprise sector are classified into a number of significant industry groups
and sub-groups by the International Standard Industrial Classification (ISIC
Rev. 3, UN, 1990, and minirevision 3.1, 2002). Table 3.1 shows a rearrangement

56 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 3 INSTITUTIONAL CLASSIFICATION

Table 3.1. International Standard Industrial Classification arranged

for the purposes of R&D statistics
ISIC Rev. 3.1 NACE Rev. I.1
Division/Group/Class Division/Group/Class

AGRICULTURE, HUNTING, FORESTRY AND FISHING 01, 02, 05 01, 02, 05

MINING AND QUARRYING 10, 11, 12, 13, 14 10, 11, 12, 13, 14

MANUFACTURING 15-37 15-37

Food, beverages and tobacco 15 + 16 15 + 16

Food products and beverages 15 15
Tobacco products 16 16

Textiles, fur and leather 17 + 18 + 19 17 + 18 + 19

Textiles 17 17
Wearing apparel and fur 18 18
Leather products and footwear 19 19

Wood, paper, printing, publishing 20 + 21 + 22 20 + 21 + 22

Wood and cork (not furniture) 20 20
Paper and paper products 21 21
Publishing, printing and reproduction of recorded media 22 22

Coke, petroleum, nuclear fuel, chemicals and products, rubber

and plastics 23 + 24 + 25 23 + 24 + 25
Coke, refined petroleum products and nuclear fuel 23 23
Coke and nuclear fuel 23 (less 232) 23 (less 23.2)
Refined petroleum products 232 23.2
Chemicals and chemical products 24 24
Chemicals and chemical products (less pharmaceuticals) 24 (less 2423) 24 (less 24.4)
Pharmaceuticals 2423 24.4
Rubber and plastics products 25 25

Non-metallic mineral products 26 26

Basic metals 27 27
Basic metals, iron and steel 271 and 2731 27.1-27.3 + 27.51/52
Basic metals, non ferrous 272 and 2732 27.4 + 27.53/54

Fabricated metal products, machinery and equipment, instruments

and transport 28-35 28-35
Fabricated metal products, except machinery and equipment 28 28
Machinery and equipment, n.e.c. 29 29
Engines and turbines, except aircraft, vehicle and cycle 2911 29.11
Special purpose machinery 292 29.3 + 29.4 + 29.5 + 29.6
Machine-tools 2922 29.4
Weapons and ammunition 2927 29.6
Office, accounting and computing machinery 30 30
Electrical machinery and apparatus n.e.c. 31 31
Electrical motors, generators and transformers 311 31.1
Electricity distribution and control apparatus
(includes semiconductors) 312 31.2

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 57

3 INSTITUTIONAL CLASSIFICATION

Table 3.1. International Standard Industrial Classification arranged

for the purposes of R&D statistics (cont.)
ISIC Rev. 3.1 NACE Rev. I.1
Division/Group/Class Division/Group/Class

Insulated wire and cable (includes optic fibre cables) 313 31.3
Accumulators, primary cells and primary batteries 314 31.4
Electric lamps and lighting equipment 315 31.5
Other electrical equipment n.e.c. 319 31.6
Radio, television and communication equipment and apparatus 32 32
Electronic valves, tubes and components 321 32.1
TV, radio transmitters and line apparatus 322 32.2
TV and radio receivers, sound and video goods 323 32.3
Medical, precision and optical instruments, watches and clocks
(instruments) 33 33
Medical appliances, instruments and control equipment 331 33.1
Instruments and appliances for measuring, checking, testing,
navigating and other purposes, except industrial process control
equipment 3312 33.2
Industrial process control equipment 3313 33.3
Optical instruments and photographic equipment 332 33.4
Watches and clocks 333 33.5
Motor vehicles, trailers and semi-trailers 34 34
Other transport equipment 35 35
Ships and boats 351 35.1
Railway and tramway locomotives and rolling stock 352 35.2
Aircraft and spacecraft 353 35.3
Transport equipment, n.e.c. 359 35.4 + 35.5

Furniture; other manufacturing, n.e.c. 36 36

Furniture 361 36.1
Other manufacturing, n.e.c 369 36.2-36.5

Recycling 37 37

ELECTRICITY, GAS AND WATER SUPPLY 40, 41 40, 41

CONSTRUCTION 45 45

SERVICES SECTOR 50-99 50-99

Wholesale, retail trade and motor vehicle repair 50, 51, 52 50, 51, 52
Wholesale of computers, computer peripheral equipment
and software 5151 51.84
Wholesale of electronic parts and equipment 5152 51.86

Hotels and restaurants 55 55

Transport, storage and communications 60, 61, 62, 63, 64 60, 61, 62, 63, 64
Telecommunications 642 64.2
Other 60-64 less 642 60-64 less 64.2

Financial intermediation (includes insurance) 65, 66, 67 65, 66, 67

58 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 3 INSTITUTIONAL CLASSIFICATION

Table 3.1. International Standard Industrial Classification arranged

for the purposes of R&D statistics (cont.)
ISIC Rev. 3.1 NACE Rev. I.1
Division/Group/Class Division/Group/Class

Real estate, renting and business activities 70, 71, 72, 73, 74 70, 71, 72, 73, 74
Renting of office machinery and equipment (including computers) 7123 71.33
Computer and related activities 72 72
Software consultancy and supply 722 72.2
Research and development 73 73
Other business activities 74 74
Architectural, engineering and other technical activities 742 74.2 + 74.3

Community, social and personal service activities, etc. 75-99 75-99

GRAND TOTAL 01-99 01-99

Source: OECD.

of ISIC Rev. 3 which is suitable for such comparisons, with a key to the
corresponding European classification, NACE Rev. 1 (Eurostat, 1990). Countries
that use a national industrial classification system rather than ISIC Rev. 3
should use concordance tables to convert their industrially classified data to
ISIC Rev. 3. Every attempt should be made to maintain the consistency of
these concordances.

The statistical unit

170. R&D is one of the activities that a business enterprise may undertake.
The business enterprise is free to organise this activity according to its
production model. Thus, core R&D may be carried out in units attached to
production units or in central units serving the whole enterprise. In most
cases, the legal entity defined in ISIC Rev. 3, paragraphs 78 and 79, is the
appropriate unit. In some cases, separate legal entities may be established to
provide R&D services for one or more related legal entities. Ad hoc R&D is
usually carried out in an operational department of a business enterprise,
such as the industrial design, quality or production department.
171. Data requirements determine the choice of the statistical unit(s).
These requirements are described in detail in Chapter 6. However, the source
and application of R&D funds is one fundamental data item. This is generally
the concern of the legal entity that controls the performance of R&D rather
than the smaller units that actually carry out the work. The latter may have to
prepare budgets and record costs, but the business’s central administration
knows the source of the funds that cover expenditures. Contracts and taxation
are principal activities of the legal entity.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 59

3 INSTITUTIONAL CLASSIFICATION

172. The enterprise, as a statistical unit, is defined as the organisational

unit of a business which directs and controls the allocation of resources
relating to its domestic operations, and for which consolidated financial and
balance sheet accounts are maintained. From these accounts, it is possible to
derive international transactions, an international investment position and a
consolidated financial position for the unit. It is therefore recommended to
use the enterprise unit as the reporting unit and, with exceptions, as the
statistical unit in the business enterprise sector. Within a group of enterprises,
it is desirable to obtain separate returns for each of the legal units performing
R&D, using estimations if necessary.
173. When an enterprise is heterogeneous with regard to its economic
activities and carries out significant amounts of R&D for several kinds of
activities, the R&D should be subdivided if the necessary information can be
obtained. In some countries, this is done by a division into statistical units
corresponding to economic units within the enterprise. In others, the R&D
activity may be broken down according to product field data.

Criteria for classification

174. Classification of these statistical units according to principal activity
should be determined by “the class of the ISIC in which the principal activity,
or range of activities, of the unit is included” (ISIC Rev. 3, paragraph 114).
175. According to ISIC, the principal activity should be determined by
computing the contribution of value added of each activity leading to the
production of goods or the rendering of services. The activity providing the
greatest contribution to the enterprise’s value added determines the
classification of that enterprise. If it is not possible to compute value added,
the principal activity can be determined either on the basis of the gross output
of the goods sold or services rendered by each activity or by the number of
persons assigned to each of these activities (ISIC Rev. 3, paragraph 115).
176. When the R&D is carried out in a legal entity specialising in R&D:

● The unit should be classified in R&D for enterprises (ISIC Rev. 3,

Division 73).
and
● Additional information should be collected for analytical
purposes and international comparisons, in order to reflect the
division into specific industries benefiting from the R&D
activities. This may be done by requesting product field data. In
practice, this means giving ISIC codes for the industries served
(described in more detail in Chapter 4).

60 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 3 INSTITUTIONAL CLASSIFICATION

3.4.3. Other institutional sub-classifications

Type of institution
177. The evolving nature of the business sector both within countries and
worldwide requires subdividing both private and public enterprises.
178. If private enterprises are broken down between independent
enterprises and enterprises belonging to a group and between national and
foreign groups, some trends in the internationalisation of industry can be
examined.
179. It is recommended, therefore, that the following classification by type
of institution be used if possible:
– Private enterprises:
❖ Enterprise not belonging to any group.
❖ Enterprise belonging to a national group.
❖ Enterprise belonging to a foreign multinational group.
– Public enterprises:
❖ Enterprise not belonging to any group.
❖ Enterprise belonging to a national group.
– Other research and co-operative institutes.
180. Public enterprises are distinguished from private enterprises on
the basis of control. The SNA 93 (paragraph 4.72) makes the following
recommendation for the definition of public non-financial corporations:
“These consist of resident non-financial corporations and quasi-
corporations that are subject to control by government units, control
over a corporation being defined as the ability to determine general
corporate policy by choosing appropriate directors, if necessary. The
government may secure control over a corporation:
– by owning more than half the voting shares or otherwise controlling
more than half the shareholders’ voting power; or
– as a result of special legislation, decree or regulation which
empowers the government to determine corporate policy or to
appoint the directors.”
181. A group must be considered as foreign when the main shareholder is a
foreign resident with more than 50% ownership and voting power, either
directly or indirectly through subsidiaries. For more information, see the OECD
Manual of Economic Globalisation Indicators (provisional title, forthcoming).

Size of institution
182. Size generally affects the extent and nature of the R&D programmes of
entities in the business enterprise sector. Size may be classified on the basis of

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 61

3 INSTITUTIONAL CLASSIFICATION

employment or on the basis of revenue or other financial items. Employment

is a less ambiguous measure and therefore preferable. This classification
should be applied to statistical units in both the manufacturing and service
industries.
183. The following size groups (according to number of employees) are
proposed:
0
1-9
10-49
50-99
100-249
250-499
500-999
1 000-4 999
5 000 and above.
These categories have been chosen for a variety of reasons, in particular for
their ability to conform to the size classification adopted by the European
Commission for small and medium-sized enterprises (which, however, also
includes a turnover or balance sheet threshold). It is therefore recommended
that, if a number of classes are dropped, the breaks at 49 and 249 employees
should be maintained so that comparable statistics can be prepared for small,
me dium-sized and large businesses. For larg e e conomies, the class
250 employees and over would be too large, so that the break at 999 employees
should also be maintained. The category 0 employee is relevant in several
countries that cover enterprises which only include the entrepreneur.

3.5. Government sector

3.5.1. Coverage
184. The government sector is composed of:

● All departments, offices and other bodies which furnish, but

normally do not sell to the community, those common
services, other than higher education, which cannot otherwise
be conveniently and economically provided, as well as those
that administer the state and the economic and social policy of
the community. (Public enterprises are included in the
business enterprise sector.)
● NPIs controlled and mainly financed by government, but not
administered by the higher education sector.

62 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 3 INSTITUTIONAL CLASSIFICATION

185. According to the SNA definition (UN, 1968; CEC et al., 1994) of
“producers of government services” (with the exception of publicly controlled
higher education institutions), this sector should include all bodies,
departments and establishments of government – central, state or provincial,
district or county, municipal, town or village – that engage in a wide range of
activities, such as: administration; defence and regulation of public order;
health, education, cultural, recreational, and other social services; promotion
of economic growth and welfare; and technological development. The
legislature, the executive, departments, establishments and other bodies of
government should be included, irrespective of their treatment in government
accounts. Government-administered social security funds are also included. It
is immaterial whether they are accounted for in ordinary or extraordinary
budgets, or in extra-budgetary funds.
186. With the exception of those administered by the higher education
sector, all non-market NPIs controlled and financed by government are
included in the government sector, irrespective of the types of institutional
units that mainly benefit from their activities. Control is the ability to
determine the NPI’s general policy or programme by having the right to
appoint the NPI’s management. Such NPIs are mainly financed by block grants
from government, and the amounts of “institutional support” are often
published in government reports or budgets. NPIs mainly financed by
government should be included in the government sector even if the
government control is not clear.
187. Units associated with the higher education sector which mainly serve
the government sector should also be included in the government sector.

3.5.2. The principal sector sub-classification

The classification list
188. The United Nations’ COFOG classification (classification of the purposes
of government) is the standard international classification for use within the
government sector. Unfortunately, it is not considered appropriate for the
classification of R&D activities. No agreement has been reached on the most
appropriate sub-classification for the government sector; therefore, no
recommendation is made. (See Chapter 4, Table 4.1 and Sections 4.4.1
and 4.5.1, for recommendations for functional distribution.)

The statistical unit

189. ISIC Rev. 3, paragraph 51, recommends that when data are combined
with those collected from legal business entities, the statistical unit should be
similar to the legal business entity.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 63

3 INSTITUTIONAL CLASSIFICATION

Criterion for classification

190. In the absence of a recognised classification list, no recommendation
can be made at this time.

3.5.3. Other institutional sub-classifications

191. The following classifications are mainly designed to reveal differences
among countries in the coverage of the government sector, which usually
result from differences in institutional arrangements.

Level of government
192. Statistical units should be classified into three categories, according to
the level of government involved, along with a fourth category for units that
cannot be distributed by level of government.
– Central and federal government units.
– Provincial and state government units.
– Local and municipal government units.
– NPIs controlled and mainly financed by government.

Type of institution
193. When important groups of units are connected both to government
and other sectors (e.g. units administered or controlled by government but
situated at, or otherwise associated with, higher education units; or units
serving industry but financed and controlled by government), it is desirable to
identify them separately when reporting to international organisations. (For
this particular classification, the statistical unit may be an establishment-type
rather than an enterprise-type unit.) Where R&D in public hospitals is
included in this sector, it is also useful to declare it separately. A useful
distinction may also be made between units for which R&D is the principal
economic activity (ISIC Rev. 3, Division 73) and the rest.

3.6. Private non-profit sector

3.6.1. Coverage
194. In line with SNA 93, the coverage of this sector was reduced
substantially in the previous revision of this Manual and now includes:

● Non-market, private non-profit institutions serving households

(i.e. the general public).
● Private individuals or households.

64 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 3 INSTITUTIONAL CLASSIFICATION

195. As a source of funds, this sector covers R&D financed by NPIs serving
households (NPSH). These provide individual or collective services to
households either without charge or at prices that are not economically
significant. Such NPIs may be created by associations of persons to provide
goods, or more often services, primarily for the benefit of members
themselves or for general philanthropic purposes. Their activities may be
financed by regular membership subscriptions or dues or by donations in cash
or in kind from the general public, corporations or government. They include
NPIs such as professional or learned societies, charities, relief or aid agencies,
trades unions, consumers’ associations, etc. By convention, this sector
includes any funds contributed directly to R&D by households.
196. As a sector of performance, PNP includes non-market units controlled
and mainly financed by NPIs serving households, notably professional and
learned societies and charities, other than those providing higher education
services or administered by higher education institutions. However, R&D
foundations managed by NPSH but having more than 50% of their running
costs covered by a block grant from government should be included in the
government sector.
197. By convention, this sector also covers the residual R&D activities of the
general public (households), which plays a very small role in the performance
of R&D. The market activities of unincorporated enterprises owned by
households, i.e. consultants undertaking R&D projects for another unit at an
economically significant price, should be included in the business enterprise
sector in line with National Accounts conventions (unless the project is
undertaken using staff and facilities in another sector, see below). Obtaining
data on such R&D may be difficult because R&D activities of individuals are
not captured in business enterprise R&D surveys. Hence, the PNP sector
should only include R&D undertaken by non-market, unincorporated
enterprises owned by households, i.e. individuals financed by their own
resources or by “uneconomic” grants.
198. Furthermore, where grants and contracts are formally awarded to
individuals who are primarily employed in another sector, such as grants
made directly to a university professor, unless such persons undertake the
R&D concerned entirely on their own time and make no use of their employing
unit’s staff and facilities, they should be included in the R&D statistics of the
employing unit. This also refers to postgraduate students in receipt of
research grants known to the research unit. It therefore follows that this sector
only includes R&D performed by individuals exclusively on their own time and
with their own facilities and at their own expense or supported by an
uneconomic grant.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 65

3 INSTITUTIONAL CLASSIFICATION

199. The following types of PNP organisations should be excluded from this
sector:
– Those mainly rendering services to enterprises.
– Those primarily serving government.
– Those entirely or mainly financed and controlled by government.
– Those offering higher education services or controlled by institutions of
higher education.

3.6.2. The principal sector sub-classification

The classification list
200. Statistical units in the PNP sector are classified into the six major fields
of science and technology suggested in UNESCO’s “Recommendation
Concerning the International Standardisation of Statistics on Science and
Technology” (1978). These fields are:
– Natural sciences.
– Engineering and technology.
– Medical sciences.
– Agricultural sciences.
– Social sciences.
– Humanities.
201. Table 3.2 shows the major fields of science, together with examples of
which sub-fields are included.
202. While the major fields of science and technology are clearly defined,
the level of disaggregation within each component field is left to each country.

The statistical unit

203. According to the SNA, the legal entity is the recommended statistical
unit for this sector. In some cases a smaller statistical unit may be appropriate
(see below).

Criterion for classification

204. The criterion for classification is the major field of science in which
most of the R&D activity is undertaken. When a major private NPI has
significant R&D activity in more than one field of science, an attempt may be
made to split the statistical unit into smaller units and classify them to
relevant major fields of science.

3.6.3. Other institutional sub-classifications

205. The role of this sector in R&D is very small. Therefore, no further
breakdown is proposed.

66 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 3 INSTITUTIONAL CLASSIFICATION

Table 3.2. Fields of science and technology

1. NATURAL SCIENCES
1.1. Mathematics and computer sciences [mathematics and other allied fields: computer sciences and other allied subjects
(software development only; hardware development should be classified in the engineering fields)]
1.2. Physical sciences (astronomy and space sciences, physics, other allied subjects)
1.3. Chemical sciences (chemistry, other allied subjects)
1.4. Earth and related environmental sciences (geology, geophysics, mineralogy, physical geography and other
geosciences, meteorology and other atmospheric sciences including climatic research, oceanography, vulcanology,
palaeoecology, other allied sciences)
1.5. Biological sciences (biology, botany, bacteriology, microbiology, zoology, entomology, genetics, biochemistry,
biophysics, other allied sciences, excluding clinical and veterinary sciences)

2. ENGINEERING AND TECHNOLOGY

2.1. Civil engineering (architecture engineering, building science and engineering, construction engineering, municipal and
structural engineering and other allied subjects)
2.1. Electrical engineering, electronics [electrical engineering, electronics, communication engineering and systems,
computer engineering (hardware only) and other allied subjects]
2.3. Other engineering sciences (such as chemical, aeronautical and space, mechanical, metallurgical and materials
engineering, and their specialised subdivisions; forest products; applied sciences such as geodesy, industrial
chemistry, etc.; the science and technology of food production; specialised technologies of interdisciplinary fields,
e.g. systems analysis, metallurgy, mining, textile technology and other allied subjects)

3. MEDICAL SCIENCES
3.1. Basic medicine (anatomy, cytology, physiology, genetics, pharmacy, pharmacology, toxicology, immunology and
immunohaematology, clinical chemistry, clinical microbiology, pathology)
3.2. Clinical medicine (anaesthesiology, paediatrics, obstetrics and gynaecology, internal medicine, surgery, dentistry,
neurology, psychiatry, radiology, therapeutics, otorhinolaryngology, ophthalmology)
3.3. Health sciences (public health services, social medicine, hygiene, nursing, epidemiology)

4. AGRICULTURAL SCIENCES
4.1. Agriculture, forestry, fisheries and allied sciences (agronomy, animal husbandry, fisheries, forestry, horticulture, other
allied subjects)
4.2. Veterinary medicine

5. SOCIAL SCIENCES
5.1. Psychology
5.2. Economics
5.3. Educational sciences (education and training and other allied subjects)
5.4. Other social sciences [anthropology (social and cultural) and ethnology, demography, geography (human, economic
and social), town and country planning, management, law, linguistics, political sciences, sociology, organisation and
methods, miscellaneous social sciences and interdisciplinary, methodological and historical S&T activities relating to
subjects in this group. Physical anthropology, physical geography and psychophysiology should normally be classified
with the natural sciences]

6. HUMANITIES
6.1. History (history, prehistory and history, together with auxiliary historical disciplines such as archaeology, numismatics,
palaeography, genealogy, etc.)
6.2. Languages and literature (ancient and modem)
6.3. Other humanities [philosophy (including the history of science and technology), arts, history of art, art criticism,
painting, sculpture, musicology, dramatic art excluding artistic “research” of any kind, religion, theology, other fields
and subjects pertaining to the humanities, methodological, historical and other S&T activities relating to the subjects
in this group]

Source: OECD.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 67

3 INSTITUTIONAL CLASSIFICATION

3.7. Higher education sector

3.7.1. Coverage
206. This sector is composed of:

● All universities, colleges of technology and other institutions of

post-secondary education, whatever their source of finance or
legal status.
● It also includes all research institutes, experimental stations
a n d c l i n ic s o p e ra ti n g un de r th e d i re c t c o n tr ol o f o r
adm ini ste re d by or as s ocia te d wit h h igh e r e ducation
institutions.

207. This is not an SNA sector. It has been separately identified by the OECD
(and by UNESCO) because of the important role played by universities and
similar institutions in the performance of R&D.
208. The above definition describes the sector’s general coverage. As it is
not backed by SNA, it is difficult to provide clear guidelines that ensure
internationally comparable reporting of data. Also, because the criteria are
mixed, it is particularly susceptible to variations in interpretation as a result of
national policy concerns and definitions of the sector.
209. The core of the sector in all countries is made up of universities and
colleges of technology. Where treatment varies, it is with respect to other post-
secondary education institutions and above all to several types of institutes
linked to universities and colleges. The main problems are considered below:
– Post-secondary education.
– University hospitals and clinics.
– “Borderline” research institutions.

Post-secondary education
210. The sector includes all establishments whose primary activity is to
provide post-secondary (tertiary level) education regardless of their legal
status. They may be corporations, quasi-corporations belonging to a
government unit, market NPIs or NPIs controlled and mainly financed by
government or by NPISHs. As noted above, the core is made up of universities
and colleges of technology. The number of units in the sector has grown as
new universities and specialised post-secondary educational institutions have
been set up and secondary level units, some of which may supply education
services at both secondary and post-secondary level, have been upgraded. If
such units supply post-secondary education as a primary activity, they are

68 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 3 INSTITUTIONAL CLASSIFICATION

always part of the higher education sector. If their primary activity is the
provision of secondary level education or in-house training they should be
allocated by sector in line with the other general rules (market or non-market
production, sector of control and institutional funding, etc.).

University hospitals and clinics

211. Inclusion of university hospitals and clinics in the higher education
sector is justified both because they are post-secondary educational
institutions (teaching hospitals) and because they are research units
“associated with” higher education institutions (e.g. advanced medical care in
clinics at universities).
212. Academic medical research is traditionally funded from many sources:
the institution’s general “block grant” (GUF); the institution’s “own funds”;
government funds or private funds, directly or indirectly (via a medical
research council, for instance).
213. Where all or nearly all activities in the hospital/medical institution
have a teaching/training component, the entire institution should be included
as part of the higher education sector. If, on the other hand, only a few of the
clinics/departments within a hospital/medical institution have a higher
education component, only these teaching/training clinics/departments
should be classified in the higher education sector. All other non-teaching/
training clinics/departments should, as a general rule, be included in the
appropriate se ctor (corporations, quasi-corporation s belong in g to a
government unit and market NPIs in the business enterprise sector; NPIs
controlled and mainly financed by government in the government sector; NPIs
controlled and mainly financed by NPISHs in the PNP sector). Care must be
taken to avoid double counting of R&D activities between the sectors
concerned.

“Borderline” research institutions

214. Traditionally, universities have been major centres of research, and
w hen countries have wished to expand their R&D in specific fields,
universities have frequently been considered appropriate locations for new
institutes and units. Most such units are principally government-financed and
may even be mission-oriented research units; others are financed by private
non-profit sector funds and, more recently, by the business enterprise sector.
215. A particular case arises when special funds are used to set up and
finance mainly basic research, which is managed by agencies that not only
award grants to universities but also have their “own” research institutes,
which may or may not be situated on university campuses. These may be
regarded as belonging to the higher education sector.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 69

3 INSTITUTIONAL CLASSIFICATION

216. One factor affecting the classification of such research institutions is

the purpose for which the research is being carried out. If the research is
predominantly to serve government’s needs, countries may decide to classify
the institution in the government sector. This is the case of “mission-oriented”
R&D institutes financed from the budget of their sponsoring ministry or
department. Alternatively, if the R&D is basic in nature and adds to the general
body of a country’s knowledge, some member countries may have opted to
classify the institutes in the higher education sector.
217. A higher education institution may have “links” with other research
institutes which are not directly concerned with teaching or which have other
non-R&D functions such as consulting, for example through the mobility of
personnel between the higher education institution and the research institute
concerned or the sharing of facilities between institutes classified in different
sectors. These institutes may be classified according to other criteria, such as
control and finance or service rendered.
218. In some countries, furthermore, borderline institutions may have a
private legal status and carry out contract research for other sectors, or they
may be government-financed research institutions. It is difficult to decide, in
such cases, whether the links between the units are strong enough to justify
including the “external” unit in the higher education sector.
219. “Science parks” located at or near universities and colleges, which host
a range of manufacturing, service and R&D entities, are a somewhat recent
development. For such groupings, it is recommended not to use physical
location and use of common resources as a criterion to classify these units in
the higher education sector. Units controlled and hosted in these parks and
mainly financed by government should be included in the government sector,
those controlled and mainly financed by the private non-profit sector should
be included in the PNP sector, while enterprises and other units serving
enterprises should be classified in the business enterprise sector.
220. Units administered by post-secondary teaching units (including
teaching hospitals), as defined above, which are not primarily market
producers of R&D, should be included in the higher education sector. This also
applies if they are mainly financed from university block grants. If they are
primarily market producers of R&D, they should be included in the business
enterprise sector despite any links with higher education units; this is
particularly relevant for science parks.
221. It is recommended that R&D expenditure and personnel of all
institutes at the borderline with the higher education sector be reported
separately.

70 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 3 INSTITUTIONAL CLASSIFICATION

3.7.2. The principal sector sub-classification

The classification list
222. Statistical units in the higher education sector, like those in the private
non-profit sector, are classified into six major fields of science and technology
as follows:
– Natural sciences.
– Engineering and technology.
– Medical sciences.
– Agricultural sciences.
– Social sciences.
– Humanities.
223. Table 3.2 gives the major science fields, together with examples of
which sub-fields are included.
224. While the major fields of science and technology are clearly defined,
the level of disaggregation within each component is left to each country’s
discretion. In the higher education sector, where detailed administrative
information is available, a detailed field of science classification can be used
as an institutional classification.

The statistical unit

225. Since the enterprise-type unit would almost invariably be involved in
more than one of the six major fields of science and technology, a smaller
statistical unit is necessary. An establishment-type unit is therefore
recommended: the smallest homogeneous unit predominantly involved in
only one of the six fields and for which a complete (or almost complete) set of
factor input data can be obtained. Depending on the size of the institution and
national terminology, the statistical unit could be a research institute, a
“centre”, a department, a faculty, a hospital or a college.

Criterion for classification

226. The statistical unit should be classified in the field of science or
technology which seems to describe most accurately its principal activity as
reflected, for example, by the occupations of most of the unit’s professional
staff. Where R&D data for this sector are estimates made by the surveying
authority, supplementary criteria, such as the institutional location of the
unit, may have to be used. Depending on the size and character of the unit, a
breakdown of the statistical unit into smaller units corresponding to several
relevant major fields of science could be used.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 71

3 INSTITUTIONAL CLASSIFICATION

3.7.3. Other institutional sub-classifications

227. For some countries, it may be helpful, for the purposes of international
comparison, to know the breakdown between public and private universities
and between universities proper and other post-secondary institutions.
228. Statistical units should therefore be classified by the most appropriate
type of main activity:
– Teaching units (e.g. faculty or departments):
❖ Public.
❖ Private.
– Research institutes or centres.
– Clinics, health centres or university hospitals.
– Other units at the borderline of the higher education sector not elsewhere
classified.

3.8. Abroad
3.8.1. Coverage
229. This sector consists of:

● All institutions and individuals located outside the political

borders of a country, except vehicles, ships, aircraft and space
satellites operated by domestic entities and testing grounds
acquired by such entities.
● All international organisations (except business enterprises),
including facilities and operations within the country’s borders.

3.8.2. The principal sector sub-classification

230. The principal sector sub-classifications are essentially designed to
classify all the R&D activities of a performing unit. However, “Abroad” occurs
in R&D surveys only as a source of funds for R&D performed by statistical units
already classified in one of the four national sectors or as a destination
for their extramural R&D expenditures. Thus, as it occurs only as a sub-item
of the R&D resource s of a statistical unit, the choice of a s tandard
sub-classification does not arise.

72 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 3 INSTITUTIONAL CLASSIFICATION

3.8.3. Other institutional sub-classifications

231. The sector may be divided into the four sectors used for domestic R&D,
plus a fifth: international organisations. The recommended classifications are:
– Business enterprise.
– Other national governments.
– Private non-profit.
– Higher education.
– International organisations.
232. When financial flows for R&D between national and foreign business
enterprise sectors are significant, they may usefully be subdivided among:
– Enterprises within the group.
– Other business enterprises.

3.8.4. Geographic area of origin or destination of funds

233. It may also be helpful to break down flows of funds to and from abroad
by geographical area as follows:
– North America: Canada, Mexico, United States.
– European Union.
– Other European OECD countries.
– Asian OECD countries: Japan, Korea.
– Oceanian OECD countries: Australia, New Zealand.
– Other European non-OECD countries.
– Other Asian non-OECD countries.
– South and Central America.
– Other Oceanian non-OECD countries.
– Africa.
234. This categorisation has been chosen to ensure that:
– All countries of the world are included and all continents singled out.
– The OECD zone may be identified separately.
– Major economic blocs (NAFTA and the EU) within the OECD zone are shown
separately.
– The list is inclusive.
235. Other groupings, such as the Nordic countries, EU candidate countries,
transition countries, etc., may also be of interest. In addition, it is important to
identify funding from the EU and from international organisations.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 73

 ISBN 92-64-19903-9
Frascati Manual 2002
Proposed Standard Practice for Surveys on Research
and Experimental Development
© OECD 2002

Chapter 4

Functional Distribution

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 75

4 FUNCTIONAL DISTRIBUTION

4.1. The approach

236. In the functional approach, the nature of the R&D activity of the
performing unit, rather than its principal (economic) activity, is examined.
The R&D resources of the performing unit are distributed to one or more
functional classes on the basis of the characteristics of the R&D itself, usually
examined at the project level but sometimes in even greater detail. The survey
approaches described in this chapter are thus unique to the field of R&D
statistics. Although in theory functional distribution is quite appropriate for
personnel data, it is generally confined to R&D expenditure.
237. The standard nomenclature used in institutional classifications may
also be used for functional distribution (e.g. field of science). However, much
nomenclature is used only for functional distribution (e.g. type of R&D). In
most cases, statistics on R&D distributed by function are already classified by
institution. For example, R&D is almost always classified by sector and sub-
sector prior to functional distribution. In fact, most functional distribution is
not appropriate for all sectors (Table 4.1).

Table 4.1. Utility of functional distributions

Breakdown by Business enterprise Government Private non-profit Higher education

Type of R&D Expenditure Recommended Recommended Recommended Recommended

for current for current for current for current
expenditure expenditure expenditure expenditure
Personnel Unlikely Unlikely Unlikely Unlikely
Product field Expenditure Recommended Unlikely Unlikely Unlikely
for current
expenditure
Personnel Possible Unlikely Unlikely Unlikely
Main field Expenditure Possible Recommended Recommended Recommended
of science Personnel Possible Possible Possible Possible
Socio-economic Expenditure Recommended Recommended Possible Possible
objective for selected
objectives only
Personnel Unlikely Unlikely Unlikely Unlikely

Source: OECD.

76 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 4 FUNCTIONAL DISTRIBUTION

The category “possible” in Table 4.1 means that the breakdown is used in
several countries. The category “unlikely” means that the breakdown is not
used in any country and its feasibility is unknown.

4.2. Type of R&D

4.2.1. Use of distribution by type of R&D
238. Breakdown by type of R&D is currently recommended for use in all four
national sectors of performance. It is usually more easily applied to R&D in the
natural sciences and engineering (NSE) than in the social sciences and
humanities (SSH). For the purposes of international comparison, the breakdown
should be based on current expenditures only. It may be applied at project level,
but some R&D projects may have to be subdivided among activities.

4.2.2. The distribution list

239. Three types of R&D may be distinguished:
– Basic research.
– Applied research.
– Experimental development.

Basic research
240.
Basic research is experimental or theoretical work undertaken
pr im ari ly to ac q uire n ew kn ow l e dg e o f th e un d e rlyi ng
foundations of phenomena and observable facts, without any
particular application or use in view.

241. Basic research analyses properties, structures and relationships with a

view to formulating and testing hypotheses, theories or laws. The reference to
no “particular application in view” in the definition of basic research is crucial,
as the performer may not know about actual applications when doing the
research or responding to survey questionnaires. The results of basic research
are not generally sold but are usually published in scientific journals or
circulated to interested colleagues. Occasionally, basic research may be
“classified” for security reasons.
242. In basic research, scientists have some freedom to set their own goals.
Such research is usually performed in the higher education sector but also to
some extent in the government sector. Basic research can be oriented or
directed towards some broad fields of general interest, with the explicit goal of
a broad range of applications in the future. One example is the public research

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 77

4 FUNCTIONAL DISTRIBUTION

programmes on nanotechnology which several countries have decided on.

Firms in the private sector may also undertake basic research, with a view to
preparing for the next generation of technology. Research on fuel cell
technology is a case in point. Such research is basic according to the above
definition as it does not have a particular use in view. It is defined in the Frascati
Manual as “oriented basic research”.
243. Oriented basic research may be distinguished from pure basic research
as follows:
– Pure basic research is carried out for the advancement of knowledge,
without seeking long-term economic or social benefits or making any effort
to apply the results to practical problems or to transfer the results to sectors
responsible for their application.
– Oriented basic research is carried out with the expectation that it will
produce a broad base of knowledge likely to form the basis of the solution
to recognised or expected, current or future problems or possibilities.
244. The separate identification of oriented basic research may provide
some assistance towards identifying “strategic research”, a broad notion often
referred to in policy making.

Applied research
245.
Applied research is also original investigation undertaken in
order to acquire new knowledge. It is, however, directed primarily
towards a specific practical aim or objective.

246. Applied research is undertaken either to determine possible uses for

the findings of basic research or to determine new methods or ways of
achieving specific and predetermined objectives. It involves considering the
available knowledge and its extension in order to solve particular problems. In
the business enterprise sector, the distinction between basic and applied
research is often marked by the creation of a new project to explore promising
results of a basic research programme.
247. The results of applied research are intended primarily to be valid for a
single or limited number of products, operations, methods or systems.
Applied research gives operational form to ideas. The knowledge or
information derived from it is often patented but may be kept secret.
248. It is recognised that an element of applied research can be described as
strategic research, but the lack of an agreed approach in member countries to
its separate identification prevents making a recommendation.

78 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 4 FUNCTIONAL DISTRIBUTION

Experimental development
249.
Experimental development is systematic work, drawing on
knowledge gained from research and practical experience, that is
directed to producing new materials, products and devices; to
installing new processes, systems and services; or to improving
substantially those already produced or installed.

250. In the social sciences, experimental development may be defined as

the process of translating knowledge gained through research into operational
programmes, including demonstration projects undertaken for testing and
evaluation purposes. The category has little or no meaning for the humanities.

4.2.3. Criteria for distinguishing between types of R&D

251. There are many conceptual and operational problems associated with
these categories. They seem to imply a sequence and a separation which
rarely exist in reality. The three types of R&D may sometimes be carried out in
the same centre by essentially the same staff. Moreover, there may be
movement in both directions. When an R&D project is at the applied research/
experimental development stage, for example, some funds may have to be
spent on additional experimental or theoretical work in order to acquire more
knowledge of the underlying foundations of relevant phenomena before
further progress can be made. Moreover, some research projects may
genuinely straddle categories. For instance, study of the variables affecting the
educational attainment of children drawn from different social and ethnic
groups may involve both basic and applied research.
252. The following examples illustrate general differences between basic
and applied research and experimental development in the natural sciences
and engineering and in the social sciences and humanities.
253. Examples from the natural sciences and engineering:
– The study of a given class of polymerisation reactions under various
conditions, of the yield of products and of their chemical and physical
properties is basic research. The attempt to optimise one of these reactions
with respect to the production of polymers with given physical or
mechanical properties (making it of particular utility) is applied research.
Experimental development then consists of “scaling up” the process which
has been optimised at the laboratory level and investigating and evaluating
possible methods of producing the polymer and perhaps articles to be made
from it.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 79

4 FUNCTIONAL DISTRIBUTION

– The study of a crystal’s absorption of electromagnetic radiation to obtain

information on its electron band structure is basic research. The study of
the absorption of electromagnetic radiation by this material under varying
conditions (for instance temperature, impurities, concentration, etc.) to
obtain given properties of radiation detection (sensitivity, rapidity, etc.) is
applied research. The preparation of a device using this material to obtain
better detectors of radiation than those already existing (in the spectral
range considered) is experimental development.
– The determination of the amino acid sequence of an antibody molecule is
basic research. Investigations undertaken in an effort to distinguish
between antibodies for various diseases is applied research. Experimental
development then consists of devising a method for synthesising the
antibody for a particular disease on the basis of knowledge of its structure
and clinically testing the effectiveness of the synthesised antibody on
patients who have agreed to accept experimental advanced treatment.
254. Examples from the social sciences and humanities:
– Theoretical investigation of the factors determining regional variations in
economic growth is basic research; however, such investigation performed
for the purpose of developing government policy is applied research. The
development of operational models, based upon laws revealed through
research and aimed at modifying regional disparities, is experimental
development.
– Analysis of the environmental determinants of learning ability is basic
research. Analysis of the environmental determinants of learning ability for
the purpose of evaluating education programmes designed to compensate
for environmental handicaps is applied research. The development of
means of determining which educational programme to use for particular
classes of children is experimental development.
– The development of new risk theories is basic research. Investigation of new
types of insurance contracts to cover new market risks is applied research.
Investigation of new types of savings instruments is applied research.
Development of a new method to manage an investment fund is
experimental development.
– The study of a hitherto unknown language to establish its structure and
grammar is basic research. Analysis of regional or other variations in the
use of a language to determine the influence of geographical or social
variables on the development of a language is applied research. No
meaningful examples of experimental development have been found in the
humanities.
255. Table 4.2 gives further examples of the distinctions between the three
types of research in the social sciences.

80 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 4 FUNCTIONAL DISTRIBUTION

Table 4.2. The three types of research in the social sciences

and humanities
Basic research Applied research Experimental development

Study of the causal relations between Study of the economic and social causes Development and testing
economic conditions and social of the drift of agricultural workers from of a programme of financial assistance
development rural districts to towns, for the purpose to prevent rural migration to large cities
of preparing a programme to halt this
development in order to support
agriculture and prevent social conflicts in
industrial areas
Study of the social structure and the Development of a model using the data Development and testing
socio-occupational mobility of a society, obtained in order to foresee future of a programme to stimulate upward
i.e. its composition and changes in consequences of recent trends in social mobility among certain social and ethnic
socio-occupational strata, social classes, mobility groups
etc.
Study of the role of the family in different Study of the role and position Development and testing
civilisations past and present of the family in a specific country of a programme to maintain family
or a specific region at the present time structure in low-income working groups
for the purpose of preparing relevant
social measures
Study of the reading process in adults Study of the reading process for Development and testing of a special
and children, i.e. investigating how the purpose of developing a new method reading programme among immigrant
human visual systems work to acquire of teaching children and adults to read children
information from symbols such as
words, pictures and diagrams
Study of the international factors Study of the specific international factors –
influencing national economic determining the economic development
development of a country in a given period with a view
to formulating an operational model
for modifying government foreign trade
policy
Study of specific aspects of a particular Study of the different aspects –
language (or of several languages of a language for the purpose of devising
compared with each other) such as a new method of teaching that language
syntax, semantics, phonetics, or of translating from or into that
phonology, regional or social variations, language
etc.
Study of the historical development – –
of a language
Study of sources of all kinds – –
(manuscripts, documents, monuments,
works of art, buildings, etc.) in order
to better comprehend historical
phenomena (political, social, cultural
development of a country, biography
of an individual, etc.)

Source: UNESCO (1984b), “Manual for Statistics on Scientific and Technological Activities”.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 81

4 FUNCTIONAL DISTRIBUTION

256. Examples from software development:

– Search for alternative methods of computation, such as quantum
computation and quantum information theory, is basic research.
– Investigation into the application of information processing in new fields or
in new ways (e.g. developing a new programming language, new operating
systems, programme generators, etc.) and investigation into the application
of in formation p roce ss in g to develop tools such as g e og raph ical
information and expert systems are applied research.
– Development of new applications software, substantial improvements to
operating systems and application programmes, etc., are experimental
development.

4.3. Product fields

4.3.1. Use of distribution by product fields
257. At present, the distribution of R&D by product fields is confined to the
business enterprise sector. It could in theory also be applied to other sectors,
but the distribution list suggested in the next section would have to be
modified to account for the different orientation of R&D carried out in non-
commercial institutions.
258. Product field analysis focuses on the actual industrial orientation of
the R&D carried out by units in the business enterprise sector. Distribution of
R&D by product fields will enhance the quality of data by distributing them
more appropriately to the relevant industries, as these are more comparable
internationally and allow for more detailed analysis. For example, R&D
expenditures by product field are better for comparisons with commodity and
production statistics than unmodified, institutionally classified data.
259. In theory, basic research, at least non-oriented basic research, cannot
be assigned to product fields. In practice, the basic research carried out by
firms is generally oriented towards a field of interest to the firm because of its
potential commercial applications. Since the product fields identified in the
next section are very broad, a firm should be able to assign even its basic
research to a field that effectively describes its orientation. It is recommended,
therefore, that all three types of R&D be considered in the product field
distribution. R&D undertaken in the expectation that it will be applied to
processes rather than products should be included in the product field in
which the process will be used.
260. At this time, it is recommended that only current intramural
expenditures should be considered for international comparisons. This is
because a number of member countries are unable to include capital
expenditures, while, on the whole, those that can are also able to report

82 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 4 FUNCTIONAL DISTRIBUTION

current expenditures separately from capital expenditures for the purpose of

international comparisons.

4.3.2. The distribution list

261. The recommended list depends on the reason for the distribution,
i.e. the intended use of the statistics. Trade data are classified by the national
equivalent of the Standard International Trade Classification (SITC) (UN, 1986);
industrial output data are classified by the national equivalent of the
International Standard Industrial Classification (ISIC) (UN, 1990). At present,
comparisons with industrial output data and with trade data are both popular
with analysts. For reasons of symmetry with the institutional classification for
the business enterprise sector, the same distribution list has been adopted
(see Table 3.1).

4.3.3. Criteria for distribution

262. There are two feasible criteria for distributing R&D by product field. In
one case, the distribution should take the nature of the product into
consideration. In the other, the distribution is based on the use of the product
in terms of the enterprise’s economic activities.

Nature of product
263. When applying the “nature of product” criterion, the R&D input is
distributed according to the type of product being developed.
264. The guidelines formerly used by the National Science Foundation to
survey applied research and experimental development in industry are good
examples of operational criteria:
“Costs should be entered in the field or product group in which the
research and development project was carried out, regardless of the
classification of the field of manufacturing in which the results are to
be used. For example, research on an electric component for a farm
machine should be reported as research on electrical machinery. Also,
research on refractory bricks to be used by the steel industry should be
reported as research on stone, clay, glass and concrete products rather
than primary ferrous metals, whether performed in the steel industry
or the stone, clay, glass and concrete industry.”
265. These guidelines should pose few problems for most R&D projects on
product development. R&D on processes may be more difficult to deal with. If
the results of the R&D will clearly be embodied in materials or equipment,
then the guidelines should be applied to those products. If not, then the
process should be allocated to the product it is destined to produce.
Furthermore, for enterprises engaged in broad R&D programmes, quite

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 83

4 FUNCTIONAL DISTRIBUTION

detailed records or consultations with R&D personnel are needed in order to

provide complete estimates.
266. The advantage of this approach is that any enterprise in any industry
carrying out R&D on a given product should select the same product field,
whatever the expected use of the product. Therefore, both inter-firm and,
especially, international data should be comparable. The main disadvantage is
that R&D on products assembled from a wide range of components, such as
aircraft, may be underestimated.

Use of product
267. The “use of product” criterion is applied in order to distribute an
enterprise’s R&D among the economic activities supported by its R&D
programme. The R&D is therefore distributed by industrial activities according
to the final products produced by the enterprise.
268. The R&D of an enterprise active in only one industry would be
assigned to the product field characteristic of that industry, except when R&D
is carried out on a product or process in order to enable the enterprise to enter
a new industry.
269. When an enterprise is active in more than one industry, the use of the
product must be considered. For example, the R&D carried out on very large-
scale integrated circuits (VLSI) could be distributed in several ways:
– For an enterprise active only in the semiconductor industry, this is R&D for
electronic components and accessories.
– For an enterprise active only in the computer industry, this is R&D for office,
computing and accounting machines.
– For an enterprise active in the semiconductor and computer industries, the
use of the VLSI will determine the choice of product field:
❖ If the VLSI is sold separately, the product field should be electronic
components and accessories.
❖ If the VLSI is included in computers sold by the enterprise, the product
field should be office, computing and accounting machinery.
270. In theory, the data derived from a functional analysis by use of product
should be exactly the same as those derived from an institutional breakdown
by industry, if the R&D by enterprises active in more than one industry is
subdivided into several institutional units. In practice, the functional
classification, which applies only to current expenditures, will be more
detailed and should distribute the activities of many firms over several
product fields, as adjustments will only be made in the institutional
classification for the most significant multi-product firms.

84 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 4 FUNCTIONAL DISTRIBUTION

271. The “use of product” approach is intended to provide R&D data as

comparable as possible with other economic statistics at the establishment
level, notably value added. It is therefore particularly useful when dealing with
enterprises active in more than one industry.
272.
It is recommended that current intramural R&D expenditure in
the business enterprise sector should be distributed by product
field for all industry groups. However, if this is not possible for all
industry groups, it is at least recommended for ISIC Division 73.
It is recommended that the product field distribution should be
based on the use of product approach (industry served for
ISIC Division 73). The classification outlined in Table 3.1 should
be used.

4.4. Fields of science and technology

4.4.1. Use of distribution by field of science and technology
273. The fields of science and technology used for functional distribution
differ in three ways from the institutional classification by major field
described in Chapter 3 (see Sections 3.6.2 and 3.7.2). First, the R&D itself is
examined, rather than the main activity of the performing unit. Second, the
resources are usually distributed at project level within each performing unit.
Third, a more detailed list of fields should be used. Such a detailed list is not
agreed; the list in Chapter 3, Table 3.2, is presented as an illustration. However,
countries are encouraged to use their detailed classifications of fields of
science. Work to develop a more detailed international classification of fields
of science for statistical use is to be undertaken. A distribution by fields of
science is most easily applied in the higher education and private non-profit
sectors. The units surveyed in the government sector may also be able to break
down their R&D activities by field of science, but this has rarely been
attempted in the business enterprise sector.
274. This classification is recommended for all R&D carried out by units in
the higher education, government and private non-profit sectors.

4.4.2. The distribution list

275. Unfortunately, no up-to-date, detailed standard international
classification of fields of science and technology, suitable for the functional
distribution of R&D activities, is available. Therefore it is recommended that
the major fields of science and technology described in Table 3.2 should be
adopted as the functional fields of a science classification system.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 85

4 FUNCTIONAL DISTRIBUTION

4.4.3. The criteria for distribution

276. Resources should be allocated to the various fields of science and
technology on the basis of the focus of R&D activities, measured in terms of
expenditure and field in which R&D personnel actually work, usually
at project level. Where appropriate, e.g. in the case of projects with a
multidisciplinary character, a breakdown of resources by several fields of
science and technology should be made.

4.5. Socio-economic objectives

4.5.1. Use of distribution by socio-economic objectives
277. This section deals with the functional analysis of the primary socio-
economic objectives of intramural R&D as reported retrospectively by the
performer. This approach should not be confused with the analysis by socio-
economic objectives of government budget appropriations or outlays for R&D
(GBAORD), which is treated in Chapter 8. (Chapter 8 deals with the objectives
of total intended government R&D expenditure – intramural and extramural –
as reported by the funder, often on the basis of budget data.)
278. Performer-based reporting of the socio-economic objectives of R&D is
most easily applied in the government and private non-profit sectors (or in a
general “institutes” survey), although individual countries have applied it in
the higher education sector and even in the business enterprise sector. It
should be applied to total intramural expenditures in all fields of science.
279. Over half of OECD countries make a detailed breakdown of R&D
expenditures by socio-economic objective in one or more sectors, and some
also use this distribution for R&D personnel data. Others, however, have not
attempted this approach.

4.5.2. Minimum recommended breakdown

280. Although a general recommendation on the utility of detailed analysis
by socio-economic objective cannot be made, it is suggested that member
countries make efforts to collect performer-reported data in all sectors for two
priority objectives:
– Defence.
– Control and care of the environment.

Defence R&D
281. Defence includes all R&D programmes undertaken primarily for
defence reasons, regardless of their content or whether they have secondary
civil applications. Thus, the criterion is not the nature of the product or
subject (or who funds the programme) but the objective. The objective of

86 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 4 FUNCTIONAL DISTRIBUTION

defence R&D is the creation or enhancement of techniques or equipment for

use by national, overseas or multinational armed forces. For example, defence
R&D includes nuclear and space R&D undertaken for defence purposes. It does
not, however, include civil R&D financed by ministries of defence, for instance
on meteorology or telecommunications. It also includes enterprise-financed
R&D for which the main applications are in the defence area.
282. At first sight, the definition of defence R&D according to objective appears
relatively straightforward. However, exactly the same R&D programme could
have either a civil or a defence objective. An example is the Canadian research on
cold-weather clothing intended for military use; because of its potential for civil
applications, this programme could have been, or could become, civil.
283. Where there is pressure to “spin off” defence R&D to civil uses, or vice
versa, the blurring of the objective may become significant. In such cases,
only the entity funding the R&D may be able to define its objective and
thus its classification as either defence or civil R&D (see also Chapter 8,
paragraphs 21-22).
284. The financing of defence R&D is increasingly internationalised and
privatised, and all sources of funds should be included. For countries with major
defence R&D efforts, a breakdown by source of funds can be informative.

Control and care of the environment

285. In recent years, policy makers’ attention has focused on all aspects of
environmental activity, and environmentally related R&D is no exception.

4.5.3. The distribution list

286. The distribution list based on NABS (see Chapter 8, Sections 8.7.3
and 8.7.4) is the same as that suggested for government R&D funding (except
for research financed from general university funds which is not appropriate
for performer-based surveys, see paragraph 288 below).
1. Exploration and exploitation of the Earth.
2. Infrastructure and general planning of land use.
3. Control and care of the environment.
4. Protection and improvement of human health.
5. Production, distribution and rational utilisation of energy.
6. Agricultural production and technology.
7. Industrial production and technology.
8. Social structures and relationships.
9. Exploration and exploitation of space.
10. Non-oriented research.
11. Other civil research.
12. Defence.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 87

4 FUNCTIONAL DISTRIBUTION

4.5.4. The criteria for distribution

287. R&D should be distributed according to the project’s primary objective.
As in the case of product field analysis, there are two approaches to
distribution. One may look at the project content itself (similar to the “nature
of product” approach) or at the end or purpose which the project is intended
to serve (similar to the “use of product” approach). The latter approach may be
the most appropriate for performer-based analysis by socio-economic
objective.
288. When this type of analysis is attempted in the higher education sector,
general university funds (GUF) (see Chapter 6, Section 6.3.3) should be
distributed among objectives and not grouped under “Non-oriented research”
(former “Advancement of research”).

88 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ISBN 92-64-19903-9
Frascati Manual 2002
Proposed Standard Practice for Surveys on Research
and Experimental Development
© OECD 2002

Chapter 5

Measurement of R&D Personnel

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 89

5 MEASUREMENT OF R&D PERSONNEL

5.1. Introduction
289. Personnel data measure the resources going directly to R&D activities.
Expenditure data measure the total cost of carrying out the R&D concerned,
including indirect support (ancillary) activities.
290. The theoretical distinction between R&D and indirect support
(ancillary) activities is discussed in Chapter 2. In practice, it is useful to
introduce additional criteria concerning the location of the activity in the
entity concerned and its relation to the R&D-performing unit, considered as
an establishment-type unit that may differ from the statistical unit.
291. In compiling R&D data, it may be difficult to isolate the R&D activities
of ancillary staff from those of other R&D staff. In theory, however, the
following activities are included in personnel and expenditure data if they are
carried out in the R&D unit:
– Performing the scientific and technical work for a project (setting up and
carrying out experiments or surveys, building prototypes, etc.).
– Planning and managing R&D projects, especially their S&T aspects.
– Preparing the interim and final reports for R&D projects, especially their
R&D aspects.
– Providing internal services for R&D projects, e.g. computing or library and
documentation work.
– Providing support for the administration of the financial and personnel
aspects of R&D projects.
292. The following are service or indirect support (ancillary) activities to be
excluded from the personnel data but to be included in the expenditure data
as overhead:
– Specific services to R&D provided by central computer departments and
libraries.
– The services of central finance and personnel departments.
– Security, cleaning, maintenance, canteens, etc.
293. The activities identified above as indirect support activities should
also be included in overhead expenditures if they are purchased or hired from
outside suppliers (see Table 5.1).

90 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 5 MEASUREMENT OF R&D PERSONNEL

Table 5.1. R&D and indirect support activities

Treatment in R&D Location in the institution carrying out Activities in each
Categories
survey the R&D category

R&D activities In R&D personnel In the R&D- R&D units (formal Direct R&D Carry out
and in R&D labour performing unit R&D) plus other experiments, build
costs units (informal prototypes, etc.
R&D)
Acquisition Drafting, typing and
and treatment of reproducing R&D
specific information reports, in-house
libraries, etc.
Specific R&D Planning and
management managing S&T
aspects of R&D
projects
Specific Bookkeeping,
administrative personnel
support administration

Indirect support Not in R&D Elsewhere in Central finance or Central R&D share
activities personnel or the performing personnel services administration of finance,
in R&D labour institution (firm, On-site consultants personnel and
costs but in “Other agency, university, general operations
current costs” etc.) (or contracted
as overhead out)
S&T-related Direct centralised R&D share
support services support activities of support provided
by computer
departments,
libraries, etc.
Other ancillary Indirect centralised Security, cleaning,
services support services maintenance,
canteen, etc.

Not involved Excluded Outside the Regional and Collection and

in performance performing national authorities, distribution of R&D
institution n.e.c. international funds
agencies, charities,
etc.

Source: OECD.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 91

5 MEASUREMENT OF R&D PERSONNEL

5.2. Coverage and definition of R&D personnel

5.2.1. Initial coverage
294.
All persons employed directly on R&D should be counted, as
well as those providing direct services such as R&D managers,
administrators, and clerical staff.

295. Persons providing an indirect service, such as canteen and security

staff, should be excluded, even though their wages and salaries are included
as an overhead cost when measuring expenditure.
296. When measuring human resources devoted to R&D, notice has to be
taken of the increased use of on-site consultants as well as the outsourcing of
R&D to other units or firms. With the greater use of consultants, human
resources devoted to R&D may be underestimated when it is difficult to
determine whether consultants are on site or part of an outsourcing
arrangement. To remedy this underestimate, it is proposed to request on-site
consultants’ full-time equivalence (FTE) on R&D in R&D surveys and to
highlight the corresponding costs in “Other current costs” in R&D survey
results. In the case of outsourcing, consultant costs clearly fall under
extramural expenditures.

5.2.2. Categories of R&D personnel

297. Two approaches may be used to classify R&D personnel: the most
commonly used is by occupation, the other is by level of formal qualification.
While both are perfectly reasonable and linked to two different UN
classifications – the International Standard Classification of Occupations
(ISCO) (ILO, 1990) and the International Standard Classification of Education
(ISCED) (UNESCO, 1997) – the differences between them lead to problems of
international comparability.
298. Each approach has advantages and disadvantages. Occupation series
reflect the present use of resources and thus are more useful for R&D analysis
more strictly defined. Furthermore, they are probably easier for employers to
provide and allow for comparisons with other employment series of
enterprises and R&D institutes. Qualification series are important for broader
analyses, for example for setting up total personnel databases and for
forecasting needs and supplies of highly qualified S&T personnel; however,
they create problems for international comparisons owing to differences in
the levels and structures of national educational systems. Both occupation

92 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 5 MEASUREMENT OF R&D PERSONNEL

and qualification series are important in the broader context of studying

human resources in science and technology.
299. The Manual therefore contains definitions for both a classification by
occupation and a classification by level of formal qualification.

The approach by occupation is however pre ferable for

inte rn ation al comparison s of th e numbers of personn el
employed in R&D.

5.2.3. Classification by occupation

Introduction
300. The standard international classification used is the International
Standard Classification of Occupations (ISCO). The following definitions of
occupations are especially designed for R&D surveys. However, they can be
linked to broad categories of ISCO-88 (ILO, 1990) as described below.

Researchers
301.
Researchers are professionals engaged in the conception or
creation of new knowledge, products, processes, methods and
systems and also in the management of the projects concerned.

302. Researchers are classified in ISCO-88 Major Group 2, “Professionals”,

and in “Research and Development Department Managers” (ISCO-88, 1237). By
convention, members of the armed forces with similar skills who perform
R&D should also be included.
303. Managers and administrators engaged in the planning and management
of the scientific and technical aspects of a researcher’s work also fall into this
category. Their rank is usually equal or superior to that of persons directly
employed as researchers and they are often former or part-time researchers.
304. Professional titles may vary from institution to institution, from sector
to sector and from country to country.
305. Postgraduate students at the PhD level engaged in R&D should be
considered as researchers. They typically hold basic university degrees (ISCED
level 5A) and perform research while working towards the PhD (ISCED level 6).
Where they are not a separate category (see Chapter 2, Section 2.3.2) and are
treated as technicians as well as researchers, this may cause inconsistencies
in the researcher series.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 93

5 MEASUREMENT OF R&D PERSONNEL

Technicians and equivalent staff

306.
Technicians and equivalent staff are persons whose main tasks
require technical knowledge and experience in one or more fields
of engineering, physical and life sciences or social sciences and
humanities. They participate in R&D by performing scientific and
technical tasks involving the application of concepts and
operational methods, normally under the supervision of
researchers. Equivalent staff perform the corresponding R&D
tasks under the supervision of researchers in the social sciences
and humanities.

307. Technicians and equivalent staff are classified in ISCO-88 Major

Group 3, “Technicians and Associate Professionals”, notably in Sub-major
Groups 31, “Physical and Engineering Science Associate Professionals”, and 32,
“Life Science and Health Associate Professionals”, and in ISCO-88, 3434,
“Statistical, Mathematical and Related Associate Professionals”. Members of the
armed forces who work on similar tasks should also be included.
308. Their tasks include:
– Carrying out bibliographic searches and selecting relevant material from
archives and libraries.
– Preparing computer programmes.
– Carrying out experiments, tests and analyses.
– Preparing materials and equipment for experiments, tests and analyses.
– Recording measurements, making calculations and preparing charts and
graphs.
– Carrying out statistical surveys and interviews.

Other supporting staff

309.
Other s upportin g staff in clude s s killed an d un skille d
craftsmen, secretarial and clerical staff participating in R&D
projects or directly associated with such projects.

310. Other R&D supporting staff are essentially found in ISCO-88 Major
Groups 4, “Clerks”; 6, “Skilled Agricultural and Fishery Workers”; and 8, “Plant
and Machine Operators and Assemblers”.

94 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 5 MEASUREMENT OF R&D PERSONNEL

311. Included under this heading are all managers and administrators
d e alin g ma inly w ith f ina nc ial an d p er son n el ma tte rs an d g e ne ra l
administration, insofar as their activities are a direct service to R&D. They are
mainly found in ISCO-88 Major Group 2, “Professionals”, and Minor Group 343,
“Administrative Associate Professionals” (except 3434).

5.2.4. Classification by level of formal qualification

Introduction
312. ISCED provides the basis for classifying R&D personnel by formal
qualification. Six classes are recommended for the purposes of R&D statistics.
They are defined exclusively by level of education, regardless of the field in
which personnel are qualified.

Holders of university degrees at PhD level (ISCED level 6)

313. Holders of doctorate degrees of university level or equivalent in all
fields (ISCED level 6). This category includes holders of degrees earned at
universities proper and also at specialised institutes of university status.

Holders of basic university degrees below the PhD level (ISCED level 5A)
314. Holders of tertiary-level degrees below the PhD level in all fields (ISCED
level 5A). This category includes holders of degrees earned at universities
proper and also at specialised institutes of university status.

Holders of other tertiary level diplomas (ISCED level 5B)

315. Holders of other post-secondary tertiary (ISCED level 5B) diplomas in
all fields. Subject matter is typically specialised, presented at a level requiring
the equivalent of full secondary level education to master it. It provides a more
practically oriented/occupation-specific education than programmes at ISCED
levels 5A and 6.

Holders of other post-secondary non-tertiary diplomas (ISCED level 4)

316. Holders of other post-secondary non-tertiary (ISCED level 4) diplomas
in all fields. This class includes holders of degrees preparing students for
studies at level 5, who although having completed ISCED level 3, did not follow
a curriculum which would allow entry to level 5, i.e. pre-degree foundation
courses or short vocational programmes.

Holders of diplomas of secondary education (ISCED level 3)

317. Holders of diplomas at the secondary level, upper stage (ISCED level 3).
This class includes not only all ISCED level 3 diplomas obtained in the

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 95

5 MEASUREMENT OF R&D PERSONNEL

Table 5.2. Standard key for ISCED levels and classes

of the Frascati Manual for R&D personnel by formal qualifications
ISCED-97 categories General coverage OECD personnel categories

6. Second stage of tertiary

education – leading Holders of university
to an advanced research degrees at PhD level
qualification

5A. Theoretically based tertiary

Holders of basic university
programmes to qualify for
degrees below the PhD
5. First stage of tertiary education entry to advanced research
Post-secondary level
– not leading to an advanced programmes
research qualification 5B. Practically oriented
or occupation-specific Holders of other tertiary
programmes degrees

Holders of other
4. Post-secondary, non-tertiary
post-secondary
education
non-tertiary diplomas

Holders of secondary
3. Upper secondary education
education diplomas
Secondary
2. Lower secondary or second
stage of basic education

1. Primary education or first stage Other qualifications

Primary
of basic education

0. Pre-primary education Pre-primary

Source: OECD.

secondary school system but also equivalent level 3 vocational diplomas

obtained from other types of educational establishments.

Other qualifications
318. This includes all those with secondary diplomas at less than ISCED
level 3 or with incomplete secondary qualifications or education not falling
under any of the other four classes.

5.2.5. Treatment of postgraduate students

319. In countries where postgraduates are not a recognised category of S&T
personnel, they are probably included in part-time teaching staff. This means
that as part of the overall calculation of higher education R&D personnel and
expenditures – by either survey or coefficients – their R&D full-time equivalent

96 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 5 MEASUREMENT OF R&D PERSONNEL

levels, their R&D costs and their sources of R&D funds are measured as for
staff employed by the higher education institution.
320. The difficulty of establishing the borderline between the R&D and
education and training activities of postgraduates (and of their teachers) in
countries where they are a recognised group are discussed in general terms in
Chapter 2 (Section 2.3.2).
321. The aim here is to present guidelines for the categories of
postgraduate students that it is both theoretically sound and practically
possible to include in R&D personnel (and hence expenditure) series.
322. As noted in Chapter 2, postgraduate students are often attached to or
directly employed by the establishment concerned and have contracts, or are
bound by similar engagements, which oblige them to do some teaching at
lower levels or to perform other activities, such as specialised medical care,
while allowing them to continue their studies and do research.
323. They can be identified according to the level of their studies. They have
completed first-stage university education (ISCED level 5A) and are studying
at the PhD level (ISCED level 6). ISCED level 6 programmes are described as
follows:
“Tertiary programmes which lead to the award of an advanced
research qualification. The programmes are therefore devoted to
advanced study and original research and are not based on course
work only.
“Classification criteria
Main criterion
It typically requires the submission of a thesis or dissertation of
publishable quality which is the product of original research and
represents a significant contribution to knowledge.
Subsidiary criterion
It prepares graduates for faculty posts in institutions offering
ISCED 5A programmes, as well as research posts in government,
industry, etc.”
324. All postgraduate students working on R&D and receiving funding for
this purpose (in the form of a salary from the university or a scholarship or
another sort of funding) should in principle be included in R&D personnel
headcounts. However, it may be necessary, for practical reasons, to reduce
coverage to those students for whom the corresponding R&D expenditures
and full-time equivalence can be estimated.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 97

5 MEASUREMENT OF R&D PERSONNEL

5.3. Measurement and data collection

5.3.1. Introduction
325. The measurement of personnel employed on R&D involves three
exercises:
– Measuring their number in headcounts.
– Measuring their R&D activities in full-time equivalence (person-years).
– Measuring their characteristics.

5.3.2. Headcount data

Reasons for the approach
326. Data on the total number of persons who are mainly or partially
employed on R&D allow links to be made with other data series, for example
education or employment data or the results of population censuses. This is
particularly important when examining the role of R&D employment in total
stocks and flows of scientific and technical personnel.
327. Headcount data are also the most appropriate measure for collecting
additional information about R&D personnel, such as age, gender or national
origin. Such data are needed to conduct analytical studies and implement
recruitment or other S&T policies aimed at reducing gender imbalances,
shortages of personnel or the effects of ageing, “brain drain”, etc. There is an
increasing demand from S&T policy makers for such data.
328. The OECD Manual on the Measurement of Human Resources devoted to S&T
– Canberra Manual (OECD/Eurostat, 1995) presents a set of guidelines aimed at
measuring the stocks and flows of scientific and technical manpower.
Researchers and technicians represent an important subset of human
resources devoted to S&T (HRST), and experience has shown that R&D surveys
are the most appropriate instrument for collecting headcount data. Population
c e n s us e s , l abour fo rc e s ur veys or pop ul at i on reg is t e rs a re use f ul
complementary data sources but cannot be used systematically to obtain R&D
personnel data.

Possible approaches and options

329. Various options are available for reporting headcount numbers:
– Number of persons engaged in R&D at a given date (e.g. end of period).
– Average number of persons engaged in R&D during the (calendar) year.
– Total number of persons engaged in R&D during the (calendar) year.
330. Insofar as possible, the approach adopted for measuring headcount
data for R&D personnel should be similar to that used for collecting other
statistical headcount series (employment, education) with which the R&D
series are likely to be compared.

98 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 5 MEASUREMENT OF R&D PERSONNEL

5.3.3. Full-time equivalence (FTE) data

Reasons for the approach
331. While data series measuring the number of R&D staff, and notably
researchers, have many important uses, they are not a substitute for a series
based on the number of full-time equivalent staff. The latter is a true measure
of the volume of R&D and must be maintained by all member countries for
international comparisons.
332. R&D may be the primary function of some persons (e.g. workers in an
R&D laboratory) or it may be a secondary function (e.g. members of a design
and testing establishment). It may also be a significant part-time activity
(e.g. university teachers or postgraduate students). To count only persons
whose primary function is R&D would result in an underestimate of the effort
devoted to R&D; to do a headcount of everyone spending some time on R&D
would lead to an overestimate. The number of persons engaged in R&D must,
therefore, also be expressed in full-time equivalents on R&D activities.

Measurement in person-years
333. One FTE may be thought of as one person-year. Thus, a person who
normally spends 30% of his/her time on R&D and the rest on other activities
(such as teaching, university administration and student counselling) should
be considered as 0.3 FTE. Similarly, if a full-time R&D worker is employed at an
R&D unit for only six months, this results in an FTE of 0.5. Since the normal
working day (period) may differ from sector to sector and even from
institution to institution, it is not meaningful to express FTE in person-hours.
334. Personnel should be measured as the number of person-years on R&D
over the same period as the expenditure series.

FTE on a fixed date

335. In some cases, it may be more practical to survey the FTE of R&D
personnel as of a specific date. If, however, there are significant seasonal
variations in R&D employment (e.g. temporary staff hired by governments at
the end of the university teaching year), allowance should be made for these
variations in order to allow for comparison with data based on FTE during a
period. Where the fixed-date approach is used and data are collected annually
for the first or last day of the expenditure period, it is recommended that
two-year moving averages should be used for comparisons with R&D
expenditure data.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 99

5 MEASUREMENT OF R&D PERSONNEL

Diversity of methods and the need for disclosure of method used

336. A number of restrictions apply to the actual measurement of FTE. It is
therefore impossible to avoid differences in the methodology used for
different countries and sectors. The most precise method, which is applied in
the higher education sector, involves carrying out time-use surveys for each
individual researcher. However, more approximate methods are often used in
practice. One method often used consists of counting the number of positions
for each category of personnel, then multiplying by appropriate R&D
coefficients. In some cases, the R&D coefficients used are founded on survey
data of some sort, while in others they are simply based on assumptions made
by those who compile the statistics.
337. To improve international comparability regardless of the
measurement methods used, the details of the methods employed should be
made public. In particular, when R&D coefficients are used, information such
as the value of coefficients, how they were obtained and how they are used in
FTE calculations should be reported with the data, notably when reporting to
international bodies (see Chapter 7, Section 7.6).

Specific problems in the higher education sector

338. The method used to measure R&D personnel should cover all
categories of personnel defined as directly contributing to R&D activities in
the sector, i.e. those actively involved in R&D and those supporting it.
339. To obtain appropriate data on R&D personnel in the higher education
sector, it may be necessary to carry out time-use surveys or studies. Such
surveys can be a source of valuable data even if they are only carried out once
every five or ten years. Annex 2 gives more details regarding time-use surveys.
340. There are two interrelated problems for measurement of R&D
personnel:
– Definition of the working time.
– Calculation of full-time equivalence.

• Definition of working time

341. The one aspect of an academic teacher’s/researcher’s workload that is

usually well-defined (although not necessarily internationally comparable) is
the number of his/her teaching hours in the academic year. Absolute working
time varies according to a number of factors, such as:
– Number of teaching hours per week.
– Demands made by examinations and student supervision on teachers’
time.
– Administrative duties, which vary according to the time of year.

100 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 5 MEASUREMENT OF R&D PERSONNEL

– Nature of R&D activities and deadlines imposed for publication and/or

presentation of results.
– Student vacation periods.
342. The working pattern of the staff therefore is very flexible, as time-use
studies have shown. It has been found that much of their professional activity
– notably R&D – is carried out outside “normal working hours” and frequently
outside the higher education institution itself.

• Calculation of full-time equivalence

343. Much attention has been devoted to defining “normal” working time,
particularly since respondents in time-use surveys frequently report much
longer working time than most similar categories of civil servants. Calculation
of full-time equivalent R&D personnel must be based on total working time.
Accordingly, no one person can represent more than one FTE in any year and
hence cannot perform more than one FTE on R&D.
344. In practice, however, it may not always be possible to respect this
principle. Some researchers, for example, may have activities in several R&D
units. This is increasingly the case of academics who also work for
enterprises. In such cases, for each individual, it may be possible to reduce the
FTE to one.
345. In carrying out surveys, the definition of R&D and of what it includes,
i.e. “normal time” and “overtime”, are very important if the respondent is to
report his/her volume of R&D accurately. The method used for the time-use
survey will have a bearing on the accuracy of FTE calculations (see Annex 2). If
the survey is based on the distribution of working hours during a specific
week, it is relatively easy to take into account R&D done outside “normal office
hours”. If the respondent must evaluate the time spent on R&D during the
whole year, it is more difficult to give correct weight to R&D (as well as to other
work-related activities) done outside “normal” hours. Also, the time of year at
which a time-use survey is carried out may have a bearing on the calculation
of the full-time equivalence.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 101

5 MEASUREMENT OF R&D PERSONNEL

5.3.4. Recommended national aggregates and variables

346.
The two recommended aggregates are for:
● The number of personnel employed in R&D measured in head
counts.
● Total FTE spent in the performance of R&D on national
territory for a given 12-month period.

These should be broken down by sector and by occupation and/or formal

qualification, as shown in Tables 5.3a and 5.3b. In case only one classification
can be provided, priority should be given to the distribution by occupation.
The other institutional classifications (and sometimes the functional
distributions) are applied within this framework.

Table 5.3a. Total national R&D personnel by sector and by occupation

Sector
Occupation Business Private Higher
Government Total
enterprise non-profit education

Researchers
Technicians and equivalent staff
Other supporting staff

Total

Source: OECD.

Table 5.3b. Total national R&D personnel by sector and by level

of qualification
Sector
Qualification Business Private Higher
Government Total
enterprise non-profit education

Holders of:
University degrees
PhD (ISCED 6)
Other (ISCED 5A)
Other tertiary diplomas (ISCED 5B)
Other post-secondary non-tertiary diplomas (ISCED 4)
Secondary diplomas (ISCED 3)
Other qualifications

Total

Source: OECD.

102 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 5 MEASUREMENT OF R&D PERSONNEL

347.
In order to understand more about the R&D labour force and
how it fits in the wider pattern of total scientific and technical
personnel, it is recommended to collect headcount data on
res earche rs and, if possible, on other categ ories of R&D
personnel, broken down by:
● Sex.
● Age.

348. To report data by age, a breakdown into six categories is

recommended:
– Under 25 years.
– 25-34 years.
– 35-44 years.
– 45-54 years.
– 55-64 years.
– 65 years and more.
The above categories are in line with the United Nations Provisional Guidelines on
Standard International Age Classifications (UN, 1982).
349. Other variables are also worth examining, such as salary levels and
national origin. The collection of such data, however, may require conducting
surveys of individuals, which is very resource-intensive. It is therefore useful
to look at other administrative sources of data, such as population registers,
social security registers, etc.
350. Different criteria are used to identify national origin: nationality,
citizenship or country of birth. Others may also be of interest, such as country
of previous residence, previous occupation or country of study at the highest
level. All have advantages and disadvantages and provide different types of
information. The combination of at least two of these criteria will give more
information. However, collection of such data for R&D personnel is still at a
preliminary stage.
351. Finally, it may be useful to collect headcount data on the educational
background of R&D personnel, i.e. field of highest qualification. Fields of study
are defined in ISCED-97 and may be related to the fields of science and
technology presented in Chapter 3, Table 3.2.

5.3.5. Cross-classified data by occupation and qualification

352. Approaches by occupation and qualification have both strengths and
weaknesses when used to classify R&D personnel. However, since each is

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 103

5 MEASUREMENT OF R&D PERSONNEL

associated with a useful body of related statistics (employment by occupation,

educational statistics by qualification), it is desirable to classify R&D
personnel by both occupation and qualification. It is recommended,
furthermore, that data should be collected – perhaps every five years – for
cross-classification between occupation and qualification on a headcount
basis, as shown in Table 5.4.

Table 5.4. R&D personnel classified by occupation

and by formal qualification
Headcount

Occupation
Qualification Technicians and Other supporting
Researchers Total
equivalent staff staff

Holders of:
University degrees
PhDs (ISCED 6)
Others (ISCED 5A)
Other tertiary diplomas (ISCED 5B)
Other post-secondary non-tertiary diplomas (ISCED 4)
Secondary diplomas (ISCED 3)
Other qualifications

Total

Source: OECD.

353. The correspondence between researchers and university graduates –

researchers are generally expected to have university-level diplomas – does
not always hold. Certain researchers have lower qualifications supplemented
by on-the-job experience. It is also increasingly common to find university
g raduates with s cience and eng ineering (NSE) degre es employed as
technicians. The correspondence is even more tenuous for the other
occupational categories. For example, other supporting staff may hold
diplomas at all levels (e.g. financial directors with university degrees in
accountancy, senior secretaries with ISCED level 5 diplomas, etc.). A cross-
classification such as the one suggested in Table 5.4 is useful for attempts to
understand another country’s R&D personnel statistics, to evaluate the
international comparability of these statistics, or, indeed, for discussing
trends in one’s own country’s R&D labour force. Furthermore, it helps to
identify the share of R&D personnel that is a subset of HRST, in particular the
share referred to as “core” in the Canberra Manual, i.e. researchers and
technicians who have completed tertiary education.

104 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 5 MEASUREMENT OF R&D PERSONNEL

354. In addition, it would be desirable to have a measure of all high-level

personnel working on R&D. The continued use of occupation and qualification
as classifications has prevented defining a single measure of this category of
personnel. Table 5.4 would therefore also provide a good basis for identifying
proxy categories of high-level personnel.

5.3.6. Regional data

355. A regional breakdown of total R&D personnel and of researchers is also
recommended for both headcounts and full-time equivalents. For EU member
states, the regional levels are given by the Nomenclature of Territorial Units
for Statistics (NUTS) classification. For other OECD member countries, regional
distribution has to be determined according to national needs. In federal
countries it might be at state level. Further details on the methods to be used
for compiling regional R&D data are found in Annex 5.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 105

 ISBN 92-64-19903-9
Frascati Manual 2002
Proposed Standard Practice for Surveys on Research
and Experimental Development
© OECD 2002

Chapter 6

Measurement of Expenditures Devoted to R&D

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 107

6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

6.1. Introduction
356. A statistical unit may have expenditures on R&D either within the unit
(intramural) or outside it (extramural). The full procedure for measuring these
expenditures is as follows:
– Identify the intramural expenditure on R&D performed by each statistical
unit (see Section 6.2).
– Identify the sources of funds for these intramural R&D expenditures as
reported by the performer (see Section 6.3).
– Identify the extramural R&D expenditures of each statistical unit (see
Section 6.4).
– Aggregate the data by sectors of performance and sources of funds to derive
significant national totals. Other classifications and distributions are then
compiled within this framework (see Section 6.7).
357. The first two steps are essential and generally suffice for undertaking
the fourth. R&D expenditure data should be compiled on the basis of
pe rforme rs’ reports of intramural expe nditures. As s upple men tary
information, the collection of extramural expenditures is desirable.

6.2. Intramural expenditures

6.2.1. Definition
358.
Intramural expenditures are all expenditures for R&D
performed within a statistical unit or sector of the economy
during a specific period, whatever the source of funds.

359. Expenditures made outside the statistical unit or sector but in support
of intramural R&D (e.g. purchase of supplies for R&D) are included. Both
current and capital expenditures are included.

6.2.2. Current costs

360. Current costs are composed of labour costs and other current costs
(see also Section 6.2.3).

108 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

Labour costs of R&D personnel

361. These comprise annual wages and salaries and all associated costs or
fringe benefits, such as bonus payments, holiday pay, contributions to pension
funds and other social security payments, payroll taxes, etc. The labour costs
of persons providing indirect services which are not included in the personnel
data (such as security and maintenance personnel or the staff of central
libraries, computer departments or head offices) should be excluded and
included in other current costs.
362. Labour costs are often the largest component of current costs.
Countries may find it useful to collect or otherwise secure labour costs by
type of personnel (e.g. researchers, technicians and equivalent staff, other
supporting staff, etc.). These extra classifications will be particularly helpful
for constructing cost indices for R&D expenditures.
363. Calculation of the salary element for postgraduate students at the PhD
level may sometimes cause a problem. Only those students who are on the
payroll of universities or R&D units (e.g. as research assistants) and/or receive
external funds for R&D (such as research scholarships) should be included in
the statistics. Sometimes, they receive less money for their work than
they would at “market value”. Only actual “salaries”/stipends and similar
expenditures associated with such students should be reported in the R&D
statistics. No inflated values should be derived.

Other current costs

364. These comprise non-capital purchases of materials, supplies and
equipment to support R&D performed by the statistical unit in a given year.
Examples are: water and fuel (including gas and electricity); books, journals,
reference materials, subscriptions to libraries, scientific societies, etc.;
imputed or actual cost of small prototypes or models made outside the
research organisation; materials for laboratories (chemicals, animals, etc.).
Costs for on-site consultants should be included in other current costs and
identified separately if possible. (See Chapter 5, Section 5.2.1, for their
treatment in personnel data). Administrative and other overhead costs
(e.g. office, post and telecommunications, insurance) should also be included,
prorated if necessary to allow for non-R&D activities within the same
statistical unit. All costs for indirect services should be included here, whether
carried out within the organisation concerned or hired or purchased from
outside suppliers. Examples of such services are: security; storage; use, repair
and maintenance of buildings and equipment; computer services; and
printing of R&D reports. Interest charges should be excluded.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 109

6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

Indirectly paid current costs

365. R&D activities may incur costs which are often not paid by the sector
but borne by institutions classified in other sectors of the economy, usually
the government sector. Two examples are discussed below.

• Rents for research facilities

366. In many countries, responsibility for “housing” public institutions

(including universities) lies with a central agency which is most likely to be
included in the government sector in R&D surveys and whose accounts would
not reflect the functional breakdown between R&D and other activities. This
may concern both the administration of ongoing accommodation and
temporary arrangements for premises and equipment and is particularly
relevant for the higher education sector.
367. In some cases, such facilities are available to institutions free of charge
or are not accounted for in the institutions’ books. To obtain a realistic cost of
R&D, all fees/rents, etc., associated with R&D should be included in
expenditure data. Where the fee or rent is charged to a unit within a sector,
this is easily done. If, however, there is no such charge, it may still be desirable,
for reasons of international comparability, to include a notional amount that
represents an actual payment known to have been made from one agency to
another in a different sector. This might serve as an estimated “market value”,
to be included in other current costs. Care must be taken to avoid “double
counting” of costs between the suppliers and the recipients of these services.
368. If actual payments are made (even if not necessarily revealed by the
R&D surveys), the national authorities should make an adjustment in their
data series – for instance, to account for the estimated market value of the
facilities concerned. This should be included under other current costs in the
receiving sector and should be subtracted, as appropriate, from the accounts
of the donating sectors concerned.

• Social security costs and pensions for R&D personnel

369. Labour costs of R&D personnel “comprise annual wages and salaries
and all associated costs or fringe benefits such as bonus payments, holiday
pay, contributions to pension funds and other social security payments,
payroll taxes, etc.” (see paragraph 361).
370. Where there is an actual provision for social security and/or pensions for
R&D personnel, these amounts should be included in R&D labour costs. They
need not necessarily be visible in the bookkeeping accounts of cost to the sector
concerned; they may often involve transactions within or between sectors. Even
when no transactions are involved, an attempt should be made to estimate these
costs. Care should be taken to avoid double counting of such expenditure.

110 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

Value added tax (VAT)

371. Data on R&D expenditure, on both a provider and a funder basis, should
be at factor cost. This means excluding VAT and similar sales taxes from the
measured cost of the R&D and specifically of R&D financed by government. Not
only will this help to make valid international comparisons, it will also help
countries in their internal analyses, for example when looking at the
opportunity cost of funds devoted to R&D or when deriving ratios using national
income and government expenditure statistics, which generally exclude VAT.
372. For the business enterprise sector, this should present very few
problems since separate recording of VAT input costs is part of standard
accounting procedures and is reclaimable if offset against any VAT charged on
outputs. In the case of the government sector, VAT on input costs is generally
reclaimable and therefore separately identifiable.
373. More difficulties may arise in the higher education and private non-
profit sectors where VAT included in goods and services purchased as part of
an R&D project may not be reclaimable; it will therefore be regarded by the
respondents as a legitimate part of their expenditures. Countries should make
every effort to exclude VAT from expenditure figures for these sectors, making
an adjustment centrally if necessary. It is recommended that figures
forwarded to the OECD should be exclusive of VAT.

6.2.3. Capital expenditures

374.
Capital expenditures are the annual gross expenditures on
fixed assets used in the R&D programmes of statistical units.
They should be reported in full for the period when they took
place and should not be registered as an element of depreciation.

375. All depreciation provisions for building, plant and equipment, whether
real or imputed, should be excluded from the measurement of intramural R&D
expenditures. This approach is proposed for two reasons:
– If depreciation (an allowance to finance the replacement of existing assets)
is included in current costs, the addition of capital expenditures would
result in double counting.
– In the government sector, no provision is normally made for depreciation of
fixed assets. Consequently, even within a country, comparisons between
sectors cannot be made unless depreciation provisions are excluded, and
aggregates for a national series cannot be compiled unless the sector totals
are put on a comparable basis.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 111

6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

376. Capital expenditures are composed of expenditures on:

– Land and buildings.
– Instruments and equipment.
– Computer software.

Land and buildings

377. This comprises land acquired for R&D (e.g. testing grounds, sites for
laboratories and pilot plants) and buildings constructed or purchased,
including major improvements, modifications and repairs.
378. The R&D share of the expenditures for new buildings is often difficult to
quantify, and many countries ignore this element of R&D expenditure (in the
higher education sector) or at best estimate it on the basis of scheduled use (see
the section on identifying the R&D content of capital expenditures below).
379. Purchase of new research equipment is often included in the cost of
new buildings and is not separately identifiable. This may result, in some
years, in underestimation of the “instruments and equipment” component in
total capital R&D expenditures.
380. Countries should maintain a consistent methodology with regard to
these costs.

Instruments and equipment

381. This covers major instruments and equipment acquired for use in the
performance of R&D including embodied software.

Computer software
382. This includes acquisition of separately identifiable computer software
for use in the performance of R&D, including programme descriptions and
supporting materials for both systems and applications software. Annual
licensing fees for the use of acquired computer software are also included.
383. In R&D surveys, however, software for own account produced as part
of R&D is included in the relevant cost category: labour costs or other current
costs.

Conventions for distinguishing between current and capital items

384. In measuring actual capital expenditure, small tools and instruments
and minor improvements to existing buildings will normally be excluded, as
in most accounting systems these items are usually carried on current cost
accounts. The boundary between “minor” and “major” items varies slightly
among countries according to their taxation practices and among different
firms and organisations in the same country according to their accounting

112 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

practices. These differences are rarely significant, and it is neither necessary

nor practical to insist on any rigid standard. Thus, national conventions will
govern allocations to current or to capital expenditures. Nevertheless, in those
countries where expenditures on very expensive prototypes (e.g. aircraft) or
equipment with a limited life (e.g. launching rockets) are considered current
costs, such conventions should always be made explicit.

Identifying the R&D content of capital expenditures

385. Occasionally, the R&D term of a fixed asset may be known at the time
of acquisition. In this case, the appropriate portion of the expenditure for the
acquisition of the asset should be attributed to R&D capital expenditures.
Similarly, when the R&D term of the asset is not known and a fixed asset will
be used for more than one activity and neither the R&D nor any of the non-
R&D activities predominates (e.g. computers and associated facilities;
laboratories used for R&D, testing, and quality control), the costs should be
prorated between R&D and other activities. This proportion could be based on
numbers of R&D personnel using the facility, compared to total personnel, or
on administrative calculations already made (e.g. the R&D budget may be
charged a certain portion of the capital cost; a certain proportion of time or
floor space may be assigned to R&D).

Sale of R&D capital goods

386. The sale/transfer of fixed assets originally acquired for R&D creates a
problem. Their disposal could be considered as a disinvestment in R&D.
However, no adjustment should be made to recorded capital expenditures. The
statistical unit’s capital R&D expenditures should not be reduced accordingly,
either currently or retrospectively (for the years in which the capital costs were
recorded). Current revisions can cause anomalies such as negative intramural
R&D expenditures. Retrospective revisions are difficult and confusing.

Libraries
387. Although payments for current purchases of books, periodicals and
annuals should be assigned to other current costs, expenditure for the
purchase of complete libraries, large collections of books, periodicals,
specimens, etc., should be included in the data under expenditure on major
equipment, especially if made when equipping a new institution (see
UNESCO, 1984b, Section 3.2.1).
388. Each country should adopt the UNESCO approach when reporting data
to the OECD. If this is not possible, a consistent methodology should be
maintained with regard to the classification of the above costs, so that it is
possible to observe changes in the pattern of such expenditure.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 113

6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

6.3. Sources of funds

6.3.1. Methods of measurement
389. R&D is an activity involving significant transfers of resources between
units, organisations and sectors. Every effort should be made to trace the flow
of R&D funds. These transfers may be measured in two ways.
390. One is performer-based reporting of the sums which one unit,
organisation or sector has received or will receive from another unit,
organisation or sector for the performance of intramural R&D during a specific
period. Funds received for R&D performed during earlier periods or for R&D
not yet started should be excluded from the sources of funds reported for the
specific period.
391. The second is source-based reporting of extramural expenditures
which are the sums a unit, organisation or sector reports having paid or
committed itself to pay to another unit, organisation or sector for the
performance of R&D during a specific period.
392. The first of these approaches is strongly recommended.

6.3.2. Criteria for identifying flows of R&D funds

393. For such a flow of funds to be correctly identified, two criteria must be
fulfilled:
– There must be a direct transfer of resources.
– The transfer must be both intended and used for the performance of R&D.

Direct transfer
394. Such transfers may take the form of contracts, grants or donations and
may take the form of money or other resources (e.g. staff or equipment lent to
the performer). When there is a significant non-monetary transfer, the current
value must be estimated, as all transfers must be expressed in financial terms.
395. Resources may be transferred in a number of ways, not all of which
may be considered direct.
396. Contracts or grants paid for the performance of current or future R&D
are clearly identifiable as a transfer of funds. Transfer of funds from the
government to other sectors is particularly important to the users of R&D
data.
397. Two categories of such government funds may be identified:
– Those that are specifically for the procurement of R&D, i.e. the results of the
R&D belong to the recipient of the output or product of the R&D, which is
not necessarily the funder of the R&D.

114 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

– Those that are provided to the performers of R&D in the form of grants or
other financial incentives, with the results of the R&D becoming the
property of the R&D performers.
398. It is recommended that, where possible, both categories of transfer of
government R&D funds should be identified in the R&D data of the business
enterprise sector. If possible, a similar breakdown should be made for
government funds to the higher education sector.
399. In theory, when a government allows a firm or university to use, free of
charge, facilities such as a wind tunnel, observatory or launching site while
carrying out R&D, the value of the service (an imputed rental) should be
identified as a transfer. In practice, the beneficiary will not normally be able to
make such an estimate, and the donor may also not be able to do so.
400. In some cases, a firm’s R&D project may be financed by loans from a
financial institution, an affiliated company or government. Loans that are to
be repaid are not to be considered transfers; by convention, loans that may be
forgiven are to be considered transfers.
401. Other government incentives for R&D in the business enterprise sector
include the remission of income taxes for industrial R&D, the payment by a
government, on demand and after audit, of a certain portion of some or all of
a firm’s R&D expenditures, bonuses added to R&D contracts to encourage a
firm’s own R&D, remission of taxes and tariffs on R&D equipment and the
reimbursement of part of a firm’s costs if it hires more R&D staff. For the
present, even when these transfers can be separately identified, they should
not be counted as direct support for R&D. The statistical units should
therefore report gross expenditures as incurred, even when their actual costs
may be reduced because of remissions, rebates or post-performance grants.

Transfers both intended and used for R&D

402. In many R&D transfers this can be taken for granted. There are
instances, however, when some clarification may be required (particularly if
there is a discrepancy between the performer’s and the funder’s report):
– In one case, a unit gives funds to another in return for equipment or
services needed for its own R&D. If the provision of this equipment or these
services does not require the second unit to carry out R&D, it cannot report
that it performed R&D funded by the first unit. For example, a government
laboratory buys standard equipment or uses an outside computer to
perform calculations required for an R&D project. The equipment supplier
or the computer service firm carries out no R&D itself and would report no
R&D funded by the government. For R&D statistics, these expenditures
should be considered by the government laboratory to be intramural capital
and intramural other current costs, respectively.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 115

6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

– In a second case, the transfers of funds are loosely described by the source
as “development contracts” for “prototypes”, but no R&D is performed by
the funder and very little by the recipient. For example, the government
places a contract with an industrial firm to “develop” a “prototype” civil
aircraft for a specific use (e.g. treatment of oil slicks). The aircraft is largely
constructed by the performer using existing materials and existing
technology, and R&D is only needed to meet the new specifications. Only
this portion of the contract should be reported by the performer as R&D
financed by the government sector, even though the funder’s accounts may
suggest at first sight that the entire contract was for R&D.
– In a third case, one unit receives money from another and uses it for R&D
although the funds were not paid out for that purpose. For example, a
research institute may finance some of its work through receipts from
royalties and profits from the sales of goods and services. Although these
funds are received from other units and other sectors, they should not be
considered as transfers for R&D but as coming from the “retained receipts”
of the performing unit itself, as the purchasers of the institute’s goods and
services did not intend to transfer funds for R&D.

6.3.3. Identifying the sources of flows of R&D funds

403. Performers are usually asked to distribute their intramural
expenditures between funds of the performing unit (own funds), funds from
other units in the same sector or sub-sector and funds from other sectors and
sub-sectors. They can usually do so relatively easily, but there are one or two
problem areas.

Sub-contracting and intermediaries

404. Problems arise when funds pass through several organisations. This
may occur when R&D is sub-contracted, as sometimes happens in the
business enterprise sector. The performer should indicate, as far as possible,
the original source of the funds for R&D. The same problems arise for EU
funding, as the funds first go to the main contractor and are then distributed
among the othe r participants (sub-contractors). In some countries,
intermediary non-performing organisations play an important role in the
financing of R&D by distributing among performers grants received from
several different sources but not “earmarked” for specific purposes. Well-
known examples are the Stifterverband für die Deutsche Wissenschaft and the
Deutsche Forschungsgemeinschaft in Germany. In such cases, while it is
acceptable to regard these organisations as the source, it is nonetheless
preferable to attempt to trace the funds to their original sources.

116 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

Public general university funds (GUF)

405. To finance their R&D activities, universities usually draw on three
types of funds:
– R&D contracts and earmarked grants received from government and other
outside sources. These should be credited to their original source.
– Income from endowments, shareholdings and property, plus surplus from
the sale of non-R&D services such as fees from individual students,
subscriptions to journals and sale of serum or agricultural produce. These
are the universities’ “own funds”. In the case of private universities, these
may be a major source of R&D funds.
– The general grant they receive from the ministry of education or from the
corresponding provincial or local authorities in support of their overall
research/teaching activities. One could argue that, as government is the
original source and has intended at least part of the funds concerned to be
devoted to R&D, the R&D content of these public general university funds
should be credited to government as a source of funds. One could also argue
that it is within universities that decisions are taken to commit money to
R&D out of a pool which contains both “own funds” as defined above and
public GUF; therefore, the sums concerned should be credited to higher
education as a source of funds. Government-financed GUF should be
credited to the public sector as a source of funds for the purposes of
international comparisons. For clarity, publicly financed GERD is divided
into two sub-categories: direct government funds and GUF.
406. The following procedures should be adopted. GUF should be separately
reported and any adjustments to the R&D cost series should take account of
real or imputed social security and pensions provisions, which should be
attributed to GUF as a source of funds. Monies from the higher education
“block grant” should be classified as GUF, and other monies generated by the
sector should be considered as “own funds”. Adjustments to other current
costs to account for real or imputed payments of rents, etc., should be
attributed to direct government funds.
407. As far as possible, the following sources of funds should be identified
in R&D surveys:
– Business enterprise sector:
❖ Own enterprise.
❖ Other enterprise in the same group.
❖ Other enterprise.
– Government sector:
❖ Central or federal government (excluding general university funds).
❖ Provincial or state government.(excluding general university funds).
❖ Public general university funds.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 117

6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

– Private non profit sector.

– Higher education sector.
– Abroad:
❖ Business enterprise:
● Enterprises within the same group.
● Other enterprises.
❖ Other national governments.
❖ Private non-profit.
❖ Higher education.
❖ EU.
❖ International organisations.

6.4. Extramural expenditures

408.
Extramural expenditures are the sums a unit, organisation or
sector reports having paid or committed themselves to pay to
another unit, organisation or sector for the performance of R&D
during a specific period. This includes acquisition of R&D
performed by other units and gran ts g iven to othe rs for
performing R&D.

409. For the acquisition of services closely related to intramural R&D

activities, the borderline between intramural and extramural expenditures is
not always clear. If these services are separate R&D projects, the expenditures
can in most cases be regarded as extramural R&D. If they are certain tasks (not
necessarily R&D as such) necessary for the intramural R&D of the unit but
contracted out, they can generally be regarded as intramural R&D expenditure
(other current costs). In principle, the same rules apply to consultants.
However, costs for on-site consultants come under other current costs (as
mentioned in paragraph 364) as their R&D activity is a direct part of the R&D
activity of the unit.
410. Data on the extramural R&D expenditures of statistical units are a
useful supplement to the information collected on intramural expenditures.
The collection of these data is therefore encouraged. These extramural
expenditure data are essential for providing statistics on R&D performed
abroad but financed by domestic institutions. They may also be helpful to
those analysing the flows of funds reported by performers, particularly if there
are gaps in survey coverage.
411. The focus of R&D data is necessarily individual countries, and it is very
difficult to track international flows of R&D funds. In the context of the

118 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

increasingly worldwide organisation of R&D, more use should be made of

analysis of extramural R&D funds to address this problem. It is therefore
recommended that some details on international flows, similar to those used
in the source of funds classification above, should be added to the
classification used to distribute extramural R&D.
412. For the distribution of extramural R&D, the following classification is
recommended:
– Business enterprise sector:
❖ Other enterprise in the same group.
❖ Other enterprise.
– Government sector.
– Private non-profit sector.
– Higher education sector.
– Abroad:
❖ Business enterprise:
● Enterprise within the same group.
● Other enterprise.
❖ Other national government.
❖ Private non-profit.
❖ Higher education.
❖ International organisations.

6.5. Reconciling differences in performer-based and source-based

reporting
413. In principle, the estimated total of R&D expenditure within a country
based on performer reports should equal the total based on reports from those
funding R&D (including funder reports to abroad). In practice, however, this is
not likely to be the case owing to sampling difficulties and reporting
differences.
414. In addition to reporting differences arising as a result of sampling error
(estimates of GERD are often obtained from sample surveys instead of surveys
of the entire population), countries have difficulty in reconciling funder and
performer data for several reasons.
415. Funders’ and performers’ views of whether the work being performed
meets the definition of R&D may differ. For example, in the US defence industry,
the eme rg enc e of n ew n on-tradition al c on tra ctors (inc luding larg e
telecommunications carriers, small high-technology firms) and increased R&D
funding of more generalised technical, analytical and professional contracts
(whose deliverables may be a small component of the overall defence R&D
project) have resulted in differing interpretations of what constitutes R&D.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 119

6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

416. The financing may be provided by an intermediary, making it difficult

for the performer to know the original source of funds (see paragraph 404). A
related problem is funding that goes outside of the funding sector but comes
back to the sector as externally funded R&D.
417. Contracts for research often extend over more than one year, with the
result that there may be timing discrepancies between funder and performer.
418. In many countries, it may be difficult to identify firms that pay for R&D
performed overseas. In fact, in cases of multinational firms, an enterprise in
one country may not know precisely how much it is funding R&D in another.
It may merely make a payment to a head office in another country for a range
of services, one of which is R&D.
419. A variant is the reconciliation of GBAORD data, which is essentially
government funder data (appropriations rather than expenditures, however),
to R&D performer data. In this case, the lack of comparability may be due to
the performance of a different amount of R&D than was expected at the
appropriations stage; it may also be due to an imprecision in the budget
a p pro p ri ati on s th a t d oes n ot a llow fo r s ep a ra te id en ti fic a ti on o f
appropriations that are specifically targeted to R&D (for more information on
GBAORD methodology, see Chapter 8).
420. In addition to the business enterprise and government sectors,
problems for reconciling funder- and performer-based R&D data arise for
other major funders of R&D, such as research councils and abroad.
421. To the extent possible, it is recommended that differences in R&D
expenditure totals between those estimated from the funders of R&D and
those estimated from the performers of R&D should be reported, and that
causal factors for the differences, if known, should be identified. It should be
recognised that such differences are not necessarily a result of inadequate or
inaccurate measurement and that providing these data will aid analytical and
statistical accuracy.

6.6. Regional distribution

422. A regional distribution of R&D intramural expenditures is also
recommended. For the EU member states, regional levels are given by the
Nomenclature of Territorial Units for Statistics (NUTS) classification. For other
OECD member countries, the regional distribution has to be determined
according to national needs. In federal countries, for example, it might be the
state level. Further details on the methods to be used for compiling regional
R&D data are found in Annex 5.

120 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 6 MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

6.7. National totals

6.7.1. Gross domestic expenditure on R&D (GERD)
423.
GERD is total intramural expenditure on R&D performed on the
national territory during a given period.

424. GERD includes R&D performed within a country and funded from
abroad but exclude s payments for R&D perform ed abroad. GERD is
constructed by adding together the intramural expenditures of the four
performing sectors. It is often displayed as a matrix of performing and
funding sectors (see Table 6.1). GERD and the GERD matrix are the basis of
international comparisons of R&D expenditures. They also provide the
accounting system within which the institutional classifications and
functional distributions may be applied.
425. It would be useful to have separate tables for defence and civil GERD,
in order to map how trends in these areas affect the level and structure of total
GERD. This is particularly true for countries with significant defence R&D
programmes. The separation is encouraged for other countries as well, as a
way to increase the comparability of data on civil R&D.

6.7.2. Gross national expenditure on R&D (GNERD)

426. The GNERD aggregate comprises total expenditure on R&D financed by
a country’s institutions during a given period. It includes R&D performed
abroad but financed by national institutions or residents; it excludes R&D
performed within a country but funded from abroad. It is constructed by
adding the domestically financed intramural expenditures of each performing
sector and the R&D performed abroad but financed by domestic funding
sectors (see Table 6.2). It gives some supplementary information on R&D co-
operation between different kinds of units.
427. To allow for the identification of R&D activities of international
org anisations , th e “Abroad” se ctor should have sub-categ ories for
international organisations, as recommended in the institutional sub-
classification (see Chapter 3, Section 3.8.3).

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 121

122

6
MEASUREMENT OF EXPENDITURES DEVOTED TO R&D
Table 6.1. Gross domestic expenditure on R&D (GERD)
Sector of performance
Funding sector Total
Business enterprise Government Private non-profit Higher education

Business enterprise Total domestic performance financed

by the business enterprise sector
Government Total domestic performance financed
by the government sector
Public general university funds (GUF) Total domestic performance financed
by public general university funds
(GUF)
Higher education Total domestic performance financed
by the higher education sector
Private non-profit (PNP) Total domestic performance financed
by the private non-profit sector
Abroad Total domestic performance financed
FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

by abroad
• Foreign enterprises
– Within the same group
– Other
• Foreign government
• European Union
• International organisations
• Other

Total Total performed Total performed Total performed Total performed GERD
in the business enterprise in the government sector in the private non-profit in the higher education
sector sector sector

Source: OECD.
FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

Table 6.2. Gross national expenditure on R&D (GNERD)

Sector of performance

National territory Abroad

Funding sector Business enterprise Total
Business Private Higher International
Government Within the same Other business Other
enterprise non-profit education organisations
group enterprise

Business enterprise Total financed by the

business enterprise
sector
Government Total financed by the
government sector
Public general Total financed by public

6
university funds (GUF) GUF

MEASUREMENT OF EXPENDITURES DEVOTED TO R&D

Higher education Total financed by the
higher education sector
Private non-profit Total financed by the
PNP sector

Total Total nationally Total nationally Total nationally Total nationally Total nationally Total nationally Total nationally Total nationally GNERD
financed financed financed financed financed financed financed financed
performed performed in performed performed performed performed performed performed
in the business the government in the private in the higher abroad abroad in other abroad abroad in other
enterprise sector non-profit education in business business in international organisations
sector sector sector enterprise enterprise organisations
within the same
group

Source: OECD.
123
 ISBN 92-64-19903-9
Frascati Manual 2002
Proposed Standard Practice for Surveys on Research
and Experimental Development
© OECD 2002

Chapter 7

Survey Methodology and Procedures

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 125

7 SURVEY METHODOLOGY AND PROCEDURES

7.1. Introduction
428. Information on R&D may be obtained from different sources, such as
annual reports of research councils or major R&D-performing institutions.
These data can only give an approximate measure of R&D efforts. Not only do
the concepts of R&D used often differ from the definitions given in this
Manual; they may also change over time. It is also very difficult to obtain all
data for the same period and to avoid double counting when tracking flows
from financial statements and other sources. For these reasons, statistics on
R&D require regular, systematic and harmonised special surveys. However,
because of lack of satisfactory records, costs of statistical surveys and the need
to restrict statistical demands on respondents, surveys cannot always provide
all the information needed.
429. Estimates are a necessary supplement to surveys (respondents must
often make estimates in order to provide the requested “survey” information).
Using ratios derived from survey data, it is possible to provide adequate
aggregate trends or totals from incomplete information without recourse to a
costly survey. Indeed, the R&D inputs of the higher education sector are often
partially, and in some countries wholly, estimated. When statistics are
released, full information on the sources and generation of the statistics
should be provided.
430. To improve international comparability, this chapter gives some
methodological guidelines for conducting R&D surveys. These are based on
identified best practices. As R&D survey methodologies and procedures are
well established in many countries, the guidelines are quite general so as to be
as widely applicable as possible.

7.2. Scope of R&D surveys

431. In theory, R&D surveys should identify and measure all financial and
personnel resources devoted to all R&D activities in all R&D units. R&D surveys
are mainly addressed to R&D-performing units, which may also finance R&D
performed in other units (this is covered by a question on extramural
expenditure). Units which only finance R&D are surveyed to some extent by
member countries. Government departments, for example, are surveyed in the
context of calculating government budget appropriations or outlays for R&D
(GBAORD) by socio-economic objectives. This chapter, however, only addresses
performer-based surveys. Statistical methodologies and other procedures have

126 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 7 SURVEY METHODOLOGY AND PROCEDURES

to be established to capture all R&D, especially for units in the business

enterprise sector with little R&D. These are described in more detail below.

7.3. Identifying target population and survey respondents

432. Only in a few member countries can the surveying agency make an
exhaustive survey of all possible R&D performers. Generally, there are many
constraints on the extent of surveys. For example, the number of respondents
may have to be restricted to keep costs down; an R&D survey may have to be
taken in conjunction with another survey with acceptable, but not ideal,
respondents; surveys of some groups may require the participation of other
agencies with different data needs and hence different questions for
respondents.
433. It is not possible to make detailed recommendations on survey methods
that would be equally relevant to all member countries, as the size and structure
of national R&D capacities vary widely. Suggestions are given for the business
enterprise, government, private non-profit and higher education sectors,
although it is recognised that some countries use different systems of sectoring
for surveying and reporting data. For example, some countries undertake surveys
of enterprises, institutes and higher education teaching establishments and
redistribute the institutes among the four standard sectors of performance.
434. Hospitals/healthcare institutions are a special category. They carry out
R&D which may relate to any of the four standard sectors of performance. Health-
related research has increased significantly in recent years, and it is worth
reviewing survey coverage to ensure that R&D data are collected for all hospitals
and healthcare units that are liable to undertake such activities, not only
university hospitals and other research-intensive hospitals but also general
hospitals and other healthcare units (ISIC 8512 and 8519). In several countries,
some categories of publicly funded hospitals/healthcare units may not be
officially permitted to use their funds for R&D but may nevertheless host some
research activities. If possible, the major research-performing units should be
surveyed following the standard national timetable. For general hospitals and
other healthcare units, benchmark surveys should be made at least every decade
and methods of estimation established for the intervening years.

7.3.1. Business enterprise sector

435. The enterprise is recommended as the main statistical unit in the
business enterprise sector (see also Chapter 3, Section 3.4.2). Some enterprises
perform R&D on a regular basis from year to year, and they may have one or
several R&D units. Other enterprises perform R&D only occasionally. They may
be engaged in a project in one year and not perform any R&D in the next. This
R&D is often performed by people from various parts of the enterprise on a

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 127

7 SURVEY METHODOLOGY AND PROCEDURES

project basis without a formal R&D organisation. In the general Frascati Manual
definition of R&D, the criterion of “creative work undertaken on a systematic
basis” is fulfilled by a project which has specific goals and a budget.

It is recommended that all enterprises performing R&D, either

continuously or occasionally, should be included in R&D surveys.

436. There are at least two feasible approaches for establishing the survey
population of the business enterprise sector. One is to take a census-based
survey of large enterprises and a sample of smaller ones belonging to a certain
population (in terms of industry and size class) from the entire sector in order
to identify R&D performers and request the information from them. The
choice of enterprises should be based on a business register of good quality. In
this approach, R&D performed in the past in the enterprise is not considered.
This is the approach followed in innovation surveys.
437. Surveys of this kind will cover a large number of enterprises and are
expensive if applied to all industries and all enterprises regardless of size. It is
therefore necessary to limit the target population in terms of size of enterprise
and industries covered. This normally leads to the systematic exclusion of very
small enterprises and enterprises in certain less R&D-intensive industries.
When the sample size is very small, estimates may be less reliable, owing to
raising factors. In practice, no member country follows this approach strictly.
438. In R&D surveys for the business enterprise sector, most member
countries use the second approach, i.e. they try to survey all enterprises
known or assumed to perform R&D. The survey is based on a register of R&D-
performing enterprises. The sources of this register include lists of enterprises
receiving government grants and contracts for R&D, lists of enterprises
reporting R&D activities in previous R&D surveys, in innovation surveys or
other enterprise surveys, directories of R&D laboratories, members of
industrial research associations, employers of very highly qualified personnel
and lists of enterprises claiming tax deductions for R&D. Several countries
only use this kind of information to identify R&D performers.
439. It is very difficult to maintain completely up-to-date registers of
enterprises that occasionally perform R&D from these sources. This may lead
to undercoverage of R&D in small or medium-sized enterprises. The effect on
total business enterprise R&D, however, is not significant, as the large R&D
performers are included in any case.
440. To improve the coverage of R&D surveys that use this approach, many
countries use a combination of these approaches, i.e. they systematically take
a census/sample survey to collect information on R&D from enterprises not

128 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 7 SURVEY METHODOLOGY AND PROCEDURES

included in registers of R&D-performing enterprises. For cost reasons, such

surveys are limited in terms of industries covered and enterprise size. The
limitations mainly concern the services sectors, as there is little experience
with surveying their R&D activities. Enterprises with a very low probability of
performing R&D should be excluded in order to reduce the response burden.
The advantage of this approach is that it considerably reduces the uncertainty
in estimating figures for the target population as compared to the pure
sampling approach described above, which does not take previous R&D into
account. Its disadvantage is the cost, which may make it difficult to apply in
bigger countries.
441. It is therefore recommended:
– To include in R&D surveys of the business enterprise sector all firms known
or supposed to perform R&D.
– To identify R&D performers not known or supposed to perform R&D by a
census/sample of all other firms in the industries listed below. In principle,
enterprises in all size classes should be included, but if a cut-off point is
necessary, it should be at ten employees.
442. The following industries should be included:

Industry ISIC Rev. 3/NACE Rev. 1

Mining 14
Manufacturing 15-37
Utilities, construction 40, 41, 45
Wholesale 50
Transport, storage and communication 60-64
Financial intermediation 65-67
Computer and related activities 72
R&D services 73
Architectural, engineering and other technical activities 742

In addition, other sectors, for example agriculture (ISIC Rev. 3, Divisions 01, 02, 05),
should be covered in countries with significant amounts of research in these
sectors.

7.3.2. Government sector

443. Units to include in surveys are:
– R&D institutes.
– R&D activities of general administrations of central or state government,
statistical, meteorological, geological and other public services, museums,
hospitals.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 129

7 SURVEY METHODOLOGY AND PROCEDURES

– R&D activities at the municipality level.

The best way to survey is to send questionnaires to all units

known or assumed to perform R&D.

444. There may be several ways of updating lists of R&D-performing units,

such as business registers, directories of R&D-performing units, research
associations, bibliometric sources, requests for updates from administrative
bodies, etc.
445. It is especially difficult to identify R&D activities at the municipality
level owing to the large number of units, the small number of likely R&D
performers and difficulties in the interpretation of the concept of R&D. Lists of
R&D performers usually do not include these units. It may be worthwhile to
make an effort to identify R&D performers in large cities.

7.3.3. Private non-profit sector

446. The sources for identifying possible survey respondents are mainly the
same as for the government sector. Register information may be less
comprehensive and could be completed by information from researchers or
research administrations. This sector may be more relevant for surveys on
R&D funding.

7.3.4. Higher education sector

447.
The surveys and estimation procedures (see below) should
cover all universities and corresponding institutions, especially
those awarding degrees at the doctorate level. Other institutions
in the sector known or assumed to perform R&D should also be
included.

448. Identification of these institutions is generally relatively easy. If

possible, it is often preferable to use smaller units, such as departments or
institutes of the university, as statistical units.

7.3.5. Hospitals
449. Some countries may find it satisfactory to include hospitals and
h ealthcare in stitutions in regular R&D surveys, using the standard
questionnaire for the sector concerned. Indeed, this may be the only option for
hospitals and other healthcare units in the business enterprise sector. In this

130 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 7 SURVEY METHODOLOGY AND PROCEDURES

case, additional guidance on the borderline between research and healthcare

and on the treatment of clinical trials might be supplied. Where university
hospitals are administratively and financially very closely integrated with
teaching establishments (see Chapter 3, Section 3.7.1), they might be treated
together for the purposes of R&D surveys/data compilation. If they are
separate units with separate accounts and administrations, they might
receive a special questionnaire directed to government hospitals (see below)
or a normal R&D questionnaire. For hospitals in the government and non-
profit sectors and university hospitals (or parts thereof) which are not
integrated with teaching establishments, a special survey may be useful. If
this is not possible, the normal R&D questionnaire may be used.
450. Whatever survey approach is taken, care should be taken to ensure
coherent treatment of R&D in units/projects under joint management by two
or more entities, by persons receiving two salaries from different entities and
by persons working at hospitals but employed by other institutions.

7.4. Working with respondents

7.4.1. Encouraging co-operation
451. The survey questionnaire must include a minimum number of basic
questions on the R&D activity in order to produce harmonised an d
comparable statistics for transmission to international organisations. Owing
to the response burden, the questionnaire should be as simple and short as
possible, logically structured and have clear definitions and instructions.
Generally, the longer the questionnaire, the lower the unit and item response
rates. For smaller units, a simplified survey questionnaire could be used. It is
highly recommended to test draft questionnaires on a sample of respondents.
Work has started to develop a harmonised OECD questionnaire for R&D
surveys in the business enterprise sector.
452. Once the survey respondent has been identified, it is necessary to
identify the best person to fill in the questionnaire. In R&D surveys, he/she is
usually in the accounting or personnel unit or in the R&D unit. Each has
advantages and disadvantages. The R&D manager can identify the R&D of the
unit according to Frascati Manual norms better but may not be able to supply
exact figures. The accountant or personnel manager has the exact figures, but
may not refer exactly to R&D as defined in the Frascati Manual. In bigger units,
the co-operation of all three types of respondents is essential. Nevertheless,
one person must co-ordinate the response. It is often useful to send the
questionnaire to the person who responded the previous year. If this is not
known, surveys should be addressed to the managing director. In big, complex
institutions like universities and large enterprises or groups of enterprises,

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 131

7 SURVEY METHODOLOGY AND PROCEDURES

it is useful to identify in advance the person responsible for providing

information and for co-ordinating information from smaller sub-units.
453. It is very important to secure the co-operation of the person in charge
of responding. Respondents are asked to spend time on a task which is often
of no direct benefit to them; they may even see completing a questionnaire on
R&D as a waste of time and money. It is the responsibility of the surveying
agency to help contributors to appreciate the potential uses of the data and to
be alert to respondents’ potential needs in terms of R&D statistics. It is also its
responsibility to respect confidential data and to ensure that users are aware
of respondents’ concerns. In the design of surveys, it should consider the need
to minimise the burden on respondents.
454. The respondent is rarely a user of the statistics, but it is important to
show what has been done with the data in order to encourage co-operation.
The respondent might receive the publication, or, if this is not feasible, a
summary. Customised information which allows the respondent to compare
his/her unit with corresponding national totals may also be useful.
455. The statistical agency should provide the respondent with technical
assistance and name, phone number, fax number and e-mail addresses for all
contact persons within the agency. The extent to which follow-up procedures
are used will depend on the level and quality of responses, the number of
units surveyed and the resources available to the surveying authority. It is
rarely feasible to contact personally all the units surveyed. One possibility is to
plan a follow-up programme for each enquiry, aiming to visit all the main
units over a given period. Another is to limit the follow-up and to check a few
entities very thoroughly. Personal contacts with respondents who require
guidance or who submit unsatisfactory returns should be encouraged.
456. Almost all respondents will have to make some estimates. Not only is
R&D a complex activity in itself, it is inextricably linked to a number of other
activities. Furthermore, an institution’s R&D may not be satisfactorily
reflected either in its organisation or in its records and accounts.
457. R&D is not only what R&D laboratories and research institutes do. It is
both less and more than this, since very few of the surveyed entities have only
one activity. The measurement of R&D inputs may be carried out in three
stages:
– Identification of all specialised R&D units and measurement of their total
activity.
– Estimates of the non-R&D portions of their activity and subtraction of these
estimates from the total.
– Estimates of the inputs used for R&D in other units and addition of these
estimates to the total.

132 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 7 SURVEY METHODOLOGY AND PROCEDURES

458. In practice, minor deviations from the strict R&D definition may be
overlooked in order to better utilise existing records or to otherwise ease the
burden on respondents. In some cases, particularly in the higher education
sector, it may be necessary to resort to very crude ratios to estimate R&D
inputs.

7.4.2. Operational criteria

459. Operational criteria that are suitable for the sector surveyed must be
developed. Thus, on questionnaires intended for the business enterprise
sector, it would be appropriate to give guidance for distinguishing between
R&D and pre-production, whereas a government questionnaire might
concentrate on the difference between R&D, on the one hand, and data
collection and information, on the other. Sector-specific examples might be
useful to guide respondents. Reference might be made to detailed examples in
this Manual. Responding units may need criteria for distinguishing between
contracts to industry for goods and services required for intramural R&D and
those awarded for the performance of industrial R&D. Criteria with the same
intent but different wording may be useful in business enterprise surveys. Nor
should differences within a sector be overlooked. For example, operational
definitions and examples appropriate for the oil and gas industry are probably
not well suited to the electrical products industry. In discussions with
respondents, supplementary criteria are often useful. Examples are given in
Chapter 2, Table 2.1.
460. During R&D surveys, respondents may find it very difficult to apply the
theoretical distinctions made in earlier chapters of this Manual to the wide
range of projects under way in their organisation. As surveying agencies are
not always able to check responses and are usually obliged to accept them as
given, it is of utmost importance that they provide the institutions surveyed
with clear explanations and guidance to complement the formal definitions in
order to ensure uniformity.
461. Four important tools for achieving this objective are:
– Explanatory notes.
– Hypothetical examples.
– Guidance to individual respondents.
– Documentation on treatment of different cases.
462. For obvious reasons, this Manual deals exclusively with the first two
t oo ls. Fo rm a l d e fi n i tio n s an d th eo re ti c al di sti n c tio n s h ave to be
complemented by the last two. To ensure that the guidance given by surveying
agencies is consistent, it is essential to develop documentation on how
difficult borderline cases have been solved. Such documentation can also

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 133

7 SURVEY METHODOLOGY AND PROCEDURES

serve as a valuable source of hypothetical examples and may help countries to

develop more uniform classification practices.

7.5. Estimation procedures

463. In the process of compiling R&D statistics, various estimation
procedures are used. Results from sample surveys have to be grossed up, using
various methods, to correspond to the total target population. Especially in
surveys of the business enterprise and government sectors, there are problems
of unit and item non-response. In the higher education sector, statistics in most
countries are based on a combination of surveys and estimation procedures.

7.5.1. Unit and item non-response

464. In practice, responses to R&D surveys are often incomplete,
irrespective of the survey method used. Two types of missing values can be
distinguished: item and unit non-responses. Unit non-response means that a
reporting unit does not reply at all. The surveying institute may not be able to
reach the reporting unit or the reporting unit may refuse to answer. For item
non-response, a unit does answer but leaves at least one question blank or
even, in an extreme case, leaves all questions but one blank.
465. Item and unit non-responses would be less of a problem if missing
values were randomly distributed over all sampling units and all questions. In
reality, however, both types of missing values are biased with respect to
certain characteristics of the population and the questionnaire. Item non-
response is more likely when the question is (or seems to be) difficult.
Examples are the breakdown of R&D investments (land and buildings and
equipment) or of R&D by type of R&D.
466. These non-responses clearly affect the comparability of the results of
national and international R&D surveys. Appropriate methods to overcome
this problem have to be developed and used. As different methods may lead to
different results, some general recommendations should be followed.
Otherwise, differences in results over time and/or among countries may arise
from using different concepts to reduce the bias of item and unit non-
responses.
467. For practical as well as theoretical reasons, one recommended way to
overcome the problem of item non-response is a group of methods called
“imputation methods” for estimating missing values on the basis of additional
information. The easiest method is to use the previous answer for the same
enterprise. Another possibility is to use statistical techniques such as “hot
decking”, using information from the same survey, or “cold decking”, using
information from previous surveys.

134 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 7 SURVEY METHODOLOGY AND PROCEDURES

468. In the case of unit non-response, past R&D data at firm level can be used
to estimate the R&D expenditures for the same firm for the current period. The
evolution of sales and or employment can be used to adapt the previous figures.
In cases where no previous R&D data at firm level are available, as R&D is a
metric variable correlated to a certain degree with sales, a recommended
method is to use the relation between the sales of the total population and the
sales of the realised sample for each cell in the sample. Another method is to
use employment as a variable. This procedure is based on the assumption that
the ratios of R&D to sales or R&D personnel to total personnel of responding and
non-responding units are identical. This assumption can be tested through
non-response analysis of a representative sample of non-responding units.
Even if the assumption is wrong, the bias introduced can be disregarded as long
as the fraction of non-responding units is fairly small.

7.5.2. Estimation procedures in the higher education sector

469.
It is recommended that information on R&D in this sector
s h o u l d b e b a s e d o n s ur v e y s o f t h e p e r f or m i n g u n i t s,
supplemented, if necessary, by estimations.

470. Often over half of the funding of R&D is given as general university
funds, not earmarked for research but given for the general functioning of the
university. The R&D share of these funds is often unknown to the universities
themselves. To determine which part should be devoted to R&D, a variety of
methods have been used:
– Central estimates not based on empirical knowledge of how time is spent
on different activities.
– Time-use surveys/studies concerning the distribution of time by various
categories of personnel.
– Time-use surveys/studies based on researchers’ own evaluation of their
working time.
471. From the time-use studies, research coefficients are derived for use in
calculating full-time equivalents on R&D (FTEs) and R&D labour costs. Other
R&D costs should primarily be estimated on the basis of purpose. For example,
the acquisition of research equipment and expenditures for a research
laboratory should be put under research, while maintenance of teaching
facilities should be put under teaching. For expenditures not clearly
attributable to either research or teaching, an estimate can be made using the
research coefficients as the basis of calculation.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 135

7 SURVEY METHODOLOGY AND PROCEDURES

472. For a more detailed discussion of various time-use survey methods

and issues related to the compilation of R&D statistics for the higher
education sector, see Annex 2.

7.6. Reporting to the OECD or to other international organisations

473. National authorities carry out R&D surveys to obtain data relevant to
national concerns in the framework of national institutional arrangements.
Discrepancies may exist between national practices and international norms
laid down in this or other manuals. Nevertheless, every effort to reduce the
impact of such discrepancies should be made when reporting these data to
the OECD or to other international organisations by making adjustments or
estimates, even if this means that the R&D data in international sources will
differ from those in national documents. If national authorities are unable to
make such adjustments on their own responsibility, they might help the
relevant organisations to make informed estimates. Where such adjustments
cannot be made, full technical notes should be submitted. Discrepancies are
generally of two kinds:
– Explicit differences in approach between national R&D surveys and that
recommended in this Manual.
– “Implicit” differences between the standard national economic or
educational classifications used in the country’s s urveys an d the
international classifications recommended in this Manual.
It is important to identify and report both kinds of discrepancy.

136 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ISBN 92-64-19903-9
Frascati Manual 2002
Proposed Standard Practice for Surveys on Research
and Experimental Development
© OECD 2002

Chapter 8

Government Budget Appropriations

or Outlays for R&D by Socio-economic
Objectives (GBAORD)

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 137

GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

8.1. Introduction
474. There are two ways of measuring how much governments spend on
R&D. The first and most accurate is to hold surveys of the units that carry out
R&D (firms, institutes, universities, etc.) in order to identify the amount
effectively spent on R&D over the previous year and the share financed by
government. The sum of the R&D spending in a national territory (see
Chapter 6, Table 6.1) is known as “government-financed gross domestic
expenditure on R&D” (government-financed GERD).
475. Unfortunately, owing to the time required to carry out such surveys
and process the results, government-financed GERD data do not become
available until between one and two years after the R&D has been carried out.
Furthermore, the R&D-performing units responding to the surveys are
sometimes unable to report on where their particular grant or contract fits
into the government’s overall S&T policy.
476. In consequence, a second way of measuring government support for
R&D has been developed using data from budgets. This essentially involves
identifying all the budget items involving R&D and measuring or estimating
their R&D content in terms of funding. These estimates are less accurate than
performance-based data but as they are derived from the budget, they can be
linked to policy through classification by “objectives” or “goals”. The
specifications of such budget-based data are described in this chapter.
Budget-based data are now officially referred to as “government budget
appropriations or outlays for R&D” (GBAORD).

8.2. Relationship with other international standards

477. The definitions discussed in this chapter are, as far as possible,
compatible with the methodologies developed by Eurostat and by Nordforsk/
Nordic Industrial Fund (Nordforsk, 1983).

8.3. Sources of budgetary data for GBAORD

478. Although details of the budgetary procedure vary from country to
country, seven broad stages can be identified:
i) Forecasts (estimates of funding before beginning of budget discussion).
ii) Budget forecasts (preliminary figures as requested by ministries,
especially for inter-ministerial discussions).

138 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

iii) Budget proposal (figures presented to the parliament for the coming
year).
iv) Initial budget appropriations (figures as voted by the parliament for the
coming year, including changes introduced in the parliamentary debate).
v) Final budget appropriations (figures as voted by the parliament for the
coming year, including additional votes during the year).
vi) Obligations (money actually committed during the year).
vii) Actual outlays (money paid out during the year).
479. Stages i)-iv) describe the government’s intentions. The data for
budgetary year y should be available as soon as possible after the end of year
y – 1. It is suggested that the preliminary GBAORD data should be based on the
first budget agreed between the government and the parliament, or stage iv).
Some countries might even base their preliminary figures on stage iii). During
the budgetary year, supplementary budgets may be voted, including increases,
cuts and reallocations of R&D funding. These are reflected in stage v). Data
should be available as soon as possible after the end of the budgetary year. It
is suggested that the final GBAORD data should be based on final budget
appropriations. Some countries may have to base their final figures on
stages vi) or vii).

8.4. Coverage of R&D

8.4.1. Basic definition
480. The basic definition is the one given in Chapter 2, Section 2.1. Basic
research, applied research and experimental development are all included but
are not identified separately.

8.4.2. Fields of science and technology

481.
The analysis covers natural sciences and engineering (NSE)
and social sciences and humanities (SSH) and makes no
distinction between the two.

8.4.3. Identifying R&D

482. As far as possible, all guidelines and conventions listed in Chapter 2 for
distinguishing R&D from non-R&D activities should be applied. Particular care
should be taken to check the real R&D content of budget items officially
described as “development contracts” or as “purchase of prototypes” as
discussed in Chapter 2, Section 2.3.4, and Chapter 6, Section 6.3.2, as well as
Annex 10.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 139

GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

483. Countries may need to develop a set of coefficients according to

discipline, institution, other criteria or a mixture of these in order to
determine the proportion of R&D in non-exclusive budget items like GUF but
also for a range of institutions which also perform activities other than R&D.
As far as possible, these coefficients should be consistent with what is
reported by these institutions as R&D in performer-based surveys.

8.5. Definition of government

484. “Government” should cover central (or federal), provincial (or state)
and local government (see Chapter 3, Section 3.5). Public enterprises are
excluded, as they are treated as part of the business enterprise sector. For the
purposes of GBAORD, however, it is recommended that:

● Central or federal government should always be included.

● Provincial or state government should be included when its
contribution is significant.
● Local government funds (i.e. those raised by local taxes) should
be excluded.

8.6. Coverage of government budget appropriations and outlays

8.6.1. Intramural and extramural expenditures
485.
GBAORD covers not only government-financed R&D performed
in government establishments but also government-financed
R&D in the other three national sectors (business enterprise,
private non-profit, higher education) as well as abroad (including
international organisations).

8.6.2. Funding and performer-based reporting

486. R&D expenditures can be reported either by the agency providing the
money (funding) or by the agency actually performing the R&D. In general, this
Manual recommends the second approach, which is used in the standard tables
in OECD surveys. However, the first approach is preferred for the GBAORD series.

GBAORD data should be based on the funder rather than the

performer.

140 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

8.6.3. Budgetary funds

487.
GBAORD clearly includes all outlays to be met from taxation or
other government revenue within the budget.

488. A problem arises for appropriations for R&D performed by government

institutions but expected to be financed from other sources. In some
countries, these may be included in the government budget, on the grounds
that the agency concerned needs government permission to spend them
(gross approach). In others they may be excluded (net approach). When
dealing with these government funds, a distinction should be made between:
– Contracts or grants from other sectors for the performance of R&D by
government institutions.
– Other government funds, such as the retained receipts of government
laboratories, receipts from levies, etc.

Net budgetary appropriations

489.
Appropriations for which corresponding revenue is expected
either from other government sources or other sectors should be
excluded from GBAORD according to the net principle.

490. If an R&D institute has a total gross budget of 10 million (including

3 million for externally financed contract research), for example, only
7 million should be counted as net budgetary appropriations for the institute,
as the 3 million is on the budget of the funder of the contract research.

Other government funds

491. No specific guidelines can be suggested, but other government funds
should generally be included in GBAORD if they are in the budget. This also
concerns social security funds, if they are voted on in the parliament during
the budgetary process.

8.6.4. Direct and indirect funding

Treatment of public general university funds
492.
GBAORD includes public general university funds (GUF).

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 141

GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

Loans and indirect funding of industrial R&D

493. As far as possible, instructions regarding both loans and indirect
funding apply (Chapter 6, Section 6.3.2). Thus, loans that may be forgiven
should be included in GBAORD, but loans that are to be repaid and indirect
support of industrial R&D via tax rebates, etc., should in principle be excluded.
Nevertheless, when such indirect support programmes are undertaken as part
of an integrated R&D policy (for example, when the sources are documented
and are included in inter-ministerial discussions of a science budget), they
may be included in GBAORD. However, indirect funding should always be
declared separately so that it can be excluded when making certain
international comparisons.

8.6.5. Types of expenditure

General coverage
494.
GBAORD includes both current costs and capital expenditure.

Money carried forward

495. In some countries it is budgetary practice to carry forward large sums
from one year to the next, sometimes including them in sums voted in
successive years.

Multi-annual projects budgeted in only one year or over several

should be allocated to the GBAORD of the year(s) in which they
are budgeted, not in the years of performance. Multi-annual
programmes that are authorised at some stage but budgeted over
several years should be allocated to the years in which they are
budgeted, not the year of authorisation.

8.6.6. GBAORD going to R&D abroad

496. Only contributions to international R&D programmes or organisations
solely or mainly concerned with R&D should be included. Contributions of a
general nature (like those to the UN, the OECD, the EU, etc.) should be
excluded. Appropriations to the following institutions should be included:
– CERN (European Organisation for Nuclear Research).
– ESA (European Space Agency).
– CGIAR (Consultative Group on International Agricultural Research).
– ESRF (European Synchrotron Radiation Facility).

142 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

– EMBO (European Molecular Biology Organisation), including EMBL, EMBC.

– IAEA (International Atomic Energy Agency).
– COST (Co-operation in Scientific and Technical Research, EU programme
accessible to non-members).
– EUREKA (European Network for Market-Oriented Industrial R&D).

8.7. Distribution by socio-economic objectives

8.7.1. Criteria for distribution
Purpose or content
497. Two approaches to distribution are possible:
– According to the purpose of the R&D programme or project.
– According to the general content of the R&D programme or project.
498. The difference between the two is illustrated by the following
examples:
– A research project on the effects on human body functions of various
chemicals which could be used as weapons: the purpose is “defence” but
the R&D content is “human health”.
– A research project to develop fuel cells to provide power in remote forest
locations financed by th e Ministry of Ag riculture : the purpose is
“agriculture, forestry and fishing” but the R&D content is “energy”.

P urpo se i s more fun d am en tal from the v iew poin t of

government policy, and this approach is used in principle for the
collection of GBAORD by socio-economic objective.

Primary and secondary objectives

499. Though some government-supported R&D programmes have only one
purpose, others may have several. For example, a government may commit
money to an aircraft project primarily for military reasons but also to
encourage export sales by the aerospace industry and even to assist spin-offs
to civil aviation. However, in reports to the OECD, R&D should be classified
according to its primary objective.

Identifying primary objectives

500. Where there are problems in identifying the primary purpose of the
fund er of the R& D or wh ere there see m to be differen ces be twe en
the “purpose” and the “content” of a programme, two principles originally

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 143

GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

developed for EU reporting using NABS (“Nomenclature for the Analysis and
Comparison of Scientific Programmes and Budgets”) may be of use:

● Direct derivation: A project which owes its existence solely to

the technical needs of another programme is directly derived
from the said programme and should be classified with it.
● Indirect spin-off: Where the results of R&D undertaken for one
purpose are subsequently reworked to give an application
relevant to another objective, this is indirect spin-off and
should be credited to the objective to which the subsequent
R&D is oriented.

8.7.2. Distribution of budgetary items

501. The allocation of R&D appropriations or outlays to socio-economic
objectives should be at the level that most accurately reflects the funder’s
purpose(s). The actual reporting level chosen will depend on practical
possibilities. The whole appropriation may be to an R&D-performing or R&D-
funding unit. In some cases, information on programme or project level may
be obtained.

8.7.3. The distribution

502. The OECD distribution list given in Section 8.7.4 is the EU classification
adopted by Eurostat for the analysis and comparison of scientific programmes
and budgets at the one-digit level (NABS) (Eurostat, 1986; 1994). The
correspondence between the NABS list and the Frascati Manual 1993 list (which
was almost identical to NABS 1986) is shown in Table 8.1 and should be used for
reporting to the OECD, even if member countries use their own classifications or
the Nordforsk classification (Table 8.2) in their national GBAORD compilations.

8.7.4. Socio-economic objectives – SEO

1. Exploration and exploitation of the Earth
503. This SEO covers research with objectives related to the exploration of
the Earth’s crust and mantle, seas, oceans and atmosphere, and research on
their exploitation. It also includes climatic and meteorological research, polar
exploration (under various SEOs, as appropriate) and hydrology. It does not
include:
– Soil improvement and land use (SEO 2).
– Research on pollution (SEO 3).
– Fishing (SEO 6).

144 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

2. Infrastructure and general planning of land use

504. This SEO covers research on infrastructure and land development,
including research on the construction of buildings. More generally, this SEO
covers all research relating to the general planning of land use. This includes
research into protection against harmful effects in town and country planning
but not research into other types of pollution (SEO 3).

3. Control and care of the environment

505. This SEO covers research into the control of pollution, aimed at the
identification and analysis of the sources of pollution and their causes, and all
pollutants, including their dispersal in the environment and the effects on
man, species (fauna, flora, micro-organisms) and the biosphere. Development
of monitoring facilities for the measurement of all kinds of pollution is
included. The same is valid for the elimination and prevention of all forms of
pollution in all types of environment.

4. Protection and improvement of human health

506. This SEO covers research aimed at protecting, promoting and restoring
human health, broadly interpreted to include health aspects of nutrition and food
hygiene. It ranges from preventive medicine, including all aspects of medical and
surgical treatment, both for individuals and groups, and the provision of hospital
and home care, to social medicine and paediatric and geriatric research.

5. Production, distribution and rational utilisation of energy

507. This SEO covers research into the production, storage, transportation,
distribution and rational use of all forms of energy. It also includes research on
processes designed to increase the efficiency of energy production and
distribution, and the study of energy conservation. It does not include:
– Research relating to prospecting (SEO 1).
– Research into vehicle and engine propulsion (SEO 7).

6. Agricultural production and technology

508. This SEO covers all research on the promotion of agriculture, forestry,
fisheries and foodstuff production. It includes: research on chemical
fertilisers, biocides, biological pest control and the mechanisation of
agriculture; research on the impact of agricultural and forestry activities on
the environment; research in the field of developing food productivity and
technology. It does not include:
– Research on the reduction of pollution (SEO 3).

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 145

GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

– Research into the development of rural areas, the construction and

planning of buildings, the improvement of rural rest and recreation
amenities and agricultural water supply (SEO 2).
– Research on energy measures (SEO 5).
– Research for the food industry (SEO 7).

7. Industrial production and technology

509. This SEO covers research on the improvement of industrial production
and technology. It includes research on industrial products and their
manufacturing processes, except where they form an integral part of the
pursuit of other objectives (e.g. defence, space, energy, agriculture).

8. Social structures and relationships

510. This SEO covers research on social objectives, as analysed in particular
by social and human sciences, which have no obvious connection with other
SEOs. This analysis includes quantitative, qualitative, organisational and
forecasting aspects of social problems.

9. Exploration and exploitation of space

511. This SEO covers all civil space research and technology. Corresponding
research in the defence field is classified in SEO 13. Although civil space
research is not in general concerned with particular objectives, it frequently
has a specific goal, such as the increase of general knowledge (e.g. astronomy),
or relates to particular applications (e.g. telecommunications satellites).

10. Research financed from general university funds

512. When reporting GBAORD by “purpose”, this class should include, by
convention, all R&D financed from general purpose grants from ministries of
education, although in some countries many of these programmes may be
relevant to other objectives. This convention has been adopted because of the
problem of obtaining suitable data and thus of comparability. Member
countries should provide the most detailed breakdown possible of the
“contents” of this class by field of science and technology and, where they are
able to do so, by objectives.

11. Non-oriented research

513. This covers all those appropriations or outlays which are earmarked
for R&D but which cannot be attributed to an objective. A supplementary
breakdown by field of science may be useful.

146 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

12. Other civil research

514. This SEO covers civil research which cannot (yet) be classified to a
particular SEO.

13. Defence
515. This SEO covers research (and development) for military purposes. It
also includes basic research and nuclear and space research financed by
ministries of defence. Civil research financed by ministries of defence, for
example in the fields of meteorology, telecommunications and health, should
be classified in the relevant SEOs.

Table 8.1. Standard key between NABS 1992

and previous OECD GBAORD objectives
NABS categories Previous OECD categories

1. Exploration and exploitation of the Earth 8. Exploration and exploitation of Earth and atmosphere
2. Infrastructure and general planning of land use 4. Development of the infrastructure
Transport and telecommunication systems (2.4 + 2.5) 4.1. Transport and telecommunications
Other infrastructure (2 less 2.4 and 2.5) 4.2. Urban and rural planning
3. Control and care of the environment 5. Sub-total environment
5.1. The prevention of pollution
5.2. Identification and treatment of pollution
4. Protection and improvement of human health 6. Health (excluding pollution)
5. Production, distribution and rational utilisation 3. Production and rational use of energy
of energy
6. Agricultural production and technology 1. Development of agriculture, forestry and fishing
7. Industrial production and technology 2. Promotion of industrial development technology
8. Social structures and relationships 7. Social development and services
9. Exploration and exploitation of space 10. Civil space
10. Research financed from general university funds 9.2. General university funds
11. Non-oriented research 9.1. Advancement of research
12. Other civil research
13. Defence 11. Defence
12. Not specified

Source: OECD.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 147

GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

Table 8.2. Standard key between NABS 1992

and Nordforsk GBAORD objectives
NABS categories Nordforsk categories

1. Exploration and exploitation of the Earth 13. Exploration and exploitation of Earth
and atmosphere
2. Infrastructure and general planning of land use
Transport and telecommunication systems (2.4 + 2.5) 4. Transport and telecommunication
Other infrastructure (2 less 2.4 and 2.5) 5. Living conditions and physical planning
3. Control and care of the environment 6. Combating pollution and physical planning
4. Protection and improvement of human health 7. Preventing and combating disease
5. Production, distribution and rational utilisation of energy 3. Production and distribution of energy
6. Agricultural production and technology 1. Agriculture, forestry, hunting and construction
and services
7. Industrial production and technology 2. Mining, trade and industry, building and
construction and services
8. Social structures and relationships 10. Education
Education, training, recurrent education and training (8.1) 9. Culture mass media and leisure
Cultural activities (8.2) 11. Working conditions
Improvement of working conditions (8.4) 8. Social conditions
Management of business and institutions, social security 12. Economic planning and public administration
systems, political structure of society, social change, social
processes and social conflicts (8 less 8.1, 8.2 and 8.4)
9. Exploration and exploitation of space 15. Space research
10. Research financed from general university funds 14. General advancement of knowledge
11. Non-oriented research 14. General advancement of knowledge
12. Other civil research
13. Defence 16. Defence

Source: OECD.

8.7.5. Principal areas of difficulty

Exploration and exploitation of space
516. This is not a purpose in its own right for most OECD countries, as such
R&D is usually undertaken for another purpose, such as non-oriented
research (astronomy) or for specified applications (e.g. telecommunications
satellites). Nevertheless, it has been maintained, as it cannot be deleted
without seriously affecting the distribution among the objectives to which it
would be reallocated for the few OECD countries that have major space
programmes.

Mining
517. Both Nordforsk and NABS agree that R&D related to prospecting
should be included in “Exploration and exploitation of the Earth”. However,
they part company on mining. According to NABS, fuel mining and extraction
belong in “Production, distribution and rational utilisation of energy”, but

148 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

mining of non-energy minerals belongs in “Industrial production and

technology”; according to the Nordforsk classification, all R&D in favour of the
m in ing in dus try sh ould be included in “Industria l production and
technology”. In the 1993 OECD distribution list, the problem of the treatment
of mining and prospecting was mentioned, and when reporting to the OECD,
“independent” member countries (i.e. those who do not use either Nordforsk
or NABS) have tended to include most or all mining R&D in “Exploration and
exploitation of the Earth” and were invited to mention specifically the
treatment of mining R&D.

Construction
518. A further difference occurs with respect to construction. Logically, if
one applies main-purpose analysis with the aid of the “derivation” convention
(see Section 8.7.1), construction R&D programmes should be broken down
according to their main aim (missile silos in “Defence”, hospitals in
“Protection and improvement of human health”, agricultural buildings in
“Agricultural production and technology”, etc., and R&D in favour of the
building industry in “Industrial production and technology”). This would leave
a residual problem of where to classify construction R&D not elsewhere
classified (n.e.c.). However, NABS has taken the approach that construction
R&D should not be considered as derived except for “defence” and “space”
programmes. According to NABS, R&D on construction materials belongs in
“Industrial production and technology”, but general construction R&D in
“Infrastructure and general planning of land use”; according to Nordforsk,
construction R&D is included in “Industrial production and technology”. The
treatment of construction R&D also appears to vary in the “independent”
countries. Here again, it is important to specify the approach used.

Production, distribution and rational utilisation of energy

519. The series of data collected and issued by the OECD Directorate for
Science, Technology and Industry for GBAORD for the objective “Production,
distribution and rational utilisation of energy”, as defined in Section 8.7.4,
should not be confused with the special series collected and issued by the
International Energy Agency (IEA) of the OECD, which covers energy research,
development and demonstration expenditures, or “RD&D”, a somewhat
broader concept.

8.8. Main differences between GBAORD and GERD data

520. Users of the GBAORD often discover and have difficulty in
understanding differences between the sums reported as:
– Total GBAORD and government-financed GERD.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 149

GOVERNMENT BUDGET APPROPRIATIONS OR OUTLAYS FOR R&D...

– GBAORD for a given objective and total R&D expenditure on the same
objectives as discussed in Chapter 4, Section 4.5. The variations in the sums
reported spring from differences in the specifications of the data.

8.8.1. General differences

521. In principle, both series should be established on the basis of the same
definition of R&D, should cover R&D in both NSE and SSH and both current
and capital expenditures.
522. They differ in two main respects. First, government-financed GERD
and GERD objectives data are based on reports by R&D performers, whereas
GBAORD is based on reports by funders. Second, the GERD-based series cover
only R&D performed on national territory, whereas GBAORD also includes
payments to foreign performers, including international organisations.
523. Differences may also occur because the periods covered are different
(calendar or fiscal years), because the money is finally spent by the performer
in a later year than the one in which it was committed by the funder, and
because the performer may have a different and more accurate idea of the
R&D content of the project concerned.

8.8.2. GBAORD and government-financed GERD

524. In addition to the general differences, government-financed GERD
should include R&D financed by central (or federal), provincial (or state) and
local government, where as GBAORD excludes local government and
sometimes also provincial government.

8.8.3. GBAORD and GERD by socio-economic objectives

525. GBAORD covers only R&D financed by government (including abroad),
whereas GERD covers all sources of funds on national territory.
526. The performer’s appreciation of the objectives of the project concerned
may differ significantly from that of the funder, notably for R&D funded from
block grants such as GUF, which should be distributed by objective in the GERD
approach.

150 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 1

Annex 1
Brief History and Origins of the Present Manual

Origins
1. Encouraged by the rapid growth of the amount of national resources
devoted to research and experimental development (R&D), most OECD
member countries began to collect statistical data in this field around 1960. In
doing so, they followed the pioneering efforts of a small number of countries,
including the United States, Japan, Canada, the United Kingdom, the
Netherlands and France. However, they encountered theoretical difficulties
when starting R&D surveys, and differences in scope, methods and concepts
made international comparisons difficult. The need for some attempt at
standardisation of the kind undertaken for economic statistics was
increasingly felt.
2. The OECD’s interest in this question dates back to the Organisation for
European Economic Co-operation (OEEC). In 1957, the Committee for Applied
Research of the European Productivity Agency of the OEEC began to convene
meetings of experts from member countries to discuss methodological
problems. As a result, an ad hoc group of experts was set up, under the
auspices of the Committee for Applied Research, to study surveys of research
and development expenditure. The Technical Secretary of the Group,
Dr. J.C. Gerritsen, prepared two detailed studies on the definitions and
methods employed to measure R&D in the government sector of the United
Kingdom and France and later of the United States and Canada. Other
members of the group circulated papers describing the methods and results of
surveys in their own countries.

First edition
3. When the Directorate for Scientific Affairs took over the work of the
European Productivity Agency in 1961, the time was ripe for specific proposals
for standardisation. At a meeting in February 1962, the Ad Hoc Group decided
to convene a conference to study the technical problems of measuring R&D. In
preparation, the Directorate for Scientific Affairs appointed a consultant,
Mr. C. Freeman, to prepare a draft document; the document was circulated to

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 151

ANNEX 1

member countries in the autumn of 1962 and revised in the light of their
comments. The “Proposed Standard Practice for Surveys of Research and
Development” (OECD, 1963) was discussed, revised and accepted by experts
from the OECD member countries at the conference, which was held in
Frascati, Italy, in June 1963.
4. Later in 1963, the OECD Directorate for Scientific Affairs invited the
United Kingdom’s National Institute for Economic and Social Research to
undertake an experimental comparison of research efforts in five western
European countries (Belgium, France, Germany, the Netherlands and the
United Kingdom), the United States and the USSR. While the study (Freeman
and Young, 1965) was based on statistics from surveys undertaken before
international standards had been decided, it also tested the first draft
definitions. It concluded that the available statistical information left a great
deal to be desired. The main improvements suggested were:
– A more rigorous conceptual separation of research and experimental
development and “related scientific activities”.
– Careful studies in the higher education sector to estimate the proportion of
time devoted to research by teaching staff and postgraduate (PhD level)
students.
– A more detailed breakdown of R&D manpower and expenditure data to
permit, inter alia, a more exact calculation of research exchange rates.
– A more systematic measurement of expenditure flows between R&D
sectors.
– More data on flows of technological payments and on international
migration of scientific personnel.
5. In 1964, following the acceptance of the Frascati Manual by member
countries, the OECD launched the International Statistical Year (ISY) on
Research and Experimental Development. Member countries returned data
for 1963 or 1964. Seventeen countries took part, many of them conducting
special surveys and enquiries for the first time (OECD, 1968).

Second edition
6. Following the publication of the Statistical Year findings, the OECD
Committee for Science Policy requested the Secretariat to prepare a revision of
the Frascati Manual in the light of the experience gained. An outline of
suggestions was circulated to member countries in March 1968. A draft
revision, incorporating most of these suggestions, was examined at the
meeting of national experts held in Frascati in December 1968. For this
revision, particular attention was paid to making the Manual conform, as far
as possible, to existing United Nations’ international standards such as the

152 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 1

System of National Accounts (SNA) and the International Standard Industrial

Classification (ISIC). A revised draft was examined by a small group of experts
in July 1969, and a revised version of the Manual was published in
September 1970 (OECD, 1970).

Third edition
7. The second revision of the Manual was influenced by two series of
events. First, by 1973, member countries had participated in four ISY surveys,
and data accuracy and comparability had benefited greatly from this
continued experience. National survey techniques had also greatly improved.
Second, in 1972 the OECD Committee for Scientific and Technological Policy
(CSTP) set up the first ad hoc review group on R&D statistics under the
chairmanship of Mr. Silver (United Kingdom) to advise it and the Secretariat
on how to make optimal use, over the short term, of the restricted resources
available for R&D statistics at the OECD while taking account of member
countries’ priorities. Member countries were asked to draw up an inventory of
their needs, and nearly all responded. In addition to giving absolute priority to
a continuation of the ISY surveys, they made a number of recommendations
touching on methodology, notably concerning the need for closer contacts
between the OECD and other international organisations.
8. As a result, the third edition of the Frascati Manual went more deeply into
some subjects and addressed new ones. The scope of the Manual was expanded
to cover research in the social sciences and humanities, and greater stress was
placed on “functional” classifications, notably the distribution of R&D by
“objectives”. A draft was discussed at a meeting of experts held at the OECD in
December 1973, and the final text was adopted in December 1974 (OECD, 1976).

Fourth edition
9. For this edition, national experts recommended undertaking only an
intermediate revision exercise, with no significant changes to be made in key
concepts and classifications. The main stress was to be placed on improving
drafting and layout. However, a number of revisions were in fact made to reflect
the recommendations of the second ad hoc review group on R&D statistics, which
met in 1976 under the chairmanship of Mr. J. Mullin (Canada), the experience
gained by the OECD Secretariat from its international surveys and analytical
reports and suggestions from national experts on R&D statistics. Revision
proposals were presented at the annual meeting of national experts in
December 1978. A small ad hoc group of experts met at the OECD in July 1979 for
more detailed discussions of a draft prepared by a consultant. A revised version
incorporating the Group’s and the Secretariat’s suggestions was discussed in
December 1979, and the text was finally adopted in autumn 1980 (OECD, 1981).

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 153

ANNEX 1

The higher education supplement

10. The higher education sector does not appear in the System of National
Accounts (SNA) adopted by the United Nations and the OECD. The OECD and
UNESCO, however, introduced the sector early in their collection of R&D statistics
because of policy makers’ interest in the role of universities and other third-level
colleges and institutions in national research efforts. Nonetheless, the problems
associated with the collection of accurate data for this sector are significant. They
were discussed at the seminar on S&T indicators for this sector held at the OECD
in June 1985. The experts felt that, while the Manual gave general guidance, it
sometimes gave insufficient practical advice. Therefore, at their annual meeting
in December 1985, the Group of National Experts on Science and Technology
Indicators (NESTI) agreed to prepare a supplement to the Frascati Manual to
address these problems and to make recommendations on improving future
survey practice. A first draft was discussed in December 1986, and the amended
text was then adopted by NESTI and, subject to some final adjustments, it was
recommended for derestriction in December 1987 (OECD, 1989b). Certain of its
recommendations are also relevant to other sectors of performance. The
supplement remains valid, although many of its recommendations were
integrated into the fifth edition of the Manual.

Fifth edition
11. By the late 1980s, it was clear that the Frascati Manual guidelines
needed to be revised to address changes in policy priorities and obtain the
data needed to inform the policy-making process. Many issues were involved,
notably developments in the S&T system and our understanding of it. Some of
these issues emerged in the context of the OECD’s Technology-Economy
Programme – TEP (e.g. internationalisation, software, transfer sciences, etc.).
Others concerned data on environmental R&D, analytical needs for R&D data
that can be integrated with other economic and industrial series and the
revisions of the international standards and classifications applied to R&D
statistics in the Manual.
12. In consequence, the Italian authorities volunteered to organise an
expert conference to discuss proposals for revisions to the Frascati Manual. The
conference took place in Rome in October 1991. It was hosted by the Italian
Ministry for Universities and Scientific Research. For the first time, experts
from the eastern European countries attended.
13. Following the conference, a draft revised version of the Manual,
incorporating much of the text of the supplement on higher education, was
formally discussed by NESTI at their April 1992 meeting. After further revision
by a small editorial group in light of the recommendations made there, the
draft was adopted early in 1993 (OECD, 1994a).

154 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 1

Sixth edition
14. The rationale for undertaking a fifth revision of the Frascati Manual
included the need to update various classifications and an increasing need for
data on R&D in the services sector, on the globalisation of R&D and on human
resources for R&D. Various benchmarking projects have also increased the
need for comparable data.
15. NESTI took the decision to revise the Frascati Manual at its 1999 meeting,
and various topics for revision were discussed at a special meeting in March 2000.
At the meeting in 2000, 19 topics were identified for further investigation. For
each of these, a small group was established, with a lead country or the OECD
Secretariat in charge of the work. The groups’ reports were discussed at a meeting
hosted by the Italian authorities in Rome in May 2001. At the subsequent NESTI
meeting in Rome, decisions were taken on the substantial revisions to be made.
Proposals for changes in wording were discussed at a meeting in October 2001.
The revised Manual was adopted in late 2002. The sixth edition of the Manual is
published in both paper and electronic versions.

Main changes in the sixth edition

16. In this edition of the Manual, an explicit effort has been made to
strengthen various methodological recommendations. As in the previous
revisions, recommendations in the National Accounts are followed as far as
possible and feasible for the purposes of R&D surveys. Some of the
recommendations made in this edition result from needs to bring R&D
statistics closer to the National Accounts.
17. Chapter 1 contains new sections on R&D in software and services, the
SNA, globalisation and co-operation on R&D and topics of special interest:
health, biotechnology, information and communication technology (ICT).
18. Chapter 2 has a new section on R&D in software, the social sciences
and services. The discussion of services is completely new and includes some
examples of R&D. The software and social sciences sections have been partly
revised to integrate material previously distributed elsewhere in the chapter.
19. In Chapter 3, the classification by type of institution in the business
enterprise sector has been changed. Sector definitions have remained
unchanged, but some recommendations are made concerning borderline
institutions in the higher education sector.
20. Chapter 4 has received additional information on the concept of basic
research. Examples of the type of R&D in the financial services industry have
been added. A more explicit recommendation is made on the use of the
product field classification in the business enterprise sector, at least for ISIC
Rev. 3, Division 73.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 155

ANNEX 1

21. Chapter 5 has been restructured into two main parts: one on coverage
and definition of R&D personnel and one on measurement issues and data
collection. The recommendation to collect headcount data in addition to FTEs
has been strengthened. Further guidelines for compiling FTEs are given. The
recommendation to report data by gender and age (with a proposed
classification by age) is new.
22. Chapter 6 offers more detailed recommendations on sources of funds
and the breakdown of extramural expenditure. The need for sources of funds
to be directly related to expenditure for R&D in a given period has been
clarified. Software acquisitions have been added as an investment item in line
with the new SNA.
23. Chapter 7 has been quite substantially modified. The main aim is to
give more specific recommendations on survey methods in the business
enterprise sector and on various estimation issues. An attempt is also made to
make the text clearer and more relevant for R&D surveys.
24. Some additional recommendations adopted by Eurostat since the last
revision of the Manual have been integrated into Chapter 8, and NABS has
been adopted as the basic classification by socio-economic objective. Several
other concepts and methodological issues have also been clarified.
25. New annexes have been introduced on R&D in some specific fields of
interest, such as ICT, health and biotechnology. One annex contains guidelines
on the regionalisation of R&D variables. The decision tree for sectoring has
been introduced in Chapter 3 and there are examples of software R&D in
Chapter 2. Most annexes in the previous version of the Manual have been
updated and further developed.

Acknowledgements
26. All the editions of the Manual have been prepared in co-operation
between experts from member countries and international organisations,
notably UNESCO, EU and Nordforsk/the Nordic Industrial Fund, and the OECD
Secretariat, especially Ms. A.J. Young and the late Y. Fabian (for the first four
editions). Particular debts of gratitude are due to the National Science
Foundation, which pioneered the systematic measurement of R&D.
27. Among those who must be mentioned in connection with the first
edition of the Manual are the late Dr. J. Perlman, Professor C. Freeman and the
French Délégation générale à la recherche scientifique et technique (DGRST).
28. The late H.E. Bishop chaired the 1968 Frascati meeting, and
Mr. H. Stead (Statistics Canada), Mr. P. Slors (Netherlands Central Bureau of
Statistics) and Dr. D. Murphy (Irish National Science Council) made major
contributions to the second edition.

156 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 1

29. Among those who helped to prepare the third version, thanks are due to
the late K. Sanow (National Science Foundation), Mr. J. Mitchell (Office of Fair
Trading, United Kingdom) and Mr. K. Perry (United Kingdom Central Statistical
Office), as well as to Mrs. K. Arnow (National Institutes of Health, United States),
Chairman of the 1973 meeting of experts, and to the chairmen of special topics,
Mr. T. Berglund (Swedish Central Statistical Office), Mr. J. Sevin (DGRST) and
Dr. F. Snapper (Netherlands Ministry of Education and Science).
30. The fourth edition owed a great deal to the work of Mr. H. Stead
(Statistics Canada). Chairing the various expert meetings involved were
Mr. G. Dean (Central Statistical Office, United Kingdom) in 1978 and Mr. C. Falk
(National Science Foundation, United States) in 1979.
31. The Higher Education Supplement was prepared by Ms. A. FitzGerald
(EOLAS – Irish Science and Technology Agency, Ireland). The section on time-
budget studies drew heavily on work by Mr. M. Åkerblom (Central Statistical
Office of Finland). The 1985 Conference on S&T Indicators for the Higher
Education Sector was chaired by Mr. T. Berglund (Statistics Sweden).
32. The fifth edition was largely prepared by Ms. A. FitzGerald (EOLAS) on
the basis of work by a large number of national experts. Particular thanks are
due to Mr. T. Berglund (Statistics Sweden), Mr. J. Bonfim (Junta Nacional de
Investigaçao Cientifica e Tecnologica, Portugal), Ms. M. Haworth (Department
of Trade and Industry, United Kingdom, Mr. A. Holbrook (Industry, Science and
Technology Canada, Canada), Mr. J.F. Minder (Ministère de la Recherche et de
la Technologie, France), Prof. F. Niwa (National Institute of Science and
Technology Policy, Japan), Dr. E. Rost (Bundesministerium für Forschung
und Technologie, Germany), Mr. P. Turnbull (Central Statistical Office,
United Kingdom), and Mrs. K. Wille-Maus (Norges allmennvitenskaplige
forskningråd, Norway). Mr. G. Sirilli (Consiglio nazionale delle ricerche, Italy)
was Chairman of the Group of National Experts on Science and Technology
Indicators during this period and also organised the Rome Conference.
33. The present sixth edition was largely prepared by Mr. M. Åkerblom
(Statistics Finland; OECD Secretariat for the drafting phase) on the basis of
work on specific topics by a large number of national experts. Particular
thanks are due to Mr. D. Byars (Australian Bureau of Statistics); Ms. D. Francoz
(Ministère de la Recherche et de la Technologie, France); Mr. C. Grenzmann
(Stifterverband, Germany); Mr. J. Jankowski (National Science Foundation,
United States); Ms. J. Morgan (ONS, United Kingdom); Mr. B. Nemes (Statistics
Canada); Mr. A. Sundström (Statistics Sweden), Mr. H. Tomizawa (NISTEP,
Japan); Ms. A.J. Young (consultant to Statistics Canada). Mr. G. Sirilli (Consiglio
nazionale delle ricerche, Italy) was Chairman of the Working Party of National
Experts on Science and Technology Indicators during this period and also
organised the Rome Conference.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 157

ANNEX 2

Annex 2
Obtaining Data on R&D in the Higher Education Sector

Introduction
1. Obtaining data on R&D in the higher education sector presents special
problems, which this annex attempts to explain in some detail. The discussion
mainly draws on methodological work from the middle of the 1980s which led
to a special supplement to the fourth edition of the Frascati Manual (R&D
Statistics and Output Measurement in the Higher Education Sector, OECD, 1989b).
2. Time-use surveys or, if these are not possible, other methods of
estimating shares of R&D (R&D coefficients) in total activities in the higher
education sector are a necessary basis for statistics. They are described below.
3. The use of coefficients based on these methods to estimate R&D
expenditure and personnel based on information on total activities in
universities is discussed next, along with some other measurement issues.

Time-use surveys and other methods of estimating shares of R&D

in total activities in the higher education sector
General
4. Member countries use various kinds of time-use surveys or other
methods to establish a basis for identifying the share of R&D in total
university activities (i.e. for calculating R&D coefficients). R&D coefficients are
fractions or proportions of the statistics covering the higher education sector’s
total resources. They serve as a tool for calculating/estimating the shares of
personnel and expenditure data attributable to R&D.
5. Caution must be exercised when using time-use surveys in higher
education surveys. Staff in institutions of higher education combine research
with a range of other duties, such as teaching, administration and
supervision. It may therefore be difficult for respondents to identify
unambiguously that part of their time (working or otherwise) that is devoted
exclusively to R&D. First, several survey methods that may help to minimise
some of the problems raised by such estimations are outlined. Then, other
ways of establishing R&D coefficients are described.

158 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 2

Methods for time-use surveys

6. When choosing the most appropriate survey method, the following
factors have to be considered:
– The resources available to the producers of statistics.
– The desired level of quality of the statistics.
– The burden that can reasonably be laid on university administrations and
individual respondents.
– Special features of the country.
7. Two methods for time-use studies can be distinguished:
– Those based on researchers’ own evaluation of the distribution of their
working time.
– Those based on estimates by the heads of university departments or
institutes.

Methods based on respondents’ own evaluation of the distribution of their

working time
8. These methods can be divided according to the period covered by the
survey:
– Surveys on the distribution of working time during the whole year.
– Surveys on the distribution of working time during one or several specified
weeks.
Surveys on the distribution of working time during the whole year by means
of partial special “rolling” surveys of a specific sample of the population every
week during the year.

• Surveys on the distribution of working time during the whole year

9. In this type of survey, questionnaires can be sent to all individual staff

members or only to a representative sample. The survey may cover the entire
higher education sector or a representative institutional sample. Respondents
are asked to estimate the allocation of their working time over a whole year
according to various categories of work-related activities. In recent surveys
undertaken by member countries, the number of activities has varied from the
two categories “research” and “other” to as many as 15 categories covering all
aspects of a working year. It is recognised that respondents may find it difficult
to recall their pattern of work and to reply accurately to the questionnaire.
10. The following is an example of a use-of-time classification but,
depending on the institutions examined, other activities may be suggested:
– Undergraduate time.
– Postgraduate course-work time.
– Postgraduate research time.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 159

ANNEX 2

– Personal research time.

– Administration.
– Unallocable internal time.
– External professional time.
11. Such questionnaires frequently also include questions on more
general topics such as respondents’ educational background, age, sex,
hindrances to R&D, membership on committees, etc.

• Surveys on the distribution of working time during one or several

specified weeks

12. Questionnaires may be sent to all staff or to a representative sample of

staff. The questionnaire is in the form of a diary in which the respondents
mark, according to the list presented, the activity that best represents the use
of each hour or half-hour of each day.
13. Staff members included in the survey may be asked to maintain
diaries for three short periods of the academic year, such as:
– A normal teaching week.
– A vacation week that falls outside a personal holiday period.
– An examination period week.

• Surveys by means of partial special surveys every week during

the year

14. It is assumed to be very difficult for academic staff to give accurate,

detailed information on how they spend their time when the questionnaire
covers more than one week. A method has therefore been developed, which
uses partial surveys of a “rolling” sample of respondents for one week to
estimate the time allocation pattern for the whole year. Sampling consists of
choosing individuals out of the total population to be surveyed and assigning
one or several particular survey week(s) to each person chosen in order to
cover the whole year. This information is then used to calculate/estimate the
corresponding R&D personnel and expenditures series.
15. The method involves the following broad steps prior to sending out the
questionnaires:
– Defining the survey population.
– Drawing a sample from the population if a full survey is not made.
– Assigning one (or several) survey week(s) to each person included in the survey.
16. Countries take various approaches to acquiring information in this
type of survey. Respondents are sometimes asked to indicate the number of
hours spent on various activities over the entire week; they are also
sometimes asked to reply for each day of the week.

160 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 2

17. While countries offer their respondents different options, the general
principle is always to list all possible work-related activities and ask
respondents to identify how much time they spend on them (in absolute or
relative terms).
18. General information of the kind referred to in paragraph 11 may also
be collected as part of the survey.
19. All survey methods based on responses from individual staff members
are comparatively expensive, and surveys of this kind are often undertaken at
rather long intervals.

Methods based on estimates by heads of university institutes

20. It is usually not possible to gather full information on R&D activities in
the higher education sector without obtaining data from the university
institutes. In most countries, R&D statistics for the higher education sector are
based on a combination of information obtained at central administrative and
institute level and information supplied by individual staff members. The
questionnaires addressed to the institutes often contain questions on certain
types of expenditures and other total resources available and on the estimated
share of R&D in these resources.
21. Several countries have found it convenient to include questions on
time use at a more aggregate level in a questionnaire addressed to the
university institutes, rather than make time-use studies of individual
researchers. The method is cheaper than those described above and puts a
less heavy burden on respondents. In this case, the questionnaires are usually
addressed to the head of the institute, who is assumed to have the knowledge
of ongoing activities needed to supply sufficiently accurate estimates.
However, consultations with individual staff members are also often
necessary to prepare the best possible estimates.

Treatment of borderline R&D activities

22. Respondents to time-use surveys need clear instructions if accurate
and comparable results are to be obtained. Therefore, the surveyor must state
very clearly which activities should be included in the R&D reported and
which should not. Clear definitions must be given in the guidelines when
respondents are asked to distribute their own activities. The recommended
guidelines given in Chapter 2 of the Manual should be followed.

Response rates
23. Methods based on estimates obtained from the university institutes
place virtually no burden on the individual researcher (or other categories of
respondents) but a modest one on the university institute itself. The diary

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 161

ANNEX 2

exercise makes rather heavy demands on the academic staff but none on the
university institute. The burden on the individual respondent is smaller in
surveys when he/she only has to indicate the distribution of time over the
whole year.
24. Response rates are generally comparatively low for diary exercises
covering one or several weeks. They are usually higher when respondents
reply for the whole year. On the other hand, response rates for surveys
addressed to the university institutes are often close to 100%.

Methods based on other sources

25. While surveys are the most systematic and accurate way of collecting
information on time use, they are not always suited to the resources and/or
needs of individual countries. They require a great deal of time and money
and can make heavy demands on the resources of producers of statistics.
Large countries, in particular, may find it difficult to carry out detailed time-
use surveys, given their many higher education institutions and researchers.
26. In addition, the formulation of education and research policy in some
countries may not require information at the level of detail available from
time-use surveys.
27. Therefore, alternative data collection methods are needed to
accommodate resource constraints and meet information needs.
28. Non-survey-based R&D coefficients are derived in a number of ways,
ranging from informed guesses to sophisticated models. Whatever the
method used, they may be an alternative to the more costly large-scale
surveys of researchers and/or higher education institutions described above.
29. The accuracy of the coefficients depends on the quality of the
judgement used in calculating them; the accuracy of the resulting estimates
depends on the quality of the data to which they are applied and the detail
available for both data and coefficients.
30. Coefficients should be prepared to match the level of detail available for
the data and needed for the statistics. They may be derived in several ways,
depending on the information available to the responsible statistical unit. It is
essential that experienced and knowledgeable persons participate in the work.
31. A variety of relevant information will normally be available.
Employment contracts may specify time allowed for some activities; the job
descriptions of some categories of employee may provide useful input. Some
institutions may have established full or partial coefficients for their own
planning or evaluations; other countries with similar education systems may
have derived relevant coefficients.

162 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 2

32. Coefficients derived for calculating overall R&D activity can sometimes
be validated by comparison with the results of time-use surveys of other
countries with similar higher education structures.
33. The use of models to derive research coefficients is a relatively new
activity resulting from the increased computerisation of information on the
higher education sector. Different models are developed by applying different
coefficients to weighted or unweighted higher education data.

Use of coefficients to estimate R&D expenditure and R&D

personnel
34. The aim of the time-use studies and other methods described above is
to obtain a basis for distributing total university resources among research,
teaching and other activities (including administration). These studies are
therefore only the first step in establishing R&D statistics. The next step is to
derive the universities’ total resources; nowadays, this is often done on the
basis of various administrative sources. The final step is to use R&D
coefficients to estimate the shares of R&D in total personnel and expenditure
resources and to break these down into more detailed categories.
35. Thus, to establish R&D statistics for the higher education sector, it is
necessary to estimate:
– The sector’s total available resources, both personnel and financial.
– The corresponding R&D expenditure by type of cost.
– The corresponding R&D expenditure by source of funds.

Total resources
36. Calculations of R&D resources are based on data on total available
resources by applying the R&D coefficients derived from time-use studies or
other sources. Total data include general university funds (GUF) and a variety
of external sources and may be derived from:
– University accounts.
– Administrative records.
– Additional breakdowns made by universities’ central administrations on
the basis of general accounts and registers.
– Surveys addressed to university institutes.
– Other statistical systems (statistics on public servants, general wage
statistics).
37. In many cases, total data are derived from various administrative
sources. The role of central administrations varies from country to country and
from level to level – nationally at the ministry of education, regionally, locally or

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 163

ANNEX 2

within the higher education institute itself. Regardless of their level, such
centres usually have a vast quantity of information as a result of their
administrative activities. The information held by central administrations,
while not necessarily specifically related to R&D, is a useful source of overall
data from which R&D data can be extracted using either estimated R&D
coefficients or R&D coefficients drawn from time-use surveys. R&D information
is sometimes available directly from central administrations. It is not
completely certain, however, that this information conforms to the definitions
of the Frascati Manual, and this limits the possibilities for using it directly.
38. The information held by central administrations in their files varies
according to the function of the particular administration. Ministries of
education may have very broad overall information, while the finance officers
of higher education institutions may have income and expenditure
information associated with individual researchers and other staff.
39. To identify the R&D in individual disciplines/fields of science may
require information at the researcher level at large institutions carrying out
research in many disciplines. Information at the level of the institution is
sufficient if its R&D is confined to a single field of science.
40. There are several advantages to collecting the data of central
administrations as part of an overall R&D data collection exercise:
– The data are consistent and unambiguous.
– There is no double counting of parameters.
– The data apply to a specific period.
– The data are easily accessible.
– The data form a useful input to the iterative process of model building.
– Use of data from secondary sources lowers the response burden on survey
respondents.
41. There are also limitations to such data, some of which, if not taken
into account, could lead to inaccuracies in the final R&D statistics:
– Incomplete specific data on R&D activities in terms of coverage of costs,
sources of funds and personnel.
– Problems of comparability between different universities.
– Data usually available at a very aggregate level.
– R&D component of general higher education statistics not separately
identified.
42. Countries have access to sufficiently detailed data on total resources
(e.g. broken down by field of science) in different ways. Differences among
universities within a given country in terms of the level of detail available may
also cause variations in countries’ ability to supply sufficiently detailed data to
the OECD.

164 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 2

43. The results of time-use studies are used to derive countries’ full-time
equivalents for R&D from data on total full-time equivalents, which in theory
can be defined in at least two different ways:
– The total amount of work done on R&D by one person in one year.
– The total number of full-time positions on R&D held by one person in one
year, with salary as the criterion.
44. The first corresponds broadly to the definition of FTE given in
Chapter 5, Section 5.3.3. In practice, the second is probably more feasible for
data collection. As in most cases it is not possible to have information on
persons who have several positions, one person may conceivably count for
more than one full-time equivalent.

Type of costs
45. According to Chapter 6, Sections 6.2.2 and 6.2.3 of the Manual, R&D
expenditures should be broken down by current and capital expenditures,
which in turn consist of labour and other current costs, on the one hand, and
instruments/equipment expenditures and land/buildings expenditures, on
the other.
46. If no data are directly available for each of these R&D components for
a certain unit, an estimate must be made on the basis of information on total
expenditure.
47. Labour costs (i.e. salaries and related social costs) usually represent
around half of total R&D expenditure in the higher education sector.
Information on total labour costs is usually available or calculated on the basis
of one or several of the following data sources:
– Point on the salary scale for each researcher, technician or other member of
the staff, and the scale itself.
– Labour costs by category of personnel and institute.
– Labour costs by category of personnel, institute, field of science or
department.
48. R&D coefficients derived from time-use studies are used directly at an
appropriate level (individual, institute, department, university) to estimate the
share of R&D in total labour costs; if necessary, adjustments should be made
to take account of the costs of various associated social security or retirement
schemes.
49. R&D coefficients can be expected to vary according to the teaching or
research discipline, the occupational category of the personnel directly
involved in R&D and the type of institution in which the activity is performed.
At the greatest level of detail, coefficients can be applied to the financial and
personnel data of individual institutions. When this is possible, coefficients

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 165

ANNEX 2

may be modified to reflect the different positions of the institutions in terms

of R&D, for example, whether they are small liberal arts colleges, technical
universities or major teaching and research universities.
50. Coefficients are typically applied in stages:
– R&D coefficients applied to different categories of staff, if possible by
discipline and institution, yield the FTE personnel estimates.
– These personnel estimates, converted to coefficients themselves, may be
applied to financial data to provide R&D expenditure estimates.
51. Information on other current costs is usually available by institute and
often concerns resources at the disposal of the institutes themselves for the
purchase of items such as documents, minor equipment, etc. The institutes
are usually asked to estimate the R&D share of these costs on the basis of
intended use. The part that is not available by institute (overhead costs such
as water, electricity, rents, maintenance, general administration, etc.) has to
be distributed among the institutional units concerned. If intended use is not
feasible as a criterion, the same distribution coefficients as for labour costs
may be used. The shares of R&D may also be determined on the basis of
conventions or the opinion of institutes.
52. Information on total investment in instruments and equipment is
usually available at the level of the institution. In many surveys, the shares of
R&D are estimated by the institutes according to the intended use of the
equipment. R&D coefficients are probably of less use for estimating the shares
of R&D in instruments and equipment than for estimating various types of
current expenditure. The share of R&D in investments in instruments and
equipment may also be based on conventions or on opinion, as for certain
types of other current costs discussed above.
53. Information on total investments in land and buildings is usually available
only at the level of the institute or the university. R&D coefficients are seldom
used to estimate the shares of R&D in these investments. Here again, the R&D
data are often estimated on the basis of the intended use of the facilities.
54. From the above, it may be concluded that R&D coefficients offer the
only way to estimate the share of R&D in labour costs, play a significant role in
estimating shares of R&D in other current costs, but are of minor importance
in calculating shares of R&D in investments in instruments and equipment or
in land and buildings.

Sources of funds
General
55. Funds for R&D in the higher education sector come from many
different sources. The main source in most member countries is traditionally

166 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 2

a proportion of the publicly funded block grant known as public general

university funds (GUF) which higher education institutions receive to support
all activities. The different activities of the staff in higher education
institutions – teaching, R&D, administration, health care, etc. – are not
specifically identified for separate payment from these grants, which, in a
general way, cover the payment of all work-related activities. In addition, R&D
funds are received in the form of grants or contracts from other sources such
as ministries, departments and other public institutions, including research
councils, from private non-profit institutions, and, in recent years,
increasingly from industry and from abroad. Some universities may also have
“own funds” (such as income from endowments, etc.).
56. Time-use studies and other methods used to identify the R&D share of
universities’ total activities usually only concern GUF, which account for the
major part of higher education R&D (HERD). External funds are often for R&D
but may be used for other purposes as well. For each project funded by
external sources, therefore, the survey respondent often has to evaluate
whether or not it funds research, if the information is not available from
central administration registers.
57. Some external funds (especially funds from foundations and research
councils) are not always fully included in the central accounting records of the
universities. Some research contracts may in fact go directly to the university
institute or to individual professors. To obtain as broad coverage as possible,
data on institutes’ external funds have in some cases to be taken from funders’
accounts (although this goes against the Manual’s principle of performer-based
reporting) or should, at least, be double-checked against those accounts.
Funder-based data usually give only expenditures, and the problem of acquiring
the corresponding R&D personnel data is therefore a tricky one.
58. Higher education institutions are looking increasingly to outside
sources to compensate for absolute cuts or levelling off of traditional GUF
resources. In particular, research links with mission-oriented ministries and
industry are being intensified, and the share of total expenditure from non-GUF
resources will ultimately increase. Such links with outside organisations may or
may not be formally identified in the accounts of the institutions and are
therefore difficult to quantify in the collection of R&D statistics. Furthermore,
these transfers of resources may be in kind (in the form of equipment and
materials) rather than money, thus creating additional measuring difficulties.
59. Accounting procedures will therefore largely determine how well the
sources of R&D income can be separately defined and identified. Producers of
R&D statistics are dependent on the detail available in such accounts. A
further complication in identifying the sources of research income is the fact
that outside organisations do not always pay the “full market cost”, however

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 167

ANNEX 2

defined, of the R&D carried out for them in institutions of higher education.
Theoretically, a part of GUF spent on administration and other extra costs for
externally financed research should be counted as research in addition to the
R&D estimated on the basis of the coefficients used above.
60. Problems of accurate coverage of R&D funding sources are common to
all member countries, but the main area of lack of international comparability
is that of distinguishing between GUF and other sources of public R&D income.

Separation of general university funds from other funding sources

61. Some of the problems of identifying what part of these grants is
attributable to R&D have already been discussed above. This identification
process is an intrinsic part of the survey methodology employed in each
country. Inconsistencies arise because different countries classify the R&D
component of GUF differently.
62. Options for classifying such public funds at the sectoral level are:
– General university funds.
– Sector’s own funds.
– Direct government funds.

• General university funds

63. A separate category of GUF has been defined for the higher education
sector to take account of the special funding mechanisms for R&D, as
compared to other sectors. Most member countries are of the view that, as
R&D forms an intrinsic part of the activities of higher education institutions,
any funds allocated to a third-level institution have an inbuilt and automatic
R&D component. On this interpretation, such funds are classified as GUF. In
adding up national totals, these data are usually included in subtotals of
public finance on the grounds that “as government is the original source and
has intended at least part of the funds concerned to be devoted to R&D, the
R&D content of these public general university funds should be credited to
government as a source of funds”, and this is the approach recommended for
international comparisons.
64. GUF should be separately reported and adjustments to the R&D cost
series should take account of real or imputed social security and pension
provisions, etc., and be credited to GUF as a source of funds.

• “Own” funds

65. “In their national publications, a few countries continue to classify the
higher education block grant of public origin not as GUF but as “own funds”,
arguing that “it is within the universities that... the decisions are taken to
commit money to R&D out of a pool which contains both ‘own funds’... and

168 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 2

public general university funds; therefore, the sums concerned should be

credited to higher education as a source of funds”.
66. In this situation, the “own funds” category is a significant source of
funds for R&D, which will be credited to higher education and not included in
public sources when adding national totals.
67. Other monies produced by the sector should be considered as “own
funds”.
68. Although national accounting practices will dictate how easily they
can be identified, such R&D income (“retained receipts”) can, notably in the
case of private universities, be a considerable source of income and should
undeniably be classified as “own funds”.

• Direct government funds

69. In addition to GUF, the government sector provides money for higher
education R&D in the form of earmarked research contracts or research
grants. This source of research income is more readily identified and does not,
in general, pose major problems for the producers of statistics, as they readily
classify it as a direct source of government funds.
70. Adjustments related to “other current costs” to account for real or
imputed payments of rents, etc., should be credited to the category of direct
government funds (see Chapter 6, Section 6.2.2 and 6.3.3 of the Manual).

Recommendations
71. To obtain the best possible international comparability of higher
education R&D statistics, it is preferable to disaggregate the sources of funds
as much as possible; this largely depends on the availability of information
from central accounting records in institutions of higher education.
72. The main problem for international comparability occurs when data for
GUF are not separately reported and are classified by different countries either
with the higher education sector’s “own funds” or with the government sector.
73. Therefore, GUF, insofar as possible, should be reported separately; if
this is not possible, the corresponding funds should be included in “funds
from the public sector” and not in the higher education sector’s “own funds”
or “other higher education funds”.
74. When reporting data to the OECD, member countries are encouraged
to indicate on which sets of expenditure and personnel data coefficients are
applied to calculate R&D data, together with the actual coefficients used.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 169

ANNEX 3

Annex 3
The Treatment of R&D in the United Nations System
of National Accounts

Introduction
1. The aim of this annex is to explain the treatment of R&D in the System
of National Accounts (SNA) to experts on S&T indicators who are unfamiliar
with SNA concepts and terminology. It deals with two topics:
– History of the relationship between the SNA and Frascati Manual systems.
– Similarities and differences between the two systems:
❖ General inclusion of R&D in the SNA.
❖ Sectors and their sub-classifications.
❖ Measuring R&D spending in the SNA.
2. References are generally to the latest 1993 version of the SNA,
prepared jointly by the Commission of the European Communities, the
International Monetary Fund, the OECD, the United Nations and the World
Bank (CEC et al., 1994). The 1968 version is only mentioned when significant
changes in treatment have occurred between the two versions.

History of the relationship between the two systems

3. The United Nations System of National Accounts was first published
in 1953. It provided a coherent framework for recording and presenting the
main flows relating to production, consumption, accumulation and external
trade. Along with the associated United Nations (UN) international
classifications, such as the International Standard Industrial Classification
(ISIC), it is the standard framework for economic statistics and analysis in
OECD member countries and is used as such by the OECD.
4. The Frascati Manual system of R&D accounts was established in 1961,
largely on the basis of work by the United States dating back to the formative
years of the SNA. The system was inspired by the SNA and adopted the idea of
dividing the economy into sectors and of measuring flows of funds between
them, but was never conceived of as part of the SNA.

170 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 3

5. Three main areas of difference have continued to exist between the

two systems:
– Economic sectors and associated classifications.
– Terminology, i.e. use of the same term for different concepts or different
terms for the same concept.
– Basic differences in accounting methods.
6. These differences between the SNA and the Frascati Manual have been
systematically reviewed on three occasions: in about 1970 and again in 1990,
when the revision of the two systems coincided, and also in the mid-1970s,
when the concept of satellite accounts for R&D was introduced.
7. On the first occasion, the revision of the SNA was completed in 1968,
prior to the main discussion of the revision of the Frascati Manual. That edition
of SNA paid very little attention to R&D. A small but outspoken group of
national R&D experts stressed the need to bring the second edition of the
Manual in line with the “new” SNA. As a result, sector definitions and
terminology were somewhat changed but differences in accounting methods
remained.
8. The relationship between the Frascati Manual and the SNA was
discussed by various international organisations, including the United
Nations Economic Commission for Europe and the European Commission. As
a result, a system of satellite accounts for R&D was developed and has been
used regularly by a number of member countries, notably France. The need for
satellite accounts was also recognised in the SNA 1993, which contains a
chapter discussing the compilation of satellite accounts in areas of particular
interest, such as R&D.
9. R&D was specifically discussed during the preparation of the SNA 1993
in the context of the possible treatment of “intangible investment” rather than
as intermediate consumption as in the SNA 1968. It was finally decided not to
treat R&D as an investment activity owing to difficulties for implementing
such a revision in practice, but the discussions did lead to the inclusion of
more specific guidelines for R&D than in the preceding version.
10. Furthermore, the national accountants who discussed the revision
became aware of the Frascati Manual, its main recommendations and the
related databases. Changes in the SNA concerning sectors and terminology
have been incorporated in the Frascati Manual wherever appropriate, but
differences in accounting practice remain.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 171

ANNEX 3

Similarities and differences in the treatment of R&D in the Frascati

Manual and the System of National Accounts
General inclusion of R&D in the SNA
11. The SNA is concerned with economic activities. The first question that
must be answered, therefore, is what constitutes an economic activity, since
this determines what falls within the scope of the national accounting system
and thus enters gross domestic product (GDP). There is no difficulty in
defining as economic those activities that result in the production of goods
and services for sale on the market. Government activities in the areas of
public administration, law and order, health, education and social services
(and activities in similar areas carried out by private non-profit organisations)
are also counted as economic, even though their output is not sold on the
market. Borderline problems do, however, arise in connection with some other
kinds of non-market activities. Notably, production of goods by households for
own consumption is included in GDP but not services, except for services of
owner-occupied dwellings. GDP includes, for example, the construction of
buildings by households and enterprises for their own use and the production
of crops and livestock for consumption on the farm. However, by convention,
the SNA does not include unpaid services rendered by household members
such as home decorating, cleaning, laundry, etc.
12. R&D is generally an economic activity as defined above. There is,
however, one category which is not; it is R&D carried out by postgraduate
students who are not employed by higher education institutions but are
supported by grants and/or their own resources. All other R&D expenditures
contained in the Frascati Manual are treated in the various accounts of the
System of National Accounts.
13. While the most recent version of the SNA gives guidelines on the
treatment of R&D, it does not systematically distinguish it in the accounts,
particularly for firms that carry out R&D for own use. This is why satellite
accounts are needed.

Sectors and their sub-classifications

Sectors
14. Both the SNA and the Frascati Manual break down institutional units
into a number of sectors. The broad correspondence is shown in Table 1.
15. Both systems use national territory on the one hand and “rest of the
world” (SNA) or “abroad” (Frascati Manual), on the other.
16. The Frascati Manual applies one set of sector definitions to all its
accounts (R&D expenditure by sector of performance, R&D expenditure by
source of funds, R&D employment). The SNA has two slightly different

172 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 3

Table 1. Summary of sectors in the SNA and in the Frascati Manual

SNA Frascati Manual

Non-financial corporations
Business enterprise sector
Financial corporations
General government Government sector
Non-profit institutions serving households
Private non-profit sector
Households
(Included in other SNA sectors) Higher education sector
Rest of the world Abroad

Source: OECD.

Table 2. Sectors and producers in the SNA

Sectors Market producers Non-market producers

Non-financial corporate sector Non-financial corporations

or quasi-corporations
Non-profit institutions (NPIs) engaging
in market production1
NPIs serving business
Financial corporate sector Financial corporations
and quasi-corporations
General government sector [Government units engaged in market Government units n.e.c.
production]2 Social security funds
NPIs mainly financed
by government n.e.c.
Non-profit institutions serving NPISH
households (NPISH)
Households Unincorporated enterprises engaged Households n.e.c. including
in market production unincorporated enterprises
engaged in production of goods
mainly or wholly for own final use

1. Supplying goods and services at an economically significant price.

2. These are treated as quasi-corporations as long as they have a separate set of
accounts.
Source: OECD.

approaches (see Table 2), and the treatment of R&D in the Frascati Manual,
especially performance, is closer to the second of these.
17. The main difference is that the Frascati Manual separates out the higher
education sector. This separation is considered very important by R&D
statisticians and policy makers, for the reasons given in Chapter 3 of the
Manual. However, this additional sector causes problems in an SNA context.
While public universities and colleges belong in the SNA government sector, the
other components of the Frascati Manual higher education sector may belong
almost anywhere in the SNA. Table 3 shows where they might be classified.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 173

ANNEX 3

Table 3. The SNA sectoring of units definitely and possibly

included in the Frascati Manual higher education sector
Market producers Non-market producers

Teaching establishments All non-financial corporations Government units PHES1

i.e. producing higher education (or quasi-corporations) PHES 1 Non-profit institutions (NPIs)
services (PHES) as a main activity Any unincorporated enterprises controlled and mainly financed
PHES 1 at an economically significant by government PHES1
price NPISHs PHES’
NPIs PHES at an economically
significant price
NPIs serving enterprises PHES 1
University hospitals providing Non-financial corporations (or quasi- Government units PHSS2 CAAHE3
healthcare services (PHSS) controlled corporations) PHSS 2 CAAHE3 NPIs controlled and mainly financed
by administered by or associated with NPIs PHSS 2 at an economically by government PHSS 2 and CAAHE3
higher education (CAAHE) and/or with significant price CAAHE3 NPIs serving households PHSS2
a significant teaching commitment
Research institutes or experimental Non-financial corporations (or quasi- Government units CAAHE3
stations CAAHE3 (“borderline” corporations) selling R&D but CAAHE3 NPIs controlled and mainly financed
research institutions) NPIs selling R&D at an economically by government but associated with HE
significant price CAAHE3 NPISHs which are CAAHE3
NPIs serving enterprises CAAHE3
Postgraduate students supported Households benefiting from subsidies
by grants

1. Providing higher education services.

2. Providing healthcare services.
3. Controlled, administrated by or associated with higher education establishments.
Source: OECD.

18. If the Frascati Manual system had no higher education sector, there
would be an almost complete match between the SNA sector classification
and the R&D sectors, as has been intended since the 1970 version of the
Frascati Manual (OECD, 1970). For example, distribution of private non-profit
(PNP) institutions among sectors in the Frascati Manual is clearly based on the
SNA; and the section of Chapter 4 of the SNA 1993 devoted to this topic
usefully supplements the discussion in Chapter 3 of this Manual.
19. Nevertheless, non-higher education units may be treated somewhat
differently in the latest versions of the Manual and of the SNA, as the Manual
adapted the original SNA definitions to reflect R&D institutional practice. The
institutions are often attributed to sectors by two different agencies which
may interpret the same instruction differently.

Classifications
20. The SNA does not always recommend the same classification as the
Frascati Manual for what the latter refers to as “sector sub-classifications”. Both
use ISIC, but the breakdown of R&D among industries may differ because of

174 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 3

variation in the unit classified and the classification criteria. In the SNA,
government outlays are broken down by the classification of the functions of
government outlays (COFOG); R&D experts have rejected this classification, in
favour of the NABS classification for GBAORD, as they have been unable to
agree on a classification for R&D performed in the government sector. In OECD
national accounts publications, the government and NPISHs are subdivided by
main types of unit (see Table 4), whereas a field of science classification is
recommended in this Manual.

Table 4. SNA classifications of government outlays and final

consumption expenditure of NPI serving households

A. Government outlays1
1. General public services (including basic research)
2. Defence
3. Public order and safety
4. Education (includes universities and colleges)
5. Health
6. Social security and welfare
7. Housing and community amenities
8. Recreational, cultural and religious affairs
9. Economic services
9.1. Fuel and energy
9.2. Agriculture, forestry, fishing and hunting
9.3. Mining, manufacturing and construction, except fuel and energy
9.4. Transportation and communication
9.5. Other economic affairs
10. Other functions
Total

B. Final consumption expenditure of non-profit institutions serving households

1. Research and science
2. Education
3. Medical and other health services
4. Welfare services
5. Recreational and related cultural services
6. Religious organisations
7. Professional and labour organisations serving households
8. Miscellaneous
Total

1. Final consumption expenditure (of which compensation of employees and other

subsidies), other current transfers and property income, gross capital formation
and other capital outlays.
Source: OECD National Accounts, Detailed Tables, Vol. II.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 175

ANNEX 3

Measuring R&D spending in the SNA

21. The SNA and the Frascati Manual also differ in the way they account for
R&D, as they rely on different conceptual frameworks. Moreover, as the
Frascati Manual also serves as a manual for the collection of data, it is more
directly influenced by feasibility concerns. The following description of the
treatment of R&D in the various accounts is largely based on quotations from
the SNA 1993 (CEC et al., 1994).

Identifying and valuing R&D in the production account

22. “Research and development by a market producer is an activity
undertaken for the purpose of discovering or developing new products,
including improved versions or qualities of existing products, or discovering or
developing new or more efficient processes of production. Research and
development is not an ancillary activity, and a separate establishment should be
distinguished for it, when possible. The research and development undertaken
by market producers on their own behalf should, in principle, be valued on the
basis of the estimated basic prices that would be paid if the research were sub-
contracted commercially, but is likely to have to be valued on the basis of the
total production costs, in practice. Research and development undertaken by
specialised commercial research laboratories or institutes is valued by receipts
from sales, contracts, commissions, fees, etc., in the usual way. Research and
development undertaken by government units, universities, non-profit
research institutes, etc., is non-market production and is valued on the basis of
the total costs incurred. The activity of research and development is different
from teaching and is classified separately in ISIC. In principle, the two activities
ought to be distinguished from each other when undertaken within a university
or other institute of higher education, although there may be considerable
practical difficulties when the same staff divide their time between both
activities. There may also be interaction between teaching and research which
makes it difficult to separate them, even conceptually, in some cases.”
(CEC et al., 1994, paragraph 6.142.)
23. The definition of R&D by the SNA as an activity aimed at developing
new products and processes differs somewhat from the Frascati Manual
definition. It characterises R&D only by its purpose, which is innovation,
whereas the major characteristic of R&D according to the Frascati Manual is the
production of new knowledge. While the two approaches are quite similar,
they do not entirely overlap. The SNA definition is closer to the Oslo Manual
definition of innovative activities, which encompasses activities such as
training and tooling up related to innovation, which are not R&D. In addition,
the SNA definition may lead the user to overlook basic research performed by
business, for which innovation is only a very indirect goal.

176 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 3

24. In terms of accounting principles, the SNA recommends measuring

gross output of R&D for market producers (firms that sell R&D) and total cost
for other producers (own account R&D). This is essentially different from the
Frascati Manual concept, which recommends measuring expenditure on R&D.
The major difference between total cost (SNA) and expenditure (Frascati
Manual) is in the treatment of fixed capital: whereas the cost approach counts
consumption of existing fixed capital, the expenditure approach counts the
expenditure (purchase) of new fixed capital. The Frascati Manual applies the
same treatment to intermediate goods, which are measured by purchases
instead of consumption (under the heading “Other current costs”). For gross
output, recommended by the SNA for market producers, it equals total cost
plus an operating surplus and adjustment for net taxes on production
(payments less subsidies) (Table 5).

Table 5. Gross output and total intramural R&D

SNA cost components Frascati Manual cost components

Similar coverage Compensation of employees = Labour costs

Intermediate consumption1 = Other current costs

Different treatment Taxes on production paid, less Subsidies included in above; taxes
subsidies received on production excluded
Consumption of fixed capital Gross capital expenditure
Operating surplus Not included

1. Intermediate consumption also includes the cost of any bought-in R&D.

Source: OECD.

25. There are other, smaller differences in the treatment of fixed capital in
the SNA and the Frascati Manual: i) in the SNA, gross fixed capital formation
(GFCF) on buildings excludes the value of the land on which they are situated,
whereas the Frascati Manual includes land and buildings as capital
expenditures, with no separate identification; ii) disposal, especially sales, of
fixed capital, is not considered in the Frascati Manual and might lead to double
counting, as part of the capital expenditure of one entity would correspond to
a reduction in the capital stock of another. This is hard to measure, and it is
likely to be small in practice.

R&D as intermediate consumption

26. The SNA 1993 gives the following instruction for the R&D of market
producers (the Frascati Manual business enterprise sector):
“Research and development are undertaken with the objective of
improving efficiency or productivity or deriving other future benefits so
that they are inherently investment – rather than consumption-type

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 177

ANNEX 3

activities. However, other activities, such as staff training, market

research or environmental protection, may have similar characteristics.
In order to classify such activities as investment type it would be
necessary to have clear criteria for delineating them from other activities,
to be able to identify and classify the assets produced, to be able to value
such assets in an economically meaningful way and to know the rate at
which they depreciate over time. In practice, it is difficult to meet all these
requirements. By convention, therefore, all the outputs produced by
research and development, staff training, market research and similar
activities are treated as being consumed as intermediate inputs even
though some of them may bring future benefits.
“As already noted, research and development is not an ancillary
activity like purchasing, bookkeeping, storage and maintenance which
tend to be found frequently in all establishments. When research and
development is carried out on a significant scale within an enterprise,
it would be desirable to identify a separate establishment for it so that
the relevant inputs and outputs could be distinguished for analytical
purposes. Because of the difficulty of obtaining price data, the output
will usually have to be valued by total costs of production, as in the
case of most other own-account production. The output produced has
then to be treated as being delivered to the establishment, or
establishments, which make up the rest of the enterprise and included
in their intermediate consumption. When there are several other
establishments, the amounts of research and development delivered
can be distributed in proportion to their total costs, or other indicator,
in much the same way that the output of head offices or other central
facilities has to be allocated.
“When an enterprise contracts an outside agency to undertake research
and development, staff training, market research or similar activities on
its behalf, the expenditures incurred by the enterprise are treated as
purchases of services used for purposes of intermediate consumption.”
(CEC et al., 1994, paragraphs 6.163-6.165.)
27. The consideration of “software and large databases” as an investment
good in the SNA 93 raises particular issues for R&D. The above text mentions
that “all the outputs produced by research and development (…) are treated as
being consumed as intermediate inputs”. This is in fact contradicted by the
capitalisation of own account software production in national accounts, as a
substantial part of own account software consists of R&D (R&D performed in
software companies and R&D on software performed in other companies).
Available figures show that a substantial and increasing share of R&D is in
programming software.

178 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 3

R&D in the expenditure account

28. The Manual distinguishes between performers and funders of R&D.
The SNA distinguishes between the producers and the users of R&D services
(expenditure account). The unit which “performs” the R&D also “produces” it.
The “funder” unit is usually, but not always, the SNA “user”.
29. The funding unit is the SNA user when the money is used to finance
intramural R&D (“own funds”) or to purchase R&D services from another unit.
The funding unit is not the SNA user if it transfers money for R&D performance
to another unit but does not receive a flow of R&D services in return, e.g. all
types of R&D grants and also indirect forms of R&D support. In this case, the
performer is the user. In the case of market producers, any grants, etc., from
government must be treated as “subsidies” (see Table 5). Problems may arise for
the R&D content of procurement contracts. In principle, the R&D is embedded
in the product as it is in other purchases of goods and services, and the SNA
user of the R&D is the producer/performer. If, however, the funding agency
places a separate R&D contract and becomes the owner of the R&D results, then
the funder is the SNA user. When a funding entity other than the government is
not the performer (extramural R&D), the Frascati Manual makes no specific
recommendation regarding the classification of such transfers of funds,
whereas the SNA proposes various categories (income from sales, subsidies,
current transfers and capital transfers), with the idea of gaining a better
understanding of the economic mechanisms at work.
30. While all R&D has a user, only part of R&D appears per se in the final
expenditure account. In the expenditure account, the vast majority of R&D is
treated as being used up in the production process and hence already
incorporated in goods and services. These are either carried forward to a
subsequent period (capital formation) or used without further transformation
to satisfy the individual or collective needs of members of the community
(final consumption). This covers all R&D financed by market producers and
R&D financed by government and non-profit institutions serving households
(NPSH) which contributes directly to the services they supply. The only R&D
activities treated per se as final consumption in the expenditure table are those
financed as a collective service by government (notably basic research) and
similar R&D financed by NPISH.

The need for satellite accounts

31. Satellite accounts are an evolving mechanism for presenting particular
topics as annexes to main national accounts.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 179

ANNEX 3

32. The characteristics of satellite accounts can be described as follows:

“Over time, satellite accounts for particular fields have come to be
associated with the following characteristics:
1. They feature data for a whole field of economic activity and provide
a framework for arraying more comprehensive information about a
field than can be shown in the main accounts.
2. They are purpose-oriented in that the criterion for a transactor’s or
transaction’s inclusion is its linkage to the field.
3. They are articulated with the main accounts and contain at least
one measure that is also in the main accounts.
4. They present information in ways that are different from the main
accounts: definitions, classifications and accounting conventions
may differ from those used in the main accounts in order to provide
the most useful presentation of information about the field. What is
counted as current or capital in the main accounts may be changed,
or the boundary of production may be moved. The definitions,
classifications and accounting conventions must be consistent
within the account, however.
5. They often contain tables that answer several questions: Who is
producing, and what are the means of production? Who is
financing? What is the result of the expense, and who is benefiting
or using the result?
6. They often encompass monetary and physical data in an integrated
fashion. Physical data may relate to production, such as the number
of persons employed in the field or the stocks of equipment.
Physical data may also relate to beneficiaries, such as the number of
persons affected by activities in the field.
The advantage of satellite accounts is that an alternative view of the
economy can be obtained without disturbing the main accounts.”
(Carson and Grimm, 1991)

180 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 4

Annex 4
R&D Related to Health, Information
and Communication Technology (ICT)
and Biotechnology

1. This annex presents three areas of R&D for which it is not possible to
derive information through direct use of the classifications recommended
elsewhere in the Manual. All three are of high policy relevance, and there is a
clear need for data on R&D related to these fields. To obtain the data, it is often
necessary to combine R&D data from various classifications or even to develop
new survey questions.

Deriving data on health-related R&D from regular R&D statistics

Introduction
2. Recently, demand for data on health-related R&D has been particularly
strong. As international comparisons are often requested, this section
provides general guidance on how to compile data on health-related R&D from
existing surveys and more general sources. In this context, “health-related”
refers not only to biomedical research but also to a wider category including
relevant R&D in the social sciences, notably on health services.
3. The aim of the exercise is to establish gross expenditure on R&D (GERD)
for health, broken down by sector of performance and source of funds. There
should be matching R&D personnel data by sector of employment. Guidance is
also given on GBAORD, as those who seek health R&D series often use this source.
Further information on international comparisons and examples of national
efforts can be found in Measuring Expenditure on Health-related R&D (OECD, 2001).
4. In principle, similar compilations could be made for other fields such
as agriculture.

General approach
5. There is demand for a data set covering all health-related R&D, but
regular R&D surveys normally break down expenditure and personnel
according to the primary aim/field/industrial activity of the unit concerned.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 181

ANNEX 4

Furthermore, classifications may not be detailed enough to identify small

categories of health-related units.
6. The process thus has to be to break out the data for categories that are
clearly health-related (core elements) and then to use various adjustment and
estimation methods to refine these data and to add in the health-related
component of other categories. This generally means starting from
institutional classes, for which a full set of data is available (sources of funds,
personnel, etc.) and then using functional data to make the necessary
adjustments. The process will vary among sectors and also among countries
because different institutional and functional breakdowns may be used and
also because data providers have special knowledge of national specificities in
the organisation of health R&D.
7. In principle, the preferred source should be GERD data reported by
performers. In practice, several sources may be used to compile health-related
R&D spending. In some countries, especially those where the collection of
GBAORD data is associated with the general R&D survey, these budget series
(particularly those for which data on first destination are compiled) may identify
central government funds for R&D on health which are not immediately visible in
the survey of performers by socio-economic objective (protection and
improvement of human health) or field of science (medical sciences). Similarly,
useful additional information and data may be gleaned from the reports of
medical charities, health research councils and funds and even from the reports
of pharmaceutical industry associations. Building up a reasonable picture of
GERD for health may involve mixing and matching data from a variety of sources.

Identifying health-related R&D in GBAORD

8. Those seeking data on government funding of health-related R&D are
often drawn to GBAORD because there is a specific category of socio-economic
objective for this topic. However, they may not realise that this category only
covers R&D whose primary purpose is the protection and improvement of
human health (NABS 4) and that funds for relevant activities may be included
in other categories.
9. The most important additional category is “General university funds
and non-oriented research”. The core coverage recommended for health in
GBAORD is therefore:
– Health.
– General university funds and non-oriented research: medical sciences.
10. Health-related research funded for other objectives, for example
military medical research, health and safety research at nuclear
establishments or support for relevant enterprise R&D as part of industrial
policy should also be included when available.

182 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 4

11. Countries that collect and report two-digit NABS data to Eurostat may
include two sub-categories of aid to industry (Table 1):
– Manufacture of pharmaceutical products (NABS 0742).
– Manufacture of medical and surgical equipment and orthopaedic
appliances (NABS 0791).

Table 1. Identifying health-related R&D in GBAORD

One-digit NABS For countries using detailed NABS

Protection and improvement of human health All
Non-oriented research Medical sciences
General university funds Medical sciences
Industrial production and technology Support for the pharmaceutical industry
Support for the medical instrument industry

Source: OECD.

12. Perhaps the most important gap is the health-related R&D included in
general university funds or non-oriented research elsewhere than in the
medical sciences, especially in the biological sciences. Where any R&D funded
by health research councils or similar research programmes is included in
non-oriented research, it may be possible to identify the health-related
element of biology to be included.
13. Health-related R&D data derived from GBAORD give an incomplete
picture of total public funding of such R&D, as GBAORD only covers the central
government budget. Some health R&D may be funded by extra-budgetary
public sources such as social security funds. Provincial and local governments
may fund health R&D, particularly when they are responsible for higher
education or for general hospitals. Where these sums are significant, an effort
should be made to add them to the data derived from GBAORD in order to
obtain a figure for total government funding of health-related R&D.

Building GERD for health

The business enterprise sector
14. There are two manufacturing ISIC categories mainly relevant to
health:
– 2423 Pharmaceuticals.
– 3311 Medical instruments.
15. Taken together, these can be considered the core components of
health-related R&D (Table 2), although medical instruments may require
special extraction in the first instance. A full set of data should be available for

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 183

ANNEX 4

Table 2. Health-related R&D from performer-reported data:

business enterprise sector
Category Source

Pharmaceutical industry (ISIC Rev. 3, 2423) Possible to derive from R&D surveys either as industry
group or product field
Medical instruments (ISIC Rev. 3, 3311) Requires special extraction from R&D surveys either as
industry group or product field
R&D on pharmaceuticals performed in other industries May be possible to derive from product field classification,
other functional classification or extramural R&D
expenditure of the pharmaceutical industry
R&D on medical instruments performed in other May be possible to derive from product field classification,
industries other functional classification or extramural R&D
expenditure of medical instruments
Private health services (ISIC Rev. 3, 851) Extract if included in the scope of R&D surveys
R&D in other industries done for private health services May be possible to distinguish if health services are
a separate product group or from extramural R&D
expenditures of private health services

Source: OECD.

each, thus making it possible to compile data on total intramural R&D by

source of funds and R&D personnel by occupation/qualification.
16. Health-related R&D also takes place in the health services industry
itself.
– 851 Human health activities, notably:
❖ 8511 Hospitals.
❖ 8519 (part) Testing laboratories, medical, analytical or diagnostic.
17. These may not be included in the R&D survey at all, especially if the
health services are mainly public. If they are included, they will probably
require special extraction. The extraction should also cover sources of funds
and R&D personnel.
18. Health-related R&D may be carried out in the services for the
pharmaceutical, medical instruments and health services industries, notably
in the R&D services industry and the IT services industry (and indeed in
medical analytical and testing laboratories). The best way of identifying the
health R&D component is via cross-tabulation of industries and product fields.
This should reveal in particular how much pharmaceutical R&D is being done
outside the industry itself and also whether firms classified in the
pharmaceutical industry are engaged in R&D on other products. Estimates
must be made of the sources of funds of the additional health-related R&D in
the services and of the R&D personnel concerned. Where product field data
are not available, other series might be examined for additional information
on health-related R&D, including R&D by field of science (medical sciences),

184 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 4

R&D by socio-economic objective (health as SEO) or extramural expenditure by

the pharmaceutical and medical instrument industries. Care should be taken
to ensure that relevant R&D by biotechnology companies is included.

R&D in the non-market sectors (government, private non-profit, higher

education)
19. Some countries undertake a standard survey of all R&D institutes, and
these are subsequently divided among sectors of performance of the Frascati
Manual. Others undertake separate surveys for each sector.

• General approach

20. This Manual proposes collecting data by field of science, such as

medical sciences, as both an institutional and a functional classification, and
by socio-economic objective, with health as a functional classification.
21. Experience shows that neither health as a socio-economic objective
nor the medical sciences as a field of science is enough to describe adequately
the field of health-related R&D. A combination is needed, as shown in Table 3.

Table 3. Identifying health-related R&D by field of science

and socio-economic objective
Fields of science and technology
Socio-economic objective Other natural sciences Social sciences
Medical/health Biological
and engineering and humanities

Protection
and improvement
of human health X X X X
Non-oriented research X ?
All other X

X = to be included.
Source: OECD.

22. The core consists of all R&D for health in the medical sciences and/or
for health as an SEO (shown bolded in Table 3). Obtaining this depends on how
the two classifications are applied in each country. In theory, where the field
of science classification matches that in Table 3, there should be little R&D for
health as an SEO which is not included under medical sciences. However, the
classification is not entirely clear for genetics, hence the column for the
biological sciences and the potential problem of identifying how much
biological R&D undertaken as non-oriented research is health-related.
23. Deriving the sources of funds and calculating the R&D personnel data
for such a combination may involve some estimations.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 185

ANNEX 4

24. This functional or semi-functional approach can be supplemented or

replaced by data sets based on national institutional classifications, local
knowledge about the national health R&D system and additional material
from funding sources. For example, a list of the core national performers of
health-related R&D in the government and private non-profit (PNP) sectors
can be established, and special extractions from their responses can be made.

• Higher education

25. Where teaching establishments receive a detailed R&D questionnaire,

health-related R&D data can be compiled in the same way as R&D for other
surveyed units. However, they often do not receive such a questionnaire and
the data are derived from responses to a simpler questionnaire or are
compiled from administrative sources. Usually, but not always, there is a
breakdown by major field of science.
26. The core category is thus medical sciences as an institutional category
for which intramural expenditure, sources of funds and R&D personnel should
be available. However, if the classification unit is large, e.g. the medical faculty,
health-related R&D in other faculties, such as biological sciences and social
sciences, may be left out. The R&D funds are usually divided into direct funds
and GUF, and details may be available on the institutional sources of the direct
funds. Where direct research funds flow to non-medical faculties from a
health research council, a programme of a health department, a medical
charity or the pharmaceutical industry, these can be added.

• Private non-profit sector

27. The recommended institutional breakdown is by field of science,

which is used in most countries that separate out the PNP sector. R&D
expenditure in the medical sciences is thus the core category, and sources of
funds and personnel data should also be available. In this sector, expenditure
on the medical sciences is generally higher than expenditure on health as an
SEO. There is not likely to be additional health-related R&D spending unless
units classified in the social sciences carry out health services R&D or general
life science units classified in the natural sciences carry out medical research.
28. Where there is no classification by field of science, the units concerned
may have to be selected individually on the basis of local knowledge. This
sector may include a significant number of research units belonging to
medical charities and should not be ignored merely because it is small overall.

• The government sector

29. The Manual does not recommend an institutional classification for

this sector and the breakdown used is often based on national administrative

186 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 4

categories. For this reason and because of international differences in the way
health-related R&D is organised in the government sector, it is particularly
difficult to propose standard methods of identifying health-related R&D in
this sector.
30. Where data are collected on both field of science and SEOs, R&D
spending on health as an SEO is often higher than spending for the medical
sciences in this sector, particularly where the medical sciences are an
institutional category and SEO is a functional category. For this sector, the core
should be all institutional units whose principal R&D activity is health as an
SEO and/or the medical sciences. Any R&D in the field and/or relevant SEO in
other institutions should be added. The additional information may be
derived from crossing institutional and functional classifications or from
other sources, for example programme descriptions in R&D budgets, annual
reports of institutions, etc.

• Special institutional problems

31. Some countries have multidisciplinary research councils with R&D-

performing units in the government or higher education sectors which are
classified under non-oriented research as an SEO and which do not break
down their expenditures on life sciences as recommended in the Manual. It is
difficult to identify the health-related component of these funds, as they are
often earmarked for basic research.
32. When obtaining funds for health-related R&D, it is useful to look at
how hospitals are treated in the national R&D survey in terms of coverage and
classification.

Aggregating GERD on health

33. In principle, GERD is obtained by adding health-related R&D in each of
the four sectors of performance. Sources of funds are found by aggregating what
each sector receives from business enterprise, government, private non-profit
(PNP), higher education and abroad for carrying out health-related R&D. At this
stage, it may be useful to check figures against any funder-reported series and
perhaps calculate a health GNERD (gross national expenditure on R&D).
Differences are to be expected, but if there are major discrepancies, for example
if medical charities report much higher research funding than appears in GERD
on health as funded by the PNP sector, further enquiries may be worthwhile.

ICT-related R&D
34. In recent years, there has been quite intensive work at the OECD by the
Working Party on Indicators of the Information Society (WPIIS) to develop
statistics and indicators for the ICT sector or, more broadly, the information

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 187

ANNEX 4

economy sector. The aim is to develop statistics and indicators that result in a
better understanding of the information economy/information society.
35. A fundamental milestone was the reaching of an agreement on a
definition of the ICT sector based on ISIC Rev. 3. This definition identifies key
industries whose main activity is producing or distributing ICT products or
services and which constitute an approximation of the “ICT-producing sector”.
It needs to be complemented by a product-based definition.
36. The list of industries belonging to the ICT sector in ISIC Rev. 3 is as
follows:
Manufacturing
3000 Office, accounting and computing machinery
3130 Insulated wire and cable
3210 Electronic valves and tubes and other electronic components
3220 Television and radio transmitters and apparatus for line telephony
and line telegraphy
3230 Television and radio receivers, sound and video recording or
reproducing apparatus, and associated goods
3312 Instruments and appliances for measuring, checking, testing,
navigating and other purposes, except industrial process equipment
3313 Industrial process control equipment
Services
5150 Wholesale of machinery, equipment and supplies (in ISIC Rev. 3.1
limited to class 5151 “Wholesale of computers, computer peripheral
equipment and software” and class 5152 “Wholesale of electronic
and telecommunication parts and equipment”)
6420 Telecommunications
7123 Renting of office machinery and equipment (including computers)
72 Computer and related activities
37. This classification is a good starting point for defining ICT-related R&D
in the business enterprise sector. In R&D surveys, data are often available only
at ISIC 2-digit classifications. This makes direct application of this list difficult.
In addition, some categories have rather limited ICT content (ISIC 3130) or are
somewhat irrelevant for R&D surveys (e.g. the wholesale or renting categories).
An operational definition of ICT-related R&D may therefore include ISIC 30, 32
and 33 (ICT-related R&D in manufacturing) and ISIC 64 and 72 (ICT-related
R&D in services).
38. The above classification needs to be complemented by a classification
that is more relevant for defining ICT-related research, i.e. a product field
classification, which is a functional classification. Work is under way to reach
an international recommendation on which product fields are to be regarded
as ICT-related. Although the product field classification is not used in R&D

188 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 4

surveys in every country, a definition in terms of product field, when agreed,

may be more relevant for defining ICT-related R&D in the business enterprise
sector. Chapter 4, Section 4.3, of the Manual now includes a more explicit
recommendation on the use of a product field classification in R&D surveys.
The product field is defined according to the final product of the enterprise.
This means, for example, that R&D performed by a car manufacturer for
software included in cars will not be considered as ICT-related R&D, as the
software is not the car manufacturer’s final product. However, if the software
is bought from an outside company, any R&D carried out by that company for
the software product is considered ICT-related R&D.
39. One problem with using the product field classification could be that
the level of aggregation of the product field classification used in R&D surveys
may be too broad to distinguish ICT-related R&D that is defined according to
very disaggregated product groups.
40. R&D that is relevant for ICT or, more broadly, for the information
economy/information society also takes place in other sectors of
performance. Here, the field of science classification is useful. However, the
field of science classification recommended in Chapter 3 of the Manual is not
very helpful for identifying ICT-related R&D. Work is to start to develop a new
classification by field of science that is applicable for statistical purposes. It
would be essential to identify under natural sciences and engineering, and
probably also social sciences, sub-fields of relevance for the ICT sector or,
more broadly, the information economy/information society. Examples
are computer hardware, communication technologies and information,
computing and communication sciences. The application of a very detailed
field of science classification certainly causes difficulties in several countries.
This will affect their possibilities for using the field of science classification to
produce information on ICT-related R&D.
41. In theory, the classification by socio-economic objective (SEO) may also
be used to distinguish ICT-related R&D. Relevant sub-classes are included at
the 2-digit level of the present NABS. However, the SEO classification is applied
at this level of detail only in some EU countries.

Biotechnology-related R&D
Introduction
42. Biotechnology is perceived as having the potential to be the next
pervasive technology of great significance for future economic development.
Work is under way at the OECD to develop a statistical framework for the
measurement of biotechnology activities and to identify more closely the
needs of users for indicators on biotechnology activities and on the effects of

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 189

ANNEX 4

biotechnology. On the basis of these considerations, a model survey of

biotechnology is under development. As an aid, a definition of biotechnology,
in terms of a single definition and a list of technologies, has been agreed as a
basis for further work and is presented at the end of this annex.

Classifications
43. Classifications are usually used to delimit a field. Because biotechnology
is a process as opposed to a product or an industry, it is not easily identifiable on
the basis of existing classifications. ISIC, the standard international classification
of economic activities, was revised during the 1980s when interest in
biotechnology was rather limited. For the moment, it is not possible to identify
specific biotechnology industries at any level of ISIC (division, group, class). Some
preliminary discussions have taken place on the possibility of identifying
biotechnology-related industries in the next major revision of the classification.
The situation is more or less the same for the central product group classification
(CPC) and the harmonised commodity description and coding system HS 2002.
44. In their present form, the more R&D-related classifications by field of
science and socio-economic objectives (SEO) are not suitable for the
identification of biotechnology. Biotechnology is related to several of the major
fields of science recommended in the Manual, including natural sciences,
engineering, medical sciences and agricultural sciences. It may be possible to
identify biotechnology on the basis of a more detailed classification by field of
science, including agreed sub-fields of the major fields of science. This has to
be investigated during work to revise the field of science classification.
45. Experience in Australia indicates possibilities for identifying
biotechnology-related R&D on the basis of a detailed field of science
classification. The Australian classification has a specific category called
“biotechnology”, but there are also relevant categories at different levels of the
classification, such as biochemistry and cell biology, genetics, microbiology,
industrial biotechnology, bioremediation, biomaterials and medical
biotechnology.
46. It will be difficult to identify biotechnology in any revised classification
by socio-economic objectives.

Model surveys
47. The only possibility for obtaining information on biotechnology R&D or
use of biotechnology is therefore to develop special surveys on biotechnology or
to ask additional questions in existing surveys, such as the R&D survey. The first
option is being explored in work to develop model surveys for biotechnology.
The second option is to obtain information on biotechnology R&D from normal
R&D surveys through use of the OECD definition of biotechnology.

190 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 4

Adding questions on biotechnology to R&D surveys

48. The following paragraphs address the issue of adding questions to an
existing R&D survey.
49. Special questions on biotechnology to be added to R&D surveys or
collected in connection with them have some limitations. These are:
– The variable should be included in the general R&D survey framework.
– Appropriate classifications should be available for describing
biotechnology-related R&D.
– The additions on biotechnology should increase the response burden only
marginally.
50. Some ten countries have experience in requesting information on the
share of biotechnology R&D in total R&D expenditure. A question is often
asked in the context of a list of interesting technologies, of which
biotechnology is one. The surveys give a single definition, a list of relevant
technologies or a combination of the two. The definitions used in various
surveys differ. To improve comparability, it is recommended to use the OECD
definitions (both the single definition and the list presented at the end of this
annex). This would be the first step towards having more comparable data on
biotechnology R&D.

The following type of question could be asked in the general R&D

survey:
Did the R&D reported above include any biotechnology R&D (see
definition)?
Yes ( )
No ( )
If yes, please provide an estimate of the share of the total intramural
R&D expenditure reported earlier that is attributable to biotechnology.
_____%.

51. To guide the respondent, the OECD definition of biotechnology should

be provided. The list-based definition may be more helpful, but both may be
needed.
52. Another question that might be considered is the share of public
funding of R&D going to biotechnology R&D. The detailed formulation of this
variable may need further elaboration.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 191

ANNEX 4

53. As the interaction between science and technology is particularly

strong in the field of biotechnology, it is also recommended to include this
kind of question for R&D surveys in the other Frascati Manual sectors. The
experience of a few countries suggests that this is feasible.
54. It is recommended to introduce a few simple questions on
biotechnology R&D in R&D surveys in as many member countries as possible
to have a broader comparable overview of the role of biotechnology in their
R&D efforts.
55. Biotechnology is a multidisciplinary field. This poses particular
problems in categorising biotechnology for survey purposes. The current
OECD definition of biotechnology is preliminary and has mainly been piloted
in R&D surveys of the business enterprise sector. For comparability, the
definition is also recommended for use in other sectors. The experience
gained from using the definition in all sectors will lead to further revisions of
the current definition.

OECD definition of biotechnology

56.
“The application of S&T to living organisms as well as parts,
products and models thereof, to alter living or non-living
materials for the production of knowledge, goods and services.”
The (indicative, not exhaustive) list-based definition is:
● DNA (coding): genomics, pharmaco-genetics, gene probes,
DNA sequencing/synthesis/amplification, genetic engineering.
● Proteins and molecules (functional blocks): protein/peptide
sequencing/s ynthesis, lipid/protein glyco–e ngineering,
proteomics, hormones and growth factors, cell receptors/
signalling/pheromones.
● Cell and tissue culture and engineering : cell/tissue culture,
tissue engineering, hybridisation, cellular fusion, vaccine/
immune stimulants, embryo manipulation.
● Process biotechnologies: bioreactors, fermentation,
biop roc e ss in g , biole a ch in g , biopul pin g , bioble ach ing ,
biodesulphurisation, bioremediation and biofiltration.
● Sub-cellular organisms: gene therapy, viral vectors.

192 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 5

Annex 5
Methods of Deriving Regional R&D Data

Introduction
1. Chapters 5 and 6 of the Manual make recommendations for breaking
down data on R&D personnel and R&D expenditure by region. This annex
briefly discusses various methods of doing so. It draws on work by Eurostat,
which has investigated the methods in greater detail. Regional data can be
derived either directly, by classifying the statistical units, or by including a
separate question on this breakdown in surveys. This annex does not discuss
details of the regional breakdown. This has to be determined according to
national or international needs for information.

Classifying the statistical units

2. In many cases it is possible and feasible to attribute statistical units to
regions on the basis of the postal address of the entity. The advantage of this
approach is that all variables will automatically be available by region.
Problems may arise if sampling is used, as the region can seldom be used as a
stratification variable. For regionalisation of R&D variables, the ideal solution
would be to have statistical units small enough to avoid their having activities
in several regions. This would in many cases imply establishment-type units.
However, this is generally not feasible from the point of view of data collection
and compiling of meaningful national aggregates by industry. The data for
R&D surveys are usually available only at the level of enterprise-type units,
and an attempt to split these units into smaller ones would create extra work
for the respondent and for the surveying agency. Sectoral aggregates by
industry would also be rather different if the establishment is used as the
statistical unit. Therefore, the Frascati Manual recommends using the
enterprise-type unit as the most appropriate for R&D surveys in all sectors
except the higher education sector.
3. Attributing large units with activities in many regions to a single
region will, however, lead to distortions in the breakdowns. It is therefore
recommended, if it is not possible to have a separate question on regional
breakdown (as described below), to have, at least for the biggest units, a

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 193

ANNEX 5

separate breakdown by region for the most important variables (R&D

expenditures, R&D personnel). If it is not possible to obtain the information
directly, it may have to be estimated on the basis of variables that can be
assumed to be closely related to R&D.
4. Depending on the method used to obtain data on the higher education
sector, the establishment unit (for example the university institutes or
corresponding units) may be more feasible. In this case, regional data can be
derived directly. Otherwise, the discussion above and in the following section
is applicable.

Asking a separate question on the regional breakdown

5. Compared with the above alternative, this option gives more precision
to regional breakdowns. It serves as a substitute when information at
establishment level is lacking. The question can be asked in many ways. The
table below indicates the information to be requested without suggesting the
formulation of the question or the priority of the variables.

R&D personnel R&D personnel

Region, municipality or establishment R&D expenditure
(headcount) (FTE)

6. Information on the regions might be asked for directly. In some

countries, however, respondents may not be aware of how the regions are
defined. An alternative is to ask for the municipalities of the sub-units and to
code them later for the appropriate regions. A third alternative is to ask for
establishment-type units and to try to identify the address of the
establishment. It is usually possible to have the variables for which
regionalisation is required at establishment level. The table needs additional
columns for data on researchers by region. With this approach, problems of
sampling may arise, as raising factors have to be applied.

194 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 6

Annex 6
Work on S&T Indicators in Other International
Organisations

1. The problems of comparing R&D data and other S&T indicators

collected in different countries with varying institutional patterns and
traditions in education and research, have been considered by other
international organisations as well as by the OECD. This annex presents an
overview of the activities undertaken to develop S&T indicators in various
international organisations.

UNESCO (United Nations Educational, Scientific and Cultural

Organisation)
2. Since 1965, the UNESCO Division of Statistics has organised the
systematic collection, analysis, publication and standardisation of data on
science and technology (S&T), and, more specifically, R&D. The first
experimental questionnaires were circulated to member states in 1966, and
standardised periodical surveys were established in 1969. From information
obtained mainly through statistical surveys on qualified S&T human
resources and R&D personnel and expenditure conducted worldwide
since 1970, a database has been built that covers at present some
100 countries and territories. These data were published regularly in the
UNESCO Statistical Yearbook (UNESCO annual until 1999) and have also been
used for special reports and analyses, such as the World Science Report.
3. The methodology used in the surveys was progressively developed
with the aid of national specialists from throughout the world and was
discussed at length during missions and at meetings convened by UNESCO.
The aim was to collect information on scientific and technological activities
(STA) in a form that allows for maximum international comparability.
Following methodological studies in 1975 and 1976, the UNESCO Secretariat
drew up, with the assistance of external specialists, a draft Recommendation
for the International Standardisation of Statistics in Science and Technology,
which was reviewed by a meeting of governmental experts and then
proposed for adoption by the UNESCO General Conference in 1978. This
Recommendation detailed international statistical standards that could be

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 195

ANNEX 6

applied by all member states, both those with advanced S&T statistical
systems and those whose systems were still being developed. Although
designed to provide standardised information on S&T activities, it
concentrated on R&D. However, it proposed a gradual extension of the
statistics beyond R&D.
4. Following the adoption of the Recommendation, two successive stages
at international level were proposed: the first, over a period of at least five
years after the adoption of the Recommendation, was to cover only R&D in all
sectors of performance, together with the stock of, and/or economically active,
qualified personnel; during the second stage, statistics were to be extended to
cover scientific and technological services (STS) and S&T education and
training, at broadly the third level (scientific and technical education and
training – STET). In 1984, UNESCO published a manual (UNESCO, 1984b) on
these international standards and issued a revised “Guide to the Collection of
Statistics on Science and Technology” (UNESCO, 1984a), for use by member
states. Throughout this work, UNESCO took account of the experience
acquired by other intergovernmental organisations such as the OECD, the
former CMEA (Council for Mutual Economic Assistance), and the OAS
(Organization of American States). Co-operation was also promoted through a
Joint Working Group of UNESCO and the ECE (United Nations Economic
Commission for Europe), which studied ways to improve and develop S&T
statistics at meetings held in 1969, 1972, 1976 and 1981.
5. Since 1976, UNESCO has also made efforts to develop a methodology
for collecting data on scientific and technological information and
documentation (STID); this work resulted in the publication of a provisional
STID Guide in 1984 (UNESCO, 1984c). Work to establish a methodology for
collecting statistics on STET began in 1981. Case studies were carried out in
various regions of the world to determine the state of S&T statistics, problems
encountered in the implementation of the Recommendation, and the need for
new S&T indicators.
6. With the changes that occurred during the 1980s and early 1990s
especially with respect to the organisation and measurement of S&T activities
in the former centrally planned economies, a special external evaluation of
UNESCO’s S&T statistical programme was carried out in 1996. The findings
and recommendations of this evaluation concluded, inter alia, that UNESCO’s
R&D statistical programme should align its methodology on that of the Frascati
Manual, and that priority should be given to further development of
international S&T indicators that respond to the needs of all countries.
7. Since the establishment of the UNESCO Institute for Statistics (UIS)
in 1999, UNESCO’s activities focus on a fundamental international review of
policy needs in S&T and of existing S&T statistical systems and capacities, in

196 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 6

close co-operation with international expert networks, the OECD and

Eurostat. The main objective of this review is to help UNESCO to redefine its
programme priorities and implementation strategies in the area of S&T
statistics. The results of this review will become available in 2003 and the
priorities and strategies will be submitted to the 32nd UNESCO General
Conference for approval, before implementation begins in 2004.

Eurostat (Statistical Office of the European Communities)

8. Eurostat, with the collaboration of EU and EEA member states
represented in Eurostat’s Working Party on R&D and Innovation Statistics,
draws up annual reports on the public financing of R&D by socio-economic
objective in member states, on the R&D appropriations of Community
institutions, the regional distribution of R&D personnel, R&D expenditure and
European patent applications. Data are collected through an annual survey of
member states and processed for presentation in comparable form. Eurostat
also collects and disseminates R&D and innovation statistics from EU
candidate countries and the Russian Federation.
9. Eurostat has been co-responsible for methodological work in various
domains. It actively participated with the OECD in the first revision of the Oslo
Manual (OECD, 1997a). Innovation survey methodology has been largely
influenced by the three Community Innovation Surveys prepared and
co-ordinated by Eurostat. Eurostat has developed a manual on the regional
aspects of R&D and innovation statistics and has developed guidelines for
collecting data on government appropriations for R&D which complement
those in the previous edition of the Frascati Manual. Eurostat also participated
actively with the OECD in the development of the Canberra Manual (OECD,
1995) on human resources for science and technology and has pioneered the
collection and publication of statistics consistent with that manual.

Nordforsk/Nordic Industrial Fund

10. Since 1968, the Nordic countries have collaborated to co-ordinate their
work in the area of R&D statistics. Until 1987, the co-operation was organised
by Nordforsk (the Nordic Co-operative Organisation for Applied Research)
which set up a special committee on R&D statistics. During this period,
various working groups discussed a number of problems related to the
production and analysis of R&D statistics, mainly with reference to inter-
Nordic data comparability. In 1974, the Committee produced a “Nordic
Manual” in the Nordic languages, which was a detailed supplement to the
Frascati Manual. Selected chapters were translated into English and presented
by Nordforsk at various meetings of experts at the OECD. The Committee also
worked on budget analysis, and relevant guidelines were published in the

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 197

ANNEX 6

Nordic languages (Nordforsk, 1983). Then, in 1986, a short report was

published on work on improved guidelines in the higher education sector
(Nordforsk, 1986).
11. In 1987, Nordforsk merged with the Nordic Industrial Fund which took
over responsibility for the Committee. The Committee continues to give high
priority to developing R&D statistics in the Nordic countries. Among other
topics, the need for projection data and a methodology for establishing such
data has been discussed. Most recently, the Committee, renamed the Nordic
Group for Development of Science and Technology Indicators, has put more
effort into the problems of measuring and evaluating the results of R&D; both
producers and users of S&T indicators are members of the group.
12. In 1989 the Nordic Industrial Fund set up a special working group for
innovation studies, which launched a Nordic survey on innovation using a
common questionnaire. It also organised several international seminars to
discuss guidelines for innovation surveys. These were the basis for
discussions at the OECD, which led to the adoption and publication of the
Oslo Manual in 1992 (OECD, 1992). In the 1990s, activities have mainly focused
on the publication of R&D statistics (every other year) and of more
comprehensive science and technology indicators (at longer intervals).

RICYT (Red Iberoamericana de Indicadores de Ciencia y Tecnología)

13. The Ibero American Network of Science and Technology Indicators
(Red Iberoamericana de Indicadores de Ciencia y Tecnología – RICYT) was created by
the Ibero American Programme of Science and Technology for Development
(Programa Iberoamericana de Ciencia y Tecnología para el Desarrollo – CYTED) on
the basis of a proposal arising from the First Ibero American Workshop on
Science and Technology Indicators held at the National University of Quilmes
in late 1994. Since it was set up, RICYT has co-ordinated its activities with the
Organisation of American States (OAS).
14. RICYT’s general aim is to promote the development of instruments for
the measurement and analysis of science and technology in Latin America,
within a framework of international co-operation, with a view to increasing
their use as a political instrument for decision making.
15. RICYT’s activities take the following forms:
– Workshops with methodological discussions on the problems of science
and technology indicators in Latin America and intensified the exchange of
information among the various members of the network. One result has
been the publication of a Latin American manual of indicators on
technological innovation, the “Bogotá Manual”.

198 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 6

– Publication of the indicators of the region in the series, “Main Ibero and
Inter-American Science and Technology Indicators” (Principales Indicadores
de Ciencia y Tecnología).
– Creation of mechanisms of mutual assistance in Latin America.
– Diffusion activities through the publication of “Indicios”, a news and
opinion bulletin, a Web page (www.ricyt.edu.ar) devoted to the network’s
activities which presents regularly updated information on the indicators,
and the edition of bibliographic material.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 199

ANNEX 7

Annex 7
Other Science and Technology Indicators

Introduction
1. As discussed in Chapter 1 of the Manual, it has become increasingly
clear that R&D statistics alone do not suffice to describe the range of inputs
and outcomes associated with scientific and technological development (see,
for example, Freeman 1987).
2. The OECD, recognising the need to facilitate the development of
indicators other than those directly associated with R&D, has prepared a
series of non-R&D methodological manuals or other guidelines (see Chapter 1,
Table 1.1). These manuals and guidelines are intended to be complementary
and, in time, to provide guidance for the collection and interpretation of data
describing the full spectrum of scientific and technological activities.
3. This annex outlines seven series of such indicators for which
guidelines are prepared or planned. Its purpose is to provide users and
producers of R&D statistics a context for setting R&D indicators within the
framework of the overall S&T system. It also outlines the sources and
availability of data in each area and describes some drawbacks relating to their
use. The indicators are presented in historical order in terms of development.
The situation is described as of 2002.

Patent statistics
Coverage
4. A patent is an intellectual property right relating to inventions in the
technical field. A patent may be granted to a firm, an individual or a public
body by a patent office. An application for a patent has to meet certain
requirements: the invention must be novel, involve a (non-obvious) inventive
step and be capable of industrial application. A patent is valid in a given
country for a limited period (20 years).
5. For purposes of international comparison, statistics on patent
applications are preferable to statistics on patents granted because of the lag

200 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 7

between application date and grant date, which may be up to ten years in
certain countries.
6. Patent indicators based on simple counts of patents filed at an
intellectual property office are influenced by various sources of bias, such as
weaknesses in international comparability (home advantage for patent
applications) or high heterogeneity in patent values within a single office.
Furthermore, differences in patent regulations across countries make it very
difficult to compare patent statistics between two (or more) patent offices.
7. To overcome the problems associated with the traditional patent
indicators (described above), the OECD has been working towards developing a
new type of patent-based indicator: patent family counts. A patent family is
defined as a set of patents taken in various countries to protect a single
invention (characterised by a first application in a country – called the priority
application – which has been extended to other offices). The advantages of
using indicators based on patent families for statistical purposes are twofold:
they improve international comparability by eliminating home advantage and
geographical influence; patents included in the patent family are of high value.
8. Patent documents contain a rich source of information on the invention
that is unavailable elsewhere and therefore constitutes a significant
complement to the traditional sources of information for measuring diffusion
of technological/scientific information (see section on bibliometrics). Patent
documents contain information on: i) technical features (such as list of claims,
technical classification, list of cited patents, etc.); ii) history of the application
(such as priority date, date of publication, date of filing in the country
concerned, date of grant, etc.); and iii) information about the inventor (such as
name and address of inventors, country of residence, name of applicants, etc.).

Use of patent statistics

9. Among the few available indicators of technology output, patent-based
indicators are probably the most frequently used. Patent-based indicators
provide a measure of the output of a country’s innovative activity: its
inventions. The scientific literature on the determinants and impact of
innovative activity increasingly uses patent data at aggregate (national) or
firm level, because of the widely recognised close relationship between
patents and innovative output. Patent data are also used to identify changes in
the structure and evolution of inventive activity in countries, industries,
companies and technologies by mapping changes in technology dependency,
diffusion and penetration.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 201

ANNEX 7

Availability
10. National and international (e.g. European Patent Office – EPO; World
Intellectual Property Organisation – WIPO) patent offices are the primary data
sources. The OECD assembles, stocks and publishes various patent-based
indicators for its member countries in the Main Science and Technology Indicators
(OECD, biannual) and the OECD Science, Technology and Industry Scoreboard
(OECD, biennial) and in the associated diskettes and CD-ROMs. The OECD
patent database also includes information on patents filed at the EPO, the
Japanese Patent Office and the United States Patent and Trademark Office
(USPTO), broken down by country of residence and technological areas.

Drawbacks
11. There are some drawbacks associated with using patent indicators for
the measurement of R&D output and/or innovative activity. Many innovations
are not patented because they are protected by other means, such as
copyright, trade secrecy, etc. The propensity to patent varies across countries
and across industries, and this makes cross-country or cross-industry
comparisons difficult. The distribution of the value of patents is skewed, as
many patents have no industrial application, hence are of little value, whereas
a relatively few are of substantial value. Given such heterogeneity, patent
counts that assume all patents to be of generally equal value are misleading.
The number of patent applications or grants per se is difficult to interpret; the
number of patents has to be used in conjunction with other indicators.

International guidelines
12. The growing role of international patent organisations contributes to
creating greater comparability of the patent data available for individual
countries, although it is still affected by the special characteristics of patents.
The OECD patent manual (“Using Patent Data as Science and Technology
Indicators – Patent Manual 1994”) (OECD, 1994b) outlines the general
guidelines for the use and interpretation of patent data as indicators of S&T.

The technology balance of payments (TBP)

Coverage
13. The TBP registers the international flow of industrial property and
know-how.
14. The following operations are included in the TBP: patents (purchases,
sales); licences for patents; know-how (not patented); models and designs;
trademarks (including franchising); technical services; finance of industrial
R&D outside national territory.

202 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 7

15. The following operations are excluded: commercial, financial,

managerial and legal assistance; advertising; insurance; transport; films,
recordings, material covered by copyright; design; software.

Use of TBP statistics

16. TBP indicators measure the international diffusion of disembodied
technology by reporting all intangible transactions relating to trade in
technical knowledge and in services with a technology content between
partners in different countries.

Availability
17. National TBP data may be collected by means of special surveys but
more often are assembled from existing records kept by central banks,
exchange control authorities, etc.
18. The OECD has assembled a database of “macro” TBP data for most of
its member countries which covers total transactions (receipts and payments)
by partner country from 1970. Data for periods since the late 1980s are
published in Main Science and Technology Indicators (OECD, biannual) and in the
associated CD-ROM. In 2000 a new international database for detailed TBP
series broken down by industry, type of operation and geographical area was
established.

Drawbacks
19. For many countries the data are available only at a rather aggregate
level. The available data do not necessarily correspond to the definition of TBP,
i.e. they may cover more or less than transactions with a technological
content. The balance is affected by sometimes non-monetary transactions
within multinational firms. There are difficulties for interpreting the data, and
the international comparability of the data may be weak.

International guidelines
20. In 1990, the OECD issued the “Proposed Standard Method of Compiling
and Interpreting Technology Balance of Payments Data – TBP Manual” (OECD,
1990). It is the second in the series of OECD manuals on science and
technology indicators.

Bibliometrics
Coverage
21. Bibliometrics is the generic term for data on publications. Originally, it
was limited to collecting data on numbers of scientific articles and other

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 203

ANNEX 7

publications, classified by author and/or by institution, field of science,

country, etc., in order to construct simple “productivity” indicators for
academic research. Subsequently, more sophisticated and multidimensional
techniques based on citations in articles (and more recently also in patents)
were developed. The resulting citation indexes and co-citation analyses are
used both to obtain more sensitive measures of research quality and to trace
the development of fields of science and of networks.

Use of bibliometric statistics

22. Bibliometric analysis uses data on numbers and authors of scientific
publications and on articles and the citations therein (as well as the citations
in patents) to measure the “output” of individuals/research teams, institutions
and countries, to identify national and international networks, and to map the
development of new (multidisciplinary) fields of science and technology.

Availability
23. Most bibliometric data come from commercial companies or
professional societies. The main general source is the set of Science Citation
Index (SCI) databases created by the Institute for Scientific Information
(United States), which Computer Horizons, Inc., has used to develop several
major databases of science indicators. Bibliometric data can also be derived
from other, more specialised databases. The OECD currently has neither the
plans, the resources, nor the competence to undertake basic data collection,
although it regularly uses bibliometric data in its analytical reports.

Drawbacks
24. The propensity to publish varies between fields of science. The utility
of bibliometric indicators is greatest for the medical sciences and certain
natural sciences. The databases are biased towards articles in English, which
may affect international comparisons.

International guidelines
25. Bibliometric methods have essentially been developed by university
groups and by private consultancy firms. There are currently no official
international guidelines for the collection of such data or for their use as
science and technology indicators. In 1989-90, the OECD commissioned a
report on the “state of the art” in bibliometrics which was published 1997 as
an STI working paper (Okubo, 1997).

204 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 7

High-technology products and industries

Coverage
26. To contribute to analyses of the impact of technology on industrial
performance, it is helpful to identify those activities and products which are
most technology-intensive, using criteria that allow for constructing special
internationally harmonised classifications. In recent years, the OECD has
developed technology classifications both by industry, which has generated
much interest and widespread application in member countries, and by
product.
27. In the industry approach, manufacturing industries are allocated to one
of four groups: “high”, “medium-high”, “medium-low” or “low” technology. Until
the late 1990s, a technology classification using ISIC Rev. 2 was widely used. It
was based on an evaluation of the ranking of three indicators of technology
intensity reflecting, to different degrees, the “technology-producer” and
“technology-user” aspects: i) R&D expenditures divided by value added; ii) R&D
expenditures divided by production; and iii) R&D expenditures plus technology
embodied in intermediate and investment goods divided by production. Since
the adoption by the OECD of ISIC Rev. 3 for presenting data by industrial activity,
the work on technology groups has been updated. However, at present, the
limited availability of ISIC Rev. 3 input-output tables (required for estimating
embodied technology) means that only the first two indicators cited above are
considered. For early results, see Annex 1 of the OECD Science, Technology and
Industry Scoreboard 2001.
28. A product approach has the advantage of allowing more detailed analysis
and identification of the technology content of products. Not all products in a
“high-technology industry” necessarily have a high technology content; likewise,
a high degree of technological sophistication may be found in products from
industries with lower technology intensities. In collaboration with Eurostat, the
OECD used detailed R&D data by product field to produce a list of high-technology
products and an associated database based on SITC Rev. 3 classes at the 5-digit
level of detail. Updating this work to generate a list based on 6-digit Harmonised
System (HS) products could be an important next step.

Use of high-technology products and industry statistics

29. When constructed, these indicators measure the technology content
of the goods produced and exported by a given industry and country, with a
view to explaining their competitive and trade performance in high-
technology markets. Such markets are characterised by rapid growth of world
demand, offer higher than average returns to trade and affect the evolution of
the structure of industry.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 205

ANNEX 7

30. Indicators on trade in high-technology products/industries were

originally designed as measures of the “output” or “impact” of R&D; they are
now seen as having wider use in the analysis of competitiveness and
globalisation.

Availability
31. Data based on the OECD definitions of high technology are published
in the OECD’s Main Science and Technology Indicators and the Science, Technology
and Industry Scoreboard. They are also used in many national publications.

Drawbacks
32. At present, the classifications do not take into account products and
industries with low R&D intensities but produced with high-technology
machinery and equipment. The classifications are based on R&D intensities
only in a certain number of OECD countries.

International guidelines
33. International guidelines do not exist, but the OECD approach to
measuring high-technology products and industries is presented and
comprehensively discussed in “Revision of the High-technology Sector and
Product Classification” (Hatzichronoglou, 1997).

Innovation statistics
Coverage
34. The OECD Proposed Guidelines for Collecting and Interpreting Innovation
Data – Oslo Manual (OECD, 1997a) defines technological product and process
innovations as those implemented in technologically new products and
processes and in significant technological improvements in products and
processes. An innovation is implemented if it has been introduced on the
market (product innovation) or used within a production process (process
innovation). Innovation involves a series of scientific, technological,
organisational, financial and commercial activities. In the various Community
Innovation Surveys (CIS) implemented by Eurostat on the basis of the Oslo
Manual, various modifications have been made to this definition.

Use of innovation statistics

35. Innovation indicators measure aspects of the industrial innovation
process and the resources devoted to innovation activities. They also provide
qualitative and quantitative information on the factors that enhance or hinder
innovation, on the impact of innovation, on the performance of the enterprise

206 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 7

and on the diffusion of innovation. A few countries have also introduced some
questions on innovation in other surveys, such as the R&D survey.

Availability
36. National data on innovation activities are generally collected by means
of surveys addressed to industrial firms on an ad hoc basis. Most OECD
member countries have organised such surveys, and the Oslo Manual is based
on their experience.
37. It is also possible to collect data on the number and nature of actual
innovations. Such information can be obtained by special surveys or
assembled from other sources, such as the technical press.
38. The first internationally comparable series of data on innovation was
collected under the auspices of the Nordic Industrial Fund. The OECD
contributed to the preparation of a list of questions proposed for inclusion in
harmonised surveys during the launching of the first Community Innovation
Survey by the European Union. The experience gained from this survey was
used in preparing the second edition of the Oslo Manual. Many OECD countries
have used the EU questionnaire as a basis for developing their own innovation
surveys. Currently (autumn 2002), the third CIS is in the data-processing stage.

Drawbacks
39. Innovation surveys suffer from some problems of quality owing to
unsatisfactory response rates in the case of voluntary surveys and differences
in the understanding of the concept of innovation among enterprises. The
ad hoc nature of national surveys is not satisfactory for users, and in many
countries innovation surveys give information on R&D that is not consistent
with information from R&D surveys.

International guidelines
40. The initial Oslo Manual (OECD, 1992) was prepared jointly by the OECD
and the Nordic Fund for Industrial Development (Nordisk Industrifond, Oslo)
in 1990 and was officially adopted by the OECD as the third in the “Frascati”
family of manuals. The manual was jointly revised with Eurostat in 1997. A
second revision may take place during the coming years.

Human resources for science and technology (HRST)

Coverage
41. The Frascati Manual discusses only the measurement of R&D
personnel. The concept of HRST is much wider and covers other categories of
personnel engaged in scientific and technological activities.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 207

ANNEX 7

42. HRST are defined in the Canberra Manual (see below) in terms of
qualifications or current occupation. In the first case, the appropriate
classification is the International Standard Classification of Education (ISCED)
(UNESCO, 1976; 1997) and, in the second, it is the International Standard
Classification of Occupations (ISCO) (ILO, 1968; 1990). Data sets and analyses
may cover only persons with university qualifications/professional
occupations or they may also include those with other post-secondary
qualifications and technical jobs. A combination of criteria and levels is
needed if supply and demand issues are to be analysed correctly.
43. An ideal database should cover total national stocks of HRST at given
points in time, broken down by employment status and by sector and type of
employment, as well as the intervening inflows (mainly educational output
and immigration) and outflows (mainly retirement and emigration). Both
stocks and flows should be broken down by field of science and technology,
age and gender and possibly also national or ethnic origin. Data on specific
categories of interest (PhDs, post-doctorates, researchers, IT professionals,
etc.) are also needed.

Use of HRST data

44. Co-ordinated sets of data on HRST can be used (when linked to
demographic statistics) to review the current and possible future supply and
use of and demand for scientific and technical personnel (at home and
abroad), with a view to evaluating the consequences for future research and
industrial performance, planning education and training, measuring the
diffusion of knowledge incorporated in human resources and assessing the
role of women (and minorities) in science and technology activities.

Availability
45. A few small OECD countries are able to maintain complete nominal
registers of all S&T graduates and their whereabouts, from which HRST data
may be produced. The National Science Foundation in the United States also
maintains a comprehensive database on the characteristics of scientists and
engineers. In most countries, however, databases on HRST have to be built up
from several sources, notably education statistics (numbers of teachers and
graduates), labour force surveys and other employment statistics, and
population censuses, supplemented by special surveys.
46. Eurostat compiles basic HRST stock data from the European
Community labour force survey and education inflow data from education
statistics, which provide fairly harmonised results. UNESCO, Eurostat and the
OECD have developed a common questionnaire to collect education statistics.
These organisations publish data on teaching personnel and on students and

208 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 7

graduates by ISCED level and field of study. The OECD hopes to build a more
detailed database and set of indicators.

Drawbacks
47. The existing statistics are quite fragmented, and the level of
aggregation is quite high owing to the use of sample surveys (e.g. the labour
force survey) as the main sources of data for the stocks of HRST.

International guidelines
48. In 1995, Eurostat and the OECD jointly released the Canberra Manual
(OECD, 1995), which includes international standards for measuring stocks
and flows of HRST. This manual is currently under review.

Information society statistics and indicators

Coverage
49. The aim is to develop indicators and analysis of the information
society for policy making and analysis. The work involves the production of
internationally comparable and policy-relevant indicators for measuring the
supply of and demand for ICT infrastructure, related services, content and
applications, in particular for electronic commerce.
50. The approach followed is that of “building blocks”. Methodological
work and data collection have proceeded in several areas at different speeds,
in a step-by-step, pragmatic way, by looking first at supply-side statistics for
the information society (statistics on the ICT sector) and then at the demand
side (ICT usage statistics).

Use of ICT sector and ICT usage statistics

51. Development and analysis of new and existing ICT indicators are an
aid to formulating policies and monitoring progress related to the information
society. ICT sector statistics help to measure the contribution of ICT-producing
industries to economic activity (e.g. value added, employment, R&D and
innovation performed, contribution to the trade balance). ICT access and
usage indicators help to identify the degree of countries’ “readiness” to adopt
new technologies and the rate of diffusion of these technologies among all
actors in the economy (businesses, households, individuals, governments).
Indicators of electronic commerce transactions are based on common OECD
definitions and measure the relative size of on-line sales and purchases and
their breakdown by type of customer and geographical destination.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 209

ANNEX 7

Availability
52. Pilot collections of ICT indicators related to the ICT sector (supply
statistics) and to ICT use and electronic commerce (demand statistics) are
under way, and information on methodologies and survey vehicles used by
member countries is been collected. The indicators are used in OECD
publications such as the Information Technology Outlook, the Communications
Outlook, and the Science, Technology and Industry Scoreboard. The OECD’s
Measuring the Information Economy (2002) looks at the role played by ICT
investment, consumption and ICT-related innovation in OECD economies; at
the size and growth of ICT activities and their contribution to employment
and international trade; at the extent to which businesses and individuals
access and use new technologies and the reasons why they do not. It has a
special focus on electronic commerce transactions and their drivers and
inhibitors.

Drawbacks
53. Drawbacks for measuring ICT usage and electronic commerce
statistics are linked both to definitional issues and to the typical structure of
member countries’ data collection programmes. The target populations and
sampling methodologies may differ in countries’ surveys of ICT use in
enterprises. This can lead to misleading international comparisons of
aggregate figures, since ICT usage statistics are very sensitive to size cut-off
and industry coverage. In surveys on ICT use in the household sector,
problems of comparability may be affected by whether the statistical unit is
the individual or the household. As relatively few businesses or individuals
currently engage in electronic commerce transactions, the statistics may not
meet the statistical standards for publication. For statistics on ICT supply,
classification is crucial. International comparability of activity-based
classifications can be hard to achieve, given the level of detail required by the
OECD definition of the ICT sector, which is based on the 4-digit classes of ISIC
Rev. 3. Problems of confidentiality are sometimes encountered when
collecting data on telecommunication services, while very few countries can
provide data on ICT wholesaling.

International guidelines
54. The methodological work entails the development of guidelines and
model surveys. Examples are: the OECD definition of ICT sector, which covers
a group of ISIC Rev. 3 manufacturing and service activities; the OECD
definitions of electronic commerce transactions and the guidelines for
implementation; the OECD model survey on ICT usage in business; the OECD
model survey on ICT usage by households/individuals. Model surveys are

210 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 7

intended to provide guidance on the measurement of ICT indicators, Internet

use and electronic commerce and are composed of separate, self-contained
modules to ensure flexibility and adaptability to a rapidly changing
environment. While the use of “core” modules allows measurement on an
internationally comparable basis, additional modules can be added to respond
to evolving or country-specific policy needs.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 211

ANNEX 8

Annex 8
Practical Methods of Providing Up-to-date Estimates
and Projections of Resources Devoted to R&D

The demand for projections of R&D data

1. Surveys are the most accurate means of measuring scientific and
technological activities. They involve a complex process, however, and there is
some delay between R&D performance, collection of data and publication. There
is therefore an increasing demand for forecasts. Both policy makers and other
users desire projections of the most useful indicators for defining, evaluating,
monitoring or introducing science and technology programmes and policies.

Types of projections covered

2. A distinction should be made between short-term, medium-term and
long-term projections. The issue of medium-term and long-term forecasts
(which may be called prospective analysis) is not addressed here. This annex
only deals with short-term projections and with attempts to estimate the
values of a few variables for very recent years or to make provisional estimates
for the current year, when survey results are not yet available.

Objective
3. This annex describes the most frequently used methods and offers a
few basic guiding principles for forecasting and projecting the values of such
variables, but it does not seek to lay down universally applicable methods (or
procedures). The special characteristics of individual countries, and indeed
sectors, each with its own determinants and pace of change, argue against
adopting standard procedures.

The variables
4. Projections are most often made for:
– R&D expenditures.
– R&D personnel.
– Technologies.

212 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 8

5. Given that there is an element of value judgement in projections

concerning technologies, no recommendations will be made on this topic.
6. The greatest demand is for indicators of recent and future trends in
R&D expenditure, specifically:
– Total national R&D expenditure [especially gross domestic expenditure on
R&D (GERD) as a percentage of the gross national product (GDP)].
– R&D expenditure by sector.
7. Projections of R&D personnel can be particularly valuable for
forecasting, as the series are usually less volatile than those for expenditures.
8. The variables discussed are not necessarily interdependent but, where
they are, careful note should be taken of the relationship, in order to check
forecasts for coherence (see paragraph 20 below).

Projection methods
Extrapolation techniques
9. Extrapolation techniques are used with time series for which R&D
variables are normally available on at least a biennial basis. Variations are
usually analysed using suitable functions (e.g. polynomial or exponential
functions).
10. When many years are taken into account, it is easier to identify
dominant trends and the fit is better. However, analysis of more recent years
may indicate “new” trends or changes in the system. Constant prices should
be used in order to clarify the trends.

Proportional projection
11. Whenever a proportional relation is believed to exist between two
variables, the following procedure should be adopted:
– The existence of the proportional relation is verified by empirical
observation, by use of correlation/regression techniques or by use of a
model.
– The proportional coefficient is calculated.
– Later values for the independent variable are obtained (by extrapolation or
from another source of information).
– The proportional coefficient is applied to this independent variable to
derive the other, dependent, variable.
12. Unless countries are undergoing rapid structural change, this
procedure can be used, for example, to estimate total R&D expenditure as a
share of GDP.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 213

ANNEX 8

13. It may also be possible to use this technique to project R&D

expenditures or personnel for individual sectors if suitable independent
variables can be found for which forecasts are available, for example in
national accounts, labour force statistics or other economic sources.

Growth rates
14. Indications of proposed or expected growth may be available for some
of the better-known variables, especially for recent years and the current year.
This is most likely to be the case for the R&D expenditure or personnel of a
particular sector. For example, company plans can be a useful input to
forecasts of R&D spending or personnel in the business enterprise sector.
15. Expert opinion can also be of great help for accurately forecasting
sectoral trends. Quite apart from their direct usefulness, such contributions
often provide information of a qualitative, and sometimes circumstantial,
nature.

Reports of R&D funders

16. While R&D data obtained from performers are generally
recommended, as being more reliable than those supplied by funders, data
from funders are often available more rapidly and can make a valuable
contribution to projecting some variables for the public sector. Government
budget appropriations or outlays for R&D (GBAORD) data can often be used to
construct provisional estimates for R&D carried out in the government sector
and in some cases in the higher education sector [using the budget forecasts
of general university funds (GUF)]. GBAORD data are of less use for projecting
R&D spending in the private non-profit (PNP) and especially in the business
enterprise sectors, as accurate figures on extramural R&D payments to these
sectors can rarely be obtained from government budgets.
17. In countries where there are linked reporting procedures for GBAORD,
government intramural and extramural R&D, and ultimately GERD, this
approach can be quite reliable. Where GBAORD is prepared separately and
only for appropriations and there is no subsequent reporting of final outlays,
it is much less accurate. Consequently, while government budgets are an
important aid in estimating certain variables, they must be used with caution.
18. Reports of non-public R&D funders should also be taken into account,
notably in the case of national funds (e.g. charities for medical research in the
case of funding by the PNP sector) and international organisations. Any major
changes in their funding behaviour may cause a discontinuity in the R&D
expenditure series.

214 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 8

Coherence and validity of projections

Dispersion of projections
19. Application of a single projection method may produce values for
subcomponents which do not add to the projected total (for example,
extrapolation of R&D spending in the four sectors of performance and of
GERD). Use of several projection methods will yield several values for the same
variable.
20. These should first be tested for coherence and plausibility, for example
by viewing trends in derived indicators, such as R&D expenditure per
researcher. Once any implausible results are excluded, averages, possibly
weighted averages, have to be calculated unless the spread is too wide.
21. It is recommended to indicate the interval, as this makes it possible to
measure discrepancies among projections obtained by different methods.

Verifying the projections retrospectively

22. If projections are made regularly, for example for annual or biennial
S&T indicators reports, retrospective R&D survey results should be used when
they become available to check the forecasts and to identify successes and
inaccuracies and the reasons for both.

Guiding principles
23. As previously noted, the special characteristics of different countries
and sectors make it impossible to select a simple methodology and
recommend its use without attention to context (particularly the performing
sector concerned). Flexibility in using methodologies is needed, and
composite approaches are acceptable and very often necessary.
24. Ideally, projections would be carried out using a single agreed
projection technique. Since this is not yet feasible, it is essential that member
countries, when publishing the results of their projections, provide relevant
documentation on how results have been obtained, regarding:
– Variables.
– Methodologies.
– Hypotheses.
– Special circumstances.
25. Compliance with this recommendation is vitally important to ensure
international comparability of member countries’ forecasts reported to the
OECD for inclusion in its databases and publications.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 215

ANNEX 8

Other guidelines
26. The ideas presented in this annex are drawn from a paper prepared by
Professor F. Niwa of the National Institute of Science and Technology Policy,
Japan, for the Expert Conference to Prepare the Revision of the Frascati Manual
for R&D Statistics held in Rome in October 1991. The paper presents a
framework, guidelines and methods for conducting R&D projections; it
suggests methods for projecting R&D expenditures at national and sectoral
levels, R&D personnel and new technologies.

216 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 9

Annex 9
R&D Deflators and Currency Converters

Introduction
1. This annex examines special methods for deflating and converting
data on R&D expenditures expressed in national currencies at current prices
to a numeraire currency.
2. Both these issues involve adjusting R&D expenditures for differences
in price levels over time (i.e. intertemporal differences) and among countries
(i.e. interspatial differences). In the case of deflators, the price differences are
intertemporal, and the question is clearly of interest both in individual
countries and for international comparisons of changes over time.

Deflation and currency conversion in the OECD’s international R&D

statistics
3. As far as possible, the same methodology should be used for both
deflation and conversion. In the absence of a full set of R&D deflators and R&D
converters for all member countries, the Manual recommends the use of the
implicit gross domestic product (GDP) deflator and GDP-PPP (purchasing
power parity for GDP), which provide an approximate measure of the average
real “opportunity cost” of carrying out the R&D.

Special R&D deflators and currency converters

4. The implicit GDP deflator and GDP-PPP are, respectively, output-based
intertemporal and interspatial deflators. This annex suggests a way to
establish special R&D deflators and PPPs either by compiling price indices
using data from price surveys of R&D (input) expenditures or by combining
proxy prices or price indices.
5. Currency converters are mainly important for international
comparisons, including, of course, those of estimated growth rates. However,
the choice of currency converters is also relevant when examining sectoral or
other breakdowns of R&D or when it is necessary to take account of relative
international variations in price levels in order to compare R&D with other
economic variables. For example, an estimate of R&D expressed as a

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 217

ANNEX 9

proportion of GDP, even if both quantities are deflated to “constant prices”

using appropriate national price indices, is still affected by differences in
relative price levels of R&D activities and all productive activities (i.e. GDP), as
compared to some international average. In other words, the ratio may be
affected by whether it is relatively expensive or inexpensive to perform R&D,
as compared to other activities.

The need for R&D deflators

6. R&D deflators are justified if it is believed that the cost of R&D has
moved in a way that is significantly different from general costs and/or if
trends in the cost of R&D have varied considerably among sectors or
industries. In general, over the long term, it is reasonable to suppose that the
implicit GDP (output) deflator would tend to increase less rapidly than a “true”
R&D (input) deflator because of productivity increases.
7. The optimal solution is to calculate special R&D deflators based on
weights and prices that are specific to R&D. The cost and complexity of
carrying out the price surveys needed for this exercise rules out using them
except for specialised analysis. The most common approach is to use weights
derived from R&D surveys combined with proxy prices.

Past OECD and national efforts

8. Work at the OECD was originally governed by five guidelines laid down
in the third edition of the Frascati Manual (OECD, 1976):
– Deflators should be produced for homogeneous sectors of the economy,
whether or not these correspond to the existing sectoral approach.
– They should be of Laspeyres form.
– In view of the relative importance of manpower in R&D activities (almost
50% of expenditure), it should receive special attention.
– Practical characteristics should take precedence over theoretical niceties.
– The best possible use should be made of existing sources of information.
9. During the 1970s, member countries and the OECD Secretariat were
active in this area, particularly in preparing deflators for the business
enterprise sector. National experts presented papers on their experience at
various meetings. Some of the methodologies were very detailed, but most
broadly followed the lines developed by the OECD in Trends in Industrial R&D in
Selected OECD Member Countries, 1967-1979 (OECD, 1979).
10. In consequence, the fourth edition of the Frascati Manual (OECD, 1981)
included a special chapter, which described some fairly simple ways of
calculating R&D deflators, using weights derived from R&D surveys and proxy
prices derived from various national or international sources. The methods

218 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 9

were presented as examples for the business enterprise sector in an imaginary

country rather than in technical form. Three methods were explained and
illustrated:
– Applying a composite index number to all expenditures using fixed
weights.
– As above, but using changing weights.
– Applying separate price indices to individual expenditure items within
subclasses of a sector.
– Further details were given on the preparation of sub-weighted indices for
labour costs. A technical presentation of the calculation of R&D deflators
was included as Annex 4.

Selection of the index-number formula

11. The recommendation to use the Laspeyres formula needs re-
examination. Hill (1988) has pointed out that theoretical advances have shown
that the index number formulas in common use (Laspeyres, Paasche, etc.)
have weaknesses with important consequences for economic analysis and
policy making. He argues for the use of chain indices, which have attractive
properties from both the theoretical and practical viewpoint and highlight the
biases of conventional fixed-weight indices of the Laspeyres or Paasche type.
12. Deflation essentially involves a comparison between situations at two
different points of time. The tendency of Laspeyres and Paasche indices to
diverge (“index number spread”) over time is well known. A chain index
should be used when the two situations being compared are dissimilar and
when linking can be achieved by passing through an intermediate point.
Ideally, the intermediate situation is one in which the pattern of relative prices
is approximated by some average of the relative prices in the two situations
being compared. In such a case, chaining reduces the index number spread
(between Laspeyres and Paasche).
13. Why chaining? In the real world, the problem faced by compilers of
index numbers is that some commodities are only found in one of the two
situations. The quantity vector is always complete (its elements are positive or
zero). However, there are many missing prices (i.e. missing commodities), and
it is impractical to suggest estimating shadow prices on a large scale, as old
products disappear as a result of obsolescence and new products appear as a
result of technological progress. This is particularly true of the commodities
likely to be included in R&D price indices.
14. The further apart the periods are, the greater the problem. The share of
total value of the expenditures in the two periods actually covered by direct
price comparisons decreases. Insisting on direct comparisons between the two

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 219

ANNEX 9

periods means accepting that price relatives can be compiled only for a small
proportion of the expenditures in both periods (in addition, the index number
spread between the Laspeyres and Paasche indices tends to be very large).
15. If a chain index is used and the amount of usable price information is
greatly increased, this is true at each link. It is also true that the amount of price
information actually used from the first and last periods will be far greater.
16. If the evolution of prices and quantities is fairly smooth, a chain-
Laspeyres will lie below a direct Laspeyres and vice versa for a chain Paasche,
thereby reducing the index-number spread. Hill describes a limiting case of a
“smooth” chain index (the “smooth” Divisia index) which eliminates the index
number problem and is quite operational.

Choosing the level of aggregation at which to deflate

17. It is possible to prepare a single R&D price index for the whole of GERD,
one for each sector or even one for individual industries in the business
enterprise sector or fields of science in the higher education sector. The choice
will depend on whether there are significant differences between the different
levels in the cost structure of R&D expenditures and whether there are
significant differences between the levels in price trends for the same cost
item. For example, it is probable that trends in wages and salaries of
researchers will be different in universities, where they are often fixed by
public-sector pay agreements, and in industrial firms. On the other hand, it is
debatable whether trends in the wages and salaries of researchers will vary
significantly between industries. The choice is also dictated by the availability
of suitable price series, whether compiled from specific price surveys or
whether proxy indices are used.

Establishing the weighting system

General
18. A simple weighting system can be derived from the recommended
breakdown by type of cost. The following shows the average breakdown in
industry in the OECD area in 1989 and 1999.

220 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 9

Percentage

1989 1999

Labour costs 43 44
Other current costs 43 45
Land and buildings 3 2
Instruments and equipment 10 9

Total 100 100

More detailed treatment of labour costs

19. Labour is typically one of the major cost items. It is therefore desirable,
whenever suitable salary price indices are available, to create a subsystem for
labour costs for each sector.

The weighting systems

20. Labour costs are not usually broken down by category of R&D
personnel, but staff and salary ratios can be used to estimate relative weights
for labour costs of different categories of personnel as follows:

Quantity ratio (%) Relative salary ratios Labour cost ratio (%)

Researchers (RSE) 50 1.00 = 50.00 59.7

Technicians 25 0.75 = 18.75 22.4
Other supporting staff 25 0.60 = 15.00 17.9

Total 100 83.75 100.0

More detailed treatment of other current costs

21. The share of other current costs has risen rapidly. Early versions of the
Manual recommended that this category should be subdivided between:
– Materials.
– Other current costs.
This distinction has since been abandoned in OECD surveys and in most
national ones. It is therefore difficult to establish a sub-weighting system.

Selecting proxy price indices

General approach
22. When it is not possible to carry out meaningful price surveys of R&D
inputs, proxy price indices for each of the classes identified in the weighting
system may be selected from the country’s national accounts or other general
sources; alternatively, an attempt can be made to identify the series whose

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 221

ANNEX 9

characteristics are most similar to R&D. As the final result will tend to be more
sensitive to the evolution of the price series than to that of the weights, the
choice of proxy price indices is the single most important decision in the
preparation of the R&D deflator and should be made with great care. It is not
possible to make firm detailed recommendations, as the amount and type of
price index data available vary from country to country. Furthermore, some
series would be relevant for a deflator for industrial R&D but not, for example,
for university R&D.

Proxies for labour costs

23. For labour costs, quantity data are usually available (number of
researchers, etc.), and two general approaches are possible: using average R&D
labour cost per total R&D person-years; using separate proxy series based on
wages and salary data. The first type of series is specific to R&D but is not very
exact if there is a significant change in the occupational qualification pattern
within the R&D labour force over the period. Given that such changes have
occurred in most member countries, it may be preferable to use the second
method. Here, it is important to select series which are as comparable as
possible with the R&D data. Thus, earnings data are generally preferable to
rates, and weekly or monthly earnings are preferable to hourly payments. The
use of salary scales as proxies for trends in labour costs poses serious
problems, notably concerning “grade drift”, changes in employers’ social
security payments and other fringe benefits, and declining “quantity” of
labour inputs owing to shorter hours and longer holidays.
24. It is usual to make a distinction between trends in the private and
public sectors. There may have to be a trade-off between breaking down
labour costs and establishing indices for separate industries. For example,
salary indices may be available for all scientists and engineers or all
technicians in industrial employment, but they may not be broken down by
individual industry. On the other hand, “average weekly wages” may be
available for these industries. The choice of method will depend on whether
the salaries of researchers move in line with those of the mass of workers in
their industry or in line with researchers in other industries.

Proxies for other current expenditures

25. This is the most difficult area to deal with. R&D surveys usually do not
reveal anything about the balance of types of expenditures included, and it is
not clear which are R&D-specific and which are industry-specific (or sector-
specific).
26. A wide range of proxy indices can be used for other current costs. For
example, the average wholesale price index for materials and supplies

222 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 9

consumed by manufacturing industry, the implicit price index of the domestic

product of industry (DPI) and the consumer price index (CPI) (excluding food
and beverages) have all been used.
27. Where indices are calculated for separate industries, it is possible to
use indices for their general input costs, but they may not be typical of R&D.
For example, it is suggested that much of the increase in current costs is due
to growth in the contracting out of support services (matching the decline in
the average number of support staff per researcher) and the greater use of
leased machinery.

Proxies for capital expenditures

28. Expenditures on land and buildings absorb a relatively low share of
R&D expenditures, and a suitable proxy index can easily be selected from the
relevant class of gross fixed capital formation (GFCF) in national accounts. The
same approach can be used for R&D expenditure on instruments and
equipment, but the extent to which such general price indices reflect changes
in R&D instrument costs is uncertain.

Currency converters for R&D

The need for special currency converters
29. Using GDP-PPPs to convert R&D expenditures to a common numeraire
currency such as the US dollar or the euro (i.e. deflating interspatially)
effectively involves adjusting to allow for differences in general price levels
between countries, not for differences in price levels for R&D. If R&D is
relatively expensive in one country, as compared with another, the use of the
GDP-PPP will distort the comparison of real expenditures on R&D.
30. As for intertemporal deflators, the ideal solution is to calculate specific
currency converters based on relative prices for R&D inputs. Once again,
carrying out the price surveys needed for this exercise (using a standard
“basket” of R&D inputs) would be both costly and complex. The more practical
solution is to use weights from R&D surveys and detailed parities from general
PPP exercises conducted by the OECD and Eurostat in the context of the
International Comparison Project (ICP) carried out under the aegis of the
United Nations Statistical Office. A major difficulty arises because the general
PPPs are calculated using a standard basket of goods and services entering
GDP or, more precisely, final demand (i.e. output), whereas R&D expenditures
are mainly inputs.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 223

ANNEX 9

Past national and OECD efforts

31. The first OECD reports on R&D statistics issued in the early 1960s used
purchasing power parities based on R&D weights and price ratios derived from
salary studies and from the 1960 benchmark calculations of general
purchasing power parities (Freeman and Young, 1965; OECD, 1968). Further
efforts were made in the late 1970s when new sets of purchasing power
parities became available. This situation was described in Chapter 7 of the
fourth edition of the Manual (OECD, 1981). Since 1990, PPPs have been
calculated every three years for OECD member countries (1993, 1996, 1999) and
annually for EU countries. Data collection for the 2002 round is under way.

The method
32. The methodology for calculating R&D PPPs should correspond to that
established in the context of the ICP.
33. The OECD and Eurostat regularly calculate PPPs for GDP (and its
expenditure components) for OECD member countries. Although the PPPs
published by the OECD are expressed in units of national currency per USD
and those published by Eurostat are in units of national currency per euro,
they are:
– Consistent (i.e. the France-Germany PPP obtained by dividing the euro PPPs
for these two countries is the same as that obtained by dividing the USD
PPPs), as “block fixity” for the EC countries has been imposed in the
calculations.
– Transitive (the PPP between countries A and B multiplied by the PPP
between countries B and C gives the PPP between countries A and C).

Choosing the level of aggregation at which to calculate R&D converters

34. Ideally, the level chosen should match that chosen for R&D deflators.
In practice, special R&D PPP rates might be calculated for the business
enterprise sector and the public sector, perhaps distinguishing government
and higher education.

The weighting system

35. As for deflators, the weighting system can be derived from the
recommended breakdown by type of cost. However, since the PPP calculations
involve the simultaneous use of the weight and price data for all the countries
included in the comparison (to ensure transitivity), it is necessary to have a
matching set of weights for all the countries in the group.

224 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 9

Choosing the proxy prices

36. Ideally, data from price surveys of a standard “basket” of R&D (input)
expenditures in each weighting category should be used. As in the case of
intertemporal price indices, such an exercise would be costly and extremely
complex and can be ruled out for all practical purposes. The next best solution
is to use proxy prices (for which the best source is the set of comparable price
data already available from the ICP), combined, if necessary, with proxy
interspatial price indices (i.e. the disaggregated parities calculated for final
expenditure components in the ICP).

Labour costs
37. No intermediate or primary input data are collected in the ICP for the
business enterprise sector, hence no data on wages and salaries. For non-
market services, however, the ICP uses input prices and thus includes data on
total employment compensation for a selected standard basket of occupations
in the public sector, notably in education, health and general government
services. This information might be supplemented by the results of
international surveys of wages and salaries of scientists and engineers or of
certain categories of business management.

Other current costs

38. Once again, the major problem is the lack of price data for
intermediate consumption, whether or not for R&D activities, in the business
enterprise sector. Certain final goods and services for which prices are
collected in the context of the ICP may also be inputs to R&D (i.e. “other
current costs”).

Capital expenditures
39. Suitable proxies for expenditures on land and buildings and on
instruments and equipment can be obtained from the ICP, subject to the
reservations already noted for estimating intertemporal R&D deflators.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 225

ANNEX 10

Annex 10
Supplementary Guidance on the Classification
of Large R&D Projects with Special Reference
to the Defence and Aerospace Industries

Introduction
1. This annex aims to provide supplementary guidance on the treatment
of large development projects in R&D statistical surveys and in returns to the
OECD. The borderline between experimental development and other industrial
activities (comprising the two overlapping groups of other innovation activities
and production and related technical activities) is described in Chapter 1,
Section 1.5.3, and Chapter 2, Sections 2.2.3 and 2.3.4 of the Manual. Chapter 1,
Section 1.5.2, and Chapter 2, Sections 2.2.2 and 2.3.3 deal with the borderline
between R&D and other related scientific and technological activities. The
distinctions are particularly difficult to establish for large, expensive
development projects in the defence and aerospace industries. The general
issues covered in this annex are nevertheless relevant to all industries.
2. Over many years, some countries have persistently had problems in
reconciling the expenditure on R&D reported by defence ministries as
contracted out to the business enterprise sector and the amount claimed as
received from government for R&D by the defence industry. In general, data
based on the government budget tend to be higher and can lead to significant
differences in the amounts of defence R&D reported in government budget
appropriations or outlays for R&D (GBAORD) and in gross domestic
expenditure on R&D (GERD). The differences have been attributed to a number
of factors, such as subcontracting and international collaborative projects, but
they have also raised questions about the correct application of the Manual’s
definitions of R&D, especially in the GBAORD series.
3. The first section of this annex compares the categories and
terminology used by the United Kingdom, the United States and France in the
defence and aerospace industries. The second section analyses examples of
defence R&D projects. Both sections also provide guidance on differentiating
between the concept of R&D as defined in the Manual and related activities
that do not count as R&D. Throughout, the term “pre-production

226 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 10

development” is used to describe non-experimental work on a defence or

aerospace product or system before it goes into production and, more
specifically, activities that are not part of scientific and technological
innovation.

Terminology and categories used in France, the United Kingdom

and the United States
4. One of the specific difficulties in applying the concepts of basic
research, applied research and experimental development to the defence and
aerospace industries is that these industries tend to have their own
terminology. This terminology differs from country to country and often cuts
across the categories used in the Manual. This section illustrates these
difficulties by comparing the Manual’s categories with terms used in the
French, UK and US defence ministries and with an industry classification used
by a major aerospace company.
5. Table 1 contains a list of the terms in common use in the defence and
aerospace industries in these countries, and Table 2 shows how some of these
terms are currently interpreted by the three countries in terms of the Frascati
Manual terminology and definition of R&D.

United Kingdom categories and terminology

6. Two categories of applied research are used in the United Kingdom’s
annual survey of government-funded R&D and are the basis of figures
reported to the OECD for GBAORD:
“Strategic research is defined as applied research which is in a subject
area which has not yet advanced to the stage where eventual
applications can be clearly specified.
“Applied research which is not strategic in nature will have quite
specific and detailed products, processes, systems, etc., as its aims.”
(Cabinet Office, 1991, Annex C, paragraphs 4-5.)
7. In an internal UK Defence Ministry study of the borderline between
R&D and pre-production development, the following non-R&D “scientific and
technical innovation” categories were identified:
– New product marketing.
– Patent work (but see below).
– Financial and organisational changes.
– Final product or design engineering.
– Tooling and industrial engineering.
– Manufacturing start-up.
– User demonstrations (but see below).

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 227

ANNEX 10

Table 1. Terminology in common use in the defence

and aerospace industries
Terminology Most likely classification1

Basic research BASIC RESEARCH

Fundamental research ..
Upstream research ..
Upstream studies ..

Applied research APPLIED RESEARCH

Demonstration model ..
Demonstration project ..
Exploratory development ..
Upstream studies ..

Experimental development EXPERIMENTAL DEVELOPMENT

Advanced development ..
Pilot plant (initially) ..
Prototype ..
Proving model ..
Proving project ..
Systems design and specification studies ..
Systems-oriented preliminary project ..
Technical demonstrations ..

Feedback R&D R&D (activity unspecified)

Research, development, test and evaluation ..

Design engineering MIXED R&D/NOT R&D

Feasibility studies R&D/pre-production
Further development R&D/pre-production
Maintenance and repairs R&D/pre-production
Project definition R&D/pre-production
Engineering development R&D/pre-production
Engineering projects R&D/pre-production
Operational development R&D/pre-production

Policy and operational studies NOT R&D

Industrial engineering Pre-production
Post-certification development Pre-production
Trial production batch Pre-production
User demonstration Pre-production
Documentation S&T innovation
Initial development S&T innovation
Manufacturing start-up S&T innovation
New product marketing S&T innovation
Patent work S&T innovation
Product engineering S&T innovation
Tooling S&T innovation
Post-design services Industrial activity
Series production Industrial activity
Related S&T activities Not R&D
S&T innovation Not R&D

1. This is only a guide. Actual classification to types of R&D as defined in this Manual
depends on the nature of the particular project and the context within which the
term is used.
Source: OECD.

228 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 10

Table 2. Current classification of French, UK and US terminology

in the Frascati Manual
Frascati Manual United Kingdom United States France

RESEARCH AND DEVELOPMENT

Basic research Basic research (O) Basic research (O) Basic research (O)
Upstream studies (O)
See also below Research
work (O)
See also Research (I)

Applied research Strategic applied research (O) Applied research (O) Applied research (O)
Demonstration project (O)
Demonstration model (I)
Specific applied research (O)
Exploratory development (O)
Development (O)
General research (I)
Preliminary project (I)
Proving project (I)
Proving model (I)
Research work (O)
Systems-oriented research (I)

Experimental Experimental development (O) Advanced technology Development (I)

development development (O)
Demonstration and Defined development (O)
validation (O) Prototype (I)
Engineering and Pilot plant (1)
manufacturing
development (O)
Management support (O)
Operational systems
development (O)

NON-R&D ACTIVITIES

Pre-production Scientific and technical S&T services (I)

development innovation (I)
S&T training
and development (I)
Other related scientific and
technical activities (O)

O = Official (Defence Ministry) terminology.

I = Industry terminology.
Source: OECD.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 229

ANNEX 10

8. The Manual (Chapter 1, Section 1.5.3) refers to a “demonstration” as

“an innovation operated at or near full scale in a realistic environment” to aid
policy or promotion, as being outside of R&D. But it is necessary to distinguish
these user demonstrations from the technical demonstrations that are part of
R&D. The French terms “demonstration project” and “demonstration model”
refer to the latter.
9. Patent work, product and design engineering, demonstrations, data
collection, testing and feasibility studies can all be part of an R&D project as
supporting work to the main project (see Chapter 2, Section 2.3.4). Likewise,
production activities can include “feedback” R&D to solve technical problems
that emerge after production has started. These are all areas in which the
distinction between “experimental development” and “pre-production
development” can be difficult and which do not necessarily follow the simple
linear model of the steps from basic research to production.
10. The UK study also identified the following non-R&D “related scientific
and technical activities”:
– General purpose data collection.
– Testing and standardisation.
– Feasibility studies.
– Policy-related studies.
– Production and related technical activities.
11. The survey concluded that “final product or design engineering”,
“feasibility studies” and “production and related technical activities” were the
areas most likely to be incorrectly included as R&D.

US categories and terminology

12. Seven categories (6.1-6.7) are defined within the US Department of
Defense (DoD) as part of its research, development, test and evaluation
(RDT&E) budget: basic research, applied research, advanced technology
development, demonstration and validation, engineering, manufacturing
development, management support and operational systems development.
All of these funds are allocated to R&D in returns to the National Science
Foundation (NSF) and hence in GBAORD returns to the OECD. Performers of
this R&D work, however, who also report to NSF (and therefore provide the
basis for GERD totals) may make different distinctions.
13. Funding for 6.1 and 6.2 activities constitute what is called DoD’s
Technology Base programme and is often referred to as the “seed corn” of
DoD’s technological capabilities. It is here that new technologies and their
potential for military application are explored and developed (sometimes over
long periods of time). Advanced technology development (6.3) activities are
meant to help technology make the transition from the laboratory to the field.

230 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 10

Taken together, activities 6.1-6.3 constitute what is called DoD’s Science and
Technology (S&T) programme.

Formal definitions for RDT&E budget activities

14. Budget Activity 6.1. Basic research is defined as systematic study
directed towards greater knowledge or understanding of the fundamental
aspects of phenomena and of observable facts, without specific applications
for processes or products in mind. It includes activities directed towards
increasing fundamental knowledge and understanding in those fields of the
physical, engineering, environmental and life sciences related to long-term
national security needs. It forms the basis for subsequent applied research
and advanced technology developments in defence-related technologies and
for new and improved military functional capabilities.
15. Budget Activity 6.2. Applied research is defined as systematic study to
gain the knowledge or understanding necessary to determine the means by
which a recognised and specific need may be met. This activity translates
promising basic research into solutions for broadly defined military needs,
short of development projects. The dominant characteristic of this category is
that it is pointed towards specific military needs with a view to developing and
evaluating the feasibility and practicability of proposed solutions and
determining their parameters.
16. Budget Activity 6.3. Advanced technology development includes all
efforts involved in the development and integration of hardware for field
experiments and tests. The results are proof of technological feasibility and
assessment of operability and producibility rather than development of
hardware for service use. Projects in this category are directly relevant to
identified military needs.
17. Budget Activity 6.4. Demonstration and validation includes all efforts
necessary to evaluate integrated technologies in as realistic an operating
environment as possible to assess the performance or cost reduction potential
of advanced technology.
18. Budget Activity 6.5. Engineering and manufacturing development
includes projects in engineering and manufacturing development for service
use which have not received approval for full-scale production. This area is
characterised by major line item projects.
19. Budget Activity 6.6. RDT&E management support includes support of
installations or operations required for general research and development use.
Included are test ranges, military construction, maintenance support of
laboratories, operation and maintenance of test aircraft and ships, and studies
and analyses in support of the R&D programme. Costs of laboratory personnel,
either in-house or contractor-operated, are assigned as a line item in the basic

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 231

ANNEX 10

research, applied research or advanced technology development programme

areas, as appropriate.
20. Budget Activity 6.7. Operational system development includes those
development projects in support of development acquisition programmes or
upgrades still in engineering and manufacturing development, but which
have received approval for production. This area also includes major system
testing and research into upgrades of existing weapon systems.
21. The US DoD reports major systems development activities (defined to
encompass budget activities 6.4 to 6.7) separately from advanced technology
development (6.3). In reporting to the OECD, all defence development activities
(6.3 through 6.7) are categorised by NSF as “experimental development”.
Most of the work categorised as advanced technology development (6.3),
demonstration and validation (6.4), engineering and manufacturing
development (6.5) is undoubtedly “experimental development”. However,
since “operational systems development” (6.7) supports development of
projects that have been “approved for production”, at least some of these
funds may be considered pre-production development and therefore fall
outside of the definition of experimental development.

French categories and terminology

22. In the French Defence Ministry the Manual’s standards are applied but
the classification of a particular project by type of activity depends on its place
in the decision-making process as well as on the nature of the work. Thus the
term “les études en amont” (upstream studies) covers basic and applied
research, including research study (straddling basic and applied) and
exploratory development (defining the operational application of new
technological developments). The term “développements décidés” (defined
developments) is used for experimental development. This includes the task
of perfecting prototypes destined for production and operational use, i.e. all
work prior to the actual start of production.
23. In the French aerospace industry, the term “research” is used to cover
both basic and applied research. The terms “development”, “prototype” and
“pilot plant” would generally fall within the Manual’s concept of experimental
development. “Scientific and technical services” and “Education and
development” would generally be excluded from R&D. However, decisions on
the precise classification of work are checked by the authorities with the
company concerned to ensure compliance with the Manual’s terminology.

232 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 10

Examples
24. This section looks at some examples of major technological development
projects in the defence and aerospace industries. The objective is to show how
the Manual’s categories may be applied and where difficulties may arise.

Example A
25. Project description:
To establish the feasibility and value of non-equilibrium device
structures and to make available the unique properties of narrow-gap
semiconductors for opto-electronics and high-speed logic functions at
ambient temperatures. If successful, the new devices will offer
substantial performance advantages over both silicon and gallium
arsenide devices for future high-speed electronic applications. The
plan is to identify useful non-equilibrium devices, to confirm some of
the key parameters of narrow-gap semiconductor materials, to use
these to predict device performance and, finally, having identified
suitable devices, to research their practical realisation and
characterise them in simple form.
26. This project is currently at the strategic applied research stage, since it
is directed at a group of applications but not a particular application. It would
have followed on from basic research that discovered non-equilibrium device
structures, probably in a university. A potential range of applications in opto-
electronics and high-speed logic functions is hypothesised, and the research
investigates the possible applications. Testing is involved “to confirm some of
the key parameters”, but this testing could well be part of the applied research
stage of exploring unknown areas merely suggested by the basic research.
27. Once suitable devices are identified, their “practical realisation” would
involve experimental development. Early prototype models to “characterise
them in simple form” could be part of this experimental development stage.
Later models and customer or user demonstration procedures (see
paragraph 7 above) would be pre-production development rather than
experimental development.

Example B
28. Project description:
X is a Short Range Air Defence (SHORAD) missile system, planned to be
evolutionary and therefore capable of responding to a developing
threat. X2 is being developed as the latest member of the X family. The
project involves development and production of the new X2 missile
and new ground equipment. The development programme is for a

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 233

ANNEX 10

large system that requires the interaction of a number of complex

technologies such as electro-optics, command links and both tracking
and surveillance radars. This will allow the operator to track more
targets, with better discrimination, and to fire multiple missiles if
required. Under single missile operation, the thermal picture from the
electro-optic (EO) tracker can be used to guide the missile all the way
to the target, but another missile cannot be fired until the EO tracker is
free again. Under multiple missile operation, a first missile may be
guided initially by the EO tracker, but then handed over to a radar
tracker for transmission to the target, thus freeing the EO tracker to
start guiding a second missile before the first has hit its target. The
programme endeavours to integrate the subsystems from high-
technology subcontractors under the guidance of a single prime
contractor.
29. The development of “Mark II” models is common in defence technology,
and it is not necessarily easy to decide how much of such development is
experimental. In this case, the difference between the single missile system and
the multiple missile system is big enough to suggest that the development of
the latter is experimental development. However, the project (as is common in
defence technology) is the development of a complex system involving different
pieces of equipment and different technologies. In theory, the project could be
broken down into a number of subprojects, some of which are subcontracted.
Some of the subprojects, involving the application of existing technology to
existing equipment, may not be R&D. A subcontractor working on such a
subproject should not count it as experimental development. The funding
organisation and the main contractor, however, may be unable to break project
expenditure down in this way.
30. Example B involves both experimental development and production. It
would be necessary to separate out the production aspect at the later stages of
the project in order to distinguish the borderline pre-production and
production elements.

Example C
31. Table 3 shows the stages of an armoured tank development
programme and a subsequent upgrade development programme.
32. Concept design appears to be at the borderline of applied research and
could be achieved at the end of an applied research project.
33. In the original development programme, detailed design and systems
integration appear to be experimental development. Systems integration
involves testing, and this is part of the experimental development stage. If the
upgrade development programme has to go through all the stages listed, the

234 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 10

Table 3. Development of an armoured tank

1. ORIGINAL DEVELOPMENT PROGRAMME

User’s operational requirements What is expected of the kit in the field
Detailed specification What the kit needs to do to achieve its role
Concept design/proof of principle The initial design to demonstrate the specification can be met
Detail design Design subsystems, identify equipment/subcontractors best
suited to achieve specification, looking first to existing kit, then
modifying existing kit and if necessary designing a new one
Systems integration Assembling all subsystems and testing to ensure all function
together as required
Trials Carry out extensive trials and testing to demonstrate
achievement of specifications
Re-design/modify Incorporate modifications identified as a result of trials
User demonstration Customer carries out own trials to ensure product meets
specifications to his satisfaction
Acceptance of design Production Build Standard agreed, Technical Data Pack
prepared
Production Series production to agreed build standard
Post-design services Modification to production build standard after entry into
service. This involves design of modifications and production
of modification kits

2. UPGRADE DEVELOPMENT PROGRAMME

Identify enhanced operational requirements What is expected of the equipment following upgrade
Detailed improvements specifications What the kit should be capable of following improvements
Systems design Design of the improved system utilising existing vehicle
equipment and new equipment from the development
programme
Systems integration Assembling all subsystems and testing to ensure all function
together as required
Trials Carry out extensive trials and testing to demonstrate
achievement of improvement specifications
Redesign/modify Incorporate modifications identified as a result of trials
User demonstration Customer carries out own trials to ensure product meets
specifications to his satisfaction
Acceptance of design Production Build Standard agreed, Technical Data Pack
prepared
Production of modification kits/upgrade vehicles Series production/modification to agreed build standard
Post-design services Modification to improved standards of material already in use.
This phase requires design of modifications and production
of modification kits

Source: OECD.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 235

ANNEX 10

probability is that a substantial improvement is involved and the work counts

as experimental development. Assuming the upgrade is work of this kind, the
systems design and systems integration stages again appear to be
experimental development.
34. There is a “feedback” situation with the trial and redesign/
modification stages. Much of this work would be experimental development.
Some of it might not be.
35. The user demonstration and acceptance of design stages appear to be
pre-production, rather than experimental development, and thus outside of R&D.
36. The post-design services stage is comparable with the re-design/
modification stage. It could involve some experimental development but in
general it would not.

Example D
37. Project description:
A fighter bomber known as QWERTY has successfully passed through
the research, technology demonstration, project design and initial
development stages to flight testing of a pre-production aircraft.
Further airframes are now required to develop and integrate the
vehicle into air offence/defence systems in order to ensure full
operational capability. This may require up to ten additional aircraft.
38. Stage one is development of the integrated air offence/defence system.
This stage involves bringing together developed components and subsystems
that have not previously been integrated in this context. It requires a large
flight test programme of the ten aircraft, which is potentially very expensive
and the main cost element prior to production. Some of the work
commissioned during this stage does not have the element of novelty
necessary for classification as R&D. Expenditure on this stage should therefore
be split between:
– Experimental development (R&D).
– Pre-production development (non-R&D).
39. The distinction between these two categories requires engineering
judgement as to when the element of novelty ceases and the work changes to
routine development of an integrated system. The description of this stage of
the project shows, once again, how difficult it can be to distinguish
experimental development from pre-production development. The need for
“engineering judgement” underlines the difficulty.
40. Stage two covers trials of the integrated air offence/defence system.
Once the system is proved to work at stage one, the development project may
move on to produce a trial production batch for operational trials. The full

236 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 10

production order depends on their success. According to the Manual, this

work is not R&D but pre-production development. However, problems may
arise during the trials, and new experimental development may be needed to
solve them. This work is described in the Manual as “feedback R&D” and
should be included as R&D.
41. Stage three concerns full production. This is not R&D.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 237

ANNEX 11

Annex 11
Correspondence between the Categories of R&D
Personnel by Occupation in the Frascati Manual
and ISCO–88 Classes

1. Table 1 below gives an indication of the ISCO-88 classes in which

researchers and other categories of R&D personnel are found. It should be read
only in one direction, e.g. researchers are found among health professionals
(ISCO-88 minor group 222) but not all health professionals are researchers.
Also, the table does not capture certain categories of R&D personnel, i.e. those
in the “Armed Forces” (ISCO-88 major group 0) and post-graduate students
who are not registered under a specific job.

238 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ANNEX 11

Table 1. Correspondence between Frascati Manual categories

of R&D personnel by occupation and ISCO–88 classes
The International Standard Classification of Occupations (ISCO) (ILO, 1990) consists
of ten major groups at the top level of aggregation, subdivided into 28 sub-major groups
(and 116 minor groups and 390 unit groups)

RESEARCHERS – ISCO-88 CLASSES (sub-major and minor groups):

21 Physical, mathematical and engineering science professionals
211 Physicists, chemists and related professionals
212 Mathematicians, statisticians and related professionals
213 Computing professionals
214 Architects, engineers and related professionals
22 Life science and health professionals
221 Life science professionals
222 Health professionals (except nursing)
23 Teaching professionals
231 College, university and higher education teaching professionals
24 Other professionals
241 Business professionals
242 Legal professionals
243 Archivists, librarians and related information professionals
244 Social science and related professionals

Plus Unit group 1237 Research and development department managers

TECHNICIANS AND EQUIVALENT STAFF – ISCO-88 CLASSES (sub-major and minor groups):
31 Physical and engineering science associate professionals
311 Physical and engineering science technicians
312 Computer associate professionals
313 Optical and electronic equipment operators
314 Ship and aircraft controllers and technicians
315 Safety and quality inspectors
32 Life safety and quality inspectors science and health associate professionals
321 Life science technicians and related associate professionals
322 Modern health associate professionals (except nursing)
Plus Unit group 3434 Statistical, mathematical and related associate professionals

OTHER SUPPORTING STAFF – ISCO-88 CLASSES (major groups):

4 Clerks
6 Skilled agricultural and fishery workers
8 Plant and machine operators and assemblers
Plus Minor group 343
Administrative associate professionals (except Unit group 3434)1
1 Legislators, senior officials and managers n.e.c.

1. Statistical, mathematical and related associate professionals (here included in

“technicians and equivalent staff”).
Source: OECD.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 239

ACRONYMS

Acronyms

CEC Commission of the European Communities

CERN European Organization for Nuclear Research
COFOG Classification of the purposes of government
DPI Domestic product of industry
EC European Community
ECE United Nations Economic Commission for Europe
EU European Union
FTE Full-time equivalence on R&D
GBAORD Government budget appropriations or outlays for R&D
GDP Gross domestic product
GERD Gross domestic expenditure on R&D
GFCF Gross fixed capital formation
GNERD Gross national expenditure on R&D
GUF General university funds
HERD Higher education R&D
HRST Human resources for science and technology
ICP International Comparison Project
ICT Information and communication technology
ILO International Labour Organisation
ISCED International standard classification of education
ISCO International standard classification of occupations
ISIC International standard industrial classification
NABS Nomenclature for the analysis and comparison of scientific
programmes and budgets
NACE General Industrial Classification of Economic Activities
within the European Union
NESTI Working party of national experts on science and technology
indicators
NPI Non-profit institution
NPSH Non-profit institutions serving households
NSE Natural sciences and engineering
NSF National Science Foundation
PNP Private non-profit
PPP Purchasing power parity
R&D Research and experimental development
RD&D Research, development and demonstration
RDT&E Research, development, test and evaluation
RSE Researchers
SCI Science Citation Index
SITC Standard international trade classification
SNA System of National Accounts
SSH Social sciences and humanities

240 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 ACRONYMS

STA Scientific and technological activities

STET Scientific and technical education and training
STID Scientific and technological information and documentation
STS Scientific and technological services
TEP Technology-Economy programme
TBP Technology Balance of Payments
UN United Nations
UNESCO United Nations Educational, Scientific and Cultural Organization
VAT Value-added tax

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 241

BIBLIOGRAPHY

Bibliography
Cabinet Office (1991),
R&D “91. Annual Review of Government Funded Research and Development, HMSO
Publications Centre, London.
Carson, C.S. and B.T. Grimm (1991),
“Satellite Accounts in a Modernized and Extended System of Economic
Accounts”, Business Economics, January.
Commission of the European Communities (CEC),
International Monetary Fund, Organisation for Economic Co-operation and
Development, United Nations and World Bank (1994), System of National
Accounts, 1993.
Eurostat (1986),
“Nomenclature for the Analysis and Comparison of Scientific Programmes and
Budgets (NABS): 1983 version”, Luxembourg.
Eurostat (1990),
“Council Regulation (EEC) No. 3037/90 of 9 October 1990 on the Statistical
Classification of Economic Activities in the European Community”, Official
Journal of the European Communities, No. L 293/1, 24 October.
Eurostat (annual),
Research and Development: Annual Statistics, Luxembourg.
Eurostat (1994),
“Nomenclature for the Analysis and Comparison of Scientific Programmes and
Budgets”, Luxembourg.
Freeman, C. and A. Young (1965),
The Research and Development Effort in Western Europe, North America and the Soviet
Union: An Experimental International Comparison of Research Expenditures and
Manpower in 1962, OECD, Paris.
Freeman, C., ed. (1987),
Output Measurement in Science and Technology: Essays in Honor of Yvan Fabian,
Elsevier Science Publishers B.V., Amsterdam.
Glennan, T.K., Jr., W.F. Hederman, Jr., L.L. Johnson and R.A. Rettig (1978),
The Role of Demonstration in Federal R&D Policy, The Rand Corporation.
Hatzichronoglou, T. (1997),
“Revision of the High-technology Sector and Product Classification”, OECD, STI
Working Paper 1997/2), Paris.
Hill, P. (1988),
“Recent Developments in Index Number Theory and Practice”, OECD Economic
Studies, No. 10 (Spring).

242 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 BIBLIOGRAPHY

International Labour Organization (1968),

International Standard Classification of Occupations (ISCO), Geneva.
International Labour Organization (1990),
International Standard Classification of Occupations: ISCO-88, Geneva.
Nordforsk (1976),
Statslige udgifter til forskning og udviklingsarbejde i de nordiske lande 1975. En budge
tanalyse, Stockholm.
Nordforsk (1983),
Retningslinier for analyse af statslige bevillinger til forskning og udviklingsarbejde,
Stockholm.
Nordforsk (1986),
R&D Statistics in the Higher Education Sector: Work on Improved Guidelines, Oslo.
OECD (1963),
“Proposed Standard Practice for Surveys of Research and Development: The
Measurement of Scientific and Technical Activities”, Directorate for Scientific
Affairs, DAS/PD/62.47, Paris.
OECD (1968),
Statistical Tables and Notes (“International Statistical Year for Research and
Development: A Study of Resources Devoted to R&D in OECD Member countries
in 1963/64”), Vol. 2, Paris.
OECD (1970),
“Proposed Standard Practice for Surveys of Research and Experimental
Development: The Measurement of Scientific and Technical Activities”, DAS/
SPR/70.40, Directorate for Scientific Affairs, Paris.
OECD (1976),
Proposed Standard Practice for Surveys of Research and Experimental Development:
“Frascati Manual”, The Measurement of Scientific and Technical Activities
Series, Paris.
OECD (1979),
Trends in Industrial R&D in Selected OECD Member Countries 1967-1975, Paris.
OECD (1981),
Proposed Standard Practice for Surveys of Research and Experimental Development:
“Frascati Manual 1980”, The Measurement of Scientific and Technical Activities
Series, Paris.
OECD (1984),
OECD Science and Technology Indicators: No. 1 – Resources Devoted to R&D, Paris.
OECD (1986),
OECD Science and Technology Indicators: No. 2 – R&D, Invention and Competitiveness,
Paris.
OECD (1989a),
OECD Science and Technology Indicators, No. 3 – R&D, Production and Diffusion of
Technology, Paris.
OECD (1989b),
R&D Statistics and Output Measurement in the Higher Education Sector: “Frascati
Manual” Supplement, The Measurement of Scientific and Technological
Activities Series, Paris.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 243

BIBLIOGRAPHY

OECD (1990),
“Proposed Standard Method of Compiling and Interpreting Technology Balance
of Payments Data: TBP Manual 1990”, The Measurement of Scientific and
Technological Activities Series, Paris.
OECD (1992),
OECD Proposed Guidelines for Collecting and Interpreting Technological Innovation Data
– Oslo Manual, Paris.
OECD (1994a),
Proposed Standard Practice for Surveys of Research and Experimental Development,
“Frascati Manual 1993”, The Measurement of Scientific and Technological
Activities Series, Paris.
OECD (1994b),
“Using Patent Data as Science and Technology Indicators – Patent Manual 1994:
The Measurement o f Scientific and Technolog ical Activities” , OC DE/
GD(94)114,1994, Paris
OECD/Eurostat (1995),
The Measurement of Human Resources Devoted to Science and Technology – Canberra
Manual: The Measurement of Scientific and Technological Activities, Paris.
OECD/Eurostat (1997a),
Proposed Guidelines for Collecting and Interpreting Technological Innovation Data – Oslo
Manual, The Measurement of Scientific and Technical Activities Series, Paris.
OECD (1997b),
Manual for Better Training Statistics – Conceptual, Measurement and Survey Issues,
Paris.
OECD (1999),
Classifying Educational Programmes, Manual for ISCED-97 Implementation in OECD
Countries, Paris.
OECD (2001),
Measuring Expenditure on Health-related R&D, Paris.
OECD (2002),
“Measuring the Information Economy”, Paris.
OECD (Biannual),
Main Science and Technology Indicators, Paris.
OECD (every second year),
Basic Science and Technology Statistics, Paris.
OECD (every second year),
OECD Science, Technology and Industry Scoreboard, Paris.
OECD (every second year),
OECD Science, Technology and Industry Outlook, Paris.
OECD (every second year),
OECD Information Technology Outlook, Paris.
Okubo, Y. (1997),
“Bibliometric Indicators and Analysis of Research Systems, Methods and
Examples”, OECD, STI Working Paper 1997/1, Paris.

244 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 BIBLIOGRAPHY

RICYT/OEC/CYTED (2001),
“Normalización de Indicadores de Innovación Tecnológica en América Latina y
el Caribe: Manual de Bogotá”, available at: www.ricyt.edu.ar/
UNESCO (1976),
“International Standard Classification of Education (ISCED)”, COM/ST/ISCED,
Paris.
UNESCO (1978),
“Recommendation Concerning the International Standardization of Statistics
on Science and Technology”, Paris, November.
UNESCO (1984a),
Guide to the Collection of Statistics on Science and Technology, Rev. 1, ST 84/WS/19,
December.
UNESCO (1984b),
Manual for Statistics on Scientific and Technological Activities, ST.84/WS/12, Paris.
UNESCO (1984c),
“Guide to Statistics on Scientific and Technological Information and
Documentation (STID) (Provisional)”, ST.84/WS/18, Paris.
UNESCO (1997),
ISCED (International Standard Classification of Education), Paris
UNESCO (Annual until 1999),
UNESCO Statistical Yearbook, Paris.
United Nations (1968),
A System of National Accounts, Studies in Methods Series F, No. 2, Rev. 3,
New York.
United Nations (1982),
Provisional Guidelines on Standard International Age Classifications, Statistical
Papers, Series M, No. 74, New York.
United Nations (1986),
Standard International Trade Classification Revision 3, Statistical Papers Series M,
No. 34, Rev. 3, New York.
United Nations (1990),
International Standard Industrial Classification of All Economic Activities, Statistical
Papers Series M, No. 4, Rev. 3, New York.

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 245

INDEX BY PARAGRAPH NUMBER

Index by Paragraph Number

Abroad
Definition of 229
Geographic area of origin or destination of funds 233-235
Government budget appropriations or outlays
for R&D (GBAORD) 496
Principal institutional sub-classification 231-232
Activities to be excluded from R&D 65-67
Administration and other supporting 26, 81-83, 131-132,
activities 289-293, Table 5.1
After-sales services and trouble-shooting Table 2.3
Applied research 64, 246-248
Definition of 245
Basic research 64, 241-242
Definition of 240
Oriented basic research 243
Bibliometrics Ann. 7 (21-25)
Biotechnology 60
OECD definition of Ann. 4 (51, 55-56)
Biotechnology-related R&D Ann. 4 (51-56)
Borderline between R&D and other activities
Administration and other supporting activities 131-132
R&D and education and training at ISCED level 6 94, Table 2.2
R&D and education and training; cases 86-88
R&D and other industrial activities 110-112
R&D and other industrial activities; cases Table 2.3
R&D and related scientific and technological
activities; cases 104, 110, 113
Budgetary funds
Government budget appropriations or outlays
for R&D (GBAORD) 487-490
Buildings as capital R&D expenditure 377-378, 385, Ann. 2 (53)

246 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 INDEX BY PARAGRAPH NUMBER

Business enterprise sector

Criteria for classification 174-176
Definition of 163
Foreign multinational enterprise 181
Health-related R&D Ann. 4 (14-18, Table 2)
Private enterprise 179
Public enterprise 179-180
Size of institution 182-183
Statistical unit 170-173
Survey methodology and procedures 435-442
Type of institution 177-179
Canberra Manual
OECD Manual on the Measurement of Human
Resources Devoted to S&T 328, Ann. 7 (48)
Capital expenditures
Computer software 382-383
Conventions for distinguishing between current
and capital items 384
Definition of 374-376
Identifying the R&D content of capital expenditures 385
Instruments and equipment 381
Land and buildings 377-380
Libraries 387-388
Sale of R&D capital goods 386
Classification of the purpose of government (COFOG) 188, Ann. 3 (20)
Classification systems for R&D 42-49
Clinical trials 130
Computer software
Capital expenditures 382-383
Control and care of the environment 285, 505
Current cost 360
Definition of 358-359
Government budget appropriations or outlays
for R&D (GBAORD) 485
Indirectly paid costs 365
Other current costs 364
Rents for research facilities 366-368
Social security costs and pensions for R&D personnel 369-370
Value-added tax (VAT) 371-373
Labour costs of R&D personnel 361-363
Defence R&D 281-284, 515, Ann. 10 (1-41)
Deflators and currency converters for R&D Ann. 9 (1-39)

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 247

INDEX BY PARAGRAPH NUMBER

Demonstration 23
Depreciation (exclusion of) 34, 374-375
Development of social system 109
Disinvestment in R&D 386
Education and training 68
Estimation procedures 463-472
EUROSTAT (Statistical Office of the European Union) Ann. 6 (8-9)
Expenditure account (R&D in SNA) Ann. 3 (28-30)
Experimental development 64
Definition of 249-250
Extramural expenditures
Definition of 408-412
Performer-based and source-based reporting 413-421
Feasibility studies 73
Feedback R&D 123
Field of science and technology (FOS) 200-202, 222-226, 273-276, Table 3.2,
Ann. 2 (42), Ann. 4 (21-22, 40, 42, 44-45)
Fiscal incentives for R&D 401, 493
Frascati Manual
Acknowledgements Ann. 1 (26-33)
Brief history and origins Ann. 1 (1-15)
Main changes in the sixth edition Ann. 1 (16-25)
Full time equivalence (FTE) 331-332, 335-337, Ann. 2 (43-44)
Calculation of full-time equivalence 343-345
Definition of working time 341-342
Fixed date for calculation 335
Higher education sector 338-340
Person-years for calculation 333-334
Functional distribution of R&D
Approach 236-237, Table 4.1
Fields of science and technology (FOS) 273-276
Product field 267-272
Socio-economic objectives (SEO) 277-280
Type of R&D 238-256
Fundamental research (see: Basic research)
Gender breakdown of R&D personnel 347
General purpose of data collection 71, 103
General university funds (GUF)
(see: Public general university funds) Ann. 2 (36)
Globalisation Indicators
OECD Manual of 181
Globalisation of R&D and R&D co-operation 39-41

248 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 INDEX BY PARAGRAPH NUMBER

Government budget appropriation or outlays 53-57, 474-496,

for R&D (GBAORD) Ann. 4 (8-10, 13, Table 1)
Treatment of multi-annual projects 495
Comparison with GERD 520-526
Government sector
Classification 188-190
Coverage of 185-187
Criterion for classification 190
Definition of 184
Government budget appropriation or outlays
for R&D (GBAORD) 484
Health-related R&D Ann. 4 (29-30)
Level of government 192
Statistical unit 189
Sub-classification 191
Survey methodology and procedures 443-445
Type of institution 193
Gross domestic expenditure 423-425, Table 6.1,
on R&D (GERD) Ann. 4 (8, 14-15, 33)
Gross domestic product (GDP) Ann. 3 (11)
Gross fixed capital formation (GFCF) Ann. 3 (25)
Gross national expenditure on R&D (GNERD) 426-427, Table 6.2
Headcount 326-328
Health service industry Ann. 4 (16-17)
Health-related R&D 58, Ann. 4 (1-33, Table 1-3)
Higher education sector
Borderline research institutions 214-221
Coverage of 207-209
Definition of 206
Estimates of R&D shares; expenditures
and personnel in surveys
Borderline R&D activities Ann. 2 (22)
Central administration data (use of) Ann. 2 (20, 37-42)
Direct government funds Ann. 2 (62, 69-70)
Estimation of shares of R&D Ann. 2 (4-5)
General university funds (GUF) Ann. 2 (61-64)
Instruments and equipment Ann. 2 (52)
Labour costs Ann. 2 (47-50, 54)
Land and buildings Ann. 2 (53)
Method based on other sources Ann. 2 (25-33)
Other current costs Ann. 2 (51, 62, 65-68)
Response rates Ann. 2 (23-24)
Source of funds Ann. 2 (55-60)

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 249

INDEX BY PARAGRAPH NUMBER

Survey procedures Ann. 2 (1-3)

Time-use surveys Ann. 2 (6-19)
Use of coefficients Ann. 2 (34-37, 43-44)
Health-related R&D Ann. 4 (23, 25-26)
Other institutional sub-classifications 227-228
Statistical unit 225
Survey methodology and procedures 447-448
High-technology products and industries Ann. 7 (26-33)
Hospitals Ann. 4 (32)
Survey methodology and procedures 449-450
Human resources for science and technology (HRST) Ann. 7 (41-48)
ICT-related R&D Ann. 4 (34-41)
Identification of R&D in software, social sciences
and service activities 25, 133-151
Indirectly paid current costs 365
Indirect supporting activities 83, 131-132, 289-293, Table 5.1
Industrial classification of economic activities
within the European Union (NACE) 169
Industrial design and drawing 124-125, Table 2.3
Industrial engineering and tooling up 126-129, Table 2.3
Information and communication technology (ICT) 59
Information society statistics and indicators Ann. 7 (49-54)
Innovation statistics Ann. 7 (34-40)
Institutional classification 152
Reporting unit 153
Sectoring 156-162, Figure 3.1
Statistical unit 154-155
Intermediate consumption (treatment of R&D in SNA) Ann. 3 (26-27)
International Standard Classification 297, 305, 323,
of Education (ISCED) Table 5.2
International Standard Classification 297, 300-301, 307,
of Occupation (ISCO) 310-311, Ann. 11 (Table 1)
International Standard Industrial 169, 174-176, 189, 261,
Classification (ISIC) Table 3.1, Ann. 4 (14, 36-38, 43, Table 2)
Intramural expenditures
Definition of 358-359
Labour costs of R&D personnel 361-363
Large-scale projects and costly “pilot plants”118-119, Ann. 10 (1-41)
Loans and indirect funding of industrial R&D
Government budget appropriation or outlays
for R&D (GBAORD) 492
Inclusion in R&D surveys 400

250 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 INDEX BY PARAGRAPH NUMBER

Local government 185, 192, 484

Measures of R&D inputs 29
Methodological manuals and documents 9, 16, Table 1.1,
of the OECD Ann. 7 (12, 20, 25, 33, 40, 48)
Mining and prospecting 106-108
Model surveys of biotechnology Ann. 4 (47-56)
National R&D efforts (see also: GERD) 38, 423-425, Table 6.1
Natural sciences and engineering
Type of R&D; examples 253
Nomenclature for the analysis and comparison
of scientific programmes and budgets (NABS) 502, Ann. 4 (8, 11, 41)
Non-profit institutions (NPI) 166-168
Nordforsk (Nordic Industrial Fund) Ann. 6 (10-12)
Objectives of R&D (see also: Socio-economic objectives)
Other current costs 364
Other industrial activities 78
Other innovation activities 79
Other related scientific and technological activities 69-77
Other supporting staff 310-311, Ann. 11 (1, Table 1)
Definition of 309
Overhead costs 26, 83, 131-132, 292-293, 296, 364
Patent and licence work 75, Table 2.3
Patent statistics Ann. 7 (4-12)
Personal education of academic staff 99-101
Pilot plants 111, 116-119, Table 2.3
Policy-related studies 76, 119
Postgraduate students at the PhD level 89-94
Post-secondary education 210
Private enterprise 164
Private non-profit (PNP) sector
Coverage of 195-199
Definition of 194
Fields of science and technology (FOS) 200-204
Health-related R&D Ann. 4 (27-28)
Statistical unit 203
Survey methodology and procedures 446
Product field 257-261
Classification; ICT-related R&D Ann. 4 (39)
Criteria for distribution 262-266
Recommendation 272
Use of product criterion 267

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 251

INDEX BY PARAGRAPH NUMBER

Production and related activities 80

Projections and up-to-date estimates of R&D Ann. 8 (1-26)
Prototypes 114-115, Table 2.3
Public enterprise 165
Public general university funds (GUF) 405-406,
as source of funds Ann. 2 (61-64)
Government budget appropriation or outlays
for R&D (GBAORD) 492
Public inspection control, enforcement of standards, regulations Table 2.3
Purely R&D-financing activities 82
R&D administration and other supporting activities 26
R&D and related activities 84-85
Criteria for distinguishing Table 2.1
R&D and technological innovation 21-24
R&D coefficients Ann. 2 (48-49, 54)
R&D expenditures 34-36, 356-357
Regional breakdown of 422
R&D facilities 37
R&D in defence and aerospace industries Ann. 10 (1-41, Table 1-3)
R&D in software development, social sciences
and humanities and in service activities 133-151
R&D personnel 30-33
Below the PhD level (ISCED level 5A) 314
Categories of 297-299
Classification by level of formal qualification 312, Table 5.2
Classification by occupation 300
Cross-classification by occupation and qualification 352-354, Table 5.4
Definition of 294-296
Diplomas of secondary education (ISCED level 3) 317
Measurement and data collection 325
Other qualifications 318
PhD level (ISCED level 6) 313
Post-secondary non-tertiary diplomas (ISCED level 4) 316
R&D and indirect support activities 289-293, Table 5.1
Recommended national aggregates and variables 346-351, Table 5.3a-b
Regional breakdown 355
Tertiary level diplomas (ISCED level 5B) 315
Treatment of postgraduate students 319-324
R&D surveys
Reliability of data and international comparability 50-52
Regional distribution of R&D 61, 355, 422, Ann. 5 (1-6)
Rents for research facilities 366-368

252 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 INDEX BY PARAGRAPH NUMBER

Reporting to the OECD or other international organisations 473

Reporting unit 153
Research and experimental development (R&D)
Definition of 17-18, 63-64
Researchers 302-305, Ann. 11 (1, Table 1)
Age of 348
Definition of 301
RICYT (Red Iberoamericana de Indicadores
de Ciencia y Technologia) Ann. 6 (13-15)
Routine software development 77
Routine tests Table 2.3
Sale of R&D capital goods 386
Sampling
Business enterprise sector 441
Higher education sector Ann. 4 (12-19)
Satellite account in the SNA Ann. 3 (31-32)
Scientific and technical information services 70
Scientific and technological activities (STA) 19-20
Scope of R&D surveys 431
Sectors
Choice of sectors 157-159
Problems of sectoring 160-162, Figure 3.1
Reasons for sectoring 156
SNA sectors and Frascati Manual Ann. 3 (14-19, Table 1-3)
Service activities
Criteria for identifying R&D 149
Examples of R&D in banking and some other service activities 150-151
Identification of R&D 25, 133-134
Problems for identifying R&D 145-148
Social sciences and humanities (SSH)
Examples of type of research 254-255, Table 4.2
General examples of R&D 143-144
Identification of R&D 25, 133-134
Inclusion of 27-28
Social security costs and pensions for R&D personnel 369-370
Socio-economic objectives (SEO) 277-288, Ann. 4 (21-22, 44, Table 3)
Agricultural production and technology 508
Control and care of the environment 505
Criteria for distribution 497-501
Defence 515
Exploration and exploitation of space 511
Exploration and exploitation of the Earth 503

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 253

INDEX BY PARAGRAPH NUMBER

Industrial production and technology 509

Infrastructure and general planning of land use 504
NABS 502, Table 8.1
Nordforsk Table 8.2
Non-oriented research 513
Other civil research 514
Principal areas of difficulty 516-519
Protection and improvement of human health 506
Protection, distribution and rational utilisation of energy 507
Research financed from general university funds 512
Social structures and relationships 510
Software
Examples of R&D 140-142
Examples of types of R&D 256
Identification of R&D 25, 133-139
In SNA Ann. 3 (27)
Source of funds
Criteria for identifying flows of R&D funds 393
Direct transfer 394-401
Identifying the sources of flows of R&D funds 403
Method of measurement 389-392
Public general university funds (GUF) 405-407
Subcontracting and intermediaries 404
Transfer both intended and used for R&D 402
Space exploration 105
Specialised health care 74, 97-98
Statistical unit (choice of)
Business enterprise sector 170-173
Government sector 189
Higher education sector 225
Private non-profit (PNP) sector 203
Strategic research
Element of applied research 248
UK definition of Ann. 10 (6)
Supervision of students 95-96
Survey methodology and procedures 428-430
System of national accounts (SNA) 13, 157-160,
and Frascati Manual Ann. 3 (1-32, Table 1-5)
Target population and survey respondents 432-450
Technicians and equivalent staff 307-308, Ann. 11 (1, Table 1)
Definition of 306
Example of tasks 308

254 FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002

 INDEX BY PARAGRAPH NUMBER

Technology balance of payments (TBP) Ann. 7 (13-20)

Testing and standardisation 72
Tooling up and industrial engineering 126-129, Table 2.3
Trial production 120-121, Table 2.3
Trouble-shooting 122, Table 2.3
Types of R&D
Criteria for 251-252
UNESCO (United Nations Educational, Scientific
and Cultural Organisation) Ann. 6 (2-7)
Universities (see also: Higher education sector)
University hospitals and clinics 211-213
Utility of functional distributions Table 4.1
Value-added tax (VAT) 371-373
Working with respondents 451-462

FRASCATI MANUAL 2002 – ISBN 92-64-19903-9 – © OECD 2002 255

OECD PUBLICATIONS, 2, rue André-Pascal, 75775 PARIS CEDEX 16
PRINTED IN FRANCE
(92 2002 08 1 P) ISBN 92-64-19903-9 – No. 52703 2002