Population Change Labor Markets And Sustainable

Growth Towards A New Economic Paradigm A Mason

And M Yamaguchi Eds download

https://ebookbell.com/product/population-change-labor-markets-
and-sustainable-growth-towards-a-new-economic-paradigm-a-mason-
and-m-yamaguchi-eds-1837002

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Chinese Research Perspectives On Population And Labor Volume 3 Policy

Changes For The Goals Of 2020 1st Edition Fang Cai

https://ebookbell.com/product/chinese-research-perspectives-on-
population-and-labor-volume-3-policy-changes-for-the-goals-
of-2020-1st-edition-fang-cai-51655250

Population Change In Southeast Asia Wilfredo F Arce Editor Gabriel C

Alvarez Editor

https://ebookbell.com/product/population-change-in-southeast-asia-
wilfredo-f-arce-editor-gabriel-c-alvarez-editor-51783518

Population Change And Economic Development In East Asia Challenges Met

Opportunities Seized Andrew Mason Editor

https://ebookbell.com/product/population-change-and-economic-
development-in-east-asia-challenges-met-opportunities-seized-andrew-
mason-editor-51943420

Population Change And Public Policy Billystrom Jivetti Md Nazrul Hoque

https://ebookbell.com/product/population-change-and-public-policy-
billystrom-jivetti-md-nazrul-hoque-21965184
Population Change And Rural Society 1st Edition David L Brown

https://ebookbell.com/product/population-change-and-rural-society-1st-
edition-david-l-brown-4388342

Population Change In The United States Socioeconomic Challenges And

Opportunities In The Twentyfirst Century 1st Edition Steve H Murdock

https://ebookbell.com/product/population-change-in-the-united-states-
socioeconomic-challenges-and-opportunities-in-the-twentyfirst-
century-1st-edition-steve-h-murdock-5238220

Population Change In The United Kingdom Tony Champion Jane Falkingham

Eds

https://ebookbell.com/product/population-change-in-the-united-kingdom-
tony-champion-jane-falkingham-eds-46209424

How Population Change Will Transform Our World 1st Edition Harper

https://ebookbell.com/product/how-population-change-will-transform-
our-world-1st-edition-harper-5845310

New Landscapes Of Population Change A Demographic World Tour Adele M

Hayutin

https://ebookbell.com/product/new-landscapes-of-population-change-a-
demographic-world-tour-adele-m-hayutin-46562886
POPULATION CHANGE, LABOR MARKETS,
AND SUSTAINABLE GROWTH
TOWARDS A NEW ECONOMIC PARADIGM
 CONTRIBUTIONS
TO
ECONOMIC ANALYSIS

281

Honorary Editors:
D. W. JORGENSON
J. TINBERGEN†

Editors:
B. BALTAGI
E. SADKA
D. WILDASIN

Amsterdam – Boston – Heidelberg – London – New York – Oxford

Paris – San Diego – San Francisco – Singapore – Sydney – Tokyo
 POPULATION CHANGE,
LABOR MARKETS,
AND SUSTAINABLE GROWTH

TOWARDS A NEW ECONOMIC PARADIGM

Edited by
Andrew Mason
University of Hawaii – Manoa and the East-West Center, USA
Mitoshi Yamaguchi
Kobe University, Japan

2007
Amsterdam – Boston – Heidelberg – London – New York – Oxford
Paris – San Diego – San Francisco – Singapore – Sydney – Tokyo
Elsevier
Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, UK

First edition 2007

Copyright Ó 2007 Elsevier B.V. All rights reserved

No part of this publication may be reproduced, stored in a retrieval system

or transmitted in any form or by any means electronic, mechanical, photocopying,
recording or otherwise without the prior written permission of the publisher

Permissions may be sought directly from Elsevier’s Science & Technology Rights
Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333;
email: [email protected]. Alternatively you can submit your request online by
visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting
Obtaining permission to use Elsevier material

Notice
No responsibility is assumed by the publisher for any injury and/or damage to persons
or property as a matter of products liability, negligence or otherwise, or from any use
or operation of any methods, products, instructions or ideas contained in the material
herein. Because of rapid advances in the medical sciences, in particular, independent
verification of diagnoses and drug dosages should be made

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library

ISBN-13: 978-0-444-53051-6
ISBN-10: 0-444-53051-7

For information on all Elsevier publications

visit our website at books.elsevier.com

Printed and bound in The Netherlands

07 08 09 10 11 10 9 8 7 6 5 4 3 2 1
CONTENTS

List of Contributors vii

Chapter 1 Introduction 1
Andrew Mason and Mitoshi Yamaguchi
Chapter 2 Evolution of Recent Economic-Demographic
Modeling: A Synthesis 5
Allen C. Kelley and Robert M. Schmidt
Chapter 3 A Century of Demographic Change and Economic
Growth: The Asian Experience in Regional and
Temporal Perspective 39
Allen C. Kelley and Robert M. Schmidt
Chapter 4 Demographic Dividends: The Past, the Present,
and the Future 75
Andrew Mason
Chapter 5 Demographic Change and Regional Economic Growth:
A Comparative Analysis of Japan and China 99
Tomoko Kinugasa, Wei Huang, and Mitoshi Yamaguchi
Chapter 6 Job Opportunities for Older Workers: When Are
Jobs Filled with External Hires? 133
Robert Hutchens
Chapter 7 Skills, Wages, and the Employment of Older Workers 161
Naoki Mitani
Chapter 8 Work-Life Balance Measures and Gender Division
of Labor 189
Akira Kawaguchi
Chapter 9 Optimal Education Policies Under an Equity–Efficiency
Trade-Off 211
Takashi Oshio
Chapter 10 Why Are Japanese Refusing to Pay the National
Pension Tax?: A Simultaneous Equation Analysis 243
Mitoshi Yamaguchi and Noriko Aoki
Author Index 265
Subject Index 269

v
This page intentionally left blank
LIST OF CONTRIBUTORS

Andrew Mason is Professor of Economics, University of Hawaii at

Manoa, and Senior Fellow, Population and Health Studies, East-West
Center, Honolulu.
Mitoshi Yamaguchi is Professor of Economics, Graduate School
of Economics, Kobe University.
Noriko Aoki is a graduate student, Graduate School of Economics,
Kobe University.
Robert Hutchens is Professor of Economics, Department of Labor
Economics, School of Industrial and Labor Relations, Cornell University.
Akira Kawaguchi is Professor of Economics, Faculty of Policy Studies,
Doshisha University.
Allen C. Kelley is Professor of Economics, Duke University.
Tomoko Kinugasa is Associate Professor of Economics, Graduate School
of Economics, Kobe University.
Naoki Mitani is Professor of Economics, Graduate School of Economics,
Kobe University.
Takashi Oshio is Professor of Economics, Graduate School of Economics,
Kobe University.
Robert M. Schmidt is Professor of Economics, University of Richmond.
Wei Huang is a Ph.D. candidate, Graduate School of Economics,
Kobe University.

vii
This page intentionally left blank
 CHAPTER 1

Introduction
Andrew Mason and Mitoshi Yamaguchi

Population growth is in decline and populations are becoming older

throughout the world. Low fertility is the common factor behind both of
these trends. In many Asian and European countries, women are averaging
fewer than two births over their reproductive span and, often, closer to
one. With fertility so low, population decline is inevitable. Ever smaller
cohorts of births are leading to ever smaller cohorts of children and young
adults, and eventually to populations in which the old greatly outnumber
the young. Reinforcing the effects of low fertility are continuing improve-
ments in life expectancy. In the past, mortality gains were achieved by
reducing death rates among infants and children. This is still true in the
world’s poorest countries, but elsewhere death rates among the young have
reached very low levels. Further gains are being achieved by reducing death
rates among the old, reinforcing the effects of fertility on population age
structure. As a consequence, the most rapidly growing demographic groups
in many countries are those in their 70s, 80s, and older.
The contributions to this volume are concerned primarily with the
economic implications of and the policy responses to these demographic
changes. The original papers were prepared as part of a Center of Excel-
lence (COE) research initiative, funded by the government of Japan and
carried out by the Graduate School of Economics Research Institute for
Economics and Business Administration at Kobe University. Leading
scholars from Japan and the United States were invited to prepare papers
on this common theme and to present them at the 4th International
Conference of the Japan Economic Policy Association, co-sponsored by
the Kobe University Center of Excellence and held at the Awaji Yumebutai
International Conference Center in Awaji, Hyogo, Japan, during 17–18
December, 2005. The papers were refereed and revised in accordance
with the reviewers’ suggestions prior to being copyedited.
The issues addressed here are of interest everywhere, but they are
particularly salient in Japan. Japan has the highest life expectancy and
POPULATION CHANGE, LABOR MARKETS, AND SUSTAINABLE GROWTH Ó 2007 Elsevier Ltd.
VOLUME 281 ISSN 0573-8555/DOI 10.1016/S0573-8555(07)81001-7 All rights reserved.

1
2 Andrew Mason and Mitoshi Yamaguchi
the oldest population of any country in the world. Its fertility rate has
dropped steadily for decades and is currently just below 1.3 births per
woman. Its population has peaked and is beginning to decline for the first
time in modern history. Its working-age population has been in decline for
a number of years. Like many other countries, Japan is struggling to
understand the implications of these demographic changes for its eco-
nomic strength. Generational issues are squarely on the front burner in
important policy debates. How can couples be encouraged to marry and
have children? Can educational reform lead to a more productive young
labor force, helping to offset their meager numbers? Can labor market
reform extend the working life of older workers without damaging the
interests of others? How should the social safety net that supports the
elderly evolve in ways that are both fair and sustainable? Of course, these
very same issues are being debated in Europe, North America, and
increasingly elsewhere around the world. The studies presented here rely
on both Japanese and international experience to address these issues.
The first four chapters provide a macroeconomic perspective that
frames the policy issues. In Chapter 2, Kelley and Schmidt provide a
conceptual framework for modeling the economic effects of demographic
change that builds on their own work and important studies by Barro;
Bloom and Williamson; Bloom, Canning, and Mulaney; and others. Using
a cross-country panel spanning the 1960–95 period, they show that demo-
graphic change in recent decades has played a pro-growth role accounting
for about 20% of per capita output growth, on average. In Asia and Europe
the positive contribution has been even greater.
In Chapter 3, Kelley and Schmidt build on the growth model presented
in Chapter 2 and focus their attention more on the future than on the past.
They provide important details about regional variation in the growth
effects of demographic change, concluding that the effects will be favorable
in Africa, Latin America, and South-Central Asia, but unfavorable in East
Asia and the West. Of particular importance are the effects of age structure
on saving and investment. They present evidence that declining youth
dependency has had strong positive effects on saving and investment, but
these favorable effects will be overwhelmed by the negative effects of rising
old-age dependency in the near future. If Kelley and Schmidt prove to be
correct, aging may lead to slower growth or even economic decline.
Mason’s analysis presented in Chapter 4 builds on earlier work with
Lee and recent empirical analysis with Kinugasa, providing an interesting
contrast to Kelley and Schmidt’s contribution. Mason relies on a simula-
tion approach and long-term historical data to explore how demographic
change influenced economic growth in India, Japan, and the United States.
His results are consistent with Kelley and Schmidt’s conclusion that
 Introduction 3
demographic change had very favorable effects on per capita income
growth during the second half of the twentieth century. With respect to
capital accumulation and the future, his findings differ. He points to
increases in life expectancy as the primary force that led to the rapid
accumulation of wealth. Moreover, his analysis implies that, although
saving rates may decline as populations age, wealth will increase relative
to GDP. In the United States, demographic factors will continue to be
moderately favorable. In Japan, however, labor productivity is likely to
rise but per capita income will drop because the share of the population in
the labor force will decline so rapidly.
Most research on population and development emphasizes the national
level, but many of the most important issues related to aging are local.
Many public services on which the elderly depend are provided by local
governments. The ability of the private sector to provide services for the
elderly depends on the resources that must be locally available. Moreover,
the age distributions of populations vary enormously by region. The young
flock to the cities, while the elderly remain in the countryside. In Chapter 5,
Kinugasa, Huang, and Yamaguchi fill this void by analyzing the influ-
ences of demographic factors on economic growth at the prefectural level
in Japan and the provincial level in China. Their most important finding is
that population size has a strong positive effect on economic growth at the
regional level. If so, population decline will have particularly severe
effects on economic growth in regions where many of the elderly are
living. This could greatly complicate efforts to formulate effective policy
responses to population aging.
Retirement behavior and policy will inevitably play a vital role in
determining the economic effects of population aging. The available evi-
dence suggests that the elderly are not only living longer, but are healthier
as well. By remaining in the labor force to a later age, older adults can
reduce the financial burden they impose on younger generations—their
adult children in some societies and taxpayers in other settings.
Chapter 6 by Hutchens and Chapter 7 by Mitani address two aspects of
the employment of older workers. Hutchens considers a difficult aspect of
the labor market for older workers: the transition into new jobs that are more
suited to their (often declining) skills. Many workers are forced to retire
from firms and seek new jobs. Many employers, however, are reluctant to
fill jobs with ‘‘outside’’ rather than ‘‘inside’’ workers—that is, workers who
are already employed by the firm. Using a survey of establishments with
information on white-collar jobs, currently held by older workers, Hutchens
provides new information about which types of jobs are likely to be filled by
outside workers and which worker characteristics influence the decision to
hire rather than to fill from within the firm.
4 Andrew Mason and Mitoshi Yamaguchi
Mitani explores two important issues that affect policy toward older
workers. The first is whether the increase in wages received by workers as
they age is matched by an increase in their skills. If not, employers will be
reluctant to retain older workers. Mitani finds that in some occupations the
match is close, but that in others it is not. The second issue, he investi-
gates, is whether the firms that retain older workers reduce their hiring of
younger workers. He finds that they do. If extending the working life of
older workers leads to the displacement of younger workers, more flexible
retirement policies are a less attractive response to population aging.
An entirely different approach to population aging is to increase the number
or the ‘‘quality’’ of young workers. As fertility rates drop ever lower, many
governments are adopting policies to encourage childbearing. In Chapter 8
Kawaguchi examines one set of such policies implemented in Japan. The
government is encouraging private firms to adopt work-life balance programs
that provide time off and other benefits to parents of newly born children. The
rates of uptake by men are very low, however, and Kawaguchi explores why
this is so and whether government subsidies should be provided to encourage
greater participation by fathers in these programs.
The total productivity of each new generation of workers depends on
both their numbers and their individual productivity. As the numbers shrink,
the importance of education and other forms of human capital investment
become all the more important. As Oshio points out in Chapter 9, however,
public spending on education typically benefits the most talented—raising
productivity but sacrificing equality. He shows how a mixed system of
public and private education can be used to solve this dilemma. The
emphasis of the public sector should be on offering public education to
low-ability individuals, while high-ability individuals are best served by
private education systems. To the extent that this has a regressive effect on
income inequality, other distributive policies can be employed to achieve
the desired distribution of income.
The final chapter in the volume, by Yamaguchi and Aoki, addresses a
remarkable development in Japan’s political economy, but one that may
be repeated elsewhere: taxpayer revolt against the public pension system.
As populations age, public pension systems will become increasingly
unattractive to taxpayers. They face higher tax rates and reductions in
pension benefits. Yamaguchi and Aoki analyze why pension compliance
rates vary and conclude that low rates of compliance are found in areas
with more university students, a higher birth rate, lower death rates, higher
unemployment, and higher rates of investment in financial securities.
These results identify some of the possible explanations for low compli-
ance and also provide clear evidence that workers are substituting personal
saving for participation in the public pension program.
 CHAPTER 2

Evolution of Recent Economic-Demographic

Modeling: A Synthesis
Allen C. Kelley and Robert M. Schmidt

Few issues in social history have attracted more attention than assessments
of the economic consequences of rapid population growth. Debates have
been vigorous and contentious. The primary evidence that has both sti-
mulated and sustained these debates is cross-country regressions and
simple correlations that expose the impacts of demographic variables on
per capita output growth. A surprising result emerges: the overall impacts
of population are generally found to be small, especially for the 1960s and
the 1970s, although some negative impacts appear to emerge for the 1980s
and possibly beyond (Kelley and Schmidt 1994). There are a number of
reasons why the cross-country studies are inconclusive, not the least being
their somewhat simple rendering of demographic processes.
Empirical analysis was enriched in the 1990s with the emergence of a
theoretical framework by Robert J. Barro that incorporates population into
convergence (or technology-gap) models. He and his collaborators con-
cluded that high fertility, population growth, and mortality all exert negative
impacts on per capita output growth (Barro 1991, 1997; Barro and Lee 1994).
We (Kelley and Schmidt 1994) extended this list to include population
density and size, which revealed positive impacts, although a net negative
assessment of combined demographic trends represented the bottom line.
Convergence modeling of demographic trends evolved further in the late
1990s through a series of papers by several Harvard economists (e.g.,
Bloom and Williamson 1997, 1998; Bloom and Canning 2001, 2003;
Bloom, Canning, and Malaney 2000; Radelet, Sachs, and Lee 2001).
Building on the Barro setup (albeit with a different choice of core variables),

This chapter has been modified slightly from an article that appeared in 2005 in the Journal
of Population Economics 18(2), 275–300. It has been included with the kind permission of
Springer Science and Business Media.

POPULATION CHANGE, LABOR MARKETS, AND SUSTAINABLE GROWTH Ó 2007 Elsevier Ltd.
VOLUME 281 ISSN 0573-8555/DOI 10.1016/S0573-8555(07)81002-9 All rights reserved.

5
6 Allen C. Kelley and Robert M. Schmidt
the Harvard framework focused on population impacts that take place as a
result of imbalanced age-structure changes over the demographic transition.
Their modeling compactly captured these impacts by just two variables:
population growth (Ngr) and working-age growth (WAgr). Such a specifica-
tion neatly ‘‘translates’’ a traditional neoclassical model formulated in per-
worker output growth into a comparable model formulated in per capita
output growth. While such a translation derives from an identity, it is
nevertheless a useful framework that provides a way of exposing some
shorter-period ‘‘population impacts’’ within the usual long-run neoclassical
framework. Such ‘‘translation’’ impacts of population in numerous Harvard
empirical studies are assessed to be sizable, especially in East Asia.
Important to appraising these empirical findings is an understanding of
the nature of the highlighted impacts of demographic translations.
Although Ngr and WAgr are primarily introduced into these models to
translate per-worker growth rates into per capita growth rates, they are
often interpreted to play a role in the determination of the per-worker
growth rate as well. Here, we build on the Harvard tradition by emphasiz-
ing the point that any per capita output growth rate can be separated into
two components: an economic production (productivity) component and a
translations component. We argue that these components are potentially
separable and that clearer insights into the multifaceted role of population
in economic growth can be gained by modeling them separately. In
particular, such a rendering allows for the possibility that Ngr and WAgr
play little or no role in explaining productivity growth once the produc-
tivity component is modeled to include additional demographic variables.
In this chapter we take a hard look at these potentially separable compo-
nents in a framework that provides reasonably clear, consistent, and
interpretable empirical results with interesting policy implications. Our
approach makes several contributions.
First it enables us to formulate an economic growth paradigm that
highlights the separability of production and translations components.
We identify quite distinct roles that population may play within each
component. We believe that this paradigm clarifies several specific ways
in which economic and demographic change interconnect. We further
believe that the modeling perspective developed in this chapter need not
be restricted to the particular empirical renderings chosen for this study.
Indeed, our paradigm provides a potential platform for introducing alter-
native treatments of population into the modeling of economic growth.
Second, we evaluate this framework empirically using the Barro core.
The results show that translations demography has little or no impact on
economic production per se, although the analysis of translations clarifies
some welfare implications of demographic change. On net, demographic
 Evolution of Recent Economic-Demographic Modeling 7
change elicits positive translations impacts over the full period in all
regions. Such impacts, a by-product of the demographic transition in
many countries, are not uniformly positive over time. Our results indicate
that the positive impacts relatively early in the demographic transition
tend to turn negative during a later phase.
Third, we show that population does matter within the productivity
component, but primarily through linkages such as youth age structure, a
variable highlighted in the literature of the 1950s–80s. These demographic
impacts on economic production, while notable, are not remarkable. World-
wide, demographic changes account for around 8% of the influences effect-
ing change in output per laborer growth, an impact with substantial regional
variance spanning 3% in Africa to 28% in Asia. Impacts of demographic
translations are also sizable (13%), but with smaller variance spanning 11%
in South America to 16% in Asia. Overall, the results place population’s
role as neither alarming nor benign. What has changed with the evolution of
modeling in the 1990s is a clearer interpretation of the channels and sizes of
demographic changes on the economy.
The remainder of the chapter has five sections. The first provides an
organizing framework that highlights the separability of impacts of demo-
graphic change on per-worker output growth from those of per capita
output growth. The second assesses alternative ways of incorporating
demographic variables into the convergence model. The third defends
our preferred empirical paradigm that meshes the Barro core, the Harvard
translations structure, and our own demographic enrichments. The fourth
examines the importance and alternative roles of population in this frame-
work, compares these results with other models in the literature, and
arrives at empirical assessments that qualify the recent literature. The
final section summarizes our conclusions.

2.1. Productivity and translations as an organizing framework

We present in equation (2.1) an initial taxonomy for organizing an assess-

ment of demographic change:
Y=Lgr ðLgr  Ngr Þ
ðY=NÞgr  |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} þ |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} ð2:1Þ
Productivity Component Translations Component
It decomposes per capita output growth [(Y/N)gr] into two components:
(1) a labor productivity component [(Y/L)gr] and (2) a translations component
[Lgr  Ngr] that converts output growth per labor hour into output growth
per person. Equation (2.1) derives from the identity Y/N  (Y/L)(L /N)
(see Bloom and Williamson 1998).
8 Allen C. Kelley and Robert M. Schmidt
Equation (2.1) is deceptively simple—it is, after all, an identity. We
would argue, however, that recasting the Harvard model into separable
productivity and translations components not only serves to clarify incon-
sistent demographic results found in previous empirical studies, but also
provides a framework upon which further advances can be based. In this
regard, note that this framework remains agnostic with respect to the
importance of both the two components to per capita GDP growth rates
and the models chosen for either component. Our study builds upon,
challenges, and enriches the modeling of the Harvard studies. As such, we
consider it a bridge between past and future demographic modeling. We
begin by considering several candidates for the translations component.
The translations component can take many forms. The one employed in
most neoclassical theoretical modeling,
 
Translation I Lgr  Ngr ¼ 0 ð2:2Þ
results in a focus on labor productivity due to the simplifying assumption
that Lgr = Ngr. Population growth has no impact on per capita economic
growth per se unless it has a direct impact on (Y/L)gr. Such a framework is
most relevant in the longer run (e.g., at or near a demographic steady state)
or during conditions of slowly evolving demographic change, and less
relevant during conditions characterized by sizable variations in Lgr and
Ngr due to major changes in mortality and fertility.
A second translation takes full advantage of the components of labor
change:
 
Translation II Lgr  Ngr ¼ ðL=LFÞgr þ ðLF=WAÞgr
ð2:3Þ
þ WAgr  Ngr
where L = total labor hours, LF = laborers available for work, and WA =
working-age population. This expression reveals, in order, the impacts of
changes in labor utilization rates (L/LF), labor force participation rates out
of the working-age population (LF/WA), working ages (WA), and popula-
tion (N). The last two terms represent population, driven by changes in
fertility, mortality, and migration. The other terms evolve from labor
market conditions and household choices. While beyond the scope of the
chapter, this translation illuminates several potentially fruitful areas for
future modeling, possibly within an endogenous framework. Consider the
last three terms. While WAgr is largely predetermined within our 10-year
growth periods, changes in labor force participation out of the working
ages [(LF/WA)gr] as well as fertility and mortality changes (Ngr) need not
be. An endogenous treatment of the interplay between fertility and female
labor force participation could be intriguing in this context. Such modeling
 Evolution of Recent Economic-Demographic Modeling 9
might incorporate macroeconomic renderings of the labor–leisure tradeoff
as well as changes in public family planning programs and family struc-
ture during development.
A third translation, advanced by the Harvard scholars, focuses on the
working-age population (WA, say ages 15–64):
 
Translation III Lgr  Ngr ¼ WAgr  Ngr ð2:4Þ
This formulation abstracts from labor force participation and employment
rates. Because changes in the age distribution can be large and account for
much of the variation in L, WA is a potentially useful proxy for labor force
effects that occur during the demographic transition. We adopt this trans-
lation for the empirical work in this chapter for two reasons. First, this
translation provides a productive comparison between empirical results
from the Harvard model and the extended model we propose. Second, and
more important, within the 10-year growth periods employed in this
chapter, WAgr is exogenous to the model whereas LFgr is not. An endo-
genous modeling of labor force growth—and in particular one highlight-
ing female labor force participation—is well beyond the scope of the
chapter, especially given the problematic quality of labor force estimates.
Common to each of these specifications is the overriding lesson that the
translations variables might have no impact on (Y/L)gr. If that were the
case, then their role would be mechanical in the sense that no econometric
estimation would be necessary to estimate their net impact on (Y/N)gr; the
net impact would be the unweighted difference between Lgr (however
approximated) and Ngr. This does not imply that their quantitative impact
need be small. Indeed, their difference can be quite large, positive or
negative, at various points during the demographic transition. We would
argue, however, that fuller modeling should include additional direct roles
for demographic change on labor productivity growth, the variable high-
lighted in neoclassical growth theory. We expand on this distinction
between direct ‘‘productivity’’ impacts and ‘‘translations’’ impacts in the
following two sections. Additionally, the fourth section provides empirical
tests for productivity impacts of the translations variables while the fifth
section presents estimates of the productivity and translations components
on changes in (Y/N)gr.

2.2. Expanding the theory: Elaborating the roles of population

Prior to elaborating the roles of population, we provide background on the

convergence framework. We then assess possible roles of population in
the productivity component of equation (2.1). Finally, we discuss another
10 Allen C. Kelley and Robert M. Schmidt
possible translations role that is not included in equation (2.1), the transla-
tion of the convergence term itself from per-worker into per capita terms.

2.2.1. Expanding theory: The convergence framework for modeling

productivity growth
The economic growth literature provides numerous ways to model pro-
ductivity growth. We focus here on the ‘‘convergence’’ or ‘‘technology-
gap’’ framework. Rooted in neoclassical growth theory, this paradigm
explores the relationships between economic growth and the level of
economic development. It focuses on the pace at which countries move
from their current economic level to their long-run, or potential, or steady-
state equilibrium level of output. (This section benefits from the presenta-
tions of Barro 1997 and Radelet, Sachs, and Lee 2001.)
The model begins by positing a convergence assumption:
Y=Lgrit ¼ c ½lnðY=Lit Þ  lnðY=Lit Þ ð2:5Þ
Here the rate of output growth per worker [(Y/L)gr] is proportional to the gap
between the logs of the long-run, steady-state (Y/L)* and the current (Y/L)
levels. The greater this gap, the greater is the gap separating physical
capital, human capital, technical efficiency, or some combination of these
three from their potential levels. Large gaps allow for ‘‘catching up’’
through (physical and human) capital accumulation and technology crea-
tion and diffusion across, and within, countries.
The rate of convergence, c, is assumed to be independent of time and
place. By contrast, potential output per worker (Y/Lit)* is specific to
country (i) and time (t). This ‘‘conditional’’ convergence allows for the
observed positive correlation between the level of development and eco-
nomic growth rates. Were Y/L* the same for all countries, the simple
correlation would be negative. Were Y/L* the same for all time periods,
the world’s economy would eventually stop growing. Potential productiv-
ity is, of course, unobservable and must be modeled. Its log is modeled as
a linear function of a vector of country- and time-specific characteristics:
lnðY=Lit Þ ¼ a þ b Zit ð2:6Þ
The actual specification of the determinants of long-run labor productivity
(i.e., the selection of Z’s) varies notably, but the basic model, which
combines equations (2.5) and (2.6), is the same across scores of empirical
studies:
Y=Lgrit ¼ a 0 þ b0 Zit  c lnðY=Lit Þ ð2:7Þ
where a0 = ac and b0 = bc.
 Evolution of Recent Economic-Demographic Modeling 11
What types of Z variables should be included as determinants of long-
run output per worker? Recognizing that a long-run, steady-state produc-
tion function lies behind Y/Lit*, one should consider factors that influence
long-run physical and human capital stocks, technology, and natural
resource stocks. Barro (1997, sect. I) additionally notes that endogenous
growth theories that include the discovery and diffusion of new technol-
ogies suggest that Y/Lit* depends upon ‘‘governmental actions such as
taxation, maintenance of law and order, provision of infrastructure ser-
vices, protection of intellectual property rights, and regulation of interna-
tional trade, financial markets, and other aspects of the economy.’’
Additionally, various authors have suggested climate, access to ports,
education, health, and many other factors as possible influences.
More subtly, equation (2.5) provides additional insight. The invariance
of the convergence parameter (c) across countries is consistent with a
neoclassical view of efficient international capital markets. Simply stated,
a neoclassical perspective models (physical and human) investment to
flow fluidly within and across countries toward highest returns (e.g., to
regions and countries with large gaps between potential and current Y/L).
However, not all countries finance investment with equal ease. Thus, a
second category of variables is added to the Z vector to ‘‘condition’’ the
convergence rate, c. These variables include country- and time-specific
factors that enhance or deter international capital flows, domestic saving,
domestic investment, or migration. Included among these are, for exam-
ple, restrictive licensing, the risk of expropriation, political conditions, the
rule of law, and migration regulations.
Conspicuous by its absence from this list is the investment share in
GDP. At first blush, investment might be the first variable one would think
to include in a model of (Y/L)gr. Indeed, Levine and Renelt (1992) surveyed
numerous empirical growth studies to identify a common set of influential
variables. They found investment rates to constitute the most robust
variable. Rather than implying that the investment rate is a viable Z
variable for predicting long-run capital-to-output ratios, however, its sig-
nificance in a convergence model suggests an incomplete set of Z vari-
ables. If the convergence hypothesis is correct, the list of Z variables is
complete, and factors enhancing or deterring the free flow of investment
have been modeled, then the investment coefficient would be rendered
largely moot. For a similar perspective, see Bloom, Canning, and Malaney
(2000); Higgins and Williamson (1997); and Kelley and Schmidt (1994).
Finally, we consider another subtlety of the model. Long-run, steady-
state productivity is specified as being time-specific. Indeed, Radelet,
Sachs, and Lee (2001) develop the convergence model from an instanta-
neous growth perspective highlighting the idea that (Y/L)*it changes for a
12 Allen C. Kelley and Robert M. Schmidt
country from one time to another. (Y/L)*it might progress or regress as, for
example, government tax policy changes. Since these models are esti-
mated over a period of years, Z variables are typically calculated as period
averages. (In some cases, beginning-of-period values are used as instru-
ments for a variable, such as population size, which may be influenced by
economic growth over a longer period.) For greater depth on the technical
details of convergence modeling, see Radelet, Sachs, and Lee (2001).

2.2.2. Expanding theory: Modeling the productivity component

to account for population
Are there roles for population among the Z variables in the convergence
model; or, put differently, how does population influence the labor pro-
ductivity component of definitional equation (2.1)? Consider first the
translations variables themselves. Do Ngr and WAgr influence productivity
growth in addition to their role of translating productivity growth into per
capita GDP growth? The potential for productivity effects of these vari-
ables has been advanced along several lines.

2.2.2.1. Measuring demographic impacts indirectly: Ngr , WAgr

Within the neoclassical model, the steady-state levels of capital and output
per worker depend upon the propensity to invest (or save in a closed
economy), the growth rate of labor (or population if it is assumed to grow
at the same rate), and the state of technology. In our context, this implies
that WAgr can play a role in the Z vector, holding the propensity to invest
and technology constant. An economy can experience capital ‘‘deepening’’
when investment outpaces WAgr or ‘‘shallowing’’ when WAgr exceeds
capital expansion. Can the economy attract sufficient investment to avoid
capital shallowing in the face of a growing workforce? Some would argue
that the answer to this question is driven largely by an institutional structure
that facilitates entrepreneurial activity and the free flow of financial capital.
For this reason, Barro’s core of Z variables (discussed below) includes
measures that represent his best efforts at proxying such impacts (govern-
ment consumption, rate of inflation, a rule-of-law index, and a democracy
index).
In a different vein, some analysts have justified the inclusion of Ngr and
WAgr to capture the impacts of ‘‘dependency’’ (the proportion of the
population or work force in the youth or aged cohorts). Specifically, the
impact of life-cycle consumption patterns on macroeconomic saving and
investment levels can be large over the demographic transition due to
swings in a country’s age structure. Since the relative rates of Ngr and
WAgr influence dependency levels, which in turn influence savings and
 Evolution of Recent Economic-Demographic Modeling 13
investment, which in turn influence the rate of an economy’s productivity
growth, Ngr and WAgr can indeed exert an effect on labor productivity.
While that is true, an issue arises whether these two variables best capture
such dependency effects. We argue probably not. The savings literature
posits that it is the current level of youth and aged dependency that
influences savings and investment rather than their rates of change (see,
for example, Kelley and Schmidt 1996 and Higgins and Williamson
1997). It would thus seem that while dependency effects are plausibly
important, measuring their impacts directly, versus capturing them indir-
ectly as correlates of the translations variables, represents a preferred
empirical methodology, allowing, of course, for estimation of direct
effects of the translations variables on labor productivity (if they exist)
as well. This more direct approach to isolating the various impacts of
population is the tack we present below.
Some authors (e.g., Bloom and Canning 2001) have justified the
inclusion of WAgr for another reason: rapid labor force growth may result
in a deterioration of labor force quality as workers with lower than average
skills and experience are hired. Again, while that is true, an issue arises
whether WAgr is the best variable for capturing these impacts. For exam-
ple, the Barro core already includes life expectancy and post-primary
educational attainment of males aged 25 and over as human-capital
measures. Of course, it is arguable whether life expectancy primarily
represents human capital. This issue is taken up in the next section,
where we specify Zs, data, and estimation.
In short, various authors have advanced arguments for the inclusion of
WAgr among the Z variables. In at least two cases, however, we would
argue that other variables capture the posited effect more directly. We
prefer to model those productivity effects explicitly and separately from
the translations variables. Nevertheless, we agree with Bloom and
Canning (2001) that the possibility that Ngr, WAgr, or both exert an
impact (Y/L)gr is a hypothesis that can and should be tested directly. A
pure translations role predicts coefficients of þ1 and 1 for WAgr and Ngr,
respectively. Deviations from those values imply either an incomplete set
of Z variables or productivity impacts. For example, each of the above
arguments posits a negative productivity impact for WAgr, implying a
coefficient WAgr lower than þ1. This is a testable hypothesis.
On the other hand, if WAgr does not differ significantly from þ1 and
Ngr does not differ significantly from 1, we would prefer to constrain
them to þ1 and 1, respectively, since this approach introduces clarity
into the modeling of population that has been absent from this literature.
Indeed, a major contribution of the Harvard translations framework is
to expose the ‘‘catch-all’’ nature of the Ngr variable included as the sole
14 Allen C. Kelley and Robert M. Schmidt
demographic measure in many studies. Introducing WAgr (with Ngr) as a
way of translating labor into per capita output growth provides a clearer
meaning for this variable. However, to the extent that Ngr and WAgr proxy
additional demographic influences, the purity and usefulness of the frame-
work are diminished. Such an outcome is minimized when those addi-
tional measures are explicitly included in the analysis. We next turn to
some of these variables.

2.2.2.2. Measuring demographic impacts directly: D1 , D2 , N , Density

What demographic variables might have a direct (Z-variable) bearing on
labor productivity growth? Of the various possibilities, two seem particu-
larly promising.
The first has already been mentioned. To the extent that a population’s
age structure (proxied here by dependency ratios: D1 = ratio of population
aged 0–14 to that of working ages 15–64 and D2 = ratio of population
aged 65þ to that of working age) influences the rate of domestic saving
(e.g., through life-cycle influences), then D1 and D2 can have both short-
and long-run influences on productivity growth. The short-run influence is
through facilitating or inhibiting the savings and investment necessary to
close the gap of equation (2.5). The long-run influence is on Y/N*. A basic
lesson of the standard neoclassical growth model is that in long-run
equilibrium, the savings level will affect the level of Y/N* but not its
rate of change. Modeling unobservable Y/N* from currently observed
variables is one challenge of convergence modeling, and dependency
rates assist in this modeling.
The second possibility is that both the scale of production (population
or labor force size) and density (population or labor force per unit of land)
can exert an impact on long-run growth. It is in agriculture where the
impacts of scale and density are most discussed. On the positive side,
higher densities can decrease per unit costs and increase the efficiency of
transportation, irrigation, extension services, markets, and communica-
tions. On the negative side, higher density may be associated with dimin-
ishing returns to land or deleterious effects of congestion. The predicted
impact of rising population density on growth is ambiguous. We might
note that the distribution of the population between urban and rural areas
does not influence this density measure. Thus, the possibility of reverse
causation, whereby rapid economic growth encourages urbanization, does
not affect our density measure. With respect to population size, scale
effects have been highlighted in earlier development studies, particularly
with reference to specialization and diversification between firms. Recent
endogenous growth models of technical change posit positive scale effects
where an R&D industry produces a nonrival stock of knowledge. Holding
 Evolution of Recent Economic-Demographic Modeling 15
density constant, we predict a positive impact of population size. Overall,
evidence on scale and density effects is mixed and sparse.

2.2.2.3. Modeling a translations component for the convergence term

Equation (2.1) includes a component (Lgr – Ngr) that translates output
growth from per-worker to per capita terms. Equation (2.5) indicates
that, the higher is the initial level of labor productivity, the slower is
economic growth. Although theory dictates the specification of this ‘‘con-
vergence’’ term in output-per-worker terms [ln(Y/Lit)], nearly all empirical
models prefer to specify it in per capita terms [ln(Y/Nit)]. To do so
properly, however, the model should include another variable to translate
the convergence term to per capita terms. When completed, equation (2.5)
would appear as equation (2.8).
Y=Lgrit ¼ c lnðY=Lit Þ  c lnðY=Nit Þ þ c lnðL=Nit Þ ð2:8Þ
Thus, another translations term, ln(L/Nit), must be included when the
convergence term is specified in per capita units. Surprisingly few empiri-
cal studies include such a term. To our knowledge, the first study to do so
was Bloom, Canning, and Malaney (2000).
Having said that, however, we prefer not to employ the per capita
variant of equation (2.8) for our primary rendering, opting for the per-
laborer variant of equation (2.5) instead. The two variants are equivalent
unless ln(L/Nit) plays a role beyond translations. While we can think of no
compelling rationale for including ln(L/Nit) in the productivity component
of the model, we can for two of its correlates. Specifically, we argued
previously that youth and aged dependency rates are plausible Z variables
to the extent that they influence the domestic savings rate. Were equation
(2.8) estimated without the dependency rates, the estimated coefficient for
ln(L/Nit) would include both translations and dependency impacts. Were
we to include the dependency rates as well, we would introduce severe
multicollinearity unnecessarily. (The simple correlations coefficient
between WA/N, our variant of L/N, and D1 or D2 are 0.98 and 0.78,
respectively.) Equation (2.5)’s rendering coupled with dependency rates as
Z variables thus provides clean translations and labor productivity inter-
pretations of age-structure effects. We will return to this issue when we
examine the empirical results.

2.3. Toward an empirical rendering of population and growth

Our preferred theoretical model is presented as equation (2.9), derived by

substituting equation (2.7) into definitional equation (2.1). This formulation
16 Allen C. Kelley and Robert M. Schmidt
highlights two sets of Z variables—an ‘‘economic’’ core (Ze) and a ‘‘demo-
graphic’’ core (Zd)—that determine output per worker (Y/WA) in the long
run. The rationale for selecting specific Zd variables was considered in the
previous section, in which we elaborated the roles of population. Here we
present the rationale for selecting specific Ze variables, describe the data, set
out estimation procedures, and present the results of alternative models that
estimate the impacts of population on economic growth.
a0 þ b0 ðZe þ Zd Þit  c lnðY=Lit Þ Lgrit  Ngrit
ðY=NÞgrit ¼ |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} þ |fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflffl} þ d i þ e  t þ "it ;
Productivity Model Translations Model
ð2:9Þ
where i provides for regional fixed effects,  t represents a period fixed
effect allowing for exogenous shocks, and "it represents an error term
following the classical assumptions.

2.3.1. Empirical rendering: Specifying Ze variables, data, and

estimation
Since the impact of population on the economy can be influenced by the
choice of Ze variables, we are cautious in selecting the Ze variables.
Furthermore, an extensive literature already exists in this area, including
the pioneering work of Barro (1997), whose core economic variables we
elect to adopt without notable modification. While one can easily imagine
alternative variables, suffice it to say that Barro’s empirical inquiries have
been expansive.

2.3.1.1. The economic Ze variables

A brief justification of the economic variables follows; see Barro (1991,
1997) for more detail. Growth in output per capita is held to be positively
related to

1. A lower initial level of productivity [ln(Y/WA)]. This convergence is

posited to be more rapid in countries with higher levels of schooling
attainment [ln(Y/WA)Ed].
2. Higher educational attainment for males of ages 25 and over [Male-
Educ], which facilitates the absorption of new technologies.
3. Higher life expectancy [ln(e0)], a proxy for better health and human
capital in general (Barro 1997). Note that the mortality patterns under-
lying life-expectancy calculations can influence life-cycle saving; for
example, rising e0 can affect saving through the financing of earlier or
later retirement in conjunction with an increasing life span (see Bloom,
Canning, and Graham 2003 and Lee, Mason, and Miller 2001). Note
 Evolution of Recent Economic-Demographic Modeling 17
also that while life expectancy is a traditional demographic variable, it
can still largely represent ‘‘health’’ if the demographic correlates of e0
(e.g., Ngr versus WAgr, youth and elderly dependency ratios) are held
constant in the empirical modeling. We lean toward this ceteris paribus
interpretation, but are open to reclassifying the impacts of e0 between
‘‘economic’’ and ‘‘demographic’’ impacts in the section on the impor-
tance of population. Suffice it to say that much work remains in future
research in decomposing the interpretation of e0 between proxies for
health, saving, and other factors. Recent work examining the influences
of life expectancy on savings, productivity, and economic growth
includes Kinugasa (2004), Kinugasa and Mason (2005), and Chapters
3 and 4 of this volume.
4. Improvement in terms of trade [TT% chg], posited to generate added
employment and income.
5. A lower rate of inflation [Inflation], leading to a more stable investment
climate and to better decisions with predictable price expectations.
6. A lower government consumption share netted of education and defense
spending [Gcons/Y], posited to release resources for more productive
private investment.
7. Stronger democratic institutions [Democracy] at low levels of democ-
racy, which promote market activity by loosening autocratic controls.
However, stronger democracies at high levels can dampen growth
if the government exerts an increasingly active role in redistributing
income. Democracy is thus entered in quadratic form, posited to rise
and then fall.
8. A stronger rule of law [Rule Law], which stimulates investment by pro-
moting sanctity of contracts, security of property rights, etc.

2.3.1.2. The demographic Zd variables

As justified in the section above, in which we elaborated the roles of
population, two demographic variables are posited as having negative
impacts: ln(D1) and ln(D2), youth and aged dependency ratios, respec-
tively. One [Density measured as 1000 population per square kilometer]
has an uncertain impact, and the other, [Size measured as ln(N)], a positive
impact on output growth per worker.

2.3.1.3. The data

The data comprise 86 countries and four growth periods (1960–70, 1970–
80, 1980–90, and 1990–95), resulting in a panel with 344 observations.
Included countries have market economies, production structures that are
not dominated by raw material exports, a population of at least one million
18 Allen C. Kelley and Robert M. Schmidt
in 1970, and reasonably reliable data. Descriptive statistics, data sources,
and a country listing are presented in the appendix.

2.3.1.4. Measuring L
We follow the Harvard studies and proxy L with WA (working-age
population, ages 15–64). The rationale for this choice is provided above
in the discussion of equations (2.3) and (2.4).

2.3.1.5. Estimation Procedures

Following Barro (1991, 1997), we employ two-stage least-squares estima-
tion, running period-specific first-stage regressions for ln(Y/WA), ln(Y/
WA)Ed, Inflation, Gcons/Y, and Democracy. All rates of change are
calculated as continuous rates over the period. Density, ln(Y/WA), ln(Y/
WA)Ed, MaleEduc, ln(e0), and ln(N) are beginning-of-period values.
Gcons/Y, Democracy, Rule Law, ln(D1), and ln(D2) are period averages.
We include regional binaries for North and Central America, South
America, Europe, Asia, and Oceania (with Africa in the intercept) to
allow for possible cross-sectional fixed effects. A country listing by
region is included in the appendix. Finally, we include period binaries
for the 1970s, 1980s, and 1990s (i.e., period fixed effects with the
1960s in the intercept) to allow for exogenous shocks. Additional
technical detail is provided in the appendix.

2.3.2. Empirical results: Choosing among alternative specifications

Table 2.1 presents five alternative regression models that account for per
capita output growth. Each shares the Barro Ze economic core, and each
illustrates an alternative demographic setup designed to distinguish
between Zd variables (productivity impacts of population) and translations
variables.
Model 1, termed Simple Demography, substitutes Ngr for the TFR (total
fertility rate) in Barro’s core. Using this model, we can ask, How do the
impacts of population growth (Ngr) change with alternative modeling
embellishments, and why?
Models 2 and 3, termed Translations Demography, append a transla-
tions component in the manner of Harvard. Both models include ln(WA/N)
and ln(WA/N)Ed to translate the theoretical convergence terms, ln(Y/
WA) and ln(Y/WA)Ed, into the analogues typically used in estimation,
ln(Y/N) and ln(Y/N)Ed. Additionally, Model 2 includes Ngr and WAgr to
translate per-worker growth rates, (Y/WA)gr, into their more typical per
capita variant, (Y/N)gr. Thus, population is not explicitly modeled as
having long-run productivity impacts. However, any of these coefficients
 Evolution of Recent Economic-Demographic Modeling 19
Table 2.1. Empirical results for five demographic specifications
Dependent variable Simple Translations Enriched
Demography Demography Demography

(Y/N)gr (Y/N)gr (Y/WA)gr (Y/WA)gr (Y/WA)gr

(1) (2) (3) (4) (5)

Productivity model
Convergence
ln(Y/N) (1–3) 1.19*** 1.18*** 1.18*** 1.31*** 1.26***
ln(Y/WA) (4–5) (4.27) (4.30) (4.30) (4.63) (4.47)
ln(WA/N) 11.56*** 11.56***
(4.77) (4.77)
ln(Y/N)Ed (1–3) 0.04 0.00 0.00 0.08 .03
ln(Y/WA)Ed (4–5) (0.30) (0.02) (0.02) (0.51) (0.18)
ln WA/NEd 0.27 0.27
(0.14) (0.14)
Ze: economic core
TT %chg 0.16*** 0.15*** 0.15*** 0.14*** 0.15***
(5.13) (5.16) (5.16) (4.92) (5.20)
Gcons/Y 0.11** 0.06 0.06 0.07* 0.07*
(2.01) (1.17) (1.17) (1.35) (1.40)
Inflation 0.04*** 0.04*** 0.04*** 0.04*** 0.04***
(4.80) (4.51) (4.51) (4.63) (4.77)
ln (e0) 6.51*** 4.49*** 4.49*** 4.40*** 5.22***
(4.92) (3.56) (3.56) (3.14) (3.99)
MaleEduc 0.38 0.17 0.17 0.23 0.20
(1.50) (0.71) (0.71) (0.95) (0.83)
Rule Law 2.42*** 1.93** 1.94** 2.04** 2.03**
(2.50) (2.15) (2.15) (2.18) (2.13)
Democracy 7.99*** 5.15** 5.16** 5.68*** 6.28***
(3.33) (2.31) (2.31) (2.49) (2.69)
Democracy2 8.41*** 6.11*** 6.12*** 6.57*** 7.10***
(3.79) (2.94) (2.95) (3.03) (3.20)
Zd: demographic core
ln(D1) 2.95*** 2.17***
(3.45) (2.99)
ln(D2) 0.01 0.32
(0.01) (0.58)
Density 0.14 0.15
(0.66) (0.69)
ln(N) 0.10 0.10
(0.99) (1.04)
Translations model
Ngr 0.29** 1.12*** 0.12 0.13
(1.71) (3.63) (0.40) (0.40)
WAgr 1.51*** 0.51* 0.23
(4.96) (1.68) (0.75)
(continued)
20 Allen C. Kelley and Robert M. Schmidt

Table 2.1. Continued

Dependent variable Simple Translations Enriched
Demography Demography Demography

(Y/N)gr (Y/N)gr (Y/WA)gr (Y/WA)gr (Y/WA)gr

(1) (2) (3) (4) (5)

Exogenous influences
Regional fixed effects
N. and C. America 0.01 0.26 0.26 0.34 0.22
(0.01) (0.62) (0.62) (0.78) (0.51)
S. America 1.24** 1.24** 1.24** 1.29*** 1.32***
(2.37) (2.57) (2.57) (2.61) (2.61)
Europe 0.91 1.12** 1.12** 0.76 0.70
(1.55) (2.04) (2.04) (1.32) (1.20)
Asia 1.47*** 1.35*** 1.35*** 1.27*** 1.21***
(3.78) (3.72) (3.72) (3.39) (3.19)
Oceania 0.87 0.61 0.61 0.96 0.89
(1.23) (0.92) (0.92) (1.44) (1.31)
Period fixed effects
Pd:1970–80 0.47 0.91*** 0.91*** 0.98*** 0.91***
(1.36) (2.68) (2.68) (2.91) (2.76)
Pd:1980–90 1.83*** 2.37*** 2.37*** 2.47*** 2.37***
(4.81) (6.41) (6.40) (6.65) (6.44)
Pd:1990–95 2.27*** 2.84*** 2.84*** 2.87*** 2.83***
(5.80) (7.48) (7.48) (7.47) (7.35)
Intercept
Constant 23.95*** 9.80 9.79 3.78 9.03
(4.59) (1.83) (1.83) (0.54) (1.41)
R Squared 0.55 0.61 0.60 0.59 0.58
Adj. R-Sq. 0.53 0.58 0.57 0.56 0.56
Std Error 1.88 1.74 1.74 1.76 1.78
# of Obs 344 344 344 344 344

Note: Descriptive Statistics, the country sample, and estimation details are provided in the
appendix. The change of dependent variable between Models 2 and 3 eliminates the need
for ‘‘translations.’’ By definition, the coefficients for Ngr and WAgr in model 3 are those of
Model 2 minus the translations values of 1 and þ1, respectively. Within the limits of
estimation precision, all other coefficients and t-values are identical.
*
denotes 10% significance,
**
denotes 5% significance, and
***
denotes 1% significance for the appropriate 1- or 2-tailed test. t-values are in parentheses.

could deviate from its translations’ prediction (e.g., WAgr or Ngr could
deviate from 1 or 1) either if it has direct long-term impacts, or if the
model is specified incompletely (e.g., if the translations variables are
correlated with omitted impacts of population).
 Evolution of Recent Economic-Demographic Modeling 21
By contrast, Model 3 specifies (Y/WA)gr as the dependent variable
[Model 2 employs (Y/N)gr]. Having removed their translations role,
Model 3’s inclusion of Ngr and WAgr provides an explicit test for direct
productivity impacts. A statistically significant coefficient for Ngr and/or
WAgr would indicate a role beyond translations, perhaps for one of the
reasons discussed above in the section outlining the roles of population.
Models 4 and 5, termed Enriched Demography, append four demo-
graphic variables [Zd = ln(D1), ln(D2), Density, ln(N)] that have the poten-
tial for influencing long-run output per worker. Additionally, Model 4
constrains two translations parameters [ln(WA/N) and ln(WA/N)Ed] to
their theoretical expectations in a translations framework, whereas Model
5 constrains all four (also Ngr and WAgr).

2.3.2.1. Overall fit and economic core results

The convergence framework appears to be a reasonable paradigm for
assessing the roles of population. The models fit the data satisfactorily
with R2’s ranging from 55 to 61%. All economic core variables are of the
expected sign, and 9 of those 12 variables are significant at the 5% level in
each of the five variants. Over half of the coefficients are significant at the
1% level. Two variables are never significant (the convergence term’s
interaction with education and the education variable itself), and one
variable (government consumption’s share in GDP) is significant at the
5% level in the simplest demographic variant and at the 10% level in the
enriched demographic variants. For the most part, the Ze variables are not
particularly sensitive to the demographic specification either in coefficient
magnitude or in statistical significance.
A qualification on the education impacts is in order. Our weak results
are not uncommon in the convergence literature. Although those results
have been found to be somewhat sensitive to the education measure used
as well as to functional form, we have chosen not to experiment with the
core model. Rather, we have adopted Barro’s formulation given our
methodological posture of building upon an arm’s-length specification
of the core economic and political variables. Having said this, we did
rerun Model 5 without the ln(Y/WA)Ed interaction as a sensitivity experi-
ment. The MaleEduc coefficient fell slightly and remained insignificant.
Coefficients and t-values of other variables changed very little.

2.3.2.2. Population results: Translations

Considering first Model 1, we observe that population growth (Ngr) has a
negative impact on per capita output growth that almost quadruples (it
declines from 0.29 to 1.12) when working-age growth (WAgr) is taken
into account (see Model 2). Presumably WAgr’s positive impacts are
22 Allen C. Kelley and Robert M. Schmidt
incorporated into the estimated coefficient on Ngr in Model 1, muddying
the measurement of population’s impact (see also Bloom and Williamson
1997). Such a result plausibly accounts for the mixed results of Ngr in
numerous empirical studies. Thus, while population growth may well have
mattered in those studies, impacts were offsetting because the frameworks
were excessively parsimonious. Kelley (1988, 1701) observes that such
results ‘‘. . . provide little prima facie information about the size and nature
of the net impact . . .’’ We further argue (Kelley and Schmidt 1995, 545) that
the insignificant result ‘‘. . . does not mean that demographic processes are
unimportant; . . . [rather, it] may imply that strong intertemporal demo-
graphic effects are offsetting.’’
Turning to the Harvard translations model (Number 2 in Table 2.1) in
which Ngr and WAgr are included to translate per-worker into per capita
growth, note that Ngr does not play a role beyond pure translation. That is,
its coefficient is not statistically different from 1 at the 10% level (the
p-value from that test, not included in the table, is 0.689). By contrast,
WAgr’s coefficient of 1.51 is notably larger than the pure translations
coefficient of þ1 and that discrepancy is significant at the 10% level
(p-value of 0.094). Model 3 makes these points more directly since it
eliminates the need for such translation by using Y/WAgr rather than
(Y/N)gr as the dependent variable. In this variant, the coefficient for Ngr
does not differ significantly from zero whereas that for WAgr is positive
(0.51) and significant at the 10% level.
This last result is quite surprising in light of the above discussion of
measuring demographic impacts indirectly, which provides three separate
rationales for including WAgr in the labor productivity model. While each
of those arguments implied a negative impact on productivity growth, the
estimated impact here is positive, quantitatively notable, and statistically
significant at the 10% level. We believe this perverse outcome to be the
result of omitted variable bias. We argued above that the savings influence
attributed to WAgr might be captured better by direct measures of the age
structure. We have done that in Model 4, which enriches the Harvard
model by adding youth and elderly dependency ratios, ln(D1) and ln(D2),
as well as density, Density, and population size, ln(N). In that model, the
perverse estimate disappears—the coefficient for WAgr drops to 0.23 and
it is insignificant at the 10% level (t-value of 0.75). Neither Ngr nor WAgr
plays any significant role beyond translations.
A surprising story emerges as well for the translations variable ln(WA/N),
included in Models 2 and 3, to translate the convergence term from ln(Y/WA)
to ln(Y/N). Its coefficient (11.56) differs statistically and substantively from
its translations prediction of the negative of the convergence term (1.18).
Although this result is not new to the literature (see, for example, Bloom,
 Evolution of Recent Economic-Demographic Modeling 23
Canning, and Malaney 2000), the striking disparity between the coefficients
has not been explained satisfactorily. As was the case with WAgr, we would
argue that this disparity is likely the result of omitted variable bias rather than
an impact of WA/N on long-run, steady-state productivity.
We have suggested a theoretical reason, in our discussion of the roles
of population, for including two correlates of ln(WA/N), the youth and
elderly dependency ratios [ln(D1) and ln(D2)], in the productivity model.
We believe that their omission from the model has resulted in a coeffi-
cient estimate for ln(WA/N) that is dramatically different from the con-
vergence coefficient. Unfortunately, we cannot include all three variables
in the model owing to the high correlations between ln(WA/N) and ln(D1)
(0.96) as well as ln(D2) (0.79). We have taken a different tack. Models 4
and 5 use ln(Y/WA) as the convergence term rather than ln(Y/N) in
conjunction with ln(WA/N). Neither variant is preferable theoretically,
and to our knowledge no empirical growth model has included ln(WA/N)
among Z vector influences on the steady-state productivity level. Having
eliminated the need for including ln(WA/N), Models 4 and 5 do include
youth and elderly dependency ratios as well as additional demographic
variables.

2.3.2.3. Demography results: Zd variables

We turn finally to the productivity roles that population may play in the
Barro convergence model. Models 4 and 5, denoted as enriched demo-
graphy, append four demographic variables to the list of variables influen-
cing an economy’s long-term productivity level. This Zd vector includes
ln(D1), ln(D2), Density, and ln(N). While each carries its predicted sign,
only ln(D1) is significant (at the 1% level). Moreover, the coefficients on
ln(D1) appear to be large. Of course, the relative importance of the various
variables cannot be determined solely by reference to their coefficients. As
a result, the following section of this chapter assesses the magnitude of
these impacts in the context of the world’s experience.
Before turning to that, however, we consider again the key theoretical
insight of our section on the roles of population and the empirical lesson of
the current section on population and growth. Population plays multi-
faceted roles in the economy. Certain influences are negative whereas
others are positive; and some are felt immediately whereas others are felt
with lags of 10, 15, 20, or more years. But all are interrelated. Conse-
quently, modeling demographic effects by any single variable such as Ngr
is overly simplistic and potentially misleading. Moreover, these influences
are separable into two distinct categories—a translations set (Ngr and
WAgr) and a productivity set [in our rendering, ln(D1), ln(D2), Density,
and ln(N)]. Furthermore, this dichotomy is pure in the sense that the results
24 Allen C. Kelley and Robert M. Schmidt
from Models 3 and 4 demonstrate no productivity influence of the transla-
tions variables. Indeed, the more demographic detail incorporated into the
model, the more sharply separable are the two models.
Additionally, the finding of a strong and significant coefficient for (D1)
provides insight into the interpretation of the translations demography in
convergence modeling. Specifically, in theoretical terms, the impacts of
the age distribution on the saving rate (S/Y ), and thus (Y/N)gr, are typically
specified in terms of the age-structure levels, such as dependency rates or
similar summary measures. Thus, while dependency is correlated with the
growth rates of population and the working-age population, the appro-
priate analytical connection is arguably better measured as dependency
levels. This point is made most forcefully for WAgr. Its productivity
impact when ln(D1) is excluded (col. 3 of Table 2.1) is estimated to be
0.51 and significant at the 10% level, two-tail. That impact falls to 0.23
and insignificance with the inclusion of the extended demography (col. 4).
In short, adding ln(D1) in the Barro core tends to provide more plausible
estimates of the translations variables. For further elaboration of saving’s
role, see Mason (1988), Higgins and Williamson (1997), and Lee, Mason,
and Miller (2001).
Finally, the insignificance of several demographic variables within this
broad country sample does not necessarily diminish their importance
within specific settings. For example, the influence of aging [ln(D2)] is
difficult, if not impossible, to identify given that aging was noteworthy in
only a few countries by the mid-1990s. Furthermore, the impacts of
density might be quite different in advanced economies such as Japan,
South Korea, Taiwan, and much of Europe than in poorer countries such
as India, Indonesia, Pakistan, and Bangladesh. Greater distinction between
demographic effects in developed and developing countries merits addi-
tional consideration. Regional and geographic factors using the parameter
estimates from this study are explored in Chapter 3 of this volume. Those
findings offer considerable encouragement to these directions of research.

2.4. Population matters: But how, and by how much?

A common methodology for assessing impacts within a regression model

is to apply estimated coefficients to period means of the corresponding
variables. However, assessing the ‘‘importance’’ of demographic trends
must account not only for the size of the coefficients and mean levels of
the variables, but also for changes in the variables over time. Such a
reckoning is compiled in Table 2.2, which applies the regression coeffi-
cients from Model 5 of Table 2.1 (our preferred rendering) to changes in
 Table 2.2. Accounting for changes in (Y/N)gr over time: Impacts of interdecadal change
World Sample North and Central America South America

Evolution of Recent Economic-Demographic Modeling

1960s 1970s 1980s Avg. 1960s 1970s 1980s Avg. 1960s 1970s 1980s Avg.
to 70s to 80s to 90s to 70s to 80s to 90s to 70s to 80s to 90s

Change in (Y/N)gr 0.84 1.73 0.28 0.77 1.15 2.44 1.02 0.86 0.27 3.58 3.31 0.00
Productivity model
Convergence 0.38 0.26 0.03 0.22 0.36 0.18 0.14 0.13 0.28 0.27 0.17 0.13
Ze: economic core 0.05 0.29 0.58 0.08 0.38 0.75 1.48 0.12 0.26 1.82 1.56 0.17
Financial 0.34 0.66 0.19 0.27 0.67 1.03 0.83 0.29 0.71 2.03 1.30 0.48
Human K: ln(e0) 0.37 0.35 0.28 0.33 0.45 0.36 0.34 0.38 0.32 0.35 0.29 0.32
Human K: Male Educ 0.06 0.09 0.07 0.07 0.06 0.11 0.08 0.08 0.02 0.09 0.04 0.05
Political 0.13 0.07 0.04 0.05 0.22 0.19 0.23 0.06 0.12 0.23 0.07 0.06
Zd: demographic core 0.09 0.26 0.18 0.18 0.16 0.34 0.22 0.24 0.15 0.26 0.19 0.20
ln(D1) 0.08 0.24 0.17 0.17 0.14 0.32 0.21 0.22 0.13 0.24 0.18 0.18
ln(D2) 0.02 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.00 0.01 0.01
Density 0.00 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
ln(N) 0.02 0.02 0.02 0.02 0.03 0.02 0.02 0.02 0.02 0.02 0.02 0.02
Translations model
Demographic translations 0.40 0.02 0.01 0.14 0.62 0.03 0.02 0.19 0.46 0.09 0.07 0.15
Ngr 0.13 0.14 0.19 0.15 0.29 0.23 0.05 0.19 0.24 0.26 0.17 0.22
WAgr 0.27 0.12 0.17 0.01 0.33 0.25 0.06 0.00 0.22 0.35 0.10 0.08
Exogenous influences
Period fixed effects 0.91 1.47 0.46 0.94 0.91 1.47 0.46 0.94 0.91 1.47 0.46 0.94

(continued)

25
 26
Table 2.2. Continued
Europe Africa Asia

1960s 1970s 1980s Avg. 1960s 1970s 1980s Avg. 1960s 1970s 1980s Avg.
to 70s to 80s to 90s to 70s to 80s to 90s to 70s to 80s to 90s

Change in (Y/N)gr 1.55 0.48 1.21 1.08 0.90 2.30 0.81 1.34 0.23 0.84 0.74 0.11
Productivity model

Allen C. Kelley and Robert M. Schmidt

Convergence 0.53 0.33 0.24 0.37 0.28 0.16 0.14 0.10 0.47 0.41 0.33 0.40
Ze: Economic core 0.39 0.38 0.09 0.03 0.15 0.26 0.07 0.16 0.77 0.07 0.82 0.55
Financial 0.67 0.21 0.13 0.20 0.20 0.80 0.47 0.49 0.20 0.43 0.32 0.03
Human K: ln(e0) 0.14 0.18 0.15 0.16 0.46 0.43 0.28 0.39 0.43 0.41 0.38 0.41
Human K: Male Educ 0.12 0.17 0.08 0.12 0.02 0.03 0.05 0.03 0.06 0.10 0.10 0.09
Political 0.01 0.18 0.01 0.06 0.43 0.08 0.06 0.09 0.08 0.02 0.01 0.03
Zd: Demographic core 0.06 0.43 0.23 0.24 0.04 0.03 0.09 0.03 0.22 0.36 0.26 0.28
ln(D1) 0.10 0.44 0.24 0.26 0.06 0.01 0.06 0.00 0.20 0.33 0.25 0.26
ln(D2) 0.05 0.01 0.02 0.03 0.00 0.00 0.00 0.00 0.02 0.02 0.03 0.02
Density 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.02 0.02 0.02
ln(N) 0.01 0.01 0.00 0.01 0.03 0.03 0.03 0.03 0.03 0.02 0.02 0.02
Translations model
Demographic translations 0.34 0.18 0.26 0.08 0.24 0.10 0.21 0.18 0.53 0.14 0.03 0.14
Ngr 0.19 0.21 0.13 0.09 0.11 0.11 0.50 0.09 0.18 0.31 0.18 0.23
WAgr 0.14 0.03 0.13 0.01 0.35 0.22 0.29 0.09 0.34 0.46 0.15 0.09
Exogenous influences
Period fixed effects 0.91 1.47 0.46 0.94 0.91 1.47 0.46 0.94 0.91 1.47 0.46 0.94
Note: ‘‘Convergence’’ includes initial economic level and its interaction with education; ‘‘Financial’’ includes terms of trade, government consumption share, and
inflation; separate ‘‘Human Capital’’ rows are included for the reader who prefers to treat life expectancy as a demographic variable; and ‘‘Political’’ includes rule of law
and democracy. Exogenous impacts include only period fixed effects and are the same for all regions since the regional fixed effects and the intercept are constant across
decades. By definition, regression impacts will add up (except for rounding) to ‘‘Change in (Y/N)gr’’ for each World Sample column but not for the regional columns. The
calculations are based on Table 2.1, col. 5, augmented by the theoretically specified coefficients for Ngr and WAgr, 1 and þ 1, respectively. Although Oceania is not
included as a separate region in this table, its three countries are included in the worldwide sample.
 Evolution of Recent Economic-Demographic Modeling 27
mean values across decades, thereby estimating impacts on interdecadal
changes in (Y/N)gr.
Table 2.2 groups impacts into broad categories denoted as ‘‘productiv-
ity model’’ (convergence, Ze: economic core, and Zd: demographic core),
‘‘translations model,’’ or ‘‘exogenous influences’’ (period fixed effects).
The table presents these results first for our ‘‘world’’ sample and for five
regions (North and Central America, South America, Europe, Africa, and
Asia) as described in the appendix. Given the extensive results in Table 2.2,
we focus initially in this section on net impacts of the major components
of the model. Next we provide additional detail by assessing contribu-
tions from the variables within those components. Finally, we summarize
the major lessons and provide bottom-line estimates of the relative
contributions of the various influences.

2.4.1. Population matters: Broad conclusions from the world

and regional samples
Table 2.2 provides some interesting insights into trends during the last
four decades of the twentieth century and in the process provides a
warning against overaggregation. For example, were we to focus exclu-
sively on the world’s ‘‘average’’ column (the equivalent of estimating
from a single 35-year cross-section) and the five variable groupings, we
would conclude that both the core demographic trends (0.18) and the
demographic translations (0.14) have dominated the impact of the eco-
nomic trends (0.08) on (Y/N)gr changes. Furthermore, at 0.94, exogenous
shocks have had several times the impact of the economic and demo-
graphic core variables. These conclusions hold across all regions
except Asia, where economic impacts (0.55) dominated the demographic
core (0.28) and translations (0.14). While interesting, these generali-
zations conceal considerable variability across the individual decades
and regions.
Consider, first, time patterns in the global aggregates. Worldwide,
positive economic impacts (Ze) between the 1980s and 1990s (0.30)
reversed two decades of growth-inhibiting trends (0.05 and 0.64). By
contrast, demographic trends have been consistently growth enhancing.
Thus, while (Y/N)gr declined notably between the 1960s and 1970s and
again between the 1970s and 1980s, those declines would have been even
larger without the ameliorating impacts of favorable demographic trends.
Although the positive impacts of changes in the demographic core (Zd)
displayed no obvious trend over time (0.09, 0.26, 0.18), a favorable net
trend in the demographic translations variables was largely exhausted in
the 1970s (0.40, 0.02, 0.03)—a fascinating finding.
28 Allen C. Kelley and Robert M. Schmidt
How do these global conclusions fare across the regions of the world?
The Americas follow the worldwide Ze pattern, with South American
countries experiencing the largest negative (1.82) and positive (1.56)
impacts. While Africa suffered negative Ze impacts throughout, positive
impacts were enjoyed in Europe in two periods and in Asia in all three
periods. By contrast, the regions followed global demographic patterns
quite well. Trends in the Zd variables contributed positively and notably to
growth throughout the period in four of the five regions. As is the case
worldwide, favorable impacts from the translations variables were largely
exhausted in the 1970s in the same four regions. Indeed, in each of those
regions, translations impacts were negative in at least one of the last two
periods. Africa displays demographic impacts different from those of the
other regions because of stubbornly high fertility rates throughout most of
the period. Correspondingly, translations impacts are smaller initially but
remain positive and are nearly as large in the last period as in the first
(0.24, 0.10, 0.21). On the other hand, Zd impacts are quite small in Africa,
starting out negative but rising gradually (0.04, 0.03, 0.09).
Finally, we briefly consider the impacts of ‘‘convergence’’ and exo-
genous influences. Since the log level of per-worker income [ln(Y/WA)] is
inversely related to (Y/N)gr, the estimated convergence impact will be
negative between any two decades of growth in ln(Y/WA) and positive
after a decline. For the world as a whole and in three of the five regions,
convergence had a negative impact during the first two periods because
economic growth dictated higher ln(Y/WA) in 1970 than 1960 and in 1980
than 1970, but a positive impact due to the widespread decline over the
1980s, resulting in a lower Y/WA in 1990 than 1980. Europe and Asia
bucked this trend, enjoying growth in every decade—implying negative
convergence impacts in all three periods. Humbling to the growth litera-
ture is the role played by the largely unexplained exogenous factors. They
exert the largest influences in Table 2.2, and they are consistently nega-
tive. (Note that exogenous impacts are the same across all regions because
regional fixed effects are held constant across all periods and thus drop out
when interdecadal changes are calculated.)

2.4.2. Population matters: Digging deeper into components

of Ze , Zd , and translations
A review of the components of these broad categories enriches this
analysis further, both temporally and spatially. Within the economic
core, trends in human capital (life expectancy and education) have been
strongly growth inducing globally (0.42, 0.44, 0.35) as well as in every
region. That consistency is not found in the financial (terms of trade,
 Evolution of Recent Economic-Demographic Modeling 29
government consumption, and inflation) and political (rule of law and
democracy) components, the combined negative impacts of which more
than offset these gains in the first two periods before enhancing the gains
with favorable impacts in the last period (0.34, 0.66, 0.19). This
worldwide trend was largely followed in North and Central America
(0.67, 1.03, 0.83) and in South America (0.71, 2.03, 1.30), but
not in Africa, which suffered throughout (0.20, 0.80, 0.47), nor in
Europe (0.67, 0.21, 0.13) or Asia (0.20, 0.43, 0.32), which enjoyed
positive impacts in one of the earlier periods. We find that the largest
negative and positive financial impacts were experienced in South Amer-
ica, followed by North and Central America (acknowledging that this
model is not designed to fully capture the Asian monetary crisis of the
late 1980s). Although the political component is the least influential
worldwide (0.13, 0.07, 0.04), this is the net result of advances and
retreats regionally. For example, whereas South America made strides in
the political arena between the 1960s and the 1970s, the region fell back in
the latter two periods (0.12, 0.23, 0.07). Africa demonstrated the
opposite pattern (0.43, 0.08, 0.06) after the colonial era ended on that
continent.
Turning to population, we see several interesting results emerge. First,
youth dependency [ln(D1)] strongly dominates other components of the
demographic core [ln(D2), ln(N), Density] in influence on (Y/N)gr trends,
and its impact has been consistently positive. This is true in every region
of the world with the exception of Africa, where, because of delayed
fertility declines, youth dependency remained high throughout the latter
half of the twentieth century. By contrast, aged dependency has had a
negligible effect to date. The impacts of aging may well reveal themselves
as the twenty-first century progresses. Large youth-dependency ratios
have already been observed, and their impacts can therefore be estimated.
By contrast, the largest elderly dependency ratios lie in the future, and
those effects cannot be estimated well from this sample.
Second, the positive economic impacts of the demographic core could
be enhanced considerably in all regions and in all periods depending upon
how much of the gains from rising life expectancy one ascribes to
demographic variables rather than to human capital. As we argued earlier,
we believe that the demographic side of life-expectancy impacts is
largely controlled for within this model by other demographic indicators.
Nevertheless, were an analyst to disagree and impute all life-expectancy
impacts to demographic variables, the effect on Zd impacts would be
substantial.
Third, within the Translations Model, declining population growth
(Ngr) positively impacted trends in (Y/N)gr relative to (Y/WA)gr throughout
30 Allen C. Kelley and Robert M. Schmidt
the last four decades of the twentieth century worldwide and in three of the
five regions of Table 2.2. The exceptions are Africa in the first two periods
for reasons discussed above and Europe in the last period. Trends in the
growth of the working-age population (WAgr) enhanced (Y/N)gr change
six times in the first two periods but not a single time since.
In short, the prevailing pattern throughout the world of a positive and
strong net translations impact between the 1960s and the 1970s, followed
by small and even negative net impacts in the latter two periods, has been
driven to date largely by initially strong but eventually declining growth
rates of the working-age population. Again, the notable exception is in
Africa where, subject to the ravages of HIV-AIDS, such a period may lie
ahead as the countries of that continent pass through the various phases of
the demographic transition.

2.4.3. Population matters: Summary lessons and bottom-line

calculations
What lessons can be learned from the rich regional detail of Table 2.2?
First, population can and has exerted quantitatively large impacts on
changes in (Y/N)gr, especially if one combines the impacts of both core
and translations demography (and even more if one includes life expec-
tancy in the demographic core).
Second, the translations impacts trace out a fairly consistent pattern
following a fertility boom and decline (for example, over the course of the
demographic transition). That pattern consists of (a) strong positive
impacts (as Ngr declines from lower fertility while WAgr rises from earlier
high fertility); and (b) small impacts, and possibly negative impacts (as Ngr
stabilizes and WAgr eventually slows due to earlier fertility declines). This
result both qualifies and elaborates the renderings of the recent literature.
The magnitude and timing of these impacts can vary dramatically across
countries and regions. Africa (with a delayed transition and positive
impacts throughout the period) and Asia (with a rapid transition in key
countries and negative impacts by the second period) provide stark con-
trasts in this regard.
Third, the demographic core variables, especially youth dependency,
change more slowly and therefore exert their influences over longer
periods than do the translations variables. For this reason we argue the
importance of a richer modeling of demographic change to an improved
understanding of economic change.
Having said all of this, we remain curious about the relative contribu-
tions of the various influences to variability in (Y/N)gr changes around the
world. Table 2.2 is inadequate for this task since (Y/N)gr changes derive
 Evolution of Recent Economic-Demographic Modeling 31
from variables that have both positive and negative impacts. As a result,
(Y/N)gr change can be quite small by comparison with its component influ-
ences. (See, for example, the 1980s–1990s in Table 2.2.) The net value
masks considerable ‘‘economic activity,’’ which can be negative as well as
positive. As an alternative approach, we have computed component shares
in total ‘‘movements’’—the sum of the unsigned impacts of all variables
(including the exogenous variables that result from period shifts in events
such as OPEC decisions and debt overhang). These movements are shown
in Table 2.3. For the world sample, core demography accounts for 8% of
total movements across the interdecadal periods. Among the other factors,
human capital (education, health) and financial and economic factors have
the most important impacts (15% apiece), followed by demographic
translations (13%), the convergence adjustments (9%), and politics (4%).
As previously noted, exogenous factors dominate interdecadal changes
(36%).
As can be seen in Table 2.3, these figures vary notably by region. They
can also vary over time and by the degree of disaggregation, justifying
caution in interpreting the results. However, the general conclusion that
‘‘population matters’’ is beyond dispute and holds everywhere: in the
aggregate it ranges from 8% of ‘‘movements’’ in the demographic core
to 21% if translations impacts are included, and to 34% if life expectancy
is interpreted not as a proxy for health but rather as largely a demographic
variable.

Table 2.3. Accounting for changes in (Y/N)gr over time: Percentage

shares in total movements
World N. & C. South Europe Africa Asia
Sample America America

Productivity model
Convergence 9 7 6 13 7 12
Ze: economic core 35 46 55 35 43 34
Financial 15 25 35 16 21 13
Human K: ln(e0) 13 11 8 5 13 12
Human K: Male Educ 3 2 1 4 1 3
Political 4 7 10 9 8 7
Zd: demographic core 8 8 5 24 3 28
Translations model
Demographic translations 13 12 11 10 17 16
Exogenous influences
Period fixed efects 36 28 23 33 31 28
32 Allen C. Kelley and Robert M. Schmidt
2.5. Conclusion

This chapter has examined various ways in which demography has been
incorporated into ‘‘convergence modeling,’’ as pioneered by Robert J.
Barro and extended by Harvard scholars and others. Our interpretation
of this literature distinguishes somewhat sharply between the impacts of
Harvard translations additions and the more traditional demographic
impacts on the economy. We have proposed, and found in our empirical
analysis, that the impacts of translations demography are best viewed as
largely ‘‘neutral’’ on economic production, although the translations fra-
mework nicely exposes effects on potential consumption (and welfare)
and, importantly, significantly clarifies the roles of other demographic
variables (e.g., dependency, size, and density).
These demographic impacts (deriving mainly from declining birth and
death rates) combine to exert positive contributions to trends in per capita
GDP growth. Worldwide, the combined impacts of demographic change
have accounted for approximately 20% of per capita output growth impacts,
with larger shares in Asia and Europe. And, in the not too distant future,
demographic change (this time deriving from low and stable death and birth
rates) will likely exert negative impacts on growth. To see how these results
can materialize, we propose that future modeling build on the type of
demographic disaggregation illustrated in this chapter, where greater dis-
tinction is made between demographic change that affects output growth per
worker and that which translates such growth into per capita terms. A
theoretical modeling perspective that synthesizes the Barro convergence
framework, augmented to include several traditional demographic variables
such as population age structure and size, and unified within the Harvard
translations framework, provides a promising and relatively clear structure
for revealing the roles of demographic change on the economy.

Appendix

Table 2.A1 presents descriptive statistics and data sources for the variables
in this study. Following Barro (1991, 1997), we use a panel with 10-year
growth periods. Our sample, grouped by region, includes the following 86
countries. Africa (26): Algeria, Benin, Cameroon, Central Africa, Ivory
Coast, Egypt, Ethiopia, Ghana, Kenya, Madagascar, Malawi, Morocco,
Niger, Nigeria, Rwanda, Senegal, Sierra Leone, South Africa, Sudan, Tan-
zania, Togo, Tunisia, Uganda, Zaire, Zambia, Zimbabwe. North and Cen-
tral America (12): Canada, Costa Rica, Dominican Republic, El Salvador,
Guatemala, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, United
 Evolution of Recent Economic-Demographic Modeling 33
Table 2.A1. Variable definitions, sources and descriptive statistics
Variable Description Source Mean Std. Min Max
Dev.

(Y/N)gr Per capita GDP (PPP) growth Trans 1.65 2.69 10.77 8.64
rate
(Y/WA)gr Per working-age GDP (PPP) Trans 1.46 2.66 11.14 7.94
growth rate
ln(Y/N) Per capita GDP (PPP, log) SH 0.85 1.03 1.35 2.89
ln(Y/WA) Per working-age GDP (PPP, log) SH, UN 1.42 0.96 0.72 3.31
ln(WA/N) Ratio of working-age to total UN 0.57 0.10 0.76 0.32
pop. (log)
ln(Y/N)  Ed Interaction: ln(Y/N)  MaleEduc Trans 0.95 1.48 0.84 10.98
ln(Y/WA)  Ed Interaction: ln(Y/WA)  MaleEduc Trans 0.86 1.36 0.91 10.14
ln(WA/N)  Ed Interaction: ln(WA/N)  MaleEduc Trans 0.09 0.14 0.14 0.83
TT %chg Terms-of-Trade, percentage WB 0.43 3.54 13.73 19.25
change
Gcons/Y Pct share of gov’t consumption WB, BL93 7.27 3.60 0.01 27.19
in GDP
Inflation Inflation rate WB 15.71 28.71 0.74 317.10
ln (e0) Life expectancy at birth (log) WB 4.07 0.21 3.46 4.37
MaleEduc Avg years post-primary educ, BL96 1.29 1.19 0.02 6.67
Males 25þ
Rule Law Index: Law & order tradition ICRG 0.56 0.24 0.10 1.00
Democracy Democracy (political rights G 0.58 0.33 0.00 1.00
index)
Ngr Population Growth Rate Trans 2.02 1.01 2.18 4.06
WAgr Working-age population Trans 2.21 1.05 1.81 4.23
growth rate
ln(D1) 100  Ratio of ages 0–14 to UN 4.14 0.41 3.14 4.67
15–64 (log)
ln(D2) 100  Ratio of ages 65þ to UN 2.14 0.51 1.21 3.31
15–64 (log)
Density Density: 1,000 population WB 0.17 0.61 0.00 5.77
per Sq. Km
ln(N) Population size (log) WB 9.37 1.26 7.06 13.65
Note: Additional definitional details are found in Kelley and Schmidt (2001). Data fills and
extrapolations were made by imposing rates of change from an alternative data set with more
complete series. For SH, WB was the primary filling source with UN and IMF as alternatives.
WB was generally filled from earlier versions, UN sources, or SH. Fills for ICRG and G are too
complicated to describe here; a description is available upon request. The ‘‘Source’’ column
from the appendix table uses the following key for data sources:
BL93 Barro and Lee’s data set used in Barro and Lee (1993).
BL96 Barro and Lee (1996) update of their education attainment series.
G Gastil (1991).
ICRG International Country Risk Guide.
SH Summers and Heston (1994) Penn World Tables, version 5.6.
Trans Transformation of variable described elsewhere in the table.
UN United Nations (1996).
WB World Bank’s (1994, 1995, 1997, 1999, and 2000) World Development Indicators
CD-ROM.
34 Allen C. Kelley and Robert M. Schmidt
States. South America (10): Argentina, Bolivia, Brazil, Chile, Colombia,
Ecuador, Paraguay, Peru, Uruguay, Venezuela. Asia (18): Bangladesh,
Hong Kong, India, Indonesia, Iran, Israel, Japan, Malaysia, Myanmar,
Nepal, Pakistan, Philippines, Singapore, South Korea, Sri Lanka, Syria,
Taiwan, Thailand. Europe (17): Austria, Belgium, Denmark, Finland,
France, Greece, Germany, Ireland, Italy, Netherlands, Norway, Portugal,
Spain, Sweden, Switzerland, Turkey, United Kingdom. Oceania (3): Aus-
tralia, New Zealand, Papua New Guinea.
Barro employs three-stage least-squares estimation, with the third stage
correcting for possible serial correlation. Since he found little evidence of
serial correlation, we opted for two-stage estimation instead for the vari-
ables identified by Barro to be endogenous: ln(Y/N), ln(Y/WA), ln(Y/
N)Ed, ln(Y/WA)Ed, Gcons/Y, Democracy and its square, and Inflation.
Following Barro, the first-stage equations include all exogenous or pre-
determined variables [ln(WA/N), TT% chg, ln(e0), MaleEduc, Rule Law,
Ngr, WAgr, ln(D1), ln(D2), Density, and ln(N)] together with 5-year lags of
ln(Y/N) or ln(Y/WA), ln(Y/N)’s or ln(Y/WA)’s interaction with contem-
poraneous education, Gcons/Y, Democracy, and Democracy’s squared
term. Finally, binaries for former colonies of Spain and Portugal and
former colonies of Great Britain and France are included as instruments
for inflation. The first-stage equations are run separately for each period.
The second-stage equation is pooled but includes period-specific binaries.
Table 2.A2 includes R2s from the first-stage regressions for our pre-
ferred variant, Model 5 from Table 2.1. As might be expected, inflation is
the hardest variable to predict with R2s ranging from 0.279 to 0.507.
Although these are low relative to the others, they are quite acceptable for
an instrumental-variables approach, and predicted inflation is consistently
significant in the second-stage regressions of Table 2.1. As might also be
expected, regressions for ln(Y/WA) and ln(Y/WA)Ed are the strongest,
partly due to the inclusion of their lagged values among the instruments.

Table 2.A2. R2s from first-stage regressions for Model 5 of Table 2.1
Endogenous Variable Regression Period

1960s 1970s 1980s 1990–95

ln(Y/WA) 0.995 0.984 0.987 0.995

ln(Y/WA)  Ed 0.994 0.993 0.996 0.994
Gcons/Y 0.740 0.508 0.556 0.520
Inflation 0.385 0.507 0.417 0.279
Democracy 0.921 0.843 0.869 0.851
Democracy2 0.946 0.863 0.884 0.912
 Evolution of Recent Economic-Demographic Modeling 35
2
With R s ranging from 0.987 to 0.996, the predicted values nearly replicate
the actual values. The full set of results from these first-stage regressions is
available upon request.
Problems of reverse causation may plague demographic variables as
well, although here the case is less clear. On the one hand, fertility rates
are likely to be more sensitive to the level than to the growth of income.
On the other hand, the length of the observations used in empirical studies
ranges from 5 to 25 years, resulting in periods sufficiently long that
the levels can change notably through growth. Our estimation uses an
intermediate period of 10 years. Consequently, we assessed the need to
instrument key demographic variables [WAgr, Ngr, ln(D1)] through the
Durbin-Wu-Hausman test (appending lagged demographic variables to the
above list of instruments). In no demographic variant was that test sig-
nificant at the 5% level. This result is consistent with that of Brander and
Dowrick (1994), who present one of the most econometrically intensive
analyses in the literature using instrumental variables for birth rates in a
production function setup. They conclude that, ‘‘there is no evidence that
the demographic variables are endogenous with respect to income growth
rates’’ (p. 18). As a result, we do not instrument any of the demographic
variables in the estimation presented in this study. As a sensitivity experi-
ment, we ran an additional set of 2SLS regressions that included first-stage
regressions for WAgr, Ngr, and ln(D1), which are thought to be most
sensitive to fertility changes. Coefficient estimates changed negligibly.
A more thorough explanation and rationale for our estimation proce-
dures can be found in Barro (1997) and Kelley and Schmidt (2001).

References

Barro, Robert J. 1991. ‘‘Economic growth in a cross section of countries’’. Quarterly

Journal of Economics 106(2): 407–44.
———. 1997. Determinants of Economic Growth: A Cross-country Empirical Study.
Cambridge, MA: MIT Press.
Barro, Robert J., and Jong-Wha Lee. 1993. ‘‘Losers and winners in economic growth’’.
Working Paper 4341. NBER, Cambridge, MA.
———.1994. ‘‘Sources of economic growth’’. Carnegie-Rochester Conference Series on
Public Policy 40: 1–46.
———. 1996. ‘‘International measures of schooling years and schooling quality’’. American
Economic Review 86(2): 218–23.
Bloom, David E., and David Canning. 2001. ‘‘Cumulative causality, economic growth, and
the demographic transition’’, in: Nancy Birdsall, Allen C. Kelley, and Steven Sinding,
eds, Population Matters: Demographic Change, Economic Growth, and Poverty in the
Developing World, pp. 165–97. New York: Oxford University Press.
———. 2003. ‘‘From demographic lift to economic liftoff: The case of Egypt’’. Applied
Population and Policy 1: 15–24.
36 Allen C. Kelley and Robert M. Schmidt
Bloom, David E., David Canning, and Bryan Graham. 2003. ‘‘Longevity and life cycle
savings’’. Scandinavian Journal of Economics 105: 319–38.
Bloom, David E., David Canning, and Pia Malaney. 2000. ‘‘Population dynamics and
economic growth in Asia’’, in: C. Y. Cyrus Chu and Ronald Lee, eds, Population and
Economic Change in East Asia. Population and Development Review, Supplement to
Vol. 26: 257–90.
Bloom, David E., and Jeffrey G. Williamson. 1997. ‘‘Demographic change and human
resource development’’, in: Emerging Asia: Changes and Challenges, pp. 141–97.
Manila: Asian Development Bank.
———.1998. ‘‘Demographic transitions and economic miracles in emerging Asia’’. World
Bank Economic Review 12(3): 419–55.
Brander, J.A., and Steven Dowrick. 1994. ‘‘The role of fertility and population in economic
growth: Empirical results from aggregate cross national data’’. Journal of Population
Economics 7: 1–25.
Gastil, Raymond D. 1991. ‘‘The comparative survey of freedom: Experiences and
suggestions’’, in: Alex Inkeles, ed., On Measuring Democracy: Its Consequences and
Concomitants, pp. 21–46. New Brunswick, NJ: Transaction Publishers.
Higgins, Matthew, and Jeffrey G. Williamson. 1997. ‘‘Age structure dynamics in Asia and
dependence on foreign capital’’. Population and Development Review 23(2): 261–93.
ICRG (International Country Risk Guide). 1982–95. ‘‘International country risk guide’’. Political
Risk Services; obtained from IRIS Center, University of Maryland, College Park, MD.
Kelley, Allen C. 1988. ‘‘Economic consequences of population change in the Third
World’’. Journal of Economic Literature 26(4): 1685–1728.
Kelley Allen C., and Robert M. Schmidt. 1994. ‘‘Population and income change: Recent
evidence’’. World Bank Discussion Papers, No. 249. Washington, DC: World Bank.
———. 1995. ‘‘Aggregate population and economic growth correlations: The role of the
components of demographic change’’. Demography 32(4): 543–55.
———. 1996. ‘‘Saving, dependency and development’’. Journal of Population Economics
9(4): 365–86.
———. 2001. ‘‘Economic and demographic change: A synthesis of models, findings, and
perspectives’’, in: Nancy Birdsall, Allen C. Kelley, and Steven Sinding, eds, Population
Matters: Demographic Change, Economic Growth, and Poverty in the Developing World,
pp. 67–105. New York: Oxford University Press.
Kinugasa, Tomoko. Life Expectancy, Labor Force, and Saving. Ph.D. dissertation. University
of Hawaii at Manoa.
Kinugasa, Tomoko, and Andrew Mason. 2005. ‘‘Why nations become wealthy: The effects
of adult longevity on saving’’. Working Paper No. 200514. Department of Economics,
University of Hawaii at Manoa.
Lee, Ronald D., Andrew Mason, and Tim Miller. 2001. ‘‘Saving, wealth, and population’’,
in: Nancy Birdsall, Allen C. Kelley. And Steven Sinding, eds, Population Matters:
Demographic Change, Economic Growth, and Poverty in the Developing World,
pp. 137–64. New York: Oxford University Press.
Levine, Ross, and David Renelt. 1992. ‘‘A sensitivity analysis of cross-country growth
regressions’’. American Economic Review 82(4): 942–63.
Mason, Andrew. 1988. ‘‘Saving, economic growth, and demographic change’’. Population
and Development Review 14 (March): 113–44.
Radelet, Steven, Jeffrey Sachs, and Jong-Who Lee. 2001. ‘‘The determinants and prospects
of economic growth in Asia’’. International Economic Journal 15(3): 1–29.
 Evolution of Recent Economic-Demographic Modeling 37
Summers, Robert, and Alan Heston,1994. PWT 5.6. Data update based on The Penn World
Table (Mark 5): An Expanded Set of International Comparisons, 1950–1988. These data
can be downloaded from http://pwt.econ.upenn.edu/Downloads/index.htm.
UNPD (United Nations Population Division). 1996. Sex and age annual data: 1950–2050
(1996 revision). Data diskettes. New York.
World Bank. 1994, 1995, 1997, 1999, 2000. World Development Indicators. CD-ROM.
Washington DC.
This page intentionally left blank
 CHAPTER 3

A Century of Demographic Change and

Economic Growth: The Asian Experience in
Regional and Temporal Perspective
Allen C. Kelley and Robert M. Schmidt

Assessing the impacts of demographic change on economic growth has

spurred an outpouring of research. The literature in the period since World
War II has vacillated between pessimism and optimism, with broad eclecti-
cism representing the median judgment. Very few empirical studies have
uncovered robust and strong macroeconomic effects of demographic change,
one way or the other—a surprise to pessimists in the Malthusian tradition.1
The 1990s witnessed a modest shift in judgment about population
consequences based on the use of convergence (or technology-gap) mod-
eling pioneered by Robert J. Barro (1991). His early empirical studies
uncovered negative consequences of high birth rates on economic growth.
In 1994 Kelley and Schmidt expanded this demography to include density
and population size, and in the late 1990s several Harvard economists
(e.g., David Bloom, David Canning, Jeffrey Sachs, and Jeffrey Williamson)
argued convincingly that whereas theorists conceptualize the eco-
nomic growth process in labor-productivity terms, empirical growth mod-
els are generally specified in per capita terms. Without an accounting
structure to translate labor-productivity impacts into per capita terms,
demographic coefficient estimates (mainly population growth) will be
biased. In that case the population-growth coefficient would capture net
demographic impacts that could be positive, negative, or neutral, depend-
ing upon time and place. Correspondingly, the Harvard rendering speci-
fied what we have labeled a ‘‘translations model’’ to accommodate these
accounting considerations. The resulting model is elegant in its simplicity,
incorporating only two demographic variables that have unambiguous

1
A history of various assessments is provided in Kelley (2001).

POPULATION CHANGE, LABOR MARKETS, AND SUSTAINABLE GROWTH Ó 2007 Elsevier Ltd.
VOLUME 281 ISSN 0573-8555/DOI 10.1016/S0573-8555(07)81003-0 All rights reserved.

39
40 Allen C. Kelley and Robert M. Schmidt
predicted coefficient values of 1 (for Ngr, population rate of growth) and
þ1 (for WAgr, working-age population rate of growth) when used to
expose population’s impact on Y/Ngr relative to Y/WAgr.
In Chapter 2 of this volume we examine the theoretical underpinnings
of the Harvard translations model and argue that, properly formulated,
its demographic terms would not be expected to have any economic-
growth impacts beyond translations accounting. Various estimates have
corroborated this interpretation. This outcome encouraged us to ‘‘enrich’’
the demographic specification. Specifically, we added a set of four
variables—youth and aged dependency (D1 and D2), density (Density),
and population size (N)—for an expanded demographic structure.
This ‘‘enriched demography’’ is then combined with the ‘‘translations
demography’’ (WAgr and Ngr) into what we term a ‘‘synthesis’’ formula-
tion. We fitted that model with Barro’s core economic and political
variables using a cross-section of 86 developed and developing coun-
tries over the period 1960–95. The results revealed the usefulness of
the synthesis framework by exposing the impacts of observed demo-
graphic change on changes in per capita GDP growth rates over the
period. We found that several aspects of demographic change influenced
economic growth, that the impacts of dependency (mainly youth) exceeded
those of the ‘‘translations’’ specifications, and that the relative and absolute
magnitudes of these separate influences varied notably over time.
The present chapter is divided into two sections. In Section 3.1,
‘‘Modeling population in development,’’ we further assess and compare
the impacts of the ‘‘translations’’ and ‘‘enriched’’ demographic structures
on growth. We begin by summarizing the theoretical structure. We next
extend the empirical period of evaluation beyond the past (1950–2000) to
include the future (2000–50). In addition, we examine regional effects.
This detail nicely exposes the dynamics of economic–demographic inter-
actions over the demographic transition. Our central finding is that the
impacts of population over the full century do not result primarily from
‘‘translations effects’’ (Ngr and WAgr), but rather from the ‘‘enriched’’
demographic structure variables—mainly from youth-dependency rates.
Section 3.2, A return to basics: Saving, investment, and capital flows,
builds upon this finding by asking the question, what are the linkages
between youth and aged dependency that would lead to sizable impacts
on economic growth in the convergence-modeling framework? These
two variables played prominent roles in the early population debates
(a la Coale and Hoover 1958), which emphasized the impacts of depen-
dency on saving and investment but have been ‘‘behind the scenes,’’
as it were, in recent convergence studies of population’s impact on
economic growth. Accordingly, using our data set, we undertake an
 The Asian Experience in Regional and Temporal Perspective 41
exploratory empirical analysis of demographic impacts on saving and
investment.2 Some intriguing results emerge. First, both youth- and
aged-dependency impacts are not only statistically significant, but are
also exceptionally large in size. Moreover, and most surprisingly, aged-
dependency impacts dominate youth-dependency impacts in the recent
past—even for developing countries—and this relationship is predicted to
continue and increase in importance at least until 2050. These patterns are
strongest for East Asia, but they are increasingly prevalent in other regions
as well.

3.1. Modeling population in development

This section of the chapter applies the methodology developed in Chapter 2

to assess the impacts of a century of demographic change on economic
growth. We first summarize the theory of the economic growth model;
then review the empirics from our estimation; and finally estimate the
impacts of demographic change on changes in per capita GDP growth.

3.1.1. Population in development: The methodology

An initial taxonomy for organizing an assessment of demographic change
is presented in equation (3.1),
Y=Lgr ðLgr  Ngr Þ;
ðY=NÞgr  |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} þ |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} ð3:1Þ
Productivity Component Translations Component
which decomposes per capita output growth (Y/Ngr) into two components:
an economic component (per-hour labor productivity growth, Y/Lgr), and
an accounting or translations component (Lgr  Ngr), which converts
output growth per labor hour into output growth per person (a possible
welfare measure).
This definition explicitly acknowledges the fact that per capita GDP
growth rates will deviate from per-worker growth rates whenever the
number of workers grows at a rate different from that of the population
as a whole. Many growth models focus exclusively on the economic
component, implicitly or explicitly assuming the equivalence of the rates
(Lgr vs. Ngr and, therefore, Y/Lgr vs. Y/Ngr). In some cases, these

2
Although there have been several studies of these relationships (Higgins 1998; Taylor
and Williamson 1994; Taylor 1995), we wish to evaluate such relationships with a country
data set, where they have been shown to be relevant.
Random documents with unrelated
content Scribd suggests to you:
back
back
back
back
back
back
back
back
back
back
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com