Personnel Economics in Imperfect

Labour Markets
This page intentionally left blank
Personnel
Economics in
Imperfect Labour
Markets

Pietro Garibaldi

1
3
Great Clarendon Street, Oxford OX2 6DP
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide in
Oxford New York
Auckland Cape Town Dar es Salaam Hong Kong Karachi
Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto
With offices in
Argentina Austria Brazil Chile Czech Republic France Greece
Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam
Oxford is a registered trade mark of Oxford University Press
in the UK and in certain other countries
Published in the United States
by Oxford University Press Inc., New York
© Pietro Garibaldi, 2006
The moral rights of the author have been asserted
Database right Oxford University Press (maker)
First published 2006
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,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above
You must not circulate this book in any other binding or cover
and you must impose the same condition on any acquirer
British Library Cataloguing in Publication Data
Data available
Library of Congress Cataloging in Publication Data
Data available
Typeset by Newgen Imaging Systems (P) Ltd., Chennai, India
Printed in Great Britain
on acid-free paper by
Biddles Ltd., King’s Lynn, Norfolk

ISBN 0–19–928066–5 978–0–19–928066–7

ISBN 0–19–928067–3 978–0–19–928067–4

10 9 8 7 6 5 4 3 2 1
to Giulia Tommaso and Alessandra
This page intentionally left blank
b P R E FAC E

THE AIM OF THE BOOK

Personnel economics, the use of economics for studying human resource issues,
is becoming a standard course in business and economics departments around
the world. Indeed, after being successfully introduced in North American busi-
ness schools by the pioneer books of Lazear (1995, 1998), the teaching of
personnel economics is now growing in Europe and in the rest of the world.
Yet, most of the traditional analysis of personnel economics assumes per-
fectly competitive labour markets, a situation in which wages are fully flexible
and dismissals can take place at no cost. Such a setting is hardly appropriate
for most European markets, where wage rigidity and wage compression are
widespread phenomena, and where employment protection legislation is very
stringent. Personnel Economics in imperfect labour markets aims at describ-
ing key personnel issues when firms and human resource managers act in
highly regulated labour markets. Few standard questions are easily identified.
Should hiring take place under temporary or permanent contracts? How can
we provide compensation related incentives when minimum wages are bind-
ing? How do we solve the employment/hours trade-off? What is the optimal
workers’ separation policy when employment protection legislation is very
stringent? Should general training be sponsored by the firm when wage com-
pression is widespread? While the underlying methods are the standard in the
field (i.e. the use of microeconomics for studying and solving human resource
problems) some of the questions and most of the answers provided by the
book are fairly original, and have so far been left to technical and special-
ized academic journals, not readily available to most students and to a broad
public.

POSITION OF THE BOOK IN THE LITERATURE

Most labour economics textbooks devote one chapter to personnel econom-
ics. Borjas (2004), Ehrenberg and Smith (2003), McConnell et al.(1999), Elliot
(1999) are just a few examples. In these chapters, the basic issues in personnel
economics are just briefly touched upon, and there is no attempt to see how
the standard issues change when labour markets are not competitive. Yet, there
are two books entirely devoted to personnel economics, and they have both
been written by Lazear. The 1995 MIT book (Personnel Economics) is a good
reference for introducing the educated scholar to the field. It is quite broad and
it is not too technical. Conversely, the 1998 Wiley book (Personnel Economics
viii PREFACE

for Managers) is thought out and written to be accessible by students of North

American business schools. While these books represent the obvious compar-
ison for personnel economics in imperfect labour markets, many chapters and
issues covered in the new book are totally absent in the two Lazear textbooks.
The reason for such a difference is that the new chapters refer to issues which
are mostly relevant for regulated labour markets, a standard situation outside
North America.

AUDIENCE
Students are the primary audience for the book. Specifically, the book is written
with three types of student in mind. First, all the material covered in the
book can be easily taught at the undergraduate level in programmes that
specialize in economics and business. Second, the book can be used in business
schools where human resource-oriented courses are taught by economists.
Third, labour economics courses taught in Europe will find many chapters
in the book a very useful reference, and can use it as a key reference in an
otherwise standard labour economics course. In addition, scholars in the field
are likely to find the book a useful reference text, and may wish to have it in
their personal library.

PREREQUISITES AND TECHNICAL LEVEL

In light of the audience in mind, the technical level required by the book is
supposed to be very modest. Ideally, readers should have taken an introductory
course in microeconomics, a semester of calculus, and an introductory course
in statistics. In practice, it will be possible to read the book even when these
prerequisites are partially or totally unfulfilled. The viability of the latter option
will rest on the various numerical examples presented in the book. In such
simple examples, all the main arguments are presented and the main results
outlined, even though they lack the rigour and the generality that the use of
algebra allows.

CASE STUDIES
An important feature of the book is the analysis of a few key case studies in
personnel economics. Such case studies rely on personnel data and analyse
key issues with an econometric perspective. They are a distinctive feature of
the book, and they should be analysed by all audiences, even though some
basic econometrics may be required. A first case study analyses the change in
overtime regulation in France in 1982. A second study refers to the introduction
of a performance pay plan in a window glass installer in the United States. A
third study analyses the role of the temporary help industry in the provision
 PREFACE ix

of training. A fourth study analyses relative performance pay in the context of

broiler production in the United States. A final study looks at the benefits of
team production in terms of firm productivity.

COURSE ORGANIZATION
There are fourteen chapters in the book. Each chapter is written with a lecture
topic in mind. Most of the chapters can be easily taught in a two-hour meeting,
even though some chapters require marginally more time. The (marginally)
more technical material is left to the appendices, so that the interested reader
can find more details if needed. A thirty-hour course, with fifteen meetings,
can easily cover most chapters of the book and leave some hours for setting
and solving problems.

ACKNOWLEDGEMENTS
The idea of this book started in the late 1990s at one of the first editions of
the European Summer Symposium in Labour Economics, a yearly CEPR con-
ference held in Amersee, Germany. Ed Lazear was one of the keynote speakers
and gave a lecture on the future and the challenges of personnel economics,
as both a research field and a teaching course. I was fascinated by the lecture,
and immediately realized that teaching ‘labour economics inside the firm’ or
‘personnel economics’ was an obvious opportunity for Bocconi, the university
that hired me back in 1999. Tito Boeri took me seriously, and immediately
gave me the opportunity to teach personnel economics in Bocconi University.
Personnel economics is now a mandatory course in the Economics and Man-
agement degree inside Bocconi. Another course has been recently introduced
at a graduate level. Without Tito, this project would have never seen the light.
I am highly indebted to him.
Teaching personnel economics to future European managers was immedi-
ately fascinating. In the first years I realized that a proper European course
had to seriously consider the institutional dimensions of the labour mar-
ket. In those years most of my research output was linked to the interplay
between labour demand and labour market institutions, notably employment
protection legislation. I realized that a proper European version of personnel
economics had to take this institutional dimension into consideration. I dis-
cussed the issue with Ed Lazear again in 2002 in Amersee. He encouraged me
to go forward with the project, and I am certainly indebted to him.
I am also indebted to my various research co-authors, since many ideas
brought forward in the various chapters have benefited from their insights,
notably Gianluca Violante, Giuseppe Bertola, Lia Pacelli, and Etienne Wasmer.
Michele Pellizzari, Marco Leonardi and Lorenzo Cappellari taught preliminary
chapters, and offered important comments.
x PREFACE

The role of students has obviously been crucial. Various versions of this
course were taught in Bocconi University, at the University of Turin, at the
State University of Milan, and at Brussels University. All these students saw
various drafts of the chapters, and accepted the inevitable mistakes that earlier
drafts incorporate. Paola Monti stands out among these students.
b CO N TE N TS

LIST OF FIGURES xvi

LIST OF TABLES xviii

1 Personnel Economics and Non-Competitive Labour Markets 1

1.1 Introduction 1
1.2 The Job in an Imperfect Labour Market 2
1.2.1 Perfect Labour Market 3
1.2.2 Imperfect Labour Market with Wage Set by the Firms 4
1.2.3 Imperfect Labour Market with Surplus Splitting 4
1.2.4 Imperfect Labour Market with Exogenous Wage 5
1.3 Minimum Wage Constraints and Union Density 5
1.4 Employment Protection Legislation 7

2 The Optimal Skill Ratio 11

2.1 Introduction 11
2.2 A General View on Personnel Policies
at the Hiring Stage 12
2.3 The Optimal Skill Ratio: Key Concepts 12
2.3.1 Skilled and Unskilled Labour Interdependent
(Interior Solution) 15
2.3.2 Skilled and Unskilled Labour Independent
(Corner Solution) 19
2.3.3 The Effects of Wage Compression 21
2.3.4 Independent Skilled and Unskilled Labour
Interacting with Costly Capital 22
Appendix 2.1. A Formal Expression of the General Problem 24

3 The Hours–Employment Trade-Off 27

3.1 Introduction 27
3.2 A Formal Classification of Labour Costs 27
3.2.1 Non-recurring Fixed Costs 28
3.2.2 Recurring Fixed Costs 29
3.2.3 Variable Costs 29
xii CONTENTS

3.3 A Theory of the Hours–Employment Trade-Off 29

3.3.1 The Isocost and the Total Labour Costs 29
3.3.2 A Formal solution to the Employment–Hours Trade-Off 33
3.4 The Overtime Premium and the Effects of Reducing the Workweek 34
3.4.1 Firm Position with the Overtime Premium 34
3.4.2 The Effect of a Reduction in the Normal Workweek 35
3.5 Reducing the Workweek: A Natural Experiment from
France 36
3.6 Theoretical Predictions and a Natural Experiment 38
3.7 The Evidence 39
Appendix 3.1. A More Formal Version of the Hours–Employment
Trade-off 41

4 Temporary or Permanent? 45

4.1 Introduction 45
4.2 Permanent Contracts with Fixed Wages 46
4.2.1 A Numerical Example 47
4.2.2 Formal Derivation 49
4.3 Temporary Contracts as a Buffer Stock 52
4.4 The Buffer Stock Model with Interim Costs 53
4.5 Temporary Contracts with Costly Turnover: Costs and
Benefits 55
4.6 A Brief Look at the Evidence 58
Appendix 4.1. The Formal Derivation of the Buffer Stock Model 60
Appendix 4.2. The Formal Derivation of the Buffer Stock Model
with Costly Interim Agency 60

5 Managing Adverse Selection in Recruiting 62

5.1 Introduction 62
5.2 Contingent Contracts 63
5.2.1 Piece Rate as a Contingent Contract 64
5.2.2 Temporary Contract with Probation Wage 67
5.2.3 Contingent Contracts with Minimum Wages 72
5.3 Use of Credentials and the Signalling Model of Education 74
5.4 Education Signalling 74
5.5 Education Signalling: A Formal Story 78
5.6 Rules of the Game 78
5.7 No Education Equilibrium (Pooling Equilibrium) 79
5.8 Education Equilibrium (Separating Equilibrium) 80
 CONTENTS xiii

6 Optimal Compensation Schemes: Foundation 82

6.1 Introduction 82
6.2 A Formal Principal–Agent Model 84
6.2.1 The General Setting 84
6.2.2 The Problem of a Risk-Neutral Agent 89
6.2.3 The Problem of the Principal with a Risk-Neutral Agent 92
6.3 The Basic Scheme with Non-Negative Payments 98
6.4 The Basic Scheme when the Agent is Risk Averse 100
6.4.1 The Agent Preferences 100
6.4.2 The Problems of the Agent 101
6.4.3 The Problem of the Principal with a Risk-Averse Agent: The Optimal
Bonus Scheme 102
Appendix 6.1. What if the Principal and the Risk-averse Agent
Could Contract on Effort 104

7 Pay for Performance with Wage Constraints 107

7.1 Introduction 107

7.2 Heterogeneous Ability: The Set-Up 108
7.3 Fixed Wage with Minimum Output 111
7.4 Pure Rent: Fixed Wage with Minimum
Output and Heterogeneous Ability 111
7.4.1 A Numerical Example 114
7.5 Performance Pay with a Two-Tier Wage System 115
7.5.1 The Formal Solution to the Two-Tier Wage System 117
7.5.2 Two-Tier System: An Example 120
7.6 Pay for Performance at Safelite 121
7.6.1 Theoretical Predictions 123
7.6.2 Basic Statistics 123
7.6.3 Basic Results 124
7.6.4 Possible Interpretations 124
7.7 Econometrics 126
7.7.1 Other Issues 127
Appendix 7.1. The Basic Franchising Scheme With the Ability
Parameter 128
Appendix 7.2. An Optimal Scheme With a Fixed Wage 129

8 Relative Compensation and Efficiency Wage 132

8.1 Introduction 132

8.2 Tournament: A Formal Story 133
8.2.1 Worker Problem 135
8.2.2 Firm Problem 136
xiv CONTENTS

8.2.3 Obtaining the Wages 137

8.2.4 Heterogeneity and Risky Strategies 138
8.3 Linear Performance Evaluation 139
8.3.1 Indirect Evidence of Relative Compensation 140
8.4 Evidence on Tournaments in Broiler Production 140
8.5 Broiler Production 141
8.6 Efficiency Wage 146
8.7 Deferred Compensation and Upward-Sloping Wage
Profile 149

9 Training and Human Capital Investment 152

9.1 Introduction 152

9.2 The Basic Human Capital Model of Schooling 153
9.3 General versus Specific Investment 156
9.4 General Training 156
9.4.1 The Basic Set-up 156
9.4.2 The Efficient Level of Training 157
9.4.3 The Solution of the Model 158
9.4.4 Investment in General Training: An Example 159
9.5 Firm-Specific Training 160
9.5.1 Surplus and Specific Rent 162
9.5.2 Rent Sharing 163
9.5.3 Firm-Specific Human Capital: An Example 166
9.6 Some Evidence on Training 166
Appendix 9.1. General Training 167
Appendix 9.2. Firm-Specific Training 169

10 Training Investment in Imperfect Labour Markets 171

10.1 Introduction 171

10.2 Firm-Sponsored Training: The Baseline Case 172
10.2.1 The Set-up 172
10.2.2 The Solution 174
10.3 A More General Framework 176
10.4 Firm-Sponsored Training in Temporary Help Firms 181
10.4.1 The Evidence: Wages and Training in the THS Industry 183
10.4.2 A Possible Story for the THS Evidence 185

11 Job Destruction 187

11.1 Introduction 187

 CONTENTS xv

11.2 Firm-Initiated Separations and Wage Cuts in Imperfect Labour

Markets 188
11.2.1 Job Destruction: An Example 191
11.2.2 Job-Preserving Wage Cuts with Asymmetric Information 192
11.3 Labour Hoarding and the Effect of Firing Taxes 193
11.4 Job Creation, Job Destruction, and Labour Hoarding 195
Appendix 11.1 Present Discounted Values and Multi-period
Jobs 200

12 Further Issues in Employment Protection Legislation 202

12.1 Introduction 202

12.2 EPL Taxes and Transfers 203
12.2.1 Calculating Taxes and Transfers 204
12.3 Outsiders, Insiders, and Tax and Transfers 207
12.3.1 Deriving the Wages with Severance Payments 208
12.3.2 Deriving the Wages with the Firing Tax 210
12.4 Threshold Effects 212
12.4.1 The Set-up of the Model 213
12.4.2 The Efficient Allocation 213
12.4.3 The Rigid System 214
12.4.4 The Role of Threshold Effects 215

13 Teams and Group Incentives 218

13.1 Introduction 218

13.2 The Team Production Problem 219
13.3 Team Norms as Remedies to the Team Production Problem 221
13.4 Teams in Reality 222
13.5 A Case Study: Production at Koret 223
13.5.1 The Impact of Teams on Productivity at Koret 225
13.5.2 Who Joins a Team? 227
13.5.3 How Does a Team Composition Affect Team
Productivity? 230
13.5.4 Discussion and Conclusion 230

Appendix A. Labour Demand at the Firm Level 232

Appendix B. Constrained Optimization 238
References 242
Index 247
b L IS T OF F I G U RES

2.1 The Isocost 14

2.2 The solution to the optimal skill ratio with interdependent production 16
2.3 The effect of an increase in γ on the optimal skill ratio 18
2.4 Optimal skill ratio with corner solution 20
3.1 Classification of labour costs 28
3.2 The solution to the hours–employment trade-off 32
3.3 The effect of the hours–employment trade-off of an increase in the size 32
of the budget (left panel) and an increase in the fixed cost (right panel)
3.4 Equilibrium position for a firm that uses and does not use overtime 35
3.5 The reduction of the normal workweek and the isocost 35
3.6 The effects of reduction in the normal workweek 36
4.1 Labour demand in good and bad times 48
4.2 Profits in good and bad times 48
4.3 Buffer stock model with costly interim workers 54
4.4 Labour hoarding of permanent workers with costly interim 55
5.1 Piece rate wage for solving adverse selection in recruitment 66
5.2 Education signalling and a separating equilibrium 77
5.3 Pooling equilibrium (education does not act as a signal) 78
6.1 Different compensation schemes as workers’ constraint 86
6.2 Worker’s indifference curve 88
6.3 The worker problem 90
6.4 Firm profits for different values of the bonus scheme 96
6.5 The optimal contract 97
7.1 Indifference curves with heterogeneous ability 110
7.2 Optimal solution with fixed wage and minimum output when ability is 112
heterogeneous
7.3 Individual self-selection for working at the firm that pays w, xmin 113
7.4 Bonus scheme with minimum wage guarantee (and minimum output 116
required)
7.5 Individual optimization with a two-tier system 117
 LIST OF FIGURES xvii

7.6 Utility levels for different individuals when a bonus scheme is 119
introduced in a firm that was offering a fixed wage with minimum
output
8.1 Broiler production 141
8.2 Variability of grower performance and grower quality: tournaments 144
8.3 Variability of grower performance and grower quality: LPE contracts 145
8.4 Deferred wages 150
9.1 A two-period model on general training 157
9.2 Wage tenure profile with general training 159
9.3 Specific investment with ex post sharing 164
9.4 Wage tenure profile with firm-specific investment 165
10.1 A two period model of firm-sponsored general training 173
10.2 Marginal benefit and marginal cost of training to the firm with and 179
without wage compression
10.3 Worker financing of general training with imperfect labor markets 180
10.4 Firm-sponsored training with a minimum wage 181
11.1 Firm-initiated separation 191
11.2 Reservation job and firing costs 195
11.3 The timing of the events 196
11.4 Determination of job creation and job destruction 197
11.5 The effects of an increase in firing costs on job creation and job 200
destruction
12.1 Labor demand in good and bad times and threshold effects 217
13.1 Average Productivity and percentage of workers employed in teams 225
A.1 Production function with constant capital 234
A.2 The marginal product and labor demand 235
B.1 The Lagrange method 238
b L IS T OF TA B L E S

1.1 Union density and wage-setting institutions in different countries 7

1.2 Strictness of employment protection legislation in different countries 10
2.1 Wage productivity premia in different countries 21
2.2 Optimal skill ratio with independent workers and costly capital 24
3.1 France: hours worked 1976–1981 37
3.2 Employment losses, by hours, in France 39
3.3 Regressions on employment losses 41
4.1 Flexible and rigid regimes 49
4.2 Profits in various regimes 50
4.3 Incidence of temporary employment in OECD countries 58
4.4 Status of temporary workers in t and t + 1 59
6.1 Profits and utility with different bonus schemes 95
7.1 Data description for the Safelite case 122
7.2 Basic result for the Safelite case 122
7.3 Regression results for Safelite cases 125
8.1 Regression results for broiler production 146
9.1 General and specific training 160
10.1 Skills training: prevalence and policies at US temporary help agencies 182
10.2 Comparison of log hourly wages of worker at training and non-training 184
establishments
10.3 Estimates of the relationship between establishment training policies 185
and worker wages
11.1 Job destruction and wage cuts 192
12.1 Tax and transfer components of firing costs in Italy 205
13.1 Summary statistics for individual and team incentives at Koret 224
13.2 Dates of team formation and average weekly productivity before and 226
after team formation
13.3 Relationship between teams and productivity 228
13.4 The impact of future team participation 229
13.5 The Impact of teams on hourly and weekly pay 230
A.1 Labour demand at the firm level 233
1 Personnel Economics and
Non-Competitive Labour
Markets

1.1 Introduction
Personnel economics analyses personnel issues inside the firm. Yet, each firm
operates inside a labour market, and the functioning of such a market has
important consequences for personnel economics. The view of personnel eco-
nomics analysed in this book is based on two key properties of the labour
markets:
• labour markets are imperfect and jobs are associated to rents;
• labour market institutions interact with personnel policies. Notably,

– wages are partly set outside the firm–worker pair (minimum wages and
collective agreements are widespread)
– job termination policies are affected by a sizeable and binding employ-
ment protection legislation.
In an imperfect labour market the traditional competitive view of the labour
market does not hold. In a competitive labour market, the wage is fully set by
market forces and firm–worker pairs have no control over its value. In addition,
in a perfect labour market the workers are just indifferent between working
for the firm or working for somebody else. The resulting equilibrium wage is
fully flexible, and whenever business conditions change, wages can be adjusted
in response.
Things are different when the labour markets are imperfect. When the
labour markets are imperfect, a job is associated with a market rent. This
means that a job brings some pure surplus to both the firm and the worker. In
other words, the firm and the worker are better off when they are together than
when they are separated, and if a given job were to be suddenly severed, both
parties would lose. The surplus can be thought of as some extra value that the
parties enjoy when they are together. Such extra value has to be split between
the firm and the employees. This brings the issue of wage determination, which
is one of the key topics in personnel policies. In the book we explore three basic
possibilities for wage determination: wage setting by the firm, wage bargaining
2 NON-COMPETITIVE LABOUR MARKETS

by the firm and the worker, and wage determination made outside the firm and
the worker, i.e. by collective agreements. The book uses these three possibilities
interchangeably.
In an imperfect labour market wages are also rigid and do not fully respond
to business fluctuations. This implies that firms have to use different instru-
ments (other than wages) to respond to changes in business fluctuations. And
when prices cannot easily move firms have to rely on quantities. The problem
is that in an imperfect labour market, interrupting an employment relation-
ship is far from straightforward. This is the case because there is pervasive and
abiding employment protection legislation.
The next section establishes the concepts introduced in this chapter in a
more formal setting, while sections 1.2 and 1.3 briefly discuss the nature of
labour market institutions introduced in these paragraphs.

1.2 The Job in an Imperfect Labour Market

The Job is defined as a firm–worker match engaged in a production setting. We
imagine that a firm and a worker have been brought together in the past. The
focus here is on a single job rather than firms.
The Value of Production is y. y indicates the value of the labour product
obtained when the firm and the worker engage in production. Let’s say that y is
measured in euros. One can think of y as the revenues from the job, so that
y is the product of a quantity of output produced times the price of output.
y corresponds also to the value of the marginal labour product to the worker.
The worker’s outside option is v. v can be thought of as the wage that the
worker would get if he or she had to leave the current job and find a position
elsewhere. It can also represent the alternative value of leisure. It is not essential
to emphasize which of the two interpretations is used, but in any case we
measure its value in euros.
The wage that the worker receives for the current assignment is w.
The firm’s marginal profit from the job is the difference between the revenue
from the job and its cost. The focus here is just on the labour cost associated
with this job, and we assume that there are no other specific costs. We can call
the profits from the job the firm surplus from being with the worker and we
shall indicate it by Sπ

Sπ = y − w

The wage w is measured in the same unit of y.

The worker surplus from being in the current job can be similarly defined
as the difference between the wage obtained in the current relationship and
 NON-COMPETITIVE LABOUR MARKETS 3

the outside option. In formula this reads

Sw = w − v
Each part is interested in its own surplus. Yet, since we are measuring each
surplus in euros, we can simply sum up the two marginal surpluses so as to
obtain an expression for the total surplus S. In formula, the total surplus from
the job is defined as
S = Sπ + Sw
S =y −v
or as the different between the marginal revenue product y and the outside
option v. In the definition of the surplus, the wage does not enter. This should
not be surprising, since the wage enters positively in the worker’s surplus and
negatively in the firm’s profit. One extra unit of wage reduces firm profits by
1 euro and increases the worker’s surplus by the same euro. This suggests that
the wage is a way to transfer utility from one party to the other. Since it acts as
a pure transfer, it does not enter into determination of the total surplus, which
is an expression for the pure value associated with a given position.

1.2.1 PERFECT LABOUR MARKET

In a perfect labour market there is no surplus from the job. This implies that
when the labour market is competitive and market forces operate without any
obstacle, the total surplus from the job is zero and
S=0
y=v
The worker is paid exactly the value of the marginal product
w=y
The worker is indifferent between working at the firm and its best alternative
w=v
Let’s discuss the consequences of these results. The firm–worker pair generates
a value of the labour product that is equal to y. In a competitive market such
value y is identical to the wage, so that we can say that the worker is the full
residual claimant of the marginal product generated by the job. But the worker
does not enjoy any pure surplus, since the wage turns out to be also equal
to its alternative use of time, which we formally indicate with v. Further, the
firm also makes zero profits, since the marginal revenue product is equal to
the wage.
4 NON-COMPETITIVE LABOUR MARKETS

1.2.2 IMPERFECT LABOUR MARKET WITH WAGE SET BY THE FIRMS

In an imperfect labour market there is surplus from the job

S>0
S =y −v
y>v

An alternative way to express the same concept is to say that in an imperfect

labour there are rents. The working of competition is not sufficient to eliminate
these pure surpluses, and firm–worker pairs enjoy such rents.
If the labour market is imperfect and the firm can freely set the wage, it will
set the wage so as to maximize profits. The firm has an interest in setting a wage
that is as low as possible, but it has to take into account the worker’s outside
option. In other words, the worker has a participation constraint which can be
written as

w≥v

which is a condition that implies that the worker will accept working for the
firm as long as the wage is larger than his or her outside option. In formula,
the firm problem is the following

Maxw Sπ = y − w
s.t . w ≥ v

This implies that the wage chosen in this case will be

w=v

and the firm will make positive profits equal to

Sπ = y − v

1.2.3 IMPERFECT LABOUR MARKET WITH SURPLUS SPLITTING

As we mentioned above, in an imperfect labour market there is surplus from
the job

S>0
y>v
 NON-COMPETITIVE LABOUR MARKETS 5

We now assume that the surplus is split between the firm and the worker,
and the worker gets a fraction β of the total surplus
wβ = v + βS
wβ = v + β(y − v)
wβ = (1 − β)v + βy
Wages are obtained as the weighted average between the worker’s outside
option and the marginal productivity on the job. The weight corresponds
to the worker’s bargaining share β.
The firm makes profits
Sπ = y − wβ
Sπ = y − v + β(y − v)
Sπ = (1 − β)(y − v)
and the worker enjoys a surplus vis-à-vis his or her outside option, since
Sw = β(y − v).

1.2.4 IMPERFECT LABOUR MARKET WITH EXOGENOUS WAGE

In an imperfect labour market there is surplus from the job
S>0
y>v
Assume that the wage is set outside the firm worker pair and equal to w̄.
The worker will accept the job if
Sw = w̄ − v ≥ 0
The firm will operate if the profits are non-negative
Sπ = y − w̄ ≥ 0

1.3 Minimum Wage Constraints and Union Density

The minimum wage is a labour market institution that sets a lower bound to
the wage paid to individual workers. While the conceptual definition is very
simple, and most countries in the world have some form of minimum wage,
the scale, eligibility, and operation details vary from country to country, so that
6 NON-COMPETITIVE LABOUR MARKETS

it is very difficult to provide a unique and comparable cross-country definition

of the minimum wage.
Some countries opt for a single national minimum rate, which can be set
on an hourly, daily, weekly, or monthly basis. Beyond the single national wage
there is often a reduced or sub-minimum rate for some groups of workers, and
notably the young. Often, sub-minimum rates do not exist de jure, but they do
exist de facto, since special employment programmes allow employers to pay
lower wages to youth workers or lower their social security charges.
In some countries, there exist minimum wage premia related to vari-
ous workers’ characteristics. For example, the minimum wage may rise with
workers’ experience, workers’ qualification, and family status. The minimum
wage may or may not be indexed to price inflation, even though such indexation
schemes exist now in a very few OECD countries.
The setting of the minimum wage also changes according to two main
criteria. In some countries, the minimum wage is unilaterally set by the gov-
ernment while in other countries it is part of a negotiation with workers and
firm representatives.
Despite the various differences that we pointed out, one can still try to
compare minimum wages across countries by measuring their value relative
to some measure of average wage. Such a ratio (often referred to as the Kaitz
index) depends on how both the numerator (the minimum wage) and the
denominator (the average wage) are measured. In principle, using median
rather than average wage provides a better measure of the denominator, since
it also takes into consideration the dispersion of earnings. Special attention
should also be paid to using the appropriate earning measure, which should
exclude any overtime and bonus payment. The OECD (1998) has compiled
indices of minimum wages for most OECD countries. High minimum wages
are considered those for which the Kaitz index is more than 60 per cent as in
France and Belgium, while it is considered low in countries in which the index
is around 30 per cent (Spain, Czech Republic).
In an imperfect labour market, trade unions play an important role in wage
determination. There are various ways to assess the importance of trade unions
in the labour market. The first way is to look at union density, i.e. the fraction
of workers registered to a trade union. Table 1.1 suggests that union density
in Europe is much larger than in the USA: 44 per cent of European workers
are registered to a trade union while in the United States the same figure is
14 per cent.
In some continental European countries trade union density is not very high
and, perhaps surprisingly, it appears larger in the United States than in France.
How is this possible when most people would think of France as the realm
of the general strike? In order to correctly estimate the importance of trade
unions it is necessary to consider the degree of trade union coverage, i.e. the
percentage of workers whose salary has been negotiated by a trade union. While
 NON-COMPETITIVE LABOUR MARKETS 7

Table 1.1. Union density and wage-setting institutions in

different countries

Densitya Coverageb Centralizationc

France 9.1 95.0 2.0

Germany 29.0 92.0 2.0
Spain 21.1 78.0 2.0
Italy 23.7 82.0 2.0
United States 14.3 18.0 1.0
Japan 24.0 21.0 1.0
European Union 43.1 82.3 1.9

a Percentage of workers belonging to a trade union.

b Percentage of workers whose wage is negotiated by a union.
c Degree of centralization in bargaining. 1 at firm level, 3 central.
Source: OECD 1999.

in the United States the difference between the trade union density and the
degree of coverage is not very important, for countries like France, Germany,
and Italy, the degree of trade union coverage is around 90 per cent. This means
that in Europe nine workers out of ten have their wage negotiated by trade
unions, even though only four out of ten are union members. Table 1.1 also
collects information on the degree of centralization of collective bargaining,
i.e. the level at which contracts are negotiated. While in the United States the
negotiation is mostly at the level of the firm, in Europe it takes place mostly at
the sectoral level. From the labour market point of view, strong trade unions
can have adverse effects on the costs of labour, and can ultimately reduce
employment. However, trade unions may also play a key role in society and
boost workers to improve workers’ morale and productivity.

1.4 Employment Protection Legislation

Employment Protection Legislation (EPL) is one of the most important insti-
tutions of the labour market. It refers to the set of norms and procedures to be
followed in cases of dismissal of redundant workers. In almost every country,
EPL forces employers to transfer to the worker a monetary compensation in
the case of early termination of a permanent employment contract. Moreover,
complex procedures have to be followed in the case of both individual and
collective lay-offs. Finally, in some countries, the final decision on the legitim-
acy of a lay-off depends on a court ruling. From the viewpoint of economic
analysis, it is very important to note that the firing decision is not only up to
the worker and/or the employer, but can involve the participation of a court,
which can be asked to assess the legal validity of the lay-off.
8 NON-COMPETITIVE LABOUR MARKETS

The most traditional dimensions of EPL are the severance payments and
advance notice. Severance payments refer to a monetary transfer from the firm
to the worker to be paid in the case of firm-initiated separation. Advance notice
refers to a specific period of time to be given to the worker before a firing can
actually be implemented. Note that the severance payment and advance notice
that are part of EPL refer to the minimum statutory payments and mandatory
rules that apply to all employment relationships, regardless of what is estab-
lished by labour contracts. Beyond mandatory payments, collective agreements
may well specify larger severance payments for firm-initiated separations. Such
party clauses, albeit important, are not considered in this report.
Another important dimension of EPL consists of the administrative pro-
cedures that have to be followed before the lay-off can actually take place. In
most countries, the employer is often required to discuss the lay-off decisions
with the workers’ representatives. Further, the legislative provisions often differ
depending on business characteristics such as firm (or plant) size and industry
of activity. As this simple introduction shows, it is obvious that the EPL is a
multidimensional phenomenon.
In most countries the legislation distinguishes between individual and
collective dismissals. Individual dismissals should be further distinguished
between economic dismissals and disciplinary dismissals, with most EPL
clauses applying only to the former case. Disciplinary dismissals (i.e. worker’s
fault dismissals), do not typically involve monetary transfers. The procedure
for collective dismissal applies to large-scale firm restructuring, and requires
the dismissal of at least a specific proportion of the workforce. When a collect-
ive dismissal is authorized by the relevant authority (often a court) the firm
can then implement a large dismissal without a large transfer. Yet, such proced-
ure requires a much tighter administrative burden, in the form of prolonged
consultation with the workers’ representatives.
From the viewpoint of economic analysis, the multidimensionality of the
EPL can be reduced to two components. The first component is simply a
monetary transfer from the employer to the worker, similar in nature to the
wage. The second, instead, is more similar to a tax, because it corresponds
to a payment to a third party, external to the worker–employer relationship.
Conceptually, the severance payment and the notice period correspond to the
transfer, while the trial costs (the fees for the lawyers, etc.) and all the other
procedural costs correspond to the tax (in Italy, the transfer part corresponds
to approximately 80 per cent of the total cost of the lay-off).
Unavoidably, the complexity of the firing regulations is costly for the
employer. It is quite difficult to quantify the total cost of a lay-off, mainly
because its exact amount depends also on the probability of the worker filing
the case to a court, and on the probability of the court invalidating the firm’s
firing. In the Italian case, in particular, if the firing decision is overruled by the
judge, the firm can be forced to take the employee back on the payroll. Despite
 NON-COMPETITIVE LABOUR MARKETS 9

the difficulty of carrying out precise calculations of the average cost of a lay-
off, a number of studies have tried to assign a reasonable value to such a cost.
Garibaldi and Violante (2006) estimate that an Italian employer with more
than fifteen employees who fires a worker and whose decision is overruled by
a court a year after the lay-off with 80 per cent probability will have to bear a
cost of fifteen months’ wages, i.e. a year and three months of wages.
From a cross-country perspective, it would be interesting to measure the
average cost of a lay-off in the various countries. Unfortunately, homogeneous
measures of such cost relative to the average wage do not exist for all countries.
In order to carry out international comparisons of the employment protec-
tion regimes, economists use the so-called method of the ‘hierarchies of the
hierarchies’. This method consists in assigning a number (say from 1 to 6) to
every country for any single feature of the protection regimes. Higher num-
bers denote more rigid regimes. Taking the average of the several components,
a single synthetic measure of the rigidity of the EPL is obtained.
The synthetic indicators, originally compiled by the OECD, are now avail-
able for transition economies as well as for Latin American countries. In order
to obtain the overall indicator of the rigidity of a country, it is necessary to
consider simultaneously (a) the rigidity of the individual firing regulation of
workers under permanent contract, (b) that of workers under temporary con-
tracts, and (c) the rigidity of collective dismissals. The average of these three
measures gives the overall indicator. Obviously, each of the three measures
is obtained, in turn, through an average of some other sub-measures. As an
example, consider the indicator for individual firings of workers under per-
manent contracts. Such a number is obtained as an average of 4 sub-indicators:
(1) the administrative procedures, (2) the length of the notice period, (3) the
amount of the severance payment (4) the severity of the enforcement (the
more or less important role of judges on firing disputes). Clearly, in order to
construct the synthetic indicators, it is necessary to consider a great number
of dimensions.
Table 1.2 displays the relative position of the six countries. The United
States turns out to be the most flexible country, while EPL is much tighter
in countries such as Italy, Spain, and Germany. The table also shows the EPL
rigidity indicators in the 1980s, thus allowing us to observe the evolution over
time of the firing regulations. As far as the ‘regular’ contracts are concerned, the
table clearly shows that the European countries have made very little change to
the firing regulations. As a matter of fact, the level of protection is practically
unchanged. Conversely, the regulation of temporary contracts has been eased
in most European countries.
These indicators are undoubtedly useful for the economic analysis, as they
are often used to empirically test the predictions of the economic models.
However, it is worth recalling the reasons why such indicators are not perfect.
First, it is very difficult to get time series of these indices (the latest OECD
10 NON-COMPETITIVE LABOUR MARKETS

Table 1.2. Strictness of employment protection legislation in different countries

Regular employment Temporary employment Overall index

1980s 1990s 1980s 1990s 1980s 1990s

France 2.3 2.3 3.1 3.6 2.7 3

Germany 2.7 2.8 3.8 2.3 3.2 2.5
Spain 3.9 2.6 3.5 3.5 3.7 3.1
Italy 2.8 2.8 5.4 3.8 4.1 3.3
Japan 2.7 2.7 ... 2.1 ... 2.4
United States 0.2 0.2 0.3 0.3 0.2 0.2
European Union 2.5 2.3 2.9 2.2 2.7 2.3

Note: Degree of centralization in bargaining. 1 at firm level, 3 central.

Source: OECD 1999.

numbers refer to three points in time). Second, the method of the ‘hierarch-
ies of hierarchies’ assigns the same weight to the various EPL sub-indicators.
An additional problem is the fact that they concern features of EPL that are
similar between countries, while they tend to ignore country-specific regula-
tions. Finally, it is most difficult to measure the degree of enforcement of the
norms. It is possible that some countries have rigid norms only softly enforced,
while in other apparently flexible countries the norms are enforced very strictly
(Boeri 1999).
2 The Optimal Skill Ratio

2.1 Introduction
Before entering into the details of personnel policies, the firm needs to have
a general view of the right balance of its workforce. Indeed, choosing the
right balance of the workforce is a key pre-requisite for sound personnel man-
agement. There are two key dimensions that should be considered: the skill
composition of the labour force and the hours/employment trade-off.
Labour is not a homogeneous quantity, and a highly skilled worker is a
very different factor of production from an unskilled worker. Highly skilled
workers are likely to be more profitable to the firm. But they are also likely to
be more costly. A trade-off emerges, and the firm has to find the right balance
between cost and productivity. In personnel economics, solving this problem
means finding the optimal skill ratio. This chapter shows that the solution to
this trade-off requires the analysis of relative costs and relative productivity,
where the word relative refers to different skill level. In general, we will see
that the best skill composition is not necessarily the cheapest skill composition
nor the most productive composition. Technological considerations are also
very important. In some firms, only one type of labour will be chosen, while in
other firms it will always be optimal to have a positive quantity of both types of
workers. This chapter will also give us the possibility of studying the impact on
the skill composition of wage compression, an important institutional feature
of imperfect labour markets.
Hiring a quantity of labour, irrespective of its skill level, involves a vari-
ety of costs, and only some of such costs vary with the hours worked. As
a consequence, total labour costs are much larger than the per hour wage
cost. Further, some of these costs do not even vary with the number of hours
worked. Thus, in real life organization a firm can adjust the quantity of a
given type of labour over the hours worked and the number of employees.
This is the hours/employment trade-off and will be the focus of the next
chapter.
The chapter is organized as follows. Section 2.1 briefly describes the general
hiring problem of a firm. Section 2.2 defines some key concepts for describing
a firm’s hiring behaviour. Section 2.2.1 describes the optimal skill ratio in the
case of a technology in which workers are interdependent in the production
process while Section 2.2.2 analyses the case of independent workers.
12 THE OPTIMAL SKILL RATIO

2.2 A General View on Personnel Policies

at the Hiring Stage
A firm in general wants to maximize profits. Let’s imagine for simplicity that the
firm is a competitive firm so that the price that the firm charges for its product
is fixed at some level. In a short-run dimension, the stock of capital can also
be thought of as fixed, so that the scale of operation is given. The key decision
concerns the amount of labour to be hired. But labour is a heterogeneous
factor of production. The dimension of heterogeneity that is more relevant
to the firm is the skill heterogeneity. Some workers are more productive than
others. In this chapter we assume that there are only two types of workers:
skilled and unskilled. We can think of skilled workers as graduate students and
unskilled workers as individuals without a university degree. The firm must
decide how much of each type of labour to hire. Further, it has to decide how
many hours per week each type of worker should work. Basically, the firm
maximization decision can be broken into three subsidiary decisions.
1. The firm decides how to produce any given amount of output. In other
words, the firm must decide the skill composition of its workforce.
2. The firm decides how to combine hours and employment for any given
amount of labour. This implies solving the hours/employment trade-off.
3. The firm decides how much output to produce.
The key personnel policies at the hiring stage are the first two. In this chapter
we analyse the first one. In the next chapter, we analyse the second one. The
pricing and quantity decision of the firm problem is certainly a key managerial
decision, but is not the topic of personnel economics, while it is the key focus
of industrial economics.

2.3 The Optimal Skill Ratio: Key Concepts

As we argued above, in this chapter we only focus on the skill combination and
we ignore the distinction between hours and employment. This means that
the cost of labour we consider is simply the wage to be paid to each employee,
without any additional cost. The problem we consider is the following.

The Firm Problem

Let us suppose that we need to produce a given quantity of output Ȳ .

The firm can potentially use two types of labour for producing any amount of output Y ,
including the output Ȳ . The two types of labour are skilled workers and unskilled workers,
 THE OPTIMAL SKILL RATIO 13

which can be thought of as graduate workers and non-graduate workers. We indicate with
L the quantity of low-skilled workers and with H the quantity of high-skilled workers. The
contribution to output stemming from each separate input, holding constant the quantity
of other input, is called marginal productivity. It is a key concept in most economics as well as
in personnel economics. The marginal product of unskilled labour is defined as the change
in output resulting from hiring additional workers, holding constant the quantity of other
inputs. The formal expression is


Y


FL =
L
H =H̄ ,K =K̄

One additional unskilled worker yields an extra amount of output equal to FL where the
notation makes clear that we are holding fixed the amount of capital and of skilled labour.1
Similarly one

additional unit of skilled labour at constant quantity of unskilled labour yields
FH = H Y


. The only assumption that we make on these two marginal products
L=L̄,K =K̄
is that they are positive. This means that an additional unit of any type of labour (holding
constant the other type) increases total output. It is an obviously reasonable assumption,
since it just says that adding more workers to the firm increases output, other things
equal.
The salary of each type of worker depends on each skill level. The firm takes as given the
wage. The salary of a graduate worker is wH while the salary of an unskilled worker is wL .
The wage refers to the hourly labour costs. Obviously, wH ≥ wL . Skilled workers are more
expensive than unskilled workers.
The firm problem is to choose the skill combination that yields output Ȳ and minimizes
costs. In other words, the firm must select the combination of H and L that yields the lowest
possible labour costs, conditional on producing total output equal to Ȳ .
1 More formally, the marginal productivity of unskilled labour can also be thought of as the partial derivative of the

production function Y = F (K̄ , L, H ) with respect to unskilled labour and indicated with YL = ∂Y
∂L = FL .

Our analysis will show that two dimensions are very important. The first
one is the relative labour costs, or the ratio of the two wages. The other key
dimension will be the relative marginal productivity of the two factors. In
addition, careful consideration must be given to the way in which the two
types of labour enter the production function, and whether the two factors are
complements or substitutes in the production process.
The formal way to think about this problem is assuming that a firm
would like to minimize labour costs subject to the constraint that the quant-
ity of output is at least equal to some value Ȳ . If we let wH be the skilled
wage and wL the low-skilled wage, the total costs of production are simply
given by
TC = wL L + wH H ,
where TC is the total labour costs. The curve that describes the combination of
skilled and unskilled labour and yields the same amount of cost is called isocost.
It is easy to study the feature of the isocost in the H − L diagram displayed in
Fig. 2.1. The isocost is just a negatively sloped line with the slope equal to the
14 THE OPTIMAL SKILL RATIO

H
Isocost:
slope – wL / wH
TC2

TC1

Figure 2.1. The Isocost

wage ratioÉ in absolute value wL

wH

H wL
=− slope of the isocost
L wH
Figure 2.1 plots two different isocosts. The curve labelled TC1 refers to a larger
total cost while the curve labelled TC2 refers to a smaller value of the cost. Note
that the two curves are parallel since the relative wage cost is the same in the
two curves.

The wage skilled premium is the percentage premium required by the market for hiring a
skilled worker versus an unskilled worker.

Given a value wH and wL we can write that wH = (1 + γ )wL so that γ is the

1
skilled premium required, and the slope of the isocost is simply equal to 1+γ .
The firm would clearly like to position itself in the lowest possible isocost,
but it has the constraint of producing the quantity of output Ȳ . A trade-off
emerges. To find the formal solution we need to discuss the actual form of the
production function. In the general case the relationship between input and
output is given by the production function

Y = F (K̄ , L, H )

where H , L, K̄ are the inputs and Y is output. The formal problem is to

minimize total costs subject to the production constraint

Min (wL L + wH H )
s.t. F (K̄ , L, H ) = Ȳ .

É To formally derive the slope of the isocost in the H − L space simply write

wL TC
H =− L+
wH wH

where − wwL represents the slope.

H
 THE OPTIMAL SKILL RATIO 15

As we will see in the rest of the chapter the solution depends on whether the
two types of labour are independent in the production process. We analyse two
cases. The first one is a case in which the two types of labour are interdependent
in the production process, while the second one is a case in which the two
factors are independent. The algebraic solution is derived in Appendix 2.1.

2.3.1 SKILLED AND UNSKILLED LABOUR INTERDEPENDENT

(INTERIOR SOLUTION)
The first case we analyse is one in which the productivity of one type of labour
depends on how many workers of the different types are around. Think of a
firm in which the production line requires both engineers (high skilled) and
blue collars (low skilled). The idea is that the productivity of a given set of
engineers depends on how many blue collars are around. And the reverse is
also true. Blue collars are more productive if there are more engineers. Further,
no output can be obtained if no engineers and/or blue collars are around. In
this case we say that the factors of production are interdependent, or imperfect
substitutes.
The formal concept that we use to illustrate the relationship between dif-
ferent input combination and output is the isoquant. The isoquant describes
the combination of skilled and unskilled labour that yields the same amount
of output. Note that the isoquant is a pure technological concept, and it has no
reference whatsoever to the cost of the two types of labour.
There are three key properties of these constant-output curves which should
be familiar from microeconomics:
1. isoquants must be downward sloping;
2. isoquants do not intersect;
3. higher isoquants are associated with higher levels of output.
There is an additional property which depends on the form of the produc-
tion function. In the case of interdependent production the following feature
holds:
4. With interdependent production isoquants are convex to the origin.
Let’s discuss these properties with the help of Fig. 2.2. The first property is
important and easy to understand. The isoquant must be downward sloping.
Let’s calculate the slope of the isoquant between points X and Y in Fig. 2.2.
In going from point X to point Y , the firm hires an additional L unskilled
workers, and each of these workers produces FL units of output.Ê Hence the
Ê The marginal product of L in a production function of the type Y = F (K , L, H ) can be indicated
either with FL or with YL .
16 THE OPTIMAL SKILL RATIO

H
Isoquant:
slope – FL / FH H
Isocost:
slope – wL / wH
x
∆H y

∆L L
L

H Equilibrium:
slope isocost
= slope isoquant

Figure 2.2. The solution to the optimal skill ratio with interdependent production

gain in output is given by the product L × FL . In going from point x to point

y, however, the firm is also getting rid of H units of skilled labour. Each of
these units has a marginal product of FH . The decrease in output is thus given
by H ×FH . Because output is constant along the isoquant, the gain in output
resulting from hiring more skilled workers must be equal to the reduction in
output attributable to the reduction in unskilled labour, or

L × FL + H × FH = 0

Rearranging the terms in the expression yields

H

FL


=− slope of isoquant
L
Y =Ȳ FH
Y =Ȳ
The slope of the isoquant is then the negative of the ratio of marginal products.
The absolute value of the slope is also called the marginal rate of technical
substitution.Ë
Ë To derive the slope of the isoquant start from the production function Y = F (L, H , K̄ ) and
consider the change in output associated with a change in L and H so that

dY = FL dL + FH dH

Since along the isoquant dY = 0 we have that

dH FL
=− .
dL FH
 THE OPTIMAL SKILL RATIO 17

The second property argues that isoquants cannot intersect. This is not
surprising, since if they did intersect, a combination of two types of labour
yields the same amount of output. Such a property cannot be consistent with
the definition of isoquant. The third property says that larger isoquants refer to
larger quantity of output. As we move to north-east in the diagram we obtain
larger output. Indeed, as we move north-east in the diagram we have a larger
amount of input, to which must necessarily correspond larger output (recall
that the marginal product is positive).
With interdependent production isoquants are convex to the origin. To
understand the convexity part think of how many skilled workers (H ) we need
in place of unskilled workers (L), and still obtain the same amount of output.
The top panel in Fig. 2.2 suggests that the fewer unskilled workers we have
relatively to high-skilled workers, the larger this substitution is. Formally, the
convexity assumption implies that we have a diminishing marginal rate of
technical substitution (or a flatter isoquant) as the firm substitutes unskilled
labour for skilled labour.
The firm wants to produce at the lowest possible isocost given the constraint
described by the isoquant. The solution is given by a tangency position, where
the slope of the isoquant is equal to the slope of the isocost. In other words,
the optimal skill ratio requires that the slope of the isocost is equal to the slope
of the isoquant. Equating the slope of the isoquant to that of the isocost we
have
wL FL
= (2.1)
wH FH
which is the fundamental condition for determining the skilled combination
of the labour force. The solution is clearly illustrated in the bottom panel of
Fig. 2.2. The firm could produce the quantity of output ȳ with other combina-
tions, but such a choice would yield a larger cost. In other words, the previous
condition says that the firm chooses the skill combination so that the relat-
ive marginal product is equal to the relative marginal costs. The optimal skill
composition crucially depends on two ratios: the ratio of marginal productiv-
ity and the ratio of wages, and in equilibrium it is obtained by equating the
two ratios.
Having determined the optimal skill ratio, we now study what happens if
there is a change in the skilled wage premium γ , a parameter that the firm
takes as given. Assume that the wage skill premium increases, and assume
that the firm needs to produce a quantity indicated by y ∗ , as indicated in
Fig. 2.3. The initial position of the firm is indicated by point A. The increase
in the skilled wage induces an anticlockwise shift of the isocost. Following
the increase in γ , the original isocost is no longer sufficient to guarantee a
production level y ∗ . This suggests that the output y ∗ will be more expensive
in the aftermath of the increase in γ . But what happen to the optimal skill
18 THE OPTIMAL SKILL RATIO

H A

1/(1 + g)
B

1/(1 + g⬘)
y = y*

Figure 2.3. The effect of an increase in γ on the optimal skill ratio

ratio? The new equilibrium position will be at point B, a point at which the
slope of the new isocost (following the increase in γ ) is identical to the slope
of the isoquant. The optimal point will correspond to a lower skill ratio, and
the firm will tend to substitute low-skilled labour for high-skilled labour. With
interdependent production, an increase (decrease) in the skilled wage premium
decreases (increases) the optimal skill ratio.
The formal proof of this is obtained in the next section, but it can also be
seen graphically by the shift of the optimal skill ratio from point A to point B
in Fig. 2.3.
Finally note that in deriving the optimal skill ratio, we have not referred
to the financial position of the firm, nor whether a firm is making or losing
money. So, the optimal skill combination is independent of firm profitability.
The previous remark is important, since it says that the firm should hold
on to its optimal skill ratio independently of the current financial conditions.
In other words, savings on skilled workers in bad times can just do damage to
the firm’s profit outlook.

A Formal Example of Interdependent Production: The Cobb–Douglas Production Function

Suppose that the capital stock is equal to 1 and that the production function is Y = L β H α with
β > 0 and α > 0. This implies that the marginal product of labour is FL = ∂Y ∂L = βL
β−1 H α
∂Y
while the marginal product of skilled workers is FH = ∂H = αL H β α−1 . From these marginal
functions, it follows immediately that the marginal product of each type of labour depends
on the number of people of the other type that are currently hired. Further, the ratio of the
marginal product is

FL βL β−1 H α
=
FH αL β H α−1
FL βH
=
FH αL
 THE OPTIMAL SKILL RATIO 19

and using the key condition one has

wL βH
=
wH α L

so that the optimal ratio is given by

 
H ∗ α wL
=
L β wH

Obviously we have wH = (1 + γ )wL so that γ is the skilled premium required. It is clear that the
optimal skill mix depends on three key parameters: α, β, and γ , as the following equation
shows:
 ∗
H α
=
L β(1 + γ )

The optimal skill ratio increases with α and decreases with γ and β.

Note that in this case we have an interior solution. It is always optimal to use some combina-
tion of skilled and unskilled labour regardless of the specific value of γ . Yet, a larger value of
γ reduces the optimal skill ratio.

2.3.2 SKILLED AND UNSKILLED LABOUR INDEPENDENT

(CORNER SOLUTION)
Suppose that a firm can still produce output with two types of labour, but the
production process of each worker is completely independent from how many
workers of the other type are around. An example in this dimension could
be a firm selling computers, where salespersons can be high skill or low skill.
Each seller has some productivity that is independent of the number of other
sellers around. For extreme simplicity, let us assume that the productivity of
each type of labour is constant so that
FL = a
FH = b
We will see that in this case only one type of labour is going to be used and the
firm will be in a corner solution. We typically think that high-skilled workers
are more productive, so that b > a. This is useful for introducing an important
concept, as we do next.

The high-skilled productivity premium is the percentage increase in marginal productivity

obtained by hiring a high-skilled worker.

If we say that skilled labour is more productive by δ per cent, then b =

a(1 + δ). In other words, δ is the productivity premium. Applying equation 2.1
20 THE OPTIMAL SKILL RATIO

H Slope isoquant: a/b H

Slope isocost: wL / wH
Hire unskilled L
a/b > wL / wH
Hire skilled H
Slope isocost: a/b < wL / wH
wL / wH Slope
isoquant: a/b

L L

Figure 2.4. Optimal skill ratio with corner solution

would yield ab = wwHL which is an equation that is independent of both H and

L. There is no reason whatsoever why the ratio ab should be equal to the ratio
of wwHL . One of the two ratios is larger. We can say that it is optimal to hire only
unskilled labour if and only if
a wL
>
b wH

This situation is also described in the graphical interpretation of Fig. 2.4, and
particularly in the top panel. Note that in this case both the isoquant and the
isocost are linear equations, so that in the equilibrium position only one type
of labour should be used. This is tectonically defined as a corner solution. In
the first panel, it is optimal to hire only unskilled workers, since ab > wwHL .
Further elaborating this inequality we can say that unskilled labour should be
used if
a wL
>
b wH
1 1
>
(1 + δ) 1+γ
γ >δ

In other words, unskilled labour should be used when the wage skill premium
is larger than the productivity skill premium. With independent workers, only
one type of labour should be used, and the solution is obtained by comparing
the value of the skilled wage premium to the productivity premium.
The condition is very intuitive. Suppose that δ = 0.05 and γ = 0.1, which
means that the market requires a premium of 10 per cent for hiring a skilled
worker while the productivity premium is only 5 per cent. In this situation,
it cannot be optimal to hire skilled workers since the proportional increase in
cost is larger than the proportional increase in productivity.
 THE OPTIMAL SKILL RATIO 21

2.3.3 THE EFFECTS OF WAGE COMPRESSION

A key feature of imperfect labour markets is the presence of wage compression
across skills. Wage compression refers to a tendency of wages to be equalized
across the skill distribution. Such a phenomenon depends on many institu-
tional reasons, and most of them are linked to the unions’ desire to have an
egalitarian wage policy. We do not want to discuss why such a phenomenon
exists. What we are interested in is the consequences of wage compression on
the optimal skill ratio.
The presence of wage compression in different countries is clearly visible
from Table 2.1. The table reports the relative expected earnings of individuals
with different degrees of education in different countries. The expected earn-
ing of a worker with below upper secondary education is normalized to 100,
and the earning of individuals with higher education is expressed as a mul-
tiple of the baseline education. The last column reports the earning ratio
of an individual with tertiary education with respect to an individual with
below secondary education. Such a ratio is very similar to the skill premium
introduced in the text (the concept of γ ). The table clearly shows that the
ratio changes markedly across countries. Whereas a country such as the
United States has a value of γ roughly equal to 3, the same ratio falls to
1.8 in countries such as Germany, and slightly above 2 in countries such as
France. Wage compression refers to a reduction of γ for given productivity
differences.
We want to understand how the presence of wage compression affects the
optimal skill ratio. At the firm level, wage compression can be described as a
force that tends to equalize wages across different skill groups. In a world with

Table 2.1. Wage productivity premia in different countries

Below upper Upper Tertiary Tertiary/below

secondary secondary education secondary γ
education education

France males 100 124.08 204.80 2.05

females 100 136.43 216.53 2.17
Germany males 100 124.60 181.39 1.81
females 100 146.86 207.73 2.08
Italy males 100 188.29 267.09 2.67
females 100 195.54 229.76 2.30
United Kingdom males 100 162.81 250.20 2.50
females 100 140.88 264.02 2.64
United States males 100 163.88 294.59 2.95
females 100 173.28 291.82 2.92

Notes: Relative expected earnings of the population with income from employment by education attainment
and gender.
Population aged 30–44 below upper secondary=100.
22 THE OPTIMAL SKILL RATIO

wage compression the wage of skilled and unskilled workers is

w̃L = wL (1 + φ)
w̃H = wH (1 − φ)
where w̃H and w̃L refer to the compressed wage and φ is the wage compression
factor. In the case of independent production, unskilled labour should be
used if
a w̃L
>
b w̃H
a wL (1 + φ)
>
b wH (1 − φ)
Since φ increases the right-hand side of this inequality, it is obvious that
unskilled labour is less likely to be used the more wages are compressed.Ì Wage
compression increases the optimal skill ratio, so that in markets where wages
are more compressed firms use technologies that are more skill intensive.
With wage compression, it is possible that a firm that under normal cir-
cumstances would hire only unskilled workers will swap the quantity of its
workforce and hire only skilled workers. We label this phenomenon overedu-
cation, since a given job is filled by individuals that are more qualified than
would be the case without wage compression.

2.3.4 INDEPENDENT SKILLED AND UNSKILLED LABOUR

INTERACTING WITH COSTLY CAPITAL
In the case of independent production, it is vital to establish whether skilled or
unskilled labour should be used. One has just to compare the skilled premium
with the productivity premium: unskilled labour should be used as long as
γ > δ. This rule is not necessarily valid when labour is independent but it
interacts with costly capital. Let’s say that the production process requires to
rent a capital machine at cost Ck . It can well be the case that even if γ > δ (so
that unskilled labour should be used) the worker/machine combination that
minimizes costs is the one that combines skilled workers with machines. To
see this let’s define the cost per unit of output of using unskilled labour as
wL + Ck
a
where the numerator is the cost of hiring 1 unit of labour with a capital machine
that costs CK while the denominator is the productivity of 1 unit of unskilled
Ì Appendix 2.1 derives the consequences of wage compression in the case of interdependent
production.
 THE OPTIMAL SKILL RATIO 23

labour. Similar, the cost per unit of output of skilled labour is

wH + Ck
b
One answer (which is similar to the rule we used in the previous section) is
that skilled labour should be used if the cost per unit of output is lower, so that
wH + C k wL + Ck
<
b a
Using the fact that wH = wL (1 + γ ) and b = a(1 + δ) the previous
condition is
wL (1 + γ ) + Ck
< wL + Ck
(1 + δ)
wL (1 + γ ) + Ck < wL (1 + δ) + Ck (1 + δ)
which becomes
 
wL
δ>γ (2.2)
wL + Ck
which is the condition for hiring skilled workers when labour interacts with
costly capital. The important thing to notice is that the term in square brackets
is less than one, and suggests that it may pay to use skilled workers even when
δ = γ , since one has to take into consideration costly capital. Let’s try to
understand the economics of this result. Let’s begin with the following:
If capital is not present in the process (so that CK = 0), skilled labour should
be used if δ > γ .
This simply says that when there is no costly capital the rule is the same as
the one analysed in the previous section.
To understand this result, let us define with γ̃ the right-hand side of equation
so that
 
wL
γ̃ = γ
wL + Ck
where it is clear that γ̃ < γ as long as Ck > 0. The following gives the condition
for hiring skilled labour:
If renting a machine
 costs Ck > 0, skilled labour should be used if δ > γ̃
where γ̃ = γ wLw+CL
.
k

How can this result be explained? The intuition is as follows. To obtain one
additional unit of output using high-skilled labour requires that b1 more unit
of capital is installed, while using unskilled labour requires a1 unit of capital to
be used. Since b > a it may pay to use skilled labour and save on costly machinery
even if γ > δ. This depends on how costly machines are and on how costly
unskilled labour is.
24 THE OPTIMAL SKILL RATIO

Table 2.2. Optimal skill ratio with independent workers and costly capital

Output/worker Labour Capital Total Cost/output

costs costs costs

Machines A
Low-skilled 4 40 6 46 11.50
High-skilled 6 64 6 70 11.67
Prod. premium 0.5
Machines B
Low-skilled 8 40 10 50 6.25
High-skilled 12 64 10 74 6.17
Prod. premium 0.5

Note: Wage rates: wl = 40, wh = 64, γ = 0.6.

Consider the example in which a firm must choose between low- or high-
skilled labour with two different technologies which we summarize by their
different cost of capital. With the new technology capital is more costly but
both skilled and unskilled workers are more productive. Let’s see whether the
optimal skill choice changes with the new or the old technology. Let wL =
40, wH = 64, and the wage productivity premium γ is 0.6. Output per worker
of low-skilled workers is yL = 4 while output per worker of high-skilled
workers is yH = 6. If capital is not relevant in the production process, then
obviously low-skilled labour should be used. With the old-type technology
production requires costly capital (CK = 8), and such cost should be taken
into account. Nevertheless, the cost per output of low-skilled labour is still
lower (11.5 < 11.67) so that unskilled labour should  be used even with this
wL
capital cost. With the new machine γ̃ = γ wL +C = 0.6 ∗ 40+10 40
= 0.48,
k
so that skilled labour should be used since δ > γ̃ . Indeed, Table 2.2 shows that
high-skilled workers are most cost effective in this case.

b APPENDIX 2.1. A FORMAL EXPRESSION OF THE GENERAL PROBLEM

Output at the firm is indicated by Y . Consider two types of labour, high-skilled and
low-skilled, which we indicate respectively with H and L. The short-run production
function takes the form
Y = F (K̄ , L, H )
where K̄ is the fixed amount of capital, L is unskilled labour, and H is skilled labour.
Suppose that for each type of labour we have to choose the hours and the number of
people so that
L = L(El , hl ) H (Eh , hh )
where El is the number of low-skilled employees and hl is the number of hours per
unskilled employee. Eh and hh are similarly defined but with respect to the skilled and
 THE OPTIMAL SKILL RATIO 25

unskilled labour. The production function can be written as

Y = F (K̄ , L(El , hl ), H (Eh , hh ))

where the total quantity of each type of labour can be obtained by changing either
employment or hours. In this chapter the distinction between El and hl (and between
Eh and hh ) was ignored, and the focus was on the choice between H and L. In the next
chapter the focus is mainly on the distinction between El and hl without considering
the distinctions between H and L. This is done for simplicity.
Interdependent Production. The problem in this case is formally studied by setting
up a Lagrangian so as to maximize the following function (see Appendix B for a simple
explanation of the Langrangian technique)

= wL L + wH H + λ[F (K̄ , L, H ) − Ȳ ]

with respect to L, H , and λ. The first-order conditions read:

wL = λFL (2.3)
wH = λFH (2.4)
F (K̄ , L, H ) = Ȳ

so that, taking the fraction between the two first-order conditions, one gets:
wL FL
= (2.5)
wH FH
which is the key condition for the optimal skill ratio.
Wage Compression with Interdependent Production. Using the wage skill premium
factor γ we have:

wH = wL (1 + γ )(1 − φ),

where φ is the wage compression factor. To see the effect of wage compression in the
case of interior production
 ∗
H α wL
=
L β wH
 ∗
H α (1 + φ)
=
L β (1 − φ)(1 + γ )
where it is easy to show that wage compression
 ∗ increases the optimal skilled ratio. To
see this note that the partial derivative of HL with respect to φ is positive. In formula
 ∗
∂ HL α 1 1
= >0
∂φ β (1 + γ ) (1 − φ)2
An example of Independent Production. The obvious example is the following
production function:

Y = [aL + bH ]z
26 THE OPTIMAL SKILL RATIO

where 0 < z < 1 can be thought of as managerial ability. Indeed, applying the standard
rule we have:

FL = azY z−1

and

FH = bzY z−1

so that the ratio of the marginal product reads

FL a
=
FH b
we typically think that high-skilled labour is more productive than low-skilled labour,
so that b > a. The case analysed in the text is one in which z = 1.
3 The Hours–Employment
Trade-Off

3.1 Introduction
In imperfect labour markets, hiring labour involves a variety of costs, and
only some of such costs vary with the hours worked. In the previous chapter
labour costs were simply defined as the per hour (or per month) wage. In
reality, a variety of labour costs exist and most of such costs are independent
of the hourly wage cost. Some of these costs are paid only upon hiring, while
other costs are paid each month, but are independent of the hours worked.
The cost of advertising a vacancy refers to the first type, while paid vacations
are an example of the second one. This means that a real life organization
can adjust the quantity of labour over two different margins, which are called
the extensive margin and the intensive margin. The extensive margin refers
to the number of people hired by the firm while the intensive margin refers
to the average hours worked per employees. Choosing the right combination
between hours per workers and the number of employees means solving the
hours–employment trade-off.
Section 3.1 presents a formal, albeit simple, classification of total labour
costs. Section 3.2 presents the basic theory of the hours–employment trade-off
and derives the firm solution to the trade-off. Section 3.3 explicitly introduces
overtime salary into analysis and derives the employment effect of an increase
in overtime pay. Section 3.4 presents a case study on the employment effect of
an increase in overtime pay, or the effect of a reduction in the normal hours
of work. The case study is derived from France in 1982, when the government
suddenly decided to reduce the number of hours from 40 to 39. The results of
this change are exactly in line with those presented in the theoretical part of
the chapter.

3.2 A Formal Classification of Labour Costs

The costs the employer incurs to acquire the services of labour fall into two
categories. First, there are fixed costs. These take two forms: (i) non-recurring
fixed costs and (ii) recurring fixed costs. Non-recurring costs are also known
28 THE HOURS–EMPLOYMENT TRADE-OFF

advertising;
Non-recurring termination costs
(training)
Fixed cost
medical insurance;
Recurring
paid vacation

Normal wage

Variable costs

Overtime wage

Figure 3.1. Classification of labour costs

as one-off costs, or set-up costs, and are costs that the firm incurs each time it
takes on a new worker; recurring fixed costs are paid by the firm on a regular
basis, as long as it continues to employ that worker, but do not vary with the
number of hours or the intensity with which an employee works. Second there
are variable labour costs. These are costs that vary according to the number of
hours, and the key distinction is the one between normal wage per hour and
overtime. The fact that most forms of employment combine elements of both
of these has led to a description of labour as a quasi-fixed factor of production.
Figure 3.1 reports a diagram with the classification that we just described.

3.2.1 NON-RECURRING FIXED COSTS

Non-recurring fixed costs are paid by the firm once over the employment
relationship. The best example of non-recurring fixed costs is the hiring and
screening costs that the firm encounters each time it tries to fill a vacancy. These
are the costs of advertising a vacancy, and securing and selecting applicants. It
is in general very difficult to measure the size of these costs. The best estimate
of these costs is the one obtained with Israeli data, and shows that the typical
hiring cost is approximately two weeks of paid salary.
An important non-recurring fixed cost is the cost associated with employ-
ment protection legislation. These costs can be very large, and are a key feature
of imperfect labour markets. They are obviously non-recurring, since they are
paid only once. Yet, rather than being paid at the beginning of the employment
relationship, they are paid at the end. Further, such costs are not paid if the
worker leaves, so that what matters at the time of hiring a worker is the probab-
ility of incurring such costs. In any event, employment protection legislation
will be analysed in some detail in Chapters 4, 7, and 8. As we will see, the costs
associated with employment protection legislation can be very large, and can
even reach 20 per cent of monthly wages in some countries.
 THE HOURS–EMPLOYMENT TRADE-OFF 29

Another example of a non-recurring fixed cost is the initial expense that

the firm faces to train workers. While these costs are certainly fixed, in this
book they are considered as investment by firms, and they will be the focus of
Chapters 9 and 10.

3.2.2 RECURRING FIXED COSTS

While non-recurring fixed costs are paid only once, there are also important
recurring fixed costs. These refer to costs that the employer must regularly
pay, independently of the actual hours worked. Employers’ contributions to
unemployment insurance funds, to health insurance, and to private pension
funds are typically paid on a recurring basis, and the level of such payments is
often independent of how long or hard the employee works. These contribu-
tions are often subject to an upper ceiling and for all those employees earning
above the upper ceiling the contributions to such funds are largely independ-
ent of the amount they earn. Other examples of recurring but fixed costs are
fringe benefits (company cars, mobile phones), the clerical and administrat-
ive costs associated with employing labour, and those parts of the personnel
and welfare offices that devote themselves to helping existing employees. Most
important of all is probably the payment for days that are not worked. Holiday
pay, both statutory (public holidays) and negotiated, and sickness pay can be
a substantial part of total labour costs.

3.2.3 VARIABLE COSTS

The distinction among variable costs is less well articulated, since such costs
are less diverse. Nevertheless, there is a key distinction between standard wages
and overtime wages. Such a difference will be also the topic of our case study.
In the case of France, an example that we will study in some detail at the end
of the chapter, the overtime premium is 25 per cent of the normal wage for the
first four hours of overtime, and 50 per cent beyond.

3.3 A Theory of the Hours–Employment Trade-Off

3.3.1 THE ISOCOST AND THE TOTAL LABOUR COSTS

In the previous chapter we emphasized the fact that labour is a very het-
erogeneous factor of production, and skilled and unskilled workers are like
two different factors of production. In this chapter we concentrate on the
employment–hours trade-off, and for simplicity we assume that there is only
30 THE HOURS–EMPLOYMENT TRADE-OFF

one type of labour. Specifically, we let E be the number of people in the firm
and h the number of hours per worker.
Let us begin by properly defining E and h. Hours h are measured as hours
per unit of time, and the convention in this respect is to measure h as hours
per week. E is simply the number of people employed by the firm. As we
saw in the previous section labour costs are very diverse. Here we provide a
very parsimonious description of total labour costs, and we will (initially) just
distinguish between fixed and variable labour costs.
The problem we want to describe is the following.

The firm has to produce a given output quantity y = ȳ.

The personnel department has to hire a given type of labour. Let’s take it that we need to
assume skilled labour. The amount of skilled labour hired must be consistent with the output
level y = ȳ.
Hiring skilled workers involves the choice of two quantities: the number of skilled individuals
(the extensive margin) and the average hours per worker (the intensive margin).
The aim of the personnel department is to minimize total costs, subject to the constraint
represented by the output level.

To solve this problem, let’s first think of the total labour costs. There are two
dimensions to consider: the number of skilled workers, which we indicate with
E, and the number of hours per worker, which we indicate with h. We assume
that each worker hired involves a fixed cost equal to F . In this respect, we are
just focusing on non-recurring fixed costs. This implies that hiring E workers
involves a fixed cost of labour equal to EF , where F is the fixed cost per worker.
Let us assume that the wage rate per hour w is initially fixed at w so that the
fixed and variable costs can be written as
Fixed Cost = EF
Variable Cost = Ewh
The variable cost is equal to the weekly wage (wh) multiplied by the number of
employees. If we indicate with C the total budget allocation then we can write
C = EF + Ewh
Let’s solve the previous condition for E to obtain
C
E=
wh + F
The previous relationship highlights a key trade-off in the firm total costs.
There exists a combination of hours (intensive margin) and people (extensive
margin) consistent with a given total cost C. In other words, the firm can
obtain total costs C by choosing a point over the curve that we label isocost.
There are two key features of the isocost that we need to emphasize. First, the
 THE HOURS–EMPLOYMENT TRADE-OFF 31

isocost is downward sloping. Second, it is non-linear. The bold curve in the

chart below describes the relationship. The downward-sloping property of the
isocost is not surprising, while the non-linearity is a new feature, and deserves
to be discussed. Consider a firm that is hiring many workers, but each worker
is working a small number of hours (point X in the figure). Because there are
many workers, a small increase in the length of the workweek requires a large
reduction in the number of workers to keep costs constant. In contrast, if the
firm is hiring few workers, an increase in hours of work need not require a
large cut in employment in order to hold costs constant. This means that the
isocost will be flatter (point Z in figure 3.2).
How do we describe the output constraint? To answer to this question we
need to discuss how hours per worker and employment are combined in the
firm so as to obtain a given amount of output. We need again the isoquant
introduced in the previous chapter, but this time it refers to a slightly different
concept. Suppose that there is a production function that describes how hours
per worker and the number of employees are combined so as to obtain a given
amount of output in the firm. Let us imagine that a firm can obtain ȳ units of
output. Such output can be obtained with different combination of hours per
worker and number of employees, and all such combinations represent a point
of the isoquant. Isoquants have the same features outlined in the previous
chapter, and we briefly report them again here.É
1. An isoquant must be downward sloping;
2. Different isoquants do not intersect;
3. Higher isoquant labours are associated with higher levels of output;
4. Isoquants are convex to the origin.
At this point we have all the elements for understanding how the firm will
operate. The goal is to minimize total costs subject to the isoquant constraint.
This means that the firm would like to be in the lowest possible isocost, con-
ditional on the isoquant. The solution is again a tangency condition and it is
given by the combination E ∗ − h ∗ in Fig. 3.2.
We ask two questions. First, what happens to the employment–hours trade-
off if the scale operation increases. Second, what happens to the employment–
hours trade-off if the fixed cost of labour increases?
Let’s consider first the increase in the scale of operation. This phenomenon
can be easily described by an increase in the budget allocation C. The figure
describes the upward shift in the budget allocation. Large budget allocation
obviously leads to an increase in the number of employees. The effect of hours
is less clear, but an obvious feature of average hours per worker observed in real
organization is that firms of very different size employ workers for a similar
É The formal expression of an isolabor is then a function.

ȳ = f (h, E)
32 THE HOURS–EMPLOYMENT TRADE-OFF

isocost
X

isoquant

A
E*
Z

h* h

Figure 3.2. The solution to the hours–employment trade-off

E A E
C

D
y⬘
B y
E** y E*

E* E**

h* h h* h** h

Figure 3.3. The effect of the hours–employment trade-off of an increase in the size of the
budget (left panel) and an increase in the fixed cost (right panel)

number of hours per week. The left-hand side of Fig. 3.3, and the formal
problem described in Appendix 3.1 captures such real life property:
Larger output leads to an increase in the number of employees. Conversely, the
choice of hours is independent of output, so that an increase in the budget does
not have an impact on the demand for hours.
This property is empirically consistent with the fact that the average hours
per worker is largely independent of the size of the firms, a property that
appears empirically true. Next consider the effect of an increase in the fixed
cost of hiring labour. The isocost shifts down, and as Fig. 3.3 shows, firms
reduce employees and increase hours.
An increase in the fixed costs of hiring increases the demand for hours.
 THE HOURS–EMPLOYMENT TRADE-OFF 33

3.3.2 A FORMAL SOLUTION TO THE EMPLOYMENT–HOURS TRADE-OFF

The firm’s labour costs are
C = E[hw + F ]
and the production function is
y = Eh β with β < 1
The firm wants to minimize costs subject to the constraint that production is
equal to a fixed value ȳ. Forming the Lagrangian one has
= E[hw + F ] + λ[ȳ − Eh β ]
whose first-order conditions are
wE = λEβh β−1
(hw + F ) = λh β
ȳ = Eh β
The first condition is the derivative of the Lagrangian with respect to h, the
second condition is the derivative with respect to E, while the last one is simply
the production constraint. Taking the fraction between the two conditions
one has
w
βh −1 =
hw + F
βhw + βF = wh
βF
h=
w(1 − β)
The latter expression gives the hours per worker as a function of the per
hour wage and the fixed cost F . To obtain an expression for the number of
employees, one can simply substitute the previous expression into the isocost
to obtain
 β
βF
ȳ = E
w(1 − β)
 −β
βF
E = ȳ
w(1 − β)
from which one can immediately see that
∂ log E
= −β
∂ log F
so that the elasticity of employment to the fixed cost is negative and equal
to −β.
34 THE HOURS–EMPLOYMENT TRADE-OFF

3.4 The Overtime Premium and the Effects of

Reducing the Workweek

3.4.1 FIRM POSITION WITH THE OVERTIME PREMIUM

In real life organization there is typically a ‘normal’ number of hours to which
the firm pays a base wage. The normal number of hours is set by law or by
collective agreements, and it is taken as given by the firm. We indicate with ho
such ‘normal’ number of hours and with w the corresponding wage.
Firms can choose to employ workers for a larger number of hours h > ho
but such rate requires an overtime wage w̃ > w.
Let us assume that the personnel department has a budget allocation equal
to C. The existence of the overtime wage implies that the shape of the isocost
changes in correspondence to the normal number of hours ho .
If the firm does not use overtime (i.e. if h ≤ ho ) the isocost is identical to
that of the previous section.

C = EF + wEh if h ≤ ho

If the firm does use overtime, two different wages apply to the firm: one for
normal hours and one for overtime hours. The isocost in this case is equal to

C = EF + wEho + w̃E(h − ho ) if h > ho

The slope of the isocost changes at the normal hours ho , with a steeper
curvature beyond the normal hours ho . Figure 3.4 plots the isocost. In cor-
respondence to the normal number of hours the isocost features a kink
point.
The isoquant for employees’ hours is the same as the one described in the
previous section, and the objective of the firm is still to minimize costs subject
to the isoquant constraint.
A firm can be in two different positions, depending on whether or not it
uses overtime.
Suppose that there are two firms, one that uses overtime and one that does
not use overtime. The situation is described in the top panel of Fig. 3.4. The
first panel describes a firm that is operating on the ‘normal’ number of hours,
and it positions itself on the kink point of the isocost. The firm in the second
panel, conversely, corresponds to a firm that is operating in the overtime range,
and positions itself to the right of the kink point ho .
 THE HOURS–EMPLOYMENT TRADE-OFF 35

E
E

h0 h h0 h

Figure 3.4. Equilibrium position for a firm that uses and does not use overtime

h0⬘ h0 h

Figure 3.5. The reduction of the normal workweek and the isocost

3.4.2 THE EFFECT OF A REDUCTION IN THE NORMAL WORKWEEK

Let’s now consider a reduction in the normal workweek ho . The first thing
to understand is what happens to the isocost when the normal week changes.
This is done in Fig. 3.5. Consider a given total cost C. For a number of hours
below the new value, the isocost does not change. For values of h above the
new minimum the isocost corresponding to C is now lower, since the firm has
to pay a larger cost w̃.
Let us first consider a firm that before the change in the normal workweek
was working exactly ho hours. In the aftermath of the change, such a firm will
find itself suddenly in the overtime range. This implies that the wage for the
last hour goes up and the isocost, in correspondence to the old normal hours,
is now steeper. Such a firm will necessarily reduce the number of employees,
and its equilibrium position, as indicated by Fig. 3.6, will shift from point A to
point B.
36 THE HOURS–EMPLOYMENT TRADE-OFF

E A

h0⬘ h0 h

E C

h0⬘ h0 h

Figure 3.6. The effects of reduction in the normal workweek

Let us now consider a firm that before the overtime change is operating
below the normal workweek. In terms of Fig. 3.6, such a firm was operating in
point C in the figure. The change in the slope of the isocost involves a part of
the constraint that does not affect the position of the firm. This suggests that
such a firm will not change its equilibrium position, and will continue to hire
the same number of workers for the same workweek.
Combining the results of the two type of firms (described by Fig. 3.6), it is
clear that the effect of a reduction in the normal workweek is a reduction in
the number of workers hired. The following summarizes the findings:
A reduction in the normal workweek induces a reduction in the number of
employees.
The next section analyses a case study on workweek reduction in France, and
shows that the theoretical predictions are largely consistent with the empirical
evidence.

3.5 Reducing the Workweek: A Natural Experiment

from France
This section discusses the reduction in the workweek that took place in France
in 1982, and shows how our analytical framework is fairly consistent with the
empirical evidence.
 THE HOURS–EMPLOYMENT TRADE-OFF 37

Table 3.1. France: hours worked 1976–1981

1976 1977 1978 1979 1980 1981

Fraction of employment
Working:
36 to 39 hours 2.2 2.1 2.5 2.4 2.6 2.4
40 hours 46.6 53.6 56.6 58.6 60.9 65.9
41 to 43 hours 18.8 18.8 19.5 19.3 17.5 15.2
44 hours 4.2 4.4 3.9 2.3 2.7 2.0
45 to 48 hours 28.2 21.1 18.6 17.4 16.4 14.5

Number of observations 5,422.0 6,133.0 6,212.0 6,123.0 6,409.0 6,509.0

Source: Crépon and Kramarz (2002)

The number of hours worked in France strongly decreased during the 1970s,
from 48 hours in 1974 to just above 40 hours in 1981 (Table 3.1). Since 1936, the
standard workweek in France was 40 hours. The French Socialist government
in 1982 reduced the workweek from 40 to 39 hours. In this section we study
the effects of such a change on the number of hirings.
François Mitterrand’s election in May 1981 induced a sudden decrease of
the standard workweek to 39 hours (16 January 1982 ordinance), and the new
39 law was imposed on 1 February 1982. The law mandated a maximum legal
workweek of 39 hours, whereas it was 40 hours previously. The overtime regula-
tion did not substantially change: the overtime premium remained 25 per cent
for the first four hours (i.e. from 40 to 43 hours), and 50 per cent above. Col-
lective agreements, specifying the terms of application of the decree, ensued in
the aftermath of the ordinance, starting with the larger firms in manufacturing
industries and spreading to smaller firms and other industries.
It is possible to show that workers directly affected by these changes—those
working 40 hours in March 1981 as well as those working overtime at the same
date—were more likely to lose their jobs between 1981 and 1982 than those
workers not affected by the changes—those working 36 to 39 hours in March
1981. The estimate of the impact of the one-hour reduction of the workweek
on employment losses varies between 2 per cent and 4 per cent.
The result is based on a comparison between workers who worked 36
to 39 hours before 1982 and workers who worked exactly 40 hours and
those who worked overtime (up to 48 hours). The result shows that workers
who were working exactly 40 hours per week in March 1981 as well as workers
who were working overtime (41 to 48 hours) per week in March 1981 were
less likely to be employed in 1982 than observationally identical workers who,
in 1981, were working 36 to 39 hours per week. The analysis uses econometric
techniques that allow the scholar to compare changes in labour market posi-
tion between 1981 to 1982, the period immediately after the implementation
of the decree, with those prevailing between 1978 and 1981, before the election
of François Mitterrand.
38 THE HOURS–EMPLOYMENT TRADE-OFF

3.6 Theoretical Predictions and a Natural Experiment

The theoretical predictions of the Mitterrand policy changes are exactly those
analysed in the previous section. Let’s think that the standard workweek is
ho = 40 and assume that we experience a sudden change to ho = 39. A
key distinction should be made between workers employed in firms that
are working 39 hours and workers employed in firms that hire workers for
40 hours. Consider first a firm that before the change is operating with a work-
week of 39 hours. What is the impact of the reduction of the workweek for
such a firm? If nothing else changes, such a firm should not be affected by the
reduction in the normal workweek.
Firms using a workweek of 39 hours before March 1982 should not be
affected by the reduction in the workweek. As a consequence, the probability
that workers lose their job is independent of the change in the length of the
workweek.
Consider now a firm that before the change in working hours is operat-
ing with a workweek of exactly 40 hours. Once the change in the workweek
is implemented, such a firm will find itself in the overtime range. This
induces a shift in the isocosts exactly like that described by Fig. 3.6. Such
a firm suddenly finds that the last hour is charged at overtime rate. How
should such a firm react to the change? We have learned from the previous
section that such a firm should reduce employment (and possibly increase
hours).
Firms using a workweek of 40 hours before March 1982 are affected by the
change in the hours, and should react by reducing the number of employees.
As a consequence, the probability that a worker loses his or her job after the
new law should increase.
This example is a wonderful situation for economics, since it can be defined
as a natural experiment, exactly like those that are undertaken by the hard
science literature. In these situations, we can compare the effect of the policy
change on the treatment group with the effect observed on the control group,
where the latter type of individuals were not treated by the experiment. In these
circumstances we can understand the impact of the policy change by looking
at the different response of the two groups.
The treatment we are considering is the reduction in the normal hours
of the workweek from 40 to 39 hours. The control group is formed by the
individuals that work 36 to 39 hours before the new law is introduced: they
are not affected by the treatment. The treatment group are those individuals
that were working 40 hours or more before the change. The outcome that
we are interested in is the probability of losing a job. For the control group,
the probability of losing a job should be independent of the change in the
workweek. Conversely, for the treatment group, the probability of losing a job
should increase. As a consequence, the difference between the two probabilities
 THE HOURS–EMPLOYMENT TRADE-OFF 39

should provide an empirical estimate of the size of the impact that we are
looking for. If the probability of losing the job for the treatment group increases
relatively to the probability of losing the job for the control group, then the
impact of the experiment is positive. This is called a natural experiment, and
it is a perfect application of our theory. Let’s see what the data said in the
French case.

3.7 The Evidence

Table 3.1 shows the proportion of full-time workers employed 36 to 39 hours
within the population of all full-time workers employed 36 to 48 hours in 1981.
This fraction is small, 2.4 per cent, but it is increasing across time. The fraction
of workers employed 40 hours is also increasing whereas the fraction of over-
time workers constantly decreases between 1976 and 1981. Table 3.2 shows
the employment to non-employment transition rates for these various cat-
egories of workers. Between 1981 and 1982, employment to non-employment
transitions are more intense for workers employed 40 hours than for those
employed 36 to 39 hours; 6.2 per cent of all workers employed 40 hours in
1981 have no employment in 1982 whereas 3.2 per cent of those employed
less than 40 hours are in the same situation, a difference of three percentage
points.
Let’s define with JL8240 the probability of a job loss for an individual employed
36−9
40 hours in 1982 and with JL82 the probability of a job loss for an individual
hired 36–9 hours in 1982. In formula, the two expressions read

40 Job losses by 40 hours workers in 1982

JL82 =
Number of Workers at 40 hours in 1981
36−9 Job losses by 36–9 hours workers in 1982
JL82 =
Number of Workers at 36–9 hours in 1981

Table 3.2. Employment losses, by hours, in France

Year t to t + 2 80–2 77–9 78–80 79–81

Job losses of workers employed:

36 to 39 hours 3.2 3.9 2.7 7.3
40 hours 6.2 4.3 5 5.5
41 to 43 hours 4.6 3.1 3.6 4
44 hours 6 5 2.1 5.8
45 to 48 hours 5.7 4 4 3
Observations 6,509 6,212 6,123 6,409

Source: Crépon and Kramarz (2002)

40 THE HOURS–EMPLOYMENT TRADE-OFF

36−9 40 is the
where JL82 is the job loss probability of the control group while JL82
job loss probability for the treatment group.
The idea is that the probability of job loss of the treatment group, in excess to
the job loss probability of the control group, describes a measure of the impact
of the policy change. In formula, a first measure of the potential impact  is
given by
40 36−9
82 = JL82 − JL82
Table 3.2 suggests that a simple measure of 82 is 3 percentage points.

The Formal Evidence

While the estimate of 82 presented above represents a first important res-
ult, there are two other important elements to be taken into account. First,
one has to consider the possibility that individuals hired in jobs that require
40 hours are systematically different from individuals hired in jobs that require
36–9 hours. In other words, we need to control for individual characteristics,
such as gender, age, education, etc. Second, we need to consider whether the
same difference calculated for other years, basically 79 or 80 , 81 , is not
statistically different. This is very important, since it would suggest that we
are experiencing an effect only in the estimate of 82 , with no effects on the
previous years. This can be done by using regression analysis and individual
data. To this end, we introduce the following notation and we label
i = all individuals that work at 36,37,38,39 or 40 hours in 1981

1 if the worker loses the job in 1982

JL82,i =
0 otherwise

The key regression is the following

JL82,i = Z81,i β + α81 40h1981 + εit
where the variable 40h1981 is a dummy variable that takes a value of 1 if the
individual was working 40 hours in 1981. In other words, α81 is an estimate of
the differences 82 controlling for individual characteristics. If one wants to
introduce also the individual working overtime the regression to be estimated is
40 ot
JL82,i = Z81,i β + α81 40h81 + α81 overtimeh81 + εit

The value of α8140 is 2.6 per cent, which suggests that the reform increases

the chance of losing one’s job if working 40 hours by 2.6 percentage points.
The coefficient is reported in the first column of Table 3.3. Finally, note that the
same regressions calculated for the previous years in which no reform was
undertaken are not statistically significant. This can be done by looking at
 THE HOURS–EMPLOYMENT TRADE-OFF 41

Table 3.3. Regressions on employment losses

80–2 77–9 78–80 79–81 Pooled

1 2 3 4 5
Non-employment at t + 2
Hours = 40 2.60 −1.26 1.29 −3.85 3.90
standard error 1.44 1.86 1.48 2.30 1.82
41<= hours<= 43 1.32 −2.40 1.06 −5.09 3.49
standard error 1.60 1.91 1.57 2.39 1.97
Hours = 44 2.50 −0.73 −0.32 −4.57 4.20
standard error 2.50 2.35 1.91 2.92 2.88
45<= hours<= 48 2.12 −2.14 1.50 −6.52 4.52
standard error 1.66 1.96 1.61 2.39 2.03

Notes: Regressions for the LFS panels of 77–9, 78–80, 79–81, and 80–2 (linear probability mod-
els). The dependent variable is non-employment in the exit year of the panel (79, 80, 81, and 82,
respectively)
Independent variables: indicator for the hours categories (only reported coefficients), industry,
region (Ile de France or other), skill level (3 categories), sex, diploma (6 categories), experience (4
categories), seniority (4 categories), labour market status (apprentice or not), and hours in first year
of the panel strictly below 40 and hours in first year of the panel strictly above 40. The population
includes all full-time workers in the private sector working between 36 and 48 hours in the median
year of each panel (78, 79, 80, and 81, respectively). Column 1 reports estimates for the panel 80–2,
columns 2, 3, and 4 report estimates for the panels 77–9, 78–80, and 79–81 respectively. Column
5 reports pooled estimates where all variables are interacted with the relevant year indicator except
for the hours categories for which we introduce pooled coefficients and coefficient specific to year
1981 (panel 80–2). These last coefficients are those reported in column 5 (pooled).
Source: Crépon and Kramarz (2002)

the other columns of Table 3.3 (those labelled 77–9, 78–80, etc.), where the
coefficient on 40 hours is not significant.

b APPENDIX 3.1. A MORE FORMAL VERSION OF THE

HOURS–EMPLOYMENT TRADE-OFF

Initially we assume that if a typical worker is asked to work additional hours, he or she
will require a higher wage rate. The firm thus faces the following wage function

w = w(h) w > 0

where the positive slope can be thought of as the result of an upward-sloping labour
supply from the worker’s standpoint. The firm’s labour costs are

Labor Cost = Ehw(h) + EF

where the simple form labour costs LC = wE follows each time we assume that F = 0
and h = 1 fixed. The production function takes the form as

y = E α hβ
42 THE HOURS–EMPLOYMENT TRADE-OFF

and the problem can be thought of as one of minimizing total costs subject to a fixed
production input ȳ. Formally, this is equivalent to having
MinE,h Ehw(h) + EF
s.t. E α h β = ȳ
So that forming the Lagrangian one has
= Ehw(h) + EF − λ[E α h β − ȳ]
whose first-order conditions are
λαE α−1 h β = [hw(h) + F ]
λβE α h β−1 = [Ew(h) + Ehw  (h)]
E α h β = ȳ
Taking the ratio of the first two we obtain
FE MCE
=
Fh MCh
αE α−1 h β hw(h) + F
α β−1
=
βE h w(h)E[1 + w(h)
h
w  (h)εw,h ]
αh hw(h) + F
=
βE w(h)E[1 + εw,h ]
where εw,h is the elasticity of w(h) with respect to h.
The general rule is
FE MCE
=
Lh MCh
where FE is the marginal increase in output following an increase in total employment
while Fh is the marginal increase in output as more hours are worked. An important
property follows:
The choice of hours is independent of the scale of operation.
The proof is straightforward from the first-order conditions, since E is immediately
removed from the equations so that h is a function of only εw,h and F , so that
αhw(h)[1 + εw,h ] = βhw(h) + βF
from which it follows that
h = h(F , εw,h )
with
∂h
>0
∂F
∂h
>
∂εw,h
 THE HOURS–EMPLOYMENT TRADE-OFF 43

while from the budget constraint one gets

C
E=
h(ε, F )w(h) + F

from which it follows that

E = E(F , ε, C)

with
∂E
<0
∂F
∂E
>0
∂F
∂E
>0
∂C

The Overtime Premium and the Effects of Reducing the

Workweek
If hs are normal working hours.

C = EF + wEh if h ≤ ho
C = EF + wEho + w̃E(h − ho ) if h > ho

Let the production function be equal to y = Eh β . The firm can be in two possible
positions. It uses overtime or it does not use overtime. The two situations are depicted
in Fig. 3.4. Let’s first study a firm that uses overtime. The formal problem is

Minh,E EF + wEho + w̃E(h − ho ) + λ[ȳ − Eh β ]

The first-order conditions are

w̃E = λβh (β−1) E

[F + who + w̃(h − ho )] = λh β

Taking the ratio of the two functions we get

w̃
βh −1 =
F + who + w̃(h − ho )
βF + βwh0 + β w̃h − β w̃h0 = h w̃
βF + βh0 (w − w̃)
h= h > ho
w̃(1 − β)
44 THE HOURS–EMPLOYMENT TRADE-OFF

Substituting this into the production function and rearranging, the optimal E reads
 
βF + βh0 (w − w̃) −β
E = ȳ
w(1 − β)
A reduction in the normal workweek induces a reduction in the number of employees.
To see this simply take the partial derivative of the log to obtain.
∂ log E
= −β
∂ log h0
4 Temporary or
Permanent?

4.1 Introduction
One of the key issues at the hiring stage concerns the type of contract that
should be offered to the firm’s workforce. In this chapter, we discuss the key
decisions in terms of the length of the contract: open-ended versus fixed-
term contract. Alternatively, as such contracts are called in Europe, the firm
has to choose between a permanent or a temporary contract. This choice is
particularly relevant when two conditions hold. First, employment protec-
tion legislation is stringent, and terminating a contract is an expensive activity.
Second, uncertainty over business conditions plays an important role. If a con-
tract can be terminated at no cost, the difference between temporary contract
and permanent contract is irrelevant. Further, if business conditions do not
fluctuate, hiring decisions are not going to be followed by separation decisions,
and the type of contract is irrelevant.
In this chapter we assume that terminating a contract involves some costs
to the employer and that business conditions can be good or bad. We study
three scenarios. First, as a baseline scenario, we consider a firm that can offer
only permanent contracts that can never be broken. Business conditions can be
good or bad, but employment cannot be adjusted in response to such changes.
We study the optimal labour demand in this condition, and we compare it
with the employment decision of a firm that is fully flexible, and can hire at
will workers on a temporary basis. We find that the rigidity imposed on the
firm affects its profitability, but has no impact on the average employment.
Indeed, the average employment of the rigid firm is the same as the average
employment of a flexible firm.
Second, we consider what happens when a firm has the possibility of com-
bining temporary and permanent contracts. The idea in this scenario is that
a buffer of temporary contracts can be used in conjunction with a stock of
permanent contracts. We show that under these conditions the constrained
firm can do as well as the flexible firm.
Third, we study the buffer stock model under an additional real life phe-
nomenon, namely the fact that hiring a temporary worker requires the use of
a costly interim agency. This implies that the cost of temporary workers also
includes the mark-up paid to temporary firms. Such a phenomenon will show
46 TEMPORARY OR PERMANENT?

that the additional cost imposed by the interim agency has important effects
on the firm’s hiring strategy. We will see that in this scenario a firm can hire
an additional amount of permanent workers (with respect to a pure buffer
model), just to save on future agency costs.
In real life labour markets, some firms offer temporary contracts while other
firms offer permanent contracts. We want to understand how this choice is
determined when terminating permanent contracts is not illegal, but requires
some costs. We thus study the trade-off between temporary and permanent
contracts when labour turnover is costly. We assume that permanent contracts
can be terminated at some cost, while temporary contracts can be terminated at
no cost. Further, we assume also that workers under temporary contracts have
a higher probability of leaving the firm, and their replacement is not necessarily
immediate. We will see that under this condition a trade-off emerges, and it is
not always optimal to hire under temporary or permanent contracts. A brief
look at the evidence completes the chapter.

4.2 Permanent Contracts with Fixed Wages

We make the following assumptions on the environment in which the firm
operates.
The production function is described by the following relation

Y = A i log L,

where L represents the quantity of labour employed and Y the output; Ai is

the productivity level;
There are business fluctuations in the productivity at the firm; A i assumes
only two values A h > A l . In every period, there is a probability p that pro-
ductivity be equal to A h and a probability 1 − p that productivity be equal to
A l ; we refer to periods in which the productivity is A h = A as good times, and
periods in which the productivity is equal to A l as bad times.
h
The marginal product of the firm is YL = AL .É Fluctuations in Ai are akin
to fluctuations in the marginal product.
The wage is fixed and equal to w and the price that the firm charges for
simplicity is set equal to 1 and does not change between good and bad times.
The key firm decision is the quantity of labour to be hired.
We consider initially two different scenarios under which the firm operates:
the flexible/temporary regime and the rigid regime. In the flexible regime, hiring

É This is obtained by taking the derivative of the production function with respect to L. Notably,
∂A i log L i
∂L = AL .
 TEMPORARY OR PERMANENT? 47

and firing can take place at no cost, and the firm can choose its employment
level after observing the realization of the value of A. The firm in the flexible
regime hires workers on a temporary basis. In the rigid regime, the firm can only
choose the average employment, so that there are only permanent contracts that
can never be broken. Firms can offer only permanent contracts, and firing is
impossible. Once employment is decided it can never be changed.
In the next section we consider an additional regime, mainly the buffer
regime, which involves the possibility that the firm combines temporary and
permanent contracts.

4.2.1 A NUMERICAL EXAMPLE

Before formally solving the problem, let us analyse a numerical example. We
assume the following values for the parameters: w = 1; A h = 10; Al = 4; and
p = 0.5.
Let’s begin with the flexible firm. Such a firm can hire and fire at no cost, and
its hiring policy is simply to choose labour so as to maximize profits period by
period. In each period, the profits are just the revenues minus costs so that

F 10 log L − WL if A = A h
 =
4 log L − WL if A = A l

The solution to the firm employment should just be such that the marginal
product is equal to the wage in good and bad times, as in a simple and static
model of labour demand. Since the wage is 1, the optimal employment is equal
to 4 in bad times and 10 in good times. Figure 4.1 also plots the marginal
product curve and the wage. Since the probability of having good and bad
times is equal to 0.5, the average employment of the flexible firm is simply
7. The firm hires six workers when business conditions improve and fires six
workers when business conditions worsen.
Consider now the rigid firm. The firm should choose a level of employment
that maximizes average profits. Such a level of employment would never be
changed. The average profits are

R = p[A h log L − wL] + (1 − p)[A h log L − wL]

R = [(1 − p)A l + pA h ] log L − wL

The value of the profits function R is easily computed in the table, and the
optimal employment level corresponds to an employment level equal to 7.
Figure 4.2 shows clearly that this is the point at which profits are maximum.
We are now in a position to derive three important empirical implications on
the effect of the protection regimes.
48 TEMPORARY OR PERMANENT?

5 Marginal productivity in good times

4
Marginal productivity in bad times
3

0
2 3 4 5 6 7 8 9 10 11 12
Employment

Figure 4.1. Labour demand in good and bad times

Flexible economy
14 in good times
12
10
Profits rigid economy
8
6
Flexible economy
4
in bad times
2
0
2 3 4 5 6 7 8 9 10 11 12 13 14 15
–2
–4
–6
Employment

Figure 4.2. Profits in good and bad times

Implication 1: the average employment of the rigid firm is the same as in

the flexible firm. In our example, the average employment of both firms is 7.
Implication 2: The volatility of employment is higher for the flexible firm.
While employment in the rigid firm is always fixed at 7, the flexible firm
hires and fires six workers when the economy switches from good to bad
periods, so that obviously its employment is more volatile.
Implication 3: The flexible firm enjoys larger average profits. To compare the
profits we need to compare the profits of the rigid economy with the average
 TEMPORARY OR PERMANENT? 49

Table 4.1. Flexible and rigid regimes

Wage Marginal Marginal Flexible economy Rigid economy

productivity productivity average profits
good times bad times Profits Profits bad times
good times

2.00 1.00 5.00 2.00 4.93 0.77 2.85

3.00 1.00 3.33 1.33 7.99 1.39 4.69
L*bad 4.00 1.00 2.50 1.00 9.86 1.55 5.70
5.00 1.00 2.00 0.80 11.09 1.44 6.27
6.00 1.00 1.67 0.67 11.92 1.17 6.54
L Rigid 7.00 1.00 1.43 0.57 12.46 0.78 6.62
8.00 1.00 1.25 0.50 12.79 0.32 6.56
9.00 1.00 1.11 0.44 12.97 −0.21 6.38
Lgood 10.00 1.00 1.00 0.40 13.03 −0.79 6.12
11.00 1.00 0.91 0.36 12.98 −1.41 5.79
12.00 1.00 0.83 0.33 12.85 −2.06 5.39
13.00 1.00 0.77 0.31 12.65 −2.74 4.95
14.00 1.00 0.71 0.29 12.39 −3.44 4.47
15.00 1.00 0.67 0.27 12.08 −4.17 3.96

Notes: Good times shifter Ah = 10; bad times shifter Al = 4; wage w = 1; probability p = 0.5.

profits of the flexible economy. Table 4.2 shows that profits are larger and
equal to 7.29 in the flexible case (the average between 13.03 and 1.55),
against 6.62 in the rigid case.

Let us now consider the buffer regime. The idea is that the firm can hire workers
under temporary contracts, and can use a combination of temporary and
permanent contracts. A stock of permanent contracts is hired in good and bad
times, while temporary workers are used as a buffer stock when conditions
improve. Table 4.2 shows that a firm that follows such a policy can reach a level
of profits identical to a flexible firm. Basically, the firm has four permanent
workers and six temporary workers when conditions improve.

4.2.2 FORMAL DERIVATION

Flexible Firm. Let’s consider first the behaviour of the firm if there were no
restrictions on the type of contracts. In other words, let’s consider the situation
when hiring and firing can take place at no cost. Under no constraint, the firm
will choose the employment level after observing the productivity level, and it
will maximize profits in the following way:

F = MaxL [A i log L − wL]

50 TEMPORARY OR PERMANENT?

Table 4.2. Profits in various regimes

Good times Bad times Average

Flexible
Employment 10.00 4.00 7.00
Profits 13.03 1.55 7.29
Rigid
Employment 7.00 7.00 7.00
Profits 12.46 0.78 6.62
Rigid with temporary
Permanent employment 4.00 4.00 4.00
Temporary employment 6.00 0.00 3.00
Profits 13.03 1.55 7.29
Rigid with costly interima
Permanent employment 5.00 5.00 5.00
Interim employment 3.00 0.00 1.50
Profits 12.04 1.44 6.74
Permanent employment 4.00 4.00 4.00
Interim employment 4.00 0.00 2.00
Profits 11.79 1.55 6.67

a Mark-up wage equal to 1.25.

which implies, after taking the derivative of the profits with respect to L

Ai
=w i = h, l,
L
Ai
Li =
w

Al
if A = A l
L= w
Ah
w if A = A h

The flexible firm chooses the level of employment after having observed
the level of productivity. Since the wage is fixed and equal to w, the flexible
firm will fire (hire) L = A w−A when the economy moves from high (low)
h l

productivity to low (high) productivity. Because the economy experiences, on

average, a fraction p of high productivity periods and a fraction (1 − p) of
periods of low productivity, the average employment in the long run will be

F (1 − p)A l + pA h
L =
w
Rigid Firm. Let us examine, now, the behaviour of the firm when it is forced to
hire only permanent contracts. In other words, an extremely strict employment
protection regime is in place, and it is not possible to change employment after
productivity changes. In this situation, the best thing the rigid firm can do is
 TEMPORARY OR PERMANENT? 51

to choose the employment that maximizes the expected value of the profits. In
formulas, this implies that the employer R will resolve the following problem:
 
R = MaxL E A i log L − wL ,

where E indicates the expected value of the profits. The being wage fixed, to
eliminate the expected value from the previous expression it is sufficient to
replace the expected value of production, so that the problem of the employer
becomes:
 
R = MaxL [(1 − p)A l + pA h ] log L − wL .

The first-order conditions allow us to derive the value of employment in the

rigid country R. as

R (1 − p)AL + pAH
L =
w
R
The value L is some average between the level of employment in the flexible
R
country during the expansions and its level during recessions. Moreover, L
coincides with A /W if the economy is always in low productivity (p = 0). If
l

lay-offs are not allowed, the firm in the rigid regime does not experience any
fluctuation as a result of variations in the productivity level.
We are now in a position to derive three important empirical implications
on the effect of the protection regimes.

Implication 1: the average employment of the rigid regime is the same as in

the flexible regime.
Implication 2:the volatility of employment is higher in the flexible regime.
Implication 3: the firm in the flexible regime is more efficient and makes
more profits.
R F
Implication 1 is immediately verified. We have seen that L = L , or that
average employment in the flexible regime is the same as for the firm in the
rigid environment. Implication 2 is also easy to show. By construction, in the R
economy there are no employment variations, while the F economy fires (hires)
h −A l
L = A W labour when the economy moves from high (low) productivity
to low (high) productivity. Implication 3 is also easy to demonstrate. It is
enough to realize that the employment level chosen by the F employer in each
period is the only level that maximizes profits in each period. Consequently,
in each period profits are higher in the flexible regime. With the same level of
employment, the firm is able to make, on average, a higher level of profits. In
other words, the firm in the flexible regime is more efficient.
52 TEMPORARY OR PERMANENT?

4.3 Temporary Contracts as a Buffer Stock

Let’s consider now an environment which is less extreme than the one before.
Specifically, let’s consider a situation in which a firm can combine permanent
and temporary contracts at the same time. The idea is that a stock of permanent
contracts can be offered alongside a set of temporary contracts that terminate
in bad business conditions. The firm’s hiring policy can be described as follows:

L perm if A = A l
L = perm
L +L temp if A = A h

The idea is that the firm can hire a stock of permanent contracts independently
of business conditions while it can hire workers under temporary contracts
when conditions are good, and dismiss such contracts when conditions are
bad. Following the algebra derived in Appendix 4.1, one can show that the
stock of permanent workers is equal to employment.
The firm has the following hiring policy:

Al
if A = A l
L = Al wAh −Al
w + w if A = A h

If temporary contracts can be used as a buffer stock, a proper combination

of temporary and permanent contracts allows the firm to reach the first best.
These results show that by combining permanent and temporary contracts
the firm is able to reach the first best. The stock of permanent contracts is
identical to the size of employment in bad times while the stock of temporary
contracts acts as a buffer stock that the firm can adjust if conditions change. In
this case the stock of workers on permanent contracts does not experience any
change in employment status, while temporary contracts are hired and fired
at every change in business conditions.
In practice there are three limitations to the policy that we just described:

1. Roll-over of temporary contracts is not always possible. Indeed, if the firm

has a sequence of good productivity shocks, it has to roll over temporary
contracts for a sequence of periods. And such a policy is not always
feasible.
2. Hiring temporary contracts often requires some specific hiring costs.
These are particularly relevant if temporary contracts are obtained
through interim agencies.
3. Workers under temporary contracts experience high turnover, and are
likely to seek permanent job opportunities elsewhere.

In the next section we analyse the role of specific additional costs associated
with temporary contracts.
 TEMPORARY OR PERMANENT? 53

4.4 The Buffer Stock Model with Interim Costs

Interim workers are technically defined as workers hired by an interim agency
and rented out to a specific firm. From the firm’s standpoint, the benefit of hir-
ing workers as interim workers is their inherent flexibility, since such workers
are technically hired by the agency. The cost associated with such a policy is
the fee that the agency imposes on the firm. In this section we study the role
of costly interim contracts in the context of our buffer stock model. We ask
what happens to the optimal combination of temporary permanent contracts
when the hiring of additional temporary workers requires a fee to be paid to
the interim agencies.
Formally, we assume that each worker hired on a temporary basis costs
w(1 + µ) where µ > 0 is the mark-up on the wage (or the fee) to be paid
to the interim agency. The firm’s hiring policy can still be described by the
employment policy of the buffer stock model

L perm if A = A l
L=
L perm + L interim if A = A h

where L perm are the permanent workers hired in bad times and L interim are
the interim workers hired when conditions turn good. The key question in
this context is how the existence of the interim fee modifies the employment
position of the firm with respect to a pure buffer stock model with µ = 0.
The firm’s position in good business conditions is not difficult to determine,
since it is simply obtained by the intersection between the marginal product in
good times (when A = A h ) and the marginal labour costs, which in this case
are equal to w(1 + µ). The position is described by point C in Fig. 4.3. In good
times the firm is hiring less labour than the case of µ = 0, indicated by point
B in the figure. This is not surprising, and it is simply linked to the increase in
the marginal labour cost.
The determination of employment in bad times is more subtle. In bad times
the firm is not going to hire interim workers. This should not be controver-
sial. The difficult question is how many permanent workers should be hired.
During bad times the firm knows that good times will eventually come in the
future, and that hiring of interim workers involves a mark-up µ on the wage.
As a way to partly avoid future hiring costs, the firm may have an incentive to
‘overhire’ permanent workers in bad times, or in any case to hire more perman-
ent workers than would be the case if µ were zero. Nevertheless, the firm is also
aware that permanent workers hired beyond L L in Fig. 4.3 are associated with
some loss, since for such workers the marginal product is lower than the wage.
A trade-off emerges in this situation, but it is likely that some overemployment
of permanent workers is going to take place. Indeed, the algebra derived in the
54 TEMPORARY OR PERMANENT?

W
MPLH
MPLL

C
w (1 + m)
D
B

LL L perm Lp + Lt LH L
= AH/ w = AH/ w
LR
=[ pAH + (1 – p)AL ]/w

Figure 4.3. Buffer stock model with costly interim workers

Appendix shows that

perm perm
L(µ>0) > L(µ=0)
perm perm
where L(µ=0) corresponds to point L L in Fig. 4.3 while L(µ>0) corresponds to
point D. This overemployment of permanent workers is called labour hoarding.
A key result is immediately obtained: The existence of turnover costs (interim
wages in these cases) induce labour hoarding of permanent workers in bad times.
The employment behaviour of the firm can be described as follows
⎧
⎨ Apwµl
if A = A l
L= w− 1−p
⎩ Ah
w(1+µ) if A = A
h

From the previous condition one can immediately see that the employment
solution collapses to a simple buffer stock model whenever µ = 0. Further,
labour hoarding increases with the size of the interim fee: The larger the interim
fee, the larger is labour hoarding.
To formally establish this point one has simply to differentiate the employ-
ment level in bad times with respect to µ, and find that the result is
positive.Ê

Ê The result yields

∂ A l pw
= >0
∂L [(1 − p)w − pwµ]2
 TEMPORARY OR PERMANENT? 55

3 Marginal productivity
in good times
2.5 Marginal
productivity
Wage with mark-up for
in bad times
temporary agencies
2

1.5

1
Labour
0.5 hoarding

0
2 3 4 5 6 7 8 9 10 11 12
Employment

Figure 4.4. Labour hoarding of permanent workers with costly interim

The numerical example in Fig. 4.4 further confirms this point. Using the
same numerical assumptions analysed in the previous section, with the addi-
tional assumption that the mark-up wage for temporary workers under interim
is equal to µ = 0.25, the marginal wage in good times is equal to 1.25, so that
the firm should hire eight workers when business conditions are good. The
interesting dimension is to study what the optimal employment level is in bad
times, and thus the number of permanent and temporary (interim) workers.

4.5 Temporary Contracts with Costly Turnover: Costs

and Benefits
The aim of this section is to consider the trade-off between a temporary and
a permanent contract when turnover is costly. Turnover costs are the costs
associated with the hiring and firing of workers. Turnover costs for workers
hired on temporary and permanent contracts are different, in at least two
dimensions. First, permanent contracts can be terminated at a cost −T while
temporary contracts can be terminated at no cost. In the previous sections we
assumed that permanent contracts last forever. Here we assume that a per-
manent contract can be broken at some cost. Conversely, temporary contracts
can be terminated at no cost as the contract expires. Second, workers under
temporary contracts have a higher probability of leaving the firm, and their
replacement is not necessarily immediate. Under this condition a trade-off
56 TEMPORARY OR PERMANENT?

emerges, and the firm will choose different types of contracts in different
circumstances. The model is described as follows.
There are two periods. A firm has to fill in a vacant job, and there is some
positive expected profit associated with the employment relationship. The key
choice of the firm is whether a temporary or a permanent contract should be
offered. A permanent contract in this context is a two-period contract, or a
long-term contract, while a temporary contract is a one-period contract. The
wage w is the same for both temporary and permanent contracts and it is set
outside the firm. This is a reasonable assumption for most imperfect labour
markets. The productivity in the first period is fixed at y. When the worker is
offered a temporary contract (which lasts one period), the job can terminate at
no cost at the beginning of period 2. When the worker is offered a permanent
contract (which is basically a two-period contract) the firm can terminate the
job at a cost equal to −T . The productivity in the second period is random.
With probability δ the productivity in the second period is very low and the job
needs to be destroyed. With probability (1 − δ) the productivity in the second
period stays at y. Jobs filled by temporary contracts terminate with probability
λ, since workers may quit and find another job somewhere else. If the worker
leaves, the firm has a chance to find a replacement. However, the firm’s ability
to replace the worker has some limit, and we assume that there is a positive
probability that a replacement is not found. We call q the probability that the
firm successfully finds another worker.
We focus on the present discounted value (PDV ) of having workers under
different types of contract. Let’s call J perm the PDV of filling the job with a
permanent job. The point of view that we take is the firm’s, but the problem is
dynamic. The formal expression of J perm reads
−δT + (1 − δ)(y − w)
J perm = y − w + ,
1+r
where in the second period, δ is the probability that the job ends because there
is a negative productivity shock. If the job is destroyed (at rate δ) the firm needs
to pay the cost T . If the job is continued (with probability 1 − δ), the firm nets
y − w in the second period. The PDV of a vacancy filled with a temporary job
is indicated by J temp and its expression reads
(1 − δ)[(1 − λ)(y − w) + λq(y − w)]
J temp = y − w + ,
1+r
where the job is continued only if the productivity in the second period remains
at y (an event which happens with probability 1 − δ). Conditional on the
productivity remaining high, and thus the job continuing, two possibilities
emerge. If the worker does not leave (which happens with probability (1 − λ))
production takes place and the firm nets y − w. If the worker leaves (which
happens with probability λ, the firm nets y − w if and only if the worker is
successfully replaced, an event that takes place with probability q.
 TEMPORARY OR PERMANENT? 57

Let’s define the convenience of the firm to offer a permanent contract as

perm = J perm − J temp
where  is just the net benefit associated with a permanent contract. The
decision rule of the firm is very simple: offer permanent contract if and only
if perm > 0.
Using the definition of J temp and J perm , one has
(1 − δ)(y − w)[λ(1 − q)] − δT
perm = (4.1)
1+r
A few conclusions immediately follow:
Offer a permanent contract if there is no uncertainty (δ = 0).
This is immediately established by looking at equation 4.1. If δ = 0 perm =
(y−w)λ(1−q)
1+r > 0. The key benefit of a temporary contract is to avoid firing costs,
which are paid conditional on productivity being bad in the second period.
As a consequence, when there is no such risk, a permanent contract should be
preferred.
Offer a permanent contract if there are no firing costs (T = 0).
This is just a corollary of the previous remark. If there are no firing costs,
there are no costs associated with a permanent contract, which basically
becomes identical to a temporary contract.
Offer a temporary contract if replacement is immediate (q = 1).
The cost associated with a temporary contract is the risk of forgoing a
profitable opportunity in the second period if conditions are good. Such a risk
fully depends on the probability of finding a replacement if productivity is high
and the worker leaves. Obviously, if another worker is immediately available
(an event which implies that q = 1) a temporary contract is to be preferred.
Formally, one can see that if q = 1 perm = −δT1+r < 0.
The role played by other parameters of the model (y, w, r, λ) can also be
discussed with a comparative static perspective. The following points are easily
established. Offering a permanent contract is more likely when
• profits are higher (y − w is larger);
• probability of quitting is larger;
• interest rate is lower.
To formally establish this result one can differentiate equation 4.1 with
respect to the various parameters.Ë The intuition is as follows. Larger net profits
y − w increase the benefit of a long-term relationship, and reduce the relative
value of firing costs. This suggests that permanent contracts should be more

Ë Notable, ∂
perm (1−δ)λ(1−q)] perm (1−δ)(y−w)(1−q)] perm perm
∂(y−w) = 1+r > 0; ∂∂λ = 1+r ; and ∂∂r = − 1+r .
58 TEMPORARY OR PERMANENT?

likely in high-productivity/high-profits jobs. Further, permanent contracts are

more likely in situations in which the turnover is higher, since the risk of
forgoing profits becomes larger. Finally, larger interest rates reduce the value
of future profits, and thus reduce the incentive to offer a permanent contract.

4.6 A Brief Look at the Evidence

Temporary contracts are now an important feature of regulated labour mar-
kets across the OECD countries in general, and across continental European
countries in particular. In 2002, 11 per cent of the total employment stock was
held by workers hired under temporary contracts. Such a number is certainly
sizeable, but it does not highlight the fact that a larger majority of new jobs
are occupied by workers on temporary contracts. This is confirmed by the
evidence reported in Table 4.3, where we report the incidence of temporary
contracts by different worker characteristics. The incidence of temporary con-
tracts is defined as the share of workers who hold a temporary job conditional
on certain observable characteristics, such as gender, education, and the type
of job. Table 4.3 clearly shows that young workers are much more likely to hold
a temporary contract. One out of four workers aged 15–24 is hired under a

Table 4.3. Incidence of temporary

employment in OECD countries

Gender
Female 12.2
Male 10.5
Age groups
15–24 25.0
25–54 8.0
55+ 9.4
Educational attainment
Low 15.7
Medium 10.4
High 9.3
Industry
Agriculture 21.9
Manufacturing 9.6
Service 10.8
Occupation
White collar 7.7
Pink collar 10.6
Blue collar 9.2
Unskilled 15.3

Source: OECD 2002.

 TEMPORARY OR PERMANENT? 59

Table 4.4. Status of temporary workers in t and t + 1

Status in 1996 Status in 1997 Status in 1999

Permanent Temporary Unemployed Permanent Temporary Unemployed

France Permanent 96.3 1.2 2.6 94.7 1.4 3.9

Temporary 20.8 56.6 22.5 37.9 41.2 20.9
Unemployed 9.5 17.2 73.3 20.9 26.2 53.0
Germany Permanent 92.8 3.3 3.8 92.3 2.5 5.2
Temporary 40.6 36.4 23.0 53.1 22.7 24.2
Unemployed 19.7 14.7 65.7 20.7 15.8 63.5
Italy Permanent 93.1 5.0 1.9 89.3 6.7 4.0
Temporary 41.3 45.9 12.7 52.2 35.2 12.6
Unemployed 8.3 9.3 82.4 15.7 17.7 66.6
UK Permanent 96.4 2.2 1.4 96.5 2.1 1.5
Temporary 56.1 34.5 67.0 27.0
Unemployed 31.4 54.7 46.9 39.3

Source: OECD 2002.

temporary contract, whereas the overall percentage is around 11 per cent. This
is consistent with the view that youth workers entering the labour market are
offered a temporary contract. The fact that women are, on average, more likely
than men to hold a temporary contract is also consistent with this view, since
women have less labour market attachment than men.
Looking at other individual characteristics, it is clear that temporary con-
tracts are more likely among individuals with low educational attachment,
employed in agriculture, and in unskilled occupations.
An important and interesting dimension to consider is to ask what hap-
pens, in a dynamic perspective, to workers hired under temporary contracts.
Are such workers likely to transit into permanent jobs, or are they more likely
to stay in a temporary contract? Table 4.4 throws some light on this question.
The table reports the status of unemployed workers as well as workers hired
on temporary and permanent contracts twelve and thirty-six months after the
first observations. The evidence refers to different European countries. The
table should be read as follows. For each country (say France) the three rows in
the table indicate the employed status of workers in 1996; where the employ-
ment status can be ‘permanent’, ‘temporary’, and ‘unemployed’ respectively.
The columns labelled status in 1997 report the labour market position of such
workers twelve months after the first observation, while the columns labelled
status in 1999 report the labour market status at a three-year distance.
Let’s start with the unemployed workers. In most countries, and France
in particular, unemployed workers are much more likely to find a temporary
contract than a permanent contract. This is obviously consistent with the idea
that labour market entrance takes place through a temporary contract. Looking
at the row on temporary contracts, Table 4.4 suggests that workers hired on
60 TEMPORARY OR PERMANENT?

temporary contracts are much more likely to remain in such a situation than
to switch to a job with a permanent contract, even at a distance of three years
from the first observation.

b APPENDIX 4.1. THE FORMAL DERIVATION OF THE BUFFER STOCK

MODEL

The maximization becomes

 = p[AH log(L perm + L temp ) − w(L perm + L temp )]
+ (1 − p)[AL log L perm − wL perm ].
The first-order conditions are
∂ pAh (1 − p)Al
=0 + =w
∂L perm L perm +L temp L perm
∂ pAh
=0 = pw
∂L temp L perm + L temp
From the second equation we get
Ah
L perm + L temp =
w
while substituting this into the first equation
pAh (1 − p)Al
+ =w
L perm +L temp L perm
(1 − p)Al
pw + =w
L perm
Al
L perm =
w
so that substituting the expression for L perm we get
Ah
L perm + L temp =
w
Ah − Al
L temp =
w

b APPENDIX 4.2. THE FORMAL DERIVATION OF THE BUFFER STOCK

MODEL WITH COSTLY INTERIM AGENCY

The maximization becomes

 = p[AH log(L perm + L temp ) − wL perm − w(1 + µ)L temp ]
+ (1 − p)[AL log L perm − wL perm ].
 TEMPORARY OR PERMANENT? 61

The first-order conditions are

∂ pAh (1 − p)Al
=0 + =w
∂L perm L perm + L temp L perm
∂ pAh
=0 = pw(1 + µ)
∂L temp L perm + L temp
From the second equation we obtain
Ah
L perm + L temp = (4.2)
w(1 + µ)
while substituting this term in the first equation we get
pAh w(1 + µ) (1 − p)Al
+ =w
Ah L perm
Al
L perm = pwµ
w− 1−p

A key result is immediately obtained.

5 Managing Adverse
Selection in Recruiting

5.1 Introduction
Throughout the hiring process, as long as a worker is not fully involved in
the firm’s activities, the firm is likely to have imperfect information about the
worker’s productivity. Such information is more likely to be known to the
worker. This implies that the hiring process is characterized by an asymmet-
ric information problem, since workers know more about their own ability
than the firm does. In the hiring process, the uninformed party is the firm,
while the informed party is the worker. This chapter analyses this asymmetric
information problem in detail, and studies which personnel policies can partly
overcome it.
The firm has a vacant job, and offers the possibility of applying for such
a job to a variety of workers. We are in a situation in which the uninformed
party (the firm) offers an option to the informed party (the worker). In the
economics of imperfect information, each time an uninformed party offers an
option to the informed party, the problem of adverse selection may arise. We
say that a firm has a problem of adverse selection each time the wrong type of
workers are attracted to the firm. Dealing with adverse selection in recruiting
means finding a set of policies that ensures that only the suitable candidates are
attracted to the firm. Dealing with the wrong type of candidates is a very costly
and time-consuming process. Conversely, a properly managed hiring strategy
can save costly resources.
There are two key ways of dealing with adverse selection in hiring. The first
way relies on wage flexibility and on the design of the labour contract. We
distinguish two such types of ‘contingent’ contract: piece rate, and temporary
contracts with a probation period. The second way of dealing with adverse
selection is the use of credentials (specify workers’ characteristics in detail).
We briefly describe these mechanisms.
The first part of the chapter introduces wage policies and contingent con-
tracts, and shows how the availability of wage flexibility can help the firm to
solve the adverse selection problem in hiring. The idea of a contingent con-
tract is to specify the wage rules of the job in such a way that only the ‘right’
type of worker is attracted to the job. We highlight these phenomena with two
types of contract that we label (i) piece rate and (ii) temporary contract with
probation wage.
 MANAGING ADVERSE SELECTION IN RECRUITING 63

Piece rate is the simplest contingent contract, with pay that is strictly based
on output: the larger the output, the larger the total pay. What the firm must
do in this case is to select a wage per piece that will be attractive only to suitable
workers. A piece rate solves the adverse selection problem if ‘good’ candidates
are attracted by the firm wage and ‘bad’ candidates do not apply. The actual
implementation of a piece rate requires output to be perfectly measurable.
In most professional jobs, notably law firms or consulting firms, output is
difficult to measure, and a pure piece rate is a very difficult mechanism to
implement. In such cases firms may use probationary periods for dealing with
adverse selection. The idea of a probationary period is that firms take ‘some
time’ to learn the ability of the workers. High-ability workers are promoted and
tenured, while low-ability workers are dismissed. To deal with adverse selection,
the firm has an incentive to reduce the probationary wage (which is paid to both
suitable and unsuitable workers) and increase the promotion wage (which is
paid only to promoted workers). Under specific circumstances, which depend
also on the firm’s ability to perfectly monitor ‘bad’ workers at the end of the
probation, this multi-period wage can offer a solution to adverse selection.
In imperfect labour markets, the wage flexibility required to implement con-
tingent contracts is often not available to individual firms. As we have argued
several times, a large part of the wage is often decided outside the individual
firm. If this is the case, firms cannot rely on sophisticated wage contracts, and
other instruments need to be implemented. The chapter considers in some
detail the use of credentials, or the possibility of linking the wage offered to
the level of education. The idea explored in the second part of the chapter
is that education, the most widely used credential, can work as a signal of a
candidate’s productivity. Firms do not observe ability, but can and do observe
whether individuals obtain a degree. If the probability of acquiring the signal
(i.e. the probability of graduation) is correlated to the performance on the job,
a signalling equilibrium may arise. In such context, firms are willing to pay a
wage premium to individuals with a degree. In turn, high-ability individuals
have an incentive to undertake the costly signal, in expectation of future wages.
In a signalling equilibrium, the credential works and the problem of adverse
selection is alleviated.
The chapter proceeds as follows. In section 5.2 we analyse the two contin-
gent contracts: the piece rate and the probationary wage. Section 5.3 analyses
credentials, and presents the model of educational signalling in a formal and
an informal way.

5.2 Contingent Contracts

In this chapter we assume that there are workers that are suitable to the firms
and workers that are not suitable. The firm has a vacancy to fill and would
64 MANAGING ADVERSE SELECTION IN RECRUITING

like to avoid applications from unsuitable workers. The productivity of each

individual is not known to the firm, while it is perfectly known to the worker.
The firm only knows that there are suitable and unsuitable candidates, or good
and bad workers, but cannot distinguish the two categories before each worker
is fully involved in the firm’s activity.
How can such a problem be resolved? The idea of a contingent contract is
very simple, and is based on writing a contract that is appealing only to the right
type of worker. If a contingent contract is successful, the problem of adverse
selection is resolved, and only the appropriate worker will come to the firm.
We distinguish between skilled and unskilled workers, where the word
skilled is not necessarily synonymous with educational background. Skilled
workers are those workers that are perfectly fit for the firm’s operation while
unskilled workers are those workers that do not perform very well within the
firm. There are two types of contract that can potentially be used. The first one
is piece rate; the second one is a probationary period. In the next subsection
we explore in turn both types of contract.

5.2.1 PIECE RATE AS A CONTINGENT CONTRACT

Piece rate is the most basic form of contingent contract, since pay is strictly on
the basis of output. A worker that is paid at piece rate is a worker that is paid
proportionally to the amount of output produced.
Let us assume that there are two types of worker, skilled and unskilled, but
let us assume that the firm cannot distinguish between workers who are skilled
for the job and workers who are not skilled for the job until they have physically
worked with the firm. Only when prospective candidates are working in the firm
is the productivity of each candidate revealed. Conversely, as long as workers
are outside the firm there is no way to find out workers’ productivity. Can the
firm use a compensation package to solve this problem? Yes. The idea is to
structure the compensation in a way that is attractive to highly skilled workers
but less attractive to unskilled workers.
The output produced by the skilled workers is larger than the output pro-
duced by unskilled workers. This is basically a definition of being skilled and
unskilled for the job. In other words, let us assume that ys > yu where ys is the
output produced by skilled workers and yu is the output produced by unskilled
workers. The firm knows that some workers produce ys while other workers
produce yu , but can not ex ante tell the type of a particular candidate. Before
proceeding, let us further assume that skilled and unskilled workers have an
outside option, which can be thought of as the value of the best alternative
available to the worker outside the firm. Let us indicate with wu the outside
wage of the skilled workers and with ws the outside wage of the skilled workers.
The table below summarizes the notation.
 MANAGING ADVERSE SELECTION IN RECRUITING 65

A piece rate is a wage per unit of output. Let us assume that the firm can
choose the piece rate. The problem is to select a value of the piece wage wp
so that only skilled workers join. To solve for wage wp the firm should look at
the problem from the point of view of each worker. By properly solving the
problem faced by each different worker, the firm can then select the appropriate
piece rate.

Skilled Unskilled

outside option ws wu
output ys yu ys > yu

We now discuss how the piece wage rate wp must be selected.

Choose wp (the piece rate wage) so that
wp ys ≥ ws skilled workers join
wp yu ≤ wu unskilled workers do not join
This implies that wp should be such that
ws
wp ≥ (5.1)
ys
wu
wp ≤ (5.2)
yu
The piece rate wage should be sufficiently high that skilled workers apply
but low enough that it does not pay for unskilled workers to apply.
Let’s discuss the two conditions of the proposition. We first focus on the
skilled workers. The total income obtained by a skilled worker who takes the
job is wp ys , since wp is the earning per piece and ys is the total number of
pieces produced. If the skilled worker does not work for the firm he or she
earns ws elsewhere. The condition wp ys ≥ ws says that the income earned
by a skilled worker inside the firm must be larger than the outside wage. The
second condition is very similar, but it refers to an unskilled worker. The firm
does not want an unskilled worker to apply for the job, and a piece rate that
satisfies wp yu ≤ wu ensures that the total income earned inside the firm is less
than the total income earned outside. wp solves the adverse selection problem
if both conditions are simultaneously satisfied. Figure 5.1 shows how the two
conditions can be simultaneous satisfied. On the one hand, the piece rate must
be large enough that skilled workers join. On the other hand, the piece rate
must be low enough that unskilled workers do not join. Any wage wp between
these two extremes (included in the two dotted regions) will solve the adverse
selection problem.
66 MANAGING ADVERSE SELECTION IN RECRUITING

Ws/Ys Wu /Yu

Skilled workers join

Unskilled workers do not join

Figure 5.1. Piece rate wage for solving adverse selection in recruitment

Obviously the condition for this scheme to work is that the two regions
in Fig. 5.1 do intersect. This is equivalent to saying that the left-hand side of
equation (5.1) is smaller than the right-hand side of equation (5.2):
ws wu
≤
ys yu
which can also be written as (inverting the ratio and changing the inequality
sign):
ys yu
≥
ws wu
ys ws
≥ .
yu wu
The last condition requires that the proportional increase in the productivity
of skilled workers is larger than the proportional increase in their respective
outside wage. Such a condition is very similar to the condition used for hiring
skilled workers in the case of independent production analysed in Chapter 2.É
Note that the piece rate is a very simple and effective way to solve the adverse
selection problem in hiring. Yet, for it to work at least three key conditions must
be satisfied.
1. Wages must be perfectly flexible and the firm must be able to apply
different wages to different workers.
2. Workers must correctly estimate their outside option. If they do not do
so (i.e. yu = ys ), there is no way you can use piece rate for hiring.
3. Output must be properly measured.
The first condition applies to any contingent contract, not only to piece
rate, and it is a serious limit to the use of these schemes in many imperfect
labour markets with serious institutional constraints. Similar comments can
É Note that if ys /yu = 1 + δ and ws /wu = 1 + γ , the condition is analogous to δ > γ .
 MANAGING ADVERSE SELECTION IN RECRUITING 67

be applied to the second condition. The last condition is probably the most
important, and deserves few comments. Pure piece rate is more likely to be
applied in occupations in which the job output is easily identified. Obvious
examples are sales jobs or basic production activities, such as those found in the
garment industry (the case study on team production of Chapter 13 provides
such an example), or traditional manufacturing jobs. For professional jobs,
such as managing consultants or law firms, piece rates are very difficult to
implement. The next section considers an alternative scenario.

Numerical Example
Assume ws = 10, wu = 5, ys = 5, and yu = 2. The conditions for attracting
skilled workers require wp ∗ 5 ≥ 10, so that wp ≥ 2 and simultaneously
wp ∗ 2 ≤ 5 or a level of wp that is less than 2.5. This means that a piece wage
between 2 and 2.5 would solve the problem. Note that in this case ys /yu = 2.5
which is greater than ws /wu = 2, and a piece rate can be found.

5.2.2 TEMPORARY CONTRACT WITH PROBATION WAGE

Consider a situation in which output cannot be properly measured so that
a piece rate contract is not really feasible. We assume that wage flexibility is
available within the firm, but output is not measurable. There is asymmetric
information still present. While workers know their own ability to perform on
the job, the firm is able to learn the potential of each worker only after some
time.
We assume that the firm is able to assess the productivity of the worker
at the end of the first period. The idea here is to offer a temporary contract
with a low initial wage, but with a confirmation clause that offers a large wage
increase. Good candidates would be attracted by the eventual wage increase,
while bad candidates, knowing ex ante that they will not be promoted, will
prefer to stay away from such a firm.
Contracts of this type are typically envisaged in law firms (or other profes-
sional firms) where young lawyers enter the firm with a temporary contract
and, once promoted, receive a large wage increase. Indeed, it is very difficult to
measure output in law firms, and this scheme seems to be far more appropriate
than a piece rate scheme. Such a scheme is also known in the literature as an
‘up or out rule’ in the sense that young workers either move up the corporate
ladder or simply leave the firm.
To make things as simple as possible, we assume that the job (or the worker’s
career) is described by two periods: a first period in which a temporary contract
is offered at a wage w1 , and a second period (when the worker is tenured) in
68 MANAGING ADVERSE SELECTION IN RECRUITING

which a tenured wage w2 is offered. In reality the second period, when the
worker is tenured, is much longer than the probationary period. While one
can certainly further complicate the problem and take into consideration the
length of the second period, in what follows we work with the simplest scheme.
There are two types of workers, skilled workers and unskilled workers, where
the word skilled refers to the ability of each worker to perform inside the firm.
Skilled and unskilled workers have different outside options, which we indicate
respectively with ws and wu . Finally, we assume that at the end of the first
period, the firm is not perfectly able to recognize a skilled worker. Specifically,
we assume that there is a probability p that an unskilled worker passes the
probationary period and it is tenured inside the firm. There is no discounting
of future wages. To summarize, we consider the following definitions
ws : skilled worker alternative wage
wu : unskilled worker alternative wage
w1 : probationary wage
w2 : post-probationary wage
p: probability that an unskilled worker passes the probationary stage
The problem of the firm is to select w1 and w2 so that only skilled workers
join the firm. We first solve the problem in assuming a perfect monitoring
technology, so that p = 0. In the second part of the section we consider the
problem with a positive probability of ‘wrong promotion’ p.

The Solution with Perfect Monitoring (p = 0)

Let us first consider a situation in which the firm is perfectly able to tell whether
the worker under temporary contract is fit or not. This is akin to assuming that
p = 0, so that at the end of the first period the firm has all the information on
each worker. To attract skilled workers into the firm we need to set the wage in
the two periods so that
w1 + w2 ≥ 2ws (5.3)
This condition is easy to interpret. It simply says that the total wage earned
within the firm by a skilled worker is larger than the cumulative wage earned
outside the firm.
To keep unskilled workers out, it is necessary that
w1 + wu ≤ 2wu (5.4)
The left-hand side is the wage earned inside the firm by the unskilled worker,
and it is the sum of the temporary wage earned in the first period plus the
wage earned in the second period. Since the worker’s contract will not be
transformed into a permanent contract, the wage in the second period is the
 MANAGING ADVERSE SELECTION IN RECRUITING 69

outside wage. The right-hand side is the cumulative wage obtained outside,
and it is simply the outside wage for two periods. The latter equation implies
that to keep unskilled workers out it suffices to have

w1 ≤ wu

or to set a wage under temporary contract lower than the unskilled worker’s
outside wage. Since the temporary wage will be offered also to permanent
workers, the condition (5.3) requires

w2 ≥ 2ws − w1

This condition says that the promotion wage must be larger than the cumu-
lative value of the outside wage (2ws ) net of the first period wage. If we satisfy
the two constraints with equality (in assuming that skilled workers do not
participate if indifferent and skilled workers do participate if indifferent), and
we indicate with w1∗ and w2∗ , the solution to this problem reads

w1∗ = wu
w2∗ = ws + (ws − wu )

The idea of the solution is simple. The first period wage should be as low
as the outside option of the unskilled workers, so that the job would not be
suitable for unskilled workers (who will be fired for sure at the end of the first
period). The confirmation wage conversely should be larger than the outside
option of the skilled worker. Indeed, in the second period a skilled worker
should get his outside wage, plus a compensation premium for the loss of
income obtained in the first period. This suggests that the high wage enjoyed
by the skilled workers in the second period corresponds to a compensation for
their income loss during the first period. Let’s define the pre-post probation
wage differential  as the difference between the wage in the second period and
the wage in the second period, so that  = w2∗ − w1∗ . The following applies:
The pre-post probation wage differential wage  increases with the outside
skilled differential ws − wu .
This is immediately confirmed by substituting in the definition of  the
expression w2∗ and w1∗ . It follows that  = 2(ws − wu ). The larger the outside
skilled differentials ws − wu the larger the compensation that the firm must
provide to the skilled worker in the second period.
An extreme form of this type of contract is the internship, a type of contract
that is often offered to just-graduate students. The idea of an internship is that
the worker can work for the firm for wage equal to 0 for an initial short period,
and at the end of the internship obtain a large wage as a permanent contract.
The firm can then set a first period wage equal to w1 = 0 and a second period
70 MANAGING ADVERSE SELECTION IN RECRUITING

wage such that

w2 ≥ 2ws
Unskilled workers will not accept the internship, since they know that they
will not be promoted, while skilled workers do accept and wait for the large
wage increase that comes with promotion.
The situation described is very simple, but it requires the firm to be perfectly
able to identify unskilled workers at the end of the temporary contract. Such
a condition is not automatically obtained, and it is important to study what
happens when the firm errs in the promotion mechanism. This problem is
taken up in the next subsection.

The Solution with Imperfect Monitoring (p < 1)

We now solve the problem by explicitly considering the case in which the firm
does err in the confirmation process, since there is a finite probability that an
unskilled worker is promoted. The condition for attracting skilled workers is
unchanged and it reads
w1 + w2 ≥ 2ws (5.5)
where the cumulative wage inside the firm must be larger than the cumulative
outside option. The condition for keeping unskilled workers out of the firm
is more complicated, since we need to take into account the probability p of a
‘wrong promotion’. The formal expression reads
w1 + pw2 + (1 − p)wu ≤ 2wu (5.6)
where the left-hand side is the cumulative expected wage inside the firm and the
right-hand side is simply the cumulative expected wage outside the firm. The
key novelty in equation 5.2 is the wage in the second period, which is obtained
as a weighted average between the promotion wage w2 and the outside wage
wu , with weights equal to the probability of ‘wrong promotion’ p. Indeed,
with probability p the bad worker is promoted and he or she enjoys the second
period wage w2 . Conversely, with probability (1 − p) the worker is detected, is
fired, and he or she enjoys the outside option wu .
If we satisfy the two constraints with equality (in assuming that skilled
workers do not participate if indifferent and skilled workers do participate if
indifferent), and we indicate with w1∗ and w2∗ , from (5.5), it follows that to keep
unskilled workers out the maximum unskilled wage is
w1∗ = 2ws − w2
and substituting this into equation (5.2) satisfied with equality one obtains an
equation for the minimum unskilled wage w2∗
2ws − w2∗ + pw2∗ + (1 − p)wu = 2wu
 MANAGING ADVERSE SELECTION IN RECRUITING 71

so that
2ws − (1 + p)wu
w2∗ =
1−p
from which it follows that the probationary wage is
2ws − (1 + p)wu
w1∗ = 2ws −
1−p
2ws − 2pws − 2ws + (1 + p)wu
=
1−p
(1 + p)wu − 2pws
=
1−p
The probationary wage is such that w1∗ < wu < ws and w2∗ > ws . In other
words the probationary wage requires that the initial wage is lower than the
outside unskilled wage while the post-probationary wage is larger than the
outside skilled wage.
First let’s check that w1∗ < wu . The latter condition is true as long as
(1 + p)wu − 2pws
< wu
1−p
(1 + p)wu − 2pws < wu − pwu
−2pws < −2pwu
ws > wu QED

Second, we need to check whether w2 > ws . The latter condition is true as

long as
2ws − (1 + p)wu
> ws
1−p
2ws − (1 + p)wu > ws − pws
ws > wu QED

Let’s define the pre-post probationary wage spread as  or, in other terms,

 = w2∗ − w1∗
2ws − (1 + p)wu (1 + p)wu − 2pws
= −
1−p 1−p
2(1 + p)(ws − wu )
=
1−p
72 MANAGING ADVERSE SELECTION IN RECRUITING

As p rises, it is difficult to detect unskilled workers, and a larger spread  is

needed to keep unskilled workers out.
To see this you need to differentiate  with respect to p to yield
∂ 2(ws − wu )(1 − p) + 2(1 + p)(ws − wu )
= >0
∂p (1 − p)2
It follows that the wage contract (w1∗ , w2∗ ) must simultaneously satisfy
(1 + p)wu − 2pws
w1∗ ≤
1−p
w2∗ ≥ 2ws − w1∗
So that once w1∗ is chosen small enough, one must choose the second period
wage accordingly.
While the probationary wage seems a very effective compensation system for
solving adverse selection issues, there are some important issues to mention.
A potential problem is what is known in the literature as rat race, or the fact
that effort and ability are likely to be substitutes. Workers put in a lot of effort
during probation, so that a firm is likely to observe bad workers working very
hard. This problem increases the chance of a mistake. In addition, effort is
likely to drop after getting a fixed job.

Numerical Example
Let us assume that wu = 2 and ws = 5 while the probability of passing the test
is p = 0.1. It is easy to check that the two conditions are
1.1 ∗ 2 − 2 ∗ (0.1) ∗ 5
w1∗ ≤
0.9
∗
w1 ≤ 1.33
w2∗ ≥ 10 − w1∗
so that if we offer w1∗ = 1.33 the corresponding w2∗ = 8.66. If the first period
wage is reduced (let’s say to w1∗ = 1) the second period wage must be increased
to 9.

5.2.3 CONTINGENT CONTRACTS WITH MINIMUM WAGES

The idea of a contingent contract, either a piece rate or a probationary period,
is to set initial wages that are so low that the application is unattractive to
unsuitable candidates. When a minimum wage is imposed on the firm, it is no
 MANAGING ADVERSE SELECTION IN RECRUITING 73

longer perfectly obvious that a firm will be able to offer contingent contracts.
We now try to understand under which circumstances a contingent contract
with minimum wage is a viable option. We analyse this issue in the case of a
probationary period, under the assumption that the minimum wage is binding,
so that the wage of the first period must be equal to a minimum wage w̄. The
question we ask is whether it is still possible to have a wage profile w̄, w2 so
that unskilled workers do not join the firm. We initially assume that the best
alternative option of an unskilled candidate is larger than the minimum wage
(wu > w̄) and we analyse the alternative case at the end of this section. What
we want to show here is that a probationary scheme will work only if the
monitoring technology is sufficiently good, so that the probability of a wrong
promotion is sufficiently small. Let’s see the key conditions.
To attract skilled workers it is necessary that
w̄ + w2 ≥ 2ws
where the first period wage w1 is now identical to the minimum wage w̄. To
keep unskilled workers out, it is necessary that
w̄ + pw2 + (1 − p)wu ≤ 2wu
which is very similar to equation 5.2, with the only key difference being that
the first period wage is no longer an endogenous variable of the firm, but it is
set equal to w̄.
If we satisfy the first constraint with an equals sign (so as to make the
skilled workers just indifferent between working for the firm and their outside
option), the promotion wage is
w2 = 2ws − w̄.
Substituting this requirement into the second equation we see that the
condition is satisfied if
wu − w̄
p<
2ws − [wu + w̄]
from which immediately follows:
1. Only if the risk of wrong promotion (i.e. promoting an unskilled worker)
is sufficiently small (p < p ∗ ) does a probationary period with minimum
u −w̄
wage solve the adverse selection problem, where p ∗ = 2ws w−[w u −w̄]
.
2. If the outside option of the unskilled worker is exactly equal to the min-
imum wage wu = w̄, there is no way to use a probationary contract to
solve the adverse selection problem.
The last remark shows that wu > w̄ is a necessary condition for the pro-
bationary period to work. Indeed, if the outside wage is smaller than the
74 MANAGING ADVERSE SELECTION IN RECRUITING

minimum wage, the probationary period requires a negative probability, a

situation which is obviously unfeasible (the probability must obviously be
between 0 and 1).
It is clear that the existence of a minimum wage greatly diminishes the
firm’s ability to use the probationary wage as a mechanism for solving the
adverse selection problem in the recruiting process. In the next section we
study situations in which firms do not use a wage contract but rely on signals
and credentials.

5.3 Use of Credentials and the Signalling Model of

Education
The idea of a credential is to specify in the job requirement specific conditions
that prospective candidates should possess for qualifying for the job. Obvious
examples of credentials are college degree, master’s, professional certificates.
Quite often, in the real world, job advertising explicitly specifies that successful
candidates must possess certain credentials, and very often such credentials
refer to educational attainments.
In this section we analyse whether the use of credentials can mitigate the
adverse selection problem. Our analysis will focus mainly on education as a sig-
nal, but it is clear that similar analysis applies to other credentials. The concept
we want to stress is the fact that under certain specific conditions a creden-
tial can profitably be used to signal a candidate’s productivity. When is such
signalling likely to work? We will see that two key conditions are necessary:
1. The credential must be correlated with job performance.
2. Obtaining the credential must be relatively easy for well-qualified work-
ers, but difficult for others. In other words the ability to perform well
in school and perform on the job are highly correlated (analytically ori-
ented jobs). The idea is that people have private information about their
own ability and they know whether the benefit of having the credential
is worth its costs.
We now analyse in some details the idea of education signalling.

5.4 Education Signalling

This section studies under which conditions the educational attainment of the
candidates may serve to signal unobservable individual ability. In this section
 MANAGING ADVERSE SELECTION IN RECRUITING 75

we assume that education attainment has no impact on individual ability, and

it is not linked to a process of skill acquisition. In reality, skill acquisition is
a dimension of the educational experience, and the chapters on training and
human capital investment will be fully devoted to such key issues in personnel
economics. In this section we are interested in the signalling value of education,
and we fully explore such a mechanism.
We thus assume that schooling makes no independent contribution to
labour productivity, which is instead determined solely by individual ability.
One may reasonably think that under these circumstances prospective employ-
ers have no direct interest in the educational attainment of workers, and profit
maximizing firms would not be willing to pay a wage premium to workers
with more schooling. Things may be very different if individual ability is not
observed by employers.
Let’s assume that individual ability is invisible to employers, and let us also
assume that individual ability is inversely correlated with the cost of schooling.
Employers would like to know the ability of potential employees, since more
capable employees are more productive inside the firm. Yet, firms are incapable
of assessing a worker’s ability. The key idea is that firms may be able to fully
learn a worker’s ability from his or her educational attainment. When such a
condition is satisfied we say that the labour market has attained a separating
equilibrium.
Let us assume that there are two types of workers, those with high ability
and those with low ability. Let us assume that there are 50 per cent of workers
that are low ability and 50 per cent of workers that are high ability. The para-
meter a describes the workers’ ability (and their marginal productivity in the
workplace), and we assume that

ah if worker has high ability

a=
al if worker has low ability

Clearly, high-ability individuals are more productive than low-ability indi-

viduals so that ah > al . Let’s consider a situation in which these two different
individuals can or cannot take a costly educational attainment. We simply
assume that schooling s is a binary variable that can take values 0 or 1,
depending on whether the individual decides to acquire education. We should
think that s = 1 corresponds to having a master’s degree while s = 0 cor-
responds to a situation without a master’s degree. Undertaking education
involves psychological costs, which we assume to be described by this simply
relationship

k
c(s = 1, a) =
a
76 MANAGING ADVERSE SELECTION IN RECRUITING

where k is a positive constant. Since a can take only two values, the cost function
is equivalent to

c(s = 1, ah ) = ak if worker has high ability

h
c(s = 1, al ) = ak if worker has low ability
l

The previous conditions imply that high-ability students have lower costs of
education, so that
cost of education for low ability > cost of education for high ability
Education attainment is observable to the firms. While firms are not able to
offer contracts that are linked to the ability of the candidates, they are able
to offer contracts that are linked to the educational attainment. A separating
equilibrium is a situation in which only high-ability individuals acquire the
educational signal and firms offer wage premia to educated individuals. If this
situation happens in the marketplace, then the use of credentials acts as a
device that solves the adverse selection problems of firms.
To see how a separating equilibrium may emerge suppose that firms expect
that only high-ability workers receive education and so they are willing to offer
a high wage to a candidate that has attained education. When this is the case
high-ability workers may have an incentive to acquire the signal in exchange
for future wage gains. Conversely, low-ability workers find such an option too
costly, and do not acquire the educational signal.
Figure 5.2 describes a signalling equilibrium. In the horizontal axis there is
the education level. Note that only two points are relevant, the origin s = 0
and the point s = 1, since we are assuming that only two types of education
are available in the marketplace. In the vertical axis we report the wage paid
for each educational level as well as the cost of education. The wage offered
by the firms is w = ah for individuals with high education and w = al for
individuals with low education attainment. The upward sloping line reports
the cost of education for the two types of workers. Low-ability individuals
have a steeper cost function, since acquiring the education signal involves a
larger cost.
How will each individual decide whether investing in education is worth
the cost? Let’s consider low-ability individuals first. If they do not acquire the
signal they enjoy a net return equal to the segment Ao in the figure, since
point A corresponds to the wage wl and the cost of education is zero. If they
acquire the signal they will get a wage wh but will have to pay the cost up to
point C so that the net value of utility will be equal to the segment BC. Since
the segment Ao is larger than BC, low-ability individuals will choose s = 0.
Let’s now consider the high-ability individual. By a similar argument the high-
ability individual should compare the segment BD (which is the value when
 MANAGING ADVERSE SELECTION IN RECRUITING 77

Wage,
cost of education
Cost for
Slope low ability
k /al
B

WH = ah

C Cost for
high ability
A Wl = alow
D
o
S=1 s, educational
level
Slope
k /ah

Figure 5.2. Education signalling and a separating equilibrium

he or she acquires education) with the segment Ao. It is clear that high-ability
individuals select the education signal. In other words, Fig. 5.2 describes a
separating equilibrium. Individuals do acquire the credentials and the firms
value such a signal in equilibrium. Under these circumstances the firm is able
to fully learn the worker type.
Things can be very different if the firm believes that an individual with
a degree has the same chances of being high or low ability. In other words,
the firm may expect that the possession of the credentials has no link to the
probability of finding a good candidate. Under such circumstances the firm
will post only a single wage and there will be no wage premium linked to better
educational attainment. This is called a pooling equilibrium, a situation that
is described in Fig. 5.3. In this situation, both individuals have no incentive
to acquire the costly signal, so that both types (high and low) choose not to
acquire the education signal. This discussion highlights the fact that firm’s
beliefs play a key role in shaping the emergence of a separating equilibrium.
The next section explicitly considers firm profits and the role played by firm’s
beliefs.Ê

Ê Note further that the nature of the separating equilibrium lends support to the claim that education
per se is useless or even pernicious, because it imposes social costs but does not increase total output.
Of course such a view is extreme, and education has a lot to do with human capital accumulation, as
Chapters 9 and 10 on training will show.
78 MANAGING ADVERSE SELECTION IN RECRUITING

Wage,
cost of education
Cost for
low ability
k/al

F Cost for
high ability
W
D
G
S=1 s, educational
o
S=0 level
k/ah

Figure 5.3. Pooling equilibrium (education does not act as a signal)

5.5 Education Signalling: A Formal Story

This section provides a formal analysis of the signalling equilibrium described
in the previous section. In game theory jargon, signalling is a way for an agent
to communicate his type under adverse selection. The idea is that education
has no direct effect on a person’s ability to be productive, but is useful for
signalling his ability to employers. Let half of the workers have the high-ability
type less and half of the workers low ability, where ability is a number denoting
the dollar value of their output. Employers do not observe the worker’s ability,
but they do know the distribution of abilities, and they observe the worker’s
education. To simplify the exposition, we will specify that the players are one
worker and two employers. The competition of employers will drive profits
down to zero, and the worker receives the gains from trade.
The worker must make two choices: the education level and his choice of
employer. The employers’ strategies are the contracts they offer giving wages as
functions of education level. The key to the model is that the signal, education,
is less costly for workers with higher ability.

5.6 Rules of the Game

Players:

A Worker and two employers

 MANAGING ADVERSE SELECTION IN RECRUITING 79

The Order of Play:

1. Nature chooses the worker’s ability a ∈ {2, 5.5}, the low and high ability
each having probability 0.5. In other words al = 2 and ah = 5.5. The
variable is observed by the worker, but not by the employers.
2. The worker chooses education level, either no education or education, so
that s can take only two values s ∈ {0, 1}. Education is observed by both
the firms and the workers.
3. Each employer offers a wage contract w(s), specifying the wage to be paid
as a function of the educational attainment.
4. The worker accepts a contract, or rejects both of them.
5. Output equals a.
Pay-off:
The worker’s utility is his wage minus his cost of education, so that
8s
Uw = w − if the worker accepts contract w(s)
a
Uw = 0 if the worker rejects both contracts
The expected profits are output minus wage costs so that
=a−w for the employer whose contract is accepted
=0 for the other employer
There are two equilibria which are of interest, one in which education is not
chosen in equilibrium, one in which education is selected only by the high-
ability type. We now focus on the two equilibria.

5.7 No Education Equilibrium (Pooling Equilibrium)

The first equilibrium we consider is a no education equilibrium, or a pooling
equilibrium in game theory jargon. In such an equilibrium both types (low
and high ability) choose no education, and the employer pays a salary which
is equal to the average productivity. The idea is that employers believe that an
educated worker has the same chance of being either high or low type, so that
they do not offer a wage premium to an educated worker. Given the employers’
beliefs, it does not pay for the workers to engage in education. Let’s first define
the average wage as the expected productivity, or as
w̄ = pah + (1 − p)al
w̄ = 0.5 ∗ ah + 0.5 ∗ al
w̄ = (2 + 5.5)/2 = 3.75
80 MANAGING ADVERSE SELECTION IN RECRUITING

Let’s define the employers’ beliefs as the probability that the type is low,
conditional on being educated as p(a = Low|s = 1), and let us assume that

p(a = low|s = 1) = 0.5

This means that conditional on having acquired education, the probability of

being of low type is the same as that for the population at large.
The no education equilibrium is such that

s(low) = s(high) = 0
w(0) = w(1) = 0.5
p(a = Low|s = 1) = 0.5

This implies that education is useless, since it does not yield any wage premium.
In this condition, the model reaches the unsurprising outcome that workers do
not bother to acquire unproductive education. Note that under this scenario
the beliefs are simple passive conjecture, since s = 1 is never observed in
equilibrium. It is clear that for the high-ability type it does not make sense for
him to educate himself, since the employer would not reward his effort through
higher wages. Further, firms make zero expected profits since the average ability
is equal to the average wage.

5.8 Education Equilibrium (Separating Equilibrium)

Let us now assume that the employer has a different set of beliefs. In particular,
let us assume that the employer believes that a worker who acquires education
is a high-ability type, so that

p(a = Low|s = 1) = 0

This leads to the separating equilibrium for which signalling is best known, in
which the high-ability worker acquires education to prove to employers that
he really has high ability. The separating equilibrium is such that

s(Low) = 0
s(High) = 1
w(0) = 2
w(1) = 5.5

How can we check that this is really an equilibrium? We need to check that the
separating contract maximizes the utility of the high and low types subject to
 MANAGING ADVERSE SELECTION IN RECRUITING 81

two constraints:
Participation constraints that the firms can offer the contract without
making losses;
Self-selection constraints that the lows are not attracted to the high contract,
and the highs are not attracted by the low contract.
The participation constraints for the employer requires that
w(0) ≤ aL = 2
w(1) ≤ aH = 5.5
but competition between employers makes the expressions above hold as
equalities, so that the employers make zero profits.
The self-selection constraint of the lows is
UL (s = 0) ≥ UL (s = 1)
8∗1
w(0) − 0 ≥ w(1) −
2
2 ≥ 5.5 − 4
2 ≥ 1.5
which is satisfied.
The self-selection constraint of the highs is
UH (s = 1) ≥ UH (s = 0)
8∗1
w(1) − ≥ w(0) − 0
5.5
5.5 − 1.45 ≥ 2
4.05 ≥ 2
which is also satisfied.
Separation is possible because education is more costly for workers if their
ability is lower. If education costs the same for both types of worker, education
would not work as a signal, because the low-ability workers would imitate the
high-ability workers. The nature of the separating equilibrium lends support
to the claim that education per se is useless or even pernicious, because it
imposes social costs but does not increase total output.
6 Optimal Compensation
Schemes: Foundation

6.1 Introduction
Choosing an appropriate compensation mechanism is probably the core prob-
lem of human resource managers, and represents the heart of personnel
economics. Loosely speaking good compensation packages must be consistent
with profit maximization on the part of firms, but they should also provide
workers with the incentives to do as well as possible. If we pay the worker
independently of his or her performance, most likely we will not provide him
or her with the right incentives to act in the interest of the firm. In the jar-
gon of the economics of imperfect information, the problem with a fully fixed
salary is one of moral hazard. The term moral hazard arises in the insurance
market and refers to the fact that a person who has insurance coverage will
have less incentive to take proper care of insured objects than a person who
does not.
Here is an example. A worker who sells computers for a small company has
to decide how much effort to put into selling the hardware. The number of
computers sold certainly depends on how hard she works, but the total sales
will also depend on other factors, like customer willingness to buy computers.
Such factors are outside the worker’s control. At the end of the month the
company will observe the number of computers sold, but will not observe how
much effort she put into the task. A very good month’s volume of sales can
certainly depend on hard work (we will call it effort in this chapter) but can also
simply depend on a surge in demand totally unrelated to effort. The question
we need to ask is, how much effort will the worker put into her task? Such
effort will obviously depend on the structure of compensation. If the worker is
paid a fixed salary independent of the computers sold, she is much more likely
to take it easy. This is because with a fixed salary the worker is fully insured
against changes in the number of computers sold. Indeed, if the pay does not
depend on how hard a worker tries, why should she bother to exercise effort?
It seems that under straight salary (which corresponds to full insurance) the
worker, who in this chapter is called the agent, is doing something that the boss
would rather not do. Obviously, it is very difficult for the firm to prove that
 OPTIMAL COMPENSATION SCHEMES 83

the worker did not work enough by simply looking at sales, since it is also
possible that the low sales were due to low customer demand. This conflict of
interest is at the heart of the moral hazard problem. What is good for the agent
is not (necessarily) good for the principal, and the principal cannot always be
on hand to monitor what the agent does.
The only way in which the firm can solve this problem is by choosing
an appropriate compensation scheme. Pay for performance, or compensation
packages that link the worker’s salary to the number of computers sold, are
likely to be the solution to the problem. In this chapter we will find that there
exists an extreme incentive scheme that is likely to be the solution to our
problem. In the extreme scheme the worker ends up making a fixed up-front
payment to the firm at the beginning of the month, and thereafter acts as a
sort of entrepreneur. In other words, the worker ‘buys’ the job from the firm,
and then acts as a residual claimant on the revenues obtained. This scheme
corresponds to a labour contract with a 100 per cent commission rate. The
salesman keeps all the revenues from the sales, and pays the firm a fixed fee.
Yet further problems exist. The first problem is whether such a scheme is
technically feasible. It is often impossible for the firm to obtain a payment from
the worker. Negative payments are simply not allowed. We will see that in this
case the firm will have to change its optimal compensation scheme.
Second, the discussion so far has not dealt with an important factor: risk.
The worker is likely to be risk averse, and she is likely to prefer income streams
that are time invariant to income streams that change dramatically from month
to month. We will see that also in this case, the optimal compensation scheme
will have to be changed slightly.
We can summarize the problem of the compensation package with the
following key questions.
The principal/firm has two problems:

Principal/firm problem Number 1: Of all the possible compensation pack-

ages that the principal could offer the agent, which one maximizes the
expected profits?
Principal/firm problem Number 2: Once the optimal compensation is
found, is it worth hiring the agent?

The worker/agent has two problems

Agent/worker problem Number 1: If the agent accepts the contract and goes
to work for this principal, how hard should she work on that job?
Agent/worker problem Number 2: Once she knows how hard it is optimal
to work, should she accept such a contract?
84 OPTIMAL COMPENSATION SCHEMES

6.2 A Formal Principal–Agent Model

6.2.1 THE GENERAL SETTING

We assume that the agent is a worker who likes higher income w and dislikes
effort. The agent quite likely dislikes income variability, so that at a given
average income, she is happier if income is more stable. We take up this issue
in section 6.4 of this chapter. In this section we assume for simplicity that the
agent is risk neutral and does not worry about income variability.
The job of the agent involves producing a good in quantity x which is sold
for a fixed price equal to p. The total revenues generated by the workers are
indicated by px, and depend on two key dimensions: luck and effort. Greater
effort (which we indicate with e) on the part of a worker implies greater output
produced and sold x, but larger x can also be the result of good luck or favourite
demand conditions. The principal observes only output px and does not know
whether the larger output x is the result of luck or the result of effort. Effort
e is unobservable, and output x is an imperfect signal of the effort elicited
by the agent.
The expected output generated by the worker is denominated by x and
depends on luck and effort from the following relationship

x = (e + η),

where e is effort generated by the worker and η is the outcome of a random

variable with zero mean (E(η) = 0) and variance equal to v. Greater effort on
the part of the worker implies greater average output x̄, but output depends
also on other factors, which can be thought of as customer demand, or luck on
the part of the worker. These other factors, however, cancel out on average. This
means that, on average, output is equal to the effort exercised by the worker
so that

x̄ = E(x) = e

In other words, the random components of output are independent of effort,

and effort determines average expected output.É Given the total revenues, we
now specify the compensation package.

É A more formal way to describe what was said in the text it is to assume that

η is a random variable with zero mean and variance v

x is randomly distributed with mean e and variance equal to v
x ∼ f (x) : x̄ = e var(x) = v
 OPTIMAL COMPENSATION SCHEMES 85

A compensation package offered to the worker is a wage contract that specifies a fixed wage
component, independent of revenues, and a variable wage component.2

We indicate the compensation package with w, where w is a linear function of

the revenue produced by the agent. In particular, we let
w = α + βx
w = α + β(e + η)
where α and β are two constants. So that the expected value of the
compensation is
w e = α + βE(x)
= α + βe
since E(η) = 0. Basically a compensation package offered by the principal
is a wage contract specifying two constants: α, a fixed wage component,
independent of revenues, and β a variable wage component.Ë
It is important to stress the difference between the wage w and its expected
value w e . The compensation wage w depends also on the outcome of the ran-
dom variable η, while the average compensation w e depends only on the effort
exercised. This makes a lot of sense. One month of high pay can be due to luck,
but on average good salaries must be linked to good performance.
One important dimension to emphasize is also that the firm observes the
output x but does not observe the effort chosen by the worker. The firm can tell
whether output produced was large or low, but it cannot really tell whether a
high output was obtained because of high effort or a large realization of η. Such
a distinction will turn out to be very important when we consider risk aversion.
There are basically three types of compensation scheme that we consider:
fixed compensation schemes, bonus schemes, and franchising schemes.
1. Purely Input Based Scheme (α > 0 and β = 0). This scheme is the
simplest one, and specifies a fixed payment per unit of time (say a month,
or a week, or a day) independently of the output produced and sold. In
formula, the compensation package can be written as
w=α
so that the wage income is completely independent of output.
Ê Note that the analysis considers only linear contracts, or contracts in which the wage is a linear
transformation of output.
Ë The compensation package has also a variance which is simply given
Var(w) = β 2 Var(x)

= β2v
The variance will play an important role when we introduce risk-averse workers.
86 OPTIMAL COMPENSATION SCHEMES

We
W e = a +pe;

W e = a + be; b < p

W e= a

Figure 6.1. Different compensation schemes as worker’s constraint

2. Bonus Scheme (0 < β < p), so that the compensation scheme is made
up of a fixed component plus a variable bonus, which is proportional
to output. The larger the coefficient β, the larger the sensitivity of the
compensation scheme to the revenues obtained.
3. Franchising Schemes (β = p; α < 0). In this scheme all the extra output
is given to the worker, so that he becomes a residual claimant in the
project. Under this pay scheme, the agent is an entrepreneur who rents a
job, and the firm makes money by choosing the fixed payment α.

The compensation schemes can also be represented graphically by a line in

the space [w e , e], where w e is the expected wage and e is effort (Fig. 6.1). Each
compensation scheme is described by a line in the space [w e , e], with different
schemes being just different lines with different slopes and different intercepts.
The intercept corresponds to the fixed part of the compensation while the slope
corresponds to the bonus. In Fig. 6.1, the horizontal line is a purely input-based
scheme while the two upward-sloping lines are the two incentive schemes, with
the slope varying according to the different degree of incentives.

Utility function. We already mentioned that the agent is a worker who

likes higher average income w e but dislikes effort.Ì A risk-neutral agent
is somebody who does not care about income variability and his utility
function is
e2
U (w e , e) = w e − δ
2
Ì As we argued above, we are now sidestepping from risk consideration and we assume that the
agent does not care about larger wage variability. We take up this important issue in section 6.4.
 OPTIMAL COMPENSATION SCHEMES 87

where δ is simply a parameter that reflects how much the worker dislikes
effort. In the utility function, 1 additional euro of income yields one addi-
tional utils, while an additional unit of effort yields a disutility equal to
δe. This implies that the marginal utility of income is constant, while the
marginal disutility of effort is increasing. At the margin, increasing effort
becomes more and more costly to the individual, while additional income
yields the same utility level.Í
Outside option. The agent has always the possibility of turning down the
principal’s job offer and working at another job (maybe as self-employed)
that yields utility level u ≥ 0.
Firm profits. The principal is a firm that maximizes expected profits. We
assume that the principal has only one worker and that there are no other
costs beyond labour costs. The expected profits can be written as expected
revenues minus expected costs, or
E[] = pE[x] − w e
= (p − β)e − α
where we make use of the definition w e = α + βe and E[x] = e. In the
previous expression p−β is the profit per unit of output sold, E[x] = e is the
expected revenues and α represents the fixed cost of the firms, independent
of output. Two remarks are in order. First, since p − β is the profit per unit
of output, the term (p − β)e can be thought of as the variable profits of the
firm. Second, the fixed cost α can obviously become a fixed revenue, since
we argued above that α can be negative.

Indifference Curves
Let’s call the indifference curve of the worker the combination of expected
income (w e ) and effort (e) that delivers to the individual the same level of
utility.Î (Figure 6.2 reports a family of indifference curves.) There are various
properties of these utility functions:
• Indifference curves are upward sloping. Higher effort, which the worker
dislikes, must be compensated by higher income in order to remain on
the same indifference curve.
• Indifference curves do not intersect. This is true with any indiffer-
ence curve. If they did intersect, then a unique combination of effort
and income would yield two different utility levels, which is obviously
not possible.
Í The indifference curves describe the behaviour of the utility function.
Î The equation of the indifference at utility level Ū for an individual that is risk neutral is
e2
E(w) = δ + Ū
2
88 OPTIMAL COMPENSATION SCHEMES

U2
E (w) U1

Figure 6.2. Worker’s indifference curve

• Indifference curves in the north-west region of the diagram correspond

to indifference curves with larger utility levels. In Fig. 6.2, the indiffer-
ence curve U2 corresponds to a larger indifference level than the curve
labelled U1 .
• The indifference curves are also convex, since more and more effort must
be compensated by an ever larger income. This gives the idea that the
marginal disutility of work is increasing. Formally, such a property comes
from the fact that effort enters the utility function with a square term.
We can also derive the slope of the indifference curve. We need first to
define the marginal utility associated with extra income and the marginal
utility associated with an extra unit of effort. Our utility function implies that
the marginal utility of an extra euro of wage income is 1 (which is exactly what
our indifference curve would say), so that


U


MUw = =1
w e
e=ē
The notation e = ē in the previous expression simply says that the marginal
utility of income is calculated holding constant effort. We can similarly define
the marginal (dis)utility of effort as the loss of utility deriving from an extra
unit of effort as


U


MUe = = −δe
e
w e =w̄
so that our utility function implies a marginal disutility that is increasing in
effort. In other words, the larger the effort that the worker exercises, the higher
the disutility associated with a further increase in effort. Having defined the
marginal disutility along an indifference curve the total change in utility must
 OPTIMAL COMPENSATION SCHEMES 89

be zero by definition so that

U

U


e + w = 0
e
E(w)=w̄ E(w)
e=ē

Rearranging the terms of the previous expression, the slope of the utility
function is
w MUe
=− = δe
e MUw

6.2.2 THE PROBLEM OF A RISK-NEUTRAL AGENT

The agent chooses effort after observing the compensation scheme. The agent
faces basically two problems/questions:
Agent Problem Number 1: If I accept the contract and go to work for this
principal, what level of effort should I choose?
Answer: I should choose the level of effort e that maximizes my expected
utility (a level of effort labelled e ∗ ).
Agent Problem Number 2: If I take the contract and choose the optimal level
of effort e ∗ , I will obtain an expected wage w e (e ∗ ) that will certainly depend
on the effort exercised. Given the effort e ∗ and the compensation w e (e ∗ ),
the expected utility will be U (w e (e ∗ ), e ∗ ). Shall I accept such a scheme?
Answer: I will take the principal’s offer only if it ensures a utility level which
is as high as the outside utility level u.

The Optimal Effort Level: The first problem of the agent

Effort is chosen so as to maximize the agent utility. Let’s analyse the level of
effort under the various contracts.
Effort under Purely Input Scheme: In the purely input scheme the com-
pensation is w = α so that w e = α. The problem of the worker is to find
the highest indifference curve given the constraint represented by the fixed
payment. The solution to such a problem is easily obtained graphically. It is
clear that the maximum indifference curve is given by the point A in the corner
solution in Fig. 6.3, so that the optimal effort is

e ∗ (purely input scheme) = 0

In other words, when wages are fixed and independent of output, the agent
optimally chooses zero effort.
Effort under the Bonus Contract: In the bonus contract w = α +βx so that
w e = α + βe. The optimal level of effort can be easily analysed graphically, as
90 OPTIMAL COMPENSATION SCHEMES

We
U1
W e = a + be; b < p

U2
B

A W e= a

e* = b/d e

Figure 6.3. The worker problem

the choice of an optimal level of effort on the part of the worker. The worker
wants to enjoy the maximum level of utility subject to the budget constraint.
The maximum point requires a tangency condition between the indifference
curve and the budget constraint. If the budget constraint is horizontal, then
clearly the worker maximizes his utility by choosing an optimal level of effort
at point A, to which corresponds a zero amount of effort. Conversely, if the
budget constraint is the one typical of a bonus scheme, then the optimal
position cannot be at zero effort, since the worker can obtain a larger level
of utility by choosing a larger amount of effort. The optimal solution of the
worker is now at point B, where the level of effort is the one corresponding to
the level of effort e ∗ = βδ . Let’s understand why.
The optimal choice of effort is simply given by the tangency between the
budget constraint and the indifference curve, so that

slope indifference curve = slope compensation

δe = β

The optimal effort choice is typically called the incentive compatibility con-
straint. As we will see later on, such a condition acts as a constraint on the firm’s
choice of the optimal compensation package, and should be always taken into
account by the firm.
The formal problem of the worker is to maximize utility with respect to e,
using the compensation scheme as a constraint

e2
Maxe : U (w e , e) = w e − δ
2
such that w e = α + βe
 OPTIMAL COMPENSATION SCHEMES 91

Substituting the constraint into the utility function, the problem is simply to
maximize with respect to effort the following function:

e2
Maxe : U (w e , e) = α + βe − δ
2
which implies that the first-order condition is

β − δe = 0

so that the optimal effort is

β
e∗ = (IC: Incentive Compatibility Constraint)
δ
from which two implications immediately follow:

To elicit hard work, workers needs incentive, or a level of β > 0.

This is immediately verified by looking at the incentive compatibility con-

straint. The optimal level of effort requires a positive value of β, or a positive
component of the bonus scheme.

Effort is maximum under the franchising scheme, i.e. when β = p.

This result is just a corollary of the previous condition. The maximum level
of effort requires the maximum possible level of the bonus scheme, which we
know to be β = p.

Is Participation Convenient? The Second Problem of the Agent

The second problem of the worker is whether participating in the contract
is convenient, since the worker can enjoy an outside opportunity that yields
expected utility equal to u.
The utility of the agent if she works with the principal is

e ∗2
U (w e , e) = α + βe ∗ − δ
2
where we already know that effort will be chosen optimally at a point in which
e ∗ = βδ . Participation is convenient if and only if the utility obtained by the
contract is larger than the outside option. In formula, this is equivalent to

U (w e (e ∗ ), e ∗ ) ≥ u (6.1)

where the notation w(e ∗ ) shows that the expected income will depend crucially
on the optimal effort e ∗ . With a general compensation package, the expected
92 OPTIMAL COMPENSATION SCHEMES

income is given by w e (e ∗ ) = α+βe ∗ which immediately implies that w e (e ∗ ) =

α + βδ . The condition 6.1 becomes
2

β2 β2
α+ −δ 2 ≥u
δ 2δ
so that
β2
α+ ≥u ((P) Participation Constraint)
2δ
The participation constraint has also a graphical interpretation. The
participation constraint requires that at the optimal level of effort the worker
enjoys a level of utility which is at least as large as the worker’s outside option,
which we labelled u. In terms of Fig. 6.3, if we say that the worker can enjoy
an outside level of utility equal to u = U2 , than at a point such as B, the agent
would be better off working for the principal (since the utility level U1 is larger
than the alternative utility level u = U2 ).
The problem of the agent is now solved and it is fully described by the two
constraints IC and P. Having solved the worker’s problem, we now turn to the
firm’s behaviour, taking into account the worker’s choices.

6.2.3 THE PROBLEM OF THE PRINCIPAL WITH A RISK-NEUTRAL AGENT

The firm/principal has two problems
Principal Problem Number 1: Of all the possible compensation packages
that could be offered to the agent, which is the one that maximizes profits?
Answer: The compensation chosen is the one that maximizes expected
profits, which implies choosing the level of α and β that maximizes the
revenues net of the compensation package.
Principal Problem Number 2: Given the answer to problem 1, is it optimal
to hire the agent?
Answer: I should hire the agent as long as profits are positive.

Principal Problem 1: The principal must now choose β and α

The problem of choosing α and β is best solved in two steps. In the first step
we consider the choice of α given a generic value of β. In the second step, we
choose the optimal level of β, taking explicitly into account the solution to the
first step.
Choice of α. Let’s first consider the optimal choice of α given β. For given
choice of β, the principal will want to pay a salary that ensures that the worker
accepts the contract. Alternatively, the firm must choose a level of α so that the
worker will choose to work, or so that the participation constraint is satisfied.
 OPTIMAL COMPENSATION SCHEMES 93

The P constraint simply says that the agent will work for this principal as long
as his utility is as high as the outside option. Thus, the smallest base pay α that
the principal can offer, given that she offers β to the worker, and still get the
agent to work for her, is the one that makes the agent just indifferent between
accepting and rejecting this offer. In other words, the principal will want to
choose an α so that the participation constraint (P) is binding:

β2
α=u− (6.2)
2δ
This condition gives us an important insight. There is a negative relationship
between the fixed payment and the bonus scheme: the fixed payment decreases
as the bonus scheme increases. There is a lot of intuition here. The principal
must provide the worker with a fixed level of utility u. Such a level can be
guaranteed with different combinations of fixed pay and bonus component,
taking into account that effort will be in any case chosen optimally. The pre-
vious expression shows that for very much larger values of the bonus scheme
(a larger β), the choice of α can even become negative. In other words, as
the bonus scheme increases, the fixed payment can becomes a payment from
the worker to the firm.
Choice of β. Let’s analyse the profits of the principals as a function of β,
given that the agent is offered a level of α indicated above. We know already
that the expected profits of the agents are expected revenues minus expected
costs or

E[] = px e − w e
= pe − α − βe,
= (p − β)e − α

Since the agent chooses effort level e ∗ to maximize his utility and since we
know that e ∗ = βδ , total profits become

β
E[] = (p − β) −α
δ
Profits depend on three terms, and all three terms depend on β. Let’s analyse
those terms:

1. Firm profit per unit of sale. This is given by the term p − β. Since β is
the bonus component, the firm gets only p − β per output sold. A larger
β obviously reduces the firm profit per unit of sale.
2. Total quantity sold depends entirely on worker effort, which we know to
be equal to βδ . Larger bonus component β implies larger effort and larger
quantity.
94 OPTIMAL COMPENSATION SCHEMES

3. Finally there is the fixed payment α. As we discussed above, notably on

equation 6.2 larger β allows the firm to reduce the fixed payment to the
worker, which can also become negative as β increases. In such a case the
fixed payment becomes a fixed revenue.

The first two elements make the variable part of profits (p − β) βδ , while the
last term is the fixed part of profits.

Total Profits = Variable Profits + Fixed Payment (Revenue)

where the term in parentheses reflects the fact that the fixed payment
may turn into a positive revenue. The problem of the firm at this point
is simply to select the value of β that maximizes profits. We begin the
exposition with a numerical example, and we then move to the formal
derivation.

A Numerical Example on the choice of β

Let’s consider the problem of the firm with the following numerical example.
The price that the firm charges is p = 2, the outside option u is equal to 0, and
δ = 1.

Total Profits = Variable Profits + Fixed Payment

β2
Fixed Payment = 1 −
2
Variable Profits = (2 − β)β

The firm must choose the level of β that maximizes total profits. The fixed
payment of the worker decreases with β, so that a larger β means that the
fixed payments from the worker increase profits. At the same time, a lar-
ger β has two effects on the variable profits. On the one hand, a larger β
increases the quantity produced. On the other hand a larger β reduces the
margin of each quantity since the firm gets only a share equal to p − β.
Indeed, a β equal to p makes the variable profits equal to 0. Table 6.1 cal-
culates all possible values of profits for different values of β. The value of
variable profits is maximum when β = 1 in the figure. But the table shows
p
that β = 1 = 2 is not the optimal profits. The firm can do better by further
increasing β (and reducing the variable profits) so as to increase the fixed pay-
ment. The maximum is clearly reached when β = p. Figure 6.4 confirms this
finding.
 OPTIMAL COMPENSATION SCHEMES 95

Table 6.1. Profits and utility with different bonus schemes

Firm Worker

Beta Var. rev. Fixed paym. Profits Worker’s ut. Fixed paym. Bonus Effort cost

0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000

0.067 0.129 0.002 0.131 0.000 −0.002 0.004 −0.002
0.133 0.249 0.009 0.258 0.000 −0.009 0.018 −0.009
0.200 0.360 0.020 0.380 0.000 −0.020 0.040 −0.020
0.267 0.462 0.036 0.498 0.000 −0.036 0.071 −0.036
0.333 0.556 0.056 0.611 0.000 −0.056 0.111 −0.056
0.400 0.640 0.080 0.720 0.000 −0.080 0.160 −0.080
0.467 0.716 0.109 0.824 0.000 −0.109 0.218 −0.109
0.533 0.782 0.142 0.924 0.000 −0.142 0.284 −0.142
0.600 0.840 0.180 1.020 0.000 −0.180 0.360 −0.180
0.667 0.889 0.222 1.111 0.000 −0.222 0.444 −0.222
0.733 0.929 0.269 1.198 0.000 −0.269 0.538 −0.269
0.800 0.960 0.320 1.280 0.000 −0.320 0.640 −0.320
0.867 0.982 0.376 1.358 0.000 −0.376 0.751 −0.376
0.933 0.996 0.436 1.431 0.000 −0.436 0.871 −0.436
1.000 1.000 0.500 1.500 0.000 −0.500 1.000 −0.500
1.067 0.996 0.569 1.564 0.000 −0.569 1.138 −0.569
1.133 0.982 0.642 1.624 0.000 −0.642 1.284 −0.642
1.200 0.960 0.720 1.680 0.000 −0.720 1.440 −0.720
1.267 0.929 0.802 1.731 0.000 −0.802 1.604 −0.802
1.333 0.889 0.889 1.778 0.000 −0.889 1.778 −0.889
1.400 0.840 0.980 1.820 0.000 −0.980 1.960 −0.980
1.467 0.782 1.076 1.858 0.000 −1.076 2.151 −1.076
1.533 0.716 1.176 1.891 0.000 −1.176 2.351 −1.176
1.600 0.640 1.280 1.920 0.000 −1.280 2.560 −1.280
1.667 0.556 1.389 1.944 0.000 −1.389 2.778 −1.389
1.733 0.462 1.502 1.964 0.000 −1.502 3.004 −1.502
1.800 0.360 1.620 1.980 0.000 −1.620 3.240 −1.620
1.867 0.249 1.742 1.991 0.000 −1.742 3.484 −1.742
1.933 0.129 1.869 1.998 0.000 −1.869 3.738 −1.869
2.000 0.000 2.000 2.000 0.000 −2.000 4.000 −2.000

Notes: Price: p = 2; outside options u = 0; δ = 1.

The Formal Solution of the Optimal Bonus Scheme

Substituting the value of α expected profits read:

E[] = (p − β)e ∗ − α
 
β β2
= (p − β) − u −
δ 2δ
pβ β2 β2
= −u+ −
δ 2δ δ
pβ β2
= −u−
δ 2δ
96 OPTIMAL COMPENSATION SCHEMES

2.500
Total profits

2.000

1.500
Profits from variable
payment

1.000 Profits from fixed

payment/revenues

0.500

0.000
0. 0
0. 7
0. 3
0. 0
0. 7
0. 3
0. 0
0. 7
0. 3
0. 0
0. 7
0. 3
0. 0
0. 7
1. 3
1. 0
1. 7
1. 3
1. 0
1. 7
1. 3
1. 0
1. 7
1. 3
1. 0
1. 7
1. 3
1. 0
1. 7
2. 3
0
00
06
13
20
26
33
40
46
53
60
66
73
80
86
93
00
06
13
20
26
33
40
46
53
60
66
73
80
86
93
00
0.

Bonus b

Figure 6.4. Firm profits for different values of the bonus scheme

The principal will maximize profits with respect to β so that the first-order
condition solves
MaxE[]β

p β
− =0 (6.3)
δ δ
β=p (6.4)
which implies that the optimal β is equal to the price. In other words, all the
revenues from the sale should go to the worker, and the profits are simply given
by the value of α.
With a risk-neutral agent, the optimal commission rate to offer is the franchising
commission.

Is the Agent worth enough to the Principal?

The answer to this very last question depends on whether total profits are
positive. If total profits are positive, then the Agent is worth enough to the
principal. If total profits turn out to be negative, it is better not to offer the
contract in the first place.
In light of the optimal β of equation (6.4), the optimal base wage α is
p2
α=u−
2δ
 OPTIMAL COMPENSATION SCHEMES 97

from which it follows that the base wage is negative as long as u = 0 (the
worker is paying for the job). Finally, the net profits of the principal are
pβ β2
E[] = −u−
δ 2δ
p2 p2
= −u−
δ 2δ
p2
= −u
2δ
which are obviously positive as long as
p2
u≤ .
2δ
If the latter condition is not satisfied, the principal does not offer the contract
to the agent. In other words, the principal has a participation constraint which
requires positive expected profits, or E[] > 0.
Summing up
For a risk-neutral agent the optimal compensation scheme of the type w =
α + βx is such that
β=p
p2
α=u−
2δ
and the agent will accept at the utility u. Further, effort is such that
p
e∗ =
δ
The optimal contract is described at point C in Fig. 6.5. The level of utility in
equilibrium is exactly equal to that of the outside option, while the incentive

E (y) = a + pe;
U=u

E (y) U2
C

a = u–p2/2d e*= p/d e

Figure 6.5. The optimal contract

98 OPTIMAL COMPENSATION SCHEMES

scheme has the slope equal to the price. The optimal effort is chosen at level
e ∗ = δ and the α is negative in the figure, so that the worker buys the job.
p

6.3 The Basic Scheme with Non-Negative Payments

In the rest of this chapter we analyse two important departures from the basic
scheme. Both departures are very realistic, and help us to understand why, in
real life labour markets, the franchising scheme analysed above is not observed
very often. The first departure we analyse is a constraint on the worker’s ability
to make negative payment to the firm. The second departure (analysed in the
next section) deals with risk aversion.
In this section we assume that the worker reservation utility is zero, so
that u = 0. In this case the optimal franchising scheme would require the
worker to make an up-front payment to the firm equal to p 2 /2δ. Let us now
assume that the worker cannot make any up-front payments to the firm. This
is a natural constraint. This situation arises for a number of reasons. First,
and most importantly, it is not obvious that a labour contract that imposes a
negative up-front payment is at all legal. Second, the worker may not have the
resources to obtain such payments and would need to rely on bank credit. In
the latter case the worker may be able to borrow some money for the payment,
but such borrowing would probably be limited. In any event, we here assume
that no borrowing can take place, or that the labour law simply makes such
payment undoable. The question is what happens to the optimal scheme in
this case.
The existence of a non-negative constraint on α makes it impossible to
implement the optimal contract, which we saw was based on having p = β
and obtaining all the profits from the fixed payment. What should the firm do
in this case? It is obvious that in this case the total profits will be just equal to
the variable profits,

Total Profits = Variable Profits

= (p − β)e ∗
β
= (p − β)
δ
and the firm must select the level of β that maximizes variable profits.
We know that the effort chosen by the worker is maximum when β = p.
But we also know that if the firm chooses β = p, the effort obtained and the
quantity produced will be maximum, but profits per unit of sale will be zero.
It is clear that in this case the optimal compensation scheme will require lower
effort. An important result follows.
 OPTIMAL COMPENSATION SCHEMES 99

When negative payments are not allowed, the optimal compensation is never
at the maximum effort, and the firm chooses a level of β < p.
Let’s call β̂ the level of bonus chosen by the firm when there is a non-negative
constraint on α. What will be the optimal level of β̂ in this case? Let’s first look
at the numerical example. Figure 6.4 immediately suggests that the optimal β̂
corresponds to β = 1 = p/2. Note also that the firm could set a larger β̂ and
still make positive profits, but such profits would be lower.
Let’s now formally obtain the optimal β̂.
β̂
Max (p − β̂) when α = 0
δ
so that the firm chooses β̂
β̂ p − β̂
− + =0
δ δ
p
β̂ =
2
When non-negative payments are ruled out, the optimal commission scheme is
the 50 per cent commission rate.
Let’s see now what the worker utility is in the case of a full variable payment
δ
U (w e (β̂), e ∗ (β̂)) = w e (β̂) − e ∗ (β̂)2
2
2
β̂ β̂
= β̂ −
δ 2δ
p2
= >0
8δ
where we recall that u = 0 in this case. The utility level is strictly positive, and
larger than the outside option. In other words, the worker gets a rent, or a value
of the utility above the reservation level, as the following remark shows:
When non-negative payments are impossible, the worker gets an extra rent and
enjoys a level of utility larger than the reservation value.
It is also clear that the firm reduces its expected profits. In the example in
the table the profits of the firms when β = 1 are much lower than the profits
under the first best scheme.

Summing up
For a risk-neutral agent with reservation utility u = 0 the optimal com-
pensation scheme with non-negative payments is of the type w = βx
100 OPTIMAL COMPENSATION SCHEMES

such that
p
β̂ =
2
α̂ = 0

Further, effort is chosen so that

p
e∗ =
2δ
The level of utility in equilibrium is strictly larger than the outside option
(which was normalized to zero in this case).

6.4 The Basic Scheme when the Agent is Risk Averse

The presence of risk aversion changes the compensation schemes. In the
baseline contract, the worker (or the agent) bears all the uncertainty asso-
ciated with the luck component of output. The firm gets a fixed payment while
the worker is obliged to observe fluctuations in her income. When the realiz-
ation of the luck component is high, the compensation increases, while it falls
when the realization of the luck component is low. As long as the worker is risk
neutral, she does not care about income variability. This is no longer true in the
case of risk aversion. When the worker is risk averse (and the firm risk neutral),
the optimal distribution of risk calls for the firm to take up some of the risk.
Indeed, as we will see, the optimal compensation implies a reduction in the
bonus component and an increase in the fixed payment. In other words, the
principal provides some insurance to the agent. Providing insurance, however,
has an impact on effort and efficiency, which we know to be maximum under
a pure bonus scheme. In other words, the optimal contract under risk aversion
is the one that solves this efficiency/insurance trade-off.

6.4.1 THE AGENT PREFERENCES

General Agent Preferences. We now assume that the agent likes higher expected
wage w e but he dislikes income variability Var(w). In this section we use a
mean-variance utility function, whereby utility increases with the expected
value w e and decreases with the variance of the compensation and with the
effort
e2
U (w, e) = w e − λVar(w) − δ
2
 OPTIMAL COMPENSATION SCHEMES 101

The difference between risk aversion and risk neutrality on the agent’s
preferences is analysed next.

A risk-neutral agent: A risk-neutral agent is somebody who does not care

about income variability, as was analysed in the first part of the chapter, so
that for a risk-neutral guy λ = 0 or

e2
U (w, e) = w e − δ
2
e2
= α + βe − δ
2
Risk-averse guy with mean variance preferences: A risk-averse guy with
mean variance preferences is somebody who does care about income
variability, so that his utility function is

e2
U (w, e) = w e − λVar(w) − δ
2
The larger λ is, the more the worker is averse to income variability.

The linear compensation that we are analysing (w = α + βx) has a variance

that is simply given by Var(w) = β 2 Var(x). Since the variance of the random
variable x is constant and equal to v the variance of the compensation is
Var(w) = β 2 v so that the utility can be written as

e2
U (w, e) = α + βe − λβ 2 v − δ
2

6.4.2 THE PROBLEMS OF THE AGENT

We now solve the two problems of the agents under the case of risk aversion
with mean variance preferences. We begin with the choice of effort under the
general bonus contract: In the bonus contract w = α+βx so that w e = α+βe
and Var(w) = β 2 v and the utility level is

e2
Maxe : U (w, e) = α + βe − λβ 2 v − δ
2
which implies that the optimal choice of effort is the condition (IC) above, or
that
β
e∗ = (IC: Incentive Compatibility Constraint)
δ
102 OPTIMAL COMPENSATION SCHEMES

The choice of the optimal effort does not depend on the presence of risk
aversion.
The second problem of the worker is whether participating in the contract is
convenient, since the worker can enjoy an outside option which yields expected
utility equal to u.
The utility of the agent is

e ∗2
U (w, e) = α + βe ∗ − λβ 2 v − δ
2
β
since e ∗ = δ it follows that participation is convenient if and only if

U (w(e ∗ ), e ∗ )) ≥ u

which implies that

β2 β2
α+ − − λβ 2 v ≥ u
δ 2δ
β2
α+ − λβ 2 v ≥ u
2δ

so that

1 − λv2δ
α + β 2[ ]≥u ((P) Participation Constraint-Risk Aversion)
2δ

The presence of risk aversion changes the participation constraint.

The presence of risk aversion makes participation in a bonus scheme a less
attractive option to the worker. A bonus scheme is associated with variable
wages, since the luck component of the output does affect the worker’s utility.
As one can see from the participation constraint required, the larger the risk
aversion component v, for given β and α, the less likely is participation in the
deal. We now move to the problem of the principal.

6.4.3 THE PROBLEM OF THE PRINCIPAL WITH A RISK-AVERSE AGENT:

THE OPTIMAL BONUS SCHEME
The principal must now chooses β̃ and α̃ where we indicate variables with a ∼
symbol to indicate that we are analysing the case of risk aversion.
 OPTIMAL COMPENSATION SCHEMES 103

Choice of α̃. Let’s first consider the optimal choice of α̃ given β̃. For given
choice of β̃ the principal will want to pay a salary so that the agent will choose
to work. The agent chooses to work for this principal as long as he is just as
well off as at his next-base opportunity. Thus, the smallest base pay α̃ that the
principal can offer, if she offers β̃, and still get the agent to work for her is

β̃ 2
α̃ + − λβ̃ 2 v = u
2δ
 
1 − λv2δ
α̃ = u − β̃ 2
2δ

Choice of β̃. Let’s study the profits of the principal when she offers β̃ and
the agent is going to work by offering a level of α̃ as indicated above. The
expected profits of the agents are

E[] = pE[x] − w e
= pe ∗ − α̃ − β̃e ∗
β̃ β̃ 2
s.t. e ∗ = and α̃ = u − + λβ̃ 2 v
δ 2δ

Substituting the two constraints in the objective functions one has

p β̃ β̃ 2
E[] = −u− − λβ̃ 2 v
δ 2δ

The principal will maximize profit with respect to β̃ so that the first-order
condition solves

p β̃
− − 2λβ̃v = 0
δ δ
p
β̃ =
1 + 2λδv

which implies that the optimal β̃ is less than the price (as long as λ > 0, υ > 0
and δ > 0).

For a risk averse guy, the optimal commission rate is never the franchising
compensation scheme, i.e. 0 < β̃ < p
104 OPTIMAL COMPENSATION SCHEMES

Summing up
With a risk-averse agent the scheme works as follows
p
β̃ =
1 + 2λδv
 
2 1 − λv2δ
α̃ = u − β̃
2δ
 
p 2 1 − λv2δ
α=u−
(1 + 2λδv)2 2δ

b APPENDIX 6.1. WHAT IF THE PRINCIPAL AND THE RISK-AVERSE

AGENT COULD CONTRACT ON EFFORT

Let’s start by defining the surplus from the deal as the sum of the utility of each party,
net of the respective outside option. This is identical to the difference between the
revenues from the deal and the workers’ cost of eliciting effort. The surplus is then
S = ( − 0) + (U − u)
δe 2
= pe − w + w − −u
2
δe 2
= pe − −u
2
and let us assume that the two parties could contract directly on effort, rather than
relying on the output x. Let’s see what the effort choice is that maximizes the surplus
from the job. In other words, the optimal effort is the one that makes the marginal
surplus zero (i.e. ∂S
∂e = 0). We call the optimal level of effort e and it is easy to see that
0

p − δe o = 0
from which it follows that
p
eo =
δ
The level of effort that maximizes the surplus is e 0 = p/δ.
So that the maximum value of the surplus is (evaluating S at e o )
p2 δp 2 p2
S(e o ) = − 2 =
δ δ 2 2δ
and the job is efficient as long as
S(e o ) > 0
p2
>u
2δ
 OPTIMAL COMPENSATION SCHEMES 105

The analysis above shows clearly that the efficient contract is identical to the contract
between the principal and a risk-neutral agent. In other words, with a risk-neutral agent,
the franchising payment implements the first best contract.
Thus, in the case of risk neutrality on the part of the agent, there is no cost linked
to the impossibility of contracting directly on effort.
The previous expression gives the joint value of the surplus, which still needs to be
allocated to the worker and to the firm. The payment w is chosen so that the worker’s
utility is identical to the outside option U = u so that

δe o2
w =u+
2

The worker will receive exactly the utility level u, while the firm will get all the
uncertainty and will have a payment equal to the residual value, in other words

p2
(e o ) = −u
2δ

and the deal will be struck as long as

p2
>u
2δ

and the principal will get all the risk.

In the case of risk aversion things change, and the solution that we analysed in the
text (the one with a bonus scheme that is less than one) does not implement the first
best. What is the cost of being unable to contract on effort? The cost of being unable
to contract on effort is given by the drop in profits with respect to those obtained with
the optimal bonus contract. The agent, once he is offered the contract α̃ and β̃, gets a
level of utility that is equal to its outside payment u. The principal profits under the
optimal bonus schemes are

∗ = pe ∗ − (α̃ + β̃e ∗ )

where

 
p2 1 − λv2δ
α̃ = u −
(1 + 2λδv)2 2δ
p
β̃ =
1 + 2λδv
β̃
e∗ =
δ
106 OPTIMAL COMPENSATION SCHEMES

So that profits become

p β̃ β̃ 2
∗ = −α−
δ δ
 
p 2 1 p2 1 − λv2δ p2 1
= −u+ −
1 + 2λδv δ (1 + 2λδv)2 2δ (1 + 2λδv)2 δ
 
p2 1 p2 −1 − 2λδv
= −u+
1 + 2λδv δ (1 + 2λδv)2 2δ
2  
p 1
= −u
2δ 1 + 2λδv
So that the cost of dealing with a risk-averse guy is
 
p2 p2 1
 o − ∗ = −
2δ 2δ 1 + 2λδv
2  
p 1
= 1−
2δ 1 + 2λδv
 
p 2 λv
=
δ 1 + 2λδv
The cost of being unable to contract on effort is equal to zero if λ = 0 and if υ = 0.
7 Pay for Performance with
Wage Constraints

7.1 Introduction
The baseline theory of optimal compensations suggests that properly designed
incentive contracts can be very effective in extracting workers’ effort. A fixed
wage scheme, where compensation is independent of output, does not provide
incentives to the worker. As we have seen in the previous chapter, in response to
a fixed wage, a utility maximizer worker would choose the minimum amount
of effort. Conversely, bonus schemes, where a fixed wage is mixed with a bonus
component proportional to output, are a much more effective compensation
scheme. We have also seen that as a benchmark limit case, when the worker is
risk neutral and no other constraints are binding, the optimal compensation
scheme is an extreme bonus scheme, with a negative fixed payment up front
and a bonus component identical to the price (we called such a scheme the
franchising scheme). In such a setting, production efficiency is maximized and
the worker becomes the residual claimant. Yet, we know that such a scheme is
hardly attainable in real life labour markets, since it requires the worker to pay
for the job.
In real life labour markets and real life organizations, it is very difficult to
implement the franchising scheme. Even if agents are risk neutral, payments
from the workers to the firms are barely conceivable. The previous chapter
has shown that the optimal franchising contract turns into a 50 per cent com-
mission rate as soon as negative payments are not allowed. In such a setting
the firms set a base pay equal to zero, and share with the worker the variable
outcome.
This chapter looks into compensation behaviour under more realistic wage
constraints. For example, firms often have to offer a simple fixed wage inde-
pendent of performance. Sometimes such a wage can be chosen by the firm
and sometimes it is negotiated outside the firm’s control, either by a minimum
wage constraint or by industry-wide collective agreements. What should the
firm do when it is forced to pay a fixed wage, independent of output? How
should such a wage be chosen? Is it possible for the firm to obtain the max-
imum profits when the wage contract must simply be a fixed (positive) wage.
The answer is yes, if three conditions hold: (i) the wage level can be optim-
ally chosen by the firm; (ii) a minimum amount of output can be observed
108 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

and imposed on the employment contract; and (iii) workers are homogeneous
vis-à-vis their ability. When one of these conditions breaks, some profits must
be given away. This chapter focuses on heterogeneity.
In this chapter we explore in detail the impact on firm performance of
worker heterogeneity. We will assume that firms face a variety of workers, and
different workers have different attitudes toward efforts. Ability is like a DNA
component known to the worker but not known to the firm. We will see that a
firm that offers a single wage with a minimum output requirement surrenders
some rent to workers, and the most efficient production cannot be realized.
Some workers in the firm enjoy a pure economic rent, and are strictly better off
than their best outside options. This is a situation that probably gets closer to
most real life firms. In most firms, workers take it easy and enjoy a specific rent.
The core of the chapter studies how the firm can improve its situation
when some of its workers are enjoying a pure economic rent. We basically
study the effect of a bonus scheme with a minimum guarantee, or a scheme
in which a minimum wage is guaranteed, but if workers do well, they can
obtain additional income. We will see that such a scheme is likely to improve
productivity vis-à-vis the simple fixed payment with minimum output. We will
show that two different effects emerge from such a move: efficiency effects and
sorting effects. The efficiency effects refer to the improvements in performance
and productivity of workers that are already in the firm. The sorting effects
refer to a change in the composition of the workforce in the firm.
The second part of the chapter shows how a minimum guarantee with
bonus scheme worked in a real life case study. We will review the Safelite case,
a window installer in the United States that introduced a pay for performance
scheme with minimum wage guarantee. The case study will clearly show that
sizeable sorting and efficiency effects are linked to the changes of remuneration
schemes.

7.2 Heterogeneous Ability: The Set-Up

The main analytical novelty in this chapter is the emphasis on workers’ het-
erogeneity. To simplify the analysis, we assume throughout the chapter that
individuals are risk neutral and there is no uncertainty. We will also assume that
the firm is producing a homogeneous product whose price is equal to 1.
The quantity of output produced by an individual is indicated by x. The
amount of production depends not only on individual effort, but also on
individual ability. We assume that individual ability differs across individuals.
Specifically, output x is related to ability by the following relationship:
x =a+e (7.1)
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 109

where a is ability and e is effort. a is a fixed individual parameter unobservable

to the firm but fully known by the agent. The previous relationship suggests
that for a given effort, an individual with larger ability can produce a larger
quantity of output. We also assume that the outside option faced by each
individual grows with the ability of the agent:

outside option = au

so that workers with larger ability have better outside options. This makes
sense, since better workers have better outside chances.
Let’s see how the preferences and the indifference curve change with ability.
The utility function is the same as the one we used in the previous chapter, so
that utility depends positively on wage income and negatively on effort.É Since
output and ability are linked by the relationship x = a  + e, the utility function
reads for an individual with ability a 

w if x < a 
U =
w − δ(x−a)
2
2 if x ≥ a 

To properly understand the utility function introduced above let us study the
indifference curve in a space [x, w]. An indifference curve in such a setting will
refer to all the combinations of x and w that give the same utility to the agent.
For this purpose, we need to define the maximum output at zero effort.
• Each individual has a maximum output level x̄ = a  that he or she can
produce without exercising any effort.
• Beyond the zero effort output level x̄ = a  , individuals need to exercise
effort to produce more output.
• Individuals that are more able, have larger output x̄ = a  .

With respect to the indifference curves in the [w, x] space, the following
properties apply
• Indifference curves are flat up to the maximum output at zero effort.
• Indifference curves are upward sloping beyond x̄ = a  .
• The slope of the indifference curve depends on ability, and less able
individuals have steeper indifference curves.
The formal definition of the indifference curve at utility level k for an
individual with ability a  is given by the following relationship

k if x < a 
w=
k + δ(x−a)
2
2 if x ≥ a 

2
É The utility function in the previous chapter was written as U = w − δe2 .
110 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

W Low ability
High ability

K1

K2

x = al x = ah x

Figure 7.1. Indifference curves with heterogeneous ability

Up to quantity level x = a  the indifference curve is flat. For obtaining further

output, effort needs to be exercised and individuals must be compensated with
further income. The associated slopeÊ is

dw 0 if x < a 
=
dx δ(x − a  ) if x ≥ a 
Figure 7.1 presents the indifference curve for two different levels of utility
k1 and k2 for two different individuals with ability al and ah . The individual
with ability al has a flat indifference curve up to x = al while the individual
with ability ah has indifference curves that are flat up to the point x = ah .
Another important dimension to consider is the slope. Consider the point
of Fig. 7.1 where the two indifference curves cross.Ë The individual with low
ability clearly has a steeper indifference curve. Economically, this means that
he or she needs to be compensated with more wages to increase his or her
output. This is because more output is more costly (in terms of effort) for the
individual with low ability.
Ê The formal proof of the indifference curve is given in what follows. Let’s recall that, in general, the
slope of the indifference curve is given by
∂U ∂U
dw + dx = 0
∂w ∂x
∂U
dw ∂x
= − ∂U
dx
∂w

If x is lower than the zero effort output the slope is zero (since ∂U∂x ), while if x is beyond the zero
output level the

dw 0 if x < a
=
dx δ(x − a) if x ≥ a

Ë The indifference curves for the same individual never cross. Yet, the indifference curves of two
different individuals can and do cross each other.
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 111

7.3 Fixed Wage with Minimum Output

Before introducing the effects of heterogeneous ability, we need to understand
the behaviour of a firm that pays a fixed wage with minimum output. We begin
by assuming that ability is constant.
The idea of a minimum level of output is simply to impose a minimum
standard of quantity, so that if a worker does not reach such a minimum
standard, the firm is not going to hire the worker, or the employment relation-
ship can be interrupted without cost. In this section we label a job [w, xmin ] as
a job that pays a fixed wage α = w and requires an output xmin .
Appendix 7.1 describes how a minimum output can be chosen optimally,
and under which conditions a firm that offers a job [w, xmin ] can do as well as
a firm that uses the franchising scheme. In practice, a firm that offers a fixed
wage with minimum output can obtain the same level of profits as an efficient
level of output only if various conditions hold. Such conditions are
• There is no uncertainty, so that the output observed reflects the effort
exercised.
• There is no heterogeneity in the workforce. In other words, the parameter
a is fixed in the population.
• The minimum level of output xmin is chosen by the firm so as to obtain
from the worker the efficient level of effort.
Note that when we talk about an efficient level of profits, we mean a level of
profits that is as high as the one that would be obtained by a firm that imple-
ments the franchising scheme, as indicated by equation 7.2 in Appendix 7.1.
Obviously the wage in this setting must be properly chosen by the firm, and
the Appendix shows how such a wage can be selected.
Things are more complicated if the wage is completely fixed and outside the
firm’s control. In this case the firm can hardly make any choice. All it can do
is to select a minimum quantity of output xmin . In other words, if the wage is
completely fixed at w̄, all the firm can do is select a quantity so that the worker
participation constraint is binding, and operate if profits are positive.

7.4 Pure Rent: Fixed Wage with Minimum

Output and Heterogeneous Ability
Let’s now explicitly introduce heterogeneous ability and let’s describe the prob-
lem of a firm that [w, xmin ] where xmin is some level of output. The firm we
are considering pays a fixed wage w regardless of ability, and requires some
level of minimum output xmin for working at the firm. Recall that ability is
112 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

High ability
W
Low ability

W*
K
K2

Xmin X

Figure 7.2. Optimal solution with fixed wage and minimum output when ability is
heterogeneous

known to the individual but it is not known to the firm. The problem can first
be described graphically.
Let us first consider the constraint faced by each worker. The constraint is
similar to the linear constraint with the fixed wage that we have analysed in
Fig. 6.1, with the only difference that the wage w is obtained only for quantity
produced larger than the selected minimum xmin . The constraint is thus a step
function, with the step that corresponds to the point at the minimum output
xmin .
Given this constraint, it is clear that no employees will ever want to produce
more than the minimum required by the firm xmin . At the same time, producing
less than xmin is not allowed. Figure 7.1 presents the indifference curve while
Fig. 7.2 describes the equilibrium with minimum output for two individuals
with different ability levels. As is clear from the picture, the individual with high
ability reaches a utility level k1 while the individual with low ability reaches
a level of utility equal to k2 . Clearly we have k1 > k2 . This makes a lot of
sense: an individual with larger ability must exercise less effort to produce an
output equal to xmin so that he will have a larger level of utility for working at
the firm. Yet, there is another key dimension to consider, namely the fact that
individuals with larger ability have also larger outside options, and may not
find it interesting working in the firm. We need to solve the problem of which
worker will self-select for the firm that we are analysing.
The utility for an individual that works for this firm with ability a is

w if a > xmin
U = δ(x−a)2
w− 2 if a ≤ xmin
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 113

The condition for working at a firm that pays according to the contract
[w, xmin ] is
δ(xmin −a)2
au ≤ w − 2 if a ≤ xmin
w if a > xmin
More able individuals find it easier and easier to produce the minimum amount
of output, and thus need to exercise lower and lower effort at the margin.
Beyond a level of utility a = xmin , individuals do not even need to exercise
effort to obtain xmin and their utility no longer increases. The outside option
increases linearly with a. This suggests that the condition above is satisfied for
al ≤ a ≤ ah
The solution is clearly described in Fig. 7.3. Individuals with utility level below
al are characterized by an outside option (the straight line in the figure) that
is larger than the utility for working at the firm. Such characteristics hold also
for individuals with ability larger than ah . While individuals in the positions al
and ah , the two extremes, are just indifferent and their participation constraint
is binding, individuals in between are enjoying a pure economic rent. The rent
emerges since the firm must obey an egalitarian wage policy.
When a firm pays a job [w, xmin ] and there is heterogeneity in the population,
a subset of workers in the population enjoys a pure surplus by working at
the firm.
The result can be summarized by saying that there are three types of worker
in the population.
1. Individual workers that are not good enough to work at the firm. Those
are the workers for which a < al ;

Outside option
Utility for working at a firm
that pays fixed wage W with
minimum output x0

u
a1 a = x0 ah

Figure 7.3. Individual self-selection for working at the firm that pays w, xmin
114 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

2. Individual workers that work at the firm that pays [w, xmin ] and are those
workers for whom a ∈ [al , ah ]; Such workers (with the exception of those
at the boundaries) enjoy a surplus inside the firm.
3. Individuals workers that are too good to work at the firm, so that a > ah .

7.4.1 A NUMERICAL EXAMPLE

Consider a firm that is forced to pay a fixed wage equal to w to all its workforce.
Assume that the utility function is

w if a ≥ xmin
U = δ(x−a)2
w− 2 if a < xmin

so that δ = 1. We also assume that u = 1/2 and that ā = 1, so that we initially

keep ability constant. We label the output associated with the optimal effort
xo and its value is

xo = ā + eo
1
xo = ā +
δ
xo = 2

since ā = δ = 1. What is the wage that the firm should fix? To find the wage
recall that the wage must satisfy

δ(xo − ā)2
w ∗ = āu +
2
1 1
= + =1
2 2
This suggests that when ability is constant and equal to ā = 1 a firm that
pays [w = 1, xo = 2] behaves efficiently. The profits of the firm are  =
2δ + ā(1 − u) = 1.
1

Assume now that workers are heterogeneous and that a is uniformly dis-
tributed between amin = 0 and amax = 4. Let’s see which workers will choose
to work at this firm that pays [w ∗ = 1, xo = 2]. The condition is

δ(xo − a)2
w∗ − ≥ au
2
(2 − a)2 a
1− ≥
2 2
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 115

whose solutions are obtained by the following quadratic equation

a 2 − 3a + 2 = 0
al = 1
ah = 2

Individuals with a < 1 are not good enough to work in the firm; individuals
with a > 2 are too good to work in the firm and will choose their outside
option. Individuals in between the two values (al , ah ) are enjoying a rent, since
they get a utility which is larger than their outside option. In other words,
heterogeneity implies that the firm that pays a fixed wage is no longer able to
induce all workers to be just to their outside option. Take an individual with
ability a = 3/2. The utility for working at the firm w = 1, xo = 2 is 7/8 while
the outside option is 3/4, so that he is enjoying a rent equal to 1/8. The firm’s
profits are still 1.

7.5 Performance Pay with a Two-Tier Wage System

Let’s consider a firm that is offering a minimum wage with minimum output
attached, or a job that pays [w, xo ]. We have seen in the previous section that
such a firm has to surrender some rent to some individuals, in particular to
the individuals that have ability strictly larger than al and strictly lower than
ah . What we want to consider is whether the firm may implement a two-
tier wage system, or a compensation structure that guarantees the minimum
wage to all the workers and gives the option of having a bonus scheme for some
individuals that are particularly productive. With such a scheme, the basic pay
that is offered to each individual is still maintained, while some workers have
the option, if they want it, to perform better. Such compensation is for example
consistent with a minimum wage to be paid to workers plus a bonus scheme
for high-performing workers.
We want to understand what will happen to this firm. In particular, we are
interested in two dimensions of the phenomenon that we label incentive effects
and sorting effect.
Incentive effects: How will the performance of the individuals who are
enjoying a rent with a fixed and egalitarian wage change?
Sorting effects: what will happen to the quality composition of the labour
force inside the firm?
Let’s consider a firm that is currently paying the following scheme [x̄, w̄]
where x̄ is the minimum output required and w̄ is the fixed wage that is paid
116 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

Bonus scheme with minimum guarantee

Compensation, w

Hourly wage, w

Minimum Output, x
output

Compensation = Max [w,a + bx]

Figure 7.4. Bonus scheme with minimum wage guarantee (and minimum output required)

to all workers that produce at least x̄. The firm is offering a performance pay
w pr that can be described by our general bonus scheme as
w pr = α + βx
where β is the incentive factor (β ≤ 1) and α is the fixed payment for zero
output (which we know can be negative). All the individuals that produce
more than x̄ can be eligible for the performance plan. Figure 7.4 describes the
constraints that each worker faces with this scheme. The fixed wage works as a
step function exactly like the one we analysed in Fig. 7.2. Yet the figure reports
also the constraint of a general bonus scheme, described by an upward-sloping
line with intercept equal to α (negative in the figure). The bold lines report the
actual compensation. As long as the individual produces the minimum output
required, a wage equal to w̄ is guaranteed, and only if the pay with the bonus
scheme is larger than such value is the bonus scheme effective. In formula, the
compensation that we are analysing is described by the following:

0 for x < x̄
Compensation =
Max[βx + α, w̄] for x > x̄
where output x̄ is required.
Let’s see which individuals would choose to produce under the bonus
scheme. We now solve the problem for the quantity x, but we know that
there is a one-to-one relationship between effort and output, as described by
x = a + e. Let’s call x pr the quantity of output that maximizes the utility. This
means that x pr is the solution to
Maxx U (w pr , x)
δ(x − a)2
U (w pr , x) = α + βx −
2
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 117

High ability

w
Low ability D

Figure 7.5. Individual optimization with a two-tier system

whose first-order condition implies the following solution

β
x pr = +a
δ

Different individuals with different levels of ability have different preferred

quantity x pr . Obviously, an individual who features x pr < x̄ would not sell
himself into the piece rate scheme but would produce at the minimum required
output. The key decision to be taken is for the individuals that under the
performance scheme produce an amount of output x pr that is larger than
the minimum required output. Assume that there is at least an individual
that features x pr > x min . The choice is described by Fig. 7.5. Without the
bonus scheme individuals with different levels of utility choose position C,
since it does not make any sense to produce more. Things are different if
the bonus scheme becomes available. In Fig. 7.5 the low-ability individual
chooses to remain at point C, since a move toward the bonus scheme would be
associated with a lower indifference curve. Things are different for the high-
ability individual, who by moving to point D in the figure increases his or her
utility.

7.5.1 THE FORMAL SOLUTION TO THE TWO-TIER WAGE SYSTEM

We want to study how ability influences the decision between the bonus scheme
or the fixed wage. An individual with ability level a will choose the bonus
118 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

scheme if

U pr w pr , x pr , a > U (w̄, x̄, a)
where we explicitly recognized that utility depends also on the ability level.
Since we know that x pr = βδ + a the utility associated with the performance
plan is
 2 
β
U pr (w pr , x pr , a) = + α + βa
2δ
which clearly shows that individuals with larger utility level a get larger util-
ity from the bonus scheme. The utility associated with operating under the
minimum guarantee (which corresponds to point C in Fig. 7.5) is

w for a > x̄
U (w̄, x̄, a) = δ(x̄−a)2
w− 2 for a < x̄
The previous condition simply says that the utility level associated with the
minimum guarantee is at most w. This is not surprising. Very able individuals
can obtain the minimum output x without any effort and therefore get a utility
equal to w. Conversely, individuals with lower ability need to exercise effort
and reach a lower utility. Since the utility with the minimum guarantee has an
upper bound level, it is clear that there exists a threshold level ã + such that for
all utility levels above the individuals are better off in the bonus scheme. In
formula, we can write that
U pr (w pr , x pr , a) > U (w̄, x̄, a) for a > ã +

The situation is clearly described in Fig. 7.6. The horizontal axis reports
different utility levels, while the vertical axis reports the utility levels. Let’s now
focus on the straight line labelled PPP and the curved lane labelled Fixed Wage.
As one can see, the curved line reaches an upper bound for ability level beyond
a = x̄ while the line labelled PPP is upward sloping. Clearly, individuals
beyond ã + are better off in the bonus scheme.
Let’s now consider the entire population of individuals with ability that
can vary from amin to amax . In the previous section we established that the
population potentially employed by the firm that was paying according to
[w̄, x̄] was described by all the individuals that belong to an interval [al , ah ]
and that with the exception of the individuals on the boundaries, there was
a pure economic rent inside the firm. The possibility of producing under the
bonus scheme, and the emergence of the threshold ã + , modifies the scheme.
In particular, the most natural assumption (consistent with Fig. 7.5 and with
the case study that will be analysed in the next section) is to have
al < ã + < ah
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 119

PPP Outside
Utility option

Fixed wage

al ã a = x0 ah

Figure 7.6. Utility levels for different individuals when a bonus scheme is introduced in a
firm that was offering a fixed wage with minimum output

The existing workforce is partitioned into various categories, which we analyse

next.

1. Individuals with ability level below al do not work in the firm.

2. Individuals with ability al ≤ a ≤ ã + operate under the minimum wage
guaranteed even if the bonus scheme is available. For these individuals
it is too costly, in terms of effort, to produce in the bonus scheme range.
Despite the incentive given to them, they would not choose the bonus
regime.
3. Individuals with ability ã + < a < ah operate under the performance
plan if they have the option to do so. These are the individuals that were
enjoying a rent under the fixed scheme, but they would respond to the
incentive given (we call this the incentive effect).
4. Individuals with ability above ah are now interested in joining the firm.
This shows that the bonus scheme increases the fraction of the workforce
for which it is attractive to work at the firm.

From the analysis of these intervals, we can also clearly distinguish between
two types of effects associated with the introduction of a bonus scheme: the
incentive effect and the sorting effect.

Incentive effects. A minimum wage scheme with a performance option

increases productivity of the existing workforce. This is a pure incentive
effect, since a given workforce improves its average effort and productivity.
Sorting effect. A minimum wage scheme with a performance option increases
the average ability of the workforce.
120 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

7.5.2 TWO-TIER SYSTEM: AN EXAMPLE

Let’s continue with the example used before. al = 1 and ah = 2 and the job
at the firm was of the type [w = 1, xo = 2]. Let’s assume that α = −1/4 and
that β = 1. Let’s first consider that the utility of the individual at the lowest
minimum output required is
δ(x̄ − a)2 3
U (w̄ = 2, x̄ = 2, a = 1) = w − =
2 2
∗
while with a bonus scheme (recalling that x ∗ = βδ + a)
 2 
β 1
U pr (w pr , x pr , a) = + α + βa =
2δ 2
so that it is clear that at the lowest support of the distribution the individual is
better off with the fixed wage. Conversely, for individuals at the upper support
U (w̄ = 2, x̄ = 2, a = 2) = 2 while U pr (w pr , x pr , a = 2) = 94 so that
individuals at the upper support level prefer the bonus scheme. The individual
who is just indifferent is the solution to
 2 
δ(x̄ − a)2 β
w− = + α + βa
2 2δ
which becomes
2a 2 − 4a + 1 = 0
whose solution above al = 1 is
√
ã + = 1 +
2
2/2
so that individuals sort in the following way:
• individuals with ability a < al do not join the firm;
• individuals with al < a < ã2+ choose minimum;
• individuals between ã2 < a < ah choose piece rate;
• individuals above ah are now interested in joining the firm.
Note that this particular type of sorting depends crucially, among other
things, on the parameter α which is the implicit cost for the job. The two
conditions that we need for this type of sorting are
U (al , w ∗ ) > U (al , PPP)
U (ah , PPP) > U (al , w ∗ )
The first condition requires α to be sufficiently large while the second condition
requires α to be sufficiently low. In other words the implicit cost of the job
must be sufficiently high that low-ability individuals do not find it convenient
but sufficiently low that some individuals would want to swap to PPP.
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 121

7.6 Pay for Performance at Safelite

Safelite is the largest car windows installer in the United States, and its
headquarters is based in Ohio. In 1994, Safelite changed the salary structure of
its car window operation. Up to January 1994, the salary structure was based
on an hourly rate, which was paid independently of the number of windows
installed. Alongside, a minimum amount of production was required in order
to continue to be employed at Safelite.
Between 1994 and 1995 the salary structure changed, and it became linked
to the number of car windows actually installed. The goal of the firm was to
increase productivity without losing any workers.
There were basically two options on the table, which critically depend on
the existence of heterogeneity in the workforce.
1. Option number 1 (homogeneous workforce): Safelite may increase the
minimum amount of output required by the firm, and leave the salary
structure based on an hourly wage (possibly increasing the hourly salary).
Our analysis has shown that if individuals are homogeneous, it is possible
to increase productivity in this way as long as the initial situation is not
the first best efficient. However, if there is heterogeneity in the workforce,
an increase in the minimum output leads to undesirable turnover at the
firm level.
2. Option number 2: (heterogeneous workforce): Introduce a compensa-
tion based on output with minimum guarantee so that more able
individuals can work harder without penalizing less able individuals.
This is exactly the two-tier compensation scheme that was analysed early
in the chapter. Such an option should induce both an incentive effect and
a sorting effect.
Safelite implemented option number 2, and introduced a plan that was
called PPP (Performance Pay Plan).
Note that Safelite could rely on a very sophisticated and digitally based
accountancy system in 1995. Its accounting system allowed management to
monitor in real time who installed what component. Since PPP has been intro-
duced over a period of nineteen months across all branches, most workers have
been observed under both regimes (fixed time input and PPP). The possibility
of observing workers before and after the policy change was introduced is an
ideal instrument for studying the effect on individual productivity linked to
the introduction of the PPP plan.
Table 7.1 reports some important basic numbers of the case study. The
guarantee wage corresponds to $11 per hour. This means that, conditional
on a minimum number of windows being installed, a worker was guaranteed
a wage equal to $11 per hour. This works as follows: if the weekly pay of
a worker turned out to be less than the guarantee, they would be paid the
122 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

Table 7.1. Data description for the Safelite case

Data description Definition Mean Standard

Variable deviation

PPP dummy A (dummy) variable equal to 1 if the worker 0.53

is on PPP during that month
Base pay Hourly wage $11.48 $2.94
Units-per-worker-per-day Average number of units of glass installed 2.98 1.53
by the given worker during the month
Regular hours Regular hours worked during the given 153 41
month
Overtime hours Overtime hours during the month 19 19
Pay Pay actually received in a given month $2,254 $882
Pay per day Actual pay per eight hours worked; this $107 $36
differs from PPP pay in that the wage
guarantee and other payments are included
Cost per unit Actual pay for a given worker divided by the $40 $62
number of units installed by that worker

Source: Lazear 2000.

Table 7.2. Basic result for the Safelite case

Variable Hourly wage Piece Rate-PPP

Mean Standard dev. Mean Standard dev.

Units-per-worker-per-day 2.7 1.42 3.24 1.59

Actual pay $2,228 $794 $2,283 $950
PPP pay $1,587 $823 $1,852 $997
Cost per unit $44.43 $75.55 $35.24 $49.00
Number of observations 13,106 15,246

Source: Lazear 2000.

guaranteed amount. Ex post, many workers ended up in the guarantee range.

On average installers were paid about $20 per unit installed. The data are
organized as follows. There were 2,755 individuals who worked as installers
over the nineteen-month period covered by the data. The unit of analysis is
a person month. Each month provides an independent unit of observation.
Over the nineteen-month period, there were more than 29,000 observations.

xit = Average numbers of glass units installed by individual i in month t

i = 1, . . . , 2,775
t = 1, . . . , 19
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 123

Note not all individuals are present in the nineteen months. Some individuals
have joined Safelite later than the first month. For example, if individual j has
joined Safelite at month 10, he or she will have observation xj1 = ., . . . xj10 = .
where the dot refers to a missing value (the individual was hired) for that
month. Similarly, if individual k leaves Safelite at month 15, he or she will
have missing values for the rest of the period. For each individual, it is
also possible to know whether he or she was working in the PPP regime or
not. Remember that the plan was phased in gradually in different branches,
so that for each individual the moment in which the PPP was operating
varies.

7.6.1 THEORETICAL PREDICTIONS

The setting of the Safelite changes fits perfectly with the introduction of a
two-tier regime, as was discussed in the previous section. There are thus two
key theoretical predictions linked to the introduction of the PPP plan.

1. Incentive effect : Output (or effort) of the workers should increase (or
not fall) as the company moves to a PPP system. Further, if there are
able individuals in the workforce, the effort increases (this is an incentive
effect).
2. Sorting effect : The average quality of the workforce increases (less able
individuals stay with the firm, but in addition it is possible to attract
better individuals into the firm) (this is a sorting effect).

Average productivity (as a consequence of the two previous effects) should

thus increase.

7.6.2 BASIC STATISTICS

Some basic characteristics of the sample are reported in Table 7.1. There
are a number of possible productivity measures. The one that most Safelite
managers look to is units-per-worker-per-day. This is the total number of
glass units per eight-hour day that are installed by a given worker. The units-
per-worker-per-day number for each individual observation relates to a given
worker in a given month. Thus, units-per-worker-per-day is the average num-
ber of units per eight-hour period installed by the given worker during the
given month. Means for actual and PPP pay reveal almost nothing about
the effects of PPP on performance and sorting. A more direct approach is
needed.
124 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

7.6.3 BASIC RESULTS

• The key basic results are reported in Table 7.2.
• The average number of windows installed (units-per-worker-per-day)
increases by 0.54 units (which corresponds to a 20 per cent increase).
• The variance of individual output increases (standard deviation increases
from 1.42 to 1.59).
• labour cost for unit of output falls from $44.43 to $35.24.
The first possible observation of the result could be the following: what
about a pure exogenous shift, linked for example to the adverse weather con-
ditions, that increases the demand for windows. At the same time as Safelite
introduced the PPP there was an exceptional increase in demand. If this was
the case, the basic effect would be the result of the demand pick-up. One way
to control for this possible demand effect is to include time dummies in the
basic average regression. The results show that productivity (measured as aver-
age number of windows installed per day), once time effects are introduced,
increases by a remarkable value of 44 per cent!!! This means that if there was a
demand effect, such an effect worked in the opposite direction.
In any event, one needs to investigate and rationalize a 44 per cent increase in
productivity. This effect can be seen by looking at the results of Table 7.3, in the
simplest specification in the first regression. The coefficient on the individual
PPP dummy is .368. Evaluated at the mean of the log of units-per-worker-per-
day, this coefficient implies that there is a 44 per cent gain in productivity with
a move to PPP.

7.6.4 POSSIBLE INTERPRETATIONS

These are our basic interpretations for the observed increase in productivity:
1. incentives: it depends on the pure incentive effect associated with the new
plan on the same labour force;
2. sorting: it depends on the fact that the average worker is different in the
two regimes, and better workers joined Safelite in the PPP regime;
3. Hawthorne effect; it is simply a short-run effect since any change inside
an organization leads to an increase in productivity.
We now analyse the three possible explanations.

The Pure Incentive Effect

To obtain an estimate of the pure incentive effect it is necessary to calculate
the effect only for the individuals that were operating under both regimes.
In other words, one should make the comparison of the productivity change
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 125

Table 7.3. Regression results for Safelite cases

Regression PPP dummy Tenure Time since PPP New hire R2

1a 0.368 0.04
s.d 0.013
2b 0.197 0.73
s.d 0.09
3c 0.202 0.224 0.273 0.76
s.d 0.009 0.058 0.018
4d 0.309 0.424 0.13 0.243 0.06
s.d 0.014 0.019 0.024 0.025

Notes: Dependent variable: ln (output-per-worker-per-day); observations: 29,837.

a
Regression 1: Dummies for month and year included.
lnxit = δt + γ PPPt + it
δt are time dummies;
γ PPPit is PPP dummy for individual i at time t
it is a white noise for individual i at time t
b
Regression 2: Dummies for month and year; worker specific dummies included (2,755 individual
workers).
lnxit = αi δt + γ PPPt + it
αi are individual dummies;
Regression 3;
c
lnxit = αi δt + γ PPPt + φ1 tenureit + φ2 timesincePPPit + it
tenureit is the tenure at Safelite for individual i at time t;
timesincePPPit is the time individual i spent in PPP;
Regression 4;
d
lnxit = αi δt + γ PPPt + φ1 tenureit + φ2 timesincePPPit + φ3 newhirei + it
newhirei is dummy if individual i was hired under PPP;
Source: Lazear 2000.

only for those individuals that were operating under both regimes. When
one performs such an exercise one finds that the increase in productivity is
equal to 22 per cent (which corresponds to 50 per cent of the total increase in
productivity).

Hawthorne Effect
The Hawthorne effect is described as the rewards you reap when you pay atten-
tion to people. Elton Mayo argued that the mere act of showing people that
you are concerned about them usually spurs them to better job performance.
Individuals’ behaviours may be altered because they know they are being stud-
ied; this was first argued in a research project (1927–32) at the Hawthorne
Plant of the Western Electric Company in Cicero, Illinois. The research, led by
the Harvard professor Elton Mayo, found that regardless of the experimental
manipulation implemented, the production of the workers under investigation
seemed to improve.
To control for the Hawthorne effect one has to control for the experience of
the worker in the firm as well as time since the individual has operated under
126 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

the PPP regime. Data suggest that the effect actually increases one year after
the introduction of the PPP (and if the change was only due to a Hawthorne
effect, we should have observed a decline).

Sorting or Learning?
The residual effect could be imputed to a pure sorting effect (better people
were employed by the firm), but we have also to consider that experience
is very important to the average productivity. One way to go about it is to
calculate the effect of the increase in productivity for the individuals hired
when the PPP was already introduced. Data suggest that the productivity of
the new hires (controlling for tenure or seniority) is much larger. In other
words, new hires are more productive.

7.7 Econometrics
The econometric model used to estimate the effect of the swap is the following
individual-based longitudinal equation

ln xit = αi + δt + γ PPPit + φ1 tenureit + φ2 timePPPit

+ φ3 new_Hirei + εit

where t = 1 . . . T are the nineteen months in which the firm is observed while
i = 1 . . . I are the 3,707 individuals who are observed. The representative
individual is the person month.

The dependent variable is ln xit , the log of output per day per person
(number of autoglass installed).
αi are individual fixed effects aimed at capturing a measure of individual
ability.
δt are time dummies aimed at capturing seasonal effects.
PPPit is an individual dummy that takes a value of 1 if the individual has
been in the PPP regime in that particular month. It’s the key variable for the
analysis of the incentive effects (row 2 in the regression table).
tenureit is the control for tenure. It is very important since productivity
increases with tenure.
timePPPit is a variable that takes the value of 1 if an individual has been in
PPP for 1 year, and 0.5 if for 6 months. It is crucial for the Hawthorne effect.
new_Hirei is an individual dummy that takes the value of 1 if the individual
has been hired after the introduction.
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 127

The various regressions in Table 7.3 report the various specifications. The
first regression includes only the PPPit dummy and it represents the estimates
of the pure PPP effect. The coefficient is positive and significant, and evaluated
at the mean of the log of units-per-worker-per-day; this coefficient implies that
there is a 44 per cent gain in productivity with a move to a PPP. Month and
year dummies are included in the regression, to account for factors, other than
the move to PPP, that can have a significant impact on productivity. In the
second regression worker-specific dummies are added. When worker dummies
are included in the regression, the coefficient drops to .197 from .368. The .197
coefficient is the pure incentive effect that results from switching from hourly
wages to piece rates. Evaluated at the means, it implies that a given worker
installs 22 per cent more units after the switch to PPP than he did before the
switch to PPP. This estimate controls for month and year effects. Regression 3
controls also for the worker tenure and the time since PPP, or the time spent
by each individual in the new regime. It shows that there should not be a
Hawthorne effect. The regression includes a variable for tenure and also one
for time that the worker has been on the PPP programme. The coefficient of
.273 on time since tenure coupled with a PPP dummy coefficient of .202 means
that the initial effect of switching from hourly wage to piece rate is to increase
log productivity by .202. After one year on the programme, the increase in log
productivity has grown to .475. The Hawthorne effect would imply a negative
coefficient on time since PPP. If the Hawthorne effect held, then the longer the
worker were on the programme, the smaller would be the effect of piece rates
on productivity. The reverse happens here. After workers are switched to piece
rates, they seem to learn ways to work faster or harder in time. The last row
of Table 7.3 provides evidence on the sorting effect, by measuring the output
of new workers hired. The variable labelled ‘New hire’ is a dummy set equal to
1 if the individual was hired after 1 January 1995, by which point almost the
entire firm had switched to piecework. The theory predicts that workers hired
under the new regime should produce more output than the previously hired
employees. Indeed, workers hired under the new regime have log productivity
that is .24 greater than those hired under the old regime, given tenure.

7.7.1 OTHER ISSUES

What about quality control? Safelite can fully control the quality of its output,
since it is very easy to count the number of windows which return to the dealer
because of a faulty installation. In the PPP plan the firm forces the guilty
workers to reinstall the faulty glass at their own expense using unpaid time
before the order is again reassigned. Actually, after the introduction of the PPP
a customer satisfaction survey suggests that there is an increase in customer
satisfaction (from 90 to 94 per cent).
128 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

b APPENDIX 7.1. THE BASIC FRANCHISING SCHEME WITH THE ABILITY

PARAMETER

Before entering the analysis of this chapter, we review the optimal franchising scheme
when ability is present in the analysis. This is easily done, once we recall from the
previous chapter the optimal linear contract w = α + βe is described by (replacing u
with au)

β
e∗ =
δ
β2
α = au −
2δ
β=p

where α is the fixed payment, β is the bonus scheme, and p is the price of output that
in this chapter we fixed at p = 1.Ì In light of the optimal scheme, the level of output
produced by an individual with ability a is indicated with x ∗ and reads

x∗ = a + e∗
1
x∗ = a + .
δ

The previous expression shows clearly that a firm that hires a more able individual will
produce more output under a franchising scheme which is done for all individuals by a
fixed payment plus an extreme bonus scheme. The profits of the firm for an individual
with ability a areÍ

E[] = px − w
1
= + ā(1 − u) (7.2)
2δ

so that also the firm makes more profits.

Ì In the previous chapter we saw that with a compensation equal to w = a +βx, the optimal amount
of effort is βδ . The chapter has also shown that the optimal choice of β from the firm’s standpoint is
β = p. Since we set p = 1, we have that β = 1 is the maximum.
Í The profits are obtained through the following steps

E[] = px − w
= p(e ∗ + ā) − α − βe,
= (p − β)e ∗ + p ā − α
 
β β2
= (p − β) + ā − au −
δ 2δ
1
= + ā(1 − u).
2δ
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 129

b APPENDIX 7.2. AN OPTIMAL SCHEME WITH A FIXED WAGE

Suppose now that the only scheme that the firm can implement is a fixed wage scheme,
so that the compensation must be equal to a fixed α that cannot vary with output.
In other words, the compensation that the firm may offer to its workers is just of the
type w = α without any bonus scheme. The firm cannot implement the franchising
scheme, since such compensation would require a wage that is negative if no output
is produced. Assume that the firm can easily observe output produced, but it cannot
observe the ability of the worker. As it will turn out, the heterogeneous ability is very
important. To show this we start asking the following question. What will be the level of
efficiency (and of profits) when a firm has to offer a fixed wage? The answer will depend
on two key conditions. What we want to show, as a benchmark case, is that the firm
that offers a fixed wage can obtain an efficient level of output if two conditions hold.
Such conditions are

• an appropriate minimum level of output xmin is chosen by the firm;

• there is no heterogeneity in the workforce. In other words, the workforce is
homogeneous.

Note that when we talk about an efficient level of profits, we mean a level of profits
that is as high as the one that would be obtained by a firm that implements the
franchising scheme, as indicated by equation 7.2.
Let’s assume for a moment that both conditions are satisfied. Homogeneous work-
force in this chapter means that the parameter a is fixed in the population, so that
a = ā, and for technical reasons we also assume that ā < 1/δ. In other words, all the
individuals have the same ability ā. Let’s see how the firm can choose a minimum level
of output xmin . The idea of a minimum level of output is simply to impose a minimum
standard of quantity, so that if a worker does not reach such a minimum standard, the
firm is not going to hire the worker, or the employment relationship can be interrupted
without cost. In this section we label a job [w, xmin ] as a job that pays a fixed wage
α = w and requires an output xmin . Let’s see how the firm operates. The problem is
solved in two steps. The first problem involves the choice of the minimum output xmin .
The second step involves the choice of the optimal wage w.
It is immediately clear that the first problem, the choice of a minimum quantity
xmin , has been already solved. With a fixed ability, all the firm should do is to select the
minimum output equal to optimal output x ∗ so that

xmin = x ∗
1
xmin = ā +
δ

The minimum output of the firm should be equal to xmin = ā + 1δ .

The second step involves the choice of the wage, and it requires looking at worker
behaviour. Let’s begin with a general question. If the firm offers the job [xmin , w] which
workers will be attracted by such a fixed wage schedule? The utility at the firm that pays
130 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS

w and requires a minimum level of output xmin is

δ(xmin − ā)2
U (w, xmin ) = w −
2
In order for the worker to participate in the market we need the level of utility (described
by the previous condition) to be larger than the worker outside option:

δ(xmin − ā)2
w− ≥ āu
2
The previous condition provides a participation constraint to the firm. Given xmin ,
the firm should choose a level of wage that makes the worker just indifferent between
working for the firm and his or her outside option (in other words the participation
constraint binds). If we call the optimal wage w ∗ the problem is solved by

1
xmin = ā +
δ
δ(xmin − a)2
w ∗ = āu +
2
The level of profits in this case isÎ

 = xmin − w ∗
1
= + ā(1 − u)
2δ
One can compare this level of profit with that obtained under a flexible system and
one immediately finds that they are the same.

If individuals are homogeneous (ā is fixed) the efficient level of profits can still be
obtained with a fixed wage plus a minimum output guaranteed.

What Happens When the Firm Cannot Choose the Wage?

If the wage is completely fixed and endogenous the firm can hardly make any choice.
All it can do is to select a minimum quantity of output xmin . The quantity is chosen by

Î The steps of algebra to obtain the results are the following

 = xmin − w ∗

1 δ(ā + 1δ − ā)2
= ā + − āu −
δ 2
1 δ 1
= − − āu + ā
δ 2 δ2
1
= + ā(1 − u)
2δ
 PAY FOR PERFORMANCE WITH WAGE CONSTRAINTS 131

the firm in such a way that the worker participation constraint is binding:

Maxxmin pxmin − w̄
δ(xmin − ā)2
s.t . w̄ − ≥ au
2
For maximization simply choose an x such that

δ(xmin − ā)2
w̄ − = āu
2
And the solution to the quadratic solution gives the minimum output. If profits are
positive the firm operates.
If the wage is completely fixed at w̄, all the firm can do is select a quantity so that
the worker participation constraint is binding, and operate if profits are positive.
A sufficient condition for the minimum level of ability al to be found is

δ(xmin − al )2
w− <0
2
To find the two levels of ability we need to distinguish two cases, which depend on
whether ah < xmin or ah > xmin . In the first case we need to solve for the following
quadratic equation:

δ(xmin − a)2
w− = au
2
δa 2 + 2a[u − δxmin ] + δxmin
2
− 2w = 0

whose solutions are


(δxmin − u) ± u − 2δ(xm u − w)
2 2
a=
δ
If the largest solution ah > xmin then ah is obtained as the solution to ah = w/u so
that

amin < al < xmin < ah < amax .

8 Relative Compensation
and Efficiency Wage

8.1 Introduction
Piece rates and bonus schemes provide incentives that operate independently of
the relationship among co-workers. Specifically, workers need not be working
with anyone else to be motivated by a piece rate scheme. Piece rate com-
pensation is based on an individual’s absolute performance rather than his
performance relative to some standard or some other individual. Yet, in real-
ity, most individual motivation is produced not by absolute reward but by
a compensation that is based on relative comparison. Managerial employees
who move up the corporate ladder do so by being better than their peers, not
necessarily by being good.
Comparisons are key in determining promotions in private firms. Since
promotions carry with them higher salaries, higher status, and perhaps more
interesting assignments, workers seek to get promotions. In this process, they
exert effort in an attempt to outperform their neighbours. Thus relative com-
pensation can provide an incentive that is as effective as a piece rate or
output-based compensation scheme based on individual performance.
There are good reasons why firms may prefer to use relative compensation
schemes. The first is that it may be easier to observe relative position than it
is to observe absolute position. Second, relative compensation differences out
common noise that risk-averse people may not like.
The most extreme form of relative compensation is a pure tournament, a
setting in which two workers compete for a promotion, an outcome that can be
described in terms of higher salaries. Workers are promoted if they do better
than their co-workers. In such a setting, the effort exercised by the workers is
proportional to the spread in prize, or to the increase in salary associated with
the promotion.
Pure tournaments are just one possible relative compensation scheme.
Indeed, the chapter introduces also the concept of a Linear Performance Eval-
uation (LPE), a situation in which workers receive a salary structure that is
the sum of two components: a fixed part plus a bonus (penalty) term, with
the bonus term linked to the average performance of the workers. There are
differences between LPE and pure tournament, particularly important when
players are heterogeneous. In LPE winning by a lot is important, since the
 RELATIVE COMPENSATION AND EFFICIENCY WAGE 133

larger the victory the larger the increase in salary. In pure tournaments, con-
versely, all that is important is winning. The implications of the two relative
compensation schemes are analysed in the context of a case study of broiler
production.
The chapter introduces also the concept of efficiency wage, a compensation
scheme whereby firms use salaries to increase workers’ effort in the context of
imperfect information. In a dynamic context, the efficiency wage mechanism
ensures that workers’ wages increase with tenure, even when productivity is
constant. Upward-sloping wage profiles appear particular relevant in real life
labour markets. Further, they form the basis of implicit contract between
workers and firms, whereby workers continuously exercise effort in view of
future wage increases.
The chapter proceeds as follows. Section 8.2 presents the basis tourna-
ment model. Section 8.3 introduces the relative compensation scheme, while
sections 8.4 and 8.5 are devoted to the case study of broiler production. Sec-
tion 8.6 deals with efficiency wages and section 8.7 with the upward-sloping
wage profiles.

8.2 Tournament: A Formal Story

Relative compensation theory, or tournament theory as it has come to be called,
is the theory used to determine the size of a rise associated with a particular
promotion. It has four essential features.
1. Prizes are fixed in advance and are independent of absolute performance.
The key is just winning the contest, not playing well, and typically the
prize awarded is independent of the difference between the two players.
In the context of the firm, this means that there are wage slots that are
fixed in advance.
2. A player receives the winner’s or loser’s prize not by being good or bad
but by being better than, or worse than, the other player. Again relative
performance rather than absolute performance is key.
3. The effort with which the worker pursues the promotion depends on the
size of the salary increase that comes with the promotion.
4. It is important to recall that the spread must be large to induce effort, but
the average prize money must be sufficiently high to attract workers to
come to the firm in the first place. Otherwise, workers will opt to enter
some other activity, since participation is not mandatory.
Consider a firm that has only two workers and sets up two jobs: boss and
operator. Workers compete against one another with the winner being design-
ated boss and the loser being designated operator. The winner receives wage
134 RELATIVE COMPENSATION AND EFFICIENCY WAGE

W1 , and the loser receives wage W2 . No wages are paid until after the contest
is completed. The probability of winning the contest depends on the amount
of effort that each individual exerts. Let the two individuals be denoted by j
and k.
Output of worker j is denoted by qj and depends on effort ej and on luck z
so that
1
qj = ej − z,
2
where z is a random variable with 0 mean. Similarly, output of worker k is
1
qk = ek + z
2
where z can simply be interpreted as relative luck with 0 mean. This implies
that expected output for worker j is E(qj ) = ej while expected output of worker
k is equal to E(qk ) = ek . Effort is chosen by the worker and it is not observed
by the firm, which observes only output.
We assume that z is uniformly distributed with zero mean over the interval
[−b, b]. Recall that if a variable is uniformly distributed over the interval
[−b, b] the following features hold:
E(z) = 0
b2
Var(z) =
3
x +b
G(x) = P(z ≤ x) =
2b
1
g (x) =
2b
where G is the cumulative distribution function and g is its density function.
Let’s define with P the probability of winning the context. For worker j, the
probability of winning the context depends on whether his output j is larger
than output k, so that
 
1 1
P(qj > qk ) = Pr ob ej − z > ek + z
2 2
= Pr ob(z < ej − ek )
= G(ej − ek )
ej − ek + b
=
2b
We are now in a position to solve the problem. We do this in two steps.
First, we model worker behaviour. Second, we solve for the firm maximization
 RELATIVE COMPENSATION AND EFFICIENCY WAGE 135

problem, taking worker behaviour into account. The second problem is thus
choosing an optimal compensation scheme to maximize profits.

8.2.1 WORKER PROBLEM

The worker utility is the expected wage minus the cost of effort. We assume that
the cost of effort is quadratic in effort, while the worker is risk neutral. In other
words, if P is the probability of winning the contest the worker’s pay-off is
δej2
U = W1 P + (1 − P)W2 − ,
2
where δ is the slope of the marginal disutility of effort. The worker chooses e
to maximize expected output so that his problem reads
δej2
Maxej W1 P(qj > qk ) + [1 − P(qj > qk )]W2 −
2
Since we know that P(qj > qk ) = G(ej − ek ) the problem becomes

δej2
Maxej W1 G(ej − ek ) + (1 − G(ej − ek ))W2 −
2
The FOC reads
∂G
(W1 − W2 ) = δej
∂ej
(W1 − W2 )g (ej − ek ) = δej
W1 − W 2
= δej
2b
∂G(e −e )
where we used the fact that j
∂ej
k
= g (ej − ek ) and that g (ej − ek ) = 2b
1

with a uniform distribution.

Note that there is a symmetrical problem for worker k which yields, after a
simple and similar simplification,
(W1 − W2 )g (ej − ek ) = δek .
Since the workers are ex ante identical and have identical first-order conditions,
there should be a symmetrical equilibrium in which both workers choose the same
level of effort, so that ej = ek = e ∗ . When ej = ek the first-order condition
becomes
(W1 − W2 )g (0) = δe ∗
136 RELATIVE COMPENSATION AND EFFICIENCY WAGE

Using the fact that the distribution is uniform, so that g (0) = 2b

1
, we obtain
the key condition
(W1 − W2 )
= δe ∗ (ICC)
2b
This condition, which determined the optimum amount of effort, acts as
an incentive compatibility constraint. From this condition, two important
conclusions follow:
Effort is larger the larger the wage spread.
Indeed an increase in the difference (W1 − W2 ) yields an increase in effort.
The more important luck is in determining the outcome, the lower is effort.
2
To see this recall that the variance of z is Var(z) = b3 . So the larger b is,
the larger is the variance of the luck component. This makes a lot of sense. If
most of your output depends on how lucky you are, why should you bother
to exercise effort? In production environments where measurements of effort
are noisy, large rises must be given in order to offset the tendency by workers
to reduce effort.
The worker has also a participation constraint. Note that under the sym-
metrical Nash equilibrium, each worker has a probability of winning the match
equal to G(0) = 1/2. This implies that, in equilibrium, a worker’s expected
utility is
W1 + W2 δe ∗2
U = −
2 2
and the worker will participate in the match (assuming that the outside option
is 0) as long as expected utility is positive, so that
W1 + W2 δe ∗2
≥ (Participation Constraint)
2 2

8.2.2 FIRM PROBLEM

The firm wants to maximize expected profit, or equivalently, profit per worker,
since the number of workers hired is assumed exogenous to this problem.
Expected output per worker is equal to E(q) = e ∗ while the average cost of
each worker is simply the average prize. This implies that the expected profits
of the firm are
W 1 + W2
E[] = e ∗ −
2
The firm tries to maximize profits by choosing W1 and W2 subject to the
participation constraint. Obviously the firm will set wages low enough to have
 RELATIVE COMPENSATION AND EFFICIENCY WAGE 137

the participation constraint satisfied with equality. In other words, the firm
problem is
W1 + W2
MaxW1 ,W2 e∗ −
2
W1 + W2 δe ∗2
s.t. =
2 2
which becomes simply
∗
δe 2
W1 ,W2 = e ∗ −
2
So that the first-order condition is
∂ ∂e ∗ ∂e ∗
= − δe ∗ =0
∂W1 ∂W1 ∂W1
 
1 e∗
= 1−
2bδ δ
∂e ∗
where we used the fact that ∂W 1
= 1
2bδ from the ICC, which suggests that when
wages are chosen optimally
1
e∗ = (efficient effort)
δ
A conclusion immediately follows:

Tournaments elicit the optimal amount of effort.

There are two ways to see this conclusion. First, firms force workers to
induce effort up to the point in which the marginal cost of effort (namely δe ∗ )
is equal to the marginal benefit (which is 1). Second, note that this condition
is the same condition that appeared in Chapter 6 on optimal wage contracts
when p = 1. Indeed, with a risk-neutral worker, you recall that the optimal
contract requires β = 1 so that optimal effort is equal to e ∗ = 1δ . This is the
same result that we have now obtained. In other words, a tournament is an
optimal compensation package, and satisfies all the conditions of the optimal
bonus scheme analysed in Chapter 6 for a risk-neutral worker.

8.2.3 OBTAINING THE WAGES

To obtain the wages, one needs to substitute the optimal level of effort into the
incentive compatibility constraint and in the participation constraint, so as to
obtain a linear system of two equations in two unknowns. In other words, the
138 RELATIVE COMPENSATION AND EFFICIENCY WAGE

system to solve is
∗
W1 + W 2 δe 2
=
2 2
(W1 − W2 )
= δe ∗
2b
where the optimal level of effort is e ∗ = 1δ . Using this result we get
W 1 + W2 δ1
= 2
2 δ 2
(W1 − W2 )
=1
2b
so that
1
(W1 + W2 ) =
δ
W1 − W2 = 2b

which gives the final wages and completes the problem

1
W1 = b +
2δ
1
W2 = −b
2δ

8.2.4 HETEROGENEITY AND RISKY STRATEGIES

If players are not identical there is a problem with tournament, since it becomes
very difficult to elicit effort from workers. Because of the natural advantage
that more able players possess, both individuals will not work hard enough.
The more able worker will tend to shirk since he is likely to win anyway, the less
able because he is likely to lose anyway. There is a way to solve this problem. It
is called a handicapping system. Such a system gives the less able player a head
start, so that makes it harder for the more able player to win.
Another feature of tournament with heterogeneous workers is linked to the
possibility of choosing strategies that affect not only the average probability of
winning, but also its variance. Imagine that two players can choose between
two strategies (call the two strategies safe and risky) that offer them the same
average probability of winning but have different variance. One can show that
with heterogeneous contest, the better individuals will avoid high-risk actions.
Choosing the risky actions (high variance) increases the chances of winning
the context but also makes it more risky. Since better players are likely to win
 RELATIVE COMPENSATION AND EFFICIENCY WAGE 139

anyway and winning by little or by a lot makes little difference, this option
offers little gain. As a result, more able players will play it safe and choose
a low-variance action. The reverse is true for less able players. Since losing
by a little or a lot is still losing and this is the likely outcome, expanding the
negative tail of the performance distribution has little cost. Less able players
should take a chance. Consequently, where tournaments encompass players of
unequal ability, there should be a negative relation between ability and variance
of performance.

8.3 Linear Performance Evaluation

To properly understand the role of tournaments and its specificity, particularly
with respect to the case studies we will examine, it is important to introduce
the concept of Linear Performance Evaluation (LPE). Under LPE players are
rewarded on the basis of relative performance, but there are some important
differences between a pure tournament and LPE. Suppose that there are many
workers and assume that the salary structure is obtained by the sum of two
components: a fixed part plus a bonus (penalty) term, with the bonus term
linked to the average performance of the workers. If an individual worker
produces and performs better than the average, he or she receives an increase
in compensation equal to m times the difference between his performance and
the average. In formula, such a reward scheme for individual i reads
δei2
Wi = W + m(ei − Q) −
2
where W is the contractually specified reward to a player with average per-
formance, m is the incremental reward (penalty) for above (below) average
performance, and Q is the average performance of all players. If individual i
views the average performance Q as fixed, he will choose ei such that
m = δei
m
ei =
δ
Notice that under LPE effort, and so performance, qi , depends positively on the
incremental reward for better relative performance, m, but is unaffected by the
average level of rewards, W . This feature holds for both LPE and tournaments.
In dimension LPE looks very much like a tournament.
The key difference between LPE and pure tournament comes about when
heterogeneity is relevant. The thing to realize is that in LPE winning by a lot
is important, while in pure tournaments all that is important is winning. This
140 RELATIVE COMPENSATION AND EFFICIENCY WAGE

implies that, unlike tournaments, LPE provides no incentives for better players
to choose conservative strategies and poorer players to choose risky ones, nor
does it provide an incentive for organizers to handicap better players. The
reason is that under LPE the incremental reward for improved performance is
the same whether a player is more able or less able. As a consequence, under LPE
there should be no relation between player ability and performance variability
nor should there be evidence of handicapping.

8.3.1 INDIRECT EVIDENCE OF RELATIVE COMPENSATION

Some evidence of the importance of relative compensation can be obtained by
examining the salary and compensation packages of corporate executives. It
seems reasonable to think that top managers’ rewards are based upon their own
firm’s performance relative to that of other firms in the same industry. Several
studies provide evidence that firms implicitly reward managers on the basis
of relative performance. An important further study is given by Gibbons and
Murphy (1992). They show that both compensation of chief executives and the
likelihood that they retain their position are related to firm performance relat-
ive to industry or relative to market. Thus, it is clear that relative performance
matters, and this is an important feature of tournaments.
Further support is found in the examination of the behaviour of professional
golfers by Ehrenberg and Bonanno (1990). Using data on professional golf
tournaments, the outcome of players’ actions and features of the environment
such as weather and course rating could be observed. The prediction of tourna-
ment theory that is tested is that increases in the differential between prizes
provide an incentive for more effort. Richer tournaments should lead to greater
effort and better performance (lower scores). The evidence appears consist-
ent with this hypothesis. As prize differentials increase, player performance
improves.

8.4 Evidence on Tournaments in Broiler Production

We now analyse a case study on relative compensation and tournament based
on an explicit business setting. It is the compensation of broiler producers,
and it is based on the research carried out by Knoeber and Thruman (1994).
Having detailed information on the structure of the tournaments and more
than one thousand observations on performance (outcome) reached by dif-
ferent players, three tests can be provided. The first test looks at the effect of a
change in the level of prizes holding prize differentials constant. In tournament
 RELATIVE COMPENSATION AND EFFICIENCY WAGE 141

theory there should be no effect on performance. This is consistent with the

discussion outlined above. The second one looks at the relation between the
riskiness of performance and player quality (ability). As we argued above, bet-
ter players should exhibit less variable performance. The third one looks at
evidence of handicapping or sorting when tournaments involve players with
different abilities. Better players should be handicapped or sorted into more
homogeneous tournaments.

8.5 Broiler Production

Broiler production is organized by firms called integrators. The integrator
hatches baby chicks. However, the integrator does not grow the chicks to broiler
size. Rather it contracts with growers to raise the chicks. Baby feed, medical
service, and advice are provided by the integrator. Chickens are owned by
the integrator. Chicken houses and labour are provided by the grower. The
birds are then transported to the integrator’s processing plant where they are
prepared for the market. The scheme in Fig. 8.1 describes the setting.
The contracts used to compensate growers provide a payment per pound
of live broiler produced. This is a piece rate. However, the size of the payment
varies among growers and is determined by relative performance. Basically the
integrator is the firm while the growers are the players. Grower performance
is measured by ‘settlement cost’ per pound of broiler produced. As we will
see, the smaller this settlement cost is, the better the grower’s productivity and
the better the grower’s performance of broiler produced. The most common
contracts in the broiler industry are LPE contracts, since they reward a grower
based upon the strength of his performance relative to the average of other
growers. But some contracts look only at performance rankings and so are
tournaments in the strict sense. The growers to whom one is compared are

Integrator (e.g. Purdue, Tyson)

chicks,
feed
broilers

Grower Grower Grower Grower

Figure 8.1. Broiler production

142 RELATIVE COMPENSATION AND EFFICIENCY WAGE

those whose flocks are harvested in the same geographical area at nearly the
same time.
An important reason for using relative performance schemes to compensate
broiler growers is that these schemes automatically difference out the effects of
common shocks to grower productivity. Since variations in temperature and
the breed of chick delivered by the integrator affect all growers, none must bear
the risk of such variation, and contracts are very simple.

The Integrator Data

The data include production information for growers under contract to one
integrator from November 1981 to December 1985. There is information on
1,174 flocks produced by 75 growers. The maximum number of flocks pro-
duced by any grower over the four-year period was 24. Of the 75 growers, 37
produced 20 or more flocks. For each flock, there is information on place-
ment and harvest dates, number of birds set, and performance by the grower.
Performance is measured by settlement cost which is defined as

settlement cost = (chicks × 12 + kilocalories × 6)/(live pounds)

This is a cost-per-live pound measure. Chicks are ‘charged’ 12 cents apiece

and calories (included in the integrator-provided ration) are charged 6 cents
per thousand. Settlement cost decreases (so that performance increases) as the
number of pounds of chicken produced per chick or per thousand calories
increases.
The contract under which broilers were produced changed in June 1984
from a pure tournament (that rewarded growers on the basis of ordinal rank)
to an LPE (that rewarded growers on the basis of performance relative to the
mean). Prior to 1984 growers whose flocks were harvested within a ten-day
period (in the same tournament) were ranked by performance (from smallest
to largest settlement cost). The ranking was then divided into quartiles, and per
pound payments to growers were determined based upon ranking. Growers
received an incremental per pound payment of 0.3 cents as they moved to the
next quartile. In the early part of the period the base pay was lowered and
ranged from 2.0 cents per pound to 2.45 in 1982. In June 1984 the form of
the contracts changed, and LPE was instituted. Now settlement costs were first
averaged. Grower pay per pound was then calculated as a base pay of 3.2 cents
plus or minus the difference between average settlement cost and the grower’s
own settlement cost. However a minimum per pound of 2.6 cents was still
guaranteed. In November 1984, per pound base pay was raised to 3.4 cents
per pound. The pay differential for better performance, however, remained the
same. There are basically four compensation regimes over the periods analysed.
 RELATIVE COMPENSATION AND EFFICIENCY WAGE 143

The tournament regimes can be expressed as

TRN1 = 2.25 + 0.3[Quart − 1] from 9-81 to 3-82
TRN2 = 2.7 + 0.3[Quart − 1] from 3-82 to 6-84
where Quart refers to the quartile in which each grower was partitioned. Under
the LPE the compensation regime was
LPE1 = Max[3.2 + (S̄ − si ); 2.6] from 6-84 to 11-84
LPE2 = Max[3.4 + (S̄ − si ); 2.8] from 11-84 to 12-85
where S refers to average settlement cost while si is the settlement of each indi-
vidual grower. Note that the remuneration increases if the settlement costs are
below the average settlement costs. Further, note the existence of a minimum
payment guarantee makes the compensation not a pure LPE scheme.

First Prediction: The Effect of Change in Base Pay Holding

Differential Constant
The settlement cost, the inverse of growers’ performance, is modelled as
Sit = µ + δi + αdaysit + βchicksit + γ1 TRN1 + γ2 TRN2 + γ3 LPE1 + γ4 LPE2
+ seasonal effects + it
where TRN1 and TRN2 refer to the period when pure tournament was in
place while LPE1 and LPE2 are the periods in which linear relative production
was in place. The ideas of these time dummies is to allow for differences in
mean performance due to changes in contract structures and level of base pay.
The variables days and chicks are numerical variables and represent the settle-
ment cost effect of longer grow-out periods (days) and larger flocks (chicks).
Numeric variables, grow-out length, and chick numbers significantly affect per-
formance. Days worsen performance by 0.055 cents per day while an additional
1,000 chicks worsen performance by 0.26 cents. The effect of these variables is
reasonable, and in line with what intuition would suggest.
The key test about the incentive effect of tournaments is on the period
dummies. Recall that periods TRN1 and TRN2 were distinguished by different
base payment amounts but had identical incremental rewards for moving up
the tournament ranks. Similarly periods LPE1 and LPE2 had different base
periods but the same marginal incentives. Nevertheless, the latter periods were
different since payment was determined by LPE rather than by tournament.
A test that looks for γ1 = γ2 and γ3 = γ4 is a test of the constancy of effort
(performance) between periods with different absolute levels of payment but
constant increments. It is not possible to reject this test, hence the results suggest
that performance is not affected by base pay.
144 RELATIVE COMPENSATION AND EFFICIENCY WAGE

Another important test is the one that looks whether performance is at all
affected by different payment periods. Indeed, mean performance varies across
periods, as suggested by the rejection of the test that γ1 = γ2 = γ3 = γ4 . In
other words growers do respond to different payment structures with different
levels of effort. Overall, this suggests that incremental rewards, and not absolute
levels, determine performance.

Grower Quality and Risk

Tournament theory has interesting predictions when heterogeneous players
do not choose only the level of effort but also the risk of the strategy chosen.
One way of thinking about the risk of the various strategies is to think that
players choose the variability. And theory predicts that a high-quality player
will adopt a low-variance strategy to assure his high ranking. Conversely, a
low-quality player will adopt a high-variance strategy. Since the data set used
contains multiple observations on individual growers (many growers have
more than twenty flocks) the settlement regression provides measures of both
mean performance and variance. The idea is to interpret the mean performance
measure as a proxy for grower quality and investigate the relationship between
quality and variability.
Figure 8.2 plots growers’ variability against grower quality during the period
in which tournaments were used to compensate growers. The measure of
variability is σi , the sample residual standard deviation for grower i from
the settlement cost regression, and each σi is calculated only from the flocks
grown during the tournament period. The measure of grower quality is the

1.4

1.2
d (std. dev. of performance)

0.8

0.6

0.4

0.2

0
–2 –1.5 –1 –0.5 0 0.5 1 1.5
d (grower quality decreases left to right)

Figure 8.2. Variability of grower performance and grower quality: tournaments

 RELATIVE COMPENSATION AND EFFICIENCY WAGE 145

2
1.8
d (std. dev. of performance)

1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
–2 –1.8 –1.6 –1.4 –1.2 –1 –0.8 –0.6 –0.4 –0.2 0 0.2 0.4 0.6 0.8 1 1.2
d (grower quality decreases left to right)

Figure 8.3. Variability of grower performance and grower quality: LPE contracts

estimate of δi , the dummy variable coefficient for grower i from the settlement
cost regression. Remember that the best growers are those with the lowest
settlement cost and the worst are those with the largest. Despite the noise
present in Fig. 8.2, it obviously slopes upward. Better growers choose less
risky strategies. During the LPE period, growers were paid based on their
performance relative to the average of all growers. The reward for improved
performance was the same for high-quality and low-quality growers. As a
consequence no differential incentive existed for low-quality growers to take
risks or for high-quality growers to play safe. In principle, there should be no
relation between σi and δi in the LPE regimes. Nevertheless, in the LPE regime
there was a minimum payment provision, so that the per pound payment
was truncated from below. The effect of this is to restore the incentive for low-
quality growers to take risks. Therefore, low-quality growers did have incentives
to adopt high-variance strategies during the LPE regimes, and this seems to be
borne out in Fig. 8.3. However, most of the positive relationship seems to be
due to the worse growers. To the left of δ = −0.4 the relationship is quite flat
while in Fig. 8.2 it is more uniform. This suggests that there was an effect of
the minimum payment provision on risk taking by low-quality growers during
the LPE period. Nevertheless, there is some qualitative difference in the two
regimes.

Handicapping
When players with different qualities interact in tournaments, organizers
should try to handicap better players, so that all players have the right incentive
146 RELATIVE COMPENSATION AND EFFICIENCY WAGE

Table 8.1. Regression results for broiler production

Dependent variable: settlement cost (cents/live weight pound)

n=1167; mean=20.8. Estimates: α =0.55 (.008); β =0.26 (0.08)
Test of payment period effects
1. Mean performance identical between periods TRN1 and TRN2
(γ1 = γ2 )and periods LPE1 and LPE2 (γ3 = γ4 ) Not rejected
2. Mean performance identical across all periods γ1 = γ2 = γ3 = γ4 Rejected
Test of grower heterogeneity:
3. Growers have identical mean performance (δ1 = δ2 = · · · = δ77 ) Rejected

to perform. There is no similar reasoning to expect handicapping in an LPE

regime, unless there is a minimum payment provision. When a minimum
payment provision is in place in an LPE regime, there is some incentive to
handicap good players, but such incentives should be lower than in a pure
tournament setting. The regressions of Table 8.1 suggest that integrators have
two possible ways to handicap players: grow-out length and flock size. Both
variables have an unfavourable effect on performance and are controlled by
the integrator, and can be used for handicapping players. We show the results
with the number of chicks with the following linear model

chicksit = ν + θ1 δi + ρ1 TRN + γ1 δi TRN + seasonal effects + it

where δi is the quality variable of grower i and the variable δi TRN looks
at the relationship between chicksit and performance under the tournament
period. If there is some handicapping we would expect a negative coefficient
on δi (negative θ1 ), with better growers (lower δi ) being given larger seasonally
adjusted flocks. This is because flock size is indeed negatively related to δi .
Indeed the coefficient θ1 is negative, and suggests that better growers are given
larger flocks. While this is consistent with handicapping it could also simply
mean that better growers have more houses. But the key test is on the interac-
tion effect between quality and the tournament periods. Since the incentive to
handicap is greater under the TRN regime, one might expect this coefficient
to be negative, suggesting that handicapping is stronger under tournaments
than under the LPE regime. If a grower is good, it should have more chicks
under tournament, so that low delta should be associated with high chicks in
the tournament dummy. The coefficient γ1 is indeed positive.

8.6 Efficiency Wage

The idea behind efficiency wage is that higher wages increase workers’ pro-
ductivity. There are several potential reasons for this phenomenon. In general,
 RELATIVE COMPENSATION AND EFFICIENCY WAGE 147

higher wages generate greater productivity through a commitment mechan-

ism. For example, the higher the wages are relative to what the worker could
get elsewhere, the less likely it is that the worker will quit. Another key reason
has to do with supervision. The higher the wage, the larger the cost of being
caught shirking. Indeed, employees realize that even though supervision may
not be detailed enough to detect shirking with certainty, if they are caught
cheating on their promises to work hard and are fired as a result, the loss of
a job paying above market wages is costly. As a result a worker may have less
incentive to shirk. The idea of high wage as a discipline device is exploited in
this section.
The worker’s utility depends positively on the compensation received and
negatively on the effort exercised. Effort is chosen by the worker and can take
two values. For simplicity we assume that e is a binary variable that can take
two values e = {0, 1}. If the worker exercises effort he or she suffers a subjective
cost equal to c/2 while he or she does not suffer any cost if e = 0. If the worker
chooses e = 0 we say that the worker shirks. The utility function is described as
c
U = compensation − e
2
where compensation is the income received by the worker.
The worker has an outside option equal to u ≥ 0.
The firm worker pair has established a job and produces a labour product
equal to y only if the worker exercises effort. If the worker does not exercise
effort the value of the labour product is 0.
Effort is observable to the worker but it is not observable to the firm. All the
firm can do is monitor the worker. We assume that the monitoring technology
is such that the firm has a probability p of finding out whether the worker
exercises effort.
If a worker exercises effort and produces the labour product y the firm offers
him or her a wage w. The wage w is set unilaterally by the firm.
A worker that is caught shirking is immediately fired and he or she gets the
outside wage u.
The compensation of the worker is then simply

w if e = 1
compensation =
pu + (1 − p)w if e = 0
The firm’s profits are defined as the difference between the value of the
labour product and the wage, and they are positive only if the worker does not
shirk. In formula, this is equivalent to

y −w if e = 1
π=
−(1 − p)w if e = 0
The firm has a clear interest that the worker does not shirk, since otherwise
its profits are going to be negative.
148 RELATIVE COMPENSATION AND EFFICIENCY WAGE

The worker will choose to exercise effort only if the utility with effort is
larger than the utility under shirk. This is equivalent to:
 c
w− ≥ (1 − p)w + pu
2
where the left-hand side is the worker utility if the worker chooses e = 1 while
the right-hand side is the utility if e = 0. Rearranging, the worker will exercise
effort if and only if
c
w ≥u+ (no shirking)
2p
The previous condition is the fundamental equation of the efficiency wage,
since it says that the worker will exercise effort only if the wage is large enough.
In this setting, the only instrument that can induce the worker to exercise effort
is a high wage. The logic goes as follows. Worker’s dislike effort but if they are
caught shirking they suffer a penalty equal to the wage loss. As a result, the
larger the wage the lower the chance that the worker shirks.
Looking at the firm’s profits, it appears that the firm would like to set the
lowest possible wage, conditional on the worker choosing e = 1. It implies that
the optimal wage is the minimum wage that satisfies the no shirking condition,
and reads
c
w =u+
2p

The efficiency wage is above the worker’s outside option.

The problem solution is the following:

• The worker chooses e = 1 and gets a wage equal to w = u + 2p

c

• The firm profits are π = y − (u + 2p

c
) and the job takes place as long as
π ≥0

From the solution to the model a first conclusion follows:

The efficiency wage increases the more difficult it is monitoring the worker

To establish this conclusion simply note that the lower p is, the larger is
the wage. The intuition is that the worker needs to be compensated more the
lower the probability of finding out whether he or she is shirking. This suggests
that the asymmetric information over effort leads to an increase in wage. To
investigate this further we can define the worker surplus as the difference
between the utility of the worker and his outside option. The worker surplus
is indicated with Sw and its expression reads

Sw = U − u.
 RELATIVE COMPENSATION AND EFFICIENCY WAGE 149

Using the efficiency wage, the total surplus reads

c c
Sw = (u + ) − − u
2p 2
c(1 − p)
Sw =
2p
from which a conclusion immediately follows:
The worker enjoys a positive rent as long as p < 1.
As we argued above, even though the firm unilaterally sets the wage, the pres-
ence of asymmetric information leads to a wage that leads to a pure economic
rent, with a utility larger than the outside option.

8.7 Deferred Compensation and Upward-Sloping

Wage Profile
We now use the model of the previous section in a dynamic setting. We assume
that there are two periods in the worker’s career.
The first period corresponds to a period in which the worker is ‘young’ and
the second period corresponds to a period in which the worker is ‘old’.
In each period the worker has a utility function that is identical to the utility
described in the previous section, and it is the sum of the compensation minus
the cost of effort. We label wy the wage in the first period, when the worker is
young, and wo the wage in the second period, when the worker is old. The cost
of effort in each period is 2c e. The young worker has a lifetime utility that is
given by the sum of the per period utility of the young and the old. In formula,
the utility of the young worker is
 c   c 
Uy + Uo = wy − ey + wo − eo
2 2
where eo is the effort chosen when the worker is old and ey that chosen when
he is young.
The other assumptions of the model are identical to those of the previous
section. In particular, effort is chosen by the worker but it is not observed
by the firm, which can monitor, in each period, the worker with a frequency
p < 1. The worker, if he does not shirk, produces in each period a value of the
labour product equal to y. The worker has an outside option in each period
equal to u. The problem is described in Fig. 8.4.
The firm sets the wage for the young and the wage for the old, wy and wo .
One of the key questions we ask in this section is whether wy is different from
wo and notably whether wy < wo .
150 RELATIVE COMPENSATION AND EFFICIENCY WAGE

Firm offers Firm offers

wage wy Monitoring wage wo Monitoring

Worker chooses Worker chooses

effort ey = {0,1} Production effort eo = {0,1} Production

Period 1 Period 2

Figure 8.4. Deferred wages

The problem should be solved backwards, starting from period 2 and going
backwards to period 1.
In the second period the problem is like a static problem, since there is only
one period to go and thereafter the game ends. This suggests that in the second
period the problem is identical to the efficiency wage model discussed in the
previous paragraph. In formula the wage for the old worker is
c
wo = u +
2p
The key problem is for the firm to choose the young wage and for the worker
to decide whether effort should be chosen or whether shirking is optimal. The
firm can take into account the second period wage and knows that in the
second period the worker will not shirk and will be paid a wage equal to wo .
Let’s focus on the worker. Suppose that the firm pays a wage wy (which is still
to be determined). The worker will choose to exercise effort if his lifetime utility
with effort is larger than the lifetime utility without effort. This is equivalent to
 c  c   c 
wy − + wo − ≥ (1 − p) wy + wo − + p2u
2 2 2
The left-hand side of the previous condition is the utility when the worker
chooses effort and receives wy when young and wo when old. The right-hand
side is the utility if the worker shirks in the first period, while he or she will not
shirk in the second period, since we have already established this. Rearranging
the previous expression one gets
c (1 − p)c
(wy + wo ) ≥ − + 2u
p 2p
and using the solution for the old wage the solution is
c (1 − p)c
wy ≥ u + − (8.1)
2p 2p
The firm’s problem is now easy to determine, since the firm will choose the
minimum young wage so that the worker does not shirk. This is identical to
 RELATIVE COMPENSATION AND EFFICIENCY WAGE 151

satisfying equation 8.1 with an equals sign so that

c (1 − p)c
wy = u + −
2p 2p
The firm defers wages, and offers to the worker an upward-sloping wage profile
with wy < wo .
The proof of the remark is simple, since it follows from a simple comparison
of the wage of the young worker and the wage of the old worker. Indeed,
wy = wo − ((1 − p)c/2p).
The reason for the upward-sloping wage profile can be understood by look-
ing at equation (8.1), which represents the no shirking constraint in a dynamic
setting. The firm knows that the worker, by shirking, loses not only the wage of
the period 1 wy but also the wage in the old phase. As a result the firm is able
to reduce the wage of the young worker. The worker on the job has incentives
to work diligently in order to qualify for the wage increase.
9 Training and Human
Capital Investment

9.1 Introduction
Lifetime education, as a form of a lifelong process of learning, is a key concept
in modern organizations. It is often argued that in a competitive and changing
world the only successful firms are those that employ workers that continuously
upgrade and adapt their skills. While such a view may look extreme, it is obvious
that lifelong learning is a key dimension in all organizations. Just think about
how difficult it is for older workers to adapt to a digital economy without the
appropriate skills. This means that education, the process of acquiring skills,
does not finish with graduation, but it continues throughout the working
careers of most workers. And most of such education takes place during the
employment relationship, in the form of training. In other words, training is a
key decision in personnel economics.
The current and the following chapter deal with on-the-job training. In this
chapter, the basic models of training are set up.These models are known in the
literature as Becker models of training. A common feature of the models of this
chapter is the assumption of a perfect labour market. In the following chapter,
we will analyse how the training decision changes when the underlying labour
market is imperfect.
To understand how economists think about training, it is first necessary
to understand how economists think about education in general. The most
important view of education is the theory of human capital. Loosely speak-
ing, human capital corresponds to any stock of knowledge or characteristics
the worker has (either innate or acquired) that contributes to his or her ‘pro-
ductivity’. This definition is very broad, but it enables us to think not only
about the years of schooling, but also about a variety of other characteristics
as part of human capital investments. Training done during the employment
relationship is one of such characteristics, and it is the core analysis of this
chapter.
Before entering into the theory of training, the chapter starts from a simple
model of human capital that enables us to understand the individual’s decision
to acquire education. The human capital approach to education (which is
common to this and the following chapter) assumes that acquiring knowledge
 TRAINING AND HUMAN CAPITAL INVESTMENT 153

and education by individuals is a process very similar to the accumulation of

physical capital by organization. This is the key concept of the theory of human
capital. Education involves first some costs (the investment phase) and some
benefits thereafter (mainly in the form of larger future wages). We will see that
such a view, extremely influential in the modern economic analysis, delivers a
variety of insights into the real life process of skill accumulation.
Human capital investment continues in the aftermath of graduation,
through on-the-job training. This is the core part of the chapter. In outlining
the basic ideas, the chapter refers to the logical distinction between general
training and specific training. General training refers to the acquisition of a
set of skills that can be used in each and every organization (or firm) active in
the labour market. Specific skills are at the opposite extreme. An investment in
specific human capital leads to the accumulation of skills that can be efficiently
applied only in the organization in which the worker is currently employed.
Such a distinction is very important for understanding the firm and worker
decision to engage in the on-the job accumulation of human capital.
In the case of general human capital, as long as the labour market is perfectly
competitive and wages fully reflect the marginal product of labour, a firm
has no incentive to invest in the general human capital of its workforce. In
such a case, the investment incentives rest on the workers, who can optimally
undertake the investment by accepting initially lower wages.
In the case of specific human capital, things are more difficult and more
complicated, since neither party has an incentive to undertake the entire cost
of the investment. Indeed, we will see that the only way by which an oppor-
tunity to invest in specific human capital is fully exploited is when both parties
share both the costs and the benefits from the investment. Specific human
capital creates firm-specific rents, and it is the basis for long-term employment
relationships.
The chapter proceeds as follows. Section 9.2 presents the baseline model
of schooling, human capital investment before the labour market. Section 9.3
defines general and specific investment, while section 9.4 presents a two-period
model of general training. Section 9.5 focuses on specific training, and formally
defines the concept of a specific rent. Section 9.6 presents some empirical
findings on training.

9.2 The Basic Human Capital Model of Schooling

Let’s consider an individual who has to decide whether to acquire some level
of education. To be specific, let us imagine that we are considering the decision
of a just-graduated student to take a degree or not. The individual has already
154 TRAINING AND HUMAN CAPITAL INVESTMENT

completed secondary education and has to decide whether or not he should

continue his investment in education. Formally, we consider an individual who
lives two periods and must decide whether or not to acquire education. His
utility is given by the difference between the lifetime income and the costs of
education. In formula, this utility function reads:

Ui = Lifetime income − costs of education

y2
Ui = y1 + − e(γ + θi )
1+r
Let’s first focus on the lifetime income, which we assume to be just two periods.
We let y1 be the income in the first period while y2 is the income in the second
period. Note that the income of the second period is discounted at the rate of
time preference r, which we assume to be identical to market interest rate r.
A single unit of income in period 2 (say 1 euro) yields a utility today of just
1/(1 + r).É We next focus on the educational choice, which we label e. Since
we are just focusing on the decision to undertake a degree, we assume that the
education variable can take only two values. In other words, we assume that
an individual can be either educated or not (either he or she has the degree
or not), and so e = {0, 1}. The disutility (or the cost of education) has two
components: γ is the tuition fee component while θi is the psychological and
subjective cost, so that individuals with larger values of θi have the largest cost
of education.
The income in the first period of life is equal to y1 = −γ if the worker
acquired education, while it is equal to the unskilled wage wu if the worker
does not undertake education. In this respect education is a full-time activity.
The wage in the second period depends on whether the individual did or did
not acquire education in the first period, so that w2 = ws (the skilled wage)
if the worker acquired education while w2 = wu (the unskilled wage) if the
worker did not educate himself or herself. The benefits of the education rest
with the increase in workers’ productivity, reflected in their higher wages.
The education decision is then solved by comparing the lifetime utility of
the individual with or without education. In other words, the individual will
É To understand the discount factor, consider the following experiment. Suppose that the market
pays an interest rate equal to r for an investiment of 1 euro between the first period and the second
period (e.g. one year). This means that 1 euro today will yield to 1(1 + r) euros tomorrow. Similarly,
one can ask what is the value today of 1 euro available in the second period. Call such value x. By
definition, it must be true that

x(1 + r) = 1

from which it follows that

1
x=
1+r
In other words, 1 euro available in the second period is equivalent to 1/(1 + r) euros immediately.
 TRAINING AND HUMAN CAPITAL INVESTMENT 155

educate himself if

utility with education ≥ utility without education

ws wu
−γ + − θi ≥ wu +
1+r 1+r
which requires
ws − wu
(θi + γ ) + wu ≤
1+r
Several important insights come from the previous inequalities:
1. The decision of whether or not to acquire human capital is identical to
that of acquiring physical capital, and it is governed by a cost–benefit
analysis.
2. The cost of education has three components: the direct tuition costs
during the education period γ , the opportunity cost of a market wage
during the education investment wu , and the psychological cost θ .
3. The benefits are expressed as net marginal benefits, and they are given by
the value of the skilled wage minus the unskilled wage.
Many implications of this simple theory follows.
Not everybody will educate themselves. People who have too large educa-
tional costs will choose not to undertake the human capital investment. To see
this assume that there is heterogeneity over values of θi , a variable that can
easily indicate inability, so that more able people are those with lower values
of θi .Ê It follows that the investment in education will be undertaken only by
people with θi ≤ θ ∗ where θ ∗ is the individual who is just indifferent to the
education investment. The formal expression of θ ∗ reads
ws − w u
θ ∗ = −γ − wu +
1+r
The psychological cost for the marginal individual is such that it just offsets
the monetary costs (in terms of tuition fees and opportunity cost and the net
benefits).
If educational costs increase (an increase in γ ) fewer people enter the edu-
cational system. To see this note that an increase γ reduces θ ∗ . If taking a
degree is more expensive, fewer people will attend the educational system.
An increase in the interest rate means less schooling. Note that from the role
of r it may also follow that there are some people that face very large interest
rates independent of ability, and will not acquire education. An additional

Ê Assume that θ can take any value between θmin and θ max and there is a continuum of individuals
distributed according to the cumulative density function F (θ ) defined over the support θmin , θ max .
156 TRAINING AND HUMAN CAPITAL INVESTMENT

interpretation of the role of financial constraint can come from interpret-

ing θ as the inability to borrow, so that individuals that have more financial
constraints have larger θ.
A larger skill premium (i.e. an increase in the difference ws − wu ) means
that a larger amount of people will undertake education.
Finally, a note on the timing of the education investment. In this simple
model, education will never be taken in the second period, since there is no
time to earn back the investment. This implies that education will always be
undertaken when people are relatively young.

9.3 General versus Specific Investment

Individuals continue to invest in human capital after the start of employment,
and we normally think of such investment as training, provided either by the
firm itself on the job, or acquired by the worker (and the firm) through voca-
tional training. While what is learned in school is typically widely usable, what
is learned through training is often less general, and can be used only in specific
industries or in specific jobs. Further, while the schooling decision is mainly
an individual decision, training decisions involve also the firm. Economists
typically distinguish between two types of training:
• firm-specific training: this provides a worker with firm-specific skills,
or skills that will increase his or her productivity only with the current
employer;
• general training: this type of training will contribute to the worker’s gen-
eral human capital, increasing his or her productivity with a range of
employers.
We now present two different models of training, one that focuses on general
training and one that focuses on specific training.

9.4 General Training

9.4.1 THE BASIC SET-UP

Consider the following very simple two-period model:
We focus on firms that are formed by single jobs. A firm and a worker are
engaged in a production opportunity that we simply call ‘job’. The job has been
already established. We do not deal here with job creation.
 TRAINING AND HUMAN CAPITAL INVESTMENT 157

Worker stays or
Production takes quits at no costs
place, net of the
Firm decides training y –ct.
training level t Wage paid Wage offer w Production
takes place, y + t

Period 1 Period 2

Figure 9.1. A two-period model on general training

In the first period there is a level of production equal to y, where y is

the value of the marginal product to the firm. In addition, there is a training
opportunity in a fixed amount. The training opportunity can either be taken
or not. Training is indicated by τ with τ = {0, 1} so that it is equal to 1 if
training is undertaken and it is equal to 0 if training does not take place. There
are two stages in the first period.
• In the first stage the firm decides whether to undertake the training oppor-
tunity. Undertaking training costs c, where c is expressed in the same unit
of the production. The total costs of training are then cτ .
• In the second stage production takes place, for a value equal to y − cτ .
In the second period there are three stages:
• In the first stage the firm makes a take it or leave it wage offer w to the
worker, and other firms compete for the worker’s labour.
• In the second stage the worker decides whether to quit and work for
another firm. We assume that the worker does not incur any cost in
the process of changing jobs, so that the labour market is perfectly
competitive.
• In the third stage production takes place. At this point the value of output
is equal to y + f τ ; in other words, if training did take place in the first
period (τ = 1) it is equal to y + f ; if training did not take place (τ = 0),
it is equal to zero. We assume that f > c.
For simplicity we assume that there is no discounting. Figure 9.1 describes
the timing of the game.

9.4.2 THE EFFICIENT LEVEL OF TRAINING

Before proceeding, let’s clearly establish that in the model undertaking training
is an efficient choice. This is ensured by the fact that f > c. Undertaking
158 TRAINING AND HUMAN CAPITAL INVESTMENT

training is an investment that has a positive net benefit (f − c > 0) and

would certainly be undertaken by an agent that weighs the costs and benefits
associated with the investment.

9.4.3 THE SOLUTION OF THE MODEL

There are two issues. First, we need to understand whether training will be
undertaken. Second, and perhaps most important from the personnel econom-
ics standpoint, we need to understand who will have the incentive to pay for it.
The model has to be solved backwards, starting from the second period and
proceeding backwards to the first one. We begin at the first stage of the second
period. In period 2 the level of training has been already decided (whether a
positive value or zero). This means that the productivity of the worker will
be y + f if training was undertaken (τ = 1) and y otherwise. Suppose for a
moment that τ = 1. If the market is competitive and if there are no costs of
changing jobs, it is clear that the wage in period 2 must be exactly y + f , or
the marginal product of labour with training. This is a key condition. As the
economists say, ‘bygones are bygones’, and at t = 2, the worker will be trying
to obtain the largest possible salary in the marketplace, independently of what
was agreed earlier on the financing of the training investment. This is obvious,
since any wage smaller than this level would leave the worker with a strong
incentive to change job right away. The solution of the second period at time
t = 2 is then very simple. If training did take place, the worker will accept
from the current firm (or from a different firm) only a wage offer equal to
w1 = y + f . If training did not take place, the wage would be equal to y, the
marginal product of labour without training.
Let’s now move back to the first period. Since in the second period the
worker will get all the benefits from the investment, the firm will not have
the incentive to sponsor the training. The firm would certainly make a loss on
sponsoring training, since it would incur a certain cost c at time t = 1 without
incurring any of the benefits at t = 2. Hence the firm will not finance the
training. But we need also to consider the worker’s incentive. Since the worker
is the full residual claimant of the increase in his or her own productivity, he or
she will have the right incentive to invest.
How can the worker finance the training? Simply through a wage cut in
the first period, so as to compensate the firm for the loss of production during
training. This is the key mechanism of the solution and an equilibrium is easily
established. The firm will offer to the worker the following package at t = 1,
τ = 1 and a wage equal to w1 = y − cτ . In this situation, the worker accepts
the deal, since he will get the full wage
w2 = y + f
 TRAINING AND HUMAN CAPITAL INVESTMENT 159

wage
Wage profile
with training
y+f

y Wage profile
without training

y–c

t=1 t=2 time

Figure 9.2. Wage tenure profile with general training

in the second period, either from the current or from another employer. The
firm makes zero profits, and has no incentive to deviate from such a situation.
To sum up, the equilibrium is described as follows. At time t = 1 the firm
chooses τ = 1 and offers a wage equal to w = y − cτ . At time t = 2 the
firm offers a wage w1 = y + τ f and the worker accepts. Figure 9.2 plots the
solution, with wage on the vertical axis against time on the horizontal axis.
With perfect competition in the labour market, the efficient level of general
training will be achieved with firms bearing none of the costs and workers
financing training by taking a wage cut in the first period of employment.
Two very important implications follow:
• Human capital investors forgo wages early on in their career in exchange
for future wage gains.
• The wage profile becomes positively sloped: workers earn less when they
are young but experience makes wages grow.
In section 9.6 we discuss some of the empirical evidence on training.

9.4.4 INVESTMENT IN GENERAL TRAINING: AN EXAMPLE

Let’s now consider a worker that is currently paid according to his marginal
product, which we set equal to 100 in Table 9.1. The worker is expected to
be hired for two years and the interest rate is zero, so that there is no issue of
discounting. There is a training opportunity in general skills that involves a loss
of productivity at time t = 1 of 15, but it generates an increase in productivity
160 TRAINING AND HUMAN CAPITAL INVESTMENT

Table 9.1. General and specific training

General training
y = 100, c = 15, f = 25
Marginal prod. Outside option Wage Firm surplus Worker surplus Total surplus
τ =0 τ =1 τ =0 τ =1 τ =0 τ =1 τ =0 τ =1 τ =0 τ =1
t = 1 100 85 100 100 100 85 0 0
t = 2 100 125 100 125 100 125 0 0

Specific training (without sharing)

y = 100, c = 15, f = 25
Marginal prod. Outside option Wage Firm surplus Worker surplus Total surplus
s=0 s=1 s=0 s=1 s=0 s=1 s=0 s=1 s=0 s=1 s=0 s=1
t = 1 100 85 100 100 100 85 0 25 0 −15 0 10
t = 2 100 125 100 100 100 100 0 25 0 0 0 25

Specific training (ex post sharing)

y = 100, c = 15, f = 25, beta = 0.5
Marginal prod. Outside option Wage Firm surplus Worker surplus Total surplus
s=0 s=1 s=0 s=1 s=0 s=1 s=0 s=1 s=0 s=1
t = 1 100 85 100 100 100 85 0 12.5 0 −2.5 0 10
t = 2 100 125 100 100 100 112.5 0 12.5 0 12.5 0 2.5
Specific training (full sharing)
y = 100, c = 15, f = 25, beta = 0.5
Marginal prod. Outside option Wage Firm surplus Worker surplus Total surplus
s=0 s=1 s=0 s=1 s=0 s=1 s=0 s=1 s=0 s=1
t = 1 100 85 100 100 100 92.5 0 5 0 5 0 10
t = 2 100 135 100 100 100 117.5 0 12.5 12.5 25

at time t = 2 of 25. The training opportunity is thus efficient. Undertaking

training increases the worker’s outside option at time t = 2 to 125. This means
that the worker can obtain in the second period a wage equal to 125, and the
firm will not have any incentive to pay for training in the first period. The only
equilibrium with training is one in which the worker takes a wage cut at time
t = 1 equal to 15 and gets a wage equal to 125 in the second period. Table 9.1
confirms this result.
There are several important lessons from this simple example, very similar
to the model above. Investing in human capital leads to a reduction in wages
early on in the career in exchange for future wage gains. Second, the wage
profile becomes positively sloped: workers earn less when they are young but
experience (or tenure) makes wages grow, exactly as is implied by this model.

9.5 Firm-Specific Training

The problem of financing general training was that the worker is the resid-
ual claimant of the investment and that any firm in the market could benefit
from such an investment. We have seen that the equilibrium is one in which
the worker takes a wage cut and finances the general training. In the case of
 TRAINING AND HUMAN CAPITAL INVESTMENT 161

firm-specific training the situation is the opposite. The current employer

is the only (or probably the main) agent that can benefit from the invest-
ment, so there is no competition from other firms to push up the worker’s
wages. Once the investment is made the current employer is the only
buyer of the specific skills, so that the firm becomes ‘monopsonistic’ vis-
à-vis those skills. The problem here is that the worker will not have
the right set of incentives to invest, since the firm will get most of the
benefits from the investment. As we will see, the solution to the firm-
specific training is quite complicated. Let’s consider the following problem.

At time t = 1, there is a specific training opportunity s that can take two values s = {0, 1}
s = 1 corresponds to a situation in which specific training is undertaken and s = 0 when
training does not take place. The worker decides on s. Undertaking training costs c in terms of
forgone production. The marginal product is thus y without training and y − c when training
is undertaken.
The wage w1 in the first period is equal to y − cs, so that undertaking training implies
obtaining a wage equal to y − c.
At time t = 2, there are three stages:

• the firm makes a wage offer to the worker;

• the worker decides to accept this wage offer and work for this firm, or take another
job without incurring any costs of changing job;
• production takes place and wages are paid. We let the productivity of the worker be
y + fs.

Since s refers to specific skills, the outside option of the worker at time t = 2 is still
equal to y.
We also assume that f > s; this means that the increase in production is strictly positive
and the training opportunity is efficient.

To solve this problem we need to proceed by backward induction. We start

from the second period. The worker will accept any wage w2 ≥ y since y rep-
resents the worker’s outside option. The firm has no incentive to offer any wage
larger than the worker’s outside option. Knowing this, the firm simply offers
w2 = y. We then move to the first period. Given the wage w2 = y in t = 2,
in the first period the worker has no investment in specific skills. Undertaking
training implies taking a wage cut in the first period which is not going to be
followed by a wage gain in the second period. The specific training is not going
to be taken, even though the investment opportunity is efficient, since f > s.
What is the problem? By investing in firm-specific skills, the worker is
increasing the firm’s profits, so that the firm would like to encourage the worker
to invest. Given the timing of the game, wages are determined by a take it or
leave it offer by the firm after the investment has taken place. In this setting, it
will always be in the interest of the firm to offer a wage to the worker exactly
identical to the outside option. This is known as a hold-up problem in eco-
nomics. In this setting, it is the worker who is held up by the firm. The worker
anticipates this hold-up problem and does not invest in his firm-specific skills.
162 TRAINING AND HUMAN CAPITAL INVESTMENT

Can the firm have the incentive to pay all the training costs at time t = 2
and still offer a wage w1 = y? Such an option would not even work since the
firm will make a certain loss at time t = 1 and it would not have any guarantee
that the worker will stay with the firm in the second period.

9.5.1 SURPLUS AND SPECIFIC RENT

For obtaining a solution, we have first to realize what is so ‘specific’ about
this investment. To do so let’s define the concept of surplus, as the difference
between the value of the job inside the relationship versus the best alternative
option. The formal definition is as follows.

The worker surplus is the present discounted value of the difference between the wages and
the outside option. The firm surplus is the present discounted value of profits (since the firm
outside option is normalized to zero).

The present discounted value can be defined at time t = 1 and at time

t = 2. Let’s look at the surplus at time t = 2. In what follows, we indicate with
Sw,2 and with Sf ,2 the worker (firm) surplus at time t = 2, where the notation
Sw,2 (Sf ,2 ) indicates the worker (firm) surplus at time t = 2. The value of the
surplus reads

Sw,2 = w2 − y
Sf ,2 = y + fs − w2

The worker surplus is just the difference between the wage paid at time t = 2
and his or her outside option, which is equal to y. The firm surplus is equal to
the marginal product at time t = 2, which is now the sum of y + fs net of the
wage to be paid to the worker. Without specifying what is the wage that is paid
to the worker, it is impossible to argue whether the surplus of each party is
positive or not. Nevertheless, no matter how such a wage is determined, once
training has taken place, the sum of the two surpluses is a positive amount as
long as s = 1. Formally

S2 = Sw,2 + Sf ,2 = fs

which is strictly positive if s = 1.

Let’s now define the surplus at time t = 1 as

Sw,1 = (w1 − y) + (w2 − y)

Sf ,1 = (y + fs − w2 ) + (y − cs − w1 )
 TRAINING AND HUMAN CAPITAL INVESTMENT 163

Once again, it is not possible to establish the value of the surplus without
specifying the wages, but we know that the total surplus at time t = 1 is
S1 = Sw,2 + Sf ,2 = s(f − c)
which is strictly positive as long as s = 1. We thus have a positive total surplus
as long as there is specific training.

Firm-Specific Rent (or Surplus): A specific rent is a positive surplus that can be exploited only
by the current relationship.

From the expression S1 and S2 it is easy to establish that firm-specific

investment generates firm-specific surplus.
This is immediately established by noting that the expressions S1 and S2 are
positive as long as s = 1. We have now understood that firm-specific training
generates firm-specific surplus. The problem is to find an institutional setting
that ensures that the surplus is exploited by the firm–worker pair. The next
section introduces the concept of rent sharing.

9.5.2 RENT SHARING

A potential solution to the lack of investment in firm-specific skills is rent
sharing. Giving some bargaining power to the worker, so that part of the extra
profits generated by the specific investment is shared between the two parties,
is part of the solution.
Let’s modify the game above by assuming that in the final period, rather
than the firm making a take it leave it offer, the worker and the firm bargain
over the firm-specific surplus generated.
At time t = 2, there are three stages:
1. The firm and the worker share the surplus generated by the firm-specific
investment, and the worker gets a fraction β of the total surplus in excess
of his or her outside option. The bargaining power β is strictly less than 1.
2. The worker decides to accept this wage offer and work for this firm, or
take another job without incurring any cost of changing jobs.
3. Production takes place and wages are paid. We let the productivity of the
worker be y + fs.
This means that in the second period, if there is specific investment, the
worker’s wage is
w2 (s = 1) = y + βf
which is the sum of the worker’s outside option (y) plus a fraction β of the total
surplus. The wage in the first period is just the marginal product of the worker,
164 TRAINING AND HUMAN CAPITAL INVESTMENT

or w1 = y − cs where s can be zero or one. Given this wage schedule, at time

t = 1 the worker will choose s = 1 if
Sw,1 (s = 1) > Sw,1 (s = 0)
which is equivalent to
y − c + y + βf > y + y
The condition that ensures that the investment is implemented is
c
f >
β
Ex post rent sharing ensures that some specific opportunities are undertaken,
but does not guarantee that each and every opportunity is undertaken.
Indeed, it is easy to show that as long as β < 1 it is possible that f > c, and
yet the investment is not undertaken.
With ex post rent sharing firm-specific investment is less than the first best.
The proof of this is easily seen with the help of Fig. 9.3. All the investments
that lie above the 45 degree lines are efficient opportunities. Yet, we know that
only the investment opportunities that lie above the dotted line f = c/β will
actually be financed. In Fig. 9.3 investment opportunities such as B and A
should be financed while investment opportunities such as C should not be
financed. As long as β < 1 there is always some investment opportunity such
as B that is not going to be financed.
Is there a way in which the optimal level of specific human capital can
be achieved? A key possibility is a full sharing of cost and benefits, so that the
worker and the firm change the organizational system so that not only the bene-
fits are shared, but also the costs. Let’s see under what conditions the worker

f = c /b

Increase in B . f=c
productivity
A .

C .

Investment cost

Figure 9.3. Specific investment with ex post sharing

 TRAINING AND HUMAN CAPITAL INVESTMENT 165

wage
Productivity
y+f

Specific
surplus y + βf Wage profile
with training

y
Outside option
y – βc

y–c

t=1 t=2 time

Figure 9.4. Wage tenure profile with firm-specific investment

will invest in training with full sharing, so that the worker pays a fraction β of
the costs and gets a fraction β of the revenues. With full sharing, the worker
chooses to invest if
y + βf − y − βc > 0
which will lead to the worker choosing the investment as long as
f >c
which is the condition that ensures that the investment is efficient.
One can look at the problem from the firm’s standpoint. The firm at time
t = 1 would like to sponsor training if
(1 − β)(y + f ) − (1 − β)(y + c) > 0
from which it follows that the firm would choose an amount of training that
satisfies f > c. An important conclusion follows.
With full sharing of costs and benefits, all profitable specific training opportun-
ities are obtained.
Further note that the wage tenure profile becomes steeper once the specific
investment is undertaken.
Note that by splitting costs and benefits, the worker and the firm have
an incentive to stay together, so that firm-specific human capital is a good
condition for establishing long-run relationships. This suggests that each time
parties engage in firm-specific investments, the relationship is going to last
longer. Note that these concepts can be applied also outside personnel eco-
nomics, such as in the economics of family. For two partners, the best example
166 TRAINING AND HUMAN CAPITAL INVESTMENT

of firm-specific investments is childbirth and child bearing. A child is an invest-

ment that is specific to the couple (the firm). Indeed, it turns out that couples
with kids last longer (they have a lower divorce rate) exactly as our theory
would predict.

9.5.3 FIRM-SPECIFIC HUMAN CAPITAL: AN EXAMPLE

Let’s consider a worker who is currently paid according to his marginal product,
which we set equal to 100 in Table 9.1. The worker is expected to be hired for two
years and the interest rate is zero, so that there is no issue of discounting. There
is a training opportunity in specific skills that involves a loss of productivity at
time t = 1 of 15, but it generates an increase in productivity at time t = 2 of
25. The training opportunity is thus efficient. Undertaking training increases
the worker’s wage at time t = 2 to 125, but his outside option remains 100.
We first consider the specific training without any rent sharing. There is
a joint surplus equal to 5 at time t = 1, but the worker has no incentive to
exploit it since his surplus is negative. In such a case all the surplus would go to
the firm. Let’s first consider the situation with ex post sharing, described in the
third row of Table 9.1. With s = 1 the worker gets a wage cut at t = 1 equal to
15 against an increase in benefit at time t = 2 equal to 12.5. This implies that
the benefit at time t = 1 is negative if s = 1. As a result, the worker will choose
not to incur the specific investment. This is also consistent with the exposition
given above, since f = 25 is lower than c/β = 30.
Let’s now consider the situation with ex ante and ex post sharing. By sharing
the costs and benefits, the worker surplus at time t = 1 is positive and equal to
2.5. This means that the worker will undertake the investment. The equilibrium
is such that the worker takes a wage cut at time t = 1 equal to 7.5 and gets a
wage increase equal to 12.5 in the second period. Table 9.1 confirms this result.

9.6 Some Evidence on Training

There are two main themes in the empirical literature on training. The first
theme refers to the incidence of training across different dimensions, and the
second is the relation between training and wages. We briefly review these
two themes.
A common finding is that younger, better educated, and male workers
are more likely to participate in training. The finding that training decreases
with age is certainly consistent with the human capital investment theory: the
pay-off period shortens with age and so will training rates. The second finding
is that better-educated workers participate more in training than less-educated
workers. This suggests that the marginal revenues to training are higher for
 TRAINING AND HUMAN CAPITAL INVESTMENT 167

more skilled workers and/or that their marginal costs are lower. Most stud-
ies find also that women receive less training than their male counterparts.
The most common argument points to the lower labour force attachment of
women, which may cause employers to invest less. Gender differences might
also come about because women hold jobs that require less training, for which
there is some evidence. Training rates are also found to increase with firm size.
Perhaps the main explanation is economies of scale in training provision from
which larger firms benefit.
Human capital theory predicts upward-sloping productivity profiles. Wage
profiles are assumed to proxy these productivity profiles. The problem, how-
ever, is that there are many other theories, besides human capital, that predict
upward-sloping wage profiles and as such it is hard to argue that this is a defin-
itive test. One would like to know to what extent wage growth correlates with
productivity growth, but the literature has not found an obvious test for such
a prediction.
It is practically very difficult to distinguish between firm-specific and general
training, since in real life most training opportunities contain a combination
of the two. In this respect, the results reviewed in this section apply to both
types of investment. Yet, one can find a range of examples for which the
baseline theory of general training appears to provide a good description.
These include some of the historical apprenticeship programmes where young
individuals worked for very low wages and then graduated to become master
craftsmen; pilots who work for the Navy or the Air Force for low wages, and
then obtain much higher wages working for private sector airlines; securities
brokers, often highly qualified individuals with MBA degrees, working at a
pay level close to the minimum wage until they receive their professional
certification; or even academics taking an assistant professor job at Harvard
despite the higher salaries in other departments.

b APPENDIX 9.1. GENERAL TRAINING

The Basic Set-up

We now consider a generalization of the simple model proposed in the main chapter.
Consider the following very simple two-period model:
In the first period there is level of production equal to y. There are two stages:

1. The firm decides the level of training τ incurring the cost c(τ ), where c(τ ) reads
cτ 2
c(τ ) =
2
and c is a positive constant. We are assuming that training can be simply measured
through the variable τ .
168 TRAINING AND HUMAN CAPITAL INVESTMENT

2. Production takes place, for a value equal to y − c(τ ).

In the second period there are three stages:
1. The firm makes a take it or leave it wage offer w to the worker, and other firms
compete for the worker’s labour.
2. The worker decides whether to quit and work for another firm. We assume that
the worker does not incur any cost in the process of changing jobs, so that the
labour market is perfectly competitive.
3. Production takes place. At this point the value of output is equal to y + f (τ )
where f (τ ) reads
bτ 2
f (τ ) = aτ −
2
and a and b are positive constants.
For simplicity we assume that there is no discounting.

The Efficient Level of Training

Before proceeding, let’s first establish what is the efficient level of training. It is simply
the level of training that maximizes the difference between training benefits and
training costs.
Maxτ f (τ ) − c(τ )
In such a case, training should be undertaken as long as the marginal benefit is equal
to the marginal cost. In other words, the level of training τ ∗ should be chosen so that
c  (τ ∗ ) = f  (τ )
where c  (τ ) = cτ and f  (τ ) = a − bτ so that the optimal level of training solves
cτ = a − bτ
a
τ∗ =
b+c

The Solution of the Model

There are now two important issues to solve. First, we need to understand whether
the efficient level of training is going to be achieved, and second and probably most
important from the personnel economics standpoint, we need to understand who will
have the incentive for paying it.
The model has to be solved backwards. At t = 2, the worker will be trying to obtain
the largest possible salary in the marketplace. The solution of the second period at
time t = 2 is then very simple. The worker will accept from the current firm (or from
a different firm) only a training opportunity that guarantee a wage w2 = y + f (τ )
somewhere in the market, where τ is the level of training.
Let’s now move back to the first period. Since in the final period the worker will get
all the benefits from the investment, it is clear that the firm will not have the incentive to
 TRAINING AND HUMAN CAPITAL INVESTMENT 169

sponsor the training. The firm would certainly make a loss on the training, since it would
incur a certain cost at time t = 1 without incurring the benefits. Since the worker is
the full residual claimant of the increase in his or her own productivity, he or she will
have the right incentive to invest. The firm will then offer to the worker the following
package at t = 0, level of training τ ∗ = a/(b + c) and wage equal to w0 = y− c(τ ∗ ).
In this situation, the worker accepts the deal, since he will get the full wage

w1 = y + f (τ ∗ )

in the second period, either from the current or from another employer.Ë

b APPENDIX 9.2. FIRM-SPECIFIC TRAINING

Let’s consider this problem:

• At time t = 1, the worker decides how much to invest in firm-specific skills,

denoted by s at the cost c(s) = cs 2 /2.
• At time t = 2, the firm makes a wage offer to the worker.
• The worker decides to accept this wage offer and work for this firm, or take
another job.
• Production takes place and wages are paid. We let the productivity of the worker
be y + f (s), where f (s) = a − (bs 2 /2).
• Since s refers to specific skills, the outside option of the worker is still y.

To solve this problem we need to proceed by backwards induction. The worker will
accept any wage w1 ≥ y since y represents the worker’s outside option. Knowing this,
the firm simply offers w1 = y. Then, given this wage w1 = y, in the previous period
the worker makes no investment in specific skills, even though there is an efficient level
of investment that satisfies c  (s) = f  (s) so that

a − bs ∗ = cs ∗
a
s∗ =
b+c
The problem is the sharing of the specific investment and the fact that the worker is
held up by the firm. The partial solution to the underinvestment in firm-specific skills
is rent sharing. Let’s modify the game above by assuming that in the final period, rather
than the firm making a take it leave it offer, the worker and the firm bargain over the
firm-specific surplus so that the worker’s wage in the first period is

w1 (s) = y + βf (s)

Ë To understand why offering exactly τ ∗ is the equilibrium one would need to consider the alternative
policy of offering any τ  = τ ∗ . For being in equilibrium, we have to be in a situation in which
no unilateral deviation increases profits. Let’s see whether there are possible deviations when the
training investment is τ  = τ ∗ . In that case the worker’s income would be lower, since by definition
y − c(τ ∗ ) + y + f (τ ∗ ) > y − c(τ ) + y + f (τ ). In this case a firm offering to the worker yo − c(τ ∗ ) −  in
the first period would attract the worker and make positive profits. This implies that there is a profitable
deviation from the policy of offering τ  = τ ∗ , so that we cannot be in equilibrium.
170 TRAINING AND HUMAN CAPITAL INVESTMENT

where β is the worker’s bargaining power. Given this wage, at time t = 0 the worker
will choose s so as to maximize with respect to s
y + βf (s) − c(s)
which gives the investment
βf  (ŝ) = c  (s)
βa − βbŝ = c ŝ
which yields the investment level
βa
ŝ =
βb + c
which is a positive amount.
Rent sharing induces a positive amount of investment in firm-specific skills, and
partly solves the underinvestment problem.
Yet, it is easy to show that as long as β < 1, the amount of investment is less than
the efficient level.
With ex post rent sharing firm-specific investment is less than the first best.
Is there a way in which the optimal level of specific human capital can be achieved?
The only way is a full sharing of cost and benefits, so that the worker and the firm
change the organizational system in such a way that not only the benefits are shared,
but also the costs. Let’s see what is the amount of training that the worker will choose
with full sharing, so that the worker pays a fraction β of the costs and gets a fraction
β of the revenues. With full sharing, the worker chooses an amount s w such that he or
she maximizes with respect to s w
y1 + βf (s w ) − βγ (s w )
which will lead the worker to choose the investment such that
βf  (s w ) − βγ  (s w ) = 0
from which it immediately follows that s w = s ∗ . Let’s see now what level of training is
going to be chosen by the firm. The firm at time 0 would like to sponsor an amount of
training that maximizes with respect to s f
(1 − β)f (s f ) − (1 − β)γ (s w )
from which it follows that the firm would choose an amount of training that satisfies
(1 − β)f  (s w ) = (1 − β)γ  (s w )

from which it follows that s f = s ∗ . An important conclusion follows:

With full sharing of costs and benefits, the efficient amount of training is obtained.
This is true, since, as we saw above, with full sharing s f = s w = s ∗ = b+c
a
.
10 Training Investment in
Imperfect Labour
Markets

10.1 Introduction
The general conclusion of the baseline model of training (or the Becker model
as it is known in the literature) developed in the previous chapter is that there
will be no firm-sponsored investment in general training. This conclusion
follows from the assumption that the labour market is competitive, so the firm
will never be able to recoup its training expenditures in general skills later
during the employment relationship.
In real life labour markets, and particularly in imperfect labour markets,
there are many instances in which firms bear a significant fraction (sometimes
all) of the costs of general training investments.
The chapter explores the key dimensions of firm-sponsored general train-
ing. We will see that in imperfect labour markets, where specific rents are
pervasive, it may well be the case that firms undertake some investment
in general training. Such a result crucially depends on the existence of an
imperfect labour market, in the sense that wages paid to workers are higher
than the corresponding outside option. Yet, such a condition is not suf-
ficient to guarantee firm-sponsored training. Beyond market imperfection,
firm-sponsored training requires skill wage compression. The intuition of this
key result is that with wage compressions firms make greater profits from
better-trained workers, and have an incentive to increase the skills of their
workforce.
A first piece of evidence comes from the German apprenticeship system.
Apprenticeship training in Germany is largely general. Firms training appren-
tices have to follow a prescribed curriculum, and apprentices take a rigorous
outside exam in their area of expertise at the end of the apprenticeship. The
industry or crafts chambers certify whether firms fulfil the requirements to
train apprentices adequately, while works councils in the firms monitor the
training and resolve grievances. At least in certain technical and business occu-
pations, the training curricula limit the firms’ choices over the training content
fairly severely. Estimates of the net cost of apprenticeship programmes to
employers in Germany indicate that firms bear a significant financial burden
172 TRAINING INVESTMENT IN IMPERFECT MARKETS

associated with these training investments. The net costs of apprenticeship

training may be as high as 4,000 euros per worker.
Another interesting example comes from the recent growth sector of the
USA, the temporary help industry, and this chapter also presents a case study.
The temporary help firms provide workers to various employers on short-
term contracts, and receive a fraction of the workers’ wages as commission.
The majority of temporary workers are in clerical and secretarial jobs. These
occupations require some basic computer, typing, and other clerical skills,
which temporary help firms often provide before the worker is assigned to
an employer. Workers are under no contractual obligation to the temporary
help firm after this training programme. Most large temporary help firms offer
such training to all willing individuals. As training prepares the workers for a
range of assignments, it is almost completely general. Although workers taking
part in the training programmes do not get paid, all the monetary costs of
training are borne by the temporary help firms, giving us a clear example of
firm-sponsored general training.
There are also many examples of firms that send their employees to col-
lege, MBA, or literacy programs, and problem-solving courses, and pay for
the expenses, while the wages of workers who take up these benefits are
not reduced. In addition, many large companies, such as consulting firms,
offer training programmes to college graduates involving general skills. These
employers typically pay substantial salaries and bear the full monetary costs of
training, even during periods of full-time classroom training.
The chapter proceeds as follows. Section 10.2 presents the baseline model of
training in imperfect labour markets, in a simple set-up that assumes that the
worker has no possibility of financing general training. Section 10.3 extends
the baseline model, assumes that both the firm and the worker can finance the
training, and studies under what conditions firm-sponsored training emerges
in equilibrium. Section 10.4 presents the case study of the US temporary help
industry, where firm-sponsored general training seems pervasive.

10.2 Firm-Sponsored Training: The Baseline Case

10.2.1 THE SET-UP

Consider the following two-period model.
In period 1, the firm decides whether or not to invest in the general training
of the worker. The variable τ can take on only two values: τ = 0 or τ = 1,
where τ = 1 corresponds to a situation in which training is undertaken. The
structure of the problem is reported in Table 10.1.
There is no discounting, and all agents are risk neutral.
 TRAINING INVESTMENT IN IMPERFECT MARKETS 173

Worker stays or
quits and gets v (t)
Firm pays
Firm decides training costs Production takes
training level t –c (t) Wage offer place, wage is
w (t) paid y + f (t)

Period 1 Period 2

Figure 10.1. A two period model of firm-sponsored general training

If the firm undertakes training in period 1 it incurs a cost c. There is no

production in the first period.
In period 2 there are three stages:
1. At the beginning of the period the worker receives a wage offer w(τ ),
where the notation w(τ ) explicitly refers to the possibility that the wage
depends on the availability of training.
2. The worker decides to stay or quit. The labour market is imperfect and
the outside option is lower than the marginal product. In particular, the
outside option v is lower than the corresponding wage so that

v(τ ) < w(τ ) ∀τ

The previous condition ensures that the worker has some costs of chan-
ging jobs, even when there is no training. As we describe below, the wage
is obtained through a simple rent sharing.
3. If the worker stays with the firm production takes place and the wage is
paid. The value of the marginal productivity is y + f τ .
Assume that all training is technologically general in the sense that the
marginal product is y + f τ in all firms.
f is greater than c, which ensures that undertaking training is potentially an
efficient opportunity.
As we argued above, the labour market is imperfect, and the value of the
outside option is less than the wage. Assume now that wages are obtained by
rent splitting, and the worker gets his outside option plus a fraction β of the
total surplus. This is equivalent to assuming that the wage w(τ ) reads

w(τ ) = v(τ ) + βS(τ )

where v(t ) is the outside option of the worker and S(τ ) is the total firm-specific
surplus.
174 TRAINING INVESTMENT IN IMPERFECT MARKETS

To obtain a formal expression for the wage we need to specify the outside
option and the total surplus.
The outside option of the worker is specified as follows:
v(τ ) = a[y + f τ ] − b 0 < a < 1; b > 0
The parameters a and b play a key role in the analysis of training in imper-
fect labour markets. While both parameters refer to the existence of labour
market rents and specific surplus, they have very different implications. As we
show next, the parameter b generates a specific rent that is independent of pro-
ductivity, while the parameter a generates a specific rent that is proportional
to the marginal productivity on the job.
Let’s now consider the surplus.
If we indicate with Sf (τ ) and Sw (τ ) the firm and the worker surplus at time
t = 2, and with w(τ ) the wage, the expression of the total surplus S(τ ) is
Sf (τ ) = y + f τ − w(τ )
Sw (τ ) = w(τ ) − v(τ )
S(τ ) = y + f τ − v(τ )
S(τ ) = (1 − a)(y + f τ ) + b
The surplus is positive as long as a < 1 and/or b > 0. We will see below that
the two types of imperfections generate very different implications. What is
crucial to our analysis is that the parameter b generates a specific rent that is
independent of productivity y + f τ , while the parameter a generates a specific
rent that is proportional to the marginal productivity on the job.

10.2.2 THE SOLUTION

The key decision is the training decision.
At time t = 1, the firm decides to undertake training if and only if the firm
surplus (or the firm profits) net of the training costs at time t = 1 is greater
with training
Sf (τ = 1) − cτ > Sf (τ = 0)
which is equivalent to
y + f − w(τ = 1) − c > y − w(τ = 0)
or to
f > c + w (10.1)
where w = w(τ = 1) − w(τ = 0) is the increase in wage associated with
training. Equation (10.1) is a key condition for firm-sponsored training in
 TRAINING INVESTMENT IN IMPERFECT MARKETS 175

imperfect labour markets. It says that the firm will finance training if the
increase in the marginal product (hence in revenue) associated with training
is larger than the cost of training augmented by the increase in wage. This
condition implies that if the wage increase associated with training is equal to
the increase in the marginal product (hence w = f ) there is no way in which
firm-sponsored training can arise in equilibrium. This is exactly what happens
in the case of a perfect labour market. Conversely, if there is no increase in
wage (hence w = 0) then firm sponsoring of general training may arise. To
obtain a solution we need to calculate the increase in wage.
Using the definition of the surplus, the wage with and without training is
w(τ = 1) = a(y + f ) − b + β[(1 − a)(y + f ) + b]
w(τ = 0) = ay − b + β[(1 − a)y + b]
so that the increase in wage associated with training is
w = f [β + a(1 − β)]
The previous condition shows that the increase in wage is proportional to f ,
but it is less than f as long as a < 1. The parameter a shows that the key
condition for general training is wage compression. The smaller a is, the more
compressed are the wages, since the wage difference across different skill levels
is larger the smaller a is. Condition (10.1) is thus identical to
f > c + f [β + a(1 − β)]
so that the final key condition is
f [1 − a](1 − β) > c (10.2)
Several important conclusions follow:
If labour markets are perfect, firms never finance general training.
This is easily established. Markets are perfect if a = 1 and b = 0. It is clear
that if a = 1 firm-sponsored training never takes place. Note that the para-
meter b does not enter in the condition (10.2). Another important conclusion
immediately follows:
If labour markets are imperfect but wages are not compressed firms do not
finance general training.
This implies that if b > 0, a condition that ensures that labour markets are
imperfect, firm-sponsored training does not arise. In other words, the key
condition for firm-sponsored training is wage compression, as the next and
last conclusion shows:
If labour markets are imperfect and wages are compressed, financing of general
training is possible.
176 TRAINING INVESTMENT IN IMPERFECT MARKETS

If a is sufficiently low it is possible that a firm may sponsor general training, as

equation (10.2) shows.
A numerical example is the following. Suppose that f = 35, c = 15, and β =
0.5. The key condition for firm-sponsored training is, from equation (10.2),
7
(1 − a) > 1
6
which is satisfied only if a < 1/7. This suggests that as long as wages are
compressed but a > 1/7 firm-sponsored training does not arise.

10.3 A More General Framework

In the previous section we studied a model in which workers could not finance
the general training. The model in this sense was very specific, particularly
when we know from the theory of training in perfect labour markets that
workers in general have an incentive to take wage cuts and finance general
training. We thus now study a more complicated version of the model in
which both the workers and the firm can finance general training. In addition,
we assume that the variable τ can take on a continuous value.
Consider the following two-period model.
In period 1, the worker and/or the employer choose how much to invest
in the worker’s general human capital, τ . There is no production in the first
period.
In period 2, the worker either quits or stays with the firm and produces
output y = f (τ ), where f (τ ) is simply a linear function of τ so that
y = δτ
where δ is the marginal benefit of training. If the worker stays he is also paid
a wage rate, w(τ ), as a function of his skill level (training) τ , or he quits and
obtains an outside wage. If the worker quits he or she obtains an outside wage
v(τ ). The cost of acquiring τ units of skill is described as a function c(τ ),
which is strictly increasing and convex, and reads:
cτ 2
c(τ ) =
2
There is no discounting, and all agents are risk neutral. Assume that all training
is technologically general in the sense that f (τ ) is the same in all firms. If a
worker leaves his original firm he will earn v(τ ) in the outside labour market.
Suppose v(τ ) < f (τ ). That is, despite the fact that τ is general human capital,
when the worker separates from the firm, he will get a lower wage than his
 TRAINING INVESTMENT IN IMPERFECT MARKETS 177

marginal product in the current firm. As in the model above, the expression
for the outside option is written as

v(τ ) = af (τ ) − b a ≤ 1 and 0 ≤ b ≤ 1

The fact that v(τ ) < f (τ ) or that b > 0 implies that there is a surplus that
the firm and the worker can share when they are together. The total surplus is
S(τ ) = f (τ ) − v(τ ). Let us suppose that this surplus will be divided by Nash
bargaining, which gives the wage of the worker as:

w(τ ) = v(τ ) + β[f (τ ) − v(τ )]

= af (τ ) − b + β[f (τ ) − af (τ ) + b]
w(τ ) = [a + β(1 − a)]δτ − b(1 − β)

where β ∈ [0, 1] is the bargaining power of the worker. Note that the equi-
librium wage rate w(τ ) is independent of c(τ ): the level of training is chosen
first, and then the worker and the firm bargain over the wage rate. At this
point the training costs are already sunk, and do not feature in the bargaining
calculations. As is typically the case with sunk costs, bygones are bygones.
Assume that τ is determined by the investments of the firm and the worker,
who independently choose their contribution, cw and cf , and the total amount
of training τ is given by c(τ ) = cw + cf . Assume that $1 investment by the
worker costs $p where p ≥ 1. When p = 1, the worker has access to perfect
credit markets and when p → ∞, the worker is severely constrained and
cannot invest at all.
More explicitly, the timing of events is:

1. In the first period the worker and the firm simultaneously decide their
contributions to training expenses, cw and cf . The worker receives an
amount of training τ such that c(τ ) = cw + cf .
2. At the beginning of the second period the firm and the worker bargain
over the wage for the second period, w(τ ), where the threat point of the
worker is the outside wage, v(τ ), and the threat point of the firm is not
to produce.
3. Production takes place.

Given this set-up, the contributions to training expenses cw and cf will

be determined non-cooperatively. More specifically, the firm chooses cf to
maximize profits:

π(τ ) = f (τ ) − w(τ ) − cf = (1 − β)[f (τ ) − v(τ )] − cf

cτ 2
π(τ ) = (1 − β)[(1 − a)f (τ )] −
2
178 TRAINING INVESTMENT IN IMPERFECT MARKETS

subject to c(τ ) = cw + cf . The first-order condition is

(1 − β)(1 − a)f  (τ ) = cτf (10.3)

(1 − β)(1 − a)δ = cτf
the worker chooses cw to maximize utility:
u(τ ) = w(τ ) − pcw = βf (τ ) + (1 − β)v(τ ) − pcw
[a + β(1 − a)]f  (τ ) = pcτw
[a + β(1 − a)]δ = pcτw (10.4)
subject to the same constraint.
Let’s start considering the case of p → ∞, so that the worker is severely
credit constrained and cannot pay for the training.
b = 0 and a = 1 is a perfect labour market. In this case there is no firm-
sponsored training. The result of no firm-sponsored investment in general
training by the firm obtains when f (τ ) = v(τ ), which is the case of a perfectly
competitive labour market. Equation (10.3) implies that τf = 0, so when
workers receive their full marginal product in the outside labour market, the
firm will never pay for training.
a = 1 and b > 0. One may think that the key condition is to have a labour
market imperfection with an outside wage that is less than the productivity
of the worker, that is v(τ ) < f (τ ). Is it enough to ensure firm-sponsored
investments in training? The answer is no. To see this, first consider the case
with no wage compression, that is, the case in which a marginal increase in skills
is valued appropriately in the outside market. Mathematically this corresponds
to a = 1 and b > 0, or to a case of an imperfect labour market without wage
compression. In this case firm-sponsored training does not happen. Indeed b
does not enter the determination of the equation. For firm-sponsored training
to happen it is not sufficient that markets are imperfect.
a < 1. The key condition for firm-sponsored training is that a < 1. This is
the definition of wage compression. When v  (τ ) < f  (τ ) it is clear that the firm
may be willing to invest in the general training of the worker. The simplest way
to see this is again to take the case of severe credit constraints on the worker,
that is, p → ∞, so that the worker cannot invest in training. Then, a < 1 is
sufficient to induce the firm to invest in training. This shows the importance of
wage compression for firm-sponsored training. The intuition is simple: wage
compression in the outside market translates into wage compression inside the
firm, i.e. it implies w  (τ ) < f  (τ ). As a result, the firm makes greater profits
from a more skilled (trained) worker, and has an incentive to increase the skills
of the worker.
To clarify this point further, Fig. 10.2 draws the marginal benefit and the
marginal cost of training with and without wage compression. When a = 1
 TRAINING INVESTMENT IN IMPERFECT MARKETS 179

Wage compression
MB, MC
ct

0<a<1
(1 – b)(1– a) d No wage
compression

a=1
t*f t

Figure 10.2. Marginal benefit and marginal cost of training to the firm with and without
wage compression

there is no wage compression and the marginal benefit is along the horizontal
axis. In this case there is no incentive from the firm to sponsor training and
τ ∗ = 0 will be chosen. A competitive labour market obviously implies this
outcome. Conversely, when 0 < a < 1 the marginal benefits cross the marginal
costs at a positive amount and the firm is willing to sponsor general training.
The intuition is that when there is wage compression and a < 1 the firm makes
more profits from more skilled workers, and is willing to invest in the general
skills of its employees.
Consider now a situation in which p is finite so that there is a solution to
both τw and τf . Let τw be the level of training that satisfies (10.4), and τf be the
solution to (10.3). It is then clear that if τw > τf , the worker will bear all the
cost of training. And if τf > τw , then the firm will bear all the cost of training
(despite the fact that the worker may have access to perfect capital markets,
i.e. p = 1).
A decrease in a is equivalent to a decrease in the price of skill in the outside
market, and would also tilt the wage function inside the firm, w(τ ), decreasing
the relative wages of more skilled workers because of bargaining between the
firm and the worker, with the outside wage v(τ ) as the threat point of the
worker. Starting from a = 1 and p < ∞, a point at which the worker makes
all investments, a decrease in a leads to less investment in training, as is clear
from equation (10.4) and Fig. 10.3. This is simply an application of the Becker
reasoning; without any wage compression, the worker is the one receiving all
the benefits and bearing all the costs, and a decline in the returns to training will
reduce his investments. As the parameter a declines further, we will eventually
reach the point where τw = τf . Now the firm starts paying for training, and
a further decrease in a increases investment in general training (from (10.3)).
180 TRAINING INVESTMENT IN IMPERFECT MARKETS

pc t
No wage
compression

MB, MC
ct

a=1 Wage
[a + b(1– a)]d compression

a=0
bd
t *w t

Figure 10.3. Worker financing of general training with imperfect labor markets

Therefore, there is a U -shaped relation between the skill premium and training,
so starting from a compressed wage structure, a further decrease in the skill
premium may increase training.
Changes in labour market institutions, such as the minimum wage, will
also affect the amount of training in this economy. To see the impact of a
minimum wage, consider the case with a = 1 and b > 0. The wage is then
w(τ ) = −b(1 − β) + δτ . We have already established that in such a case there
should not be firm-sponsored training. Yet, for very low values of τ such a
wage can even be negative. Suppose that there is a minimum wage w̄ that is
binding for values of τ lower than τ̄ . In other words, the wage is

w̄ if τ ≤ τ̄
w(τ ) =
−b(1 − β) + δτ if τ > τ̄

In this situation firm-sponsored training may arise, as indicated in Fig. 10.4.

Imposing a minimum wage distorts the wage structure and encourages the firm
to invest in skills up to τ̄ , as long as c(τ̄ ) is not too high. This is because the firm
makes higher profits from workers with skills τ̄ than workers with skills τ = 0.
This is an interesting comparative static result, since the standard Becker model
with competitive labour markets implies that minimum wages should always
reduce training. The reason for this is straightforward. Workers take wage cuts
to finance their general skills training, and minimum wages will prevent these
wage cuts, thus reducing training. Therefore, an empirical investigation of the
relationship between minimum wage changes and worker training is a way of
finding out whether the Becker channel or the wage compression channel is
 TRAINING INVESTMENT IN IMPERFECT MARKETS 181

No wage compression
with minimum wage

MB, MC ct

t* t
w

Figure 10.4. Firm-sponsored training with a minimum wage

more important. Empirical evidence suggests that higher minimum wages are
typically associated with more training for low-skill workers.

10.4 Firm-Sponsored Training in Temporary

Help Firms
The temporary help service (THS) industry supplies its workers to client sites
on an as-needed basis, charging the client an hourly fee that typically exceeds
the wage paid to the THS worker by 35 to 65 per cent. David Autor has
used the case study of the THS industry as a well-documented example of
employer-financed general training. In this section we review such a case study.
THS employment grew rapidly throughout the 1990s, accounting for about
10 per cent of net US employment growth over the decade. As of 2001,
approximately 1 in 35 US workers was an employee of the THS industry.
Further, given turnover rates exceeding 350 per cent, the industry’s point in
time employment is likely to substantially understate the number of workers
who have contact with it annually. Up to the mid-1990s, the job skills required
by THS firms (primarily clerical) were essentially static without any signific-
ant amount of training. Things changed with the proliferation of workplace
computing technology that generated demand for new and rapidly shifting
expertise that could be mastered quickly. As is documented in Table 10.1, from
the mid-1990s training is a pervasive industry feature. Of 1,002 US THS estab-
lishments surveyed by the Bureau of Labor Statistics (BLS) in 1994, 78 per cent
offered some form of skills training, and 65 per cent provided computer skills
training. (Computerized tutorials are the most common form of instruction
182 TRAINING INVESTMENT IN IMPERFECT MARKETS

Table 10.1. Skills training: prevalence and policies at US temporary help agencies

Training provided Training policies

All skills training (multiple policies possible)
Any 78% ‘Up-front’: All/Volunteers
White-collar workers 56% trained 66%
Clerical/sales workers 81% Establishment selects
Blue-collar workers 59% trainees 34%
Computer skills training Client requests and pays 36%
Any 65% No training 22%
White-collar workers 27% Training methods used
Clerical/sales workers 74% (if training given)
Blue-collar workers 14% (multiple methods possible)
‘Soft’ skills training Computer-based tutorials 82%
Any 70% Classroom work, lectures 45%
White-collar workers 52% Written self-study materials 52%
Clerical/sales workers 70% Audiovisual presentations 47%
Blue-collar workers 58% Other 14%
Detailed training subject frequencies by major occupation group
Any White collar Clerical/sales Blue collar
Word processing 63% 23% 75% 13%
Data entry 58% 19% 69% 11%
Computer programming
languages 22% 12% 23% 1%
Customer service 41% 27% 47% 12%
Workplace rules/on-job
conduct 66% 55% 68% 60%
Interview and résumé
development skills 30% 31% 32% 13%
Communications skills 14% 15% 14% 10%

(82 per cent), while 52 per cent of establishments provide workbook exercises
and 45 per cent provide classroom-based training.)
Almost without exception, training is given prior to or between assignments
during unpaid hours with all fixed and marginal costs paid by the THS firm.
It is reported that 44 per cent of all skills training is given ‘up front’ to allow
workers to qualify for their first assignments. Trainees are not contractually
bound to take or retain a job assignment afterwards, nor would such a contract
be enforceable. While THS firms are prone to overstate the efficacy and depth
of their training, evidence of its value is found in the fact that several leading
firms sell the same training software and courses to corporate customers that
they provide for free to their workers.
These basic facts run counter to the baseline human capital model of train-
ing outlined in the previous chapter. In the competitive case analysed by Becker,
workers pay for general skills training by accepting a wage below their mar-
ginal product during training. The threat of poaching or hold-up ensures that
workers earn their full post-training marginal product, and hence up-front
general skills training should never be provided by the employer. By contrast,
the case study under scrutiny reveals that THS firms routinely provide train-
ing up front during unpaid hours, and hence the opportunity for workers to
 TRAINING INVESTMENT IN IMPERFECT MARKETS 183

offset costs through a contemporaneously lower training wage is essentially

non-existent.
David Autor (2001) used a unique survey for analysing the relationships
among wages, training, and competition at temporary help establishments.É
Conducted in 1994, the survey enumerates employment, wages, training offer-
ings, and training policies at 1,033 temporary help establishments in 104
Metropolitan Statistical Areas (MSAs). The sample comprises an estimated
19 per cent of all THS establishments employing twenty or more temporary
workers in 1994 and 34 per cent of all THS employment. One of the questions
in the survey was whether establishments offer skills training to each ‘collar’
in eight subject categories: word processing, data entry, computer program-
ming languages, workplace rules and on-the-job conduct, customer service
skills, interview and résumé development skills, communications skills, and
other. The core of the regression analysis that follows relies on computer skills
because they are well defined, hold market value, and clearly constitute general
skills training.

10.4.1 THE EVIDENCE: WAGES AND TRAINING IN THE

THS INDUSTRY
In the baseline model of training, wages should be higher in establishments that
provide general training. Since training is general, the worker is the residual
claimant of the training provided, and he or she should obtain a wage gain
in the aftermath of training. Table 10.2 reports a bivariate comparison of
mean log wages at training and non-training establishments in the nine major
occupational groups in the sample (three in white collar, two in clerical/sales,
four in blue collar). The comparison is striking. In eight of nine occupations,
mean wages are lower at training establishments, with an average occupational
wage difference of minus 6.4 log points. Since workers receive training during
non-work hours, the wage differentials estimated above do not reflect ‘training
wages’ in the conventional sense of Becker. Rather, they indicate that workers at
establishments providing up-front training receive lower wages while assigned to
client sites, either before or after they have received training (or both).
To make a more formal comparison, David Autor estimated the following
equation:
Wij = a + δTij + γ1 Ej + γ2 Oi + γ3 Rj + eij
where Wij is the natural logarithm of hourly wages of individual i at estab-
lishment j. Tij is a vector of establishment training variables, Ej is a vector of
establishment characteristics, Oi is a vector of major occupation indicators,
É Occupational Compensation Survey of Temporary Help Supply Services (OCS hereafter).
184 TRAINING INVESTMENT IN IMPERFECT MARKETS

Table 10.2. Comparison of log hourly wages of worker at training and non-training
establishments

Log hourly wages Training Non-training

No. workers No. workers
Free No Difference No. estabs. No. estabs.
training training

White collar
All 2.66 2.79 −0.13 10,497 13,034
(0.04) (0.05) (0.06) 360 270
Professional specialty 3.05 3.17 −0.13 2,918 5,016
(0.02) (0.03) (0.04) 200 170
Technical 2.41 2.45 −0.05 5,805 6,554
(0.04) (0.05) (0.06) 274 213
Accountants and 2.72 2.77 −0.06 1,774 1,464
auditors (0.04) (0.06) (0.07) 187 134
Clerical/sales
All 2.01 2.09 −0.09 156,419 17,925
(0.01) (0.03) (0.03) 693 166
Clerical and 2.02 2.10 −0.08 145,997 16,957
administrative (0.01) (0.02) (0.03) 690 164
support
Marketing and sales 1.84 1.97 −0.13 10,422 1,328
(0.03) (0.08) (0.09) 435 42
Blue collar
All 1.76 1.78 −0.02 85,756 50,257
(0.01) (0.01) (0.02) 461 294
Precision production, 1.89 1.97 −0.08 8,193 6,142
craft, and repair (0.04) (0.04) (0.06) 216 162
Operators, assemblers, 1.79 1.82 −0.03 19,867 12,851
and inspectors (0.02) (0.02) (0.03) 310 187
Transport, material 1.89 1.92 −0.03 1,884 1,809
movement (0.06) (0.05) (0.08) 186 126
Handlers, equipment 1.72 1.71 0.01 55,812 29,445
cleaners, and labourers (0.01) (0.01) (0.02) 445 252

Rj is a vector of 103 MSA indicators, and eij is a random error term assumed
to be composed of a person-specific and establishment-specific component.
The parameter of interest is δ, the wage differential at training estab-
lishments. Due to the inclusion of narrow MSA and occupation indicators,
δ effectively measures wage differentials among local THS establishments
potentially competing for the same workers and supplying labour to the same
customers.
The first three columns of Table 10.3 presents wage models for the full
sample. The initial specification estimates the training wage differential with
an indicator variable that is equal to one if the establishment provides computer
skills training. The coefficient on this variable indicates that wages at training
establishments are on average 2.0 log points lower, which is significant at the
5 per cent level. To probe alternative explanations for this wage differential,
the second column introduces two additional controls. The first is the log of
 TRAINING INVESTMENT IN IMPERFECT MARKETS 185

Table 10.3. Estimates of the relationship between establishment training policies and
worker wages

Pooled estimates Fixed effects estimates

(1) (2) (3) (4) (5) (6)

Any training −0.020 −0.019 −0.035 −0.034

provided (0.010) (0.010) (0.0179) (0.0176)
Up-front training −0.025 −0.049
provided (0.010) (0.019)
Firm selects 0.005 −0.026
trainees (0.013) (0.040)
Client requests/pays 0.003 0.061
for training (0.012) (0.039)
Log of −0.026 −0.025 −0.020 −0.022
establishment size (0.004) (0.004) (0.007) (0.007)
Log of THS 0.051 0.050 0.023 0.024
employment in (0.012) (0.011) (0.013) (0.013)
MSA-collar
Firm fixed effects No No No Yes Yes Yes
R2 0.62 0.62 0.62 0.54 0.54 0.54
n 333,888 333,888 333,888 201,314 201,314 201,314

establishment size. Because large establishments typically provide more con-

sistent THS assignments, workers at these establishments may accept lower
hourly wages. And since large establishments are substantially more likely to
offer training, it is plausible that the observed training–wage relationship in
part reflects a size–wage differential. The second control introduced is the log
of THS employment in the major occupation (‘collar’) in the MSA. The results
do not change.
The empirical evidence provided suggests that training firms pay lower
wages to workers at client firms. And these lower wages do not reflect training
wages, since training is given up front. The result is thus in full contrast with the
standard human capital model for general training. The next section describes
a possible rationalization of this finding.

10.4.2 A POSSIBLE STORY FOR THE THS EVIDENCE

Autor has also provided a model in which not only firms sponsor general
training, but also workers in THS training firms receive lower wages than
workers in firms that do not provide training. The key idea of the model
developed by Autor is that firms offer general skills training to induce self-
selection and perform subsequent screening of worker ability.
The model has three periods. There are a large number of THS firms, some
of which offer skills training, and some of which do not. These are referred
to as training and non-training firms. All firms and workers are risk neutral,
186 TRAINING INVESTMENT IN IMPERFECT MARKETS

and there is no discounting between periods. In the first period, workers may
select to work at either a training or non-training firm. Training firms provide
general skills training to the workers whom they hire during the first period.
Non-training firms do not. At the end of the first period, a fraction of the
workers at each THS firm quits for exogenous reasons to enter the second-hand
market. In addition, workers may quit their first period THS firms voluntarily
to enter the second-hand market. Workers in the second-hand market are hired
by other THS firms. At the beginning of the third period, all workers are hired
by clients into the permanent sector.
The model can generate several equilibria. The equilibrium of empirical rel-
evance, and consistent with the facts above, is a separating equilibrium in which
workers with high-ability beliefs self-select to receive training while those with
low beliefs do not. The separating equilibrium works as follows. In period 1
wages are identically zero for trainees and non-trainees and expected period 3
wages are higher for trainees. Although all workers would forgo some earnings
to receive training, workers with high beliefs will forgo more because their
expected period 3 gains are larger. At a separating equilibrium, the expected
period 3 wage gain for high-belief workers offsets their training wage penalty
in period 2, while for low-ability belief workers it does not. Consequently, at a
separating equilibrium, wages at training firms are lower than at non-training
firms.
The separating equilibrium depends critically upon two features of the
model. The first is the complementarity between training and ability. Because
training and ability are complements, high-belief workers apply to training
firms, and low-belief workers apply to non-training firms. Training therefore
serves as a self-selection device. If training and ability were not complements,
either all workers or no workers would choose training. A separating equi-
librium would be infeasible. The second critical feature of the model is that
training elicits private information about worker ability. If training firms did
not acquire private information about worker ability, competitive markets
would ensure that each trainee received his marginal product after train-
ing. Hence, the dual roles played by training in the model, self-selection
and information acquisition, are complementary. By inducing self-selection
of high-ability workers, training improves the firm’s worker pool. By revealing
private information about worker ability, training then allows the firm to profit
from this pool.
While the model is of course stylized, these private-information-based
results appear consistent with the personnel policies of THS firms. After
initial training and testing, THS workers are normally first placed at lower-
wage, lower-skill assignments and subsequently given better placements as
they demonstrate success. Workers who test and train successfully and per-
form well at assignments advance more rapidly while workers who perform
poorly are rarely offered placements. Consequently, poor workers turn over
disproportionately while good workers frequently remain.
11 Job Destruction
11.1 Introduction
Sooner or later, every firm experiences bad times. And in bad times,
management often needs to reduce the size of the workforce. The choice of
downsizing is obviously a key concern for personnel managers and for person-
nel economics. The firm needs not only to choose which workers should be
targeted for lay-off, but also to realize that reducing the workforce will have an
impact on the remaining workers.
Job destruction is a subtle business, even in labour markets in which employ-
ment protection legislation is not particularly stringent, and employer-initiated
legislation is largely admissible. The first key concept that the chapter high-
lights is the difference between consensual and non-consensual separation,
where the former refers to those separations that, at a given wage, are agreed
by both parties, while the latter refers to those separation decisions that are
not shared by both parties. Over the job destruction decision, each party (i.e.
the firm and the worker) looks at his or her surplus from the job, and wants to
preserve a job as long as the difference between the value of the job net of his
or her outside option is positive. When the surplus of each party is negative
there is obviously no point in continuing the job, and job destruction should
take place without particular difficulties.
Whenever a non-consensual separation emerges, the firm has to consider
whether it is possible to preserve jobs through wage cuts. Wage cuts, or an
agreed reduction in future wages, are a job-preserving possibility that should
always be contemplated. The chapter shows that as long as the joint surplus
from the job is positive, there is always a profitable wage cut that can preserve
the job. The issue is whether such wage cuts will be implemented in reality.
There are, indeed, a number of reasons that explain why job-preserving wage
cuts are barely observed in practice. A first issue to consider is whether reducing
wages is allowed by the existing legislation. But there are other more subtle
issues to consider, such as the effects of wage cuts on the remaining workers. In
practice, firms often opt for job destruction rather than job cuts, even though
there are important exceptions.
In simple and static views of labour demand, job destruction should take
place each and every time the wage falls short of the value of the marginal
productivity. In dynamic settings, and especially when employment protection
legislation is stringent and firm-initiated separation is costly, things are more
188 JOB DESTRUCTION

difficult. Labour hoarding is defined as a situation in which a firm hires a

given quantity of labour even though the current marginal profit associated
with such labour is negative. In an imperfect labour market, labour hoarding
is a common phenomenon. An obvious reason for the occurrence of labour
hoarding is the existence of employment protection legislation: when firing
is costly, firms hold on to marginal losses just to delay the payment of firing
costs. This is the key message of the second part of the chapter.
The final part of the chapter shows that labour hoarding emerges also when
there is no employment protection legislation if two conditions hold: (i) there
is surplus from a job; and (ii) there is uncertainty over the value of labour
product. In a simple two-period model of job creation and destruction this
type of labour hoarding naturally emerges.
The chapter proceeds as follows. Section 11.2 formally defines firm-initiated
separation and distinguishes between consensual and non-consensual sep-
aration. It also discusses the conditions under which a profitable wage cut
may exist. Section 11.3 defines labour hoarding and considers the optimal
job destruction decision when firm-initiated separation involves a fixed cost.
Section 11.4 presents the two-period model of job creation and destruction
under uncertainty, and shows that labour hoarding emerges, in such a situation,
even though firing is costless.

11.2 Firm-Initiated Separations and Wage Cuts in

Imperfect Labour Markets
Let’s focus on a single job that has been opened in the past, and let us use our
standard notation for considering the value of the marginal productivity to
the firm as Jt and the wage paid to the worker as Wt . The outside option to the
worker is indicated with Ut .É The worker’s surplus from the job is
Sw = Wt − Ut
A worker is clearly interested in continuing to work with a firm as long as
Sw ≥ 0, or as long as the value of the wage is larger than the outside option.
Let’s define the firm’s surplus from the job as
Sf = Jt − Wt
so that a firm wants to continue to hire a worker as long as Sf ≥ 0, or as long
as the marginal product is larger than the wage. For the purpose of this section
we assume that the wage has already been set by collective agreements.
É In this section we use capital letter to indicate Jt , Wt , and Ut since, as we show in Appendix 11.1,
such values may refer to present discounted values.
 JOB DESTRUCTION 189

We have not formally specified what is the time horizon of the job. This
was not an omission, since the concepts we are highlighting hold for both
static jobs (i.e. 1 period) and multi-period jobs. In the former case, the wage
Wt is simply the period wage. In the latter case, the variable Wt refers to the
present discounted value of wages. Similar distinction can be made for Jt and
for Ut and Appendix 11.1 reports the present discount interpretation of our
discussion.
In any event, the total surplus from a single job can be written as

S = Sf + Sw
= (Jt − Wt ) + (Wt − Ut ) (11.1)
= Jt − Ut

We are now in a position to define consensual separation.

Consensual Separation: A separation is consensual as long as both Sf < 0

and Sw < 0. In other words, when the surplus is negative S < 0, or when
Ut > Jt .

This works as follows. A worker wants to stop working for a firm as soon as
Sw < 0 while a firm wants to fire the worker as soon as Sf < 0. If Wt < Ut and
simultaneously Jt < Wt , then necessarily Ut > Jt . In this case there is no point
in continuing with this job, and it does not matter whether it is the firm that
fires the worker or whether it is the worker himself that quits. In either case
the job is going to be terminated.
Things are more complicated when the firm surplus is negative for the firm
while positive for the worker. In this case the firm has an interest in terminat-
ing the job and proceeding to a firm-initiated separation. Nevertheless, such
separation may not be consensual.

Firm-initiated separation is not consensual if Sw > 0 and Sf < 0.

When the worker surplus is positive (i.e. Sw > 0), there are no individual
incentives to quit the job, and the worker will hold on to the job as long as the
firm-initiated separation is not completed. If the firm has full authority over
the continuation of the job, and there are no separation costs to be paid, the
firm will initiate a separation procedure. The key question we address in the rest
of this section is the following. Is it possible for the worker to accept a sufficiently
large wage cut so that the firm-initiated separation is avoided? The remark shows
that such a worker–firm transfer exists only when the total surplus is positive. A
job-preserving wage cut exists when firm-initiated separation is not consensual
but the surplus is positive.
190 JOB DESTRUCTION

To see the last conclusion, let’s consider a situation in which separation is

not consensual, so that
Wt − U t > 0
Jt − Wt < 0
The worker is willing to accept a wage cut  (where  is defined in absolute
terms) as long as the cut  leaves him or her with a positive surplus. In formula,
 must be such that
(Wt − ) − Ut > 0 (Worker)
 < Wt − Ut (11.2)
The previous condition simply says that the wage cut should be lower than the
initial worker surplus. Simultaneously, a firm is potentially willing to continue
operation after the wage cut as long as
Jt − (Wt − ) > 0 (Firm)
 > W t − Jt (11.3)
which simply says that the wage cut should be large enough to compensate the
firm for the initial loss. Both conditions (Firm and Worker) hold only when
the right-hand side of the (worker) condition is larger than the right-hand side
of the (firm) condition so that
Wt − Ut > Wt − Jt
or when
Ut < Jt
The previous condition is the final proof, since it shows that as long as the job
has a positive value, a profitable wage cut can be potentially implemented.
Another issue to consider is whether such wage cuts are possible in practice,
since collective agreements specify wages and often do not allow such wage
cuts to be implemented. This suggests that we may end up with inefficient
separation. A firm-initiated separation is inefficient whenever a profitable job-
preserving wage cut is not implemented. The inefficiency is linked to the fact that
a job that has strictly positive surplus ends up being destroyed. Conversely, a
separation that takes place when the total surplus is negative is always efficient.
So far we have only discussed firm-initiated separation, but it is clear that
one can also discuss the case of worker-initiated separation. A worker-initiated
separation is a situation in which the worker surplus is negative. In this case, we
say that the worker would like to quit his current job. The interesting condition
arises when the worker has a negative surplus while the firm enjoys a positive
surplus.
 JOB DESTRUCTION 191

Consensual separation
Sw < 0 and Sf < 0

Firm-initiated
separation Efficient
Sf < 0 S = Sf + Sw < 0

Non consensual
separation
Sw > 0 and Sf < 0

Inefficient
S = Sf + Sw > 0

Figure 11.1. Firm-initiated separation

Worker-initiated separation is not consensual if Sw < 0 and Sf > 0.

In this scenario, rather than wage cuts, what we need to consider is whether
there exists a profitable wage increase that can preserve jobs. For symmetry,
it is clear that a job-preserving wage increase exists when the total surplus is
positive.
A job-preserving wage increase is possible when the worker-initiated
separation is not consensual but the surplus is positive.
The proof is analogous to what was described in the case of the wage cut.

11.2.1 JOB DESTRUCTION: AN EXAMPLE

Consider a firm that is hiring ten workers and needs to restructure and down-
size its labour force. In other words, the firm needs to decide which workers
should be kept and which workers should be offered a wage cut. The various
workers act separately and the firm is able to distinguish the job of each and
every worker. Table 11.1 reports the present discounted values of wages, outside
option, and the production values for each of the ten workers. The firm is losing
money on three workers (i.e. numbers 3, 4, and 9) and these three workers are
the obvious candidates for a separation. On worker 3 the firm loses 8 while
the worker enjoys a surplus of 5. Obviously, a wage cut up to 5 is profitable for
this worker. But such a wage cut is not enough to compensate the firm. The
total surplus is indeed negative and a separation is efficient. A similar situation
exists for worker number 9, since the firm is losing 20 while the worker has
192 JOB DESTRUCTION

Table 11.1. Job destruction and wage cuts

Worker Wages Outside Production Firm Worker Separation Total Action

surplus surplus surplus

1 70 60 70 0 10 10
2 75 65 80 5 10 15
3 40 35 32 −8 5 Firm initiated −3 Separation
4 85 74 80 −5 11 Firm initiated 6 Wage cut
5 90 85 100 10 5 15
6 25 30 25 0 −5 Quit −5 Separation
7 65 70 71 6 −5 Quit 1 Wage raise
8 55 45 65 10 10 20
9 90 75 70 −20 15 Firm initiated −5 Separation
10 60 55 70 10 5 15

a surplus of 15. On worker 4, the situation is more complicated, since the firm
is losing 5 but the worker enjoys a surplus of 11. Here a job-preserving wage
cut is possible, since the total surplus is positive. Indeed, a wage cut equal to 6
would make the job viable.
The situation of workers 6 and 7 is also interesting. On worker 6, the firm
is indifferent while the worker has a negative surplus. This separation is not
firm initiated but worker initiated, and we typically called it a quit. There is
no surplus and the job is going to be destroyed. On worker 7, conversely, a
profitable wage rise is possible.

11.2.2 JOB-PRESERVING WAGE CUTS WITH ASYMMETRIC

INFORMATION
The discussion and formal analysis of the previous section suggests that in
many instances, and notably when there is positive surplus, wage cuts could
save jobs in an efficient manner. Yet, such wage cuts are rarely observed in
reality. In this section we try to discuss why.
There are various concerns. First, there is the issue of collective agreements.
Such wage cuts may not be possible by law, and in any case agreement with
unions or works councils is necessary. This section focuses on another element
that prevents the emergence of wage cuts, notably the presence of asymmetric
information. The key point of this section is that when asymmetric information
is pervasive, job destruction is often a preferred option to a wage cut. There
are two key reasons.

1. State of Demand. In a large organization, the overall state of demand is

known better to management than to the workers. There is asymmetric
information over the state of business, and the firm is better informed
than workers. In this setting, workers will not find credible the demand for
a cut in wages because the state of demand is low. The problem is that the
 JOB DESTRUCTION 193

firm could just act strategically in order to increase profits. By pretending

to be in difficult business conditions, the firm may be able to increase
profits. Conversely, reducing employment is a much more credible way
to signal the state of demand, since firms will also incur output losses.
The result is that job destruction is the only credible option.
2. Internal labour markets. In most organizations, the internal labour mar-
ket of firms implies an implicit contract in which workers expect a rising
wage profile, so that workers expect that conditional on good perform-
ance, future wage gains will be attained. Such a contract will be briefly
described in Chapter 12. An overall wage cut would work as a breach
of this implicit contract from the point of view of remaining workers.
Conversely, employment reductions are consistent with such a contract.
This discussion suggests that downsizing ends up being a better option for
the remaining workers. And since the firm acts as a forward-looking agent, it
becomes the option to be pursued.
When the separation is worker initiated, in the sense that the worker is
threatening to quit, the firm has to consider whether to implement a wage
raise. In the case of perfect information, the situation is simple and it is the one
described above. When information is asymmetric, often firms opt for a policy
of not responding to wage increases obtained by employees outside the firm.
The key reason to avoid matching a wage offer is the risk of inducing the rest
of the workers to search more actively for a job. And if most workers behave in
this way, it becomes very difficult for the firm to verify the credibility of each
and every threat to quit.

11.3 Labour Hoarding and the Effect of Firing Taxes

In the previous discussion we have assumed that job termination is possible
without the firm incurring any costs. Yet, in most imperfect labour markets job
termination is a costly process and it involves firing costs. This section explicitly
considers the situation of a firing tax to be paid by the firm in the case of a
firm-initiated separation.Ê We basically assume that firm-initiated legislation
is admissible, but it involves a fixed cost equal to F . The presence of the firing
tax changes the outside option of the firm, which now involves the cost F . If
the firm keeps the job open it will get a value that is equal to the difference
between the value product and the wage, while if the firm terminates the job it
will have to pay the firing tax, so that the surplus of the firm is now

Sf = J − W − (−F )

Ê As we discuss in the next chapter the firing costs involve a monetary transfer to the worker as well
as a tax to be paid outside the firm–worker pair.
194 JOB DESTRUCTION

where S  refers to the fact that we are now explicitly considering the firing cost.
The surplus of the worker is unchanged and it is equal to Sw = W − U .
To understand the effect of firing costs let us assume that there are many
jobs in the firm and that different jobs can have different values of J . In other
words, we now assume that J can take any value from a minimum of zero up
to a maximum of J max . Let’s try to understand which jobs will be kept open by
the firm.
Let’s first analyse the case in which there are no firing costs, so that F = 0,
and let’s ask what is the minimum value of J for the firm to keep a job open.
The answer is simple, since the firm will want to continue to operate any job
that yields at least the wage rate. Let’s call this value the reservation productivity
J ∗ or a value of J that makes the firm indifferent between keeping the job open
and closing it down. The reservation productivity when there are no firing
costs is the value of J that makes the surplus equal to zero
Sf ( J ∗ ) = 0
J∗ = W
When there are no firing costs the firm’s firing decision is as follows: keep open
any job that yields a marginal value that is at least as high as the wage rate. This
is also the simple condition of labour demand in a static setting.
Let’s now consider what happens when the firm has to pay a firing cost for
implementing a job destruction. The firm will want to keep a job open as long
as the continuation value is larger than the firing cost. Let’s indicate with J ∗∗
the value of the marginal productivity to the firm when there are firing costs
so that

Sf (J ∗∗ ) = 0
J ∗∗ − W + F = 0
J ∗∗ = W − F
where it immediately follows that J ∗∗ < J ∗ . The implication of this finding is
obvious. When there are firing costs to be paid, the firm holds on to jobs that
yield a value that is lower than the wage rate. This is called the labour hoarding
effect of firing costs. In order to avoid the firing costs, the firm is willing to accept
a loss on the marginal job and keeps open a job that yields less than the wage
rate. This effect is clearly visible from Fig. 11.2.
Firing costs reduce the number of jobs that are destroyed.
This conclusion can be easily verified, since the larger the firing cost, the
lower the reservation productivity. Firing costs induce firms to hold on to
jobs that have lower marginal value. This is not surprising, since we have
imposed a cost on the job destruction decision. Yet, in a general view of the
 JOB DESTRUCTION 195

S=J–W+F
S = J–W

J **
F

J
–(W–F )
J*

–W

Figure 11.2. Reservation job and firing costs

firm’s personnel policies, firing costs may certainly reduce job destruction, but
they are also likely to have an impact on job creation. The rest of the chapter
describes a two-period model of job creation and job destruction.

11.4 Job Creation, Job Destruction, and Labour

Hoarding
We now consider the following two-period model of job creation and destruc-
tion. Assume that a firm and a worker face a potential opportunity to form a
job. The salary to be paid in each period is equal to w and for simplicity we
assume that it is constant in both periods. The job involves some uncertainty
over the value of the labour product. We assume that in each period the value
of the labour product (i.e. the value of the marginal product of labour) is
akin to a random draw from a continuous distribution G(ε) defined over the
support ε ∈ [εmin, εmax ]. The values of the labour product in the two periods
are independent events, so that there is no link between the realization of ε
in the first period and in the second period. In each period there are different
stages, as is clearly indicated in Fig. 11.3.
1. In period 1 the firm first observes the realization of the productivity value
ε, and then decides whether or not to create the job (this is called the
job creation decision). If the firm creates the job, production in the first
period takes place and the job moves to the second period.
2. At the beginning of the second period the firm observes the new real-
ization of the productivity ε and it then decides whether or not the job
196 JOB DESTRUCTION

Job creation Job destruction

decision decision

Production Production
Firm observes ε takes place Firm observes ε takes place

Period 1 Period 2

Figure 11.3. The timing of the events

should be destroyed (the job destruction decision). Job destruction at this

point involves a firing cost equal to −F . If the firm destroys a job it has
to pay F . If the job goes on production takes place and the job ends.
3. The firm has a discount rate equal to r. This means that 1 euro at time 2
yields a value at time 1 equal to (1/1 + r).

There are two key decisions that the firm should make. At period 1 the firm
should decide which realizations of ε should lead to a job formation. In period
2, the firm should decide which jobs should be destroyed. The problem is
solved backwards. We first solve the job destruction decision at the beginning
of the second problem, and then, given this decision, we move to the first
problem and solve for the job creation decision.
We begin with the job destruction condition. At the beginning of the second
period the firm observes a value of ε, so that if the firm operates it enjoys a
marginal profit equal to ε − w. If the firm destroys the job it has to pay the
fixed cost equal to −F . It is clear that the job destruction decision is given by
the solution to the following problem

Destroy if ε − w < −F
Continue if ε − w > −F

This is the same problem as the one that we discuss in the previous para-
graph, and it is clear that the firm decision will be described by a reservation
productivity ε d such that

εd = w − F

The value of εd is a reservation productivity and it is called the job destruction

margin in the rest of the section. All jobs whose value of ε falls below ε d
should be destroyed. Note clearly that the job destruction decision is completely
 JOB DESTRUCTION 197

εc
JD

Jobs
created JC

εc
Jobs not
created A

εd = w – F εd

Jobs destroyed Jobs continued

Figure 11.4. Determination of job creation and job destruction

independent from what happens in the first period. Bygones are bygones, and
the firm decision is completely forward looking.
Two results are immediate:
1. Job destruction increases if the wage goes up. To see this one can notice
that an increase in the wage increases the reservation productivity εd .
2. Job destruction is reduced if firing costs go up. To see this one can
notice that an increase in the firing cost −F increases the reservation
productivity ε d .

The job creation decision is more difficult to write down, but it delivers
very important insights on hiring decisions in the context of firing costs. Let
us indicate with 1 (ε) the present discounted value, evaluated at time 1 of
profits at the beginning of the first period, once the firm has observed the
realization of the productivity. Its expression reads
1  
1 (ε) = (ε − w) + Expected Profits in Period 2
1+r
 
1
1 (ε) = (ε − w) + Max[z − w, −F ]dG(z)
1+r
The previous expression should be read in the following manner. The left-hand
side is the presented discounted profits for a job with realization in the first
period equal to ε. The term in brackets ε − w is simply the operational profits
of the first period. The term 1/1+r is the discount factor, while the term inside
198 JOB DESTRUCTION

the integral is the expected profits from the second period, taking into con-
sideration the job destruction decision. This expression, whose formal value is
given by the term inside the brackets, should be read as follows. In the second
period the productivity will be equal to z, where z can take any value from the
support of the distribution G. But we know that, conditional on observing z,
the firm will operate as long as the expected profits are larger than the firing
costs. This is reflected by the max operator. The integral symbol simply sums
up all possible values of z, with weights given by the distribution G(z).
At time 1 there are no firing costs to be paid, since the job has not yet been
established. If the job is not going to be created, the firm does not face any cost,
so that the outside option at the moment of creation is simply zero. The job
creation decision is then described by the following simple condition:
Create if 1 (ε) > 0
Not Create if 1 (ε) < 0
This suggests that the decision to create will be regulated by a reservation
productivity ε c such that
1 (ε c ) = 0
εc is the creation margin, and needs to be solved with a few steps of algebra.

Solving for the Job Creation Condition

Using the definition of the job creation margin, the job creation margin is the
solution to
 
1
εc − w + Max[z − w, −F ]dG(z) = 0
1+r
The first thing we can do is to get rid of the max operator inside the integral.
The job destruction is obtained by the reservation productivity ε d , and we
know that for values of z below ε d the job is destroyed while for values above
ε d the job continues its operation. We can then remove the max operator, by
writing down two integrals for different values of z. The expression reads:
 εmax
c 1 1
ε −w =− (z − w)dG(z) + FG(εd )
1 + r εd 1+r
The previous expression gives the job creation as a function of the firing cost
and the job destruction decision. It has important economic implications. The
left-hand side is the operational profits for the marginal job, and it is equal
to two terms in the right-hand side. The first term in the right-hand side is
a negative term (the integral is a positive function, since z is always above w
for values of z above εd ) while the second term is positive. When F = 0 the
 JOB DESTRUCTION 199

right-hand side is obviously negative. This suggests that there is a tendency of

the firm to take a loss in the first period, and to open up those jobs that yield
a negative loss in the first period, but are offset by positive expected profits in
the second period. This is a form of labour hoarding. There is labour hoarding
each time the firm optimally holds on to current losses from hiring a worker.
When F > 0, the right-hand side tends to increase. This suggests that the
larger the firing costs, the larger the operational profit of the marginal job. This
makes a lot of sense, since the firm anticipates future firing costs and opens up
jobs that yield larger profits. This will be confirmed in the comparative static
below.
The full solution of the job creation margin can be obtained through a
diagram in which on the two axes are reported the job creation and the job
destruction margins [εc , εd ]. The job destruction curve is a vertical line that
crosses the axis at the point in which εd = w − F . The job creation curve
is the locus that describes the combination of ε c , εd that yields zero expected
profits. The slope of the job creation is obtained by a simple differentiation
which shows that

dεc g (ε d ) d
= [ε − w + F ]
dε d 1+r

In other words, the job creation margin is akin to a quadratic function that
features a minimum in correspondence to a point of the job destruction curve.
To understand this property recall that along the job creation curve (ε c ) = 0.
Consider a point on the job creation curve to the left of the job destruction
curve. In such a point, an increase in the job destruction margin (which is
lower than w − F ) increases profits in the second period, and such an increase
must be compensated by a decrease in ε c for the expected profits to be zero.
The opposite happens to the right of the maximum.

The Effect of Change in Firing Costs

The effect of change in firing costs is immediately obtained, and it is described
by a shift to the left of the job destruction curve. The job creation shifts upward
as can be immediately seen by differentiating the job creation curve with respect
to F at given εd . The new situation is described in Fig. 11.5, and the following
important result is immediately derived.

An increase in firing costs reduces job destruction in the second period but it
also reduces job creation in the first period.

This conclusion is the fundamental effect of firing costs on job creation and
destruction.
200 JOB DESTRUCTION

εc
JD

εc

JC

εd = w–F εd

Figure 11.5. The effects of an increase in firing costs on job creation and job destruction

b APPENDIX 11.1 PRESENT DISCOUNTED VALUES AND

MULTI-PERIOD JOBS

Employment relationships often involve a time horizon which is larger than

a single period. To discuss issues linked to multi-period labour demand, we
need to introduce some key concepts:
ut : Value at time t of the alternative use of time. This is the value of the
outside option at time t .
wt : wage profile; it describes the value of the wage at each period t .
yt : productivity profile, the value of the productivity at time t .
Let’s consider a worker that potentially can be hired for T periods, and
thereafter he or she retires. We call t the tenure of the worker or the years of
experience with the firms. 0 is the age of youth (newly hired) workers while T
is retirement age.
Let’s define three key present discounted values. Note that we use capital
letters for present discounted values and lower-case letters for flow payments
at period t . At time t < T , Wt , or the presented discounted values of wages,
reads:
wt +1 wt +2 wT +t
Wt = wt + + + ... +
(1 + r) (1 + r)2 (1 + r)T

T
wk+t
W1 = ;
(1 + r)k
k=1
 JOB DESTRUCTION 201

Ut is the present discounted value of the worker’s alternative at time t so that


T
uk+t
Ut =
(1 + r)k
k=1

Finally, Jt is the present discounted value of the productivity stream of the

worker to the firm.

T
yk+t
Jt =
(1 + r)k
k=1

When the firm problem is dynamic and multi-period, the surplus of the firm
is the difference between the present discounted value of productivity and the
presented discounted value of wages. If we label t the present discounted
value of profits (or Sft ) we have
yt +2 − wt +2 y T − wT
t = yt − wt + ... +
1+r (1 + r)T
which can be written as

T
yt +k − wt +k
t =
(1 + r)k
k=1

Further we know that the surplus is positive if


T
yt +k  wt +k
T
t = − ≥0
(1 + r)k (1 + r)k
k=1 k=1
t = Jt − Wt ≥ 0.
12 Further Issues in
Employment Protection
Legislation

12.1 Introduction
The previous chapters in the book studied the role of employment protection in
the context of temporary contract and in the case of job destruction. We learnt
two key messages. First, employment protection biases the firm choice toward
temporary contract. Second, employment protection reduces job destruction
and increases labour hoarding. Yet, when job creation is also explicitly con-
sidered, employment protection also depresses job creation. While these are
indeed the key messages in the literature of employment protection legislation,
other important issues need to be considered.
In real life labour markets, employment protection legislation is a multi-
dimensional institution. Its provisions are often very complex, and the
legislation not only forces the firm to pay specific sums of money to the dis-
missed worker, but it also forces the firm to follow specific procedures, which
inevitably involve deadweight losses to both parties. In other words, employ-
ment protection legislation involves both transfer inside the firm–worker pair
as well as taxes paid outside the match. Section 12.2 documents that the largest
part of the legislation refers to transfer. But the difference between taxes and
transfer is economically very important when wage determination is explicitly
considered. In previous chapters, the role of employment protection legisla-
tion was always analysed at a given wage. In reality, wages are likely to respond
to the existence of EPL. Indeed, the model of section 12.3 shows that when
wages are endogenously determined by rent sharing, the transfer component
of EPL is likely to be allocative neutral. This means that employment protec-
tion legislation may have a large impact on wage profiles, but little impact on
job destruction. Such a conclusion is not true in the case of the tax component
of EPL, whose effects cannot be fully absorbed the wages.
Another dimension of EPL in real life labour markets is the different degree
of stringency across firms of different sizes. Large firms have typically a degree
of EPL that is tighter than the degree of EPL imposed on small firms. This
differences is also economically important, since it may affect the scale of
 EMPLOYMENT PROTECTION LEGISLATION 203

operation of each firm. In other words, firms may have an incentive to operate
below a given EPL threshold just as a way to save on EPL provisions beyond
the threshold. These issues are analysed in section 12.4, in a simple model of
EPL with threshold effects.

12.2 EPL Taxes and Transfers

Employment Protection Legislation (EPL) is one of the most important insti-
tutions of the labour market. It refers to the set of norms and procedures
to be followed in cases of dismissals of redundant workers. In almost all
countries, EPL forces employers to transfer to the worker a monetary com-
pensation in case of early termination of a permanent employment contract.
Moreover, complex procedures have to be followed in case of both individual
and collective lay-offs.
EPL is obviously a multidimensional institution. The most traditional
dimensions of EPL are the severance payments and advance notice. Severance
payments refer to a monetary transfer from the firm to the worker to be paid
in case of firm-initiated separation. Advance notice refers to a specific period
of time to be given to the worker before a firing can actually be implemen-
ted. Note that the severance payment and advance notice that are part of EPL
refer to the minimum statutory payments and mandatory rules that apply to
all employment relationships, regardless of what is established by labour con-
tracts. Beyond mandatory payments, collective agreements may well specify
larger severance payments for firm-initiated separations. Such party clauses,
albeit important, are not considered in this report.
Another important dimension of EPL consists of the administrative pro-
cedures that have to be followed before the lay-off can actually take place. In
most countries, the employer is often required to discuss the lay-off decisions
with the workers’ representatives. Further, the legislative provisions often differ
depending on business characteristics such as firm (or plant) size and industry
of activity. As this simple introduction shows, it is obvious that the EPL is a
multidimensional phenomenon.
From the viewpoint of economic analysis, the multidimensionality of the
EPL can be reduced to two components. The first component is simply a
monetary transfer from the employer to the worker, similar in nature to the
wage. The second, instead, is more similar to a tax, because it corresponds
to a payment to a third party, external to the worker–employer relationship.
Conceptually, the severance payment and the notice period correspond to the
transfer, while the trial costs (the fees for the lawyers, etc.) and all the other
procedural costs correspond to the tax.
204 EMPLOYMENT PROTECTION LEGISLATION

12.2.1 CALCULATING TAXES AND TRANSFERS

Decomposing the total firing cost between tax and transfer components is an
exercise that requires detailed knowledge of the country-specific institutions.É
Garibaldi and Violante provide estimates of the transfer and tax components in
the statutory firing cost for Italy, one of the countries with the strictest employ-
ment protection legislation (OECD 1999). The estimates show that transfers
significantly exceed taxes. In this section we briefly review such calculations.
In the Italian legislation, an employer-initiated lay-off against an individual
employee is legitimate only when it satisfies a ‘just cause’. The Italian civil law
foresees that individual dismissals are legal only under two headings: justified
objective motive, i.e. ‘justified reasons concerning the production activity, the
organization of labour in the firm and its regular functioning’, and justified
subjective motives, i.e. ‘a significantly inadequate fulfilment of the employee’s
tasks specified by the court’. The first case involves events which are outside the
employee’s control, while the second case requires misconduct on the part of
the worker. The worker has always the right to appeal against the firm’s decision,
and the final judgment ultimately depends on the court’s interpretation of
the case. If the worker does not appeal against the firing decision, or if the
separation is ruled fair, the legislation does not impose any firing cost on the
firm.Ê Conversely, when the separation is ruled unfair and illegitimate, the
court imposes a specific set of transfers and ‘taxes’ on the firm, which we
analyse next.Ë
The total costs associated with this procedure can be calculated ex post, i.e.
when the firm has already learnt that the job is no longer viable, as well as
ex ante, i.e. when the job is still viable and production is going on.
Ex post firing cost. Specifically, we start by considering a situation where
an employer-initiated individual separation against a blue-collar worker with
average tenure (eight years) in a firm with more than fifteen employees is ruled
unfair by the judge after a twelve-month trial, the average length of a labour
trial in Italy. This firing cost is therefore ex post with respect to the court’s
decision. Although, this is not the exact counterpart of the cost in the firm’s
hiring and firing decision, it is a useful starting point.
First of all, the worker will be granted the forgone wages from the separation
day up to the court ruling (i.e. twelve months under our assumptions), while

É This requirement goes well beyond the information published by the OECD (1999). Possibly, for
this reason we are not aware of any other study trying to make this comparison.
Ê The union to which the worker is affiliated usually bears all the legal costs in this case.
Ë Concerning the definition of a legitimate separation, the Italian EPL does not make any difference
in terms of firm size. Yet, the maximum compensation to which unlawfully fired workers are entitled
varies with firm size in two important dimensions. For small firms (with less than fifteen employees),
the choice between a full reinstatement and a severance payment rests with the firm. Further, for
workers employed in firms with less than fifteen employees the maximum severance payment that can
be obtained in court is limited to six months’ wages.
 EMPLOYMENT PROTECTION LEGISLATION 205

the firm will pay the forgone social insurance contributions augmented by a
penalty for delayed payment. In addition, the worker may choose between a
severance payment of fifteen months or the right of being reinstated by the
firm that unlawfully fired him.Ì In over 95 per cent of the cases, the worker opts
for the former option. Finally, all the legal costs will be paid by the firm. Thus,
if we let n be the number of months that it takes to reach a court decision,
w the gross monthly wage, τ s the social security contributions, τ h the health
insurance contribution, φ the penalty rate on forgone contributions, sp the
mandatory severance payments for unfair dismissal, and lc the total legal cost,
the total ex post firing cost FC is

FC = nw + (τ s + τ h + φ)nw + sp + lc.
The pure transfer component paid by the firm to the worker is
S = nw + ατ s nw + sp,
where α is the share of the social security contributions that is rebated to
the worker in the form of increased future pensions, in which case such a
payroll contribution should be counted as transfer inside the match. The tax
component is
T = (1 − α)τ s nw + (τ h + φ)nw + lc.
Table 12.1 provides an estimate of the size of FC as well as of the components T
and S in the total firing costs when α = 0, the share that minimizes the transfer

Table 12.1. Tax and transfer components of firing costs in Italy

Components of firing costs

Total Tax Transfer

Forgone wages nw 12 0 12
Health insurance τ h w 1 1 0
Social security contributions τ s w 4 4 0
Sanctions for delayed payments φ w 3 3 0
Legal costs lc 6 6 0
Severance payments sp 15 0 15
Ex post firing costs FC 41 14 27
Share 100 34 66

Cost in out of court agreement pu (S + T/2) 17 0 17

Total ex ante firing costs F̃C̃ 18.75 3.5 15.25

Share 100 19 81

Ì See Ichino (1996) for the legal sources of this binding rule. Note that the number reported by the
OECD (1999: table 2.A.2, p. 95) on the statutory severance payment in Italy is erroneous, since it refers
instead to the mandatory deferred wage scheme (TFR), a very different institution.
206 EMPLOYMENT PROTECTION LEGISLATION

component, i.e. the case that makes the tax component as large as possible. The
estimate suggests that the total ex post cost is over forty months’ wages, and
the transfer component of the total firing costs amounts to 66 per cent.
Ex ante firing cost. The above computation results in an impressively high
firing cost, but we should keep in mind that it is based on the worst possible
scenario for the firm: once the case has been taken to court and the judge has
reached a verdict favourable to the worker. Obviously, ex ante the firm–worker
pair does not know with certainty whether the separation will be ruled unfair
by the tribunal: let pu denote the probability of such event. Many employer-
initiated separations are not settled in court. Firms and workers often find
a satisfactory settlement out of court and strike a deal before the full trial is
over. In the case of an out of court agreement, the parties can save any court
penalties that may eventually be imposed by a judge, and all the legal costs
linked to the trial. In particular, if the two parties bargain in a symmetrical
Nash fashion on the settlement, the joint maximization problem will solve
 1  1
max S − pu S 2 −S + pu (S + T ) 2 ,

S

where we denote by Ŝ the point of agreement between firm and worker. Notice
that we have assumed—as common practice in Italy—that the labour union
will pay the legal costs in cases where the lay-off is ruled fair. The solution gives
Ŝ = pu (S + (T /2)) which is an amount larger than the expected transfer the
worker would receive, but smaller than the total cost (transfer plus tax) the
firm would pay in case the firing is ruled unfair. The intuition is that half of
the tax becomes part of the settlement. For the purpose of our analysis, it is
important to remark that in this case the entire firing cost for the firm is a
transfer to the dismissed worker.
Let pa be the probability of agreement out of court. If we ignore discounting,
the ex ante (with respect to the court’s verdict) expected firing cost FC  is
 
T   

FC = pa pu S + + 1 − pa pu FC + (1 − pu )CL , (12.1)
2

where CL is the firing cost incurred by the firm when the judge rules the firing
legitimate. Since, as we explained above, in the Italian legislation CL = 0, the
expected transfer component is
 
T 

S = pa pu S + + 1 − pa pu S (12.2)
2

while the expected tax component is



T = 1 − pa pu T . (12.3)
 EMPLOYMENT PROTECTION LEGISLATION 207

Galdon-Sanchez and Guell (2000), using data based on actual court sen-
tences, compute
 that in Italy the probability of reaching an out of court
agreement pa is roughly
 0.50, and the probability of the individual lay-off
being ruled unfair pu is also approximately 0.50. With these probabilities,
 falls to eighteen months’ wages. However,
using the estimates of Table 12.1, FC
for the sake of our analysis, what matters is the fact that the share of the transfer
rises to over 80 per cent of the total.

12.3 Outsiders, Insiders, and Tax and Transfers

Let us now consider a two-period model. A firm and a worker face a job
opportunity that lasts two periods. The value product in the first period is
equal to εh . The value of the labour product in the second period is random.
It can be equal to εh with probability p and it is equal to εl with probability
1 − p. We assume that εl is a large negative number so that the firm always
fires the worker when the realization in the second period is εl . In the second
period the worker is entitled to employment protection legislation. The second
period is called the ‘insider phase’, since the worker is entitled to employment
protection legislation (EPL). EPL takes two different forms. On the one hand,
there is a severance payment to be paid to the worker that we indicate with the
letter T . On the other hand there is firing tax that we call F . Even though both
payments are technically made by the firm, there is a key difference between
the two institutions. The severance payment is paid to the worker while the
firing tax is paid outside the match. In the first period the worker is not entitled
to employment protection legislation. We thus call such a phase the outsider
phase. Note that there is no explicit firing in the first period. Yet, if the parties
do not agree on a wage negotiation the worker is not entitled to any form of
EPL.
We will assume for most of the section that wages are negotiated through
rent sharing, and that the worker gets a fraction β of the total surplus. The
surplus that each party enjoys must be properly defined in each period, since
it changes between the outsider and insider phase.
• In period 1 the job is already formed and the productivity is εh . The
firm and the worker must agree on an outsider wage which we label w0 .
Production takes place and the worker, at the end of period 1, becomes an
insider worker.
• At the beginning of period 2 the realization of the labour product is known
to both parties. If the value product is εl the job ends and the firm must
pays both the transfer T and the tax F . The insider workers gets the transfer
T while the payment F is dissipated. If the value product is equal to εh ,
208 EMPLOYMENT PROTECTION LEGISLATION

the party must agree on a salary that we label wi . At that point production
takes place.
• We assume that there is no discount rate. This means that 1 euro at time
2 yields a value at time 1 equal to 1 euro.

12.3.1 DERIVING THE WAGES WITH SEVERANCE PAYMENTS

The problem is solved backwards. We first solve the wage in the second period,
when the worker is an insider. Next we solve for the wage in the first period
when the worker is an outsider, taking into account the wage of the insider.
Let’s first focus on severance payments only, so that in this section we assume
that F = 0. We begin in the second period and assume that the job continued,
so that ε = εh . In the second period, the value of the profit to the firm if a
wage agreement is found is εh − wi while if no agreement is found the firm
needs to pay the severance payment T . The firm surplus at time t = 2 is

Sf ,2 = εh − wi − (−T )

Let’s now consider the worker position at time t = 2. The worker surplus from
a wage agreement is

Sw,2 = wi − (T + u)

where u is the value of the outside option for 1 period and T is the severance
payment. The total surplus is simply

S2 = εh − u

from which follows an obvious important conclusion.

The severance payment does not enter in the total surplus.
To see this point simply notice that in the determination of the expres-
sion the severance payment does not enter. This is not surprising, since the
severance payment is just a pure transfer. Nevertheless, the previous remark
will have important consequences for the effect of severance payment on job
destruction. Having derived the surplus, the wage of the insider is obtained in
such a way that the worker gets a fraction β of the total surplus. This means
that the worker surplus is

wi : wi − (T + u) = βS2

from which it follows, after a simple substitution, that

wi = (1 − β)u + βεh + T
 EMPLOYMENT PROTECTION LEGISLATION 209

The wage of the insider is a simple weighted average between the worker outside
option and the productivity of the match. In addition, the severance payment
increases the wage of the insiders. This means that insiders’ wages go up with
respect to a situation without severance payment.
Let’s now move to the first period. The cumulative expected profits from
employment at time 1 for the firm are equal to
1 = εh − wo + p(εh − wi ) − (1 − p)T
where εh − wo are the operational profits in the first period. Note that the
second period expected profits are random. In the second period two things
can happen: with probability p the job continues and the firm gets εh − wi (wi
is the insider wage), while with probability (1 − p) the job is destroyed and the
firm pays the severance payment to the worker.
Let’s now turn to the workers’ value of employment at time 1, which can be
written as
W1 = wo + pwi + (1 − p)(u + T ).
The total value of the outside option, which we indicate with U , is simply equal
to the total sum of the per period outside option u, so that U = 2u.
As we mentioned above, at time t = 1 the worker is an outsider and he or
she is not entitled to severance payment. This means that the total surplus to
the worker is
Sw,1 = W1 − U1 .
As for the firm, if agreement is not reached, there is no severance payment
involved so that Sf ,1 = 1 . This means that the total surplus at time t = 1 is
S1 = W1 − U1 + 1
which, after a simple substitution of the functions, leads to
S1 = εh − u + p(εh − u)
The previous expression simply says that the total surplus from the match
is the difference between the value of the labour product and the worker’s
outside option in period 1, plus the same difference in period 2 multiplied
by the probability that the job continues (which is equal to p). Note that the
severance payment does not enter in the determination of the surplus also in
the first period. To obtain the insider wage consider that
W − U = βSi
which is an expression that says that the worker gets a fraction β of the total
surplus. The total outsider wage is then
wo + pwi + (1 − p)T − u − u = βεh − βu + βp(εh − u)
wo = u(1 − β) + βεh − T
210 EMPLOYMENT PROTECTION LEGISLATION

The wage of the outsider worker is reduced by the full amount of the severance
payment. While it is true that the insider workers get an increase in their wage
equal to T , the analysis has also shown that such a wage increase is fully prepaid
by the outsider worker. This has also a further important consequence.
When wages are flexible, the severance payment has no impact on firm profits.
In other words, the severance payment with wage flexibility is neutral.
This can be seen by showing that the severance payment has no impact on
the firm’s expected profits. One simply substitutes the outsider wage and the
insider wage into the firm’s profits to obtain
1 = εh − wo + p(εh − wi ) − (1 − p)T
= εh − u(1 − β) − βεh + T + pεh − p(1 − β)u − pβεh − pT − (1 − p)T
= (εh − u)(1 − β) + p(1 − β)(εh − u)
From the previous conclusion it follows immediately that the firm’s hiring
decision is not affected by the presence of severance payments. If the cumu-
lative profits are unchanged, the firm clearly does not change its hiring policy,
even though the hiring decision is not explicitly modelled in this section.

12.3.2 DERIVING THE WAGES WITH THE FIRING TAX

In this section we carry out a similar exercise without severance payments with
a firing tax. In other words we can now assume that F > 0 while T = 0. In the
second period, the value of the profit to the firm if a wage agreement is found
is εh − wi while if no agreement is found the firm needs to pay the firing tax
F . This implies that the firm surplus at time t = 2 is
Sf ,2 = εh − wi − (−F )
Let’s now consider the worker position at time t = 2. The worker surplus at
time t = 2, following a wage agreement, is
Sw,2 = wi − u
where u is the value of one period of unemployment. The total surplus is
simply
S2 = εh − u + F
from which follows an obvious important conclusion. The firing tax enters in
the total surplus. This is not surprising, since the firing tax enters into the threat
point of the firm and does not enter into that of the worker. Nevertheless, it will
have important consequences for the effect it will have in the labour market.
 EMPLOYMENT PROTECTION LEGISLATION 211

Having derived the surplus, the wage of the insider is obtained so that the
worker gets a fraction β of the total surplus. This means that
wi : wi − u = βS2
from which it follows, after a simple substitution, that
wi = (1 − β)u + βεh + βF
The wage of the insider is then a simple weighted average between the worker’s
outside option and the productivity of the match. In addition, the firing tax
increases the wage of the insiders by a factor that is proportional to the workers’
bargaining power. This means that insiders’ wages go up with respect to a
situation without firing tax.
Let’s now move to the first period. The expected profits from employment
at time 1 for the firm are equal to
1 = εh − wo + p(εh − wi ) − (1 − p)F
where εh − wo are the operational profits in the first period. The second period
expected profits are random. In the second period two things can happen:
with probability p the job continues and the firm gets εh − wi (wi is the insider
wage), while with probability (1 − p) the job is destroyed and the firm pays
the firing tax. Let’s now turn on the workers’ value of employment at time 1,
which can be written as
W1 = wo + pwi + (1 − p)u.
The total value of the outside option, which we indicate with U , is simply equal
to the total sum of the per period outside option u, so that U = 2u.
As we mentioned above, at time t = 1 the worker is an outsider and he or
she is not entitled to the firing tax. This means that the total surplus to the
worker is
Sw,1 = W1 − U1 .
As for the firm, if agreement is not reached, there is no severance payment
involved so that Sf ,1 = 1 . This means that the total surplus at time t = 1 is
simply
S1 = W1 − U1 + 1
which, after a simple substitution of the functions, leads to
S1 = εh − u + p(εh − u) − (1 − p)F
The previous expression simply says that the total surplus from the match is
the difference between the value of the labour product and the outside option
212 EMPLOYMENT PROTECTION LEGISLATION

in period 1, plus the same difference in period 2 multiplied by the probability

that the job continues (which is equal to p). In addition, the presence of
the firing tax reduced the surplus at time t = 1. To obtain the insider wage
consider that

W − U = βSi

so that

wo + pwi + (1 − p)u − u − u = βεh − βu + βp(εh − u) − β(1 − p)F

wo = u(1 − β) + βεh − βF

The wage of the outsider worker is reduced by a fraction β of the firing tax.
While the insider worker gets an increase in his wage equal to βF , the analysis
has also shown that such a wage increase is fully prepaid by the outsider worker.
This has also a further important consequence.

Even when wages are flexible, the firing tax has impact on firm profits, and
it is not neutral.

This can be seen by showing that the firing tax reduces expected profits.
One simply substitutes the outsider wage and the insider wage into the firm’s
profits to obtain

1 = εh − wo + p(εh − wi ) − (1 − p)F
= (εh − u)(1 − β) + p(1 − β)(εh − u) − F (1 − β)(1 − p)

The previous expression clearly shows that the firing tax F negatively affects
the present discounted profits.

12.4 Threshold Effects

EPL is traditionally modelled as a firing tax on labour shedding, and the ori-
ginal theoretical framework is the dynamic labour demand under uncertainty.
Bentolila and Bertola (1990) characterize the optimal employment strategy
of a monopolistic firm subject to idiosyncratic shocks and firing costs, hold-
ing wages fixed. Most of this literature takes EPL as given, and looks at the
employment effect of different degrees of job security provisions. A very simple
exposition of the Bertola–Bentolila model is that of Schivardi (2000). The
present section studies employment dynamics when EPL is binding only for
firms larger than a given size. We introduce threshold effects in a toy model of
 EMPLOYMENT PROTECTION LEGISLATION 213

labour demand. This section proceeds as follows. First, we solve for the efficient
allocation, next we show the properties of the model with an extreme form
of EPL. Finally, we introduce threshold effects, and derive the main empirical
predictions on firm level dynamics.

12.4.1 THE SET-UP OF THE MODEL

We assume that there is a continuum of firms of mass 1, and that wages are exo-
genously fixed and equal to w. Each firm hires only labour and produces and
sells a homogeneous output with a convex production function y = f (α, ε, l),
where α is a stochastic shifter of labour demand, l is the quantity of labour
employed, and ε is a fixed firm-specific parameter heterogeneous across firms.
The shifter parameter α is an index of business conditions at each firm. It can
take two different values, α = ab in bad business conditions and α = ε (with
ε > ab ) in good business conditions. Firms are subject to an i.i.d. (identical
and independently distributed) idiosyncratic shock and in each period there is
a probability p that business conditions are bad and a probability (1 − p) that
business conditions are good. The parameter ε differs across firms, and is dis-
tributed according to the distribution function F (x) = Prob(ε < = x), where F
is continuous with no point mass and defined over the support  ∈ [ab , εmax ].
This implies that firms are identical when business conditions are idiosyncrat-
ically bad, but differ in their profit schedule when business conditions are good.
Since firms differ only for their idiosyncratic parameter ε, in what follows we
index firms simply by ε. Firms are dislocated in islands, there is no entry or
exit, and profits exist in good and bad times as long as ab > w. In this respect,
the analysis is left at the partial equilibrium level. The model is stationary and
we do not need to explicitly keep track of the time index t , even when we
introduce EPL. If the production function is quadratic in labour, firm’s profit
for a type-ε firm can be written as
1
(α, ε, l) = αl − l 2 − wl
2
where

a with probability p
α= b
ε with probability 1 − p

12.4.2 THE EFFICIENT ALLOCATION

Assume now that each type-ε firm can choose the optimal employment level
after observing the realization of the shock α, and assume that hiring and firing
214 EMPLOYMENT PROTECTION LEGISLATION

can take place at no cost. Firm optimal employment behaviour is obtained

simply by maximizing profits in each period, so that the firm continuously sets
the marginal product equal to the wage, or

l = ab − w if α = ab
l ∗ (ε) = b
lg (ε) = ε − w if α = ε
This implies that a type-ε firm, in steady state, spends a fraction p of its time
in bad business conditions with l ∗ = lb and a fraction (1 − p) in good business
conditions with l ∗ = lg (ε), where the asterisk refers to the efficient allocation.
In this situation firms shed all labour in excess of lb when business conditions
turn bad and hire up to lg (ε) = ε − w when business conditions turn good.
Expected profits for a type-ε firm are
p (1 − p)
E∗ (ε) = [ab − w]2 + [ε − w]2
2 2
Profits are obviously increasing in ε.

12.4.3 THE RIGID SYSTEM

Assume now that EPL is so strict that firing is impossible. A type-ε firm
will then choose a level of employment that maximizes average profits, and will
keep its employment constant at all time. In other words, a type-ε firm will
choose a level of employment to maximize average expected profits:
   
R 1 2 1 2
 (ε, l) = p ab l − l − wl + (1 − p) εl − l − wl
2 2
where R (ε, l) are the profits for a type-ε firm in the rigid system. If we indicate
with l R (ε) the result of the maximization, its expression reads
l R (ε) = pab + (1 − p)ε − w
Confronting the rigid and the efficient allocation, an important implication
immediately follows.
Average employment for a type-ε firm in the efficient and in the rigid alloc-
ation is identical. The result is obtained by simple inspection of l R (ε), which
can be written as l R (ε) = p(ab − w) + (1 − p)(ε − w). But then l R (ε) is
the average level of employment of a type-ε firm in the efficient allocation.
Further, profits are larger in the efficient allocation, as long as p is different
from 0 and 1.
To obtain the latter result simply observe that profits in the rigid system are
1
R (ε) = [pab + (1 − p)ε − w]2
2
 EMPLOYMENT PROTECTION LEGISLATION 215

which is an expression that is always lower than E∗ (ε) as long as p is strictly

positive and less than one. In addition, one can also observe that firm employ-
ment in the rigid system is less volatile than in the efficient allocation, since
firms never hire and fire. These results are the standard implications of the
EPL literature with fixed wages, and are just reported for introducing threshold
effects, to which we turn next.

12.4.4 THE ROLE OF THRESHOLD EFFECTS

Assume now that the rigid regime is enforced only for an employment level
larger than l thr , where l thr is an exogenous threshold specified by the legislation.
The only restriction we impose is that l thr > ab − w, otherwise the problem is
not even interesting. In this setting, once a firm grows beyond the employment
level l thr firing becomes impossible, while it can take place at no cost for
employment levels less than or equal to l thr . With threshold effects, some type-
ε firms have the option to permanently fluctuate in the flexible fringe of the
firm size distribution, or in the interval l ∈ [lb , l thr ], where lb is the efficient
level of employment when business conditions are bad. We label these type of
firms ‘constrained firms’, and their formal definition follows.

Constrained Firm: A type-ε firm with efficient employment allocation in good business con-
ditions larger than the threshold (lg (ε) > l thr ) is constrained when it employs l = lb in bad
business conditions and l = l thr in good business conditions.

Thus, a constrained firm never passes the threshold, sheds labour up to

lb when business conditions turn bad, and hires up to l thr when business
conditions are good, and features an average employment level l SC = plb +
(1 − p)l thr . A constrained firm follows a stay-small policy, since in good times
it is reluctant to grow beyond the threshold. Expected profits of a constrained
firm are
   
SC 1 2 thr 1 thr2 thr
 (ε, l) = p ab l − l − wl + (1 − p) εl − l − wl
2 2
while its employment behaviour is

∗ l = ab − w if α = ab
l (ε) = b
lg (ε) = l thr if α = ε
so that the average level of profits is increasing in the idiosyncratic level ε.
With threshold effects, some firms have to choose between a rigid allocation
and a stay-small policy. In the former case they have an employment base
larger than the threshold, they permanently employ l R (ε), and never fire. In
216 EMPLOYMENT PROTECTION LEGISLATION

the latter case, they permanently fluctuate inside the flexible fringe of the size
distribution.
To complete our description, we need to characterize the conditions that
ensure that constrained firms exist in equilibrium. In general, a type-ε firm
will be constrained and will follow a stay-small policy as long as its average
profits are higher than the average profits from the rigid system, or when
ESC (ε) > ER (ε). Among other things, this condition clearly depends on
the specific value of the idiosyncratic parameter ε, as we show in our next
result.
Firms in the interval ε ∈ [ε∗ , ε∗∗ ] are constrained, where ε∗ = l thr + w
and ε ∗∗ is a positive number larger than ε∗ . In light of this result, the firm
size distribution is partitioned in three intervals. Firms with idiosyncratic
component ε lower than ε ∗ are totally efficient and do not interact in any
way with the threshold (their employment level in good times is lower than
the threshold). Firms with idiosyncratic component in the interval [ε ∗ , ε∗∗ ]
are constrained, and in good times bunch with employment l thr = ε∗ − w.
Finally for idiosyncratic values of ε larger than ε ∗∗ , firms are rigid and hire
l R (ε).
To prove this result one needs simply to introduce the function
z(ε) = E SC (ε) − E R (ε)
whose expression reads
 
p 2 thr 1 thr 2
z(ε) = (ab − w) + (1 − p) εl − l − wl
2 2
1
− [pab + (1 − p)ε − w]2
2
First note that the threshold is irrelevant for those firms for which lg (ε) <
l thr , which is a condition that is satisfied as long as ε < ε∗ , with ε ∗ = l thr + w.
Type-ε firms with idiosyncratic component below ε ∗ are totally efficient and
do not interact in any way with the threshold.
Second, note that z(ε∗) = p(1 − p)[ab − w − l thr ]2 > 0 and that z(ε∗ ) > 0
so that firms with ε > ε∗ are certainly constrained. To find the upper support
of the interval [ε∗ , ε∗∗ ] one needs to solve the quadratic equation z(ε) = 0
whose largest root reads
ε∗ − (pab )1/2
ε∗∗ =
1 − p 1/2
It is immediately clear that ε∗∗ > ε ∗ strictly, so that all firms in the interval
[ε∗ , ε∗∗ ] are constrained. Conversely, for ε > ε∗∗ z(ε) < 0 and firms choose
the rigid system.
 EMPLOYMENT PROTECTION LEGISLATION 217

MPL, w

ab
A B
D
C
w

lb lthr lR(e) lg(e) l

Figure 12.1. Labor demand in good and bad times and threshold effects

An important result easily follows.

A type-ε firm that is constrained has an average employment level that is
lower than the average employment level in the efficient allocation and in the
fully rigid system.
From the definition of constrained firms it follows immediately that lg (ε) >
l thr for all ε ∈ (ε ∗ , ε∗∗ ] where lg (ε) = ε − w. Since l R (ε) − l SC (ε) = (1 −
p)[ε − ε ∗ ], it is obvious that all constrained firms with ε ∈(ε∗ , ε∗∗ ] feature
average employment that is lower than the average employment in the rigid
system.
This result is important, since it shows that one of the standard predictions
of traditional EPL models, summarized in the first result above, no longer holds
when constrained firms exist. The last result is further summarized by looking
at Fig. 12.1, where we report the optimal employment level for a type-ε firm
under three regimes: the efficient allocation, the rigid system, and a stay-small
policy. Points A and B in the figure refer to the employment level under the
efficient allocation, when the firm switches its employment level between lb and
lg (ε). Point C refers to the employment level under the rigid system, and l R (ε)
is the amount of labour that the firm permanently employs, independently of
business conditions. When a firm is constrained, its employment level shifts
between point A in bad times and point D in good times. Clearly, the average
between A and D is lower than the employment level associated with point C.
13 Teams and Group
Incentives

13.1 Introduction
Team-organized production is a key feature of many organizations. Most job
advertisements explicitly look for workers that possess team-specific skills, and
are motivated to work in teams. Indeed, team production and group-based
incentives pose a great challenge to personnel economics. As it turns out, a
simple approach to the problem would argue that organized team production
is likely to lead to a reduction in productivity. This is the so-called 1/N th
problem, whereby effort put forward by workers under team production is
likely to be less than what would be exerted under individual-based incentives.
If this is indeed the case, all we can do under team production is to find ways to
preserve the right incentives, even though it appears that overall productivity
will hardly be larger than what it would have been if individual-based incentives
were in place.
Yet, to fully understand team production, and all the emphasis that real
life organizations do place on teams, we have to move forward and expli-
citly recognize that there exist benefits and costs linked to team production.
After all, if so many firms are interested in workers able to work in teams,
it has to be the case that team-specific skills (such as communication skills)
and the possibility of mutual learning among team members may increase
the total surplus generated by the workforce. The personnel economics lit-
erature is still growing in this dimension. Many good questions exist, but
answers are often limited. In this context, we prefer to offer a case study
approach.
Does the adoption of a team increase or decrease productivity? How does
team composition affect productivity? Are teams more productive if members
are homogeneous or heterogeneous? As we argued above, personnel economics
has no full answer to this problem. What we can do is to bring empirical
research to the topic, and analyse in close detail a case study on the adoption
of team production in the garment industry. Surprisingly, and contrary to
what the 1/N th problem would predict, we will see that in the specific study
analysed, individual productivity under group incentives has increased, rather
than decreased. This suggests that there is still a lot to be learnt in the area of
team production.
 TEAMS AND GROUP INCENTIVES 219

13.2 The Team Production Problem

We begin by presenting the simple team production problem, or the 1/N th
problem, as it is often defined in the literature. Consider a group of individuals
of size N that work together on a specific project. The price of the output is
normalized to 1 and the production is such that the total revenue produced by
the entire team depends on the effort of every team member via the simple,
additive production function:
Q = e1 + e2 + e3 + · · · + eN
where ei is the effort of individual i.
We can think of a team of software developers engaged on a project with a
time deadline. There is an agreed payment and a target delivery date for the
final product, but each day of delay reduces the fee received by the entire team
by $X . Total output is the simple sum of the effort of all individuals.
Note that, according to the above example, (i) all team members are equally
productive, and that by assumption (ii) the production function is linear, so
that there are no complementarities between the team members’ efforts. While
complementarities make team production more advantageous, neither of these
assumptions affects the main results we will come up with below.
Finally, assume that each agent’s disutility of effort is given by:
ei2
C(e) = ∀i
2
We now ask two questions in turn. First, what is the efficient effort level for
each member of this team? Second, given a simple revenue-sharing compens-
ation scheme, how hard will each team member actually work (the ‘worker’s
problem’)?
Efficient effort level. To calculate the efficient effort of each and every team
member we have to consider the benefits and the costs associated with increas-
ing effort of each agent. Total output is given by i ei while the total cost of this
effort is i C(ei ). This suggests that the efficient effort level is the solution to
  e12 e2
Maxe1 ,e2 ,...,en = ei − C(ei ) = e1 + · · · + eN − − ··· − n
2 2
i i

where the maximization is to be taken for each and every member

1 − ei = 0 ∀i
Therefore the efficient outcome is for each worker to supply at the point in
which the marginal benefit of effort to aggregate team production (which is 1)
is equal to the marginal cost of effort, which in this case is equal to ei .
220 TEAMS AND GROUP INCENTIVES

Worker’s problem. Beside the efficient level, we know that each individual
solves their own maximization problem by maximizing their utility. The utility
of the individual depends, as always, on the compensation scheme and on the
cost of effort, so that
ei2
Ui = compensationi −
2
The most obvious compensation scheme is one in which the total output from
the team is shared proportionally among its team members. In other words,
each worker maximizes his or her individual utility, given a compensation
schedule, or ‘output sharing rule’. Further suppose this sharing rule implies the
e
following compensation, compensationi = N i ∀i. In other words the team
members agree to share the total revenues of the team equally. In this situation,
how hard will people actually work?
Individual i’s pay-off: Yi = e1 +eiN+...eN ; Individual i’s disutility of effort:
e12
C(ei ) = 2. So the individual’s problem is:

e 1 + e i + . . . eN e2
max − i
e1 N 2
with the following first-order condition:
1
− ei = 0
N
An Individual’s (privately) optimal effort is therefore given by e1∗ = 1/N .
How can the previous conclusion be explained, and why is the individual
optimal effort so different from the efficient level? The intuition is as follows.
While the worker is still bearing the full cost of his or her additional effort,
the compensation scheme implies that the marginal benefit of one additional
unit of effort is only 1/N . This is called the ‘free rider problem’ or the ‘1/N
problem’. The total effort of all agents is N × N1 = 1 instead of N , which is the
efficient level.
Note that this result has nothing to do with the assumption that the group’s
output is shared equally among its members. It is true for any fixed sharing
rule, where worker i receives a share γi of the total, as long as γi = 1 (all the
workers’ shares must add up to 100 per cent).
Under the above rule in our example each worker’s effort will equal γi , and
the sum of all the workers’ efforts will equal γi = 1, rather than N which is
needed for efficiency. Summarizing our result:
Any group compensation rule that shares group output according to a fixed
rate induces sub-optimal effort due to free riding. The problem increases in
severity as the group size, N , increases.
 TEAMS AND GROUP INCENTIVES 221

The free rider problem associated with team production is quite common
in reality, even outside strict personnel economics examples. Among students,
when group projects are associated with a common grade, it is very difficult to
induce efforts by all members, and there is always a tendency to rely on other
people’s effort and ability. A somewhat similar problem takes place when a
group of friends decide to share restaurant bills. Each individual has a tendency
to consume expensive meals, so as to rely on other friends’ contributions.

13.3 Team Norms as Remedies to the Team

Production Problem
In the above example the individual reward was assumed to be 1/N Q or γ Q
where γ can be any share as long as γi = 1. Such a revenue-splitting game
can be thought of as a special case of our linear compensation scheme, which
is based on a fixed amount α plus a bonus component β. Think of these as
special cases of a linear compensation schedule based on the group’s output
which can be written as
compensationi = α + βQ
We can ask what should be the value of β that induces the efficient level of
effort, which we know should be e ∗ = 1. In other words the problem becomes
one of finding β to induce an efficient effort E ∗ = 1 by every team member.
The individual problem is
e12
Max α + β[e1 + · · · + en ] −
2
Hence individual i’s problem yields
ei = β
Therefore to get the worker to choose e ∗ = 1, we need to set β = 1. This
is not surprising, since we know very well that effort is maximum when the
individual is the full residual claimant, as the following conclusion highlights.
To induce efficient effort by every member of a team, each team member’s
individual compensation must increase by 1 for every 1 increase in the
group’s output. Each individual must be the full residual claimant.
This makes clear how difficult it is to find such a scheme. The difficulty with
the previous comment is that the output of all individuals must increase by $1
for every $1 increase in the group’s output. It sounds quite hard.
222 TEAMS AND GROUP INCENTIVES

One way to obtain this scheme is that the ‘firm’ or some agreed-on team
member sets a target output M , which we can call team norm M .
• Everyone is paid zero if the output M is not reached, and the firm keeps
the difference.
• The output is split evenly among team members if the output norm is
reached, while the pay is zero otherwise.
• If the team norm M is the output when all individuals produce at the
efficient level e ∗ = 1, then each worker gets paid $1 and the firm breaks
even.
As you can see, establishing a team norm is the way to keep the incentive
right for workers, since that marginal increase of producing at the efficient
level is exactly $1. But the very difficult task is to establish the proper norm,
and to enforce it.

13.4 Teams in Reality

Many firms use teams. Using teams involves both costs and benefits. The
existing literature argues that teams are desirable for three main reasons: (i)
to make possible gains in complementarities in production among workers;
(ii) to facilitate gains from specialization by allowing workers to accumulate
task-specific human capital; (iii) to encourage gains from knowledge transfer.
Yet teams also have costs. The most important cost is the free rider problem,
which was analysed above. So it is difficult to say ex ante what is the impact
of teams on output. Let’s see some key questions. Does the adoption of a
team increase or decrease productivity? How does team composition affect
productivity? Are teams more productive if members are homogeneous or
heterogeneous? In a case study of the garment industry at Koret it is possible
to answer some of these questions.
Worker heterogeneity Theoretically, worker heterogeneity in teams has two
advantages. It facilitates mutual learning and can influence the group produc-
tion norm. (i) Mutual learning suggests that more able workers are able to
teach less able workers to be more productive. If the mutual learning effect
is significant, then teams that are initially more heterogeneous in ability will
perform better. (ii) Team norm. A relationship between worker heterogeneity
and team performance could also be the result of forming a team norm, a
concept that we formally defined above.
Worker decision to join a team Let us imagine a context in which workers are
given the possibility of joining a team, and let us also imagine a setting in which
non-team jobs are available in the workplace. Then each worker will solve a
 TEAMS AND GROUP INCENTIVES 223

cost–benefit analysis over the decision to enter team production. Obviously, the
highest-ability worker will join a team only if he or she obtains an additional
source of surplus from team production. In terms of output performance,
high-ability workers are bound to suffer, since they are going to share pro-
duction with less productive co-workers. This suggests that for a high-ability
worker to join the team, there must be additional reasons, which can be concep-
tualized in two ways: (i) productivity gains may derive from multi-skill abilities;
(ii) socialization within the team may compensate high-ability workers in
terms of income. As we see in the case study below, these phenomena do exist in
reality.

13.5 A Case Study: Production at Koret

The case study is based on weekly productivity reports from a Koret gar-
ment manufacturing facility in Napa, California. The establishment produces
‘women’s lowers’ including pants, skirts, shorts, etc. Prior to 1995, produc-
tion was organized with individual piece rate. Over 1995–7, production
organization shifted to team production. The organization change was intro-
duced in response to demand by retailers that companies make just-in-time
deliveries. Such demands required a more flexible production system, and
many firms in the industry responded to such demands with team-based
production.
Garment production is done in three stages. First, cloth is cut into pieces
that conform to garment patterns. Second, garments are assembled by sewing
pieces together. Third, garments are finished by pressing. The case study we
are considering focuses on the sewing operation.
The change that we analyse is a passage from progressive bundling sys-
tem production (PBS) to module production (MP). In PBS sewing operations
are broken down into distinct operations. PBS is a piece rate scheme while
MP is a team-based remuneration. Sewers are paid on the basis of individual
piece rates according to a standard set for the operation they undertake. The
standards are usually set by the management in accordance with the uni-
ons. Quality is evaluated by randomly selecting six out of forty garments and
the sewer’s name is recorded. In this setting piece rate is appropriate and
possible.
In 1994 the plant manager began experimenting with flexible teams, which
in the garment industry are called module production (MP). The manager
asked for volunteers. After joining a team sewers could return to PBS if they
preferred. The data are described in Table 13.1. In module production, each
team includes six or seven team members who work in a U-shaped work space
of 4 × 8m. The close proximity of workers and machines reportedly facilitated
224 TEAMS AND GROUP INCENTIVES

Table 13.1. Summary statistics for individual and team incentives

at Koret

Variable Overall (1) Not in team (2) Team member (3)

Productivity
25% 80.12 71.55 87.58
50% 99.33 86.8 105.45
75% 118.3 108.5 120.38
Weekly earnings
25% $197.52 $159.64 $226.66
50% $274.02 221.77 $301.03
75% $350.26 291.51 $369.06
Hours/week
25% 28.52 24 30.01
50% 34.4 32 35.72
75% 38.8 38.33 38.86
Age
25% 30 30.9 30.1
50% 37 39.7 35.9
75% 45.4 48 44.3
Observations 20,626 6,688 13,938

team members’ ability to identify bottlenecks and changes in productivity.

Teams (or modules) are compensated with a group piece rate for each operation.
The team’s net receipts are shared and divided equally. Worker productivity in
PBS and MP is measured in the same standards and it is comparable.
The case study is based on personnel data of employees at Koret from
1 January 1995 until 31 December 1997. The data consist of weekly inform-
ation on worker productivity, pay, hours worked, and team membership for
all individuals employed at Koret. There are no data on individual education.
It is possible to use individual data observed under both regimes to assess the
impact of teams on productivity. The productivity variable is measured as effi-
ciency relative to the standard described, with numbers above 100 indicating
performance above the standard level. Figure 13.1 plots median weekly pro-
ductivity at the plant from the first week of 1995 (week 0) to the last week of
1997 (week 156). In addition, the fraction of plant workers engaged in team
production is also presented. The figure shows that median productivity at Koret
increases after most Koret workers are working in teams. This is clearly visible in
the Figure after week 70. The picture shows also that there is substantial cyclical
variation in productivity.É
Table 13.1 presents summary statistics for the person-week data, overall
and by membership status. Consistent with Fig. 13.1, columns 2 and 3 indicate
that productivity was higher under team production rather than lower, as

É In technical terms the presence of cyclical variation implies that a proper multivariate analysis
should include also time dummies.
 TEAMS AND GROUP INCENTIVES 225

100 120
90
110
80
Percent of workers in team

100

Worker productivity
70
60
90
50 Median weekly
worker productivity 80
40
30 70
20
Fraction of workers in team 60
10
0 50
0 26 52 78 104 130 156
Number of weeks since 1/1/1995

Figure 13.1. Average Productivity and percentage of workers employed in teams

would be predicted if free riding were dominant. Table 13.2 reports average
weekly worker productivity for individuals prior to joining the team and team
productivity after joining the team. Productivity increased in fourteen of the
twenty-three teams for which data are available. Teams formed in 1995 are the
most likely to show an increase in productivity, suggesting that workers with
greater collaborative skills joined the early teams.
The case study analysed three questions in detail:
1. Did the use of teams lead to higher productivity, contrary to what might
be predicted by simple models of free riding?
2. Which type of workers (in terms of ability) joined the teams? Is there
an adverse selection into teams (with less able workers joining teams) as
the free riding problem would suggest? Or maybe there are collaborative
skills that were not used in PBS?
3. How does team composition affect team productivity?

13.5.1 THE IMPACT OF TEAMS ON PRODUCTIVITY AT KORET

Let yit be the natural log of productivity of worker i in week t at Koret.
A worker’s productivity is modelled as
yit = Xit β + αTEAMit + it (13.1)
where the indicator variable TEAMit equals one if worker i is a member of
a team in week t and zero otherwise. The variables included in the vector X
226 TEAMS AND GROUP INCENTIVES

Table 13.2. Dates of team formation and average weekly productivity

before and after team formation

Team Date of Average weekly productivity

team formation (1)
Individuals prior to Team (for weeks 21+)b (3)
joining teama (2)

1 12 March 1994 97.8 114.3

2 7 January 1995 82.9 122.6
3 28 January 1995 79.4 97.6
4 28 January 1995 94.0 106.0
5 28 January 1995 117.8 118.9
6 28 January 1995 89.4 88.3
7 29 April 1995 89.6 107.8
8 7 October 1995 122.6 115.6
9 28 October 1995 127.4 131.3
10 13 April 1996 85.6 83.6
11 30 March 1996 100.4 111.8
12 13 April 1996 87.3 109.3
13 13 April 1996 94.6 106.1
14 13 April 1996 85.6 91.2
15 18 May 1996 78.3 76.8
16 22 June 1996 81.1 82.6
17 20 July 1996 81.7 122.9
18 13 April 1996 92.6 95.5
19 13 April 1996 86.1 79.7
20 10 August 1996 127.5 114.4
21 7 December 1996 …c 139.1
22 18 January 1997 94.0 80.0
23 1 February 1997 89.2 70.9
24 15 March 1997 92.1 61.2
25 6 September 1997 76.9 …c

a Entries in col. (2) are calculated by averaging the individual person week productivity values of
workers who subsequently join the particular team (individuals are weighted by the length of time
they spent on the team).
b Team averages in col. (3) are calculated after excluding the first 20 weeks.
c Team 21 consisted of almost all new hires, and so the pre-team productivity data are not available.

include individual age (and squared), monthly data on US women’s retail

apparel sales (as well as future sales up to six months), and monthly dummies.
The key parameter is α and the results are reported in the table.
Contrary to the predictions of moral hazard models that emphasize free
riding, the estimate of the productivity parameter α was positive. After the
adoption of teams, productivity increase was 18 per cent.Ê There are potentially
two factors that can explain this finding. First, for any particular individual,
joining a team may lead to an increase in productivity. Second, high-ability
workers may systematically choose to join teams. The first effect would be a
pure team productivity effect while the second one would be an adverse selec-
tion effect. To distinguish between these two potential explanations, consider

Ê Using quantile regression, the productivity increased to roughly 21%.

 TEAMS AND GROUP INCENTIVES 227

the following specification of the above equation

yit = θi + Xit β + αTEAMit + ηit (13.2)
where θi represents a set of unobserved characteristics that influence pro-
ductivity of worker i under both PBS and MP. In this new specification the
term α offers the productivity impact of team production once we control for
individual specific characteristics. The results of the new regressions suggest
that, on average, a particular productivity increases by roughly 14 per cent after
joining a team. Basically, the estimate of α in model (13.1) falls to 14 per cent in
model (13.2). Of the 18 per cent early increase in productivity associated with
teams at Koret, one-fifth reflects the systematic selection of high-ability work-
ers under the PBS system. Overall, the finding is consistent with the utilization
of collaborative skills but also with bargaining and mutual learning, which
appear (in this particular case) to more than offset any possible free riding.

Were early teams more productive?

One possible drawback of the model above is that it considers the productivity
of all teams to be the same, independent of the data of team formation. Since
team formation was voluntary, one may expect that early teams were more
productive since workers with higher collaborative skills might be the first to
participate in teams. The estimate of this effect is done by adding the following
three variables to the model above:
yit = θi + Xit β + αTEAMit + δ1 TEAM 95 + δ2 TEAM 96 + δ3 TEAM 97 + ηit
There is substantial heterogeneity across teams in their impact on output.
Workers on teams formed in 1995 had double the impact on productivity.
Since these effects could be due to team tenure effects, with early teams charac-
terized by more experience with team membership, the estimates in Table 13.3
distinguish among the impact on productivity of different teams at different
weeks. The results are confirmed, and do reflect a team effect. Overall, the
estimates suggest that the introduction of teams led to higher productivity
for teams formed in the first year and that the productivity declined as more
workers in the firm engaged in modular production. They suggest that workers
with higher levels of collaborative skills that were productive in a team setting
tended to participate in teams first.

13.5.2 WHO JOINS A TEAM?

The results of the previous section suggest that there are important self-
selection mechanisms that operate under team formation at Koret. This is
not particularly surprising. The surprising result, however, is that it seems that
228 TEAMS AND GROUP INCENTIVES

Table 13.3. Relationship between teams and productivity

Dependent variable: Ln(productivity)

No individual dummies Individual dummies

fixed effects

OLS Median OLS Median

A. All teams grouped together

TEAM 0.178 0.211 0.136 0.116
(0.053) (0.044) (0.035) (0.032)
B. Teams grouped by year of formation
TEAM95 0.221 0.234 0.177 0.169
(0.052) (0.048) (0.034) (0.038)
TEAM96 0.06 0.092 0.088 0.072
(0.07) (0.061) (0.044) (0.038)
TEAM97 −0.09 −0.094 −0.072 −0.082
(0.12) (0.186) (0.065) (0.068)
C. Team effects by year of formation and team tenure
TEAM95
Weeks 1–10 0.394 0.221 0.31 0.11
(0.183) (0.175) (0.161) (0.095)
Weeks 11–20 0.331 0.286 0.241 0.167
(0.071) (0.055) (0.074) (0.055)
Weeks 21+ 0.201 0.237 0.183 0.204
(0.05) (0.043) (0.046) (0.039)
TEAM96
Weeks 1–10 −0.087 −0.073 −0.046 −0.081
(0.118) (0.118) (0.081) (0.06)
Weeks 11–20 −0.019 0.057 −0.023 0.039
(0.084) (0.074) (0.065) (0.04)
Weeks 21+ 0.079 0.118 0.15 0.131
(0.063) (0.059) (0.048) (0.047)
TEAM97
Weeks 1–10 −0.365 −0.235 −0.277 −0.18
(0.092) (0.169) (0.102) (0.144)
Weeks 11–20 0.064 0.128 0.101 0.161
(0.086) (0.357) (0.121) (0.19)
Weeks 21+ −0.034 −0.165 0.002 −0.043
(0.166) (0.279) (0.075) (0.102)

Notes: Number of observations is 20,627 person weeks. Standard errors are in

parentheses. Each regression also includes a constant, age of the worker and, its
square, dummies for each month, and cyclical variable measuring women’s retail
garment sales.

more productive workers, rather than less productive, were the first to join
teams. The 1/N problem would suggest that most able workers have the most
to lose from team membership. One can examine the issue in more detail. To
investigate the self-selection into teams, one can estimate a ‘pre-team’ regres-
sion for the 151 non-team workers employed in week 0, when teams were not
yet formed. This simple regression takes the form

yi0 = Xi0 β + γ FTEAMi0 + εi0

 TEAMS AND GROUP INCENTIVES 229

Table 13.4. The impact of future team participation

Dependent variable: Ln (productivity)

Variable OLS (1) Median (2)

A. All future teams pooled

FTEAM (joins team in the future) 0.156 0.145
−0.07 −0.129

B. Future teams formed in 1995 and 1996

FTEAM95 (joins team in 1995) 0.206 0.154
−0.078 −0.131
FTEAM96 (joins team in 1996) 0.061 −0.015
−0.093 −0.186

Notes: Number of observations is 151. Standard errors are in parentheses

(robust standard errors for OLS). Standard errors for the median regressions
are bootstrapped using 500 replications. All regressions include a constant
and age and its squares.

where yi0 is the ln(productivity) of worker i in the first week of January 1995
(week 0) and the indicator variable FTEAMio equals one if worker i joins a team
by mid-1996 and zero if she does not. Basically the coefficient γ captures the
productivity under PBS of the individuals that later decided to enter team pro-
duction. Because future team membership cannot affect current productivity,
γ measures the selection into teams. Panel A of Table 13.4 indicates that work-
ers who joined teams in future are approximately 15 per cent more productive
in non-team work in January 1995. This result goes against the free rider prob-
lem. It is not that low-productivity workers are sorted into teams because of
the adverse selection problem. Just the opposite occurs. Either technical ability
is somewhat positively correlated with collaborative skills or teams offer non-
pecuniary benefits that are larger for high-ability workers. In addition, it may
be that the disutility imposed by a team norm on low-productivity workers was
so large as to discourage many of them from joining teams at the beginning.
The big puzzle is what happens to earnings. Regressions of the log of hourly
pay on team membership (controlling for demographics and cyclical factors)
in Table 13.5 show that while average pay increased after the introduction of
teams, workers at the top end of the pay distribution experienced an 8 per cent
reduction in hourly pay under teams. The overall increase in average pay is
not surprising, especially since we saw that overall productivity increased. The
surprise is the fall in wage at the top of the distribution. This suggests that
non-pecuniary benefits of team membership, such as more control over the
work environment and less repetition, are important factors for many workers.
Further analysis shows also that the pay results are not caused by the fact
that high-ability workers participated in teams only for a few periods and then
quit the firm when they learned that their pay was unlikely to rise.
230 TEAMS AND GROUP INCENTIVES

Table 13.5. The impact of teams on hourly and weekly pay

OLS 0.05 0.25 0.5 0.75 0.95

Variable −1 −2 −3 −4 −5 −6

A. Dependent variable: ln (hourly pay)

TEAM 0.158 0.101 0.233 0.222 0.148 −0.083
−0.035 −0.044 −0.38 −0.045 −0.046 −0.039
B. Dependent variable: ln (weekly pay)
TEAM 0.26 0.285 0.348 0.312 0.24 0.026
−0.039 −0.076 −0.042 −0.041 −0.039 −0.063

Note:The number of observations is 20,193. Standard errors are in parentheses (robust standard errors
for OLS regressions). Each regression also includes a constant, age of the worker and its square, dummies
for each month.

13.5.3 HOW DOES A TEAM COMPOSITION AFFECT TEAM

PRODUCTIVITY?
While it is obvious that teams with high-ability members should be on average
more productive, an interesting question concerns the links between hetero-
geneity and productivity. The question here is whether worker heterogeneity
increases or decreases team productivity. Theoretically, it may be the case that
better workers may be able to impose a higher team norm output level, espe-
cially when there are non-team jobs at the plant. Similarly, if mutual learning is
more important, heterogeneity may lead to an increase in productivity. Thus,
the ability composition of teams must be carefully defined.
The estimates show that more heterogeneous teams are more productive,
with average ability held constant, since the coefficient on the ratio of the max-
imum to the minimum individual productivity of team members is positive
and significant. When one is forming a team, it appears better to have a mix of
high-ability and low-ability workers rather than a set of workers with identical
technical abilities. It may be that high-ability workers are able to enforce a
higher team norm by exerting their bargaining leverage. On the other hand,
the results may reflect mutual learning, in which more able workers are able to
teach their less able teammates to be more productive.

13.5.4 DISCUSSION AND CONCLUSION

The case study at Koret can be summarized by four key findings. First, the
adoption of teams at Koret improved worker productivity by 14 per cent on
average, even after we accounted for the self-selection of high-productivity
workers under the PBS production system onto teams. Second, this productiv-
ity improvement was greatest for the teams that formed earliest and diminished
 TEAMS AND GROUP INCENTIVES 231

as more workers in the firm engaged in modular production. Third, high-

ability workers not only tended to join the teams first, but were no more likely
to subsequently leave the firm after joining a team. Finally, more heterogeneous
teams are more productive, when average ability is held constant.
These results obviously suggest that group piece rate incentives achieved
higher productivity than individual piece rate incentives. Free riding does not
appear to be the dominant behaviour response at Koret. The lack of free rid-
ing may be simply due to the ease of peer monitoring within teams at Koret.
However, the possibility of straightforward mutual monitoring cannot explain
why productivity increased significantly.
These results are certainly consistent with the view that workers have both
technical and collaborative skills (such as communication and leadership skills
and flexibility), the latter of which were not utilized under PBS production. In
this respect, team production may have expanded production possibilities by
utilizing collaborative skills.
It is also difficult to escape the conclusions that workers received some non-
pecuniary benefit by participation in teams: some high-ability workers joined
teams despite an absolute decrease in pay. Teams offered more varied and less
repetitive work and also reduced variation in weekly pay. The results are also
consistent with the predictions of bargaining and mutual learning models.
More heterogeneous workers are more productive, and ‘stars’ are influential in
raising team productivity.
Overall, these arguments suggest that it may be feasible for a firm to reduce
turnover and increase production by introducing team production, even if
some workers earn lower pay after joining the team. Overall, this chapter
suggests that team production is a complex behavioural phenomenon that
presents a challenge because it involves multiple mechanisms.
b APPENDIX A LABOUR DEMAND AT THE FIRM LEVEL

PRODUCTION FUNCTION
The production function is the technology that the firm uses in the production process.
For simplicity, in Appendix A we assume that there are only two factors of production
(two inputs): the number of employee-hours hired by the firm (L) and capital (K ).
We write the production function as

q = f (L, K )

where q is the firm output, and f is the technological relationship that transforms
inputs into output. The production function specifies how much output is produced
by any combination of labour and capital. With respect to the labour input, there are
two key assumptions to be discussed in detail. First, the number of employee-hours
L is given by the product of the number of workers hired times the average number
of hours worked per person. Chapter 3 considers the distinction between the number
of workers hired and the number of hours worked. In this appendix we simply refer
to labour input L as the number of workers hired. Second, the production function
assumes that different types of workers can be aggregated into a single input that we call
labour. In fact workers are heterogeneous, and Chapter 2 discusses issues of workers’
heterogeneity.

MARGINAL PRODUCT
The marginal product of labour, which we denote as MPL , is defined as the change in
output resulting from hiring an additional worker, holding constant the quantities of
all other inputs. The MPL is formally given by

q


MPL =
L
K

where the notation |K should be read as holding capital constant.É

É The marginal product of capital is similarly defined as

q


MPk =
K
L
 LABOUR DEMAND AT THE FIRM LEVEL 233

The marginal product of labour is the slope of the total curve, which can be thought
as the relationship between output and labour holding capital constant. There are two
important properties of the marginal product

1. The marginal product is non-negative. This property is a minimum basic require-

ment that we assume to be satisfied throughout the book. Adding more of one
input, holding constant the other input, never leads to a reduction in output,
and in general it increases. This condition simply means that adding one extra
employee inside the firm, at given capital stock, increases output. While such an
assumption is most of the time reasonable, one can envisage conditions in which
it does not hold, mainly because extra workers lead to increased congestion inside
the firm.
2. Output increases at a decreasing rate. The assumption that the marginal product
of labour declines follows from the law of diminishing returns: as more and more
workers are added to a fixed capital stock, the gains from specialization decline
and the marginal product of workers declines. While we will often assume that
the law of diminishing returns operates over some range of employment, various
chapters in the book assume that the marginal product is constant.

It is easy to understand how to calculate the marginal product of labour by using a

numerical example. The baseline data are reported in Table A.1, where we summarize
the firm’s production when it hires different numbers of workers, holding capital
constant. If the firm hires one worker, it produces 11 units of output. The marginal
product of the first worker hired, therefore, is 11 units. If the firm hires two workers,
production rises to 21 units of output, and the marginal product of the second worker
is 10 units. Figure A.1 illustrates the total product curve. This curve describes what
happens to output as the firm hires more workers. The total product curve is obviously
upward sloping. The marginal product of labour is the shape of the total product
curve, and it is plotted in Fig. A.2. In our example, output first rises at an increasing
rate as more workers are hired. This implies that the marginal product of labour is
rising. Eventually, output increases at a decreasing rate. In our example, the marginal
product of the fourth worker declines further at 8 units.

Table A.1. Labour demand at the firm level

Employment Output Marginal prod Wage1 Wage2

0.00 0.00 — 5.00 8.00

1.00 11.00 11.00 5.00 8.00
2.00 21.00 10.00 5.00 8.00
3.00 30.00 9.00 5.00 8.00
4.00 38.00 8.00 5.00 8.00
5.00 45.00 7.00 5.00 8.00
6.00 51.00 6.00 5.00 8.00
7.00 56.00 5.00 5.00 8.00
8.00 60.00 4.00 5.00 8.00
9.00 63.00 3.00 5.00 8.00
10.00 65.00 2.00 5.00 8.00
234 APPENDIX A

70.00

60.00

50.00

40.00
Output

30.00

20.00

10.00

0.00
1 2 3 4 5 6 7 8 9 10
–10.00 Employment

Figure A.1. Production function with constant capital

PROFIT MAXIMIZATION
We assume that the firm’s employment decision maximizes profits. The firm’s profits
are given by:
π = pq − wL − rK
where p is the price at which the firm can sell its output; w is the wage rate (that is,
the cost of hiring an additional worker) and r is the price of capital. In this appendix
we assume that the firm is perfectly competitive, both in the output market and in
the input market. This simply means that the firm takes as given the price p, the wage
w, and the cost of capital r. Further, such quantities are independent of how much
the firm produces and hires. If we substitute the production function into the profit
equation we obtain
π = pf (L, K ) − wL − rK
Since p, w, and r are constant, the previous equation makes clear that the only actions
the firm can take to influence profits involve the decisions of how much labour and
capital to hire. The firm maximizes profits by hiring the right amounts of these inputs.

EMPLOYMENT DECISION IN THE SHORT RUN

The short run is a time span in which the stock of capital is constant at Ko . The firm
profit in the short run is then given by
π = pf (L, Ko ) − wL − rKo
this implies that in the short run the only variable that the firm can ‘play around with’
is the number of workers it hires. In this dimension, the key variable to the firm is the
 LABOUR DEMAND AT THE FIRM LEVEL 235

dollar value of what an additional worker produces. Its expression in formula is

VMPL = pMPL
Since we apply the law of diminishing returns and the price is constant, it is clear that
the VMPL will eventually decline.
In our example, to obtain the dollar value of what an additional worker produces,
we multiply the marginal product of labour times the price of the output. If the price is
equal to $1, then the value of the marginal product is identical to the marginal product.
The competitive firm can hire all the labour it wants at a constant wage of w dollars.
Since each additional worker yields to the firm an amount equal to VMPL it is obvious
that the optimal employment decision in the short run is to hire up to a point in which
the wage rate is equal to the VMPL and the law of diminishing returns applies. In
formula, the employment decision is
VMPL = w and VMPL is decreasing
While the first condition is fairly intuitive (as long as an additional worker yields to
the firm more than his cost he should be hired) the second condition is less obvious.
Indeed, if an additional worker yields as much as he costs, but adding another worker
increases profits more than it costs (i.e. if the VMPL is increasing), then it cannot be
optimal to stop hiring. In Fig. A.2 the second condition is satisfied.
Note that it is possible to obtain the optimal hiring condition also with the help of
a little algebra. If we assume that the production function is differentiable, we can say
that the problem of the firm is to maximize profits with respect to labour. In formula,
this is equivalent to
MaxL π = pf (L, Ko ) − wL − rKo

12.00
Marginal product High wage
11.00
10.00 Low wage
Wage, marginal product

9.00
8.00
7.00
6.00
5.00
4.00
3.00
2.00
1.00
0.00
1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00
Employment

Figure A.2. The marginal product and labor demand

236 APPENDIX A

The first-order condition requires putting the marginal profits to zero

π
=0
L


q

L
p −w =0
L
K L
which implies that
VMPL = w
This is just a first-order condition for profit maximization. To be sure that the quantity
chosen is optimal, it is also necessary that less the second-order condition be negative,
which is like saying that VMPL is decreasing. Let’s now turn to our numerical example:
it is clear that for our competitive firm it is not optimal to hire more than seven workers
when the wage is 5.

LABOUR DEMAND
The demand for labour tells us what happens to the firms’ employment as the wage
changes, holding capital constant. From Fig. A.1, it is clear that the optimal quantity of
labour hired depends on the wage rate. If the wage rate increases to 8, the amount of
labour hired falls, while if the wage rate falls, the amount of labour hired increases. All
this happens as long as the marginal product of labour is downward sloping. It follows
that the firm’s demand for labour in the short run is equivalent to the downward-sloping
segment of its value marginal product of labour schedule.

THE DEMAND FOR LABOUR IN THE LONG RUN

In the long run both labour and capital can be set optimally. This implies that not
only labour but also capital should be set optimally. In formula this yields that the
conditions are
VMPL = w
VMPK = r
which can be expressed as
pMPL = w
pMPK = r
which can be written (using the fraction of the to conditions to eliminate the price) as
MPL MPK
=
w r
pMPL = w
The first condition is the new key condition for long-run profit maximization. The
left-hand side yields the output of the last dollar spent on labour. Indeed an additional
 LABOUR DEMAND AT THE FIRM LEVEL 237

worker yields MPL and costs w so that the ratio gives the output of the last dollar
spent on labour. Similarly, the right-hand side gives the output yield of the last dollar
spent on capital. The condition can be used to obtain the capital–labour ratio. Once
the capital–labour ratio is obtained we can obtain the actual level of labour from the
second condition, which is identical to the short-run maximization.
b APPENDIX B CONSTRAINED OPTIMIZATION

Constrained optimization lies at the heart of many problems in personnel economics.

The central mathematical problem is that of maximizing or minimizing (as in the
case of the optimal skill ratio of Chapter 2 and of the hours–employment trade-off of
Chapter 3) a function of several variables, where these variables are bound by some
constraining equations. The typical problem is
maximize f (x1 , x2 )
where x1 and x2 must satisfy the constraint
h(x1 , x2 ) = C
The function f is called the objective function while h is the constrained function.
The geometric solution to the problem is easy and should be familiar from inter-
mediate microeconomics. First, draw the constraint in the x1 x2 place, the thick line in
Fig. B.1. Then, draw a representative sample of the level curves of the objective func-
tion f (we typically call the level curves indifference curves). Geometrically, our goal
is to find the highest-valued level curve f which meets the constraint set. The highest
level curve of f (consistent with the constraint) cannot cross the constraint curve C;
if it did, nearby higher-level sets would cross too, as occurs in point b in Fig. B.1.
This highest-level set of f must touch C (so that the constraint is satisfied) but must
otherwise lie on one side of C (since it cannot cross over C). Another way of saying
this is that the highest-level curve of f to touch the constraint set C must be tangent
to C at the constrained max. This situation occurs at point (x1∗ , x2∗ ) in Fig. B.1.

x2

x*

h(x1, x2) = C

x1

Figure B.1. The Lagrange method

 CONSTRAINED OPTIMIZATION 239

The fact that the level curve f is tangential to the constraint set C at the constrained
maximizer (x1∗ , x2∗ ) means that the slope of the level set of f equals the slope of the
constraint curve C at (x1∗ , x2∗ ). The slope of the level set f at (x1∗ , x2∗ ) is
∂f
∂x1
− ∂f
(x1∗ , x2∗ )
∂x2

and the slope of the constraint set at (x1∗ , x2∗ )

∂h
∂x1
− ∂h
(x1∗ , x2∗ )
∂x2

The fact that these two slopes are equal at (x1∗ , x2∗ ) means
∂f ∂h
∂x1 ∂x1
∂f
(x1∗ , x2∗ ) = ∂h
(x1∗ , x2∗ )
∂x2 ∂x2

which can be written as

∂f ∂f
∂x1 ∂x2
∂h
(x1∗ , x2∗ ) = ∂h
(x1∗ , x2∗ )
∂x1 ∂x2

To avoid working with (possibly) zero denominators in the previous expression, let
us indicate with µ the common value of the two quotients in the previous expression
so that
∂f ∂f
∂x1 ∂x2
∂h
(x1∗ , x2∗ ) = ∂h
(x1∗ , x2∗ ) = µ
∂x1 ∂x2

and the previous condition can be written as

∂f ∗ ∗ ∂h ∗ ∗
(x1 , x2 ) − µ (x , x ) = 0 (B.1)
∂x1 ∂x1 1 2
∂f ∗ ∗ ∂h ∗ ∗
(x1 , x2 ) − µ (x , x ) = 0 (B.2)
∂x2 ∂x2 1 2
We basically have now three unknowns (x1, x2 , µ), so that we now need three equations,
one more than what is listed above. But we have the third equation and it is simply the
constraint h(x1 , x2 ) = C. Including the constraint equation with the two equations
in (B.1) and (B.2) yields a system of three equations in three unknowns that read
∂f ∗ ∗ ∂h ∗ ∗
(x , x ) − µ (x , x ) = 0
∂x1 1 2 ∂x1 1 2
∂f ∗ ∗ ∂h ∗ ∗
(x , x ) − µ (x , x ) = 0 (B.3)
∂x2 1 2 ∂x2 1 2
h(x1 , x2 ) = C.
240 APPENDIX B

There is a convenient way of writing the system. Form the Lagrangian function as

L(x1, x2 , µ) + µ[C − h(x1 , x2 )]

and find the critical values of the Lagrangian L by computing

∂L
=0
∂x1
∂L
=0
∂x2
∂L
=0
∂µ
One can easily see that the result of this process is precisely the system of equation (B.3).
Note also that since µ just multiplies the constraint in the definition of L, the equation
∂L/∂µ is just the constraint h(x1 , x2 ) = C. This new variable µ which multiplies the
constraint is called a Lagrange multiplier.
This process is somewhat magical. When we want to maximize a function in an
unconstrained problem, we simply solve for its critical points by setting its first-order
partial derivatives equal to zero. However, by the introduction of the Lagrange multi-
plier µ into the constrained problem, we have transformed a two-variable constrained
problem to the problem of finding the critical points of a function L(x1, x2 , µ) of one
more variable. In other words, we have reduced a constrained optimization problem
in two variables to an unconstrained problem in three variables.É The cost of this oper-
ation is the inclusion of a new and artificial variable µ. Yet, in economic application µ
has a clear economic meaning. In most economic applications the multiplier measures
the sensitivity of the optimal value of the objective function to changes in the right-
hand sides of the constraints and, as a result, it provides a natural measure of the value
of scarce resources in a maximization problem.
The problem:

maximize f (x1 , x2 )
subject to h(x1 , x2 ) = c

whose solution is x1∗ , x2∗ is solved (as long as x1∗ , x2∗ is not a critical point of h) by setting
up a Lagrange function

L(x1, x2 , µ) = f (x1 , x2 ) + µ[C − h(x1 , x2 )]

and the solution x1∗ , x2∗ is obtained by finding a critical point x1,∗ x2∗ , µ∗ of the Lagrangian
so that
∂L ∂L ∂L
= 0, = 0, =0
∂x1 ∂x2 ∂µ

É The process would not have worked if both ∂h/∂x1 and ∂h/∂x2 were zero at the point x1∗ x2∗ . This
implies that we need to make the assumption that these partial derivatives are not zero at the constrained
maximizer. In a case where the constraint is linear this additional condition is automatically satisfied.
 CONSTRAINED OPTIMIZATION 241

If we were minimizing f instead of maximizing f on the constraint set C we would

have used the same arguments that we used above, so that the previous remark
holds whether we are maximizing or minimizing f on C. The difference in the
second-order condition we describe below.
The second-order condition is obtained by studying the determinant of a matrix
that is called the bordered Hessian.
Second-order condition for a maximization. Suppose that . x1,∗ x2∗ , µ∗ are a critical
point of the Lagrangian so that
∂L ∂L ∂L
= 0, = 0, =0
∂x1 ∂x2 ∂µ
To check whether x1,∗ x2∗ is a local max of f (x1 , x2 ) on h(x1 , x2 ) one needs to check that
⎧ ∂h ∂h
⎫
⎪
⎪ 0 ∂x ∂x ⎪
⎪
⎨ ∂h ∂2L
1
∂2L
2 ⎬
det ∂x1 ∂x12 ∂x1 ∂x2 >0
⎪
⎪ ⎪
⎪
⎩ ∂h ∂2L ∂2L ⎭
∂x2 ∂x1 ∂x2 ∂x2 2

at x1,∗ x2∗ , µ∗
If the problem is one of minimizing a function, the second-order condition involves
finding a determinant that is negative, so that in the previous condition the determinant
of the 3 by 3 matrix should be negative.
Second-order condition for a minimization. Suppose that . x1,∗ x2∗ , µ∗ are a critical
point of the Lagrangian so that
∂L ∂L ∂L
= 0, = 0, =0
∂x1 ∂x2 ∂µ
To check whether x1,∗ x2∗ is a local min of f (x1 , x2 ) on h(x1 , x2 ) one needs to check that
⎧ ∂h ∂h
⎫
⎪
⎪ 0 ∂x1 ∂x2 ⎪
⎪
⎨ ∂h ∂2L ∂2L
⎬
det ∂x1 ∂x12 ∂x ∂x <0
⎪
⎪
1 2
⎪
⎪
⎩ ∂h ∂2L ∂2L ⎭
∂x2 ∂x1 ∂x2 ∂x2 2

at x1,∗ x2∗ , µ∗
b R E F E R E N C ES

Abowd, J. M., Kramarz, F., Margolis, D. N., and Philippon, T. (2000), The Tale of Two Countries:
Minimum Wage and Employment in France and the United States, IZA working paper.
Acemoglu, D. (1997), ‘Training and Innovation in an Imperfect Labour Market’, Review of
Economic Studies, 64: 445–64.
Acemoglu, Daron, and Pischke, Jorn-Steffen (1998), ‘Why Do Firms Train: Theory and Evidence’,
Quarterly Journal of Economics, 113: 79–199.
—— (1999), ‘The Structure of Wages and Investment in General Training’, Journal of Political
Economy, 107: 539–72.
Adam, Paula, and Canziani, Patrizia (1998), Partial De-regulation: Fixed Term Contracts in Italy
and Spain, Centre for Economic Performance Discussion Paper 306.
Akerlof, G. A., and Katz, L. (1989),‘Workers’ Trust Funds and the Logic of Wage Profiles’, Quarterly
Journal of Economics, 104/3: 525–36.
Alvarez, F., and Veracierto, M. (2001),‘Severance Payments in an Economy with Frictions’, Journal
of Monetary Economics, 47: 477–98.
Autor, David (2001), ‘Why Do Temporary Help Firms Provide Free General Skills Training?’,
Quarterly Journal of Economics, 116/4 (Nov.): 1409–48.
Autor, D., Katz, L., and Krueger, A. (1988), ‘Computing Inequalities: Have Computers Changed
the Labor Market’, Quarterly Journal of Economics, 113: 1169–213.
Baker, George (1992), ‘Incentives Contracts and Performance Measurement’, Journal of Political
Economy, 100: 598–614.
—— (1994),‘The Internal Economics of a Firm: Evidence from Personnel Data’, Quarterly Journal
of Economics, 109/4: 891–919.
—— and Holmstrom, Bengt (1995),‘Internal Labor Markets: Too Many Theories, Too Few Facts’,
AER, 85/2: 255–9.
—— Gibbons, Robert, and Murphy, Kevin (1984), ‘Subjective Performance Measures in Optimal
Incentive Contracts’, Quarterly Journal of Economics, 109/4: 921–55.
—— Jensen, Michael, and Murphy, Kevin (1988), ‘Compensation and Incentive: Practice vs.
Theory’, Journal of Finance, 43: 593–616.
Barron, J. M., Black, D. A., and Loewenstein, M. A. (1987), ‘Employer Size: The Implications
for Search, Training, Capital Investment, Starting Wages, and Wage Growth’, Journal of Labor
Economics, 5/1: 76–89.
—— (1993), ‘Gender Differences in Training, Capital, and Wages’, Journal of Human Resources,
28/2: 343–64.
Barron, J. M., Berger, M. C., and Black, D. A. (1997), ‘How Well do we Measure Training?’ Journal
of Labor Economics, 15/3: 507–28.
Barron, J., Berger, M., and Black, D. (1999), ‘Do Workers Pay for On-the-Job Training?’ Journal
of Human Resources, 34/2: 235–52.
Bartel, A. (1995), ‘Training, Wage Growth, and Job Performance: Evidence from a Company
Database’, Journal of Labor Economics, 13: 401.
 REFERENCES 243

Barton, H. Hamilton, Nickerson, Jack A., and Owan, Hideo (2003), ‘Team Incentives and
Worker Heterogeneity: An Empirical Analysis of the Impact of Teams on Productivity and
Participation’, Journal of Political Economy, 111/3: 465–97.
Becker, G. S. (1962), ‘Investment in Human Capital: A Theoretical Analysis’, Journal of Political
Economy, 70: 9–49.
Becker, Gary (1964), Human Capital, Chicago: University of Chicago Press.
Becker, G. (1993), Human Capital: A Theoretical and Empirical Analysis with Special Reference to
Education, 3rd edn. Chicago: University of Chicago Press.
Bentolila, Samuel, and Bertola, Giuseppe (1990), ‘Firing Costs and Labor Demand: How Bad is
Eurosclerosis?’, Review of Economic Studies, 57: 381–402.
Bertola, Giuseppe (1990), ‘Job Security Employment and Wages’, European Economic Review,
851–66.
—— (2004), ‘A Pure Theory of Job Security Provisions’, Review of Economic Studies.
—— and Rogerson, R. (1997), ‘Institutions and Labor Reallocation’, European Economic Review,
1147–71.
Blanchard, O. (1998), Employment Protection and Unemployment, mimeo, MIT, http://econ.www.
mit.edu/faculty/blanchar/index.htm.
Blanchard, Olivier, and Landier, Augustin (2000), ‘The Perverse Effects of Partial Labor Market
Reform: Fixed Duration Contracts in France’, Economic Journal, 112: F214–F244.
Boeri, T. (1999), ‘Enforcement of Employment Security Regulations, on the Job Search and
Unemployment Duration’, European Economic Review, 43: 65–89.
Boeri, Tito (1999), ‘Enforcement of Employment Security Regulations, On-the-Job Search and
Unemployment Duration’, European Economic Review, 43: 65–89.
Booth, A. (1993), ‘Private Sector Training and Graduate Earnings’, Review of Economics and
Statistics, 75/1: 164–70.
Booth, Alison, Francesconi, Marco, and Frank, Jeff (2002), ‘Temporary Jobs: Stepping Stones or
Dead Ends?’, Economic Journal, 112: F189–F213.
Borjas, G. (2004), Labor Economics, 3rd edn. New York: McGraw-Hill.
Brown, J. N. (1989), ‘Why do Wages Increase with Tenure? On-the-Job Training and Life-Cycle
Wage Growth Observed within Firms’, American Economic Review, 79: 971–91.
Brown, C. (1990), ‘Firms’ Choice of Method of Pay’, Industrial and Labor Relations Review, 43/3:
165–82.
Burda, M. (1992), ‘A Note on Firing Costs and Severance Payments in Equilibrium Unemploy-
ment’, Scandinavian Journal of Economics, 94/3: 479–89.
Calmfors, L., and Hoel, M. (1988), ‘Work Sharing and Overtime’, Scandinavian Journal of
Economics, 90: 45–62.
Cobb, C., and Douglas, P. (1928), ‘A Theory of Production’, American Economic Review, Papers
and Proceedings, 18: 139–65.
Crépon, Bruno, and Kramarz, Francis (2002), ‘Employed 40 Hours or Not-Employed 39: Lessons
from the 1982 Mandatory Reduction of the Workweek’, Journal of Political Economy, 110/6:
1355–89.
Davis, Steven, Haltiwanger, John, and Schuh, Scott (1996), Job Creation and Destruction,
Cambridge, Mass.: MIT Press.
Ehrenberg, R. G. (1971), ‘Heterogeneous Labor, the Internal Labor Market, and the Dynamics of
the Employment–Hours Decision’, Journal of Economic Theory, 3: 85–104.
244 REFERENCES

Ehrenberg, Ron, and Bonanno, Michaeal (1990),‘The Incentive Effects of Tournaments Revisited:
Evidence from PGA Tour’, Industrial Labor Relation Review, 43: 74–89.
Ehrenberg, R., and Smith, R. (2003), Modern Labor Economics: Theory and Public Policy, 8th edn.
New York: Addison Wesley Longman Inc.
Elliot, R. F. (1999), Labor Economics: A Comparative Text, New York: McGraw-Hill.
Galdon-Sanchez, J., and Guell, M. (2000),‘Let’s Go to Court! Firing Costs and Dismissal Conflicts’,
Industrial Relations Sections, Princeton University, working paper no. 444.
Garibaldi, Pietro (1998),‘Job Flow Dynamics and Firing Restrictions’, European Economic Review,
42: 245–75.
—— (2004), ‘Search Unemployment with On the Job Search’, Macroeconomic Dynamics.
—— and Pacelli, Lia (2004), ‘Employment Protection and the Size of Firms’, Giornale degli
economisti.
Garibaldi, P., and Violante, G. (2005),‘The Employment Effects of Severance Payments with Wage
Rigidities’, Economic Journal, November.
Gibbons, Robert (1987), ‘Piece Rate Incentives Scheme’, Journal of Labor Economics, 5: 413–29.
—— (1996), ‘Incentives and Careers in Organization’, in David Kreps and Ken Wallis (eds.),
Advances in Economics and Econometrics: Theory and Applications. Cambridge: Cambridge
University Press, vol. ii, ch. 1.
Gibbons, R. A. (1998), ‘Incentives in Organizations’, Journal of Economic Perspectives, 12: 115–32.
Gibbons, R., and Murphy, K. J. (1992), ‘Optimal Incentive Contracts in the Presence of Career
Concerns’, Journal of Political Economics, 100: 468–506.
Goux, D., Maurin, E., and Pauchet, M. (2001), ‘Fixed Term Contract and the Dynamics of Labor
Demand’, European Economic Review, 45: 533–52.
Grubb, D., and Wells, W. (1993), ‘Employment and Regulations and Patterns of Work in OECD
Countries’, OECD Economic Studies, 7–58.
Guell, M. (2000), Fixed-Term Contracts and Unemployment: An Efficiency Wage Analysis, Industrial
Relations Sections, Princeton University, Working Paper 433.
Hamermesh, D. (1993), Labor Demand, Princeton: Princeton University Press.
Hart, R. (1987), Working Time and Employment, Boston: Allen and Unwin.
Holmstrom, Bengt (1979),‘Moral Hazard and Observability’, Bell Journal of Economics, 10: 75–91.
—— (1982), ‘Moral Hazard in Teams’, Bell Journal of Economics, 13: 324–40.
Hunt, J. (1999), ‘Has Work-Sharing Worked in Germany?’, Quarterly Journal of Economics, Feb.:
117–48.
Hunt, Jennifer (2000), ‘Firing Costs, Employment Fluctuations and Average Employment: An
Examination of Germany’, Economica, 67: 177–202.
Ichino, P. (1996), Il lavoro e il mercato, Milan: Mondadori.
Ichino, Pietro (1996), Lo stato e il mercato. Milan: Mondadori.
Kandel, Eugene, and Lazear, Edward P. (1982),‘Peer Pressure and Partnerships’, Journal of Political
Economy, 100: 801–17.
Knoeber, Charles, and Thruman, Walter (1994), ‘Testing the Theory of Tournaments: An
Empirical Analysis of Broiler Production’, Journal of Labour Economics, 12/2: 155–79.
Kuhn, P. (1986), ‘Wages, Effort, and Incentive Compatability in Life-Cycle Employment
Contracts’, Journal of Labor Economics, 4/1: 28–49.
 REFERENCES 245

Lazear, E. P. (1979), ‘Why is there Mandatory Retirement?’, Journal of Political Economy,

1261–84.
—— (1981),‘Agency, Earnings Profiles, Productivity, and Hours Restrictions’, American Economic
Review, 71/4: 606–20.
—— (1988), ‘Employment at Will, Job Security, and Work Incentives’, in Robert Hart (ed.),
Employment, Unemployment and Labor Utilization, Boston: Unwin Hyman.
—— (1990),‘Job Security Provisions and Employment’, Quarterly Journal of Economics, 699–726.
Lazear, Edward (1989), ‘Salaries and Piece Rates’, Journal of Business, 59/3: 405–31.
Lazear, Edward P. (1995), Personnel Economics, Cambridge: MIT Press.
—— (1998), Personnel Economics for Managers, New York: Wiley.
—— (2000), ‘Performance Pay and Productivity’, American Economic Review, 90: 1346–61.
Leuven, Edwin (2004), The Economics of Training: A Survey of the Literature, Amsterdam:
Department of Economics, University of Amsterdam.
Ljungqvist, L. (2002), ‘How Do Layoff Costs Affect Employment?’, Economic Journal, 112:
829–53.
Lynch, L. (1992),‘Private Sector Training and the Earnings of Young Workers’, American Economic
Review, 82: 299–312.
MacLeod, B., and Malcolmson, J. (1993), ‘Investments, Holdup and the Form of Market
Contracts’, American Economic Review, 83: 811–37.
McLaughlin, K. (1991),‘A Theory of Quits and Layoffs with Efficient Turnover’, Journal of Political
Economy, 99: 1–29.
McConnell, Campbell R., Brue, Stanley L., and Macpherson, David A. (1999), Contemporary
Labor Economics, New York: McGraw-Hill.
Mortensen, D. T., and Pissarides, C. A. (1994), ‘Job Creation and Job Destruction in the Theory
of Unemployment’, Review of Economic Studies, 61: 397–415.
—— (1998), ‘New Developments in Models of Search in the Labor Market’, in O. Ashenfelter and
D. Card (eds.), Handbook of Labor Economics, Amsterdam: North Holland.
OECD (1996), Employment Outlook, Paris: Organization for Economic Development.
—— (1997), Employment Outlook, Paris: Organization for Economic Development.
—— (1998), Employment Outlook, Paris: Organization for Economic Development.
—— (1999), Employment Outlook, Paris: Organization for Economic Development.
—— (2002), Education at a Glance, Paris: Organization for Economic Cooperation and
Development.
—— (2002), Education at a Glance. Paris: Organization for Economic Development.
—— (2003), Employment Outlook, Paris: Organization for Economic Development.
O’Flaherty, B., and Siow, A. (1995), ‘Up-or-Out Rules in the Market for Lawyers’, Journal of
Labour Economics, 13: 709–35.
Oi, W. (1962), ‘Labor as a Quasi-Fixed Factor’, Journal of Political Economy, 70/6: 538–55.
Parsons, D. (1972), ‘Specific Human Capital: An Application to Quit Rates and Layoff Rates’,
Journal of Political Economy, 80/6: 1120–43.
Pissarides, C. A. (2000), Equilibrium Unemployment Theory, Cambridege, Mass.: MIT Press.
Prendergast, C. A. (1999), ‘The Provision of Incentives in Firms’, Journal of Economic Literature,
37: 7–63.
246 REFERENCES

Rosen, S. (1969), ‘Short-Run Employment Variation on Class-I Railroads in the U.S., 1947–1963’,
Econometrica, 36: 511–29.
Schivardi, Fabiano (2000), ‘Labor Market Rigidities, Unemployment and Growth’, Giornale degli
economisti e annali di economia, 115–41.
Shapiro, Carl, and Stiglitz, Joseph (1984), ‘Equilibrium Unemployment as a Discipline Device’,
American Economic Review, 74: 434–44.
Spence, Michael (1973), ‘Job Market Signaling’, Quarterly Journal of Economics, 87: 355–74.
Spence, A. M. (1974), Market Signaling: Informational Transfer in Hiring and Related Screening
Processes, Cambridge, Mass.: Harvard University Press.
Trejo, S. J. (1991), ‘The Effects of Overtime Pay Regulation on Worker Compensation’, American
Economic Review, 81/4: 719–40.
Wassmer, Etienne (1999), ‘Competition for Jobs in a Growing Economy and the Emergence of
Dualism in Employment’, Economic Journal, 109/457: 349–71.
Yashiv, Eran (2000), ‘The Determinants of Equilibrium Unemployment’, American Economic
Review, 90/5: 1297–322.
b INDEX

ability 68, 72, 74, 128, 131, 141, 225 negative relationship between fixed
average 80, 119, 231 payment and 93
complementarity between training and optimal 95–6, 102–4, 137
186 option for particularly productive
education useful for signalling 78 individuals 115
heterogeneous 108–10, 111–15, 129 preference for 120
high/low 63, 75–81, 110, 112, 117–20, reduction in 100
186, 223, 226, 227, 229, 230, 231 two types of effects associated with 119
private information about 186 bordered Hessian matrix 241
screening of 185 borrowing 98, 156
technical 229, 230 bottlenecks 224
tendency to rely on other people’s 221 broiler production 140–6
unequal 139 budget allocation 31, 34
workers know their own 62, 67 budget constraint 43, 90
accountancy 121 buffer stock model 45, 46, 47
advance notice 8, 203 formal derivation of 49–51, 60–1
adverse selection 62–81, 225, 229 interim costs 53–5
agencies 46, 53 temporary contracts 52
interim 52, 60–1 business conditions:
apprenticeship programmes 167, 171–2 fluctuations 2, 45, 46, 50, 51
asymmetric information 62, 67, 148, 149 good/bad 45, 46, 47, 48, 49, 52, 53, 54,
job-preserving wage with 192–3 55, 187, 213, 214, 215, 217
Autor, David 181, 183

California 223
backward induction 161 capital 232
bad candidates/workers 63, 64, 67, 70, 72, constant 236
186 costly, independent labour interacting
bad times, see business conditions with 22–4
bank credit 98 stock of 12, 13, 18, 233, 234
bargaining power 163, 177, 211 see also human capital
baseline (Becker) model 100, 153, 167, capital markets 179
171, 172–80 cheating 147
facts that run counter to 182 chief executives 140
Belgium 6 classroom-based training 172, 182
Bentolila, S. 212 clerical and secretarial jobs 172, 181
Bertola, G. 212 Cobb-Douglas Production
BLS (US Bureau of Labor Statistics) 181 Function 18
blue collars 15, 204 coefficients 40–1, 124, 127, 145, 146,
Bonanno, M. 140 184, 230
bonus schemes 6, 85, 86, 99, 101, 105, 108, collaborative skills 225, 227, 229, 231
116, 117, 132, 139, 221 collective agreements 1, 2, 8, 34, 37, 190,
effort under 89–91 192, 203
extreme 107, 128 wage set by 188
individuals better off in 118 collective bargaining 7
248 INDEX

commission rate 83, 107, 172 separation 189

optimal 96, 99, 103 settlement 141, 142, 143, 145
compensation 7, 64, 72, 132–51 shared 164, 165
consistent with minimum wage 115 social 77 n., 81
individual utility depends on 220 splitting benefits and 165, 166
linear 221 subjective 154
maximum 204 n. sunk 177
monetary 203 team 222
optimal 82–106, 107, 135, 137 trial 203
premium for loss of income 69 turnover 54, 55–8
simple revenue-sharing scheme 219 see also firing costs; labour costs; training
see also bonus schemes; fixed costs
compensation; franchising; piece courts 7, 8, 204, 205
rates settlement out of 206, 207
competition 4, 78, 183 credentials 62, 63, 63, 74, 76, 77
complementarities 219 credit 177, 178
computing technology 181 cumulative distribution function 134
confirmation process 69, 70 customer satisfaction 127
conflict of interest 83 Czech Republic 6
constant-output curves 15
constrained firms 215–17
constrained optimization 238–41 deadweight losses 202
decision rules 57
contracts 76, 78, 79, 85, 142
deferred compensation 149–51
baseline 100
degrees 63, 75, 155, 167
bonus 89–91, 101, 105
demand 83, 84
contingent 62, 63–74
exceptional increase in 124
implicit 133, 193
new and rapidly shifting expertise 181
linear 85 n.
overall state of 192–3
LPE 141
totally unrelated to effort 82
optimal 97, 98, 100, 105, 107, 137
se also labour demand
separating 80–1
density function 134
short-term 172
cumulative 155 n.
see also permanent contracts; temporary discount 189
contracts; termination of contracts discount factor 154
convexity assumption 17, 31 discount rate 196
corner solution 19–20, 89 dismissals 7, 8, 9, 206
cost-benefit analysis 223 legality of 204
cost function 76 norms and procedures to be
costs 105 followed 203
advertising 27, 28 unfair 205
agency 46, 53 see also firing; separation
changing jobs 173 distribution function 213
education 76, 79, 81, 153, 154 disutility 87, 88, 135, 219, 229
expected 87, 93 downsizing 187, 191, 193
hiring 52, 53, 55
legal 204 n., 205, 206
marginal 167, 178, 219 econometrics 37, 126–7
monetary 155 economies of scale 167
opportunity 155 education 79, 156, 166
procedural 203 costs and benefits 153, 154, 155
psychological 154, 155 different degrees of 21
relative 11, 14 human capital approach to 152
screening 28 lifetime 152
 INDEX 249

linking wage offered to level of 63 symmetrical 135

low attachment 59 training and 160, 172, 175
more costly for workers of lower ability utility in 100
81 see also Nash equilibrium
signalling value of 74–8 expansions 51
unproductive 80 expected earnings 21
utility with(out) 155 expected profits 51, 80, 93, 95–6, 103, 197
efficiency 100, 129, 220, 224 constrained firm 215
maximized 107 cumulative 209
efficiency effects 108 firing tax reduces 212
efficiency wage 133, 146–9, 150 larger than firing costs 198
efficient allocation 213–14, 215, 217 maximized 83, 87, 92, 136, 214
efficient opportunities 164 positive 97, 199
potential 173 random 209, 211
effort 72, 85, 93, 97, 99, 108, 112, 123, 150 reduced 99
average 119 severance payment has no impact on
cost of 135, 149 210
demand totally unrelated to 82 zero 199
disutility of 219 extensive margin 27, 30
efficient 219, 221
efficient level of 111, 137
impact on 100 factors of production 11, 232
incentive for 107, 140 heterogeneous 12, 29
lower and lower at the margin 113 interdependent 15
marginal benefit/cost of 137, 219, 220 quasi-fixed 28
marginal (dis)utility of 87, 88, 135 financial constraints 156
need to exercise 118 firing 7, 51, 57, 69, 147
noisy measurements of 136 advance notice before 8
observable 147 impossible 214
output depends on 134 legitimate 206
proportional to spread in prize 132 no cost 47, 49, 213–14
salaries used to increase 133 regulations 8, 9
tendency by workers to reduce 136 role of judges on disputes 9
tendency to rely on other people’s 221 ruled unfair 206
test of constancy of 143 firing costs 55, 57, 188, 193, 194, 196, 197,
utility depends negatively on 109 205, 212
very difficult to elicit 138 effect of change in 199–200
worker dislike of 84, 86, 87, 148 ex ante 204, 206
zero 109, 110 expected profits larger than 198
see also optimal effort future 199
egalitarian wage policy 21, 113 impressively high 206
Ehrenberg, R. 140 statutory 204
employers’ contributions 29 firing taxes 193–5, 207
EPL (employment protection legislation) deriving wages with 210–12
1, 2, 7–10, 51, 202–17 firm-initiated separation 187, 188–93, 204
cost associated with 28 monetary transfer paid in case of 203
strict(est) 50, 204, 214 firm-specific training 156, 160–6, 167,
stringency of 45, 187, 202 169–70
equilibrium 18, 20, 35, 36, 112, 158, 166 first-order conditions 25, 33, 42, 60, 61, 91,
constrained firms exist in 216 96, 103, 116, 137, 178, 220, 236
pooling 77, 79–80 identical 135
separating 75, 76, 77, 80–1, 186 fixed compensation/payment schemes
signalling 63, 76 85–6, 93, 94, 98, 108, 128
250 INDEX

fixed compensation/payment schemes graduation 63

(Cont.) Guell, M. 207
increase in 100
negative 107
see also fixed wages handicapping system 138, 140, 141, 145–6
fixed costs 30, 31, 33, 193 Hawthorne effect 124, 125–6, 127
effect of an increase in 32 health insurance contributions 205
non-recurring 27–9 heterogeneity 12, 29, 108–10, 111–15, 129,
recurring 27, 28, 29 132, 144, 155
fixed wages 85, 107, 117, 118, 120, 212, 215 advantages of 222
exogenous 213 options which critically depend on 121
minimum output with 111–16, 130 productivity and 230, 231
optimal scheme with 129–31 risky strategies and 138–9
permanent contracts with 46–51 substantial, across teams 227
flexible firm 45, 46, 49–50 ‘hierarchies of the hierarchies’ method 9,
average employment 47, 51 10
volatility of employment 48, 51 hiring 32, 37, 51, 92, 127, 166, 191
formal derivations 49–51, 60–1, 94 business conditions turn good 214
France 21, 27, 29, 59 cost of 11, 22, 27, 28, 52
union density 6, 7 decision not affected by presence of
workweek reduction 36–7 severance payments 210
franchising schemes 85, 86, 96, 103, 105 firm optimally holds on to current losses
basic, with ability parameter 128 from 199
effort maximum under 91 general view on personnel policies 12
firm cannot implement 129 important effects on strategy 46
interim workers 53
optimal 98, 128
less labour in good times 53
profits 111
no cost 47, 49, 213–14
very difficult to implement 107
one of the key issues 45
free rider problem 220–1, 222, 225, 226,
skilled workers 14, 19, 20, 30
227, 231
very important insights on decisions
fringe benefits 29
197
future wages 63, 68, 133, 153, 159, 160
see also recruiting
agreed reduction in 187
hold-up problem 161, 182
holiday pay 29
homogeneous workforce 121, 129, 130
Galdon-Sanchez, J. 207 hours-employment trade-off 11, 27–44
game theory 78, 79 solving 12
see also tournaments human capital investment 152–70, 176,
Garibaldi, P. 9, 204 222
garment industry 67, 218
see also Koret
gender differences 167 ICC (incentive compatibility constraint)
general training 153, 156–60, 167–9 90, 91, 101, 136, 137
firm-sponsored 171, 172–6, 178, 179, idiosyncratic values 215, 216
180, 181–6 Illinois 125
Germany 7, 9, 21 imperfect information 62, 82, 133
apprenticeship training 171–2 imperfect labour markets 1, 6, 56, 63, 152
Gibbons, R. 140 firm-initiated separations and wage cuts
good candidates 63, 64, 67, 115, 186 188–93
credentials no link to probability of firm-worker match and 2–5
finding 77 important institutional feature of 11
good times, see business job termination 193
conditions key feature of 21, 28
 INDEX 251

labour hoarding a common job creation 195–9

phenomenon 188 job destruction 187–201, 202, 211
training investment in 171–86 effect of severance payment on 208
imperfect substitutes 15 termination policies 1
incentive effect 115, 119, 123, 126, 143 job loss probability 38–40
pure 124–5, 127 job opportunities 52
incentives 82, 86, 97–8, 107, 132, 140, 145, job security 212
146, 171 just cause 204
extreme 83 justified objective/subjective motives 204
group 218–31
investment/training 153, 158, 160,
161, 162 Kaitz index 6
marginal 143 Knoeber, C. 140
worker skills 178 knowledge transfer 222
see also ICC Koret 222–31
income 65, 86, 110, 147, 223
additional 108 labour costs 2, 12, 33, 41, 87, 124
average 84 formal classification of 27–9
compensation premium for loss of 69 hourly 13
expected 87, 91–2 marginal 53
fluctuations in 100 relative 13
lifetime 154 total 11, 29–32
marginal utility of 87, 88 variable 28, 29, 30
stable 84 see also fixed costs
time invariant streams 83 labour demand:
utility depends positively on 109 dynamic, under uncertainty 212
variability of 100, 101 firm level 232–7
see also salary; wages long-run 236–7
indexation schemes 6 multi-period 200
indifference curves 87–9, 90, 109–10, 112, optimal 45
117, 238 simple and static model of 47, 187, 194
information acquisition 186 stochastic shifter of 213
informed parties 62 toy model of 212–13
inherent flexibility 53 labour hoarding 54, 188
insider phase 207–10 firing taxes and 193–5
insurance 82, 100 job creation, job destruction
integrators 141–3, 146 and 195–9
intensive margin 27, 30 labour markets 46, 98, 107
interest rates 58, 154, 155 changes in position 37
zero 159, 166 EPL in 202
interior solution 15–19, 25 internal 193
internship 69–70 one of the most important institutions
investment, see education; human capital; of 203
training one of the most important institutions
isocosts 13–14, 17–18, 20, 33, 34, 35, 36 of 7
shift induced in 38 separating equilibrium 75
total labour costs and 29–32 temporary contracts an important
isoquants 15–16, 17, 18, 20, 31, 34 feature of 58
Italy 7, 8–9, 204, 206, 207 upward-sloping wage profiles relevant in
133
youth workers entering 59
job advertising 74, 218 see also imperfect labour markets;
costs 27, 28 perfect labour markets
252 INDEX

labour product 147, 149, 209, 211 maximization decision 12, 60

random value 207 maximization problem 134–5, 220, 240
realization is known to both parties 207 joint 206
uncertainty over value of 195 Mayo, Elton 125
labour supply 184 microeconomics 15, 238
Lagrangian technique 25, 33, 42, 240 minimum guarantee 118, 121
Latin American countries 9 minimum output 117, 120, 129–30, 131
law firms 67 fixed wage with 111–16, 130
law of diminishing returns 233, 235 increase leads to undesirable turnover
lay-offs: 121
administrative procedures before 203 minimum wage with 115
choosing which workers should be observed and imposed on employment
targeted for 187 contract 107–8
costs of 8–9 obtained without any effort 118
employer-initiated 204 minimum payment provision 145, 146
legitimacy of 7 minimum wages 1, 148
not allowed 51 constraints and union density 5–7
ruled fair 206 contingent contracts with 72–4
ruled unfair 207 guaranteed 108, 115, 119
Lazear, E. P. vii–viii impact of 180
leisure 2 more training for low-skill workers 181
LPE (Linear Performance Evaluation) pay level close to 167
132–3, 139–46 setting 6
luck 84, 85, 100, 102 unskilled 70–1
output depends on 134, 136 misconduct 204
Mitterrand, François 37, 38
monetary transfer 8, 193 n., 203
mandatory payments 8, 203 monitoring technology 73, 147
manufacturing industries 37 imperfect 70–2
traditional jobs 67 perfect 68–70
marginal benefit 137, 155, 176, 178, 179, monopolistic firm 212
219, 220 monopsonistic firm 161
marginal product of labour 3, 15 n., 16, 18, moral hazard 82, 83, 226
26, 46, 47, 53, 157, 158, 159, 161, morale 7
162, 163, 177, 232–6 motivation 132
accepting a wage below, during MP (modular production) 223, 224,
training 182 227, 231
full, in outside labour market 178 MSAs (Metropolitan Statistical Areas) 183,
larger than cost of training 175 184–5
larger than wage 188 multi-skill abilities 223
outside option lower than 173 multivariate analysis 224 n.
positive 17 Murphy, K. J. 140
unskilled 13 mutual learning 218, 222, 227, 230, 231
value is akin to random draw 195
wages fully reflect 153
marginal productivity 5, 17, 75, 173, Nash bargaining 177, 206
188, 194 natural experiment 39
percentage increase in 19 new technology 24
relative 13 non-pecuniary benefits 229
specific rent proportional to 174 non-linearity 31
wage falls short of value of 187 non-negative payments 98–100
marginal rate of technical substitution norms 203
16, 17 team 221–2
 INDEX 253

OCS (Occupational Compensation Survey better 109

of Temporary Help Supply cumulative 70
Services) 183 n. different 68
OECD countries 6, 9, 58, 204 n. exactly equal to minimum wage 73
Ohio 121 firing tax and 193
older workers 149–50, 151 good individuals and 115
on-the-job training 152–86 larger ability and 112
one-off costs 28 lower than marginal product 173
opportunity cost 155 normalized to zero 100
optimal effort 89–91, 92, 98, 104, 128, 136, present discounted value of 191
137, 138, 220 skilled/unskilled labour and 68, 69
output associated with 114 specified 174
risk aversion and 101–2 surplus and 5, 162, 163
optimal skill ratio 11–26 training increases 160
output 2, 78, 81, 85, 86, 93, 122, 128, 176, utility and 91, 92, 93, 97, 99, 100, 102,
219, 221 104, 105, 113, 115, 130, 136, 148–9
ability and 109, 110 wages higher than 4, 161, 171, 188
choice of hours independent of 32 outsider phase 207–10
compensation independent of 107 overeducation phenomenon 22
contribution to 13 overemployment 53–4
cost per unit of 22–3, 24 overtime 6, 37, 38
decrease in 16 normal wage per hour and 28
depends on effort and luck 134, 136 overtime premium 34–6, 43–4
easily identified 67
expected 84, 134, 135, 136
firm can easily observe 129 paid vacations 27
firm decides how much to produce 12 participation constraint 4, 81, 92–3, 97,
gain in 16 136–7
homogeneous 213 binding 107, 111, 113, 130, 131
individual, variance of 124 presence of risk aversion changes 102
input and 14, 232 PBS (progressive bundling system) 223,
isoquants and higher levels/larger 224, 227, 231
quantity of 15, 17 PDV (present discounted values) 56, 162,
losses of 193 188 n., 189, 191
luck component of 100, 102 and multi-period jobs 200–1
marginal increase in 42 penalties 139, 148
maximum 109 court 205, 206
measuring 67, 127 perfect information 193
pay based strictly on 63, 64 perfect labour markets 1, 3
profit per unit of 87 training in 152, 153, 157, 159, 175, 176,
random components of 84 178
sharing 220 performance 63, 146
skilled labour must be consistent with absolute 132, 133
30 average 132, 139
team impact on 222, 227 credential must be correlated with 74
wage income completely independent of good 85, 193
85 impact of worker heterogeneity on 108
zero 110, 116 mean 143, 144
see also minimum output output 223
output constraint 31 pay for 83, 107–31
outside options 2–3, 87, 94, 147, 166, 169, relative 133, 139, 140, 141, 142, 145
177, 187, 200, 208, 209, 211 reward for 140, 145
best 108 riskiness of 141
254 INDEX

performance (Cont.) production functions 14, 15, 25–6, 33, 41,

worker paid independently of 82 43–4, 46, 232
worsening 143 convex 213
see also LPE linear 219
permanent contracts 45, 52, 53, 57, 59 partial derivative of 13
firing indicators for workers under 9 short-run 24
fixed wages and 46–51 see also Cobb-Douglas
large wages as 69 productivity 15, 46, 49, 64, 119, 121, 124,
reduced incentive to offer 58 125, 152, 154, 156, 158, 161, 163,
trade-off between temporary and 55 209, 211
two-period/long-term 56 average 123, 126
personnel economics: bad 57
constrained optimization and 238 better 141
great challenge to 218 business fluctuations in 50, 51
heart of 82 candidate, signal of 63
key concern for 187 constant 133
key policies at hiring stage 12 cost and 11
non-competitive labour markets and credential can profitably be used to
1–10 signal 74
training a key decision in 152 effect of piece rates on 127
piece rates 62–3, 64–7, 117, 120, 132, 141, expected 79
223 firm is able to assess 67
effect on productivity 127 fixed 56
group 224, 231 heterogeneity and 230, 231
pilots 167 high 50, 51, 56, 57
poaching threat 182 higher wages increase 146–7
poor workers, see bad candidates/workers imperfect information about 62
PPP (Performance Pay Plan) 121, 123, 124, independent 19
126, 127 loss of 159, 166
preferences 109 low 50, 51, 56
agent 100–1 negative 56
principal-agent problem 83–98, 100, present discounted value of 201
102–6 random 56
private information 186 relative 11
prizes 132, 133, 140 reservation 194, 196, 197, 198
average 136 scheme that is likely to improve 108
probationary period 62, 63, 64, 67–72, 73 skilled workers 66
requires negative probability 74 trade unions and 7
production 2, 56, 108, 125, 157, 161, 163, wage growth and 167
173, 177, 196, 204, 207, 208 see also marginal productivity; team
basic activities 67 productivity
efficiency maximized 107 productivity premium 19, 20, 22, 24
expected value of 51 productivity profile 200
flexible 223 productivity shocks 52, 56, 142
forgone 161 professional jobs 67
independent 19, 22 profit maximization 12, 47, 49–50, 51, 75,
interdependent 15–19, 25–6 94, 96, 103, 177, 234
interior 25 expected profits 83, 87, 92, 136, 214
minimum amount required to continue firm optimal employment behaviour
to be employed 121 obtained by 214
team 67, 218–31 first-order condition for 236
tournaments in 141–2 good compensation packages must be
see also factors of production consistent with 82
 INDEX 255

long-run 236 rewards 125, 221

optimal compensation scheme for 135 incremental 139, 140, 143, 144
profitability 18, 45 performance 142, 145
profits 3, 48, 57, 105–6, 114, 128, 148, 161, top managers 140
171, 178, 179, 180, 192–3 rigid firm 45, 46, 47, 50, 214–15
average 49, 214, 215, 216 average employment 48, 51
cumulative 210 average profits 49
damage to outlook 18 risk aversion 83, 85, 98, 100–6, 132
discounted 197 risk neutrality 86, 87, 100, 105, 107, 108,
efficient level of 111, 129, 130 135, 137, 176, 185
firing tax has impact on 212 agent problem 89–92, 99, 101
good and bad times 213 principal problem, with agent 92–8
marginal 2, 188, 196 risky strategies 138–9, 140, 144
negative 147
non-negative 5
operational 197, 198, 199, 209, 211 Safelite 121–6
positive 4, 92, 96, 99, 111 salary 83, 92, 103, 208
present discounted value of 162, 201, based on hourly rate 121
212 constant 195
risk of forgoing 58 corporate executives 140
shared 163 equal to average productivity 79
some must be given away 108 fixed 82
variable 87, 94, 98 good 85
zero 78, 80, 98, 159 graduate workers 13
see also expected profits increase associated with promotion 132
promotion 63, 69 structure obtained by sum of two
increase in salary associated with 132 components 139
size of rise associated with 133 substantial 172
wrong 68, 70, 73 trying to obtain largest possible 158
unskilled workers 13
sales jobs 67
quality control 127 Schivardi, F. 212
quality evaluation 223 second-order conditions 236, 241
securities brokers 167
self-selection 112
rat race problem 72 constraints 81
recessions 51 important mechanisms under team
recruiting 62–81 formation 227
regression analysis 40, 124, 126, 127, 144, inducing 185, 186
145, 146, 227, 228–9 separation:
quantile 226 n. consensual 187, 189
relative compensation 132, 133 efficient 191
indirect evidence of 140 fair 204
rents 4, 99, 119 illegitimate 204
firm-specific 153 inefficient 190
jobs associated to 1 legitimate, definition of 204 n.
pure 108, 111–15, 118, 149 non-consensual 187, 189
sharing 163–6, 173, 207 obvious candidates for 191
specific 162–3, 171, 174 unfair 204, 206
surplus 162–3 see also firm-initiated separation;
restructuring 191 worker-initiated separation
large-scale firm 8 set-up costs 28
revenue-splitting games 221 settlement costs 141, 142, 143, 145
256 INDEX

severance payments 8, 9, 203, 207, 211 negative 187, 189, 190, 191, 192
choice between full reinstatement and positive 189, 190, 191, 192, 201
204 n., 205 specific 174
deriving wages with 208–10 worker gets fraction of total 207, 208,
mandatory 205 209, 211
shifter parameter 213 zero 194
shirking 138, 147, 148, 149, 150, 151 synthetic indicators 9
shocks:
idiosyncratic 212, 213 tangency condition 17, 31, 90
productivity 52, 56, 142 taxes and transfers 203–12
sickness pay 29 see also firing taxes
signals 63 team production problem 218, 219–20
educational 74–8 team norms as remedies to 221–2
imperfect 84 team productivity 218, 222, 223, 224,
skill acquisition 75, 153 225–7, 229
skill composition 11, 12 ‘stars’ influential in raising 231
skill premium 156, 180 team composition and 230
productivity 20 technological considerations 11
see also skilled wage premium temporary contracts 45, 46, 49, 53, 60
skilled labour 12–13, 64, 65, 70 as buffer stock 52
amount hired must be consistent with costly turnover and 55–8
output 30 identifying unskilled workers at the end
attracting 67, 68, 70, 73 of 70
independent 19–20, 22–4 probation period/wage and 62, 67–72
interdependence of unskilled and 15–19 women more likely to hold 59
outside options 68, 69 temporary help industry, see THS
wage compression 22 tenure 63, 67–8, 126, 127, 160
skilled wage premium 14, 17, 18, 19, 22, 25 average 204
larger than productivity skill premium wages increase with 133
20 termination of contracts 45, 46, 52, 56
smaller firms 37, 202, 204 n. early 203
social security 6, 205 tertiary education 21
socialization 223 threat point 177
sorting effect 108, 115, 119, 120, 123, 124, threshold effects 212–17
127, 141 Thruman, W. 140
pure 126 THS (temporary help service) industry
Spain 6, 9 172
specialization gains 222, 233 firm-sponsored training 181–6
specific training 153 tournaments 132–46
standard deviation 124, 144 trade unions 192, 223
statutory payments 203 density 6–7
stay-small policy 215, 216 desire to have egalitarian wage policy 21
students 69 training 152–86
subsidiary decisions 12 training costs 158, 161, 167, 174, 175
substitution 208, 209, 210, 211 already sunk 177
sunk costs 177 difference between training benefits and
surplus 2–3, 4–5, 113, 114, 148–9, 162, 175, 168
188, 194, 210, 218, 223 financing by taking a wage cut 159, 180
divided by Nash bargaining 177 firm pays all 162, 171, 172, 179
firm-specific 163, 173 marginal 178
joint 166 opportunity to offset 182–3
joint value of 105 sharing 165, 166, 170
marginal 104 total 157
 INDEX 257

transfers 203–12 wage compression 180–1

transition economies 9 definition of 178
turnover 181 effects of 21–2
costs of 46, 54, 55–8 importance for firm-sponsored training
feasible to reduce by introducing team 178
production 231 interdependent production with 25
high 52 key condition for
undesirable 121 general/firm-sponsored training
175–6
marginal cost of training with and
unemployment 210 without 178–9
uninformed parties 62 skill and 11, 171
United States 6–7, 9, 21, 121–6 wage constraints 107–31
team production 222–31 wage cuts 158, 160, 220
THS employment 172, 181–6 effects on remaining workers 187
units-per-worker-per-day 123–4 incentive to take 176
unpaid hours 182 job-preserving 187, 189, 190, 192–3
unskilled labour 11, 12–13, 59, 64, 65 overall 193
difficult to detect 72 preventing 180, 192
firm required to be perfectly able to profitable 187, 188, 190, 191
identify 70 training costs financed by 159, 180
independent 19–20, 22–4 willingness to accept 190
interdependence of skilled and 15–19 wage determination 1–2, 161, 162
keeping out 68, 69, 70, 72, 73 explicitly considered 202
marginal product of 13 wage differentials:
optimal to hire only 20 pre-post probation 69, 71
outside options 68, 69 THS establishments 184–5
wage compression 22
wage flexibility 1, 62, 63, 67, 212
up-front payments 98
severance payment with 210
‘up or out’ rule 67
wage functions 179
utility 79, 80–1, 104, 110, 112, 113, 117,
education level 78, 79
120, 135
associated with performance plan 118 wage offers 158, 159, 162, 173
compensation, effort and 147 accepted 161, 163
depends on ability level 118 key reason to avoid matching 193
expected 89, 91, 102, 136 take it or leave it 157
lifetime 149, 150, 154 wage policy 21, 113
marginal 87, 88 wage premia 76, 77, 79, 80
maximized 89, 90, 93, 107, 116, 178, 220 firms willing/not willing to pay 63, 75
outside options and 91, 92, 93, 97, 99, see also skilled wage premium
100, 102, 104, 105, 113, 115, 130, wage profiles 133, 159, 160, 167
136, 148–9 EPL may have a large impact on 202
reservation 98, 99 rising 193
strictly positive 99 wage setting 1, 4, 136–7, 149
with(out) education 155 by collective agreements 188
utility functions 86, 87, 88, 89, 91, 101, unilateral 147
109, 114, 147, 149, 154 wages 13, 56, 110
mean-variance 100 alternative 68
average 79, 80
base 96–7
Violante, G. 9, 204 confirmation 69
vocational training 156 cumulative 68, 69, 70
volatility of employment 48, 51 deferred 151
258 INDEX

wages (Cont.) probationary 63, 67–72, 74

deriving with firing tax 210–12 promotion 63, 69, 70, 73
equilibrium 1, 177 relation between training and 166
exogenous 5 relative 179
expected 70, 89, 100, 135 rigid 2
forgone 159, 204 skilled 71, 154, 155
guaranteed 121–2 temporary 68, 69, 183–5
hourly 121, 127, 183, 185 tenured 67–8
identically zero 186 two-tier 115–20, 121, 122
initial 67, 71, 72 unskilled 71, 154
insider 208–9, 210, 211, 212 variable 85, 86, 102
interim 54 see also efficiency wage; fixed wages;
large 36, 67, 68, 69, 70 future wages; minimum wages
marginal 55 Western Electric Company 125
marginal product lower than 53 women 59
negotiated through rent sharing 207 training 167
no competition to push up 161 worker-initiated separation 190, 192, 193
obtaining 137–8, 173 not consensual 191
optimal 129, 130, 148 works councils 171, 192
outside 64, 66, 68–9, 71, 176, 178, 207, workweek reduction 34–7, 38, 43–4
209–10, 212
overtime 29, 34
period 189
piece rate 63, 64–5, 67, 127 youth/young workers 59, 67, 149–50, 151,
post-probationary 68, 71 200
presented discounted value of 201 earnings 159, 160, 167