ADVANCED
MACKOECONOMICSThis book was set in Lucida Bright by Publication Services, Inc.
The editor was Lucille Sutton;
the production supervisor was Friederich W. Schulte.
The cover was designed by Delgado Des
the cover illustration was drawn by Shane Kelley
Project supervision was done by Publication Services. inc
R. R. Donnelley & Sons Company was printer and binder.
McGraw-Hill
A Division of The McGraw-Hill Compames
ADVANCED MACROECONOMICS
Copyright © 1996 by The McGraw-Hill Companies, Inc. All rights reserved. Printed in the
United States of America. Except as permitted under the United States Copyright Act of
1976, no part of this publication may be reproduced or distributed in any form or by any
means, or stored in a data base or retrieval system, without the prior written permission
of the publisher.
This book is printed on acid-free paper.
7890DOC DOC 90987
ISBN 0-07-053667-8
Library of Congress Cataloginj
Romer, David.
Advanced macroeconomics / David Romer.
p. cm. — (McGraw-Hill advanced series in economics)
Includes bibliographical references and index.
ISBN 0-07-053667-8
1. Macroeconomics. 1. Title. Il. Series.
HBI72.5.R66 1996
339—de20 95-3728ABOUT THE AUTHOR
David Romer is professor of economics at the University of California,
Berkeley. He received his A.B. from Princeton University, where he was vale-
dictorian, and his Ph.D, from M.LT. He has been on the faculty at Princeton
and has been a visiting faculty member at M.LT. and Stanford. He is also a
Research Associate of the National Bureau of Economic Research and serves
on the editorial boards of several economics journals. His main research in-
terests are monetary policy, the foundations of price stickiness, empirical
evidence on economic growth, and asset-price volatility. He is married to
Christina Romer, who is also an economist, and has two children, Katherine
and Paul.CONTENTS
Acknowledgments
Introduction
Chapter 1
THE SOLOW GROWTH MODEL
Theories of Economic Growth
Assumptions
The Dynamics of the Model
The Impact of a Change in the Saving Rate
Quantitative Implications
‘The Solow Model and the Central Questions of Growth
Theory
Empirical Applications
Problems
Chapter 2 BEHIND THE SOLOW MODEL:
INFINITE-HORIZON AND
OVERLAPPING-GENERATIONS
MODELS
Part A THE RAMSEY-CASS-KOOPMANS MODEL
ea)
—
eS
24
=
2.6
=
=
=
Assumptions
The Behavior of Households and Firms
The Dynamics of the Economy
Welfare
‘The Balanced Growth Path
The Effects of a Fall in the Discount Rate
The Effects of Government Purchases
Bond and Tax Finance
‘The Ricardian Equivalence Debate
xix
uw
w
12
15
18
23
26
34
38
39
se
40
46
aE
ape
=
64
66xii ~=CONTENTS
Part B_ THE DIAMOND MODEL 72
2.10 Assumptions 72
2.11 Household Behavior 73
2.12 The Dynamucs of the Economy 75
2.13 The Possibility of Dynamic Inefficiency 81
2.14 Government in the Diamond Model 85
Problems 88
Chapter 3 BEYOND THE SOLOW MODEL: NEW
GROWTH THEORY 95
Part A RESEARCH AND DEVELOPMENT MODELS 96
3.1 Framework and Assumptions 96
3.2 The Model without Capital 98
3.3. The General Case 104
3.4. The Nature of Knowledge and the Determinants of
the Allocation of Resources to R&D il
3.5 Endogenous Saving in Models of Knowledge
Accumulation: An Example 118
3.6 Models of Knowledge Accumulation and the Central
Questions of Growth Theory 121
3.7 Empirical Application: Population Growth and
Technological Change since 1 Million a. 122
Part B_ HUMAN CAPITAL 126
3.8 Introduction 126
3.9 A Model of Human Capital and Growth 128
3.10 Implications 132
3.11 Empirical Application: Physical and Human Capital
Accumulation and Cross-Country Differences in
Incomes 137,
Problems 140
Chapter 4 REAL-BUSINESS-CYCLE THEORY 146
4.1 Introduction: Some Facts about Economic Fluctuations 146
4.2 Theories of Fluctuations 150
4.3. A Baseline Real-Business-Cycle Model 152
4.4 Household Behavior 154
4.5 A Special Case of the Model 158
4.6 — Solving the Model in the General Case 16447
48
aes
4.10
CONTENTS
Implications
Empirical Application: The Persistence of Output
Fluctuations
Additional Empirical Applications
Extensions and Limitations
Problems
Chapter 5 TRADITIONAL KEYNESIAN THEORIES
5.1
a
El
5.4
=
5.6
OF FLUCTUATIONS
Introduction
Review of the Textbook Keynesian Model of Aggregate
Demand
The Open Economy
Alternative Assumptions about Wage and Price Rigidity
Output-Inflation Tradeoffs
Empirical Application: Money and Output
Problems:
Chapter 6 MICROECONOMIC FOUNDATIONS
Part A
6.1
6.2
63
64
Part B
6.5
66
67
68
69
Part C
6.10
6.11
6.12
6.13
OF INCOMPLETE NOMINAL
ADJUSTMENT
‘THE LUCAS IMPERFECT-INFORMATION MODEL
Overview
The Case of Perfect Information
The Case of Imperfect Information
Implications and Limitations
STAGGERED PRICE ADJUSTMENT
Introduction
A Model of Imperfect Competition and Price-Setting
Predetermined Prices
Fixed Prices
The Caplin-Spulber Model
NEW KEYNESIAN ECONOMICS,
Overview
Are Small Frictions Enough?
The Need for Real Rigidity
Empirical Applications
xiii
168
175
180
183
190
195
195
197
205
214
222
232
236
241
242
242
243
246
250
256
256
caTE
262
265
7
276
276
278
ea
eaxiv CONTENTS
6.14 Coordination-Failure Models and Real Non-Walrasian
Theories
6.15 Limitations
Problems
Chapter 7 CONSUMPTION
71
=
rae
7.4
rae
7.6
Consumption under Certainty: The Life-Cycle/
Permanent-Income Hypothe:
Consumption under Uncertain
Hypothesis
Empirical Application: Two Tests of the Random-Walk
Hypothesis
The Interest Rate and Saving
Consumption and Risky Assets
Alternative Views of Consumption
Problems
: The Random-Walk
Chapter 8 INVESTMENT
8.1
8.2
8.3
84
85
8.6
8.7
88
Investment and the Cost of Capital
A Model of Investment with Adjustment Costs
Tobin's q
Analyzing the Model
Implications
‘The Effects of Uncertainty: An Introduction
Financial-Market Imperfections
Empirical Applications
Problems
Chapter9 INFLATION AND MONETARY POLICY
cal
es
9.3
9.4
ce
96
9.7
9.8
Introduction
Inflation, Money Growth, and Interest Rates
Monetary Policy and the Term Structure of Interest
Rates
The Dynamic Inconsistency of Low-Inflation Monetary
Policy
Addressing the Dynamic-Inconsistency Problem
Some Macroeconomic Policy Issues
Seignorage and Inflation
The Costs of Inflation
Problems
294
300
302
309
310
316
a
BP
328
332
341
345
345
348
353,
354
358
364
369
380
384
388
388
389
Be
398
403
412
420
429
433CONTENTS xv
Chapter 10 UNEMPLOYMENT 439
10.1 Introduction: Theories of Unemployment 439
10.2 A Generic Efficiency-Wage Model 441
10.3 A More General Version 446
10.4 The Shapiro-Stiglitz Model 450
10.5. Implicit Contracts : 461
10.6 Insider-Outsider Models 465
10.7 Hysteresis 469
10.8 Search and Matching Models 473
10.9 Empirical Applications 481
Problems 486
References 494
Name Index 523
Subject Index 528Section 1.7
Section 2.7
Section 2.13
Section 3.7
Section 3.11
Section 4.8
Section 4.9
Section 5.6
Section 6.4
Section 6.13
Section 6.14
Section 7.1
Section 7.3
Section 7.5
Section 7.6
Section 8.8
Section 9.3
Section 9.5
Section 10,9
EMPIRICAL APPLICATIONS
Growth Accounting
Convergence
Saving and Investment
Investment, Population Growth, and Output
Wars and Real Interest Rates
Are Modern Economies Dynamically Ffficient?
Population Growth and Technological Change Since
1 Million B.c
Physical and Human Capital Accumulation and
Cross-Country Differences in Incomes
The Persistence of Output Fluctuations
Calibrating a Real-Business-Cycle Model
Productivity Movements in the Great Depression
Money and Output
International Evidence on Output-Inflation Tradeoffs
The Average Inflation Rate and the Output-Inflation
Tradeoff
Supply Shocks
Microeconomic Evidence on Price Adjustment
Experimental Evidence on Coordination-Failure Games
Understanding Estimated Consumption Functions
Campbell and Mankiw's Test of the Random-Walk
Hypothesis Using Aggregate Data
Shea's Test of the Random-Walk Hypothesis Using
Household Data
The Equity-Premium Puzzle
Liquidity Constraints and Aggregate Saving
Buffer-Stock Saving
The Investment Tax Credit and the Price of Capital Goods
Cash Flow and Investment
The Response of the Term Structure to Changes in the
Federal Reserve's Federal-Funds-Rate Target
Central-Bank Independence and Inflation
Contracting Effects on Employment
Interindustry Wage Differences
XVii
26
27
31
32
61
84
122
137
175
180
182
232
253
289
291
293
297
312
319
322
330
338
339
380
381
396
409
481
484ACKNOWLEDGMENTS
This book owes a great deal to many people. The book is an outgrowth of
courses I have taught at Princeton, M.LT., Stanford, and especially Berkeley.
1 want ta thank the many students in these courses for their feedback, their
patience, and their encouragement.
Four people provided detailed, thoughtful, and constructive comments
on almost every aspect of the book: Laurence Ball, A. Andrew John, N. Gre-
gory Mankiw, and Christina Romer. Each of them significantly improved the
book, and | am deeply grateful to them for their efforts.
In addition, Susanto Basu, Matthew Cushing, Charles Engel, Mark Gertler,
Mary Gregory, A. Stephen Holland, Gregory Linden, Maurice Obtsfeld, and
Robert Rasche made valuable comments and suggestions concerning some
or all of the book. Jeffrey Rohaly not only prepared the superb Solutions
Manual to accompany the book, but also read the page proofs with great
care and made many corrections. Teresa Cyrus helped with the preparation
of some of the tables and figures. Finally, the editorial staff at McGraw-Hill
and the production staff at Publication Services, Inc., especially Leon Jeter,
Victoria Richardson, Scott Schriefer, Scott Stratford, and Lucille Sutton, did
an excellent job of turning the manuscript into a finished product. | thank
all of these people for their help.
xixINTRODUCTION
Macroeconomics is the study of the economy as a whole. It is therefore
concerned with some of the most important questions in economics. Why
are some countries rich and others poor? Why do countries grow? What
are the sources of recessions and booms? Why is there unemployment, and
what determines its extent? What are the sources of inflation? How do gov-
ernment policies affect output, unemployment, and inflation? These and
related questions are the subject of macroeconomics.
This book is an introduction to the study of macroeconomics at an ad-
\anced level. It presents the major theories concerning the central questions
of macroeconomics. Its goal is to provide both an overview of the field for
students who will not continue in macroeconomics and a starting point for
students who will go on to more advanced courses and research in macro-
economics and monetary economics.
The book takes a broad view of the subject matter of macroeconomic:
it views it as the study not just of aggregate fluctuations but of other fea-
tures of the economy as a whole. A substantial portion of the book is de-
soted to economic growth, and separate chapters are devoted to theories of
the natural rate of unemployment and to theories of inflation. Within each
part, the major issues and competing theories are presented and discussed.
Throughout, the presentation is motivated by substantive questions about
the world. Models and techniques are used extensively, but they are treated
as tools for gaining insight into important issues, not as ends in themselves.
The first three chapters are concerned with growth. The analysis focuses
on two fundamental questions: Why are some economies so much richer
than others, and what accounts for the huge increases in real incomes over
time? Chapter 1 is devoted to the Solow growth model, which is the basic
reference point for almost all analyses of growth. The Solow model takes
technological progress as given and investigates the effects of the division
of output between consumption and investment on capital accumulation
and growth. The chapter presents and analyzes the model and assesses its
ability to answer the central questions concerning growth.
Chapter 2 relaxes the Solow model's assumption that the saving rate is
exogenous and fixed. It covers both a model where the set of households2 INTRODUCTION
in the economy 1s fixed (the Ramsey model) and one where there is turnover
(the Diamond model).
Chapter 3 presents the new growth theory, The first part of the chapter
explores the sources of the accumulation of knowledge, the allocation of
resources to knowledge accumulation, and the effects of that accumulation
on growth. The second part investigates the accumulation of human as well
as physical capital.
Chapters 4 through 6 are devoted to short run fluctuations—the year-to-
year and quarter-to-quarter ups and downs of employment, unemployment,
and output. Chapter 4 investigates models of fluctuations where there are
no imperfections, externalities, or missing markets, and where the economy
1s subject only to real disturbances. This presentation of real-business-cycle
theory considers both a baseline model whose mechanics are fairly trans-
parent and a more sophisticated model that incorporates additional impor-
tant features of fluctuations.
Chapters 5 and 6 then turn to Keynesian models of fluctuations. These
models are based on sluggish adjustment of nominal wages and prices, and
emphasize monetary as well as real disturbances, Chapter 5 takes the exis:
tence of sluggish adjustment as given It first reviews the closed-economy
and open-economy versions of the traditional /S-LM model. It then inves-
tigates the imphcations of alternative assumptions about price and wage
rigidity, market structure, and inflationary expectations for the cychcal be-
havior of real wages, productivity, and markups, and for the relationship
between output and inflation.
Chapter 6 examines the fundamental assumption of Keynesian models
that nominal wages and prices do not adjust immediately to disturbances,
The chapter covers the Lucas imperfect-information model, models of stag-
gered adjustment of prices or wages, and new Keynesian theories of small
frictions in price-setting. The chapter concludes with a brief discussion of
theories of fluctuations based on coordination failures and real non-
Walrasian features of the economy.
The analysis in the first six chapters suggests that the behavior of con-
sumpton and investment is central to both growth and fluctuations. Chap-
ters 7 and 8 therefore investigate the determinants of consumption and in-
vestment in more detail. In each case, the analysis begins with a baseline
model and then considers alternative views, For consumption, the baseline
1s the hfe-cycle/permanent-ncome hypothesis; for mvestment, its q theory.
The fmal two chapters are devoted to mflation and unemployment.
Chapter 9 begins by explammg the central role of money growth m causing
inflation and by mvestigating the effects of money growth on inflation,
interest rates, and the real money stock. The remainder of the chapter
considers two sets of theories of the sources of high money growth: theo-
ries emphasizing output-inflation tradeoffs (particularly theories based on
the dynamic inconsistency of low-inflation monetary policy), and theories
emphasizing governments’ need for revenue from money creation.INTRODUCTION 3
The main subject of Chapter 10 is the determinants of an economy's nat-
ural rate of unemployment. The chapter also investigates the impact of fluc-
tuations in labor demand on real wages and employment. The main theories
considered are efficiency-wage theories, contracting and insider/outsider
theories, and search and matching models.'
Macroeconomics is both a theoretical and an empirical subject. Because
of this, the presentation of the theories is supplemented with examples of
relevant empirical work. Even more so than with the theoretical sections, the
purpose of the empirical material is not to provide a survey of the literature;
nor is it to teach econometric techniques. Instead, the goal is to illustrate
some of the ways that macroeconomic theories can be applied and tested.
The presentation of this material is for the most part fairly intuitive and
presumes no more knowledge of econometrics than a general familiarity
with regressions. In a few places where it can be done naturally, the empir-
ical material includes discussions of the ideas underlying more advanced
econometric techmiques.
Each chapter concludes with an extensive set of problems. The problems
range from relatively straightforward variations on the ideas in the text to
extensions that tackle important new issues. The problems thus serve both
as a way for readers to strengthen their understanding of the material and
as a compact way of presenting significant extensions of the ideas in the
text?
The fact that the book is an advanced introduction to macroeconomics
has two main consequences. The first is that the book uses a series of for-
mal models to present and analyze the theories. Models identify particular
features of reality and study their consequences in isolation. They thereby
allow us to see clearly how different elements of the economy interact and
what their implications are. As a result, they provide a rigorous way of in-
\ estigating whether a proposed theory can answer a particular question and
whether it generates additional predictions.
The book contains hterally dozens of models. The main reason for this
multiplicity is that we are interested in many issues. The features of the
economy that are crucial to one issue are often unimportant to others.
Money, for example, is almost surely central to inflation and is probably
not central to long-run growth. Incorporating money into models of growth
would only obscure the analysis. Thus instead of trying to build a single
The chapters are largely independent. The growth and fluctuations sections are almost
entirely self-contained (although Chapter 4 builds moderately on Part A of Chapter 2). There
1» also considerable independence among the chapters in each section. New growth theory
«Chapter 3) can be covered either before or after the Ramsey and Diamond models (Chapter
2), and Keynesian models (Chapters 5 and 6) can be covered either before or after real
busmness-cycle theory (Chapter 4). Finally, the last four chapters are largely self-contained
‘although Chapter 7 rehes moderately on Chapter 2, Chapter 9 relies moderately on Chapter
5, and Chapter 10 relies moderately on Chapter 6).
*A solutions manual prepared by Jeffrey Rohaly 1s available for use with the book.4 INTRODUCTION
model to analyze all of the issues we are interested in, the book develops a
series of models.
An additional reason for the multiplicity of models 1s that there 1s con-
siderable disagreement about the answers to many of the questions we will
be examining When there 1s disagreement, the book presents the leading
views and discusses their strengths and weaknesses. Because different the-
ories emphasize different features of the economy, again it is more enlght-
ening to mvestigate distinct models than to build one model incorporating
all of the features emphasized by the different views.
The second consequence of the book’s advanced level is that it presumes
some background in mathematics and economics. Mathematics provides
compact ways of expressing ideas and powerful tools for analyzing them.
The models are therefore mamly presented and analyzed mathematically.
The key mathematical requirements are a thorough understanding of single-
variable calculus and an introductory knowledge of multivariable calculus.
Tools such as functions, logarithms, derivatives and partial derivatives,
maximization subject to constramt, and Taylor-series approximations are
used relatively freely. Knowledge of the basic ideas of probability—random
variables, means, variances, covariances, and independence—is also as-
sumed.
No mathematical background beyond this level 1s needed. More advanced
tools (such as simple differential equations, the calculus of variations, and
dynamic programmung) are used sparingly, and they are explained as they are
used. Indeed, since mathematical techniques are essential to further study
and research in macroeconomics, models are sometimes analyzed in more
detail than 1s otherwise needed in order to illustrate the use of a particular
method.
In terms of economics, the book assumes an understanding of microeco-
nomics through the intermechate level. Familiarity with such ideas as profit-
maximization and utlity-maxmuzation, supply and demand, equilibrium,
efficiency, and the welfare properties of competitive equihbria 1s presumed.
Little background in macroeconomics itself 1s absolutely necessary. Read-
ers with no prior exposure to macroeconomics, however, are hkely to find
some of the concepts and termmology difficult, and to find that the pace
1s rapid (most notably in Chapter 5). These readers may wish to review an
intermediate macroeconomics text before beginning the book, or to study
such a book in conjunction with this one.
The book was designed for first-year graduate courses in macroeco-
nomucs. But it can be used in more advanced graduate courses, and (either
on its own or m conjunction with an intermediate text) for students with
strong backgrounds in mathematics and economics in professional schools
and advanced undergraduate programs. It can also provide a tour of the
field for economists and others working in areas outside macroeconomics.Chapter 1
THE SOLOW GROWTH MODEL
1.1 Theories of Economic Growth
Standards of living differ among parts of the world by amounts that almost
defy comprehension. Although precise comparisons are difficult, the best
available estimates suggest that average real incomes in such countries as
the United States, Germany, and Japan exceed those in such countries as
Bangladesh and Zaire by a factor of twenty or more. There are also large
differences in countries’ growth records. Some countries, such as South Ko-
rea, Turkey, and Israel, appear to be making the transition to membership
in the group of relatively wealthy industrialized economies. Others, includ-
ing many in South America and sub-Saharan Africa, have difficulty simply
in obtaining positive growth rates of real income per person. Finally, we see
vast differences in standards of living over time: the world is much richer
today than it was three hundred years ago, or even fifty years ago.
The implications of these differences in standards of living for human.
welfare are enormous. The real income differences across countries are as-
sociated with large differences in nutrition, literacy, infant mortality, life
expectancy, and other direct measures of well-being. And the welfare con-
sequences of long-run growth swamp any possible effects of the short-run
fluctuations that macroeconomics traditionally focuses on. During an av-
erage recession in the United States, for example, real income per person.
falls by a few percent relative to its usual path. In contrast, the productiv-
ity slowdown—the fact that average annual productivity growth since the
1970s has been about 1 percentage point below its previous level—has re-
duced real income per person in the United States by about 20 percent rel-
ative to what it otherwise would have been. Other examples are even more
startling. If real income per person in India continues to grow at its postwar
average rate of 1.3 percent per year, it will take about two hundred years
for Indian real incomes to reach the current U.S. level. If India achieves 3
percent growth, the process will take less than one hundred years. And if it
achieves Japan's average growth rate, 5.5 percent, the time will be reduced
to only fifty years. To quote Robert Lucas (1988), “Once one starts to think
about [economic growthl, it is hard to think about anything else.”6 Chapter 1 THE SOLOW GROWTH MODEL
The first three chapters of this book are therefore devoted to economic
growth. We will investigate several models of growth. Although we will ex-
amine the models’ mechanics in considerable detail, our ultimate goal is
to learn what insights they offer concerning worldwide growth and income
differences across countries.
This chapter focuses on the model that economists have traditionally
used to study these es, the Solow growth model.! The Solow model is
the starting point for almost all analyses of growth. Even models that depart
fundamentally from Solow’s are often best understood through comparison
with the Solow model. Thus understanding the model is essential to under-
standing theories of growth.
The principal conclusion of the Solow model is that the accumulation
of physical capital cannot account for either the vast growth over time in
output per person or the vast geographic differences in output per person
Specifically, suppose that the mechanism through which capital accumula-
tion affects output is through the conventional channel that capital makes a
direct contribution to production, for which it is paid its marginal product.
Then the Solow model implies that the differences in real incomes that we
are trying to understand are far too large to be accounted for by differences
in capital inputs. The model treats other potential sources of differences in
real incomes as either exogenous and thus not explained by the model (in
the case of technological progress, for example), or absent altogether (in the
case of positive externalities from capital, for example), Thus to address the
central questions of growth theory we must move beyond the Solow model.
Chapters 2 and 3 therefore extend and modify the Solow model. Chapter
2 investigates the determinants of saving and investment. The Solow model
has no optimization in it; it simply takes the saving rate as exogenous and
constant. Chapter 2 presents two models that make saving endogenous and
potentially time-varying. In the first, saving and consumption decisions are
made by infinitely-lived households; in the second, they are made by house-
holds with finite horizons.
Relaxing the Solow model’s assumption of a constant saving rate has
three advantages. First, and most important for studying growth, it demon-
strates that the Solow model's conclusions about the central questions of
growth theory do not hinge on its assumption of a fixed saving rate. Second,
it allows us to consider welfare issues. A model that directly specifies rela-
tions among aggregate variables does not provide a way to judge whether
some outcomes are better or worse than others: without individuals in the
model, we cannot say whether different outcomes make individuals bet-
ter or worse off. The infinite-horizon and finite-horizon models are built
up from the behavior of individuals, and can therefore be used to discuss
welfare issues. Third, infinite- and finite-horizon models are used to study
'The Solow model—sometimes known as the Solow-Swan model—was developed by
Robert Solow (Solow, 1956) and T. W. Swan (Swan, 1956).1.2 Assumptions 7
many issues in economics other than economic growth; thus they are valu-
able tools.
Chapter 3 investigates more fundamental departures from the Solow
model. Its models, in contrast to Chapter 2's, provide different answers than
the Solow model does to the central questions of growth theory. The models
depart from the Solow model in two basic ways. First, they make technolog-
ical progress endogenous. We will investigate various models where growth
‘occurs as the result of conscious decisions on the part of economic actors
to invest in the accumulation of knowledge. We will also consider the deter-
minants of the decisions to invest in knowledge accumulation.
Second, the models examine the possibility that the role of capital is con-
siderably larger than is suggested by considering physical capital's share in
income. This can occur if the capital relevant for growth is not just physical
capital but also human capital. It can also occur if there are positive exter-
nalities from capital accumulation, so that what capital earns in the market
understates its contribution to production. We will see that models based
on endogenous technological progress and on a larger role of capital pro-
vide candidate explanations of both worldwide growth and cross-country
ancome differences.
We now turn to the Solow model.
1.2 Assumptions
Inputs and Output
The Solow model focuses on four variables: output (Y), capital (K), labor
+L), and “knowledge” or the “effectiveness of labor” (A). At any time, the
economy has some amounts of capital, labor, and knowledge, and these are
combined to produce output. The production function takes the form
Y(t) = F(K(t), A(OL(O), aD
where t denotes time.
Two features of the production function should be noted. First, time
does not enter the production function directly, but only through K, L,
and A. That is, output changes over time only if the inputs into production
change. In particular, the amount of output obtained from given quantities
of capital and labor rises over time—there is technological progress—only
if the amount of knowledge increases.
Second, A and L enter multiplicatively. AL is referred to as effective
labor, and technological progress that enters in this fashion is known as
labor-augmenting or Harrod-neutral This way of specifying how A enters
*If knowledge enters in the form Y = F(AK,1), technological progress is capital
augmenting. If it enters in the form Y = AF(K, L), technological progress is Hicks-neutral.8 Chapter 1 THE SOLOW GROWTH MODEL
together with the other assumptions of the model, will imply that the ratio
of capital to output, K/Y, eventually settles down. In practice, capital-
output ratios do not show any clear upward or downward trend over
extended periods. In addition, building the model so that the ratio is eventu-
ally constant makes the analysis much simpler. Assuming that A multiplies
Lis therefore very convenient.
The central assumptions of the Solow model concern the properties of
the production function and the evolution of the three inputs into produc-
tion (capital, labor, and knowledge) over time. We discuss each in turn.
Assumptions Concerning the Production Function
The model's critical assumption concerning the production function is that
it has constant returns to scale in its two arguments, capital and effective la-
bor. That is, doubling the quantities of capital and effective labor (for exam-
ple, by doubling K and Z with A held fixed) doubles the amount produced.
More generally, multiplying both arguments by any nonnegative constant c
causes output to change by the same factor:
F(cK,cAL) = cF(K,AL) forall c = 0. ~ (1.2)
The assumption of constant returns can be thought of as combining
two assumptions. The first is that the economy is big enough that the gains
from specialization have been exhausted. In a very small economy, there are
probably enough possibilities for further specialization that doubling the
amounts of capital and labor more than doubles output. The Solow model
assumes, however, that the economy is sufficiently large that, if capital and
labor double, the new inputs are used in essentially the same way as the
existing inputs, and thus that output doubles.
The second assumption is that inputs other than capital, labor, and
knowledge are relatively unimportant. In particular, the model neglects
Jand and other natural resources. If natural resources are important, dou-
bling capital and labor could less than double output. In practice, however,
the availability of natural resources does not appear to be a major con-
straint on growth. Assuming constant returns to capital and labor alone
therefore appears to be a reasonable approximation.’
The assumption of constant returns allows us to work with the produc-
tion function in intensive form. Setting c = 1/AL in equation (1.2) yields
(K\_ Lae ;
F (52) = Srwan. (1.3)
%Growth accounting, which is described in Section 1.7, can be used to formalize the
‘argument that natural resources are not very important to growth. Problem 1.10 investigates
a simple model where natural resources cause there to be diminishing returns to capital and
Jabor. Finally, Chapter 3 examines the implications of increasing returns.1.2 Assumptions 9
K /ALis the amount of capital per unit of effective labor, and F(K, AL)/AL is
Y/AL, output per unit of effective labor. Define k ~ K /AL, y = Y/AL, and
f(k) = F(k, 1). Then we can rewrite (1.3) as
y = fk). (4)
That is, we can write output per unit of effective labor as a function of
capital per unit of effective labor.
To see the intuition behind (1.4), think of dividing the economy into AL
small economies, each with 1 unit of effective labor and K / AL units of capi-
tal. Since the production function has constant returns, each of these small
economies produces 1/AL as much as is produced in the large, undivided
economy. Thus the amount of output per unit of effective labor depends
only on the quantity of capital per unit of effective labor, and not on the
overall size of the economy. This is what is expressed mathematically in
equation (1.4). If we wish to find the total amount of output, as opposed to
the amount per unit of effective labor, we can multiply by the quantity of
effective labor: Y = ALf(k).
The intensive-form production function, f(k), is assumed to satisfy
(0) = 0, f’(k) > 0, f"(k) < 0.4 It is straightforward to show that f'(k) is the
marginal product of capital: since F(K, AL) = ALf(K/AL), 4F(K /AL)/aK =
ALf'(K /AL\(1/AL) = f'(k). Thus these assumptions imply that the marginal
product of capital is positive, but that it declines as capital (per unit of
effective labor) rises. In addition, f(+) is assumed to satisfy the Inada con-
ditions (Inada, 1964): limg—o f’(k) = 0, limy.. f’(k) = 0. These conditions
«which are stronger than is needed for the model's central results) state
that the marginal product of capital is very large when the capital stock
is sufficiently small and that it becomes very small as the capital stock
becomes large; their role is to ensure that the path of the economy does not
diverge. A production function satisfying f’(*) > 0, f’"(*) < 0, and the Inada
conditions is shown in Figure 1.1
A specific example of a production function is the Cobb-Douglas:
F(K,AL) = KAD", O
Note that with Cobb-Douglas production, labor-augmenting, capital-augmenting, and.
-neutral technological progress (see n. 2) are all essentually the same. For example, to
rewrite (1.5) so that technological progress 1s Hicks-neutral, simply define A = A'~*; then
(KL,1.2 Assumptions 11
The Evolu
in of the Inputs into Production
The remaining assumptions of the model concern how the stocks of labor,
knowledge, and capital change over time. The model is set in continuous
time; that is, the variables of the model are defined at every point in time.©
The initial levels of capital, labor, and knowledge are taken as given.
Labor and knowledge grow at constant rates:
Lt) = nko), (1.8)
A(t) = gA(t), (1.9)
where n and g are exogenous parameters and where a dot over a vari-
able denotes a derivative with respect to time (that is, X(t) is shorthand
for dX(v)/de). Equations (1.8) and (1.9) imply that L and A grow exponen-
tially. That is, if 1(0) and A(0) denote their values at time 0, (1.8) and (1.9)
imply L(t) = Le", AW) = A(es.?
Output is divided between consumption and investment. The fraction
of output devoted to investment, s, is exogenous and constant. One unit of
output devoted to investment yields one unit of new capital. In addition,
existing capital depreciates at rate 6. Thus:
K(t) = sY(t) = 6K(0). (1.10)
Although no restrictions are placed on n, g, and 6 individually, their sum
1 assumed to be positive. This completes the description of the model.
Since this is the first model (of many!) we will encounter, a general com-
ment about modeling is called for. The Solow model is grossly simplified
ma host of ways. To give just a few examples, there is only a single good;
government is absent; fluctuations in employment are ignored; production
1s described by an aggregate production function with just three inputs;
and the rates of saving, depreciation, population growth, and technolog-
ical progress are constant. It is natural to think of these features of the
model as defects: the model omits many obvious features of the world, and
surely some of those features are important to growth. But the purpose of
a model is not to be realistic. After all, we already possess a model that
1s completely realistic—the world itself. The problem with that “model
that it is too complicated to understand. A model's purpose is to provide
©The alternative is discrete time, where the variables are defined only at specific dates
‘usually ¢ = 0, 1,2,...). The choice between continuous and discrete time is usually based on
convenience. For example, the Solow model has essentially the same implications in discrete
as in continuous time, but is easier to analyze in continuous time.
"Yo verify this, note that Lit) = L(0)e" implies that L(t) = L(OJe""n = nL(t) and that the
mnitial value of Lis L(0)e°, or £(0) (and similarly for A).12. Chapter | THE SOLOW GROWTH MODEL
insights about particular features of the world. If a simplifying assump-
tion causes a model to give incorrect answers to the questions it is being
used to address, then that lack of realism may be a defect. (Even then, the
simplification—by showing clearly the consequences of those features of
the world in an idealized setting—may be a useful reference point.) If the
simplification does not cause the model to provide incorrect answers to the
questions it is being used to address, however, then the lack of realism is
a virtue: by isolating the effect of interest more clearly, the simplification
makes it easier to understand.
1.3. The Dynamics of the Model
We want to determine the behavior of the economy we have just described.
The evolution of two of the three inputs into production, labor and knowl-
edge, is exogenous. Thus to characterize the behavior of the economy we
must analyze the behavior of the third input, capital.
The Dynamics of k
Because the economy may be growing over time, it turns out to be conve-
nient to focus on the capital stock per unit of effective labor, k, rather than
the unadjusted capital stock, K. Since k = K / AL, we can use the chain rule
to find®
i) Kw)
= OL > fame oo + LACH)
ko
(1.11)
Kw Kit) Lt) K(_ Atty
© ADL ALO LQ ~ ALO A)
K/ALis simply k. From (1.8) and (1.9), L/L and A/A are n and g. K is given
by (1.10). Substituting these facts into (1.11) yields
SY(t) ~ 8K(t)
——— — k(t)n — k(t)g
KO = AOL)
(1.12)
Yio
AOL) ~ BRO ~ nk(O) — gki0.
®That is, since k is a function of K, I, and A, each of which are functions of t, then1.3. The Dynamics of the Model 13
Break-even investment
(n+ g+a)k
sftky
Actual investment
Investment per
unit of effective labor
K k
FIGURE 1.2 Actual and break-even investment
Finally, using the fact that Y/ AL is given by f(k), we have
Kio) = sf(k(W) - (n + g + 8)K(O. (1.13)
Equation (1.13) is the key equation of the Solow model. It states that the
rate of change of the capital stock per unit of effective labor is the differ-
ence between two terms. The first, sf(k), is actual investment per unit of
effective labor: output per unit of effective labor is f(k), and the fraction of
that output that is invested is s. The second term, (n +g +8)k, is break-even
mvestment, the amount of investment that must be done just to keep k at
its existing level. There are two reasons that some investment is needed
to prevent k from falling. First, existing capital is depreciating; this capital
must be replaced to keep the capital stock from falling. This is the 5k term
in (1.13). Second, the quantity of effective labor is growing. Thus doing
enough investment to keep the capital stock (K) constant is not enough to
keep the capital stock per unit of effective labor (k) constant. Instead, since
the quantity of effective labor is growing at rate n + g, the capital stock
must grow at rate n +g to hold k steady. This is the (n + g)k term in (1.13).9
When actual investment per unit of effective labor exceeds the invest-
ment needed to break even, k is rising. When actual investment falls short
of break-even investment, k is falling. And when the two are equal, k is
constant
Figure 1.2 plots the two terms of the expression for k as functions of k.
Break-even investment, (n +g +6)k, is proportional to k, Actual investment,
sf(k), is a constant times output per unit of effective labor.
Since f(0) = 0, actual investment and break-even investment are equal at
k = 0. The Inada conditions imply that at k = 0, f"(k) is large, and thus that
°The growth rate of a variable, X, refers its proportional rate of change, X/X. It is easy
to verify that the growth rate of the product of two variables, X,:, is the sum of their
growth rates, X;/X; + X2/X:. Similarly, the growth rate of the ratio of two variables, X; /X,
1s the difference of their growth rates, X /X; — Xp/X. Thus, the growth rate of k = K/AL is
K/K -(AjA + L/1).1t follows that keeping k constant requires K /K =n +g.14 Chapter 1 THE SOLOW GROWTH MODEL
FIGURE 1.3. The phase diagram for k in the Solow model
the sf(k) line is steeper than the (n + g + 5)k line. Thus, for small values of
k, actual investment is larger than break-even investment. The Inada con-
ditions also imply that f’(k) falls toward zero as k becomes large. At some
point, the slope of the actual investment line falls below the slope of the
break-even investment line. With the sf(k) line flatter than the (n + g + )k
line, the two must eventually cross, Finally, the fact that f’”(k) < 0 implies
that the two lines intersect only once for k > 0. We let k* denote the value
of k where actual investment and break-even investment are equal.
Figure 1,3 summarizes this information in the form of a phase diagram,
which shows k as a function of k. If k is initially less than k*, actual in-
vestment exceeds break-even investment, and so k is positive—that is, k is
rising. If k exceeds k*, k is negative. Finally, if k equals k*, k is zero. Thus,
regardless of where k starts, it converges to k*.!°
The Balanced Growth Path
Since k converges to k*, it is natural to ask how the variables of the model
behave when k equals k*. By assumption, labor and knowledge are growing
at rates n and g, respectively. The capital stock, K, equals ALK; since k is
constant at k*, K is growing at rate n +g (that is, K/K equals n + g). With
both capital and effective labor growing at rate n +g, the assumption of
constant returns implies that output, Y, is also growing at that rate. Finally,
capital per worker, K/L, and output per worker, Y/L, are growing at rate g.
Thus the Solow model implies that, regardless of its starting point,
the economy converges to a balanced growth path—a situation where each
“If k is initially zero, it remains there. We ignore this possibility in what follows.1.4 The Impact of a Change in the Saving Rate 15
variable of the model is growing at a constant rate. On the balanced growth
path, the growth rate of output per worker is determined solely by the rate
of technological progress.
The balanced growth path of the Solow model fits several of the major
stylized facts about growth described by Kaldor (1961). In most of the ma-
jor industrialized countries over the past century, it is a reasonable first
approximation to say that the growth rates of labor, capital, and output are
each roughly constant. The growth rates of output and capital are about
equal (so that the capital-output ratio is approximately constant) and are
larger than the growth rate of labor (so that output per worker and capital
per worker are rising). The balanced growth path of the Solow model has
these properties.
1.4 The Impact of a Change in the
Saving Rate
The parameter of the Solow model that policy is most likely to affect is the
saving rate. The division of the government's purchases between consump-
tion and investment goods, the division of its revenues between taxes and
borrowing, and its tax treatments of saving and investment are all likely to
affect the fraction of output that is invested. Thus it is natural to investigate
the effects of a change in the saving rate.
For concreteness, we will consider a Solow economy that is on a bal-
anced growth path, and suppose that there is a permanent increase in s.
Jn addition to demonstrating the model's implications concerning the role
of saving, this experiment will illustrate the model’s properties when the
economy is not on a balanced growth path.
The Impact on Output
The increase in s shifts the actual investment line upward, and so k* rises.
This is shown in Figure 1.4. k does not immediately jump to the new value
of k*, however. Initially, k is equal to the old value of k*. At this level, actual
investment now exceeds break-even investment—more resources are being
devoted to investment than are needed to hold k constant—and so k is
positive. Thus k begins to rise. It continues to rise until it reaches the new
value of k*, at which point it remains constant.
The behavior of output per worker, Y/L, is something we are likely to
be particularly interested in. Y/L equals Af(k). When k is constant, Y/L
grows at rate g, the growth rate of A. When k is increasing, Y/L grows
both because A is increasing and because k is increasing. Thus its growth
rate exceeds g. When k reaches the new value of k*, however, again only
the growth of A contributes to the growth of Y/L, and so the growth rate
of ¥ /Lreturns to g. Thus a permanent increase in the saving rate produces a16 Chapter 1 THE SOLOW GROWTH MODEL
(n+ g+ak
Investment per unit of effective labor
Kou Kew k
FIGURE 1.4 The effects of an increase in the saving rate on investment
temporary increase in the growth rate of output per worker: k is rising for
atime, but eventually it increases to the point where the additional saving
is devoted entirely to maintaining the higher level of k.
These results are summarized in Figure 1.5. f& denotes the time of the
increase in the saving rate. By assumption, s jumps at time fo and remains
constant thereafter. k rises gradually from the old value of k* to the new
value. The growth rate of output per worker, which is initially g, jumps
upward at fq and then gradually returns to its initial level. Thus output per
worker begins to rise above the path it was on and gradually settles into a
higher path parallel to the first.!!
In sum, a change in the saving rate has a level effect but not a growth
effect: it changes the economy's balanced growth path, and thus the level
of output per worker at any point in time, but it does not affect the growth
rate of output per worker on the balanced growth path. Indeed, in the Solow
model only changes in the rate of technological progress have growth ef-
fects; all other changes have only level effects.
The Impact on Consumption -
If we were to introduce households into the model, their welfare would de-
pend not on output but on consumption: investment is simply an input into
production in the future. Thus for many purposes we are likely to be more
interested in the behavior of consumption than in the behavior of output.
"The reason that Figure 1.5 shows the log of output per worker rather than its level is
that when a variable is growing at a constant rate, a graph of the log of the variable as a
function of time is a straight Line. That is, the growth rate of a variable is the derivative with
respect to time of the log of the variable: d In(X)/dt = (1/X) dX jdt = X/X.1.4. The Impact of a Change in the Saving Rate 17
Growth
rate
of ¥/L
g
ny/L
ty
FIGURE 1.5. The effects of an increase in the saving rate
Consumption per unit of effective labor equals output per unit of effec-
tive labor, f(k), times the fraction of that output that is consumed, 1 - s.
Thus, since s changes discontinuously at f and k does not, initially con-
sumption per unit of effective labor jumps downward, Consumption then
rises gradually as k rises and s remains at its higher level. This is shown in
the last panel of Figure 1.5.
Whether consumption eventually exceeds its level before the rise in s is
not immediately clear, Let c* denote consumption per unit of effective labor
on the balanced growth path. c* equals output per unit of effective labor,
{(k*), minus investment per unit of effective labor, sf(k*). On the balanced18 Chapter 1 THE SOLOW GROWTH MODEL
growth path, actual investment equals break-even investment, (n + g + 5)k*.
Thus,
c* = f(k*)—(n +g + d)k*. (1.14)
k* is determined by s and the other parameters of the model, n, g, and 4;
we can therefore write k* = k*(s,n, g, 6). Thus (1.14) implies
Ak* (5g 8)
as
= [f'(k*(s,n,g,5) —(n +g + 8} (1.15)
We know that the increase in s raises k*. Thus whether the increase
raises or lowers consumption in the long run depends on whether f’(k*)—
the marginal product of capital—is more or less than n + g + 5, Intuitively,
when K rises, investment (per unit of effective labor) must rise by n + g + 5
times the change in k for the increase to be sustained. If f'(k*) is less than
n+g +6, then the additional output from the increased capital is not enough
to maintain the capital stock at its higher level. In this case, consumption
must fall to maintain the higher capital stock. If f’(k*) exceeds n +g +6, on
the other hand, there is more than enough additional output to maintain k
at its higher level, and so consumption rises.
f’(k*) can be either smaller or larger than n + g + 6. This is shown in
Figure 1.6. The figure shows not only (n + g + 8)k and sf(k), but also f(k).
On the balanced growth path, consumption equals output less break-even
investment; thus the distance between f(k) and (n +g + 6)k. In the top
panel, /’(k*) is less than n + g + 8, and so an increase in the saving rate
Jowers consumption even when the economy has reached the new balanced
growth path. In the middle panel, the reverse holds, and so an increase in s
raises consumption in the long run.
Finally, in the bottom panel, f’(k*) just equals n + g + 5—that is, the
f(k) and (n + g + 5)k lines are parallel at k = k*. In this case, a marginal
change in s has no effect on consumption in the long run, and consumption
is at its maximum possible level among balanced growth paths. This value
of k* is known as the golden-ruie level of the capital stock. We will discuss
the golden-rule capital stock further in Chapter 2. Among the questions we
will address are whether the golden-rule capital stock is in fact desirable
and whether there are situations in which a decentralized economy with
endogenous saving converges to that capital stock. Of course, in the Solow
model, where saving is exogenous, there is no more reason to expect the
capital stock on the balanced growth path to equal the golden-rule level
than there is to expect it to equal any other possible value.
1.5 Quantitative Implications
We are often interested not just in a model’s qualitative implications, but
in its quantitative predictions. If, for example, the impact of a moderate1.5 Quantitative Implications 19
fk)
H (ne g+8k
Output and investment
per unit of effective labor
Output and investment
fk)
(n+ gtk
sftk)
K k
FIGURE 1.6 Output, investment, and consumption on the balanced growth
path
increase in saving on growth remains large after several centuries, the result
that the impact is temporary is of limited interest.
For most models, including this one, obtaining exact quantitative results
requires specifying functional forms and values of the parameters; it often20. Chapter 1 THE SOLOW GROWTH MODEL
also requires analyzing the model numerically. But in many cases, it is possi-
ble to learn a great deal by considering approximations around the long-run
equilibrium. That is the approach we take here.
The Effect on Output in the Long Run
‘The long-run effect of a rise in saving on output is given by
BE = prey RG G.8),
as as )
where y* = f(k*) is the level of output per unit of effective labor on the
balanced growth path. Thus to find ay*/s, we need to find dk* /as. To do
this, note that k* is defined by the condition that k = 0; thus k* satisfies
Sf(K*(S,n, g,8)) = (n +g + AK*(5,n,G, 6). (1.17)
Equation (1.17) holds for all values of s (and of n, g, and 4). Thus the deriva-
tives of the two sides with respect to s are equal:!*
| ee ak*
SPUN SE + FU) = (n +9 + IS (1.18)
where the arguments of k* are omitted for simplicity. This can be rear-
ranged to obtain!
ake fk)
“as neg +8)- RY 7
Substituting (1.19) into (1,16) yields
of ie) (1.20)
Two changes help in interpreting this expression. The first is to convert it
to an elasticity by multiplying both sides by s/y*. The second is to use the
fact that sf(k*) = (n + g + 6)k* to substitute for s, Making these changes
gives us
"This technique is known as implicit differentiation. Even though (1.17) does not ex-
plicitly give k* as a function of s, n, g, and 6, it still determines how k* depends on those
variables. We can therefore differentiate the equation with respect to s and solve for ak* /s.
“We saw in the previous section that an increase in s raises k*. To check that this is
also implied by equation (1.19), note that n +g + 6 is the slope of the break-even investment
line and that sf’(k*) is the slope of the actual investment line at k*, Since the break-even
investment line is steeper than the actual investment line at k* (see Figure 1.2), it follows
that the denominator of (1.19) is positive and thus that k*/as > 0.1.5 Quantitative Implications 21
"(k*)
(1 +g + dK*P(k*)
© FR Min +9 + 8) (n+ 9 + KP PR)
ea
RP IF)
L—[k* fk *)/f(K*
k~f'(k*)/f(k*) is the elasticity of output with respect to capital at k =
«<7. Denoting this by ax(k*), we have
8 ayt __ax(k*)
yt as V—ag(k*)"
If markets are competitive and there are no externalities, capital earns
-s marginal product. In this case, the total amount received by capital (per
nut of effective labor) on the balanced growth path is k*f’(k*). Thus if
1.6 The Solow Model and the Central
Questions of Growth Theory
The Solow model identifies two possible sources of variation—either over
time or across parts of the world—in output per worker: differences in cap-
ital per worker (K /L) and differences in the effectiveness of labor (A). We
have seen, however, that only growth in the effectiveness of labor can lead
to permanent growth in output per worker, and that for reasonable cases
the impact of changes in capital per worker on output per worker is modest.
As a result, only differences in the effectiveness of labor have any reason-
able hope of accounting for the vast differences in wealth across time and
space. Specifically, the central conclusion of the Solow model is that if the
returns that capital commands in the market are a rough guide to its con-
tributions to output, then variations in the accumulation of physical capital
do not account for a significant part of either worldwide economic growth
or cross-country income differences.
There are two problems with trying to account for large differences in in-
comes on the basis of differences in capital. First, the required differences
in capital are far too large. Consider, for example, a tenfold difference in
output per worker. Output per worker in the United States today, for in-
stance, is on the order of ten times larger than it was a hundred years ago,
and than it is in India today. Recall that ax is the elasticity of output with
respect to the capital stock. Thus accounting for a tenfold difference in out-
put per worker on the basis of differences in capital requires a difference of
a factor of 10!/** in capital per worker. For ax = }, this is a factor of a thou-
sand. Even if capital's share is one-half, which is well above what data on
capital income suggest, one still needs a difference of a factor of a hundred.
There is no evidence of such differences in capital stocks. One of the
stylized facts about growth mentioned in Section 1.3 is that capital-output
the doubling time of a variable with positive growth is 70 divided by the growth rate). Thus
in this case the half-life is roughly 70/(4%%/year), or about eighteen years. More exactly, the
half-life, t*, is the solution to e-*” = 0.5, where Ais the rate of decrease. Taking logs of both
sides, tf? = —In(0.5)/A = 0.69/2.
“These results are derived from a Taylor-series approximation around the balanced
growth path. Thus, formally, we can rely on them only in an arbitrarily small neighbor-
hood around the balanced growth path. The question of whether Taylor-series approxima-
tions provide good guides for finite changes does not have a general answer. For the Solow
model with conventional production functions, and for moderate changes in parameter val-
ues (such as those we have been considering), the Taylor-series approximations are generally
quite reliable,24 Chapter 1 THE SOLOW GROWTH MODEL
ratios are roughly constant over time. Thus the U.S. capital stock per worker
is roughly ten times larger than it was a hundred years ago, not a hun-
dred or a thousand times larger. Similarly, although capital-output ratios
vary somewhat across countries, the variation is not great. For example, the
capital-output ratio appears to be two to three times larger in the United
States than in India; thus capital per worker is “only” about twenty to thirty
times larger in the United States. In sum, differences in capital per worker
are far smaller than those 1.2eded to account for the differences in output
per worker that we are trying to understand.!®
The second difficulty is that attributing differences in output to differ-
ences in capital without differences in the effectiveness of labor implies
immense variation in the rate of return on capital (Lucas, 1990a). If markets
are competitive, the rate of return on capital equals its marginal product,
f’(k), minus depreciation, 6. Suppose that the production function is Cobb-
Douglas (see equation [1.5]), which in intensive form is f(k) = k*. With this
production function, the elasticity of output with respect to capital is simply
a. The marginal product of capital is
f'(k) = ak"
(1.27)
= ay bie,
Equation (1.27) implies that the elasticity of the marginal product of capital
with respect to output is —(1 - @)/a. If « = },a tenfold difference in output
per worker arising from differences in capital per worker thus implies a
hundredfold difference in the marginal product of capital. And since the
return to capital is f“(k) — 8, the difference in rates of return is even larger.
Again, there is no evidence of such differences in rates of return. Di-
rect measurement of returns on financial assets, for example, suggests only
moderate variation over time and across countries. More tellingly, we can
learn much about cross-country differences simply by examining where the
holders of capital want to invest. If rates of return were larger by a factor
of ten or a hundred in poor countries than in rich countries, there would be
immense incentives to invest in poor countries. Such differences in rates of
return would swamp such considerations as capital-market imperfections,
government tax policies, fear of expropriation, and so on, and we would ob-
serve immense flows of capital from rich to poor countries. We do not see
such flows.!7
‘One can make the same point in terms of the rates of saving, population growth, and
so on that determine capital per worker. For example, the elasticity of y* with respect to s
is ax /(1 — ax) (see [1.22]). Thus accounting for a difference of a factor of ten in output per
worker on the basis of differences in s would require a difference of a factor of a hundred
ins if ax = } and a difference of a factor of ten if a, = }. Variations in actual saving rates
are much smaller than this.
One can try to avoid this conclusion by considering production functions where capi-
tal's marginal product falls less rapidly as k 's than it does in the Cobb-Douglas case. This1.6 The Solow Model and the Central Questions of Growth Theory 25,
Thus differences in physical capital per worker cannot account for the
differences in output per worker that we observe, at least if capital's con-
tribution to output is roughly reflected by its private returns.
The other potential source of variation in output per worker in the Solow
model is the effectiveness of labor. Attributing differences in standards of
living to differences in the effectiveness of labor does not require huge dif
ferences in capital or in rates of return. Along a balanced growth path, for
example, capital is growing at the same rate as output; and the marginal
product of capital, f’(k), is constant,
The Solow model's treatment of the effectiveness of labor is highly in-
complete, however. Most obviously, the growth of the effectiveness of labor
is exogenous: the model takes as given the behavior of the variable that it
identifies as the driving force of growth. Thus it is only a small exaggeration
to say that we have been modeling growth by assuming it.
More fundamentally, the model does not identify what the “effectiveness
of labor” is; it is just a catchall for factors other than labor and capital that
affect output. To proceed, we must take a stand concerning what we mean.
by the effectiveness of labor and what causes it to vary. One natural possi-
bility is that the effectiveness of labor corresponds to abstract knowledge.
To understand worldwide growth it would then be necessary to analyze the
determinants of the stock of knowledge over time. To understand cross-
country differences in real incomes, one would have to explain why firms
in some countries have access to more knowledge than firms in other coun-
tries, and why that greater knowledge is not rapidly transferred to poorer
countries.
There are other possible interpretations of A: the education and skills of
the labor force, the strength of property rights, the quality of infrastructure,
cultural attitudes toward entrepreneurship and work, and so on. Or A may
reflect a combination of forces. For any proposed view of what A represents,
one would again have to address the questions of how it affects output, how
it evolves over time, and why it differs across parts of the world.
The other possible way to proceed is to consider the possibility that cap-
ital is more important than the Solow model implies. If capital encompasses
more than just physical capital, or if physical capital has positive external-
ities, then the private return on physical capital is not an accurate guide to
capital's importance in production. In this case, the calculations we have
done may be misleading, and it may be possible to resuscitate the view that
differences in capital are central to differences in incomes.
These possibilities for addressing the fundamental questions of growth
theory are the subject of Chapter 3.
approach would encounter two major difficulties. First, since the marginal product of capital
would be similar in rich and poor countries, capital's share would be much larger in rich
countries. Second, and similarly, real wages would be only slightly larger in rich than in
poor countries, These implications appear grossly inconsistent with the facts.26 Chapter 1 THE SOLOW GROWTH MODEL
1.7 Empirical Applications
Growth Accounting
In the Solow model, long-run growth of output per worker depends only on
technological progress. But short-run growth can result from either tech-
nological progress or capital accumulation. Thus the model implies that
determining the sources of short-run growth is an empirical issue. Growth
accounting, which was pioneered by Abramovitz (1956) and Solow (1957),
provides a way of tackling this subject.
To see how growth accounting works, consider again the production
function Y(t) = F(K(t), A(¢)L(t)). This implies
MOK . FOipn 4 FO Ko, (1.28)
YO= KO at) A(t)
oY /aL and dY/A denote [dY / (ALA and [4Y /4(AL)L, respectively. Divid-
ing both sides by Y(t) and rewriting the terms on the right-hand side yields
Kavi Kin | Lo a¥(o Le) | Aw
Y(t) aK(t)K(t) Y(t) al(t) Li) ¥(e)
cent ® agen ll
= ox (OR + alr + R(t).
Here a(t) is the elasticity of output with respect to labor at time t,
ax(t) is again its elasticity with respect to capital, and R(t) =
[A(O/ ¥ (OMA (0)/ ACOA) ACO]. Subtracting L(t)/ L(t) from both sides and
using the fact that a:(t) + ex(t) = 1 (see Problem 1.7, at the end of this
chapter) gives us an expression for the growth rate of output per worker:
Fo _ LW fo be
vi) Lo ~ oo i0| we CL
The growth rates of Y, K, and L are straightforward to measure. And we
know that if capital earns its marginal product, ax can be measured using
data on the share of income that goes to capital, R(t) can then be mea-
sured as the residual in (1.30). Thus (1.30) provides a way of decomposing
the growth of output per worker into the contribution of growth of capital
per worker and a remaining term, the Solow residual. The Solow residual
is sometimes interpreted as a measure of the contribution of technological
progress. As the derivation shows, however, it reflects all sources of growth
other than the contribution of capital accumulation via its private return.
This basic framework can be extended in many ways (see, for exam-
ple, Denison, 1967). The most common extensions are to consider differ-
ent types of capital and labor and to adjust for changes in the quality of1.7 Empirical Applications 27
inputs. But more complicated adjustments are also possible. For example,
if there is evidence of imperfect competition, one can try to adjust the data
on income shares to obtain a better estimate of the elasticity of output with
respect to the different inputs.
Growth accounting has been applied to many issues. For example, Young
(1994) uses detailed growth accounting to argue that the unusually rapid
growth of Hong Kong, Singapore, South Korea, and Taiwan over the past
three decades is almost entirely due to rising investment, increasing labor-
force participation, and improving labor quality (in terms of education), and
not to rapid technological progress and other forces affecting the Solow
residual.'®
To give another example, growth accounting has been used extensively
to study the productivity slowdown—the reduced growth rate of output per
worker-hour in the United States and other industrialized countries that be-
gan in the early 1970s (see, for example, Denison, 1985; Baily and Gordon,
1988; Griliches, 1988; and Jorgenson, 1988). Some candidate explanations
that have been proposed on the basis of this research include slower growth
in workers’ skills, the disruptions caused by the oil-price increases of the
1970s, a slowdown in the rate of inventive activity, and the effects of gov-
ernment regulations.
Convergence
An issue that has attracted considerable attention in empirical work on
growth is whether poor countries tend to grow faster than rich countries.
‘There are at least three reasons that one might expect such convergence.
First, the Solow model predicts countries converge to their balanced growth
paths. Thus to the extent that differences in output per worker arise from
countries being at different points relative to their balanced growth paths,
one would expect the poorer countries to catch up to the richer. Second, the
Solow model implies that the rate of return on capital is lower in countries
with more capital per worker. Thus there are incentives for capital to flow
from rich to poor countries; this will also tend to cause convergence. And
third, if there are lags in the diffusion of knowledge, income differences can
arise because some countries are not yet employing the best available tech-
nologies. These differences might tend to shrink as poorer countries gain
access to state-of-the-art methods.
Baumol (1986) examines convergence from 1870 to 1979 among the 16
industrialized countries for which Maddison (1982) provides data. Baumol
regresses output growth over this period on a constant and initial income;
that is, he estimates
Inf(¥/N)ia979] — In{(¥ /N)jas70] = @ + BIn[(Y/N)iaszol + &- (1.31)
‘Other authors examining the same issue, however, argue for a larger role for the resid-
ual. See, for example, Page (1994).28 Chapter 1 THE SOLOW GROWTH MODEL
Here In(¥/NN) is log income per person, ¢ is an error term, and i indexes
countries." If there is convergence, b will be negative: countries with higher
initial incomes have lower growth. A value for b of —1 corresponds to
perfect convergence: higher initial income on average lowers subsequent
growth one-for-one, and so output per person in 1979 is uncorrelated with
its value in 1870. A value for b of 0, on the other hand, implies that growth
is uncorrelated with initial income and thus that there is no convergence.
The results are
In{(Y /N)ia979] ~ In[(Y /N)i,1870] = 8.457 - 0.995 Inl(Y/N),i870],
(0.094) (1.32)
R? =0.87, see. = 0.15,
where the number in parentheses, 0.094, is the standard error of the re-
gression coefficient. Figure 1.7 shows the scatterplot corresponding to thi
regression.
The regression suggests almost perfect convergence. The estimate of b
is almost exactly equal to —1, and it is estimated fairly precisely; the two-
standard-error confidence interval is (0.81, 1.18). In this sample, per capita
income today is essentially unrelated to per capita income a hundred years
ago.
+ United States
Denmark,
Switzerland
Belgium
Log per capita income growth
1870-1979
56 38 60 62 64 66 68 70 72 74 76
Log per capita income in 1870
FIGURE 1.7. Initial income and subsequent growth in Baumol’s sample (from
De Long, 1988; used with permission)
Baumol considers output per worker rather than output per person. This choice has
little effect on the results.1.7 Empirical Ap
29
De Long (1988) demonstrates, however, that Baumol’s finding is largely
spurious. There are two problems. The first is sample selection. Since his-
torical data are constructed retrospectively, the countries that have long
data series are generally those that are the most industrialized today. Thus
countries that were not rich a hundred years ago are typically in the sample
only if they grew rapidly over the next hundred years. Countries that were
rich a hundred years ago, in contrast, are generally included even if their
subsequent growth was only moderate. Because of this, we are likely to see
poorer countries growing faster than richer ones in the sample of countries
we consider even if there is no tendency for this to occur on average.
The natural way to eliminate this bias is to use a rule for choosing the
sample that is not based on the variable we are trying to explain, which
1s growth over the period 1870-1979. Lack of data makes it impossible
to include the entire world. De Long therefore considers the richest coun-
tries as of 1870; specifically, his sample consists of all countries at least
as rich as the second poorest country in Baumol’s sample in 1870, Finland.
This causes him to add seven countries to Baumol’s list (Argentina, Chile,
East Germany, Ireland, New Zealand, Portugal, and Spain), and to drop one
«Japan).2°
Figure 1.8 shows the scatterplot for the unbiased sample. The inclusion
of the new countries weakens the case for convergence considerably. The
2.6
24
.
2.2 2
Fast Germany # +
1870-1979
18) Spain + +
16 Ireland + cal
le + + +New Zealand
rab Cle * portugal + :
Argentina
Log per capita income growth
1
%o 62 64 66 68 70 7
Log per capita income m 1870
FIGURE 1.8 Initial income and subsequent growth in the expanded sample
(from De Long, 1988; used with permission)
20since a large fraction of the world was richer than Japan in 1870, 1t 18 not possible
to consider all countes at least as rich as Japan. In addition, one has to deal with the fact
that countries’ borders are not fixed. De Long chooses to use 1979 borders. Thus his 1870
income estimates are estimates of average comes in 1870 in the geographic regions defined
by 1979 borders.30 Chapter 1 THE SOLOW GROWTH MODEL
regression now produces an estimate of b of —0.566, with a standard er-
ror of 0.144. Thus accounting for the selection bias in Baumol’s procedure
eliminates about half of the convergence that he finds.
The second problem that De Long identifies is measurement error. Es-
timates of real income per capita in 1870 are imprecise, Measurement er-
ror again creates bias toward finding convergence. When 1870 income is
overstated, growth over the period 1870-1979 is understated by an equal
amount; when 1870 income is understated, the reverse occurs. Thus mea-
sured growth tends to be lower in countries with higher measured initial
income even if there is no relation between actual growth and actual initial
income.
De Long therefore considers the following model:
Inf(Y/N)i1979] - In[(Y/N)iag70]* = @ + bIn[(Y/N)iiszol* + €, (1.33)
In{(Y/N),1870] = In(¥/N)ias70l* + ui, (1.34)
where In{(Y /N)ig70|* is the true value of log income per capita in 1870 and
In{(¥ /.N)1s70] is the measured value. ¢ and u are assumed to be uncorrelated
with each other and with In[(Y/N)iszal*.
Unfortunately, it is not possible to estimate this model using only data
on In{(¥/N)iszo] and In{(Y /N)1979]. The problem is that there are different
hypotheses that make identical predictions about the data. For example,
suppose we find that measured growth is negatively related to measured
initial income. This is exactly what one would expect either if measurement
error is unimportant and there is true convergence or if measurement error
is important and there is no true convergence. Technically, the model is not
identified.
De Long argues, however, that we have at least a rough idea of how good
the 1870 data are, and thus have a sense of what is a reasonable value for
the standard deviation of the measurement error. oy = 0.01, for example,
implies that we have measured initial income to within an average of 1 per-
cent; this is implausibly low. Similarly, o, = 0.50—an average error of 50
percent—seems implausibly high. De Long shows that if we fix a value of
ou, We can estimate the remaining parameters.
Even moderate measurement error has a substantial impact on the re-
sults. For the unbiased sample, the estimate of b reaches 0 (no tendency
toward convergence) for o, ~ 0.15, and is 1 (tremendous divergence) for
ou ~ 0.20. Thus plausible amounts of measurement error eliminate most
or all of the remainder of Baumol’s estimate of convergence.
It is also possible to investigate convergence for different samples of
countries and different time periods. Figure 1.9 is a convergence scatter-
plot analogous to Figures 1.7 and 1.8 for virtually the entire non-Communist
world for the period 1960-1985. As the figure shows, there is little evidence
of convergence. We return to the issue of convergence at the end of Chapter 3.1,7. Empirical Applications 31
207
8
8 +
Sash + +
ms 1s t
8 + + +
a ys
= 10+ 5. 7 ty
z ve + tat
5 + t
5 tt +
3 ty + +t + +
+ 4 ee te
ost . + o oe wh
++ +#4 +
5 ete SFr + +
g a. . .
5: tf
ee
g 7) RY Se
5 7
& +H
2-05 ae
é
6
3 6 7 8 9 1
Log income per capita in 1960 (1985 international prices)
FIGURE 1.9 Initial income and subsequent growth in the postwar period (data
from Summers and Heston, 1991)
Saving and Investment
Consider a world where every country is described by the Solow model and
where all countries have the same amount of capital per unit of effective
labor. Now suppose that the saving rate in one country rises. If all of the ad-
ditional saving were invested domestically, the marginal product of capital
in that country would fall. There would therefore be incentives for residents
of the country to invest abroad. Indeed, in the absence of any impediments
to capital flows, the investment resulting from the increased saving would
be spread uniformly over the whole world; the fact that the rise in saving
occurred in one country would have no special effect on investment there.
Thus there would be no reason to expect countries with high saving to also
have high investment.
Feldstein and Horioka (1980) examine the association between saving
and investment rates. They find that, contrary to this simple view, saving
and investment rates are strongly correlated. Specifically, Feldstein and Ho-
rioka run a cross-country regression for 21 industalized countries of the
average share of investment in GDP during the period 1960-1974 on a con-
stant and the average share of saving in GDP over the same period. The
results are
F——S—SsCSe = Ph (1.35)
(0.018) (0.074)32 Chapter 1 THE SOLOW GROWTH MODEL
where again the numbers in parentheses are standard errors. Thus, rather
than there being no relation between saving and investment, there is an
almost one-to-one relation.
There are various possible explanations for Feldstein and Horioka’s find-
ing (see Obstfeld, 1986, for a discussion). One possibility, suggested by Feld-
stein and Horioka, is that significant barriers to capital mobility exist. In this
case, differences in saving and investment across countries would be asso-
ciated with rate of return differences.
Another possibility is that there are underlying variables that affect both
saving and investment. For example, high tax rates can reduce both saving
and investment (Barro, Mankiw, and Sala-i-Martin, 1995). Similarly, countries
whose citizens have low discount rates, and thus high saving rates, may
provide favorable investment climates in ways other than the high saving;
for example, they may limit workers’ ability to form strong unions.
Finally, the strong association between saving and investment can arise
from government policies that offset forces that would otherwise make sav-
ing and investment differ. Governments may be averse to large gaps be-
tween saving and investment—after all, a large gap must be associated with
a large trade deficit (if investment exceeds saving) or a large trade surplus
(if saving exceeds investment). If economic forces would otherwise give rise
to a large imbalance between saving and investment, the government may
choose to adjust its own saving behavior or its tax treatment of saving or
investment to bring them into rough balance.
In sum, the strong relationship between saving and investment differs
dramatically from the predictions of a natural baseline model. Whether thi:
difference reflects major departures from the baseline (such as large barri-
ers to capital mobility) or something less fundamental (such as underlying
forces affecting both saving and investment) is not known.
Investment, Population Growth, and Output
According to the Solow model, saving and population growth affect output
per worker through their impact on capital per worker. A country that saves
more of its output has more capital per worker, and hence more output per
worker; a country with higher population growth devotes more of its saving
to maintaining its capital-labor ratio, and so has less capital and output per
worker.
‘The model makes not just qualitative but quantitative predictions about
the impact of saving and population growth on output. We saw in Section
1.5 that the elasticity of output on the balanced growth path with respect
to s is a/(1 — a), where a is capital’s share. Similarly, one can show that its
elasticity with respect to n + g +8 is —a/(1 — a) (see Problem 1.5). Thus,?!
‘One can also derive (1.36) by assuming that the production function 1s Cobb-Douglas;
in this case, no approximations are needed (see Problem 1.2).1.7. Empirical Applications 33
Iny* =a+ In(n + g +6). (1.36)
Mankiw, D. Romer, and Weil (1992) estimate equation (1.36) empirically
using cross-country data. Their basic specification is,
Iny, =a + bllns, ~In(y +9 + 8) +e, (1.37)
«here i indexes countries. Finding empirical counterparts for y, s, and n
5 fairly straightforward. Mankiw, Romer, and Weil measure y as real GDP
per person of working age in 1985, s as the average share of real private
and government investment in real GDP over the period 1960-1985, and n
as the average growth rate of the population of working age over the same
period.?? Finally, g + 4 is set to 0.05 for all countries.
The results for the broadest set of countries considered by Mankiw,
Romer, and Weil are:
Inyi = 6.87 + 1.48[lns, — n(n + 0.05),
(0.12) (0.12) (1.38)
R°=0.59, — s.e.e. = 0.69.
Saving and population growth enter in the directions predicted by the
model and are highly statistically significant, and the regression accounts
for a large portion of cross-country differences in income. In this sense, the
model is a success.
There is one major difficulty, however: the estimated effect of saving
and population growth is far larger than the model predicts. The estimate
of b = 1.48 implies & = 0.60 (with a standard error of 0.02).?3 Thus the
relationship between saving and population growth and real income is far
stronger than the model predicts for reasonable values of the capital share,
and the data are grossly inconsistent with the hypothesis that a is in the
vicinity of one-third. Thus, Mankiw, Romer, and Weil's results confirm the
conclusion that the Solow model cannot account for important features of
cross-country income differences.
“The data are from the Summers and Heston (1988) cross-country data set.
mers and Heston (1991) for a more recent version.
“Finding estimates and standard errors for parameters that are nonlinear functions of
regression coefficients is straightforward. In the case of (1.36)-(1.38), solving b
tor a yields a = b/(1 +b). The estimate of & = 0.60 is thus obtained by computing
b) = 1.48/(1 + 1.48). In addition, a first-order Taylos 's approximation of
around b ~ b yields « = [b/(1 + b)] +[1/(. + 6)" (b~ b). Thus the difference between the true
4 and @ is approximately 1/(1 + b)*, or 0.16, times the difference between the true b and
b. The standard error of a is therefore approximately 0.16 times the standard error of b,
or 0,16(0.12) = 0,02. (Because of the nonlinearity and the use of approximations, the formal
econometric justification for these procedures reltes on asymptotic theory. See, for example,
Greene, 1993, Section 10.3.3; or Judge et al., 1985, Section 5.
See Sum-34 Chapter 1 THE SOLOW GROWTH MODEL
Problems
1.1. Consider a Solow economy that is on its balanced growth path. Assume for
simplicity that there is no technological progress. Now suppose that the rate
of population growth falls.
(a) What happens to the balanced-growth-path values of capital per worker,
output per worker, and consumption per worker? Sketch the paths of these
variables as the economy moves to its new balanced growth path.
(b) Describe the effect of the fall in population growth on the path of output
(that is, total output, not output per worker).
1.2.
Suppose that the production function is Cobb-Douglas.
(a) Find expressions for k*, y*, and c* as functions of the parameters of the
model, s, n, 5, g, and «.
(b) What is the golden-rule value of k?
(c) What saving rate is needed to yield the golden-rule capital stock?
1.3. Consider the constant elasticity of substitution (CES) production function, Y =
[Kio + (Apye-Weyele-D, where 0 < @ < = and a = 1. (is the elasticity of
substitution between capital and effective labor. In the special case of + 1,
the CES function reduces to the Cobb-Douglas.)
(a) Show that this production function exhibits constant returns to scale.
(b) Find the intensive form of the production function.
(c) Under what conditions does the intensive form satisfy f'(*) > 0, f’"(*) < 0?
(d) Under what conditions does the intensive form satisfy the Inada condi-
tions?
1.4, Consider an economy with technological progress but without population
growth that is on its balanced growth path. Now suppose there is a one-time
jump in the number of workers.
(a) At the time of the jump, does output per unit of effective labor rise, fall,
or stay the same? Why?
(b) After the initial change (if any) in output per unit of effective labor when
the new workers appear, is there any further change in output per unit of
effective labor? If so, does it rise or fall? Why?
(c) Once the economy has again reached a balanced growth path, is output per
unit of effective labor higher, lower, or the same as it was before the new
workers appeared? Why?
1.5. Find the elasticity of output per unit of effective labor on the balanced growth
path, y*, with respect to the rate of population growth, n. If ax(k*) = 4,9 =
2%, and 6 = 3%, by about how much does a fall in n from 2% to 1% raise y*?
1.6. Suppose that, despite the political obstacles, the United States permanently
reduces its budget deficit from 3% of GDP to zero. Suppose that initially s =
0.15 and that investment rises by the full amount of the fall in the defic
Assume that capital’s share is 5.1.8.
19.
1.10.
Problems 35
(a) By about how much does output eventually rise relative to what it would
have been without the deficit reduction?
(b) By about how much does consumption rise relative to what it would have
been without the deficit reduction?
(c) What is the immediate effect of the deficit reduction on consumption?
About how long does it take for consumption to return to what it would
have been without the deficit reduction?
Factor payments in the Solow model. Assume that both labor and capital
are paid their marginal products. Let w denote aF(K,AL)/aL and r denote
aF(K, ADI aK.
(a) Show that the marginal product of labor, w, is Alf(k) — kf’(k)l-
(b) Show that if both capital and labor are paid their marginal products, con-
stant returns to scale implies that the total amount paid to the factors of
production equals total output. That is, show that under constant returns,
wL + rK = F(K, AL).
(c) Two additional stylized facts about growth listed by Kaldor (1961) are
that the return to capital (r) is approximately constant and that the shares
of output going to capital and labor are each roughly constant, Does a
Solow economy on a balanced growth path exhibit these properties? What
are the growth rates of w and r on a balanced growth path?
(d) Suppose the economy begins with a level of k less than k*. As k moves
toward k*, is w growing at a rate greater than, less than, or equal to its
growth rate on the balanced growth path? What about r?
Suppose that, as in Problem 1.7, capital and labor are paid their marginal
products. in addition, suppose that all capital income is saved and all labor
income is consumed. Thus K = [@F(K,AL)/4K|K ~ 8K.
(a) Show that this economy converges to a balanced growth path,
(b) Is k on the balanced growth path greater than, less than, or equal to the
golden-rule level of k? What is the intuition for this result?
The Harrod-Domar model. (See Harrod, 1939, and Domar, 1946.) Suppose
the production function is Leontief, Y(t) = minke K(t), c.e%"L(t)], where cx,
cz, and g are all positive. As in the Solow model, i(t) = n£(t) and K(t) = s¥(t)—
8K (t). Finally ime Cx K (0) = c,L{0).
(a) Under what condition does ex K(t) = cre L(t) for all 7 If cx, CLs 9, 5. 8s
and n are determined by separate considerations, is there any reason to
expect that this condition holds?
(b) If cye% L(t) is growing faster than cx K(f) (and if the excess labor is as-
sumed to be unemployed), what happens to the unemployment rate over
time?
(c) If cx K(0) is growing faster than c,e%" (7) (and if the excess capital is as-
sumed to be unused), what happens to the fraction of the capital stock
that is used over time?
Natural resources in the Solow model. At least since Malthus, some have
argued that the fact that some factors of production (notably land and naturala
Lu.
1.12.
Chapter 1 THE SOLOW GROWTH MODEL
resources) are available in finite supply must eventually bring growth to a
halt. This problem asks you to address this idea in the context of the Solow
model
Let the production function be Y = K*(ALPR'-*-*, where R is the
amount of land. Assume a > 0, 8 > 0, and @ + B <1. The factors of pro-
duction evolve according to K = s¥ ~ 8K, A= gA, L, and R = 0,
(a) Does this economy have a unique and stable balanced growth path? That
is, does the economy converge to a situation in which each of Y, K, I,
A, and R are growing at constant (but not necessarily equal) rates? If so,
what are those growth rates? If not, why not?
(b) tn light of your answer, does the fact that the stock of land is constant
imply that permanent growth is not possible? Explain intuitively.
Embodied technological progress. (This follows Solow, 1960, and Sato,
1966.) One view of technological progress is that the productivity of capital
goods built at ¢ depends on the state of technology at t and is unaffected by
subsequent technological progress. This is known as embodied technological
progress (technological progress must be “embodied” in new capital before
it can raise output). This problem asks you to investigate its effects.
(a) Asa preliminary, let us modify the basic Solow model to make technolog-
ical progress capital-augmenting rather than labor-augmenting. So that
a balanced growth path exists, assume that the production function is
Cobb-Douglas: Y(t) = [A(t)K (A)}*L(t)!"*. Assume that A grows at rate yu:
A(t) = wA(t).
Show that the economy converges toa balanced growth path, and find
the growth rates of ¥ and K on the balanced growth path. (Hint: show that
we can write ¥ /(A®L) as a function of K /(A¢L), where @ = a/(1 ~ a). Then
analyze the dynamics of K /(4*L).)
(b) Now consider embodied technological progress. Specifically, let the pro-
duction function be Y(t) = J(*L(c)!*, where J(t) is the effective capital
stock. The dynamics of J(t) are given by Jit) = sA(t)¥(t)— a(t). The pres-
ence of the A(t) term in this expression means that the productivity of
investment at t depends on the technology at t.
Show that the economy converges to a balanced growth path. What
are the growth rates of Y and J on the balanced growth path? (Hint: let
Jit) = J(d)/ Alt). Then use the same approach as in (a), focusing on J/(A%L)
instead of K /(A*L).)
(c) What is the elasticity of output on the balanced growth path with respect
tos?
(d) In the vicinity of the balanced growth path, how rapidly does the economy
converge to the balanced growth path?
(e) Compare your results for (c) and (d) with the corresponding results in the
text for the basic Solow model.
Consider a Solow economy on its balanced growth path. Suppose the growth-
accounting techniques described in Section 1.7 are applied to this economy.Problems 37
(a) What fraction of growth in output per worker does growth accountng:
attribute to growth in capital per worker? What fraction does it attribute
to technological progress?
(b) How can you reconcile your results in (a) with the fact that the Solow
model imphes that the growth rate of output per worker on the balanced
growth path is determined solely by the rate of technological progress?
+ 13. (a) In the model of convergence and measurement error in equations (1.33)-
(1.34), suppose the true value of b 1s -1, Does a regression of
In(¥/N)1o75 ~ ICY /N)igzo ON a constant and In(Y /N)iszo yield a biased
estimate of b? Explain.
(b) Suppose there 1s measurement error in measured 1979 income per capita
but not in 1870 income per capita. Does a regression of In(¥ /N)ia70
In(Y /N)iszo on a constant and In(¥/N)is7o yield a biased estimate of b?
Explam.Chapter 2
BEHIND THE SOLOW MODEL:
INFINITE-HORIZON AND
OVERLAPPING-GENERATIONS
MODELS
This chapter investigates two models that resemble the Solow model but in
which the dynamics of economic aggregates are determined by decisions at
the microeconomic level. Both models continue to treat the growth rates of
labor and knowledge as exogenous. But the models derive the evolution of
the capital stock from the interaction of maximizing households and firms
in competitive markets. As a result, the saving rate is no longer exogenous,
and it need not be constant.
The first model is conceptually the simplest. Competitive firms rent
capital and hire labor to produce and sell output, and a fixed number of
infinitely-lived households supply labor, hold capital, consume, and save.
This model, which was developed by Ramsey (1928), Cass (1965), and
Koopmans (1965), avoids all market imperfections and all issues raised
by heterogeneous households and links among generations. It therefore
provides a natural benchmark case.
The second model is the overlapping-generations model developed by
Diamond (1965). The key difference between the Diamond model and the
Ramsey-Cass-Koopmans model is that the Diamond model assumes that
there is continual entry of new households into the economy. As we will
see, this seemingly small change has important consequences.
382.1 Assumptions 39
PartA The Ramsey-Cass-Koopmans
Model
2.1 Assumptions
Firms
There is a large number of identical firms. Each has access to the production
function Y = F(K, AL), which satisfies the same assumptions as in Chapter
i. The firms hire workers and rent capital in competitive factor markets,
and sell their output in a competitive output market. Firms take A as given;
as in the Solow model, A grows exogenously at rate g. The firms maximize
profits. They are owned by the households, so any profits they earn accrue
to the households.
Households
There is also a large number of identical households. The size of each house-
hold grows at rate n. Each member of the household supplies one unit of
tabor at every point in time. In addition, the household rents whatever cap-
xal it owns to firms, It has initial capital holdings of K(0)/H, where K(0)
1s the initial amount of capital in the economy and H is the number of
households. For simplicity, in this chapter we assume that there is no de-
preciation. The household divides its income (from the labor and capital it
supplies and, potentially, from the profits it receives from firms) at each
point in time between consumption and saving so as to maximize lifetime
anlity.
The household's utility function takes the form
v= f_emuccen te at 1)
C(t) is the consumption of each member of the household at time ft. u(*)
is the instantaneous utility function, which gives each member's utility at
a given date. L(t) is the total population of the economy; L(t)/H is there-
fore the number of members of the household. Thus u(C(t))L(t)/H is the
household’s total instantaneous utility at t. Finally, p is the discount rate;
the greater is p, the less the household values future consumption relative
to current consumption.!
‘One could also write utility as f(°) e-*'u(C(t))dt, where p’ = p—n. Since L(t) = L(O)e"",
‘his expression equals the expression in equation (2.1) divided by 1{0)/H, and thus has the
same implications for behavior.