The Growth of Nations
N. Gregory Mankiw; Edmund S. Phelps; Paul M. Romer
Brookings Papers on Economic Activity, Vol. 1995, No. 1, 25th Anniversary Issue. (1995), pp.
275-326.
Stable URL:
http://links.jstor.org/sici?sici=0007-2303%281995%291995%3A1%3C275%3ATGON%3E2.0.CO%3B2-Y
Brookings Papers on Economic Activity is currently published by The Brookings Institution.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained
prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in
the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/journals/brookings.html.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic
journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers,
and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take
advantage of advances in technology. For more information regarding JSTOR, please contact support@jstor.org.
http://www.jstor.org
Thu Jun 21 06:44:47 2007
N. GREGORY M A N K I W
Harvard University
The Growth of Nations
AVERAGEINCOMES in the world's richest countries are more than ten
times as high as in the world's poorest countries. It is apparent to anyone
who travels the world that these large differences in income lead to large
differences in the quality of life. Less apparent are the reasons for these
differences. What is it about the United States, Japan, and Germany that
makes these countries so much richer than India, Indonesia, and Nigeria?
How can the rich countries be sure to maintain their high standard
of living? What can the poor countries do tojoin the club?
After many years of neglect, these questions are again at the center of
macroeconomic research and teaching. Long-run growth is now widely
viewed to be at least as important as short-run fluctuations. Moreover,
growth is not just important. It is also a topic about which macroeconomists,
with their crude aggregate models, have something useful to say.
My goal here is to assess what we now know about economic growth.
The scope of this paper is selective and, to some extent, idiosyncratic.
The study of growth has itself grown so rapidly in recent years that it
would take an entire book to discuss the field thoroughly.' In this paper,
I do not try to lay out the many different views in the large literature on
economic growth. Instead, I try to present my own views, as cogently
as I can, on what we know about the growth of nations.
Textbook Neoclassical Theory
Most students of economics begin their study of long-run growth with
the neoclassical model of capital accumulation. When discussing what
we know about growth, this model is the natural place to start.
I am grateful to Laurence Ball, John Leahy, Edmund Phelps, Jordan Rappaport,
Michael Rashes, Paul Romer, Jaume Ventura, David Weil, and Martin Weitzman for helpful
comments, and to the National Science Foundation for financial support.
1. Indeed, Barro and Sala-i-Martin(1995) havejust finished such a book.
276 Brookings Papers on Economic Activity, 1:I995
Overview
Robert Solow provided the most basic version of the neoclassical theory
of growth.?The centerpiece of the model is the production function
( 1 ) Y = F ( K ,AL),
where Yis output, K is capital, L is labor, and A is a measure of the level
of technology. A L can be seen as the labor force measured in efficiency
units, which incorporates both the amount of labor and the productivity
of labor as determined by the available technology. If we assume that
the production function has constant returns to scale, we can write the
production function as
(2) Y = f(k)>
where y = YIAL,k = KIAL, andf(k) = F(k,1).This production function
relates output per efficiency unit of labor to the amount of capital per
efficiency unit of labor.
The neoclassical model emphasizes how growth arises from the accumulation
of capital. The capital stock per efficiency unit, k, evolves according
to
where s is the rate of saving, n is the rate of population growth, g is the
rate of growth in technology, 6 is the rate at which capital depreciates,
and a dot over a variable denotes change per unit of time. The model
takes s, n , g, and 6 as exogenous.
As long as the production function is well behaved, the economy approaches
a steady state over time. The steady state is defined by
(4) k = 0,
or, using a star to denote a steady-state value,
(5) sf(k") = (n + g + 6)kX.
In the steady state, income per efficiency unit, y* = f(kX),is constant.
Income per person grows at rate g, and total income grows at rate
(n + g).
N. Gregory Mankiw
Predictions
One strength of Solow's version of the neoclassical growth model is
that, despite its simplicity, it has many predictions. Since these predictions
are so well-known, I will not formally derive them here. But, in
evaluating the usefulness of the model in explaininggrowth experiences,
it is worth stating some of these predictions once again:
1. In the long run, the economy approaches a steady state that is independent
of initial conditions.
2. The steady-state level of income depends on the rates of saving and
population growth. The higher the rate of saving, the higher the steadystate
level of income per person. The higher the rate of population
growth, the lower the steady-state level of income per person.
3. The steady-state rate of growth of income per person depends only
on the rate of technological progress; it does not depend on the rates of
saving and population growth.
4. In the steady state, the capital stock grows at the same rate as income,
so the capital-to-income ratio is constant.
5. In the steady state, the marginal product of capital is constant,
whereas the marginal product of labor grows at the rate of technological
progress.
These predictions are broadly consistent with experience. If factors
of production earn their marginal product, then the last prediction can
be tested with data on factor prices. As a first approximation, in the U.S.
economy the real wage grows at about the same rate as income per person,
and the profit rate exhibits little trend. These facts are consistent
with prediction 5. Similarly, the capital-to-income ratio exhibits little
trend, which is consistent with prediction 4.
The dependence of steady-state income on saving and population
growth rates (prediction 2) also appears to be consistent with experience.
In cross-country data income per person is positively correlated
with saving rates and negatively correlated with population growth
rates. Moreover, these correlations are quite strong: a regression of income
per person on these two variables alone, using a sample of ninetyeight
countries, yields an adjusted R2of 59 p e r ~ e n t . ~It is possible that
3. Mankiw,Romer, and Weil(1992, p. 414).
278 Brookings Papers on Economic Activity, 1:l995
reverse causality is part of the story here; but at the very least, these
correlations do not give any reason to doubt the model.
The other two predictions are less easily evaluated. The inability of
saving to affect steady-state growth (prediction 3) might appear inconsistent
with the strong correlation between growth and saving across
countries. But this correlation could reflect the transitional dynamics
that arise as economies approach their steady states. The independence
of the steady state from initial conditions (prediction I) is closely related
to the debate over convergence, which I discuss below.
The simplicity of the neoclassical model, together with its ability to
yield substantive and seemingly reasonable predictions, has given it a
prominent place in the macroeconomist's toolbox. Whenever practical
macroeconomists have to answer questions about long-run growth, they
usually begin with a simple neoclassical growth model. For example,
when the 1994Economic Report of the President discusses the benefits
of higher national saving, it does so by offering numerical simulations
from a standard growth model of the sort Solow presented almost forty
years earlier.4
Household Behavior
So far I have followed Solow's rendition of the neoclassical model in
treating the saving rate as an exogenous parameter. This approach is extraordinarily
useful. It allows us to abstract from household behavior in
order to highlight the roles of capital accumulation, population growth,
and technological progress. In other words, by sweeping the saving decision
under the rug, we can concentrate our attention on the production
side of the model.
If we want to add an explicit analysis of household behavior, we can
choose from two basic approaches. We can follow Paul Samuelson and
Peter Diamond and model the economy as composed of a series of overlapping
g-nerations, each with a finite lifetime.5Or we can follow Frank
Ramsey, David Cass, and Tjalling Koopmans and model the economy
with a single, infinitely lived representative c o n ~ u m e r . ~Both of these
4. Economic Report of the President, 1994, pp. 85-87.
5. See Samuelson (1958)and Diamond (1965).
6. See Ramsey (1928), Cass (1965), and Koopmans (1965).
N. Gregory Mankiw 279
approaches parsimoniously turn the Solow model into a rigorous general-equilibrium
model.
These two models of household behavior differ in some important respects.
For example, in the overlapping-generations model, the economy
can accumulate too much capital. In particular, the economy can
reach a steady state with the capital stock greater than what Edmund
Phelps called the Golden Rule level.' This outcome is not possible in the
representative-consumer model, for the representative consumer would
never choose an allocation of resources that is dynamically ineffi~ient.~
Togive another example, in the overlapping-generations model, government
debt can alter the saving rate and the capital stock. In the representative-consumer
model, Ricardian equivalence holds, and so government
debt does not crowd out capital.
For most of the issues addressed by neoclassical growth theory, however,
the two approaches to household behavior yield similar results. In
both the overlapping-generations and representative-consumer models,
the economy reaches a steady state with a constant saving rate. The
steady-state saving rate is higher when consumer preferences exhibit
more patience. In both models, a constant saving rate can arise even out
of steady state for specific functional forms and parameter values. The
steady states in these models are much the same as the steady states in
the Solow model. In particular, the five predictions listed above continue
to hold.
In my view, neither of these standard approaches to modeling household
behavior is fully satisfactory, for neither holds up under empirical
scrutiny. The overlapping-generations model assumes that all saving is
for life-cyclepurposes; yet bequests are a large part of wealth accumulation9
The representative-consumer model assumes that all consumers
look into the infinite future when deciding how much to save; yet many
people leave no bequests and, therefore, are not economically linked to
future generations. Both models of household behavior are based on the
premise that people smooth consumption over their own lifetimes; yet,
in the world, consumption smoothing is far from perfect.I0
7. Phelps (1961).
8. Although important as a matter of theory, excessive capital accumulation is not a
practical concern for policymakers. Actual economies appear to have less capital than the
Golden Rule level. See Abel and others (1989).
9. Kotlikoff and Summers (1981).
10. Campell and Mankiw (1989).
Brookings Papers on Economic Activity, 1:I995280
For the purposes of this paper, these issues regarding household behavior
are tangential. Little would be added (and much generality would
be lost) by tying the analysis to either the overlapping-generations model
or the representative-consumer model. Therefore, like Solow, I will abstract
from household behavior and take the saving rate as given.
I do not mean this choice to suggest that explaining saving is uninteresting
or unimportant. Indeed, as I hope to make clear, the saving rate
is a crucial variable to understand. Yet the models of saving routinely
employed in the growth literature do not take us very far toward this
goal. Until better models of household behavior are developed,the most
we can do is discuss some general lessons about economic growth that
apply regardless of what determines national saving.
Theoretical Objections
Is the neoclassical model a good theory of economic growth? Although
this question is largely empirical, one might also answer it along
theoretical lines. Before turning to the data, therefore, let us consider
some possible theoretical objections one might lodge against the neoclassical
growth model.
One might object to the model on the grounds that it does not, in the
end, shed light on economic growth. In the steady state of the neoclassical
model, all growth is due to advances in technology, but technological
progress is taken as exogenous. It might seem that the model unravels
the mystery of economic growth simply by assuming that there is economic
growth. Indeed, this critique helped to motivate the recent theories
of endogenous growth, which I discuss later.
The persuasiveness of this objection to the neoclassical model depends
on the purpose of growth theory. If the goal is to explain why standards
of living are higher today than a century ago, then the neoclassical
model is not very illuminating. In my view, however, the goal is not to
explain the existence of economic growth. That task is too easy: it is obvious
that living standards rise over time largely because knowledge expands
and production functions improve.
A more challenging goal is to explain the variation in economic growth
that we observe in different countries and in different times. For this
purpose, the neoclassical model's assumption of constant, exogenous
technological change need not be a problem. Even with this assumption,
N. Gregory Mankiw 281
the model predicts that different countries will reach different steadystate
levels of income per person, depending on their rates of saving and
population growth. And it predicts that countries will have different
rates of growth, depending on each country's initial deviation from its
own steady state. Thus, the assumption of constant, exogenous technological
change does not preclude addressing many of the central issues
of growth theory.
To use the neoclassical model to explain international variation in
growth requires the assumption that different countries use roughly the
same production function at a given point in time. To some writers, this
assumption is preposterous." In poor countries, workers dig ditches
with shovels. In rich countries, they use bulldozers. Common sense
seems to suggest that we abandon the premise of a common production
function.
In my view, this objection to the neoclassical model is also not persuasive.
The production function should not be viewed literally as a description
of a specific production process, but as a mapping from quantities
of inputs into a quantity of output. To say that different countries have
the same production function is merely to say that if they had the same
inputs, they would produce the same output. Different countries with
different levels of inputs need not rely on exactly the same processes for
producing goods and services. When an economy doubles its capital
stock, it does not give each worker twice as many shovels. Instead, it
replaces shovels with bulldozers. For the purposes of modeling economic
growth, this change should be viewed as a movement along the
same production function, rather than as a shift to a completely new production
function.
In summary, various theoretical objections can be advanced against
the neoclassical growth model. Yet none is compelling. More important
is the empirical question: Can the model help to explain the wide variation
in economic experiences observed throughout the world?
Three Problems for Neoclassical Growth Theory
The neoclassical model has come under attack in recent years as providing
an empirically inadequate theory of growth. I now turn to some
11. See, for example, Grossman and Helpman (1994).
Brookings Papers on Economic Activity, 1:I995282
of the problems that arise when this model is brought to the data. In particular,
I consider three reasons to doubt the validity of the neoclassical
model, at least as it has been traditionally interpreted.
The approach I take here is one of calibration rather than estimation.
The problems that I highlight do not hinge on subtle issues of theory or
econometrics. They are, I hope to show, robust implications of the neoclassical
model that are evaluated fairly easily. The issue at hand is not
whether the neoclassical model is exactly true. The issue is whether the
model can even come close to making sense of international experience.
Problem 1: The Magnitude of International Differences
Suppose, for the moment, that all economies were in their steady
states. The neoclassical model predicts that different countries should
have different levels of income per person, depending on the various parameters
that determine the steady state. To see these predictions, consider
the two steady-state conditions:
(6) sy* = (n + g + S)k*.
and
(7)
Equation 6 says the saving must equal break-even investment (the
amount of investment required to keep k constant). Equation 7 is the
production function.
To see the predicted variation in income per person, differentiate this
system and solve for dy" to obtain
(8) dy*/y* = [cw/(l - a)][ds/s - d(n + g + S)/(n + g + a)],
where a =f'(k*)k*/'f(kx). If the factors of production earn their marginal
product, then cw is the steady-state capital share. Notice that this equation
does not require that the production function be Cobb-Douglas. If
it is Cobb-Douglas, then cw is a constant production function parameter.
Equation 8 is easily calibrated with data from the national income
accounts. A standard estimate of the capital share is one-third, so
a/(l - a) is one-half. The equation now shows the magnitude of income
differences that the neoclassical model can explain. In particular, it says
that differences in rates of saving will lead to differences in income that
N. Gregory Mankiw 283
are proportionately half as large. If one country's saving rate is four
times that of another country, its steady-state income will be about twice
as large.
The equation also has numerical implications for the impact of population
growth. For example, suppose that (g + 6)is 5 percent per year. (As
I discuss later, this figure is about right.) Then, a decrease in population
growth from 3 to 1percent per year reduces (n + g + 6)from 8 to 6 percent.
If aI(l - a) is one-half, then steady-state income moves proportionately
half as much as (n + g + 6). Hence, a country with 1 percent
population growth will have steady-state income 1.15[(816)"*]times the
steady-state income of a country with 3 percent population growth.
These calculations show the first defect of the neoclassical model: it
does not predict the large differences in income observed in the real
world. A comparison of rich and poor countries finds saving rates that
differ by about a multiple of four and population growth rates that vary
by about 2 percentage points.I2The above calculations, therefore, indicate
that the model can explain incomes that vary by a multiple of
slightly more than two. Yet income per person varies by a multiple of
more than ten. There is much more disparity in international living standards
than the neoclassical model predicts.
These findings might once again call into question the assumption that
all countries operate with the same production function. Perhaps poor
countries have not only low saving and high population growth, but also
poor production technologies. But it should be clear that the magnitude
of the unexplained differences makes this explanation unsatisfactory.
My calculations above indicate that the neoclassical model leaves a multiple
of five in income per person unexplained. If differences in the production
function are the reason, then poor countries must be using a
technology that is vastly inferior to that of rich countries. That is, poor
countries could be producing much more output without increasing the
quantities of their capital or their labor. If this were the case, the incentive
to imitate technology used by rich countries would be tremendous.j 3
Of course, imitating technology is not necessarily so easy. To adopt
12. See Mankiw, Romer, and Weil(1992,data appendix).
13. It may be helpful to state this point in units of time. I f technological change enhances
productivity by 2 percent per year, and ifrich countries are five times as productive
as poor countries, then poor countries must be using a production function that is about
eighty years out ofdate.
284 Brookings Papers on Economic Activity, 1:I995
the best available technology, an economy may need a skilled labor
force. In this case, however, it would not be enough simply to assert that
there are international differences in production functions without further
analysis. Instead, one is naturally led to think about the role of human
capital in economic growth. I take up this issue later.
Problem 2: The Rate of Convergence
Much of the recent debate over economic growth has centered on the
issue of convergence. Convergence has usually been defined as a tendency
of poor economies to grow more rapidly than rich economies.
Convergence in this sense might more properly be called mean re-
version.
Whether convergence is found in the data depends on the sample being
examined. Samples that include relatively homogeneous economies,
such as the countries of the OECD or the states of the United States,
typically yield evidence of convergence.14 Yet more diverse samples
give the opposite result. In large samples of countries, such as the data
set compiled by Robert Summers and Alan Heston, a country's initial
level of income per person is not correlated with its subsequent growth
rate. l5
As I have described the neoclassical model, it does not necessarily
predict convergence. If countries are in different steady states, then rich
countries remain rich, and poor countries remain poor. On the other
hand, if all countries have the same steady state and differ only in initial
conditions, then the model does predict convergence. Those who reject
the neoclassical model on the grounds that it predicts convergence,
which does not occur in large samples of countries, appear to be assuming
this very special case of identical steady states.
More generally, the neoclassical model predicts that each economy
converges to its own steady state, which in turn is determined by its saving
and population growth rates. This prediction has been called conditional
convergence. To test for conditional convergence, various authors
have run regressions of growth rates on initial income, including
variables to control for determinants of the steady state.I6Most studies
14. See, for example, Dowrickand Nguyen (1989),and Barro and Sala-i-Martin (1991).
15. Summers and Heston (1991);Romer (1987).
16. See, for example, Barro (1991),Mankiw, Romer, and Weil(1992),and Levine and
Renelt (1992).
N. Gregory Mankiw 285
have found evidence of conditional convergence at a rate of about 2 percent
per year. That is, each country moves 2 percent closer to its own
steady state each year and, by implication, moves halfway toward its
steady state in thirty-five years. This rate of conditional convergence is
fairly robust to the sample being examined.
Although conditional convergence is qualitatively consistent with the
neoclassical model, the model begins to have problems once again when
we turn to its quantitative predictions. According to the model, income
converges to its steady-state level as follows:
(9) j, = -A(y - y*),
where
These equations are derived in the appendix. For the purposes at hand,
the key parameter is the rate of convergence, A. This parameter measures
how quickly a deviation from steady state dissipates over time. If
n, g , and 6 are measured as percent per year, then A indicates the percentage
of the deviation from steady state that is eliminated each year.I7
This formula is easily calibrated. In the United States, for example,
the capital share, a,is about one-third, and the rate of population
growth, n, is about I percent per year. The average rate of growth of income
per person is about 2 percent per year, which gives us a value for g.
If we take the capital consumption allowance and divide it by the capital
stock, we obtain an estimate of the depreciation rate, 6, of 3 percent per
year. Together with the equation above, these estimates give a predicted
rate of convergence, A, of 4 percent per year. At this rate, an economy
would go halfway toward its steady state in seventeen and one-half
years.
Hence, although the model does predict the conditional convergence
found in the empirical literature on economic growth, it does not predict
the rate of convergence that these studies estimate. In particular, the
model predicts convergence at about twice the rate that actually occurs.
In practice, economies do regress toward their conditional mean, but
17. The rate of convergence depends on the steady-state capital share, a,but not on
the specific form of the production function. If the production function happens to be
Cobb-Douglas, then a is a production function parameter. If it is not Cobb-Douglas, then
a will depend on the saving rate as well as production function parameters.
286 Brookings Papers on Economic Activity, 1:I995
only slowly. An economy's initial condition matters for much longer
than the model says it should.
Problem 3: Rates of Return
A third critique of the neoclassical model emphasizes the predicted
differences in rates of return. If poor countries are poor because they
have small capital stocks, then the marginal product of capital should be
high. We should, therefore, observe higher profit rates and higher real
interest rates in poor countries. Moreover, capital should be eager to
flow from rich to poor countries.
There is some evidence for return differentials of this sort. Because
the profit rate is capital income divided by the capital stock, it also
equals the capital share of income divided by the capital-income ratio. If
one accumulates data on investment to obtain data on capital stocks,
one finds that capital-income ratios are more than twice as large in rich
as in poor co~ntries.'~Unless rich countries have capital shares that are
also more than twice as large (which appears not to be the case), they
must have lower profit rates. This seems qualitatively consistent with
the neoclassical model.
Persistent return differentials are not necessarily puzzling, even in the
presence of capital mobility, if the differentials are modest. Investment
abroad, particularly in poor countries, involves greater information
costs and greater risk of expropriation. Many American investors are
only beginning to see the virtues of international diversification, even in
other developed economies. Investment in the so-called emerging markets
is still rare. Those investors who are now entering these markets,
often through mutual funds, do expect to earn higher returns.
The neoclassical model runs into trouble when we turn from qualitative
to quantitative predictions about rates of return. To see the magnitude
of the predicted return differentials, consider the following equations
(which hold both in and out of steady state):
The first equation is the production function. The second equation says
that the gross return to capital, R , is the marginal product of capital,
18. See, for example, Mankiw, Romer, and Weil(1992, p. 431).
N. Gregory Mankiw 287
which equals the first derivative of the production function. Differentiating
these equations, with some manipulation, yields
(12) dRIR = [fl/Cf')2]dyiy.
Additional algebra allows substitution forf,f ,andf'in order to obtain
a more easily interpreted expression:
where a is the capital share and cr is the elasticity of substitutionbetween
capital and labor.19This equation shows how the return to capital varies
with the level of income.
As equation 13 illustrates, it is impossible to establish a quantitative
prediction about return differentials without saying something about the
elasticity of substitution between capital and labor. A common approach
is to assume that the production function is Cobb-Douglas, so
that u = 1.20Imposing also the standard value for a of one-third gives
That is, the return to capital moves proportionately twice as much (in
the opposite direction) as the level of income. Because poor countries
have about one-tenth the income of rich countries, they should have returns
to capital that are about one hundred times as large. In particular,
since the profit rate is about 10 percent per year in rich countries, it
should be about 1,000percent per year in poor countries. Detailed data
are not necessary to reject this prediction. Moreover, the return differential
is solarge that the failure of capital to move toward poor countries
cannot be explained by invoking information costs or political risk.
This prediction, however, relies crucially on the assumption of a
Cobb-Douglas production function. The size of the predicted return differentials
can be reduced by assuming a larger elasticity of substitution
between capital and labor. If the elasticity is four, for example, the return
to capital in a poor country should be only 3.2 [10"2]times as large
as in a rich country, in contrast to the previous multiple of 100.Although
this figure is still large, it is empirically much more pla~sible.~'
19. Straightforward algebra establishes that a = f'klf and a = [(f')V(f'fl][l- f/(f'k)].
20. See, for example, Lucas (1990).
21. The calculations in this paragraph are indicative o f how the elasticity o f substitution
matters, but they should not be taken too literally. These calculations rely on a loglinear
approximation that is exact if u equals one, but becomes less accurate as a departs
from one,for a is no longer a constant in this case.
288 Brookings Papers on Economic Activity, 1:I995
A large elasticity of substitution is not as outlandish as it might first
appear. Cobb-Douglas is often considered a reasonable assumption because
factor shares in U.S. data are roughly constant over time. Yet, in
the neoclassical model, factor shares are constant in the steady state for
any production function. Factor shares may be roughly constant in U.S.
data merely because the U.S. economy has not recently been far from
its steady state. The approximate constancy of the capital-to-income ratio
over time suggests that the U.S. economy has indeed been in a steady
state.
In judging the plausibility of a large elasticity of substitution, keep in
mind that an economy's elasticity depends not only on each industry's
technology, but also on the ability to move resources among industries.
For example, a large elasticity of substitution might arise because of international
trade. In traditional Heckscher-Ohlin trade theory, international
trade in goods equalizes factor prices in countries with different
factor endowments. The implication of this factor-price-equalization
theorem for growth theory is that trade increases an economy's ability
to substitute capital and labor.22When a country's endowment of capital
increases, it can increase exports of capital-intensive goods and increase
imports of labor-intensive goods without altering the returns to either
capital or labor. Each country has a production function that is, in effect,
linear in capital and labor. In other words, as international trade works
to equalize factor prices around the world, it drives the effective elasticity
of substitution in each economy toward infinity.
There are, however, limits to how high the elasticity of substitution
can plausibly be set when trying to explain the return to capital. This can
be seen by considering the return to the other factor of productionlabor.
In the neoclassical model, the real wage, W, is
(15) W = f(k) - f'(k)k. 

This equation, together with the production function, y = f(k),implies 

The wage is positively related to income per person. In contrast to capital,
labor should be eager to migrate from poor to rich countries, since
larger capital stocks raise labor productivity and real wages. Of course,
these predictions about wages and migration are consistent with experi-
22. See Ventura (1994).
N. Gregory Mankiw 289
ence. The magnitude of the predicted effects, however, diminishes as
the elasticity of substitution rises. Although the elasticity of substitution
may be greater than its traditional Cobb-Douglas value of one, international
experience on factor returns cannot be explained merely by positing
that this elasticity is very large.
In summary, the neoclassical growth model can be evaluated by examining
how the return to capital differs across countries. The size of
the predicted differentials, however, depends on the production function.
The larger the elasticity of substitution between capital and labor,
the smaller the return differentials. For the traditional Cobb-Douglas
production function, the return differentials predicted by the neoclassical
model are vastly larger than are observed in the world.
A New View of Capital
The neoclassical growth model emphasizes the accumulation of capital
over time. To use the model to shed light on international experience,
one must have some interpretation of the term capital and some way of
measuring the return to capital. Traditionally, capital is thought to be
tangible-it includes the economy's stock of equipment and structures.
The return to capital is the profit received by the owners of equipment
and structures.
Over the past decade, a new view of capital has emerged. According
to this view, the return to capital is a much larger fraction of national
income than has been traditionally believed. This new view alters the
interpretation of the neoclassical growth model and, by doing so, greatly
expands its scope and applicability.
The Key Role of the Capital Share
The previous section discussed three problems that arise when the
neoclassical growth model is used to understand international experience.
First, the model predicts less variation in income than is observed
across countries. Second, the model predicts a faster rate of convergence
to the steady state than most studies estimate. Third, the model
predicts greater variation in rates of return across countries than is empirically
plausible.
290 Brookings Papers on Economic Activity, 1:I995
Recall that these conclusions were derived from the following equa-
tions:
dy*ly" = [al(l - a)][ds/s- d(n + g + 6)/(n+ g + a)],
Looking at these equations for a minute yields an important insight: The
capital share, a ,plays a key role in each of the threeproblems. The reason
is that the capital share determines the shape of the production function.
To be precise, one can show that the elasticity of the average product
of capital,f(k)lk, with respect to amount of capital, k , is ( a - I). The
larger the capital share, the less rapidly the average product declines.
Thus, a larger value of a implies that changes in saving have greater effects
on steady-state income, that the transition to the steady state is
slower, and that the return to capital varies less with income.
In my earlier calculations, I used a conventional estimate for the capital
share of one-third. This estimate comes from the national income accounts.
In the U.S. economy, capital receives about one-third of gross
income, and labor receives about two-thirds. Approximately the same
is true in the other developed economies for which good data on factor
income are available.
Suppose, however, that the capital share were two-thirds rather than
one-third. (For now, do not ask how this could be true. I will turn to that
question in a moment.) Let us see how this increase in the capital share
would affect each of the problems with neoclassical growth theory.
Consider first the predicted differences in steady-state income. An
increase in the capital share, a , from one-third to two-thirds raises
a/(l - a ) in the first equation from one-half to two. Income now moves
proportionately twice as much as the rate of saving, rather than half as
much. If one country has four times the saving rate as another, it would
have sixteen times as much income per person, rather than only two
times as much. Thus, the model can now explain variation in income of
the magnitude observed.
This increase in the capital share also raises the predicted impact of
population growth, n , on steady-state income. Suppose that (g + 6) is
5 percent per year, and we are comparing one country with population
growth of 1 percent per year and another with population growth of 3
percent per year. In this case, ( n + g + 6)is 6 percent in the first country
N. Gregory Mankin) 291
and 8 percent in the second. If aI(1 - a ) is one-half, then income in the
country with low population growth is 1.15 [(816)'12]times income in the
country with high population growth. Yet if aI(1 - a ) is two, the predicted
multiple is 1.78 [(816)2].Thus, with a larger capital share, differences
in population growth are predicted to lead to much larger differences
in income.
Next, consider the predicted rate of convergence. If the capital share,
a , is two-thirds rather than one-third, the implied convergence parameter,
A, is half as large. For the other parameter values used earlier,
(n + g + 6) = 0.06, and A = 0.02. Thus, the model now predicts the rate
of convergence estimated in the empirical literature.
Finally, consider the implied differences in the return to capital between
a rich country and a poor country with incomes that differ by a
multiple of 10. For concreteness, suppose that the elasticity of substitution,
a,is one. The last equation says that the return moves proportionately
twice as much as income if the capital share is one-third. Yet the
return moves proportionately half as much as income if the capital share
is two-thirds. Thus, the predicted return differential falls from a multiple
of 100 [lo2]to a multiple of 3.16 [10'12].And if the capital share is larger
than two-thirds, the predicted return differential is smaller still.
Once again, the elasticity of substitution, a,is significant. If this elasticity
is four rather than one, then the last equation (with a = 213) says
that the return to capital moves proportionately one-eighth as much as
income. In this case, the return in a poor country is predicted to be only
1.33[lO'ls]times the return in a rich country. That is, if the return to capital
is 10percent per year in rich countries, then it should be 13 percent
per year in poor ones.
The predicted return differentials now appear unobjectionable. It is
not at all implausible to think that the emerging economies offer returns
of 3 percentage points per year higher than is available in developed
economies. Keep in mind that, for the purpose of these comparisons,
the relevant return is gross of taxes and political risks. Foreign investors
in emerging economies might require compensation of several percent
per yearjust to compensate them for the risk of e~propriation.~~
Hence, each of the three problems with the neoclassical model of
23. The night this paragraph was written, the television show "60 Minutes" ran a segment
on Nigerians posing as government officials to defraud American businessmen, and
the Nigerian government's failure to prosecute the con men. Viewers of this show were
left in no doubt as to why capital does not flow from rich to poor countries.
292 Brookings Papers on Economic Activity, 1:1995
growth would disappear if the capital share were much higher than is
conventionally understood. As a theoretical matter, this resolution is attractive
for its parsimony: it solves three serious problems by changing
the value of one parameter. The question is, why should one believe that
the capital share is actually so high?
Capital with Externalities
One way to raise the capital share above one-third is to argue that
there are positive externalities to capital. That is, some of the benefits to
capital accumulation may accrue not to the owners of capital, but to others
in society. If new ideas arise as capital is built, for example, and if
these ideas enter the general pool of knowledge, then even if capital receives
only one-third of income, in some sense it deserves credit for
more.
One of the recent advances in growth theory, due largely to Paul
Romer, has been the ability to formalize this idea.24Suppose that each
individual firm, i, in an economy has the production function
where 0 is a parameter that each firm takes as given. Also suppose that,
because of externalities to capital, the technology available to each firm
is determined by the average firm's level of capital, k, so that
Then the production function for the economy is
In this case, the parameter a measures the role of capital in each firm's
production function; it determines the fraction of income paid to the
owners of capital. Yet the role of capital in the economy's production
function is measured by (a + b). For the purposes of calibrating the
neoclassical growth model, the capital share, a , should be set equal to
(a + b).
For the economy's capital share, a , to be much larger than the share
of income received by the owners of capital, the externalities to capital
must be large. To raise the capital share from one-third to two-thirds,
24. Romer(1986, 1987).
N . Gregory Mankiw 293
the externality (as measured by the parameter b) must be about as large
as the direct benefit (as measured by the parameter a).That is, the owners
of capital must be paid only half of the social return from their invest-
ments.
In evaluating the plausibility of such an externality, it is important to
note that the externality must be geographically limited if it is to help
explain differences across countries. One commonly cited external benefit
to capital accumulation is the creation of ideas. Yet since many
ideas, like public goods, are neither rival nor excludable, they flow
freely around the world. If externalities to capital are to make the neoclassical
model conform to experience, the parameter b should be taken
to measure the benefits to capital that are external to the firm and yet
stop at the border.
It is hard to know how large such externalities could be. The idea that
capital conveys positive externalities is plausible. It is easier to be skeptical
whether capital conveys local externalities of the magnitude necessary
to save the neoclassical model from the three problems discussed
above.
Human Capital
A second argument for a larger capital share posits that capital is a
much broader concept than is suggested by the national income accounts.
In the national income accounts, capital income includes only
the return tophysical capital, such as plant and equipment. More generally,
however, we accumulate capital whenever we forgo consumption
today in order to produce more income tomorrow. In this sense, one of
the most important forms of capital accumulation is the acquisition of
skills. Such human capital includes both schooling and on-the-job
training.
When applying the neoclassical model to understand international experience,
it seems best to interpret the variable k as including all kinds
of capital. Thus, the capital share, a, should include the return to both
physical and human capital. Yet, the return to human capital is not part
of capital income in the national income accounts. Instead, it is part of
labor income. Therefore, if we use the national income accounts to calibrate
the neoclassical model, we are likely to substantially underestimate
the capital share.
294 Brookings Papers on Economic Activity, 1:I995
To gauge the true capital share, it is necessary to decide how much of
labor income should be credited to human capital. To do this, we might
look at the minimum wage, which is roughly the return to labor with minimal
human capital. In the United States the minimum wage today is
about one-third of the average wage. This fact suggests that the return
to human capital is about two-thirds of labor income, or almost one-half
of national income.
Another way to estimate the human capital share of income is to look
at the return to schooling. A large literature in labor economics finds that
each year of schooling raises a worker's wage by at least 8 percent.
Moreover, the average American has about thirteen years of schooling.25Together
these facts imply that the average worker earns almost
three times as much as he would without any human capital. Again, this
suggests that about two-thirds of the average worker's earnings is the
return to his education, and that human capital earns almost one-half of
national income.
Adding this estimate of the human capital share to the physical capital
share of one-third, we find that the income from all forms of capital is
about 80 percent of national income. Hence, when calibrating the neoclassical
model, the capital share, a, should be set at about 0.8. For reasons
already discussed, a parameter value of this magnitude makes the
neoclassical model conform much more closely to international expe-
rience.
Reinterpreting capital to include human capital, however, creates a
new problem: Data from the national income accounts correspond less
well to the variables in the model. Most of the accumulation in human
capital takes the form of forgone wages of students in school or workers
in training. The national income accounts do not include this expenditure
on human capital in either measured investment or measured GDP.
Moreover, most direct expenditures on human capital, such as on teachers
and books, enter the national income accounts as consumption
rather than investment.
Focusing on human capital does help to resolve an earlier problem:
the international comparison of rates of return. There is a large literature
estimating the returns to schooling around the world. George Psacharopoulos
summarizes results from over 60 countries.26The findings are
25. Psacharopoulos (1985, table 3) and StatisricalAbstrnct of the United States, 1993,
table 230.
26. Psacharapoulos (1985).
N . Gregory Mankin) 295
consistent with the neoclassical model and its prediction of diminishing
marginal product. The measured return to schooling is consistently
larger in poor countries than in rich countries.
Acknowledging the role of human capital can also help explain why
capital does not flow from rich to poor countries. Because human capital
does not serve well as collateral, borrowing to finance human capital investment
is often difficult, even in rich countries with developed financial
institutions. It is not surprising that resources do not flow into poor
countries to help them finance investment in human capital. Moreover,
once we acknowledge this capital market imperfection, the failure of
physical capital to flow is less puzzling. Because human and physical
capital are complementary inputs in production, imperfections in the financing
of human capital impede the international movement of physical
capital as well.27
Incorporating human capital investment into the neoclassical model
also raises the proportion of international variation that the model can
explain. As mentioned earlier, a cross-country regression of income per
person on saving and population growth rates yields an R2of 59percent.
If a measure of saving in the form of human capital is added to this regression,
the R2rises to 78 percent.28Put simply, most international differences
in living standards can be explained by differences in accumulation
of both human and physical capital.
Hence, broadening the meaning of capital to include human as well as
physical capital can help make the neoclassical growth model consistent
with international experience. In this case, one can argue for a large capital
share without invoking externalities of any sort. Of course, externalities
to capital, either physical or human, could still exist. If so, the
implied capital share is higher still. But externalities are not necessary
for the capital share to be large. Either with or without externalities, the
role of capital in economic growth appears much greater than traditionally
has been assumed.
Theories of Endogenous Growth
So far, my attention has centered on the neoclassical growth model. I
have taken the conservative position that we should begin with this stan-
27. Barro, Mankiw, and Sala-i-Martin (1995).
28. Mankiw, Romer, and Weil(1992, p. 420).
296 Brookings Papers on Economic Activity, 1:I995
dard model, see what problems arise when we go to the data, and then
try to make the minimal changes necessary to make the model work. If
we are willing to give the capital share a value much larger than traditionally
has been assumed, the major problems with this model disappear,
and the model becomes a useful framework for understanding international
differences in economic growth.
Yet much of the recent literature on economic growth has taken a
more radical approach. The neoclassical model, even with capital interpreted
broadly, implies that growth in income per person eventually approaches
g, the exogenous rate of technological progress. Although the
model can explain international differences in growth rates as the result
of convergence to different steady states, it cannot explain the persistence
of economic growth throughout most of the world. Persistent
growth is built into the neoclassical model in a way that is simple but not
terribly illuminating. The goal of much recent work in growth theory,
therefore, has been to develop models of persistent growth that avoid
the assumption of exogenous advances in technology. Hence, this work
goes by the name endogenous growth theory.
The Basic Model
The idea behind endogenous growth theory can be seen most easily
by considering the production function, Y = AK. This production function
has the property of constant returns to the accumulated factor. If
we double the amount of capital, we double the amount of output. To
see what this implies for economic growth, consider the accumulation
equation:
(20) K = s Y - S K . 

This equation, together with the Y = AK production function, implies 

YIY = KIK = sA - 6.
As long as sA > 6, income grows forever, even without the assumption
of exogenous technological progress.
Thus, a simple change in the production function can dramatically alter
the predictions about economic growth. In the neoclassical model,
saving leads to growth temporarily, but eventually the economy approaches
a steady state in which growth is independent of the saving
N . Gregory Mankiw 297
rate. By contrast, in this endogenous growth model, saving leads to
growth forever.
The endogenous growth model can be viewed as a limiting case of the
neoclassical model. As I showed above, the rate of convergence to the
steady state in the neoclassical model depends on the capital share. As
the capital share, a ,goes to one, the rate of convergence, h,goes to zero.
The basic endogenous growth model is, in essence, the limiting case of
the neoclassical model in which convergence is so slow as to be nonexistent.
In the endogenous growth model, the transitional dynamics of
the neoclassical model last forever.
This limiting case has two notable properties. First, differences in saving
rates across countries lead to increasingly large differences in income
over time. Second, large difference~in income are not associated
with differences in the return to capital. Thus, the world can contain
great disparities in income without any incentive for capital to move
from rich to poor countries.
In some sense, the endogenous growth model is not really novel.
Solow himself pointed out in his original article that if the production
function were not well behaved, his model might not have a steady
state.29Solow noted that, in this case, capital accumulation can lead to
perpetual growth. The innovation in endogenous growth theory is to
make this case canonical.
The question is, how do we interpret the variable Kin the production
function, Y = AK? If K is seen as including only the economy's stock of
plant and equipment, then it is natural to assume diminishing returns.
That is probably why Solow chose not to emphasize the possibility of
endogenous growth in his model. Yet if we interpret K more broadly,
then the assumption of constant returns to capital is more palatable.
Thus, the literature on endogenous growth has often relied on capital
with externalities and human capital when making the case for constant
returns.
The most appealing way of interpreting the endogenous growth model
is to view knowledge as a type of capital. It is clear that scientific discoveries
build on previous scientific discoveries. Knowledge is used to produce
knowledge. Compared to other forms of capital, the production of
knowledge seems less likely to exhibit diminishing returns. Indeed, as
Paul Romer and, more recently, Michael Kremer have emphasized, if
29. Solow (1956, p. 77).
298 Brookings Papers on Economic Activity, 1:I995
we look over very long spans of history it appears that growth has accelerated
somewhat over time.30The production of knowledge might even
exhibit increasing returns.
It is worthwhile to distinguish here between knowledge and human
capital. Although the two terms are sometimes treated as synonyms,
there is an important difference. Knowledge refers to society's understanding
about how the world works. Human capital refers to the resources
expended transmitting this understanding to the labor force. Put
crudely, knowledge is the quality of society's textbooks; human capital
is the amount of time that has been spent reading them. Even if the accumulation
of knowledge does not encounter diminishing returns, it seems
likely that the accumulation of human capital does. The best case for endogenous
growth, therefore, relies on knowledge rather than human
capital as the source of perpetual growth.
Bells and Whistles
The Y = AK model is the simplest example of endogenous growth,
but it does not do justice to the large literature on endogenous growth
theory. The literature has moved two steps further. The first step was to
develop models with more than one sector of production; for instance,
one sector might produce goods and services while another sector produces
innovations in technology. The second step was to develop
models that incorporate the microeconomic decisions behind the research
process. Both are steps in the direction of offering a more explicit
description of how economic growth arises from the accumulation of
knowledge.
To see what can be learned from models with more than one sector,
consider the following example of endogenous growth based on the
work of Hirofumi Uzawa, Robert Lucas, and Casey Mulligan and Xavier
Sala-i-Martir~.~'The economy has two sectors: manufacturing firms
and research universities. Firms produce goods and services, which are
used for consumption and capital accumulation. Universities produce a
factor of production called knowledge, which is then freely used in both
sectors. The economy is described by the production function for firms,
30. See Romer (1986)and Kremer (1993).
31. See Uzawa (1965), Lucas (1988),and Mulligan and Sala-i-Martin (1993).
N. Gregory Mankiw 299
the production function for universities, and the capital accumulation
equation:
where u is the fraction of the labor force in universities, His the stock of
knowledge, and g' 2 0. The rest of the notation is standard. As before,
the production function for the manufacturing firms is assumed to be homogeneous
of degree one in its two arguments.
This model is a cousin of the Y = AK model. In particular, this economy
exhibits constant returns to scale in the accumulated factors of production.
If we double the inputs of capital and knowledge, we double the
output of both sectors. Like the Y = AK model, this model can generate
perpetual growth without the assumption of exogenous shifts in the production
function.
This model is also a cousin of the neoclassical growth model. For any
constant u, knowledge, H, grows at the constant rate g , and the model
is exactly the one Solow analyzed in 1956.Thus this model has the transitional
dynamics of the neoclassical model. When the stock of capital is
low relative to the stock of knowledge (that is, when KIHL is below its
steady-state value), the economy's production of goods and services
grows more quickly. In contrast to the Y = AK model, this endogenous
growth model can explain conditional convergence.
There are two key decision variables in this model. As in the neoclassical
model, the fraction of output used for saving and investment determines
the stock of capital. In addition, the fraction of labor used in universities
determines the stock of knowledge. Both s and u affect the level
of income, although only u affects the long-run growth rate. Thus, this
model of endogenous growth takes a small step toward showing which
societal decisions determine the rate of technological change.32
Yet even this endogenous growth model tells only a rudimentary story
about the creation of knowledge. If one thinks about research for even a
32. For a discussion of decisionmaking and dynamics in this kind of endogenous
growth model, see Barro and Sala-i-Martin (199.5,chap. 5).
Brookings Papers on Economic Activity, 1:I995300
moment, three facts become apparent. First, even though knowledge is
largely a public good, much research is done in firms that are driven by
the profit motive. Second, research is profitable because innovations
give firms temporary monopolies, either because of the patent system or
because there is an advantage to being first. Third, when one firm innovates,
other firms build on that innovation in order to produce the next
generation of innovations. These (essentially microeconomic) facts are
not easily connected with the (essentially macroeconomic) growth
models discussed so far.
Some recent work on endogenous growth is aimed at incorporating
these facts about research and development by melding the theory of
monopolistic competition into the theory of growth.33One virtue of this
class of endogenous growth models is that it offers a detailed description
of the process of innovation. In linking technological change and market
power, these models harken back to themes that Joseph Schumpeter
emphasized many years ago.34The recent formalization is valuable both
for clarifying these old ideas and for providing a framework in which to
examine the welfare properties of equilibria.
Whither Endogenous Growth?
Like many theories, the theory of endogenous growth has its place but
has been oversold by its advocates. Its value is twofold. First, it helps
explain the existence of worldwide technological progress, which the
neoclassical growth model takes as given. Second, it offers a more realistic
description of research and development.
Yet for practical macroeconomists trying to understand international
differences, the payoff from endogenous growth theory is not clear.
Models that emphasize unmeasurable variables such as knowledge are
hard to bring to the data. It is not surprising, therefore, that these models
have appealed to more theoretically inclined economists, and that there
have been few attempts to evaluate these models empirically.
Even though knowledge is undeniably important for economic
growth, theories of the creation of knowledge may be of little help in explaining
international differences in growth rates. Knowledge, as op-
33. For example, see Romer (1990), Grossman and Helpman (1991), and Aghion and
Howitt (1992).
34. Schumpeter (1942).
N. Gregory Mankiw 301
posed to capital, travels around the world fairly quickly. State-of-the-art
textbooks are available in the poorest countries. Even when a firm has
some monopoly power over an innovation, this lasts only a short time,
after which the innovation becomes a worldwide public good. For understanding
international experience, the best assumption may be that
all countries have access to the same pool of knowledge, but differ by
the degree to which they take advantage of this knowledge by investing
in physical and human capital.
It is ultimately an empirical issue whether capital or knowledge is
more important in explaining international differences in economic
growth. In a recent paper, Alwyn Young sheds light on this question by
examining in detail the spectacular growth of Hong Kong, Singapore,
South Korea, and Taiwan.35From 1966to 1990income per person grew
by more than 7percent per year in each of these countries. Young shows
that this exceptional growth can be traced to large increases in measured
factor inputs: labor-force participation, educational attainment, and the
capital stock. Growth in total factor productivity-which measures the
rate of improvement over time in the production function-was not unusually
high. For understanding these growth miracles, endogenous
growth theory appears unnecessary.
Empirical Studies of Economic Growth
One hallmark of recent work on economic growth, in contrast to the
work of the 1950sand 1960s, is its empirical emphasis. Largely because
of the important work of Summers and Heston, international data suitable
for cross-sectional analysis are now available for most countries.36
These data have allowed systematic examination of the differences between
economies that have experienced rapid growth and those that
have not.
Some Findings
The typical empirical paper on economic growth chooses a sample of
countries and then runs a cross-sectional regression. On the left-hand
35. Young(1995).
36. See Summers and Heston (1991).
302 Brookings Papers on Economic Activity, 1:I995
side is each country's average growth rate over a long period. On the
right-hand side is a set of variables expected to determine that growth
rate. The variables change from study to study, and the interpretations
of the results differ, but the basic setup is usually pretty much the
Although it is impossible to summarize this vast literature, some of
the findings are worth noting:
-A low initial level of income is associated with a high subsequent
growth rate when other variables are held constant. This is the finding of
conditional convergence, discussed earlier.
-The share of output allocated to investment is positively associated
with growth.
-Various measures of human capital, such as enrollment rates in primary
and secondary schools, are positively associated with growth.
-Population growth (orfertility) is negatively associated with growth
in income per person.
-Political instability, as measured by the frequency of revolutions,
coups, or wars, is negatively associated with growth.
-Countries with more distorted markets, as measured by the black
market premium on foreign exchange or other impediments to trade,
tend to have lower growth rates.
-Countries with better developed financial markets, as measured,
for instance, by the size of liquid assets relative to income, tend to have
higher growth rates.
Each of these findings has been confirmed independently in several
studies. Most of them depend to some extent on which other variables
are included in the regression. This sensitivity to specificationis not surprising
in light of the serious multicollinearity problem (which I discuss
below).
It is easy to reconcile these results with existing theories of growth. In
terms of the neoclassical model, these regressions show a tendency for
convergence toward a steady state that is determined by the control
variables. And, as discussed above, endogenous growth models with
37. For a sampling from the huge literature estimating cross-country growth regressions,
see Kormendi and Meguire (1985), Dowrick and Nguyen (1989), Barro (1991), De
Long and Summers (1991, 1992a, 1992b, 1993), Murphy, Shleifer, and Vishny (1991), Fischer
(1991, 1993),Mankiw, Romer, and Wei1(1992), Levine and Renelt (1992), and King
and Levine (1993a, 1993b).
N . Gregory Mankiw 303
more than one sector are consistent with a similar interpretation. Even
the signs on the control variables make sense. What economist doubts
that growth is fostered by high investment, widespread education, low
population growth, political stability, free markets, and well-developed
financial institutions?
Yet despite their success in uncovering some interesting and interpretable
regularities, cross-country growth regressions are not unassailable
as a tool for learning about the determinants of growth. Three problems
in particular affect this entire literature.
The Simultaneity Problem
The most obvious problem with cross-country growth regressions is
simultaneity-the fact that the right-hand-side variables are not exogenous,
but are jointly determined with the growth rate. For example,
there is a strong, positive correlation between investment and growth.
But does this imply that high investment causes high growth, that high
growth causes high investment, or that some third variable causes both
high investment and high growth? The same question can be posed for
population growth, human capital, political stability, and all the other
variables commonly used in growth regressions.
The standard econometric solution to simultaneity is to follow the advice
of the Cowles Commission: find exogenous variables to use as instruments.
The problem is that there are few, if any, such variables in
cross-country data sets.
When looking for instruments, it is easy to fall prey to temptation.
Some economists are tempted to treat political variables, such as the
number of revolutions and coups, as if they were exogenous. But political
scientists are tempted to call economic growth exogenous and put
political instability on the left-hand side. Some economists are tempted
to treat lagged variables as exogenous. But a variable is not necessarily
exogenous just because it is re determined.^^ For empirical work on
economic growth, the advice of the Cowles Commission does not seem
feasible.
38. Anyone who thinks that predetermined variables are naturally good instruments
should ponder the following problem: A person regresses the quantity of apples on the
price of apples, instrumenting with yesterday's price of apples. Has he identified the supply
curve or the demand curve for apples? (The answer, in most cases, is neither.)
304 Brookings Papers on Economic Activity, 1:I995
Where does this conclusion leave us? Very simply, with a bunch of
correlations among important endogenous variables. The cross-country
data can never establish, for instance, the direction of causality between
investment and growth. Yet the correlation between investment and
growth is an interesting and useful fact. It says that any explanation of
international experience must be consistent with these two variables
moving strongly together. In other words, correlations among endogenous
variables can rule out theories that fail to produce the correlations,
and they can thereby raise our confidence in theories that do produce
them, but these correlations can never establish causality beyond a reasonable
doubt.
The Multicollinearity Problem
Another problem with interpreting cross-country regressions is multicollinearity-the
strong correlation among all the right-hand-side variables.
In some ways, multicollinearity is more a pernicious problem than
simultaneity, for its effects are less obvious.
There is no doubt that multicollinearity is severe in cross-country regressions.
High-growth countries have higher rates of investment,
higher enrollment in primary schools, higher enrollment in secondary
schools, higher rates of literacy, lower rates of population growth, more
developed financial markets, and fewer revolutions and coups than lowgrowth
economies. As a rough approximation, those countries that do
things right do most things right, and those countries that do things
wrong do most things wrong.
At first, multicollinearity might not seem like a problem, for multiple
regression is supposed to deal with it automatically. Under standard
econometric assumptions, multicollinearity reduces the precision with
which coefficients are estimated, but it does not bias the coefficients or
standard errors. Thus, it might appear that multicollinearity does not
contaminate the inferences drawn from cross-country regressions.
Yet there are two reasons to doubt that multiple regression does in
fact deal well with multicollinearity. The first reason is that multiple regression
treats each country as if it were an independent observation.
For the reported standard errors to be correct, the residual for Canada
must be uncorrelated with the residual for the United States. If country
residuals are in fact correlated, as is plausible, then the data most likely
contain less information than the reported standard errors indicate. That
N.Gregory Mankiw 305
is, statistical significance is overstated. The reported standard errors,
therefore, cannot be relied upon to warn us about the problem of multi-
collinearity.39
The second reason to doubt that multiple regression deals well with
multicollinearity is the prevalence of measurement error in international
data sets. Everyone who uses cross-country data admits that many variables
are crude proxies at best. Yet the implications of this fact are rarely
acknowledged. It is a standard econometric result that measurement error
tends to bias downward the coefficient on the variable measured with
error. At the same time, it can bias upward the coefficients on variables
correlated with the variable measured with error, for these other variables
stand in for the error-prone variable. Hence, when growth rates
are regressed on a group of crude proxy variables that exhibit substantial
multicollinearity, the set of coefficients largely reflects the differing
measurement errors in the right-hand-side variables.
Let us consider a specific example. In a series of provocative papers,
Bradford De Long and Lawrence Summers argue that the social return
to equipment investment is very high.40This finding is based largely on
cross-country growth regressions with equipment investment among
the control variables. From a large coefficient on equipment investment,
they conclude that equipment investment yields substantial externalities,
and that government policy should target this form of capital accu-
mulation.
De Long and Summers also find in their sample that secondary school
enrollment has an insignificant negative coefficient, but they choose not
to emphasize this result. They note that secondary school enrollment is
"not a very good proxy" for investment in human capital. They go on to
explain that "it is premature to conclude that education is not important:
education almost surely is important. Instead, the lack of significance of
our human capital investment proxies in our cross-national regressions
should most likely be attributed to the large divergence between measured
schooling and actual skills learned."41
This interpretation makes sense. But do the implications of measure-
39. De Long and Summers (1991)tried to address the possibility of spatial correlation
between the residuals in cross-country growth regressions and were surprised that they
found little evidence of it. More recently, Elliot (1993)has found such evidence. He reports
that standard errors corrected for spatial correlation are substantially larger than the uncorrected
standard errors usually used.
40. De Long and Summers (1991, 1992a, 1992b, 1993).
41. De Long and Summers (1992b, pp. 116-17).
Brookings Papers on Economic Activity, 1:I995306
ment error stop at the schooling coefficient? Almost certainly not. If investment
in human capital is important but hard to measure, then it is
possible that equipment investment is also serving as a proxy for investment
in human capital. Because equipment requires workers to operate
it, economies with highly skilled workers may attract more equipment
investment than economies with less skilled workers. Thus, it is hard to
know whether the coefficient on equipment investment reflects a truly
large rate of return, or merely the combination of measurement error
and multicollinearity.
I am not picking on De Long and Summers because their work is particularly
egregious. Quite the opposite is true. These authors are keenly
aware of the pitfalls in interpreting cross-country growth regressions
and they do more than most to try to avoid them. But they, like others
running these regressions, are limited by the available data.
The Degrees-of-Freedorn Problem
One goal of empirical work on economic growth is to establish the
conditions that are associated with rapid growth. Is the key to growth a
high rate of equipment investment? Political stability? Well-developed
financial institutions? Stable money growth? Low inflation? More engineers
than lawyers? Or some variable yet to be inve~tigated?~~
There is no limit to the possibilities-and that itself raises one more
problem. There are only about one hundred countries on which to run
cross-country growth regressions. As a matter of econometric logic, one
hundred observations can estimate only one hundred coefficients (and
far fewer than that with any degree of accuracy). There are too few degrees
of freedom to answer all the questions being asked
Of course, no individual study ever includes one hundred variables on
the right-hand side of a cross-country growth regression. Yet including
only a subset of variables does not help matters much. Itjust means that
the results of the study are contingent upon what variables the study
chooses to exclude.
There is no easy solution to this degrees-of-freedom problem. In the
end, there appears to be little choice but to admit the limitations of cross-
42. See, respectively, De Long and Summers (1991, 1992a, 1992b, 1993),Barro (1991),
King and Levine (1993a, 1993b), Kormendi and Meguire (1985), Fischer (1991, 1993),and
Murphy, Shleifer, and Vishny (1991).
N.Gregory Mankiw: 307
country growth regressions. It is not that we have to stop asking so many
questions about economic growth. We just have to stop expecting the
international data to give us all the answer^.^'
The Future of Growtlz Empirics
Although I applaud the empirical emphasis in recent work on economic
growth, I am not sanguine about the future of this work. There
are two problems working in concert: the subtlety of the theories and the
limitations of the data.
When the field of economic growth was revived about a decade ago,
some people argued that the international data offered a way of rejecting
some theories in favor of others. The neoclassical model predicts convergence,
whereas the Y = AK endogenous growth model does not.
Since large samples of countries exhibit little evidence of convergence,
the data seemed to favor the endogenous growth model.
In actuality, things are not so simple. If different countries are allowed
to have different steady states, then the neoclassical model predicts conditional
convergence, which is found in the data. Conditional convergence,
however, is also consistent with more sophisticated endogenous
growth models that exhibit some form of transitional dynamics. Once
we admit that growth theories have subtle implications, we are forced to
conclude that cross-country regressions cannot easily distinguish
among them.
Using these regressions to decide how to foster growth is also most
likely a hopeless task. Simultaneity, multicollinearity, and limited degrees
of freedom are important practical problems for anyone trying to
draw inferences from international data. Policymakers who want to promote
growth would not go far wrong ignoring most of the vast literature
43. One approach to the degrees-of-freedom problem is to move from cross-sectional
to panel data. See, for example, Islam (1995) and Barro and Sala-i-Martin (1995, chap. 12).
Including more frequent observations on each country increases the number of degrees of
freedom. Yet the amount of information being added is not obvious, for the new observations
are not independent of the old ones. Moreover, this information comes at the cost
of introducing another problem: the business cycle. Distinguishing growth effects from
business-cycle effects is difficult, for many of the determinants of long-run growth (such
as investment) fluctuate strongly over the business cycle. Given the traditional view that
short-run fluctuations and long-run growth are fundamentally different phenomena, it is
not at all apparent that the benefit of panel data is worth the cost.
308 Brookings Papers on Economic Activity, 1:1995
reporting growth regressions. Basic theory, shrewd observation, and
common sense are surely more reliable guides for policy.
Where Do We Stand?
The purpose of economic theory is to take a complicated world, abstract
from many details, and express the key economic relationships in
a way that enhances understanding. From this standpoint,the neoclassical
model is still the most useful theory of growth we have. It will continue
to be the first growth model taught to students and the first growth
model used by policy analysts.
The modern interpretation of this model, however, differs from that
of twenty years ago. There is an increasing consensus that the role of
capital in economic growth should be interpreted more broadly. Either
because capital accumulation conveys positive externalities, or because
labor income is largely the return to human capital, or for both reasons,
the capital share is likely much greater than the traditional estimate of
one-third derived from the national income accounts. If the neoclassical
model is to explain the international experience, a capital share of at
least two-thirds seems necessary.
Recent work in endogenous growth has shown how to model technological
progress, which is exogenous in the neoclassical growth model.
Although endogenous growth models are often presented as alternatives
to the neoclassical model, they can also be viewed as complements. Endogenous
growth models provide a plausible description of worldwide
advances in knowledge. The neoclassical growth model takes worldwide
technological advances as given and provides a plausible description
of international differences.
If one accepts the neoclassical conclusion that the accumulation of
capital (broadly interpreted) is the key to international differences in
economic growth, one is naturally led to ask why some countries save
and invest so much more than others. Little progress has been made in
answering this important question. Recent work on economic growth
has emphasized innovations in the production side of the model, appending
standard theories of the household to explain the saving decision.
Much of the growth literature continues to rely on the most basic
representative-consumer model to explain saving, even though the con-
N. Gregory Mankiw 309
sumption literature has rejected this model of household behavior. At
this point, advances in the study of consumption and saving may do
more to further understanding of economic growth than advances in
growth theory as conventionally practiced.
The implications of recent work on economic growth for policymakers
are far from clear. Obviously, if capital accumulation is the key to
growth, then accumulating capital more rapidly will raise the growth
rate. Just as obviously, political stability will raise growth by encouraging
investment both by domestic residents and from abroad. Yet some
recent work on economic growth suggests that a more activist government
could be beneficial. If certain types of capital yield large positive
externalities, then policymakers should try to direct resources in this direction.
The problem is that economists have not produced a persuasive
way of measuring the magnitude of these externalities. Relying on estimates
from cross-country regressions (or on thejudgments of the political
process) will likely lead to haphazard policy, which is surely worse
than no policy at all. Without a solution to this measurement problem,
modern growth theory does not offer any clear policy prescriptions. In
my view, policymakers who want to foster economic growth would do
well to heed the first rule for physicians: do no harm. This may seem like
a modest conclusion from an ambitious literature. But sometimes modesty
is all that economists have a right to offer.
A P P E N D I X
The Rate of Convergence to the Steady State
THISAPPENDIX derives the rate of convergence to the steady state in the
Solow version of the neoclassical growth model. Begin with the capital
accumulation equation:
(-41) dkldt = sf(k) - (n + g + 6)k.
Now take the first-order Taylor expansion of the right-hand side of this
equation around the steady-state capital stock, k*. This yields
(-42) dkldt = [sf'(k*) - (n + g + 6)](k - k*).
310 Brookings Papers on Economic Activity, 1:I995
Substitute for s using the steady-state condition sf(k*) = (n + g + 6)k*
This gives
If capital earns its marginal product, then f'(k*)k*if(k*) is the steadystate
capital share, a. Therefore,
(-44) dkidt = - A(k - k*),
where
To show that income converges to its steady-state level at the same rate
as capital, note that y = f(k). Therefore, the following first-order approximations
hold:
(A6) y - Y * = f'(k*)(k - k*).
By substitution,
(A7) dyidt = - AO, - y*).
Thus, income converges to its steady state at rate A. Notice that this result
does not depend on any specific functional form for the production
function,f(k).
In interpreting the neoclassical growth model, some authors have asserted
that convergence follows from the property of diminishing marginal
product. This interpretation is not correct, at least for this most basic
version of the model. The second derivative of the production
function,f',measures the extent of diminishing marginal product; this
plays no role in determining the rate of convergence, A. It is more accurate
to say that convergence arises here from the property of diminishing
average ,:roduct. As the capital stock rises, depreciation rises proportionately,
but income and investment rise less than proportionately;
thus, a higher level of capital leads to a lower growth rate of capital.
Comments
and Discussion
Edmund S. Phelps: This paper gives some noteworthy support to the
wisdom that the phenomenon of poor countries in the world can be explained
by the importance of human capital. Having little human capital,
they invest little in tangible capital. If virtually all income is attributable
to this capital, human and nonhuman, a dearth of human capital has a
devastating effect on national potential output. Moreover, the importance
of human capital (on top of the already measured importance of
nonhuman capital) in the production function is a serious drag on the
speed with which a low-income economy can climb to its steady-state
income level; and it adds mightily to the sensitivity of the steady-state
income level to a country's saving-to-income ratio.
Becoming very bold, Mankiw adds that not much of the difference in
income from country to country is to be accounted for by disparities in
technological knowledge. Textbooks, blueprints, and chemical formulas
travel fast over the world, evidently at low transmission costs. So in
the author's view, it is ultimately human capital that holds back a country-though
he agrees with the recent econometric studies findingthat a
country can boost its growth path with "political stability, free markets,
and well-developed financial [intermediaries]."
For me this paper is a model of exposition and is destined to have a
long life. Yet I see places in which Mankiw's analysis badly needs to be
corrected or supplemented. One of these is his apparent conception of
the function of human capital, defined as the resource cost incurred in
teaching and learning the world's stock of knowledge. He appears to see
it entirely as a factor of production, analogous to the stocks of tangible
capital. Thus he posits an elasticity of true national income with respect
to that human stock, to be measured by labor's share of true income,
31 1
312 Brookings Papers on Economic Activity, 1:I995
that is entirely analogous to the capital elasticity of output. This is a very
static and deterministic view.
The alternative view is that all or most persons in the labor force could
forget everything they had learned beyond the ninth grade, say, without
putting much of a dent in today's output. First of all, most schooling is
learning how to learn-which fosters the ability to understand a description
of an innovative technique, such as a new tool or a new chemical,
or understand legislation setting out regulations or prohibitions affecting
some industrial activity. Thus education facilitates the adoption and dissemination
of technical advances and, more generally, the exploitation
of market opportunities. This theme is developed in a paper by Richard
Nelson and myself.' Second, much of our learning is precautionary, and
seemingly redundant, because we do not know when we are young, and
our opportunity costs are low, whatjob or sequence ofjobs will be most
in demand over our working life. The best econometric evidence for this
view so far is the finding that the stock of human capital contributes negatively
to a country's productivity level but positively to the rate of improvement
in its producti~ity.~Certainly in communist Eastern Europe,
where the demand for innovation was weak, having massive human capital
did not appear to help much at all. There is also microeconomic evidence
in longitudinal studies finding that entrepreneurs show outside returns
from additional e d ~ c a t i o n . ~
One consequence of this alternative model concerns the importance
of human capital. The disadvantage posed by having low human capital,
this model says, is that it impedes the ability to implement promptly and
widely the successive advances in the best-practice technology. The
country is always behind the curve. Thus we may assume that a country
is farther behind the best-practice frontier, the lower is its human capital
and the faster that the frontier is advancing. If only technological progress
would stop, the returns on much of the human capital would drop,
and the countries disadvantaged in human capital could then converge
to the frontier (perhaps very nearly at Mankiw's original 4 percent rate).
The alternative view also has implications for the demand for human
capital. Why is it that several countries have, in only a few short decades,
experienced a rapid accumulation of human capital-the Asian
1. Nelson and Phelps (1966).
2. Benhabib and Spiegel(1994).
3. Evans and Leighton (1989).
N. Gregory Mnnkiw 313
miracle economies-while othei-countries at about the same place in the
poverty ranking have not? Surely the answer is the emergence of entrepreneurship,
encouraged and sanctioned by the government. To an important
degree, I suggest, a dearth of human capital is found in those
countries where there is a low demand for it, its reward being meager
because the entrepreneurs who might introduce best-practice techniques
and enter new markets are not permitted or emboldened to do so
by the government and the prevailing economic philosophy.
This observation leads to another area where I feel the paper falls a bit
short. A student in Budapest or Moscow could come away from this paper
with the impression that, although noninterference with exchange
rates and other niceties of liberal economic policy may be rather important
for a government to observe, whether ownership and control of
industry is primarily capitalist or socialist should not be a vital consideration,ina
country's strategy for a high rate of growth. The c-word barely
appears in the paper. Unless I missed it, the postwar experience of the
socialist countries in Eastern Europe, Africa, and Asia is not considered
significant.
If the new wave of research on economic growth is to graduate to a
really useful endeavor, it has to introduce the factors that have become
prominent in discussions of the road back to capitalism in eastern Europe:
tax rates on enterprise profits and payrolls, the size of the public
enterprise sector, red tape and corruption in the government's licenses
and contracts to the private sector, impediments to shareowners' exercise
of enterprise control, and various other property rights.
This gap in the paper is frustrating. We Western economists should be
sending messages to countries where corporate ownership and control
are now crucial issues. Yet if this paper is a guide, present-day mainstream
research on economic growth is cut off from the searching analysis
of the crucial contributions of key capitalist institutions for economic
growth that has been touched off by the events in eastern Europe in the
1990s. However, it is not too late to start filling this gap.
Paul M. Romer: Greg Mankiw and I agree on many issues concerning
growth, but it will be more useful if I focus here on the areas in which we
disagree. Our most obvious disagreement is apparently over a statement
of fact. Mankiw argues that technology is a public good that is available
everywhere in the world. I argue that there is ample evidence that this
314 Brookings Papers on Economic Activity, 1:I995
assertion is wrong. But our disagreement here is not really about the
facts. Mankiw's position is not that his claim is literally true, but that
it is close enough for macroeconomics. What constitutes close enough
depends on what one is trying to accomplish-getting the answers right
or catering to a target audience.
Our differing positions on the nature of technology are derived from
a more basic disagreement about strategies for constructing macroeconomic
models of growth. Mankiw believes that the neoclassical model
built on the foundation of the public-good assumption is so useful in the
classroom and in policy debates that the burden of proof should rest on
those who support a richer model. Even in the face of strong evidence
against the public-good model, he would apparently be reluctant to consider
an extension. I believe that an unnecessary reliance on this neoclassical
model has hampered clear thinking about growth, particularly
among macroeconomists and the students and policymakers who listen
to them. Even in the absence of strong evidence against this model, we
must explore an extended model that forces us to think more carefully
about the economics of technology and knowledge.
The differences between our modeling strategies may themselves be
the result of different beliefs about the ultimate objectives that economists
should pursue. I believe that our fundamental goals are, first, to
uncover important truths, and then, to communicate them to outsiders.
The order in this two-step process is important. We should start by using
observation and logic to decide what those truths are, without thinking
ahead to the reception that awaits our findings. Once our results are in
hand, we should communicate them to the relevant outsiders, without
catering or condescending to them.
From this point of view, it follows almost immediately that we should
work with an extended theoretical framework that lets us take technology
seriously. It costs little to adopt an extended model because these
kinds of models have a mathematical structure that is only slightly more
complicated than that of the public-good model. On the benefit side,
technological change is an extremely important force in modern economic
life, one that we would surely like to understand better. The extended
model forces us to be precise in our reasoning about intangible
inputs like technology, and it encourages us to adopt a broader perspective
when we look at the evidence concerning growth. Because the pub-
N. Gregory Mankiw 315
lic-good model of technology can be nested as a special case of this extended
model, working with the extension keeps all of our intellectual
options open.
There is, however, another set of beliefs about what it is that economists
should do. According to this view, we should think ahead to the
reaction of our audience when we engage in research. I am frequently
warned that the models I use, and the results I describe, could be used
to justify bad government policies. The implication is that economists
should filter their results, keeping in mind how they might be used in the
political process. From this point of view, a model that takes technology
seriously poses risks that are not present in the public-good model. The
public-good model used by Mankiw implies that the optimal government
policy (at least for a small developing country) is laissez faire. As a result
it is unlikely to provide support for the wrong kinds of policies.
A similar strategic calculation could presumably apply to the reaction
that a more sophisticated treatment of technology would provoke
among students. Precisely because it does not try to capture any of the
subtle issues that arise when we treat technology and knowledge as economic
goods, the public-good model is familiar and unthreatening to the
median student. If we plan with this student's reaction in mind, the intellectual
power of a broader perspective is a disadvantage rather than an
advantage. It raises new issues, some of which are not yet resolved. Any
discussion of these issues will inevitably leave many loose ends. If our
strategy in doing research is to cater to the demands of a textbook market
that values simplicity,familiarity, and decisive answers over all else,
a model that treats technology seriously may indeed be something to
avoid.
It is within this context that the balance of my comments must be
placed. I will point to empirical failures of the public-good model of technology.
Many of the points I raise are not new. Jan Fagerberg provides
auseful discussion of the history of objections to this approach to modeling
growth.' As the persistence of this debate suggests, a discussion of
the evidence by itself is unlikely to resolve the differences of opinion on
what is a good model of growth. Whether the problems noted below are
minor issues that a theory of growth can skip over or whether they are
1. Fagerberg (1994).
Brookings Papers on Economic Activity, 1:I995316
decisive evidence against the public-good model may depend entirely on
one's views about what the goal of growth theory, or economics more
generally, should be.
The recent history of the public-good model of cross-country differences
in wages and income is a story of strategic retreat. This kind of
model gets the signs right for many questions about growth, but careful
examination eventually shows that it fails to explain the magnitudes observed
in the data. As Mankiw explains in this paper, and as I have argued
elsewhere, the first stage in this retreat came with the recognition
that a model of the form Y, = AKj*Ljl-* cannot explain the cross-country
data for values of the parameter a that are close to capital's share in
total i n ~ o m e . ~
This finding provoked two different responses. The first was to allow
for the possibility that the technology parameter, A, could vary across
countries. Early versions of the endogenous growth models let A vary
because of spillover effects from investment in physical capital or human
capital. More recent models have proposed more complicated
mechanisms for producing variation in A, such as research and development,
or trade in intermediate inputs in production. But whatever the
cause of the differences in technology, these models attribute an important
part of the cross-country variation in wages and incomes to variation
in the technology used in different countries.
The other response was to leave A the same in all countries and to add
an additional input, Hj (for human capital), that covaries with K,. The
message of Mankiw's paper with David Romer and David Weil is that
this, by itself, is e n ~ u g h . ~There is no need to consider the possibility
that the technology might also vary across countries. As they show, a
model of the form Y, = AKj113H,1'3Lj1'3can be made to fit the cross-country
data. But as Mankiw recognizes in subsequent work and reiterates in
this paper, there are important quantitative problems with this model as
well, when one looks beyond the national income accounts data used in
the cross-country regression^.^ If H and K covary across countries, the
rate of return to physical capital will be much higher in poor countries
than in rich countries.
The next retreat from the neoclassical strategy of treating each nation
2. See Romer (1994).
3. Mankiw, Romer, and Weil(1992).
4. See Barro, Mankiw, and Sala-i-Martin (1994).
N . Gregory Mankiw 317
as a closed economy with the same public-good technology is therefore
to allow the rate of return on physical capital to be equalized across
countries through a process of international borrowing and lending. The
relative scarcity of H can then be used to explain why total income and
wages for unskilled workers are both lower in poor countries.
As Mankiw observes, this amended version of the Mankiw, Romer,
and Weil model gets the signs right. The rate of return to investments
in human capital does seem to be higher in poor countries, just as this
approach predicts. What he fails to note is that the implied magnitudes
are wildly inconsistent with the available evidence. Using the baseline
model with exponents of one-third on the three main inputs, and
allowing for free mobility of physical capital, an easy calculation leads
to the following simple result. In the poorest countries, where the wage
for unskilled labor is one-tenth the wage for unskilled labor in the United
States, the wage for skilled labor will be ten times larger than the wage
for skilled labor in the United States. Thus if the ratio of the skilled wage
to the unskilled wage in the United States is two, the ratio of the skilled
wage to the unskilled wage in the poor country will be two hundred! Because
the cost of education is the forgone unskilled wage, and the return
to education is the differential between the skilled and the unskilled
wage, the implied rate of return to education in poor countries should be
larger than the return in the United States by a factor of one hundred,
rather than by the factor of two or three that is found in the data.5Moreover,
as Robert Lucas has emphasized, we can also use evidence about
migration to test our models of g r ~ w t h . ~Here the public-good model that
Mankiw proposes does not even get the signs right. The net flow of
skilled workers is from poor countries to rich countries, rather than from
rich to poor.
The new fallback position for the neoclassical model that Mankiw introduces
in this paper is to suggest that the elasticity of substitution between
capital and labor could be four or ten, instead of the value of one
implied by a Cobb-Douglas specification. He raises this possibility only
in the context of a model with two factors of production and leaves the
exploration of the model with three factors of production for future research.
I will interpret his suggestion by treating one of the two inputs in
5. See Psacharopoulos (1985) for a description of the empirical results on rates of
return.
6. Lucas (1988).
318 Brookings Papers on Economic Activity, 1:I995
a CES production function as unskilled labor and letting the other input
be a composite of physical capital and human capital. In this setting a
higher elasticity of substitution helps the model fit the data along some
dimensions, but it hurts in others. As Mankiw emphasizes, a higher elasticity
of substitution can lead to large differences in income per capita
without inducing large differences in rates of return to physical and human
capital. This reduces the amount by which the wage for the scarce
skilled workers in poor countries exceeds the wage for the abundant
skilled workers in rich countries. But this change also leads to reductions
in the predicted difference between the wages for unskilled workers
in rich and poor countries.
To get an order of magnitude estimate of the effect that this change in
the elasticity of substitution can have on wages for the unskilled workers,
let us accept the rough estimate that the share of total income accruing
to physical capital and human capital in the United States is about
0.8. Then we can calibrate a CES production function with an elasticity
of substitution of four between unskilled labor and the composite of human
and physical capital. This implies that the wage for unskilled workers
in a country that has zero human and zero physical capital is about
60 percent of the wage for unskilled workers in the United States. This
fraction is far too high to be consistent with the evidence on cross-country
variation in wages for low-skilled workers.
This kind of result should come as no surprise. Mankiw justifies his
high elasticity of substitution by invoking the arguments that lead to factor
price equalization. In the limit, where the elasticity of substitution is
infinite, wages for unskilled workers will be the same all over the world,
regardless of the local stock of human and physical capital. The point for
the purpose of this discussion is that an elasticity of four goes a long way
toward infinity.
The basic conclusion that emerges from this account is simple. The
neoclassical assumption that the aggregate level of technology is the
same in all countries is inconsistent even with the macroeconomic data
on growth and development. Fitting the public-good model of technology
to the these data is like squeezing a balloon. You can make it smaller
in one place, but problems always pop out somewhere else.
The case against the public-good model becomes much stronger
when one looks at the microeconomic evidence. Formal comparisons of
productivity levels routinely uncover wide variation among firms in the
N. Gregory Mankiw 319
same manufacturing industry. Even in a service industry such as retailing,
firms such as K-Mart and Wal-Mart use very different technologies
to provide their service, with very different outcomes in terms of profitability
and returns on equity. These persistent differences are difficult to
explain if the technology that each uses is a public good.
Furthermore, even a cursory look at the details of the development
experience suggest that the process of technology transfer by foreign
firms has been important in many countries. This process of transfer is
also responsive to the incentives created by the host government. For
example, when Mauritius pursued the traditional policy of erecting high
tariff barriers to encourage import-substituting local manufacturing, its
only exports were in agriculture. Once it had created an export processing
zone that let foreign firms earn profits by making use of local labor,
garment assembly firms from Hong Kong located production there,
and exports of garments from Mauritius to the United States and Europe
grew dramatically. The garment assembly industry did not exist prior to
the creation of the export processing zone in 1970. By 1990almost onethird
of all employment on the island was in this industry.
The impediment to the development of a garment assembly industry
on Mauritius before 1970was not a level of savings that was too low to
finance the purchase of sewing machines. Nor was it a level of education
too low for workers to be able to operate such machines. The problem
was that the relevant technology was not a freely available public good.
Until the foreign entrepreneurs arrived, no one in Mauritius knew
enough about the garment business to begin production there. This
knowledge did not leak in from Hong Kong. It was brought in when entrepreneurs
were presented with an economic environment that let them
earn a profit on the knowledge that they pos~essed.~If the public-good
model does not apply to an industry as basic as garment assembly,
where could it apply?
We have overwhelming evidence that technology is not a public
good. We also have formal models of growth that let us take account of
this fact. The puzzle for me is why many economists still resist doing so
in their teaching and in their research. The only conjecture that I can offer
is the one outlined above. These economists may be paying too much
attention to how a particular model will be received and used by out-
7. See Romer (1993) for more details concerning this case
320 Brookings Papers on Economic Activity, 1:1995
siders, and too little attention to what they think is true. Mankiw may be
right that the neoclassical model "will continue to be the first growth
model taught to students and the first growth model used by policy analysts."
Nevertheless, as economists, we should not settle for this. Our
goal should be to make them have second thoughts about a question that
is as interesting and as important as the one addressed here: What
causes growth and development?
General Discussion
There was a lively discussion of the relative importance to growth of
conventional inputs, like physical capital, human capital, and labor, and
intangible factors such as knowledge. Several participants criticized the
paper's assumption that knowledge is identical across countries. James
Duesenberry argued that the process by which modern techniques are
mastered is more complicated than simply sending people to school and
handing them blueprints when they graduate. It typically involves many
kinds of learning, including experience and interaction with foreigners.
Robert Gordon emphasized the importance of organizational capital,
citing Paul Romer's example of Mauritius. If organizationalcapital were
not important, the management consultant industry would not exist.
And Nordhaus noted that technologies differ in their rates of diffusion
across space and time; he felt that we do not know very much about the
process. Barry Bosworth and John Haltiwanger noted that the variability
of total factor productivity (TFP) supports these observations. Even
after human capital is accounted for, there is tremendous variation in
TFP across countries. In fact within narrowly defined industries in the
United States itself there are large productivity differences across
plants. Moreover, these differentials are persistent; the most productive
plants in 1995were generally the most productive in 1985,and often also
in 1975. William Brainard added that productivity differentials exist
even within plants; different assembly lines are often of different vintages,
for example.
Severalparticipants thought that the paper gave short shrift to institutions
and government policy. Duesenberry and William Branson found
it curious that the study of "growth" emphasized theoretical models, national
production functions, and cross-country regressions but failed to
embrace the study of "development," which emphasizes the development
of product, labor, and financial markets and the many ways in
N. Gregory Mankiw 321
which the government can affect efficiency;for example, overvalued exchange
rates, parallel markets, subsidies, rationing, state enterprises,
corruption, volatile inflation, interest rate regulation, and capital allocation.
Several members of the panel presented evidence on the importance
of government policy. Nordhaus suggested that the narrow concentration
on factor inputs as the source of growth was rejected by what
he called the socialist experiment. Although Eastern European countries
were relatively well endowed with physical and human capital after
World War 11, this just seemed to help them go downhill. According to
research by Fred Bergsten, in the U.S.S.R., Hungary. Poland, and Yugoslavia
output per worker in 1975 was 30 percent lower than in the
Western countries, after accounting for differences in capital and land
per worker and adjusting for labor quality. Bosworth noted that, while
socialist economies clearly underperformed capitalist economies, it has
been more difficult to demonstrate the benefits of liberalizing measures
in economies that have pursued a middle road. For instance, he interpreted
the evidence as indicating that trade promotion has been a better
strategy than trade liberalization.
Jeffrey Frankel thought that government policy toward trade and
openness was worthy of special comment. Although human and physical
capital explain much of the variation in GDP, trade explains a significant
portion of the residual. The positive correlation between growth
and openness is a robust finding of cross-country studies, and is confirmed
in the paper by Sachs and Warner in this volume. Frankel also
argued that causation has convincingly been shown to run from trade to
growth, as there now exist relatively good instrumental variables for
trade. One of the channels through which trade aids growth is by facilitating
the transfer of technology.
Haltiwanger noted that an important component of productivity
growth is the reallocation of resources to the more productive plants,
implying that policies interfering with resource mobility can have significant
effects. Bosworth pointed out that some countries have had negative
TFP growth for periods lasting many years, a fact that the neoclassical
model cannot account for. In addition to government policy and
technological diffusion, Bosworth emphasized the importance of macroeconomic
stability, reminding the Panel of Arthur Okun's dictum that
"one recession can wipe out a thousand Harberger triangles." He noted
that a number of countries show negative TFP growth over certain periods.
These data points, which represent the loss of precious ground
322 Brookings Pupers on Economic Activity,1:1995
gained, can be attributed largely to economic crises. One secret of success
of the Asian economies has been their ability to avoid periods of
declining output.
While agreeing that institutions and technological diffusion cannot be
ignored as sources of growth, and that the neoclassical model sheds no
light on the effects of science policy, trade policy, the socialist experiment,
and the like, Mankiw asserted that doing so was not its purpose.
He defended the model as being good at what it is supposed to do -explain
differences in standards of living across countries and time. A regression
with per capita income levels on the left and saving rates, population
growth rates, and human capital variables on the right has an R2
around 0.78. He noted that the neoclassical model seems to work well in
explaining the experience of the newly industrializing countries (NICs)
in Asia. Alwyn Young's research indicates that the NICs grew mainly
through the accumulation of physical and human capital, rather than increases
in TFP.
Gordon echoed Romer's comments about wage differentials. He
wondered how it is that an individual migrating to the United States with
the same human capital as in his home country can work with about the
same physical capital as in his home country and raise his standard of
living by a factor of ten. He reasoned that to explain this, one needs to
allow for complementarity among factors of production. Mankiw suggested
that the model's difficulty in explaining facts like these reflects an
oversimplified production function. He believed that human capital is
fundamentally different from physical capital, implying the need for a
production function that explicitly included both.
Benjamin Friedman noted that the correlation between saving rates
and income per capita is poor. Many high-income countries have low
saving rates, and vice versa; a notable comparison is the low-saving
United States with the high-saving China. But he was less pessimistic
than Mankiw about the accuracy of the neoclassical growth model in
predicting rates of convergence. The estimated rate is about 0.02, while
the predicted rate is 0.04; given the usual downward bias in regression
coefficient estimates, and the inherent difficulty of fitting models to
facts, Friedman thought the model actually gets pretty close.
There was debate about whether human capital helps to explain much
of the variation in economic growth. Frankel and Bosworth both noted
that adding human capital to growth regressions significantly improves
their explanatory power. Branson mentioned research by Krueger in the
N. Gregory Manki~v 323
1960s that estimated that close to 60 percent of the difference between
developed and less developed countries is attributable to human capital.
However, Nordhaus said that, using Denison's technique to measure
the impact of education on productivity growth and income growth, one
finds that it explains very little of the differences across countries. He
preferred Denison's approach for its "internal consistency ." Richard
Cooper suggested that comparison of national capital-to-output ratios
provides evidence on this point. The rank correlation between per capita
income and capital stock per unit of output is nearly perfect. The only
outlier whose capital-to-output ratio is lower than expected is the United
States, suggesting more efficient use of capital. He dismissed human
capital as afull explanation because, in that case, American human capital
per worker would need to be way out of line with that of other rich
OECD countries, including Germany, the United Kingdom, France,
and Japan.
Gordon expressed skepticism about the importance of investment externalities
in some models of endogenous growth. In particular, he questioned
the plausibility of the De Long and Summers thesis that equipment
investment plays a special role. In the United States from 1936to
the present, the ratio of equipment to structures has increased steadily
from 111 to 311. There has been little correlation between this ratio and
growth, as the increasing importance of equipment has continued unabated
during both fast and slow periods. Nordhaus stated that although
the evidence is clear on externalities to R&D, he knew of no evidence
for the existence of returns to physical capital that are not captured by
the firm.
Some discussants offered suggestions for future empirical research.
Duesenberry thought that researchers working with cross-country data
should pay more attention to important events at the microeconomic
level. By using microeconomic data, he thought it possible to avoid
some of the identification problems that plague macroeconomic studies.
Nordhaus cautioned that researchers defining human capital as an input
need to account for it on the output side also; they need to take care to
count production of human capital not as consumption, but as investment.
He thought most empirical studies get the accounting of human
capital wrong, with Dale Jorgenson's recent work a noteworthy exception.
Friedman suggested that it was important to be explicit in empirical
research about whether economies are to be treated as if they are at their
steady states.
Brookirlgs Papers on Economic Activity, 1:I995324
References
Abel, Andrew B., and others. 1989."Assessing Dynamic Efficiency: Theory and
Evidence." Review ofEconomic Studies 56 (1): 1-19.
Aghion, Philippe, and Peter Howitt. 1992. "A Model of Growth Through Creative
Destruction." Econometrica 60(2): 323-51.
Barro, Robert J. 1991. "Economic Growth in a Cross Section of Countries."
Quarterly Journal of Economics 106(2):407-43.
Barro, Robert J., N. Gregory Mankiw, and Xavier Sala-i-Martin. 1995. "Capital
Mobility in Neoclassical Models of Growth." American Economic Review
85(1): 103-15.
Barro, Robert J., and Xavier Sala-i-Martin. 1991. "Convergence across States
and Regions." BPEA, 1:1991, 107-58.
. 1995.Economic Growth, New York: McGraw-Hill.
Benhabib, Jess, and Mark M. Spiegel. 1994."The Role of Human Capital in Economic
Development: Evidence from Aggregate Cross-Country Data." Journal
of Monetary Economics 34(2):143-73.
Campbell, John Y., and N. Gregory Mankiw. 1989. "Consumption, Income,
and Interest Rates: Reinterpreting the Time-Series Evidence." In NBER
Macroeconomics Annual 1989, edited by Olivier Jean Blanchard and Stanley
Fischer. Cambridge, Mass.: MIT Press.
Cass, David. 1965."Optimum Growth in an Aggregative Model of Capital Accumulation."
Review of Economic Studies 32(3): 233-40.
De Long, J. Bradford, and Lawrence H. Summers. 1991. "Equipment Investment
and Economic Growth." Quarter'ly Journal of Economics 106(2):
445-502.
. 1992a. "Equipment Investment and Economic Growth: How Strong Is
the Nexus?" BPEA, 2:1992, 157-99.
. 1992b. "Macroeconomic Policy and Long-Run Growth." In Policiesfor
Long-Run Economic Growth. Kansas City, Mo.: The Federal Reserve Bank
of Kansas City.
. 1993. "How Strongly Do Developing Countries Benefit from Equipment
Investment?" Journal of Monetary Economics 32(3): 295-415.
Diamond, Peter A. 1965. "National Debt in a Neoclassical Growth Model."
American Economic Review 55(5): 1126-50.
Dowrick, Steve, and Duc-Tho Nguyen. 1989. "OECD Comparative Economic
Growth 1950-85: Catch-Up and Convergence." American Ecorzomic Review
79(5): 1010-30.
Elliot, Graham. 1993. "Spatial Correlation and Cross Country Regressions."
Unpublished paper. Harvard University.
Evans, David S., and Linda S. Leighton. 1989. "Some Empirical Aspects of Entrepreneurship."
American Economic Review 79(3): 519-35.
Fagerberg, Jan. 1994. "Technology and International Differences in Growth
Rates." Journal of Economic Literature 32(3): 1147-75.
N. Gregory Mankiw 325
Fischer, Stanley. 1991. "Growth, Macroeconomics, and Development." In
NBER Macroeconomics Annual 1991, edited by Olivier Jean Blanchard and
Stanley Fischer. Cambridge, Mass.: MIT Press.
. 1993. "The Role of Macroeconomic Factors in Growth." Journal of
Monetary Economics 32(3): 485-512.
Grossman, Gene M., and Elhanan Helpman. 1991. Innovation and Growth in
the Global Economy. Cambridge, Mass.: MIT Press.
. 1994. "Endogenous Innovation in the Theory of Growth." Journal of
Economic Perspectives 8(1): 23-44.
Islam, Nazrul. 1995. "Growth Empirics: A Panel Data Approach." Quarterly
Journal of Economics 1lO(4) (forthcoming).
King, Robert G., and Ross Levine. 1993a. "Finance and Growth: Schumpeter
Might Be Right." Quarterly Journal of Economics 108(3): 717-37.
. 1993b. "Finance, Entrepreneurship, and Growth: Theory and Evidence."
Journal of Monetary Economics 32(3): 513-42.
Koopmans, Tjalling C. 1965. "On the Concept of Optimal Economic Growth."
In Study Week on the Econometric Approach to Development Planning.
Amsterdam: North-Holland.
Kormendi, Roger C., and Philip G. Meguire. 1985. "Macroeconomic Determinants
of Growth: Cross-Country Evidence." Journal of Monetary Economics
16(2): 141-63.
Kotlikoff, Laurence J., and Lawrence H. Summers. 1981. "The Role of Intergenerational
Transfers in Aggregate Capital Accumulation." Journal ofPoliticnl
Economy 89(4): 706-32.
Kremer, Michael. 1993. "Population Growth and Technological Change: One
Million B.C. to 1990." Quarterly Journal ofEconomics 108(3): 681-716.
Levine, Ross, and David Renelt. 1992. "A Sensitivity Analysis of Cross-Country
Growth Regressions." American Economic Review 82(4): 942-63.
Lucas, Robert E., Jr. 1988. "On the Mechanics of Economic Development."
Journal of Monetary Economics 22(1): 3-42.
. 1990. "Why Doesn't Capital Flow from Rich to Poor Countries?" American
Econornic Review, Papers and Proceedings 80(2): 92-96.
Mankiw, N. Gregory, David Romer, and David N. Weil. 1992. "A Contribution
to the Empirics of Economic Growth." Q~rarterlyJournal of Economics
107(2): 407-37.
Mulligan, Casey B., and Xavier Sala-i-Martin. 1993. "Transitional Dynamics in
Two-Sector Models of Endogenous Growth." Quarterly Journal ofEconomics
108(3):739-73.
Murphy, Kevin M., Andrei Shleifer, and Robert W. Vishny. 1991. "The Allocation
of Talent: Implications for Growth." Quarterly Journal of Economics
106(2): 503-30.
Nelson, Richard R., and Edmund S. Phelps. 1966. "Investment in Humans,
Technological Diffusion and Economic Growth." American Economic Review,Papers
and Proceedings 56(2): 69-75.
Brookings Papers on Economic Activity, 1:I995326
Phelps, Edmund. 1961. "The Golden Rule of Accumulation: A Fable for
Growthmen." American Economic Review 51(4): 638-43.
Psacharopoulos, George. 1985. "Returns to Education: A Further International
Update and Implications." Journal of Human Resources 20(4): 583-604.
Ramsey, F. P. 1928. "A Mathematical Theory of Saving." Economic Journal 38
(152): 543-59.
Romer, Paul M. 1986. "Increasing Returns and Long-Run Growth." Journal of
Political Economy 94(5): 1002-37.
. 1987. "Crazy Explanations for the Productivity Slowdown." In NBER
Macroeconomics Annual 1987, edited by Stanley Fischer. Cambridge,
Mass.: MIT Press.
. 1990. "Endogenous Technological Change." Journal of Political Economy
98(5, part 2): S71-S102.
. 1993. "Two Strategies for Economic Development: Using Ideas and
Producing Ideas." In Proceedings of the World Bank Annual Conference on
Development Economics 1992, edited by Lawrence H. Summers and
Shekhar Shah. Washington: World Bank.
. 1994. "The Origins of Endogenous Growth." Journal of Economic Perspectives
8(1): 3-22.
Samuelson, Paul A. 1958. "An Exact Consumption-Loan Model of Interest with
or without the Social Contrivance of Money." Journal of Political Economy
66(6): 467-82.
Schumpeter, Joseph A. 1942. Capitalism, Socialism, and Democracy. New
York: Harper.
Solow, Robert M. 1956. "A Contribution to the Theory of Economic Growth."
Quarterly Journal of Economics 70(1): 65-94.
Summers, Robert, and Alan Heston. 1991. "The Penn World Table (Mark 5): An
Expanded Set of International Comparisons, 1950-1988." Quarterly Journal
of Economics 106(2):327-68.
Uzawa, Hirofumi. 1965. "Optimum Technical Change in an Aggregative Model
of Economic Growth." International Economic Review 6(1): 18-3 1.
Ventura, Jaume. 1994. "Growth and Interdependence." Unpublished paper.
Harvard University (November).
Young, Alwyn. 1995."The Tyranny of Numbers: Confronting the Statistical Realities
of the East Asian Growth Experience." Quarterly Journal of Econonzics
110(3) (forthcoming).
You have printed the following article:
The Growth of Nations
N. Gregory Mankiw; Edmund S. Phelps; Paul M. Romer
Brookings Papers on Economic Activity, Vol. 1995, No. 1, 25th Anniversary Issue. (1995), pp.
275-326.
Stable URL:
http://links.jstor.org/sici?sici=0007-2303%281995%291995%3A1%3C275%3ATGON%3E2.0.CO%3B2-Y
This article references the following linked citations. If you are trying to access articles from an
off-campus location, you may be required to first logon via your library web site to access JSTOR. Please
visit your library's website or contact a librarian to learn about options for remote access to JSTOR.
[Footnotes]
2
A Contribution to the Theory of Economic Growth
Robert M. Solow
The Quarterly Journal of Economics, Vol. 70, No. 1. (Feb., 1956), pp. 65-94.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28195602%2970%3A1%3C65%3AACTTTO%3E2.0.CO%3B2-M
3
A Contribution to the Empirics of Economic Growth
N. Gregory Mankiw; David Romer; David N. Weil
The Quarterly Journal of Economics, Vol. 107, No. 2. (May, 1992), pp. 407-437.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199205%29107%3A2%3C407%3AACTTEO%3E2.0.CO%3B2-5
5
An Exact Consumption-Loan Model of Interest with or without the Social Contrivance of
Money
Paul A. Samuelson
The Journal of Political Economy, Vol. 66, No. 6. (Dec., 1958), pp. 467-482.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28195812%2966%3A6%3C467%3AAECMOI%3E2.0.CO%3B2-Z
http://www.jstor.org
LINKED CITATIONS
- Page 1 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
5
National Debt in a Neoclassical Growth Model
Peter A. Diamond
The American Economic Review, Vol. 55, No. 5, Part 1. (Dec., 1965), pp. 1126-1150.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28196512%2955%3A5%3C1126%3ANDIANG%3E2.0.CO%3B2-L
6
A Mathematical Theory of Saving
F. P. Ramsey
The Economic Journal, Vol. 38, No. 152. (Dec., 1928), pp. 543-559.
Stable URL:
http://links.jstor.org/sici?sici=0013-0133%28192812%2938%3A152%3C543%3AAMTOS%3E2.0.CO%3B2-R
6
Optimum Growth in an Aggregative Model of Capital Accumulation
David Cass
The Review of Economic Studies, Vol. 32, No. 3. (Jul., 1965), pp. 233-240.
Stable URL:
http://links.jstor.org/sici?sici=0034-6527%28196507%2932%3A3%3C233%3AOGIAAM%3E2.0.CO%3B2-H
7
The Golden Rule of Accumulation: A Fable for Growthmen
Edmund Phelps
The American Economic Review, Vol. 51, No. 4. (Sep., 1961), pp. 638-643.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28196109%2951%3A4%3C638%3ATGROAA%3E2.0.CO%3B2-I
8
Assessing Dynamic Efficiency: Theory and Evidence
Andrew B. Abel; N. Gregory Mankiw; Lawrence H. Summers; Richard J. Zeckhauser
The Review of Economic Studies, Vol. 56, No. 1. (Jan., 1989), pp. 1-19.
Stable URL:
http://links.jstor.org/sici?sici=0034-6527%28198901%2956%3A1%3C1%3AADETAE%3E2.0.CO%3B2-K
9
The Role of Intergenerational Transfers in Aggregate Capital Accumulation
Laurence J. Kotlikoff; Lawrence H. Summers
The Journal of Political Economy, Vol. 89, No. 4. (Aug., 1981), pp. 706-732.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28198108%2989%3A4%3C706%3ATROITI%3E2.0.CO%3B2-J
http://www.jstor.org
LINKED CITATIONS
- Page 2 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
11
Endogenous Innovation in the Theory of Growth
Gene M. Grossman; Elhanan Helpman
The Journal of Economic Perspectives, Vol. 8, No. 1. (Winter, 1994), pp. 23-44.
Stable URL:
http://links.jstor.org/sici?sici=0895-3309%28199424%298%3A1%3C23%3AEIITTO%3E2.0.CO%3B2-R
12
A Contribution to the Empirics of Economic Growth
N. Gregory Mankiw; David Romer; David N. Weil
The Quarterly Journal of Economics, Vol. 107, No. 2. (May, 1992), pp. 407-437.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199205%29107%3A2%3C407%3AACTTEO%3E2.0.CO%3B2-5
14
OECD Comparative Economic Growth 1950-85: Catch-Up and Convergence
Steve Dowrick; Duc-Tho Nguyen
The American Economic Review, Vol. 79, No. 5. (Dec., 1989), pp. 1010-1030.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28198912%2979%3A5%3C1010%3AOCEG1C%3E2.0.CO%3B2-T
15
The Penn World Table (Mark 5): An Expanded Set of International Comparisons,
1950-1988
Robert Summers; Alan Heston
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 327-368.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C327%3ATPWT%285%3E2.0.CO%3B2-D
16
Economic Growth in a Cross Section of Countries
Robert J. Barro
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 407-443.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C407%3AEGIACS%3E2.0.CO%3B2-C
16
A Contribution to the Empirics of Economic Growth
N. Gregory Mankiw; David Romer; David N. Weil
The Quarterly Journal of Economics, Vol. 107, No. 2. (May, 1992), pp. 407-437.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199205%29107%3A2%3C407%3AACTTEO%3E2.0.CO%3B2-5
http://www.jstor.org
LINKED CITATIONS
- Page 3 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
16
A Sensitivity Analysis of Cross-Country Growth Regressions
Ross Levine; David Renelt
The American Economic Review, Vol. 82, No. 4. (Sep., 1992), pp. 942-963.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199209%2982%3A4%3C942%3AASAOCG%3E2.0.CO%3B2-J
18
A Contribution to the Empirics of Economic Growth
N. Gregory Mankiw; David Romer; David N. Weil
The Quarterly Journal of Economics, Vol. 107, No. 2. (May, 1992), pp. 407-437.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199205%29107%3A2%3C407%3AACTTEO%3E2.0.CO%3B2-5
24
Increasing Returns and Long-Run Growth
Paul M. Romer
The Journal of Political Economy, Vol. 94, No. 5. (Oct., 1986), pp. 1002-1037.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28198610%2994%3A5%3C1002%3AIRALG%3E2.0.CO%3B2-C
25
Returns to Education: A Further International Update and Implications
George Psacharopoulos
The Journal of Human Resources, Vol. 20, No. 4. (Autumn, 1985), pp. 583-604.
Stable URL:
http://links.jstor.org/sici?sici=0022-166X%28198523%2920%3A4%3C583%3ARTEAFI%3E2.0.CO%3B2-0
26
Returns to Education: A Further International Update and Implications
George Psacharopoulos
The Journal of Human Resources, Vol. 20, No. 4. (Autumn, 1985), pp. 583-604.
Stable URL:
http://links.jstor.org/sici?sici=0022-166X%28198523%2920%3A4%3C583%3ARTEAFI%3E2.0.CO%3B2-0
27
Capital Mobility in Neoclassical Models of Growth
Robert J. Barro; N. Gregory Mankiw; Xavier Sala-I-Martin
The American Economic Review, Vol. 85, No. 1. (Mar., 1995), pp. 103-115.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199503%2985%3A1%3C103%3ACMINMO%3E2.0.CO%3B2-O
http://www.jstor.org
LINKED CITATIONS
- Page 4 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
28
A Contribution to the Empirics of Economic Growth
N. Gregory Mankiw; David Romer; David N. Weil
The Quarterly Journal of Economics, Vol. 107, No. 2. (May, 1992), pp. 407-437.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199205%29107%3A2%3C407%3AACTTEO%3E2.0.CO%3B2-5
29
A Contribution to the Theory of Economic Growth
Robert M. Solow
The Quarterly Journal of Economics, Vol. 70, No. 1. (Feb., 1956), pp. 65-94.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28195602%2970%3A1%3C65%3AACTTTO%3E2.0.CO%3B2-M
30
Increasing Returns and Long-Run Growth
Paul M. Romer
The Journal of Political Economy, Vol. 94, No. 5. (Oct., 1986), pp. 1002-1037.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28198610%2994%3A5%3C1002%3AIRALG%3E2.0.CO%3B2-C
30
Population Growth and Technological Change: One Million B.C. to 1990
Michael Kremer
The Quarterly Journal of Economics, Vol. 108, No. 3. (Aug., 1993), pp. 681-716.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199308%29108%3A3%3C681%3APGATCO%3E2.0.CO%3B2-A
31
Optimum Technical Change in An Aggregative Model of Economic Growth
Hirofumi Uzawa
International Economic Review, Vol. 6, No. 1. (Jan., 1965), pp. 18-31.
Stable URL:
http://links.jstor.org/sici?sici=0020-6598%28196501%296%3A1%3C18%3AOTCIAA%3E2.0.CO%3B2-Y
31
Transitional Dynamics in Two-Sector Models of Endogenous Growth
Casey B. Mulligan; Xavier Sala-I-Martin
The Quarterly Journal of Economics, Vol. 108, No. 3. (Aug., 1993), pp. 739-773.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199308%29108%3A3%3C739%3ATDITMO%3E2.0.CO%3B2-3
http://www.jstor.org
LINKED CITATIONS
- Page 5 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
33
Endogenous Technological Change
Paul M. Romer
The Journal of Political Economy, Vol. 98, No. 5, Part 2: The Problem of Development: A
Conference of the Institute for the Study of Free Enterprise Systems. (Oct., 1990), pp. S71-S102.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28199010%2998%3A5%3CS71%3AETC%3E2.0.CO%3B2-8
33
A Model of Growth Through Creative Destruction
Philippe Aghion; Peter Howitt
Econometrica, Vol. 60, No. 2. (Mar., 1992), pp. 323-351.
Stable URL:
http://links.jstor.org/sici?sici=0012-9682%28199203%2960%3A2%3C323%3AAMOGTC%3E2.0.CO%3B2-%23
35
The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth
Experience
Alwyn Young
The Quarterly Journal of Economics, Vol. 110, No. 3. (Aug., 1995), pp. 641-680.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199508%29110%3A3%3C641%3ATTONCT%3E2.0.CO%3B2-8
36
The Penn World Table (Mark 5): An Expanded Set of International Comparisons,
1950-1988
Robert Summers; Alan Heston
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 327-368.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C327%3ATPWT%285%3E2.0.CO%3B2-D
37
OECD Comparative Economic Growth 1950-85: Catch-Up and Convergence
Steve Dowrick; Duc-Tho Nguyen
The American Economic Review, Vol. 79, No. 5. (Dec., 1989), pp. 1010-1030.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28198912%2979%3A5%3C1010%3AOCEG1C%3E2.0.CO%3B2-T
37
Economic Growth in a Cross Section of Countries
Robert J. Barro
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 407-443.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C407%3AEGIACS%3E2.0.CO%3B2-C
http://www.jstor.org
LINKED CITATIONS
- Page 6 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
37
Equipment Investment and Economic Growth
J. Bradford De Long; Lawrence H. Summers
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 445-502.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C445%3AEIAEG%3E2.0.CO%3B2-A
37
The Allocation of Talent: Implications for Growth
Kevin M. Murphy; Andrei Shleifer; Robert W. Vishny
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 503-530.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C503%3ATAOTIF%3E2.0.CO%3B2-P
37
A Contribution to the Empirics of Economic Growth
N. Gregory Mankiw; David Romer; David N. Weil
The Quarterly Journal of Economics, Vol. 107, No. 2. (May, 1992), pp. 407-437.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199205%29107%3A2%3C407%3AACTTEO%3E2.0.CO%3B2-5
37
A Sensitivity Analysis of Cross-Country Growth Regressions
Ross Levine; David Renelt
The American Economic Review, Vol. 82, No. 4. (Sep., 1992), pp. 942-963.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199209%2982%3A4%3C942%3AASAOCG%3E2.0.CO%3B2-J
37
Finance and Growth: Schumpeter Might be Right
Robert G. King; Ross Levine
The Quarterly Journal of Economics, Vol. 108, No. 3. (Aug., 1993), pp. 717-737.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199308%29108%3A3%3C717%3AFAGSMB%3E2.0.CO%3B2-4
39
Equipment Investment and Economic Growth
J. Bradford De Long; Lawrence H. Summers
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 445-502.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C445%3AEIAEG%3E2.0.CO%3B2-A
http://www.jstor.org
LINKED CITATIONS
- Page 7 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
40
Equipment Investment and Economic Growth
J. Bradford De Long; Lawrence H. Summers
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 445-502.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C445%3AEIAEG%3E2.0.CO%3B2-A
42
Equipment Investment and Economic Growth
J. Bradford De Long; Lawrence H. Summers
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 445-502.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C445%3AEIAEG%3E2.0.CO%3B2-A
42
Economic Growth in a Cross Section of Countries
Robert J. Barro
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 407-443.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C407%3AEGIACS%3E2.0.CO%3B2-C
42
Finance and Growth: Schumpeter Might be Right
Robert G. King; Ross Levine
The Quarterly Journal of Economics, Vol. 108, No. 3. (Aug., 1993), pp. 717-737.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199308%29108%3A3%3C717%3AFAGSMB%3E2.0.CO%3B2-4
42
The Allocation of Talent: Implications for Growth
Kevin M. Murphy; Andrei Shleifer; Robert W. Vishny
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 503-530.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C503%3ATAOTIF%3E2.0.CO%3B2-P
43
Growth Empirics: A Panel Data Approach
Nazrul Islam
The Quarterly Journal of Economics, Vol. 110, No. 4. (Nov., 1995), pp. 1127-1170.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199511%29110%3A4%3C1127%3AGEAPDA%3E2.0.CO%3B2-2
http://www.jstor.org
LINKED CITATIONS
- Page 8 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
3
Some Empirical Aspects of Entrepreneurship
David S. Evans; Linda S. Leighton
The American Economic Review, Vol. 79, No. 3. (Jun., 1989), pp. 519-535.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28198906%2979%3A3%3C519%3ASEAOE%3E2.0.CO%3B2-Y
1
Technology and International Differences in Growth Rates
Jan Fagerberg
Journal of Economic Literature, Vol. 32, No. 3. (Sep., 1994), pp. 1147-1175.
Stable URL:
http://links.jstor.org/sici?sici=0022-0515%28199409%2932%3A3%3C1147%3ATAIDIG%3E2.0.CO%3B2-9
2
The Origins of Endogenous Growth
Paul M. Romer
The Journal of Economic Perspectives, Vol. 8, No. 1. (Winter, 1994), pp. 3-22.
Stable URL:
http://links.jstor.org/sici?sici=0895-3309%28199424%298%3A1%3C3%3ATOOEG%3E2.0.CO%3B2-H
3
A Contribution to the Empirics of Economic Growth
N. Gregory Mankiw; David Romer; David N. Weil
The Quarterly Journal of Economics, Vol. 107, No. 2. (May, 1992), pp. 407-437.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199205%29107%3A2%3C407%3AACTTEO%3E2.0.CO%3B2-5
5
Returns to Education: A Further International Update and Implications
George Psacharopoulos
The Journal of Human Resources, Vol. 20, No. 4. (Autumn, 1985), pp. 583-604.
Stable URL:
http://links.jstor.org/sici?sici=0022-166X%28198523%2920%3A4%3C583%3ARTEAFI%3E2.0.CO%3B2-0
References
http://www.jstor.org
LINKED CITATIONS
- Page 9 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
Assessing Dynamic Efficiency: Theory and Evidence
Andrew B. Abel; N. Gregory Mankiw; Lawrence H. Summers; Richard J. Zeckhauser
The Review of Economic Studies, Vol. 56, No. 1. (Jan., 1989), pp. 1-19.
Stable URL:
http://links.jstor.org/sici?sici=0034-6527%28198901%2956%3A1%3C1%3AADETAE%3E2.0.CO%3B2-K
A Model of Growth Through Creative Destruction
Philippe Aghion; Peter Howitt
Econometrica, Vol. 60, No. 2. (Mar., 1992), pp. 323-351.
Stable URL:
http://links.jstor.org/sici?sici=0012-9682%28199203%2960%3A2%3C323%3AAMOGTC%3E2.0.CO%3B2-%23
Economic Growth in a Cross Section of Countries
Robert J. Barro
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 407-443.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C407%3AEGIACS%3E2.0.CO%3B2-C
Capital Mobility in Neoclassical Models of Growth
Robert J. Barro; N. Gregory Mankiw; Xavier Sala-I-Martin
The American Economic Review, Vol. 85, No. 1. (Mar., 1995), pp. 103-115.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199503%2985%3A1%3C103%3ACMINMO%3E2.0.CO%3B2-O
Optimum Growth in an Aggregative Model of Capital Accumulation
David Cass
The Review of Economic Studies, Vol. 32, No. 3. (Jul., 1965), pp. 233-240.
Stable URL:
http://links.jstor.org/sici?sici=0034-6527%28196507%2932%3A3%3C233%3AOGIAAM%3E2.0.CO%3B2-H
Equipment Investment and Economic Growth
J. Bradford De Long; Lawrence H. Summers
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 445-502.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C445%3AEIAEG%3E2.0.CO%3B2-A
http://www.jstor.org
LINKED CITATIONS
- Page 10 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
National Debt in a Neoclassical Growth Model
Peter A. Diamond
The American Economic Review, Vol. 55, No. 5, Part 1. (Dec., 1965), pp. 1126-1150.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28196512%2955%3A5%3C1126%3ANDIANG%3E2.0.CO%3B2-L
OECD Comparative Economic Growth 1950-85: Catch-Up and Convergence
Steve Dowrick; Duc-Tho Nguyen
The American Economic Review, Vol. 79, No. 5. (Dec., 1989), pp. 1010-1030.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28198912%2979%3A5%3C1010%3AOCEG1C%3E2.0.CO%3B2-T
Some Empirical Aspects of Entrepreneurship
David S. Evans; Linda S. Leighton
The American Economic Review, Vol. 79, No. 3. (Jun., 1989), pp. 519-535.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28198906%2979%3A3%3C519%3ASEAOE%3E2.0.CO%3B2-Y
Technology and International Differences in Growth Rates
Jan Fagerberg
Journal of Economic Literature, Vol. 32, No. 3. (Sep., 1994), pp. 1147-1175.
Stable URL:
http://links.jstor.org/sici?sici=0022-0515%28199409%2932%3A3%3C1147%3ATAIDIG%3E2.0.CO%3B2-9
Endogenous Innovation in the Theory of Growth
Gene M. Grossman; Elhanan Helpman
The Journal of Economic Perspectives, Vol. 8, No. 1. (Winter, 1994), pp. 23-44.
Stable URL:
http://links.jstor.org/sici?sici=0895-3309%28199424%298%3A1%3C23%3AEIITTO%3E2.0.CO%3B2-R
Growth Empirics: A Panel Data Approach
Nazrul Islam
The Quarterly Journal of Economics, Vol. 110, No. 4. (Nov., 1995), pp. 1127-1170.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199511%29110%3A4%3C1127%3AGEAPDA%3E2.0.CO%3B2-2
http://www.jstor.org
LINKED CITATIONS
- Page 11 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
Finance and Growth: Schumpeter Might be Right
Robert G. King; Ross Levine
The Quarterly Journal of Economics, Vol. 108, No. 3. (Aug., 1993), pp. 717-737.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199308%29108%3A3%3C717%3AFAGSMB%3E2.0.CO%3B2-4
The Role of Intergenerational Transfers in Aggregate Capital Accumulation
Laurence J. Kotlikoff; Lawrence H. Summers
The Journal of Political Economy, Vol. 89, No. 4. (Aug., 1981), pp. 706-732.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28198108%2989%3A4%3C706%3ATROITI%3E2.0.CO%3B2-J
Population Growth and Technological Change: One Million B.C. to 1990
Michael Kremer
The Quarterly Journal of Economics, Vol. 108, No. 3. (Aug., 1993), pp. 681-716.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199308%29108%3A3%3C681%3APGATCO%3E2.0.CO%3B2-A
A Sensitivity Analysis of Cross-Country Growth Regressions
Ross Levine; David Renelt
The American Economic Review, Vol. 82, No. 4. (Sep., 1992), pp. 942-963.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28199209%2982%3A4%3C942%3AASAOCG%3E2.0.CO%3B2-J
A Contribution to the Empirics of Economic Growth
N. Gregory Mankiw; David Romer; David N. Weil
The Quarterly Journal of Economics, Vol. 107, No. 2. (May, 1992), pp. 407-437.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199205%29107%3A2%3C407%3AACTTEO%3E2.0.CO%3B2-5
Transitional Dynamics in Two-Sector Models of Endogenous Growth
Casey B. Mulligan; Xavier Sala-I-Martin
The Quarterly Journal of Economics, Vol. 108, No. 3. (Aug., 1993), pp. 739-773.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199308%29108%3A3%3C739%3ATDITMO%3E2.0.CO%3B2-3
http://www.jstor.org
LINKED CITATIONS
- Page 12 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
The Allocation of Talent: Implications for Growth
Kevin M. Murphy; Andrei Shleifer; Robert W. Vishny
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 503-530.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C503%3ATAOTIF%3E2.0.CO%3B2-P
The Golden Rule of Accumulation: A Fable for Growthmen
Edmund Phelps
The American Economic Review, Vol. 51, No. 4. (Sep., 1961), pp. 638-643.
Stable URL:
http://links.jstor.org/sici?sici=0002-8282%28196109%2951%3A4%3C638%3ATGROAA%3E2.0.CO%3B2-I
Returns to Education: A Further International Update and Implications
George Psacharopoulos
The Journal of Human Resources, Vol. 20, No. 4. (Autumn, 1985), pp. 583-604.
Stable URL:
http://links.jstor.org/sici?sici=0022-166X%28198523%2920%3A4%3C583%3ARTEAFI%3E2.0.CO%3B2-0
A Mathematical Theory of Saving
F. P. Ramsey
The Economic Journal, Vol. 38, No. 152. (Dec., 1928), pp. 543-559.
Stable URL:
http://links.jstor.org/sici?sici=0013-0133%28192812%2938%3A152%3C543%3AAMTOS%3E2.0.CO%3B2-R
Increasing Returns and Long-Run Growth
Paul M. Romer
The Journal of Political Economy, Vol. 94, No. 5. (Oct., 1986), pp. 1002-1037.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28198610%2994%3A5%3C1002%3AIRALG%3E2.0.CO%3B2-C
Endogenous Technological Change
Paul M. Romer
The Journal of Political Economy, Vol. 98, No. 5, Part 2: The Problem of Development: A
Conference of the Institute for the Study of Free Enterprise Systems. (Oct., 1990), pp. S71-S102.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28199010%2998%3A5%3CS71%3AETC%3E2.0.CO%3B2-8
http://www.jstor.org
LINKED CITATIONS
- Page 13 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.
The Origins of Endogenous Growth
Paul M. Romer
The Journal of Economic Perspectives, Vol. 8, No. 1. (Winter, 1994), pp. 3-22.
Stable URL:
http://links.jstor.org/sici?sici=0895-3309%28199424%298%3A1%3C3%3ATOOEG%3E2.0.CO%3B2-H
An Exact Consumption-Loan Model of Interest with or without the Social Contrivance of
Money
Paul A. Samuelson
The Journal of Political Economy, Vol. 66, No. 6. (Dec., 1958), pp. 467-482.
Stable URL:
http://links.jstor.org/sici?sici=0022-3808%28195812%2966%3A6%3C467%3AAECMOI%3E2.0.CO%3B2-Z
A Contribution to the Theory of Economic Growth
Robert M. Solow
The Quarterly Journal of Economics, Vol. 70, No. 1. (Feb., 1956), pp. 65-94.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28195602%2970%3A1%3C65%3AACTTTO%3E2.0.CO%3B2-M
The Penn World Table (Mark 5): An Expanded Set of International Comparisons, 1950-1988
Robert Summers; Alan Heston
The Quarterly Journal of Economics, Vol. 106, No. 2. (May, 1991), pp. 327-368.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199105%29106%3A2%3C327%3ATPWT%285%3E2.0.CO%3B2-D
Optimum Technical Change in An Aggregative Model of Economic Growth
Hirofumi Uzawa
International Economic Review, Vol. 6, No. 1. (Jan., 1965), pp. 18-31.
Stable URL:
http://links.jstor.org/sici?sici=0020-6598%28196501%296%3A1%3C18%3AOTCIAA%3E2.0.CO%3B2-Y
The Tyranny of Numbers: Confronting the Statistical Realities of the East Asian Growth
Experience
Alwyn Young
The Quarterly Journal of Economics, Vol. 110, No. 3. (Aug., 1995), pp. 641-680.
Stable URL:
http://links.jstor.org/sici?sici=0033-5533%28199508%29110%3A3%3C641%3ATTONCT%3E2.0.CO%3B2-8
http://www.jstor.org
LINKED CITATIONS
- Page 14 of 14 NOTE:
The reference numbering from the original has been maintained in this citation list.