PRL ch4

4
Theories of
Economic Growth
E
conomists have long puzzled over the question of economic growth. What is it
that makes some countries rich while others remain poor? Formal studies of
this question date back at least to the eighteenth-century writings of Scottish
social philosopher Adam Smith. The search for answers continues to dominate
economic thinking. In a lecture presented in 1985, Nobel Prize–winning economist
Robert Lucas noted the then-rapid economic growth of Indonesia and Egypt and
slow growth of India, famously asking,
Is there some action a government of India could take that would lead the Indian
economy to grow like Indonesia’s or Egypt’s? If so,what exactly? If not, what is
it about the “nature of India” that makes it so? The consequences for human
welfare involved in questions like these are simply staggering: Once one starts to
think about them it is hard to think of anything else.1
More than a quarter century later, economists Dani Rodrik, Arvind Subramanian,
and Francesco Trebbi noted,
Average income levels in the world’s richest and poorest nations differ by a factor
of more than 100. Sierra Leone, the poorest economy for which we have national
income statistics, has a per-capita GDP of $490, compared to Luxembourg’s
$50,061. What accounts for these differences, and what (if anything) can we do
1
Robert E. Lucas, “On the Mechanics of Economic Development,” Journal of Monetary Economics 22,
no. 1 (July 1988), 3–42.
89
90 [CH. 4] THEORIES OF ECONOMIC GROWTH
to reduce them? It is hard to think of any question in economics that is of greater

intellectual significance, or of greater relevance to the vast majority of the world’s
population.2
Two observations emerge from the juxtaposition of these strikingly similar

quotations: (1) Long-run economic growth may be the single most fundamental
determinant of human welfare around the world, and (2) despite substantial efforts
and significant progress toward solving the puzzle of economic growth during the
27 years that separate these comments, we remain far from a complete and policy-
relevant understanding of the deep determinants of growth.
This chapter explores key contributions to the theory of economic growth. We
began to explore these issues in the last chapter by examining some of the basic
processes and patterns that characterize economic growth in low-income coun-
tries. We emphasized that growth depends on two processes: the accumulation of
assets (such as capital, labor, and land), and making those assets more productive.
Saving and investment are central, but investments must be productive for growth
to proceed. Our approach was largely empirical, as we examined much of the data
on growth and some of the key findings from research on the determinants of
growth across countries. We saw that government policy, institutions, political and
economic stability, geography, natural resource endowments, and levels of health
and education all play some role in influencing economic growth. We emphasized
that growth is not the same as development, but it remains absolutely central to the
development process.
This chapter develops these ideas more formally by introducing the underly-
ing theory and the most important basic models of economic growth that influ-
ence development thinking today. These models provide consistent frameworks
for understanding the growth process and provide a theoretical foundation for the
empirical approach we took in the last chapter. Here, we identify specific mathemati-
cal relationships between the quantity of capital and labor, their productivity, and the
resulting aggregate output. It is important that these models also explore the process
of accumulating additional capital and labor and increasing their productivity, which
shifts the model from determining the level of output to the rate of change of output,
which of course is the rate of economic growth.
As we begin to examine the models, it is useful to consider the words of Robert
Solow, the father of modern growth theory, who once wrote: “All theory depends on
assumptions that are not quite true. That is what makes it theory. The art of success-
ful theorizing is to make the inevitable simplifying assumptions in such a way that
2
Dani Rodrik, Arvind Subramanian, and Francesco Trebbi, “Institutions Rule: The Primacy of Institu-
tions over Geography and Integration in Economic Development,” Journal of Economic Growth, 9, no. 2
(2004), 131–65.
THE BASIC GROWTH MODEL 91
the final results are not very sensitive.”3 The best models are simple, yet still manage
to communicate powerful insights into how the real world operates. In this spirit, the
models presented here make assumptions that clearly are not true but allow us to
simplify the framework and make it easier to grasp key concepts and insights. For
example, we begin by assuming that our prototype economy has one type of homo-
geneous worker and one type of capital good that combine to produce one standard
product. No economy in the world has characteristics even closely resembling these
assumptions, but making these assumptions allows us to cut through many details
and get to the core concepts of the theory of economic growth.
THE BASIC GROWTH MODEL
The most fundamental models of economic output and economic growth are
based on a small number of equations that relate saving, investment, and popula-
tion growth to the size of the workforce and capital stock and, in turn, to aggregate
production of a single good. These models initially focus on the levels of invest-
ment, labor, productivity, and output. It then becomes straightforward to examine
the changes in these variables. Our ultimate focus is to explore the key determinants
of the change in output—that is, on the rate of economic growth. The version of the
basic model that we examine here has five equations: an aggregate production func-
tion, an equation determining the level of saving, the saving-investment identity, a
statement relating new investment to changes in the capital stock, and an expression
for the growth rate of the labor force.4 We examine each of these in turn.
Standard growth models have at their core a production function. At the indi-
vidual firm or microeconomic level, a production function relates the number of
employees and machines to the size of the firm’s output. For example, the production
function for a textile factory would reveal how much more output the factory could
produce if it hired (say) 50 additional workers and purchased five more looms. Pro-
duction functions often are derived from engineering specifications that relate given
amounts of physical input to the amount of physical output that can be produced
with that input. At the national or economywide level, production functions describe
the relationship of the size of a country’s total labor force and the value of its capital
stock with the level of that country’s gross domestic product (GDP; its total output).
These economywide relationships are called aggregate production functions.
3
Robert Solow, “A Contribution to the Theory of Economic Growth,” Quarterly Journal of Economics 70
(February 1956), 65–94.
4
This five-equation presentation is based on teaching notes compiled by World Bank economist
Shantayanan Devarajan, to whom we are indebted.
Our first equation is an aggregate production function. If Y represents total

output (and therefore total income), K is the capital stock, and L is the labor supply,
at the most general level, the aggregate production function can be expressed as
Y = F (K, L) [4–1]
This expression indicates that output is a function (denoted by F) of the capi-

tal stock and the labor supply. As the capital stock and labor supply grow, output
expands. Economic growth occurs by increasing either the capital stock (through
new investment in factories, machinery, equipment, roads, and other infrastruc-
ture), the size of the labor force, or both. The exact form of the function F (stating
precisely how much output expands in response to changes in K and L) is what dis-
tinguishes many different models of growth, as we will see later in the chapter. The
other four equations of the model describe how these increases in K and L come
about.
Equations 4–2 through 4–4 are closely linked and together describe how the
capital stock (K) changes over time. These three equations first calculate total sav-
ing, then relate saving to new investment, and finally describe how new investment
changes the size of the capital stock. To calculate saving, we take the most straightfor-
ward approach and assume that saving is a fixed share of income:
S = sY [4–2]
In this equation, S represents the total value of saving, and s represents the average
saving rate. For example, if the average saving rate is 20 percent and total income is
$10 billion, then the value of saving in any year is $2 billion. We assume that the sav-
ing rate s is a constant, which for most countries is between 10 and 40 percent (typi-
cally averaging between 20 and 25 percent), although for some countries it can be
higher or lower. China’s savings rate in 2008 (along with those of several large oil
exporters) exceeded 50 percent, while several countries (including Mozambique,
Guinea, the Seychelles, and Georgia) reported savings rates less than 5 percent of
GDP. Actual saving behavior is more complex than this simple model suggests (as we
discuss in Chapter 10), but this formulation is sufficient for us to explore the basic
relationships between saving, investment, and growth.
The next equation relates total saving (S) to investment (I). In our model, with
only one good, there is no international trade (because everyone makes the same
product, there is no reason to trade). In a closed economy (one without trade or
foreign borrowing), saving must be equal to investment. All output of goods and
services produced by the economy must be used for either current consumption
or investment, while all income earned by households must be either consumed or
saved. Because output is equal to income, it follows that saving must equal invest-
ment. This relationship is expressed as
S = I [4–3]
THE BASIC GROWTH MODEL 93
We are now in a position to show how the capital stock changes over time. Two main
forces determine changes in the capital stock: new investment (which adds to the
capital stock) and depreciation (which slowly erodes the value of the existing capital
stock over time). Using the Greek letter delta (!) to represent the change in the value
of a variable, we express the change in the capital stock as !K, which is determined
as follows:
!K = I - (dK) [4–4]
In this expression d is the rate of depreciation. The first term (I) indicates that the
capital stock increases each year by the amount of new investment. The second term
-(d * K) shows that the capital stock decreases every year because of the deprecia-
tion of existing capital. We assume here that the depreciation rate is a constant, usu-
ally in the range of 2 to 10 percent.
To see how this works, let us continue our earlier example, in which total income
is $10 billion and saving (and therefore investment) is $2 billion. Say that the value of
the existing capital stock is $30 billion and the annual rate of depreciation is 3 percent.
In this example, the capital stock increases by $2 billion because of new investment
but also decreases by $0.9 billion (3 percent * $30 billion) because of depreciation.
Equation 4–4 puts together these two effects, calculating the change in the capi-
tal stock as !K = I - (d * K) = $2 billion - (0.03 * $30 billion) = $1.1 billion.
Thus the capital stock increases from $30 billion to $31.1 billion. This new value of
the capital stock then is inserted into the production function in equation 4–1, allow-
ing for the calculation of a new level of output, Y.
The fifth and final equation of the model focuses on the supply of labor. To keep
things simple, we assume that the labor force grows exactly as fast as the total popu-
lation. Over long periods of time, this assumption is fairly accurate. If n is equal to the
growth rate of both the population and the labor force, then the change in the labor
force (!L) is represented by
!L = nL [4–5]
If the labor force consists of 1 million people and the population (and labor force) is
growing by 2 percent, the labor force increases annually by 20,000 (1 million * 0.02)
workers. The labor force now consists of 1.02 million people, a figure that can be
inserted into the production function for L to calculate the new level of output. (If we
divide both sides of equation 4–5 by L, we can see directly the rate of growth of the
labor force, !L>L = n.)
These five equations represent the complete model.5 Collectively, they can be
used to examine how changes in population, saving, and investment initially affect
5
Note that because the model has five equations and five variables (Y, K, L, I, and S) it always can be
solved. In addition, there are three fixed parameters (d, s, and n), the values of which are assumed to be
fixed exogenously, or outside the system.
the capital stock and labor supply and ultimately determine economic output. New
saving generates additional investment, which adds to the capital stock and allows
for increased output. New workers add further to the economy’s capacity to increase
production.
One way these five equations can be simplified slightly is to combine equa-
tions 4–2, 4–3, and 4–4. The aggregate level of saving (in equation 4–2) determines
the level of investment in equation 4–3, which (together with depreciation) deter-
mines changes in the capital stock in equation 4–4. Combining these three equa-
tions gives us
!K = sY - dK [4–6]
This equation states that the change in the capital stock (!K) is equal to saving (sY)
minus depreciation (dK). This expression allows us to calculate the change in the
capital stock and enter the new value directly into the aggregate production function
in equation 4–1.
THE HARROD-DOMAR GROWTH MODEL
As we have stressed, the aggregate production function (shown earlier as equation

4–1) is at the heart of every model of economic growth. This function can take many
different forms, depending on what we believe is the true relationship between the
factors of production (K and L) and aggregate output. This relationship depends on
(among other things) the mix of economic activities (for example, agriculture, heavy
industry, light labor-intensive manufacturing, high-technology processes, services),
the level of technology, and other factors. Indeed, much of the theoretical debate
in the academic literature on economic growth is about how to best represent the
aggregate production process.
THE FIXED-COEFFICIENT PRODUCTION FUNCTION

One special type of a simple production function is shown in Figure 4–1. Output in
this figure is represented by isoquants, which are combinations of the inputs (labor
and capital in this case) that produce equal amounts of output. For example, on the
first (innermost) isoquant, it takes capital (plant and equipment) of $10 million and
100 workers to produce 100,000 keyboards per year (point a). Alternatively, on the
second isoquant, $20 million of capital and 200 workers can produce 200,000 key-
boards (point b). Only two isoquants are shown in this diagram, but a nearly infinite
number of isoquants are possible, each for a different level of output.
The L-shape of the isoquants is characteristic of a particular type of produc-
tion function known as fixed-coefficient production functions. These production
THE HARROD-DOMAR GROWTH MODEL 95
Quantity of capital (million $)
b Isoquant II
$20 (200,000
keyboards)
a c Isoquant I
$10 (100,000
keyboards)
100 200
Quantity of labor (person-years)
F I G U R E 4 – 1 Isoquants for a Fixed-Coefficient Production Technology

With constant returns to scale, the isoquants will be L-shaped and the production function
will be the straight line through their minimum-combination points.
functions are based on the assumption that capital and labor need to be used in a
fixed proportion to each other to produce different levels of output. In Figure 4–1,
for the first isoquant, the capital–labor ratio is 10 million:100, or 100,000:1. In other
words, $100,000 in capital must be matched with one worker to produce the given
output. For the second isoquant, the ratio is the same: $20 million:200, or 100,000:1.
This constant capital–labor ratio is represented in Figure 4–1 by the slope of the ray
from the origin through the vertices (a and b) of the isoquants. These vertices rep-
resent the least cost and hence most efficient mix of capital and labor to produce
a given quantity of output. In the case of fixed coefficients, this mix is the same for
every quantity of output.
With this kind of production function, if more workers are added without invest-
ing in more capital, output does not rise. Look again at the first isoquant, starting at
the elbow (with 100 workers and $10 million in capital). If the firm adds more work-
ers (say, increasing to 200 workers) without adding new machines, it moves hor-
izontally to the right along the first isoquant to point c. But at this point, or at any
other point on this isoquant, the firm still produces just 100,000 keyboards. In this
kind of production function, new workers need more machines to increase output.
Adding new workers without machines results in idle workers, with no increase in
output. Similarly, more machinery without additional workers results in underused
machines. On each isoquant, the most efficient production point is at the elbow,
where the minimum amounts of capital and labor are used. To use any more of either
factor without increasing the other is a waste.
The production technology depicted in Figure 4–1 also is drawn with constant
returns to scale, so if capital is doubled to $20 million and labor is doubled to 200
workers, output also exactly doubles to 200,000 keyboards per year.6 With this further
assumption, two more ratios remain constant at any level of output: capital to output
and labor to output. If keyboards are valued at $50 each, then 100,000 keyboards are
worth $5 million. In this case, in the first isoquant, $10 million in capital is needed to
produce $5 million worth of keyboards, so the capital–output ratio is $10 million:
$5 million, or 2:1. In the second isoquant the ratio is the same ($20 million:$10 mil-
lion, or 2:1). Similarly, for each isoquant the labor–output ratio is also a constant,
in this case equal to 1:50,000, meaning that each worker produces $50,000 worth of
keyboards, or 1,000 keyboards each.
THE CAPITAL–OUTPUT RATIO AND

THE HARROD-DOMAR FRAMEWORK
The fixed-coefficient, constant-returns-to-scale production function is the center-
piece of a well-known early model of economic growth that was developed inde-
pendently during the 1940s by economists Roy Harrod of England and Evsey Domar
of MIT, primarily to explain the relationship between growth and unemployment
in advanced capitalist societies.7 It ultimately focuses attention on the role of capi-
tal accumulation in the growth process. The Harrod-Domar model has been used
extensively (perhaps even overused) in developing countries to examine the relation-
ship between growth and capital requirements. The model is based on the real-world
observation that some labor is unemployed and proceeds on the basis that capital is
the binding constraint on production and growth. In the model, the production func-
tion has a very precise form, in which output is assumed to be a linear function of
capital (and only capital). As usual, the model begins by specifying the level of out-
put, which we later modify to explore changes in output, or economic growth. The
production function is specified as follows:
Y = 1>v * K or Y = K>v [4–7]
where v is a constant. In this equation, the capital stock is multiplied by the fixed
number 1>v to calculate aggregate production. If v = 3 and a firm has $30 million
in capital, its annual output would be $10 million. It is difficult to imagine a simpler
6
In a constant-returns-to-scale production function, if we multiply both capital and labor by any num-
ber, w, output multiplies by the same number. In other words, the production function has the following
property: wY = F(wK, wL).
7
Roy F. Harrod, “An Essay in Dynamic Theory,” Economic Journal (1939), 14–33; Evsey Domar, “Capital
Expansion, Rate of Growth, and Employment,” Econometrica (1946), 137–47; and Domar, “Expansion and
Employment,” American Economic Review 37 (1947), 34–55.
production function. The constant v turns out to be the capital–output ratio because,
by rearranging the terms in equation 4–7, we find
v = K>Y [4–8]
The capital–output ratio is a very important parameter in this model, so it is worth

dwelling for a moment on its meaning. This ratio essentially is a measure of the pro-
ductivity of capital or investment. In the earlier example in Figure 4–1, it took $10 mil-
lion in investment in a new plant and new equipment to produce $5 million worth of
keyboards, implying a capital-output ratio of 2:1 (or just 2). A larger v implies that more
capital is needed to produce the same amount of output. So, if v were 4 instead, then
$20 million in investment would be needed to produce $5 million worth of keyboards.
The capital–output ratio provides an indication of the capital intensity of the pro-
duction process. In the basic growth model, this ratio varies across countries for two
reasons: either the countries use different technologies to produce the same goods
or they produce a different mix of goods. Where farmers produce maize using trac-
tors, the capital–output ratio will be much higher than in countries where farmers
rely on a large number of workers using hoes and other hand tools. In countries that
produce a larger share of capital-intensive products (that is, those that require rela-
tively more machinery, such as automobiles, petrochemicals, and steel), v is higher
than in countries producing more labor-intensive products (such as textiles, basic
agriculture, and foot wear). In practice, as economists move from the v of the model
to actually measuring it in the real world, the observed capital–output ratio can also
vary for a third reason: differences in efficiency. A larger measured v can indicate
less-efficient production when capital is not being used as productively as possible. A
factory with lots of idle machinery and poorly organized production processes has a
higher capital–output ratio than a more-efficiently managed factory.
Economists often calculate the incremental capital–output ratio (ICOR) to
determine the impact on output of additional (or incremental) capital. The ICOR
measures the productivity of additional capital, whereas the (average) capital–output
ratio refers to the relationship between a country’s total stock of capital and its total
national product. In the Harrod-Domar model, because the capital–output ratio is
assumed to remain constant, the average capital–output ratio is equal to the incre-
mental capital–output ratio, so the ICOR = v.
So far, we have been discussing total output, not growth in output. The produc-
tion function in equation 4–7 easily can be converted to relate changes in output to
changes in the capital stock:
!Y = !K>v [4–9]
The growth rate of output, g, is simply the increment in output divided by the total
amount of output, !Y>Y . If we divide both sides of equation 4–9 by Y, then
g = !Y>Y = !K>Yv [4–10]

Finally, from equation 4–6, we know that the change in the capital stock !K is equal
to saving minus the depreciation of capital (!K = sY - dK). Substituting the
right-hand side of equation 4–6 into the term for !K in equation 4–10 and simplify-
ing8 leads to the basic Harrod-Domar relationship for an economy:
g = (s>v) - d [4–11]
Underlying this equation is the view that capital created by investment is the main
determinant of growth in output and that saving makes investment possible.9 It rivets
attention on two keys to the growth process: saving (s) and the productivity of capital
(v). The message from this model is clear: Save more and make productive invest-
ments, and your economy will grow.
Economic analysts can use this framework either to predict growth or to calculate
the amount of saving required to achieve a target growth rate. The first step is to try
to estimate the incremental capital–output ratio (v) and depreciation rate (d). With a
given saving rate, predicting the growth rate is straightforward. If the saving (or invest-
ment) rate is 24 percent, the ICOR is 3, and the depreciation rate is 5 percent, then the
economy can be expected to grow by 3 percent (because 0.24>3 - 0.05 = 0.03).
How does this model work in practice? Consider Indonesia, which from 2002
to 2007 had an investment rate of about 30 percent and recorded a GDP growth just
under 5.5 percent per year. Assuming a depreciation rate of 5 percent, the implied
ICOR was approximately v = 2.86.10 Would these figures have helped the Indo-
nesian government predict the 2007–08 growth rate? In 2008, the investment rate
was 29 percent, so the Harrod-Domar model would have predicted growth of
5.1 percent (g = 0.29>2.86 - 0.05). The actual growth rate in 2007–08 was 6.0 per-
cent, within sight of the prediction but not highly accurate.
STRENGTHS AND WEAKNESSES OF THE

HARROD-DOMAR FRAMEWORK
The basic strength of the Harrod-Domar model is its simplicity. The data requirements
are small, and the equation is easy to use and to estimate. And, as we saw with the
example of Indonesia, the model can be somewhat accurate from one year to the next.
Generally speaking, in the absence of severe economic shocks (such as a drought, a
financial crisis, or large changes in export or import prices), the model can do a reason-
able job of estimating expected growth rates in most countries over very short periods
of time (a few years). Another strength is its focus on the key role of saving. As discussed
8
Substituting equation 4–6 into 4–10 leads to g = (sY - dK)>Y * 1>v, which can be simplified to
g = (s - d * K>Y) * 1>v. Since K>Y = v, we have g = (s - dv) * 1>v, which leads to g = s>v - d.
9
For an important early contribution to the discussion of the importance of capital accumulation to the
growth process, see Joan Robinson, The Accumulation of Capital (London: Macmillan, 1956).
10
Because g = s>v - d, then v = s>(g + d). For Indonesia between 2002 and 2007, v = 0.30>(0.055 +
0.05) = 2.86.
in Chapter 3, individual decisions about how much income to save and consume
are central to the growth process. People prefer to consume sooner rather than later,
but the more that is consumed, the less that can be saved to finance investment. The
Harrod-Domar model makes it clear that saving is crucial for income to grow over time.
The model, however, has some major weaknesses. One follows directly from the
strong focus on saving. Although saving is necessary for growth, the simple form of
the model implies that it is also sufficient, which it is not. As pointed out in Chap-
ter 3, the investments financed by saving actually have to pay off with higher income
in the future, and not all investments do so. Poor investment decisions, changing
government policies, volatile world prices, or simply bad luck can alter the impact of
new investment on output and growth. Sustained growth depends both on generat-
ing new investment and ensuring that investments are productive over time. In this
vein, the allocation of resources across different sectors and firms can be an impor-
tant determinant of output and growth. Because (for simplicity) the Harrod-Domar
assumes only one sector, it leaves out these important allocation issues.
Perhaps the most important limitations in the model stem from the rigid
assumptions of fixed capital-to-labor, capital-to-output, and labor-to-output ratios,
which imply very little flexibility in the economy over time. To keep these ratios
constant, capital, labor, and output must all grow at exactly the same rate, which is
highly unlikely to happen in real economies. To see why these growth rates must all
be the same, consider the growth rate of capital. If the capital stock grew any faster or
slower than output at rate g, the capital–output ratio would change. Thus the capi-
tal stock must grow at g to keep the capital–output ratio constant over time. With
respect to labor, in our original five-equation model, we stipulated (in equation 4–5)
that the labor force would grow at exactly the same pace as the population at rate n.
Therefore, the only way that the capital stock and the labor force can grow at the
same rate is if n happens to be equal to g. This happens only when n = g = s>v - d,
and there is no particular reason to believe the population will grow at that rate.
In this model, the economy remains in equilibrium with full employment of
the labor force and the capital stock only under the very special circumstances that
labor, capital, and output all grow at the rate g. On the one hand, if n is larger than g,
the labor force grows faster than the capital stock. In essence, the saving rate is not
high enough to support investment in new machinery sufficient to employ all new
workers. A growing number of workers do not have jobs and unemployment rises
indefinitely. On the other hand, if g (or s>v - d) is larger than n, the capital stock
grows faster than the workforce. There are not enough workers for all the available
machines, and capital becomes idle. The actual growth rate of the economy no lon-
ger is g, as the model stipulates, but slows to n, with output constrained by the num-
ber of available workers.
So, unless s>v - d (or g) is exactly equal to n, either labor or capital is not fully
employed and the economy is not in a stable equilibrium. This characteristic of the
Harrod-Domar model has come to be known as the knife-edge problem. As long as
g = n, the economy remains in equilibrium, but as soon as either the capital stock
or the labor force grows faster than the other, the economy falls off the edge with con-
tinuously growing unemployment of either capital or labor.
The rigid assumptions of fixed capital–output, labor–output, and capital–labor
ratios may be reasonably accurate for short periods of time or in very special cir-
cumstances but almost always are inaccurate over time as an economy evolves and
develops. Each of these varies among countries and, for a single country, over time.
Consider the incremental capital–output ratio. The productivity of capital can change
in response to policy changes, which in turn affects v. Moreover, the capital inten-
sity of the production process can and usually does change over time. A poor country
with a low saving rate and surplus labor (unemployed and underemployed workers)
can achieve higher growth rates by utilizing as much labor as possible and thus rela-
tively less capital. For example, a country relying heavily on labor-intensive agricul-
tural production will record a low v. As economies grow and per capita income rises,
the labor surplus diminishes and economies shift gradually toward more capital-
intensive production. As a result, the ICOR shifts upward. Thus a higher v may not
necessarily imply inefficiency or slower growth. ICORs can also shift through market
mechanisms, as prices of labor and capital change in response to changes in supplies.
As growth takes place, saving tends to become relatively more abundant and hence
the price of capital falls while employment and wages rise. Therefore, all producers
increasingly economize on labor and use more capital and the ICOR tends to rise.
Consider again the example of Indonesia. The ICOR changed from approxi-
mately 2.4 during the 1980s, to 4.1 during the 1990s, to 3.6 during 2000–09, reflect-
ing a trend toward more capital-intensive production processes. To continue to use
the 1980s ICOR in 2009 would have been very misleading and betrayed a significant
misunderstanding of the growth process. The structure of the economy had changed
substantially during that time period and, with it, the ICOR. Thailand provides a sim-
ilar example, as described in Box 4–1.
As a result of these rigidities, the Harrod-Domar framework tends to become
increasingly inaccurate over longer periods of time as the actual ICOR changes and,
with it, the capital–labor ratio. In a world with fixed-coefficient production functions,
little room is left for a factory manager to increase output by hiring one more worker
without buying a machine to go with the worker or to purchase more machines for
the current workforce to use. The fixed-proportion production function does not
allow for any substitution between capital and labor in the production process. In the
real world, of course, at least some substitution between labor and capital is possible
in most production processes. As we see in the next section, adding this feature to the
model allows for a much richer exploration of the growth process.
A final weakness of the Harrod-Domar model is the absence of any role for pro-
ductivity growth—the ability to produce increasing quantities of output per unit of
input. In Figure 4–1, increased factor productivity can be represented by an inward
shift of each isoquant toward the origin, implying that less labor and capital would
B O X 4–1 E C O N O M I C G R O W TH I N TH A I L A N D
In the 1960s, Thailand’s agrarian economy depended heavily on rice, maize, rub-
ber, and other agricultural products. About three-quarters of the Thai population
derived its income from agricultural activities. Gross domestic product (GDP) per
capita in 1960 (measured in 2005 international dollars) was around $1,200—less
than one-tenth the average income in the United States. Life expectancy was
53 years and the infant mortality rate was 103 per 1,000 births. Few observers
expected Thailand to develop rapidly.
However, since the mid-1960s, the Thai economy has grown rapidly (if not
always steadily), benefiting from relatively sound economic management and a
favorable external environment. The government regularly achieved surpluses on
the current account of its budget and used these funds (plus modest inflows of
foreign assistance) to finance investments in rural roads, irrigation, power, tele-
communications, and other basic infrastructure. At least until the mid-1990s,
the government’s fiscal, monetary, and exchange rate policies kept the macro-
economy relatively stable with fairly low inflation, despite the turbulent period of
world oil price shocks in the 1970s and 1980s. Beginning in the 1970s, the
government began to remove trade restrictions and promote the production of
labor-intensive manufactured exports. These products found a ready market in
the booming Japanese economy of the 1980s and provided a growing number
of jobs for Thai workers.
Thailand’s ability to make investments and deepen its capital stock
depended on its capacity to save. The country’s saving rate averaged about
20 percent in the 1960s, already high for developing countries, and increased
steadily over time to an average of 35 percent in the 1990s, falling to about
32 percent during the 2000s. These high saving rates, combined with rela-
tively prudent economic policies, supported very rapid economic growth and
development.
Thailand’s development experience has been far from completely smooth,
however. In mid-1997, a major financial crisis erupted. Huge short-term offshore
borrowing combined with a fixed exchange rate and weak financial institutions
led to a collapse of a real estate bubble, rapid capital flight, a substantial depre-
ciation of the Thai baht, and a deep recession (see Chapter 13). In some ways,
Thailand had become the victim of its own success, with its rapid growth attract-
ing significant numbers of investors looking to gain quick profits, who rapidly
fled once the bubble began to collapse. After two years of negative growth (with
GDP falling 10 percent in 1998), the economy began to recover and growth
rebounded to 3.9 percent between 1999 and 2007.
Over the longer period between 1960 and 2007, per capita growth averaged
4.5 percent, so that the average income in Thailand is now more than eight times
higher than it was in 1960. By 2009, life expectancy grew to 69 years, infant
mortality fell to 12 per thousand, and adult literacy reached 93 percent. During
this period, the structure of the economy changed significantly. By 2009, manu-
facturing accounted for 34 percent of GDP, up from just 14 percent in 1965,
while the share of agricultural production dropped commensurately. The com-
position of exports shifted away from rice, maize, and other agricultural com-
modities toward labor-intensive manufactured products, which now account for a
large majority of all exports. As the Harrod-Domar model predicts, Thailand’s high
saving rate and resulting capital accumulation was accompanied by a dramatic
increase in output (and income) per capita. Contrary to the Harrod-Domar model,
however, the ICOR did not remain constant. As the stock of capital grew and the
economy shifted toward more capital-intensive production techniques, the ICOR
increased from 2.6 in the 1970s to nearly 5 by the early 2000s. The rising ICOR
indicated that, as the Thai economy expanded and the level of capital per worker
increased, an ever-larger increment of new capital was required to bring about a
given increase in total output.
be needed to produce the same amount of output. The simplest way to capture this
in the Harrod-Domar framework is to introduce a smaller ICOR, but of course, this
would contradict the idea of a constant ICOR.
Despite these weaknesses, the Harrod-Domar model is still used to a surpris-
ingly wide extent. Economist William Easterly documented how the World Bank and
other institutions use the model to calculate “financing gaps” between the amount of
available saving and the amount of investment supposedly needed to achieve a target
growth rate.11 He shows how simplistic and sometimes careless use of the model can
lead to weak analysis and faulty conclusions. In essence, analysts enamored by the sim-
plicity of the model tend to overlook its shortcomings when applying it to the real world.
The Harrod-Domar model provides some useful insights but does not take us
very far. The fixed-coefficient assumption provides the model with very little flexibil-
ity and does not capture the ability of real world firms to change the mix of inputs in
the production process. The model can be reasonably accurate from one year to the
next (in the absence of shocks), and it rightly focuses attention on the importance
11
See William Easterly, “Aid for Investment,” The Elusive Quest for Growth (Cambridge: MIT Press,
2001), chap. 2; and Easterly, “The Ghost of the Financing Gap: Testing the Growth Model of the Interna-
tional Financial Institutions,” Journal of Development Economics 60, no. 2 (December 1999), 423–38.
THE SOLOW (NEOCLASSICAL) GROWTH MODEL 103
of saving. But it is quite inaccurate for most countries over longer periods of time
and implies that saving is sufficient for growth, although it is not. Indeed, in the late
1950s, Domar expressed strong doubts about his own model, pointing out that it was
originally designed to explore employment issues in advanced economies rather
than growth per se and was too rigid to be useful for explaining long-term growth.12
Instead, he endorsed the new growth model of Robert Solow, to which we now turn
our attention.
THE SOLOW (NEOCLASSICAL) GROWTH MODEL
THE NEOCLASSICAL PRODUCTION FUNCTION

In 1956, MIT-economist Robert Solow introduced a new model of economic growth
that was a big step forward from the Harrod-Domar framework.13 Solow recognized
the problems that arose from the rigid production function in the Harrod-Domar
model. Solow’s answer was to drop the fixed-coefficients production function and
replace it with a neoclassical production function that allows for more flexibility
and substitution between the factors of production. In the Solow model, the capital–
output and capital–labor ratios no longer are fixed but vary, depending on the relative
endowments of capital and labor in the economy and the production process. Like
the Harrod-Domar model, the Solow model was developed to analyze industrialized
economies, but it has been used extensively to explore economic growth in all coun-
tries around the world, including developing countries. The Solow model has been
enormously influential and remains at the core of most theories of economic growth
in developing countries.
The isoquants that underlie the neoclassical production function are shown
in Figure 4–2. Note that the isoquants are curved rather than L-shaped as in the
fixed-coefficient model. In this figure, at point a, $10 million of capital and 100
workers combine to produce 100,000 keyboards, which would be valued at $5 mil-
lion (because, as stated earlier, keyboards are priced at $50 each). Starting from
this point, output could be expanded in any of three ways. If the firm’s managers
decided to expand at constant factor proportions and move to point b on isoquant II
to produce 200,000 keyboards, the situation would be identical to the fixed propor-
12
Evsey Domar, Essays in the Theory of Economic Growth (Oxford: Oxford University Press, 1957).
13
The two classic references of Solow’s work are his “A Contribution to the Theory of Economic
Growth” and “Technical Change and the Aggregate Production Function,” Review of Economics and Sta-
tistics 39 (August 1957), 312–20. For an excellent and very thorough undergraduate exposition of the Solow
and other models of economic growth, see Charles I. Jones, Introduction to Economic Growth (New York:
W. W. Norton and Company, 2001). In 1987, Solow was awarded the Nobel Prize in economics, primarily
for his work on growth theory.
Quantity of capital (million $)
d
$22
b
$20 c
$18
Isoquant II
(200,000
a keyboards)
$10
Isoquant I
(100,000
keyboards)
100 200
Quantity of labor (person-years)
F I G U R E 4 – 2 Isoquants for a Neoclassical Production Technology

Instead of requiring fixed factor proportions, as in Figure 4–1, output can be achieved with
varying combinations of labor and capital. This is called a neoclassical production function.
Note that the isoquants are curved rather than L-shaped.
tions case of Figure 4–1. The capital–output ratio at both points a and b would be 2:1,
as it was before ($10 million of capital produces $5 million of keyboards at point a,
and $20 million of capital produces $10 million of keyboards at point b). Note that
the Solow model retains from the Harrod-Domar model the assumption of constant
returns to scale, so that a doubling of labor and capital leads to a doubling of output.
But by dropping the fixed-coefficients assumption, production of 200,000 keyboards
could be achieved by using different combinations of capital and labor. For example,
the firm could use more labor and less capital (a more labor-intensive method), such
as at point c on isoquant II. In that case, the capital–output ratio falls to 1.8:1 ($18
million in capital to produce $10 million in keyboards).
Alternatively, the firm could choose a more capital-intensive method, such as at
point d on isoquant II, where the capital–output ratio would rise to 2.2:1 ($22 million
in capital to produce $10 million in keyboards). The kinds of tools that policy makers
use to try to decrease or increase the capital–output ratio are discussed in depth in
several chapters later in this text.
THE BASIC EQUATIONS OF THE SOLOW MODEL

The Solow model is understood most easily by expressing all the key variables in
per-worker terms (for example, output per worker and capital per worker). To do
so, we divide both sides of the production function in equation 4–1 by L, so that it
takes the form
y = f (k )
F I G U R E 4 – 3 The Production Function in the Solow Growth Model

The neoclassical production function in the Solow model displays diminishing returns to
capital so that each additional increment in capital per worker (k) is associated with smaller
increases in output per worker ( y).
Y>L = F(K>L, 1) [4–12]
The equation shows that output per worker is a function of capital per worker.14 If we
use lower-case letters to represent quantities in per-worker terms, then y is output
per worker (that is, y = Y>L) and k is capital per worker (k = K>L). This gives us the
first equation of the Solow model, in which the production function can be written
simply as
y = f (k) [4–13]
Solow’s model assumes a production function with the familiar property of dimin-
ishing returns to capital. With a fixed labor supply, giving workers an initial amount
of machinery to work with results in large gains in output. But as these workers are
given more and more machinery, the addition to output from each new machine gets
smaller and smaller. An aggregate production function with this property is shown
in Figure 4–3. The horizontal axis represents capital per worker (k), and the vertical
axis shows output per worker (y). The slope of the curve declines as the capital stock
increases, reflecting the assumption of the diminishing marginal product of capital.
14
We can divide both sides by L because the Solow model (like the Harrod-Domar model) assumes
the production function exhibits constant returns to scale and has the property that wY = F(wK, wL). To
express the Solow model in per-worker terms, we let w = 1>L.
Each movement to the right on the horizontal axis yields a smaller and smaller
increase in output per worker. (By comparison, the fixed-coefficient production func-
tion assumed in the Harrod-Domar model would be a straight line drawn through
the origin.)
The first equation of the Solow model tells us that capital per worker is funda-
mental to the growth process. In turn, the second equation focuses on the determi-
nants of changes in capital per worker. This second equation can be derived from
equation 4–615 and shows that capital accumulation depends on saving, the growth
rate of the labor force, and depreciation:
!k = sy - (n + d)k [4–14]
This is a very important equation, so we should understand exactly what it means.

It states that the change in capital per worker (!k) is determined by three things:
• The !k is positively related to saving per worker. Because s is the saving rate
and y is income (or output) per worker, the term sy is equal to saving per
worker. As saving per worker increases, so does investment per worker, and
the capital stock per worker (k) grows.
• The !k is negatively related to population growth. This is shown by the
term -nk. Each year, because of growth in the population and labor force,
there are nL new workers. If there were no new investment, the increase in
the labor force would mean that capital per worker (k) falls. Equation 4–14
states that capital per worker falls by exactly nk.
• Depreciation erodes the capital stock. Each year, the amount of capital per
worker falls by the amount -dk simply because of depreciation.
Therefore, saving (and investment) adds to capital per worker, whereas labor
force growth and depreciation reduce capital per worker. When saving per capita, sy,
15
To derive equation 4–14, we begin by dividing both sides of equation 4–6 by K so that
!K>K = sY>K - d
We then focus on the capital per worker ratio, k = K>L. The growth rate of k is equal to the growth rate
of K minus the growth rate of L:
!k>k = !K>K - !L>L
With a little rearranging of terms, this equation can be written as !K>K = !k>k + !L>L. We earlier
assumed that both the population and the labor force were growing at rate n, so !L>L = n. By substitu-
tion we obtain
!K>K = !k>k + n
Note that in both the first equation of this footnote and the equation just given, the left-hand side is
equal to !K>K . This implies that the right-hand sides of these two equations are equal to each other, as
follows:
!k>k + n = sY>K - d
By subtracting n from both sides and multiplying through by k, we find that
!k = sy - nk - dk or !k = sy - (n + d)k
is larger than the amount of new capital needed to compensate for labor force growth
and depreciation, (n + d)k, then !k is a positive number. This implies that capital
per worker k increases.
The process through which the economy increases the amount of capital per
worker, k, is called capital deepening. Economies in which workers have access to
more machines, computers, trucks, and other equipment have a deeper capital base
than economies with less machinery, and these economies are able to produce more
output per worker.
In some economies, however, the amount of saving is just enough to provide
the same amount of capital to new workers and compensate for depreciation. An
increase in the capital stock that just keeps pace with the expanding labor force
and depreciation is called capital widening (referring to a widening of both the
total amount of capital and the size of the workforce). Capital widening occurs
when sy is exactly equal to (n + d)k, implying no change in k. Using this termi-
nology, equation 4–14 can be restated as saying that capital deepening (!k) is
equal to saving per worker (sy) minus the amount needed for capital widening
[(n + d)k].
A country with a high saving rate can easily deepen its capital base and rapidly
expand the amount of capital per worker, thus providing the basis for growth in out-
put. In Singapore, for example, where the saving rate has averaged more than 40 per-
cent since the early 1980s, it is not difficult to provide capital to the growing labor
force and make up for depreciation and still have plenty left over to supply existing
workers with additional capital. By contrast, Kenya, with a saving rate of about 15 per-
cent, has much less saving to spare for capital deepening after providing machines to
new workers and making up for depreciation. As a result, capital per worker does not
grow as quickly, and neither does output (or income) per worker. Partly because of
this large difference in saving rates, output per person in Singapore grew by an aver-
age of 4.9 percent per year between 1960 and 2009, while Kenya’s growth averaged
about 0.34 percent.
We can summarize the two basic equations of the Solow model as follows. The
first [ y = f (k)] simply states that output per worker (or income per capita) depends
on the amount of capital per worker. The second equation, !k = sy - (n + d)k,
says that change in capital per worker depends on saving, the population growth
rate, and depreciation. Thus, as in the Harrod-Domar model, saving plays a central
role in the Solow model. However, the relationship between saving and growth is not
linear because of diminishing returns to capital in the production function. In addi-
tion, the Solow model introduces a role for the population growth rate and allows for
substitution between capital and labor in the growth process.
Now that we are equipped with the basic model, we can proceed to analyze the
effects of changes in the saving rate, population growth, and depreciation on eco-
nomic output and economic growth. This is accomplished most easily by examining
the model in graphical form.
THE SOLOW DIAGRAM

The diagram of the Solow model consists of three curves, shown in Figure 4–4. The
first is the production function y = f (k), given by equation 4–13. The second is a
saving function, which is derived directly from the production function. The new
curve shows saving per capita, sy, calculated by multiplying both sides of equation
4–13 by the saving rate, so that sy = s * f (k). Because saving is assumed to be a
fixed fraction of income (with s between 0 and 1), the saving function has the same
shape as the production function but is shifted downward by the factor s. The third
curve is the line (n + d)k, which is a straight line through the origin with the slope
(n + d). This line represents the amount of new capital needed as a result of
growth in the labor force and depreciation just to keep capital per worker (k) con-
stant. Note that the second and third curves are representations of the two right-
hand terms of equation 4–14.
The second and third curves intersect at point A, where k = k0. (Note that, on
the production function above the sy curve, k = k0 corresponds to a point directly
above A where y = y0 on the vertical axis.) At point A, sy is exactly equal to (n + d)k,
so capital per worker does not change and k remains constant. At other points along
the horizontal axis, the vertical difference between the sy curve and the (n + d)k line
determines the change in capital per worker. To the left of point A (say, where k = k1
(n + d )k
y = f (k )
y2
y0
y1 sy
k1 k0 k2 k
F I G U R E 4 – 4 The Basic Solow Growth Model Diagram

In the basic Solow diagram, point A is the only place where the amount of new saving, sy, is
exactly equal to the amount of new capital needed to compensate for growth in the workforce and
depreciation (n + d)k. Point A is the steady state level of capital per worker and output per worker.
and on the production function y = y1), the amount of saving in the economy per
person (sy) is larger than the amount of saving needed to compensate for new work-
ers and depreciation [(n + d)k]. As a result, the amount of capital per person (k)
grows (capital deepening) and the economy shifts to the right along the horizontal
axis. The economy continues to shift to the right as long as the sy curve is above the
(n + d)k curve, until eventually the economy reaches an equilibrium at point A. In
terms of the production function, the shift to the right implies an increase in out-
put per worker, y (or income per capita), from y1 to y0. To the right of point A (say,
where k = k2 and y = y2), saving per capita is smaller than the amount needed for
new workers and depreciation, so capital per worker falls and the economy shifts to
the left along the horizontal axis. Once again, this shift continues until the economy
reaches point A. The shift to the left corresponds to a decline in output per worker
from y2 to y0.
Point A is the only place where the amount of new saving, sy, is exactly equal
to the amount of new capital needed for growth in the workforce and depreciation.
Therefore, at this point, the amount of capital per worker, k, remains constant. Saving
per worker (on the vertical axis of the saving function) also remains constant, as does
output per worker (or income per capita) on the production function, with y = y0.
As a result, point A is called the steady state of the Solow model. Output per capita
at the steady state (y0) is alternatively referred to as the steady state, long run, or
potential level of output per worker.
It is very important to note, however, that all the values that remain constant are
expressed as per worker. Although output per worker is constant, total output con-
tinues to grow at rate n, the same rate the population and workforce grow. In other
words, at the steady state GDP (Y) grows at the rate n, but GDP per capita (y) is con-
stant (average income remains unchanged). Similarly, although capital per worker
and saving per worker are constant at point A, total capital and total saving grow.
CHANGES IN THE SAVING RATE AND POPULATION

GROWTH RATE IN THE SOLOW MODEL
Both the Solow and Harrod–Domar models put saving (and investment) at the core
of the growth process. In the Harrod-Domar model, an increase in the saving rate
translates directly (and linearly) into an increase in aggregate output. What is the
impact of a higher saving rate in the Solow model?
As shown in Figure 4–5, increasing the saving rate from s to s" shifts the saving
function sy up to s"y, without shifting either the production function or the capital
widening line (n + d)k. The increase in the saving rate means that saving per worker
(and investment per worker) now is greater than (n + d)k, so k gradually increases.
The economy shifts to a new long-run equilibrium at point B. In the process, capital
per worker increases from k0 to k3 and output per worker increases from y0 to y3.
The aggregate economy initially grows at a rate faster than its steady-state growth
(n + d )k
y = f (k )
y3
s'y
y0
B
sy
k0 k3 k
F I G U R E 4 – 5 An Increase in the Saving Rate in the Solow Model

An increase in the saving rate from s to s" results in an upward shift in the capital deepening
curve, so that capital per worker increases from k0to k3.
rate of n until it reaches point B, where the long-run growth rate reverts to n. Thus,
the higher saving rate leads to more investment, a permanently higher stock of capi-
tal per worker, and a permanently higher level of income (or output) per worker. In
other words, the Solow model predicts that economies that save more have higher
standards of living than those that save less. (The increase in per capita income, how-
ever, is smaller than for a similar increase in s in the Harrod-Domar model because
the Solow model has diminishing returns in production.) Higher saving also leads to
a temporary increase in the economic growth rate as the steady state shifts from A to
B. However, the increase in the saving rate does not result in a permanent increase in
the long-run rate of output growth, which remains at n.
The Solow diagram also can be used to evaluate the impact of a change in the
population (or labor force) growth rate. An increase in the population growth rate
from n to n" rotates the capital widening line to the left from (n + d)k to (n" + d)k,
as shown in Figure 4–6. The production and saving functions do not change. Because
there are more workers, savings per worker (sy) becomes smaller and no longer is
large enough to keep capital per worker constant. Therefore, k begins to decline, and
the economy moves to a new steady state, C. More workers also means that capital
per worker declines from k0to k4and saving per worker falls from sy0 to sy4. Output
per worker (or income per capita) also declines, from y0 to y4. Thus, an increase in
the population growth rate leads to lower average income in the Solow model. Note,
however, that the new steady-state growth rate of the entire economy has increased
(n' + d ) k (n + d )k
y = f (k )
y0
y4 sy
sy 0
A
sy 4 C
k4 k0 k
F I G U R E 4 – 6 Changes in the Population Growth Rate in the Solow Model

An increase in the rate of population growth from n to n" causes the capital widening curve to
rotate to the left. Equilibrium capital per worker drops from k0to k4.
from n to n" at point C. In other words, with a higher population growth rate, Y needs
to grow faster to keep y constant.16 By contrast, a reduction in the population growth
rate rotates the (n + d)k line to the right and leads to a process of capital deepening,
with an increase in both k and in the steady-state level of income per worker, y. How-
ever, the relationship between population growth and economic growth is not quite
so simple, as described in Box 4–2.
The Solow growth model (as described to this point) suggests that growth rates
differ across countries for two main reasons:
• Two countries with the same current level of income may experience
different growth rates if one has a higher steady-state level of income than
the other. To the extent that two countries with the same current level
of income have different aggregate production functions, saving rates,
population growth rates, or rates of change in productivity (described later),
their steady-state income levels will differ and so will their growth rates
during the transition to their respective steady states.
• Two countries with the same long-run steady-state level of income may
have different growth rates if they are in different points in the transition to
16
A similar exercise can be used to determine the impact of an increase in the depreciation rate, d. Such
an increase results in a reduction in k and y to a lower steady-state income per capita. The subtle differ-
ence between an increase in n and an increase in d is that the latter case does not lead to a change in the
long-run growth rate of Y, which remains equal to n.
B O X 4–2 P O P U L ATI O N G R O W TH A N D E C O N O M I C G R O W TH
The inverse correlation between population growth and economic growth sug-
gested by the Solow model has a lot of intuitive appeal. Countries such as Belize,
Burkina Faso, Liberia, Niger, and Yemen have some of the world’s fastest rates
of population growth. They are also among some of the poorest nations in the
world. But closer inspection of the Solow model reveals other predictions about
how population growth may affect economic growth, and a further examination
of empirical trends suggests a much more complex relationship.
Figure 4–6 illustrates that, in the Solow model, an increase in the rate of pop-
ulation growth lowers the steady-state level of income, y. However, y refers to
output per workerr, whereas the more common measure of aggregate economic
welfare is output per person, y *. These two measures of output, of course, are
related to one another, as follows:
y * = y * (N/Pop
p)
where N equals the number of workers, and Pop p is the total population. Differ-
ences in the level of output per capita, therefore, depend on both the amount
of output per worker and the ratio of workers to total population. Growth in per
capita output, similarly, depends on growth in output per worker and the growth
in the worker-to-population ratio.
The Solow model suggests that more rapid population growth reduces capi-
tal deepening and hence reduces growth in output per worker. But the effect
of population growth on the ratio of workers to total population is more com-
plex. It depends on the age structure of the population. Because of rapid popula-
tion growth, most developing countries have a young age structure, with a larger
share of younger people than is the case in developed nations that have grow-
ing populations of more elderly people. As a result of previously higher popula-
tion growth rates, many developing nations today are experiencing an increase in
their ratio of workers to total population. This positive effect of a changing age
structure on per capita incomes, sometimes referred to as a demographic gift,
can play a positive and large role in determining economic growth rates.
The impact of population growth on economic growth goes beyond its effects
on capital widening and a nation’s age structure. In the Solow model, saving and
technological change are considered exogenous. But population growth can
affect these parameters as well. The net effect of population growth on eco-
nomic growth is therefore an empirical matter. Econometric investigations of the
impact of population growth on economic growth generally show no systematic
relationship (see Chapter 7). Population growth influences many aspects of eco-
nomic growth and development, not only those described by the Solow model.
the steady state. For example, consider two countries that are identical in
every way except that one has a higher saving rate than the other and, so,
initially has a higher steady-state level of income. At the steady states, the
country with the higher saving rate has a higher level of output per worker,
but both are growing at the rate n. If the country with the lower saving rate
suddenly increases its saving to match the other country, its growth rate will
be higher than the other country until it catches up at the new steady state.
Thus, even though everything is identical in the two countries, their growth
rates may differ during the transition to the steady state, which may take
many years.
TECHNOLOGICAL CHANGE IN THE SOLOW MODEL

The Solow model, as described to this point, is a powerful tool for analyzing the inter-
relationships between saving, investment, population growth, output, and economic
growth. However, the unsettling conclusion of the basic model is that, once the econ-
omy reaches its long-run potential level of income, economic growth simply matches
population growth, with no chance for sustained increases in average income. How
can the model explain the historical fact reported in Chapter 2 that many of the
world’s countries have seen steady growth in average incomes since 1820? Solow’s
answer was technological change.17 According to this idea, a key reason why France,
Germany, the United Kingdom, the United States, and other high-income countries
have been able to sustain growth in per capita income over very long periods of time
is that technological progress has allowed output per worker to continue to grow.
Technological progress is a key driver, but not the only driver, of productivity
growth. Historically, most technological innovation has originated in today’s devel-
oped countries, where technological progress has played a central role in explaining
productivity growth. Technological progress is also important for today’s developing
countries; yet, most such progress in developing countries is adopted and adapted
from developed countries. In developing countries, the productivity-enhancing
effects of technological innovation have played a smaller role in driving productiv-
ity growth than in developed countries. In the former, productivity growth has also
resulted from improvements in physical infrastructure, increased education of the
labor force, and improvements in regulatory environments and incentives. Absent
these (and related) factors, technology adoption in developing countries tends to be
more limited and less effective in enhancing factor productivity. With these impor-
tant caveats, we proceed to introduce productivity growth into the Solow model using
technological progress as a shorthand.
17
See Solow, “Technical Change and the Aggregate Production Function.” For an early discussion about
the relationship between capital accumulation and technological progress, see Joan Robinson, Essays in
the Theory of Economic Growth (London: Macmillan, 1962).
To incorporate an economy’s ability to produce more output with the same

amount of capital and labor, we slightly modify the original production function and
introduce a variable, T, to represent technological progress, as follows:
Y = F(K, T * L) [4–15]
In this specification, technology is introduced in such a way that it directly enhances

the input of labor, as shown by the specification in which L is multiplied by T. This
type of technological change is referred to as labor augmenting.18 As technology
improves (T rises), the efficiency and productivity of labor increases because the
same amount of labor can now produce more output. Increases in T can result from
improvements in technology in the scientific sense (new inventions and processes)
or in terms of human capital, such as improvements in the health, education, or
skills of the workforce.19
The combined term T * L is sometimes referred to as the amount of effective
units of labor. The expression T * L measures both the amount of labor and its effi-
ciency in the production process. An increase in either T or L increases the amount of
effective labor and therefore increases aggregate production. For example, an insur-
ance sales office can increase its effective workforce by either adding new workers
or giving each worker a faster computer or better cell phone. An increase in T differs
from an increase in L, however, because the rise in aggregate income from new tech-
nology does not need to be shared with additional workers. Therefore, technological
change and productivity growth more broadly allow output (and income) per worker
to increase.
The usual assumption is that technology improves at a constant rate, which we
denote by the Greek letter theta (u), so that !T>T = u. If technology grows at 1 percent
per year, then each worker becomes 1 percent more productive each year. With the
workforce growing at n, growth in the effective supply of labor is equal to n + u. If
the workforce (and population) grows by 2 percent per year and technology grows by
1 percent per year, the effective supply of labor increases by 3 percent per year.
To show technological change in the Solow diagram, we need to modify our
notation. Whereas earlier we expressed y and k in terms of output and capital per
worker, we now need to express these variables in terms of output and capital per
effective worker. The change is straightforward. Instead of dividing Y and K by L as
18
Two other possibilities are capital-augmenting technological change [Y = F(T * K, L)], which
enhances capital inputs, and Hicks-neutral technological change [Y = F(T * K, T * L)], which enhances
both capital and labor input. For our purposes, the specific way in which technology is introduced does
not affect the basic conclusions of the model.
19
Keep in mind, however, that, while these two broad categories of improvements in technology have
similar general effects in this aggregate model, their true effects are somewhat different in the real world.
Technological change in the mechanical sense or from the spread of a new idea can be shared widely
across the workforce and considered a public good. Improvements in human capital, by contrast, are spe-
cific to individual workers and are not necessarily widely shared. However, both have the effect of aug-
menting the supply of labor and increasing total output.
ye
(n + d + θ )k
ye = f (ke )
ye 0
sye
sye 0
A
ke0 ke
F I G U R E 4 – 7 The Solow Model with Technical Change

In the Solow model with technical change, the equilibrium level of effective capital per worker
(Ke0) is determined by point A, the intersection of the effective capital widening curve
(n + d + u) and the effective saving curve (sye).
previously (to obtain y and k), we now divide each by (T * L). Thus output per
effective worker (ye) is defined as ye = Y>(T * L). Similarly, capital per effective
worker (ke) is defined as ke = K>(T * L).20
With these changes, the production function can be written as ye = f (ke) and
saving per effective worker expressed as sye. With effective labor now growing at the
rate n + u, the capital accumulation equation (4–14) changes to
!ke = sye - (n + d + u)ke [4–16]
The new term (n + d + u)ke is larger than the original (n + d)k, indicating that
more capital is needed to keep capital per effective worker constant.
These changes are shown in Figure 4–7, which looks very similar to the basic
Solow diagram, with only a slight change in notation. There still is one steady-state
point, at which saving per effective worker is just equal to the amount of new capi-
tal needed to compensate for changes in the size of the workforce, depreciation, and
technological change in order to keep capital per effective worker constant.
One change, however, is very important. At the steady state, output per effective
worker is constant, rather than output per worker. Total output now grows at the rate
n + u, so that output per actual worker (or income per person) increases at rate u.
Note that this is consistent with the earlier notation. If there is no technological change (our earlier
20
assumption), so that T = 1 (and remains unchanged), then ye = y and ke = k.

With the introduction of technology, the model now incorporates the possibility of an
economy experiencing sustained growth in per capita income at rate u. This mecha-
nism provides a plausible explanation for why the industrialized countries never
seem to reach a steady state with constant output per worker but instead historically
have recorded growth in output per worker of between 1 and 2 percent per year.
STRENGTHS AND WEAKNESSES OF THE SOLOW FRAMEWORK

Although the Solow model is more complex than the Harrod-Domar framework,
it is a more powerful tool for understanding the growth process. By replacing the
fixed coefficients production function with a neoclassical one, the model provides
more reasonable flexibility of factor proportions in the production process. Like the
Harrod-Domar framework, it emphasizes the important role of factor accumulation
and saving, but its assumption of diminishing marginal product of capital provides
more realism and accuracy over time. It departs significantly from the Harrod-Domar
framework in distinguishing the current level of income per worker from the long-
run steady-state level and focuses attention on the transition path to that steady state.
The model provides powerful insights into the relationship between saving, invest-
ment, population growth, and technological change on the steady-state level of out-
put per worker. The Solow model does a much better job, albeit far from perfect, of
describing real world outcomes than the Harrod-Domar model.
A particularly important contribution of the model is the simple yet powerful
insights it provides into the role of technological change and productivity growth
in the growth process. For policy makers, key questions then become how to best
acquire new technologies and how to magnify their potential contributions to pro-
ductivity growth through complementary investments and policies. For most low-
income countries, while some domestic innovation is possible, for many industries
it is probably most cost-effective for entrepreneurs to acquire the bulk of their new
technologies from other countries (one of the benefits of “globalization”) and adapt
them to local circumstances. The willingness of entrepreneurs to make such invest-
ments, however, may depend on a number of factors that government can influence.
Investments in education (for example, improvements in the quality of the labor
force) may enable firms in developing countries to make better use of imported tech-
nologies. Investments in physical infrastructure might motivate technology adoption
by better connecting firms to markets. Improved public institutions to protect prop-
erty rights and extend the rule of law may also provide powerful incentives for firms
to invest in improving their own productivity. However, the model’s focus on the role
of factor accumulation and productivity (including technology) as the proximate
determinants of the steady state raises a new set of questions that the model does not
answer.
The model’s most troubling limitation is that Solow specified productivity
growth as exogenous (that is, determined independently of all the variables and
D I M I N I S H I N G R ETU R N S A N D TH E P R O D U CTI O N F U N CTI O N 117
parameters specified in the model). He did not spell out exactly how it takes place
or how the growth process itself might affect it. In this sense, productivity growth
has been called “manna from heaven” in the Solow model. This is an appropriate
abstraction for the purpose of explaining the theoretical role of technological change.
In practice, policy makers in developing countries need to know the sources of this
manna.
What are the more fundamental determinants of factor accumulation and pro-
ductivity that affect the steady state and the rate of economic growth? The empirical
evidence in Chapter 3 suggests that the most rapidly growing developing countries
share certain common characteristics: greater economic and political stability, rela-
tively better health and education, stronger governance and institutions, more export
oriented trade policies, and more favorable geography. Box 4–3 provides an estimate
of the quantitative importance of these factors in East Asia’s rapid growth relative to
other countries. In the language of the Solow model, these characteristics operate
through factor accumulation and productivity to help determine the precise shape
of the production function and the steady-state level of output per worker. Changing
any of these factors—say, encouraging more open trade—changes the steady-state
level of output per worker and therefore the current rate of economic growth as the
economy adjusts to the new steady state. Thus, the model helps us focus attention
on these more fundamental influences on the steady state and the growth rate, but
it does not provide a full understanding of the precise pathways through which these
factors influence output and growth.
Certain characteristics of the neoclassical production function embedded in the
Solow model also lead to one of the model’s broadest and most problematic empiri-
cal predictions—that initially poorer countries will grow more rapidly than initially
wealthier countries and eventually catch up. To understand this theoretical predic-
tion, we begin by reviewing the neoclassical production function introduced in
Chapter 3.
DIMINISHING RETURNS AND THE PRODUCTION

FUNCTION
For output and income to continue to grow over time, a country must continue to
attract investment and achieve productivity gains. But as the capital stock grows, the
magnitude of the impact of new investment on growth may change. Most growth
models are based on the assumption that the return on investment declines as the
capital stock grows. We illustrate this aspect of the aggregate neoclassical produc-
tion in Figure 4–8. The shape of this production function reflects the important but
common assumption of diminishing returns to capital or, more precisely, a dimin-
ishing marginal product of capital (MPK). This property is indicated by the gradual
B O X 4–3 E X P L A I N I N G D I F F E R E N C E S I N G R O W T H R AT E S
Many recent studies have shown that the initial levels of income, openness to trade,
healthy populations, effective governance, favorable geography, and high saving
rates all contribute to rapid economic growth. But which are most important? One
study sought to explain differences in growth during the period 1965–90 among
three groups of countries: 10 East and Southeast Asian countries (in which per
capita growth averaged 4.6 percent), 17 sub-Saharan African countries (in which
growth averaged 0.6 percent), and 21 Latin American countries (in which growth
averaged 0.7 percent).a
Policy variables explained much of the differences in growth rates. The East
and Southeast Asian countries recorded higher government saving rates, were
more open to trade, and had higher-quality government institutions. Together,
the differences in these policies accounted for 1.7 percentage points of the
4.0 percentage point difference between the East and Southeast Asian and
sub- Saharan African growth rates, and 1.8 percentage points of the difference
between East and Southeast Asia and Latin America. Openness to trade stood
out as the single most important policy choice affecting these growth rates.
Initial levels of income also were important, as the Solow model predicts.
Because the Latin American countries had higher average income (and therefore
greater output per worker) than the East Asian countries in 1965, the Solow
model would predict somewhat slower growth in Latin America. Sure enough,
this study estimates that Latin America’s higher initial income slowed its growth
rate by 1.2 percentage points relative to East and Southeast Asia, after con-
trolling for other factors. By contrast, the sub-Saharan African countries had
lower average initial income, indicating that (all else being equal) these coun-
tries could have grown 1.0 percentage point faster than the East and Southeast
Asian countries, rather than the actual outcome of 4.0 percentage points slower.
This suggests that the other factors had to account for a full 5.0 percentage
point difference in growth rates between East and Southeast Asia and sub-
Saharan Africa.
Initial levels of health, as indicated by life expectancy at birth, were a major
factor contributing to sub-Saharan Africa’s slow growth. Life expectancy at birth
averaged 41 years in sub-Saharan Africa in 1965, compared to 55 years in East
aSteven Radelet, Jeffrey Sachs, and Jong-Wha Lee, “The Determinants and Prospects for Eco-
nomic Growth in Asia,” International Economic Journal 15, no. 3 (Fall 2001), 1–30. These results
are summarized in the Asian Development Bank’s study Emerging Asia: Changes and Challenges
(Manila: Asian Development Bank, 1997), 79–82.
and Southeast Asia. The study estimates that this reduced sub-Saharan Africa’s
growth rate by 1.3 percentage points relative to East and Southeast Asia. By
contrast, because average life expectancy in Latin America in 1965 was almost
the same as in East and Southeast Asia, health explains little of the difference in
growth between these regions.
Favorable geography helped East and Southeast Asia grow faster. The com-
bination of fewer landlocked countries, longer average coastline, fewer countries
located in the deep tropics, and less dependence on natural resource exports all
favored Asia. Taken together, these factors accounted for 1.0 percentage point
of East and Southeast Asia’s rapid growth compared to sub-Saharan Africa, and
0.6 percentage points relative to Latin America. Differences in initial levels of
education and the changing demographic structure of the population accounted
for the remaining differences in growth rates across these regions.
Of course, this simple accounting framework does not fully explain the com-
plex relationships that underlie economic growth. Each of the variables in the
study captures a range of other factors that affect growth rates. For example,
differences in government saving rates probably reflect differences in fiscal pol-
icy, inflation rates, political stability, and many other factors. Because of lack of
sufficient data, the analysis omits several factors (such as environmental deg-
radation) that may be important. And it certainly does not begin to explain why
different policy choices were made in different countries. As a result, studies like
these should be seen as a first step to understanding growth, rather than as a
precise explanation for the many complex differences across countries.
flattening (or declining slope) of the curve as capital per worker grows. It is important,
however, to bear in mind the distinction between diminishing returns to individual
factors of production holding constant the use of all other factors and the returns
to scale of the entire production function. The production functions illustrated in
Figure 4–8 exhibit diminishing returns to capital per worker but (by assumption) still
have constant returns to scale. This latter characteristic means that if producers dou-
bled all inputs (rather than just one input) then total output would also double.
Looking at the production function in Figure 4–8, we can see that at low lev-
els of capital per worker (such as point a), new investment leads to relatively large
increases in output per worker. But at higher levels of capital per worker (such as
point b), the same amount of new investment leads to a smaller increment in output.
Each addition of a unit of capital per worker (moving to the right along the x-axis)
yields smaller and smaller increases in output per worker. More generally, giving the
same number of workers more and more machinery yields smaller and smaller addi-
tions to output.
Production
function I
Output per worker
b MPK 2
1
MPK 1
a
1
Capital per worker (thousands)
F I G U R E 4 – 8 Diminishing Marginal Product of Capital

Along production function I, an addition of 1 unit of capital per worker at point a yields
a much larger increase in output per worker than the same investment at point b. The
incremental output produced by adding an additional unit of capital is called the marginal
product of capital (MPK).
The assumption of a diminishing marginal product of capital has many impli-

cations, but three are particularly important for developing countries. Consider
countries located toward the left on the x-axis of Figure 4–8, such as at point a. These
countries have both relatively small amounts of capital per worker and low levels of
output per worker. The latter means, by definition, that these countries are relatively
poor. By contrast, countries toward the right have both higher levels of capital per
worker and more output per worker, the latter implying that they are relatively rich.
In general, low-income countries tend to have much less capital per worker than do
richer countries. Therefore, if all else is equal between the two countries—a crucial
qualifier—new investment in a poor country will tend to have a much larger impact
on output than the same investment in a rich country. The three key implications are
as follows:
• If all else is equal, poor countries have the potential to grow more rapidly
than do rich countries. In Figure 4–8, a country located at point a has the
potential to grow more rapidly than does a country at point b because the
same investment will lead to a larger increase in output.
• As countries become richer (and capital stocks become larger), growth
rates tend to slow. In other words, as a country moves along the production
function from point a to point b over a long period of time, its growth rate
tends to decline.
• Because poor countries have the potential to grow faster than do rich
countries, they can catch up and close the gap in relative income. To the
extent this happens, income levels between rich and poor countries would
converge over time.
These are very powerful implications. It is important to recognize that they rest
on the assumption that all else is equal between the two countries, in particular that
countries have the same rates of savings, population growth, and depreciation, and
hence the same steady-state level of capital and income per capita. As this interpreta-
tion of the Solow model is not conditional on countries differing from one another in
these key parameters, it is known as unconditional convergence. For all else to be
equal, both countries also have to be operating along the same production function,
have access to the same technology. If they are not, the predictions for rich and poor
countries do not necessarily hold. For instance, a given poor country might actually
be operating along different, and perhaps much flatter, production function than the
one shown in Figure 4–8 if it does not have access to the same technology as reflected
in the production function in Figure 4–8. In that case, each new investment (at a given
level of the capital stock) would produce less output on the margin than that same
additional investment would produce on the steeper production function. In that case,
the poor country might not be expected to grow faster than the rich country and may
never catch up. The phrase ceteris paribus (all else being equal), which is much used
and often overlooked in economics, is of great importance in the convergence debate.
THE CONVERGENCE DEBATE

If it were true that poorer countries could grow fast while richer countries experience
slower growth, poorer countries (at least those in which all else is equal) could begin
to catch up and see their income levels begin to converge with the rich countries. Has
this actually happened?
The short answer is that it has for some countries but not for most. Consider the
example of Japan. In the 1960s, Japan’s income per capita was only about 35 percent
of average U.S. income, and it had a much smaller capital stock, giving it the potential
for very rapid growth. (We use the United States as the benchmark for convergence
because it has among the highest per capita incomes in the world and is usually con-
sidered the global technological leader.) Indeed, Japan’s GDP growth rate exceeded
9 percent during the 1960s. By the time Japan had reached 70 percent of U.S. per cap-
ita income in the late 1970s, its GDP growth rate had slowed to about 4 percent. As
its income continued to grow, its growth rate fell further, and growth was very slow
after Japan reached about 85 percent of U.S. income in the early 1990s. Japan’s expe-
rience illustrates the preceding three points very well: (1) When it was relatively poor,
it could grow fast; (2) as its income increased, its growth rate declined; and (3) as a
result, its income converged significantly toward U.S. income. People who boldly
predicted in the 1960s and 1970s that Japan could grow at 7 to 9 percent per year
indefinitely—and many people did—ignored the impact of diminishing returns of
capital on long-term growth rates.
Japan is not the only country whose income has converged with the world
leaders since 1960. Look again at the group of rapidly growing countries shown in
Table 3–1. All these countries were relatively poor in 1960, and all grew by an average
of between 3 and 6 percent per capita for nearly 50 years. Rich countries cannot grow
that fast over a period of many years (in the absence of a continuous infusion of new
technology), but poor countries can because they start with low levels of capital.
However, being poor and having low levels of capital per worker by no means guar-
antees rapid growth. As the upper sections of Table 3–1 show, many low-income coun-
tries recorded low growth. Not only did these countries not catch up but they fell further
behind and their incomes diverged even more from the world leaders. The point is that
low-income countries have the potential for rapid growth, if they can attract new invest-
ment and if that new investment actually pays off with a large increment in output.
Looking beyond the experience of a few individual countries, is there a general
tendency for poor countries to grow faster and catch up with the richer countries?
Broadly across all countries, the short answer is no. Figure 4–9 shows the initial level
of per capita income in 1960 and subsequent rates of growth from 1960 to 2009 for
6 Botswana
Singapore
Growth Rate of GDP per capita, 1960—2009 (percent)
Turkey
USA
2 Pakistan
Argentina
0
Madagascar
6 7 8 9 10
Log GDP per capita in1960 (2005 International $)
F I G U R E 4 – 9 Gross Domestic Product (GDP) Growth, Unconditional

See Table 3–1 for data.
Source: Penn World Tables 7, http://pwt.econ.upenn.edu/php_site/pwt_index.php, accessed June 2011.
the 38 countries listed in Table 3–1. If it were true that poor countries were growing
faster than rich countries, the graph would show a clear downward slope from left
to right. Poor countries would record a high rate of growth (and appear in the upper
left part of the figure) and rich countries would display a slower growth rate (and be
in the bottom right). But there is no clear pattern evident in the figure. For example,
income per capita in 1960 was approximately the same in Botswana, Pakistan, and
Madagascar; yet while Botswana subsequently grew at nearly 6 percent per year,
Pakistan’s subsequent growth was only 2.3 percent per year and Madagascar’s
income per capita actually declined. Similarly, Pakistan, Turkey, and the United
States all grew at roughly the same rate from 1960 to 2009, yet they began the period
with quite different levels of income. The only part that seems accurate is that almost
all the rich countries display relatively slow growth rates, as expected. Such results
have been documented in many studies using larger samples and more sophisticated
statistical techniques. The empirical fact is clear: There has been no general tendency
for poor countries to catch up to the world leaders. If anything, the opposite has been
true. As we saw in Figure 2–1, for the last two centuries, the gap between the richest
and the poorest regions of the world has grown, implying a divergence of incomes for
these countries.
However, the simple graph in Figure 4–9 does not really do justice to the predic-
tions of convergence, which are based on the critical assumption that all else is equal
across countries. This assumption clearly is not true for all countries in the world.
Instead, if convergence were to occur, we should expect to find it among countries
that share some broad key characteristics, such as a similar underlying production
function and similar rates of saving, population growth, depreciation, and technol-
ogy growth. Some of the poor countries in Figure 4–9, for example, have low saving
rates or very little growth in technology compared with other countries and, there-
fore, have much less potential for rapid growth. To see if the convergence predic-
tions of the Solow model hold under these stricter conditions, we have to dig a little
deeper.
Figure 4–9 demonstrates the lack of unconditional convergence, unconditional in
the sense that it imposes the assumption that all countries share the same key param-
eters (depreciation, savings, and population growth rates in particular). Digging
deeper into the Solow model’s predictions, however, we consider the implications
for convergence of allowing each country to take on its own individual combina-
tion of these key parameters. This allows each country to have its own individual
steady-state level of capital and income per capita. Once we allow that, the question
is whether initially poorer countries grow faster as they approach their own steady
state. Figure 4–10 applies statistical techniques to relax the previous assumption that
all countries have the same rates of population growth. Looking at the data from the
same countries as in Figure 4–9, but conditional on the countries’ population growth
rates, we begin to see the predicted negative association between initial income lev-
els and subsequent growth rates. This is known as conditional convergence.
4
Botswana
Singapore
Growth Rate of GDP per capita, 1960—2009 (percent)
Pakistan
Turkey
0
USA
Argentina
—2 Madagascar
—2 —1 0 1 2
Log GDP, 1960 (2005 International $)
F I G U R E 4 – 10 Gross Domestic Product (GDP) Growth, Conditional on Population Growth

Sources: Penn World Tables 7, http://pwt.econ.upenn.edu/php_site/pwt_index.php, accessed June 2011; World
Bank, “World Development Indicators,” http://databank.worldbank.org.
Another approach is to pick a group of countries from all income levels that are
similar in their policy choices, geographic characteristics, or some other variable.
Economists Jeffrey Sachs and Andrew Warner, for example, examined the evidence
for convergence among all countries that had been consistently open to world trade
since 1965.21 Open economies, as discussed in Chapter 3, are similar in that they
have similar (global) markets for their products, purchase their inputs on world
markets, and can acquire new technology relatively quickly from other open econo-
mies through imports of new machinery and their connections to global production
networks. What if we include countries’ openness to trade among the conditioning
variables? Figure 4–11 conditions GDP growth on countries’ rates of savings and
population growth in addition to the proportion of time between 1960 and 2009 that
they were open to trade. Clearly, these additional conditioning variables sharpen the
inverse correlation between initial income and subsequent growth as indicated by
the tighter fit around the trend line.22 Initially wealthier countries may still grow faster
in absolute terms than initially poorer countries if the former are converging toward
Jeffrey D. Sachs and Andrew Warner, “Economic Reform and the Process of Global Integration,”
21
Brookings Papers on Economic Activity, 26, no. 1 (1995), 1–118.

22
Tracing the countries highlighted in Figure 4–9 through Figures 4–10 and 4–11, we note that, despite
the additional conditioning variables, Botswana and Singapore consistently grew faster than predicted for
their initial incomes (indicated by their positions above the trend lines), whereas Madagascar and Argen-
tina consistently grew more slowly than predicted.
BEYO N D SO LOW: N EW APPROACH ES TO G ROWTH 125
2
Botswana
Growth Rate of GDP per capita, 1960—2009 (percent) 1

Turkey
USA
0 Singapore
Madagascar
Pakistan
—1
—1 Argentina
—2
—3
—2 —1 0 1 2
log GDP, 1960 (2005 International $)
F I G U R E 4 – 11 Gross Domestic Product (GDP) Growth, Conditional on Openness, Savings,

and Population Growth
Sources: Penn World Tables 7, http://pwt.econ.upenn.edu/php_site/pwt_index.php, accessed June 2011; World
Bank, “World Development Indicators,” http://databank.worldbank.org.
much higher steady-state levels of capital per capita. That is why we do not observe
unconditional convergence across rich and poor countries’ levels of income. On the
other hand, if each country is allowed to have its own steady-state level, we do see
faster growth among initially poorer countries, as predicted by the Solow model and
its implication of conditional convergence.
BEYOND SOLOW: NEW APPROACHES TO GROWTH
A new generation of models takes off where Solow left off, by moving beyond the
assumptions of an exogenously fixed saving rate, growth rate of the labor supply,
workforce skill level, and pace of technological change. In reality, the values of these
parameters are not just given but are determined partially by government policies,
economic structure, and the pace of growth itself. Economists have begun to develop
more-sophisticated models in which one or more of these variables is determined
within the model (that is, these variables become endogenous to the model).23
23
The seminal contributions to the new growth theory are Paul Romer, “Increasing Returns and Long-
Run Growth,” Journal of Political Economy 94 (October 1986), 1002–37; Robert Lucas, “On the Mechan-
ics of Economic Development,” Journal of Monetary Economics 22 (January 1988), 3–42; and Paul Romer,
“Endogenous Technological Change,” Journal of Political Economy 98 (October 1990), S71–S102.
These models depart from the Solow framework by assuming that the national
economy is subject to increasing returns to scale, rather than constant returns to
scale. A doubling of capital, labor, and other factors of production leads to more than
a doubling of output. To the extent this occurs, the impact of investment on both
physical capital and human capital would be larger than suggested by Solow.
How can a doubling of capital and labor lead to more than a doubling of output?
Consider investments in research or education that not only have a positive effect on
the firm or the individual making the investment but also have a positive “spillover”
effect on others in the economy. This beneficial effect on others, called a positive
externality, results in a larger impact from the investment on the entire economy.
The benefits from Henry Ford’s development of the production line system, for
example, were certainly large for the Ford Motor Company, but they were even larger
for the economy as a whole because knowledge of this new technique soon spilled
over to other firms that could benefit from Ford’s new approach.
Similarly, investments in research and development (R&D) lead to new knowl-
edge that accrues not only to those that make the investment but to others who even-
tually gain access to the knowledge. The gain from education is determined not just
by how much a scientist’s or manager’s productivity is raised by investment in his or
her own education. If many scientists and managers invest in their own education,
there then will be many educated people who will learn from each other, increasing
the benefits from education. An isolated scientist working alone is not as productive
as one who can interact with dozens of well-educated colleagues. This interaction
constitutes the externality. In the context of the “sources of growth” analysis intro-
duced in Chapter 3, such externalities suggest that the measured contribution of
physical and human capital to growth may be larger than that captured by the Solow
framework. Among other implications, this outcome could account for a significant
portion of the residual in the Solow accounting framework, meaning that actual TFP
growth is smaller than many studies have suggested.
Another important implication is that economies with increasing returns to scale
do not necessarily reach a steady-state level of income as in the Solow framework.
When the externalities from new investment are large, diminishing returns to capi-
tal do not necessarily set in, so growth rates do not slow, and the economy does not
necessarily reach a steady state. An increase in the saving rate can lead to a perma-
nent increase in the rate of economic growth. These models can explain continued
per capita growth in many countries without relying on exogenous technological
change. Moreover, they do not necessarily lead to the conclusion that poor coun-
tries will grow faster than rich countries because growth does not necessarily slow
as incomes rise, so there is no expectation of convergence of incomes. Initial dispari-
ties in income can remain, or even enlarge, if richer countries make investments that
encompass larger externalities.
Because growth can perpetuate in these models without relying on an assump-
tion of exogenous technological change, they often are referred to as endogenous
SUMMARY 127
growth models. They are potentially important for explaining continued growth in
industrialized countries that never reach a steady state, especially those engaged in
R&D of new ideas on the cutting edge of technology.
For developing countries, the new models reinforce some of the main messages
of the Solow and Harrod-Domar models. Like their forerunners, these models show
the importance of factor accumulation and increases in productivity in the growth
process. The potential benefits from both of these sources of growth are even greater
in endogenous growth models because of potential positive externalities. The core
messages of saving, investing in health and education, using the factors of produc-
tion as productively and efficiently as possible, and seeking out appropriate new
technologies are consistent across all these models.
The applicability of endogenous growth models to developing countries is ques-
tionable, however, because many low-income countries can achieve rapid growth by
adapting the technologies developed in countries with more advanced research capac-
ities rather than making the investments in R&D themselves. For many low-income
countries, the Solow model’s assumptions of exogenous technological change and
constant returns to scale in the aggregate production function may be more appropri-
ate. Productivity growth in such countries may also depend as much on appropriate
complementary investments and policies as on technological change itself.
SUMMARY
• Formal growth models provide a more precise mechanism to explore

the contributions to economic growth of both factor accumulation
and productivity gains. These models allow us to understand better
the implications of changes in saving rates, population growth rates,
technological change, and other related factors on output and growth.
• The Harrod-Domar model assumes a fixed-coefficients production
function, which helps simplify the model but introduces strict rigidity in
the mix of capital and labor needed for any level of output. In this model,
growth is directly related to saving in inverse proportion to the incremental
capital–output ratio.
• The Harrod-Domar model usefully emphasizes the role of saving, but at
the same time overemphasizes its importance by implying that saving (and
investment) is sufficient for sustained growth, which it is not. Also, the
model does not directly address changes in productivity. In addition, the
model’s assumption of a fixed ICOR leads to increasing inaccuracy over
time as the structure of production evolves and the marginal product of
capital changes.
• The Solow model improves on some of the weaknesses in the Harrod-

Domar framework and has become the most influential growth model in
economics. The model allows for more flexibility in the mix of capital and
labor in the production process and introduces the powerful concept of
diminishing marginal product of capital. It allows for exploration of the
impact on growth of changes in the saving rate, the population growth rate,
and technological change. The model helps provide a deeper understanding
of a much wider range of growth experiences than the Harrod-Domar
model. Nevertheless, the model does not provide a full explanation for
growth. It does not provide insights into the more fundamental causes of
factor accumulation and productivity growth and, as a one-sector model,
does not address the issue of resource allocation across sectors.
• The Solow model has several powerful implications, including (1) poor
countries have the potential to grow relatively rapidly; (2) growth rates tend
to slow as incomes rise; (3) across countries that share important common
characteristics, the incomes of poor countries potentially can converge with
those of the rich countries; and (4) acquiring new technology is central to
both accelerating and sustaining economic growth.
• Some poor countries have achieved rapid growth and seen their incomes
converge with those of richer countries, but many have not and have seen
their incomes fall further behind. There is no evidence of unconditional
convergence across countries, but there is strong evidence of conditional
convergence, in which countries sharing certain characteristics are able
to achieve rapid growth and begin to catch up with the richer countries.
We find evidence of conditional convergence in the inverse relationship
between initial income and subsequent growth rates, when we account for
countries differing in key parameters and hence in their steady-state levels.
• In the 1990s, new theories of economic growth gained popularity. These
theories sought to fill in key gaps in the neoclassical model. In particular,
these models sought to explain the rate of technological progress, which
Solow had identified as the determinant of long-term growth in per capita
income but that remained exogenous in his model. These new growth
theories are thus known as endogenous growth models.

PRL ch4

Uploaded by

Copyright:

Available Formats

PRL ch4

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PRL ch4

Uploaded by

Copyright:

Available Formats

4

to reduce them? It is hard to think of any question in economics that is of greater

Two observations emerge from the juxtaposition of these strikingly similar

THE BASIC GROWTH MODEL

Our first equation is an aggregate production function. If Y represents total

This expression indicates that output is a function (denoted by F) of the capi-

THE HARROD-DOMAR GROWTH MODEL

As we have stressed, the aggregate production function (shown earlier as equation

THE FIXED-COEFFICIENT PRODUCTION FUNCTION

Quantity of capital (million $)

F I G U R E 4 – 1 Isoquants for a Fixed-Coefficient Production Technology

THE CAPITAL–OUTPUT RATIO AND

Y = 1>v * K or Y = K>v [4–7]

The capital–output ratio is a very important parameter in this model, so it is worth

g = !Y>Y = !K>Yv [4–10]

STRENGTHS AND WEAKNESSES OF THE

THE SOLOW (NEOCLASSICAL) GROWTH MODEL

THE NEOCLASSICAL PRODUCTION FUNCTION

Quantity of capital (million $)

F I G U R E 4 – 2 Isoquants for a Neoclassical Production Technology

THE BASIC EQUATIONS OF THE SOLOW MODEL

F I G U R E 4 – 3 The Production Function in the Solow Growth Model

Y>L = F(K>L, 1) [4–12]

This is a very important equation, so we should understand exactly what it means.

THE SOLOW DIAGRAM

F I G U R E 4 – 4 The Basic Solow Growth Model Diagram

CHANGES IN THE SAVING RATE AND POPULATION

F I G U R E 4 – 5 An Increase in the Saving Rate in the Solow Model

F I G U R E 4 – 6 Changes in the Population Growth Rate in the Solow Model

TECHNOLOGICAL CHANGE IN THE SOLOW MODEL

To incorporate an economy’s ability to produce more output with the same

In this specification, technology is introduced in such a way that it directly enhances

F I G U R E 4 – 7 The Solow Model with Technical Change

!ke = sye - (n + d + u)ke [4–16]

assumption), so that T = 1 (and remains unchanged), then ye = y and ke = k.

STRENGTHS AND WEAKNESSES OF THE SOLOW FRAMEWORK

DIMINISHING RETURNS AND THE PRODUCTION

Capital per worker (thousands)

F I G U R E 4 – 8 Diminishing Marginal Product of Capital

The assumption of a diminishing marginal product of capital has many impli-

THE CONVERGENCE DEBATE

F I G U R E 4 – 9 Gross Domestic Product (GDP) Growth, Unconditional

F I G U R E 4 – 10 Gross Domestic Product (GDP) Growth, Conditional on Population Growth

Brookings Papers on Economic Activity, 26, no. 1 (1995), 1–118.

Growth Rate of GDP per capita, 1960—2009 (percent) 1

F I G U R E 4 – 11 Gross Domestic Product (GDP) Growth, Conditional on Openness, Savings,

BEYOND SOLOW: NEW APPROACHES TO GROWTH

• Formal growth models provide a more precise mechanism to explore

• The Solow model improves on some of the weaknesses in the Harrod-

You might also like