Cross-Country Variations in National Economic Growth Rates: The
Role of "Technology"
J. Bradford De
University of California at Berkeley and National
Bureau of Economic Research
A .pdf version of this essay
Technology is both more and less important than the
conventional wisdom recognizes as a determinant of differences across
national economies in productivity levels. "Technology" in the sense
of total factor productivity is more important because of the
strong endogeneity of population growth and investment rates that
magnifies small total factor productivity differentials manyfold in
steady state. Thus the apparent paradox of "conditional
convergence"--national economies that seem to move toward their
steady-state growth paths--coupled with the continuing divergence of
relative national GDP per capital levels in the world economy.
By contrast, technology proper is less important: much if not
most differences in total factor productivity are only tenuously or
not at all related to mastery of technology in the sense of the
internal combustion engine or the freeze-drying process. Robert Solow
(1957) called shifts in total factor productivity "technical change";
his doing so may not have helped economists think clear thoughts over
the past forty years.
I want to praise "technology" as the important factor in the
relative growth performance of nation-states' economies.
I want to argue that the conventional wisdom substantially
understates the role of total factor productivity differences in
explaining differentials across nation-state economies in GDP per
capita. "Technology" in this sense is more important, because
of the strong endogeneity of population growth and investment rates.
Rich economies are economies in which children are much more
"consumption" than "investment" goods, and that have completed their
demographic transitions to a régime of low fertility and low
population growth. Thus an economy that, initially, finds itself with
a small total factor productivity advantage will see that advantage
magnified into a larger advantage in output per capita as it
converges to a steady-state growth path with lower population growth
and a higher capital-output ratio.
Similarly, a rich economy is an economy in which the price of capital
goods is relatively low: in a rich economy a given share of national
product saved translates into a greater real investment effort than
if the economy had the world's average relative price structure. This
channel magnifies differences in total factor productivity into
larger differences in output per capita working through the
steady-state capital output ratio.
Researchers in economic growth have been puzzled by the apparent
combination of "conditional convergence" with absolute divergence.
Economies appear to be moving toward their individual steady-state
growth paths by about two percent per year. Yet the spread of
relative output per capita levels across the world continues to
A naive interpretation of this pattern would suggest that at some
time in the past nation-states' savings and population growth
rates--and thus their output per capita levels--were closer together
than they are now, that some shock drove savings and population
growth rates apart, and that since then the world's distribution of
relative incomes has diverged as economies have traversed toward
their steady-state growth paths. But what was this shock that drove
savings and population growth rates apart? The evolution of the
world's cross-country distribution of income and productivity is much
more understandable once one recognizes the endogeneity of factor
accumulation, and that relatively poor countries have low investment
and high population growth rates in large part because they are
But I also have a caveat: there is also a sense in which I want to
bury technology. Robert Solow's (1957) article is entitled
"Technical Change and the Aggregate Production Function." Certainly
since 1957, and perhaps before, economists have used "technical
change" and "technology" as shorthand ways of referring to shifts in
the aggregate production function. Yet much of difference seen across
nations in aggregate total factor productivity has little to do with
technology--in the sense of knowledge of the internal
combustion engine, continuous-casting, the freeze-drying process, or
anything that would be recognizable in a model like that of Caballero
and Jaffe (1993). Technology properly so-called is the
ultimate source of our enormous material wealth today relative to our
counterparts of a century or so ago: economic growth over the past
century in the United States is built on our knowledge today
of the internal combustion engine, continuous-casting, freeze-drying,
and all of our other technologies. Yet differences across
nation-states in total factor productivity seem to be related
tenuously, or not at all, to technology.
` Robert Solow may not have done us a big favor when he convinced us
to call shifts in the aggregate production function "technical
change"; his doing so may not have helped economists to think clear
thoughts over the past forty years.
As best as we can determine from badly flawed data, the economic
history of the past century and a quarter is a history not of
"convergence" but of "divergence": the different countries and
peoples of the world have not drawn closer together in relative
living standards, but have drifted further apart.
Figure 1 below shows the distribution of world real GDP per
capita--by percentage of world population, not by nation-state--in
1993 and in 1870, as best as it can be estimated. 1993 estimates of
real GDP per capita are purchasing-power-parity concept estimates,
measured in the "international dollar" concept that pegs U.S. GDP per
capita to its current-dollar value, but that attempts to use the
relative price structure not of the advanced industrial economies but
of the "world average" economy. They are taken from the 1995 World
1870 estimates of real GDP per capita are my own extensions and
modifications of those found in Angus Maddison's (1995) Monitoring
the World Economy; by and large they are constructed by
"backcasting" individual nation-specific estimates of real GDP per
capita growth rates.
Thus there are a very large number of caveats attached to figure
* Because estimates of 1870 GDP per capita are "backcast," errors in
estimating 1993 GDP per capita are necessarily included in estimated
1870 GDP per capita as well.
* The individual nation-specific estimates of growth rates underlying
the backcasting are of widely variable quality; they do not use the
* Most of the nation-states of today's world did not exist in 1870.
Estimates for 1870 cover roughly the same area then that the
nation-state occupies now.
* Figure 1 suppresses all variability in productivity and real GDP
per capita inside of nation-states: everyone in China is
assumed to have the 1993 purchasing-power-parity concept real GDP per
capita of $2,330.
* Estimates of even 1993 purchasing-power-parity concept real GDP per
capita for developing countries are very uncertain. This especially
applies to China which, as the World Bank team politely puts it in a
footnote, has a GDP per capita estimate that is "subject to more than
the usual margin of error."
* The entire enterprise of computing purchasing-power-parity concept
real GDP per capita levels may be seriously biased; it may fail to
incorporate appropriate allowances for quality differences between
products produced in industrialized and developing economies.
Certainly purchasing-power-parity concept estimates of relative
living standards east and west of the Iron Curtain made in the 1980s
appear, in retrospect, to have wildly exaggerated levels of
productivity and material wealth in the former Soviet Union's sphere
* Estimates of 1870-1993 real GDP per capita growth are unlikely to
adequately incorporate changes in quality and in the scope of
products that are produced. The thought experiment that underlies
constant-dollar cross-time comparisons implicitly involves taking the
output produced at a particular date, moving it across time to the
base year, and selling it in the base year at the base year's market
prices. But suppose you gave me the $2,763 dollars-- the estimate of
U.S. GDP per capita in 1870--and told me "by the way, you can only
spend this sum on products that existed and quality levels that were
produced in 1870." Under these stringent restrictions on what I could
purchase, I might well value that sum as worth much less than $2,763
of today's dollars.
* Figure 1--plotting approximate GDP per capita by percentile of the
world's population--looks significantly different in some respects
from figure 2, which plots GDP per capita in 1870 and 1993 by
percentile of the world's number of nation-states. Nation-state based
calculations show a nearly uniform distribution of log GDP per capita
levels over the observed range, especially for 1993. Population-based
calculations show a non-uniform distribution with a pronounced upper
tail: the difference, of course, springs from the two very large
population nation states of China and India, which are now and were
in 1870 relatively poor.
Nevertheless, figure 1 is the best we can do at present.
What are the principal lessons of figure 1? I believe that there are
The first is the extraordinary pace of real economic growth over the
past century. The highest GDP per capita level attained in 1993 (for
the United States) was some $24,470 1993-level international dollars;
the highest GDP per capita level attained in 1870 (for Australia) was
some $4,108 1993-level international dollars. Using this particular
metric, the United States today is some six times as wealthy in a
material-product real-income sense as was Australia in 1870 (and the
United States today is some nine times as well-off as was the United
States in 1870).
I stress that this pace of growth is not only very large but also
extraordinarily larger than in any previous century that we know of.
If 1870-1993 growth were simply a continuation of pre-1870 growth
trends, then in 1600 the richest economy in the world would have had
a real GDP per capita level of some $110 a year--far too low to
support human life.4
The twentieth century (extended back to 1870) has seen at least a
sixfold multiplication of real GDP per capita at the leading edge of
the world's economies; the previous century and a quarter had seen
perhaps a doubling during the period of the classical industrial
revolution (see Crafts, 1985; Mokyr, 1985). But before that? Perhaps
the most prosperous economy of the mid-eighteenth century (probably
the Netherlands) held a fifty-percent edge over the most prosperous
economy of the mid-fifteenth century (probably the city-states of
northern Italy). But perhaps not.
And looking more than five hundred years into the past it is hard to
see any significant advance in living standards or average
productivity levels. Human populations appear to be in a
near-Malthusian equilibrium, in which population growth quickly
removes the margin for any significant increase in living standards
(see Kremer, 1993; Livi-Bacci, 1992; Malthus, 1798). It is not clear
that a French peasant of the seventeenth century was any better off
than an Athenian peasant of the fourth century B.C.
The second important lesson of figure 1 is the extremely uneven pace
of economic growth over the past century. Because the relatively poor
economies of the world have not yet completed their demographic
transitions to a régime of relatively low fertility, the
poorest economies have been the fastest growing over the past
century. International migration has not proceeded at a particularly
fast pace. Thus the distribution of economic growth appears more
uneven and less widely distributed in figure 1, which plots GDP per
capita by percentile of the world's population, than in figure 2
which plots GDP per capita by nation-state.
But in both figures the line plotting the world's has rotated
clockwise about the bottom right corner: the richest economies today
have some six to nine times the GDP per capita of their counterparts
in 1870; the economy containing the median today has perhaps four
times the GDP per capita of its counterpart in 1870; the poorest
economies are little advanced over their counterparts of 1870.
To put this lesson another way, the strong economic growth of the
past century--the rise in the geometric average output per capita
level in the world from some $760 to some $3150 1993 international
dollars per year--has been accompanied by a substantial increase in
variance as well. In 1870 the standard deviation of log GDP per
capita across the world's population was some 0.53; today it is 1.00.
The range from one standard deviation below to one standard deviation
above the mean in log GDP per capita took up the interval from $450
to $1310 international dollars in 1870; the same interval runs from
$1160 to $8510 international dollars today.
The third lesson is that by and large the economies that were rich in
relative terms in 1870 are rich in relative terms today, and that the
economies that were poor in relative terms in 1870 are poor in
relative terms today.
Barro and Sala-i-Martin (1995) draw a distinction between what they
call ][[sigma]]-divergence and [[beta]]-divergence: they call
"[[sigma]]-divergence" the case where the variance of a distribution
grows in spite of a tendency for any given element to revert toward
the mean over time; they call "[[beta]]-divergence" the case where
the variance of the distribution would continue to widen even in the
absence of all shocks--when there is no systematic regression toward
The world since 1870 has exhibited not only [[sigma]]-divergence but
also [[beta]]-divergence: the world's distribution has a greater
spread today because there has been a systematic tendency for the
relatively rich to grow faster than the relatively poor, and not
because shocks to individual nation-states' GDP per capita levels
have dominated regression to the mean. Table 1 documents this by
reporting simple regressions of nation-states' log GDP per capita
levels in 1993 on the level of 1870. If two economies' log GDP per
capita levels were separated by an amount X in 1870, they were
separated by 1.542(X) in 1993.
The degree of [[beta]]-divergence is slightly attenuated when
continent dummies are added to the right hand side. The continent
dummies have the standard pattern: strongly positive for North
America, strongly negative for Africa. More interesting, perhaps, is
that there is some evidence that GDP per capita levels have tended to
converge over the past century and a quarter, if attention is
confined to those economies that were in the richer half of the
sample in 1870.
The fact that the world's distribution of income and productivity
levels across nation states has been diverging goes oddly with a
large number of studies (see Cogley and Spiegel, 1996; Mankiw, Romer,
and Weil, 1992) that find evidence for "conditional convergence":
gaps between an economy's aggregate income and productivity level and
the level corresponding to the steady-state growth path predicted by
its investment and population growth rates shrink over time by some
two to three percent per year.
A naive interpretation of this pattern would suggest that at some
time in the past nation-states' savings and population growth rates
must have been more equal. If not, then how did the more concentrated
distribution of output per capita levels in the past ever arise? And
it would suggest that there was some massive shock that drove savings
and population growth rates apart and that since then the world's
distribution of relative incomes has diverged, as economies have
traversed toward their steady-state growth paths. But there are no
candidates for such a shock: the Industrial Revolution initially did
not lower but raised population growth rates in the most heavily
III. Endogenous Factor Accumulation
Barro (1991) and Mankiw, Romer, and Weil (1992) were among the first
to stress the existence of conditional convergence in the
post-World War II cross-section of the economic growth rates of
nation-states' economies. Mankiw (1995) interprets this as indicating
that the straightforward Solow growth model is working better and
better as time passes: it is becoming more and more the case that
differences across nations in relative GDP per capita levels are
reflections of the differences in steady-state capital intensity
implied by their rates of factor accumulation and population
Yet the appearance of conditional convergence--a coefficient of
between -2 and -3 percent per year when the growth rate is regressed
on the difference between an economy's initial GDP per capita level
and the steady-state level implied by its investment and population
growth rates--fits oddly with the fact, documented in the previous
section, of unconditional divergence. How can economies be traversing
toward their steady states and at the same time drawing
further and further apart in relative GDP per capita levels?
A naive interpretation of this pattern would suggest that at some
time in the past nation-states' savings and population growth rates
must have been much more closely bunched together than they are
today. This would mean that at that time in the past economies'
steady-state and actual output per capita levels were bunched
together more closely than they are today. And that some economic
shock or series of shocks has since driven their respective savings
and population growth rates apart--and thus that the world's
distribution of relative incomes has diverged since, as the world's
economies have traversed toward their--now distantly
separated--steady-state growth paths.
But this naive interpretation has a central problem: what was this
shock that drove savings and population growth rates apart? The
principal candidate would be the Industrial Revolution. But the
Industrial Revolution saw not a fall but a sharp rise in population
growth rates in the most heavily affected economies (see Livi-Bacci,
1992). And today very little is left of Rostow's (1957) bold
hypothesis that the key to the Industrial Revolution was a sharp rise
in investment as a share of national product (see Crafts, 1985;
Mokyr, 1985). The shifts in investment and population growth rates
brought about by the Industrial Revolution do not go the right
Other candidates for a shock to drive economies' investment and
population growth rates sharply enough away from one another to
generate the observed divergence seen over the past century are
simply absent. The overwhelming bulk of divergence in GDP per capita
over the past century and a quarter has been due to the uneven spread
of the Industrial Revolution, and to differences in relative national
rates of total factor productivity growth.
But why, then, the finding of conditional convergence, and the strong
positive association of GDP per capita levels with investment rates
and the negative association with population growth rates?
Population Growth and the Demographic Transition
One reason is the endogeneity of population growth. Sometime between
the fifteenth and the eighteenth centuries the human race passed
through what we all hope was its last "Malthusian" episode, in which
rising population and limited agricultural resources led to
nutritional deficits, higher than average mortality, and population
stagnation. Since then the pace of productivity improvement in
agriculture has kept ahead of agricultural resource scarcity and
population growth (that has carried the world's population from one
to six billion so far). Nutrition has been relatively high by
historical standards, natural fertility high as well, and natural
In the past, the richest human populations appear to have also seen
the fastest population growth. But starting perhaps in eighteenth
century France a new pattern began to emerge, in which increases in
GDP per capita led not to greater fertility and faster population
growth but to lower fertility and slower population growth. The
number of girls born per potential mother fell, and population growth
Figure 4 shows this pattern at work in the United States over the
past two centuries: as GDP per capita has grown, the rate of natural
increase of the U.S. population has fallen steadily. Once U.S. GDP
per capita grew beyond the $2000 or so 1993 dollars level, fertility
began to drop sharply enough to offset the declines in mortality that
also accompanied better medicine and rising material prosperity. The
rate of population growth, excluding net immigration, is now little
over one percent per year--far below the 3.5 percent per year in
natural population increase seen in the first half-century of the
The pattern of rising material prosperity and falling natural
population increase has had only one significant interruption in the
United States in the past two centuries. The Great Depression of the
1930s saw a very sharp fall in childbearing, and a reduction in
natural population growth in the 1930s to only 0.7 percent per year.
In what Richard Easterlin (1982) sees as a delayed positive response
to the Great Depression that balanced out the birth deficit of that
decade, births rose to a level not seen since the nineteenth century
in the "baby boom" of the 1950s.
The pattern of increasing material wealth and slowing population
growth seen in the United States is completely typical of the pattern
that has so far been followed by all nations that have successfully
industrialized. Each tripling of GDP per capita is associated with an
approximately one percentage point per year fall in the rate of
natural population increase.
To my knowledge, no one has ever argued that falling population
growth has any sources other than in the increasing material
prosperity of the United States, and in the changes in social and
economic organization that have followed from the United States's
growing material wealth. A richer country has more literate women,
and literate women--worldwide--are very interested in effective birth
control. In a poorer country the average level of education is low,
and children can be put to work at a relatively early age, thus
augmenting the production resources of the household. In a richer
country the average level of education is high, and children are a
major drain on household cash flow for nearly two decades.
Children in relatively poor, low-productivity economies are much more
like an "investment" good than are children in rich,
high-productivity economies: they are a way to augment the economic
resources of the household in a time span of a decade or so. By
contrast, children in relatively rich, high-productivity economies
are more like a "consumption" good.
Thus we would expect--and we do see--a substantial correlation
between high GDP per capita and low population growth arising not so
much because low population growth leads to a higher steady-state
capital-output ratio but because of the demographic transition: the
changes in fertility that have so far been experienced in every
single industrialized economy.
The Relative Price of Investment Goods
Begin with the large divergence between purchasing power parity and
current exchange rate measures of relative GDP per capita levels. The
spread between the highest and lowest GDP per capita levels today,
using current exchange rate-based measures, is a factor of 400; the
spread between the highest and lowest GDP per capita levels today
using purchasing power parity-based measures is a factor of 50. If
the purchasing power parity-based measures are correct, real exchange
rates vary by a factor of eight between relatively rich and
relatively poor economies. And the log GDP per capita level accounts
for 80 percent of the cross-country variation in this measure of the
real exchange rate, with each one percent rise in GDP per capita
associated with an 0.34 percent rise in the real exchange rate.
Real exchange rates are such as to make the prices of traded
manufactured goods roughly the same in the different nation-states of
the world, putting to one side over- or undervaluations produced by
macroeconomic conditions, tariffs and other trade barriers, and
desired international investment flows. Thus the eight-fold
difference in real exchange rates between relatively rich and
relatively poor economies is a reflection of an approximately
eight-fold difference in the price of easily-traded manufactured
goods: relative to the average basket of goods and prices on which
the "international dollar" measure is based, the real price of traded
manufactures in relatively rich countries is only one-eighth the real
price in relatively poor countries.
This should come as no surprise. The world's most industrialized and
prosperous economies are the most industrialized and prosperous
because they have attained very high levels of manufacturing
productivity: their productivity advantage in unskilled service
industries is much lower than in capital- and technology-intensive
And a low relative price of technologically-sophisticated
manufactured goods has important consequences for nation-states'
relative investment rates. In the United States today machinery and
equipment account for half of all investment spending; in developing
economies--where machinery and equipment, especially imported
machinery and equipment is much more expensive--it typically accounts
for a much greater share of total investment spending (see Jones,
1994; DeLong and Summers, 1991).
Consider the implications of a higher relative price of capital goods
for a developing economy attempting to invest in a balanced mix of
machinery and structures. There is no consistent trend in the
relative price of structures across economies: rich economies can use
bulldozers to dig foundations, but poor economies can use large
numbers of low-paid unskilled workers to dig foundations. But the
higher relative price of machinery capital in developing countries
makes it more and more expensive to maintain a balanced mix: the
poorer a country, the lower is the real investment share of GDP that
corresponds to any given fixed nominal savings share of GDP.
Table 2 shows the consequences--the gap between nominal savings and
real investment shares of GDP--that follow from the high relative
price of machinery and equipment in poor countries that wish to
maintain a balanced mix of investment in structures and equipment.
For a country at the level of the world's poorest today--with a real
exchange rate-based GDP per capita level of some $95 a year--saving
20% of national product produces a real investment share (measured
using the "international dollar" measure) of only some 5% of national
In actual fact poor economies do not maintain balanced mixes
of structures and equipment capital: they cannot afford to do so, and
so economize substantially on machinery and equipment. Thus here are
two additional channels by which relative poverty is a cause slow
growth: first, relative poverty is the source of a high real price of
capital, a low rate of real investment corresponding to any given
nominal savings effort, and a low steady-state capital-output ratio;
second, to the extent that machinery and equipment are investments
with social products that significantly exceed the profits earned by
investors (see DeLong and Summers, 1991), the price structures in
relatively poor developing economies lead them to economize on
exactly the wrong kinds of capital investment.
The standard Solow (1956) and Swan (1956) growth model, written in
per worker terms and expressed in logs, contains the production
where y is output per worker, k is capital per worker, [[alpha]] is
the capital share in the production function, and [[tau]] is the log
of total factor productivity. If the economy has a constant
investment rate I, a constant population growth rate n, and has labor
efficiency growth and depreciation rates g and d, then in steady
state at any point in time output per worker will be given by:
Suppose, however, that we take account of the feedback from GDP per
capita levels on population growth rates:
where n is that portion of ln(n+g+d) that is not accounted for by the
combination of the dependence of population growth on output and the
background rates of labor efficiency growth and depreciation. The
pattern of the demographic revolution from the United States's
historical experience suggests that the parameter [[phi]] is, over
the relevant range, approximately equal to 0.2.
And suppose we take account of the feedback from GDP per capita
levels to the real investment share:
where s is the economy's nominal savings share, pk is the real price
of capital goods, [[eta]] is the deviation of the price of capital
goods from what would have been predicted given the level of real
output, and [[theta]]--the elasticity of capital goods prices with
respect to output--is roughly equal to 0.3 over the range relevant
for developing economies.
Combining (2), (3), and (4) produces an expression for the
steady-state level of output allowing for the endogeneity of
population growth rates as a result of the demographic transition,
and for the dependence of the relative price of investment on output
Equation (5) allows us to calculate, for various possible values for
the share [[alpha]] of produced capital goods in the production
function and for the chosen values of [[phi]] and [[theta]], the
impact on the level of the steady-state growth path of a shift in the
exogenous component of savings, capital goods prices, population
growth, or total factor productivity. Because they enter
symmetrically into equation (5) the effects of the first three are
Table 3 reports that--with a produced factor inputs share in the
production function of 0.4--a one percent increase in the savings
rate (or a one percent fall in the exogenous component of capital
goods prices) carries with it a one percent increase in the
steady-state level of output. But a one percent increase in total
factor productivity raises the steady-state level of output by fully
2.5 percent. Growth-accounting decompositions would, if applied to
such an economy, attribute only one percent of the higher level of
output to higher total factor productivity--less than one-sixth of
the total effect. The growth accounting decomposition is not wrong,
but incomplete: to the extent that the higher capital stock is a
result of higher total factor productivity reducing the relative
price of capital, and to the extent that higher total factor
productivity pushes an economy further along its demographic
transition to low population growth, exogenous shifts in total factor
productivity have effects orders of magnitude greater than growth
accounting procedures suggest, even without any powerful
externalities in the production function.
As interesting, perhaps, is the case in which there are
externalities to investment--whether in infrastructure, in research
and development, in human capital, or in machinery and equipment--and
in which the true capital share [[alpha]] in the production function
is substantially greater than the 0.4 found in the usual
specifications of the Solow model. The true capital share cannot get
as high as 0.67 without triggering explosive paths for output per
capita, in which very small boosts to total factor productivity set
in motion patterns of population growth reduction and investment
increase that converge to no steady state at all, but simply grow
until the log-linear approximations in equations (3) and (4) break
It is difficult to look at the cross-country pattern of growth over
the past century without thinking that the determinants of the
steady-state growth paths toward which countries converge must be
nearly singular. What difference between Canada and Argentina in 1870
would have led anyone to forecast their--now more than two and a
half-fold--difference in GDP per capita? Or the twenty-fold gap
between Taiwan and India? Recognizing the endogeneity of the
demographic transition and of investment has the potential to help us
understand why the economic history of the past century and a quarter
has proceeded as it did, without requiring assumptions of external
effects that seem--perhaps--implausibly large.
The endogeneity of the demographic transition, and of investment,
also helps make sense of the odd combination of global divergence
together with "conditional convergence." To the extent that
relatively low productivity today is a cause of an economy's
attraction to a low steady-state growth path, it is less necessary to
look for shocks in the past that both (a) pushed economies away from
their long-run growth paths, and (b) pushed economies' GDP per capita
levels together, if we want to account for the evolution of the
world's distribution of income.
But I do have one important caveat: I want to praise "technology," as
it appears in the aggregate economic growth models; but there is a
sense in which I also want to bury it.
Robert Solow's (1957) article is entitled "Technical Change and the
Aggregate Production Function." Certainly since 1957 economists have
used "technical change" and "technology" as shorthand ways of
referring to shifts in the aggregate production function.
Yet do we really want to do this? Much of the difference seen across
nations in aggregate total factor productivity seems to have little
to do with technology--in the sense of knowledge of the
internal combustion engine, continuous-casting, or the freeze-drying
Consider Greg Clark's (1987) excellent study of productivity in the
cotton textile industry circa 1910. Table 4 reports some of
Clark's calculations, most strikingly the seven-fold difference in
labor productivity found between mills in the United States and
cotton mills in the region of China near Shanghai.
The most striking thing about this seven-fold differential--the point
of Clark's article--is that all of these mills used the same
technology, if that word has any meaning. Japanese, Chinese,
and Indian cotton mills had no local source of capital goods. So they
bought and imported textile machinery made in the same machine shops
near Liverpool as did British manufacturers. The United States
produced its own textile machinery; Belgium, France, Germany, Austria
produced textile machinery as well. But everyone else imported the
capital goods--and in many cases, according to Clark, paid British
mechanics to assemble and install it as well.
Yet with the same technology--the same machinery, the same
production process, the same automated transformation of raw
materials by metal and chemistry into final product--Clark found
differences in labor productivity that reached three-to-one even
comparing the United States to Italy, a country with a very long
history of textile production.
The key to the difference in labor productivity is found in the last
column of table 4: staffing levels. In the United States, one
operative took care of three machines. In China, two operatives took
care of one machine. Add this six-fold difference in staffing levels
to the perhaps fifteen percent lower output per machine-hour near
Shanghai to obtain an arithmetic explanation of the seven-fold
difference in output per worker.
Since Clark wrote his article, a cottage industry has sprung up to
try to explain how all of these textile mills could still be
operating on the same production function. Perhaps the extra workers
in the Asian mills were substituting for a poorer quality of raw
materials? After all, poorer-quality raw materials would lead to more
breaks, snarls, and machine stoppages that would have to be
corrected. Perhaps the extra workers in the Asian mills allowed the
machines to run faster? Perhaps the extra workers allowed the
machines to run with less downtime? None of these attempts to
establish that these textile mills were working on the same
production function, with Asian mills getting increased output (or
diminished other inputs) in return for their higher staffing levels
has yet been convincing.
The turn-of-the-last-century cotton textile industry did exhibit very
large differences in productivity across countries, yes. But the
differences do not seem to be readily attributable to differences in
anything I would call technology.
Or consider the McKinsey Global Institute's (1993) study of
manufacturing productivity in the United States, Germany, and
Japan--a study carried out with the assistance of Martin Baily and
Robert Solow. As best they could estimate, Japanese manufacturing
productivity in 1990 varied from 33 percent of the U.S. level in food
processing to 147 percent of the U.S. level in consumer electronics.
German manufacturing productivity varied from 43 percent of the U.S.
level in food processing to 91 percent of the U.S. level in
It is hard to understand--if we are going to attribute these
productivity differences to differences in technology--how
Japanese businesses can be so successful at learning and developing
technologies for making automobile parts, and so inept and learning
and developing technologies for making frozen fish.
True differences in technology are surely a greater factor in
comparisons between countries further apart in the world distribution
of GDP per capita than Germany, Japan, and the United States:
developing economies do use last generation's or last century's
procedures and practices because they cannot afford the capital goods
that embody today's, because they do not have the mechanics to
maintain this today's, or because they have different factor price
structures that make it more costly to use today's best practice. But
even identical technologies can yield very different productivities.
There is a lot more going on.
Thus the moral of this paper is that "technology" is a more and a
less important factor in accounting for relative national levels of
prosperity than the conventional wisdom suggests.
"Technology"--in the sense of differences in total factor
productivity--is more important because of the strong
endogeneity of population growth and capital investment rates.
Countries that are rich have low rates of population growth: they
have completed their demographic transitions to a régime in
which fertility is relatively low because children have become more
"consumption" than "investment" goods. Countries that are rich also
have relatively low prices of capital goods--hence a given share of
national product saved implies a higher investment to GDP ratio.
Hence being rich tends to make a nation-state's capital-output ratio
Thus small differences in total factor productivity can translate
into large differences in productivity levels and living standards,
once the feedback from a richer economy to higher investment and
lower population growth rates is taken into account. And studies that
examine the impact of total factor productivity differences on output
per capita that hold savings and population growth rates constant
understate the true long-run impact of raising total factor
On the other hand, technology--in the sense of knowledge of
the internal combustion engine, continuous-casting, or
freeze-drying--is much less important in accounting for differences
across nations. Many differences in total factor productivity are
related tenuously, or not at all, to differences in
technology. All of the textile factories at the turn of the
last century were equipped with the same or similar machines, many of
them from the same machine shops in Lowell, Massachusetts or
This should not be taken to imply that technology proper is
unimportant in long-run economic growth. It is very important in
those particular industries that are near the active edge of
technological expansion and are intensive in research and
development. Indeed, better technology today is the sole important
reason why we today have six to twenty times the standard of living
of our predecessors in 1870. But it has much less to do with the
sources of aggregate productivity differences across nations.
The last wave of research on aggregate growth theory called forth an
effort, by Abramovitz (1956, 1986) and Denison (1967) among others,
to try to decompose aggregate total factor productivity differences
into more interesting and meaningful components. It is too bad that
the current wave of research on aggregate growth has failed to
generate a corresponding effort.
Moses Abramovitz, "Resource and Output Trends in the United States
since 1870," American Economic Review 46:2 (May 1956), pp.
Moses Abramovitz, "Catching Up, Forging Ahead, and Falling Behind,"
Journal of Economic History (1986), pp. 385-406.
Robert Barro, "Economic Growth in a Cross Section of Nations,"
Quarterly Journal of Economics 106:2 (May 1991), pp.
Robert Barro and Xavier Sala-i-Martin, Economic Growth (New
York: McGraw Hill, 1995).
Ricardo Caballero and Adam Jaffe, "How High Are the Giants'
Shoulders?: An Empirical Assessment of Knowledge Spillovers and
Creative Destruction in a Model of Economic Growth," NBER
Macroeconomics Annual (1993), pp. 15-74.
Timothy Cogley and Matthew Spiegel,
N.F.R. Crafts, British Economic Growth during the Industrial
Revolution (Oxford: Oxford University Press, 1985).
J. Bradford DeLong, "Productivity Growth, Convergence, and Welfare:
Comment," American Economic Review 78:5 (December 1988), pp.
J. Bradford DeLong and Lawrence H. Summers, "Equipment Investment
and Economic Growth," Quarterly Journal of Economics 106:2
(May 1991), pp. 445-502.
Edward Denison, Why Growth Rates Differ (Washington: Brookings
Steven Dowrick and Duc-Tho Nguyen, "OECD Comparative Economic Growth
1950-85: Catch-Up and Convergence," American Economic Review
79:5 (December 1989), pp. 1010-1030.
Charles Jones, "Economic Growth and the Relative Price of Capital,"
Journal of Monetary Economics 34 (December 1994), pp.
Charles Jones, "Time Series Tests of Endogenous Growth Models,"
Quarterly Journal of Economics (May 1995), pp. 495-525.
Charles Jones, "R&D-Based Models of Economic Growth," Journal
of Political Economy (August 1995), pp. 759-84.
Michael Kremer, "Population Growth and Technological Change: One
Million B.C. to 1990," Quarterly Journal of Economics 108
(August 1993), pp. 681-716.
Simon Kuznets, "The Meaning and Measurement of Economic Growth," in
Barry Supple, ed., The Experience of Economic Growth (New
York: Random House, 1963).
Simon Kuznets, "Modern Economic Growth," American Economic
Review 63:3 (June 1973), pp. 247-58.
Ross Levine and David Renelt, "A Sensitivity Analysis of
Cross-Country Growth Regressions," American Economic Review
82:4 (September 1992), pp. 942-63.
W. Arthur Lewis, The Evolution of the International Economic
Order (Princeton: Princeton University Press, 1978).
Massimo Livi-Bacci, A Concise History of World Population
(Oxford: Blackwell, 1992).
Angus Maddison, Monitoring the World Economy, 1820-1992
(Paris: OECD, 1995).
Thomas R. Malthus, An Essay on the Principle of Population
N. Gregory Mankiw, David Romer, and David Weil, "A Contribution to
the Empirics of Economics Growth," Quarterly Journal of
Economics 107:2 (May 1992), pp. 407-37.
N. Gregory Mankiw, "The Growth of Nations," Brookings Papers on
Economic Activity (1995).
Edwin Mansfield, "How Rapidly Does New Industrial Technology Leak
Out?" Journal of Industrial Economics 34:2 (December 1985),
McKinsey Global Institute, Manufacturing Productivity
(Washington: McKinsey Global Institute, 1993).
Joel Mokyr, "Introduction" to Mokyr, ed., The British Industrial
Revolution (Totowa, N.J.: Rowman and Allanheld, 1985).
William Nordhaus, "Do Real Output and Real Wage Measures Capture
Reality? The History of Lighting Suggests Not" (New Haven: Cowles
Discussion Paper 1078, 1994).
Lant Pritchett, "Divergence, Bigtime" (Washington: World Bank
Discussion Paper, 1994).
Danny Quah, "Galton's Fallacy and Tests of the Convergence
Hypothesis," Scandinavian Journal of Economics 95:4 (Fall
1993), pp. 427-43.
Paul Romer, "Increasing Returns and Long Run Growth," Journal of
Political Economy 94:5 (October 1986), pp. 1002-1037.
Walt Whitman Rostow, The Stages of Economic Growth (New York:
Harper and Row, 1957).
Robert Solow, "A Contribution to the Theory of Economic Growth,"
Quarterly Journal of Economics 70:1 (February 1956), pp.
Robert Solow, "Technical Change and the Aggregate Production
Function," Review of Economics and Statistics 39:3 (August
1957), pp. 312-320.
Robert Summers and Alan Heston, "The Penn World Table (Mark 5): An
Expanded Set of International Comparisons," Quarterly Journal of
Economics 106:2 (May 1991), pp. 327-68.
Trevor Swan, "Economic Growth and Capital Accumulation," Economic
Record 32 (November 1956), pp. 334-61.
Alan Taylor and Jeffrey Williamson, "Convergence in the Age of Mass
Migration" (Cambridge: NBER Working Paper 4711, 1994).
Jeffrey Williamson, "Globalization, Convergence, and History,"
Journal of Economic History 56:2 (June 1996).
World Development Report 1995: Workers in an Integrating World
(New York: Oxford University Press, 1995).