China's Economic Growth 1978-2025

World Development Vol. 36, No. 10, pp.
16651691, 2008
2008 Elsevier Ltd. All rights reserved
0305-750X/$ - see front matter
www.elsevier.com/locate/worlddev
doi:10.1016/j.worlddev.2007.09.013
Chinas Economic Growth 19782025: What

We Know Today About Chinas Economic
Growth Tomorrow
CARSTEN A. HOLZ *
Princeton University, NJ, USA
Hong Kong University of Science and Technology, Kowloon, Hong Kong
Summary. This paper examines Chinas future growth prospects and the potential drivers of future growth using two approaches. It rst asks in how far Chinas recent economic development
matches standard growth patterns identied in development economics and trade theory. Second,
GDP is decomposed into income components, which in turn are explained, for the reform period,
by the quantity and quality of labor; future GDP can then be re-composed from future labor data
available today. Overall, Chinas economic growth is likely to continue at current rates through
2015 before it gradually slows. Such a growth has numerous implications.
2008 Elsevier Ltd. All rights reserved.
Key words economic growth, growth accounting, growth forecasts, development theories, education, China
1. INTRODUCTION
The rapid economic growth of China since
the beginning of economic reforms in 1978
has captured the imagination of Western commentators and researchers. The responses range
from outright pessimism about Chinas future
to fear of a strong China. Brown (1995) wondered who will feed China. Chang (2001) announced the coming collapse of China.
Henderson (1999) saw China as on the brink,
while Terrill (2003) wrote of the illusory nature of the market in most of the Chinese economy and that a crash looms because the
Leninist core of the regime is unchanged from
Maos construction of it in Yanan six decades
ago (pp. 329, 313).
At the other end of the spectrum are those who
envisage a strong China. Murray (1998) described China as the next superpower. A number
of authors view an all-powerful China as a threat
(Gertz, 2000, or Timperlake et al., 2002). 1 News
items imploring the devastation Chinese competitors are inicting on US industries, from
kitchenware and car tires to electronic circuit
boards and the futility of trying to match the
China price have become common fare. 2
What is undeniable is Chinas rapid economic

growth over the past 27 years, from the beginning of economic reforms in 19782005, of,
measured in gross domestic product (GDP),
on average 9.6% per year. In economic size, China in 2005 was surpassed only by the United
States, Japan, and Germany. 3 Its share in global growth 19952002 has been estimated at
25%, compared to 20% for the United States. 4
While Chinas future economic growth has
received much attention in the popular literature, research literature is scarce and more limited in scope. Douglas (2002) documented the
transformation in four industries in China
and predicted the imminent emergence of
world-class Chinese brands. Zeng and Williamson (2003) examined the international growth
prospects of Chinese rms (positive). Carter
and Rozelle (2001) asked if China will become
* The work described in this paper was supported by a
Grant from the Research Grants Council of Hong Kong
(Project HKUST6423/05H). I gratefully acknowledge
feedback received from discussants and participants at a
dozen seminars and conference presentations in 2003
07. Final revision accepted: September 24, 2007.
1665
1666
WORLD DEVELOPMENT
a major force in world food markets (yes, if

structural transformation occurs).
A number of studies try to explain Chinas
past economic growth. Much of the transition
literature focuses on institutional changes to explain past reform successesusually, if only
implicitly, measured as economic growth (e.g.,
Qian, 2000, 2003, or Woo, 1994, 1999). Lin
et al. (2003) explained Chinas reform success
with the adoption of a comparative-advantage
following strategy, allowing full play to Chinas plentiful endowment with labor. Others
analyze Chinas past economic growth within
an aggregate production function framework
(Wang & Yao, 2003; Wu, 2004; Young,
2003). Chow and Li (2002) obtained GDP values through 2010 by extending their 195298
estimation using variations of past factor input
growth rates; Chow (2002) further included an
estimate of the year 2020 GDP. A 1997 World
Bank policy analysis of current economic issues
is supplemented by an estimate of Chinas GDP
in 2020 based on a closed-economy model. A
systematic study of future economic growth in
China that makes full use of hard facts about
the future that are available today, however,
is still lacking. 5
Answers to the question of what we know today about the economic size of China 10 or 20
years from now are relevant not only for the
popular discourse on the new superpower and
for China threat theorists and military strategists, but also for economists. For example,
conventional economic wisdom holds that free
trade benets, by the laws of comparative
advantage, all countries in the long run. Thus,
a growing China, and growing trade between
China and the United States, benets the United States. But as Samuelson (2004) within the
framework of standard trade theory showed,
a productivity gain in one country in the production of the good in which the other country
previously had a comparative advantage may,
in fact, cause a permanent loss in per capita real
income in the other country (though not relative to autarky); references to China, as the
country with a productivity gain, abound.
This paper examines Chinas future economic
growth and some potential drivers of this
growth. The rst step is a linear extrapolation
of the stable, past growth trend into the future.
A number of scenarios are played through in a
comparison with the United States. But extrapolation is not a particularly convincing research
tool. Why should past trends continue into the
future?
One approach is to examine how China fares

with respect to standard growth patterns identied by development economics and trade theory. These are structural changes, catching
up, and factor price equalization. Chinas past
economic growth ts well with all three. Furthermore, Chinas reform period growth, within these three growth patterns, matches that of
Japan, Korea, and Taiwan at an earlier stage of
their development. Obviously, these four countries dier, as do the domestic and international
circumstances under which they experienced a
particular stage of development. But the dierences need not be systematic with respect to the
impact on the dependent variable economic
growth. In as far as these growth patterns have
already played out for Japan, Korea, and Taiwan in the past, they bode well for Chinas future economic growth.
A second approach is to decompose GDP
into its income components and to explain
these, for the reform period, with the quantity
and quality of labor. We know with near-certainty the size of Chinas working age population through 2015, and can project with high
reliability for several years after. In as far as
gains in high school and university enrollment
are unlikely to be reversed, a bottom line scenario for the quality of Chinas future laborers
can be developed. Future GDP growth can
then be re-composed, using the established relationships, from data on the future quantity and
quality of labor available today. One result of
the decomposition of GDP growth is the potential importance of future labor quality in driving continued economic growth in China. A
separate section directly compares education
levels and education trends in China and the
United States.
Projections cannot predict the future with
certainty. Potential complications abound,
from political risks to non-performing loans
in Chinas banking system, loss-making stateowned enterprises, pilfered pension funds, local
government budget decits, ruralurban
inequality, and environmental degradation.
This paper works within these limitations. It
does not attempt to provide an exhaustive list
of complications, examine each of them, quantify their potential impact, and to attribute a
likelihood to their occurrence. It predicts Chinas future economic growth under the
assumption that current and future problems
continue to be resolved as they arise, and that
events of catastrophic dimensions do not
occur. 6
CHINAS ECONOMIC GROWTH 19782025
2. EXTRAPOLATION OF PAST GROWTH

INTO THE FUTURE
Chinas year 2005 nominal GDP of 18.308
trillion RMB at the ocial average annual exchange rate of 8.1917 RMB per USD translates
into USD 2.235 trillion, compared to the year
2005 US GDP of 12.456 trillion. 7 The US
economy in 2005, thus, was almost six times
larger than the Chinese economy. However,
such a comparison neglects to take into account
the diering price levels in the two countries.
The Penn World Table 6.2 (PWT 6.2), which
reports comparable GDP data in international
dollars, assumes that the general price level in
the United States in 2004, the most recent year
covered, was 3.86 times higher than the Chinese
price level. Applying the 2004 price adjustment
factor to the 2005 GDP data, the US economy,
today, in purchasing power terms, is only 43%
larger than Chinas economy.
1667
In what year will Chinas GDP exceed that of

the United States, assuming that the past real
growth rates continue into the future? The answer, reported in Table 1, is obtained by applying the average annual real growth rate of the
period 19782005 of China and the United
States to their respective year 2005 GDP values,
and then noting the year in which Chinas GDP
value exceeds that of the United States 8 (the
initial year, 1978, reects the beginning of Chinas economic reforms). In the rst row shown
in Table 1, it is assumed that the PWT 6.2 price
adjustment factor of 3.86 (in 2004) continues to
apply into the indenite future. In this scenario,
Chinas GDP begins to exceed that of the United States, in purchasing power terms, in 2011.
If one further incorporates a 3% annual appreciation of the RMB, reecting the most recent
exchange rate developments, the year when
Chinas GDP exceeds that of the United States,
in purchasing power terms, moves further to
Table 1. First year in which Chinas GDP exceeds US GDP
1.
2.
3.
4.
5.
6.
Price adjustment?
Aggregate
Per capita GDP
Exchange rate adjustment?
GDP
National
China coast
China 5 prov.
US price level = 3.86 * PRC price level

+3% annual RMB appreciation
US price level = 3* PRC price level
No price adjustment
2011
2009
2015
2012
2033
2024
2035
2026
2040
2029
2057
2041
2024
2019
2027
2021
2041
2031
2021
2017
2023
2019
2036
2028
Base year for the projections is 2005. Annual growth rates used in the projections are the average annual real growth
rates of China and the United States in 19782005. The price adjustment factor of 3.86 is that of 2004, the most
recent year for which the PWT 6.2 provides data (The values of 200004 are: 4.217306, 4.223668, 4.264024, 4.186857,
3.859263).
China coast refers to the coastal provinces of China (in comparison to the national average in the United States); it
includes Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, and
Guangxi (the small province Hainan was omitted due to a lack of data for 1978 when Hainan was part of Guangdong
province). China 5 prov. refers to the ve fastest-growing coastal provinces in China (Jiangsu, Zhejiang, Fujian,
Shandong, and Guangdong). The share of the coast in Chinas population in mid-2005 is 42.5%, and in annual GDP
66.2%; the corresponding shares of the ve provinces are 25.9% and 43.3%.
Nominal and real GDP values of 1978 and 2005 are those published by the National Bureau of Statistics in China
and by the Bureau of Economic Analysis in the United States. Per capita GDP values are obtained by dividing
aggregate GDP by the mid-year population. United States population data for the year 2005 is the ocial estimate.
Chinese provincial-level population data of 1978 are based on the July 1, 1982 and the November 1, 2000 censuses,
with mid-1978 values obtained by extrapolation using the implicit average annual population growth rate of 1982
2000.
The average annual real growth rate of economy-wide GDP in China in 19782005 is 9.6358%; in the United States,
it is 2.9687%. The average annual growth rate of real GDP per capita in China is 8.3840%, and in the United States it
is 1.8821%. The average annual population growth rate in China is 1.1549% and in the United States it is 1.0665%.
The average exchange rate in 2005 is 8.1917 yuan RMB/USD.
Sources: PWT 6.2. Census 2000 Major Figures, pp. 16, 31. Statistical Yearbook 1991, p. 79; 2006, pp. 57, 60, 99, 734.
Bureau of Economic Analysis, at <https//www.bea.gov/bea/dn/home/gdp.htm> (choose Current-dollar and real
GDP, XLS; accessed on 30 Nov. 06). US population data are from <http://www.census.gov/popest/archives/1990s/
popclockest.txt> and for 200005 from <http://www.census.gov/popest/states/NST-ann-est.html> (both accessed on
November 30, 2006).
1668
WORLD DEVELOPMENT
the present, to 2009 (second row shown in

Table 1).
The price adjustment factor in the case of
China carries substantial uncertainty (Heston, 2001, p. 1). The multiplicative factor of
3.86 could be considerably o in either direction. 9 A price adjustment factor of three would
imply a cross-over year of 2015 or, with RMB
appreciation, of 2012 (third and fourth rows
shown in Table 1). At the extreme, making no
price adjustments whatsoever, Chinas GDP
equals that of the United States 25 years in
the future, in 2033, or, with RMB appreciation,
in 2024 (fth and sixth rows shown in Table 1).
In terms of per capita GDP, the year when
China surpasses the United States is 2050
years in the future (see additional columns
shown in Table 1). But what is far in the future
for the average Chinese person moves closer to
the present for the richer areas of China. One to
two decades from now, coastal China (as dened in the notes to Table 1), accounting for
42% of Chinas population in 2005 and exceeding the US population by 87%, may well be as
rich, per capita, as the average US citizen.
Focusing only on the ve fastest-growing provinces in coastal China, with a joint population
in 2000 that is 11% larger than the US population, moves the cross-over year a few years further to the present. In other words, by around
2020, a share of Chinas population that exceeds the size of the US population could enjoy
the same average standard of living as the United States; in addition, China has a poorer hinterland with roughly three times the population
in the richer areas. 10
In these calculations, all variables are held
constant except those explicitly allowed to vary.
In particular, in the extrapolation of aggregate
GDP, the use of an average annual past real
growth rate reects the product of the real
GDP growth rate per capita and the population
growth rate. In the short run, such as one decade, changes in the population growth rate are
likely to be minor, but this may not hold in the
long run. Similarly, the exchange rate was held
constant in three scenarios, as was the price
adjustment factor throughout (at the three different levels). 11 The further into the future, the
less valid are the assumptions. A more sophisticated approach would, furthermore, focus on
the employed instead of the population. But
none of these adjustments is likely to have a signicant impact in the short run, and the extrapolations at the aggregate level suggest that
much will change in the short run. 12
Extrapolations do not constitute a reliable research tool. The following two sections employ
theories and tools of economics to examine
Chinas future growth and its potential determinants.
3. DEVELOPMENT THEORIES AND
ASIAN PRECEDENTS
Explanations of Chinas economic growth
tend to focus on economic transition. The success of the reform process is explained by transition facts and strategies, where success is
usually taken to imply economic growth (or a
rise in living standards). For example, Woo
(1994) listed as crucial the creation of non-state
rms in every sector of the economy, a high
savings rate, good initial conditions (such as a
small extent of central planning, unemployment
in the countryside that could be taken up by
township and village enterprises, or a limited
social security net), historical conditions, and
the Chinese Diaspora. Qian (2000, 2003) ascribed much of Chinas reform success to the
unorthodox economic policy measures adopted
by Chinas leadership; the key to Chinas economic growth was the unleashing of incentives
and competition while making reform interestcompatible for those in power.
Yet these explanations of past economic reform successes (and thus economic growth) lack
a counterfactual. Wing-Thye Woo oers crosscountry comparisons, but the number of explanatory variables appears larger than the number
of comparison countries. Qian Yingyi makes
an in part historical argument, but oers no
explicit time series evaluation that examines the
status before and after a particular reform measure was implemented. These explanations, thus,
are not as strong as one might wish them to be. 13
If one takes the view that economic transition
in total has caused Chinas past economic
growth, then one could argue that the key elements of transition have been in place since
the early 1990s (price and domestic trade liberalization, the entry of private enterprises) and
that therefore the gains from transition have already been exhausted. Consequently, economic
growth should since have slowed, which it did
not, or it could slow any time soon. Alternatively, one could argue that past transition measures impact on economic growth over an
extended period of time, or that transition is
as yet incomplete, with further measures to
go, in which cases the gains continue.
Growth patterns identied by economic theories of development and trade perhaps oer rmer ground. Economic transition could then be
viewed as the removal of constraints that prevented well-known development patterns from
unfolding. Furthermore, if Chinas reform period growth patterns match those exhibited by
other countries at stages in their economic
development similar to Chinas development level in the reform period, then Chinas future
growth patterns may also match those of the
other economies later. The argument has two
foundations: one is a focus on standard growth
patterns, the other is a cross-country comparison with countries with which a comparison is
likely to be meaningful.
The growth patterns are structural change,
catching up, and factor price equalization. 14
These three patterns are not independent of
each other, nor do they hold irrespective of
the larger economic environment; they do not
always come with perfectly clear-cut causality
(though one causality, perhaps the most plausible one, is used to present each pattern). The
patterns have the advantage that they do not
rely on individual transition measures and that
they have been identied by development economics and trade theory as of relevance. In as
far as China is at the very early stages of these
accepted development patterns, with good reasons for these patterns to continue to unfold,
future economic growth is likely.
The comparison countries are Japan, Korea,
and Taiwan. These three countries are relevant
for China only if they do not dier systematically with respect to the relationship between
key variables. While these four countries dier,
as do the domestic and international circumstances under which they experienced a particular stage of development, that does not
necessarily invalidate the assumption of a similar relationship between key variables.
The choice of countries to compare China to is
a subjective choice, motivated only by the desire
to make comparisons across a relatively homogeneous group of countries. China may share
some economic growth patterns with Japan,
Korea, and Taiwan due to cultural similarities,
geographic location, similar economic development strategies, or, in the case of Japan, relatively large size of the domestic economy.
Limiting the analysis to Japan, Korea, and Taiwan allows the careful compilation of the necessary data from each individual countrys
statistical oce, with a very few missing data
points obtained from the World Bank Develop-
1669
ment Indicators database, the International

Financial Statistics (IFS), and the PWT. 15 An
attempt was made to cover the years 1950
2005 but data are often available only since the
1960s or 1970s. The earlier in the development
process of Japan, Korea, and Taiwan, the closer
to the case of China the initial conditions may
have been. Data on China are for the years of
the reform period (the years since 1978).
Throughout, the variable to be explained is
real GDP growth per laborer, because GDP is
produced only by the active working population, namely laborers. 16
(a) Structural change
As labor shifts from low-productivity agriculture to higher-productivity industry and services, economy-wide real GDP per laborer,
that is, real (partial) labor productivity, increases, if only because those laborers who have
shifted sectors now produce a multiple of their
former output value. Figure 1 shows just how
big, and increasing, the labor productivity differences are between the three economic sectors
in China. Consequently, one would expect to
see relatively high aggregate (economy-wide)
labor productivity growth in those years when
a relatively large number of laborers shift out
of agriculture.
Figure 2 shows that this is indeed the case.
Each data point reects a particular country
in a particular year. In the years with a high
absolute reduction in the share of laborers in
agriculture, the growth rate of (real, likewise
below) labor productivity is high. 17 This relationship holds equally for all four countries
and is remarkably similar for China, Taiwan,
and Korea. The pattern of decline over time
in the share of laborers in agriculture is also
similar across all countries, with Japan the rst
to have embarked on this process, followed by
Taiwan, Korea, and then China (Figure 3). 18
At Chinas 19782005 rate of decline in the
share of laborers in agriculture from just above
70% to 45%, with an annual reduction in the
share of laborers in agriculture by, on average,
an absolute value of 0.01 every year (thus, e.g.,
from 0.7 to 0.69 in one year), China has another 35 years to go before its agricultural labor
share reaches the 10% level at which the agricultural labor shares of Japan, Korea, and Taiwan stabilize. But this implies that China faces
another 35 years of structural change as a
source of economic growth, a long time span
in the context of the extrapolations.
WORLD DEVELOPMENT
Value-added per laborer (yuan RMB, current prices)
1670
50000
45000
Primary
40000
Secondary
Tertiary
35000
30000
25000
20000
15000
10000
5000
0
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
Growth rate of real GDP per laborer (in %)
Figure 1. Sectoral labor productivity in China. Employment data are mid-year values (as average of end-year values).
All employment data (sectoral, total) during 19891990 experience a statistical break; end-year 1990 total employment
exceeds that of 1989 by 17.03%. Ocial (unadjusted) employment data are used. The secondary sector comprises
industry and construction.
China
Taiwan
Korea
-0.035
-0.030
-0.025
-0.020
-0.015
-0.010
14
Japan
13
Korea
12
11
Taiwan
10
China
9
8
7
6
5
4
3
2
1 Japan
0
-0.005 -10.000 0.005 0.010
-2
-3
-4
Annual absolute change inshare of agriculture in employment
Figure 2. Structural change and labor productivity growth. Japan 19562005, Korea 19712005, Taiwan 19672005,
and China 19792005. (Pre-1991 employment data for China are adjusted as explained in appendix.) Two data points
are omitted in order to keep the chart compact: Korea 1972 (0.0224, 0.9448) and 1998 (0.0125, 4.6146).
The pattern of structural change is likely to

continue to unfold in China. Rural areas still
suer from underemployment in agriculture;
the number of agricultural laborers per land
area is more than one hundred times higher
than in the U.S. School leavers in rural areas almost invariably do not enter agriculture, and
this is unlikely to change. The development of
rural industries is ocially encouraged. Urban
income, including for unskilled labor, is still
Share of agriculture in employment
0.8
0.7
Japan
Korea
Taiwan
China
1671
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1953 1957 1961 1965 1969 1973 1977 1981 1985 1989 1993 1997 2001 2005
Figure 3. Share of agricultural laborers in economy-wide employment. Employment data are average annual values (in
the case of China, mid-year values). Values for Japan only comprise agriculture (presumably farming and animal
husbandry) and forestry; they exclude shery due to the lack of time series data.
signicantly higher than the income in many

rural areas, creating incentives for ruralurban
migration. The constraints that limited rural
urban labor ows in the pre-reform period have
been gradually relaxed and ocial policy is to
relax them further. The development of rural
land markets is on the political agenda, and,
once completed, will likely remove another
impediment to ruralurban migration. 19
(b) Catching up
A latecomer to economic development has
the advantage that production techniques and
technologies that have already been invented
can be copied. Taking the United States as
the leader in research and development, and
proxying the level of technological development
by labor productivity, the distance between a
particular countrys labor productivity (in
USD) and US labor productivity serves as a
measure of the potential scope for catching
up. One would expect to see relatively high real
GDP growth per laborer in a country whose
distance to the leading country (the United
States) is relatively high, that is, when the potential for catching up is large.
Figure 4 conrms the hypothesis. Japan
experienced high growth rates when its labor
productivity was 1050% of US labor productivity. As the gap closed, growth rates in Japan
fell. 20 Korea and Taiwan exhibit constant high
growth rates of labor productivity at an aver-
age 4% to 5% throughout all years. These two

countries level of economic development has
so far remained in the range of 5% to 50% of
the US level, a range in which Japan also exhibited high and constant labor productivity
growth rates. Korea and Taiwan may yet have
to experience the slow-down that comes when
the potential gains from catching up are exhausted. China has remained within such a narrow range of (low) development, with labor
productivity at only 1.2 to 3.4% of the US level
(using the ocial exchange rate to translate
Chinese yuan RMB values into USD values),
that a meaningful pattern is not yet discernible
(Figure 5). At Chinas highest level of catching
up so far, the growth rate of labor productivity
was a relatively high 9%.
Independent of the choice of data, China is at
a very low development level compared to the
United States. It appears to be at a stage of economic development where other Asian countries started out more than 30 years ago. 21
Even if China were to move along the pattern
only as far as Korea and Taiwan have, the
extrapolations presented earlier still apply.
The pattern of catching up is likely to continue to unfold in China as Chinas integration
into the world economy continues. Foreign
rms bring their knowledge of production processes if not their innovative capacity to China.
Chinese rms buy foreign rms abroad or set
up factories abroad, gaining more immediate
access to foreign knowledge. At the same time,
WORLD DEVELOPMENT
1672
14
13
12
11 China
10
9
8
7
6
5
4 Korea
3
2
1
0
-10.0 0.1
-2
-3
Japan
Korea
Taiwan
China
Taiwan
Japan
0.2
0.3
0.4
0.5
0.6 0.7
0.8
0.9
1.0
1.1
1.2
1.3
1.4
GDP per laborer (in USD) per U.S.GDP per laborer (in USD)
Figure 4. Catching up and labor productivity growth. Japan 19602005, Korea 19712005, Taiwan 19672005, and
China 19792005.
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0.012 0.014 0.016 0.018 0.020 0.022 0.024 0.026 0.028 0.030 0.032 0.034
GDP per laborer (in USD) per U.S.GDP per laborer (in USD)
Figure 5. Catching up and labor productivity growth in China, 19792005.
foreign knowledge becomes more accessible

within China due to the emergence of a Chinese
generation that is both signicantly better educated than previous generations and increasingly uent in English. A growing contingent
of Chinese has studied abroad. 22
(c) Factor price equalization
The factor price equalization theorem (or
Heckscher-Ohlin-Samuelson theorem) states
that factor prices, such as skill-specic wages,
should equalize between two countries as
long as a range of assumptions are met, 23
that is, the skill-specic wage rate of one

country divided by that of the other country
should equal unity. The slightly less restrictive, relative version of the factor price
equalization theorem focuses on the relative
prices of factors of production. Thus, for
example, a country with labor that is cheap
relative to capital should see demand for its
labor rise. As underemployed laborers
become fully employed (or labor shifts out
of low-productivity agriculture), labor productivity rises.
In the absence of a reliable price of capital,
the price of investment goods is used as a
proxy. The price of investment goods of any

particular country, relative to the United
States, is available in the PWT.
Figure 6 reveals a clear downward trend for
Japan. Relative factor price equalization, at
1040% of the US level, appears to have little
eect on Korea and Taiwan, which experience
continuously stable labor productivity growth,
with perhaps a slight decline in the most recent
years when approaching 50% of the US level.
The observations for China are within such a
narrow range, at only 27% of the corresponding US ratio, that a clear pattern does not yet
emerge. (If the China data are presented in a
chart of their own, omitted here, that chart
looks similar to Figure 5.) Chinas level of relative factor price equalization is well below the
early levels of the other three Asian countries.
If China were again at the beginning of a
long-run trajectory, then the patterns of the
other three Asian countries suggest that Chinas potential for economic growth from relatively low labor costs will continue to exist for
another 30 years. 24
Chinas relative wages are currently still so
low that they are likely to remain internationally attractive for many years to come. Access
to world trade translates the low relative wages
into investment in labor-intensive industries in
China and thus into industrialization and ultimately labor productivity growth. As long as
China participates in world trade, there appears
14
13
12
11
China
10
9
8
7
6
5
4
3
2
1
0
-1 0.0
0.1
-2
-3
-4
-5
1673
no reason why the gradual process of factor

price equalization should not continue to unfold in China.
This section applied three explanations of
economic growth from development economics
and trade theory to the case of China. Chinas
past economic growth matches these standard
development patterns, as does economic
growth in Japan, Korea, and Taiwan. In as
far as Chinain comparison to Japan, Korea,
and Taiwanis located at the early stages of
each pattern, there remains much scope for future gains in labor productivity and thereby
growth.
These broad development patterns are obviously not etched in stone and ignore numerous
other factors that may impact on labor productivity growth. One example is short-run deviations from the trend (as exhibited in the
gures) due to such factors as business cycles
or discrete decisions on the systemic features
of the economy. Similarly, what was the case
for Japan, or Korea, or Taiwan may not hold
for China, or may not occur at the same stage
of development in China as it has in the other
countries. One example would be labor mobility (but the restrictions on ruralurban labor
mobility in China are falling), or weaknesses
in the legal system that may limit adoption of
new technologies. Nevertheless, as the gures
attest, these broad patterns tend to play out
in the long run. 25
Japan
Korea
Taiwan
China
Taiwan
Japan
Korea
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Relative wage rates, divided by relative investment prices (PWT)

Figure 6. Relative factor price equalization and labor productivity growth. Japan 19602004, Korea 19712004, Taiwan
19672004, and China 19792003. For China, 2004 income approach GDP data (with the component labor
remuneration, needed to derive wages) are not available. The relative investment prices (only) are from the PWT, with
2004 as the nal year for which data are available. Wages and investment prices are relative to the United States.
1674
WORLD DEVELOPMENT
4. GROWTH ACCOUNTING
26
Growth accounting decomposes GDP growth

into growth of variables other than GDP, that
is, growth of GDP is explained by the
weighted growth rates of other variables. 27 If
GDP growth and the growth of the variables
into which it is decomposed exhibited a stable
relationship in the past, and if this stable relationship is likely to continue into the future,
then information about the future values of
the variables into which GDP has been decomposed allows derivation of future GDP growth.
One way to decompose GDP is to assume the
existence of an aggregate production function
and to explain GDP growth by the growth of
factor inputs and of a residual, with the weights
of factor input growth given by their estimated
output elasticity. Estimation of an economywide production function for China in 1978
2002 yields insignicant coecients, that is, in
the typical Cobb-Douglas production function,
the output elasticities are not constant over
time. Perhaps for good reason, authors from
Wang and Yao (2003) to Young (2003) did
not actually estimate a production function,
but assumed weights. They assumed constant
weights and set them approximately equal to
the factor shares; this implies the assumptions
of constant output elasticities (despite the evidence to the contrary) and perfect competition
(unlikely to be met in a planned economy). 28
A dierent growth decomposition that is free
of assumptions follows from the denition of
GDP on the income side:
GDP labor remuneration + depreciation +
net taxes on production + operating surplus.
The data on these income components are
available for China, for all years of the reform
period, at the provincial but not at the national
level. Summing across provinces, and applying
the sum provincial shares of the dierent income
components to national GDP yield an approximation of national income components. 29
Net taxes on production reect taxes on production, such as the value added tax, less production subsidies, such as subsidies for policy
losses, price subsidies for the grain system,
and tax refunds for exporting enterprises. The
largest part of this income component is likely
to depend on value added. Net taxes on production are in the following taken to be a function of value added (GDP).
Depreciation depends on the value of the not
yet fully depreciated xed assets. It is in the following taken to be a function of capital.
The operating surplus approximately reects

economy-wide business prot. 30 It is the return
on equity. Consolidating balance sheets across
the economy and eliminating the nancial sector so that deposits at banks combined with
bank loans to enterprises translate into household and rm claims on production units, operating surplus constitutes the return on xed
assets, intangible assets, and inventories. Since
inventories are not productive, operating surplus is largely a function of xed and intangible
assets. With data on land values not available
(and presumably even today much smaller than
the value of xed assets), only xed assets remain as feasible argument.
The national income identity, that is, the definition of GDP in the income approach, can
then be rewritten as
P t Y t wt Lt dt K t st P t Y t st K t ; or
wt
Kt
Y st Y t 1 st Lt dt st ;
Pt
Pt
1
2
where P denotes the price level, Y the real

GDP, w the wage rate, L the quantity of labor,
K the nominal xed asset stock (capital), and
dt, st, and st the depreciation rate, the net tax
rate on production, and the surplus rate. Taking derivatives with respect to time, and then
dividing by GDP less net taxes on production,
while expanding right-hand side terms, yields
dY st
dt
Y st
dY t
dt
1 st
Y t dsdtt

Y t 1 st Y t 1 st
w
t
dP t
dt
wt
Pt
Lt
wt
Pt
Y st
ddt
dst
t
wt Lt dL
dt
dt K t dt
dt
P t Y st Lt dt P t Y st st
d
Kt
Pt
t
K t st
dt P t

:
s dt st K t
Pt Y t
Y st
Pt
Using hats to denote growth rates and abbreviating the shares of labor, depreciation, and
operating surplus in GDP less net taxes on
production by ast ; d st , and est lead to the income
growth identity
dst
dt
Yb st Yb t
1 st

bt
s w
ast b
L t d st ^
dt
at
Pt
!
bt
K
s
s
s
:
et ^st d t et
Pt
Like the behavioral growth accounting equation derived from an assumed aggregate production function, the income growth identity
represents one form of GDP decomposition,
only now in the form of a denitional identity.
Compared to the traditional growth accounting
equation, the income version newly introduces
the growth rates of the real wage rate, of the
depreciation rate, of the surplus rate (or rental
rate of capital), and of the net tax rate on production; in exchange, there is no room for a
residual. 31
In the period 19782002, nominal GDP
deated by the implicit deator as rst
published grew by 827%, and nominal GDP
less net taxes on production, deated by the implicit deator as rst published, grew by
814%. 32 The dierence of only thirteen percentage points implies that the changes in the
net rate of taxes on production are negligibly
small. The income decomposition, using the
average income shares of 19782002, that is,
weighting factor growth with the mean of the
rst and last periods income share values, is
as follows:

bt
s
s w
b
Y t at
ast b
L t d st ^
dt
Pt
!
bt
K
s
s
s
et ^st d t et
Pt
814% 0:5981 495% 0:5981 60%
0:1370 16% 0:2649 50%
0:4019 1185%
296% 36% 2%13%
476% 797%
In the period 19782002, growth in constantprice capital accounted for approximately
60% of real GDP growth (476/797) and growth
in the real wage for 37%; growth in the quantity
of labor contributed just below 5%, with the
remainder due to very small positive and negative contributions of changes in the depreciation rate and in the surplus rate.
The income growth identity only holds with
perfect accuracy in the instantaneous case when
growth rates are innitesimally small and the
factor shares at a given point of time therefore
are exactly applicable. Measuring growth rates
over 24 years introduces a discrepancy of 814%
versus 797% due to the need to apply average
factor shares. The discrepancy is small, that
is, using mid-period weights has little impact
on the outcome.
1675
The share of labor remuneration in nominal

GDP less net taxes on production was rather
constant in the period 19782002, with a mean
of 0.5981 and a standard deviation of 0.0110
(Figure 7); consequently, the weight of real capitalunity less the labor share, or the sum of
the shares of depreciation and operating surpluswas also near-constant.
The shares of depreciation and of the operating surplus varied over time, but they appear
individually only in the products with the
growth rate of the depreciation rate and of
the surplus rate, and these products are relatively small. The product of the depreciation
share and depreciation rate growth in the income decomposition of GDP contributed only
0.25% of GDP growth in 19782002 (2% divided by 797%), and the product of surplus
share and surplus rate growth only (negative)
1.63%. These two terms, thus, can be ignored
without any signicant impact on the explanation of GDP.
The growth decomposition of 19782002 suggests that when forecasting future GDP
growth, (i) the growth of GDP less net taxes
on production is a good proxy for real GDP
growth and (ii) future GDP growth can be recomposed solely from the growth rates of the
future real wage, labor, and real capital, that is
Yb t Yb st

bt
w
s
b
at
L t d st est
Pt

bt
w
ast
b
L t 1 ast
Pt
!
bt
K
Pt
!
bt
K
:
Pt
The future growth rates of the quantity of labor are readily available if one is willing to
make a few, plausible assumptions. Those aged
15 in 2015 (the minimum working age) were already born at the time of the year 2000 population census, that is, forecasts through 2015 are
highly reliable, and in as far as trends in reproduction rates have been stable in recent years,
forecasts through later years are likely to only
gradually lose reliability. 33
What remains is to obtain future growth
rates of the real wage and of real capital. Ideally, these growth rates are related as much as
possible to variables whose future values are already known todaythe quantity of labor, the
quality of labor, and age measures. When the
unit of analysis is the individual, these are natural explanatory variables for the real wage. 34
1676
WORLD DEVELOPMENT
0.65
0.60
0.55
0.50
0.45
0.40
Labor remuneration
Depreciation
Operating surplus
0.35
0.30
0.25
0.20
0.15
0.10
1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002
Figure 7. Income factor shares in GDP less net taxes on production, 19782003. The values of all income components
are by necessity derived from provincial-level data published in GDP 195295, GDP 19962002, and for the years since
the mid-1990s in the corresponding years Statistical Yearbook. 2004 values are not available. For each year, the share of
each income component in GDP less net taxes on production is obtained as the sum across provinces of the values of a
particular income component divided by the sum across provinces of GDP less net taxes on production.
Since they are the only hard facts that we know

about the future, they are also used to explain
real capital. This is plausible at least for the
quality of capital, because more sophisticated
capital is likely to require a higher level of labor
quality. To some extent, it may even be plausible for the quantity of capital, if agriculture requires little education and little capital, and
those who move into other sectors with a higher capital-labor ratio are the most educated.
Examining three dozen potentially wage-related variables in the period 19782002, from
average years of schooling and share of laborers with a particular level of schooling (primary, lower middle, upper middle, college,
university, post-graduate) to age characteristics
and potentially relevant non-labor variables,
such as the share of foreign direct investment
in total domestic investment, most of these series are unit root processes.
The solution then is to nd a long-run relationship using the time series data of the period
19782002. Table 2 reports two long-run relationships embedded in vector error correction
(VEC) models, one with variables in levels,
the other with variables in logarithms. The
same procedure was used for real capital and
those results are also reported in Table 2. 35
The rst signicant long-run relationship
(cointegrating equation) reported in the table
is between the real wage, the share of laborers
with upper middle school education, and the
share of laborers with college level education

or above. For simplicity of exposition, the
relationship is expressed with the wage as
left-hand side variable. The coecient signs
are plausible and the coecients are highly
signicant. Forecast error variance decomposition shows that much of the movement in the
wage series is due to its own shocks, but over
a 10-year forecast horizon, 22% of the forecast
error variance of the wage series is due to
shocks to the share of laborers with college
level education or above. 36 The other three
cointegration equations reported in the table
also have plausible coecient signs and the
coecients are highly signicant. 37 The two
cases with variables in natural logarithms
come with strong Granger causality from the
education variables to the wage and capital
series, and over a 10-year forecast horizon
the forecast error variance of capital is mostly
accounted for by shocks to one of the education variables. 38
The rst four data columns of Table 3 report
the average annual growth rates (over ve-year
periods) that are obtained by tting the four
cointegration equations of the previous table
to the forecast labor quantity and quality values through 2025. When the cointegration
equations are estimated in logarithms, the
resulting growth rates tend to be higher, especially far into the future. The forecast quantity
of labor series (fth data column) has a falling
1677
Table 2. Cointegration equations, 19782002/03

Dependent variable
Regression number
Constant
C. Average years of schooling
minus 5
All variables
in levels
All variables in
natural logarithms
A. Wage
I
B. Capital
II
A. Wage
III
B. Capital
IV
4387.96
4144.38
6.2426
0.7171
(12.84)
7.6163
0.2829
(3.18)
D. Average years of schooling

minus 5, squared
3126.44
(23.20)
1.3593
(5.02)
E. Share of laborers with primary

school education
F. Share of laborers with upper
middle school education
39066.18
(4.07)
G. Share of laborers with college

level education or above
152080.0
(18.80)
Lag length of the VEC

Trace, maximum eigenvalue test (%)
Normality (10% level)
Granger causality Y: X (Z%)
2
1, 5
OK
None
13973.42
(2.55)
238.63
(2.39)
2
1, 1
OK
H:D(10)
Variance decomposition period 3
A: 97/1/2
F: 34/66/0
G: 1/78/21
A: 66/12/22
F: 16/81/3
G:19/60/21
B: 87/11/2
D: 19/77/5
H: 83/1/17
B: 25/70/5
D: 23/72/5
H:33/45/22
1
1, 1
C 1%, E 10%
A: E(0.1)
C: E(5)
A: 82/3/15
C: 0/94/6
E: 0/61/38
A: 65/4/31
C: 11/89/10
E:13/60/27
23
22
24
H. Number of laborers
I. Trend
Variance decomposition period 10
Number of observations
0.0946
(15.23)
2
5, 5
OK
B:I(1)
B: 87/13
C: 8/92
B: 29/71
C: 6/94
22
Wage per laborer: labor compensation in the national income accounts divided by the number of laborers including military
personnel (in yuan per laborer); deated by the GDP deator (in 2000 prices).
Capital: measure of capitals contribution to production, in billion yuan, deated by the GDP deator (in 2000 prices).
Number of laborers is in billion.
Values in parentheses underneath coecient estimates are asymptotic t-values (with standard errors corrected for degrees of freedom). All coecients are signicant at the 1% level, except the ones of H and I in the second data column (signicant at the 5% level).
Trend: the cointegrating vector always includes a constant (as does the VAR); depending on the results of the trace test and the
maximum eigenvalue test, a trend is included or not included.
The logic behind the use of the variable average years of schooling minus ve (squared in one case) is that a primary school
education is standard, and wage dierentiation then depends only on the number of years of schooling in excess of primary school
education. While primary school lasts for six years, the 1978 average years of schooling is just below 6 years; in order to have a
positive variable throughout (necessary to take logarithms), 5 years is used as a benchmark.
Trace, maximum eigenvalue test: signicance level of the existence of one eigenvector.
Normality in VEC equations: Jarque-Bera values are used (based on Cholesky (Lutkepohl) orthogonalization). When normality is
rejected for one of the VEC equations at the 10% level, the signicance level is stated together with a letter which identied the
variable that, in rst dierences, is the dependent variable of the particular equation.
Granger causality Y:X (Z%) means that Y is Granger-caused by X (at the 0.1%, 1%, 5%, or 10% signicance level).
Variance decomposition: a Cholesky ordering of variables as listed above in the particular table column was used, and the forecast
error variance decomposed annually over periods of 3 and 10 years. Values are percentages, ordered as the variables appear in the
column. For example, if the three variables involved are A, C, and D, and the variance concerned is that of A, then A/C/D % of the
variance of A is explained by variation in A, C, and D. Percentages may not add up to 100 due to rounding.
1678
WORLD DEVELOPMENT
Table 3. Growth forecasts (average annual growth rates in %)
Individual variables
200005
200510
201015
201520
202025
200025
Cumulative
200025
GDP
Wage
Capital
Wage
Capital
Labor
Eqn. (5)
Labor share
assumed xed
(I)
(II)
(III)
(IV)
(V)
I, II, V
III, IV, V
I, V
III, V
4.15
8.18
7.76
6.36
4.63
6.20
6.98
6.37
5.63
5.40
4.14
5.70
5.18
6.29
7.81
11.79
14.22
9.01
10.91
10.85
10.75
10.71
10.54
10.75
0.83
0.71
0.20
0.80
0.62
0.06
5.80
7.84
7.06
5.50
4.06
6.05
8.09
8.83
9.42
10.85
12.02
9.83
5.01
8.95
7.98
5.51
3.98
6.27
6.05
7.05
8.03
10.90
13.51
9.07
350.04
299.69
763.49
1184.75
1.58
333.77
942.78
357.15
777.14
The numbers IIV refer to the previous table. All calculations are done annually before averaging to ve-year or 25year values. The antilogarithm was applied to the annual forecast values of the wage rate (III) and capital (IV) before
calculating growth rates.
GDP calculated via the denition of the labor share, with an assumed xed labor share in GDP including net taxes
on production (of 0.5138, year 2000), uses one of the two wage series plus the labor series to calculate annual GDP
and then the ve-year average GDP growth rate. (Alternatively, one could simply add up the wage and labor growth
rates as provided in the table.) GDP calculated using Eqn. (5), in addition, uses one of the two capital series.
When calculating the year 2001 GDP growth rate using Eqn. (5), the actual year 2000 labor share in GDP less net
taxes on production (0.5985) is used to weight the 2001 (over 2000) growth rates of the wage, of labor, and of capital
to obtain 2001 GDP. The 2001 labor share is then calculated from the 2001 labor remuneration and GDP values, and
used in the calculation of the 2002 GDP growth rate, etc. Using variables I+II, the minimum labor share in the
period 200025 is 0.5741, the maximum labor share is 0.6336; using variables III+IV, the two values are 0.4615 and
0.5985.
For comparison, the average annual GDP growth rate in 19782003 was 9.37% (using the ocial real GDP growth
rates) or 9.72% using the latest revised nominal GDP data combined with the rst published, implicit GDP deator.
positive growth rate that eventually (by 2014)

turns negative.
Future GDP growth can be obtained by summing the weighted growth rates of wages, capital, and labor (sixth and seventh data
columns). In each year, the previous-year values
of the tax-adjusted labor share, and of one
minus this share, are used as weights; that is,
weights are allowed to change in accordance
with the growth forecast.
In 200005, depending on which cointegration equation is used (with variables in levels
or in logarithms), the resulting average annual
GDP growth rate is 5.80% or 8.09% (sixth
and seventh data columns shown in the table).
With actual real growth in this period of
approximately 9.5%, the second series, with
the cointegration based on variables in logarithms, seems more appropriate. In the periods
200510 and 201025, the two series yield
rather similar estimates, in the range of 7.06
9.42%. At this level of economic growth, the
extrapolations of Table 1 suggest that China
will surpass the US economy in size, in purchasing power terms, by around 2010.
Further into the future, the predictions diverge signicantly, with one GDP growth series
falling o, and the other rising. Due to the recent explosion in education, the out-of-sample
forecasts use education values that are far from
sample average values. The comparison of the
two GDP growth series shows well the degree
of uncertainty involved in predicting further
into the future than 2015.
An alternative approach to predicting GDP
growth is to assume a constant labor share.
This appears plausible given the worldwide
experience that labor shares tend to be constant
over time. 39 In China, the share of labor in
GDP (now not less net taxes on production)
has been near-constant throughout the reform
period, with a mean of 0.5210 and a standard
deviation of 0.0111 in 19782002, that is, the
coecient of variation is only 2%. Given the
denition of the labor share as labor remuneration divided by GDP, that is, as
wt
Lt
at P t ;
Yt
assuming the labor share at to be constant

yields the growth rate of real GDP as
^t
w
Lt:
7
Yb t b
Pt
That is, under the assumption of a constant labor share, real GDP growth, by denition, is
the sum of real wage growth and labor growth.
Neither capital series nor weights are needed.
For the period 19782002, the relationship between the real wage and the quantity and quality of labor was identied in Table 2 and can be
applied to the data on the future quantity and
quality of labor constructed through 2025. Table 3 in the last two columns reports the resulting real GDP growth rates. They carry no new
or dierent insights but conrm the ndings
following the more elaborate decomposition
of GDP growth by income categories.
The above growth forecasts rely solely on
facts about the future quantity and quality of
laborers in China, facts that are near-certain today, and some assumptions. The assumptions
are that the past relationship between wage
and labor variables (and, in the case of the income growth identity, also between capital
and labor variables) continues to hold in the future, that the annual change in the net tax rate
on production continues to be relatively small,
and that, in the case of the income growth identity, the products of the depreciation share and
depreciation growth rate and of the surplus
share and surplus growth rate continue to be
relatively small.
The projections, by necessity, reect only
long-run trends. The model does not allow for
business cycles or any other sources of volatility. Koren and Tenreyro (2007) in a cross-country study found that as countries develop, their
production structure moves from volatile to
less volatile sectors and the volatility of country-specic macroeconomic shocks falls, both
of which would add credibility to the long-run
trends obtained here for China.
The long-run trends in the income growth
identity are driven by the number of laborers
and human capital formation. Other variables,
from average age to non-demographic variables, do not exhibit a long-run stable relationship with the real wage (or capital) in 1978
2002. With little change in Chinas labor force
over the next 20 years, the quality of labor becomes of key interest. The next section examines in more detail the most recent
developments in human capital formation in
China.
1679
5. HUMAN CAPITAL
Change in human capital formation has been
rapid in reform period China. A comparison
with the United States puts Chinas human capital measures into perspective, while the development of education in China over time
reveals the scale of changes underway.
Table 4 compares the educational level of the
Chinese versus US population as reported in
the population censuses of 1990 and 2000 in
both countries. 40 As a percentage of the total
population, a far higher proportion of the US
population has achieved a secondary or tertiary
level of education. For example, in 1990, 75.2%
of the US population had completed high
school (or above), but only 9.52% of the Chinese population had; 20.3% of the US population had completed a bachelors degree (or
above), but only 0.56% of the Chinese population had. But because the US population in
1990 was only 28% of the size of the Chinese
population (in 2000, 24%), the dierence
shrinks by a factor of approximately four once
total population numbers are considered.
During 19902000, China narrowed the relative gap. While in 1990 the number of US citizens with high school education was 1.65 times
the number in China, by 2000 this ratio had
fallen to 1.17 (last column shown in Table 4).
For the BA degree, the ratio fell from 7.59 to
4.33. At the level above the bachelors degree,
the ratio in 2000 was still 21.42 in favor of
the United States (with no such gure available
for 1990).
The population total of China masks drastic
dierences in education levels of dierent age
cohorts. Figure 8 shows that in 2000 more than
half the population aged 65 or above had a level of education below primary school, while of
the age cohort 2029 years, only 2.06% had; for
some form of secondary education, the two percentages are 10.55% (age 65 or above) versus
69.44% (aged cohort 2029), and for tertiary level education 1.50% versus 7.53%. The age cohorts in between reveal a smooth transition
from the less educated older generation to the
ever better educated younger age cohorts. 41
The Chinese age cohort of 2029 years in
2000, at the primary and secondary school levels, comes close to the United States population
average, but in tertiary level education still falls
far short.
The picture changes drastically if one considers current education data. In terms of currentyear new enrollment at the BA level (including
1680
WORLD DEVELOPMENT
Table 4. Educational level (or higher) of population aged 25 or older (Census data)
This degree or above
in %
Absolute number
(mio.)
United States China United States

2000
High school
Some college
College-level associate degree
Bachelors degree
Advanced degree
Reference: total population aged 25 or higher
1990
High school
Some college
College-level associate degree
Bachelors degree
Advanced degree
Reference: total population aged 25 or higher
80.4
51.8
16.5
4.3
1.4
0.1
24.4
8.9
75.2
45.2
20.3
7.2
9.52
1.51
0.56
n.a.
146.498
94.386
44.460
16.217
182.212
119.469
71.809
32.250
11.439
158.868
Ratio United
States/China
China
125.137
1.17
32.457
10.258
0.757
760.480
2.91
4.33
21.42
0.24
72.421
1.65
11.517
4.249
n.a.
571.589
6.24
7.59
n.a.
0.28
For the classication, see appendix. Advanced degree holders in the United States refer to the three degrees above
B.A., in China to Masters or Doctorate degree holders. No data on advanced degree holders in China for 1990 are
available. If Bachelors degree and above in the case of the United States were compared to College-level associate
degree and above in China, the ratio of the United States to China in 1990 is 2.80 (32.250/11.517), and in 2000 it is
1.37 (44.460/32.457).
United States data report degree obtained, while Chinese data include those who have already obtained the
particular degree as well as those who currently study for the particular degree. Since the US age classication is
limited to those aged 25 and older (and the Chinese classication has been reduced here to the US classication), the
number of Chinese who are currently studying toward a particular degree should be small, except possibly in the case
of advanced degrees.
Sources: United States: Educational Attainment 2000. China: Census 2000, pp. 593602; Census 1990, Vol. 1, p. 380,
Vol. 2, pp. 112f. and 132f.
80
60
40
20
%
No school/ lit. class
Primary school
20-29
Secondary school
40-49
Tertiary level
30-39
50-60
60-64
65+
Figure 8. Share of educational level in individual age groups, China, 2000 (in %). No school / lit. class refers to no
schooling or only basic literacy class (below primary school level). Source: Census 2000 (nationality), pp. 125127.
college-level associate degrees) China has now

caught up with the United States (Figure 9).
In the reform period, new enrollment in China
rst languished, but began to expand gradually
in the early 1980s and then rapidly after 1998.
In 1998, new enrollment was at approximately
twice its 1978 level. In the course of the following seven years, during 19982005, new enrollment quintupled. Since 2001, new enrollment
in Chinese BA programs at (only) regular institutions of higher education exceeds the number
of freshmen in the United States. In 2005, Chinas new enrollment gure was twice that of the
1681
United States, and in 2006 it was scheduled to

rise by another 5%. 42
Data on the joint category of MA and PhD
new enrollment in China (also Figure 9) follow
the same trend as new enrollment for the BA/
college-level associate degree. Lacking data on
new enrollment at the graduate level in the United States, and therefore assuming that the
1998 US MA and PhD graduation number of
476,174 and the 2005 number of 609,200 are a
good indicator of current new enrollment, US
enrollment numbers at the graduate level were
at least seven times the Chinese gure of
5.5
0.55
5.0
China: BA+ Associate degree
0.50
4.5
U.S.: Freshmen (2- and 4-year programs)
0.45
China: MA+ PhD --- right scale
0.40
3.5
0.35
3.0
0.30
2.5
0.25
2.0
0.20
1.5
0.15
1.0
0.10
0.5
0.05
0.0
1978
1981
1984
1987
1990
1993
1996
1999
2002
million
million
4.0
0.00
2005
Figure 9. Tertiary level new enrollment, China, and United States.

BA in the case of China refers to (current-year) new enrollment in regular institutions of higher education, that is,
includes BA and college-level associate degree programs of regular institutions of higher education; no data on new
enrollment in college-level associate degree programs of non-regular institutions of higher education are available. For
the United States, only data on rst-time freshmen are available; the coverage is unclear, but seems to be 2-year and 4year programs, presumably associate and BA programs, and possibly including professional degree programs. Stable
graduation gures in US MA/PhD programs suggest new enrollment in these programs of at least 400,000 to 500,000 per
year (assuming no drop-outs). In China, in 2005, 54.35% of total enrollment in regular institutions of higher education
was in regular BA programs (benke), while the rest was in college-level associate degree programs (zhuanke); but only
46.86% of new enrollment in regular institutions of higher education was in regular BA programs. In the United States, in
2004, the latest year for which the data are available, 62.10% of total fall enrollment in degree-granting institutions was
in 4-year (versus 2-year) programs, and 56.54% of total rst-time freshmen enrollment in undergraduate programs was
in 4-year programs.
Data for the United States for the years through 1995 cover institutions of higher education, and since 1996 degreegranting institutions (which include some additional institutions, primarily 2-year colleges, and exclude a few higher
education institutions that did not award associate or higher degrees); for rst-time freshmen enrollment, both sets of
data are available for 1996 and 1997: data in the latter classication are 34 percentage points larger than the data in the
earlier classication.
Sources: China: Statistical Yearbook 1990, pp. 708f.; 2003, pp. 720f.; 2004, pp. 7802; 2005, pp. 692f.; 2006, pp. 798,
800f.United States: <http://nces.ed.gov/programs/digest/d05/tables/dt05_179.asp>, and . . ./dt195.asp, accessed on Nov.
15 2006.
1682
WORLD DEVELOPMENT
72,508 in 1998, but perhaps only one-third

above the Chinese gure of 364,831 in 2005.
Graduation numbers mimic enrollment numbers with a time lag (Figure 10). The absolute
number of undergraduate level graduates in
China began to exceed that of the United States
in 2004, and by 2005 it was 50% higher. 43
Based on the enrollment numbers of 2005
(and the signicant drop-out rates in the United
States), by 2008/09 the number of undergraduate level graduates in China should be 2.5 times
the US number. At the MA and PhD level
(with only joint data available for China), the
United States in 1998 had a 10-fold lead in
absolute graduation numbers (Figure 10). 44
But by 2005, the gap narrowed to a threefold
lead, and by 2008, given the 2005 enrollment
numbers, may shrink to a less than 3040%
lead. 45
The explosion in education is only happening
since the late 1990s. Much of the upgrading in
the education level of Chinese laborers may
not be felt until some years later, when fresh

graduates have gained some job experience,
and not all the additional graduates may be
as qualied as the average graduate of earlier
generations.
Head count data do not necessarily translate
into comparable accumulation of human capital because they reveal nothing about the quality of education. 46 One may suspect that
education in China contains much rote learning or is dominated by a cult of authority,
but how this translates into GDP growth at
Chinas current stage of economic development
is unclear. Education in China through the secondary school level may not be all that inferior
to the United States. Even at the tertiary level,
a provincial railway ministry college in China
could match a US community college. At the
MA and PhD level, the conclusion is likely to
be less ambiguous; but at that level Chinese students probably make good use of the US institutions.
3.2
2.8
China: BA + Associate degree

U.S.: BA + Associate & Professional degree
2.4
China: MA + PhD
U.S.: MA + PhD
million
2.0
1.6
1.2
0.8
0.4
0.0
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
Figure 10. Tertiary level graduates, China and United States.

For the coverage of BA in China, see the previous gure. BA in the case of the United States refers to associate
degrees, BA degrees, and rst professional degrees. If rst professional degrees had been included with MA/PhD, in 2005
the United States BA category would have been (projected to be) smaller by 3.92%, and the US MA/PhD category
would have been larger by 13.95%. US data for 2005 are ocial projections. Data for the United States for the years
through 1995 cover institutions of higher education, and since 1996 degree-granting institutions. (Also see notes to
Figure 9.)
In China, in 2005, 47.78% of the BA + Assoc. graduates received a BA degree, the others a college-level associate
degree. In the United States, in 2005, BA degrees are projected to outnumber associate degrees 2:1 (1,416,000 versus
668,000).
Sources: China: Statistical Yearbook 1990, pp. 708f.; 2003, pp. 720f.; 2004, pp. 780-2; 2005, p. 693.; 2006, pp. 798,
801.United States: <http://nces.ed.gov/programs/digest/d05/tables/dt05_246.asp>, accessed on Nov. 15 2006.
6. CONCLUSIONS AND IMPLICATIONS

Extrapolation of Chinas reform period economic growth into the future suggests that the
size of Chinas economy will exceed that of
the United States, in purchasing power parity
terms, by about 2010. Per capita, the point of
time when China catches up with the United
States is much further into the future, 20 to
50 years from now, although the coastal areas,
especially in the past the fastest growing ve
provinces, with a population exceeding that of
the United States, may catch up in as little as
one to two decades.
Chinas economic development in the reform
period ts well with the broad development
patterns of structural change, catching up,
and factor price equalization, not least in comparison with other Asian countries earlier in
their economic development. Following all
three patterns, China faces another 30 years
of continued growth. Re-composing Chinas
economic growth from growth in income components indicates economic growth during
200515 in the range of 79%, high enough
for China to catch up with the United States,
in purchasing power terms, within about ve
years.
In the growth decomposition, human capital
had signicant predictive power. Common
sense similarly suggests that if talent is randomly distributed among the world population
and if Chinas education system is able to identify the brightest students, then China has a larger pool of talent to draw from than any other
country in the worldfor every genius in the
United States, China potentially has four. 47 If
innovation depends on the agglomeration of
talent (geographically, nationally, culturally,
or linguistically), then China is in an excellent
position to innovate and grow. The explosion
in tertiary level education in China may yet
supply the productivity gains that Paul Samuelson identies as the potential causes of reductions in the United States per capita real
income. 48
Chinas talent may initially be focused on a
few projects or industries, such as the state-supported space program. In industry, it is only
when prots can no longer be made by using
existing technologies or upgrading to known
technologies that pressure for indigenous innovation sets in. Because not all industries sectors
develop at the same pace, China is unlikely to
technologically advance on all fronts simultaneously.
1683
Demographics also matters in terms of market sizeand 80% of Chinas population lives
in the densely populated Eastern part of the
country. Size of the domestic market should allow unprecedented variety and economies of
scale. Economies of scale lead to low cost. A
large number of suppliers suggest a high degree
of competition and pressure on prices; eventually, decreasing prot margins create incentives
to innovate in order to either lower costs further or be able to oer a more dierentiated
product with higher prot margins. The price
wars of the late 1990s are evidence of the degree
of price pressures Chinese markets are capable
of. 49
One immediate implication of Chinas economic growth on peoples livelihood is that
one-fth of the world population will continue to experience signicant improvements
in their living standard. A share of Chinas
population that exceeds the size of the US
population will enjoy living standards close
to the level of developed countries in the near
future. In the Chinese hinterland, the rise out
of poverty continues. Consumption waves
similar to those previously experienced in
the West are unfolding in China, albeit with
some variations, such as the early introduction of the cell phone. In production, with
rst labor shortages in Guangdong province
in 2005, wages of unskilled labor have begun
to rise, and labor standards are increasingly
being enforced. From the creation of capital
markets to the establishment of a countrywide social security system, what has taken
the West centuries has been condensed into
decades in China.
Internationally, the growth of China implies
that the center of world economic activity shifts
to an Asia that besides China also includes India, South East Asia, Japan, and Korea.
Shocks to the United States and European
economies, or stagnation of these economies,
will have less and less impact on Asia that continues to integrate. In 2005, Chinas ratio of
exports of goods and services to GDP was
37%, compared to 10% for the United States. 50
Of Chinas exports of goods in 2005, 21% went
to the United States and 22% went to Europe,
but 48% went to Asia; 67% of Chinas imports
came from Asia. 51
The West plays an important entrepreneurial
role in China. In 2004, the most recent year for
which the data are available, 19% of gross
output value of industrial enterprises with annual sales revenue in excess of 5 m yuan
1684
WORLD DEVELOPMENT
RMB (USD 0.6 m) was produced by foreignfunded enterprises, a gure which excludes an
additional 11% in enterprises funded by Hong
Kong, Macao, and Taiwanese entrepreneurs, 52
that is, non-domestic entities control one-third
of Chinese industry. To the extent that Western
pension funds invest in these rms, often multinational corporations, Western citizens directly
benet from Chinas continued economic
growth.
The foreign-funded industrial enterprises
alone, with their approximately one-fth share
of industrial output in China, accounted for
58% of Chinas exports in 2005 (and 59% of
Chinas imports). 53 This suggests that Chinas
exports to the West largely reect the West
importing from itself, where the Wests
nationals, or rms incorporated in the West,
use Chinese labor in their production
processes.
Chinas economic growth aects relative
prices around the world, thereby promoting
an increasing division of labor and specialization worldwide. This implies structural
changes in countries around the world. The
restructuring extends beyond the traditional
realms of manufacturing, to services; for example, the newly rich in Chinas coastal provinces
are potential tourists for the West, while backroom banking services may gradually shift
from the West to China. Chinas current
strength in unskilled labor and mass production implies a wage constraint for unskilled
laborers employed in Western industries producing tradables.
Chinas economic growth is in some ways
a rerun of the industrial revolution in the
UK two hundred years earlier, but at a
much swifter pace. It is accompanied by
simultaneous changes in the patterns of production, ownership, and sources of income in
economies around the world. The concept of
a national economy, and, following from
that, of a new economic superpower, appears far-fetched; China is integrated into
the world economy to an extent that is similar to that of a European country within
Europe. What emerges is one world economy
organized along dimensions of comparative
advantage, where the construct of a nation
is, in the economic realm, reduced to issues
of enterprise registration, regulation, and taxation.
One may wonder if China will not, one
day, be overwhelmed by obstacles to further
economic growth, ranging from bad loans in

the banking system to an under-funded pension insurance scheme, increasing income
inequality, the lack of a rural health care system, bankrupt local governments, and environmental pollution. In 1998, Lardy stressed
the large economic problems and the unprecedented potential for social unrest due to the
ever more indebted state-owned enterprises,
the extent of nonperforming loans, and a decline in government revenue. By 2005, the
state-owned enterprises have largely been reformed, the problem of nonperforming loans
has been addressed, and government revenue
is rising. There is a rationale for the Chinese
leadership to not try to shoot for optimal
transition in every respect (not feasible, and
too costly), but to let problems emerge and
in part resolve themselves, and to step in only
when the need for action becomes urgent.
Where Western observers tend to predict Chinas doom, Chinas leadership muddles
through by trial and error. The same pattern
appears to be repeated yet again when it
comes to inequality, social security, or environmental pollution.
Income inequality has now reached the top of
the agenda of policymakers and rst signs have
emerged that some aspects of it may be about
to go into reverse. 54 Reform of the urban
health care and pension system is underway.
The systematic construction of a rural social
security system that includes health care and
pension has begun.
The current ve-year plan (200610) lists
obligatory targets to reduce the use of water
(per unit of value added by 30%) and the volume of euents (overall reduction by 10%,
with specic targets to be set by the State
Environmental Protection Administration).
The Five-Year Plan also stipulates an obligatory reduction of energy use per unit of
GDP by 20%. 55 Environmental protection,
energy reduction, and land conservation targets have been written into cadres evaluation
system as obligatory. With little success in
environmental preservation and energy conservation in the rst year of the current FiveYear Plan (2006), the year 2007 has started
with a plethora of detailed plans, regulations,
and exhortations that include targets handed
down to individual enterprises. There can be
no doubt that drastic action is being taken.
How successful that action will be remains to
be seen.
1685
NOTES
1. Broomeld (2003) provided an overview over the
China threat literature. Lubman (2004) documented
the negative views of China in the US congressional
debates.
2. The quotes are from The China Price in Business
Week, 6 December 2004 (issue 3911), pp. 102.
3. Statistical Yearbook (2006, pp. 734, 1025).
4. See The Economist, November 15, 2003, p. 67,
quoting the Bank Credit Analyst, a Canadian research
rm.
5. Perkins and Rawski (2006) perhaps came closest.
They rely on the growth accounting equation derived
from the Cobb-Douglas production function, using
income shares as weights. Their labor calculations are
similar to those made here; their future capital series are
assumed series.
6. This paper is also not concerned about how meaningful the concept of GDP data is for measuring wellbeing or standard of living. One may further
question if value added created in China contributes as
much to well-being as does the value added created in
the United States, or vice versa. Since the use of GDP as
a measure of economic size, or, in per capita terms, as a
proxy for standard of living, dominates in the literature
as well as in popular discussions, this paper focuses on
GDP.
7. See Statistical Yearbook (2006, pp. 57, 734); <http://
www.bea.gov/bea/newsrel/gdpnewsrelease.htm> (accessed November 30, 2006).
8. The GDP data are those published by National
Bureau of Statistics in China (for China) and those
published by the Bureau of Economic Analysis in the
United States (for the United States). The PWT also
include GDP values; in the case of China these incorporate manipulations of the ocial values in accordance
with adjustments to ocial values suggested by Maddison. Holz (2006a) argued that Maddisons adjustments
are not justied (and questions Angus Maddisons
(2006) reply in Holz (2006c)).
9. The PWT 6.1 for the year 2000 (the most recent year
covered in the PWT 6.1) report a factor of 4.32; in the
PWT 6.2, for 2000, it is 4.22. This suggests a rather small
margin of error, except that the PWT 6.2 values seem to
incorporate very little new information.
10. One could allow the future real growth rate to dier
from the past one. For example, assuming a 7% real
growth rate for China (instead of the 197805 average of
9.64%) and a 3% real growth rate for the United States
(instead of 2.97%), Chinas GDP exceeds that of the
United States by 2015; with a 3% RMB appreciation this
changes to 2011. Per capita, reducing the growth rate by
three percentage points results in the years 2059 countrywide, 2033 for the coastal provinces, and 2027 for the
ve fastest-growing provinces (or, with RMB appreciation, 2035, 2023, and 2020). In contrast, increasing
Chinas aggregate real growth rate to 12% (or increasing
the per capita growth rate by two percentage points)
yields 2010 (2029, 2020, 2018) or, with RMB appreciation, 2009 (2023, 2017, 2015). The impact of such
changes to the growth rate is small in the short, but
substantial in the long run.
11. One particular alternative would be the projections
based on nominal values combined with assumptions
about the price adjustment factor and/or the exchange
rate.
12. An appendix with further details on extrapolations
is available, as all appendices are, at <http://ihome.ust.hk/~socholz/Growth/Appendices-Holz-ChinaGrowth-20253July07.pdf>.
13. In a dierent approach, Woo (1999) examined the
results of sector-level reforms in China, thus breaking
down the case of China into several sectoral observations (which, however, are not independent). Rawski
(1999) traced the interaction between dierent reforms;
in total, these are taken to add up to the reform success.
The World Bank (1997, p. 19) listed a variety of
strengths and advantages considered to promote
economic growth, and a variety of risks and challenges considered to hamper economic growth.
14. The rst two are standard fare in development
economics; see, for example, Cypher and Dietz (1997,
pp. 27, 267, 403). The third is a basic theorem in trade
theory as a corollary of the Heckscher-Ohlin theorem;
see, for example, Salvatore (1998), chapter 5.
15. The individual data sources, with explanations, are
documented in an appendix, which also oers some
further thoughts on the inclusion of other countries (see
note 12 for the web location of the appendix).
16. Total and sectoral employment data for China
experience a statistical break during 198990; total
1686
WORLD DEVELOPMENT
employment, following the 1990 population census,

increases by 17.03%. In the following, total employment
data for the years prior to 1990 are adjusted; for details,
see Holz (2006d), for a summary explanation see an
appendix. Sectoral employment data and the share of
employment in the primary sector are the ocial
values.
17. Figure 2 suggests that the typical annual rate of
structural change contributes approximately three to
four percentage points to Chinas annual growth rate.
18. While doubts are occasionally raised in the literature about the accuracy of the Chinese data on the
number of laborers in agriculture, suggesting that the
ocial data constitute an over-count, the 1996 agricultural census (with a distinction between pure farmers
and part-time farmers) as well as the 1990 and 2000
population censuses (with unambiguous denitions that
should lead to a proper counting of migrant laborers)
implies that the ocial primary sector labor data, used
here, are, if anything, an underestimate.
19. Among the three patterns, the direction of causality
is perhaps most ambiguous in the case of structural
change. One could probably also argue for labor
productivity growth to cause structural change; a more
specic argument would be for labor productivity
growth in non-agriculture that is dierent from labor
productivity growth in agriculture to cause structural
change. On the other hand, if much of ruralurban
migration happens in response to a reduction in
constraints on such migration, then the past and
ongoing relaxation of these constraints would explain
migration and, that, in turn average economy-wide labor
productivity growth (also see Holz, 2006b). No matter
which direction the causality runs (probably in both), on
both accounts there is pressure for the pattern to
continue to unfold.
20. A second-degree polynomial is imposed on the
trend lines shown in Figure 4 and in the following
gures. In contrast, a linear trend was used in Figure 2
because one extreme outlier in the cases of Taiwan and
Japan would have resulted in (misleadingly) sharp
inverse-U shaped curves for these two countries. The
sensitivity of second-degree polynomials to some types
of outliers is also seen in the upturn of the Japanese
trend shown in Figure 4.
21. If the scope for catching up were measured by the
labor productivity dierences in the industrial sector
only (rather than economy-wide), Chinas highest level
of catching up would probably be slightly higher, but
not yet in the range where the corresponding growth
rates of real GDP per laborer fall o signicantly.
22. As Abramovitz (1986) elaborated, catching up

depends, more broadly, on the social capability to
catch up and on the factors controlling the rate of
realization of the potential for catching up (facilities for
the diusion of knowledge, conditions facilitating or
hindering structural change, and macroeconomic and
monetary conditions). In all these respects, Chinas
capabilities have been (and are still) improving.
23. These assumptions are two countries; two commodities (goods); two factors of production; same,
constant returns to scale technology in use in both
countries; across both countries one commodity is labor
intensive, the other capital intensive; incomplete specialization in production in both countries; tastes are equal
in both countries; perfect competition in both goods and
factor markets in both countries; perfect labor mobility
within each country but no international factor mobility;
no transportation costs, taris, or other obstructions to
the free ow of international trade; all resources are fully
employed in both countries; international trade between
the two countries is balanced (Salvatore, 1998, p. 110).
24. A chart with annual wages relative to the United
States, based on the absolute factor price equalization
theorem, is very similar to Figure 4 (with GDP per
laborer as a measure of catching up) and omitted; this is
due to the fact that labor shares in GDP tend to be
constant over time and similar across countries. (Wages
are obtained by dividing the labor remuneration component of GDP by the number of laborers.)
25. All three patterns also well predict Japans relative
stagnation since the 1990s.
26. While all other sections of this paper use the most
updated data available, including the revised 1978/93
2004 production approach GDP values following the
2004 economic census, this section uses the pre-economic census national accounts data. This is because
revised income approach data, which are at the core of
this section, are not available. An appendix explains the
issue of data availability in more detail and explores the
potential implications of using pre-economic census data
(none, or only very minor ones). Two separate appendices provide all data used in this section.
27. Barro and Sala-i-Martin (1995, p. 352) stressed that
growth accounting does not . . . constitute a theory of
economic growth because it does not attempt to explain
how the changes in inputs and the improvements in total
factor productivity relate to elementssuch as aspects
of preferences, technology, and government policies
that can reasonably be viewed as fundamentals. Even if
such relationships could be specied, the next question
that arises is what explains preferences, technology, and

government policies? Trying to trace economic growth
back to original causes presents a problem of innite
regress. The decision on when to stop, that is, which
level to view as ultimate explanation, appears arbitrary.
28. For the results of production function estimations
with the data used here, see appendix.
29. Provincial gross value added in the income
approach in most years, for most provinces, exactly
equals provincial gross value added in the production
approach, Chinas ocial approach to calculating GDP.
But the sum of provincial gross value added only closely,
not perfectly matches Chinas national GDP value; the
National Bureau of Statistics does not solely rely on
provincial data when deriving the national GDP value.
In addition, a very few provinces do not report income
components for some of the early reform years. As long
as the discrepancy between sum provincial GDP and
national GDP does not reect a particular bias in the
size of one or more of the dierent income components
(and there is no reason why it should), the procedure
used here yields accurate national income component
values. The provincial data are from GDP 195295 for
the years 197895, and from GDP 19962002 for the
years 19962002; since the mid-1990s, each issue of the
annual Statistical Yearbook also reports the data of the
previous year.
30. Business prot equals prot net of investment
returns and non-business revenues. For details on the
link between the operating surplus and prot (or
business prot), see Holz (2003, p. 88f). The denition
of operating surplus in China follows the standard
System of National Accounts promulgated by the
United Nations. In the case of individual-owned enterprises with seven or fewer employees (getihu) and of
agricultural households, labor remuneration incorporates the operating surplus.
31. What is called technological progress (the residual) in the traditional growth accounting equation is
now measured as (and is identical to the) weighted rate
of change in the real wage, in the depreciation rate, and
in the surplus rate, plus a particular form of change in
the net tax rate on production. Harberger (1998, p. 1)
postulated a growth equation that attributes to each
factor a marginal product measured by its economic
reward in the form of pDy w
DL
q dDK R,
with the variables denoting initial general price level,
change in output (GDP), initial real wage, change in
labor input, initial real rate of return to capital, rate of
real depreciation of capital, change in capital stock, and
the residual interpreted as real cost reduction. Compared to the income growth identity (and incorporating
1687
the price level in the coecients of capital rather than in

capital), the real cost reduction, once divided by output,
is identical to the weighted growth rates of real wages,
the real depreciation rate, the real rate of return, and a
particular form of change in the rate of net taxes on
production.
32. Since the right-hand side variables are deated
using the implicit GDP deator as rst published, the
reported real growth rates of GDP are obtained by
deating nominal GDP by the same deators. Nominal
GDP values for all years are taken from a recent source
(see table of data in the appendix) and may include
revisions of earlier years nominal GDP; in contrast,
ocial real GDP growth rates of earlier years are
typically not revised. Ocial real GDP grew by 759%.
33. Annual data on laborers by age and education for
the period 19782003 are obtained/calculated from the
1982, 1990, and 2000 population censuses together with
the 1987 and 1995 1% population sample surveys.
Projections for the years 200025 make use of the year
2000 census data and past patterns and trends. An
appendix on labor data provides general explanations on
the derivation of the labor data; all calculations and
estimations are explained in detail in Holz (2005).
34. Young (2003), for example, regressed individuals
wages on gender, age (for which he has a total of 11
categories), and education (total of 4 categories).
35. Details on the estimation of the VEC models are
reported in an appendix.
36. Similarly, the largest share of the forecast error
variance of the share of laborers with upper middle
school education is due to its own shocks, but 60% of the
forecast error variance of the share of laborers with
college level education or above is explained by variation
in the share of laborers with upper middle school
education. The forecast error variance decomposition
depends, in this particular instance, much on the
Cholesky ordering of the variables. Reversing the order,
63% of the variance of the wage is explained by variation
in the share of laborers with upper middle school
education.
37. Non-education variables are typically not cointegrated with the wage rate. Age, or age squared, does not
have a unit root. The dierence in the choice of variables
for the cointegration equations in levels versus in
logarithms is necessitated by the fact that some variables
have unit roots when the variable is measured in levels,
but not when it is measured in logarithms, and vice
versa. For example, in levels, average years of schooling
minus ve is not I(1) but I(2), while the share of
1688
WORLD DEVELOPMENT
laborers with college level education or above in

logarithms is I(2). Average years of schooling is I(2)
independent of if the variable is in levels or in
logarithms.
38. In the case of the rst cointegration equation, with
the wage in levels, a dierent cointegration equation that
involves the share of laborers with primary school
education rather than the share of laborers with upper
middle school education exhibits strong Granger causality from both education variables to the wage series.
In a visual comparison of the actual wage series and
these two tted series, the tted series reported in the
table with the share of laborers with upper middle school
education better matched the actual series.
39. This is a standard assumption in production
function based growth accounting, where, furthermore,
the constant labor share, and one minus the labor share,
is assumed to equal output elasticities.
40. The United States and Chinese education systems
are not perfectly comparable. An appendix establishes
the equivalence of dierent degrees in the two countries
and provides additional details.
41. The educational level of the age cohorts in the late
1940s may be exaggerated in that some upper middle
school graduates in China during the Cultural Revolution may have enjoyed no more than a lower middle
school education under the label of an upper middle
school education.
42. The 5% increase is reported in a news item in China
Infobank (<http://www.chinainfobank.com>) of May
26, 2006, titled 2006 nian zhongguo gaoxiao de
zhaosheng guimo zai qunian 504 wan ren de jichushang
zai zengjia 5% (The 2006 tertiary level new enrollment
gure is 5% higher than last years basis of 5.04 m).
43. This assumes a narrow denition of undergraduate
level graduates in China and a broad denition for the
case of the United States (See notes to Figure 10).
44. If the US graduates with a rst professional degree
are included with the US gure of MA and PhD
recipients, the comparison with China is slightly further
in favor of the United States. But with Chinese medical
doctors and lawyers primarily holding a BA degree (in
the medical doctors case a 5-year BA program), it seems
inappropriate to include graduates of medical and law
schools in the US number but not in the Chinese
number.
45. These data ignore Chinese students studying
abroad because no specic degree and program comple-
tion data are available for these. Assuming Chinese

students abroad were all studying in MA and PhD
programs, their number in 2002 was equivalent to
approximately half of the year 2005 domestic MA and
PhD graduates; the 2005 number of returning students,
after completing studies, was one-fth the 2005 number
of domestic MA and PhD graduates. (For the data
sources, see those listed in Figure 9). At the high school
level, the number of Chinese graduates has well exceeded
their US counterpart in absolute numbers throughout
the reform period. In 2005, China had 2.69 times more
high school graduates than the United States. For
additional data, see appendix.
46. The head count comparison is furthermore incomplete because it excludes education in China outside the
regular institutions of higher education, while the US
data make no distinction between ocial and unofcial education. For example, adult education outside
the regular university system in China is not included in
the tertiary level education data presented. In 2001,
316,367 persons graduated with a degree from a tertiary
level adult education institution. This number compares
to 3,068,000 graduates from regular institutions of
higher education (at undergraduate level) and corresponds to just below 1& of the population age 2540 on
November 1, 2000 (Census 2000, Vol. 1, pp. 2157;
Education Yearbook 2002, p. 107).
47. There is little doubt that the Chinese education
system is indeed capable of identifying the brightest
children, with lower middle school to university admission now routinely based on entrance examinations.
Heckman (2005) expressed concern about nancial
hurdles for children from low-income families.
48. Why has China not overtaken the United States
a long time ago? The immediate answer is Chinas choice
of economic system prior to 1978. In terms of education,
with uniform wages and access to higher-level education
severely limited, neither were incentives for human
capital accumulation in place nor the opportunities
available. For example, tertiary level education almost
ceased during the Cultural Revolution. In 1971, a since
1949 low of 600 students are reported to have graduated
from a regular institution of higher education, down
from 186,000 in 1965; only by 1977 did the gure again
reach the 1965 level, with 194,000 graduates. (Statistical
Yearbook 1990 p. 709.) This means that among the
older generations, only the generation born in or before
the 1940s had some access to tertiary level education.
49. Holz (2003), Chapter 8, provided evidence of the
price wars of the late 1990s among state-owned enterprises.

50. See Statistical Yearbook 2006, pp. 57, 76, 734. The
share in France was 26%, in Germany 40%, in the UK
26%, and in the relatively small South Korea 43%. Data
for all countries except China are from the IFS. The
value of Chinas exports of goods and services is not
reported in the IFS, only the net export value. The data
on net exports and GDP for China in the IFS are
virtually identical to those reported in Chinas balance
of payments and national income accounts in the
Statistical Yearbook series; the balance of payments
data on goods, in turn, are near identical to the importexport data in the trade section of the Statistical
Yearbook.
51. For the Data, see Statistical Yearbook 2006, pp.
7403.
52. Statistical Yearbook 2006, p. 505. These data are
from the economic census 2004. Value added data for
2004 are not available. The data for 2005 do not come
with an ownership breakdown; in 2005, industrial
enterprises with annual sales revenue in excess of 5 m
yuan RMB accounted for 93.86% of total industrial
1689
value added (pp. 57, 511). Based on pre-economic census

(unrevised) values, foreign-funded enterprises plus Hong
Kong/Macao/Taiwan enterprises together in 2005 accounted for 28.57% of the value added produced in
industrial enterprises with annual sales revenue in excess
of 5 m yuan RMB (Statistical Abstract 2006, p. 133).
53. Statistical Yearbook 2006, pp. 734, 751. Foreignfunded plus Hong Kong/Macao/Taiwan enterprises
accounted for 70.84% of total exports by industrial
enterprises with annual sales revenue in excess of 5 m
yuan RMB in 2005 (Statistical Abstract 2006, p. 133).
54. See, for example, Riskin (2007) on recent decreases
in rural inequality and in urban inequality.
55. Work on a green GDP has begun. Qi (2004)
presented an overview of progress toward the calculation of green GDP for China. No economy-wide gure
appears as yet available, but work by the National
Bureau of Statistics and Chinas National Environmental Protection Oce has begun to put a framework for
green GDP into place within 36 years.
REFERENCES
Abramovitz, M. (1986). Catching Up, Forging Ahead,
and Falling Behind. Journal of Economic History,
46(2), 385406.
Barro, R. J., & Sala-i-Martin, X. (1995). Economic
growth. New York, NY: McGraw-Hill.
Broomeld, E. V. (2003). Perceptions of Danger: The
China Threat Theory. Journal of Contemporary
China, 12(35), 265284.
Brown, L. R. (1995). Who will feed China? Wake-up call
for a small planet. New York: W.W. Norton & Co..
Carter, C. A., & Rozelle, S. (2001). Will China Become a
Major Force in World Food Markets?. Review of
Agricultural Economics, 23(2), 319331.
Census 1990 (1993). Zhongguo 1990 nian quanguo renkou
pucha ziliao (National population census material
1990). Compiled by the State Council population
census oce and the National Bureau of Statistics
population division. Beijing: Zhongguo tongji chubanshe.
Census 2000 (2002). Zhongguo 2000 nian quanguo renkou
pucha ziliao (National population census material
2000). Compiled by the State Council population
census oce and the National Bureau of Statistics
population, society, and science division. Beijing:
Zhongguo tongji chubanshe.
Census 2000 Major Figures (2001) 2000 nian di wu ci
quanguo renkou pucha zhuyao shuju (Major gures
from the 2000 population census of China). Beijing:
Zhongguo tongji chubanshe.
Census 2000 (nationality) (2003). 2000 nian renkou pucha

zhongguo minzu renkou ziliao (National population
census material 2000 by nationality). Issued by the
National Bureau of Statistics population, society,
and science division. Beijing: Zhongguo tongji chubanshe.
Chang, G. G. (2001). The coming collapse of China. New
York: Random House.
Chow, G. C. (2002). Chinas economic transformation.
Malden, MA: Blackwell Publishing.
Chow, G., & Li, K.-W. (2002). Chinas Economic
Growth 19522010. Economic Development and Cultural Change, 51(1), 247256.
Cypher, J. M., & Dietz, J. L. (1997). The process of
economic development. London: Routledge.
Douglas, A. (2002). Meet the New China. International
Journal of Business, 7(3), 3549.
Education Yearbook (Annual). Zhongguo jiaoyu nianjian (China education yearbook). Beijing: Zhongguo
jiaoyu chubanshe.
Educational Attainment 2000 (2004). At <http://
www.census.gov/prod/2003pubs/c2kbr-24.pdf> Accessed 1.01.04.
GDP 195295 (1997). Zhongguo guonei shengchan zongzhi hesuan lishi ziliao 19521995 (Historical data on
Chinas gross domestic product 19521995). Dalian:
Dongbei caijing daxue chubanshe.
GDP 19962002 (2004). Zhongguo guonei shengchan
zongzhi hesuan lishi ziliao 19962002 (Historical data
1690
WORLD DEVELOPMENT
on Chinas gross domestic product 19962002). Beijing: Zhongguo tongji chubanshe.

Gertz, B. (2000). The China threat how the peoples
republic targets America. Washington D.C: Regnery
Publishing.
Harberger, A. C. (1998). A Vision of the Growth
Process. American Economic Review, 88(1), 1
32.
Heckman, J. J. (2005). Chinas Human Capital Investment. China Economic Review, 16(1), 5070.
Henderson, C. (1999). China on the brink: The Myths
and realities of the Worlds largest market. New
York: McGraw-Hill.
Heston, A. (2001). Treatment of China in PWT 6. Draft,
December 2001. <http://pwt.econ.upenn.edu/Documentation/China.PDF> Accessed 16.12.03.
Holz, C. A. (2003). Chinas industrial state-owned enterprises: Between protability and bankruptcy. Singapore: World Scientic Publishing Co.
Holz, C. A. (2005). The quantity and quality of labor in
China 197820002025. Mimeo, Hong Kong University of Science and Technology. <http://ihome.ust.hk/~socholz>.
Holz, C. A. (2006a). Chinas Reform Period Economic
Growth: How Reliable Are Angus Maddisons
Estimates? Review of Income and Wealth, 52(1),
85119.
Holz, C. A. (2006b). Why Chinas Rise Is Sustainable.
Far Eastern Economic Review, 169(3), 4146.
Holz, C. A. (2006c). Chinas Reform Period Economic Growth: How Reliable Are Angus
Maddisons Estimates? Response to Angus Maddisons Reply. Review of Income and Wealth, 52(3),
471475.
Holz, C. A. (2006d). Measuring Chinese productivity
growth, 19522005. Mimeo, Hong Kong University
of Science and Technology. <http://ihome.ust.hk/
~socholz>.
IFS (Annual). International nancial statistics yearbook.
Washington, DC: International Monetary Fund.
Koren, M., & Tenreyro, S. (2007). Volatility and
Development. The Quarterly Journal of Economics,
122(1), 243287.
Lardy, N. R. (1998). Chinas unnished economic revolution. Washington D.C.: The Brookings Institution.
Lin, Justin Yifu, Cai, Fang, & Li, Zhou (2003). The
China miracle: Development strategy and economic
reform (revised edition). Hong Kong: The Chinese
University Press.
Lubman, Stanley (2004). The dragon as demon: Images
of China on capitol hill. Journal of Contemporary
China, 13(40), 541565.
Maddison, A. (2006). Do Ocial Statistics Exaggerate
Chinas GDP Growth? A Reply to Carsten Holz.
Review of Income and Wealth, 52(1), 121
126.
Murray, G. (1998). China: The next superpower. Dilemmas in change and continuity. Richmond, Surrey:
China Library (Curzon Press).
Perkins, D. H., & Rawski, T. G. (2006). Forecasting
Chinas Economic Growth over the Next Two
Decades. Draft, 29, 2006, June.
PWT6.1. Heston, A., Summers, R., & Aten, B. (2002).

Penn World Table Version 6.1, Center for International Comparisons at the University of Pennsylvania (CICUP). <http://pwt.econ.upenn.edu/>.
PWT6.2 Heston, A., Summers, R., & Aten, B. (2006),
Penn World Table Version 6.2, Center for International Comparisons of Production, Income and
Prices at the University of Pennsylvania. <http://
pwt.econ.upenn.edu/>.
Qi Yuanjun (2004). Guoneiwai luse GDP yanjiu de zongti
jinzhan (Overall progress in research on green GDP in
China and abroad). Jingji yanjiu cankao (88), 259,
34.
Qian, Yingyi (2000). The Process of Chinas Market
Transition (19781998): The Evolutionary, Historical, and Comparative Perspectives. Journal of Institutional and Theoretical Economics, 156(1), 151171.
Qian, Yingyi (2003). How Reform Worked in China. In
D. Rodrik (Ed.). Search of prosperity: Analytic
narratives on economic growth (pp. 297333). Princeton, NJ: Princeton University Press.
Rawski, T. G. (1999). Reforming Chinas Economy:
What Have We Learned? The China Journal, (41),
139156.
Riskin, C. (2007). Has China Reached the Top of the
Kuznets Curve?. In V. Shue, & C. Wong (Eds.),
Paying for progress in China (pp. 2945). Milton
Park, UK: Routledge.
Salvatore, D. (1998). International economics. Upper
Saddle River, NJ: Prentice Hall.
Samuelson, P. (2004). Where Ricardo and Mill Rebut
and Conrm Arguments of Mainstream Economists
Supporting Globalization. Journal of Economic Perspectives, 18(3), 135146.
Statistical Abstract (Annual). Zhongguo tongji zhaiyao
(China Statistical Abstract). Beijing: Zhongguo tongji chubanshe.
Statistical Yearbook (Annual). Zhongguo tongji nianjian
(Statistical Yearbook of China). Beijing: Zhongguo
tongji chubanshe.
Terrill, R. (2003). The new Chinese empire: And what it
means for the United States. Basic Books.
Timperlake, E., & Triplett, W. C. Jr., (2002). Red dragon
rising: Communist Chinas military threat to America.
Washington, D.C.: Regnery Publishing.
Wang, Yan, & Yao, Yudong (2003). Sources of Chinas
Economic Growth 19521999: Incorporating Human
Capital Accumulation. China Economic Review,
14(1), 3252.
World Bank Development Indicators. Cross-country
database. <http://www.worldbank.org/data>.
World Bank (1997). China 2020. Washington, DC:
World Bank.
Woo, W. T. (1994). The Art of Reforming Centrally
Planned Economies: Comparing China, Poland, and
Russia. Journal of Comparative Economics, 18(3),
276308.
Woo, W. T. (1999). The Real Reasons for Chinas
Growth. The China Journal, (41), 115137.
Wu, Yanrui (2004). Chinas economic growth: A miracle
with Chinese characteristics. London and New York:
RoutledgeCurzon.

Young, A. (2003). Gold into base metals: Productivity
growth in the Peoples Republic of China during the
reform period. Journal of Political Economy, 111(6),
12201261.
1691
Zeng, M., & Williamson, P. J. (2003). The Hidden

Dragons. Harvard Business Review, 81(10), 9299.
Available online at www.sciencedirect.com

China&#39;s Economic Growth 1978-2025

Caricato da

Informazioni sul documento

Descrizione originale:

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

China&#39;s Economic Growth 1978-2025

Caricato da

Copyright:

Formati disponibili

World Development Vol. 36, No. 10, pp.

Chinas Economic Growth 19782025: What

What is undeniable is Chinas rapid economic

a major force in world food markets (yes, if

One approach is to examine how China fares

CHINAS ECONOMIC GROWTH 19782025

2. EXTRAPOLATION OF PAST GROWTH

In what year will Chinas GDP exceed that of

Table 1. First year in which Chinas GDP exceeds US GDP

Per capita GDP

Exchange rate adjustment?

US price level = 3.86 * PRC price level

the present, to 2009 (second row shown in

CHINAS ECONOMIC GROWTH 19782025

ment Indicators database, the International

Value-added per laborer (yuan RMB, current prices)

Growth rate of real GDP per laborer (in %)

Annual absolute change inshare of agriculture in employment

The pattern of structural change is likely to

CHINAS ECONOMIC GROWTH 19782025

Share of agriculture in employment

signicantly higher than the income in many

age 4% to 5% throughout all years. These two

Growth rate of real GDP per laborer (in %)

Growth rate of real GDP per laborer (in %)

foreign knowledge becomes more accessible

that is, the skill-specic wage rate of one

CHINAS ECONOMIC GROWTH 19782025

Growth rate of real GDP per laborer (in %)

proxy. The price of investment goods of any

no reason why the gradual process of factor

Relative wage rates, divided by relative investment prices (PWT)

Growth accounting decomposes GDP growth

The operating surplus approximately reects

where P denotes the price level, Y the real

CHINAS ECONOMIC GROWTH 19782025

The share of labor remuneration in nominal

Since they are the only hard facts that we know

share of laborers with college level education

CHINAS ECONOMIC GROWTH 19782025

Table 2. Cointegration equations, 19782002/03

D. Average years of schooling

E. Share of laborers with primary

G. Share of laborers with college

Lag length of the VEC

Variance decomposition period 3

Variance decomposition period 10

positive growth rate that eventually (by 2014)

CHINAS ECONOMIC GROWTH 19782025

assuming the labor share at to be constant

This degree or above

United States China United States

CHINAS ECONOMIC GROWTH 19782025

college-level associate degrees) China has now

United States, and in 2006 it was scheduled to

China: BA+ Associate degree

U.S.: Freshmen (2- and 4-year programs)

China: MA+ PhD --- right scale

Figure 9. Tertiary level new enrollment, China, and United States.

72,508 in 1998, but perhaps only one-third

not be felt until some years later, when fresh

China: BA + Associate degree

Figure 10. Tertiary level graduates, China and United States.

China's Economic Growth 1978-2025

China's Economic Growth 1978-2025