Professional Documents
Culture Documents
China's Economic Growth
China's Economic Growth
China's Economic Growth
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
The University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to
Economic Development and Cultural Change
http://www.jstor.org
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms
Gregory C. Chow
Princeton University
Kui-Wai Li
City University of Hong Kong
I. Introduction
This article attempts to account for Chinas economic growth in terms of
labor, capital, and total factor productivity by estimating a Cobb-Douglas
production function using official Chinese data. It is an extension of Gregory
C. Chows earlier work in 1993 and has two purposes: to find out whether
the parameters of the production function have changed and to use the production function to forecast GDP growth up to 2010.1 Official data on labor
force and national output are readily available, although there was a change
in the national income accounting system in 1994 from using an old and
narrower definition of national income to using a measure of GDP that conforms to the standard definition of national income accounting. The estimation
of a capital stock series is discussed in Section II. Section III provides estimates
of the parameters of a Cobb-Douglas production function. Section IV presents
the decomposition of growth of aggregate output into its three components
and provides projections of GDP up to 2010. Section V contains some concluding remarks.
II. Data on Capital Stock and Output
The construction of our capital stock series is based on Chows 1993 study
(tables 6 and 7), in which several capital stock series were used to estimate
aggregate production functions, and very similar results were obtained.2 We
have chosen the series with an initial capital stock of 2,213 (all such numbers
in 100 million yuan, and hereafter) at the end of 1952 because, as with the
construction of that series, we continue to use aggregate net investment in
this study to construct a capital stock series for the entire economy. That series
provides a value of capital stock of 14,112 at the end of 1978.
If net investment (called accumulation in official statistics) data in
2002 by The University of Chicago. All rights reserved.
0013-0079/2003/5101-0012$10.00
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms
248
constant prices were available, we could construct a capital stock series from
1978 onward simply by adding them to the initial capital of 14,112 at the
end of 1978. There are two problems, however. First, published accumulation data are in current prices, not constant prices. From 1952 to 1978, prices
of investment goods remained almost constant, and accumulation in current
prices can be treated as accumulation in constant prices, as in Chow.3 After
1978, when economic reform started, prices of investment goods began to
change, but an appropriate price index is not readily available. Second, after
1994, official Chinese national income statistics were changed from national
income available, which equals consumption plus accumulation, to the
new GDP, which equals final consumption expenditure plus gross capital
formation plus net export of goods and services. The coverage of the new
GDP and GNP is broader than the former national income statistic by the
inclusion of some service items that were previously excluded.
For the period 197892, we tentatively use a rate of depreciation equal
to 4% (to be revised below) and apply the equation
K t p 0.96K t1 RGI t ,
(1)
where real gross investment RGIt at period t is obtained by using the following
national income accounting identity in real terms based on GDP data in the
Statistical Yearbook of China (hereafter SYC ):4
GDP p Consumption Gross Investment
Net Export of Goods and Services.
(2)
(3)
(4)
Gross investment (GI) is found in SYC. Net investment (NI) equals GI less
total provincial depreciation (Dep), as estimated above.
We first estimated the implied depreciation rates for 199398 by solving
equation (1) for the coefficient of K t1 given the initial estimates of capital
stock and RGI. This yields 0.9549, 0.9520, 0.9492, 0.9450, 0.9394, and
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms
249
(5)
(6)
Table 2 summarizes the least squares estimates of equations (5) and (6).
The estimates of the capital and labor coefficients in Chow based on the
sample period 195280, excluding 195869, are 0.6353 and 0.3584, respectively, showing a constant return to scale as they sum to 0.9937.13 For this
period, China did not have technological progress, as can be detected by
adding a trend variable.14 Using equation (6) and after having detected no
trend in the above sample period, Chow found a capital coefficient of 0.6317
with a standard error of 0.0219, which is very similar to our coefficient of
0.6284 in table 2.15 The regressions in table 2 suggest that there is an average
increase in total factor productivity of about 2.6% per year from 1978 to 1998.
This is a new result to be added to Chows conclusion that technological
progress was absent in 195280, as the estimate of the trend coefficient is
0.0065 with a standard error of 0.0187.16 Residuals of equation (6) are shown
in the last column of table 1.
The sum of the point estimates of the coefficients of ln K and ln L is very
close to one, supporting the assumption that their parameter values actually sum
to one. The F(1, 31) statistic for testing this assumption is only 0.0458, strongly
supporting this assumption and, thus, the use of equation (6). It can also be
seen from table 2 that the trend coefficients of 0.0262 and 0.0263 estimated
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms
TABLE 1
Economic Data
Year
RGDP
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
799.32
911.22
963.98
1,025.52
1,170.20
1,222.95
1,492.32
1,614.62
1,591.44
1,119.04
1,046.31
1,158.21
1,349.25
1,577.85
1,846.42
1,712.93
1,601.03
1,910.37
2,354.79
2,520.24
2,592.18
2,807.20
2,839.17
3,074.97
2,993.44
3,226.84
3,624.10
3,899.53
4,203.96
4,425.03
4,823.68
5,349.17
6,160.97
6,990.89
7,610.61
8,491.27
9,448.03
9,832.18
10,209.09
11,147.73
12,735.09
14,452.91
16,283.08
17,993.66
19,718.73
21,454.67
23,129.01
2.0729
2.1364
2.1832
2.2328
2.3018
2.3771
2.6600
2.6173
2.5880
2.5590
2.5910
2.6640
2.7736
2.8670
2.9805
3.0814
3.1915
3.3225
3.4432
3.5620
3.5854
3.6652
3.7369
3.8168
3.8834
3.9377
4.0152
4.1024
4.2361
4.3725
4.5295
4.6436
4.8197
4.9873
5.1282
5.2783
5.4334
5.5329
6.3909
6.4799
6.5554
6.6373
6.7199
6.7947
6.8850
6.9600
6.9957
Dep
3,989.12
5,406.88
7,094.10
8,781.32
10,486.41
11,981.24
IPD
Residual
.088602
.014768
.016008
.006048
.0670675
.0518618
100.00
103.56
107.47
109.88
109.76
110.94
116.39
128.23
134.05
140.88
158.00
171.98
181.68
193.92
209.17
239.64
287.17
324.99
344.26
347.07
343.27
2,212.993
2,380.993
2,575.993
2,760.993
2,977.993
3,210.993
3,589.993
4,147.993
4,648.993
4,843.993
4,942.993
5,125.993
5,388.993
5,753.993
6,223.993
6,527.993
6,825.993
7,182.993
7,800.993
8,484.993
9,132.993
9,873.993
10,614.993
11,444.993
12,192.993
13,024.993
14,111.993
14,882.124
15,735.359
16,569.286
17,653.459
18,991.843
20,627.591
22,598.034
24,876.884
27,413.644
30,524.741
33,773.284
36,805.533
40,115.037
44,131.925
50,105.391
56,732.505
64,013.641
71,700.806
79,542.496
87,764.476
.0115272
.0140107
.006527
.0159559
.025393
.000783
.073868
.045429
.013125
.007434
.0054
.024562
.017425
.0046278
.0539845
.0841383
.0721603
.0837293
.0859975
.0293702
.066794
.06427
.021606
.005646
.004742
.001545
.012354
.023415
.038171
Note.RGDP p GDP (100 million, 1978 yuan); L p labor force (100 million); Dep p
depreciation (100 million, current yuan); IPD p implicit price deflator; K p capital stock (100
million, 1978 yuan).
250
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms
251
TABLE 2
Chinas Aggregate Production Function: 195298
Intercept
1.7451
(.4580)
1.6612
(.1960)
ln K
ln L
.6136
(.0772)
.4118
(.1996)
ln (K/L)
Trend
R 2/s
.9979/.0465
.6284
(.0258)
.0263
(.0025)
.0262
(.0024)
.9946/.0459
Note.Figures in parentheses are standard errors of the estimate, K p capital stock (100
million, 1978 yuan), L p labor force (100 million), s is the standard error of the regression,
and R2 is the adjusted R square.
with and without making this assumption are almost identical. The estimate of
the trend coefficient is thus robust against the assumption of constant returns
to scale in aggregate production. We also test whether the exponent of capital
(and of labor under constant returns) remains unchanged after 1978 in the context
of equation (6). The null hypothesis is that, except for the trend term that starts
in 1979, the production function remains unchanged after 1978. We find the
sum of squares A of residuals from equation (6) using all observations 195298,
excluding 195869, with 32 degrees of freedom (df) to be 0.06727. The prereform regression with 12 df has a sum B of squared residuals that is equal to
0.02283. The sum C of squared residuals for the 197898 period with 18 df is
0.03924. The F(2, 30) statistic for testing the stability of both the intercept and
the first coefficient of equation (6) equals 1.2566.17 Thus, the null hypothesis
that the parameters of the Cobb-Douglas production function remained the same
for the two subperiods cannot be rejected even at the 20% level (the critical
value at the 20% level is 1.70). Such parameter stability increases our confidence
in using the equation for forecasting GDP growth for another decade.
By examining the Durbin-Watson (D-W) statistic of the residuals of
equation (6), which equals 0.638 (shown in the last column of table 1), we
have found a significant positive serial correlation. Estimating a first-order
autoregression for the residuals yields a coefficient of 0.6234 with a standard
error of 0.1263. The D-W statistic of the residuals of this autoregression is
1.633, indicating no remaining positive correlation. The Cochrane-Orcutt procedure is applied to estimate equation (6), yielding a coefficient of 0.5577
(0.0468) for ln (K/L) and a trend coefficient of 0.03028 (0.0040), with standard
errors in parentheses. This regression will be used to account for past output
growth and to predict future output growth in the next section.
IV. Accounting for Chinas Economic Growth
In table 1, we find that real GDP in China grew from 799.32 (100 million,
1978 yuan) in 1952 to 23,129.01 in 1998, implying an average annual exponential rate of growth of 0.07315, or 7.6%, per year. Before economic
reform began in 1978, there was no increase in total factor productivity, and
the average annual exponential rate of growth from 1952 to 1978 (with real
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms
252
GDP equal to 3,624.10) was only 0.05814, or 6.0%, per year. From 1978 to
1998, the exponential rate of growth was 0.09267, or 9.7%, per year. The
exponential rate of 0.09267 is decomposed into the exponential rate of growth
of capital of 0.09138 times its coefficient 0.5577 (or 0.05096) plus the exponential rate of growth of labor of 0.02776 times its coefficient 0.4423 (or
0.01228) plus the exponential rate of growth of productivity of 0.0303. The
three components add up to 0.09352, which is slightly different from the
observed increase of 0.09267 because the regression representing the CobbDouglas production function does not explain real GDP in the beginning and
terminal years (1978 and 1998) exactly. According to the estimated production
function of the 0.09352 exponential rate of growth explained, 0.051 results
from the increase in capital, 0.012 results from the increase in labor, and
0.0303 results from the increase in total factor productivity. The importance
of capital accumulation and increase in productivity in accounting for Chinas
economic growth in the postreform period of 197898 is a major conclusion
of this article. While capital accumulation is very important, accounting for
0.051/0.09352, or 54%, of the growth, productivity increase is also important,
accounting for 0.03/0.09352, or 32%, of the growth, leaving only 13% to
labor. (The sum of the three differs from 100% because of rounding.)
This study has implications for Chinas future growth. Let us consider
what the growth in the decade following 1999 would be if growth in total
factor productivity were to remain constant at 0.0303, to be reduced by half
to 0.0151, and to be reduced to zero. We can simulate the path of Chinas
real GDP by using the estimated production function (eq. [6]) corrected for
autoregressive residuals and the capital formation equation (eq. [3]); the labor
and investment functions can be constructed as follows:
L t p (L t1 ) # (1 0.0281),
(7)
It p 0.3373RGDPt1.
(8)
The labor growth rate in the 197898 period is 0.0281. In equation (8) the
coefficient 0.3373 is the ratio of the sum of net investments to the sum of
real GDP in 197898. Note that the coefficient 0.0281 in equation (7) tends
to be an overestimate because of the tight birth-control policy introduced in
1980 and that the coefficient 0.3373 in equation (8) may be an underestimate,
as suggested by the higher investment rates in later years. The effect of the
former on the future growth rate is small because labor accounts for only
0.012 of the exponential growth rate of GDPreducing it to half only means
an effect of 0.006 on the exponential growth rate. If we were to adjust the
rate of investment upward to, for example, 0.036, the stock of capital would
be increased only slightly in the next decade since investment is a small
fraction of capital, and proportional increase in GDP would be about 0.6 as
much as the increase in capital stock. We will use equations (3), (6), (7), and
(8) for the simulation exercise.
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms
253
TABLE 3
Simulated Real GDP: 19992010
Year
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Exponential growth
Annual growth
TFP p .03028
TFP p .01514
TFP p .0
25,307.96
27,698.33
30,320.99
33,198.89
36,357.27
39,823.88
43,629.25
47,806.94
52,393.89
57,430.70
62,962.05
69,037.09
.0836
.0872
24,927.69
26,853.72
28,915.72
31,122.82
33,484.79
36,012.00
38,715.53
41,607.18
44,699.52
48,005.94
51,540.71
55,319.04
.0664
.0687
24,553.12
26,035.12
27,576.83
29,180.15
30,847.00
32,579.40
34,379.41
36,249.16
38,190.85
40,206.75
42,299.20
44,470.60
.0495
.0507
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms
254
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms
255
measurement errors, and the loss of degrees of freedom when more parameters
are used. For the purpose of forecasting, we do not need to forecast variables
that determine changes in human capital, technology, allocation efficiency,
and sectorial composition.
Under the assumption of constant returns to scale, which is strongly
supported by the data, we need only two parameters, the elasticity of output
with respect to capital and the coefficient of trend, which are estimated to be
about 0.6 and 0.03, respectively. The 0.6 figure is well supported empirically,
as it was found in works by Chow and also by N. Gregory Mankiw, David
Romer, and David N. Weil, who estimated the classic Solow growth model
with a Cobb-Douglas production function using data for approximately 70
countries.21 If the 0.6 figure is correct, by regressing log(output/labor) minus
0.6 times log(capital/labor) on time, one finds total factor productivity to have
zero growth from 1952 to 1978 and to grow at an average exponential rate
of approximately 0.03 from 1978 to 1998. Hence, the estimate of the second
parameter is also well supported. Using these two parameters alone, we have
been able to obtain a reasonably accurate decomposition of output growth in
China. For forecasting the near future, the 0.6 figure for the coefficient of
capital is unlikely to change, and one may allow for a slightly different trend
in the growth of total factor productivity. Since China still has plenty of
opportunity to adopt new technology from abroad and to carry out additional
institutional reforms, it is unlikely that the 0.03 figure will be reduced substantially from its past trend by as much as a half in the period ending 2010.
Thus, the TFP p 0.01514 column in table 3 can serve as a very conservative
estimate of Chinas output growth even after allowing for the limited effect
of a possible reduction in the growth of the labor force (labors contribution
to the exponential growth rate was only 0.012, historically).
There has been a debate on whether substantial growth in total factor
productivity occurred in selected East Asian countries (see, e.g., Alwyn Young
and Paul Krugman) and whether the lack of such would inhibit future growth
in these countries.22 As far as China is concerned, this article has reached the
following conclusions. There was a substantial total factor productivity growth
at the annual rate of about 0.03 in China during the period 197898. Even
if total factor productivity growth is to be reduced somewhat, however, in the
next decade the Chinese economy would still manage to grow at a substantial
rate of at least 7% because of the expected high rate of capital formation of
over 30% of GDP and the high capital elasticity of about 0.6.
Notes
* The authors are indebted to the excellent assistance of Queenie Y. P. Wu of
the APEC Study Center, City University of Hong Kong; the valuable comments from
an anonymous referee; participants in a seminar at Yale University, May 2000; Jin
Zhang; and Anloh Lin. Financial support from the Center for Economic Policy Study
at Princeton University to Gregory Chow and from the University Grants Committee
of the City University of Hong Kong to Kui-Wai Li is gratefully acknowledged.
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms
256
This content downloaded from 39.57.1.42 on Sat, 26 Mar 2016 18:00:50 UTC
All use subject to http://about.jstor.org/terms