1

Export and Economic Growth in China: A Demand-Oriented Analysis

Justin Yifu Lin*
Peking University and Hong Kong University of Science and Technology

Yongjun Li
Peking University


Abstract: Many studies, based on the accounting identify of gross domestic product
(GDP), found that the contribution of foreign trade to Chinas economic growth over
the past 20 years was very small. In this paper, we re-examine the issue and find that
those studies underestimate the contribution of exports to GDP growth by overlooking
the indirect impacts of exports on domestic consumption, investment, government
expenditures and imports. We propose a new estimation method and find that a ten
percent increase in exports resulted in a one percent increase in GDP in the 1990s in
China, when both direct and indirect contributions are considered.

Key words: Economic growth, International trade, Export, Growth Accounting











________________________________
*Correspondence author. Please send the correspondences regarding this manuscript
to Justin Yifu Lin at China Center for Economic Research, Peking University, Beijing
100871, China; e-mail: jlin@ccer.pku.edu.cn
2
Export and Economic Growth in China: A Demand-Oriented Analysis
Justin Yifu Lin and Yongjun Li

Since Chinas initiation of the economic reforms and the adoption of its open door
policy, foreign trade in China has experienced rapid growth. In 1978, the total value
of both imports and exports equaled US$ 20.64 billion. By 2001, this number
increased to US$ 509.8 billion. On average, the annual growth rate of foreign trade
has been as high as 15%, 5.5 percentage points higher than Chinas 9.5% annual GDP
growth rate over the corresponding period. The relationship between Chinas foreign
trade and economic growth has become a hot issue among academics and China
watchers. So far, the discussions in this area can be roughly divided into two
categories. One set of discussions focuses on the causality between exports and
economic growth, i.e., seeking answers to the question, is economic growth propelled
by exports or vice versa? The other discussion focuses on the empirical analysis of the
contribution of foreign trade to economic growth. These studies generally assume that
export growth has pushed forward economic growth, and they estimate foreign trades
contribution to economic growth accordingly. This paper belongs to the second
category.
Previous estimations of the contribution of export all share a common
shortcoming. They estimate only the direct impact of export and ignore the indirect
impacts, which include consumption, investment, government expenditure and
imports. Because of this shortcoming, previous estimations should not be used to
guide for policy formulation, or the policies may be misleading. For example, a
typical conclusion is that exports have not generated much economic growth since the
start of the reforms. Following this argument, economists tend to emphasize
domestic demand and to overlook the importance of exports. In contrast to this
argument, the analysis in this paper demonstrates that exports have always been and
will continue to be a major contributor to Chinas economic growth.

I. The Issues of Traditional Estimations

Previous estimations use the following accounting identity as the starting point
and analyze the contributions of exports accordingly:
( ) (1) Y C I G X M = + + +
where M X G I C Y , , , , , represent, respectively, national income, consumption,
investment, government expenditure, exports and imports. Differentiating (1) with
respect to time gives:
) 2 ( ) ( M X G I C Y

+ + + =
where
dt
dY
Y =

and other terms are computed similarly. Equation (2) can be

rewritten in the following form:
3
) 3 (
Y
NE
NE
E N
Y
G
G
G
Y
I
I
I
Y
C
C
C
Y
Y

+ + + =
where M X NE = signifies net exports;
NE
E N
G
G
I
I
C
C
Y
Y

, , , , signify the
growth rates of the corresponding variables and
Y
NE
Y
G
Y
I
Y
C
, , , signify, respectively,
the ratios of consumption, investment, government expenditure and net exports to
national income. When the quantity of any component is changed, its direct impact on
the growth rate of national income is given by equation (3). Therefore, the direct
impact of changes in net exports can be estimated accordingly. In particular,
Y
NE
NE
E N


(or
Y
E N

), the fourth term on the right side of equation (3), signifies the part of GDP
growth that is attributed to net exports. In the literature, this term is usually regarded
as a measure of the impact of foreign trade growth on the GDP growth rate.
Accordingly,
Y
Y
Y
E N

(or
Y
E N

) measures the proportional contribution of net exports

in the overall GDP growth rate and is usually called the contribution of foreign trade
to GDP growth rate.
In China, government expenditure, G, is not singled out in the calculation of gross
domestic product through an expenditure approach due to data constraints. Moreover,
in the GDP calculation through expenditure approach, C represents final consumption,
which includes both household consumption and government consumption; I
represent total capital formation, which includes both enterprise investment and the
part of government expenditure that is directly used for capital formation. As such, we
rewrite equation (3) as:

1 1 1
1 1 1 1 1 1 1
(4)
t t t t t t t
t t t t t t t
Y C C I I NE NE
Y C Y I Y NE Y



= + +
where
1
=
t t t
Y Y Y and , ,
t t t
C I NE are calculated accordingly. In line with
(4), the national income identity through the expenditure approach (i.e., equation (1))
can be rewritten as:
(5) Y C I NE = + +
Table 1 presents the results of estimations based on equation (4).



4

Table 1Decomposition of the GDP Growth Rate
(unit%)
Decomposition of GDP
Growth Rate
Decomposition of GDP
Growth Rate Year
GDP
Growth
Rate C I NE
Year
GDP
Growth
Rate C I NE
1981 5.32 5.72 -0.96 0.57 1991 8.83 5.19 3.27 0.37
1982 12.11 6.79 3.7 1.63 1992 12.68 7.72 6.66 -1.7
1983 9.52 6.19 4.07 -0.74 1993 16.44 6.43 13.36 -3.36
1984 12.38 7.47 5.73 -0.82 1994 13.95 7.33 3.12 3.5
1985 11.39 7.62 8.44 -4.67 1995 9.74 5.31 3.9 0.53
1986 10.24 5.51 3.33 1.4 1996 10.25 7.06 2.54 0.65
1987 10.66 5.41 2.63 2.62 1997 8.76 4.74 2 2.01
1988 11.26 7.59 4.9 -1.23 1998 8.1 5.31 2.43 0.36
1989 2.88 2.3 0.71 -0.13 1999 6.72 5.59 2.08 -0.95
1990 5.32 1.23 0.03 4.06 2000 8.56 5.68 3.08 -0.19
Source: The Gross Domestic Product by Expenditure Approach in various
issues of China Statistical Yearbook
Note: 1. Real GDP is obtained by dividing nominal GDP by CPI, taking 1978 as
the base year, which is used throughout this paper.
2. Because we use the data from Gross Domestic Product by Expenditure
Approach, the annual economic growth rate is a bit different than
reports utilizing data from GDP by Sectors. For instance, in Table 1,
the economic growth rate was 8% in 2000; while under the latter method,
the figure should be 8.56%. The reason is that in 2000, GDP through the
expenditure approach was larger than that by Sectors.

In the estimation of foreign trades contribution to GDP, the method presented
above has two characteristics: 1) the calculation process is simple; 2) data are easily
available. For these two reasons, this has been the most commonly used method for
Chinese economists when they analyze foreign-trade-related issues (for instance, Zhu
(1998), Wang (1998), Foreign Trade Policy Research Group (1999) and Peng (1999)).
The contribution of foreign trade estimated in this way is, however, usually very
low. For example, the Foreign Trade Policy Research Group (1999) concluded, the
contribution of net export is very limited...except for a few years (like 1990 and 1994),
when the foreign trade surplus increased dramatically. Only in those few years, was
net exports contribution relatively large (more than 3 percentage). Results of this
kind have fueled policy recommendations arguing for more emphasis on domestic
demand. Zhu (1998) held that Chinas economic development in the last two decades
is mainly propelled by domestic demand like investment and consumption instead of
export.
Furthermore, some economists found that the foreign trade balance and GDP
growth are negatively related, after they compared the time series data of GDP and
net exports (Zhang & Hu (1999)). That is to say, in years that see high GDP growth
5
rate, the estimated contribution of net exports is usually low.
The discovery of the negative relationship has challenged the theoretical
foundation of the estimation method. According to the national income identity (i.e.,
equation (1) or (5)), GDP is the sum of consumption, investment, government
expenditures and net exports. Therefore, net exports and GDP should be positively
related. In fact, only under the following two exceptional conditions will they be
negatively related: 1) every time net exports increase, there comes an exogenous force
that could affect other variables in the identity, which finally leads to a decrease in
GDP; 2) when independent variables in the identity are correlated. For instance, an
increase in investment and consumption may lead to an increase in imports and
consequently a decrease in net exports but the GDP as a whole still increases.
It is difficult to believe that an unpredictable exogenous force would take effect
whenever net exports increase. Therefore, the second condition might be the major
reason for the negative relation between net exports and GDP growth. This is just
the explanation given by Zhang & Hu(1999). But the key issue is that since we have
realized the existence of the correlation among consumption, investment, imports and
exports, we should also acknowledge the limitations of the traditional estimation
method, because it only illustrates the direct quantitative relationship between the net
exports and national income but it fails to reflect the relationship between net exports
and other variables. Therefore, the estimated low contribution might be attributable to
overlooking such correlations and thus should not be used as the foundation for
theoretical research and policy discussion.
Table 2 presents the estimated relationships between exports and the other three
variables (consumption, investment and imports). The simple regression method is
applied in the process. As Table 2 indicates, in China, the elasticity of consumption
with respect to exports is 0.86, that of investment is 0.47 and of imports 0.63. In other
words, increases in exports not only affect imports significantly but such increases
also have great impacts on consumption and investment. Although the impact of
exports on imports is comparatively larger than that on consumption and investment,
we are in no position to neglect the latter two effects. Therefore, the previous method
based on the national income identity has estimated only the impact of exports on
imports and has taken the result as the overall effect of exports on economic growth,
thus overlooking the indirect effects of exports on economic growth through
consequent changes in consumption and investment. For this reason, this method
actually underestimated the impact of exports on economic growth.









6
Table 2 The Impact of Exports on Imports, Consumption and Investment

Model Regression Result
0 1
( ) ( ) Log M Log X = + +

0
2
( ) 0.86 ( )
(10.64)
113.22 0.8618 1.534
m
Log M Log X e
F Adj R dw
= + +
= = =

0 1
( ) ( ) Log C Log X = + +

0
2
( ) 0.47 ( )
(7.505)
56.32 0.7545 1.526
c
Log C Log X e
F Adj R dw
= + +
= = =

0 1
( ) ( ) Log I Log X = + +

0
2
( ) 0.63 ( )
(15.88)
252.24 0.9331 1.714
i
Log I Log X e
F Adj R dw
= + +
= = =

Note:
(1) The data set covers data for the years 1979 through 2000; the sample size is
therefore 22.
(2) Serial correlation exists in all the simple regression models above. We
assume that errors follow an (1) AR model and estimate and coefficients
on other explanatory variables in turn. An iterative approach is followed to
ensure that estimations converge. Results presented in this table are obtained
in this way.

The limitations of the traditional estimation method inspired us to re-estimate the
impact of foreign trade on economic growth. As illustrated in the previous paragraphs,
the economic roles of exports and imports in economic growth are quite different.
Therefore, the subsequent analysis will mainly focus on the relationship between
exports and economic growth. Both the direct and indirect impact of export will be
examined so that we can understand the overall impact of exports on economic
growth in China. For this reason, the rest of this paper is divided into two parts: 1) we
propose a new method to investigate the relationship between exports and economic
growth; then, 2) we estimate the overall impact of exports on economic growth, and
compare our result with traditional estimations.

. Theoretical Analysis of the Relationship Between Exports and Economic
Growth

The relationship between exports and economic growth has always been a hot
issue among economists. The debate on Export-led Economic Growth is just one
7
offshoot of this discussion. Generally speaking, there are two approaches to
addressing the issue:
The first approach is to study the contribution of exports to economic growth
through analysis of the supply side of the economy. This approach originates from the
neo-classical economic growth theory. According to the approach, the major source of
economic growth lies in two areas: 1) factor input increases and 2)
efficiencyimprovements. Correspondingly, analysis from this approach often regards
exports as a factor that can affect technological progress or other factors that are
related to economic efficiency. In practice, the contribution of exports is thought to be
included in the residuals of growth accounting. The new growth theory endogenizes
the mechanism through which exports affect economic growth. For instance,
Grossman and Helpman (1990) proposed a two-country growth model with
endogenous technological progress. In their model, exports help to promote
technology and knowledge diffusion and thus accelerate economic growth.
How to introduce exports into the production function is the main problem
involved in econometric analysis that follows the neoclassical approach. Some
analysts directly include exports in the production function as the third variable beside
labor and capital. Others use more sophisticated methods. For example, Feder (1983)
divided the economy into two sectors: the export sector and the domestic sector.
Because the export sector has to cope with foreign producers and consumers more
often than the domestic sector does, it is more efficient than domestic sector. In order
to capture the diffusion process of technology and knowledge, Feder introduced the
output of the export sector (total exports) into the production function of the domestic
sector as an element that could affect its economic efficiency.
The second approach is to study the contributions of exports through analysis on
the demand side of the economy. Since this is the approach adopted in this paper, we
will make a more thorough review in the following discussions.
This demand-side approach is also called demand oriented analysis or
Post-Keynesian analysis. According to traditional Keynesian theory, an increase in
exports is one of the factors that can stimulate increases in demand and thus will
surely lead to increases in outputs. However, this approach has not been used widely.
According to McCombie and Thirlwall (1994), this is because of the remnant of Says
Law in peoples mind. Most people believe that the major constraints of modern
economic growth lie on the supply side instead of on the demand side. In other words,
only increases in factor inputs and improvements in economic efficiency can
stimulate economic growth.
Proponents of demand-oriented analysis disagree with this view
1
. While analyzing
the waxing and waning process of British industry, Kaldor (1972) pointed out that,
contrary to the traditional view which attributed the rate of industrial development in
England to the rate of saving and capital accumulation and to the rate of technical
progress due to invention and innovation, more recent evidence tends to suggest that
Britains industrial growth was export-led from a very early date. Kaldor also
pointed out that, there can be little doubt that throughout the nineteenth century and

1
McCombie and Thirlwall (1994) elaborated on the issues of neoclassical growth theory.
8
also in the present century, right up to the Second World War, Great Britains
economic growth was closely dependent on the growth of her export. Given the fact
that her share of the world market was bound to decline continually, , it was quite
inevitable that both the growth of production and the accumulation of capital should
be much lower in Britain than in the countries that were subsequently
industrialized .
Thirlwall, McCombie and others later developed this view into a new theoretic
framework that analyzes the relationship between exports and economic growth
2
.
Briefly speaking, the theory has the following characteristics: (1) contrary to what
most people believe, the Keynesian model can be used to analyze long-term
phenomena like economic growth; (2) exports are an autonomous component of
demand; (3) the role that exports plays in an open economic model is as important as
investment in a closed economic model; (4) the role of the balance of payments as a
constraint on economic growth is important.
The conclusions derived from above principles of Thirlwall and McCombies
theory revert to the Harrod Foreign Trade Multiplier and the Hicks Super-Multiplier.
According to Thirlwall (1979), when the balance of payment constraint was
introduced in the model, the conclusion was:
3

t t ft dt
Bt
x e p p
y
+ + +
=
) )( 1 (

Where
t Bt
x y , denote the equilibrium growth rate of economy and exports
under the constraint of balance of payments, respectively;
t
e ,
dt
p ,
ft
p denote
changes in the exchange rate, domestic prices and foreign prices, respectively;
, denote the price elasticity of imports and exports, in turn and denotes the
income elasticity of imports. According to this equation, increases in exports ( 0 >
t
x )
will boost the economy ( 0 >

t
x
).
McCombie (1985) did not include the constraint of balance of payment; rather, he
derived the results directly from traditional Keynesian model. One result is as follows:
( )( ) M X G I C
K
Y
M X G I C

+ + + =
1

where Y

denotes GDP growth; M X G I C

, , , , denote changes in autonomous
consumption, investment, government expenditure, exports and imports, respectively,
C
denotes the ratio of consumption to GDP, while
I
,
G
,
X
,
M
denote the ratio

2
See Dixon and Thirlwall (1975), Thirlwall (1979, 1980, 1987), McCombie (1985), McCombie and Thirlwall
(1994, 1997, 1999).
3
The equation has been revised for convenience under the condition that it will not change any conclusions of the
author.
9
of investment, government expenditure, exports and imports to GDP, respectively.
According to this model, an increase in exports ( 0 > X

) will increase GDP by

(
K
X
X

).
Based on empirical studies, Thirlwall (1979) thought he could assume that, in the
long term, the relative price of a commodity sold in the domestic market to that sold
in the foreign market would not change much if prices were measured in the same
currency. That is equivalent to assuming that 0 =
t ft dt
e p p over the long run.
Under this assumption, the conclusions of Thirlwall are equivalent to Harrods theory
of the Foreign Trade Multiplier. The only difference is that Harrods multiplier theory
is more convincing in the Thirlwall framework. As for McCombies (1985)
conclusions, if we assume that the autonomous changes in consumption, investment,
government expenditure and imports are not important, the outcomes also revert to
the Harrod Multiplier. In fact, among all the assumptions, the insignificance in the
autonomous changes of government expenditure may be the only one that needs
further clarification. To clarify the issue, McCombie (1985) cited the conclusion of
the Cambridge Economic Policy Group (CEPC): government expenditure is
endogenously restricted by government revenue. Because revenue from taxation is the
major source of income for the government, and because revenue from taxation is
proportional to national income, changes in government expenditures are after all
restricted by changes in national income. That is to say, autonomous changes are
unimportant.
The econometric tests of Thirlwall and others theories from this approach usually
depend on the following two econometric models. The only difference between the
two lies in whether or not the model includes government expenditure as an
autonomous variable.
t t t
x gdp + + =
or
t t t t
g x gdp + + + =
where g x gdp , , denote the growth rate of GDP, exports and government
expenditures, respectively
4
. These tests clearly illustrate the theoretical essence of
Thirlwalls theory: i.e., to regard exports as an important exogenous variable that can
significantly affect economic growth.
The theoretical model we construct here follows Thirlwalls approach. Our
primary concern is how to measure exports indirect impact on GDP growth via its
influences on consumption, investment and government expenditure. Since China has
a large amount of foreign exchange reserves, we have not introduced the balance of
payment constraint as a factor that affects economic growth. Instead, we start directly
from the traditional Keynesian model and explicitly include the indirect impact of

4
See Atesoglu (1994) and Somez (1996).
10
exports in the model. As in Thirlwalls model, our model emphasizes that exports
should be considered as an exogenous and autonomous element stimulating economic
growth. It is also quite similar to the model in McCombie (1985), which also started
directly with the traditional Keynesian model.
The model we construct is a simultaneous equations model. It includes four
equations: the national income identity by expenditure approach, the consumption
function, the investment function and the import function. The specifications of the
functions are as follows:
Consumption Function: we assume that consumption is determined by residents
permanent income and that their expectations of permanent income adjust in an
adaptive way. Let
p
t
Y denote residents permanent income, then:

1
(1 ) (1 ) (0 1) (6)
p k
t t t t k
Y Y Y Y

= + + + + < <
By applying the Koyck transformation to (6) (see Gujarati, 1995, pp 594-611), a
linear model of the consumption function can be obtained. That is:

0 1 2 1
(7)
t t t t
C Y C

= + + +
Investment Function: the theory on the investment function is the area that is most
difficult and most hotly debated in macroeconomics. Theoretically speaking,
investment is closely related to entrepreneurs expectations of the future.
Unfortunately, economists so far have failed to appropriately include this expectation
in their economic models. In our model, we simply assume that investment depends
on interest rates and the size of economy. As in the consumption function, a linear
model is used to approximate the investment function, i.e.,

0 1 2
(8)
t t t t
I Y R = + + +
where
t
R denotes the real interest rate at time t.
In order to facilitate comparison of the results with the traditional ones that
measure only the direct impact of exports, consumption and investment in the new
model actually include both resident and government expenditures. Again, we have to
find whether or not autonomous expenditures by the government are important.
According to CEPGs conclusions, government expenditure cannot be regarded as a
kind of autonomous demand, because government expenditure is a function of
taxation, which is itself a function of national income. Whether or not this assumption
applies to China is an issue that needs verification.
Figure 1 shows the total government revenue and expenditures for the years 1978
through 2000. It is evident that government expenditures and government revenue are
closely related. Although evidence shows that the revenue of the central government
in proportion to GDP is declines over time, many researchers still maintain that the
total revenue of the government as a proportion of GDP may not be declining. The
reason is that the extra budgetary revenue of the government is increasing rapidly (see
Hu (2001), chapter 7). For the reasons above, we accept the assumption that
11
government expenditure is a function of national income.


Figure 1: Government Revenue and Expenditures
(Unit= 100 million yuan)
0
500
1000
1500
2000
2500
3000
3500
1
9
7
8
1
9
7
9
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
Total Revenue Total Expenditure
Partial Expenditure

Source: China Statistical Yearbook 2001
Note: Partial Expenditure is the total expenditure less transfer payments.

Import Function: we assume imports are determined by domestic demand and
foreign exchange rates. Because each component of domestic demand may have a
different influence on imports, we specify the import function as follows. Again, it is
assumed to take a linear form.

0 1 2 3 4
(9)
t t t t t t
M C I X ER = + + + + +
where
t
ER denotes the real effective exchange rate at time t.
Since we have already taken exports as an autonomous variable, it is not included
in the model. This needs further clarification. According to Thirlwall (1994), the
export function is assumed to take the form:

Z
E P
P
b X
f
d
|
|
.
|

\
|
=
where X and Z denote, respectively, the total sum of exports and GDP of countries all
around the world. E denotes the price of foreign currency (measured in domestic
currency).
d f
P and P denote domestic and foreign prices, respectively;
denotes the price elasticity of exports and denotes the income elasticity of exports.
Then, Thirlwall assumes that commodity prices measured in a single currency will
not change much over the long run, i.e.,
d
f
P
E P
approximates a constant. The export
function can thus be simplified to

Z b X

= , which says that exports only depend on

12
the overall income levels of the world. Therefore, we only need to test the long-term
fluctuations in the real effective exchange rate of the RMB (the Chinese currency) and
the result will tell us whether or not Thirlwalls assumptions apply to China. In other
words, these computations will verify whether we can reasonably regard exports as an
exogenous variable.
Figure 2 shows the trend of fluctuations in the nominal effective exchange rate
(NEER) of RMB and the real effective exchange rate (REER) of RMB estimated by
the IMF. We find that the REER became approximately stable only after 1994, so
Thirlwalls assumption may not apply to China. This may be due to the overestimated
RMB exchange rate under the planned economy. As the economy underwent the
transition from a planned economy to a market economy, the exchange rate
depreciated dramatically and finally converged to the equilibrium level.

Figure 2: Fluctuations in Nominal and Real Effective Exchange Rates of RMB
0
50
100
150
200
250
300
350
400
450
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
NEER REER

Source: Financial Statistic Yearbook 2001, IMF, Washington.

Although we cannot directly use Thirlwalls assumptions, we can use econometric
methods to test if the variables in question significantly influence exports and
determine the feasibility of regarding exports as an exogenous variable. The
econometric model is derived from Thirlwalls export function. Data for the
worldwide GDP come from World Economic Outlook Database (2001) published by
the IMF; other data come from China Statistical Yearbook. The result is presented as
follows
5
:
2
ln( ) 2.06ln( ) 0.32ln( )
(3.703) ( 1.145)
0.6602 2.389 20
X WGDP REER
Adj R dw n
= + +

= = =

This result indicated that the real effective exchange rate (REER) is insignificant
(both statistically and economically). Therefore, it is safe to conclude that, in China,

5
In order to deal with the multicollinearity problem, we regressed on the difference of the data, which means that
no intercept estimate is reported in the result.
13
exports are significantly influenced by demand from the world market instead of by
fluctuations in exchange rates. The impact of changes in domestic and foreign prices
is also insignificant. Consequently, we can reasonably define exports as an exogenous
variable.
After combining the equations (5), (6), (7), (8) altogether, we have a multivariable
linear equation system. Through estimation of the structural form of the equations, we
can obtain estimates for the reduced form of the model:

0 1 1 2 3 4
0 1 1 2 3
0 1 1 2 3
0 1 1 2 3
(10)
t t t t t yt
t t t t t t ct
t t t t t t it
t t t t t t mt
Y C R X ER e
C C R X ER e
I C R X ER e
M C R X ER e

= + + + + +

= + + + + +

= + + + + +

= + + + + +

The contribution of exports to GDP can be obtained from the estimates of the
reduced form. Since
3
t
t
dY
dX
= and the direct impact of export increases on
economic growth is
1
1 1
t t
t t
X X
X Y

(or
1
t
t
X
Y

), the overall contribution of export increases

to economic growth is
1
3
1 1
t t
t t
X X
X Y

(or
3
1
t
t
X
Y

). The result also enables us to

obtain the impact of export increases on consumption, investment and imports. This is
one of the reasons that we estimate the structural equations instead of using the
reduced form.

III. Improved Estimations

We use the regression model as described in part II. Data used are for the years
1979 through 2000. The sources of the data are as follows: data for GDP, investment,
consumption, imports and exports are obtained from various editions of China
Statistical Yearbook and are converted to real values by the consumer price index;
interest rates are the loan rates for fixed asset investment with one year maturity; they
are taken from China Financial Yearbook, and are also converted to real interest rates
by the consumer price index. The exchange rate is the real effective exchange rate for
RMB estimated by the IMF.
Before estimating this linear simultaneous-equations model, we select from the
various estimation methods, including Ordinary Least Squares (OLS), Two-Stage
Least Squares (2SLS), Three-Stage Least Squares (3SLS) and Seemingly Unrelated
Regression (SUR). Generally speaking, when the sample size is small in the
estimation of a simultaneous-equations model, 2SLS is more advisable than 3SLS.
Results of some Monte Carlo experiments show that SUR estimators may not be more
efficient than OLS estimators when the sample size is small (see Gao, 1998, p512).
Since we have a data set that only covers the years 1979 through 2000, the sample
size is 22 = n , which is definitely a small sample size. We consequently adopted
14
the 2SLS method.
Firstly, we estimate equation system (10) using the 2SLS method (which will be
called Estimation I from now on). Table 3 presents the estimates of the structural
equations. It shows that all explanatory variables are significant; the F value and R
2
value for each equation are all satisfactory. The DW-test for the error terms in the
equations differs when a different estimation method is applied, but most DW
statistics are satisfactory
6
.

Table 3: Estimates of the Structural Equations

Note: Corresponding t Statistics in Parentheses.


With estimates from Estimation I in hand, we then can obtain the reduced-form
estimates. Since the method is similar, only the result of the first equation is
presented:
1
697.43 1.54 30.25 1.06 2.28 Y C R X ER

= + + +
which indicates that 1 06 . 1 > =
dX
dY
. This is contrary to our intuition; this result
may exaggerate the impact of exports on economic growth. Furthermore, the
coefficent estimate corresponding to consumption in the import function is negative,
which is inconsistent with economic theory. The multicollinearity problem might be

6
We have not made any corrections for serial correlation when the estimation failed to pass the DW-test. The
reasons will become clear as readers continue with this paper. In the following paragraphs, through comparisons
among estimates obtained from different methods, we will assess the impact of export increases on economic
growth. If the estimation could pass the DW-test when all other estimation methods are used, then, for the equation
that failed to pass the DW-test, the impact of serial correlation on the estimates can be figured out approximately.
Equations
Explanator
y Variables
Esimation I Esimation II Esimation III
C
-1
0.68 4.623 0.67 4.623 0.67 4.623
F=8852.67 F=8852.67 F=9601.6
R
2
=0.9990 R
2
=0.9990 R
2
=0.9990
R -53.72 -4.618-53.72 -4.618-53.72 -4.618
F=1674.3 F=1674.3 F=1674.2
R
2
=0.9949 R
2
=0.9949 R
2
=0.9949
C -0.32 -2.924 -- --
I 0.67 3.119 -- --
X 0.38 1.890 0.71 17.97 0.80 23.40
ER -1.33 -1.750-2.28 -3.13 --
F=241.53 F=412.69 F=547.6
R
2
=0.9847 R
2
=0.9798 R
2
=0.9682
0.41 56.05
Statistics
Statistics
Consumptio
n Function
Investmen
Function
Import
Function
Statistics
Y 0.41 56.05 0.41 56.05
Y 0.22 1.931 0.22 2.866 0.22 2.866
15
the source of this inconsistency. As a result, we analyzed the correlation coefficients
for explanatory variables C, I, X and ER (see Table 4). All the correlation coefficients
for C and I, C and X, as well as I and X exceed 0.9, which shows that there is serious
multicollinearity in the model.
The existence of multicollinearity suggests that the Estimation I estimates might
be incorrect. To find a more stable estimate, we have to deal with the multicollinearity
problem. Since the model we use is a linear simultaneous-equations model, the
remedy becomes too complicated. We therefore circumvent the issue. Instead of
coping directly with multicollinearity in the original model, we estimate another two
simultaneous-equations model. The only difference lies in the different specification
of the import function. They are as follows:

0 1 2
(11)
t t t t
M X ER = + + +
and
0 1
(12)
t t t t
M X = + +

Table 4. Correlation Coefficients Between Explanatory Variables in the Import
Function


Of course, if the structural equations are correctly specified, when the export
function is substituted by equation (11) or (12), the new estimates will be biased. (If
the export function is substituted by (11), we call the estimation from the new model
Estimation II and if substituted by (12), we call it Estimation III). However, it can
be predicted that Estimation II will overestimate the impact of exports on economic
growth, while Estimation III will underestimate the impact. The reasons are: (1)
according to the analysis above, the greater the impact of exports on imports, the less
the impact of exports on economic growth will be; (2) From economic theory, we
know that, in an export function, increases in consumption, investment and exports
will have positive impacts on imports, while changes in the real effective exchange
rate will have negative impact. If the positive impacts of consumption and investment
are excluded, the impact of these two variables might be attributed to exports and thus
overestimate exports impact on imports; similarly, if the negative impact of the
exchange rate is excluded, exports impact on imports might be underestimated.
Estimates for Estimation II and Estimation III are also presented in Table 3.
As in Estimation I, the reduced-form estimates can be inferred from the structural
equation estimates. For the same reason, only the estimates for the first equation of
the reduced form are listed below.
For Estimation II:
Correlation Coefficient C I X ER
C 1 0.9896 0.984 -0.7223
I 0.9896 1 0.9847 -0.7373
X 0.984 0.9847 1 -0.7074
ER -0.7223 -0.7373 -0.7074 1
16
1
2241.00 1.80 142.22 0.77 6.03 Y C R X ER

= + + +
For Estimation III:
1
784.28 1.80 142.22 0.54 Y C R X

= + +
The estimates are consistent with our previous judgments. The greater the impact
of exports on imports in the structural equation, the less the impact of exports on
economic growth in the reduced form will be. Among the estimates of the structural
equations, in Estimation III, the estimate for dM dX stabilizes at 0.8, which is quite
a high level. We are inclined to accept that the estimate for dY dX of the reduced
form in Estimation III represents the minimum value of the corresponding coefficient
across different estimations.
In light of the reduced-form estimates of Estimations I, II and III, and for the
reasons analyzed above, we conclude that the real value for dY dX is between 0.52
and 0.77. We might as well assume that 0.54 dY dX =
7
, which is a result from
Estimation III. The contribution rate of an increase in exports to economic growth can
be calculated in the way described above and the results are presented in Table 5.




















7
Correspondingly, 0.80 dM dX = . The average elasticity of imports with respect to exports can be obtained
with this figure. The computed result is 0.81, which is similar to the estimate for the elasticity in Table 2. This also
illustrates that the estimate of 0.54 dY dX = is very close to the minimum value of the estimate for the
corresponding variable.
17

Table 5
Improved Estimation of Contributions of Exports and Foreign Trade to GDP Growth
(Unit: %)

Year
GDP
Growth
Rate
Export
Growth
Rate
Exports
Contribu
tion to
GDP
Growth
Rate
Foreign
Trades
Contribu
tion to
GDP
Growth
Rate
Year
GDP
Growth
Rate
Export
Growth
Rate
Exports
Contribu
tion to
GDP
Growth
Rate
Foreign
Trades
Contribu
tion to
GDP
Growth
Rate
1981 5.32 32.56 1.05 1.19 1991 8.83 20.09 1.77 1.42
1982 12.11 12.69 0.51 1.93 1992 12.68 13.28 1.29 -0.94
1983 9.52 4.8 0.2 -0.63 1993 16.44 -1.35 -0.13 -3.44
1984 12.38 26.24 1.02 -0.21 1994 13.95 64.56 5.34 6.67
1985 11.39 26.48 1.16 -3.98 1995 9.74 5.57 0.67 0.92
1986 10.24 27.96 1.39 2.22 1996 10.25 -4.65 -0.53 0.33
1987 10.66 29.26 1.69 3.62 1997 8.76 19.62 1.95 3.17
1988 11.26 7.16 0.48 -0.95 1998 8.1 2.95 0.32 0.55
1989 2.88 1.72 0.11 -0.07 1999 6.72 8.51 0.89 -0.43
1990 5.32 44.5 2.85 5.75 2000 8.56 26.59 2.81 1.48

With the data in Table 5, we can calculate the elasticity of GDP to export growth
in corresponding years. In the 1990s, the value is about 0.1. In other words, for every
10% increase in exports, the GDP will increase by 1 percent. The corresponding value
for the 1980s is comparatively lower. The increasing tendency of exports as a
proportion to GDP might be the reason for this phenomenon (see Figure 3). In the
early 1980s, the proportion of exports to GDP was lower than 10% (which was only
5.1% in 1979), whereas in the late 1990s, it stabilized around 20% (for instance, it
was 22.9% in 2000). As the proportion of exports in GDP increases, the impact of
exports on economic growth will surely become more and more significant. This also
proves that increases in exports are of great significance to Chinas economy.
Finally, in order to compare these results with the results from traditional methods,
we redefined the contribution of foreign trade to economic growth. The new
definition includes both the direct and indirect impacts of exports on economic growth:
the contribution of export on economic growth is exports direct contribution to
economic growth plus its indirect contribution via the induced increase in
consumption and investment, i.e.,
( ) { } 1 t t t
dC dI
X NE Y
dX dX

+ + . Estimation of
the structural equations provides us with corresponding estimates for
dX
dC
and
dX
dI
. Calculation of the contribution rate of foreign trade thus becomes much more
18
convenient.
The contribution of foreign trade is also included in Table 5. Estimates obtained
from the new method show a larger contribution of foreign trade to GDP growth than
those obtained from traditional method in years that export increased because the new
method includes the indirect impact of exports on GDP growth via consumption and
investment. For instance, in 1999 and 2000, the contribution of foreign trade to
economic growth is negative when the traditional method is utilized. In those two
years, imports increased by a large amount and the increase in imports exceeded that
of exports. In comparison with 1998, exports increased by 8.51% while imports
increased by 20.89%. When the new estimation method is utilized, although its
contribution is still negative (-0.43%), its negative effect is much smaller than that
from the traditional method (-0.95%). As for the year 2000, the contribution of foreign
trade is 1.48% when the new method is used, whereas it is -0.19% using the
traditional method. In 2000 although imports increased more than 30%, exports
increased by about 27%. In light of exports impact on consumption and investment,
the positive contribution of foreign trade to economic growth for 2000 is credible.

Figure 3: The Proportion of Export to GDP (%)













Source: China Statistical Yearbook 2001


IV. Conclusion

The traditional method used to estimate the contribution of foreign trade to GDP
growth does not distinguish between the different roles of imports and exports in
economic growth. Moreover, it fails to account for the interactions among various
economic variables and thus tends to underestimate the contribution of foreign trade.
The key of our estimation improvement is our emphasis on exports indirect impact
via its influences on consumption and investment. Only when this indirect impact of
exports is included will it be possible for us to have a correct general view of the
contribution of foreign trade to economic growth in China. According to our
0.00
5.00
10.00
15.00
20.00
25.00
1
9
7
9
1
9
8
0
1
9
8
1
1
9
8
2
1
9
8
3
1
9
8
4
1
9
8
5
1
9
8
6
1
9
8
7
1
9
8
8
1
9
8
9
1
9
9
0
1
9
9
1
1
9
9
2
1
9
9
3
1
9
9
4
1
9
9
5
1
9
9
6
1
9
9
7
1
9
9
8
1
9
9
9
2
0
0
0
19
estimation, a 10% increase in exports will lead to a 1% increase of GDP on average.
Discussions in this paper are not based on mathematical wizardry. Different views
may have different implications for economic policy. We argue that to emphasize the
role of domestic demand does not conflict with the fact that foreign trade is
important for Chinas economic growth. Furthermore, in order to adjust the industrial
structure in China so that it becomes consistent with the comparative advantages of
Chinas economy, development of foreign trade is a factor that is of crucial
importance.
With our improved estimation method, the impact of exports on economic growth
is determined through econometric methods. We acknowledge that the specification
of the model and the selection of the estimation method might seriously affect the
results. Clearly, further improvement in these two areas is expected.

20

References

Atesoglu (1994), An application of a Kaldorian export-led model of growth to the
United States, Applied Economics, 26, 479-483.
Dixon, R. and Thirwall A. P. (1975), A model of regional growth-rate difference on
Kaldorian lines, Oxford Economic Papers, July, 1975
Feder, G.(1983), On export and economic growth, Journal of Development
Economics 12, 59-73.
Foreign Trade Policy Project Team (1999), Contribution of Foreign Trade to
Economic Growth and the Basic Orientation in Adjusting Foreign Trade Support
Policy, Finance and Trade Economics, June 1999.
Grosman, G. and E. Helpman. (1990),Comparative advantage and long-run growth,
American Economic Review 80(4): 796-815.
Gujarati, D. N. (1995), Basic Economitrics, McGraw-Hill, Inc
Gao, H. et al, editors, (1998), SAS/ETS Software: Users Guide, China Statistics Press
Kaldor, N. (1972), Capitalism and industrial development: Some lessons from
Britains experience, in Kaldor, N., Further essays on applied economics,
Holmes & Meler Publishers, Inc, 1978.
McCombie G. S. L. (1985), Economic growth, the Harrod foreign trade multiplier
and Hicks super-multiplier, Applied Economics, 17, 55-72.
McCombie, J. and A. Thirlwall. 1994. Economic Growth and the Balance of
Payments constraint. London, St. Martins.
McCombie, J. and A. Thirlwall. 1997. The dynamic Harrod Foreign Trade Multiplier
and the Demand oriented Approach to Economic Growth: an evaluation.
International Journal of Applied Economics,11(1): 5-26.
McCombie, J. and A. Thirlwall. 1999. Growth in an international context: a Post
Keynesian view. In: Deprez, J. and J. Harvey. Foundations of International
Economics: Post Keynesian Perspectives. London, Routledge.
Peng, Fuwei (1999), The Role of Foreign Trade in the Chinas Economy , Journal
of International Trade, Jan. 1999. Vol. 193.
Somez(1996), A demand-oriented explanation of economic growth in Germany,
Journal of Post-Keynsian Economics, Vol. 19, (1), 101.
Thirlwall, A. P.(1979), The balance of payment constraint as an explanation of
international growth rate difference, Banca Nazionale del Lavoro Quarterly
Review, 1979
Thirlwall, A. P.(1980), Regional problem are balance of payments problems,
Regional Studies, 5, 1980
Thirlwall, A. P.(1987), Keynes, economic development and the developing
countries, in A. P. Thirlwall (ed.), Keynes and economic development, London:
Macmillan, 1987
Wang, Z. (1998), Rethinking the Role of Foreign Trade in the Chinas Economy,
Intertrade, Oct. 1998, Vol.202.
Zhang, X. and J. Hu (1999), Behind Free Trade: Import and Chinas Economic
21
Development, Intertrade, April 1998. Vol. 208.
Zhu, W. (1998), The Puzzle of Chinas Export-Oriented Development Strategy,
Strategy and Management, May 1998. Vol. 30.