Proceedings of the 15 th Annual Conference of the Association for Chinese Economics Australia (ACESA) Development of Financial Intermediation and Economic Growth Chinese Experience CHEN Hao 1 CERDI, Université d’Auvergne 65, boulevard François-Mitterrand 63000 Clermont-Ferrand, France Tel: 0033 660189655 E-mail: [email protected]Abstract Using Chinese provincial data from 1985 to 1998 and applying recent GMM techniques developed for dynamic panels, this paper examines how the development of financial intermediation influences China’s economic growth during the post-1978 reform period. Our econometric results show that China’s financial intermediation development contributes to its rapid economic growth through two channels: the substitution of loans for state budget appropriation and the mobilization of households savings, but not through loan expansion since loan distribution by financial intermediaries is inefficient. Deep financial sector reform aiming at correcting this inefficiency is desirable, and is expected to sustain China’s economic development in the future. Key Words: financial intermediation; financial development; economic growth; China. 1 Correspondence to: CHEN Hao CERDI, Université d’Auvergne 65, boulevard François-Mitterrand 63000 Clermont-Ferrand, FRANCE E-mail: [email protected]Tel: 0033 660189655 Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 1 -
29
Embed
Development of Financial Intermediation and Economic ...mams.rmit.edu.au/vau5brg8hx2sz.pdf · of financial intermediation influences China’s economic growth ... most case studies
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
By construction, (Yi,t-1 - Yi,t-2) and (εi,t - εi,t-1) are correlated. OLS estimation of
equation 2 will not give an unbiased and consistent estimate of β. Hence, we must
find valid instruments for (Yi,t-1 - Yi,t-2).
Assuming that (a) the error terms are not serially correlated,
E[εi,t εi,s] = 0 for i = 1, …, N and s ≠ t
and that (b) the initial conditions Yi,1 are predetermined,
E[Yi1 εi,t] = 0 for i = 1, …, N and t ≥ 2
Arellano and Bond (1991) proposes the following moment restrictions
E[Yi,t-s (εi,t - εi,t-1)] = 0 for t = 3, …, T and s ≥ 2
Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 12 -
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
(ACESA)
Since the values of Yi,t lagged two periods or more are correlated with (Yi,t-1 - Yi,t-2),
but not with (εi,t - εi,t-1), they are valid instruments for equation 2.
With regard to the explanatory variables Xi,t, there are three possible situations:
If the explanatory variables Xi,t are strictly exogenous (i.e., the explanatory variables
are assumed to be uncorrelated with all past, present and future values of the error
term), then all the past, present and future values of Xi,t are valid instruments for
equation 2.
If the explanatory variables Xi,t are predetermined (i.e., the explanatory variables are
assumed to be correlated with past values of the error term, but uncorrelated with
current and future values of the error term), then the values of Xi,t lagged one period
or more are valid instruments for equation 2.
If the explanatory variables Xi,t are endogenous (i.e., the explanatory variables are
assumed to be correlated with past and present values of the error term, but
uncorrelated with future values of the error term), then the values of Xi,t lagged two
periods or more are valid instruments for equation 2.
However, Blundell and Bond (1998) argues that when the lagged dependent and the
explanatory variables are persistent over time, lagged values of these variables are
only weak instruments for the first-differenced equation. And the first-differenced
GMM estimator is expected to have a large finite sample bias and poor precision in
simulation studies. Blundell and Bond (2000) confirms this statement by showing that
in the case of weak instruments, the first-differenced GMM estimator will be biased
towards the Within groups estimator. To reduce the potential biases and imprecision,
Arellano and Bover (1995) and Blundell and Bond (1998) suggest estimating a system
that combines the set of equations in first-differences (equation 2) with the additional
set of equations in levels (equation 1). For the regression in differences, the
instruments are the same as above. For the regression in levels, the instruments are the
suitably lagged differences of corresponding variables. Assuming that (a) the
differences of the explanatory variables are uncorrelated with the individual-specific
effects,
Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 13 -
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
(ACESA)
E[∆Xi,t ηi] = 0 for i = 1, …, N and t = 2, …,T
and (b) ∆Yi2 are uncorrelated with the individual-specific effects,
E[∆Yi2 ηi] = 0 for i = 1, …, N,
then, if the Xi,t are strictly exogenous or predetermined, ∆Yi,t-1 and ∆Xi,t are valid
instruments for the levels equations; if the Xi,t are endogenous, ∆Yi,t-1 and ∆Xi,t-1 are
valid instruments for the levels equations.
The consistency of the GMM-System estimator depends on the validity of the
assumption of no serial correlation of the error term, and on the validity of the
instruments, This can be tested by two specification tests proposed by Arellano and
Bond (1991), Arellano and Bover (1995) and Blundell and Bond (1998). One is a
Sargen test of over-identifying restrictions, which can test the overall validity of the
instruments. Another is the m2 statistic, which tests the presence of second-order
serial correlation in the first-differenced error term. Failure to reject the null
hypotheses of both tests provides evidence to suggest that the no serial correlation
assumption and the instruments are valid.
B. Indicators of the development of Financial intermediation
We use three indicators to measure the three aspects of the development of financial
intermediation: (i) the ratio of state banking sector’s loans outstanding relative to
GDP (bank)6; (ii) the ratio of households savings deposits in financial intermediaries
relative to GDP (savings), and (iii) the share of fixed asset investment financed by
domestic loans relative to that financed by state budgetary appropriation
(loan/budget). Following the analysis of Section 2, savings and loan/budget are
expected to enter growth regressions positively and significantly, while bank not
significantly.
6 In China, the statistics concerning credit to the private sector are not available. Moreover, at the
provincial level, total loans outstanding of financial system is available on a consistent basis only after
1989. However, at the national level, the state banking sector accounts for more than 75% of total loans
outstanding of financial system during the post-1978 reform period. So we use the aggregate lending of
state banking sector as the indicator showing the loan expansion aspect of financial intermediation
development.
Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 14 -
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
(ACESA)
C. Data and Model
The panel consists of data for 28 Chinese provinces over the period 1985-19987. The
data on education are drawn from Démurger (2001), all other data come from the
China Statistical Yearbook (various years), the Comprehensive Statistical Data and
Materials on 50 Years of New China (1999), the Almanac of China Finance and
Banking (various years) and the China Regional Economy: A Profile of 17 Years of
Reform and Opening-up (1996).
To assess the impact of the development of financial intermediation on economic
growth, we introduce the financial variables into the traditional growth regression
framework. Our analysis consists in estimating the following growth equation:
Yi,t = α + βYi,t-1 + γXi,t + δFi,t + ηi + εi,t
(3)
where Y is the logarithm of real per capita GDP, X is the set of traditional growth
determinants (investment, population growth, education and infrastructure), F is the
indicators of the development of financial intermediation (bank, savings and
loan/budget), η is the unobserved province-specific effect, ε is the error term, and the
subscripts i and t represent province and time respectively.
Regarding the set of control variables, we introduce the ratio of fixed asset investment
to GDP as a proxy for physical capital (investment), the share of population with at
least secondary schooling as a proxy for human capital (education), the density of
roads as a proxy for infrastructure (infrastructure) and the annual population growth
rate (population growth). All these control variables are assumed weakly exogenous8.
Besides, all financial variables—bank, savings, and loan/budget—are assumed to be
endogenous, since some theorists argue that the relationship between finance and
growth is reciprocal: finance favors growth and growth in turn spurs financial
development9. Hence we must control for the endogeneity of financial variables to
avoid potential biases induced by simultaneity.
7 Due to the data unavailability, Tibet and Hainan are excluded from the sample. 8 The empirical results are similar when these control variables are assumed strictly exogenous. 9 See Greenwood and Smith (1997).
Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 15 -
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
(ACESA)
Table 1 presents descriptive statistics and correlations for the dependent variable and
financial variables. All these variables exhibit a large variation. Savings and
loan/budget are positively and significantly, but bank is negatively and significantly
correlated with the growth rate. Savings is positively and significantly correlated with
bank and loan/budget, and bank and loan/budget are negatively and insignificantly
correlated.
Table 1: Descriptive Statistics and Correlations
Descriptive Statistics
Economic growth Bank Savings Loan/budget
Mean 0.06 0.83 0.43 4.31
Maximum 0.31 1.63 1.14 19.48
Minimum -0.11 0.38 0.12 0.36
Std. Dev. 0.06 0.22 0.18 3.57
Observations 406 403 403 346
Table 1 (continued)
Correlations
Economic growth Bank Savings Loan/budget
Economic growth 1.000
Bank -0.188 1.000
(0.000)
Savings 0.103 0.476
(0.042) (0.000)
Loan/budget 0.273 -0.074 0.423 1.000
(0.000) (0.177) (0.000)
p-values are reported in parentheses
Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 16 -
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
(ACESA)
D. Results
Arellano and Bond (1991) and Blundell and Bond (1998) argue that, although a two-
step estimator is more efficient than a one-step estimator, Monte Carlo studies show
that the efficiency gain is small while the asymptotic errors associated with the two-
step estimators may be seriously biased downwards. Thus asymptotic inference from
one-step standard errors may be more reliable. We therefore report the one-step
parameter estimates for GMM-System estimator (Table 2 and Table 3).
In order to identify the global impact of the development of financial intermediation
on economic growth, we exclude at first the variable investment from the regression.
Table 2 reports the results. In the columns 1, 2 and 3, we introduce respectively the
three variables of financial intermediation. As we expect, bank does not enter the
growth regression significantly, while the coefficients on the other two financial
variables, savings and loan/budget, are positive and strongly significant. Since we
have controlled for the endogeneity of these three variables, the results suggest that
the development of financial intermediation has a causal and positive impact on
growth through the channels of the mobilization of households savings and the
substitution of loans for state budget appropriation. Furthermore, this impact is
economically large. For example, Inner Mongolia province’s value of loan/budget
over the period 1985-1998 was 1.77, while the mean value for the whole country was
4.30. Therefore if exogenous factors had pushed Inner Mongolia province’s value of
loan/budget to the country’s mean, Inner Mongolia province would have witnessed its
annual growth rate increased by 2.13 percentage points. With regard to other
variables, a lower population growth rate and a more developed infrastructure favor
economic growth, which confirms our expectations. However, the coefficient on the
human capital variable (education) is not significant. Finally, turning to the test
statistics, neither the Sargen test nor the m2 statistics provide evidence that rejects the
validity of the instruments and the no serial correlation assumption.
Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 17 -
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
(ACESA)
Table 2: Development of Financial Intermediation and Economic Growth:
Without Investment (GMM-System, dependent variable: logarithm of
real GDP per capita)
1 2 3
Constant 0.131 0.417 0.409
(0.742) (0.078) (0.196)
Initial GDP per capita 0.899*** 0.926*** 0.852***
(0.000) (0.000) (0.000)
Population growth -0.042*** -0.035*** -0.042***
(0.000) (0.000) (0.000)
Education 0.143 -0.001 0.104
(0.086) (0.988) (0.132)
Infrastructure 0.097*** 0.015 0.109**
(0.004) (0.734) (0.020)
Bank 0.023
(0.328)
Savings 0.061***
(0.004)
Loan/Budget 0.024***
(0.000)
Sargen Test 1.000 1.000 1.000
M2 0.116 0.130 0.097
Observations 333 333 287
Provinces 28 28 27
Note: In the regression, the right-hand-side variables are included as log(variable);
p-values in parentheses, ** (***) indicates statistical significance at the 5 (1)
percent level;
For the regressions including the variables loan/budget, Fujian province is
excluded from the sample due to missing data.
Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 18 -
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
(ACESA)
To separate the productivity effects of financial intermediation from its investment
effects, we now introduce the variable investment into the regression. Table 3 shows
the results. Investment has a positive and significant coefficient, which supports the
theoretical prediction that physical capital formation contributes to economic growth.
Bank is still insignificant while the coefficients on savings and loan/budget remain
significant and decline only slightly, from 0.061 to 0.051 and from 0.024 to 0.018,
respectively. It appears that the impact of financial intermediation on growth runs
through its impact on investment and productivity, but mainly through the latter. With
regard to other variables, the results are similar to those of Table 2, except that
education enters two of three regressions significantly (see columns 1 and 3). The
human capital spurs growth through its impact on total factor productivity, but may
hamper physical capital formation since education demands economic resources and
reduces resources available for investments in physical projects. As a result, its global
effects are ambiguous as shown in Table 2. Finally, both the Sargen test and the m2
statistics give support to our model.
Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 19 -
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
(ACESA)
Table 3: Development of Financial Intermediation and Economic Growth: With
Investment (GMM-System, dependent variable: logarithm of real GDP
per capita)
1 2 3
Constant 0.927*** 0.772*** 0.856***
(0.004) (0.006) (0.007)
Initial GDP per capita 0.835*** 0.884*** 0.804***
(0.000) (0.000) (0.000)
Investment 0.075*** 0.067*** 0.074***
(0.001) (0.005) (0.003)
Population growth -0.032*** -0.029*** -0.039***
(0.000) (0.000) (0.000)
Education 0.283*** 0.073 0.184**
(0.000) (0.488) (0.019)
Infrastructure 0.086** 0.031 0.120**
(0.023) (0.545) (0.016)
Bank -0.035
(0.177)
Savings 0.051**
(0.036)
Loan/Budget 0.018***
(0.003)
Sargen Test 1.000 1.000 1.000
M2 0.112 0.082 0.077
Observations 333 333 287
Provinces 28 28 27
Note: In the regression, the right-hand-side variables are included as log(variable);
p-values in parentheses, ** (***) indicates statistical significance at the 5 (1)
percent level;
For the regressions including the variables loan/budget, Fujian province is
excluded from the sample due to missing data.
Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 20 -
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
(ACESA)
For comparative purposes, we present the results using OLS levels estimator and
Within groups estimator in Table 4 and Table 5. In comparison with Table 2 and
Table 3, the main difference consists in the coefficient on the lagged dependent
variable. It seems that the OLS levels estimator gives an estimate biased upwards
while the Within groups estimator is biased downwards, which conforms with the
theoretical arguments of Hsiao (1986) and Nickell (1981). The GMM-System
estimate of this coefficient lies comfortably above the corresponding Within Groups
estimate, and below the corresponding OLS levels estimate, which can be regarded as
a signal that the GMM-System estimator is probably preferable. Moreover, the use of
OLS levels estimator and Within groups estimator makes the variable bank enter the
regressions with a negative coefficient, which is significant in seven of eight
regressions. In contrast, the GMM-System estimator always gives the variable bank
an insignificant coefficient. As we have shown in Section 2, the central government
considers financial intermediation as a means to tax rich and dynamic regions and to
subsidize poor and stagnant regions, which may lead to an artificially imposed
causality from economic development to financial intermediation. It seems that OLS
levels estimator and Within groups estimator suffer from the bias induced by the
endogeneity of the variable bank, while the GMM-System estimator manages to avoid
this bias. Finally, with regard to the other financial intermediation variables, savings
and loan/budget always has a positive and strongly significant coefficient. The use of
alternative estimators does not change our conclusion concerning the role of financial
intermediation in the process of economic growth in China.
Chen, H., ‘Development of Financial Intermediation and Economic Growth: Chinese Experience’. - 21 -
Proceedings of the 15th Annual Conference of the Association for Chinese Economics Australia
(ACESA)
Table 4: Development of Financial Intermediation and Economic Growth:
Without Investment (OLS and Within)
OLS OLS OLS Within Within Within
Constant -0.154 -0.217** -0.036
(0.203) (0.027) (0.630)
Initial GDP per capita 1.006*** 1.011*** 0.995*** 0.839*** 0.890*** 0.827***