The Impact of Bank and Non-Bank Financial Institutions on Local Economic Growth in China* and

Transcription

1 The Impact of Bank and Non-Bank Financial Institutions on Local Economic Growth in China* by Xiaoqiang Cheng Katholieke Universiteit Leuven Department of Economics Naamsestraat 69 BE-3000 Leuven Belgium and Hans Degryse CentER - Tilburg University and CESIfo P.O. Box NL-5000 LE Tilburg The Netherlands h.degryse@uvt.nl * The authors thank seminar participants of the LICOS-Monetary Economics workshop in Leuven, and Lijan Sun, Vincenzo Verardi, Ellen Vanassche, and Patrick Van Cayseele for useful comments. Financial assistance from FWO-Flanders, NWO-The Netherlands, and the Research Council of the University of Leuven is gratefully acknowledged.

2 The Impact of Bank and Non-Bank Financial Institutions on Local Economic Growth in China Abstract This paper shows that finance spurs growth, even in a country where the economy shows a high growth rate, such as China. Employing data of 27 Chinese provinces over the period , we study whether two different types of institutions banks and non-bank financial institutions have (a significantly different) impact on local economic growth. Our findings show that banks outperform non-bank financial institutions. Only bank loans exert a statistically and economically significant positive impact on local economic growth. This effect becomes more pronounced when the banking sector is less concentrated. Key Words: growth, financial development, Chinese provinces, banks JEL-codes: E44, G21 1

3 I. Introduction In recent years, a growing literature has explored the role of finance for economic growth. Cross-country studies provided convincing evidence. For example, King and Levine (1993a) employ data on 77 countries over the period , to document that the level of financial development determines long-run economic growth, capital accumulation, and productivity growth. Levine and Zervos (1998) refine this and find that initial stock market liquidity and banking development are both positively correlated with future rates of economic and productivity growth in a sample of 42 countries over the period Most cross-country studies however do not include China, one of the most important developing countries in the world, when addressing the finance-growth nexus. China maintained during the last two decades an average real growth rate of about 9 percent per year. Our paper aims to contribute to this finance-growth literature by looking at the impact of bank and non-bank financial institutions on the growth rate of Chinese provinces over the period Our motivation to distinguish banks and non-bank financial institutions is that they differ in their geographical scope, size and organization, and efficiency. Identifying the effects of financial development on economic growth is a challenging task. The difficulties stem from the following two aspects (see e.g. Rajan and Zingales (1998)). First, omitted variables that drive both financial development and growth may imply biased results. Second, financial development may react to the expectation of enhanced future economic growth; hence finance may lead rather than cause growth. Jayaratne and Strahan (1996) tackle the first problem by keeping effects other than financial development constant. They use financial deregulation in the early 1970s in 35 U.S.-states as an exogenous shock to local financial development. They find that in the 30 years after the deregulation, the economy grew faster in the deregulated states than in the other states. They also test the hypothesis of deregulation happening only due to expectation about the future needs of financing. They reject this hypothesis by observing that the loans after deregulation did not explode. Therefore, they attribute the relatively faster economic growth in those deregulated states to the improvements in loan quality. Following a similar difference in difference approach, Rajan and Zingales (1998) examine the differential effect of financial development across industries within a country. Based on the theoretical predictions that financial markets and institutions help firms or industries to overcome problems of moral hazard and adverse selection, they test whether industries that are more dependent on external financing grow relatively faster in countries that are more financially developed. Their findings are consistent with the hypothesis that a more developed financial sector spurs relatively faster growth of capital-intensive industries, and provide the smoking gun for the debate on causality. Other research focuses on finding proper instruments to extract the exogenous part of financial development when trying to settle the issue of causality. La Porta et al (1998) 2

4 link the legal origin of a country to its financial development. Their empirical results suggest that a variety of legal origins (British, French, German or Scandinavian laws) differing in protecting the rights of both shareholders and creditors and in the efficiency of legal enforcement, reasonably lead to different levels of financial development. Based upon the above legal origin-finance instruments, a substantial body of empirical work further testifies that financial development promotes economic growth in cross-country, industry and firm level analysis (see e.g. Levine, Loayza, and Beck, 2000 and Demirgüç-Kunt and Maksimovic, 1998). Except for instruments such as legal origins, economists also rely on improved econometric techniques to instrument endogenous variables. Authors employ the dynamic system GMM panel estimator proposed by Arellano and Bover (1995), to extract the impact of financial development on economic growth by controlling for potential endogeneity. Which identification strategy do we follow? As our data cover 27 provinces within China, we are able to more adequately control for institutional, legal and cultural factors that may commonly affect the entire Chinese financial system (see Carbo Valverde et al (2005) for a similar application to Spain). Moreover, we focus on the effects of two different types of financial institutions banks and non-bank institutions on provincial growth. Thus we examine whether differently developed financial institutions contribute differently to per capita GDP growth within a province. The rationale for this approach stems from the following reasoning. Theory argues that financial institutions efficiently allocate capital to where it can generate better returns and therefore promote growth. Based on recent balance sheet data as well as reported data on non-performing loans (NPLs), we find that the Chinese banks perform better than the Chinese non-bank institutions. This may stem from banks having a higher developed human capital stock and being more technologically advanced. Typically, banks also have a larger geographical scope than non-bank financial institutions. One could argue that if finance passively follows rather than causes growth, those differences between banks and non-banks financial institutions in enhancing economic growth are unimportant. Then we would observe a similar correlation 1 between the financial institutions development and economic growth. However, if finance in China spurs growth, then endogeneity is still a problem. To deal with this, we employ the dynamic system GMM estimator proposed by Arellano and Bover (1995) to control for such endogeneity. Banks in our study include the five biggest commercial banks in China: the four biggest state-owned commercial banks and one national commercial bank 2 (Bank of Communications). Non-bank financial institutions mainly include rural credit cooperatives, and local trust and investment companies. Interesting is that those five banks, and especially the four state-owned banks are known for their prudence in granting loans to small private companies (Allen et al., 2005 and Boyreau-Debray, 1 If finance simply follows rather than causes growth, we should interpret the relationship, if any, between finance and growth as correlation rather than causation (impact). 2 The difference between state-owned banks and national commercial banks is illustrated in table a1 in Appendix. 3

5 2002). As a comparison, most of non-bank institution loans have been extended to the non-state-owned sector (Xie, 1998). We study the impact of bank loans extended by these two types of financial institutions on the provincial real per capita GDP growth, using a dataset covering 27 provinces over the period Our findings highlight that the size of bank loans shows a significantly positive impact on local growth, both statistically and economically. As a comparison, non-bank financial institutions, while granting most of their loans to the non-state-owned sector, seem to be less important for local growth. How to reconcile these results with Allen et al. (2005), who argue that growth in China mainly stems from the private sector? One potential explanation is that banks may enjoy a better pool of borrowers as they have a larger geographical scope, face fewer restrictions in attracting deposits and therefore can establish stronger bank-firm relationships, and finance companies with almost all sizes. Non-bank financial institutions may have a restricted choice due to their smaller nature. We further find that the positive impact of bank loans is more pronounced in provinces exhibiting a less concentrated banking sector. Of course we are not the first to deal with the finance-growth nexus in China; however, no conclusive results have been reached yet. One strand of papers argues that financial development matters for economic growth in China. For example, Li and Liu (2001) show that growth of provincial aggregate output is positively and significantly related to the growth of banking sector loans and self-raised funds, employing data from 25 provinces for the period They attribute the positive correlation to the improvement in the efficiency of capital reallocation during the liberalization in both financial and real sectors. However, the authors do not employ instrumental variables and therefore do not control for potential endogeneity. Another strand of papers holds the opinion that China is a counterexample to the current findings in the finance and growth literature. More specifically, they question whether financial development plays an important role in economic growth in China, as they observe the coexistence of its weak legal and financial systems and fast economic growth. Allen et al. (2005) examine closely the relationship between law, finance and growth in China. Their analysis suggests that the relatively poor legal system and the underdeveloped financial sector contribute little to the growth of the private sector, which is known as the most important component of the fast growth of the Chinese economy. Hence, Allen et al. (2005) argue that there exist other financing channels for the private sector than those of financial institutions. We focus on the most recent time period that incorporates some financial deregulation. We find that the financial development of banks contributes to growth, suggesting that not only reputation and relationships are important. The rest of our paper is organized as follows. Section II introduces the Chinese financial system and its development indicators. Section III presents the theoretical background our study is based on as well as the empirical framework we employ. The data description and empirical results are presented in section IV. The last section concludes. 4

6 II. The Chinese financial system and its development indicators 2.1. The Chinese financial system We first provide a brief description of the Chinese financial structure. 3 We explicitly focus on the differently developed financial institutions in China banks and non-bank financial institutions, rather than the stock markets. The Chinese financial system is dominated by financial institutions, especially banks. 4 [Insert Figure 1 here] Figure 1 presents the structure of the Chinese financial institutions at the end of It shows that financial institutions in China can clearly be divided into two categories: banks and non-bank financial institutions. In the banking sector (labeled as banks ), there are three policy banks focusing on policy-oriented loans and fifteen commercial banks, out of which four state-owned banks dominate the whole banking sector. Among the eleven national and regional banks, Bank of Communications 5 is the biggest one, of which the largest shareholder is China s Finance Ministry. Sometimes researchers also loosely call the four state-owned banks and the Bank of Communications, the five biggest state-owned banks. The non-bank financial sector consists of urban and rural credit cooperatives, trust and investment companies, financial companies and other institutions. Banks are hierarchically organized while non-bank institutions are generally following a decentralized form. This hierarchical structure mainly stems from their size. For example, Industrial and Commercial Bank of China, the largest bank among the four state-owned banks, has 37,039 branches all over the country. As a comparison, there are in total 50,745 rural credit cooperatives in the whole country. Typically, a rural credit cooperative does not belong to any headquarter, is independent from other rural credit cooperatives, and is active in one province only. [Insert Figure 2 here] As Figure 2 describes, the total assets of four state-owned banks, which were approximately 7,122 billion RMB 6 at the end of 1994, cover around 78 percent of the total assets of the financial sector. The other banks are relatively smaller. As the fifth 3 Table a1 in Appendix introduces the functions of the main Chinese financial institutions. 4 For example, at the end of 1994, the ratio of the stock market capitalization to total assets of financial institutions was approximately 6.7%. Although the importance of stock markets has increased somewhat since the early 1990s, the scale and the importance of the financing channels of the stock markets are not comparable to those of financial institutions (Allen et al., 2005). In this paper, we assume that stock markets have no significantly different impacts on different provinces. Hence employing a fixed effects panel model and incorporating time dummy variable in our analysis may well control for the impact of stock markets. 5 The Bank of Communications has been publicly listed in Hong Kong Stock Exchange since June RMB=Renminbi (in 2000, 1 US $ = 8.3 RMB) 5

7 biggest bank in China, Bank of Communications occupied more than half of the total assets of all national and regional banks at the end of The total assets of non-bank financial institutions together take 16 percent of the assets of all financial institutions. The market share of rural credit cooperatives is 7 percent, which is comparable to that of trust and investment companies. [Insert Figure 3 here] Figure 3 shows the importance of the different financial institutions for State-owned banks still dominate but their market share declined towards 68 percent. National and regional banks gained market share towards 15 percent. Rural credit cooperatives increased whereas trust and investment companies decreased in market share. We argue that the two types of financial institutions banks and non-bank financial institutions differ in several dimensions. First, they show a diverging geographic scope. Banks are bigger players than non-bank financial institutions. Most banks in China are regional or national players, and some of them are even international players. A non-bank financial institution, in contrast, is typically only present in one province. Second, banks may be technological more advanced. Banks often pay higher salaries and offer better career opportunities to young graduates. Therefore, banks may attract higher quality personnel. Banks also benefit more easily from technological spillovers, as they often recruit experts having overseas working experience. Third, large banks branches may benefit from expert credit systems developed centrally. Although the hierarchical structure also has clear disadvantages and may imply a focus on hard information as argued by Stein (2002), banks in emerging countries may still benefit from such organizational structure as it helps in reducing asymmetric information problems. Isolated non-bank financial institutions are more likely to suffer from asymmetric information in the Chinese financial system where there is no third-party credit rating agency. Recent balance sheet data as well as reported data on non-performing loans (NPLs) show that banks perform better than non-bank financial institutions. [Insert Tables 1 and 2 here] This is illustrated in Tables 1 and 2. Table 1 shows the operating costs of different types of Chinese financial institutions. These numbers reveal that the ratio of operating costs to assets is lower for banks than for non-bank financial institutions. Table 2 provides some data on the non-performing loans (NPLs) in the Chinese financial sector. Although the average NPL ratio is high relative to other countries, banks have a lower NPL ratio than non-bank financial institutions. The numbers presented in Table 1 and 2 suggest that banks are more efficient than non-bank financial institutions. 6

8 Having a clear image of how bank-firm relationships are established in China may help us to understand the role of finance for growth. Two surveys discussed in the Appendix (survey a1 and a2) provide evidence on how firms may choose between different financial institutions. They show that firms apply first for credit at banks before turning to other sources of finance, such as loans from non-bank financial institutions Bank and non-bank financial development indicators Typical indicators of financial development employed in cross-country studies are Liquid Liabilities, Commercial-Central Bank and Private Credit (see e.g. Levine, Loayza and Beck (1999)). However, in our provincial-level study, none of the above-mentioned indicators is available. We therefore construct three financial development indicators at province level. We discuss these three indicators for banks and non-bank financial institutions, respectively. Indicators of financial development of banks Following Boyreau-Debray (2002), we construct Bank Size at province level. Bank Size equals the ratio of the savings in the banking system to local GDP. Bank size is a measure of financial depth of the local banking sector. In our study, we construct Bank Size for each of the 27 provinces in China. A second indicator is Bank Credit, which equals the credit extended by banks to local enterprises over local GDP. This indicator measures the financial resources provided by banks to provincial entities needed for economic growth. Finally, we construct a measure Bank Concentration, following Boyreau-Debray (2002). We compute Bank Concentration by the Herfindahl-Hirschman Index (HHI), employing bank market shares in the deposit market. We include this measure to proxy for the competitiveness of the banking sector. Before 1980, there were only 3 banks in China and each of them enjoyed a different segment of the deposit market. After 1984, the number of banks in the market increased and banks began to compete for deposits under the permission of the central government. It has been argued that HHI may show little correlation with the degree of competition (see e.g. Cetorelli (1999)). A preferred indicator for banking competition would be the Rosse-Panzar (1987) statistic computed for individual provinces. However, local bank level data, such as details of balance sheet of individual banks, prevent us from computing the Rosse-Panzar statistic. Indicators of financial development of non-bank financial institutions 7 According to the two surveys, at the end of 2002, banks in China seemed more likely to discriminate the borrowers in their sizes rather than their ownership. This may stem from the fact that the Chinese banks are used to extent loans on the basis of collateral. Hence in the starting-up period, without sufficient collateral, small companies are less likely to be financed by banks. This situation improves once the company becomes larger. The survey results do not necessarily contradict the findings of Allen et al (2005), given that most small firms are young and private-owned in China. 7

9 In a similar fashion as for the bank indicators, we construct Non-bank Size, Non-bank Credit and Non-bank Concentration for non-bank financial institutions. Descriptive statistics on all the development indicators are discussed in Section IV. III. Theoretical background and empirical framework 3.1 Theoretical background A rich literature has studied the relationship between finance and growth. There are two schools of thought, which hold contrasting views on the importance of finance to growth. On the one hand, the idea that finance contributes little to growth on both the rate of investment in physical capital and changes in investment is mainly based on the result of Solow s (1956, 1957) analysis. On the other hand, other theorists argue that financial development plays a key role for growth by observing the potential channels through which financial intermediation provides external financing and monitors the behavior of entrepreneurs. In their view, differences in the quantity and quality of services provided by financial institutions partly explain why countries grow at different rates (Goldsmith, 1969, Mckinnon, 1973, and Shaw, 1973). Most of the studies emphasize the effect of financial constraints and/or financial intermediation on long-term growth. For example, Greenwood (1990), Levine (1991), and Sussman (1993) analyze the impact of financial intermediation on growth in an AK-style model without making a distinction between investing in technology and investing in physical or human capital accumulation. Recent studies, such as King and Levine (1993b), Galetovic (1996), Blackburn and Hung (1998), and Aghion, Howitt and Mayer-Foulkers (2003) examine the relationship between finance and growth in the context of innovation-based background. For instance, Aghion, Howitt and Mayer-Foulkers (2003) show why the existence of technology transfers is not sufficient to put all countries on parallel long-run growth rate paths. They find that that it is not just financial constraints that make some countries poor but rather that financial constraints inhibit a technological transfer and thus lead to an ever-increasing technology gap. Following King and Levine (1993b), we illustrate briefly how financial development affects technology innovation and hence possibly influences the long-run growth rate. Their endogenous growth model focuses on the connections between finance, entrepreneurship and economic growth. Financial institutions in this model play an important role in both monitoring and financing potential entrepreneurs in the initiation of innovative activities and launching of new products to the market. [Insert Figure 4 here] 8

10 Figure 4 shows the channels through which financial intermediation contribute to economic growth. Initially, in the entrepreneurial selection procedure, the financial intermediary monitors the whole set of candidates in the market and picks up the potential entrepreneurs with the ability to manage innovations in the intermediate goods production technology. Second, the financial intermediary finances the innovative activities. If the entrepreneurs are successful, they will enjoy the monopoly profit by producing the unique intermediate product at a lower cost than their rivals but charging the same price. However, to produce intermediate goods the successful entrepreneurs need external financing. The financial intermediary hence evaluates and finances again those entrepreneurs while it can pay back the consumers (savers) the interest according to its evaluation of the profitability of those entrepreneurs. Requiring the input of intermediate goods and labor, the production of final goods is also affected by the innovative success -- the productivity increases with the technological progress. Of course, the final goods production in aggregate influences the consumers, who also provide the labor in this model, by affecting their optimal choice of intertemporal substitution in consumption. Again, as most neo-classic models predict, the intertemporal substitution elasticity and time preferences of labor together with real return rates (interest rates in this model), are positively correlated with the aggregate growth. Moreover, the equilibrium conditions of the model show that the growth rate is not only affected by productivity, which is partly decided by the probability that a candidate is a potential entrepreneur, but also negatively impacted by the cost of monitoring. The model reveals the following potential relationship between finance and growth. First, finance supports innovations and hence increases the productivity which is positively correlated with growth. Second, increases of the efficiency of financial sector, such as a decrease in the cost of monitoring, will increase the real rate of return and thus lead to a higher growth rate in the future. Third, the model also suggests a reverse channel of causation where distortions in the innovative sector lower the demand for financial services and retard financial development. 3.2 Empirical framework To estimate the potential impact of financial development on economic growth, consider a Cobb-Douglas production function at the individual level, α y = k x, (1) where y equals real per capita GDP, k equals real per capita physical capital stock, x equals other determinants of per capita growth, and parameter. Taking the logarithm of (1) yields, is a production function ln y = α ln k + ln x. (2) As most neo-classical R&D models predict, for example King and Levine (1993), the 9

11 growth of x comes from technological innovation. First-difference of (2) reads, GYP = α ( GK) + PROD Where GYP is the growth rate of real per capita GDP, GK is the growth rate of real per capita capital stock and PROD is the growth rate of everything else. If we assume that the hours worked per worker are relatively stable in our sample range, PROD should provide a reasonable conglomerate indicator of technology growth. If there is any key relationship between technology growth and financial development, for instance, efficiency, the contemporaneous impact of finance on growth hence can be estimated by, GPY = a + a GK + a FI + ε 3 t o 1 t 2 t t where FI t is the financial development indicator at time t. For an empirical application of equation (3) to China s local province growth, we base our estimation on panel data from different provinces over the period The advantage of using panel data is that we can estimate the corresponding relationship even in a relatively short period. The fixed effects model derived from equation (3) controlling for time effects can be written as GPY = a GK + a FI + a CON + δ U + φv + ε i, t 1 i, t 2 i, t 3 i, t i i t t i, t i= 1 t= 1 I T where FI i,t is the financial development indicator of either banks or non-bank financial institutions in province i at time t. U i is a set of province dummy variables, V t is the set of time dummy variables, and δ i and φ t are the vectors of coefficients. CON refers to the conditioning informational set, which contains FDI, which equals the ratio of Foreign Direct Investment to GDP, and Investment, which is the ratio of total investment to GDP. These two variables in our study provide an approximate measure of other potential financing resources that are relevant for local economic growth. Panel models are the most prevailing econometric approach in recent studies of cross-country or within-country finance and growth issues. For instance, King and Levine (1993a) use a within model to study predetermined financial development and growth. Li and Liu (2001) employ both within and random-effects models to investigate the relationship between financial development and growth in China. Boyreau-Debray (2002) uses a dynamic panel model to study the impact of finance on growth using provincial data in China. In order to reveal the relationship between financial development and future economic growth, we introduce the lagged financial development indicators in our panel regression, 10

12 I T 8 i, t = 1 i, t + 2 i, t i, t + δi i + φt t + εi, t i= 1 t= 1 GPY a GK a FI a CON U V Here equation (4) can be estimated by OLS in general, assuming that the lagged FI is exogenous and there is no heteroskedasticity and serial autocorrelation in error term. However, problems pop up when those assumptions are violated. For example, heteroskedasticity or serial autocorrelation in the error term is often observed in panel analysis. Fortunately, recent econometricians have solved this problem by introducing robust standard errors or by first differencing the data. In our analysis, heteroskedasticity is detected. We report the results of regression (4) employing robust standard errors. (4) Equation (4) reveals the correlation between finance and growth, without controlling for the potential endogeneity problem. However, as argued before, by comparing the correlations between the loans granted by the two types of financial institutions and local economic growth, we can identify whether in China finance simply follows growth or not. Once our identification shows that finance matters for growth in China, we will further explore to what extent financial development promotes economic growth. In doing so, we have to deal with the endogeneity of financial development which is well documented in the finance and growth literature (see for example, Goldsmith (1969) and King and Levine (1993b)). Introducing lagged FI in regression (4) cannot fully solve the problem. It is reasonable to argue that financial development can react to the expectation of future economic growth. The most prevailing way to control for the endogeneity problem is to introduce instrumental variables. In this paper, we try to analyze the impact of finance on growth by employing the dynamic system GMM estimator developed by Arellano and Bover (1995). The dynamic panel model requires the lagged dependent variable to enter to right-hand side of the regression. For example, regression (4) can be extended to a dynamic panel regression as follows, GPY = a GPY + a GK + a FI + a CON + δ U + φv + ε i, t 0 i, t 1 1 i, t 2 i, t 1 3 i, t i i t t i, t i= 1 t= 1 First differences of (5) read, I T (5) GPY GPY = a ( GPY GPY ) + a ( GK GK ) i, t i, t 1 0 i, t 1 i, t 2 1 i, t i, t 1 + a ( FI FI ) + a ( CON CON ) 2 i, t 1 i, t 2 3 i, t i, t 1 + ( ε ε ) i, t i, t 1 (6) 8 Here we control for the contemporaneous effects of conditioning variables, such as FDI and Investment, following the traditional finance and growth literature (see e.g. King and Levine (1993a)). As a robustness test, we also model the finance and growth relationship by controlling for the lagged value of conditioning variables, as conventional growth theory suggests. Our results remain robust. 11

13 A system estimator jointly estimates the regression in levels (5) and the regression in differences (6). In order to correct for endogeneity, Arellano and Bover (1995) suggest employing the lagged first differences of the explanatory variables as instruments for the equation in levels (5) and the lagged values of the explanatory variables in levels as instruments for the equation in differences (6). The crucial assumptions therefore are that the lagged differences of variables are good instruments for explaining subsequent levels and the lagged levels of variables are good instruments for explaining subsequent first differences. Rejection of the Sargan test of over-identifying restrictions at 5% level questions the validity of those instruments. Moreover, we test whether the error term of regression (6), ε i, t ε, is i, t 1 second-order serially autocorrelated. Accepting the null hypothesis of no second-order serial autocorrelation supports the assumption of the moment condition of (6). IV. Data Description and Empirical Results 4.1. Data Description Our dataset contains annual growth rates of real per capita GDP, real per capita capital stock, FDI and investment ratios in 27 provinces of China over the period Lagged financial development indicators are also included in our dataset from 1994 to The financial development indicators in our study are calculated employing the statistics data reported by Almanac of China s Finance and Banking. The Almanac documents the provincial data of annual savings and loans of 5 banks only: 4 state-owned banks and the Bank of Communications, the biggest bank of the national commercial banks. At the end of 1994, those 5 banks represent approximately 96 percent of the total assets of the banking sector. From 1994 onwards, Almanac of China s Finance and Banking reports the provincial data of savings and loans of rural credit cooperatives and of some selected trust and investment companies, financial companies, and other non-bank financial institutions. 10 Only the non-bank financial institutions that are considered to be large enough have their data reported in the Almanac, whereas smaller institutions remain uncovered. This may introduce a sample selection bias in that provinces with many small institutions may have an underestimated size of the non-banking sector. However such selection bias should be taken care of by our province dummies in as far the reporting bias remains constant over our sample period within a province. 9 Data reasons prevent us to include three provinces (Hubei, Tibet and Hainan). 10 The data of urban credit cooperatives are also reported but not for every year. We therefore decided to exclude urban credit cooperatives from our sample. 12

14 We construct the financial development indicators of non-bank institutions from the annual provincial data of rural credit cooperatives and other reported non-bank financial institutions. The computation of the non-bank concentration based on the aggregate data of Almanac of China s Finance and Banking also induces some problems. While rural credit cooperatives, like other non-bank financial institutions, are isolated from each other, Almanac of China s Finance and Banking reports the province level aggregate for all rural credit cooperatives jointly in every province. However in reality rural credit cooperatives are not integrated into one entity. We will deal with this measurement error by using instruments. [Insert Table 3 here] Table 3 provides the summary statistics of our data. We present the time averages for the 27 provinces. There exists a wide range of values across provinces. As Table 3 illustrates, the highest average annual real per capita GDP growth rate equals 10.2 percent (Zhejiang province). The lowest real per capita GDP average growth rate equals 5.7 percent (Yunnan province). Similarly, Beijing on average has the highest values of both Bank Size and Bank Credit, while Shandong province on average has the lowest levels of Bank Size and Bank Credit. Non-bank financial institutions suffer the poorest development in Qinghai province, while Shanxi on average has the highest Non-bank Size and Guangdong enjoys average the highest Non-bank Credit. The financial development indicators for China are relatively high compared to those for other countries (see also Allen et al. (2005)). Both Bank Size and Bank Credit outweigh those of non-bank financial institutions. For example, the average ratios of Bank Size and Bank Credit -- measured as bank savings and bank loans to GDP respectively -- across provinces are and 0.683, while the average ratios of non-bank savings and loans to GDP across provinces are and only. 4.2 Empirical results [Insert Table 4 here] Table 4 presents the results of different versions of equation (4). The left panel (4a,b and c) introduces the results including the bank financial development indicators in the regression. Bank Size and Bank Credit are significantly positively correlated with future economic growth. The middle panel (4d,e and f) presents the results where non-bank financial development indicators enter the regression. Only Non-bank Size is positively correlated with growth. While most of the Non-bank Credit is granted to the non-state-owned sector, those loans exhibit little correlation with future growth. The right panel (4g and h) shows the results for the regressions where both bank and non-bank indicators enter the specification. Including both Bank Credit and Non-bank Credit into one regression (4h) confirms the robustness of the results. Both Bank Size and Non-bank Size are significantly positively correlated with future growth (column 13

15 4g), which is in line with most finance and growth literature. However, neither Bank Concentration nor Non-bank Concentration predicts future growth. One possible explanation is that the HHI-measure of banking concentration is not adequate to capture the degree of competition (see e.g. Cetorelli (1999)). Table 4 also shows that the per capita capital stock growth is not statistically significant. An explanation is that it is much easier for people to move across provinces (or states) within a country than to move across countries. Hence an empirical application using local data of a country may suffer from the problem that the provincial population is quite unstable over time. Therefore, the insignificance of the coefficient of per capita capital stock growth may be due to the fluctuation of local population. We hence test whether the growth of aggregate capital stock is correlated with aggregate economic growth. [Insert Table 5 here] Table 5 presents the results of regressing provincial aggregate GDP growth on aggregate capital stock growth and financial development indicators. Aggregate capital stock growth seems to be significantly positively correlated with aggregate GDP growth, after the population fluctuation has been removed. Bank Credit continues to be positively correlated with future growth, which indicates robustness of the previous results. However, Bank Size now turns insignificant whereas Non-bank Size becomes significant in (5g). Tables 4 and 5 show a significant different correlation of the financial development indicators of banks and non-bank financial institutions with growth. In particular, Bank Credit is highly correlated with real per capita GDP growth even though bank loans are more focused on the state-owned sector. 11 As a comparison, although non-bank loans are mostly extended to the non-state-owned sector, Non-bank Credit is largely irrelevant in explaining growth. This remarkable difference between banks and non-bank institutions suggests that the loans of the financial sector do not simply follow growth. The reasoning is that when finance would simply follow growth, both Bank Credit and Non-bank Credit should show high correlations with GDP growth. Moreover, in that case, Non-bank Credit should exhibit a higher correlation with GDP growth as the non-state-owned sector is the most important contributor to the growth of China. Our results contradict the arguments related to finance simply following growth. As banks compared to non-bank financial institutions have a wider geographical scope, are technologically more advanced, and may have been more affected by deregulatory financial reforms 12, we may identify the causation between 11 This situation may have improved recently. Survey 1a in Appendix shows that around half of the bank loans were extended to the non-state-owned sector at the end of Table a2 in the Appendix provides an overview of the financial reforms in China. For example, the series of reforms in 1994 such as separating policy lending from commercial banks and liberalizing the interest rates within bounds are known as deregulations. Moreover, that PBOC abolished the credit plan requirement for commercial banks in 1998 is also a good example of deregulation in the Chinese banking sector. Some other events, such as opening of the Chinese financial sector in 2006, trigger the Chinese government to improve the efficiency of 14

16 finance and growth. So by looking at bank and non-bank financial institutions, we follow a similar difference in difference approach as employed by Jayaratne and Strahan (1996). In order to further control for the endogeneity of financial development, we adapt the dynamic system GMM estimator as mentioned in the previous section. [Insert Table 6 here] Table 6 reports the impact of financial development on economic growth when using the dynamic system GMM estimator. Bank loans significantly spur future economic growth, both economically and statistically. For example, if Shandong, the province now receiving the least bank credit enjoyed as much bank credit as Beijing, where the most bank credit is extended, ceteris paribus, Shandong s growth rate would increase approximately 8 percent per year, which is huge. Column 6h displays the results when we include Bank Credit and Non-bank Credit in one regression. Again, the impact of Bank Credit appears to be positive and significant. Bank Size does not show any significant impact. The impact of Non-bank Size appears to be different in column 6d and column 6g, questioning the robustness of this variable. The coefficients of Bank Concentration and Non-bank Concentration are not significant. The fact that the null hypotheses of both the Sargan test and the second-order serial autocorrelation tests cannot be rejected at the 5 percent level approves the validity of the results of dynamic panel regressions 13. Neither FDI nor Investment in our study shows a significant impact on growth 14. This result may stem from the fact that those two variables in our dataset do not show enough variation over time and hence do not show a statistical significance in our panel regressions. In order to extract more information from Bank Concentration, we introduce the interaction between Bank Concentration and Bank Credit (Bank Concentration x Bank Credit) as a new regressor in the panel regression. [Insert Table 7 here] Column 7a in Table 7 reports the results when the interaction term Bank Concentration x Bank Credit is included. When Bank Credit still appears a significantly positive impact on future growth, the interaction term shows a strongly negative impact. If Bank Concentration captures the degree of banking competition, a higher Bank Concentration should negatively impact future growth. The significantly banks. 13 The null of the Sargan test of the regression reported in column 6f cannot be rejected at 5% but can still be rejected at 10%. 14 Next to FDI and Investment, we also control for other variables, such as Schooling, that might matter for growth. Our results are robust to the inclusion of this variable. However, Schooling in our data only is available for 6 years. Therefore we don t report the results in detail in this paper. 15

17 negative coefficient of the interaction term indeed implies that Bank Concentration matters. Furthermore, in columns 7b and 7c we report the results for two sub-samples containing the above and below median level of Bank Concentration, respectively. The impact of Bank Credit in the provinces where bank assets are less concentrated is much more pronounced 15. V. Conclusion Is the finance-growth nexus at work in an economy exhibiting a high growth rate? In this paper we address this question by providing empirical evidence on the impact of financial development on the local growth of Chinese provinces over the recent period Exploiting within variation in China by employing data on 27 provinces, we are able to more adequately control for institutional, legal and cultural factors that may commonly affect the Chinese financial system. We find that the finance-growth nexus also applies to the economic growth of Chinese provinces. But what type of financial institutions contributes to the Chinese finance-growth nexus? We look at the impact of two types banks and non-bank financial institutions. The reasoning is that banks, relative to non-bank financial institutions, have a wider geographical scope, have a larger size, and are often more hierarchically organized. Bank branches are also well integrated and may benefit from centrally developed technology and expert credit systems. We find that the amount of bank loans exerts a statistically and economically significant positive impact on local economic growth 16. In contrast, even though non-bank financial institutions focus more on the non-state-owned sector, we still find that the amount of non-bank financial institution loans shows little contribution to local growth. We also find that bank loans in provinces with a less concentrated banking sector exert a more pronounced impact on local growth, showing that competition pronounces the finance-growth nexus. In general, our findings challenge the view that China is a significant counterexample to the current findings in the finance-growth literature. Our focus on a recent time period and the difference between banks and non-bank financial institutions shows that the finance-growth nexus also applies to the economic growth of Chinese 15 The sub-sample of high concentration contains the seven most developed provinces and six of the poorest provinces in China. This implies that the concentration of the banking sector in individual provinces in our sample seems not to be driven by the expectation of future economic growth. 16 As the two surveys in Appendix document, the Chinese banks seem to be less prone to grant loans when firms are in the starting-up stage but become the most important loan providers once the firms survive and become larger. Hence the Chinese banks may discriminate the two different channels modeled by King and Levine (1993b), as illustrated in Figure 4. Specifically, banks are more likely to drop the channel of financing potential entrepreneurs but adopt the channel of evaluating and financing intermediate goods monopoly. 16

18 provinces. 17

19 References Aghion, P., Howitt, P., and Mayer-Foulkers, D., 2003, The Effect of Financial Development on Convergence: Theory and Evidence, Quarterly Journal of Economics, 120: Allen, F., Qian, J., and Qian, M., 2005, Law, Finance and Economic Growth in China, Journal of Financial Economics, 77: Arellano, M., and Bover, O., 1995, Another Look at the Instrumental-Variable Estimation of Error-Components Models, Journal of Econometrics, 68: Bekaert, G., and Harvey, C. R., Chronology of Economic, Political and Financial Events in Emerging Markets: China. Blackburn, K., and Hung, V. T. Y., 1998, A Theory of Growth, Financial Development and Trade, Economica 65: Boyreau-Debray, G., 2003, Financial Intermediation and Growth: Chinese Style, World Bank Policy Research Working Paper, Carbó Valverde,S., López del Paso, R., and Rodríguez Fernández, F.,2004, Banks, Financial Innovations and Regional Growth, European Integration, Financial Systems and Corporate Performance (EIFC) Working Paper Cetorelli, N.,1999, Competitive Analysis in Banking. An Appraisal of the Methodologies. Economic Perspectives, First Quarter: Demirgüç-Kunt, A., and Maksimovic, V., 1998, Law, Finance and Firm Growth, Journal of Finance, 53: Galetovic, A., 1996, Specialization, Intermediation, and Growth, Journal of Monetary Economics, 38: Goldsmith, R. W., 1969, Financial Structure and Development, New Haven, CT: Yale University Press. Greenwood, J., and Jovanovic, B., 1990, Financial Development, Growth, and the Distributionof Income, Journal of Political Economy, 98: Jayaratne, J., and Strahan, P.E., 1996, The Finance-Growth Nexus: Evidence from Bank Branch Deregulation, Quarterly Journal Of Economics, 111: King, R.G., and Levine, R., 1993a, Finance and Growth: Schumpeter Might Be Right, Quarterly Journal of Economics, 108: King, R.G., and Levine, R., 1993b, Finance, Entrepreneurship, and Growth: Theory and Evidence, Journal of Monetary Economics, 32: La Porta, R., Lopez-de-Silanes, F., and Shleifer, A., and Vishny, R. W., 1998, Law and Finance, Journal of Political Economy, 106: Levine, R., 1991, Stock Markets, Growth, and Tax Policy, Journal of Finance, 46: Levine, R., 2004, Finance and Growth: Theory and Evidence, NBER Working Papers, Levine, R., Loayza, N., and Beck, T., 2000, Financial Intermediation and Growth: Causality and Causes, Journal of Monetary Economics, 46: Levine, R., and Zervos, S., 1998, Stock Markets, Banks, and Economic Growth, American Economic Review, 88: Li, K., and Liu, T., 2001, Impact of Liberalization of Financial Resources in China's Economic Growth: Evidences from Provinces, Journal of Asian Economics, 12: Miller, M. H., 1988, Financial Markets and Economic Growth, Journal of Applied Corporate Finance, 11,

20 McKinnon, R.I., 1973, Money and Capital in Economic Development, Washington, DC: Brookings Institution. Panzar, J. C., & Rosse, J. N., 1987, Testing for "Monopoly" Equilibrium, Journal of Industrial Economics, 35: Rajan, R. G., and Zingales, L., 1998, Financial Dependence and Growth, American Economic Review, 88: Schumpeter, J. A., 1912, Theories der Wirtschaftlichen Entwicklung. Leipzig: Dunker & Humblot, (The Theory of Economic Development, 1912, translated by R. Opie, 1934, Cambridge, MA: Harvard University Press). Shaw, E.S., 1973, Financial Deepening in Economic Development, New York, NY. Oxford University Press. Solow, R.M., A Contribution to the Theory of Economic Growth, Quarterly Journal of Economics, 70: Solow, R.M., 1957, Technological Change and the Aggregate Production Function, Review of Economics and Statistics, 39: Stein, J.C., 2002, Information Production and Capital Allocation: Decentralized vs. Hierarchical Firms, Journal of Finance, 57: Sun, L., 2003, The Fragility in China s Financial Systems, Finance and Trade Economics (in Chinese), 268: Sussman, O., 1993, A Theory of Financial Development, In: Finance and Development: Issues and Experience, edited by A. Giovannini, Cambridge, UK: Cambridge University Press. Xie, P., 1998, Zhongguo Fei Yinhang Jinrong Jigou Yanjiu ( A Study of the Chinese Non-bank Financial Institutions), Economics and Finance, 1998 (3), (4), and (5). Xie, P., 2001, Zhongguo nongcun xinyongshe tizhi gaige de zhenglun (The system reform of China s rural credit cooperatives), Financial Research, 2001(1). Zhang, Q., 2003, Zhongguo Nongcun Jinrong Xianzhang Yu Zhengce Fengxi (Analysis of the Chinese Rural Financial System), Report for Asian Development Bank,