计量经济学作业Microsoft Word 文档

Financial development and economic growth:Evidence from ChinaQi LIANG a,b,*, Jian-Zhou TENG c,daDepartment of Finance, School of Economics, Nankai University, Tianjin, 300071, ChinabGraduate School of Commerce and Management, Hitotsubashi University, Kunitachi, 186-8601 Tokyo, JapancGraduate School of Economics, Hitotsubashi University, Kunitachi, 186-8601 Tokyo, JapandSchool of Mathematics and Statistics, Northeast Normal University, Changchun, 130024, China Received 17 April 2005; accepted 26 September 2005China Economic Review 17 (2006) 395–411AbstractThis paper investigates the relationship between financial development and economic growth for the case of China over the period 1952–2001. After considering the time series characteristics of the dataset, a multivariate vector autoregressive (VAR) framework is used as an appropriate specification and the long-run relationship among financial development, growth and other key growth factors is analyzed in a theoretically based high dimensional system by identification of co-integrating vectors through tests of over-identifyingrestrictions. The empirical results suggest that there exists a unidirectional causality from economic growth to financial development, conclusions departing distinctively from those in the previous studies.D 2005 Elsevier Inc. All rights reserved.JEL classification: C32; O11; G28Keywords: Financial development; Economic growth; Multivariate V AR; Causality; China1. IntroductionThe economic growth of China is remarkable since the outset of the reform program in early 1980s. A great number of theoretical and empirical studies have explored the sources of economic growth at both national and provincial levels (e.g., Borensztein & Ostry, 1996; Chen& Feng, 2000; Chow, 1993; Chow & Li, 2002; Wu, 2000; Yu, 1998), and ongoing debate is mainly concerned with which source, factor accumulation or productivity improvement, is the key growth-driving factor. However, unfortunately, the role of financial development in economic growth has till recently often been ignored, with a conspicuous lack of studies being made to theoretically examine and empirically determine this. The effects of the financial sector on real economy can hardly be over-emphasized, as Goldsmith (1969, p. 390) puts bone of the most important problems in the field of finance. . . is the effect that financial structure and development have on economic growthOn one hand, it is difficult to investigate various aspects of the finance-growth nexus since simply examining the correlations among them, which are utilized in most cross-country studies,can lead to spurious estimations due to a number of limitations inherent in the cross-sectional technique.1Besides, it is well known that correlations reveal nothing about causation. On the other hand,the majorities of the existing time-series studies applying only bivariate causality tests between indicators of financial development and growth variables (e.g., Bell & Rousseau, 2001; Caldero´n & Liu, 2003; Demetriades & Hussein, 1996)2 also suffered from the omitted variables bias and could lead to erroneous causal inferences, since any causality test between financial development and economic growth which excludes other decisive growth determinants from the system and analyzes only a financial development indicator and an output variable, is very likely to be mis- specified, and little reliance could be placed on the results of such studies. There are, to the best knowledge of authors, only a few studies using multivariate causality test in the examination of finance-growth nexus. For example, Luintel and Khan (1999) propose an approach through which long-run relationship between financial development and economic growth is evaluated in a theoretically based multivariate V AR model—a framework in which the analysis in the present paper relies heavily. By identifying the finance-growth nexus in a co-integrating framework through tests of over-identifying restrictions, Luintel and Khan find bi-directional causality between financial development and economic growth in all the sample countries. Christopoulos and Tsionas (2004) also investigate the long-run relationship between financial development and economic growth in a multivariate V AR framework, though the examination is carried out via panel unit root tests and panel co-integration analysis in a panel-based vector error correction model (VECM). For all the developing countries in their sample, Christopoulos and Tsionas find unidirectional causality from financial depth to growth. These two reviewed studies are subject to limitations such that the capital formation is the only additional growth-determining variable incorporated in the framework. Moreover, stock-flow problem inherent in the indicator measurement of financial development is not finely dealt with. Additionally, focusing exclusively on one country instead of a number of nations is advantageous in that the econometric findings can be related to the prevailing institutional structure (Bell & Rousseau, 2001), as Chandavarkar (1992, p. 134) argues that the relationship between finance and growth bmerits systematic testing on a country wide basis over sufficiently long periodsQ.This paper aims to examine the relationship between financial development and economic growth for the case of China over the period 1952–2001. Specifically, what kind of role has financial sector played to the economic growth process? What is the nature and direction of the1 Quah (1993) focuses on the lack of balanced growth paths across countries; Pesaran and Smith (1995) explore theheterogeneity of slope coefficients; Demetriades and Hussein (1996) dwell on the model misspecification resulted fromomitted variable bias, sample selection bias and inappropriate weighting of countries; while Arestis and Demetriades(1997) and Luintel and Khan (1999) emphasize the problem of causality.2 Bell and Rousseau (2001) examine the finance-growth nexus in a tri-variate VAR framework using macroeconomic,monetary and financial indicators, their specification, however, could hardly be free of omitted variable bias since M1, ameasurement of monetary, is also a proxy indicator of financial development.----------------------- Page 3-----------------------Q. Liang, J.-Z. Teng / China Economic Review 17 (2006) 395–411 397 relationship between financial development and economic growth? What kind of effect, positive or negative, has financial development exerted on economic growth? What type of causality, unidirectional or bi-directional, exists in finance-growth nexus? We attempt to answer these questions empirically and try to shed some light on the roles of financial development as well as other conditional variables in determination of economic growth. After considering the time series characteristics of our dataset and several confirmed sources-of-China’s-growth such as capital stock and international trade, a theoretically based multivariate VAR framework is usedas an appropriate specification and whether proxy measurement of financial development is associated with long-run economic growth is identified in a co-integrating framework through tests of over-identifying restrictions. We find that there exists in China over the period 1952–2001 a unidirectional causality from economic growth to financial development, results that depart distinctively from those in the existing literatures.The rest of the paper is organized as follows. Section 2 gives a brief review of existing literature on the various aspects of the finance-growth nexus. Section 3 describes system specification, indicator measure and descriptive statistics in this work. Section 4 discusses the methodological issues in the multivariate V AR framework. Section 5 presents the empirical results and relevant discussions. Section 6 concludes the whole paper.2. Review of literatureEver since the pioneering contributions of Schumpeter (1911), and more recently of Patrick (1966), Goldsmith (1969), MacKinnon (1973) and Shaw (1973), the relationship betweenfinancial development and economic growth remains an important subject in economicliterature. Different aspects of this relationship have been analyzed extensively in boththeoretical and empirical studies for single-country and cross-nations as well as at industry and firm levels through cross-sectional, time-series and dynamic panel techniques.Schumpeter (1911) points out the role of financial intermediaries in mobilizing funds, evaluating and selecting projects, managing risk, monitoring entrepreneurs and facilitatingtransactions as the critical elements in fostering technical innovation and economic growth. Under the assumption that the size of a financial system is positively correlated with the supply and quality of financial services, Goldsmith (1969) documents a positive correlation between the financial development and economic performance in his 35-country sample. MacKinnon (1973) and Shaw (1973) propose that government intervention on the development of financial systems is impediment in the process of economic growth. In this paradigm, financial development is seen as exerting positive effects on economic growth. By contrast, a few economists hold skeptical view on the decisive role played by financial development in the economic growth process. Robinson (1952, p. 52) questions this one-way causality, declaring that bby and large, it seems to be the case that where enterprise leads finance followsQ. Lucas (1988) believes that economists badly over- stress the significance of finance. Chandavarkar (1992, p. 134) notes bnone of the pioneers of development economics. . .even list finance as a factor of developmentQ. In this paradigm,financeis of little importance and only responds passively to economic growth. To the contrary, Lewis (1955) implies that initially economic growth facilitates the formation of financial markets and then matured financial markets promote economic growth, thus postulating a bi-directional relationship between financial development and economic growth. Similar to Lewis (1955), Patrick (1966) concludes that a supply-leading relationship exists in the early stage of economic development as causation runs from financial development to economic growth while a demand- following relationship prevails in the later stage as the direction of causation is reversed.----------------------- Page 4-----------------------398 Q. Liang, J.-Z. Teng / China Economic Review 17 (2006) 395–411 Obviously, there is no general consensus among economists on the relationship between financial development and economic growth. Therefore, as a practical way to resolve thecontroversy, empirical study is called for. A large body of empirical studies, mainly usingcross-sectional approaches, has found that the level of financial development is a goodpredictor of economic growth (e.g., Beck, Levine, & Loayza, 2000; Gregorio & Guidotti, 1995; King & Levine, 1993; Levine, 2002). A number of recent literatures on endogenous growth also favor the positive role of financial intermediaries played to economic growth (e.g., Amable & Chatelain, 2001; Bencivenga & Smith, 1991; Bencivenga, Smith, & Starr, 1995; Benhabib & Spiegel, 2000). These researchers support the point of views that financialdevelopment may raise savings rate, stimulate investment, avoid premature liquidations ofcapital, reduce the cost of external finance, enhance the efficiency of capital allocation and insure more productive technological choices, all factors that in turn lead to high economic growth. Alternatively, various models of joint determination of real and financial sectors (e.g., Deidda & Fattouh, 2002; Odedokun, 1996; Rioja & Valev, 2004) present a non-linearrelationship between financial development and economic growth, indicating financialdevelopment is not associated with higher growth rates at all levels of economic development. Although the existence of a positive relationship between financial development and economic growth even after allowing for other growth determinants has been recognized, the empirical findings do not settle the issue of causality. Contrary to the cross-section studies, time-series approaches employing the V AR framework offer an opportunity to analyze the causalitypattern and its evolution over the time between indicators of financial development andeconomic growth. Most of the time-series studies find either unidirectional causality fromfinance to growth (e.g., Bell & Rousseau, 2001; Christopoulos & Tsionas, 2004; Fase &Abma, 2003) or bi-directional causality (e.g., Caldero´n & Liu, 2003; Demetriades & Hussein,1996; Luintel & Khan, 1999).3 These studies also demonstrate that causality patterns varyacross countries and, therefore, highlight the dangers of statistical inference based on cross- section country studies. Likewise, numerous endogenous growth models also support a bi- directional relationship between financial development and economic growth (e.g., Berthelemy& Varoudakis, 1996; Greenwood & Jovanovic, 1990; Greenwood & Smith, 1997).In general, theoretical models and empirical analyses have provided conflicting predictions and implications about both the impacts of financial development on economic growth and the repercussions of overall financial development on economic performance.Since theextraordinary economic achievement in China offers a great opportunity to test theSchumpeterian hypothesis empirically, we believe that the present study could be instructiveand complementary to the existing literature on finance-growth nexus.3. System specification, indicators measurements and descriptive statisticsThis section describes: (i) the system specification which makes up the theoretical premise in the multivariate VAR framework utilized to identify the co-integrating vectors among variables and to test the causality between financial development and economic growth; (ii) measures of economic growth, financial development and other decisive growth determinants; (iii)descriptive statistics and correlations among variables.3 Demetriades and Hussein (1996) find evidence that economic growth systematically causes financial development inquite a few sample countries.----------------------- Page 5-----------------------Q. Liang, J.-Z. Teng / China Economic Review 17 (2006) 395–4113993.1. System specificationFollowing the standard literature and improving upon the theoretical postulate of Luintel and Khan (1999), we specify economic growth relationship as:Y ¼f ðK ;R;Z Þð1Þwhere Y is the real per capita GDP; K is the real per capita physical capital stock; R is the real interest rate (deflated by inflation), whose positive coefficient might capture indirectly theproductivity effects on Y; Z is the vector of other decisive growth determinants. In the present study, the only element of the vector Z is TR (trade ratio).4 Likewise, we specify a financial development relationship as:FD ¼ g ðY ;RÞð2Þwhere FD is the indicator of financial development, Y and R are the same as explainedabove. It used to be common in the literature to employ some measures of the money stockover GDP as a proxy for the financial development. This proxy, however, poses significant problems of interpretation because monetary aggregates: (i) measure more the extent ofmonetization rather than of financial development, especially for the developing economies; (ii) make no differentiation of liabilities among financial institutions; (iii) cannot represent the actual volume of funds channeled to the productive sector (e.g., Demetriades& Hussein,1996; Gregorio & Guidotti, 1995; Luintel & Khan, 1999). Therefore, under the hypothesisthat the size of financial intermediaries is positively related to the provision and quality of financial services, we utilize Bank Credit Ratio (BCR) as the indicatorof financialdevelopment, which equals the values of domestic credit by banking institutions divided by GDP.5 Besides BCR, we also consider in the robustness test Deposit Liabilities Ratio (DLR), a traditional measure of financial development that equals the ratio of total deposit liabilities of banking institutions to GDP. Although theory suggests that income (GDP) exerts a positiveeffect on financial development (thus, g V is positive), it provides an ambiguous answer to theYeffects of R . Conventional analysis shows that the effect of R on savings depends on the relative strength of income and substitution effects. However, the repressionists school represented by MacKinnon (1973) and Shaw (1973) propose that R is positively correlated with savings indeveloping countries, asserting that the positive substitution effect dominate the negative income effect.4 Many researchers include international trade (or export) as a factor in their studies onfinance-growth nexus (e.g.,Beck, 2002; Beck & Levine, 2004; Odedokun, 1996). Additionally, the significant contribution of international trade toChina’s economic growth is an indisputable fact. Thus, we have concluded that it is indispensable to include internationaltrade in our multivariate V AR framework to investigate a robust relationship between financial development andeconomic growth.5 It is better to use Private Credit (utilized in e.g., Beck et al., 2000; Levine. Loayza, & Beck, 2000) as an indicator toproxy the financial development, which restricts the credit by financial intermediaries to the private sector. However, it ishard, if not impossible, to construct it in a long time series data sense for a transition economy like China. Additionally,Beck et al. (2000) point out that it would be valuable to incorporate measures of securities market development inempirical research on the relationship between financial development and economic growth. However, stock marketsstarted in China only in 1990, if we simply add measures of banking institutions and stock market together, it might bringabout structural break problem to the whole indicator series of financial development. Moreover, although stock marketsin China have been growing very fast, their scale and importance are still not comparable to those----------------------- Page 6-----------------------400 Q. Liang, J.-Z. Teng / China Economic Review 17 (2006) 395–4113.2. Indicators measurementsBefore describing the indicators measurements, we present model data sources. Our data are extracted from Comprehensive Statistical Data and Materials on 50 Years of New China,China’s National Income, 1952–1995, and recent issues of China Statistical Yearbook (2003, 2004) and Almanac of China’s Finance and Banking (2003).Following the standard practice (e.g., Luintel & Khan, 1999; Levine et. al., 2000), our indicator for economic growth is LNY, the natural logarithm of the real per capita GDP.6 The real per capita GDP is measured as a ratio of real GDP to total population and real GDP is measured as nominal GDP divided by GDP deflator (1952=100). Since all the data sources have not provided the GDP deflator, we have to estimate it, through a newly constructed implicit deflator, which is calculated by both the current value of GDP and GDP index.As mentioned above, financial development is proxied by either BCR or DLR.7 We address the stock-flow problem inherent in these two indicators. Following Levine et al. (2000), we deflate the end-of-year items of the balance sheet of banking institutions by their corresponding end-of- year CPI and deflating the GDP series by the annual CPI, then we average the real banking institution balance sheet items in year t and t 1 and divide this average by real GDP measuredin year t. Since each data source has not provided complete CPI series covering the period 1952–2001, we estimate new CPI series (1952=100) using information from all of them.Besides measures of economic growth and financial development, our model also includes an array of conditioning information to control other factors associated with either economic growth or financial development, which are real interest rate (R) measured by the bank deposit rate deflated by the inflation, physical capital stock (LNK) by the natural logarithm of real per capita fixed capital formation and trade ratio (TR) by the total values of exports and importsin year t as a share of GDPt. We construct our physical capital stock series from investment flows. Specifically, we use a value of 175 billion yuan for the aggregate physical capital stockat the beginning of 1952 at 1952 price as in Chow (1993) and Wang and Yao (2003), adopt an average rate of depreciation rate of 5% as in Perkins (1988) and Wang andYao (2003),estimate the real investment flow as the gross investment divided by investment deflator and finally construct our real physical capital stock series following perpetual inventory method. Investment deflator is updated from both the index provided by Hsueh and Li (1999) covering the 1952–1990 periods and the index provided by China Statistical Yearbook starting from 1991.8 Like GDP figures, we convert the physical capital stock series into real term using this new index (1952=100).3.3. Descriptive statistics and correlationsWe use annual data spanning from 1952 to 2001.9 It is now known that the span of the data is much more important than the number of observations and there is no gain in the V ARframework by switching from low frequency to high frequency data (e.g., Campbell & Perron,6 Real GDP per capita figure is superior to total real GDP figure, because some of the errorsof the level of GDP and of population tend to be offsetting (Heston, 1994).7 The contemporaneous correlation coefficient between BCR and DLR is 0.9346, significant at 1% level.8 Hsueh and Li (1999) data, supported by the State Statistical Bureau of PRC, is the most complete and reliable nationalaccount data available so far (Wang & Yao, 2003). Therefore, it is not unreasonable to combine both sources to estimate acomplete index as they use similar approaches to construct their own indexes.9 Data of series (LNY, BCR, LNK, R, TR, DLR) are not reported to save space but are available upon request.----------------------- Page 7-----------------------Q. Liang, J.-Z. Teng / China Economic Review 17 (2006) 395–411401Table 1Summary statistics (1952–2001; 1952=100)LNY BCR LNK TR R (%) DLRMean 5.9595 0.5757 6.8545 0.1750 2.9270 0.4696 Median 5.6951 0.5081 6.7032 0.1052 3.0000 0.3331 Std. Dev. 0.8402 0.2464 0.8064 0.1244 5.8214 0.3138 Start value 4.7717 0.1570 5.7636 0.0951 11.70000.1356End value 7.5964 1.0874 8.3258 0.4335 1.5500 1.3737 1991; Hakkio & Rush, 1991). Table 1 presents descriptive statistics. It is evident that LNY, BCR, LNK, TR and DLR have increased steadily during the past 50 years. The respective average annual growth rates are 5.76%, 4.03%, 5.23%, 3.14% and 4.84%. Data also show that LNY correlates high positively with BCR (0.8915), LNK (0.9909), TR (0.9350) and DLR (0.9116). However, although the contemporaneous correlations between LNY and both BCR and DLR are very high, the correlations between the growth rate of LNY and the level of BCR and DLR are negative (0.0599) and pretty low (0.08) respectively. This phenomenon, similar to thosefindings in other studies (e.g., Luintel & Khan, 1999), suggests that the relationship between financial development and economic growth in China is a long-run one.4. Econometric methodology-multivariate V AR frameworkSuppose that the level of Yt can be represented as a non-stationary p -th order vectorautoregression equation:Y ¼ a þ U Y þ U Y þ . . . þ U Y þ U Y þ e ð3Þt 1 t1 2 t2 p 1 tp þ1 p tp tAccording to Hamilton (1994), this V AR(p ) model can be reparameterised as:DY ¼ a þ W DY þ W DY þ . . . þ W DY þ qY þ eð4Þt 1 t1 2 t2 p 1 tp þ1 t1 twhere Y =[LNY, BCR, LNK, R, TR] (or Y =[LNY, DLR, LNK, R, TR] in the robustness test) is t ta 5 1 vector of the first-order integrated variables; W = tU + U +. . .+ Ub for s =1, 2,s s+1 s+2 p. . ., p 1 and q = U(1); Wi are 5 5 coefficient matrices; et is a vector of normally and independently distributed error terms. Johansen (1988) and Johansen and Juselius (1992) derive the trace test and maximal eigenvalue test to identify the existence and number of distinct co- integrating vector in the V AR framework and Osterwald-Lenum (1992) tabulates appropriate critical values. If there exist r(0br b 5) co-integrating vectors, it implies q is rank-deficient,then q can be decomposed as: q = pcV, where p and cV . Thus, Eq. (4) can bere-written(5 r) (r 5)as:DY ¼ a þ W DY þ W DY þ . . . þ W DY þ pcVY þ eð5Þt 1 t1 2 t2 p 1 tp þ1 t1 twhere the components of p are the error correction coefficients indicating the speed ofadjustment towards long-run equilibrium and the rows of cV can be interpreted as the distinct co- integrating vectors. Pesaran and Shin (2002) suggest identification of co-integrating vectors through tests of r2 +k (k z 1) restrictions, where r2 is the just-identifying restriction proposed by Johansen (1991) and k is the over-identifying restriction. Each vector requires at least rrestrictions and one of them should be the normalization restriction. Restrictions must be based on economic theory so that the identified co-integrating vectors could be interpreted as----------------------- Page 8-----------------------402 Q. Liang, J.-Z. Teng / China Economic Review 17 (2006) 395–411 economically meaningful long-run relationships. For an illustration, we assume that there are two co-integrating vectors in our V AR model, which are normalized as economic growth and financial development relationships, respectively. Following Pesaran and Shin (2002), we needat least five theoretical plausible restrictions on c vector in order to find out long-run relationships among variables:0DLNYt 1 0p11 p12 1 0LNYt1 1B C B C B CDBCRt p21 p22 BCRt1B C B C 1 0 c13 c14 c15 B CB C B C B CDLNKt ¼ p31 p32 LNKt1 :B C B C c21 1 0 c24 0 B C@ A @ A @ ADRt p41 p42 Rt1DTRt p51 p52 TRt1ð6ÞTwo normalization restrictions are straightforward, i.e., the coefficients of LNY(c11) andBCR(c22) take a value of unity in the first and second vector, respectively. Theoretically, financial development affects economic growth indirectly. Besides, economic theory does not postulate any direct effect from physical capital stock and international trade to financial development. Thus, the other three restrictions are generated by setting the coefficient of c12 to zero in the first co-integrating vector and setting the coefficients of c23 and c25 to zero in the second co-integrating vector. In addition, in order to identify meaningful co-integration vectors, Wickens (1996) points out that the error correction coefficients (in this case, p11 andp22) must be statistically significant and their signs must be negative.After identifying co-integrating vectors, we test causality between financial development and economic growth. Johansen and Juselius (1992) point out that a test of zero restrictions on p is the test of weak exogeneity while Hall and Milne (1994) further show that weak exogeneity in a co-integrated system equals the long-run causality. If the null hypothesis p12=0 is rejected, then the economic growth vector is not weakly exogenous with respect to the financial development vector implying financial development does cause economic growth in the long run. Likewise, if the null hypothesis p21 =0 is rejected, then the financial development vector is not weakly exogenous with respect to the economic growth vector implying economic growth does cause financial development in the long run. If the null hypothesis p12=0 \p21=0 is rejected, it implies a bi-directional causality relationship between financial development and economic growth in the long run.5. Empirical resultsThe examination of long-run relationship among financial development, economic growth and other key growth factors is carried out in three steps. Since a VAR framework dependson the time series characteristics of the dataset, we initially investigate the order of integration of the variables using standard tests for the presence of unit roots. Next, the number of co-integrating vectors is tested using the Johansen maximum likelihood approach while the economically meaningful co-integrating vector is identified through tests of over- identifying restrictions. Finally, causal relationship between financial development and economic growth is evaluated through tests of weak exogeneity. Additionally, we substitute BCR with DLR to our multivariate V AR framework to test the robustness, and providediscussions of our results.----------------------- Page 9-----------------------Q. Liang, J.-Z. Teng / China Economic Review 17 (2006) 395–4114035.1. Unit roots testWe choose to implement KPSS test (Kwiatkowski, Phillips, Schmidt, & Shin, 1992) of stationary hypothesis upon all series under consideration.10 KPSS test statistics indicates that LNY, BCR, LNK and TR are both level non-stationary and trend non-stationary, all significant at 1% level except for the 10% significant for BCR in the trend stationary test. The first differences of LNY, BCR, LNK and TR are tested to be stationary. As to the real interest rate, KPSS test statistics shows that R is level non-stationary but trend stationary. Tests on the first difference indicate R to be stationary. We have decided to treat R as I(1) since economic theory does not postulate a time trend in the real interest rate. Thus, we have concluded that all series are integrated of order one, or I(1).。