How agglomeration in the financial services industry influences economic growth: Evidence from Chinese cities

Liang, Lin; Lin, Shanglang; Li, Yong Working Paper How agglomeration in the financial services industry influences economic growth: Evidence from Chinese cities Economics Discussion Papers, No. 2014-6 Provided in Cooperation with: Kiel Institute for the World Economy – Leibniz Center for Research on Global Economic Challenges Suggested Citation: Liang, Lin; Lin, Shanglang; Li, Yong (2014) : How agglomeration in the financial services industry influences economic growth: Evidence from Chinese cities, Economics Discussion Papers, No. 2014-6, Kiel Institute for the World Economy (IfW), Kiel This Version is available at: https://hdl.handle.net/10419/92415 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/3.0/ Received January 16, 2014 Accepted as Economics Discussion Paper January 29, 2014 Published February 10, 2014 © Author(s) 2014. Licensed under the Creative Commons License - Attribution 3.0 Discussion Paper No. 2014-6 | February 10, 2014 | http://www.economics-ejournal.org/economics/discussionpapers/2014-6 How Agglomeration in the Financial Services Industry Influences Economic Growth: Evidence from Chinese Cities Lin Liang, Shanglang Lin, and Yong Li Abstract This paper empirically tests the effect of financial knowledge spillovers on agglomeration in China’s financial services industry and examines the external effects on cities’ economies. The authors apply hierarchical linear modeling to examine a data set that comprises 276 Chinese cities and draw the following conclusions. Firstly, they find that agglomeration in the financial services industry and the Jacobs spillovers of industry diversification both promote financial knowledge spillovers in terms of industry specialization. Secondly, agglomeration in this studied industry has a significant positive influence on a city’s economic growth, while financial knowledge spillovers have a significant but negative effect on a city’s economic growth. Thirdly, the tendency towards agglomeration in the financial services industry in a few major cities is clear and the clustering significantly influences cities’ boundaries. Finally, China’s financial services industry is limited by a serious degree of regulation and is dominated by the main banking institutions. JEL G20 O4 Keywords Financial services industry agglomeration; industry specialization; knowledge spillovers; city economies; hierarchical linear modeling Authors Lin Liang, School of Economics and Management, Tongji University; 1239 Siping Road, Shanhai, P.R. China, lianglintj201[email protected] Shanglang Lin, School of Economics and Management, Tongji University, P.R. China Yong Li, School of Economics and Management, Tongji University, P.R. China Citation Lin Liang, Shanglang Lin, and Yong Li (2014). How Agglomeration in the Financial Services Industry Influences Economic Growth: Evidence from Chinese Cities. Economics Discussion Papers, No 2014-6, Kiel Institute for the World Economy. http://www.economics-ejournal.org/economics/discussionpapers/2014-6 1. Introduction The modern financial services industry has certain notable characteristics compared with other market sectors. Firstly, it tends to be more dependent on the development of the local economy as well as on local market capacity. Secondly, its investment into physical capital is relatively low and it rather relies on the provision of financial information and on human capital, the use of which has become increasingly intensive. The combination of these features leads the financial services industry to produce more obvious spatial agglomeration effects compared with other sectors. Western academics began to pay attention to agglomeration in the financial services industry in the 1970s. For example, although Kindleberger (1973) and Gehrig (1998) defined the concept of a financial center from different angles, they agreed that financial centers are typical of agglomeration in this industry. Porteous (1995) explained the development of regional financial centers through the concepts of ‘path dependence,’ ‘asymmetric information,’ and the ‘information hinterland.’ Krugman (1991a) proposed that agglomeration in the financial services industry is clearer than that in the manufacturing industry because knowledge spillovers are typically external and that such a development was an important driving force behind the agglomeration of London’s financial services industry. Hall and Appleyard (2009) further pointed out that business knowledge, highly skilled financiers, and the financial labor market were all important factors in the knowledge spillovers in London’s financial center. Finally, Keeble and Nachum (2002) stated that knowledge-intensive industries such as the financial sector should aim to explore the benefits of agglomeration from the aspects of knowledge accumulation and the innovation environment. Industry agglomeration has begun to occur in China over the past three decades. Wen (2004), for instance, pointed out that many of China’s manufacturing industries are highly geographically concentrated in several coastal regions and that this geographical concentration has increased since the economic reform. Lu and Tao (2009) also used a large firm-level data set for the period 1998– 2005 and found that the extent of industry agglomeration in China’s manufacturing industry increased steadily throughout the sample period. Lu (2010) found that the primary sector and private firms are more spatially concentrated than the secondary and tertiary sectors and public firms, respectively. In the same vein, Li et al. (2012) determined a positive correlation between industry agglomeration and firm size for Chinese manufacturing firms from 1998 to 2005 and showed that firms are more likely to become larger by being located with a number of larger firms compared with a larger number of firms. Moreover, researchers have found that protectionism among China’s various regions hampers the geographical concentration of manufacturing industries. In this regard, Bai et al. (2004) argued that less geographic concentration is found in industries where the past tax-plus-profit margins and the shares of state ownership are high, reflecting stronger local government protection in these industries. 2 However, while previous studies of industry agglomeration focus on manufacturing in terms of agglomeration degree, influencing factors, and the role of government in the agglomeration process, there has been little research into agglomeration in the financial services industry. In particular, no studies have thus far explored financial knowledge spillovers and the relation between agglomeration in the financial services industry and economic growth. The present study therefore contributes by bridging a gap in the literature. Economic researchers began to pay attention to the correlation between financial development and economic growth in the 1990s. For example, King and Levine (1993a, 1993b, 1993c) studied the effect of financial development on economic growth from the aspect of financial function, while Rousseau and Sylla (1999) investigated changes to the US financial system from 1790 to 1840 and found that the real power of modern economic growth comes from financial changes. We build on these findings by asking the following research questions: Does there exist a relationship between agglomeration in China’s financial services industry and regional economic growth? What effect on economic development is produced by financial agglomeration and financial knowledge spillovers? And what role does the Chinese government play in financial agglomeration and financial knowledge spillovers? Specifically, this paper analyzes these relations according to the theories of agglomeration economics, new economic geography, and service economics. Methodologically, it tests four hypotheses by using hierarchical linear modeling (HLM) based on a data set that comprises 279 Chinese cities. We predict that agglomeration in the financial services industry promotes financial knowledge spillovers. At the same time, agglomeration and knowledge spillovers in the financial services industry benefit a city’s economic growth. The remainder of the paper is organized as follows. Section 2 reviews the relevant literature on financial agglomeration and knowledge spillovers. Section 3 introduces the research design. Section 4 explains the data source and describes the statistics. Section 5 puts forward the research hypotheses and model structure. Section 6 analyzes the estimation results and discusses the implications. Section 7 concludes the findings and offers suggestions for the further research. 2. Literature review Enterprises that produce identical or related financial products are concentrated in a specific area of the financial services industry, which reduces investment in financial products and transaction costs and thus produces a scale agglomeration effect. Marshall (1890) explained industry clustering by using the theory of external economies of scale. He claimed that specialized division of labor promotes economic growth in the region in which the industry is concentrated. Later, Krugman (1991b), the founder of modern location theory, summarized three types of gains from specialized agglomeration: a shared labor market, non-tradable intermediate inputs, and the production function conversion that follows from knowledge spillovers. Marshall (1890), Arrow 3 (1962), and Romer (1986) believed that regional clusters can promote knowledge diffusion among different firms in the same industry, thereby improving R&D and innovation. Davis (1990) found that financial agglomeration can promote mutual learning and technological innovation as well as reduce transaction costs through knowledge spillovers. Aydogana and Lyon (2004) also pointed out that central gathering places and two-way communication can realize knowledge spillovers. Finally, according to Sternberg (1996), an informal communication network can effectively build open information exchanges and an innovation environment, strengthening the knowledge flow, accelerating the transformation of know-how, improving overlaps among knowledge sources, and promoting cluster integration. Agglomeration in the financial services industry and industry diversification influence the type and degree of financial knowledge diffusion and spillover. In a general sense, knowledge spillovers can come from both enterprises in the same industry as well as those in different industries. Knowledge spillovers from the same industry are called MAR externalities in academic circles (Marshall 1890; Arrow 1962; Romer 1986, 1990). For MAR externalities, knowledge spillovers result from information exchange among enterprises in the same industry (e.g., the exchange of production information or flow of professional and technical personnel between enterprises). By contrast, knowledge spillovers that come from different enterprises are called Jacobs externalities (Jacobs 1969). This type of spillover mechanism expands the scope of knowledge spillovers through interaction with other industries. Further, the diversification of geographical agglomeration can promote the enterprise’s innovation behavior. Although the contribution of knowledge spillovers to economic growth has been acknowledged, evidence on the influence of MAR and Jacobs externalities is inconclusive. Glaeser et al. (1992), for example, used data on six large industries in 170 US cities in 1956 and 1987 to verify the effect of Jacobs externalities, but found that MAR externalities negatively affect regional economic growth. Moreover, Henderson, Kuncorn, and Turner (1995) used data on 224 major metropolitan areas in the US in 1970 and 1987 and found MAR externalities but no Jacobs externalities in mature capital-intensive industries. Similar research and different conclusions have also appeared in Italy (Cainelli and Leoncini 1999, Forni and Paba 2002), France (Combes 2000), Spain (De Lucio, Herce, and Goicolea 2002), and the Netherlands (Van Soest, Cerking, and Van Oort 2002) owing to the different country- and regional-level characteristics, degrees of industry organization and development, samples and variables, and validation methods. 3. Research design 3.1 The agglomeration in the financial services industry at the Chinese city level This paper uses the comprehensive index evaluation method in order to analyze 4 agglomeration in the financial services industry at the Chinese city level (CFAG hereafter). We select banking, insurance, and securities to represent the financial services industry because these three areas account for more than 90% of financial enterprises in China. Table 1 provides more details on the index system. The analytic hierarchy process is then used in order to determine the weight of each index, while the Weaver index method is applied to calculate the sequence and key elements of the different indexes (Table 2). (Table 1) (Table 2) 3.2 Industry specialization and diversification Based on the methods of Glaeser et al. (1992) and Feldman and Audretsch (1999), this paper adopts employment distribution in order to measure the characteristics of industry diversification (CDIV) and industry specialization (CSPE) as follows: The index definition of industry diversification is CDIVi= 1 − ∑pij 219 j=1 (3.1) Where Pij represents the employment of industry j in city i as a proportion of the city’s total employment. The value of CDIV is thus 0～1. If the index is high and the Pij value is low, it indicates that the employment trend is dispersive in different industries, which means that the industry distribution in the city is diverse. Further, there are 19 industry groupings according to the China City Statistical Yearbook (J=1–19). The index definition of industry specialization is CSPEI=piP ⁄ (3.2) Where pi represents a city’s financial industry employment as a proportion of its total employment and P represents the proportion of national financial industry employment as a proportion of total employment. The value of CSPE is thus 0～1. If the index is high, it indicates that the city’s proportion of financial industry employment is higher than the national average. This measure represents that the level of specialized financial services in the city is higher than the national average. 3.3. The application of HLM According to Kreft, De Leeuw, and Aiken (1995) and Krasnikov, Jayachandran, and Kumar (2009), HLM is a statistical method for processing nested data. Social science data often have a hierarchical structure, which not only describes the individual variables, but also shows the higher-level variables formed by individuals. All the variables analyzed in the general regression 5 method are at the same level. The premise of the general regression method is therefore random error independence and homogeneity of variance among all variables. For multilayer nested data, the analysis of all the variables in a level includes both individual factors and repeated measurement factors. Hence, the hypothesized premise of the general regression method is not always true because the results of the data have an unreasonable or incorrect interpretation. HLM can divide the random variation into two parts by defining different levels of the model: the first level of individual differences, which are independent of each other, and the second level between the variables that are independent of each other. Therefore, HLM also includes the variation caused by different levels. Because this test used herein relates to two levels of data, namely the city nested in the province, the differences between these two levels should be taken into account in the empirical testing. Cities in the same province may not be independent of each other because they share policies, have a unified approach to administrative management, and similar cultural traditions, climatic conditions. Thus, the test errors are divided into two parts. One is the error among the individual differences of different cities; here, the measurement error is assumed to be independent of each other. Secondly, the error at the provincial level among the different provinces is assumed to be independent of each other. 4. Data sources and descriptive statistics 4.1 Data sources This empirical study used a sample of 279 prefecture-level cities belonging to China’s 25 provinces in 2011 (excluding the four municipalities directly under the central government and the less prefecture-level cities of Qinghai and missing data in Tibet). Because the financial services industry is mainly concentrated in cities, the spatial range of our single sample data set did not include counties. The data were derived from the relevant 2012 statistical yearbooks. The empirical test of the main variables involved CFAG, CDIV, CSPE, and the city’s economic growth index (CECO), which is the log of the city’s GNP in 2012. 4.2 Control variables In addition to the main variables above, we used separate control variables for the province level and city level (Table 3). There were five control variables at the city level. High productivity is one of the effects of industry agglomeration. Andersson and Loof (2011) suggested a learning effect in that agglomeration enhances productivity. The higher the degree of openness of the city, the greater is FDI and the higher is the agglomeration of the financial services industry. Zhao, Zhang, and Wang (2004) argued that the higher the position of a financial enterprise in an information center, the lower is the cost of the information obtained. Therefore, information infrastructure is important for financial agglomeration. Meanwhile, human capital in the city is good for industry 6 development and higher wages attract more financial enterprises. There were also five control variables at the province level. The financial output contribution rate (PFVC) and financial value of the location (PAPS) improve the city’s economic development and thus financial agglomeration. Moreover, R&D expenditure input intensity (PRD) and the patents granted rate (PAT) reflect knowledge spillovers. In addition, in order to separate the provincial- and city-level agglomeration in the financial services industry, we used the Herfindahl index (PHPS) of provincial financial services in the regression equation. The higher this index, the more concentrated the financial services industry in this province is in a few cities (Table 3). (Table 3) 4.3 Descriptive statistics In order to calculate the correlations between the province-level variables and city-level variables, we follow Fu et al. (2010) by disaggregating the former into the latter. The results are presented in Table 4. We find that CFAG is positively correlated with CSPE (r=0.211, p<0.01) and CECO (r=0.605, p<0.01) and that CSPE and CDIV are positively correlated (r=0.368, p<0.01). Further, the correlation between CSPE and CECO is significant but negative as is that between CDIV and CECO. These results provide partial support for the research hypotheses, which are tested in more detail by using HLM in Section 5. Here, we introduce independent variables into two models (A and B) in turn in order to eliminate the influence of multicollinearity. The L1, L2, and mixed-model equations are available on request. (Table 4) 5. Research hypotheses and model structure 5.1 Research hypotheses According to literature review and study methodology, we thus propose the following four research hypotheses: Hypothesis 1: Agglomeration in the financial services industry is positively related to the effect of specialized knowledge spillovers. Hypothesis 2: Jacobs spillovers based on industry diversification are positively related to financial knowledge spillovers. Hypothesis 3: Agglomeration in the financial services industry is positively related to the city’s economic growth. 7 Hypothesis 4: Financial knowledge spillovers are positively related to the city’s economic growth. 5.2 Test model structure The hypothesis test focuses on two models. Model A tests the industry specialization and knowledge spillovers caused by agglomeration in the financial services industry, in which financial knowledge spillovers is the dependent variable and agglomeration in the financial services industry and industry diversification are the independent variables. Model B tests whether agglomeration in the financial services industry and financial knowledge spillovers significantly affect the city’s economic growth. Here, the city’s GDP is the dependent variable, while agglomeration in the financial services industry, financial knowledge spillovers, and industry diversification are the independent variables. The specific models are described below. Model A: 𝐋𝟏:CSPE = β0j + β1jCFAG + β2jCDIV + β3jCPRO + β4jCOPE + β5jCIE + β6jCSH + β7jCWA + εij ; 𝐋𝟐:β0j = γ00 + γ01PFVC + γ02PAPS + γ03PHPS + γ04PRD + γ05PAT + μ0j β1j = γ10 + μ1j ;β2j = γ20 + μ2j；β3j = γ30 + μ3j；β4j = γ40 + μ4j；β5j = γ50 + μ5j；β6j = γ60 + μ6j；β7j = γ70 + μ7j (5.1) Model B: 𝐋𝟏:lnCECO = β0j + β1jCFAG + β2jCDIV + β3jCSPE + β4jCPRO + β5jCOPE + β6jCIE + β7jCSH + β8jCWA + εij ; 𝐋𝟐: β0j = γ00 + γ01PFVC + γ02PAPS + γ03PHPS + γ04PRD + γ05PAT + μ0j β1j = γ10 + μ1j；β2j = γ20 + μ2j；β3j = γ30 + μ3j；β4j = γ40 + μ4j；β5j = γ50 + μ5j；β6j = γ60 + μ6j；β7j = γ70 + μ7j;β8j = γ80 + μ8j (5.2) The first-level (L1) data in Models A and B are the city samples and the second level (L2) data are the province samples. The L1 model is similar to the general regression model, while the economic explanations of the coefficients of the variables are also similar. Although the intercept and slope of the regression equation are not assumed to be constant, they act as explanatory variables of in L2 regression equation (Gu 2010). In order to simplify the analysis, however, this 8 http://www.marketingpower.com/AboutAMA/Pages/AMA%20Publications/AMA%20Journal s/Journal%20of%20Marketing/TOCs/SUM_2009.6/Impact_of_Customer_Relationship.aspx. Kreft, I., J. de Leeuw, and L. Aiken. (1995). The Effect of Different Forms of Centering in Hierarchical Linear Models. Multivariate Behavioral Research 30(1): 1–22. URL: http://www.tandfonline.com/doi/abs/10.1207/s15327906mbr3001_1#.UtULvvQW300. Krugman, P. (1991a). Geography and Trade. Cambridge, MA: MIT Press. Krugman, P. (1991b). Increasing Returns and Economic Geography. Journal of Political Economy 99(3): 483–499. URL: http://www.jstor.org/discover/10.2307/2937739?uid=3738256&uid=2&uid=4&sid=21103335 902433. Leoncini, R., and G. Cainelli. (1999). Externalities and Long-term Local Industrial Development: Some Empirical Evidence from Italy. Revue d’Economie Industrielle 90(90): 25–39. URL: http://www.persee.fr/web/revues/home/prescript/article/rei_0154-3229_1999_num_90_1_176 2. Lu, J. (2010). Agglomeration of Economic Activities in China: Evidence from Establishment Censuses. Regional Studies 44(3): 281–297. URL: http://www.tandfonline.com/doi/abs/10.1080/00343400802195139#.UtUM9_QW300. Li, D., Y. Lu, and M. Wu. (2012). Industrial Agglomeration and Firm Size: Evidence from China. Regional Science and Urban Economics 42(1/2): 135–143. URL: http://www.sciencedirect.com/science/article/pii/S0166046211000937. Lu, J., and Z. Tao. (2009). Trends and Determinants of China’s Industrial Agglomeration. Journal of Urban Economics 65(2): 167–180. URL: http://www.sciencedirect.com/science/article/pii/S0094119008001125. Marshall, A. (1890). Principles of Economics. 8th ed. London: Macmillan. Porteous, D. (1995). Geography of Finance: Spatial Dimensions of Intermediary Behavior. Aldershot: Avebury. Romer, P. (1986). Increasing Returns and Long-run Growth. Journal of Political Economy 94(5): 1002–1037. URL: http://www.jstor.org/discover/10.2307/1833190?uid=3738256&uid=2&uid=4&sid=21103335 902433. Romer, P. (1990). Endogenous technology change. Journal of Political Economy 98(5): S71– S102. URL: 15 http://www.jstor.org/discover/10.2307/2937632?uid=3738256&uid=2&uid=4&sid=21103335 902433. Rousseau, P., and R. Sylla. (1999). Emerging Financial Markets and Early US Growth. NBER Working Paper, No. 7448. URL: http://www.nber.org/papers/w7448 Sternberg, R. (1996). The Nature of Creativity. Creativity Research Journal 18(1): 87–98. URL: http://www.tandfonline.com/doi/abs/10.1207/s15326934crj1801_10#.UtUPF_QW300. van Soest, D. P., S. Cerking, and F. van Oort. (2002). Knowledge Externalities, Agglomeration Economies and Employment Growth in Dutch Cities. Center Discussion Paper, Tilburg University. URL: http://arno.uvt.nl/show.cgi?fid=4341. Wen, M. (2004). Relocation and Agglomeration of Chinese Industry. Journal of Development Economics 73(1): 329–347. URL: http://www.sciencedirect.com/science/article/pii/S0304387803001408. Zhao, S., L. Zhang, and D. Wang. (2004). Determining Factors of The Development of a National Financial Center: The Case of China. Geoforum 35(5): 577–592. URL: http://www.sciencedirect.com/science/article/pii/S0016718504000053. 16 Tables Table 1: Evaluation index of the agglomeration in the financial services industry at the city level Industry Sub-industry (Ai) Evaluation index system (Bi) CFAG Banking (BAG) Financial assets state (B1, %), Financial employment contribution rate (B2, %), Financial employment location (B3), Deposit income ratio (B4, %), Deposit-loan difference (B5), Loan-to-deposit ratio (B6,%) Insurance (IAG) Property insurance premium income (I1), Property insurance density (I2, Yuan/per), Property insurance depth (I3, %), Personal insurance premium income (I4), Personal insurance density (I5, Yuan/per), Personal insurance depth (I6, %) Securities (CAG) Number of listed companies/ten thousand people (C1), Share of the total transaction volume/ten thousand people (C2, millions of dollars), Shares of IPO raised capital/ten thousand people (C3, millions of dollars) Note: The calculation of city-level agglomeration in the financial services industry does not include the securities industry. 17 Table 2: Explanation of the main variables in the evaluation index Variable Function Calculation method B1 Financial assets state Deposits and loans are the most important instruments in the financial industry This index measures the financial development of cities. B1=Deposits balance in city + Loans balance in city /city GDP that year. B2 Financial employment contribution rate; B3 financial employment location These two indexes measure the city’s financial employment contribution to national financial employment and concentration degree. B2 = City’s financial employment/national financial employment; B3 = (city’s financial employment/city’s employment)/(national financial employment/national employment). B4 Deposit-income ratio; B5 Deposit-loan difference; B6 Loan-to-deposit ratio These three indexes measure the city’s financial reserves capacity and financial capital supply into demand service ability B4 = Deposits balance in city/city GDP that year; B5 = Deposits balance in city-Loans balance in city at the end of the year; B6 = Loans balance in city/deposits balance in city at the end of the year. I2 Property insurance density; I5 Personal insurance density This index measures the level of the city’s insurance industry development and the degree of people participating in it City’s insurance premium income/city’s total population that year. I3 Property insurance depth; I6 Personal insurance depth This index measures the status of the city insurance industry in the national economy. It depends on the national overall economic development level and the insurance industry development speed. City insurance premium income/city’s GDP that year. 18 Table 3: Explanation of the control variables Level 1: City level Control variable Calculation method CPRO Productivity City’s output/city’s inputs COPE Openness degree Foreign capital used in the city/city’s GDP that year CIE Information infrastructure Postal and telecommunication income/city’s GDP CSH Human capital The number of normal schools/city’s total population CWA Wages The city’s annual average wage Level 2: Province level PFVC Financial output contribution rate Provincial financial services industry output/national financial services industry output PAPS Financial value of the location (Provincial financial services industry output/provincial GDP)/(national financial services industry output/national GDP). PHPS Herfindahl index n × ∑(pi p)2 n i=1 n is the number of cities belonging to this province, pi is the industry employment of the city’s financial services, P is the industry employment of provincial financial services. PRD R&D expenditure input intensity Provincial R&D expenditure/provincial GDP PAT Patents granted rate Three kinds of domestic patents granted by the province/total number of patents granted 19 Table 4: Correlations and Statistical description of All Variables in the Study Statistical description of variables Variables CPRO COPE CIE CSH CWA PFVC PAPS PHPS PRD PAT CFAG CDIV CSPE lnCECO Mean 3.28 1.86 6.49 0.26 0.38 3.07 0.82 2.04 1.16 4.00 1.18 0.82 1.15 6.09 SD 1.83 2.09 4.87 0.11 0.08 3.11 0.28 0.65 0.53 6.65 0.50 0.09 0.54 1.06 Minimum 0.76 0.00 0.70 0.70 0.19 0.41 0.40 1.25 0.41 0.08 -0.05 0.42 0.14 3.67 Maximum 19.04 16.25 29.12 0.91 0.74 11.26 1.70 3.58 2.17 26.82 2.29 0.92 3.22 9.35 Number 25 279 25 Correlation coefficient between variables Variables CPRO COPE CIE CSH CWA PFVC PAPS PHPS PRD PAT CFAG CDIV CSPE lnCECO COPE 0.094 CIE -0.34** -0.165** CSH -0.145* -0.269** 0.274** CWA 0.349** 0.192** -0.286** -0.267** PFVC -0.306 0.138 -0.047 -0.068 0.389 PAPS 0.116 -0.009 0.032 0.150 0.183 0.552** PHPS 0.042 0.086 0.016 -0.002 0.435 0.114 0.002 PRD -0.363 0.060 0.187 -0.079 0.280 0.740** 0.219 0.315 PAT -0.238 0.028 -0.023 -0.080 0.261 0.937** 0.512** 0.039 0.731** CFAG 0.072 0.276** -0.058 -0.257** 0.426** 0.192 -0.089 -0.365 0.110 0.160 CDIV -0.031 -0.240** 0.240** 0.224** -0.262** -0.282 -0.132 -0.445* -0.319 -0.180 -0.05 CSPE 0.078 -0.150* 0.348** 0.092 -0.040 -0.141 0.113 -0.252 -0.290 -0.066 0.211** 0.368** lnCECO 0.417** 0.373** -0.536** -0.433** 0.532** -0.110 -0.095 -0.124 -0.128 -0.131 0.605** -0.199** -0.243** Notes: ** and * denote 1%, 5% significance respectively. 20 Table 5: Estimation Results of the HLM Model: Model A Variables CSPE Ma-1 Ma-2 Ma-3 Ma-4 Intercept 1.13**(0.03) 1.13**(0.04) 1.13**(0.03) 1.13**(0.04) Level 1 control variables CPRO 0.08**(0.02) COPE -0.03*(0.01) CIE 0.044**(0.01) CSH 0.001(0.36) CWA -0.36(0.51) 0.08**(0.02) -0.04*(0.02) 0.04**(0.007) 0.15(0.40) -0.79(0.48) 0.07**(0.02) -0.02(0.01) 0.04**(0.008) -0.05(0.33) -0.43(0.43) 0.07**(0.02) -0.02*(0.01) 0.03**(0.007) 0.02(0.34) -0.49(0.42) Level 2 control variables PFVC 0.03(0.03) PAPS -0.017(0.15) PHPS -0.13*(0.06) PRD 0.22(0.15) PAT -0.02*(0.01) 0.03(0.04) -0.018(0.21) -0.13*(0.07) 0.22(0.15) -0.02*(0.02) 0.03(0.03) -0.02(0.15) -0.13*(0.06) 0.23(0.15) -0.02*(0.01) 0.03(0.04) -0.02(0.21) -0.13*(0.07) 0.23(0.15) -0.02*(0.01) Independent variable(level 1) CFAG CDIV 0.29**(0.07) 0.22**(0.05) 1.64**(0.26) 0.22**(0.05) 1.59**(0.27) Sigma square 0.23 Tau 0.02 Chi-square 40.27**(19) Pseudo R2 change(Level 1) 0.22 0.02 42.83**(19) 0.043 0.20 0.02 45.67**(19) 0.091 0.20 0.02 45.67**(19) 0.091 Note: The values in brackets are standard errors for the corresponding estimates. Level 1 = city level; Level 2 = province level. **and * denote 1% and 5% significance, respectively. 21 Table 6:Estimation Results of the HLM Model: Model B Variables lnCECO Mb-1 Mb-2 Mb-3 Mb-4 Mb-5 Mb-6 Intercept 6.00**(0.05) 6.00**(0.05) 6.00**(0.06) 6.02**(0.06) 6.02**(0.06) 6.02**(0.06) Level 1 control variables CPRO 0.08**(0.02) COPE 0.08**(0.02) CIE -0.08**(0.01) CSH -1.19*(0.60) CWA 3.94**(1.02) 0.11**(0.03) 0.07**(0.02) -0.17**(0.01) -0.19*(0.59) 3.85**(1.04) 0.09*(0.03) 0.08**(0.02) -0.07**(0.01) -1.43*(0.56) 4.06**(1.01) 0.08**(0.02) 0.03*(0.01) -0.08**(0.01) - 0.59(0.54) 2.29**(0.67) 0.08**(0.02) 0.03(0.01) -0.08**(0.01) -0.56 (0.51) 2.24**(0.71) 0.12**(0.02) 0.01(0.01) -0.06**(0.01) -0.62(0.45) 2.02**(0.70) Level 2 control variables PFVC 0.14**(0.04) PAPS -0.6**(0.18) PHPS -0.25*(0.09) PRD 0.15 (0.17) PAT 0.001(0.01) 0.14**(0.04) -0.6**(0.18) -0.25*(0.09) 0.15 (0.17) 0.001(0.01) 0.14**(0.04) -0.6**(0.18) -0.25*(0.09) 0.14 (0.17) 0.001(0.01) 0.14**(0.04) -0.59**(0.18) -0.25*(0.09) 0.08 (0.17) 0.004*(0.01) 0.14**(0.04) -0.59**(0.18) -0.25*(0.09) 0.08 (0.17) 0.004(0.01) 0.14**(0.04) -0.59**(0.18) -0.25*(0.10) 0.05 (0.18) 0.005(0.01) Independent variable(level 1) CFAG CDIV CSPE -0.25*(0.07) 1.43**(0.62) -0.27**(0.10) 1.13**(0.05) 1.13**(0.05) -0.20 (0.51) 1.25**(0.05) 0.64(0.62) -0.54**(0.10) Sigma square 0.47 Tau 0.05 Chi-square 43.40**(19) Pseudo R2 change(Level 1) 0.46 0.05 44.70**(19) 0.021 0.44 0.05 46.98**(19) 0.043 0.26 0.08 81.44**(19) 0.409 0.26 0.08 81.15**(19) 0.409 0.20 0.10 108.10**(19) 0.231 Note: The values in brackets are standard errors for the corresponding estimates. Level 1 = city level; Level 2 = province level. **and * denote 1% and 5% significance, respectively. 22 Please note: You are most sincerely encouraged to participate in the open assessment of this discussion paper. You can do so by either recommending the paper or by posting your comments. Please go to: http://www.economics-ejournal.org/economics/discussionpapers/2014-6 The Editor © Author(s) 2014. Licensed under the Creative Commons Attribution 3.0.