Crony Banking and Local Growth in China

The rise of city commercial banks (CCBs) in Chinese cities provides a unique opportunity to study the finance and growth nexus at the city level. Given the notorious inefficiency of China's "Big Four" state banks, policymakers attempted to correct the situation in 1995 through the creation of a new kind of local bank designed to promote local growth by lending to small and medium-sized enterprises. Using 1990-2009 panel data for 283 prefectural-level cities and four provincial-status municipalities, we find that the establishment of CCBs significantly reduced local economic growth overall. We suggest this outcome stems from the ability of firms to bribe local government officials to obtain credit from their local CCBs. In our proposed model for crony banking relations, large firms spend disproportionately larger amounts of time and bribe money cultivation relations with local officials involved in CCB lending decisions, so we expect large firms to have easier access to credit than small firms even if it results in inefficient lending. Using data on 206,771 firms for 1999-2007, we find that cities with CCBs had significantly lower overall growth rates. Small firms, in particular, were negatively impacted by the presence of CCBs, while large firms benefited from their presence. In the cities with CCBs, large firms, even those with relatively poor return-on-assets ratios, obtained more credit than small firms in aggregate. Using data from the 2005 World Bank Business Environment Survey, we find that an increase in a firm's crony relations with the government, measured in terms of the average number of days a month top managers of the firm spend interacting with government officials, increases the likelihood a firm will be granted bank credit. This effect was quite distinct for cities with CCBs.


Introduction
China's state-owned banks, particularly the "Big Four" (Bank of China, China Construction Bank, Industrial and Commercial Bank of China, and China Agricultural Bank) are often accused of inefficient lending practices (Allen, Qian and Qian, 2005), especially lending to state-owned enterprises. In 1995, the China Bank Regulatory Commission (CBRC) sought to rectify some of this inefficiency through the creation of small local banks to lend to small and medium-sized enterprises (SMEs) for the purpose of promoting local economic growth.
Operations of these local city commercial banks (CCBs) 1 were confined by law to their designated home cities until 2006. This decade during which CCBs were under regulatory constraints provides a unique opportunity to study the heavily debated relationship between finance and growth, 23 because the Big Four state-owned banks maintain branches in every Chinese city.
CCBs were initially created through the merger and restructuring of more than 5,000 urban credit cooperatives. CCBs introduced a more aggressive approach to banking than that of the cooperatives they absorbed (see Section 2). For example, the Bank of Shanghai was founded through the merger of 98 urban credit cooperatives at the end of 1995. Just a year after the Bank of Shanghai was created, its total assets had risen by 89.3% and its total loan stock had increased by 82.8%.
By the end of 2012, CCBs had been established in 161 of China's 288 main cities (includes four provincelevel municipalities). Their assets accounted for 9.24% of total domestic banking sector assets. CCB impact at the local scale was proportionately greater, however, as CCBs were only found in about 60% of cities and their activities largely concentrated on the urban core. 4 A report by KPMG (2007) to foreign investors notes that CCBs were widely perceived at that time to be heavily influenced by local government. This observation is hardly surprising, given that local governments on average held roughly 70% stakes in their local CCBs. The KPMG report raises the possibility that local government officials might be tempted to use their influence over local bank lending decisions to further their own personal interests. News in recent years about widespread corruption in local-level banking suggests this temptation is hard to resist. For example, the newspaper Can Kao Xiao Xi observed in late 2014 that most financial corruption news in the previous two years in some way involved local banks. Among the more standout stories was the tale of a local bank manager in Shenmu, Aiai Gong, who managed to amass personal real estate holdings worth approximately $2 billion on her annual salary of less than $10,000. In another story, the president of the Chengdu CCB was sentenced to death for soliciting bribes. Despite the bad press and harsh punishments for corruption, local banks continued to focus their lending on large inefficient state-owned To address possible endogeneity concerns, we use the percentage of neighboring cities that have established CCBs as our instrumental variable (IV) for the CCB establishment probability in a target city. CCBs were first established in politically important cities, i.e. provincial capitals, the four municipalities with provincial status, and the five cities with sub-provincial status. We thus consider this initial wave of CCB creation as exogenously determined. We adopt the instrumental variable IV in accordance with the policy diffusion literature. Simmons and Elkins (2004), for example, argue that neighboring regions are more likely to adopt the same policy due to factors such as altered payoff, reputation concern, and learning. We see this policy diffusion mechanism clearly at work in Figure 1 in the spread of CCBs. Our first stage regression presented below confirms a very significant relation that further validates our IV choice. Both macro and micro regressions show that establishment of CCBs reduced growth significantly.
We apply the same estimation methodology to a firm-level dataset that includes all SOEs and all non-SOEs with annual sales greater than RMB 5 million. After keeping firms with four consecutive years of appearance and controlling firm characteristics, we find that the presence of a CCB in a surveyed firm's city significantly reduced its growth rate. This negative effect was strongest for small firms.
We next develop a model that might explain for this counter-productive (at least from the central government's standpoint) result, arguing that the establishment of local banks alters local firm size distribution and consequently depresses the economic growth rate. Ostensibly, CCBs were created to foster higher growth rates by lending to small firms. As these local banks were largely extensions of local government, however, they were susceptible to capture, especially by large firms that could pay disproportionately larger bribes and devote more time to capture of local government officials. A bribe need not be monetary, but might include implicit benefits such as reputation enhancement, future employment, or the power to influence corporate decisions. Local officials taking bribes also prefer to keep their number of donors limited to avoid the risk of getting caught and punished, as well as provide adequate attention to bribe-paying patrons. This behavior comports with capture theory for democracies as developed by Bardhan and Mookherjee (2000), whereby rich voters increase campaign spending to influence campaign outcomes. Local banks controlled or heavily influenced by local governments are expected to be more inclined to lend to large firms, even when these firms are less productive than their smaller counterparts due e.g. to a decreasing return to scale for capital. The seminal work on local banks by Guiso, Sapienza and Zingales (2004) asserts that local financial development can contribute to local growth when local financial development makes it easier for small firms to obtain loans. Local development policy in the case of China's CCBs, however, seems to have achieved the opposite result.
We test our theory using data for 206,771 firms during 1999-2007, and find that large firms in cities with a CCB presence had easier access to credit, even if large firms on average were less efficient than their smaller counterparts in terms of return-on-assets (ROA), an economic efficiency measure. We also use the firm-level dataset from the 2005 World Bank Business Environment Survey, and specifically responses to the question of how much time top managers dedicate to interacting with government officials.The greater the amount of interaction, the higher the value we give to our crony relationship measure. We find that increased crony relations with government officials was associated with increased access to bank financing both in terms of improved chances of loan approval and larger loan amounts. This effect particularly strong in cities with a CCB presence. In cities with CCBs, the fact that firms tended to be larger on average may also evidence the relation of size to loan access.
The rest of the paper proceeds as follows. Section 2 offers a brief literature review. Section 3 introduces China's banking sector, especially CCBs. Section 4 describes the data and our methodology. Section 5 present the results. Section 6 develops a model to explain the results. Section 7 provides further evidence to test the mechanism. Section 8 concludes.

Literature review
There is a large body of literature on the effects of financial development on economic and firm growth.
In China's case, some studies find that the development of financial sector, including the banking sector is significantly and positively associated with economic growth (Ljungwall and Li, 2007;Zhang, Wang and Wang, 2012). Others find no significant relation, or even a negative relation, between the Chinese financial sector and economic growth (Boyreau-Debray, 2003;Chang, Jia and Wang, 2010). This negative relation may reflect distortions caused by a state-owned banking system reluctant to lend to SMEs (and even if SMEs are, in fact, key drivers of economic growth). For firm-level data, Ayyagari, Demirgüç-Kunt and Maksimovic (2010) and Allen, Qian and Qian (2005) also have different findings regarding the importance of formal or bank finance in China.
The CBRC says that CCBs were designed to lend to SMEs to take advantage of the small bank advantage in lending to small firms. CCBs operating solely in their own home cities for our observation period provide a natural experiment in whether these newly established CCBs positively affected local economic growth. From a bank competition viewpoint, CCB entry in a local banking sector should increase competition among banks and thereby boost the local growth rate.
Studies dealing with the finance and growth nexus have their theoretical roots in the work of Joseph Schumpeter (Shumpeter, 1934), who argued that finance contributes to growth because banks can identify and loan to the most innovative and promising firms. Financing of innovative firms promotes technological innovation and consequently promotes economic growth. A survey article by Levine (2005) lists the channels of finance contributing to economic growth. These channels include producing information and allocating capital, monitoring firms and exerting corporate governance, diversifying and managing risk, mobilizing and pooling savings, and easing the exchange of goods and services. Robinson (1952) challenged the Schumpeterian view, arguing that financial sector development follows economic growth. Lucas (1998) asserted that researchers overstate the role of finance in economic growth.
Goldsmith (1969) finds a higher level of financial development is associated with higher growth after investigating 35 countries from 1860 to 1963. This finding is confirmed by King and Levine (1993), who use four different financial development indicators and expand the number of countries to 77 for the period 1960-1989.
To deal with concerns about reverse causality of finance and growth, Levine, Loayza and Beck (2000) use a country's legal and accounting system as their instrumental variable and still find financial development led to economic growth for a sample of 71 countries for the period 1960-1995. Jayaratne and Strahan (1996 use a difference-in-difference estimation similar to the one we use here to study a wave of bank deregulation in the US. They find that the changes positively affected economic growth at the state level. Regarding studies of CCBs in China, Ferri (2009) demonstrates that the efficiency of CCBs strongly depends on provincial economic growth (without addressing the obvious endogeneity problem, however). Zhang, Wang and Qu (2012) find that CCB performance and risk-taking are positively related to provincial-level law enforcement, which, considering China's poor law enforcement, motivates our study on how CCBs affect local city growth and the business environment.
Che and Qian (1998) studies how local government ownership of firms can contribute to growth when firms escape the central government's grasping hand and manage to secure property rights despite China's weak property protections. Notably, ownership of banks that can channel deposits to private firms is distinguished.
Bank owners can easily instigate corrupt behavior in the private economy (which was already beginning to flourish in our sample period).

An introduction to city commercial banks
Urban credit cooperatives are the precursors to CCBs. The first urban credit cooperative was established in the city of Zhumadian in Henan Province in 1979. Table 1 shows the number of urban cooperatives from 5 1987 to 1998. There were 1,615 urban credit cooperatives in China at the end of 1987. That number increased to 5,229 at the end of 1994. 5 In the early 1990s, many urban credit cooperatives faced financial problems, including large stocks of nonperforming loans. In 1995, the State Council released a document to set up city cooperative banks (later redesignated city commercial banks) in 35 major cities through the merger and reorganization of existing urban credit cooperatives. In the same year, the first city cooperative bank was set up in Shenzhen. Some 60 cities  CCBs are quite different from urban credit cooperatives. First, urban credit cooperatives are not banks.
They are classed by the government as "cooperative financial organizations" and subject to the Regulation of Urban Credit Cooperatives, which e.g. sets strict deposit-taking and loan-issuing limits. Under the Regulation of Urban Credit Cooperatives, 9 "Deposits from non-cooperative members should not exceed 40% of all deposits, and deposits from any single individual non-cooperative members should not exceed RMB 150,000. Loans to any single client should not exceed RMB 500,000, and loans to non-cooperative members could not exceed 40% of all loans." Other regulations deny urban credit cooperatives access to the interbank market and the right to trade in government bonds or issue financial bonds. Such rules limit the abilities of urban credit cooperatives to serve local banking needs.

Data and methodology
We mainly use difference-in-difference model to estimate the effect of CCB establishment on city and firm growth. In addition to the manually collected data detailed below, we draw on the CEIC China premium database for city-level data, the China Annual Census of Enterprises for firm-level data, and various statistical yearbooks.

Model and variables
We estimate the following difference-in-difference model, where i and t denote city and year, α t and β i control time and city fixed effects. The dependent variable Economic Growth i,t is specified into two measures. The first is GRGDP i,t , which measures the GDP growth rate for city i at time t. The second is GRGDP P C i,t , which measures the GDP per capita growth rate. The key explanatory variable, CCB i,t , is a dummy indicating whether city i at year t owns a CCB or not, which is equal to 1 if yes, 0 otherwise. Following Berger et al. (2005), a dynamic time variable CCBY EAR i,t indicating the number of years CCB has been in the city is included to measure the long-term impact of the CCB.
Our control variables X i,t are as follows. LOAN i,t is the ratio of total loans in all local financial institutions to GDP. LnGDP (P C) i,t−1 , the logarithm of real local GDP (per capita) in previous year, is used to control for the economic convergence effect. F AI i,t is fixed asset investment divided by GDP. F DI i,t is total utilized foreign direct investment divided by GDP. F ISCAL i,t is the ratio of local government expenditure to GDP.
GRP OP i,t is the population growth rate. EDU i,t is percentage of population with secondary school education or higher. These variables are summarized in Table 3.
As macro growth paths for different cities vary, we also consider a nonlinear trend with the following specifications: where CCBEstablishY ear i denotes the number of years the CCB has existed city i (if there is a CCB).

Data
China has four levels of government administration: provincial level, prefectural level, county level, and village level. The provincial-level division includes 23 provinces, five autonomous regions, four municipalities and two special administrative regions (SARs). The four municipalities with provincial-level status are Beijing, Shanghai, Tianjin, and Chongqing. Provinces and autonomous regions are made up of prefectures. There were 284 prefectural-level cities in China at the end of 2011. 10 After dropping Lhasa, 11 our sample (see Table 4) consists of 283 prefectural-level cities and the four major municipalities. For convenience, "cities" here refers to both prefectural-level cities and municipalities. The prefectural-level city economic data is limited to years before 2001 as data for some of our key control variables are missing. Thus, the sample period chosen begins in 2001 and ends in 2011, the most recent year of available statistical data. After 2006, CCBs start to operate in cities other than their home cities. The CCBs venturing into the national market consist mainly of CCBs established in the first wave of the mid-1990s. In any case, only a few cross-city CCBs existed before 2008, and even then main operations remained focused on the home city. To study the lagged effect of CCBs, we extend our observation period through 2011.
Data for prefectural-level cities are from CEIC China premium database. Missing values are manually added from China Statistical Yearbook for Regional Economy and statistical yearbooks of provinces and prefectural-level cities. The descriptive statistics of macro variables are summarized in Table 5.
Some 157 of 284 prefectural-level cities and all four municipalities has CCBs at the end of 2011. After mergers and acquisitions, a total of 144 CCBs remain. The establishment year of CCBs is manually collected from public information, including local yearbooks, official websites and annual reports of CCBs. The dummy variable CCB is set to 1, the year after the CCB's establishment, because of the possible lag effect.

Micro: firm-level growth 4.2.1 Model and variables
A similar model to equation (1) is estimated to test the effect of establishment of a CCB on local firm growth, where i, j, t, denote firm, the city where the firm is located, and year, respectively, α t , and β i control year and firm fixed effects. Firm growth can be measured by GRSALES i,j,t and GRASSET i,j,t (see Table 6).
GRSALES i,j,t is annual growth rate of sales of firm i, located in city j in year t. GRASSET i,j,t is annual growth rate of asset of firm i in city j at year t. The key variable, dummy CCB j,t , indicates whether a CCB has already been established in city j at year t, and is equal to 1 if yes, and 0 otherwise.
Control variables Xi,j,t are listed as follows. STATECAP is the percentage of state-owned paid-up capital.
In addition, SOE is a dummy variable which indicates whether the firm is an SOE or not. It equals 1 if over 50% of the firm's shares are state-owned. The size of firm is controlled by ASSETi,j,t and SMEi,j,t . ASSET is the logarithm of the firm's total assets. SME is a set of two dummies classifying firms into three groups, small firms (fewer than 300 employees, annual sales of less than RMB 30 million, or total assets of less than RMB 40 million), medium-sized firms (fewer than 2,000 employees, annual sales of less than RMB 300 million, or total assets of less than RMB 400 million) and large firms. In addition, years the firm has been active is also used as 10 China Statistical Yearbook, 2011. 11 Lhasa is dropped due to limited statistical data. a control variable. Firms are divided into three groups according to their growth stages: start-ups (in business five years or less), growth phase (6 to 20 years in business), and mature (21 years or more).

Data
The data used for firm-level analysis are taken from the Annual Census of Enterprises produced by the Chinese National Bureau of Statistics from 1999 to 2007. The census includes all SOEs and non-SOEs with sales over RMB 5 million. It included 160,733 firms in 1999 and 335,076 in 2007. The data contains all the information from the three accounting statements (balance sheet, profit and loss, and cash flow). Only firms with records of at least four consecutive years are kept. Firms with total assets, total output, fixed asset, paid-in capital are 0 and total staff less than 8 (lack of credible accounting system) are dropped, and observations with sales growth rate and asset growth rate ranked in top and bottom 0.5% of the sample (16,231 out of 223,002) are dropped. In the end, our sample (see Table 7) consists of 206,771 firms from 40 industries 12 (mainly manufacturing) and 947,536 observations. The summary statistics of the sample are reported in Table 8.
As shown above, about three-quarters of firms are located in cities with CCBs. More than 90% of firms are non-SOEs and SMEs. As for firm ages, the majority of firms are in the growth phase of development.

Endogeneity
The endogeneity problem of CCBs in the above equation is hardly severe. After all, how likely is that a city government would set up a CCB for the purpose of lowering the city's growth rate? Of course, a CCB might be established to mitigate a potential slowdown in city growth, or there might be some omitted variables affecting growth rate and establishment of CCB simultaneously, such as city governance, all of which contribute to endogeneity problem for the above equations that we estimate. Using an instrumental variable (IV) method to solve the potential endogeneity problem, we see obvious clustering among CCBs in Figure 1. We find that for every province, CCB was first established at its capital cities, except five provinces which have one sub-provincial status city each. Based on the clustering observation and policy diffusion argument detailed by Simmons and Elkins (2004), we use the percentage of neighboring cities in the same province having established CCB to instrument for the CCB dummy in our regressions. We use neighboring cities in the same province as they are more likely to share the same financial and economic policies under the same provincial government leadership.
We run a two-stage regression for our endogeneity problem. The first stage takes the following form, while the second stage is the main equation described above.
where i and t denote city and year respectively. CCB i,t is what we used in equation (1) , the key explanatory variable. N eighbor i,t is the percentage of neighboring cities in the same province having established CCBs by year t. Alternatively, we use the percentage of cities that had established CCBs by year t.
5 Empirical results 5.1 City commercial bank and city macro performance: city-level data 5.1.1 City commercial bank establishment and city GDP growth We can see from Table 10 that the effect is stronger for urban core growth rate, where CCB mainly operated in the core city area. We can also see from Tabel 11 that our negative results still hold under alternative time trend specifications.
The possible lag effect has already been taken into account when constructing the dummy variable CCB, it could conceivably take longer than a year for CCB effects to emerge. Thus, we also include the lagged values of CCB to test for this possible longer lag effect. The sample period is still 2001 to 2011 and the lagged CCB is obtained manually based on the number of years of the CCB has existed.  -2009, 1999-2008, 1998-2007, 1997-2006 and 1996-2005 are set as the experimental group. As a result, nearly all cities with CCBs are covered in experimental group among these regressions (only the Shenzhen and Shanghai CCBs were established before 1996). The results still show a significant negative relation between lagged CCB and local GDP growth rate for different lagged-effects and experiment group members.
We next use the GDP per capita growth rate instead of GDP growth rate as our dependent variable. Table   13 shows the results. We can see that establishment of CCBs (CCB) and their lagged effects (CCB −1 , CCB −2 , CCB −3 ) are all significantly negative with regard to the GDP per capita growth rates of cities. We also replace the dummy variable CCB with CCBYEAR in Reg2 to test the long term effect of CCB, as one might conjecture that there might be a learning curve for CCB to be effective, and the results hold.

CCBs and firm growth: micro-level evidence
Tables 14-17 show how establishment of a CCB affects the growth rates of local firms. We use data from the China Annual Census of Enterprises from 1999 to 2007, which includes 206,771 firms with at least four years of consecutive observations. The Hausman test strongly proves the fixed effect model, and we use a pooled regression model as our robustness check.
From Table 14, we see firm growth rates significantly declined after CCB establishment, which is robust to the commonly used controls including firm characteristic variables and in all the regression (controlling for both year effects and firm fixed effects). Compared to firms in cities without CCBs, the establishment of a CCB results in a decline in the growth of firm of around two percentage points, or about a third of standard deviations for the sales growth rate.
Other variables have the expected signs, including negative signs for SOE and STATECAP that indicate government intervention has negative consequences. Both size dummy and logarithm of firm assets indicate that large firms enjoy a higher sales and asset growth rate, possibly because of greater amounts of available capital and easier access to banking finance than for small firms. Firms in the start-up phase have a higher sales and asset growth rates than firms in the growth and mature phases. Table 15 replaces sales growth rate with asset growth rate of firms as the dependent variable and generates the same result from all the four regressions, i.e. CCB establishment has a significant negative effect on the firm asset growth rate.
While it is clear that the political justification for CCBs of creating local banks that would contribute to SME growth better than the nationwide bank has not been realized, it is also clear that impact of CCBs on firms depends on firm size. Thus, we divide firms into SME and large firm groups, and then divide the SME group into small and medium sized groups. We can see from Table 16 that SMEs and small firms experienced lower growth rates in the cities with a CCB.
To check the robustness of our results, a regression using pooled OLS with same observations is estimated and similar results are obtained (Table 17). Industry and regional (province) dummies are added to control the possible industrial and regional specification. Our results that a CCB presence leads to an approximately two percentage point decline in the firm sales growth rate and a one percentage point drop in the asset growth rate remain quite robust. Table 18 shows the heterogeneous impact on firm growth of a CCB presence, where CCBSM ALL = CCB * (DummySmall). We can see that the establishment of a CCB reduces small-firm growth rates most.

Heterogeneous impact of CCB presence on firm growth
We suggest that when CCB lending policies are dictated by city officials, the tendency to prefer large firms is enhanced for several reasons. First, SOEs tend to be large firms with political connections to the city government.
Second, city officials may prefer large private firms as lending targets as they can provide sufficiently large bribes or because city officials have their personally connected through such favors as providing jobs for their family members. Finally, city officials might want to advance their careers or the city's reputation by attracting large brand-name firms to the city.

Endogeneity
Although, as mentioned earlier, it is hard to imagine why a local government might want to establish a CCB to lower the local economic growth rate, some endogeneity issues might exist. We adopt the neighboring IV estimation as mentioned in the methodology section.
The first stage regression (Table 19) displays strong predictive power of establishing CCB when there are more CCBs in neighboring cities. We use the percentage of neighbors with their own CCB, i.e. N EIGHBOR, in our first two regressions. We then use the percentage of cities in a province have established their own CCB, i.e. P ROV P ERCEN T , as an alternative measure. Groups are defined at the provincial level as cities are under the administration of provincial government officials.
We can see from the second-stage regression (Table 20) that CCB still has a negative impact on a city's GDP and GDP per capita growth rate. We therefore conclude that the negative impact on growth from establishing a CCB is quite robust.

A simple model of crony relations
Our model of crony relations is based on Bai, Hsieh and Song (2014). As the local government controls the bank, firms pay bribes to local politicians to obtain credit. Larger firms, which have more total assets than small firms, can pay bigger bribes and receive preferential treatment compared to small firms. Moreover, as officials caught taking bribes face punishment, the size of the bribe has to be big enough for the official to justify taking the risk. However, due to decreasing return to scale on capital, larger firms are less efficient than smaller firms. Therefore, the existence of a local bank that indulges in crony lending distorts the firm size distribution in that city through less-than-optimal allocation of credit. Small productive firms suffer and the overall growth rate of the local economy is reduced.
Nationwide banks are more efficient in credit allocation as politicians in central government face greater punishment and thus higher risk for bribe-taking than their local government counterparts. Indeed, there is considerable anecdotal evidence that local government officials in China tend to be more corrupt than central government officials. Therefore, nationwide banks, especially the Big Four banks controlled by the central government, are likely to restrain their crony relations with select large firms, making these nationwide stateowned banks generally more efficient than local banks.
We assume that as long as firms pay a sufficiently large bribe, the minimum loan any firm can obtain is I.
There are a continuum of firms [0, 1], with firm size uniformly distributed on U [0, K].
Government i's problem can thus be stated as i denotes city, W denotes total welfare, f denotes firms, ϕ denotes the fraction of output contributed to bribe, α < 1 denotes decreasing return to scale, K firm f in city i's total asset, I if denotes firm f 's total credit obtained from local bank, G(i) denotes public goods consumption. B denotes the government's budget, δ denotes the substitutability between government spending and local bank subsidy and δ >> 1 indicating that subsidizing the local bank is quite costly. The first budget constraint indicates the government budget balance.
The second indicates total loans obtained by firms in city i is equal to the bank's total credit. For the last constraint, F denotes the risk of punishment for accepting bribes. The left hand side is the bribe obtained from 13 firm f . The last constraint indicates the bribe must be sufficiently larger than the risk of being caught and punished.
FOC solutions for f ∈ [f H , 1] are So we obtain Proposition: ∂Ii ∂Fi < 0. A lower value for F i , i.e. a higher level of corruption tolerance, encourages crony lending and more corruption, i.e. I i , leading to lower growth. Nationwide banks, in turn, do a better job of promoting growth as they are more selective in their crony relations and face a higher value for F . This proposition has the implication that local government controlled banks lead to more crony lending which consequently led to lower growth rate.
The social planner's problem can be stated as Proposition: Output is higher under the social planner's problem when crony lending is lower, i.e. Y s > Y c .

Moreover, Ys
Is > Yc Ic . We suggest from this proposition that under imperfect institutions, distortions created by local bank entities are larger than central government controlled banks.

Crony relations and bank behavior
We start this section with further assessment of our firm data for 1999-2007. As Table 21 shows, large firms obtain more benefits when there is a CCB operation in their city, even when they are less efficient in terms of a lower return on assets (ROA) than their smaller counterparts. Regression 1 indicates that, on average, firms overall are less likely to get bank loans in cities with a CCB presence. While we might expect a city with more banking institutions to have more lending overall and consequently more lending to firms, it appears that only large firms overcame this negative effect and obtained more loans. This could be explained by a CCB focus on lending to large firms that stems from the fact that they are structurally forced to compete with Big Four banks for deposits. Those that fail to attract deposits are left with less to lend. Regression 2 shows that cities with CCBs have less firm investment overall, but large firms overcome this negative effect. Regression 2 is a direct a consequence of Regression 1. Regression 3 shows that the assets of the top 100 firms in cities with CCBs grew faster on average than firms in cities without CCBs. Regression 4 shows that, even though these large firms obtained loans, they had lower ROAs than their smaller counterparts.
To test whether firms with more crony relations with government officials have a higher probability of obtaining loans or obtain a disproportionally large amount of loans compared to other firms, we use the widely adopted World Bank Business Environment Survey data conducted in 2005 in 120 Chinese cities. The survey involved face-to-face interviews with top managers and business owners, and used the Global Methodology and uniform sampling. The responses were anonymous to bolster their reliability and authenticity. The survey is designed to reveal characteristics of the regional business environment. It includes categories such as regulations and taxes, corruption, crime, informality, gender, finance, infrastructure, innovation and technology, trade, workforce, firm characteristics, and obstacles to performance. Not only are the data quite representative, but 2005 is a good year for our purposes as CCBs had been around for a decade, so the effects from the presence of CCBs had accumlated but few CCBs had yet to launch operations outside their home cities.
For our key crony measure, we use the following survey question: How many days does the GM or Vice GM spend on government assignments and communications per month? (Government agencies include e.g. the Tax Administration, Customs, Labor Bureau, and Registration Bureau. Assignments refer to handling the relationship with government officials, consolidating and submitting various reports and statements, etc.).
We estimate the following equations to test our model prediction.
where i, j denote firm and located city respectively. There are two measures for variable Loan i,j . One corresponds to a dummy variable question regarding whether firm i has obtained loans: Does your company have loans from banks or other financial institutions? (1) Yes (2) No. The other corresponds to the interest payments firms made to banks in 2014. We will use a Probit model to estimate the first and simple OLS regression to estimate the second. Crony i,j corresponds to the above survey question regarding how many days a month top managers in firm i interact with government officials in city j. X are control variables including firm age, share of state ownership, share of foreign ownership, export status, and production capacity.
From Table 22, we can see how increased crony relations alter the firm size distribution through bank loan access. Since local officials can influence CCB lending, we can see from Regressions 1 and 4 that firms that spend more time interacting with officials (i.e. have a higher crony index value) have a higher chance of obtaining bank loans, and from Regressions 2 and 5 that firms with a higher crony index obtain a disproportionately large amount of bank loans. These effects are strongest for firms in cities with a CCB presence. We also find that a high crony index value correlates with larger firm size. This effect was also strongest in cities with a CCB 15 presence.

Conclusions
Using panel data for all major Chinese cities (except Lhasa) from 2001 to 2011, CCB presence was found to have a negative effect on local economic growth (even if the CBRC's stated purpose for creating CCBs was to enhance local growth by giving local SMEs better access to credit). Moreover, firm-level data for 206,771 firms from 1999 to 2007 show that CCBs had a negative effect on firm growth, particularly in the case of small firms.
Both results are quite robust. Considering the possible endogeneity problem, we borrowed from the policy diffusion literature using the percentage of neighbor cities that established CCBs as the instrument variable and found the results to still be negative.
To explain this outcome, we argue that local governments, which are directly and indirectly involved with C-CB operations, may be open to bribes. Large firms can provide adequately large bribes to local government officials to entice them to risk sanctions. The resulting inefficient lending, in turn, reduces growth. We then investigate our theory using two datasets. The first is comprehensive and disnguishes firms by size. The second is a representative sample to which we attach a firm-level crony relations measure. We find that in cities with a CCB, large firms enjoyed disproportionately easier access to bank credit even with lower ROAs than their smaller counterparts. Crony relationships with government officials facilitated firm access to bank financing and large firms tended to be more likely to cultivate crony relationships.   , 1987, -1998, Year 1987, 1988, 1989, 1990, 1991, 1992 Number of urban credit cooperatives 1,615 3,265 3,409 3,421 3,518 4,001 Year 199319941995199619971998 Data source: Almanac of China's Finance and Banking, 1990Banking, -1999               Standard errors robust to heteroskedasticity and autocorrelation are in parentheses. *, ** and *** denote significance at 10%, 5% and 1% levels respectively. 30 Standard errors robust to heteroskedasticity are in parentheses. *, ** and *** denote significance at 10%, 5% and 1% levels respectively. Standard errors robust to heteroskedasticity are in parentheses. *, ** and *** denote significance at 10%, 5% and 1% levels respectively. Standard errors robust to heteroskedasticity are in parentheses. *, ** and *** denote significance at 10%, 5% and 1% levels respectively.    Note: Standard errors robust to heteroskedasticity are in parentheses. *, **, and *** denote significance at 10%, 5%, and 1% levels, respectively. Note: Standard errors robust to heteroskedasticity are in parentheses. *, **, and *** denote significance at 10%, 5%, and 1% levels, respectively.   Note: Standard errors robust to heteroskedasticity are in parentheses. *, **, and *** denote significance at 10%, 5%, and 1% levels, respectively. is conjectured to have more bank loans is more productive. Standard errors robust to heteroskedasticity and autocorrelation are in parentheses. *, ** and *** denote significance at 10%, 5%, and 1% levels, respectively. 38