Banking Paper May 2018

How Does Greater Bank Competition Affect Borrower Screening? Evidence from a Natural Experiment Based on China’s WTO Entry Thomas J. Chemmanur∗

Jiaqi Qin† Yan Sun‡ Xiang Zheng¶

Qianqian Yu§

May 2018

Abstract Using banking sector reforms that accompanied China’s accession to the WTO and a large sample of Chinese private firms, we empirically analyze the relation between greater bank competition and the screening of potential borrowers. Our results are summarized as follows. First, the sensitivity of bank credit to prior borrowing firm performance increases after China’s entry to the WTO. This increase in sensitivity is larger in more bank-dependent industries and smaller in Chinese regions with a higher level of financial development. Second, the increase in the sensitivity of bank credit to firm performance is much greater for state-owned firms compared to private firms without state ownership. Third, the effect of bank credit on subsequent firm productivity and performance is greater for loans given after China’s WTO entry compared to those given prior to WTO entry. Overall, the results of our empirical analysis suggest that the stringency of bank screening in China increases with greater competition in the banking sector.

JEL Classification: G21; G32 Keywords: Bank Competition; Borrower Screening ∗

Professor of Finance and Hillenbrand Distinguished Fellow, Finance Department, Fulton Hall 336, Carroll School of Management, Boston College, Chestnut Hill, MA 02467, Tel: (617) 552-3980, Fax: (617) 552-0431, Email: [email protected]. † Professor of Finance, Department of Financial Management, Business School, Nankai University, Tel: +86-22-23509480, Email: [email protected]. ‡ Associate Professor of Accounting, John Cook School of Business, Saint Loius University, St. Louis, MO 63108, Tel: (314) 977-3818, Email: [email protected]. § Assistant Professor of Finance, Perella Department of Finance, Lehigh University, Bethlehem, PA 18015, Tel: (610) 758-2959, Email: [email protected]. ¶ Ph.D. Candidate, Finance Department, Fulton 154B, Carroll School of Management, Boston College, Chestnut Hill, MA 02467, Tel: (617) 552-2049, Email: [email protected].

1

Introduction

An important economic role of financial intermediaries such as banks is to allocate capital investments optimally to firms in an economy through efficient screening of borrowers. An interesting question in this context is how the efficiency of screening by banks changes as the banking sector becomes more competitive. The theoretical banking literature has been surprisingly ambiguous regarding this question. For example, symmetric information models such as Klein (1971) predict that, when there is less competition in the banking sector, banks charge higher interest rates leading to a decrease in loan supply. Models incorporating asymmetric information, however, often yield the opposite prediction. In particular, the model of Petersen and Rajan (1995) shows that greater competition in the credit market imposes constraints on the ability of a borrowing firm and the lender to intertemporally share surplus by maintaining lending relationship, leading to the prediction that only higher quality firms will receive loans as the banking sector becomes more competitive.1 In the model of Broecker (1990), as bank competition increases, potential borrowers are able to pass banks’ test more easily, so that the average creditworthiness of borrowers decreases. More recently, Marquez (2002) shows that greater bank competition leads to greater dispersion of information about potential borrowers, thus reducing banks’ screening ability. Several papers in the literature have attempted to address the above question empirically. Some studies have used bank-level data to analyze the effect of bank deregulation and mergers on small business lending (see, e.g., Berger, Saunders, Scalise, and Udell (1998) and Sapienza (2002)). While these studies provide evidence on how banks change their loan portfolio in response to bank deregulation and mergers among banks, they do not provide evidence on the effect of greater bank competition on borrower screening. More recently, two studies 1

The model of Boyd and De Nicolo (2005) generates a similar prediction but through a risk-shifting channel. They introduce a loan market into a setting broadly similar to that of Allen and Gale (Chapter 8, 2000). In their setting, less competition in the banking sector increases the interest rates charged to borrowers, so that profits of borrowers go down, inducing them to seek more risk (i.e., engage in risk-shifting).

1

have empirically analyzed the real effects of greater bank competition. The first study is Bertrand, Schoar, and Thesmar (2007), who investigate how deregulation of the French banking industry in the 1980s affected the real behavior of borrowing firms. The second study is Zarutskie (2006), who analyzes the financial and real effects of bank competition by studying the firm level effects of bank deregulation (the Riegle-Neal Interstate Banking and Branching Efficiency Act of 1994) in the U.S. However, despite such empirical analyses focused on developed economies, the effect of greater banking competition on firm-behavior remains ambiguous, especially in the context of developing economies with a large fraction of firms in the state-owned sector. The objective of this paper is to fill the above gap in the literature. We analyze the real effects of greater bank competition by studying the effect of China’s entry into the World Trade Organization (WTO) in 2001 on bank lending. We make use of a large sample on private firms collected through annual surveys by the National Bureau of Statistics (NBS) in China. Subsequent to China’s WTO entry, the banking sector in China became much more competitive, so that this is a natural setting to study the effect of greater bank competition on their screening behavior of borrowers. China is not only the second largest economy in the world but also a developing country with high economic growth in the past 30 years. Understanding how bank competition plays a role in the Chinese highgrowth environment may generate implications that may be useful in understanding the effect of greater bank competition in other developing countries as well. Moreover, compared to developed economies, China also has the rather unique feature of having a very high portion of the economy controlled by state-owned enterprises (SOEs): it is interesting to study how greater competition in the banking sector affects bank lending to SOEs in China. Our findings can be summarized as follows. First, we find that the sensitivity of bank credit to borrowing firm performance increases after China’s entry to the WTO: i.e., greater bank competition leads to more stringent screening of borrowers. Further, this increase in sensitivity is larger in industries with a higher level of bank dependence and smaller in regions

2

with a higher level of financial development. Second, the sensitivity of bank credit to SOE performance is significantly greater after China’s entry to WTO: this increase in sensitivity is much greater for lending to SOEs compared to that for firms without state ownership. Third, the effect of bank credit on subsequent firm productivity and performance is greater for loans given after China’s WTO entry compared to those given prior to WTO entry. This result also supports the notion that the screening effectiveness of banks goes up after China’s WTO entry (as competition in the banking sector increases). These empirical results are broadly consistent with the implications of the theoretical model of Petersen and Rajan (1995), in the sense that the average quality of borrowing firms in China goes up as competition in the banking sector becomes greater. Our paper is related to two different strands in the empirical banking literature.2 The first strand is the literature that examines how changes in bank competition affect borrower behavior. Bertrand, Schoar, and Thesmar (2007) present evidence that, following bank deregulation, French banks were less willing to bail-out poorly performing firms, and firms in the more bank-dependent sectors were more likely to undertake restructuring activities. Zarutskie (2006) analyzes bank deregulation in the U.S. in the 1990s and shows that newly formed firms, characterized by higher informational asymmetry, use less bank credit and invest less when bank competition increases, while such negative effects are gradually reduced as firms become older and finally reverse sign.3 Cetorelli (2004) shows that deregulation of the banking industry in EU countries enhances small firm entry into nonfinancial industries.4 Black and Strahan (2002) study the effects of banking deregulation on entrepreneurship, and 2

Our paper is also related, though more distantly, to the literature documenting the effects of other financial intermediaries such as venture capitalists on the performance of private firms: see, e.g., Chemmanur, Krishnan, and Nandy (2011). This literature has shown that venture capital financing provides firm help to make their firm more efficient. 3 One paper that studies the effect of foreign bank entry into China is Lin (2011). Lin (2011), however, focuses on public firms. The findings in his paper suggest that less opaque firms and non-state-owned firms benefit more from foreign bank entry. In a similar vein, Gormley (2010) studies the impact of foreign bank entry into India after a policy change. He finds that, on average, firm are less likely to get credit after the policy change but profitable firms are more likely to secure bank credit. 4 However, Beck, Demirgüç-Kunt, and Maksimovic (2004) provide international evidence showing that more banking competition reduces financial obstacles for small firms only in countries with low levels of economic and institutional development.

3

show that the rate of new incorporations increases following the deregulation of branching restrictions.5 Our analysis complements the above literature by empirically analyzing the effects of greater bank competition on bank screening in a high growth developing economy such as China even in the context of developing economies like China.6 Our results suggests that bank screening becomes more stringent after bank competition increases. Further, we show that this effect is more pronounced for state-owned firms and firms in the more bankdependent industries.7 Second, this paper contributes to the literature on the relationship between financial development and bank competition. It’s well-accepted that financial development affects economic growth (see, e.g., Jayaratne and Strahan (1996) and Rajan and Zingales (1998)). But the relationship among financial development, bank competition, and the efficiency of bank screening has not been studied before in the literature.8 We contribute to this literature by showing that the effect of bank competition on borrower screening is weaker in Chinese regions with a higher level of financial development. The remainder of the paper is organized as follows. Section 2 briefly describes the institutional details of the banking reform in China. Section 3 discusses the underlying theory and develops testable hypotheses. Section 4 presents our data and sample selection proce5

Cetorelli and Strahan (2006) show that, in markets with concentrated banking, potential entrants face greater difficulty gaining access to credit than in markets in which banking is more competitive. Krishnan, Nandy, and Puri (2014) use the Longitudinal Research Database (LRD) of the U.S. Census Bureau to analyze how increased access to bank financing following interstate banking deregulation affects the total factor productivity (TFP) of entrepreneurial firms. Kerr and Nanda (2009) use the Longitudinal Business Database (LBD) of the U.S. Census Bureau to analyze how U.S. branch banking deregulation impacts entrepreneurship rates and incumbent firm displacement. 6 One paper that also looks China’s entry into the WTO as a plausibly exogenous shock to China’s banking system is Qian, Strahan, and Yang (2015). Unlike our paper, they use China’s entry into the WTO as an exogenous shock to loan officer’s incentives to produce information rather than to the competitiveness of China’s banking sector and use proprietary data from a single large Chinese state-owned bank to analyze how such incentives affect ex ante loan pricing and future loan performance. 7 Our paper is also related, though more distantly, to the theoretical and empirical literature on firms’ choice between bank loans and public debt. See, e.g., Diamond (1989) and Chemmanur and Fulghieri (1994) for theoretical analyses; see, e.g., Hoshi, Kashyap, and Scharfstein (1990) for empirical evidence. 8 A recent paper that argues that China is an important exception to the finance-growth nexus (since it has experienced rapid economic growth over three decades despite an inefficient banking sector) is Lin et al. (2015). They try to unravel the puzzling relationship between financial development and economic growth in China by specifying an empirical model to disentangle the effect of bank ownership structure and size structure.

4

dures, and variable definitions. Section 5 presents our empirical tests and results. Section 6 concludes.

2

Banking Reform in China and China’s WTO Entry

The banking system in China has experienced several structural changes since the Chinese government decided to initiate the economic reform of 1978 (Guariglia and Poncet (2008)), where the objective was to transform the sector from being state-owned and monopolistic to a multi-ownership and competitive system. These reforms can be categorized into several phases. Initially, the People’s Bank of China (PBC) not only served as the central bank but also performed the function of a commercial bank. PBC still served as the central bank after 1978, but its functions as a commercial bank were replaced by four state-owned commercial banks: the Bank of China (BOC), the Agriculture Bank of China (ABC), the Construction Bank of China (CBC), and the Industrial and Commercial Bank of China (ICBC). These four banks were authorized and designated to provide separate services for different sectors and began to compete with each other after 1985. The Chinese government recapitalized the above four state-owned commercial banks and many other commercial banks were established, leading to greater competition in banking sector starting from 1994. China was officially accepted as a member of the WTO on December 11th 2001, which marked a new phase in China’s reform of its banking sector to prepare for foreign competition. At the time of WTO entry, the Chinese government agreed to a five-year transition period in which they would enact additional reforms annually to ease restrictions on the banking sector in order to fully comply with WTO rules by the end of 2006.9 To prepare for the complete opening of the banking sector to foreign banks, domestic banks expanded substantially at the level of branches. Therefore, it is likely that bank 9

Berger, Hasan, and Zhou (2009) analyze the efficiency of Chinese banks over 1994 – 2003 and argue that minority foreign ownership is conducive to increased Chinese bank efficiency.

5

competition became stronger than before with the rapid expansion of bank branches. This is also supported by the evidence in various papers using other proxies to measure bank competition. For example, Liang, Xu, and Jiraporn (2013) uses the Profit Elasticity (PE) approach to measure the level of bank competition in China during 1996-2008, and finds that competition increased during this period, especially after China’s accession to the WTO. In addition to the increase in bank branches, China undertook several reforms to improve the competitive advantage of its banking sector after entering the WTO. First, China established the China Banking Regulatory Commission (CBRC) on April, 2003, which has 36 bureaus in 31 provinces and 5 special cities. One of the main objectives for establishing the CBRC was to maintain the legal and steady operation of the banking sector through uniform and effective supervision. Therefore, the CBRC was expected to assess, monitor, and cut down the overall risk of the banking sector, leading banks to keep effective internal controls, encouraging fairer competition among banks, and enhancing the international competitiveness of domestic banks. Second, the four state-owned commercial banks were successfully transformed from being wholly state-owned banks to shareholding companies through recapitalization and IPOs. According to the statistics provided by Leigh and Podpiera (2006), the Construction Bank of China (CBC) allowed the Bank of America to hold 8.5% of its shares and Temasek to hold 6% of its shares as strategic investors. The Bank of China (BOC) introduced a consortium led by the Royal Bank of Scotland to hold 9.6% of its shares, UBS to hold 1.6% of its shares, and Temasek to hold 4.8% of its shares as strategic investors. In 2006, the Industrial and Commercial Bank of China (ICBC) introduced Goldman Sachs, Allianz, and American Express as strategic investors, holding a combined 8.5% of its shares. Finally, all of the above banks made IPOs of their shares toward the end of 2006 on the Hong Kong Stock Exchange. The third reform introduced by China was decentralization: imposing greater responsibilities on individual loan officers in charge of different steps in the lending process. Before 2002, each step of the lending process was conducted without a clear designation of individual

6

responsibilities, where a free-rider problem could easily arise. After 2002, each of the lending steps within a branch were divided into five clearly defined roles: (initial) investigation; verification; deliberation and discussion; approval; and post-loan monitoring. This reform in the lending process was expected to improve loan quality by increasing the incentives of loan officers to exert more effort to produce hard as well as soft information relevant to the loan. The above reforms were not triggered by any specific group of firms but rather were aimed at improving the competitiveness of all large state-owned banks ahead of the entry of foreign competition in the banking sector. In summary, these reforms provide a plausibly exogenous shock to China’s banking sector, which we will exploit in our empirical analysis of the effects of greater bank competition on the efficiency of borrower screening.

3

Theoretical Background and Testable Hypotheses

The theoretical banking literature has diverging implications for the effect of greater bank competition on loan quality. One strand in the literature argues that a higher level of bank competition will lead to more stringent screening, thereby increasing the average quality of loans made by them. In particular, the asymmetric information model of Petersen and Rajan (1995) shows that greater competition in the credit market imposes constraints on the ability of a borrowing firm and the lender to intertemporally share surplus by maintaining lending relationship. This leads to the prediction that only higher quality firms will receive loans as the banking sector becomes more competitive.The model of Boyd and De Nicolo (2005) generates a similar prediction but through a risk-shifting channel. They introduce a loan market into a setting broadly similar to that of Allen and Gale (Chapter 8, 2000). In their setting, less competition in the banking sector increases the interest rates charged to borrowers, so that profits of borrowers go down, inducing them to seek more risk (i.e., engage

7

in risk-shifting).10 Overall, the above papers suggest that a higher level of bank competition leads to more stringent bank screening (H1A). On the other hand, Marquez (2002) argues that with more competing banks, borrowerspecific information becomes more dispersed, since each bank becomes informed only about a smaller pool of borrowers. This reduces banks’ screening ability, resulting in more low-quality borrowers obtaining financing. Broecker (1990) develops a model in which banks compete by announcing the interest rates at which they are willing to provide credit to those applicants who pass the banks’ tests. The proportion of applicants who pass the test of at least one bank increases with the number of banks providing credit, so that the average credit-worthiness of borrowers decreases as a result of greater bank competition. Freixas, Hurkens, Morrison, and Vulkan (2007) extend the analysis of Broecker (1990) and show that an increase in the number of active banks increases credit risk, thereby lowering loan quality. Cetorelli and Peretto (2012) and Petriconi (2015) also show that bank competition decreases the incentive of banks to screen, leading to lower loan quality and loan standards. Overall, this strand of the theoretical banking literature suggests that a higher level of bank competition leads to less stringent bank screening (H1B). The effect of bank competition, either positive or negative, will be amplified for industries that are more dependent on bank credit. Therefore, if H1A holds, we would expect the positive effect of bank competition on bank screening to be stronger in industries with a higher level of bank dependence (H2A). On the one hand, if H1B holds, we would expect the negative effect of bank competition on bank screening to be also stronger in industries with a higher level of bank dependence (H2B). Since the level of financial development is significantly different across geographic regions in China, we would expect that the relationship between bank competition and loan screening to vary across regions based on their level of financial development. For example, when the 10

Chen (2005) argues that greater bank competition caused by banking market liberalization results in a lower opportunity cost of screening loan applicants. This would induce banks facing entry threats to invest in better screening instead of relying on collateral requirements, thereby improving loan quality.

8

region is more financially developed, there will be more non-bank financial intermediaries that firms are able to borrow from. Therefore, we would expect the effect of bank competition to be weaker for regions that are more financially developed. Consequently, if H1A holds, we would expect the positive effect of bank competition on bank screening to be weaker for firms in regions with a higher level of financial development (H3A). On the other hand, if H1B holds, we would expect the negative effect of bank competition on bank screening to be also weaker for firms in regions with higher levels of financial development (H3B). It is well documented that state-owned enterprises (SOE) in China are very inefficient in spite of the large amounts of implicit subsidies they have received in the form of lowinterest loans (Lin, Cai, and Li (1998), Bai, Lu, and Tao (2006)). On the one hand, greater bank competition and concurrent structural reform of state-owned banks may largely reduce subsidized loans and lead to much more stringent screening for SOEs. If this is the case, we would expect the positive effect of bank competition on bank screening to be stronger for SOEs (H4A). On the other hand, if greater bank competition leads to less stringent screening, then SOEs will be the favored clients for banks to compete for providing loans as competition increases. If this is the case, we would expect the negative effect of bank competition on bank screening to be stronger for SOEs (H4B). We now turn to the effect of bank competition on subsequent borrowing firm performance. If greater bank competition indeed leads to more stringent bank screening (i.e., if H1A holds), then we would expect better investment projects to get funded, which will subsequently lead to better firm performance (H5A) for firms that received funding from banks under the more competitive banking environment. On the other hand, if bank competition leads to less stringent bank screening (i.e. if H1B holds), then we would expect the opposite to be true, that is, worse investment projects will get funded as bank competition increases, which will subsequently lead to worse firm performance (H5B).

9

4

Data and Summary Statistics

Our main data source is the National Bureau of Statistics (NBS) database of China, which is similar to the Longitudinal Research Database (LRD) maintained by the U.S. Bureau of the Census.11 Financial development data as measured by the index of financial industry competition comes from Fan, Wang, and Zhu (2003). Our firm-level sample covers the period from 1999 to 2007. Our sample ends in 2007 in order to avoid any potential contamination effects of the global financial crisis which began in 2008. We construct a dummy variable, Af ter, to capture the effect of increased bank competition as a result of China entering WTO. Af ter is equal to one for observations from 2003 to 2007 and zero otherwise.12 We apply the following filters to the database to construct our sample (see Table A.1 in the Appendix for more details regarding our sample construction). First, we drop firms in non-manufacturing industries as our primary focus is on firms in manufacturing industries. Second, we drop firms controlled by persons or entities from Hong Kong, Taiwan, Macao, or foreign countries, as they are less affected by China’s banking competition. Third, we drop conglomerates from our sample, since they are likely to be self-financed through internal capital markets and thus less affected by bank competition (see, e.g., Zarutskie (2006)). Fourth, we drop public firms and venture capital-backed firms, since they have more access to capital and are thus less affected by bank competition (see, e.g., Chemmanur, Krishnan, and Nandy (2011) or Zarutskie (2006)). Finally, we drop firms with insufficient information. To mitigate the effects of outliers, we winsorize all continuous variables at the level of 1% and 99%. Our results are robust to winsorization. Table 1 reports our final sample distribution by year and industry. To measure the 11

See Brandt, Van Biesebroeck, and Zhang (2012) and Brandt, Van Biesebroeck, and Zhang (2014) for a more detailed discussion of this database. 12 Since it takes time for accession to WTO to be effective on bank competition, we choose 2003 as the cutoff year.

10

availability of bank credit, we construct a dummy variable, Availability, which is equal to one if the change of a firm’s long-term debt is positive and zero otherwise.13 The percentage of firms receiving bank credit (i.e., with Availability = 1) exhibits a decreasing trend from 1999 to 2007. The maximum fraction of firms receiving bank credit for a given year is 23.2% in 1999. Industries with the largest fractions of firms receiving credit include tobacco, pharmaceutical, and alcohol, beverages, and refined tea industries. Table 2 reports the summary statistics for main variables used in our empirical analyses. We define amount of bank credit, Amount, as the logarithm of change in long-term debt, where we require the change in long-term debt to be greater than zero. The price of bank credit is determined by two variables: interest cost (loan spread) and transaction cost.14 We define Interest Cost, T ransaction Cost, and T otal Cost of bank credit, as the ratio of interest payment, total financial expense minus interest payment, and total financial expense to total debt. As reported in Table 2, 14.4% of our sample firms received bank loans during the sample period. From the results of univariate comparisons of all variables before and after 2002, we find that the percentage and amount of bank loan that firms receive went down after 2002. Further, Interest Cost, T ransaction Cost, T otal Cost, ROA, ROAM A , Investment, SalesGrowthM A , and P roductivity are significantly higher after 2002. However, the level of Ln(Asset), Leverage, Ln(Age), and F ixedAsset decreased significantly after 2002. 13

We assume that a firm receives bank credit in year t if and only if the change of its long-term debt is greater than zero. One may be concerned that the change of long-term debt does not necessarily come from bank credit, as long-term debt includes credits from other financial institutions (e.g., investment companies) as well. However, as Cull, Xu, and Zhu (2009) point out, private firms in China have little access to long-term debt from non-bank institutions, which supports the validity of our measure. 14 PBC sets the loan spread by specifying a target credit growth for each bank. Banks are allowed to offer interest rates only within a small range under PBC regulations so that we also study the transaction cost in additional to interest cost (Chen, Liu, and Su (2013); Jiang, Jiang, Huang, Kim, and Nofsinger (2017)).

11

5

Empirical Tests and Results

5.1

The Relation between Bank Competition and Banking Screening

To test our key hypotheses H1A and H1B, we run the following regressions:

A MA 0 yi,t = α + β1 ROAM i,t−1 + β2 Af ter + β3 Af ter × ROAi,t−1 + φ Xi,t−1 + γt + ηj + δk + i,t . (1)

In the above, the dependent variable yi,t includes the availability, amount, and various cost of bank credit.15 We use the moving average of return on assets (ROA) in the past three years to measure firm performance, and denote it as ROAM A .16 Our control variables (Xi,t−1 ) include the following: Ln(Asset), defined as the natural logarithm of firm assets; Leverage, defined as the ratio of debts over assets; F ixed Asset, defined as the ratio of fixed assets over total assets; and Ln(Age), defined as the natural logarithm of firm age plus one. γt , ηj , and δk denote year, industry, and province fixed effects, respectively. We report the results of the Availability and Amount regressions in Table 3. In all four regressions, the coefficient of ROAM A is positive and statistically significant, suggesting that firms with better performance have a higher probability of getting bank credit and getting larger amounts of bank credit. The negative coefficient of Af ter suggests that, after China’s entry into the WTO, firms are less likely to get bank credit and are likely to get a smaller amount of bank credit. The positive coefficients of Af ter × ROAM A in Columns (2) and (4) suggest that the sensitivity of the amount of bank credit to firm performance is significantly stronger after bank competition increases as a result of China’s entry into the WTO. In terms of economic magnitude, a one standard deviation increase in ROAM A leads to a 1.3% 15

Note that we use the logit model when the dependent variable is a binary variable throughout the paper. We replace ROAM A by the past two-year average ROA and the past year ROA, when the past three-year average ROA and the past two-year average ROA are unavailable. 16

12

increase in the probability of getting bank credit and a 5.8% increase in the amount of bank credit after 2002. We also report the results of the cost regressions with Interest Cost, T ransaction Cost, and T otal Cost as dependent variables in Table 4.17 The negative coefficient of Af ter × ROAM A in Columns (2), (4), and (6) suggests that better firms experience a significantly lower cost of bank credit after 2002 using all three cost measures that we discussed above. The positive coefficient of ROAM A in model (1), (3), and (5) may seem to be puzzling at first, since better performing firms are expected to enjoy a lower cost of bank credit. However, if we look at the results in model (7) and (8), we find that better performing firms are more likely to have long-term debt and the positive correlation gets even stronger after China’s entry to WTO. As long-term debt usually has a higher interest cost than short-term debt provided that term structure is upward sloping, this explains why better performing firms have higher interest costs. In terms of economic magnitude, a one standard deviation increase in ROAM A leads to a 3.4 , 3.8, and 4.1 basis point decrease respectively in interest cost, transaction cost and total cost of getting per unit bank credit after 2002. Therefore, the above results are significant both statistically and economically. In sum, the above findings provide support for our hypothesis H1A, i.e., greater bank competition leads to more stringent bank screening in the sense that firms with better past performance are more likely to receive bank credit and have a lower cost of bank credit after 17

Note that we choose to use the Tobit model for the T ransaction Cost regressions (as reported in Columns (3) and (4) of Table 4), since a significant fraction of observations have a zero transaction cost. In untabulated analysis, we rerun these regressions using the OLS model and results are qualitatively similar.

13

China’s entry into the WTO.18

5.2

The Effect of Bank Dependence

We now test the hypotheses H2A and H2B by constructing a dummy variable, BankDependence, to measure the level of bank dependence. Specifically, BankDependence equals one if the average book leverage for a two-digit industry in which the firm belongs to during 1998-2002 is larger than the median of all industries, and zero otherwise. We define this measure using the window of 1998-2002 in order to avoid the potential influence of bank competition after 2002. With the above definition, we divide the sample into two groups based on BankDependence. We first run regression model (1) in two subsamples of firms (i.e., firms with a higher level of bank dependence and those with a lower level of bank dependence) and compare the difference in the coefficient of Af ter × ROAM A across these two subsamples. Further, we use the following triple interaction regression to formally test whether the effect of bank competition on bank screening is stronger for firms in more bank-dependent industries: A MA 0 yi,t = α + β1 ROAM i,t−1 + β2 Af ter + β3 Af ter × ROAi,t−1 + φ Xi,t−1 + β4 BankDependence A +β5 BankDependence × ROAM i,t−1 + β6 BankDependence × Af ter A 0 +β7 BankDependence × Af ter × ROAM i,t−1 + φ Xi,t−1 + γt + ηj + δk + i,t ,

(2)

Our coefficient of interest is the coefficient of the triple interaction, β7 , which captures the 18 We also test the robustness of the results in Table 3 using alternative measures. First, we use alternative proxies to measure various dependent variables. For the availability of bank credit, the Availability measure in the analysis we presented earlier captures all rounds of bank credit availability. To mitigate the potential effect of the first round of credit on subsequent rounds of credit, we also consider only the first round of credit and drop all observations in years after the first credit. Our results are robust to this alternative definition of availability of bank credit. We also measure the availability from the perspective of the stock of bank credit to capture whether a firm ever got bank credit before. Further, we also use interest payment to indicate the availability of bank credit. For the amount of bank credit, we use the amount of bank credit that a firm first ever receives, or the stock of long-term debt. For the cost of bank credit, we also measure the cost as the logarithm of difference between financial expense and interest payment. Second, we use different proxies to measure firm performance: ROAM A adjusted by industry median; ROA adjusted by industry median; ROAM A in the past two years adjusted by industry median; mean of ROA across all the years available adjusted by industry median; Return on Equity (ROE) or Return on Sales (ROS). In all cases, our results above remain qualitatively unchanged.

14

statistical significance of the difference in the coefficients of the double interaction in the two groups. Table 5 reports the results of the above regressions.19 Columns (1) and (2) show that the coefficients of the double interaction, Af ter × ROAM A , are 0.214 for firms in industries with a higher level of bank dependence and 0.146 for firms in industries with a lower level of bank dependence, where the coefficient is significant at 10% level for the former group but insignificant for the later group. Column (3) shows that the coefficient of the triple interaction, BankDependence × Af ter × ROAM A , is positive and significant, indicating that the coefficient of the double interaction for firms in industries with a higher level of bank dependence is significantly larger than that for firms in industries with a lower level of bank dependence. This confirms the prediction of H2A: the sensitivity between Availability and ROAM A after the WTO is even stronger for firms in industries with a higher level of bank dependence. Similarly, the significantly positive coefficient of the triple interaction in Column (6) also confirms the prediction of H2A. The positive coefficient of triple interaction in Column (9) indicates that better performing firms in the industries with a higher level of bank dependence will incur only a higher interest cost after 2002. However, this coefficient of the triple interaction in Column (9) is statistically insignificant.

5.3

The Effect of Financial Development

We then move on to testing our hypotheses H3A and H3B. We run the same regressions as in model (2) but replaced BankDependence with F inDevelopment. F inDevelopment is a dummy variable capturing the level of financial development (Fan, Wang, and Zhu (2003)) in a specific province. It is equal to one if the index of financial industry competition in a province in which the firm is headquartered in China during 1998-2002 is greater than the provincial median for a given year and zero otherwise. We report the results of these 19

We do not control for industry fixed effects here since the definition of bank dependence is time invariant for each industry.

15

regressions in Table 6.20 Different from our earlier results on bank dependence, Columns (3) and (6) show that the coefficient of the triple interaction is negative and significant, indicating that the coefficient of the double interaction for firms in regions with a higher level of financial development is significantly lower than that for firms in regions with a lower level financial development. This supports H3A: the sensitivity of Availability (Amount) to ROAM A after bank competition increases (as a result of China’s entry into the WTO) is smaller for firms in regions with a higher level of financial development. When we compare the effect of competition on interest cost across regions with different levels of financial development, we find that the decrease in interest cost after WTO is only significant in the more financially developed area. This finding indicates that, after banking competition increases, banks in more financially developed regions mainly compete on the price of bank credit. In sum, our empirical results support for our hypothesis H3A, i.e., the relationship between bank competition and loan screening after is weaker for regions with a higher level of financial development.

5.4

Credit Availability for State-Owned Enterprises (SOE)

To investigate whether bank competition has different effects on SOE firms versus nonSOE firms, we use a dummy variable, SOE, to indicate whether the firm is a state-owned enterprise. SOE equals one if the firm is state-owned and zero otherwise. Table 7 reports our results using the regressions in regression model (2) by replacing BankDependence with SOE. As shown in Columns (1) and (2), we find that the increasing sensitivity of Availability to ROAM A after bank competition increases (as a result of China’s entry into the WTO) is mainly concentrated in SOE firms. Further, the insignificant coefficient of ROAM A in Column (1) indicates that there is no significant association between firm performance and the probability of getting bank credit before 2002 for SOE firms. Columns (3) and (6) show that the coefficient of the triple interaction, SOE ×Af ter ×ROAMA , is posi20

We do not control for the province fixed effects here since the definition of financial development is time-invariant for each province.

16

tive and significant, suggesting that the coefficient of the double interaction, Af ter×ROAM A , for SOE firms is significantly higher than that for non-SOE firms. The negative coefficients of Af ter × ROAM A in Columns (7) and (8) indicate that the decreases in interest costs are mainly concentrated in non-SEO firms. This may be because, prior to China’s entry into the WTO, the government may have directed banks to lend to SOEs at concessional interest rates, a practice which was stopped or less used after China’s entry into the WTO (since the banking sector becomes more competitive). Consistent with this, we find that the coefficient of the triple interaction term (SOE × Af ter × ROAM A ) in Column (9) is positive and significant. In sum, the results of Table 7 support our hypothesis H4A, i.e., greater bank competition leads to more stringent screening for SOE firms than that for non-SOE firms.

5.5

The Effect of Bank Competition on Subsequent Firm Performance

We test our hypotheses H5A and H5B in the following two steps. First, we test whether the quality of projects financed through bank credit after 2002 is better by investigating whether firms receiving bank credit after 2002 exhibit stronger sensitivity between investment and growth opportunities. We construct a dummy variable, Credit, which equals one if the first time that a firm receives a bank credit is during 2003-2007 and zero otherwise. To isolate the effect of bank credit on investment quality, we drop observations that receive bank credit during both 19972002 and 2003-2007 and observations that never receive bank credit during our sample period. We then use the following model to test our hypotheses: A MA Investmenti,t = α + β1 SalesGrowthM i,t−1 + β2 Credit + β3 Credit × SalesGrowthi,t−1

+φ0 Xi,t−1 + γt + ηj + δk + i,t ,

(3)

where Investment measures the level of firm investment and is defined as the change of

17

non-current assets divided by last year’s total assets. SalesGrowthM A is the moving average of annual sales growth in the past three years.21 The control vector X includes a list of firm characteristics: Ln(Asset), Leverage, Ln(Age), ROAM A , and lagged Investment. γt , ηj and δk denote year, industry, and province fixed effects, respectively. We expect the coefficient of the double interaction in model (3) to be positive, i.e., firms receiving bank credit after 2002 exhibit stronger sensitivity of their investment to growth opportunities compared to firms receiving bank credit before 2002. To mitigate the potential sample selection bias, we conduct a propensity score matching analysis to construct a control group for the treated group above. We select firms that never received bank credit in forming the control group. In particular, we use the following model to calculate the propensity score. A 0 T reati,t = α + β1 SalesGrowthM i,t−1 + φ Xi,t−1 + i,t ,

(4)

where T reat = 1 for the sample of firms used to test regression model (3) and T reat = 0 for the sample of firms that never received bank credit. After calculating the propensity score for each firm-year observation, we implemented the one-to-one nearest neighbor matching methodology year by year. The matching is conducted without replacement within a year as well as across years. We assign the matching year to be the event year for all firms in the control group and label it with the same dummy variable Credit. Thus, each firm is assigned a group dummy variable, T reat, which represents either the treated group or the control group, and a time dummy variable, Credit, which denotes receiving bank credit after or before 2002. For the control sample, Credit is a counter-factual variable capturing what would happen to a treated firm if it did not receive bank credit. We When the past three-year moving average is unavailable, SalesGrowthM A is replaced by the past twoyear moving average, or replaced by annual growth in sales of last year when the past two-year moving average is unavailable. 21

18

use the following model to capture the differences between the treated and control group: A MA Investmenti,t = α + β1 SalesGrowthM i,t−1 + β2 Credit + β3 Credit × SalesGrowthi,t−1 A +β4 T reati + β5 T reati × SalesGrowthM i,t−1 + β6 T reati × Credit A 0 +β7 T reati × Credit × SalesGrowthM i,t−1 + φ Xi,t−1

+γt + ηj + δk + i,t

(5)

If H5A holds, we would expect the coefficient of the triple interaction, β7 , to be significantly positive. Table 8 reports the regression results testing the sensitivity of investment to growth opportunity. In Column (1), we show that higher SalesGrowthM A leads to more investment. In Column (2), we find that such an effect is stronger for firms that received bank credit during 2003-2007. Such an effect is insignificant for firms in the control group when we assign them a pseudo time of receiving bank credit. Further, the positive and significant coefficient of the triple interaction in Column (4) shows that the coefficient of the double interaction in Column (2) is significantly higher than that in Column (3), suggesting that the investment-growth sensitivity is stronger for firms that received bank credit during 2003-2007. In the second step, we move on to investigating the dynamics of productivity and financial performance around the first new loan to test H5A and H5B. The identification strategy is as follows: if banks conduct stricter screening after 2002, firms receiving credit after 2002 should have better productivity and performance subsequent to receiving credit compared to firms receiving credit before 2002. Specifically, we use the following model: yi,t = α +

TX =2

β1T P re[T ]i,t +

T =1 TX =3 + β5T Credit T =0

TX =3

β2T P ost[T ]i,t + β3 Credit +

T =0

TX =2

β4T Credit × P re[T ]i,t

T =1

× P ost[T ]i,t + φ0 Xi,t−1 + γt + ηj + δk + i,t ,

(6)

where the dependent variable, y, includes P roductivity and ROA. P roductivity is the logarithm of capital productivity, which is defined as sales divided by fixed assets. ROA is the 19

annual returns on asset. P re[S] takes a value of one if the observation is S year(s) before receiving the first new credit and zero otherwise, where S = 0, 1, or 2. P ost[T ] takes the value of one if the observation is T year(s) after receiving the first new credit and zero otherwise, where T = 1, 2, or 3. Table 9 reports the results of the above tests. Column (1) reports results for firms receiving bank credit during 1999-2002. The coefficients of the time variables are all significantly positive and increase over years, suggesting that firms’ productivity increases over time. These results indicate that banks are able to select high productivity firms and their bank credit helps firms to further improve their productivity. Column (2) reports the result for firms receiving bank credit during 2003-2007: the magnitude of the coefficient of time variables are greater than that in Column (1) for five out of six time dummies. Column (3) reports the result testing the differences in the time variable between Columns (1) and (2). Five out of the six coefficients of the double interactions are significantly positive, indicating that firms receiving bank credit during 2003-2007 exhibit significantly higher productivity both before and after receiving bank credit compared to firms receiving bank credit during 1999-2002. Columns (4) to (6) reports the results using ROA as the dependent variable: we find results consistent with those reported in Columns (1) to (3).22 This, again, supports our hypothesis H5A: greater bank competition leads to more stringent screening, which enables banks to select better firms and leads to better firm performance subsequent to receiving credit.

5.6

Additional Robustness Tests

In this section, we present several additional robustness tests. First, regarding our main regression results presented in Table 3, we rerun these regression by controlling for firm fixed effects: we present the results in Table A.2 in the Appendix. The coefficient of Af ter × ROAM A remains qualitatively unchanged. Second, we also analyze firm performance subsequent to receiving credit using the sub22

We also use labor productivity and ROE as alternative proxies for firm performance and our results continue to hold.

20

sample of firms that have received at least one loan (bank credit) before 2002 and one loan after 2002 in order to focus more on the monitoring effect of bank credit. The results of this analysis are presented in Table 10. Here we assume that borrowing firm’s quality does not change over time. By controlling for firm fixed effects, we are able to compare the effect of the same firm having received multiple rounds of bank credits. The positive and significant coefficients of Af ter in Columns (1) and (3) in Table 10 suggest that the bank credit received by a firm after 2002 has a greater effect on a firm’s performance in the following year. We further include two dummies (CreditBef ore and CreditAf ter) to capture the bank credit that a firm receives before and after, respectively, China’s entry into the WTO. The positive and significant coefficients of CreditBef ore in Columns (2) and (4) suggest that firms perform better after they have received the first bank credit prior to 2003. However, it is more interesting to note that the coefficient of CreditAf ter is positive and significantly greater than that of CreditBef ore, suggesting that the effect of bank credit on subsequent firm performance is much greater if a borrowing firm received a second bank loan after China’s WTO entry. The results in Table 10 support the notion that the greater bank competition that resulted from China’s entry into the WTO leads Chinese banks into becoming better monitors of borrowing firms as well as being better able to screen borrowing firms.

6

Conclusion

In this paper, we use the banking sector reforms that accompanied China’s accession to the WTO and a large sample of Chinese private firms, to empirically analyze the relation between greater bank competition and the screening of potential borrowers. Our results may be summarized as follows. First, the sensitivity of bank credit to prior borrowing firm performance increases after China’s entry to the WTO. This increase in sensitivity is larger in more bank-dependent industries and smaller in Chinese regions with a higher level of financial development. Second, the increase in the sensitivity of bank credit to firm performance is 21

much greater for state-owned firms compared to private firms without state ownership. Third, the effect of bank credit on subsequent firm productivity and performance is greater for loans given after China’s WTO entry compared to those given prior to WTO entry. Overall, the results of our empirical analysis suggest that the stringency of bank screening in China increased with greater competition in the banking sector arising form China’s entry into the WTO.

22

References Allen, Franklin and Douglas Gale (Chapter 8, 2000), Comparing financial systems. MIT press. Bai, Chong-En, Jiangyong Lu, and Zhigang Tao (2006), “The multitask theory of state enterprise reform: Empirical evidence from China.” The American Economic Review, 96, 353–357. Beck, Thorsten, Asli Demirgüç-Kunt, and Vojislav Maksimovic (2004), “Bank competition and access to finance: International evidence.” Journal of Money, Credit and Banking, 627–648. Berger, Allen N, Iftekhar Hasan, and Mingming Zhou (2009), “Bank ownership and efficiency in China: What will happen in the world’s largest nation?” Journal of Banking & Finance, 33, 113–130. Berger, Allen N, Anthony Saunders, Joseph M Scalise, and Gregory F Udell (1998), “The effects of bank mergers and acquisitions on small business lending.” Journal of Financial Economics, 50, 187–229. Bertrand, Marianne, Antoinette Schoar, and David Thesmar (2007), “Banking deregulation and industry structure: Evidence from the French banking reforms of 1985.” The Journal of Finance, 62, 597–628. Black, Sandra E and Philip E Strahan (2002), “Entrepreneurship and bank credit availability.” The Journal of Finance, 57, 2807–2833. Boyd, John H and Gianni De Nicolo (2005), “The theory of bank risk taking and competition revisited.” The Journal of Finance, 60, 1329–1343. Brandt, Loren, Johannes Van Biesebroeck, and Yifan Zhang (2012), “Creative accounting or creative destruction? Firm-level productivity growth in Chinese manufacturing.” Journal of Development Economics, 97, 339–351. Brandt, Loren, Johannes Van Biesebroeck, and Yifan Zhang (2014), “Challenges of working with the Chinese NBS firm-level data.” China Economic Review, 30, 339–352. Broecker, Thorsten (1990), “Credit-worthiness tests and interbank competition.” Econometrica, 429–452. Cetorelli, Nicola (2004), “Real Effects of Bank Competition.” Journal of Money, Credit and Banking, 543–558. Cetorelli, Nicola and Pietro F Peretto (2012), “Credit quantity and credit quality: Bank competition and capital accumulation.” Journal of Economic Theory, 147, 967–998. Cetorelli, Nicola and Philip E Strahan (2006), “Finance as a barrier to entry: Bank competition and industry structure in local us markets.” The Journal of Finance, 61, 437–461. 23

Chemmanur, Thomas J and Paolo Fulghieri (1994), “Reputation, renegotiation, and the choice between bank loans and publicly traded debt.” Review of Financial Studies, 7, 475– 506. Chemmanur, Thomas J, Karthik Krishnan, and Debarshi K Nandy (2011), “How does venture capital financing improve efficiency in private firms? A look beneath the surface.” The Review of Financial Studies, 24, 4037–4090. Chen, Xiaofen (2005), “Financial liberalization, competition, and bank loan quality.” Journal of Economic Integration, 109–122. Chen, Yunling, Ming Liu, and Jun Su (2013), “Greasing the wheels of bank lending: Evidence from private firms in China.” Journal of Banking & Finance, 37, 2533–2545. Cull, Robert, Lixin Colin Xu, and Tian Zhu (2009), “Formal finance and trade credit during China’s transition.” Journal of Financial Intermediation, 18, 173–192. Diamond, Douglas W (1989), “Reputation acquisition in debt markets.” Journal of Political Economy, 97, 828–862. Fan, Gang, Xiaolu Wang, and Hengpeng Zhu (2003), “Neri index of marketization of china’s provinces.” National Economic Research Institute, Beijing. Freixas, Xavier, Sjaak Hurkens, Alan D Morrison, and Nir Vulkan (2007), “Interbank competition with costly screening.” The BE Journal of Theoretical Economics, 7. Gormley, Todd A (2010), “The impact of foreign bank entry in emerging markets: Evidence from India.” Journal of Financial Intermediation, 19, 26–51. Guariglia, Alessandra and Sandra Poncet (2008), “Could financial distortions be no impediment to economic growth after all? Evidence from China.” Journal of Comparative Economics, 36, 633–657. Hoshi, Takeo, Anil Kashyap, and David Scharfstein (1990), “Bank monitoring and investment: Evidence from the changing structure of japanese corporate banking relationships.” In Asymmetric information, corporate finance, and investment, 105–126, University of Chicago Press. Jayaratne, Jith and Philip E Strahan (1996), “The finance-growth nexus: Evidence from bank branch deregulation.” The Quarterly Journal of Economics, 111, 639–670. Jiang, Fuxiu, Zhan Jiang, Jicheng Huang, Kenneth A Kim, and John R Nofsinger (2017), “Bank competition and leverage adjustments.” Financial Management. Kerr, William R and Ramana Nanda (2009), “Democratizing entry: Banking deregulations, financing constraints, and entrepreneurship.” Journal of Financial Economics, 94, 124– 149. Klein, Michael A (1971), “A theory of the banking firm.” Journal of Money, Credit and Banking, 3, 205–218. 24

Krishnan, Karthik, Debarshi K Nandy, and Manju Puri (2014), “Does financing spur small business productivity? Evidence from a natural experiment.” The Review of Financial Studies, 28, 1768–1809. Leigh, Lamin and Richard Podpiera (2006), The rise of foreign investment in China’s banks: Taking stock. 6-292, International Monetary Fund. Liang, Qi, Pisun Xu, and Pornsit Jiraporn (2013), “Board characteristics and Chinese bank performance.” Journal of Banking & Finance, 37, 2953–2968. Lin, Huidan (2011), “Foreign bank entry and firms’ access to bank credit: evidence from china.” Journal of Banking & Finance, 35, 1000–1010. Lin, Justin Y, Xifang Sun, and Harry X Wu (2015), “Banking structure and industrial growth: Evidence from China.” Journal of banking & Finance, 58, 131–143. Lin, Justin Yifu, Fang Cai, and Zhou Li (1998), “Competition, policy burdens, and stateowned enterprise reform.” The American Economic Review, 88, 422–427. Marquez, Robert (2002), “Competition, adverse selection, and information dispersion in the banking industry.” The Review of Financial Studies, 15, 901–926. Petersen, Mitchell A and Raghuram G Rajan (1995), “The effect of credit market competition on lending relationships.” The Quarterly Journal of Economics, 110, 407–443. Petriconi, Silvio (2015), “Bank competition, information choice and inefficient lending booms.” Bocconi University Working Paper. Qian, Jun QJ, Philip E Strahan, and Zhishu Yang (2015), “The impact of incentives and communication costs on information production and use: Evidence from bank lending.” The Journal of Finance, 70, 1457–1493. Rajan, Raghuram G and Luigi Zingales (1998), “Financial Dependence and Growth.” The American Economic Review. Sapienza, Paola (2002), “The effects of banking mergers on loan contracts.” The Journal of Finance, 57, 329–367. Zarutskie, Rebecca (2006), “Evidence on the effects of bank competition on firm borrowing and investment.” Journal of Financial Economics, 81, 503–537.

25

Table 1: Sample Distribution by Year and Industry This table reports the distribution of our sample during 1999-2007 by year and by industry. Availability equals one if a firm’s change in long-term debt in year t is greater than zero and zero otherwise.

Availability=1

Total

Year

N

Pct(%)

N

1999 2000 2001 2002 2003 2004 2005 2006 2007

10,526 9,495 8,471 9,851 9,790 9,638 12,565 12,299 14,624

23.2% 20.9% 19.1% 16.6% 15.2% 14.7% 11.8% 11.0% 11.1%

45,423 45,461 44,262 59,372 64,340 65,534 106,409 112,342 132,074

N

Pct(%)

N

Processing of food from agricultural products Foods Alcohol, beverages, refined tea Tobacco Textiles Textiles, clothing, and apparel industry Leather, fur, feather, and related products Processing of timber, manufacture of wood, bamboo, etc. Furniture Paper and paper products Printing Articles for culture, education, etc. Petroleum processing Chemical materials and products Pharmaceutical Chemical fibers Rubber Plastics Non-metallic mineral products Smelting and processing of ferrous metals Smelting and processing of non-ferrous metals Metal products General purpose machinery Special purpose machinery Transportation equipment Electrical machinery and equipment Computers, communication, and other electronic equipment Measuring instruments Other manufacturing Comprehensive use of waste resources

6,252 2,259 1,739 226 8,249 2,044 1,135 1,867 673 3,849 2,097 581 914 8,874 3,168 568 1,162 3,616 10,848 2,409 2,004 4,257 8,636 4,597 5,007 3,754 2,375 2,606 1,395 98

17.2% 17.1% 19.3% 22.8% 13.4% 8.4% 9.0% 14.3% 10.4% 16.2% 14.9% 9.0% 17.5% 16.1% 22.1% 16.1% 13.8% 11.5% 17.5% 14.0% 14.4% 10.8% 14.2% 15.8% 15.2% 12.1% 13.0% 14.7% 11.2% 17.5%

36,414 13,202 9,009 990 61,369 24,284 12,562 13,035 6,452 23,706 14,076 6,473 5,224 55,131 14,320 3,523 8,440 31,311 62,118 17,192 13,948 39,578 60,814 29,121 32,902 31,035 18,207 17,766 12,454 561

Total

97,259

14.4%

675,217

Two-digit Industry

26

Table 2: Summary Statistics This table presents the descriptive statistics for the sample. Af ter equals one for firm-years during 2003-2007 and zero otherwise. Availability equals one if a firm’s change in long-term debt in year t is greater than zero and zero otherwise. Amount is the natural logarithm of the change in long-term debt plus one. Interest Cost represents interest payment per unit of bank credit, and is defined as interest payment over total liability. T ransaction Cost represents transaction cost of bank credit, and is defined as financial expense minus interest payment over long-term debt. T otal Cost represents total cost that is associated with per unit of of bank credit, and is defined as financial expense over total liability. ROA is the annual return on asset, defined as net income divided by assets. ROAM A is a firm’s moving average of annual return on assets in the past three years. Investment, the dependent variable, is divided the change of non-current assets by the lag total assets. SalesGrowthM A is the moving average of sales growth in the past three years. Standard errors are clustered at firm level and reported in parentheses. P roductivity is the logarithm of capital productivity, defined as sales divided by fixed asset. Ln(Asset) is the natural logarithm of firm assets. Leverage is defined as book debts to assets ratio. Ln(Age) is the natural logarithm of firm age plus one. F ixed Asset is defined as fixed assets divided by total assets. Tests for differences are based on t-statistics (two-tailed). ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% levels, respectively.

VARIABLES

N

Mean Total

After Availability Amount Interest Cost Transaction Cost Total Cost ROA ROAMA Investment SalesGrowthMA Productivity Leverage Ln(Asset) Ln(Age) FixedAsset

675,217 675,217 291,191 524,464 524,464 524,464 675,217 675,217 423,180 423,180 263,778 675,217 675,217 675,217 675,217

0.712 0.144 2.001 0.036 0.011 0.046 0.072 0.068 0.041 0.198 1.325 0.578 9.560 2.037 0.348

After=0

After=1

0.197 2.076 0.041 0.010 0.051 0.063 0.063 0.030 0.132 1.111 0.598 9.626 2.278 0.368

0.123 1.956 0.034 0.011 0.044 0.076 0.071 0.044 0.218 1.498 0.569 9.534 1.940 0.341

27

S.D.

Min

0.453 0.351 3.282 0.066 0.049 0.082 0.116 0.100 0.173 0.299 1.138 0.238 1.211 0.883 0.204

0 0 0 0 0 0 -0.119 -0.108 -0.546 -0.458 -1.497 0.016 7.012 0 0

p50

Max

Difference 0.075*** 0.121*** 0.007*** -0.000* 0.007*** -0.013*** -0.007*** -0.015*** -0.087*** -0.387*** 0.029*** 0.091*** 0.338*** 0.027***

1 1 0 1 0 9.926 0.021 1 0 1 0.026 1 0.035 0.836 0.036 0.564 0.001 0.796 0.153 1.198 1.289 5.456 0.608 0.974 9.418 13.110 1.946 3.912 0.324 0.868

Table 3: The Sensitivity of Bank Credit to Prior Firm Performance This table reports estimation results testing the sensitivity between availability/amount of bank credit and firm performance. Availability equals one if a firm’s change in long-term debt in year t is greater than zero and zero otherwise. Amount is the natural logarithm of the change in long-term debt plus one. ROAM A is a firm’s moving average of annual return on assets in the past three years. Af ter equals one for firm-years during 2003-2007 and zero otherwise. Ln(Asset) is the natural logarithm of firm assets. Leverage is defined as book debts to assets ratio. Ln(Age) is the natural logarithm of firm age plus one. F ixedAsset is defined as fixed assets divided by total assets. Standard errors are clustered at firm level and reported in parentheses. ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% levels, respectively.

Availability

Amount

(Logit)

(OLS)

(1) ROAMA

(2)

0.639*** (13.95)

After After × ROAMA Ln(Asset)

0.297*** (80.25) Leverage 0.079*** (4.17) Ln(Age) 0.088*** (17.90) Fixed Asset 0.754*** (35.73) Constant -4.287*** (-86.22) Year FE Yes Two-digit Industry FE Yes Province FE Yes Adjusted R2 Observations 675217

28

(3)

(4)

0.556*** 0.846*** 0.488*** (8.06) (11.32) (4.57) -0.791*** -0.485*** (-50.57) (-17.14) 0.131* 0.579*** (1.66) (4.53) 0.297*** 0.345*** 0.344*** (79.91) (55.65) (55.29) 0.079*** -0.452*** -0.451*** (4.17) (-14.52) (-14.47) 0.087*** -0.221*** -0.223*** (17.80) (-29.28) (-29.43) 0.754*** 0.492*** 0.489*** (35.67) (14.27) (14.15) -4.278*** -0.398*** -0.364*** (-85.33) (-5.00) (-4.54) Yes Yes Yes Yes Yes Yes Yes Yes Yes 0.024 0.024 675217 291191 291191

Table 4: The Sensitivity of Bank Credit-related Cost to Prior Firm Performance This table reports estimation results testing the sensitivity between cost of bank credit and firm performance. Interest Cost represents interest payment per unit of bank credit, and is defined as interest payment over total liability. T ransaction Cost represents transaction cost of bank credit, and is defined as financial expense minus interest payment over long-term debt. T otal Cost represents total cost that is associated with per unit of bank credit, and is defined as financial expense over total liability. Long-term Debt equals one if a firm’s long-term debt in year t is greater than zero and zero otherwise. ROAM A is a firm’s moving average of annual return on assets in the past three years. Af ter equals one for firm-years during 2003-2007 and zero otherwise. Ln(Asset) is the natural logarithm of firm assets. Leverage is defined as book debts to assets ratio. Ln(Age) is the natural logarithm of firm age plus one. F ixedAsset is defined as fixed assets divided by total assets. Standard errors are clustered at firm level and reported in parentheses. ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% levels, respectively.

ROAMA

Interest Cost

Transaction Cost

Total Cost

Long-term Debt

(OLS)

(Tobit)

(OLS)

(Logit)

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

0.097∗∗∗ (0.002)

0.120∗∗∗ (0.004) -0.008∗∗∗ (0.000) -0.034∗∗∗ (0.004) -0.004∗∗∗ (0.000) -0.035∗∗∗ (0.001) 0.001∗∗∗ (0.000) 0.021∗∗∗ (0.001) 0.090∗∗∗ (0.001) Yes Yes Yes 0.107 524464

0.062∗∗∗ (0.002)

0.141∗∗∗ (0.003)

0.455∗∗∗ (0.005) 1.168∗∗∗ (0.023) 0.336∗∗∗ (0.006) 1.332∗∗∗ (0.025) -5.558∗∗∗ (0.061) Yes Yes Yes

0.347∗∗∗ (0.082) -1.220∗∗∗ (0.015) 0.248∗∗∗ (0.089) 0.454∗∗∗ (0.005) 1.169∗∗∗ (0.023) 0.336∗∗∗ (0.006) 1.330∗∗∗ (0.025) -5.541∗∗∗ (0.061) Yes Yes Yes

524464

524464

0.168∗∗∗ (0.005) -0.007∗∗∗ (0.000) -0.041∗∗∗ (0.005) -0.007∗∗∗ (0.000) -0.061∗∗∗ (0.001) 0.000 (0.000) 0.028∗∗∗ (0.001) 0.136∗∗∗ (0.002) Yes Yes Yes 0.153 524464

0.503∗∗∗ (0.055)

-0.000 (0.000) -0.038∗∗∗ (0.001) -0.003∗∗∗ (0.000) 0.001 (0.001) 0.071∗∗∗ (0.002) Yes Yes Yes

0.088∗∗∗ (0.004) 0.011∗∗∗ (0.001) -0.038∗∗∗ (0.004) 0.000 (0.000) -0.038∗∗∗ (0.001) -0.003∗∗∗ (0.000) 0.002∗∗ (0.001) 0.070∗∗∗ (0.001) Yes Yes Yes

524464

524464

After 29

After × ROAMA -0.004∗∗∗ (0.000) Leverage -0.035∗∗∗ (0.001) Ln(Age) 0.001∗∗∗ (0.000) Fixed Asset 0.021∗∗∗ (0.001) Constant 0.093∗∗∗ (0.001) Year FE Yes Two-digit Industry FE Yes Province FE Yes 2 Adjusted R 0.107 Observations 524464 Ln(Asset)

-0.007∗∗∗ (0.000) -0.061∗∗∗ (0.001) 0.000 (0.000) 0.028∗∗∗ (0.001) 0.139∗∗∗ (0.001) Yes Yes Yes 0.153 524464

Table 5: The Effects of Bank Dependence on the Sensitivity of Bank Credit to Prior Firm Performance This table reports estimation results testing the effect of bank dependence on the sensitivity between bank credit and firm performance. Availability equals one if a firm’s change in long-term debt in year t is greater than zero and zero otherwise. Amount is the natural logarithm of the change in long-term debt plus one. Interest Cost represents interest payment per unit of bank credit, and is defined as interest payment over total liability. ROAM A is a firm’s moving average of annual return on assets in the past three years. Af ter equals one for firm-years during 2003-2007 and zero otherwise. BankDependence equals one if the mean of total debt ratio (total debt divided by total asset) of a two-digit industry during 1998-2002 is larger than the median of all industries, and zero otherwise. All regressions include Ln(Asset), Leverage, Ln(Age), and F ixedAsset as control variables. Standard errors are clustered at firm level and reported in parentheses. ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% levels, respectively.

Availability BankDependence =1 (1) ROAMA After 30

After × ROAMA

0.285*** (0.103) -0.773*** (0.021) 0.214* (0.116)

BankDependence BankDependence × After BankDependence × ROAMA BankDependence × After × ROAMA Constant Control Variables Year FE Province FE Adjusted R2 Observations

-4.747*** (0.066) Yes Yes Yes 372419

=0 (2)

Amount Whole Sample (3)

BankDependence =1 (4)

=0 (5)

0.670*** 0.913*** 0.533*** 0.350** (0.093) (0.089) (0.158) (0.145) -0.804*** -0.801*** -0.468*** -0.494*** (0.023) (0.019) (0.038) (0.042) 0.146 0.045 0.703*** 0.473*** (0.108) (0.107) (0.188) (0.174) 0.030** (0.015) 0.021 (0.019) -0.799*** (0.129) 0.259* (0.156) -3.989*** -4.431*** -0.909*** 0.006 (0.069) (0.048) (0.093) (0.101) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 0.027 0.021 302237 674656 162872 128085

Interest Cost Whole Sample (6)

BankDependence =1 (7)

=0 (8)

Whole Sample (9)

0.726*** 0.111*** 0.127*** 0.136*** (0.138) (0.006) (0.005) (0.005) -0.482*** -0.009*** -0.008*** -0.008*** (0.033) (0.000) (0.001) (0.000) 0.362** -0.034*** -0.032*** -0.034*** (0.172) (0.006) (0.006) (0.006) 0.053** -0.000 (0.024) (0.000) -0.001 -0.001* (0.030) (0.000) -0.492** -0.032*** (0.195) (0.008) 0.459* 0.003 (0.252) (0.009) -0.551*** 0.073*** 0.105*** 0.088*** (0.069) (0.001) (0.002) (0.001) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 0.024 0.095 0.107 0.104 290957 290748 235371 526119

Table 6: The Effects of Financial Development on the Sensitivity of Bank Credit to Prior Firm Performance This table reports estimation results testing the effect of bank dependence on the sensitivity between bank credit and firm performance. Availability equals one if a firm’s change in long-term debt in year t is greater than zero and zero otherwise. Amount is the natural logarithm of the change in long-term debt plus one. Interest Cost represents interest payment per unit of bank credit, and is defined as interest payment over total liability. ROAM A is a firm’s moving average of annual return on assets in the past three years. Af ter equals one for firm-years during 2003-2007 and zero otherwise. F inDevelopment equals one if the index of financial industry competition in a province during 1998-2002 is larger than the median of the index of financial industry competition of a given year and zero otherwise. All regressions include Ln(Asset), Leverage, Ln(Age), and F ixed Asset as control variables. Standard errors are clustered at firm level and reported in parentheses. ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% levels, respectively.

Availability FinDevelopment =1 (1) ROAMA After 31

After × ROAMA

0.801*** (0.077) -0.804*** (0.018) -0.063 (0.089)

FinDevelopment FinDevelopment × After FinDevelopment × ROAMA FinDevelopment × After × ROAMA Constant Control Variables Year FE Two-digit Industry FE Adjusted R2 Observations

-4.606*** (0.050) Yes Yes Yes 532187

=0 (2)

Amount Whole Sample (3)

FinDevelopment =1 (4)

=0 (5)

Interest Cost Whole Sample (6)

FinDevelopment =1 (7)

=0 (8)

Whole Sample (9)

-0.378** -0.395*** 0.645*** 0.050 -0.023 0.141*** 0.088*** 0.093*** (0.157) (0.151) (0.115) (0.249) (0.236) (0.005) (0.008) (0.008) -0.798*** -0.791*** -0.473*** -0.552*** -0.500*** -0.010*** -0.009*** -0.010*** (0.031) (0.022) (0.032) (0.061) (0.039) (0.000) (0.001) (0.001) 0.671*** 0.692*** 0.452*** 0.902*** 1.081*** -0.048*** -0.009 -0.012 (0.183) (0.181) (0.139) (0.306) (0.304) (0.005) (0.010) (0.010) -0.093*** -0.116*** -0.000 (0.017) (0.028) (0.000) -0.014 0.012 0.000 (0.021) (0.035) (0.001) 1.208*** 0.690*** 0.046*** (0.167) (0.257) (0.010) -0.770*** -0.674** -0.035*** (0.201) (0.334) (0.011) -4.408*** -4.490*** -0.501*** -0.551*** -0.431*** 0.084*** 0.100*** 0.088*** (0.085) (0.045) (0.079) (0.145) (0.073) (0.001) (0.003) (0.001) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 0.020 0.027 0.022 0.101 0.077 0.095 143030 675217 224213 66978 291191 421472 104965 526437

Table 7: The Effects of State Ownership on the Sensitivity of Bank Credit to Prior Firm Performance This table reports estimation results testing the effect of state ownership on the sensitivity between bank credit and firm performance. Availability equals one if a firm’s change in long-term debt in year t is greater than zero and zero otherwise. Amount is the natural logarithm of the change in long-term debt plus one. Interest Cost represents interest payment per unit of bank credit, and is defined as interest payment over total liability. ROAM A is a firm’s moving average of annual return on assets in the past three years. Af ter equals one for firm-years during 2003-2007 and zero otherwise. SOE means state-owned enterprises, which equals one if the firm is state-owned and zero otherwise. All regressions include Ln(Asset), Leverage, Ln(Age), and F ixedAsset as control variables. Standard errors are clustered at firm level and reported in parentheses. ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% levels, respectively.

Availability

ROAMA After 32

After × ROAMA

SOE (1)

Non-SOE (2)

-0.444 (0.306) -0.878*** (0.053) 1.425*** (0.414)

0.566*** (0.072) -0.740*** (0.018) 0.098 (0.082)

-4.790*** (0.146) Yes Yes Yes Yes

-4.226*** (0.054) Yes Yes Yes Yes

49196

626021

SOE SOE × After SOE × ROAMA SOE × After × ROAMA Constant Control Variables Year FE Two-digit Industry FE Province FE Adjusted R2 Observations

Amount Whole (3)

SOE (4)

0.610*** 1.641*** (0.071) (0.531) -0.786*** -0.925*** (0.016) (0.093) 0.062 1.459* (0.082) (0.786) 0.010 (0.018) -0.019 (0.028) -1.068*** (0.297) 1.699*** (0.420) -4.280*** -2.859*** (0.050) (0.254) Yes Yes Yes Yes Yes Yes Yes Yes 0.051 675217 34991

Non-SOE (5) 0.410*** (0.111) -0.359*** (0.031) 0.475*** (0.132)

-0.100 (0.085) Yes Yes Yes Yes 0.021 256200

Interest Cost Whole (6)

SOE (7)

0.589*** 0.091*** (0.110) (0.011) -0.447*** -0.011*** (0.029) (0.001) 0.399*** -0.002 (0.131) (0.019) 0.095*** (0.030) -0.217*** (0.046) -0.380 (0.507) 1.959** (0.784) -0.379*** 0.051*** (0.080) (0.003) Yes Yes Yes Yes Yes Yes Yes Yes 0.025 0.058 291191 37671

Non-SOE (8)

Whole (9)

0.115*** (0.004) -0.010*** (0.000) -0.031*** (0.005)

0.118*** (0.004) -0.010*** (0.000) -0.033*** (0.005) -0.006*** (0.000) 0.001*** (0.001) -0.052*** (0.011) 0.034* (0.019) 0.088*** (0.001) Yes Yes Yes Yes 0.108 526437

0.090*** (0.001) Yes Yes Yes Yes 0.109 488766

Table 8: A Propensity Score Matching Analysis of the Effect of Bank Competition on Investment-growth Sensitivity This table reports the sensitivity between investment and growth opportunity based on propensity score matching. The dependent variable, Investment, defined as the change of non-current assets divided by last year’s total assets. SalesGrowthM A is the moving average of sales growth in the past three years. Credit equals one if a firm’s first time of increase in longterm debt is during 2003-2007 and zero otherwise. T reat equals one for firms that have ever received bank credit, and zero otherwise. Here control group (T reat = 0) is selected based on propensity score matching method. All regressions include Ln(Asset), Leverage, Ln(Age) and Investment(lagged) as control variables. Standard errors are clustered at firm level and reported in parentheses. ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% levels, respectively.

Investment Treat Sample (1) SalesGrowthMA Credit Credit × SalesGrowthMA Treat Treat × Credit Treat × SalesGrowthMA Treat × Credit × SalesGrowthMA Constant Control Variables Year FE Two-digit Industry FE Province FE Adjusted R2 Observations

0.0681*** (0.0022)

Treat Sample (2)

Control Sample (3)

Whole Sample (4)

0.0567*** 0.0624*** 0.0621*** (0.0038) (0.0033) (0.0033) 0.0162*** 0.0111*** 0.0159*** (0.0014) (0.0015) (0.0013) 0.0135*** 0.0003 0.0003 (0.0045) (0.0042) (0.0042) 0.0001 (0.0013) -0.0017 (0.0017) -0.0059 (0.0050) 0.0134** (0.0061) 0.0885*** 0.0808*** 0.1667*** 0.1116*** (0.0065) (0.0066) (0.0083) (0.0051) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 0.039 0.041 0.042 0.041 116,226 116,226 110,196 226,422

33

Table 9: The Effect of Bank Credit on Subsequent Firm Productivity and Performance This table reports the dynamics of capital productivity and financial performance around the first time of bank credit. P roductivity is the logarithm of capital productivity, defined as sales divided by fixed asset. ROA is the annual return on asset, defined as net income divided by assets. Credit equals one if a firm’s first time of increase in long-term debt is during 2003-2007 and zero otherwise. P re[i] (P ost[i]) takes a value of one if the observation is i = 0, 1, 2 (=1,2,3) years before (after) the year receiving the first new credit and zero otherwise. All regressions include Ln(Asset), Leverage and Ln(Age) as control variables. Standard errors are clustered at firm level and reported in parentheses. ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% levels, respectively.

Productivity

Pre[2] Pre[1] Post[0] Post[1] Post[2] Post[3]

Credit=0 Sample (1)

Credit=1 Sample (2)

Whole Sample (3)

Credit=0 Sample (4)

Credit=1 Sample (5)

Whole Sample (6)

0.113*** (0.011) 0.213*** (0.013) 0.266*** (0.015) 0.338*** (0.017) 0.410*** (0.018) 0.486*** (0.018)

0.102*** (0.008) 0.228*** (0.009) 0.275*** (0.011) 0.358*** (0.014) 0.445*** (0.017) 0.528*** (0.021)

0.004*** (0.001) 0.009*** (0.001) 0.013*** (0.002) 0.015*** (0.002) 0.018*** (0.002) 0.020*** (0.002)

0.010*** (0.001) 0.019*** (0.001) 0.026*** (0.001) 0.030*** (0.002) 0.032*** (0.002) 0.039*** (0.003)

4.766*** (0.110) Yes Yes Yes No Yes 0.829 122,057

5.256*** (0.092) Yes Yes Yes No Yes 0.760 141,721

-0.022* (0.013) -0.055*** (0.015) -0.107*** (0.018) -0.132*** (0.021) -0.140*** (0.023) -0.122*** (0.025) -0.143*** (0.024) -0.007 (0.015) 0.038** (0.017) 0.092*** (0.020) 0.137*** (0.024) 0.199*** (0.027) 0.229*** (0.030) 4.559*** (0.037) Yes Yes No Yes Yes 0.341 263,778

0.378*** (0.012) Yes Yes Yes No Yes 0.708 122,240

0.378*** (0.011) Yes Yes Yes No Yes 0.613 142,093

-0.001 (0.001) -0.002 (0.001) 0.002 (0.002) -0.001 (0.002) -0.000 (0.002) -0.001 (0.002) -0.009*** (0.002) 0.003* (0.002) 0.006*** (0.002) 0.010*** (0.002) 0.010*** (0.002) 0.009*** (0.003) 0.013*** (0.003) 0.319*** (0.004) Yes Yes No Yes Yes 0.263 264,333

Credit Credit × Pre[2] Credit × Pre[1] Credit × Post[0] Credit × Post[1] Credit × Post[2] Credit × Post[3] Constant Control Variables Year FE Firm FE Two-digit Industry FE Province FE Adjusted R2 Observations

ROA

34

Table 10: The Effect of Repeated Bank Credit on Subsequent Firm Productivity and Performance This table reports the effects of repeated bank credit on productivity and Performance. P roductivity is the logarithm of capital productivity, defined as sales divided by fixed asset. ROA is the annual return on asset, defined as net income divided by assets. Af ter equals one for firm-years during 2003-2007 and zero otherwise. CreditBef ore equals one for years after firms receive the first bank credit before the WTO and zero otherwise. CreditAf ter equals one for years after firms receive the first bank credit after the WTO and zero otherwise. Ln(Asset) is the natural logarithm of firm assets. Leverage is defined as book debts to assets ratio. Ln(Age) is the natural logarithm of firm age plus one. F ixedAsset is defined as fixed assets divided by total assets and Ln(Loan) is the natural logarithm of total loans. Standard errors are clustered at firm level and reported in parentheses. ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% levels, respectively.

Productivityt+1 (1)

(2)

ROAt+1 (3)

(4)

After

0.248*** (0.008)

CreditBefore

0.188*** 0.009*** (0.008) (0.001) 0.216*** 0.014*** (0.008) (0.001) -0.174*** -0.192*** -0.007*** -0.008*** (0.011) (0.011) (0.001) (0.001) -0.056** -0.047* -0.034*** -0.034*** (0.026) (0.026) (0.004) (0.004) 0.014** 0.004 0.000 -0.000 (0.007) (0.007) (0.001) (0.001) -1.211*** -1.204*** 0.004 0.004 (0.035) (0.035) (0.004) (0.004) 0.001* -0.013*** 0.000*** -0.001*** (0.001) (0.001) (0.000) (0.000) 3.307*** 3.450*** 0.146*** 0.154*** (0.113) (0.115) (0.014) (0.014)

CreditAfter Ln(Asset) Leverage Ln(Age) FixedAsset Ln(Loan) Constant Firm FE Adjusted R2 Observations

Yes 0.786 59729

0.015*** (0.001)

Yes 0.787 59729

35

Yes 0.630 59804

Yes 0.631 59804

A

Appendix Table A.1: Sample Construction Observations 2,225,945 -177,205 -464,486 -57,555 -3,046 -167,968 -680,468 675,217

Firms in the sample Full Sample from National Bureau of Statistics Firms in non-manufacturing industries Non-Mainland(China) firms Conglomerates Public firms and firms backed by VC Firms with sales less than RMB 5 million Firms with missing values in the variables we use Final Sample

36

Table A.2: The Sensitivity of Bank Credit to Prior Firm Performance This table reports estimation results testing the sensitivity between availability/amount of bank credit and firm performance. Availability equals one if a firm’s change in long-term debt in year t is greater than zero and zero otherwise. Amount is the natural logarithm of the change in long-term debt plus one. ROAM A is a firm’s moving average of annual return on assets in the past three years. Af ter equals one for firm-years during 2003-2007 and zero otherwise. Ln(Asset) is the natural logarithm of firm assets. Leverage is defined as book debts to assets ratio. Ln(Age) is the natural logarithm of firm age plus one. F ixedAsset is defined as fixed assets divided by total assets. Standard errors are clustered at firm level and reported in parentheses. ***, **, and * indicate statistical significance at the 1%, the 5%, and the 10% levels, respectively.

Availability ROAMA

(1)

(2)

(3)

(4)

0.015 (1.01)

-0.031 (-1.54) -0.095*** (-22.39) 0.069*** (3.66) -0.000 (-0.11) -0.122*** (-24.60) 0.005*** (2.88) -0.006 (-0.99) 0.276*** (14.42) Yes Yes 0.165 675217

0.290 (1.40)

-0.152 (-0.61) -0.890*** (-20.89) 0.691*** (3.15) 0.023 (0.88) -1.962*** (-28.61) 0.017 (0.88) -0.098 (-1.20) 3.490*** (13.05) Yes Yes 0.043 291191

After After × ROAMA Ln(Asset) Leverage Ln(Age) Fixed Asset Constant Year FE Firm FE Adjusted R2 Observations

Amount

0.000 (0.25) -0.123*** (-24.76) 0.005*** (2.99) -0.006 (-0.94) 0.267*** (14.05) Yes Yes 0.165 675217

37

0.032 (1.23) -1.969*** (-28.73) 0.019 (0.98) -0.094 (-1.15) 3.381*** (12.78) Yes Yes 0.043 291191