Within the banking sector itself, although it is still dominated by majority-state owned big five"banks which hold about 47%o of industry assets and roughly the same share of total deposits in the sector, contestability of the sector has also increased (PBC FSr Report, 2012). Admittedly, bank competition is to a significant extent constrained by relatively high entry barriers and interest rate regulation The existing literature on the efficiency of the Chinese banking sector has mixed findings Banks undergoing a foreign acquisition or public listing are found to have better pre event performance, but little performance changes were recorded after changes of ownership(Lin and Zhang, 2009). On the other hand, there is evidence that the Chinese banking system has benefited from the entry of foreign investors through higher profitability and incr reased efficiency in the banking system( Garcia-Herrero and Santabarbara, 2008). City commercial banks were found to outperform state-owned commercial banks, suggesting diversity in terms of ownership is key to better banking in China(ferri, 2009). Changes in ownership also has an impact on the lending behavior of banks, as lending by state-owned banks were found to be less prudent than lending by joint-Stock banks Jia, 2007) However, the benefits and efficiency improvements are not distributed evenly across banks and lenders. Smaller, less-regulated financial institutions appear more commercially oriented and gained more market shares in some areas after the reforms (Podpiera, 2006). Higher contestability in the banking sector helps alleviate financing constraints for small and medium enterprises( Chong et al., 2012). Joint-stock banks and city commercial banks were also found to have gained higher total factor productivity growth than state-owned banks in recent years( Chang et al, 2012) 3. A Stylized Model on Dual-Track Interest Rates System The setting of the model is similar to the framework in Chen et al. (201 1)and He and Wang(2012), which extend the model developed in Freixas and Rochet(2008). As in He and wang(2012), we introduce a dual-track interest rates system and focus on fund flows between the regulated banking system(the first track) and non-regulated money and bond market(the second track). In the banking sector, the central bank influence bank lending using the benchmark deposit rate and the rrr. In addition, bank lending can be affected by changes in the market price of funds due to arbitrage between the banking sector and capital markets. In this sense, the market price of funds in the money and bond markets is a shadow price to bank lending. In contrast to He and Wang(2012) which focuses on how monetary policy changes affect market interest rates, the new model in this paper pays more attention to how monetary policy and market interest rate affect both the price and quantity of bank lending. In order The"big five"banks are Industry and Commercial bank of China(ICBc), Bank of China(boc) Construction Bank of China(CBC), Agricultural Bank of China(ABC) and Bank of Communications BOCOM)
6 Within the banking sector itself, although it is still dominated by majority-state owned “big five” banks which hold about 47% of industry assets and roughly the same share of total deposits in the sector, contestability of the sector has also increased (PBC FSR Report, 2012).4 Admittedly, bank competition is to a significant extent constrained by relatively high entry barriers and interest rate regulation. The existing literature on the efficiency of the Chinese banking sector has mixed findings. Banks undergoing a foreign acquisition or public listing are found to have better preevent performance, but little performance changes were recorded after changes of ownership (Lin and Zhang, 2009). On the other hand, there is evidence that the Chinese banking system has benefited from the entry of foreign investors through higher profitability and increased efficiency in the banking system (Garcia-Herrero and Santabarbara, 2008). City commercial banks were found to outperform state-owned commercial banks, suggesting diversity in terms of ownership is key to better banking in China (Ferri, 2009). Changes in ownership also has an impact on the lending behavior of banks, as lending by state-owned banks were found to be less prudent than lending by joint-stock banks (Jia, 2007). However, the benefits and efficiency improvements are not distributed evenly across banks and lenders. Smaller, less-regulated financial institutions appear more commercially oriented and gained more market shares in some areas after the reforms (Podpiera, 2006). Higher contestability in the banking sector helps alleviate financing constraints for small and medium enterprises (Chong et al., 2012). Joint-stock banks and city commercial banks were also found to have gained higher total factor productivity growth than state-owned banks in recent years (Chang et al., 2012). 3. A Stylized Model on Dual-Track Interest Rates System The setting of the model is similar to the framework in Chen et al. (2011) and He and Wang (2012), which extend the model developed in Freixas and Rochet (2008). As in He and Wang (2012), we introduce a dual-track interest rates system and focus on fund flows between the regulated banking system (the first track) and non-regulated money and bond market (the second track). In the banking sector, the central bank influence bank lending using the benchmark deposit rate and the RRR. In addition, bank lending can be affected by changes in the market price of funds due to arbitrage between the banking sector and capital markets. In this sense, the market price of funds in the money and bond markets is a shadow price to bank lending. In contrast to He and Wang (2012) which focuses on how monetary policy changes affect market interest rates, the new model in this paper pays more attention to how monetary policy and market interest rate affect both the price and quantity of bank lending. In order 4 The “big five” banks are Industry and Commercial bank of China (ICBC), Bank of China(BOC), Construction Bank of China(CBC), Agricultural Bank of China(ABC) and Bank of Communications (BOCOM)
to analyse whether the impact is symmetric across firms of different size and across different phases of monetary policy stance, we introduce several cases with specific SSumptions into the model. In addition, loan demand and deposit supply functions are specified in the model to facilitate the computation of derivatives A Stylized Model Similar to previous studies, we assume N independent banks in the competitive banking sector and that n is sufficiently large so that each bank is a price taker. Each bank takes deposits(d)from households and makes loans(l) to firms in the loan market. Each bank has to submit required reserves to the central bank according to rrr(a) set by the PBC. In addition, each bank can buy central bank bills(Bi), on which the interest rate is set by the PBC (exogenous to each bank), and each bank can also invest in bonds or other financial products(NR, in the money and bond markets The key feature of the dual-track interest rate system is there exist a deposit-rate ceiling and a lending rate floor imposed by PbC in the banking sector. The deposit ceiling is in general considered binding while the lending-rate floor is not binding in most cases unable to maximize their profits as they do in a free market. In other words, the depos a ( Feyzioglu et. Al, 2009; He and Wang, 2012). The binding price control means banks market can not be cleared by market forces when the deposit rate ceiling is binding When the deposit ceiling is binding in the banking sector, bank i maximizes its profit as follows. I= Max(r L +raD+B+n D-C(D, L)(1) st.r≤rb Where r is the lending rate, r, is the deposit rate, r is the deposit rate ceiling, r, is the interest rate paid on required reserves, and r, is the market rate in the non-regulated market. C(D, L) is the bank's managing cost, which is a function of deposits and loans NR is the net position of bank i in the non-regulated market, which is given by NR1=D1-L-0D1-B;(2) Given that the deposit rate is binding and that the lending rate is not binding, the profit maximization function changes as follows I,= Max(L+raD(a)+,B,+r -PD ()-C(D, L))(3) Note that here the deposit function is determined solely by the supply of savings, and therefore, D is a function solely of r. In the capital wholesale market, the supply function S(r, Ir)is also a function of r, where r is exogenous and is determined by
7 to analyse whether the impact is symmetric across firms of different size and across different phases of monetary policy stance, we introduce several cases with specific assumptions into the model. In addition, loan demand and deposit supply functions are specified in the model to facilitate the computation of derivatives. A Stylized Model Similar to previous studies, we assume N independent banks in the competitive banking sector and that N is sufficiently large so that each bank is a price taker. Each bank takes deposits ( ) Di from households and makes loans ( ) Li to firms in the loan market. Each bank has to submit required reserves to the central bank according to RRR (α ) set by the PBC. In addition, each bank can buy central bank bills ( ) Bi , on which the interest rate is set by the PBC (exogenous to each bank), and each bank can also invest in bonds or other financial products ( ) NRi in the money and bond markets. The key feature of the dual-track interest rate system is there exist a deposit-rate ceiling and a lending rate floor imposed by PBC in the banking sector. The deposit ceiling is in general considered binding while the lending-rate floor is not binding in most cases (Feyzioglu et. Al, 2009; He and Wang, 2012). The binding price control means banks are unable to maximize their profits as they do in a free market. In other words, the deposit market can not be cleared by market forces when the deposit rate ceiling is binding. When the deposit ceiling is binding in the banking sector, bank i maximizes its profit as follows: { ( , )} , , l i r i b i nr i d i i i Li Di Bi Πi = Max r L + rαD + r B + r NR − r D −C D L (1) st. b d d r ≤ r Where l r is the lending rate, d r is the deposit rate, b d r is the deposit rate ceiling, r r is the interest rate paid on required reserves, and nr r is the market rate in the non-regulated market. ) ( , C Di Li is the bank’s managing cost, which is a function of deposits and loans. NRi is the net position of bank i in the non-regulated market, which is given by NRi = Di − Li −αDi − Bi (2) Given that the deposit rate is binding and that the lending rate is not binding, the profit maximization function changes as follows: { ( ) ( ) ( , )} , , i i b d s i b b i nr i d b d s l i r i Li Di Bi Πi = Max r L + rαD r + r B + r NR − r D r −C D L (3) Note that here the deposit function is determined solely by the supply of savings, and therefore, s D is a function solely of b d r . In the capital wholesale market, the supply function ( , ) nr b d S r r is also a function of b d r , where b d r is exogenous and is determined by
the central bank. Based on the above simple model, we try to answer our research questions one by one It can be proved that, under this scenario the equilibrium loan rate and loan size can be written as follows(proofs can be found in Appendix a): AD+N N+8 L-N(AD-A'2 (5) Where AD is aggregate demand for loans in the economy, S, is managing cost in the banking sector and a, is firms'price sensitivity for banking loans. From the above two equations, we can see that four factors could affect loan rate and loan size: r,(the market interest rate in free capital market), AD, 8 and A. Interestingly, changes of monetary policy instruments such as benchmark deposit rate or rrr do not enter loan equations directly, however, they could affect loan pricing and loan size indirectly since monetary policy changes will affect the market interest rate(Proofs can be found in Appendix A) Theoretical predictions When the deposit rate ceiling is binding and lending rate floor is not binding, loan rate increases with the market interest rate, while loan quantity decreases with the market interest rate. Monetary policy instruments can also affect bank lending through th market interest rate: loan rate increases with the benchmark deposit rate and RRR, while the loan quantity decreases when the Pbc raises the benchmark deposit rate and rRR 4. Empirical Analysis 4.1 Empirical Specification The goal of empirical models is to test the theoretical predictions and the theoretical model provides a good guideline for empirical specification. However the reality is much more complicated than that in the simple model. First, we need to identify the most likely scenario in the real world: the deposit rate ceiling is binding while the lending rate floor is not binding. Even though the credit quota may be imposed on the banking sector when necessary, it is generally believed that the PbC tends to use it as little as possible, especially in recent years. Therefore, our empirical models are based on the simple scenario without credit quota, although the potential impact of credit quota will be discussed in the caveat. Now we discuss empirical factors impacting bank lending one by a) Policy instruments and the market interest rate The first theoretical prediction illustrates that when the deposit rate ceiling is binding and lending rate floor is not binding, loan rate increases with the market interest rate in the
8 the central bank. Based on the above simple model, we try to answer our research questions one by one. It can be proved that, under this scenario the equilibrium loan rate and loan size can be written as follows (proofs can be found in Appendix A): L l L nr l N AD Nr r δ λ δ + + = * (4) L l l nr N N AD r L δ λ λ + − = ( ) * (5) Where AD is aggregate demand for loans in the economy, L δ is managing cost in the banking sector and λl is firms’ price sensitivity for banking loans. From the above two equations, we can see that four factors could affect loan rate and loan size: nr r (the market interest rate in free capital market), AD , L δ and λl . Interestingly, changes of monetary policy instruments such as benchmark deposit rate or RRR do not enter loan equations directly, however, they could affect loan pricing and loan size indirectly since monetary policy changes will affect the market interest rate (Proofs can be found in Appendix A). Theoretical predictions When the deposit rate ceiling is binding and lending rate floor is not binding, loan rate increases with the market interest rate, while loan quantity decreases with the market interest rate. Monetary policy instruments can also affect bank lending through the market interest rate: loan rate increases with the benchmark deposit rate and RRR, while the loan quantity decreases when the PBC raises the benchmark deposit rate and RRR. 4. Empirical Analysis 4.1 Empirical Specification The goal of empirical models is to test the theoretical predictions and the theoretical model provides a good guideline for empirical specification. However the reality is much more complicated than that in the simple model. First, we need to identify the most likely scenario in the real world: the deposit rate ceiling is binding while the lending rate floor is not binding. Even though the credit quota may be imposed on the banking sector when necessary, it is generally believed that the PBC tends to use it as little as possible, especially in recent years. Therefore, our empirical models are based on the simple scenario without credit quota, although the potential impact of credit quota will be discussed in the caveat. Now we discuss empirical factors impacting bank lending one by one. a). Policy instruments and the market interest rate The first theoretical prediction illustrates that when the deposit rate ceiling is binding and lending rate floor is not binding, loan rate increases with the market interest rate in the
free capital markets, while loan quantity decreases with the market interest rate. The central bank can also impact bank lending through the market interest rate: loan rate increases with the benchmark deposit rate and rrR, while the loan quantity decreases when the PBC raises the benchmark deposit rate and RRR. The benchmark deposit rate is well defined in reality: we use one-year deposit rate set by PBC as the benchmark deposit rate in empirical models. However, the role of rrR has changes a lot because the liquidity situation in Chinese banking system is overwhelmed by large foreign exchange purchases in recent years and rrr has been used as a main ool to drain the liquidity released from the sterilization. Therefore, we also include foreign exchange purchases in our empirical models to better identify the impact from the RRR In addition to shocks from the central bank, it is easy to see that bank lending could be impacted by various shocks from financial markets and the real economy through the market interest rate. For example, large fund-raising events like large IPOs in the stock market withdraw a lot of liquidity from the market, which could push the market interest rate higher and affect bank lending. Shocks from the real economy such as capital inflows/outflows also affect the market interest rate and bank lending which suggests the market interest rate includes more information other than monetary policy changes Therefore, in this empirical analysis we also include the market interest rate from non regulated markets after controlling for monetary policy changes. In the empirical model, we use 7-day Repurchase Agreement interest rate(7-day Repo) to represent the market interest rate in the free capital market since the 7-day repo rate is the most widely used indicator for capital price in non-regulated market b). Aggregate loan demand It is easy to see that both price and quantity of loans depend positively on aggregate loan demand because better economic conditions increase the profitability of projects and hence increase the demand for credit(Kashyap et al., 1994). Empirically, we use fixed asset investment growth as the proxy to represent loan demand in banking sector since most loans in our sample are for medium-long term investment projects. In reality the nominal interest rate on loans is also affected by macroeconomic variables such as money supply and inflation. Therefore, we also include money supply and inflation in the empirical models to control for macroeconomic conditions c). Credit risk The management of risk is a major issue for loan making as banks have to control various risks(credit risks, liquidity risks and market risks etc. ) that are inherent in their business Among these risks, credit risk is the most important in bank lending and banks charge borrowers different risk premium according to their features. Measuring credit risk means evaluating the probability of default by a particular borrower after taking into account various risk diversification and hedging arrangements In theory, issuing central bank bills could impact the market interest rate, however, the impact is not significant empirically(He and Wang, 2012) 6 The level of risks related to bank loans also depends on many institutional arrangements, and a
9 free capital markets, while loan quantity decreases with the market interest rate. The central bank can also impact bank lending through the market interest rate: loan rate increases with the benchmark deposit rate and RRR, while the loan quantity decreases when the PBC raises the benchmark deposit rate and RRR. The benchmark deposit rate is well defined in reality: we use one-year deposit rate set by PBC as the benchmark deposit rate in empirical models. However, the role of RRR has changes a lot because the liquidity situation in Chinese banking system is overwhelmed by large foreign exchange purchases in recent years and RRR has been used as a main tool to drain the liquidity released from the sterilization. Therefore, we also include foreign exchange purchases in our empirical models to better identify the impact from the RRR.5 In addition to shocks from the central bank, it is easy to see that bank lending could be impacted by various shocks from financial markets and the real economy through the market interest rate. For example, large fund-raising events like large IPOs in the stock market withdraw a lot of liquidity from the market, which could push the market interest rate higher and affect bank lending. Shocks from the real economy such as capital inflows/outflows also affect the market interest rate and bank lending, which suggests the market interest rate includes more information other than monetary policy changes. Therefore, in this empirical analysis we also include the market interest rate from nonregulated markets after controlling for monetary policy changes. In the empirical model, we use 7-day Repurchase Agreement interest rate (7-day Repo) to represent the market interest rate in the free capital market since the 7-day Repo rate is the most widely used indicator for capital price in non-regulated markets. b). Aggregate loan demand It is easy to see that both price and quantity of loans depend positively on aggregate loan demand because better economic conditions increase the profitability of projects and hence increase the demand for credit (Kashyap et al., 1994). Empirically, we use fixedasset investment growth as the proxy to represent loan demand in banking sector since most loans in our sample are for medium-long term investment projects. In reality the nominal interest rate on loans is also affected by macroeconomic variables such as money supply and inflation. Therefore, we also include money supply and inflation in the empirical models to control for macroeconomic conditions. c). Credit risk The management of risk is a major issue for loan making as banks have to control various risks (credit risks, liquidity risks and market risks etc.) that are inherent in their business. Among these risks, credit risk is the most important in bank lending and banks charge borrowers different risk premium according to their features. Measuring credit risk means evaluating the probability of default by a particular borrower after taking into account various risk diversification and hedging arrangements.6 5 In theory, issuing central bank bills could impact the market interest rate, however, the impact is not significant empirically (He and Wang, 2012). 6 The level of risks related to bank loans also depends on many institutional arrangements, and a
In theory it can be proved that the risk premium charged by banks increases with the debt-to-asset ratio and maturity of loans( Freixas and Rochet, 2008). In practice, many banks employ various models to measure credit risks. Those models generally have multiple indicators outlining various aspects of the risk related to borrowing firms The most commonly used indicators include total asset, total employment, liquid asset ratio, debt-to-asset ratio, profit margin and equity-to-debt ratio etc. In addition, some qualitative indicators are also included: The sector of the firm, the area where the firm is located, the ownership of the firm etc(Mu, 2007). Of course, the riskiness of loans is affected by the existence of collateral, which will be considered in loan pricing On the other hand various contract features are also used to mitigate credit risk in bank lending practice to enhance their ability to monitor borrowers over the course of the relationship(Strahan, 1999). Loan size is one of these features, which limits the banks potential exposure to credit risk of a specific borrower. Other features such as maturity and collateral of loans also play important roles in reducing credit risk in lending practice d ). Bank efficiency and price sensitivity of loans According to Equations(4)and (5), both loan rate and loan amount are also affected by efficiency of bank lending, which can be largely measured by managing costs of banks such as screening, monitoring and branching costs etc. dr N(AD-2r) >0 since ad-2r>0 and. >r d6(N+824)2 dL -N2(AD-1rnr) o8(N+84) <0 since AD-n, >0, and n,>0 (7) From Equation(6)and(7)it is easy to see that loan pricing is positively correlated to banks' managing costs and the opposite applies to loan amount. However, in the empirical analysis it is hard to get detailed data about managing cost of screening and monitoring since this data is usually commercially confidential. The way we deal with this issue is to include dummies variables for banks so as to control for divergent efficiency across them. Another factor impacting loan pricing is price sensitivity of loans, which is clearl negatively related to loan rate from Equation(4). Intuitively it makes perfect sense that banks could charge a firm with higher prices when loan demands from firms are insensitive to loan rates. In the empirical study we assume the price sensitivity of loans can in general be captured by a firm' s characteristics and dummies representing firm's industries 4.2 What determines loan rate and loan size? comprehensive discussion about this can be found at Chapter 8 in Freixas and Rochet (2008). ompanies like Standard Poor's, Moody's Analytics, Fitch Ratings, and Dun and Bradstreet provide such services. Most Chinese commercial banks have adopted quantitative credit risk evaluation models, for example, ICBC use s rating system similar to Standard Poors to evaluate credit risk from borrowers(Mu
10 In theory it can be proved that the risk premium charged by banks increases with the quasi debt-to-asset ratio and maturity of loans (Freixas and Rochet, 2008). In practice, many banks employ various models to measure credit risks.7 Those models generally have multiple indicators outlining various aspects of the risk related to borrowing firms. The most commonly used indicators include total asset, total employment, liquid asset ratio, debt-to-asset ratio, profit margin and equity-to-debt ratio etc. In addition, some qualitative indicators are also included: The sector of the firm, the area where the firm is located, the ownership of the firm etc (Mu, 2007). Of course, the riskiness of loans is affected by the existence of collateral, which will be considered in loan pricing. On the other hand, various contract features are also used to mitigate credit risk in bank lending practice to enhance their ability to monitor borrowers over the course of the relationship (Strahan, 1999). Loan size is one of these features, which limits the bank’s potential exposure to credit risk of a specific borrower. Other features such as maturity and collateral of loans also play important roles in reducing credit risk in lending practice. d). Bank efficiency and price sensitivity of loans According to Equations (4) and (5), both loan rate and loan amount are also affected by efficiency of bank lending, which can be largely measured by managing costs of banks such as screening, monitoring and branching costs etc. 2 ( ) ( ) L l l nr L l N r N AD r δ λ λ δ + − = ∂ ∂ >0 since l l AD − λ r >0 and l nr r > r (6) 0 ( ) ( ) 2 < + − − = ∂ ∂ L l l l nr L N L N AD r δ λ λ λ δ since l l AD − λ r >0, and λl >0 (7) From Equation (6) and (7) it is easy to see that loan pricing is positively correlated to banks’ managing costs and the opposite applies to loan amount. However, in the empirical analysis it is hard to get detailed data about managing cost of screening and monitoring since this data is usually commercially confidential. The way we deal with this issue is to include dummies variables for banks so as to control for divergent efficiency across them. Another factor impacting loan pricing is price sensitivity of loans, which is clearly negatively related to loan rate from Equation (4). Intuitively it makes perfect sense that banks could charge a firm with higher prices when loan demands from firms are insensitive to loan rates. In the empirical study we assume the price sensitivity of loans can in general be captured by a firm’s characteristics and dummies representing firm’s industries. 4.2 What determines loan rate and loan size? comprehensive discussion about this can be found at Chapter 8 in Freixas and Rochet (2008). 7 Companies like Standard & Poor's, Moody's Analytics, Fitch Ratings, and Dun and Bradstreet provide such services. Most Chinese commercial banks have adopted quantitative credit risk evaluation models, for example, ICBC use s rating system similar to Standard & Poor's to evaluate credit risk from borrowers (Mu, 2007)