Information disclosure and bank risk taking behavior under partially implicit deposit insurance system: Evidence from China

Transcription

1 Inforation disclosure and bank risk taking behavior under artially ilicit deosit insurance syste: Evidence fro China a WANG Zongrun, a WAN Yuanyuan, b Jin Yanbo, a ZHOU Yanju a Business school, Central South University, Changsha, China b Deartent of Finance, Financial Planning, and Insurance, California State University, Northridge, 18111, CA, USA Abstract: We exaine the role of inforation disclosure on banks risk-taking in China. China s banking syste has an ilicit deosit insurance syste that big banks, esecially the state-owned banks, would be rescued by governent when in trouble. We use a gae theory odel to analyze the interaction echanis between banks risk-taking behavior and inforation disclosure under this artially ilicit guarantee. We find that under such syste, banks tend to take excessive risk. In addition, the ilicit guarantee weakens the arket onitoring effect through disclosure. On the other hand, we show that banks without 100% ilicit guarantee tend to take less risk. Using data fro 30 coercial banks in China, we find strong eirical suort to our findings. Key words:disclosure;bank risk taking;artially ilicit deosit insurance 1. Introduction Inforation disclosure is an iortant eans of arket disciline. In the banking industry, investors and deositors alike use a bank s inforation disclosure to gauge its risk and ake investent decisions. However, under an exlicit or ilicit deosit insurance syste, they ay not have adequate incentive to do so. This would weaken the arket discilinary function of the disclosure. In China, the Chinese governent has rescued any banks that are considered "too big to fail". 1 This suggests that although China has not yet established an exlicit deosit insurance syste, the ilicit deosit insurance syste exists. This would no doubt iact investors interest to risk inforation. Eirically, researches (Zhang & 1 Soe exales of rescues rovided by governent include: 1). the establishent of four big governent-owned asset anageent coanies to reove banks non-erforing loans off their balance sheet and into the custody of these asset anageent coanies; 2). Issuance of secial governent bond to rovide caital injection into the banks; 3). Direct funding suort when the banks went IPO; and 4). The olicies that allow banks to ileent debt-for-equity swa and debt-for- caital swa. 1

2 He, 2005; Li & Han, 2008) and Fitch s suort rating(aendix 1) suggest that the robability of a bailout deends on the tye and size of the bank. In this aer, using Gae Theory, we exaine the link between inforation disclosure and banks risk-taking behaviour, under such artially ilicit deosit insurance syste. We show that banks tend to take excessive risk under ilicit deosit insurance. However, for banks without 100% ilicit insurance, ore inforation disclosure could lead to less risk-taking. On the other hand, ilicit deosit insurance daens the relation between inforation disclosure and risk-taking. We use data fro 30 coercial banks in China to eirically test our theory. The results are consistent with the theoretical conclusion. More secifically, we find that large banks and banks with ore governent stake tend to take ore risk. On the other hand, we see a stronger relationshi between the level of inforation disclosure and risk-taking aong saller banks and banks with less or no governent stake. This suggests that the arket discilinary role of inforation disclosure works well when the ilicit insurance is not guaranteed. However, such role is weakened when investors are certain that banks are known to have ilicit governent insurance. Our aer contributes to the literature by develoing a theory that incororates the ilicit artial deosit insurance into the relation between inforation disclosure and banks risk taking. We also eirically test the roositions and find strong suort to the theory. More secifically, we find that ilicit insurance syste daens the role inforation disclosure lays on banks risk-taking. This has olicy ilications for the evolving Chinese banking regulation. Mandatory inforation disclosure ay enhance the arket function that inforation disclosure lays on banks risk taking behaviour. Our aer is organized as follows. In section 2, we discuss literature review. Section 3 describes the gae theory odel and section 4 shows the eirical analysis. Finally, section 5 concludes. 2. Literature Review The otential for arket disciline has been recognized by the Basel Coittee on Banking Suervision, which has roosed that bank caital regulation ought to be coleented by ore arket disciline through the use of external credit assessent agencies. In order to infor and facilitate effective arket disciline, the 2

3 Coittee has highlighted the need for ore inforation disclosure. On the other hand, as one of the eans of arket disciline, the iortance of inforation disclosure is also suorted by a growing eirical literature, which finds that disclosure tends to irove credit arket erforance. The ore inforation disclosed to the ublic, the stronger the arket disciline is. For exale, Mayes & Hale (2001) shows that iroving bank disclosure is helful for investors to gauge banks risk, and encourage banks to ileent risk anageent and irove efficiency. Hirtle(2007)studies the relationshi between the bank inforation disclosure and its subsequent effect on risk and erforance. The results suggest that the greater disclosure is associated with ore efficient risk taking and thus iroves risk-return tradeoff. Deirgüç-Kunt et. al. (2008) find that banks receive ore favorable Moody s financial strength ratings in countries with better coliance with Basel Core Princiles related to inforation rovision. By easuring bank soundness through Moody s ratings and Z-score, and the level of inforation disclosure through the coliance with Basel Core Princiles, the study shows that bank soundness is ositively related to inforation disclosure. Hakiah & Ibrahi (2011) rovides a anel data analysis of the interrelationshi aong disclosure, risk and erforance in Islaic banks. Using leverage as the easure of risk and ROA as the easure of bank rofitability, the study shows that both rofit and risk are significantly related to inforation disclosure. Houston et al. (2010) finds that the benefits of inforation sharing aong creditors aear to be universally ositive. Greater inforation sharing leads to higher bank rofitability, lower bank risk, a reduced likelihood of financial crisis and higher econoic growth. Soe scholars consider the effect of inforation disclosure on bank risk taking under deosit insurance syste,. Vaez-Zadeh et al.(2002)uses a sile odel for the deterination of the otial level of insurance coverage, it is shown that the otial coverage is higher for develoing countries coared to develoed countries; a condition that is broadly satisfied by revailing deosit insurance ractices around the world. Diaond and Rajan (2000) identifies overlooked consequences of having regulatory caital requireents and deosit insurance. Deirgü-Kunt and Huizinga( 2004) suggests that exlicit deosit insurance reduces required deosit interest rates, while at the sae tie it lowers arket disciline on bank risk taking. They use a newly constructed data set of deosit insurance design features to exaine how different design features affect deosit interest rates and arket disciline. Xu et al.( 3

4 2009 ) eirically studies the effect of banks inforation disclosure on their risk-taking behaviour based on the data of 14 riary coercial banks fro 2000 to It indicates that the function of inforation disclosure as reducing the risk-taking behaviour of banks deends on the corresonding institutional foundation and arket environent. The arket disciline actions of creditors can coletely serve as reducing the risk-taking behaviors of banks only under the higher financial arketization degree and the colete and effective risk inforation disclosure of banks. Bourgain et al.(2012)finds a new channel through which deositors can exercise ressure to control risk-taking. Deositors can reallocate their savings away fro their hoe country to the ore rotective syste of a develoed econoy. They find that sufficiently high financial oenness is necessary for a ositive link between financial transarency and safe risk anageent. And they test the relationshi between disclosure, financial oenness and bank risk-taking for a anel of 258 banks fro the MENA region and Turkey. Vauhkonen(2012)considers the iact of andatory inforation disclosure on bank safety in a satial odel of banking coetition, in which a bank s robability of success deends on the quality of its risk easureent and anageent systes. Wilson et al.(2012)finds that the rudential regulation of banks in New Zealand relies heavily on the ublic disclosure of risk inforation. This work reorts a significant relationshi between deosit risk reius and disclosure risk indicators, suggesting New Zealand's disclosure regie is effective in oderating excessive risk-taking in banks. But, as the relationshi is found to be strongest before ublication, it cannot be attributed directly to arket-disciline. Instead, it is suggested self-disciline. Wu & Bowe(2010)analyses inforation fro a corehensive sale of Chinese banks over the eriod. They find joint-equity banks that disclose ore inforation to the ublic aintain larger caital ratios; and banks that release ore transarent financial inforation hold ore caital against their nonerforing loans. Wu & Bowe ( 2012 ) investigates the relationshi between inforation disclosure and deositor behaviour in the Chinese banking sector. they show that banks with ore transarent inforation disclosure is ore likely to exerience growth in its deosit base and ore able to attract funds by offering higher interest rates. 4

5 3. The odel 3.1 Assutions Consider two tyes of banks: G and F, where Bank G (such as the five state-owned banks) is certain to be rescued by the governent when in trouble, and Bank F ay or ay not be rescued by the governent. As a result,bank F has coarative disadvantage in attracting deositors. In order to coete with Bank G, Bank F needs to increase the interest rates on savings deosits to attract deositors. Denote the interest rate offered by Bank G as l G, lg (0,1) ;the interest rate offered by Bank F as l F,and lf lg l,where l [0,1) and lf (0,1). l is the interest rate reiu coared to that offered by Bank G. It can also be viewed as the coensation to the deositors because of the lack of ilicit guarantee fro the governent coared to Bank G. Let x be the roortion of the deosit a deositor allocates to Bank F, where x [0,1]. The total aount of the deosit is R, hence the aount deosited to Bank F is Rx.The aount of deosit allocated to Bank G is then R(1 x).we assue that deositors have reference for risk, 2 if the exected revenue are identical obtained fro Bank F and Bank G. As the interest rate offered by Bank F is higher, the incoe which investors can receive is higher, so investors choose Bank F. Bank F chooses a strategy s, s { M, }. stands for excessive risk-taking, M indicates sound risk-taking strategy. If action is chosen, bank yields r with robability, (0,1) and is zero otherwise. If action M is chosen, bank yields r M with robability of 1, r r M.That is, we assue that with excessive risk-taking, Bank F can also achieve higher incoe. In addition, if the bank is in trouble, Bank F deositors can receive the coensation k.that is, k can be viewed as the roxy of 2 We adit that this is a weakness of the odel. However, with this assution the solutions to the roble are ore tractable. 5

6 the governent suort received by Bank F, k [0,1). On the other hand, Since Bank G has the ilicit guarantee fro the governent, we assue that Bank G yields r with robability of 1, r ( r, r ). Here r deends on the level of the bank s risk taking strategy. 3.2 Decision Process x [0,1]. Ste1.Bank F chooses the risk strategy M or ; Ste2.Bank F chooses the degree of disclosure; Ste3.Deositors choose the roortion of deosits allocated to Bank F: x, 3.3 Deositors We assue that the Bank F has both infored and uninfored deositors. In addition, the fraction of infored deositors is denoted ( 0 1).We use the araeter as a roxy for the degree of inforation disclosure. I indicates infored deositors,u indicates uninfored deositors. The exected utility of infored deositors of Bank F becoes: M U F (1 lf ) (1 ) k if s= ( xi) 1+ lf if s= M The exected utility of the uninfored deositors of Bank F becoes: U ( x ) (1 l ) (1 ) k if s {, M} F U F (1) (2) We assue that since deositors of Bank F had no inforation of the bank, they assue Bank F takes excessive risk when aking investent decisions. The strategy of infored deositors of Bank F is detailed below: Since the deositors of Bank G receive an incoe of 1 l with the robability G of 1. If bank takes the action M,Sincel F l,deositors all choose Bank F, G x 1;If bank takes the action : M I If (1 lf ) (1 ) k 1 l,that is: (1 )(1 lg G l,deositors all 6

7 choose Bank F, x 1; I (1 )(1 lg If l,deositors all choose Bank G, xi 0; The strategy of uninfored deositors of Bank F becoes: If If (1 )(1 lg l,deositors all choose Bank F, xu 1; (1 )(1 l l,deositors all choose Bank G, xu 0; 0 G Cobining the investent function by calculating x x (1 ) x,we get the following results: I U In the first case,when (1 )(1 lg l : x xi (1 ) xu 1, M M M x x (1 ) x 1. That is, if the interest rate offered by Bank F is high I U enough, the rate reiu relative to Bank G can coensate the risk taken by deositors. Regardless if bank F adots action or M,deositors will all choose Bank F. (1 )(1 lg In the second case,if l : F; x x (1 ) x 0,when Bank F adots, no deositors would choose I U M M M x x (1 ) x,if F adots M, a roortion of of deositors I U would choose F. That is, only deositors with the inforation of Bank F would choose to deosit in Bank F. Under these circustances, if Bank F increases the degree of disclosure, ore deositors becoe infored, which increase the total deosits to Bank F. 3.4 Bank s strategy decision If Bank F takes excessive risk, the revenue of Bank F is F,Bank F needs to 7

8 choose the roer l to attract deosits and axiize rofit. Max l F = Rx ( r- lf) (3) If (1 )(1 lg l, x 1. F Since 0, we can conclude that when l (1 )(1 lg l, rofit is axiized; if (1 )(1 l l, x 0.That is, no deositors would choose F. 0 G Therefore, Max =0 l F As a result, Bank F would choose (1 )(1 lg l,and the revenue is (1 )(1 lg R( r-l G- ). M F. If Bank F choose sound risk-taking strategy M, the revenue of Bank F is M M Max l F = x R( rm- lf) (4) If (1 )(1 lg M l, x 1,Therefore, when (1 )(1 lg l, bank F s rofit is axiized; When (1 )(1 l l,since Bank F adots 0 G sound risk taking strategy, x M, and l 0. Bank F would be able to not only attract deositors at the sae interest rate, but also achieve rofit axiization: M Max l F = R( rm- lg) Based on the two cases stated above, when Bank F adots sound risk-taking strategy, the otial rofit function is (1 )(1 lg M ax{ R( rm-lg- ), R( rm- lg)}.when F - F is ositive, Bank F 8

9 would adot M. That is, If (1 )(1 ) (1 )(1 ) ax{ ( - - l G k G M G ), ( M- G)} ( Rr l Rr l Rr l k lg ) (5) bank F adots sound risk-taking strategy rm lg (1 )(1 lg (1 )(1 lg In other words,if r l 2 G,bank F chooses sound risk taking rm lg (1 )(1 lg (1 )(1 lg strategy.if r l 2 G, bank F can only increase the degree of disclosure to let investors know bank s risk-taking strategy, so that they can realize rofit axiization when they adot sound risk-taking strategy. Since (1 )(1 l = Rx ( r -l - ), under this condition, the bigger the k is, G F G the ore revenue the bank will get, indicating that ilicit insurance syste would otivate Bank F to take excessive risk; Bank G gains r with the robability of 1, r ( r, r ).as inforation disclosure does not influence the risk taking strategy. As a M result, Bank G tends to take excessive risk to obtain high rofit. We can then derive roosition 1: Proosition 1:Under the governent ilicit guarantee,banks tend to take excessive risk; Fro equation (5) we find that (1 )(1 lg r ( lg ) r l tends to choose sound risk taking strategy. This leads us to roosition 2: 9 M G (6). Bank F Proosition 2:With ore inforation disclosure, banks without 100% ilicit guarantee fro the governent tend to adot sound risk-taking strategy. Moreover, fro equation (6), we see that the higher k is, the higher will be.that is, with ore ilicit guarantee fro the governent, banks would need to disclose ore inforation to ake the sound risk-taking strategy otial.. Therefore, governent s ilicit guarantee can weaken the arket disciline of inforation

10 disclosure. Fro this, we derive roosition 3: Proosition 3:Ilicit guarantee fro the governent would weaken the arket discilinary function of inforation disclosure. 4. Eirical analysis In this section, we use data fro Chinese banks to eirically test the roositions discussed in the revious section. 4.1 Sale selection Our sale consists of 5 state-owned banks, 7 national joint-equity coercial banks and 18 urban coercial banks 3. The data sans fro 2006 to Data on inforation disclosure, financials and the Fitch s suort rating are obtained fro Bankscoe database; all other acroeconoic data are collected fro statistical bulletins and the website of the Peole s Bank of China Disclosure Indexes Following the ethodologies develoed in revious studies 5 in inforation disclosure easureent, we select the following ites fro Bankscoe to easure the level of bank disclosure:balance sheet(non-earning Assets, Off-Balance-Sheet Ites Interbank Ratio 6 );incoe stateent(net Incoe, Net Interest Margin, ROAA (Return On Average Assets), ROAE(Return On Average Equity), Cost To Incoe Ratio), Credit risk rofile (Loan Loss Reserves, Loan Loss Provisions, Iaired Loans / Gross Loans) and caital sufficiency(total Caital Ratio, Tier 1 Ratio, Total Caital, Tier 1 Caital).For each ite, S=1 if the bank discloses the inforation, S=0 otherwise. We then calculate the coosite inforation index as the 3 The banks are classified according to the classification of China Banking Regulatory Coission. 4 The link is htt:// Peole s bank of China is China s central bank. 5 See Bauann &Nier (2004), Loes &Rodrigues (2007), Dahawy (2009), Hakiahand Lbrahi (2010) and Zhang and Liao(2010). 6 The interbank ratio is the ratio of oney lent to other banks divided by oney borrowed fro other banks. A ratio greater than one indicates that the bank is a net lacer in the interbank arket and is therefore ore liquid. By contrast, if the ratio is below one, the bank is a net borrower in the interbank arket and heavily reliant on interbank deosits to fund its assets. (Banking Coetition and Caital Ratios, by Klaus Schaeck and Martin Čihák, IMF working aer 2007) 10

11 equally-weighted average of the 15 ites. Thus, the axiu score a bank could have on disclosure is 15. And the range on inforation disclosure index is [0,1]. The calculation ethod is dislayed below: 15 1 DIS Si (1) 15 Where DIS is the disclosure index, S i is the disclosure on each ite. i Bank risk taking variables Following Deirgüç-Kuntet. al. (2008), we use Z-score as a coosite index of bank risk-taking behaviour. Z-score easures the robability of bankrutcy (Laeven and Levine,2009),and is widely used in eirical researches on banks (Agoraki et. al.,2011). Banks with higher Z-scores are ore robust in dealing with return volatilities and are considered less likely to encounter bankrutcy. The Z-score is defined as follows: E ( ROA) Zscore A (2) ( ROA) Where E indicates equity, A indicates total assets,roa= net incoe/ total assets, ( ROA) is the standard deviation of ROA, couted fro 2006 to We use leverage (Leverage = (equity + Loan Loss Reserves -non-erforing loans)/ total assets) to evaluate bank risk anageent. The higher the leverage is, the lower the risk the bank takes. According to Delis and Kouretas (2011), credit risk is the ost iortant risk for a bank. We use non-erforing loan ratio(non-erforing loan aount / gross loan aount) to easure a bank s risk-taking behaviour. Since low quality loans reduce a bank s interest incoe and increase its rovision for loan iairent, banks with ore iaired assets are ore likely to be bankrut (Wheelock and Wilson, 2000). We use Non-Perforing Loans Ratio (NPLR) to easure asset quality. 11

12 Table 1 describes the variables constructed in easuring bank risk taking. Table 1. Bank risk-taking: Variable Definitions Measures of VariableN Variable Construction bank ae risk-taking E q u ity Zscore Zscore ( ( R O A ) ) / ( R O A ) Leverage Non-erforing loan ratio LEV NPLR T o t a l A s s e t s (equity + loan reserve requireent - non-erforing loans)/total assets non-erforing loans/ total loans Research Design We use both ultivariate and univariate analysis to investigate the relationshi between disclosure and bank risk-taking. First, we include other control variables and uses ultivariate regressions to exaine the relation between disclosure and bank risk taking. In the ultivariate regressions, we build siultaneous equations odels and use Generalized Least Squares (GLS) ethod to estiate the araeters. Next, we exclude the state-owned banks and divide the banks into two grous according to their level of disclosure, and test the differences between grous using univariate analysis. 5. Eirical results 5.1. Multivariate Regression Model Setu In regression analysis, we use the variables that easure bank risk taking as deendant variables, and the one-year lagged risk disclosure index (DIS t-1 ) as the indeendent variable to test the relationshi between bank risk taking and disclosure. Since bank risk taking is related to both the individual bank characteristics and the acroeconoic factors, we choose control variables based on both levels. Below we discuss the choice of control variables in ore details. 12

13 Bank-level variables 1. Percentage of governent stake. Although China has been transitioning toward arket econoy in recent decades, governent and governent agencies are still heavily involved in the banking sector in the econoy. This ay have a significant iact on bank risk taking. The governent acts as both olicyaker and bank shareholders. This would increase the robability of a governent bailout. This also reduces banks incentives to anage risk. To exaine this issue, we use the ercentage of governent stake to easure the degree of governent involveent in banks. 2. Bank size. (SIZE) In addition, larger banks tend to be ore diversified, have less credit constraints and ore in-deth anageent. Therefore, bank size can affect bank soundness (Falkenstein et al.,2000). On the other hand, since any large banks are considered too big to fail in China, investors do not ay uch attention to banks risk disclosure since they believe the banks will be rescued by the governent if needed. We use the natural logarith of total assets to easure bank size (SIZE). 3. Fitch s suort rating. In China, banks can receive various suorts fro the central and local governent, based on their ownershi structure. Naturally, governent suort can irove bank soundness by reducing its risk and increasing its caital adequacy ratio, everything else held equal. To easure how likely a bank will receive governent suort, we use Fitch s suort rating fro Bankscoe database. The rating scale is fro 1-5, where 1 eans that the bank is ost likely to receive governent suort and 5 eans the bank is least likely to receive governent suort. 4. In the Aendix, we list the suort rating of all 13 banks in our sale with data available during Interestingly, we find that on average, governent suort to banks is rising over tie. Also, larger banks tend to 13

14 receive ore governent suort: aong the 13 banks, the correlation coefficient between the likelihood of governent suort and bank size is Besides, the correlation coefficient between the likelihood of governent suort and the ercentage of governent stake is 0.803, suggesting that banks with large governent stakes tend to receive higher suort. Esecially, for the five state-owned banks 7 in China, the Fitch suort rating were alost always equal to 1 (ost likely to receive governent suort) between 2004 and The evidence suggests that the robability of a governent bailout is highly correlated with bank size and governent stake. Since the ajority of the banks in our sale do not receive Fitch s suort rating, this aer uses the ercentage of governent stake and bank size to easure the robability of a bailout. Pre-IPO or Post-IPO. (LIST) Post-IPO banks can receive tiely feedbacks fro the caital arket. That hels the ake infored decisions which ay iact their risk taking behaviour. We use a duy variable that equals 0 for re-ipo years, and 1 for ost-ipo years.. 5. Loan Loss Provisions ratio (LLP) The ratio is easured by loan loss rovisions divided by total loans. This ratio is an iortant indicator of the adequacy of a bank s loan loss reserve. This reflects whether a bank has sound risk anageent ractice and whether its risk is controllable (Houston et al., 2010). 6. Fixed assets ratio (FA) This ratio equals fixed assets divided by total assets. It easures how efficiently a bank uses its assets. The lower this ratio, the ore efficient the bank is using its assets. This indicator controls for the difference in the structure of 7 The five state-owned banks are: Industrial and Coercial Bank of China, China Construction Bank, Bank of China, Bank of Counications, and Agricultural Bank of China. For the Agricultural Bank of China, data is only available for year

15 bank business.(bourgain et al.,2012) We use this ratio to easure for bank s asset usage efficiency. In addition to bank-level variables, we also include the following acroeconoic variables that ay iact a bank s risk-taking behaviour. Macroeconoic factors 1. GDP growth rate (GDPG) Eirical research shows that GDP growth rate is negatively related to banks robability of bankrutcy (Evrensel, 2008). Due to the cyclical nature of its business, bank risks can be attenuated by high econoic growth (Davis and Kari, 2008). We use the year-to-year growth rate in GDP (GDPG t ) to easure the level of econoic growth. Due to the varying econoic develoent level in each region, for state-owned banks and national coercial banks we choose the national GDP growth rate; For regional coercial banks, we use regional GDP growth rate. 2. M2 growth rate. (MG) Changes in the onetary olicy can affect a bank s credit suly and its risk-taking behaviour. Higher growth in oney suly ay increase coetition between banks, leading to ore risk taking. Monetary olicy iarts a direct iact on deosits and that deosits act as the driving force of bank lending. A reforulation of the bank lending channel is roosed that works riarily through the iact of onetary olicy on banks balance sheet strength and risk ercetion (Disyatat, 2011). In this aer, we use broad oney M2 growth to easure onetary olicy. Table 2 shows the descrition of exlanatory variables. 15

16 Table 2. Descrition on Exlanatory Variables Variable Nae Descrition DIS t-1 One-year lagged disclosure index Governentstake the ercentage of governent stake LIST Pre-IPO, LIST 0 ; Post-IPO, LIST 1 SIZE Natural log of total assets LLP LLP Non erforing loans reserve /total loans FA GDPG MG FA =fixed assets/ total assests Growth rate in GDP The growth of M Descritive Statistics Table 3 resents the suary statistics on all variables. There are several observations that are worth entioning. Firstly, we see that there is a huge variation on the non-erforing loan ratio (NPLR), fro 0.33% to 23.5%. This shows that the quality of bank assets varies substantially across different banks. Secondly, the average of disclosure index is coarative high. However, the iniu of that is ,indicating that soe banks have low levels of disclosure. Besides, the iniu of leverage is ,indicating that soe banks take excessive risk. Table 3. Suary Statistics Bank risk taking N Miniu Maxiu Mean Standard Deviation Zscore LEV NPLR (%) DIS Governentstake LIST SIZE LLP FA GDPG MG Regression Results Using the roxies of banks risk-taking as deendent variables, we test the following regression equation: Risk taking DIS Bank Controls Macro Controls it 0 1 it 1 2 it 3 it (3) Where, i indicates bank,t indicates tie, Bank Controls indicates the bank 16

17 level control variable, Macro Controls indicates the acro level control variables. We use Generalized Least Squares (GLS) ethod to estiate the araeters. Results are resented in Table 4. We see that there is a highly significant relationshi between the level of inforation disclosure and banks risk-taking behaviour. Disclosure is ositively related to Z-score and leverage, indicating that banks with ore inforation disclosure is less risky, since banks with higher Z-scores and higher leverage (using our definition) is less likely to encounter bankructy. Moreover, the level of disclosure is negatively related to non-erforing loan ratio. The results are consistent with Proosition 2, which suggests that banks with ore inforation disclosure tend to take less risk. In addition, the results show that larger banks or banks with higher governent stake tend to take ore risks. This is consistent with Proosition 1, that banks with ilicit governent guarantee tend to take ore risk. As discussed before, larger banks and banks with higher governent stake are shown to be ore likely to receive bailout when in trouble. Aong the control variables, ost-ipo firs tend to take less risk. This is consistent with the arket discilinary role of inforation disclosure, since listed banks are under ore ublic scrutiny and are therefore ore likely to adot sound risk-taking strategy. There is also soe evidence that M2 growth is ositively related to bank risk-taking, consistent with the rediction. 17

18 Table 4. Inforation Disclosure and Risk-taking Exlanatory Variable Intercet *** Zscore LEV NPLR ( ) DIS(-1) *** ( ) Governentstake ( ) SIZE *** ( ) LIST ( ) LLP ( ) FA ( ) GDPG * ( ) MG ** ( ) *** ( ) ** ( ) *** ( ) *** ( ) *** ( ) *** ( ) ** ( ) ( ) *** ( ) *** ( ) *** ( ) *** ( ) ( ) * ( ) ( ) ( ) ( ) ( ) R-squared Adjusted R-squared Observations Note: Standard errors are corrected for fir-level correlation and for heteroskedasticity. T-statistics are reorted in the arentheses. *, ** and *** denote significance at the10%, 5% and 1% levels, resectively (two-tailed test). Proosition 3 suggests that governent ilicit guarantee would weaken the arket discilinary function on inforation disclosure. In order to test this roosition, we construct two interaction variables (DIS(-1)*governent stake, and DIS(-1)*SIZE) to test the joint effect of the likelihood of ilicit governent guarantee and disclosure on bank risk-taking. As discussed before, both governent stake and bank size serve as roxies for the robability of receiving governent bailout. The Regression results are resented in Table 5. The results are consistent with Proosition 3. We see that ore disclosure is related to less risk-taking, as in Table 4. However, this relationshi is weakened when banks have a higher likelihood of ilicit guarantee. For exale, in Panel A, when governent stake is used as the 18

19 roxy for governent ilicit guarantee, in the NPLR equation, the coefficient on the interaction ter is ositive and significant, oosite to the coefficient on disclosure. In Panel B, when bank size serves as the roxy for ilicit guarantee, the Z-score regression has a ositive coefficient on disclosure but negative coefficient on the interaction ter. Again, this suggests that for banks that have ore ilicit guarantee, the iact of inforation disclosure on risk taking is daened. Table 5. Inforation Disclosure and Risk-taking: the effect of ilicit governent guarantee Panel A. Governent Stake as the Proxy for ilicit guarantee Exlanatory Variable Zscore LEV NPLR Intercet *** ( ) DIS(-1) ** ( ) Size *** ( ) LIST ( ) LLP * ( ) FA ( ) DIS(-1) *Governentstake ( ) GDPG * ( ) MG * ( ) *** ( ) ** ( ) *** ( ) *** ( ) *** ( ) ** ( ) *** ( ) ( ) ** ( ) ** ( ) *** ( ) ( ) ( ) ( ) ( ) *** ( ) ( ) ( ) R-squared Adjusted R-squared Observations Note: Standard errors are corrected for fir-level correlation and for heteroskedasticity. T-statistics are reorted in the arentheses. *, ** and *** denote significance at the10%, 5% and 1% levels, resectively (two-tailed test). 19

20 Panel B. Bank Size as the Proxy for ilicit guarantee Exlanatory Zscore LEV NPLR Variable Intercet *** * *** ( ) DIS(-1) *** ( ) Governentstake ( ) LIST * ( ) LLP * ( ) FA ( ) DIS(-1)*SIZE *** ( ) GDPG * ( ) MG * ( ) ( ) *** ( ) *** ( ) *** ( ) *** ( ) ** ( ) *** ( ) ( ) *** ( ) ( ) *** ( ) *** ( ) * ( ) ( ) ( ) ( ) ( ) ( ) R-squared Adjusted R-squared Observations Note: Standard errors are corrected for fir-level correlation and for heteroskedasticity. T-statistics are reorted in the arentheses. *, ** and *** denote significance at the10%, 5% and 1% levels, resectively (two-tailed test) Univariate Analysis As entioned before, the five state-owned banks have alost 100% ilicit guarantee by the governent. As such, they ay have different risk-taking incentives. As a robustness check, in this section we exclude these five banks and exaine the risk-taking behavior of all other coercial banks in the sale. We use univariate analysis to coare bank risk based on low and high level of disclosure. Secifically, we divide the sale into two grous, one with higher than average disclosure, and the other with lower than average disclosure. We then exaine whether there is any difference in the two grous risk-taking behavior. Table 6 resents the results. We see that banks with higher level of disclosure tend to have lower non-erforing loan ratio, an indication of low risk. The difference is statistically significant. The difference on Z-score and leverage is not significant. 20

21 Variable High-level disclosure banks Table 6 Univariate Analysis Low-level t-statistics(mean) disclosure Difference banks -Value Zscore LEV NPLR Conclusion Inforation disclosure is an iortant eans of arket disciline. In a fully functioning arket econoy, investors ay close attention to a bank s inforation disclosure in order to onitor its risk-taking behavior. This would revent banks fro taking excessive risks. However, if investors know that the banks are very likely to be rescued by governent when in trouble, they would have less incentive to onitor the disclosure. This weakens the function of disclosure on banks risk-taking. Although China does not have an exlicit deosit insurance syste, there sees to be an ilicit guarantee syste. Big banks, esecially the five state-owned banks, are widely exected to be bailed out when in trouble. Other banks ay also receive artial hel when a financial crisis occurs. In other words, Chinese banks are under a artially ilicit guarantee syste. This ay have different ilications on the relation between inforation disclosure and banks risk-taking behavior. In this aer, using Gae Theory, we analyze the interaction echanis of disclosure and banks risk-taking behavior under artially ilicit deosit insurance guarantee. We show that with ilicit insurance, banks tend to take excessive risk. In addition, with ore inforation disclosure, banks without 100% ilicit guarantee tend to take less risk. Moreover, the ilicit guarantee would daen the arket disciline that disclosure exerts on banks risk-taking behavior. Using data fro 30 Chinese coercial banks fro , we show that banks with ore disclosure tend to take less risk. However, this relation is strongest for saller banks and banks with little governent stake. In other words, when banks are not exected to be bailed out by governent, inforation disclosure serves as an effective onitoring syste, leading to less risk-taking. On the other hand, we show that big banks and banks with ore governent stake tend to take ore risk. Furtherore, the negative relation between the level of disclosure and risk-taking is daened for larger banks and bank with higher 21

22 governent stake. The evidence suggests that this ilicit deosit insurance syste discourages the investors fro onitoring banks risk-taking fro inforation disclosure. In order for the disclosure to serve as an effective eans of arket disciline, olicy akers need to ake changes in the banking regulation. One of the solutions is to ove to an exlicit deosit insurance syste. On the other hand, ore andatory disclosure can be roosed to force banks to disclose ore inforation. Both easures could otentially hel rotect deositors interest and revent a bank run in a financial crisis. Acknowledgent This research described in this aer was substantially suorted by a rogra (No. NCET ) for New Century Excellent Talents in University, grants (No , No , and No )fro the National Natural Science Foundation of China, and a roject(2011jq025) fro the National Basic Research Progra of China. 22

23 Reference Arnaud Bourgain, Patrice Pieretti, Skerdilajda Zanaj,2012.Financial oenness, disclosure and bank risk-taking in MENA countries. Eerging Markets Review 13 : Agoraki, M.E.K., Delis, M.D.,Pasiouras, F.Regulations, Coetition and bank risk-taking in transition countries. Journal of Financial Stability 1: Bauann,U.,Nier,E Disclosure,Volatility and transarency: An eirical investigation into the value of bank disclosure. Econoic Policy Review 9: DG Mayes, L Hale, A Liuksila. Iroving banking suervision DW Diaond, RG Rajan, A theory of bank caital. The Journal of Finance 6: Deirguc-Kunt, Asli & Huizinga, Harry,2004. Market Disciline and Deosit Insurance. Journal of Monetary Econoics 51: Davis, E.P., Kari, D., Coaring early warning systes for banking crises. Journal of Financial Stability 2: Delis M.D. and G.P.Kouretas,2011, Interest Rates and Bank Risk-Taking. Journal of Banking and Finance 35: Deirgüç-Kunt, A.,Detragiache,E., Tresse,T., Banking on the rinciles: Coliance with Basel Princiles and bank soundness. Journal of Financial Interediation 4: Dahawy,K.2009.Coany Characteristics and Disclosure Level :The case of Egyt. International Research Journal of Finance and Econoics 34: Disyatat P.,2011. The Bank Lending Channel Revisited, Journal of Financal Econoics 93: Evrensel,A.Y., Banking crisis and financial structure: a survival-tie analysis.international Review of Econoics and Finance 4: Falkenstein, E., Boral, A., Carty, L.V., Risk Calc for rivate coanies: odeling ethodology. Moody s KMV Coany. Hirtle, B., Public Disclosure, Risk, and Perforance at Bank Holding Coanies. Federal Reserve Bank of New York Staff Reorts. Hakiah,W., Lbrahi,W, Disclosure, Risk and Perforance in Islaic Banking: A Panel Data Analysis. International Research Journal of Finance and Econoics72: Hannan.,and Hanweck., Bank Insolvency Risk and the Market for large certificates of Deosit. Journal of Money, Credit and Banking 20: Joel F. Houston, Chen Lin, Ping Lin, Yue Ma, Creditor rights, inforation sharing, and bank risk taking. Journal of Financial Econoics 3: Jukka Vauhkonen, 2012.The Iact of Pillar 3 Disclosure Requireents on Bank Safety. Journal 23

24 of Financial Services Research 41: Loes,P.T., Rodrigues L.L.,2007.Accounting for financial instruents: An analysis of the deterinants of disclosure in the Portuguese stock exchange. The International Journal of Accounting 1: Laeven,L, Levine,R., Bank governance, regulation and risk taking. Journal of Financial Econoics 2: Reza Vaez-Zadeh, Danyang Xie and Edda Zoli. MODIS: a arket-oriented deosit insurance schee. IMF Willia Wilson, Lawrence Rose, John Pinfold, Moderating risk in New Zealand retail banks: Disclosure as an alternative regulatory regie. Journal of Banking Regulation13:4-23. Wheelock, D.C., Wilson, P.W., Why do banks disaear? The deterinants of US bank failures and acquisitions. The Review of Econoics and Statistics 82: Xu You-chuan,2009. Inforation Disclosure, Market Disciline and Risk-taking Behaviors of Banks. Journal of Finance and Econoics 35: (in Chinese) Yuliang Wu, Michael Bowe, Inforation Disclosure, Market Disciline and the Manageent of Bank Caital: Evidence fro the Chinese Financial Sector. Journal of Financial Services Research 38: Yuliang Wu, Michael Bowe, Inforation disclosure and deositor disciline in the Chinese banking sector. Journal of International Financial Markets, Institutions & Money22:

25 Aendix: Fitch s suort rating Bank Shenzhen Develoent Bank Liited Shanghai Pudong Develoent Bank Hua Xia Bank Co., Ltd China Minsheng Banking Cor.LTD China Merchants Bank Industrial Bank Co.LTD Bank of Beijing Co., Ltd Agricultural Bank of China Bank of Counications Industrial and Coercial Bank of China CHINA EVERBRIGHT BANK CO., LTD China Construction Bank Bank of China Liited CHINA CITIC BANK (Data source:bankscoe, 2012) 25