銀行是否能夠藉由創造流動性增加績效?

全文

(1) . 國立高雄大學經營管理研究所碩士論文. Can Bank Improve Performance by Creating Liquidity? 銀行是否能夠藉由創造流動性增加績效？ . 研究生：宋遠揚撰指導教授：陳怡凱博士中華民國九十九年七月 . .

(2) . Acknowledgements The person who I want to thank the most is my adviser, Prof. Yi-Kai Chen. Since I was totally a novice in this field, he gave me knowledge and expertise, and he pulled me up with patience when I confront problems every time, even sometimes he chose to sacrifice his rest time to solve my questions. It’s my luck to meet such an enthusiastic teacher. Next I must be grateful to my family who always support me and tolerate the emotion what I give because I often shout at them when I feel frustrated and they just call me by chance. It makes me especially embarrassed that I seldom go back home this semester, though it just takes 20 minutes from school to home. I also want to thank to my dear friends and my girlfriend who accompany by me, and help me discuss some miscellaneous problems and bring me laugh when I feel upset. Congratulations that we all graduate, now it’s your time! At last, maybe the most I should appreciate, is my computer, I can’t imagine that what I can do if it shuts down suddenly. I can’t finish this study without his help. No matter sunny or rainy, days or nights, he spends his life to help me research and sing for me (though it can’t sing itself). In sum, no word except thanks can explain my feelings, thank you all and finally I am graduated.. Yuan-Yang, Sung Institute of Business and Management, National University of Kaohsiung July, 2010. I .

(3) . 銀行是否能夠藉由創造流動性增加績效？指導教授：陳怡凱博士國立高雄大學金融管理學系學生：宋遠揚國立高雄大學經營管理研究所摘要銀行傳統業務在金融體系中扮演著做為流動性備份的重要功能，它提供了有流動性需求的使用者們一個融資的管道。但近年來像非利息收入這樣的非傳統金融業務因其高獲利的特性而日漸興盛，甚至一些銀行開始將重心放至非傳統金融業務，而逐漸放棄了原本的貸放業務。這個現象使我們產生一些想法，當銀行逐漸放棄傳統業務，是不是代表銀行做為流動性備份的功能已經無法與績效畫上等號了呢？本研究導入 Berger & Bouwman (2009)所建構的流動性創造測度，研究流動性創造與績效之間的關聯性。研究結果發現整體而言流動性創造不僅能夠增加銀行績效並穩定獲利，還可以增加風險調整後的報酬。依照金融體系區分後結果又有所不同。在市場基礎下，流動性創造能夠增加績效、降低獲利的波動性，進而提高風險調整後的報酬；而在銀行基礎下，流動性創造則會損害銀行的獲利，但仍具有穩定銀行獲利的能力，對於風險調整後的報酬則沒有顯著的影響。概括而論，儘管對不同金融體系而言，流動性創造不一定能夠提高獲利，但其穩定獲利的機能仍是非常重要的。關鍵字：流動性創造、績效、工具變數 . II . .

(4) . Can Bank Improve Performance by Creating Liquidity?. Advisor: Dr. Yi-Kai Chen Department of Finance National University of Kaohsiung Student: Yuan-Yang Sung Institute of Business and Management National University of Kaohsiung. Abstract In general, bank’s traditional lending business plays an important role as a liquidity backup of the financial system. However, the non-traditional business, such as activities in non-interest incomes, has risen since its high profitability. Even some banks pay more attention on it than the traditional business. This phenomenon inspires us to examine whether the function of liquidity backup has no longer linked with bank performance anymore. This study introduces the sense of liquidity creation by Berger & Bouwman (2009) to conduct our research. The results show that the liquidity creation can not only improve the performance and stabilize the profitability, but improve the risk-adjusted return. Evidences are differ from the financial systems. Liquidity creation may enhance bank profit abilities in market-based system but is harmful in bank-based system. Though the liquidity creation cannot bring profitability to bank in the bank-based system, the stabilizing function still works causing a positive risk-adjusted return. Hence the importance of liquidity creation as a stabilization mechanism is undoubted.. Keywords: Liquidity Creation, Performance, Instrument Variable. III .

(5) . Table of Contents Chapter 1 Introduction ................................................................................................... 1 Chapter 2 Literature Review .......................................................................................... 4 2.1 Bank Performance Determinants ..................................................................... 4 2.1.1 Internal Factors of Bank Performance .................................................. 4 2.2.2 External Factors of Bank Performance ................................................. 5 2.2 Liquidity Creation ............................................................................................ 6 2.2.1 The Origin of Liquidity Creation .......................................................... 7 2.2.2 Measurement of Liquidity Creation ...................................................... 8 2.2.3 Determinants of Liquidity Creation ...................................................... 9 Chapter 3 Data and Methodology ................................................................................ 11 3.1 Data Description ............................................................................................ 11 3.2 Variables ......................................................................................................... 11 3.2.1 Liquidity Creation Computation ......................................................... 12 3.2.2 Control Variables ................................................................................ 14 3.2.3 Determinants of Liquidity Creation .................................................... 15 3.3 Methodology .................................................................................................. 16 3.3.1 Panel unit root test .............................................................................. 16 3.3.2 Two stage least squares (2SLS) .......................................................... 18 Chapter 4 Empirical Results ........................................................................................ 21 Chapter 5 Conclusions & Implications ........................................................................ 25 Appendix: Survey Questions of Bank Regulation and Supervision ............................ 41. IV .

(6) . List of Tables. Table 1 Liquidity classification (sorted in Bankscope categories) ......... 30 Table 2 Variable definition ........................................................................ 32 Table 3 Financial system classifications ................................................... 33 Table 4 Correlation coefficients ................................................................ 34 Table 5 Descriptive statistics ..................................................................... 35 Table 6 Unit root test results ..................................................................... 35 Table 7 Endogenous regression result ...................................................... 36 Table 8 IV Regression results for full sample .......................................... 36 Table 9 IV regression results split by financial structure ....................... 38. V .

(7) . Chapter 1 Introduction Financial intermediaries allocate the resources across space and time, in an uncertain environment (Bodie & Merton, 1995; Levine, 1997), this sentence describes the function of banking system perfectly. And one of the important functions is to backup the sources of liquidity(Corrigan, 1982, 2000). But recent years the traditional business is recessing due to the high profitability of non-traditional business. This carries out an inspiration that if the traditional business is no more indispensible for banks. We discuss this topic by introducing liquidity creation by Berger and Bouwman (2005, 2009), they constructed this measurement to calculate the liquidity that bank provides to the financial system and found a positive relationship between liquidity creation and bank market value. However, there’s no further research about the effect of liquidity creation on bank performance recently, so this study keeps an eye on the relationship between bank performance and liquidity creation. They also mentioned that the liquidity creation and its corresponding risk are not perfectly matched, hence we would further discuss about the risk-adjusted performance. Previous researches discussed the bank performance can be mainly divided into two parts, for one is internal factors and external factors is the other (Athanasoglou, Delis, & Staikouras, 2006; Bourke, 1989). Internal determinants are factors that are mainly influenced by a bank’s management decisions and policy objectives, such profitability determinants are the level of liquidity, provisioning policy, capital adequacy, expenses management and bank size. On the other hand, the external determinants, both industry-related and macroeconomic, are variables that reflect the economic and legal environment where the credit institution operates. In a word, with regard to the bank performance, previous studies overlooked the importance of the. 1 .

(8) . behavior of creating liquidity for a long time. More specifically, banks inject liquidity into the financial market by transforming illiquid assets to liquid liabilities; on the contrary, banks destroy liquidity by transforming illiquid assets to liquid one. From this perspective we can easily point out a view that liquidity creation may enhance bank performance because the return for bearing the liquidity risk heightens conduct a risk premium and the business activities is also increasing. But in the risk-adjusted return, the relationship is unsure. Bank could obtain a premium since it suffers the liquidity risk, but it could also bare a discount because it suffers “too much” risk. Consider to the liquidity creation determinants, Berger and Bouwman (2009) also found that the bank size is a critical factor to the effect of bank capital on liquidity creation. That is, a higher level of bank capital may enhance the ability to create liquidity for large banks, but higher bank capital may impede the financial fragility-crowding out effect, and then reduce the liquidity creation for small banks. From paragraphs above, clearly the liquidity creation cannot be determined exogenously, hence we know that if we want to examine the effect of bank liquidity creation on performance, we should eliminate the influence of other factors or we will get a biased relationship. Therefore we use instrumental variable (IV) to solve this problem, and we expect that our proposition that the liquidity creation is positively linked with bank performance would be verify. Furthermore, we divide our sample as two subsamples by financial structure since many scholars thought that the bank activities is different relative to the financial market in bank-based countries versus market-based countries. To achieve our purpose, this study is organized as follow. At first we use several simple paragraphs to state our knowledge background, motivation, and researching 2 .

(9) . purpose. Then we discuss the literatures about the determinants of our major topics liquidity creation and performance. In Chapter 3, the data will be shown and we organize our model and conduct some pretests, after that we use IV regression to separate the irrelevant effects in Chapter4, and the empirical results would be present in this chapter. Then the following is our conclusion.. 3 .

(10) . Chapter 2 Literature Review. 2.1 Bank Performance Determinants Bank performance, or bank profitability, typically measured by the return on assets (ROA) and the return on equity (ROE), and even net interest margin (NIM). Previous researches discussed about the determinants of bank performance are miscellaneous, but they can be roughly divided into two parts, for one is internal factors and external factors is the other (Athanasoglou, et al., 2006; Bourke, 1989). Internal determinants are factors that are mainly influenced by a bank’s management decisions and policy objectives, and the opposite, the external determinants, are variables that reflect the economic and legal environment where the credit institution operates. Here we are going to summarize them as follow.. 2.1.1 Internal Factors of Bank Performance Bank size is generally used to capture potential economies or diseconomies of scale in the banking sector. In previous studies, some studies have found scale economies (Akhavein, et al. 1997; Bourke, 1989; Molyneux & Thornton, 1992; Bikker & Hu, 2002; Goddard, et al., 2004), and some are not(Berger, et al., 1987; Boyd & Runkle, 1993; Miller & Noulas, 1997; Athanasoglou, et al., 2006). Eichengreen & Gibson (2001), suggest that the effect of a growing bank’s size on profitability may be positive up to a certain limit. Beyond this point the effect of size could be negative due to bureaucratic and other reasons. Hence, the size-profitability relationship may be expected to be non-linear.. 4 .

(11) . Capitalization has been demonstrated to be important in explaining the performance of financial institutions, its impact on bank profitability is ambiguous. As lower capital ratios suggest a relatively risky position, one would expect a negative coefficient on this variable (Berger, 1995). However, a higher levels of equity would decrease the cost of capital, causing a positive impact on profitability (Molyneux & Forbes, 1995). Moreover, capital may also reduce the expected costs of financial distress, including bankruptcy (Berger, 1995). Indeed, most studies that use capital ratios as an explanatory variable of bank profitability (Bourke, 1989; Molyneux & Thornton, 1992; Goddard, et al., 2004) observe a positive relationship.. 2.2.2 External Factors of Bank Performance External to the bank, several factors have been suggested as impacting on profitability. For the first is the regulation, banks’ activities is regulated by government and differed from countries, it determined the capital and loan loss provisioning decision directly (Demirgüç-Kunt & Huizinga, 1999). Barth, et al. (2004) noted that the regulation restricts bank activities in many dimensions and separate the regulatory into several domains. They discovered that there’s no consistency among the relationship between bank performance and regulatory variables, but addressed the influence actually exists. Bank profitability is sensitive to macroeconomic conditions. Generally, higher economic growth encourages banks to lend more and permits them to charge higher margins, as well as improving the quality of their assets. Neely & Wheelock (1997) use per capita income and suggest that this variable exerts a strong positive effect on bank earnings. Demirgüç-Kunt & Huizinga (2000) and Bikker & Hu (2002) attempted to identify possible cyclical movements in bank profitability - the extent to which 5 .

(12) . bank profits are correlated with the business cycle. Their findings suggest that such correlation exists, although the variables used were not direct measures of the business cycle. Most studies (Bourke, 1989; Molyneux & Thornton, 1992) observe a positive relationship between inflation and bank performance. The financial structure represents the role which bank plays in a country, and it is easy to understand why the structures are so different. In a bank-based country such as Germany and Japan are considered to play a leading role in mobilizing savings, allocating capital, overseeing the investment decisions of corporate managers, and in providing risk management vehicles. Relatively, securities markets share center stage with banks in terms of getting society's savings to firms, exerting corporate control, and easing risk management, for example, England and USA. The position which bank occupies in a financial system is connected to its performance deeply. Demirgüç-Kunt & Levine (1999) constructed three indices by comparing relative bank size, activity, and efficency to market, and combined these indices by taking average of them into an integrated structure index. They compare the index to the sample mean and countries beyond this are more market-based.. 2.2 Liquidity Creation The liquidity creation is the function that bank creates liquidity to financial market through accepting liquid assets and making illiquid loans. This procedure of liquidity creation is not just restricted to bank activities, former research considered it as a behavior that transforms the assets into liabilities, so the liquidity creation may be accomplished at the firm level without the necessity for bank intermediation (Gorton & Pennacchi, 1990). Firms can just simply split the cash flows of their asset portfolios. 6 .

(13) . by issuing both equity and debt in a stock or equity market. In this study we mainly focus on the bank activities and performance, therefore we don’t take this view into account. Since the definition of liquidity creation is not so clear, we thus discuss its original senses by introducing the origin of liquidity in the first section.. 2.2.1 The Origin of Liquidity Creation It is necessary to distinguish the liquidity versus liquidity creation. Though they are highly correlated, they are not all the same. Initially, the analyses of Patinkin (1965), Tobin (1965), and Niehans (1978) provided insights into characterizing the liquidity of assets, and inspired Diamond & Dybvig (1983) to discover the importance of liquidity, therefore the definition of liquidity had been broadly discussed for a long time. Thadden (1999) concluded three prominent views as follows: we view an asset as a liquid asset if it can be bought or sold quickly at low transaction costs and a reasonable price (Biais et al, 1997); the liquidity also refers to the availability of instruments that can be used to transfer wealth across periods (Bryant, 1980; Holmström & Tirole, 1998), that means we can get money by liquidating assets over periods; Diamond & Dybvig (1983) also defined an asset is liquidity if it allows agents to consume intertemporally as they would like to, it describes the function best that banks prepare liquidity for being withdrawn anytime. Even though the opinions are widely divided, but in generally we refer to the ability that how fast and how good an asset converts into cash when mentioning the word “liquidity”. Compare to the definitions of liquidity, the liquidity creation is essentially the process that bank provides liquidity for any natural or legal person for drawing anytime, and make the illiquid loans so that agents can ensure that bank won’t call back its money anytime. In short, the liquidity creation is that bank accepting 7 .

(14) . short-term, liquid deposits and make long-term, illiquid loans (Deep & Schaefer, 2004). Consider the category rather than the maturity, Berger & Bouwman (2005, 2009) extended the definition and thought liquidity creation as the procedure that bank transforms its illiquid assets into liquid liabilities. In fact it’s more like extension from the view of Diamond & Dybvig (1983); bank creates liquidity for customers to be used. Through this process, or this function, bank injects liquidity into the financial market and stimulates the economic activities.. 2.2.2 Measurement of Liquidity Creation Though the function of creating liquidity had been brought up since 1965, there was no measurement for quantifying it until the liquidity creation by Berger & Bouwman (2009). Allen (1981) indicated that the pure liquidity creation takes place when the banking system guarantees the client that he will be able to borrow money whenever he likes. However, they did not offer a comprehensive method to measure the liquidity creation; it just represents the loan-deposit matching relationship. Even if Diamond & Dybvig (1983) provided an insight of the process of liquidity creation, they only take this idea into account but didn’t develop the measurements. The liquidity creation measurement was originated from liquidity transformation by Deep & Schaefer (2004). They constructed the liquidity transformation gap (LT gap) to calculate the magnitude of liquidity transformation, and defined it as the difference between liquid liabilities and assets held by a bank, scaled by its total assets, and the liquid reffered to the maturity that below one year.. LT gap . Liquid Liabilities  Liquid Assets Total Assets. (1). The intuition is that a bank financed in large part by liquid deposits and that 8 .

(15) . holds mostly illiquid loans (and thus a small proportion of liquid assets) performs a significant amount of liquidity transformation and would have a high LT gap value. Berger & Bouwman further constructed liquidity creation and wider the criteria because they thought the domain of liquidity transformation which Deep & Schaefer discussed was too narrow. On one hand they argued that the catagory should be the most important determinant of liquidating velocity rather than the maturity since assets such as residential mortgage may have a long maturity but easily be sold if one can securitize it. On the other hand Deep and Schaefer exclude the off-balance sheet activities because they don’t involve in the transformation process and occupied a relatively small part in bank assets, but Berger & Bouwman found that the off-balance sheet activities exactly have influence on the liquidity creation indeed and included this part. Hence they extended the liquidity creation measurement to two dimensions: category/maturity (cat/mat) and on/off-balance sheet (nonfat/fat), and chose the catagory-off-balance sheet (cat-fat) combination rather than the maturity/on-balance sheet (mat-nonfat) one. Here we follow the definition and measurement of liquidity creation by Berger & Bouwman (2005) to examine our study. They evaluate the assets on the basis of ease, cost, and time, and categorize them into several groups. They calssfied all bank activities as six categories: liquid assets, semi-liquid assets, illiquid assets, liquid liabilities, semi-liquid liabilities, and illiquid liabilities. Then the liquid creation is computed by using these data. We will discuss it in later chapter.. 2.2.3 Determinants of Liquidity Creation. Though there are not so many studies about the liquidity creation, some scholars still devoted to discover the determinants of liquidity creation. Berger & Bouwman 9 .

(16) . (2005, 2009) found some interesting conclusions. On one hand, for large banks a higher level of bank capital may enhance the ability to create liquidity according to the risk absorption theory. This theory is based on two stands that liquidity creation exposes bank to the liquidity risk (Allen & Gale, 2004; Allen & Santomero, 1998; Diamond & Dybvig 1983) but the bank capital expands bank’s risk-bearing capacity (Bhattacharya & Thakor, 1993; Repullo, 2004; von Thadden, 1999). On the other hand, higher bank capital may impede the financial fragility-crowding out effect, and then reduce the liquidity creation for small banks. A fragile capital structure encourages the bank to commit to monitoring its borrowers, and hence allows it to extend loans. Additional equity capital makes it harder for the less-fragile bank to commit to monitoring, which in turn hampers the bank’s ability to create liquidity. Capital may also reduce liquidity creation because it “crowds out” deposits (Gorton and Winton, 2000). We can infer that the bank size and bank capital may be important factors in determining liquidity creation. Besides, the liquidity intuitionally represents the potentials that bank operates a liquidity creation just like the chips for the gambler; that is, if bank holds more liquidity, the possibility or the ability that create/destruct more liquidity. Berger & Bouwman (2009) also indicate that the macroeconomic changes would affect the liquidity creation so they add it as control variables. To conclude, liquidity creation could be mainly interpreted by these items, hence we use these terms to construct our endogenous model in later chapter.. 10 .

(17) . Chapter 3 Data and Methodology. 3.1 Data Description. We choose the G10 countries as our criteria since these countries hold most of the banks in the world, and their financial system are also well-developed for decades, thus the data would more obtainable. The G10 geographic region include Belgium, Canada, France, Germany, Italy, Japan, Netherland, Switzerland, Sweden, United Kingdom, United States of America, 11 countries and 6968 observations in total. In this paper we compute liquidity creation by the measure in Berger & Bouwman (2005) among these countries. The observations span over 15 years, from 1995 to 2008 unbalanced data, and the samples are drawn from the Bankscope Database and composed of three types: commercial banks, cooperative banks, and saving banks. These banks survive on deposits and lendings, so it is meaningful to discuss about their liquidity creation.. 3.2 Variables. According to the literatures, we use the generally adopted measurement, return on assets (ROA), here we divide the performance measurement into three part to capture the whole picture of bank performance, first we use three year average ROA (AVROA) form year t-2 to t to mitigate the effect of extreme value. Second we measure the volatility of ROA by counting the corresponding standard deviation (sigavroa). The last one we count the Sharpe Ratio for these years (sharpROA), which equals the AVROA divide by sigavroa., and we describe our explanatory variables in 11 .

(18) . following sections.. 3.2.1 Liquidity Creation Computation. Berger & Bouwman (2005, 2008, 2009) constructed a two dimension measurement of category/maturity (cat/mat) and on/off-balance sheet activities (fat/nonfat) to compute liquidity creation. Our major variable, liquidity creation, is computed by the cat-fat combination that differs from the mat-nonfat measurement such as LT gap (Deep & Schaefer, 2004). Step 1 is to classify all activities in balance sheet and off-balance sheet as liquid, semi-liquid or illiquid and step 2 is given weights to the categories then computes it by the model they addressed. In Step 1, we classify all the activities in the balance sheet. In the asset side, cash, due from banks, and securities can transform in financial market, thus we sort these items to liquid assets. On the contrary some items cannot be sold quickly (e.g., loan to company, hire purchase, lease, and problem loans), we classify these item as illiquid assets. Additionally, for mortgage, loans to bank and government, the liquidity is between cash and loans so we classify these items as semi-liquid assets. Similarly, we classify liabilities and off-balance activities as liquid, semi-liquid and illiquid. The original classification is shown in, besides we set up a new one due to the category restrictions of Bankscope Database as Table 1. But here is still something different from past literature, for example, the mortgages are classified in illiquidity asset by Berger and Bowuman (2005), but we sort in semi-liquid assets. We think the character of mortgage is between the liquidity and illiquidity. And we categorized all of the demand deposits and saving deposits by liquidity liabilities. In step2, we adopt the same weight used by Berger & Bouwman (2009) but. 12 .

(19) . follow the categories we prepared in Table 1. According to liquidity creation theory, they suggest that when banks transform $1 of illiquid assets into $1 of liquid liabilities, $1 of liquidity is created, and when $1 of liquid assets transforms into $1 of illiquid liabilities, $1 of liquidity is destructed. To achieving the constraints, they assigned a weight of 1/2 for illiquid assets and liquid liabilities, it represents that both the source and the use of funds affect the liquidity creation. Where the transformation can present in formula “1/2 * $1 liquid liabilities + 1/2 * $1 illiquid assets = $1 liquid creation” as maximum created. Similarly, they applied a weight of -1/2 to liquid assets, illiquid liabilities and equity. Then the equation “-1/2 * $1 illiquid liabilities -1/2 * $1 liquid assets = $-1 liquidity creation” describes as maximum destroyed. The semi-liquid items were denoted as an intermediate weight to 0, and then gave a weight of 1/2 to illiquid financial guarantees. Then sum up the liquidity creation and liquidity destruction, the complete formula was set: LC = 1/2 * illiquid assets + 0 * semi-liquid assets + 1/2 * liquid liabilities. (2). -1/2 * liquid assets + 0 * semi-liquid liabilities -1/2 * illiquid liabilities and equity +1/2 *illiquid financial guarantees +1/2*semi-liquid financial guarantees. It’s obvious that the liquidity creation ties on the lending business closely. This creation may lead a premium since banks have to suffer the illiquidity of these actions, but the risk-adjusted return is unsure. More specifically, bank could obtain a premium since it suffers the liquidity risk, but it could also bare a discount because it suffers “too much” risk.. 13 .

(20) . 3.2.2 Control Variables. Here are all variable definitions and we list them all in Table 2. The size factor is commonly adopted to be the natural logarithm of total asset (lnTA). Similarly, the total equity to assets ratio (TETA), is widely used in the empirical research as the key capital ratio. The liquidity is often measured by the ratio of liquid assets to total assets (L_TA), but however, the data is not available. To substitute, we use the ratio of total loan to total asset as our liquidity measurement (Athanasoglou, et al., 2006). In order to capture the effect of the macroeconomic environment, we follow the literatures and use annual percent change of GDP (GDPC), we further add GDP annual percent change of last year (GDPCt-1) to capture the lagged effects. For the regulation variable, we follow Barth et al. (2004) and measure the regulatory in three dimensions. The official supervisory power index (OSP) represents the power of supervisory institution; the private monitoring index (PMI) includes information on the degree to which bank regulations force banks to disclose accurate information to the public and induces private sector monitoring of banks; and the overall bank activities and ownership restrictiveness (BAR) reflects the restriction that bank encountered in the financial system, includes the regulations on security, insurance, real estate activities, and the restrictions on bank owning/controlling a financing firm. We use these three indices to proxy a relative severity among countries. But since the country indices are time-invariant variables and cannot be applied in fixed-effect model, hence we compare the fixed-effect and random-effect model and found that the fixed-effect model would be better. In replacement, we also use the interaction terms with GDPC named GDPCXOSP, GDPCXPMI, GDPCXBAR respectively as our regulation variable. Owing to the interaction term are highly correlated to each other because they are originated from GDPC and would 14 .

(21) . lead to the colinearity, hereby we only use one term in each model. So there will be four models which use gdpcxosp, gdpcxpmi, gdpxxbar, and no regulation variable in each panel. Finally, numerous literatures argued that the position of bank would be different among bank-based and market-based countries. Bank occupies a crucial role as mobilizing savings, allocation capital, overseeing the investment decisions of corporate managers and providing risk management vehicles; but in market-based countries the securities markets share this center stage with banks (Demirgüç-Kunt & Levine, 1999). Hence we follow them and use their structure index to proxy the financial structure. They measure the relative importance of bank or market finance by the relative size, activity, and efficiency. We classify countries with values of the Structure index above (below) the sample mean as market-based (bank-based) financial systems as Table 3. Further, financial structure (d_fs) is a dummy variable that takes the value 0 for market-based systems and 1 for bank-based systems.. 3.2.3 Determinants of Liquidity Creation. According to our literature, the size and capital are the first two important factors that affect the bank liquidity creation (Berger & Bouwman, 2009). We here use the natural logarithm of total asset (lnTA) and the total equity to assets ratio (TETA), and these two factor is just discuss in prior section. In intuition, the holding of liquidity is directly correlated to the bank liquidity creation. No matter how it influences, a higher level of bank liquidity implies that bank can operate more instruments to construct or destruct the liquidity offers in the financial system. Here we also adopt the ratio of loan to total assets (L_TA). 15 .

(22) . mentioned in prior section. At last, we use the change in economic growth (GDPC) and lagged change in economic growth (GDPCt-1) to capture the macroeconomic changes and the lagged effect respectively. Then all the correlation coefficients and summary statistics are shown in Table 4 and Table 5.. 3.3 Methodology. At first we have to confirm whether the data is stationary by unit root test, if the data is non-stationary, we can convert it to a stationary series through process of differencing. Then we apply the two stage least square (2SLS) to solve the endogeneity of liquidity creation.. 3.3.1 Panel unit root test. We prepare the unit root test for the panel data, the test is to examine if the data follow a stationary stochastic. That is, the generated probability distribution is independent of time, so the distribution would not change by time. To describe briefly, a stationary time series will just be affected temporally by some environmental influence, and it may return to its original path (or average) by time; but a nonstationary time series may accumulate the influence thus deviates from the path. All tests are based on the following Augmented Dickey-Fuller regression:. Yi ,t   i   i ,t  i Yi ,t 1   i ,t , i  1, 2,..., N , t  1, 2,..., T. (3). Where  i and  i ,t allow for fixed and unit specific time trends for each i.  i ,t 16 .

(23) . is error term and  i ,t ~ IID(0,  2 ) . The null hypothesis of unit root is:. H0 : i  0, i. (4). Thus under the null hypothesis, all series in the panel are nonstationary processes, and the alternative hypothesis is different for each test. Levin et al. (2002) test may be viewed as a pooled Augmented Dickey-Fuller test. Besides, the test considered an alternative hypothesis where the autoregressive coefficient is homogeneous across groups:. H1: i    0, i. (5). Im et al. (2003) extends the Levin et al. (2002) framework to allow heterogeneity in the value of the autoregressive coefficient. Under the alternative, parts of the series in the panel are assumed to be stationary. This is in contrast to the Levin et al. test, which presumes that all series are stationary under the alternative hypothesis. In this test, it has individual i  1,2,..., N1 where Yi ,t is stationary and individual i  N1  1, N1  2,..., N where Yi ,t is nonstationary. The alternative hypothesis is specified as:. H 1: i  0, i  1, 2,..., N1 and i  0, i  N1  1, N1  2,..., N. (6). However, the test assumes that T is the same for all cross-section units. So the test is applied only for balanced panel data. So the test is applied only for balanced panel data.. Here we introduce the Fisher P test proposed by Maddala and Wu (1999), this test statistics is derived from Augmented Dickey-Fuller regression and the null hypothesis is that the data is non-stationary. It combines the p-values in all cross-sectional ADF test results into one single term, thus one can easily comprehend 17 .

(24) . whether the data is stationary or not. For the most important thing, the method doesn’t restrict the data to be balanced, so this method fits to conduct our research. The statistic of Fisher P test is given by: N. P  2 ln( pi ) ~  2 (2 N ) i 1. (7). Where pi is the p-value of the test statistic of the ADF test in each group.. 3.3.2 Two stage least squares (2SLS). 2SLS is used to solve the endogeneity problem. When there is an endogenous variable in the model, the error term would correlate to the endogenous variable and causes a bias in regression results. The concept of this method is to view the endogenous variable as a function of instrumental tools and substitute the function with the endogenous variable, thus the endogenous problem would be solved. Since the results form IV regression and OLS estimators are really different, and here we test if the endogeneity does exactly exist by Davidson-MacKinnon test first. Many econometrics texts discuss the issue of “OLS vs. IV” such as the well-known Hausman test, which involves estimating the model via both OLS and IV approaches and comparing the resulting coefficient vectors. But in finite samples, the precision matrix (defined as the difference between the two estimated variance-covariance matrices of the parameter estimates) may not be positive definite. A less problematic approach is the auxiliary regression framework of Davidson and MacKinnon1 (1993), which may always be computed. Consider the following model: 1. For more detailed econometrics see “Estimation and Inference in Econometrics” Davidson and MacKinnon (1993, p.235) and “Econometric Analysis of Cross Section and Panel Data” Wooldridge (2002, chapter.5). 18 .

(25) . y1  0  1 y2   2 z1  3 z2  u1. (8). The auxiliary regression approach involves estimating the reduced form (first–stage) regression for . y2   0   1 z1   2 z2   3 z3   4 z4  u2. We assume that z1 , z2 , z3 , z4. (9). is exogenous and now we are concerning with. that y2  u1 . Exogeneity of the z s implies that the residuals from OLS estimation of equation u2 will be a consistent estimator of u2 , and we would rewrite equation (8) in following equation: y1   0  1 y2   2 z1   3 z2 + u2  v1. (10). Then the significance of  represents the existence of exogeneity. While there are multiple endogenous variables, the generalized test for restricted and unrestricted models would become an F-test,. F. SSRu  SSRur n  k  ~ Fm, n  k SSRur m. (11). where m is the number of endogenous variables and k is the number of total variables. As they discussed, the test may be applied to a subset of the endogenous variables, treating those not specified as endogenous. The null hypothesis states that an ordinary least squares (OLS) estimator of the same equation would yield consistent estimates: that is, any endogeneity among the regressors would not influence on OLS estimates. The rejection of the null hypothesis represents that the instrumental variables are meaningful. In our model the liquidity creation is not an exogenous variable since the time we regress factors (capital, size, credit risk) on bank performance, it may affect the 19 .

(26) . liquidity creation at the meanwhile. Our model is constructed as follow. Performance  f ( LC , size, capital , gdpgrowth, regulation) where LC  g ( size, capital , liquidity, gdpgrowth). (12). Next we regress for the endogenous function and check if the instrumental variables are able to interpret the endogenous variable or not, then we use IV regression to verify our purpose, the effect of liquidity creation on risk-adjusted performance. Additionally, we also use the Hausman test to find out which is better among fixed-effect and random-effect, but here we only show the better one for avoiding complicated results.. 20 .

(27) . Chapter 4 Empirical Results At first we test if the variables are stationary by Fisher Pλ test, and Table 6 shows that all the test statistics reject null hypothesis of non-stationary. Next we test for the instrumental variables for liquidity creation as Table 7, and results show that all the coefficient of size, teta, and GDP are significant. The R-square reaches 0.4306 represents that these factors do exactly explain almost half of the total variance, it implies this model fits to explain our endogenous variable, LnLC. Then we test for our purpose which is mention in chapter 3 (see Table 8). We regress the model for avroa, sigavroa, sharproa both in panel regression and panel IV regression models respectively. Note that the Davidson-MacKinnon F test statistics state the results of Davidson-MacKinnon exogeneity test, and the significance means this model is fitting to use IV regression. To avoid being too complex, we only present panel regression in place of panel IV regression. Likewise, if the Hausman Chi2 is significant, we would only present the fixed-effect result in tables. In Panel A of Table 8, we state both the regression with and without regulation variables; clearly we can observe that coefficient of LnLC is positively significant in most results. This finding infers that the liquidity creation benefits bank performance. Viewing the control variables, the SIZE and TETA term reflects a consistent result in each model; both are positively related to the bank performance, which supports the point of literatures of scale economies and risky operation of capital. Besides, the GDPC coefficients also suggest that the change in GDP couldn’t be reflected on performance right away but with lags on the next period. Among the regulatory variables, evidence shows that except the private monitoring power, both official supervisory power and restrictions on bank activities would cause a negative effect on. 21 .

(28) . performance, this implies the government actions lead to more barriers than private sectors. Next going to the Panel B of Table 8, this regresses on ROA volatility and finds an interesting implication. One could notice that LnLC is negatively linked with the ROA volatility; it refers that the liquidity creation can stabilize the ROA volatility and can lower the risk of bank. Size factor is negative significant to ROA volatility, this indicates that larger banks have more capabilities to control its risk and also suggest that scale economies do exactly exist. But the capital coefficients refuse to support the risk absorption theory, a possible interpret is that bank holds more capital and leads to a relative risky operation since the abundance of money, thus obtain a premium of risk. Regulation coefficient terms show that the monitoring degree of private sector and restrictions on bank activities would decrease bank risk. Panel C of Table 8 presents the results for risk-adjusted ROA. The coefficients of LnLC are all positive and significant at significant level = 0.1, it means that the liquidity creation performs a positive risk-adjusted return. But we found that size is negatively related to risk-adjusted ROA. Most capital coefficients are negatively significant at significant level = 0.1, implies though capital can enhance bank profitability, but consider to the risk-adjusted return, the risky operation would lead to a worsen return. Since financial system is a crucial factor that determines the importance of bank in a country, therefore we next divide our subsamples in bank-based and market-based. Table 9 exhibits the subsample analysis result split by financial structure, in Panel A first we find the endogeneity assumptions is not supported in bank-based countries. LnLC is negatively significant at significant level of 0.01 in bank-based countries while a positive one at significant level of 0.1 in market-based countries. This 22 .

(29) . represents that liquidity creation would performs better in market-based countries than bank-based countries; this may be owing to the bank position in bank-based system cause them to compete with each other and lead to a relative small profit margin. Size is not significant in market-based systems but presents a better result in bank-based system, this indicates that scale economies only exists in bank-based countries so that it would be more powerful to compete with others. The capital effect remains positive in all situations, which states that banks can obtain profits by using capital. Additionally, the regulation variables provide mixed results, and most coefficients are not significant and sensitive to profitability. In Panel B of Table 9 the function of profitability stabilization of liquidity creation still holds both for bank-based and market-based countries, but size factor doesn’t seem to be so. Evidence show that scale economies points is only supported in bank-based countries and capital bring either bank-based or market-based countries risks. With regard to official supervisory power, the effect seems like very different. Coefficients are positively related to sigavroa in bank-based countries, but negative in market-based ones. This means the supervisory of authority would cause lower the ROA volatility of banks in market-based countries, but risen the volatility in bank-based countries in contrast. The monitoring degree of private sector also works better in market-based countries than in bank-based. Then the restriction coefficients are only significant in bank-based countries. We think it might because bank activities are well-behaved in bank-based countries, that is, banks in market-based countries tend to operate more risky assets so that the risk control effect of all variables are obvious than in bank-based countries. Next go to Panel C of Table 9, the result is noticeable, LnLC is positively related to risk-adjusted performance in market-based system but not significant in bank-based system, so liquidity creation fits to operate in a market-based system since it can 23 .

(30) . lower the risk effectively and bring about a positive risk-adjusted ROA. Compare to the bank-based system, though bank occupies a relative important situation, it still cannot improve its risk-adjusted return. It may be the case that banks in bank-based countries operate lesser risky instruments so that the liquidity creation wouldn’t bring any excess benefits. These finding suggests that it could be a non-linear relationship between liquidity creation and bank performance since liquidity creation can moderate risk to a certain extent, but has no effects or worse for more. Consider to the risk-adjusted return, scale diseconomies exist instead. Capital coefficients are mostly insignificant may be because that the risk just offsets the profit it brings. Results also indicate that all the regulation variables affect risk-adjusted return significantly in bank-based system.. 24 .

(31) . Chapter 5 Conclusions & Implications Creating liquidity is one of the crucial functions of bank, and this function is linked with bank traditional lending business. But the increasing non-traditional business leads to the decrease of traditional business recently. We wonder if the traditional business is negligible and introduce the measurement of liquidity creation to examine the relationship of liquidity creation on both risk-unadjusted and risk-adjusted performance. This study researches on relationship among liquidity creation and bank performance for 11 advanced countries across 14 years from 1995 to 2008. Since the liquidity creation cannot be determined exogenously, we apply panel two stage least square (2SLS) method. We found that the liquidity creation is positively related to ROA and risk-adjusted ROA and negative to ROA volatility. The results shows that banks can improve both performance and risk-adjusted performance by creating liquidity, and liquidity creation can also reduce bank risk. Furthermore, the results also present that scale economies exist so that larger banks link with a higher performance and a lower risk level. But additional capital would rather increase performance and risk, this result does not support the risk absorption theory. In the subsample analysis, since bank-based and market-based systems are very different, then we would divide our sample by financial systems. Evidence shows that the results from financial systems are unsurprisingly different; liquidity creation could improve bank performance in market-based system but harm in bank-based system, and consider to the risk-adjusted ROA, liquidity creation only works in market-based system. But the stabilization ability on ROA both exists in bank-based and market-based systems, this point out the importance of stabilization of liquidity 25 .

(32) . creation. Though liquidity creation cannot carry profits for a part of banks, but the effectiveness of lowering risk is doubtless. This study measures the bank traditional lending business and finds that the traditional business can lower the risk effectively and heighten the risk-adjusted return. This implies that even the non-traditional business is profitable, but never overlooks the risk hidden behind and the traditional business can help reduce bank risk thus obtain a better risk-adjusted return. For a firm which pursues high profit and low risk at the meanwhile, it should not abandon the traditional business since its risk-downing nature.. 26 .

(33) . Reference Akhavein, J. D., Berger, A. N., & Humphrey, D. B. (1997). The Effects of Megamergers on Efficiency and Prices: Evidence from a Bank Profit Function. Review of Industrial Organization, 12(1), 95-139. Allen, F., & Gale, D. (2004). Financial Intermediaries and Markets. Econometrica, 72(4), 1023-1061. Allen, F., & Santomero, A. M. (1997). The Theory of Financial Intermediation. Journal of Banking & Finance, 21(11-12), 1461-1485. Allen, W. A. (1981). Intermediation and Pure Liquidity Creation in Banking Systems. BIS working paper(5). Athanasoglou, P. P., Delis, M. D., & Staikouras, C. K. (2006). Determinants of Bank Profitability in the South Eastern Euorpean Region. Working Paper(47). Barth, J. R., Caprio, G., & Levine, R. (2004). Bank Regulation and Supervision: What Works Best? Journal of Financial Intermediation, 13(2), 205-248. Beck, T., Demirgüç-Kunt, A., & Levine, R. (2009). Financial Institutions and Markets across Countries and over Time - Data and Analysis. Policy Research Working Paper, No. 4943. Berger, A. N. (1995). The Relationship between Capital and Earnings in Banking. Journal of Money, Credit and Banking, 27(2), 432-456. Berger, A. N., & Bouwman, C. H. S. (2005). Bank Capital and Liquidity Creation. Working Paper. Berger, A. N., & Bouwman, C. H. S. (2008). Financial Crises and Bank Liquidity Creation. Working Paper. Berger, A. N., & Bouwman, C. H. S. (2009). Bank Liquidity Creation. Review of Financial Study, Vol. 22(No.9), hhp014. Berger, A. N., Hanweck, G. A., & Humphrey, D. B. (1987). Competitive Viability in Banking: Scale, Scope and Product Mix Economies. Journal of Monetary Economics, 20, 501-520. Bhattacharya, S., & Thakor, A. V. (1993). Contemporary Banking Theory. [doi: DOI: 10.1006/jfin.1993.1001]. Journal of Financial Intermediation, 3(1), 2-50. Biais, Bruno, Foucault, Thierry, & Hillion, P. (1997). Microstructure des Marchés Financiers: Institutions, Modèles et Tests empiriques. Presses Universitaires de France. Bikker, J. A., & Hu, H. (2002). Cyclical patterns in profits, provisioning and lending of banks and procyclicality of the new Basel capital requirements. BNL Quarterly Review, 221, 143-175. Bodie, Z., & Merton, R. C. (1995). A Functional Perspective of Financial Intermediation. Financial Management, 24(2). 27 .

(34) . Bourke, P. (1989). Concentration and other Determinants of Bank Profitability in Europe, North America and Australia. Journal of Banking and Finance, 13(65-79). Bryant, J. (1980). A Model Of Reserves, Bank Runs, and Deposit Insurance. Journal of Banking and Finance, 4, 335-344. Cooper, M. J., Jackson, W. E., & Patterson, G. A. (2003). Evidence of predictability in the cross-section of bank stock returns. Journal of Banking and Finance, 27, 817-850. Corrigan, E. G. (1982). Are Bank Special? The Region. Federal Reserve Banks of Minneapolis. Corrigan, E. G. (2000). Are Bank Special? A Revisitation. The Region. Federal Reserve Banks of Minneapolis. Davidson R. & MacKinnon J. G (1993). Estimation and Inference in Econometrics. Oxford University Press Deep, A., & Schaefer, G. (2004). Are Banks Liquidity Transformers? Faculty Research Working Papers Series. Demirgüç-Kunt, A., & Huizinga, H. (1999). Determinants of Commercial Bank Interest Margins and Profitability: Some International Evidence. The World Bank Economic Review, 13(2), 379-408. Demirgüç-Kunt, A., & Huizinga, H. (2000). Financial Structure and Bank Profitability. Policy Research Working Paper, No. 2430. Demirguc-Kunt, A., & Levine, R. (1999). Bank-Based and Market-Based Financial Systems: Cross-Country Comparisons. World Bank Policy Working Paper No. 2143. Diamond, D. W., & Dybvig, P. H. (1983). Bank Runs, Deposit Insurance, and Liquidity. The Journal of Political Economy, Vol. 91(No. 3), 401-419. Duca, J. V., & McLaughlin, M. M. (1990). Developments Affecting the Profitability of Commercial Banks. Federal Reserve Bulletin, 76(7), 477-499. Eichengreen, B., & Gibson, H. D. (2001). Greek Banking at the Dawn of the New Millennium. CERP Discussion Paper, 2791. Goddard, J., Molyneux, P., & Wilson, J. O. S. (2004). Dynamics of Growth and Profitability in Banking. Journal of Money, Credit, and Banking, 36(6), 1069-1090. González, F. (2005). Bank regulation and risk-taking incentives: An international comparison of bank risk. [doi: DOI: 10.1016/j.jbankfin.2004.05.029]. Journal of Banking & Finance, 29(5), 1153-1184. Gorton, G., & Pennacchi, G. (1990). Financial Intermediaries and Liquidity Creation. The Journal of Finance, 45(No.1), 49-71. Holmström, B., & Tirole, J. (1998). Private and public supply of liquidity. Journal of Political Economy, Vol. 106, 1-40. 28 .

(35) . Huh, C., & Kim, S. B. (1994). Financial Regulation and Banking Sector Performance: A Comparison of Bad Loan Problems in Japan and Korea. Economic Review. Levine, R. (1997). Financial Development and Economic Growth: Views and Agenda. Journal of Economic Literature, Vol. 35(No. 2), 688-726. Miller, S. M., & Noulas, A. G. (1997). Portfolio Mix and Large-bank Profitability in the USA. Applied Economics, 29(4), 505-512. Molyneux, P., & Forbes, W. (1995). Market Structure and Performance in European Banking. Applied Economics, 27(2), 155-159. Molyneux, P., & Thornton, J. (1992). Determinants of European bank profitability: A note. Journal of Banking & Finance, 16(6), 1173-1178. Neely, M. C., & Wheelock, D. C. (1997). Why Does Bank Performance Vary across States? Federal Reserve Bank of St. Louis Review, 27-40. Niehans, J. (1978). The Theory of Money. Baltimore: Johns Hopkins Univ. Press. Patinkin, D. (1965). Money, Interest, and Prices: An Integration of Monetary and Value Theory (2d ed.): New York: Harper & Row. Repullo, R. (2004). Capital Requirements, Market Power, and Risk-taking in Banking. Journal of Financial Intermediation, 13(2), 156-182. Tobin, J. (1965). The Theory of Portfolio Selection. London: Macmil- lan. von Thadden, E. L. (1999). Liquidity Creation through Banks and Markets: Multiple Insurance and Limited Market access. European Economic Review, 43(4-6), 991-1006. Wooldridge, J. M. (2002). Econometric Analysis of Cross Section and Panel Data. MIT press. Wooldridge, J. M. (2006). Introductory Econometrics: A Modern Approach. Thomson South-Western, (3). 29 .

(36) . Table 1 Liquidity classification (sorted in Bankscope categories). This table is modified from the categories of Berger and Bouwman (2009) since the categories are different from these in Bankscope database. Assets Illiquid (weight=1/2). Semi-liquid (weight=0). Liquid (weight=-1/2). HP / Lease. Mortgages. Deposits with Banks. Other Loans. Loans to Banks. Cash and Due from Banks. Loans to Group Companies / Associates. Loans to Municipalities / Government. Due from Other Banks. Loans to Other Corporate. Due from Other Credit Institutions. Overdue Loans. Total Securities. Restructured Loans. Treasury Bills. Other non-performing Loans. Other Bills. Trust Account Lending. Bonds. Other Lending. CDs. Intangible Assets. Equity Investments. Other Non Earning Assets. Other Investments. Land and Buildings Total Fixed Assets. 30 .

(37) . Table 1 Liquidity classification (cont’d). Liabilities plus Equity: liquid (weight=1/2). Semi-liquid (weight=0). Illiquid (weight=-1/2). Deposits - Demand. Time Deposit. Subordinated Debt. Deposits – Savings. Commercial Paper. Other Liabilities. Banks Deposits. Other Funding. Total Equity. Municipalities / Government Deposits. Other Securities. Other Bonds. Commercial Deposits. Debt Securities. Mortgage Bonds. Other Deposits. Securities Loaned. Convertible Bonds. Off-balance sheet activities: Illiquid (weight=1/2). Semi-liquid (weight=0). Acceptances. Documentary Credits. Documentary Credits Guarantees Other. 31 .

(38) . Table 2 Variable definition This table summarize all variable used and the expected sign denotes the expected results of specific variable for regression on avroa.. Variable. Symbol. Expected Sign. Definition. avroa. --. 3 year average of return on average asset from year t-2 to t. sigavroa. --. 3 year standard deviation corresponding to AVROA. sharproa. --. Sharpe ratio of ROA = AVROA/sigavroa. Liquidity creation lnlc. +. Ln(liquidity creation). Size. size. ?. Ln(total asset). Capital. teta. +. total equity/total asset. gdpc. +. annual change(%) in GDP. gdpct_1. +. lagged GDP change. osp. ?. index for official supervisory power. pmi. ?. private monitoring index. bar. ?. bank activity regulatory. l_ta. --. gross loan/ total asset. d_fs. --. dummy for financial structure, 0 = Market-based, 1 = Bank-based. Performance. Macroeconomic. Regulation Liquidity Financial Structure. 32 .

(39) . Table 3 Financial system classifications This table shows the corresponding financial systems of the countries in our sample criteria and the data source is originated from Demirgüç-Kunt & Levine (1999). Financial System. Country. Bank-based. Belgium France Germany Italy Japan. Market-based. Canada Netherland Sweden Switzerland United Kingdom United States of America. 33 .

(40) . Table 4 Correlation coefficients . avroa avroa. sigavroa sharproa. lnlc. size. teta. gdpc. gdpct_1 gdpcxosp gdpcxpmi gdpcxbar. 1.0000. sigavroa 0.2153 1.0000 sharproa 0.1038 -0.3485. 1.0000. lnlc. -0.1477 -0.1588. 0.0502. 1.0000. size. -0.0385 -0.0284. -0.0284. 0.3161 1.0000. teta. 0.4645 0.3274. -0.0591. -0.4069 -0.2349 1.0000. gdpc. 0.3983 0.0674. 0.0588. -0.1232 -0.0524 0.2503 1.0000. gdpct_1. 0.3869 0.0717. 0.0547. -0.1184 -0.0346 0.2476 0.8408 1.0000. gdpcxosp 0.4183 0.0544. 0.1035. -0.0930 -0.0010 0.2107 0.9147 0.7559. 1.0000. gdpcxpmi 0.4444 0.0821. 0.0749. -0.1233 -0.0239 0.2615 0.9795 0.8458. 0.9527. 1.0000. gdpcxbar 0.4552 0.0738. 0.0700. -0.0913 -0.0450 0.2585 0.9485 0.8482. 0.8933. 0.9596. 34 . 1.0000.

(41) . Table 5 Descriptive statistics This table shows the descriptive statistics of all time-variant variables.. Variable. Obs. Mean. Std. Dev. Min. Max. avroa. 46050. 0.558549 0.663699 -1.38333 5.036667. sigavroa. 45692. 0.25529. sharproa. 45234. 7.689502 9.675609 -1.0505. lnlc. 63107. 11.43905 5.94134. teta. 63083. 8.999253 10.04948 0. size. 63112. 13.61435 1.820165 6.471707 21.82674. l_ta. 62775. 58.1515. gdpc gdpct_1. 63112 63112. 1.176872 1.307741 -1.810955 5.22 1.220063 1.317528 -1.810955 5.22. 0.426366 0.005774 3.810372 59.17843. -20.7909 20.88027. 20.47498 -0.23. 100 99.99. gdpcxosp 63112. 11.40942 14.0274. -23.5424 45.74921. gdpcxpmi 63112. 7.404327 9.223923 -14.4876 31.32. gdpcxbar 63112. 2.311303 3.486737 -5.97615 13.01324. Table 6 Unit root test results Fisher Pλ test is proposed by Maddala and Wu (1999) and test for the null hypothesis that the specific variable is nonstationary, and we test for all time varying variables include dependent and independent variables.. Variable. df.. Chi2. Prob > chi2. avroa. 10384. 1.76e+04. 0.0000. sigavroa. 10372. 2.08e+04. 0.0000. sharproa. 10334. 2.81e+04. 0.0000. lnlc. 12848. 1.96e+04. 0.0000. teta. 12838. 2.32e+04. 0.0000. size. 12848. 1.41e+04. 0.0000. l_ta. 12782. 2.17e+04. 0.0000. 35 .

(42) . Table 7 Endogenous regression result This model use fixed-effect model to test for the relationship of endogenous variable (LnLC) and instrument variables.. lnlc. Coef.. Std. Err.. t. P>t. _cons. -4.087***. 0.405228. -10.08. 0. size. 0.741***. 0.02917. 25.42. 0. teta. -0.125***. 0.002947. -42.36. 0. gdpc. 0.036*. 0.018408. 1.93. 0.053. gdpct_1. 0.059***. 0.016658. 3.53. 0. l_ta. 0.112***. 0.001482. 75.59. 0. Rsq=0.4306 Note: ***, ** and * denote significance at the 1%, 5% and 10% levels respectively.. Table 8 IV Regression results for full sample In this table we estimate the effects of dependent variables on performance by two stage least square (2SLS), and from panel A to C shows the result for unweighted 3 year average (AVROA), 3 year sigma of ROA (sigavroa) and Sharpe Ratio of ROA (sharproa).. Panel A: Regress for AVROA (N=46045) _cons. 0.110*. 0.060. 0.108. 0.132**. lnlc. 0.011***. -0.001. 0.011***. 0.012***. size. 0.013**. 0.029***. 0.013**. 0.011**. teta. 0.020***. 0.017***. 0.020***. 0.020***. gdpc. 0.001. 0.019*. -0.016. 0.032***. gdpct_1. -0.032***. -0.033***. -0.033***. -0.031***. gdpcxosp. -0.002*. gdpcxpmi. 0.003. gdpcxbar. -0.017***. R-sq. 0.1156. Hausman Chi2. 2,924.770*** 3,473.860*** 2,575.310*** 2,517.570***. Davidson-MacKinnon F 39.053***. 0.1063 38.396***. 36 . 0.1284 38.036***. 0.0683 43.233***.

(43) . Panel B: Regress for sigavroa (N=45685) _cons. 0.591***. 0.609***. 0.594***. 0.608***. lnlc. -0.015***. -0.015***. -0.015***. -0.015***. size. -0.016***. -0.018***. -0.017***. -0.018***. teta. 0.006***. 0.006***. 0.006***. 0.006***. gdpc. 0.009***. -0.016. 0.038***. 0.034***. gdpct_1. 0.009***. 0.009***. 0.010***. 0.010***. gdpcxosp. 0.002***. gdpcxpmi. -0.005**. gdpcxbar. -0.014***. R-sq. 0.0729. 0.0713. 0.0706. 0.0706. Hausman Chi2. 415.650***. 434.560***. 441.330***. 449.720***. 53.738***. 52.365***. 51.322***. Davidson-MacKinnon F 54.500***. Panel C: Regress for sharpROA (N=45685) _cons. 14.945***. 13.736***. 14.874***. 14.602***. lnlc. 0.098*. 0.092*. 0.090*. 0.090*. size. -0.538***. -0.455***. -0.526***. -0.508***. teta. -0.025*. -0.022. -0.026*. -0.024*. gdpc. -0.419***. 1.328***. -1.312***. -0.974***. gdpct_1. -0.254***. -0.262***. -0.277***. -0.273***. gdpcxosp. -0.169***. gdpcxpmi. 0.146***. gdpcxbar. 0.298***. R-sq. 0.0002. 0.0042. 0.0016. 0.0015. Hausman Chi2. 37.090***. 535.380***. 916.700***. 1,310.790***. 3.085*. 3.173*. 3.127*. Davidson-MacKinnon F 3.789*. Note: *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. The Hausman Chi2 and Davidson-MacKinon F is using to compare the IV versus OLS regression and fixed-effect versus random-effect, and only the fitting one would be shown in table.. 37 .

(44) . Table 9 IV regression results split by financial structure Same as table 7, this model is estimated by two stage least square (2SLS) both in bank-based and market-based countries, and from panel A to C shows the result for unweighted 3 year average (AVROA), 3 year sigma of ROA (sigavroa) and Sharpe Ratio of ROA (sharproa) respectively.. Panel A: Results for unweighted 3 year average ROA Market-Based(N=12738). avroa. _cons lnlc size teta gdpc gdpct_1 gdpcxosp gdpcxpmi gdpcxbar. 0.733*** 0.031*** -0.010 0.015*** 0.031*** -0.089***. R-sq. 0.0601. 0.721*** 0.031*** -0.009 0.015*** 0.053** -0.088*** -0.002. 0.715*** 0.031*** -0.009 0.015*** 0.101 -0.088***. Bank-Based(N=33185) 0.703*** 0.031*** -0.007 0.015*** 0.138*** -0.089***. -0.099 -0.001* 0.025*** 0.024*** -0.006** -0.018***. -0.100 -0.001* 0.025*** 0.024*** -0.005 -0.018*** 0.000. -0.009. -0.120* -0.001* 0.027*** 0.025*** -0.058*** -0.019*** 0.009***. -0.050***. Hausman Chi2 273.400*** Davidson-MacKinnon F 78.635***. -0.093 -0.001* 0.024*** 0.024*** -0.002 -0.018***. -0.002. 0.0581. 0.059. 0.0294. 0.1648. 0.1653. 0.1857. 0.1597. 302.910*** 78.569***. 277.340*** 78.673***. 490.460*** 81.546***. 1098.81*** 2.546. 1039.04*** 2.627. 926.19*** 3.758*. 968.47*** 2.492. Note: *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. The Hausman Chi2 and Davidson-MacKinon F is using to compare the IV versus OLS regression and fixed-effect versus random-effect, and only the fitting one would be shown in table.. 38 .

(45) . Panel B: Results for 3 year sigma ROA Market-Based(N=12853). sigavroa. _cons lnlc size teta gdpc gdpct_1 gdpcxosp gdpcxpmi gdpcxbar. 0.489*** -0.023*** -0.001 0.004*** -0.001 0.046***. R-sq. 0.0735. 0.451*** -0.023*** 0.003 0.004*** 0.059*** 0.048*** -0.005***. 0.431*** -0.023*** 0.004 0.004*** 0.217*** 0.048***. Bank-Based(N=32694) 0.490*** -0.023*** -0.001 0.004*** -0.005 0.046***. 0.636*** -0.011*** -0.026*** 0.008*** 0.009*** -0.002. 0.680*** -0.010*** -0.029*** 0.008*** -0.049*** 0.000 0.006***. -0.029***. 0.662*** -0.010*** -0.029*** 0.008*** 0.074*** 0.000 -0.011***. 0.002. Hausman Chi2 252.480*** Davidson-MacKinnon F 44.008***. 0.701*** -0.010*** -0.032*** 0.008*** 0.051*** 0.000. -0.024***. 0.0766. 0.0779. 0.0736. 0.0478. 0.0325. 0.0512. 0.0488. 252.350*** 44.198***. 251.350*** 44.031***. 252.700*** 44.117***. 148.910*** 15.758***. 511.360*** 13.257***. 10.860* 13.130***. 340.400*** 12.301***. Note: *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. The Hausman Chi2 and Davidson-MacKinon F is using to compare the IV versus OLS regression and fixed-effect versus random-effect, and only the fitting one would be shown in table.. 39 .

(46) . Panel C: Regress for sharpROA Market-Based(N=12754). sharproa. _cons lnlc size teta gdpc gdpct_1 gdpcxosp gdpcxpmi gdpcxbar. 19.957*** 0.197*** -0.781*** -0.029 -0.194 -0.876***. R-sq. 0.031. 19.707*** 0.197*** -0.757*** -0.028 0.267 -0.860*** -0.041. 19.965*** 0.197*** -0.782*** -0.029 -0.228 -0.876***. Bank-Based(N=32297) 19.635*** 0.200*** -0.742*** -0.025 0.912** -0.875***. 12.720*** -0.001 -0.360** -0.013 -0.449*** -0.080. 10.945*** 0.004 -0.279* -0.008 2.044*** -0.138* -0.250***. 0.005. 11.763*** 0.001 -0.278* -0.005 -3.012*** -0.121* 0.455***. -0.521***. Hausman Chi2 62.540*** Davidson-MacKinnon F 7.039***. 11.339*** 0.049*** -0.305*** -0.054*** -1.347*** 0.044. 0.659***. 0.0272. 0.031. 0.0232. 0.0012. 0.0204. 0.0039. 0.0003. 299.540*** 7.023***. 189.320*** 7.039***. 72.660*** 7.385***. 170.03*** 0.333. 498.84*** 0.002. 149.35*** 0.012. 189.28*** 0.005. Note: *, ** and *** denote significance at the 10%, 5% and 1% levels respectively. The Hausman Chi2 and Davidson-MacKinon F is using to compare the IV versus OLS regression and fixed-effect versus random-effect, and only the fitting one would be shown in table.. 40 .

(47) . Appendix: Survey Questions of Bank Regulation and Supervision. I. Based on Barth, Caprio, and Levine (2004), the survey questions used to construct the official supervisory power:. 1.. Does the supervisory agency have the right to meet with external auditors to discuss their report without the approval of the bank? Yes/No. 2.. Are auditors required by law to communicate directly to the supervisory agency any presumed involvement of bank directors or senior managers in explicit activities, fraud, or insider abuse? Yes/No. 3.. Can supervisors take legal action against external auditors for negligence? Yes/No. 4.. Can the supervisory authority force a bank to change its internal organizational structure? Yes/No. 5.. Are off-balance sheet items disclosed to supervisors? Yes/No. 6.. Can the supervisory agency order the bank’s directors or management to constitute provisions to cover actual or potential losses? Yes/No. 7.. Can the supervisory agency suspend the directors’ decision to distribute dividends? Yes/No. 8.. Can the supervisory agency suspend the directors’ decision to distribute bonuses? Yes/No. 9.. Can the supervisory agency suspend the directors’ decision to distribute management fees? Yes/No. 10. Can the supervisory agency legally declare—such that this declaration supersedes the rights of bank shareholders—that a bank is insolvent? Yes/No 11. Does the Banking Law give authority to the supervisory agency to intervene—that is, suspend some or all ownership rights—a problem bank? 41 .

(48) . Yes/No 12. Regarding bank restructuring and reorganization, can the supervisory agency or any other government agency do the following? Yes/No 13. Supersede shareholder rights? Yes/No 14. Remove and replace management? Yes/No 15. Remove and replace directors? Yes/No II. Based on Barth, Caprio, and Levine (2004), the survey questions used to construct the private monitoring index:. 1.. Whether bank directors and officials are legally liable for the accuracy of information disclosed to the public? Yes/No. 2.. Whether banks must publish consolidated accounts? Yes/No. 3.. Do banks must be audited by certified international auditors? Yes/No. 4.. Whether 100% of the largest 10 banks are rated by international rating agencies? Yes/No. 5.. Are off-balance sheet items are disclosed to the public? Yes/No. 6.. Whether banks must disclose their risk management procedures to the public? Yes/No. 7.. Whether accrued, though unpaid interest/principal enter the income statement while the loan is still non-performing? Yes/No. 8.. Is subordinated debt is allowable as part of capital? Yes/No. 9.. Whether there is no explicit deposit insurance system and no insurance was paid the last time a bank failed? Yes/No. 42 .

(49) . III. Based on Barth, Caprio, and Levine (2004), the survey questions used to construct overall bank activities and ownership restrictiveness:. Bank activities restrictiveness: 1. What is the level of regulatory restrictiveness for bank participation in securities activities (the ability of banks to engage in the business of securities underwriting, brokering, dealing, and all aspects of the mutual fund industry)? 2.. What is the level of regulatory restrictiveness for bank participation in insurance activities (the ability of banks to engage in insurance underwriting and selling)?. 3.. What is the level of regulatory restrictiveness for bank participation in real estate activities (the ability of banks to engage in real estate investment, development, and management)? Unrestricted = 1: full range of activities can be conducted directly in the bank. Permitted = 2: full range of activities can be conducted, but some or all must be conducted in subsidiaries. Restricted = 3: less than full range of activities can be conducted in the bank or subsidiaries. Prohibited = 4: the activity cannot be conducted in either the bank or subsidiaries.. Ownership restrictiveness: 1. What is the level of regulatory restrictiveness for bank ownership of nonfinancial firms? Unrestricted = 1: a bank may own 100 percent of the equity in any nonfinancial firm Permitted = 2: a bank may own 100 percent of the equity of a nonfinancial firm, but ownership is limited based on a bank’s equity capital Restricted = 3: a bank can only acquire less than 100 percent of the equity in a nonfinancial firm Prohibited = 4: a bank may not acquire any equity investment in a nonfinancial firm. Source: World Bank guide questions, which is available from World Bank research (Bank Regulation and Supervision) or Barth, Caprio, and Levine (2004).. 43 .

(50)