Disclosure and the cost of debt financing in banking: Empirical evidence from the European Union

Transcription

1 Disclosure and the cost of debt financing in banking: Empirical evidence from the European Union Oliver Müller * Abstract: Employing panel data on 181 quoted credit institutions across the 27 member states of the European Union for the period from 1997 to 2009, this study provides empirical evidence that more detailed information disclosure reduces the bank s cost of debt capital. This result remains robust when controlling for endogeneity concerns and accounting for the unbalanced structure of the underlying dataset. Moreover, results from further analyses clearly indicate that the nexus between information disclosure and cost of debt in banking is affected by national banking market structures as well as regulatory and supervisory strategies. Our findings reveal that banks either operating in concentrated banking markets that are not contestable or in countries having imposed excessive regulations are penalized with higher costs on their debt capital when refusing to disclose comprehensive information. JEL classification: G 14, G 18, G 21, L 22 Keywords: Disclosure, Cost of debt, Banking regulation, Market structure * Correspondence address: University of Bochum, Department of Economics, Bochum, Germany, oliver.mueller@rub.de.

2 1. Introduction In response to the recent financial crisis, supervisory and regulatory authorities around the world have proposed various regulatory resolutions being determined to avoid or mitigate financial meltdowns in the future and to foster the stability of the banking system. Most recently, the members of the Basel Committee on Banking Supervision have agreed on a new regulatory standard (Basel III), which primarily enhances capital adequacy requirements for banks and aims at reducing systemic shocks by increasing the resilience at the individual bank level (BIS, 2011). While these measures focus on making banks regulatory capital more sensitive to risk and improving the supervisory review process, little attention has been paid to the advancement of public disclosure requirements in the banking industry. This is surprising, not only since the improvement of market discipline through disclosure marked a milestone in the Basel II framework (Pillar III), but even the majority of empirical studies on disclosure in financial markets stress the relevancy of empowering private monitoring and forcing banks to disclose more detailed information on their risk-taking behavior. For instance, BARTH ET AL. (2004) find that regulatory and supervisory practices promoting accurate information disclosure while providing incentives for private agents to exert corporate control foster bank performance and stability. GELOS and WEI (2005) provide evidence that emerging market funds privilege portfolio investments in more transparent countries and are less likely to exit these countries during a financial crisis. BECK ET AL. (2006) argue that supervisory strategies that focus on empowering private monitoring and force banks to disclose accurate information improve firm s access to capital markets and NIER and BAUMANN (2006) show that stronger market discipline through disclosure results in larger capital buffers of the banking industry. Finally, DEMIRGÜÇ-KUNT ET AL. (2008) even postulate that countries intended to upgrade banking regulation and supervision should give priority to information disclosure over other elements being associated with Basel II. 2

3 This paper evaluates the consequences of the availability and the level of detail of information disclosed in banks annual reports on the cost of their debt capital. The study of this issue is motivated by economic theory suggesting that enhanced disclosure requirements make use of the market s disciplining forces and thereby may reduce information asymmetries between bank managers and market participants. This is based on the assumption that well informed providers of capital will reward a credit institution s risk-conscious strategy and effective risk control in demanding lower risk premiums when making their investment decisions and will penalize risky behavior respectively (BLISS, 2004; FLANNERY, 2001). Since risk premiums are part of the bank s cost of debt funding, this coherency may act as a strong incentive and encourage bank managers to efficiently monitor and manage the bank s risk positions. Our analysis contributes to this field of research and complements and extends previous empirical studies on information disclosure by three specific aspects. First, while previous studies analyze the impact of disclosure on the cost of debt for non-financial firms (MIL- LER and PUTHENPURACKAL, 2002; SENGUPTA, 1998) or do not distinguish between nonfinancial and financial firms (FRANCIS ET AL. (2005)), we explicitly focus on credit institutions and thereby account for their role in resource allocation between surplus and deficit sectors in an economy. Second, to the best of our knowledge, this is the first study that explicitly examines the nexus between bank-level disclosure and debt capital funding costs for banking institutions located in the EU 27. While previous studies either refer to a single country (e.g. SENGUPTA, 1998) or a broader but not further defined set of countries (GELOS and WEI, 2005; FRANCIS ET AL., 2005; MILLER and PUTHENPURACKAL, 2002), the analysis at hand provides novel evidence by exclusively focusing on the EU s 27 member states. The EU is particularly suitable and interesting for this investigation since the structure and regulation of national banking markets is still heterogeneous although national and European legal systems are gradually becoming intertwined, with the primacy of Community law prevailing. Hence, we 3

4 additionally control for banking market structures and characteristics in the member states of the European Union as well as national regulatory and supervisory strategies. Finally, with regard to the timeframe of this investigation covering the period from 1997 to 2009, this study implicates the recent financial crisis and its aftermath on banks debt funding costs. Employing panel data on 181 quoted commercial, savings and cooperative banks as well as bank holding companies across the 27 member states of the European Union for the period from 1997 to 2009, our analysis initially provides empirical evidence that more detailed information disclosure in published accounts decreases the bank s cost of debt capital. Our baseline finding is reinforced when controlling for endogeneity concerns and accounting for the unbalanced structure of the underlying dataset. Moreover, results from further analyses reveal that the link between banks cost of debt and the amount of information disclosed varies among national banking market structures and characteristics as well as different regulatory frameworks across the member states. The remainder of the paper is structured as follows. While Section 2 contains a discussion of the related literature, Section 3 outlines the sample selection and describes the variables and methodology employed in this study. Section 4 presents the empirical results of the baseline regressions as well as further sensitivity analyses. Finally, Section 5 concludes the paper. 2. Related literature This paper contributes to the rich strand of empirical literature on the determinants of debt capital s cost of funding. Results of related studies clearly point to the existence of causal effects of borrower risk (STRAHAN, 1999), covenants (SMITH and WARNER, 1979), (geographic) diversification of assets (DENG ET AL., 2007), the borrower s takeover vulnerability (CHAVA ET AL., 2009), accounting quality (BHARATH ET AL., 2008), auditor choice (PITTMAN and FORTIN, 2004), lending relationship (BHARATH ET AL., 2011), product market competi- 4

5 tiveness (VALTA, 2011), national laws and institutions (QIAN and STRAHAN, 2008), creditor rights and its enforceability (BAE and GOYAL, 2009) or political rights (QI ET AL., 2010) on the costs of debt capital. Although some kind of analogy to research questions raised in these studies does exist, the number of studies explicitly dealing with the nexus between information disclosure and the funding cost of debt capital in the banking industry is still small. To begin with, in an influential study, SENGUPTA (1998) studies the relationship between a firm s overall disclosure practices and its cost of debt financing, proxied by the yield to maturity of corporate bonds, while controlling for issue characteristics and market conditions. He uses an index obtained from the annual volumes of the Report of the Financial Analysts Federation Corporate Information Center (FAF) to measure the disclosure quality of 103 non-financial firms for the period from 1987 to 1991 and provides empirical evidence of an inverse relationship and concludes that lenders take into consideration the firm s disclosure quality when demanding default risk premiums. Moreover, his results indicate that this causal relationship is stronger the higher market uncertainty surrounding the firm. MILLER and PUTHENPURACKAL (2002) study the influence of investor protection and information disclosure, measured by two indices assessing creditor rights and rule of law at the country-level, on the pricing of 260 fixed-rate corporate Yankee bonds issued by firms headquartered outside the United States from 1987 to They find that investors demand a higher premium on bond issues from firms located in countries disregarding more detailed information disclosure in their legislation. This relationship is especially reinforced for firsttime issuers suggesting that asymmetric information is a major determinant of investor s risk premiums. More recently, FRANCIS ET AL. (2005) investigate if the disclosure policy of 484 firms located in 34 non-u.s. countries around the world is an instrument for mitigating information asymmetries. Their study reveals empirical evidence that firms with greater external financing 5

6 needs have higher voluntary disclosure levels, proxied by disclosure scores retrieved from the Center for International Financial Analysis and Research (CIFAR) database. Moreover, even when controlling for further institutional characteristics measured at the country-level, their results reveal that firms pursuing a more expanded disclosure policy constantly have better access to debt funding at lower costs. 3. Empirical analysis 3.1. Data and sources Our empirical analysis entails 181 quoted commercial, savings and cooperative banks as well as bank holding companies across the 27 member states of the European Union for the period 1997 to 2009 following the creation of the European single banking license. While notes on variables and data sources are presented in Table 1, Table 2 reports descriptive statistics for the entire dataset. Correlations among variables either used in the baseline or further regressions are provided in Tables 4 and Cost of debt In contrast to related empirical work on determinants of the cost of debt funding of (non-) financial firms (e.g. BOUBAKRI and GHOUMA, 2010; DENG ET AL., 2007; DE BONDT, 2005), we do not employ bond yield-spreads as a proxy for the bank s funding costs since it does not account adequately for the more diversified funding base of financial firms compared to the non-financial sector. Not only raise financial institutions short-term and long-term wholesale debt capital on the capital or interbank market, but banks traditionally engage in asset transformation and serve as intermediaries between savers and borrowers (FREIXAS and ROCHET, 2008). Accounting for the specifics of the banking industry s debt structure, our dependent variable is, similar to PITTMAN and FORTIN (2004), the average interest rate paid on the bank s debt, calculated as the annual interest expense divided by the bank s average inter- 6

7 est-bearing liabilities during the year. We retrieve this variable from the BankScope database, which is compiled by Fitch Ratings and contains comprehensive micro-level information on public and private banks around the globe. Due to the high degree of institutional and regional heterogeneity in the sample and the long-term horizon of our analysis over different business cycles, our dependent variable is widely spread around the mean of 3.79 percent (Table 2) Disclosure index Following NIER and BAUMANN (2006) and their rich work on the effects of disclosure, we construct a disclosure index which represents the availability and the level of detail of publicly available information on the bank s risk-taking behavior. We assess the amount of information disclosed in the banks annual reports based on published accounts in the BankScope database serving as our data source. In total, 18 subindices are defined that are either related to the sources of risk in terms of interest rate, market, credit and liquidity risk or to the risk-bearing ability of the bank. Thus, the index provides detailed information about the amount of information disclosed regarding the current conditions and future prospects of each bank included in the analysis. For each subindex, a 1 is assigned if there is an entry in the corresponding category in the BankScope database and a 0 otherwise. The overall disclosure index is equal to the sum of all dummy variables and ranges between 0 and 18 with higher index values corresponding to a higher level of disclosure. As represented by the minimum and maximum values of 3 and 17 respectively and a standard deviation of 2.37, there is considerable dispersion around the mean of the disclosure index of The average value of the index increases from just about 9 in 1997 to more than 12 in 2009 and it heavily varies among and within the banks included in our sample. While detailed information on the construction of the index and its subindices is provided in Table 3, Figure 1 within the Appendix provides a box plot of the disclosure index for the period of interest. 7

8 Two major advantages favor the use of an index to measure the availability and quantity of data disclosed. First, in contrast to related studies that control for disclosure at the country-level and thus rely on cross-country differences alone (e.g. DEMIRGÜÇ-KUNT ET AL., 2008; BARTH ET AL., 2004; MILLER and PUTHENPURACKAL, 2002), the index evaluates the degree of disclosure at the bank-level. Hence, it is a more accurate and more meaningful measure since it explicitly accounts for variation over time within banks and the degree of heterogeneity among banks which may stem from different business models, ownership structures or managerial skills. Second, although we rely on data retrieved from banks annual reports, we do not proxy disclosure by the accounting regime (e.g. LEUZ and VERRECCHIA, 2000). Since even high-quality reporting standards allow alternative accounting treatments, a proxy variable serves as an indirect measure of disclosure at best Further control variables When examining the relationship between a bank s level of disclosure and its cost of debt, it is imperative to control for further bank-specific and macroeconomic characteristics that are likely to affect our left-hand side measure and hence help mitigate omitted variable bias. The following well-accepted variables control for variation in debt pricing attributable to bank-specific characteristics other than the bank s level of disclosure. We retrieve bankspecific variables from aggregating consolidated balance sheet data from BankScope database per country and year. In order to avoid noise resulting from extreme outliers of our bankspecific variables, we discard observations outside the 5th and 95th percentiles respectively of 1 We are aware of the fact that the index measures the availability and the amount of information disclosed in annual reports, but neither the content of information nor any information in the market provided by third parties such as (financial) regulatory authorities or rating agencies. NIER (2005) provides a discussion on the use of indices in empirical studies on the effects of disclosure and stresses the advantages of these indices over proxy measures for disclosure. 8

9 the distribution of each variable. To begin with, we control for the bank s size, measured as the log of its total assets, and expect an inverse relationship between size and the bank s cost of debt. Large banks may be able to reap economies of scale in raising external funds (YILD- IRIM and PHILIPPATOS, 2007), benefit from the ability to better diversify their income and asset structure (ALTUNBAS ET AL., 2007; HUGHES ET AL., 2001) and reduce information asymmetries between firm insiders and capital markets (BOOTH ET AL., 2001; RAJAN and ZINGALES, 1995). Next to size, the bank s cost of debt capital may depend on its overall asset quality, proxied by the share of the bank s loan loss provisions to total assets. Related literature suggests ambiguous theories on the nexus between loan loss provision and its effects on investors behavior. On the one hand, cost of debt may increase with decreasing asset quality since a low quality of the bank s loan portfolio may boost loan loss provisions in order to cover expected losses (WALL and KOCH, 2000) and hence to guarantee the solvency and capitalization of the bank. On the other hand, KANAGARETNAMA ET AL. (2005); AHMED ET AL. (1999) and BEAVER ET AL. (1989) hypothesize that an increase in loan loss provisions may indicate that the bank s managers calculate future earnings prospects to be sufficiently strong to withstand a hit to earnings in the form of additional loan loss provisions, regarded as a positive signal to investors. Furthermore, we control for the bank s funding structure, measured as the share of the bank s gross loans to its total deposits. Since deposits are considered to be a more stable form of funding and high loan-to-deposit ratios may indicate possible liquidity problems of banks that are more dependent on having constant access to money and capital markets compared to banks being able to fund their loans through deposits, we expect this control variable to be positively related to the bank s interest rate on debt capital. 2 2 In contrast to related studies (e.g. REEB ET AL., 2001), we do not control for a bank s credit rating since the major rating agencies claim to assess the level of information disclosed in the rating process. 9

10 Turning to macroeconomic characteristics that are likely to impact the bank s cost of debt, we follow SENGUPTA (1998) and MILLER and PUTHENPURACKAL (2002) and control for the risk-free interest rate, proxied by Moody's seasoned Aaa corporate bond yield and expect a positive relationship between increasing market interest rates and cost of debt in banking. In addition, cost of debt may increase in rising inflation rates, defined as the annual percentage change in consumer prices, since inflation may erode the value of debt and make raising debt capital more difficult (FRANCIS and OSBORNE, 2009). Finally, we include the annual GDP growth rate to capture the state of the economy within the business cycle and predict a negative impact on our dependent variable since economic growth reflects the future potential of investors and creditors investments (BOUBAKRI and GHOUMA, 2010). Because of the high degree of heterogeneity in the level of economic strength among the member states of the European Union, our macroeconomic controls vary quite a bit Empirical model In order to study the impact of bank-level disclosure on the cost of debt capital, we estimate the following fixed-effects model on panel data: y it = α i + Σ ß k x it,k + Σ ß l x jt,l + ε it, where y it represents the cost of debt capital of bank i in year t. While the vector x it,k includes the disclosure index as well as further bank-specific determinants k of the interest rate on debt, x jt,l describes both macroeconomic characteristics measured at the country-level l and employed in the baseline regression as well as variables proxying national banking market structures and regulatory strategies in further analyses. Moreover, the model includes banklevel fixed effects α i accounting for unobserved time-invariant measures that may affect the 3 For reasons of clarity, we discuss control variables employed in regressions other than the baseline model in Sections 4.2 and

11 bank s interest rate on debt capital. ε it is the independent and identically distributed (iid) random error term. In addition, to control for heterogeneity in funding costs across financial institutions due to regional or institutional aspects, the model includes cluster-robust standard errors at the bank-level. Estimating the model with fixed effects is a consequent strategy for two reasons. First, the relationship between a bank s level of disclosure and its cost of debt capital may be characterized by endogeneity, resulting in biased and inconsistent estimates. From a theoretical point of view, endogeneity may either result from simultaneity or omitted unobservable firmspecific factors. With regard to the first source, related studies have already provided evidence that simultaneity is not likely to bias the result of OLS estimation on disclosure and cost of debt (HAIL, 2002; WELKER, 1995). 4 Hence, in the context of disclosure, two sources of unobservable firm characteristics may distort estimation results, namely (1) the costs of disclosure and (2) entrepreneurial skills of the bank s management. To begin with costs, VERRECCHIA (2001) discusses an equilibrium in which the bank s management does not fully inform about the firm s perspectives but shares favorable information with market participants. The discretionary choice of managers to share or to withhold specific information in order to maximize the value of the firm is, in turn, associated with direct or indirect costs, i.e. rising or falling costs of (debt) capital. Turning to the bank management s entrepreneurial skills, it seems intuitive that highly-qualified managers may be better able to make decisions that mitigate inefficiencies, facilitate entrepreneurial development (e.g. with regard to capital structure adjustments) and avoid the disutility of liquidation ( threat-of-liquidation effect ; SCHMIDT, 1997). Investors may appreciate these advantages and adjust their expected return requirements in terms of lower risk costs as part of the bank s interest costs on its debt. 4 Since we cannot rule out that simultaneity may bias our results, we lagged our variable of main interest by one period. In Section 4.1, we provide further robustness checks on simultaneity. 11

12 Addressing the endogeneity bias in the relationship between cost of debt capital and disclosure, NIKOLAEV and VAN LENT (2005) provide econometric evidence that collecting multiple observations for each firm in the sample and applying a fixed effects model yields consistent and unbiased results, but only on condition that changes over time within each firm are driving the relation of interest. Hence, as reported in Table 6, we calculate a transition probabilities matrix, which provides information on the probability of a bank changing its level of disclosure from year t to year t+1. In our sample, the average probability of staying at the same level of disclosure does not exceed 50 percent indicating a substantial over-time variation in each bank s policy of disclosure. Hence, the estimation of a model with fixed effects is appropriate for our sample of listed credit institutions. Second, from an econometric point of view, the issue of correlated errors is the major determinant in discriminating between fixed and random effects models. The random effect assumption is that the individual specific effect is uncorrelated with the independent variables whereas the fixed effects model assumes correlation between the individual effect and the exogenous measures. Since we introduce cluster-robust standard errors at the bank-level and the Hausman test (HAUSMAN, 1978) is inappropriate under heteroscedasticity, we employ a generalization of the Hausman approach proposed by ARELLANO (1993) to test for the appropriateness of our model specification. Adopting this approach, the null hypothesis of no correlation between the individual specific effect and the independent variables is rejected at ρ < 0.00, suggesting that estimating a fixed effects model is appropriate in our case. 4. Empirical results We summarize major results of our baseline regression in Table 7, regression specification (1). Specifications (2) and (3) are robustness checks addressing endogeneity concerns discussed in Section 3.2 and specification (4) considers the unbalanced structure of the underlying dataset. While regressions controlling for banking market structures and characteristics 12

13 in the European Union are presented in Table 8, regressions addressing the nexus between disclosure and the national regulatory legislation under which listed banks operate are reported in Table Main findings and robustness checks As presented in Table (7), specification (1), the disclosure index enters the regression significantly negative at the five percent level, indicating that the costs of listed banks debt capital inversely depend on the amount of information disclosed. This finding supports the theoretical argument that greater disclosure lowers information asymmetries between creditors and borrowing banks (BLISS, 2004; FLANNERY, 2001) and is in line with previous empirical findings on the nexus between disclosure and the cost of debt of (non-) financial firms (MILLER AND PUTHENPURACKAL, 2002; SENGUPTA, 1998). Based on the empirical model employed, an increase in the disclosure index by one level results in a decrease in the average interest rate paid on the bank s debt of 8.13 basis points. Among the control variables employed in this study, neither of our bank-specific measures size, asset quality and funding structure enters the baseline regression statistically significant, although showing the expected signs. Turning to macroeconomic controls and introducing the risk-free interest rate, the variable enters the regression statistically significant at the 0.01 level providing evidence for a positive relationship between market-based interest rates and the bank s cost of debt. This result is not surprising and it is precisely because the investor can obtain the risk-free interest rate with no risk, it is implied that taking on additional risk will be rewarded with an interest rate higher than the risk-free rate. Turning to inflation, the control variable enters our baseline regression significantly positive at the one percent level revealing that rising inflation rates lead to an increase in the bank s cost of debt capital. This result is in line with FRANCIS and OSBORNE (2009) who provide empirical evidence that growing inflation rates erode the nominal value of debt and 13

14 hence make it more difficult for firms to raise additional debt capital. Moreover, FAVERO and GIAVAZZI (2005) call attention to the role of tight monetary conditions and show that active policies such as the Taylor rule lead nominal interest rates to increase in response to inflation, which in turn accelerates the growth of nominal debt resulting in rising costs of debt capital. Finally, our measure for a country s business cycle enters the regression significantly negative at the one percent level indicating that the bank s cost of debt cyclically decreases during economic upturns while they increase in periods of economic contraction. This finding reinforces arguments produced by BOUBAKRI and GHOUMA (2010) who state that investors are sensitive to economic growth since it reflects the future potential of their investments. In addition, YILDIRIM and PHILIPPATOS (2007), SCHURE ET AL. (2004) and DEMIRGÜÇ-KUNT and MAKSIMOVIC (1998) provide empirical evidence that countries with relatively high GDP growth rates are characterized by more efficient banking institutions. Since improved efficiency lowers a bank s risk of default, we consequently suggest that creditors may adjust their risk premiums accordingly leading to a decrease in the bank s cost of debt capital. By means of regressions (2) to (4) we investigate the robustness of our results when addressing endogeneity concerns discussed in Section 3.2 and controlling for missing values in our unbalanced panel dataset. To begin with endogeneity concerns, HAIL (2002) and WELKER (1995) propose that simultaneity is not likely to bias results of OLS estimation on disclosure and cost of debt. However, since we lagged the disclosure index by one period as a precaution, we introduce the index simultaneously in specification (2). Furthermore, although NIKOLAEV and VAN LENT (2005) provide econometric evidence that applying a fixed effects model when estimating the impact of disclosure on the costs of debt yields consistent and unbiased results, we initially address this econometric problem by eliminating our bank-specific control variables in regression specification (3). In both cases findings from our baseline regression are reiterated suggesting that main results are robust and not biased due to bankspecific endogeneity. As regards the absolute coefficient value of the disclosure measure, it 14

15 turns out to be slightly higher in specification (2) and marginally lower in specification (3) as compared to the baseline regression. We finally control for any bias that may result from the unbalanced structure of our dataset due to missing observations. Hence, we build a balanced sample which is compiled by discarding the firm year if any missing observations are encountered in our period of interest. As reported in specification (4), main findings remain robust, but the estimate of our variable of main interest is slightly lower compared to the baseline regression suggesting that a onelevel increase in the disclosure index reduces a bank s cost of debt by 7.48 instead of 8.13 basis points Disclosure and banking market structures and characteristics Although the European Commission pursues a pro-active competition policy to identify and help tackle (regulatory) barriers in the Single Market in order to promote employment and economic growth (EUROPEAN COMMISSION, 2010, 2005), banking market structures remain heterogeneous across member states. Hence, in this Section, we examine the relationship between a bank s level of disclosure and its cost of debt while controlling for the structures and characteristics of national banking markets in the European Union. Results are presented in Table 8. To be upfront with it, our main finding of an inverse relationship between disclosure and the bank s cost of debt capital is reiterated among all regression specifications and the coefficients of our variable of main interest range between and basis points. We initially control for the intensity of competition in national banking markets by employing the H-statistic proposed by PANZAR and ROSSE (1987). The H-statistic is suitable and advantageous to measure competition since it (1) is estimated cross-sectionally for each country and year in the sample using bank-level data and (2) captures competitive behavior of other market participants serving as a measure of direct competitive conduct. Following SCHAECK ET AL. (2006), the H-statistic is a measure of the sum of the elasticities of the re- 15

16 duced-form revenues from interest-generating and non-interest generating activities with respect to factor prices in terms of operating and administrative, interest as well as personnel expense. Hence, H 0 indicates a monopoly in which higher factor costs correspond with a decrease in revenues since the bank is operating at the price elastic portion of its demand function; 0 < H < 1 describes monopolistic competition under which the bank faces an inelastic demand curve and hence interest and non-interest revenues grow underproportionally as factor costs increase, and finally H = 1 indicates perfect competition with free market entry and thus revenue increasing at the same rate as the increase in costs. As presented in regressions (1) and (2) in Table 8, we adopt two different approaches to estimate the intensity of competition in European banking markets. Introducing the H-statistic (1) restricted to interest generating activities, this variable enters specification (1) significantly negative at the 0.1 level indicating that banks operating in monopolistic or perfectly competitive loan markets are able to slash their costs of debt by increasing the amount of information disclosed in their annual reports. Since increasing competition may precipitate (cost and profit) efficiency among European banks through shifts in output ( competitionefficiency hypothesis ; DEMSETZ 1973; SCHAECK and ČIHÁK 2006) and may help prevent managers from enjoying a quiet life (BERGER and HANNAN, 1998), we suggest that market participants reward corporate development and adjust cost of debt requirements in turn. Turning to the H-statistic (2), which additionally accounts for non-interest generating activities, the variable enters regression (2) with a negative sign, but statistically insignificant. VENNET (2002) argues that competition will lead to more concentrated banking markets through the liquidation of inefficient competitors. Hence, we refer to IO-related research in banking within the structure-conduct-performance paradigm and include the degree of local concentration measured by the fraction of assets of a country s total banking system s assets held by the largest five domestic banks. As shown in specification (3), the variable enters our regression significantly positive at the one percent level indicating that higher cost of debt 16

17 capital are imposed on banks refraining from disclosing information while operating in concentrated banking markets. This result is complementary to CHORTAREAS ET AL. (2011), YILD- IRIM and PHILIPPATOS (2007) and BERGER and HANNAN (1998) who suggest that a lack of market discipline in concentrated markets results in a reduction of a bank s profit and efficiency. If this coherency is true market participants may adjust their risk premiums in order to compensate for information asymmetries and uncertainty stemming from the inability to efficiently monitor the credit institution s portfolio risks. In a next step, we introduce our measures for competition (H-statistic (1) and (2)) and concentration simultaneously and follow BAUMOL (1982) who argues that even highly concentrated markets might be contestable ( theory of contestable markets ). As presented in specifications (4) and (5), our proxies for competition in the banking market enter both regressions significantly negative at the 0.05 level while our measure for concentration turns out to be positively significant at the one percent level. This finding reveals that banks operating in concentrated banking markets which are not contestable may pay higher risk premiums on their debt capital when refusing to disclose comprehensive information in their annual reports. If, however, concentrated banking markets are contestable and banks decide to increase the level of detail of information on their risk-taking behavior, creditors may reward the lower level of information asymmetries in adjusting required rates of return. Since competition may either stem from the banking industry itself or from other capital market participants, we additionally control for credit growth, i.e. the amount of credit provided by the banking sector relative to GDP, and for the degree of disintermediation which proxies the development of and the competitive pressure induced by national capital markets and is measured by the country s turnover ratio of stocks traded. To begin with credit growth, the variable enters regression (6) significantly positive at the one percent level suggesting that banks refraining from increased disclosure may have to pay higher interest rates on their debt capital during a period of excessive lending. This finding corresponds with DELL ARICCIA 17

18 and MARQUEZ (2006) who postulate that an expansion in credit increases the sensitivity of bank s profits to aggregated shocks and hence may reduce its ability to service debt adequately. The estimate of disintermediation (specification (7)), in turn, is positive and statistically significant at the one percent level revealing that banks shifting from traditional intermediation to more complex structured products and hence depending on a constant access to capital markets for the purpose of refinancing face higher cost of debt, especially at low levels of disclosure. Not only reiterates this result the positive sign of the estimate of the funding structure in our baseline regression, but it is also in line with BIKKER and SPIERDIJK (2008) providing evidence that capital market competition spurs banks to focus on more sophisticated and thus riskier products with less price competition. Finally, we control for the ownership structure of banks operating in national European banking markets and distinguish between the share of government-owned and foreignowned institutions respectively as published in BARTH ET AL. (2008a). Referring to specification (8), governmental ownership enters the regression with a positive sign, but the estimate remains insignificant suggesting that disclosure is not particularly more important in markets characterized by a high concentration of state bank ownership with regard to the bank s cost of debt. We carefully suggest that investors may consider capital provided to banks owned by the state as safe investment and hence pay less attention to their disclosure policy. In contrast, foreign ownership enters regression (9) significantly negative at the ten percent level indicating that costs of debt capital are inversely linked with the presence of foreign banks. Since a large share of foreign-owned banks may be the result of the national banking market s contestability, this result reconfirms our previous findings on competition presented in specifications (1) to (2) and (4) to (5). 18

19 4.3. Disclosure and the regulatory environment across the European Union Over the last years, policy makers such as the International Accounting Standards Board (IASB) or the Basel Committee on Banking Supervision have placed emphasis on enhanced disclosure and market-oriented reporting (e.g. IFRS 7, Basel Framework for the International Convergence of Capital Measurement and Capital Standards (Basel II)), thereby changing the conditions under which banks operate. Hence, we examine the relationship between a bank s level of disclosure and its cost of debt while controlling for the regulatory environment and accounting regimes in the member states of the European Union and present major results in Table 9. Again, our main finding of a negative relationship between disclosure and a bank s cost of debt is reconfirmed among all regression specifications with coefficient values ranging from to basis points. Turning to the regulatory environment in specification (1), we initially introduce a dummy variable which takes on the value of one for all the years t in our sample following the incorporation of Basel II into national law in the bank s home country j and 0 otherwise. The measure enters the regression significantly positive at the one percent level providing evidence that Basel II capital regulations have increased the bank s cost of debt capital as compared to banking regulation under Basel I. However, banks providing detailed information on their risk-taking behavior and risk-bearing ability through higher levels of disclosure may counteract rising cost of regulation. In order to shed a brighter light on the nexus between disclosure, national regulation and supervision and the cost of bank s debt beyond the implementation of Basel II, we build six indices assessing the scope and the development of financial regulation at the countrylevel and over time. All indices, namely entry restrictions, activity restrictions, diversification, capital regulation, official supervisory power and private monitoring, are built from combined data retrieved from three World Bank surveys on bank regulation and supervision initially conducted in 1998/2000 and updated in 2003 and 2007 (BARTH ET AL., 2008a). These 19

20 surveys aimed at collecting detailed and comprehensive information on the regulation and supervision of credit institutions around the globe as well as data on bank structures and deposit insurance schemes. 5 The construction of each index is explained in detail in Table 1 within the Appendix. To begin with entry restrictions, the index rates a country s entry into banking requirements and enters regression specification (2) significantly positive at the ten percent level suggesting that the more legal submissions are required to obtain a banking license, the higher the interest rate banks have to pay on their debt capital. In line with CLAESSENS and LAEVEN (2004), who argue that excessive entry restrictions may primarily act as an impediment to competition in local banking markets, this result reiterates our finding of an inverse relationship between competition and cost of debt as presented in Section 4.2. A similar picture is presented when turning to activity restrictions which assesses the extent to which banks overall activities are restricted. As shown in specification (3), the estimate is positive and statistically significant at the 0.01 level indicating that restricting banks from engaging in security, insurance or real estate activities is strongly associated with higher cost of debt capital. This finding corresponds with BARTH ET AL. (2008b) who argue that activity restrictions may reduce competition in the banking sector and limit economies of scope, both resulting in inefficient allocation of capital. If this is true, capital market investors may adjust their return requirements with regard to the inability of banks to expand their business base and to increase operational efficiency. Rating legal regulations that allow banks to diversify income streams and loan portfolios domestically and internationally, diversification enters specification (4) significantly positive at the weak ten percent level revealing that the existence of explicit guidelines for asset diversification allowing banks to make loans abroad increases debt funding costs. This result 5 BARTH ET AL. (2001) provide comprehensive and in-depth information on the structure of the World Bank survey. 20

21 may be surprising, but is in accordance with previous (empirical) findings in the banking literature suggesting that investors have less information about more diversified (banking) corporations since they are more opaque. In this context, LAEVEN and LEVINE (2007) propose that the market value of financial conglomerates engaged in multiple activities is lower than the sum of the disaggregated values of the single entities and DEYOUNG and RICE (2004) provide evidence that well-managed banks expand more cautiously into non-interest revenue generating activities. Introducing capital regulation, the measure evaluates the magnitude of regulatory capital and its relationship to the bank s overall asset portfolio as required by national banking regulation. Presented in regression (5), this estimate is positive and statistically significant at the one percent level indicating that greater capital stringency coincides with higher funding costs of the bank s debt capital. This result is consistent with theoretical and empirical findings proposing that although higher capital requirements may internalize inefficiencies resulting from a gambling for resurrection strategy, they may harm the bank s charter value and, in turn, increase the bank s risk-appetite in order to achieve target rates of return (e.g. on equity capital; GALE, 2010; HELLMANN ET AL., 2000). Moreover, related literature stresses significant costs incurred with the implementation of a regime with higher capital requirements which may increase banks marginal costs (COSIMANO and HAKURA, 2011; ANGELINI ET AL., 2011; BIS, 2010). Given the extent of information asymmetries prevailing in the banking industry, both public and private institutions may help verify the information disclosed in banks annual reports, thereby reducing asymmetric information between market participants. Accordingly, we control for the strength of the public supervisory authority (official supervisory power) and the extent to which private monitoring and supervision is pursued at the national level (private monitoring). Official supervisory power enters specification (6) significantly positive at the five percent level suggesting that greater power of supervisory authorities increases the bank s 21

22 cost of debt, which may be explained by rising cost of regulation being associated with prudential public supervision that have to be covered by the bank s profit. In contrast, the estimate of private monitoring (regression (7)) is negative and statistically significant at the 0.1 level indicating that, in line with disclosure at the bank-level, policies that promote private monitoring at the country-level help reduce the bank s interest rate paid on its debt capital. Against the background of these results, we carefully suggest that excessive public supervision may be seen as an involvement by the government that hinders efficient capital allocation by banks, whereas authorities adopting private monitoring strategies within national regulatory frameworks may enhance bank operations. Since these six individual modules combine to form a national regulatory and supervisory strategy in the member states of the European Union, we control for the entire package of regulations and include the six indices in the same regression presented in specification (8). Not only is our baseline finding of an inverse relationship between disclosure and the bank s cost of debt reconfirmed, but five out of the six indices enter the regression statistical significantly and with the expected signs. The only exception from this is our measure of diversification which turns out to be insignificant. We finally consider national accounting regimes since bank regulators in the European Union mandate fair value accounting for financial institutions for the purpose of assessing a bank s regulatory capital. Moreover, prior to the European Union s agreement to apply International Accounting Standards (IAS) and International Financial Reporting Standards (IFRS) for listed companies from January 1, 2005 onwards, some banks in our sample have prepared their accounts in accordance with domestic GAAP. In order to mitigate any possible bias resulting from different objectives in accounting standards, we introduce a dummy variable which takes on the value of one if financial accounts are prepared using IAS/IFRS and 0 otherwise. 22

23 Reported in specification (9), the estimate enters our regression significantly negative at the 0.05 level suggesting that banks disclosing detailed information on their risk-taking behavior while reporting under IAS/IFRS pay lower interest rates on debt capital. This result complements studies by LI (2010) and LEUZ and VERRECCHIA (2000) providing empirical evidence that the (mandatory) adoption of IFRS lowers firms cost of equity capital and by HORTON ET AL. (2010) revealing that consensus forecast errors decrease for firms mandatorily adopting IFRS relative to forecast errors for other firms due to information and comparability effects. With regard to our sample of credit institutions located in the European Union we suggest that international accounting standards may reduce information asymmetries between capital market participants when replacing domestic GAAP of lower quality that are likely to have been applied especially in Eastern European transition economies. Moreover, common accounting standards may simplify the comparability of banks headquartered in different countries within the European Union thereby enhancing competition in the Single Market. 5. Conclusion Employing panel data on 181 quoted commercial, savings and cooperative banks as well as bank holding companies across the 27 member states of the European Union for the period from 1997 to 2009, this study provides empirical evidence that more disclosure of information in published accounts reduces the bank s cost of debt capital. Our baseline finding is reinforced when controlling for endogeneity concerns and accounting for the unbalanced structure of the underlying dataset. Moreover, results from further sensitivity analyses reveal that the nexus between banks cost of debt and the amount of information disclosed varies among national banking market structures and characteristics as well as different regulatory and supervisory strategies. Against the background of our empirical results, we deduce the following policy implications. To begin with, disclosure may be an appropriate way to reduce information asym- 23

24 metries between bank managers and investors, who in turn may reward lower levels of uncertainty by adjusting required rates of return. Consequently, the disclosure of bank-level information may provide an opportunity for bank managers to cut debt capital funding costs which may provide new entrepreneurial opportunities in turn. Moreover, empirical results reveal that there is a great reliance on disclosure if banking markets are either highly concentrated or not competitive. Since both characteristics of national banking market structures may coincide with high levels of information asymmetries, we propose that policy makers ensure the contestability of local banking markets within the European Union. Finally, this study provides evidence that enhanced disclosure of bank-level information may complement regulatory measures in terms of capital requirements and risk management. We suggest that supervisors strengthen the role of disclosure within the regulatory framework in order to make use of the disciplining forces of the market and to enable credit institutions to (partially) offset cost of regulation. Nevertheless, although the index employed in this study is an accurate measure of disclosure at the micro-level, it neither assesses the content of information disclosed nor account for any information published by third parties. Moreover, disclosing information imposes direct and indirect costs on banks that need to be justified and offset by accompanying benefits. Dealing with these subjects for discussion is left for future research. 24