The Voluntary Adoption of International Accounting Standards and Loan Contracting around the World

Transcription

1 The Voluntary Adoption of International Accounting Standards and Loan Contracting around the World By Jeong-Bon Kim, Judy S. L. Tsui and Cheong H. Yi Current Version March 2007 Kim is at Concordia University and The Hong Kong Polytechnic University. Both Tsui and Yi are at The Hong Kong Polytechnic University. We thank Jong-Hag Choi, Annie Qiu, Byron Song, Haina Shi, Yoon S. Zang, and participants of Ph.D. and DBA Research Seminars at The Hong Kong Polytechnic University for their useful comments on the earlier version of the paper. We acknowledge financial support for this research obtained through the 2006 Competitive Earmarked Research Grant (CERG) of the Hong Kong SAR Government. Correspondence: Jeong-Bon Kim, John Molson School of Business, Concordia University, 1455 de Maisonneuve Blvd. West, Montreal, PQ, H3G 1M8, Canada ( Phone: Ext. 8933).

2 The Voluntary Adoption of International Accounting Standards and Loan Contracting around the World Abstract Using a sample of non-us borrowers from 30 countries over the period, this paper investigates the effect of the voluntary adoption of International Accounting Standards (IAS) on the price and non-price terms of bank loan contracts and the mix of domestic vs. foreign lenders who participate in loan deals. Our results reveal the following. First, we find that lenders charge significantly lower loan rates to IAS adopters than they do to non-adopters. The rate difference between the two groups amounts to nearly 25 basis points. Second, we find that lenders impose more favorable or less restrictive non-price terms on IAS adopters than they do on non-adopters. In particular, our results show that IAS adopters have a larger amount of loan facility, and are less likely to have restrictive covenants in their loan contracts, compared with nonadopters. Finally, we find that voluntary IAS adoption by borrowers attracts more suppliers of loans, and this increase in the number of lenders is due to IAS adopters attracting more foreign lenders from the international loan market. In conclusion, our results, taken as a whole, support the view that voluntary IAS adoption improves the contracting efficiency in the market for private debts such as bank loans by enabling lenders to assess borrowers credit quality more accurately and improving lenders familiarity with borrowers in the international loan market. Keywords: International accounting standards, loan spreads, debt covenants, lender mix. 1

3 The Voluntary Adoption of International Accounting Standards and Loan Contracting around the World 1. Introduction In 1973, representatives of the professional accounting bodies from major developed economies 1 reached an agreement to establish the International Accounting Standards Committee (IASC) with no statutory or regulatory mandate given by political jurisdictions. Since then, the IASC and its successor, the International Accounting Standards Board (IASB), have issued a total of 41 International Accounting Standards (IAS) and a total of 7 International Financial Reporting Standards (IFRS), respectively, in an effort to harmonize financial reporting standards around the world. Since the IASC was restructured into the IASB in 2001, this private sector-based voluntary effort for developing a common language of business has made significant progress as manifested in the 2002 cooperative agreement between the IASB and the Financial Accounting Standards Board (FASB) to work together to develop high quality, fully compatible financial reporting standards that could be used for domestic and cross-border reporting (Schipper 2005, p. 102). IAS and IFRS (hereafter IAS for convenience) have now emerged as the most popular financial reporting model in the world (Barth et al. 2005; Covrig et al. 2007) as thousands of companies around the world voluntarily adopted the standards. Further, IAS has increasingly received wide support from securities regulators across different political jurisdictions as culminated in the European Union (EU) s decision to mandate all companies listed on organized securities exchanges in EU countries to prepare their financial statements in accordance with IAS starting from January 1, Countries which participated in the agreement are Australia, Canada, France, Germany, Japan, Mexico, Netherlands, the UK/Ireland and the US. 2

4 Several studies have examined economic consequences of IAS adoption, and provide evidence suggesting that financial disclosures under IAS are, in general, of higher quality than those under domestic accounting standards. In particular, these studies find that voluntary IAS adoption leads to less accounting flexibility and smaller analysts forecast errors (Ashbaugh and Pincus 2001), higher market liquidity and trading volume (Leuz and Verrecchia 2000), higher earnings response coefficients (Bartov et al. 2004), and better accounting quality in terms of less aggressive earnings management, more timely recognition of economic losses and greater value relevance of accounting amounts (Barth et al. 2005), a convergence of accounting amounts under IAS with those under US GAAP (Barth et al. 2006), and greater investment flows by attracting more foreign mutual funds (Covrig et al. 2007). A major argument in favor of accounting standards harmonization via IAS is that IAS adoption enables firms to get easier access to external financing, in particular, by facilitating external financing from the global equity and debt markets and cross-border investment flows. Surprisingly, however, previous research has paid little attention to examining the effect of IAS adoption on the cost of equity or debt financing. Given that private debts such as bank loans are the most important source of external financing to most firms around the world, this paper aims to provide systematic evidence on the effect of voluntary IAS adoption on the price and non-price terms of bank loan contracts as well as on the lender mix (the composition of foreign vs. domestic lenders). To do so, we construct a sample of non-us borrowers from 30 countries who engaged in loan deals during the nine-year period from 1997 to 2005, and then compare various features of loan contracts between IAS adopters and non-adopters to address the following questions. 3

5 First, we investigate whether, and how, voluntary IAS adoptions by borrowers lead lenders to charge lower loan rates. Voluntary IAS adoptions, which give rise to greater and higher-quality disclosures (e.g., Ashbaugh and Pincus 2001), provide researchers with an ideal setting for evaluating economic consequences of a borrower s commitment to better disclosure. In this paper, we argue that the voluntary IAS adoption reduces ex ante information uncertainty faced by lenders and/or information asymmetries between borrowers and lenders. As a result, lenders are better able to assess borrowers credit quality, and thus, to save ex post costs associated with monitoring borrowers credit quality and re-negotiating contractual terms when credit quality changes. 2 We predict and find that, after controlling for borrower-specific credit risk, loan-specific characteristics, and country-level factors, lenders charge a lower loan rate to borrowers who voluntarily adopt IAS (hereafter IAS-adopters) than they do to non-adopters. The rate difference between the two groups amounts to nearly 25 basis points, and is not only statistically significant but also economically significant as well. We also find that the loan rate-reducing effect of voluntary IAS adoption holds, irrespective of a country s property rights, creditor rights, credit market development, and economic development. Second, we investigate whether voluntary IAS adoptions have an impact on the nonprice terms of loan contracts: loan size, maturity, securitization, and restrictive covenants. Commercial banks and other lenders use loan size and maturity to ration credit in equilibrium (Chava et al. 2005). Studying the effect of IAS adoption on loan size and maturity is interesting and important because it provides useful insights into how the 2 An important feature of private debts such as bank loans that distinguishes from public debts such as public bonds is the ability of lenders to renegotiate the loan terms when credit quality changes after the loans are granted. For public bonds, the re-contracting costs are, in general, prohibitively high due to diverse, widespread, arms-length bondholders, compared with those for private debts. 4

6 voluntary IAS adoption plays a role in credit rationing by lenders. Moreover, lenders often require loans to be secured by collateral and/or impose protective covenants in an effort to reduce the agency cost of debt (Smith and Warner 1979; Bradley and Roberts 2004). In this paper, we also examine whether voluntary IAS adoptions influence the presence of loan securitization and restrictive covenants in loan contracts to provide evidence on the effect of IAS adoption on the overall design of loan contacts. To the extent that higherquality disclosures via IAS adoption alleviate information asymmetries between lenders and borrowers and facilitate more efficient monitoring, we expect that lenders impose more favorable or less restrictive non-price terms on borrowers who use IAS than they do on borrowers who use local accounting standards. Overall, our results show that IAS adopters enjoy more favorable non-price terms than non-adopters. In particular, we find that firms that apply IAS, on average, have a larger amount of loan facility, and are less likely to have restrictive covenants in their loan contracts, compared with firms using local accounting standards. With respect to loan maturity and the likelihood of loans being secured by collateral, however, we find no significant difference between IAS users and non-ias users. Finally, we investigate whether voluntary IAS adoption by borrowers leads to an increase in the number of lenders and a change in the lender mix, i.e., the composition of domestic vs. foreign lenders who participated in loan deals. To the extent that voluntary IAS adoption mitigates information problems faced by lenders participating in syndicate loans, and enhances lenders familiarity with borrowers accounting standards in the international loan market, one can predict that the voluntary IAS adoption increases the number of participant lenders and attracts more foreign lenders. Consistent with the 5

7 prediction, we find that the voluntary IAS adoption attracts more suppliers of loans, and this increase in the number of lenders is due to IAS adopters attracting more foreign lenders from the international loan market. In summary, our results support the view that voluntary IAS adoption improves the contracting efficiency in the market for private debts such as bank loans by enabling lenders to assess borrowers credit quality more accurately and improving lenders familiarity with accounting standards adopted by borrowers in the international loan market. Our study adds to the existing literature in the following ways. First, this paper is, to the best of our knowledge, the first study that investigates the impact of IAS adoption on the price and non-price terms of loan contracts and the lender mix. We provide direct evidence that voluntary IAS adoption leads lenders to charge lower loan rates, increase loan size, impose less restrictive covenants, and attracts more foreign lenders in each loan deal. Second, our study contributes to the loan contracting literature as well. Our finding is consistent with the notion that the commitment to higher quality disclosures via IAS adoption mitigates ex ante information risk faced by lenders and/or information asymmetries between lenders and borrowers, and thus lowers loan rates. Previous studies in the loan pricing literature examine a variety of borrower-specific factors determining various features of private debt contracting (e.g., Strahan 1999; Esty and Megginson 2003). However, no previous research has investigated how a commitment to a better reporting strategy such as voluntary IAS adoption improves the efficiency of private debt contracting. Finally, recent studies by Bharath et al. (2006) and Kim et al. (2006) provide evidence that banks take into account the quality of financial reporting, proxied by accrual quality and audit quality, respectively, when assessing borrowers credit risk. However, the focus of 6

8 these studies is on loan contracts in the US where the quality of financial reporting is considered the highest, and thus a voluntary commitment to a better reporting strategy is likely to be of second-order importance. Thus, international (non-us) evidence reported in this study sheds right on the role of greater and higher-quality disclosures in private debt contracting under financial reporting environments that are significantly different from the US. The remainder of the paper is structured as follows. In section 2, we develop research hypotheses. Section 3, we specify an empirical model for hypothesis testing. In Section 4, we describe our sample and data sources, present descriptive statistics on major research variables, and conduct univariate tests. In Section 5, we present the results of various multivariate tests. The final section concludes the paper. 2. Hypothesis Development 2.1. The effect of IAS adoption on bank loan contracting Economic theory suggests that higher quality accounting information and disclosure effectively reduce information asymmetries between corporate insiders and outsiders and thereby lowers the cost of capital. A firm s decision to voluntarily switch from local GAAP to IAS is an important strategic commitment that typically causes an increase in the quantity and quality of accounting disclosures in most financial reporting regimes (Ashbaugh and Pincus 2001; Covrig et al. 2007). This commitment is costly, and thus credible, because it is difficult for IAS adopters to reverse the decision once made, and IAS adoption requires nontrivial efforts and resources on the part of preparers of financial statements and their auditors. Higher quality disclosures via IAS alleviate the degree of 7

9 uncertainty faced by lenders concerning borrowers ability to pay the interests and principal of bank loans. This reduction in ex ante information risk faced by lenders leads to lowering the cost of external financing (e.g. Diamond and Verrecchia 1991; Baiman and Verrecchia 1996). From an ex post standpoint, higher quality disclosures via IAS reduce costs associated with monitoring borrowers performance or credit quality and renegotiating contractual terms subsequent to credit quality changes. It is thus likely that voluntary IAS adoption enables lenders to charge a lower loan rate to borrowers in equilibrium. Recently, Leuz and Verrecchia (2005: LV) provide another reason why high quality information reduces the cost of external financing. They analyze the role of information in aligning the interests of firms and outside capital suppliers with respect to capital investment decisions, and establish an inverse relation between the quality of performance reports to outside capital suppliers and a firm s cost of external financing. Their analysis shows that high-quality reports improve the coordination between firms and capital suppliers with respect to capital investment decisions. On the other hand, poor-quality reports lead to a misaligned capital investment due to the impaired coordination. Anticipating this, rational capital suppliers demand a higher risk premium to firms with poor-quality reports. The LV theory suggests that higher quality disclosures via IAS adoption give rise to a saving in the cost of coordination between borrowers and lenders, which in turn enables lenders to charge lower loan rates to IAS adopters than they do to non-adopters. Recently, Francis et al. (2005) report that firms with better accrual quality pay lower interest rates on borrowing, a finding consistent with the above view. 8

10 In sum, we predict that lenders charge lower loan rates to IAS adopters than they do to non-adopters, because voluntary IAS adoption leads to: (1) lowering ex ante information risk faced by lenders and ex post monitoring and re-contracting costs; and (2) improving the coordination between lenders and borrowers with respect to capital investment decisions. We therefore hypothesize in alternative form: H1: Loan spreads, measured by loan rates in excess of a benchmark rate, are lower for borrowers who voluntarily use IAS than those who do not, other things being equal. Bank loan contracts include not only price terms, but also non-price terms such as loan size, maturity, securitization, and restrictive covenants. Lenders use various nonprice terms (as well as price terms) when designing loan contracts in an attempt to mitigate information problems and potential conflicts between lenders and borrowers. Faced with information problems, lenders may control their risk exposure to low quality borrowers by limiting the size of loans and/or shortening the maturity of loans (Strahan 1999). To the extent that voluntary IAS adoption reduces information uncertainty or the associated information asymmetry faced by lenders, lenders are better able to assess borrowers credit quality. As a result, for IAS adopters, lenders are faced with lower ex ante information asymmetry than they are for non-adopters. One may therefore expect that lenders offer more favorable contractual terms in terms of loan size and maturity to IAS using firms than they do to firms using local GAAP. H2: Loan size is larger, and loan maturity is longer, for borrowers who voluntarily use IAS than those who do not, other things being equal. The presence of collaterals and restrictive covenants in loan contracts. Lenders may also be associated with information problems faced by lenders. Lenders are more likely to 9

11 require collaterals or the inclusion of various restrictive covenants in bank loans to borrowers with opaque information or requiring intense monitoring (e.g., Rajan and Winston 1995; Bradley and Roberts 2004). Higher quality accounting information via IAS will lower ex post costs associated with monitoring borrowers credit quality and renegotiating contractual terms in response to credit quality changes. Thus, we expect that lenders are less likely to require loan securitization and restrictive covenants for IAS users than for local GAAP users. 3 H3: The likelihood of loans being secured by collateral and restrictive covenants being imposed on borrowers is greater for those who use IAS than for those who do not, other things being equal The effect of IAS adoption on the number of lenders and the mix of domestic vs. foreign lenders We now turn our attention to the effect of voluntary IAS adoption on the number of lenders participating in each loan and the lender mix (the composition of foreign vs. domestic lenders). Dennis and Mullineaux (2000) show that fewer lenders are involved in loans to borrowers with severe information problems. Sufi (2006) also demonstrates that loans to opaque borrowers have less participant lenders. These studies suggest that credible financial reports of borrowers may mitigate adverse selection and moral hazard problems among syndicate loan participants, thereby attracting more participants in a syndicate. 3 Evidence shows that IAS adoption not only increases the quantity and quality of financial disclosures, but also reduces accounting flexibility by restricting a firm s choice of accounting measurement methods (e.g., Ashbaugh and Pincus 2001). Ashbaugh and Pincus report that this reduced accounting flexibility improves the ability of analysts to forecast future earnings more accurately. Bharath et al. (2006) provide evidence suggesting that lenders use more stringent (non-price) contractual terms for borrowers with poor reporting quality. IAS adoption may cause a decrease in the agency cost of debt to the extent that this reduced accounting flexibility via IAS adoption increases reporting quality and thus enables lenders to save ex post costs associated with loan monitoring and re-contracting. To this extent, lenders are also likely to offer more favorable non-price terms, or impose less restrictive covenants, for IAS adopters than for non-adopters. 10

12 Based on the above evidence, we expect that more lenders are involved in loans to firms that use IAS. We also expect that voluntary IAS adoption attracts more foreign lenders into a syndicate by increasing lenders familiarity with a borrower. On one hand, IAS-based reporting makes it relatively easier for borrowers to explain and communicate their financial results and credit quality to foreign lenders in a more user-friendly way. On the other hand, IAS-based reporting makes it less costly for foreign lenders to assess borrowers credit risk prior to loan origination and to monitor credit quality and renegotiate the contractual terms subsequent to credit quality changes. The home-bias literature in international finance suggests that foreign investors are faced with higher information costs than domestic investors when making portfolio decisions, and thus prefer to invest in firms they are familiar with (Chan et al. 2005; Kang and Stulz 1997; Dalhquist and Robertsson 2001; Covrig et al. 2006; Kim and Yi 2005). These studies provide evidence suggesting that an increase in a firm s exposure to foreign investors (for example, by making them more familiar with the firm) expands investor base by attracting more foreign investors. In the international loan market one may therefore expect that an increase in a borrower s visibility or lenders familiarity via IAS adoption draws more foreign lenders by alleviating their perceived uncertainty about the borrower. In sum, we predict that the enhanced credibility of financial reports and the improved familiarity via IAS increase the number of participant lenders and attract more foreign lenders into a syndicate. To provide empirical evidence on this unexplored issue, we test the following hypothesis in alternative form: 11

13 H4: The number of lenders and the percentage of foreign lenders who participate in each loan are greater for borrowers who voluntarily use IAS than those who do not, other things being equal. 3. Empirical Model To test our hypotheses, H1 to H4, we specify the following regression model linking various features of loan contracts, one by one, with our test variable and control variables: Loan Feature t = α 0 + α 1.DIAS t-1 + α 2. Borrower-specific Controls t-1 + α 3. Loan-specific Controls t + α 4.Country-level Controls (1) + (Industry Dummies) + (Year Dummies) + error term where the dependent variable, Loan Feature, denotes one of the proxies for price and nonprice terms of loan contracts, the number of lenders, or the mix of foreign vs. domestic lenders, and empirical definitions of all variables are summarized in Appendix I. To test H1, we estimate Eq. (1) using the price term, Spread, as the dependent variable. The Spread variable is measured by the drawn all-in spread in basis points. This all-in spread represents the interest rate charged by lenders (plus the annual fee and the upfront/maturity fee) over the benchmark rate, i.e., LIBOR. We measure the cost of loan using a spread over LIBOR because most loans in the international loan market are priced in terms of the floating rate in excess of LIBOR. Commercial banks and other lenders typically assess the risk of a loan based upon the information on the business nature and performance of borrowing firms, and then set a markup over a prevailing benchmark rate such as LIBOR to compensate for the credit risk. The Spread variable thus reflects lenders perceived level of risk on a loan facility provided to a specific borrower. 12

14 To test H2, we use, as the dependent variable, two non-price terms of loan contracts, i.e., the size and maturity of loan facility, denoted by LoanAMT and Maturity, respectively. The LoanAMT variable is measured by the natural log of the amount of each loan facility granted to a borrower. The Maturity variable is measured by the natural log of the loan maturity period which is defined as the difference in months between the loan origination date and the maturity date. To test H3, we first estimate Eq. (1) using, as the dependent variable, three indicator variables, i.e., the probabilities of a loan being secured by collateral, financial covenants being imposed, and general (non-financial) covenants being imposed, denoted by DSecured, DFinCov and DGenCov, respectively. These indicator variables take the value of 1 for secured loans, loans with at least one financial covenant included, and loans with at least one general covenant included, respectively, and 0 otherwise. When one of these indicator variables is used as the dependent variable, Eq. (1) is estimated using the probit regression procedure. In addition, we also construct a covenant index (CovIndex) as explained below, and then estimate Eq. (1) using CovIndex as the dependent variable. For our international sample, loan covenants included in the loan contracts are classified into three broad categories: (1) the requirement of loan securitization by collateral; (2) financial covenants that are typically linked to accounting numbers; (3) general covenants which include all other non-financial covenants such as restrictions on prepayment, 4 dividend payment, and voting rights. To obtain a composite measure of the strength of various covenants included in the loan contract, we construct the covenant index by assigning the value of 1 for a secured loan (DSecured = 1), for a loan facility with financial covenants 4 The prepayment restriction includes asset sweep, excess cash flow sweep, debt issue sweep, equity issue sweep, and insurance proceeds. 13

15 (DFinCov = 1) and for a loan facility with general covenants (DGenCov = 1), and then adding up the values for each loan facility to obtain our empirical measure of the covenant index. We estimate Eq. (1) by running a Poisson regression of CovIndex on DIAS and other control variables. Finally, to test H4, we estimate Eq. (1) using the number of lenders who participated in each loan facility, denoted by NLender, and the composition of foreign vs. domestic lenders, which is measured by the ratio of foreign lenders to NLender and is denoted by %Foreign. To determine whether a lender is foreign or domestic, we check manually whether commercial banks and other financial institutions who participated in each loan facility are headquartered in the same country where borrowers are headquartered. We identify the nationality of the headquarter office of each bank participating in each loan facility using The Bankers Almanac Our test variable, DIAS is a dummy variable which equals 1 if the borrower voluntarily adopted IAS in fiscal year t - 1 (i.e., a year immediately before loans are made) during the period when loans are made in year t during our sample period, ; and 0 otherwise. Recall that the EU mandated all listed firms to adopt IAS starting from January 1, 2005, while some of these firms voluntarily adopted IAS prior to As the lagged DIAS is used in Eq. (1), we effectively link various measures of Loan Feature in 2005 to DIAS in 2004 in our regression. 5, 6 In Eq. (1), the coefficient on DIAS 5 Since the IAS adoption dummy, i.e., DIAS, (as well as all borrower-specific, financial statements variables) is measured in year t 1 and the dependent variable, i.e., Loan Feature, is measured in year t, there is no two-way causation between DIAS (our test variable) and Loan Feature (our dependent variables). This approach mitigates a concern over reverse causality in Eq. (1) with respect to the Loan Feature-DIAS relation. 6 To correct for a potential self-selectivity problem, we also use the Heckman-type, two-stage treatment effects model. In the first stage, we run a probit model that links a firm s IAS adoption to explanatory variables and then obtain the inverse Mills ratios. Following Barth et al. (2005), we include firm size, leverage, cash flows, sales growth, percentage change in common stock and percentage change in total debt in the probit IAS-adoption model. We then estimate the probit model using the maximum likelihood 14

16 captures the difference in the value of each dependent variable representing Loan Feature between IAS adopters and non-adopters. When Loan Feature is one of Spread, DSecured, DFinCov, DGenCov, or CovIndex, a negative coefficient on DIAS (i.e., α 1 < 0) is consistent with H1 and H3, while when Loan Feature is either LoanAMT or Maturity, a positive coefficient on DIAS is consistent with H2. When Loan Feature is NLender or %Foreign, a positive coefficient on DIAS is consistent with H4. To isolate the effect of voluntary IAS adoption on Loan Feature from the effect of borrower-specific characteristics, we include four borrower-specific control variables, ROA, Size, MB, and Leverage, that are known to affect borrowers credit quality and thus the price and non-price terms of loan contracts. In addition to these four variable, we also consider an additional borrower-specific variable, namely asset maturity (ASM) when Eq. (1) is estimated using loan maturity (Maturity) as the dependent variable. All borrowerspecific variables are measured in a year immediately before loan deals are made. Previous research on bank loan contracts shows that several loan-level characteristics are related to the loan rate charged by lenders (e.g. Strahan 1999; Dennis et al. 2000; Bharath et al. 2006). To control for potential confounding effects of these loan characteristics on our results, we include five loan-specific variables, that is LoanAMT, Maturity, NLender, DForCurr, and DPPricing. 7 Here, LoanAMT and NLender are as defined earlier. DForCurr is a dummy variable that equals 1 for a loan facility quoted in foreign currency and 0 otherwise. DPPricing is a dummy variable that equals 1 for a loan procedure and obtain the inverse Mills ratio. In the second stage, we include in Eq. (1) the inverse Mills ratio as an additional control variable to correct for potential self-selection biases. Though not tabulated, the results from the two-stage treatment effects model are qualitatively identical with the results reported in the paper. 7 As will be further explained later on, when one of these loan-specific variables (e.g., LoanAMT) is used as the dependent variable, the same variable is, of course, not used as an independent variable. 15

17 facility with performance pricing options and 0 otherwise. In addition to these loanspecific variables, we also consider an additional loan-specific variables, namely the term loan indicator (TLoan) when Eq. (1) is estimated using loan maturity (Maturity) as the dependent variable. Since we do not have a clear theory predicting the directional effect of these variables on the price and non-price terms of loan contracts, we do not predict the signs of the coefficients on these loan-specific variables. All loan-specific variables are measured in the same year when loan deals are made. Previous research suggests that a country s protections of property rights and creditor rights and a country s credit market development influence bank loan contracting. For example, Bae and Goyal (2003) examine the effect of various institutional variables on loan spreads and find that a country s property rights protection is the most important institutional variable determining loan spread. 8 Esty and Megginson (2003) find that a country s creditor rights protection is an important factor determining the size and structure of loan syndicates in the syndicated project finance loans. In estimating Eq. (1), we consider four country-level variables, that is PRights, CRights, CMktDev, and LGDP, representing the levels of a country s property rights protection, creditor rights protection, credit market development, and economic development, respectively. 9 Both property rights and creditor rights protections are measured using the property rights index and the creditor rights index, respectively, developed by La Porta et al. (1998). The level of a country s credit market development in year t is measured by the amount of credits 8 In their cross-country regressions of loan spreads on country-level determinants of loan spreads, Bae and Goyal (2003) consider several country-level variables, including property rights, creditor rights, language, religion, and legal origin. Overall, they find property rights are the most important determinant of the loan rates across different regression specifications. 9 We also consider other institutional variables considered in Bae and Goyal (2003), but find that they are, overall, insignificant in our regressions. 16

18 supplied by financial intermediaries to the private sector in year t deflated by a country s GDP in year t. 10 LGDP denotes the natural log of a country GDP per capita in year t. Finally, we include Industry Dummies and Year Dummies to control for potential differences in various proxies for the Loan Feature variable across industries and over years. 4. Sample and Data 4.1. Sample and data sources The initial list of our sample consists of all firms that are included in the Worldscope database and the Loan Pricing Company s Dealscan database during the sample period, The data on a firm s IAS adoption and all borrower-specific variables for the period are obtained from Worldscope. The Dealscan database is an online database which contains a variety of historical bank loan data and other financial arrangements. 11 The database includes the loan data starting from 1986, and expands its coverage over time, in particular, after We select 1997 as the starting year of our sample period because there are few IAS-adopters that are included in the Dealscan database prior to Our sample period ends in 2005 because loan-related data are available to us only up to year 2005 and to exclude all EU firms that are mandated to adopt IAS starting from January 1, CMktDev and LGDP are measured using the data obtained from the International Monetary Fund (IMF). 11 Other papers using the LPC Dealscan database include Strahan (1999), Bae and Goyal (2003), Bharath et al. (2006), Asquith et al. (2005), Ivashina et al. (2005), and Kim et al. (2006). 12 Note that all EU listed firms are mandated to prepare their financial statements in accordance with IAS starting in January As shown in Eq. (1), empirical measures of our dependent variables (Loan Feature) in 2005 are linked to voluntary IAS adoption (DIAS) in As such, all EU firms that are mandated to adopt IAS in 2005 are effectively excluded from our sample. 17

19 The loan data in the Dealscan database are compiled for each deal and facility. Each deal, i.e. a loan contract between a borrower and bank(s) at a specific date, may have only one facility or have a package of several facilities with different price and non-price terms. 13 We consider each facility as a separate observation in our sample because many loan characteristics and loan spreads vary across facilities, and require that all loan facilities in our sample are senior debts. 14 We then match the loans with borrowers financial statement data in Worldscope, using the ticker symbol and name of each borrower. This procedure leads to a substantial reduction in the number of available loan facilities because many borrowers included in the Dealscan database are subsidiaries of public firms, private firms and government entities rather than publicly traded companies, and some public companies are not covered by Worldscope (Strahan 1999; Dichev and Skinner 2002). We require that all the relevant annual financial statements data needed to compute all borrower-specific characteristics be available in the fiscal year immediately before the loan initiation year. As shown in Panel A of Table 1, we obtain a sample of 2,425 facility-year observations from 30 countries after applying the above selection procedures. Out of 2,425, 166 observations are from borrowers that voluntarily adopted IAS. 15 As shown in column 1 of Panel A, the number of facility-year observations in the total sample of borrowers with both IAS adopters and non-adopters is widely distributed across countries, ranging 13 For instance, a deal may comprise a line of credit facility and a term loan with longer maturity. 14 This selection criterion is similar to those used by Bharath et al. (2006) and Kim et al. (2006). 15 The percentage of IAS adopters in our sample, which is about 6.8%, is greater than that in the sample of Covrig et al. (2007). They use a total sample of 24,592 firm-years with both IAS adopters and non-adopters in the period from 29 countries to examine the effect of IAS adoption on foreign mutual fund holdings in the global equity market. In their total sample, the percentage of IAS adopters is about 5% (See their Table 1). It should be noted that that their focus is on the global equity market while our focus is on the international market for private debts, primarily loans by commercial banks and other institutional lenders such as investment banks and insurance companies. 18

20 from the lowest of 1 for Austria to the highest of 477 for the United Kingdom. Column 2 of Panel A shows the distribution of facility-years using IAS across countries which ranges from 0 for 16 countries to the largest of 65 for Germany. 16 Columns 3 and 4 of Panel A show the property rights index and the creditor rights index by country, respectively, that are developed by La Porta et al. (1998). Similar to Morck et al. (2000) and Bae and Goyal (2003), a country s property rights index is measured by adding three indices from La Porta et al. (1998) representing the extent of government corruption, the risk of expropriation by the government, and the risk of the government repudiating contracts. Each of the three indices ranges from 0 to 10, and thus the property rights index ranges from 0 to 30 with high values indicating more respect for private property rights. As shown in column 4, the property rights index ranges from the lowest of 12.9 for Philippines to the highest of 29.6 for Norway. A country s creditor rights index is measured by adding four dummy variables representing no automatic stay on assets, secured credit first, restrictions for going into reorganization, and current management does not stay in the reorganized firm. It is measured in such a way that creditor rights are better protected in a country with a higher value of the creditor rights index. As shown in column 4, the creditor rights index ranges from the lowest value of 0 for France and Philippines to the highest value of 4 for six countries including Hong Kong, Netherlands, Singapore, and the UK. 16 We have also estimated all regressions reported in the paper after excluding observations from 16 countries with no IAS adoptors. Though not report, we find that the results using this reduced sample are qualitatively similar to those reported in the paper. Note that Covrig et al. (2007) also include in their sample observations from 9 (out of 29) countries with no IAS adopters when examining the effect of IAS adoption on foreign mutual fund holdings. 19

21 Columns 5 and 6 of Panel A show the mean levels of a country s credit market development and GDP per capita in US dollars. As shown in Column 5, the average amount of credits supplied by financial intermediaries varies widely across countries ranging from 20% of GDP for Israel and Turkey and to 164% of GDP for Switzerland. As expected, Column 6 shows that average GDP per capita during our sample period, , varies widely across countries, ranging from US$508 for India to US$40,412 for Switzerland. Panel B of Table 2 reports that the yearly distribution of 166 IAS adopters and 2,259 non-ias adopters. As shown in Panel B, the number of IAS adopters, overall, increases over the years with a slight decline only in The number of non-ias adopters in our sample also increases over the years with a decline in 2001, which reflects an increasing trend in the Dealscan coverage over the years. Panel C of Table 2 presents the distribution of IAS adopters and non-adopters across 8 different industries, and reveals that both IAS adopters and non-adopters are most heavily concentrated in the manufacturing industry. [INSERT TABLE 1 ABOUT HERE!] 4.2. Descriptive statistics and univariate tests Table 2 presents descriptive statistics for all borrower-specific and loan-specific variables considered in this study, separately, for the IAS adopters and the IAS nonadopters, and performs univariate tests for the mean and median differences between the two groups. As shown in Panel A, the mean (median) drawn all-in spread (Spread) is about 52 (36) basis points for the IAS adopters while it is about 103 (75) basis points for non- 20

22 adopters. The mean and median differences are both significant at less than the 1% level, which is in line with H1. Consistent with H2, the amount of loan facility (LoanAMT) is significantly larger for IAS adopters than non-adopters. Note, however, that IAS adopters have a shorter loan maturity (Maturity) than non-adopters, which is inconsistent with H2. A comparison of DSecured, DFinCov, DGenCov, and CovIndex between the two samples reveals that IAS adopters are less likely to have their loans secured, and to have restrictive covenants, which is consistent with H3. When compared with non-adopters, IAS adopters have not only more lenders but also more foreign lenders who participate in each loan facility, a finding consistent with H4. Finally, we find that IAS adopters are more likely to have their loans quoted in foreign currency (DForCurr), and are less likely to have a term loan (TLoan), compared with non-adopters. We find, however, that there is no significant difference in the likelihood of loans with performance pricing options between the two samples. As shown in Panel B, the mean profitability (ROA) is not significantly different between IAS-adopters and non-adopters with the same median ROA of 5% for both samples. The mean borrower size (Size) is not significantly different between the two samples, though its median is significantly larger for IAS-adopters than for non-adopters at the 5% level. The growth potential, measured by the market-to-book ratio (MB) is, on average, smaller for IAS adopters, compared with non-adopters, though its median is not significantly different between the two samples. Both mean and median of the debt-to-total asset ratio (Leverage) are not significantly different between the two groups. Both mean and median asset maturity (ASM) is significantly shorter for IAS adopters than for nonadopters. 21

23 [INSERT TABLE 2 ABOUT HERE!] Table 3 reports a Pearson correlation matrix. Our test variable, DIAS, is negatively correlated with the drawn all-in spread (Spread), which is consistent with H1. Note that DIAS is positively correlated with LoanAMT, which is consistent with H2, but it is negatively correlated with Maturity, which is inconsistent with H2. We find that DIAS is negatively correlated with DSecured, DFinCov, DGenCov, and CovIndex, which is in line with H3. Consistent with H4, DIAS is positively correlated with NLender and %Foreign. Consistent with our priors, Spread is negatively correlated with ROA, Size, and LoanAMT. A significantly negative correlation of Spread with DForCurr and DPPricing suggests that borrowers with foreign currency loans and performance pricing options in their loans are likely to pay lower loan rates. The correlation between LoanAMT and NLender is This high correlation is not surprising given that large loans are often provided through a loan consortium or syndicate with multiple lenders. [INSERT TABLE 3 ABOUT HERE!] 5. Results of Multivariate Tests 5.1. Tests for the effect of IAS adoption on loan spread To test H1, we estimate Eq. (1) using Spread as the dependent variable. Table 4 presents the results of the OLS regressions in Eq. (1) using the full sample of 2,425 facility-years with both IAS adopters and non-adopters over the period. Reported t-values are computed using standard errors adjusted for heteroskedasticity and 22

24 clustering at the firm level. 17 In column 1, we estimate Eq. (1) after excluding the countrylevel control variables but including country dummies. In columns 2 and 3, we estimate Eq. (1) after including only one of the two institutional variables, i.e., PRights and CRights, while in column 4, we include both of them. As shown in columns 1 to 4, the coefficients on DIAS are significant with an expected negative sign at less than the 1% level across all cases after controlling for all other factors. These significantly negative coefficients on DIAS strongly support H1. The coefficient on DIAS captures the loan rate difference in basis points between IAS adopters and non-adopters. The results in columns 1 to 4 show that the magnitude of the DIAScoefficient ranges from about 20 basis points in column 1 to about 25 basis points in column 3. This suggests that the IAS adopters, on average, pay lower loan rates than the non-adopters by more than 20 basis points even after controlling for borrower-specific and loan-specific characteristics and country-level factors. This loan rate difference is economically significant as well. Overall, the above results suggest that voluntary IAS adoption mitigates ex ante information risk faced by lenders and ex post loan monitoring and re-contracting costs, which in turn translates into significantly lower loan rates charged to IAS adopters. With respect to the estimated coefficients on borrower-specific variables (Panel B), the following is apparent. First, the coefficients on both ROA and Size are highly significant with an expected negative sign across all cases. This is consistent with the view that lenders consider both large and high-roa borrowers as having less credit risk or better 17 We also estimated Eq. (1) using the weighted least squares (WLS) procedure with an equal weight assigned to each country to address a concern over potential problems arising from unequal distribution of samples across different countries. Though not reported for brevity, we found that the WLS results are qualitatively identical with those reported in the paper. 23

25 capacity to repay the loan, and thus charge lower loan rates to such borrowers. Second, the coefficient on MB is insignificant, albeit positive, across all cases. Borrowers with high growth potential (as reflected in high MB) may have a lower credit risk or higher credit quality because they have a greater ability to generate future cash flows, compared with borrowers with low growth potential. In such case, the coefficient on MB should be positively significant. On the other hand, borrowers with high growth potential could be viewed as having a higher risk because cash flows of high growth firms tend to be more volatile over time than those of low growth firms. These two opposing effects may cancel out each other, leading us to observe an insignificant coefficient on MB. Finally, the coefficient on Leverage is highly significant with an expected positive sign, which is consistent with evidence reported in many other studies (e.g., Bharath et al. 2006; Kim et al. 2006). High-leverage firms are likely to have higher default risks, and thus have relatively poor credit quality, compared with low-leverage firms. To compensate for this potential credit risk, banks are likely to charge a higher loan rate for high-leverage firms than for low-leverage firms. With respect to the estimated coefficients on loan-specific variables (Panel C), the following is noteworthy. The coefficient on LaonAMT is highly significant with an expected negative sign across all cases, suggesting that lenders charge lower loan rates on large loans than they do on small loans. The coefficient on Maturity is insignificant except for column 1. The coefficient on NLender is significantly negative in columns 2 and 4, suggesting that loan rates decreases as more lenders participate in a loan deal. The coefficients on DForCurr are significantly negative except for column 1 with its magnitude ranging from to This suggests that loan rates are significantly 24

26 lower for foreign currency loans than for local currency loans by more than 11 basis points. The coefficient on DPPricing is insignificant across all cases, indicating that for our international sample, no significant difference in loan rates exists between loans with and without performance pricing options. With respect to country-level control variables, we find that the coefficients on PRights are significant, ranging from in column 4 to in column 2. The significantly negative coefficients on PRights suggest that borrowers from countries with strong property rights pay lower loan rates than borrowers from countries with weak property rights. We find, however, that the coefficients on CRights and CMktDev are insignificant. This is consistent with Bae and Goyal (2003) who report that the extent of a country s property rights protection is the most (and the only in most cases) significant institutional variable determining loan rates among several other institutional variables they consider. In the next two subsections, when we examine the effect of IAS adoption on various non-price terms of a loan contract and the lender mix, we therefore report the results of regressions that include only PRights, but not CRights, or CMktDev, along with LGDP. In sum, consistent with our hypothesis H1, the coefficients on DIAS are significantly negative across all cases, after controlling for all borrower-specific and loanspecific characteristics and country-level factors. Our results reported in Table 4, taken together, suggest that voluntary IAS adoption enables borrowers in the international loan market to save a significant amount of borrowing cost. [INSERT TABLE 4 ABOUT HERE!] 25