Credit Risk and IFRS: The Case of Credit Default Swaps

Gauri Bhat*, Jeffrey L. Callen**, Dan Segal***

October 2011

*Olin Business School, Washington University in St. Louis, One Brookings Drive, MO 63105, USA; bhat@wustl.edu **Rotman School of Management, University of Toronto, 105 St. George St., Toronto Ontario, Canada M5S 3E6; callen@rotman.utoronto.ca *** Arison School of Business, Interdisciplinary Center Herzliya, Israel 46510; dsegal@idc.ac.il

Acknowledgements Bhat gratefully acknowledges the Center for Research in Economics and Strategy (CRES) at the Olin Business School for financial support. Callen gratefully acknowledges the Humanities and Social Sciences Research Council of Canada for financial support. We thank Nicole Jenkins, Richard Carizosa and workshop participants at the University of Waterloo, University of Missouri at Columbia, Washington University in St Louis, and participants at the 21st Annual Conference on Financial Economics and Accounting and the 2011 Utah Winter Accounting Conference for helpful comments.

Credit Risk and IFRS: The Case of Credit Default Swaps

Abstract Theory predicts that accounting information transparency affects credit spreads. Given that one of the putative benefits of International Financial Reporting Standards (IFRS) is transparency of accounting information, motivates this study to evaluate the impact of IFRS on the pricing of credit spreads in the over-the-counter Credit Default Swap (CDS) market. Using a difference in differences methodology, with a matched IFRS-U.S. sample, we find evidence that while earnings are a significant determinant of credit risk, adoption of IFRS did not change the CDS-earnings relation. Because IFRS induced transparency did not affect all firms in the same manner, we examine two dimensions of transparency. First, at the country level, we examine institutional factors that have been shown to affect the quality of accounting information. We find that institutional factors have an impact on the CDS-earnings relation but there is no change in the CDS-earnings relation as a result of IFRS adoption. Second, we examine the impact of IFRS on firms with low earnings versus firms with high earnings. Consistent with the prior literature, we find that the CDS-earnings relation is non-linear; earnings are more credit risk informative for firms with low earnings. However, we find no change in the CDS-earnings relation in the pre and post adoption periods across levels of earnings.
Key words: Credit Default Swaps, Credit Risk, IFRS

JEL Classification: M40, M41, G13, G20 Data Availability: All data are publicly available.

1. Introduction
This study evaluates the impact of International Financial Reporting Standards (IFRS) on the pricing of credit risk in the over-the-counter Credit Default Swap (CDS) market. The CDS is essentially a pure credit default instrument that provides a far less noisy measure of credit risk by comparison to corporate bonds or credit ratings. The intent of this study is to compare the credit risk information conveyed by IFRS earnings relative to local GAAP earnings for countries which adopted IFRS. Our study is guided by the theoretical CDS pricing models of Merton (1974) and Duffie and Lando (2001). The latter model provides justification for the relevance of noisy accounting information as a determinant of CDS prices. Although accounting plays no role in the former, the Merton model motivates the need to control for other determinants of CDS spreads. We focus our analysis on accounting earnings as a proxy for the firms uncertain wealth and noisy asset dynamics. To the extent that earnings are a meaningful proxy, the hybrid CDS pricing model of Duffie and Lando (2001) predicts an inverse relation between CDS spreads and earnings. Indeed, Callen, Livnat, and Segal (2009) show that U.S. GAAP earnings are an important determinant of CDS spreads. In their levels analysis, they find that a one percent increase in earnings (normalized by total assets) reduces CDS spreads by a non-trivial nine percent. We measure the informativeness of earnings with respect to credit risk by the earnings coefficient in a regression of CDS spreads on normalized earnings controlling for other potential determinants of the spread. The more negative the coefficient, the more informative the credit risk information conveyed by earnings. We use this coefficient to test the relation between financial statement variables and CDS spreads, rather than simple correlations, in order to allow for a richer, more powerful analysis.1 In their seminal study, Duffie and Lando (2001) show theoretically that CDS spreads are negatively correlated with the transparency of accounting information

Simple correlations potentially obfuscate the transition to IFRS if only because the volatility of many variables increased in the post-IFRS period consistent with the focus of IFRS on fair value accounting (Barth, Cram, and Nelson 2001).

because transparency reduces the noise about the reference firms asset structure and wealth dynamics. The adoption of IFRS provides a unique research opportunity to examine the impact of accounting information transparency on the relation between credit risk and financial statement information, particularly earnings, because the switch to IFRS for most firms was exogenously mandated by accounting regulators, thereby mitigating the potential impact of confounding endogenous events.2 In addition, a relatively large number of firms had to switch to IFRS, thereby providing a reasonable sample size for testing purposes. Earnings under IFRS are potentially more transparent than the earnings under previous local GAAP for the following reasons. First, IFRS is principles based rather than rules based. Thus, not only does IFRS discourage financial statement engineering but it encourages companies to adapt their reporting to changes in the business environment, so that financial information under IFRS better reflects economic reality. Second, the reporting requirements under IFRS are more comprehensive than those under many local GAAPs, especially for cash flows, pension obligations, leases and liabilities of uncertain timing and amount. Cash flow information, in particular, is helpful to creditors to assess whether the firm will be able to generate sufficient cash flows to service its debt.3 Third, IFRS emphasizes greater use of fair value accounting than most of the local GAAPs. Fair value information provides early warnings to investors and regulators of changes in current market expectations, especially in an environment characterized by decline in asset prices and increase in risk. These early warning signals are particularly relevant in the analysis of credit risk because under historical cost, the recognition of decline in asset prices depends on management discretion and the extent of conditional conservatism (Linsmeier 2011). Finally, one set of standards across countries promotes uniformity and comparability, and together with the increased transparency should allow for better assessment of credit risk. To quote the credit rating agency Moodys:

Our sample excludes firms that adopted IFRS voluntarily before it was mandated.

Firms in Italy and Spain were not required to include a cash flow statement in their annual reports prior to

the adoption of IFRS.

the adoption of European commission-endorsed IFRS from 2005 by nearly all publicly traded companies incorporated in the European union should improve the efficiency and transparency of Moodys analytical processes. (Moodys. Special Comment. 2004) However, any discretion accompanying a principles based approach is bound to be plagued by inconsistent interpretation and application, and potentially compromises comparability. Indeed, the International Accounting Standards Board (IASB) recognizes that allowing choices in accounting methods hinders comparability. Furthermore, any increase in accounting information transparency as a result of IFRS adoption is potentially contingent on country level institutional factors that complement accounting standards and shape financial reporting incentives. The extant prior literature documents that institutional factors are associated with the quality of accounting information (e.g. Ahmed, Neel, and Wang 2010, Chen, Tang, Jiang and Lin 2010, Alford, Jones, Leftwich and Zmijewski 1993, Ali and Hwang 2000, Bartov and Goldberg 2001, Bushman and Piotroski 2006, Ball, Kothari and Robin 2000). Thus, we examine the impact of variation in institutional factors, including the system of laws (code vs. common law a proxy for quality of financial statement information)4, the quality of securities law enforcement, the pervasiveness of earnings management, the extent of differences between local GAAP and IFRS, and the country-level degree of conservatism, measured by differential timeliness,5 on the relation between CDS spreads and earnings. We perform three distinct tests to examine the impact of IFRS on the credit risk informativeness of earnings. First, we examine the association of earnings with CDS spreads under both local GAAP (i.e. pre-IFRS) and IFRS, and test whether the adoption

Code-law accounting affords managers more latitude in timing income recognition thereby obfuscating the economic performance of the firm. For example, firms create earnings reserves in good years through excessive impairment charges and provisions, and use these reserves in bad years. The low quality of accounting information is attributed primarily to governance practices in code-law countries which are based on the stakeholder model, whereby the different stakeholders such as employees, government, lenders and shareholders are able to affect accounting choices. In particular, one of the main accounting incentives of the various stakeholders is to reduce the volatility of net income, thereby creating a strong incentive to smooth earnings (Ball, Kothari and Robin 2000). 5 While differential timeliness (DT) is also related to country characteristics such as code vs. common law (Ball, Kothari and Robin 2000, Bushman and Piotroski 2006) and strength of securities law enforcement (Bushman and Piotroski 2006), we treat DT as an independent feature of the accounting system.
4

of IFRS changed the association between these variables. Second, we examine whether the association of earnings with CDS spreads subsequent to the adoption of IFRS depends on the country level institutional factors described above. Third, we also test for the potential asymmetry (non-linearity) of CDS spreads to high versus low profitability in light of the evidence in Callen, Livnat and Segal (2009) of a non-linear relation between CDS spreads and profitability. We gauge the impact of IFRS on the credit informativeness of earnings using three samples of firms referenced in CDS contracts: a sample of firms that adopted IFRS; a sample of U.S. firms; and a matched sample consisting of both IFRS adopters and U.S. firms. The latter sample allows us to control for potential time-period effects (Meyer 1995, Bertrand, Duflo, and Mullainathan 2004) and other potential correlated omitted variables (Cram, Karan and Stuart 2009). Consistent with the findings in Callen, Livnat and Segal (2009) and Das, Hanouna, and Sarin (2009), we find that earnings are informative about credit risk in IFRS countries both before and after the adoption of IFRS. However, there is no statistical difference in the pre and post adoption period, indicating that IFRS had no impact on the credit informativeness of earnings. These results are robust to the inclusion of other determinants of credit risk, namely size, leverage, credit rating, volatility of stock returns, risk free interest rates, and quarter, industry and country fixed-effects. The results of our second set of tests which condition on institutional factors show that while earnings are (not) informative about credit risk in pre and post adoption periods in common (code) law countries, in countries with strong (weak) legal enforcement, in countries with low (high) earnings management, and in countries with low (high) differences between local GAAP and IFRS, the adoption of IFRS did not change the CDS-earnings relation. The results of our third test confirm the asymmetry (non-linearity) in CDS spreads to high and low earnings, but there is no change in the post-IFRS period compared to the pre-IFRS period. We contribute to the extant literature relating to IFRS by studying the impact of the adoption on the credit markets. The prior literature in this area predominantly focuses
5

on the equity markets (Hail and Leuz 2007), but the credit market is no less important for several reasons. First, the information needs of the equity market may differ from those of the credit market. Exclusive reliance on the equity market to quantify the impact of IFRS ignores the fact that the credit market represents a significant source of financing for the public firms. The CDS market, which is a subset of the credit market, is a multitrillion-dollar market. Second, we provide empirical evidence to show whether adoption of IFRS is potentially significant for creditors and credit rating agencies whose primary concern is downside risk. Finally, this paper provides additional evidence on the credit informativeness of earnings using an international sample. While prior literature discusses the significance of accounting for the credit markets, the empirical evidence on the credit informativeness of earnings is fairly limited.6 In what follows, Section 2 provides the literature review and develops the hypotheses based primarily on the Merton (1974) and Duffie-Lando (2001) models of CDS pricing. Section 3 describes the data. Section 4 presents the empirical results. Section 5 concludes.

2. Literature Review and Hypotheses Development7

The extant research on IFRS adoption is concerned almost exclusively with equity markets (see for example Daske, Hail, Leuz and Verdi 2007 and Li 2010). But, as the financial crisis of 2008 has shown, debt markets are no less crucial than equity markets for the functioning of the financial system in general and the financing of public corporations in particular. The issue that we address is whether IFRS conveys better information about the firms credit risk, proxied by the level of CDS spreads, than local GAAP. In theory, the credit risk information conveyed by IFRS earnings could be evaluated through debt markets, such as the corporate bond market, rather than through the CDS market. In the absence of arbitrage opportunities, contractual features (such as embedded options, covenants, and guarantees), and market frictions, the CDS spread and

The information needs of the credit market help to shape the properties of the accounting numbers (Watts 2003) especially because the primary influence on financial reporting comes from debt markets, not equity markets (Ball 2006). 7 We do not summarize the burgeoning literature on IFRS and equity markets in this study.
6

the corporate bond yield spreadthe difference between the bond yield and the risk-free rateare necessarily identical for floating rate corporate debt (Duffie 1999). Nevertheless, it is precisely because of these latter factorscontractual features and market frictionsthat CDS instruments offer several advantages over corporate bonds (and other debt instruments) for analyzing the determinants of credit risk. First, the finance literature has shown that corporate bond spreads include factors unrelated to credit risk, such as systematic risk unrelated to default (Elton, Gruber, Agrawal and Mann 2001) and especially illiquidity (Longstaff, Mithal and Neis 2005). Huang and Huang (2002) conclude that less than 25 percent of the credit spread in corporate bonds is attributable to credit risk. Second, interest rate risk drives fixed-rate corporate bond yields for fixed rate debt quite independently of credit risk. Third, in contrast to CDS instruments, corporate bonds are replete with embedded options, guarantees, and covenants. Heterogeneity in these features potentially distorts the relationship between accounting numbers and credit risk in cross-sectional studies. Even more problematic is that they may generate a spurious relation between earnings and credit risk. For example, the positive relation between earnings and corporate bond prices could be driven by earnings-based covenants rather than by credit risk per se. With lower earnings, earningsbased covenants are more likely to be binding, increasing the probability of technical bankruptcy and concomitant expected transactions (renegotiation) costs, thereby leading to reduced bond prices. In contrast, except in rare cases, technical default is not a credit event in CDS contracts and thus has little impact on CDS spreads. Fourth, the available empirical evidence indicates that credit risk price discovery takes place first in the CDS market and only later in the bond and equity markets (Blanco, Brennan and Marsh 2005; Zhu 2006; Daniels and Jensen 2005; Berndt and Ostrovnaya 2008). The bond markets lagged reaction potentially distorts empirical studies relating earnings to bond prices. Fifth, unlike corporate bond yield spreads, no benchmark risk-free rate need be specified for CDS spreads minimizing potential misspecification of the appropriate risk-free rate proxy(Houweling and Vorst 2005). Sixth, CDS rates are closely related to the par value of the reference bond, whereas corporate bond values (including their taxability characteristics) are affected by coupons. Heterogeneity in coupon rates potentially
7

distorts the relationship between earnings and credit risk in cross-sectional studies. Finally, bond yield spreads are affected by tax differentials in bond pricing. Elton et al. (2001) document a tax premium of 29 to 73 percent of the corporate bond spread, depending on the rating.8 2.1 Earnings and CDS Spreads Of the information provided by financial statements that relates to credit risk, we focus our analysis on earnings. Earnings can be used by investors to estimate the reference entitys economic performance and true asset (wealth) dynamics, another important determinant of credit risk. More specifically, increased profitability of the firm, as measured by current accounting earnings, should reduce its credit risk since, with increased profitability, the reference entity is wealthier and less likely to default. Moreover, accounting studies have shown that current earnings are a good predictor of future earnings (Finger 1994; Nissim and Penman 2001), future cash flows (Dechow, Kothari and Watts 1998; Barth, Cram, and Nelson 2001) and firm equity performance (Dechow 1994). In other words, an increase (decrease) in earnings portends an increase (decrease) in current and future operating and equity performance and, hence, a reduced (increased) probability of bankruptcy. Also, earnings comprise a significant portion of the short-term change in firm assets (via clean surplus) and, therefore, provide information to investors about the firms asset and wealth dynamics, crucial variables in the estimation of credit risk (Duffie and Lando 2001). Consistent with these arguments, Callen, Livnat and Segal (2009) and Das, Hanouna and Sarin (2009) show earnings to be an important determinant of CDS spreads, indicating that earnings provide relevant information in assessing firm wealth and asset dynamics.

One may still be tempted to argue that since CDS instruments are derivatives whose price depend on the value of the underlying debt, the role of accounting information in determining the CDS spread is unclear because the prices and volatilities of the underlying financial instruments are observable. However, as shown by Duffie and Lando (2001), even noisy accounting information is relevant for the pricing of CDS because accounting provides information about the firms wealth and asset dynamics and, hence, about the probability of the occurrence of credit events such as bankruptcy. Also, unlike equities, even the largest corporate bonds do not trade very often and bond prices are often not observable. In addition, published bond prices are often stale or simply interpolated.
8

Although Callen, Livnat and Segal (2009) and Das, Hanouna and Sarin (2009) show that earnings are relevant for assessing credit risk, the findings in these studies are based primarily (but not exclusively) on U.S. reference entities employing U.S. GAAP. Given the relatively high quality accounting standards together with effective regulation and securities laws enforcement in the U.S., one cannot generalize these findings to an international setting where countries differed in their accounting standards prior to IFRS adoption, and differ subsequently in securities law enforcement, quality of financial information, and extent of earnings management. These considerations lead to our first hypothesis: H1a: CDS spreads are negatively correlated with earnings both pre- and postIFRS adoption. Proponents of IFRS argue that a single set of high quality accounting standards enhances the transparency and comparability of financial statements. IFRS requirements are potentially important for evaluating the firms credit risk because they are likely to result in more transparent statements that provide better information about the firms ability to pay back its debts. Examples of some of these IFRS requirements include: comprehensive reporting on cash flows; disclosure of significantly more information about pension obligations than under local GAAP; disclosure of scale and extent of offbalance-sheet obligations such as leased assets; information about changes in the amounts set aside for liabilities of uncertain timing or amount; consolidation of special purpose entities used for securitization transactions when the substance of the relationship indicates that they are controlled by the transferor of the asset; and stricter rules for gain recognition from asset securitization, resulting in more frequent treatment of securitization as financing transactions. More transparent statements are also likely to provide debt markets with better estimates of the potential for the firm to recover from a credit event such as bankruptcy and better estimates of the amounts that debt holders might recover in case of bankruptcy. Enhanced comparability should allow for easier and/or better assessment of credit risk, especially when the reference entity and the firm ensuring itself against the credit risk of the reference entity via CDS instruments reside in different local GAAP jurisdictions.
9

The [limited] evidence in the literature appears to support these conjectures. Specifically, Beneish, Miller and Yohn (2009) find that IFRS adopting countries attract more debt investment, and have a lower extent of debt home bias. They also find that their result is contextual and is driven by adopting countries that have weaker investor protection and higher financial risk. They argue that their findings indicate that IFRS adoption reduces the agency costs of debt in countries with less developed investor protection and greater financial risk, consistent with IFRS providing more transparent information. Florou and Kossi (2009) examine whether the mandated introduction of IFRS affects the source and cost of corporate debt. They find that mandatory IFRS adopters are more likely to issue public bonds than to borrow privately. They also find that mandatory IFRS adopters pay lower bond yield spreads but not in the case of private loans. They argue that their findings are consistent with IFRS enhancing the quality and comparability of accounting information. Their findings also suggest that mandatory IFRS adoption is especially beneficial for public bond investors, who rely much more on financial statements and have much less monitoring and renegotiating privileges compared to private lenders. Finally, they show that the positive consequences of IFRS for debt financing are present only in countries with stricter enforcement rules, higher control of corruption and lower financial risk. Christensen, Lee and Walker (2009) show that mandatory IFRS affects debt contracting. Their results suggest that as a result of the change to IFRS, the likelihood of debt covenant violations increases, requiring costly renegotiations between lenders and debtors. These arguments lead to the following hypothesis: H1b: CDS spreads are more negatively correlated with earnings post-IFRS adoption than pre-IFRS adoption. 2.2 Institutional differences, Conditional Conservatism and CDS Spreads Skeptics often question whether IFRS can credibly provide more transparent information and raise concerns that the one size fits all approach simply distorts economic, political and regulatory differences among firms in different jurisdictions. Furthermore, the implementation of IFRS crucially depends on the effectiveness of
10

regulation which is likely to depend in turn on the underlying economic and political institutions influencing the incentives of the managers and auditors responsible for financial statement preparation (e.g., Ball, Robin and Wu 2003). Empirical evidence highlights the role of institutions in moderating the impact of IFRS adoption on economic constructs. Specifically, Daske et al. (2007) and Li (2010) show that IFRS has an impact on cost of equity and valuation only in countries with strong enforcement of securities laws. In a recent working paper, Kosi, Pope and Florou (2010) find a significant increase in the credit relevance of financial statement information to firm ratings for mandatory IFRS adopters. The increase is greater than for a matched sample of U.S. firms, and is more pronounced in countries with strong enforcement regimes and higher discrepancies between local standards and IFRS. Thus, IFRS may not improve the accounting quality of credit risk relevant information uniformly across firms and countries because of additional factors such as legal and political systems and incentives of financial reporting (Soderstrom and Sun 2011). Earnings opacity, the reverse of earnings transparency, is a result of complex interactions among managerial motives, accounting standards, and enforcement of accounting standards. On one hand managers may intentionally manipulate earnings leading to less transparent earnings, while on the other hand less transparent earnings may be a result of lax enforcement or poor accounting standards. Because managements motives are not observable, earnings transparency at the country level is inherently difficult to measure. We gauge accounting information transparency at the country level using factors that affect the quality of the accounting information. Specifically, we use the following factors in the analysis: origin of the legal system, level of enforcement, level of earnings management, degree of conditional conservatism, and a measure which quantifies differences between local GAAP and IFRS. Accounting standards in code law countries are primarily influenced by governmental priorities. Political influence on accounting standard setting in code law countries makes accounting earnings a measure by which to divide profits among various stakeholders such as governments, shareholders, banks, and labor unions (Ball, Kothari
11

and Robin 2000). As a consequence, earnings in code law countries are arguably less informative about credit risk than earnings in common law countries both pre and post IFRS adoption. Enforcement of the legal system also affects accounting quality directly, through enforcement of accounting standards and litigation against managers and auditors. This enforcement role of legal systems is especially important in the context of IFRS for two reasons. First, the IASB issues IFRS but does not have the power to enforce the proper application of the standards. It is the local legal system of each country where firms are listed that is responsible for enforcement (Schipper, 2005). Second, IFRS are principlesbased, which means that auditors and accountants need to use judgment and discretion rather than detailed standards in order to adapt these principles to specific situations (Ball, 2006). Thus, earnings in countries with high legal enforcement may be more reflective of credit risk than earnings in countries with low legal enforcement pre and post IFRS adoption. Earnings management is a measure of earnings opacity, the opposite of earnings transparency. While it is difficult to measure earnings transparency directly at the country level, we focus on distributional properties of reported accounting numbers across countries and across time as captured by the Leuz et al. (2003) measure of earnings management at the country level. We focus on this measure because past literature has identified the existence of earnings management as weakening the link between accounting performance and the true economic performance of a firm. Thus, one would expect earnings in countries with high earnings management to be less informative about credit risk than earnings in countries with low earnings management. We use the difference between local GAAP and IFRS at the country level to identify countries which stand to benefit the most from the transition to IFRS. If the difference between the local GAAP and IFRS is large, the IFRS induced transparency would be high for that country, the underlying assumption being that IFRS is more transparent.

12

Conditional conservatism should be of particular importance to CDS investors (and bondholders) who are far more concerned with earnings decreases than earnings increases (Callen, Livnat and Segal 2009). As emphasized by Watts (2003), conditional conservative accounting is demanded by debt holders because it reduces managements ability to artificially increase earnings and asset values. Specifically, when future negative cash flow shocks are anticipated, conservative accounting requires the firm to recognize future losses immediately in income, resulting in a concomitant reduction in asset values. Hence, the degree of conditional conservatism has a direct impact on the informativeness of earnings in assessing credit risk. The more conditionally conservative the firm, the more likely is the firms accounting to act as a trip wire regarding anticipated future negative cash flow shocks that may reduce the firms ability to pay back its debt. Thus, one should expect ex ante that the more conservative the firm, the more informative is its earnings about firm credit risk and, hence, the more negative the relation between earnings and CDS spreads. The relation between conditional conservatism and IFRS is complex. On one hand, one of the major characteristics of IFRS is their anti-conservative nature. The IASB has explicitly rejected the notion of conservatism in accounting and indicated a preference for neutral accounting (Watts 2003). This is evident in several IFRS standards that reduce conservatism by design. For instance, IFRS disallow the completed contract method, allow for fair value accounting for investment properties, allow for impairment reversals, and for the revaluation of PP&E and biological assets. Indeed, Ahmed, Neel and Wang (2010) show that firms exhibited more conservative accruals and more timely loss recognition in the pre-adoption period. Thus, if IFRS yield financial statements that are less conservative by comparison to local GAAP, we should expect a weaker inverse relation between credit spreads and IFRS earnings relative to local GAAP earnings. On the other hand, the use of fair value accounting under IFRS provides an early warning signal of declining asset prices and, hence, credit risk. The induced transparency under IFRS could potentially have a more significant impact on firms who are less conservative under local GAAP because of the faster recognition of losses under fair value accounting. These considerations lead to our next set of hypotheses:
13

H2a: The relation between CDS spreads and earnings, both pre- and postIFRS adoption depends on country-wide institutional differences such as code law versus common law, regulation enforcement, earnings management, local GAAP-IFRS differences and conditional conservatism. H2b: Change in the association between CDS spreads and earnings following the adoption of IFRS depends on country-wide institutional differences such as code law versus common law, regulation enforcement, earnings management, local GAAP-IFRS differences and conditional conservatism. Although the IFRS literature tend to find that institutional factors affect the quality of earnings, a caveat is warranted regarding the effect of institutional factors on the credit risk informativeness of earnings. The extant empirical literature which looks at the relation between institutional factors and earnings quality focuses on the equity markets as end-users. These findings do not necessarily extend to credit markets in general or to the credit risk informativeness of earnings (as measured by CDS spreads) in particular. 2.3 Non-linearity in CDS Spreads and the earnings relation While earnings may provide credit risk relevant information in general, the nonlinear payoff function of debt holders, and by extension CDS holders, suggests that CDS prices will have a stronger reaction to accounting information that presages potential bankruptcy than to information that presages additional profits. Extant evidence of such nonlinearities in the corporate bond market is mixed. Datta and Dhillon (1993) find that corporate bond yields do not react more to (unexpected) losses than to (unexpected) profits whereas Easton, Monahan and Vasvari (2009) find just such an asymmetry. More related to our study, Callen, Livnat and Segal (2009) provide evidence supporting this non-linearity in the CDS market; they show that CDS spreads react more to earnings of firms with low profitability.9 Following the discussion above related to the differences

As pointed out by Callen et al. (2009), firms that reference CDS instruments tend to be large and successful with few loss quarters. Therefore, we measure asymmetry with reference to the median level of earnings rather than profits and losses.
9

14

between the U.S. and the international setting, one cannot simply generalize the U.S. findings on the non-linear relation between CDS prices and earnings to the international milieu. These considerations lead to our last set of hypotheses: H3a: The relation between CDS spreads and earnings, both pre- and postIFRS adoption, is greater in absolute value for firms with earnings below the median than for firms with earnings above the median. H3b: Change in the association between CDS spreads and earnings following the adoption of IFRS is greater in absolute value for the firms with earnings below the median than for firms with earnings above the median.

3. Sample Data and Univariate Empirical Results

CDS data, currency exchange rates and interest rates are collected from Thomson DataStream Navigator. Thomson has Credit Market Analysis (CMA) data covering CDS contracts for 70 countries from 2003 through 2008 (see Panel A of Table 1). Deleting CDS indices and keeping CDS contracts of reference entities in countries which adopted IFRS results in an initial sample of 1,132 firms. We match each of the tickers on Thomson Financial to obtain financial statement variables. We find a ticker match for 782 firms, out of which 392 are U.S. firms and 390 are firms from across 40 other countries. Out of these 390 firms, we identify 211 firms in 17 countries that adopted IFRS in 2005. For each firm-quarter we obtain the price of CDS contracts with 5-year maturity issued 45 days after the fiscal-quarter-end. If there are no CDS contracts issued on the 45th day after fiscal quarter-end, we utilize the first CDS contract issued in the range from 42 to 48 days after the quarter-end. The spread for each CDS contract is derived from mid-market quotes contributed by investment banks and default-swap brokers. For each CDS contract, we collect data on its seniority (senior or subordinated), and the currency of the underlying debt, which in turn determines the currency of the CDS

15

contract.10 We convert the CDS quotes to U.S. dollars (USD) using the exchange rate as of the CDS quote date. We obtain quarterly financial statement data required to compute market value of equity, ROA, and leverage from the Worldscope database. We download the financial information in USD wherever available; otherwise, we convert the variables to USD using the exchange rate as of the fiscal quarter-end. When available we use shortterm credit ratings from S&P to proxy for credit ratings; otherwise, we use long-term credit ratings. Variable definitions are listed in Appendix A. We impose the following restrictions on the sample: positive leverage (measured as short term debt plus long term debt scaled by market value of equity plus total liabilities), non-missing value for each of the following: market value of equity, return on assets (computed as income before extraordinary items scaled by total assets), standard deviation of stock returns (computed on a rolling basis using the most recent 12 monthly returns with at least 6 data points), and credit rating. In addition, all sample CDS contracts in this study are limited to senior debt both because there are very few junior CDS contracts in our initial sample and because their pricing determinants are very different from senior contracts. Also, given the paucity of voluntary adopters in our data, we eliminate voluntary adopter observations from our sample. These restrictions reduce the sample size to 1,699 firm-quarters of 155 firms across 17 countries. For our control sample, we find 5,295 firm quarters of 308 U.S. firms, which meet the same criteria above. Finally, we remove firm-quarters for which we could not find a U.S. match (see below), and end up with a matched sample of 1,663 CDS contract quarters for the two groups, namely, a treatment group of 155 IFRS firms and a control group of 279 U.S. firms. To mitigate the effect of outliers, all continuous variables are winsorized at the top and bottom 1%. Table 1, Panel A summarizes the sample selection and distribution of the IFRS and U.S. samples. Panel B lists the number of firms and the number of firm-quarter observations pre- and post-IFRS adoption by country. Most of the data are from six countries: Germany, the U.K., France, Sweden, Italy and Australia. The data also indicate

Restructuring clauses are only available from 2008. As a result we do not control for this variable in the analysis.
10

16

that there are significantly more CDS contracts and, hence, sample data, post-IFRS adoption by comparison to pre-IFRS adoption. This is consistent with the world-wide secular increase in CDS usage. (Insert Table 1 about here) To alleviate concerns regarding time series effects and potential correlated omitted variables (Meyer 1995, Bertrand, Duflo, and Mullainathan 2004), we use a matched sample for our empirical tests, where each IFRS CDS observation is matched with a U.S. CDS observation. We match each IFRS firm-quarter in the sample with a U.S. domiciled firm-quarter utilising propensity score matching. For each fiscal quarter, we estimate the propensity score using CDS determinants: profitability, size, leverage, credit ratings and stock return volatility. We select the U.S. match based on the closest propensity score without replacement. Our final matched sample comprises 1,132 firm quarters in the post period and 531 firm quarters in the pre period. (Insert Table 2 about here) Table 2, Panel A presents descriptive statistics of the main variables used in the analysis for the IFRS and U.S. matched samples by pre- and post-IFRS adoption periods. These variables include the CDS spread (CDS_PRM), Return on Assets (ROA), Log Market Values (SIZE), Leverage (LEV), Standard Deviations of Returns (SD_RET), the countrys risk-free rate of interest (SPOT), and S&P firm ratings (RATEN). CDS spreads are significantly higher in the post adoption period for the IFRS firms and the U.S. firms, indicating an increase in the average level of credit risk. Possible explanations for these results include the overall impact of the credit crisis and the increase in the number of reference entities in the post-IFRS period. Comparing the means and the medians by period, the data show significant variation between the pre and post adoption periods for IFRS firms as well as U.S. firms for the other variables. Specifically, IFRS firms are larger in the post period, are more profitable, have lower leverage, higher quality credit rating (lower credit rating number) and experience greater volatility in equity returns. In addition, interest rates are also higher in the post period. The U.S. firms exhibit a similar pattern in all variables. Despite the fact that we have matched the two groups using
17

propensity score on profitability, size, leverage, credit ratings and stock return volatility for each firm quarter, we find that the mean ROA differs significantly between the IFRS firms and U.S. firms in the pre and post period, and the mean SIZE and LEV differ in the post period. To address issues arising out of imperfection in the matching process, we use the control variables in our regressions to soak up the effect of the matched pair differences (Cram, Karan and Stuart 2009). Finally, the CDS spreads across the two samples have similar means in the pre- and post-IFRS periods. However, the median spread in the U.S. is higher (lower) in the pre (post) periods. Panels B lists the means for the main variables for the 1,663 firm quarters in the matched sample by country. On the whole, the data show significant variation across countries. Panel C shows the institutional variables by country. CODE is equal to 1 if the country has a code law system, and zero otherwise. Out of the 17 countries in our sample, only 4 countries are common law countries. The RULE column measures the strength of country legal enforcement based on the year 2005 proxy from Kaufmann, Kraay, and Mastruzzi (2007). The EM column measures the extent of earnings management as measured by the country-wide aggregate earnings management score from Leuz, Nanda, and Wysocki (2003). The DIFF column measures the extent of accounting differences between local GAAP and IFRS for each country based on the summary score of 21 key accounting dimensions as computed by Bae, Tan, and Welker (2008). The fifth column lists average differential timeliness (DT) during the sample period. DT is accounting conservatism measured using the Basu (1997) differential timeliness (DT) metric at the country level following Ball, Robin and Sadka (2000). We estimate DT for each country in the period prior to the adoption of IFRS (2003-2004) and in the period after the adoption (2006-2008) using the entire cross-section of firms for which we could find the required data on Datastream. Panel D presents the correlations between the CDS spread and the primary determinants of the spread pre- and post-IFRS adoption. All correlations are highly significant at the 1% level and the signs pre- and post-IFRS are identical. With one exception, these correlations are also consistent with the underlying theory (Merton 1974; Duffie and Lando 2001). Specifically, the CDS spread is correlated positively with
18

leverage, volatility, and the ratings variable (the higher the ratings number, the lower the credit rating); the spread is correlated negatively with profitability, size, and the SPOT rate except in the post period for the IFRS sample. We elaborate more on the positive correlation between the spread and SPOT in the post period for the IFRS sample below.

4. Multivariate Empirical Results

We analyze the impact of IFRS on CDS premia and the relation between CDS premia and earnings using a levels analysis. Lack of sufficient appropriate cross-country data, primarily CDS related data, precludes us from applying changes and event study analyses to the issue at hand.11 Section 4.1 estimates the general relation between CDS and ROA (normalized earnings) pre and post-IFRS. Section 4.2 examines whether the relation depends on country characteristics. Section 4.3 explores whether the relation is non-linear. Section 4.4 presents additional tests. 4.1 Main Regressions Based on the predictions of the Merton (1974) and Duffie and Lando (2001) models, CDS spreads should be decreasing with the reference firms size and ratings quality and increasing with the reference firms leverage, the volatility of its stock returns, and the risk-free rate of interest in the economy. In addition, we control for industry, quarter and country fixed-effects, where applicable. We estimate the regressions using the following five samples: full IFRS, full U.S., matched IFRS only (consisting of IFRS firms for which we could find a U.S. match), matched U.S. only (consisting of U.S. firms that were matched to an IFRS firm), and a matched sample of IFRS-U.S. firms. The matched sample compares the change in credit risk informativeness pre and post adoption for IFRS firms with the credit risk informativeness of a control group of U.S. firms (difference in difference with a matched sample design). (Insert Table 3 about here)

11

Only U.S.databases provide sufficient CDS data for such analyses. See Callen, Livnat and Segal (2009).

19

Table 3, Panel A shows the panel data regression coefficient estimates for the full IFRS sample (column 1), the matched IFRS sample (column 2), the full U.S. sample (column 3) and the matched U.S. sample (column 4). The fifth column (sixth column) shows the regression for the matched sample of IFRS and U.S. firms (excluding financial crisis quarters). In all six regressions, we use Ordinary Least Squares and control for industry and quarter fixed-effects. We also control for country fixed effects in all regressions involving IFRS firms. Statistical significance is based on robust standard errors corrected for firm and time clustering (Petersen 2009). Because the ratings numbers are inversely related to the quality of the firms rating, the CDS spread should increase with the credit rating (labelled as RATEN). ROA is separated into pre- and postIFRS adoption periods for the IFRS and U.S. samples. For example, ROA_PRE_US (ROA_POST_US) equals ROA for U.S. sample firms if the fiscal quarter-end is in the period prior to (after) the IFRS adoption period and zero otherwise. The results of F-tests comparing pre and post coefficient estimates are given below the estimated regressions. The results are reasonably consistent with the Merton and Duffie-Lando model predictions for the full IFRS sample as well as for the full U.S. sample.12 Specifically, the coefficients for ratings, leverage and volatility are positive and significant whereas the coefficient for size is negative and significant. The coefficient on the risk-free rate (SPOT_IFRS) is positive and significant which is the only estimated coefficient in this regression inconsistent with theory. We checked the correlation of CDS spreads with the spot rate by country to see if the results are driven by specific countries, but we obtain positive correlations across all countries. One plausible explanation for the positive correlation is that when interest rates are low, the cost of financing is also low leading to a lower probability of default.13 The coefficients on the earnings variables are all negative and highly significant (p-value<0.01), consistent with the negative relation between the spread and profitability. Comparing the coefficients on earnings across periods for the IFRS sample reveals that there is no significant difference across the pre and post

12

p-values in the tables are at the two-tailed level. However, we refer to one-tailed significance results in the text when the theory or the empirical evidence in the literature dictates a directional relation between a variable and the CDS spread. Unless explicitly noted otherwise, significance is two-tailed. 13 Offsetting this explanation is the possibility that lower interest rate regimes are often symptomatic of a recessionary economy and, therefore, of increased credit risk.

20

periods, indicating that the adoption of IFRS did not change the credit informativeness of earnings. Specifically, the coefficient on ROA_POST is not statistically different from the coefficient on ROA_PRE for the IFRS samples (p-value=0.94 for full IFRS sample and p-value=0.91 for matched IFRS sample) and the matched samples (p-value=0.923 in the matched regression and p-value=0.243 in the matched regression excluding the crisis period). Similar results are obtained for the U.S. sample, with one exception. The regression using the full U.S. sample shows a significant decrease (p-value=0.022) in the coefficient on ROA indicating that the credit risk informativeness of earnings declined for the U.S. sample over the two periods. However the coefficients are not statistically different in any of the regressions involving the matched U.S. sample. Overall, these results indicate that although earnings is informative about credit risk both for IFRS adopters and U.S. firms, the adoption of IFRS did not have a significant impact on the credit risk informativeness of earnings for IFRS firms. 4.2 Institutional differences, conditional conservatism and CDS spreads The regressions in Table 3 do not account for certain country and firm-level factors that might be related to the credit risk informativeness of earnings, especially when contrasting pre- and post-IFRS adoption periods. Table 4 replicates Table 3 for the matched sample, controlling separately for code versus common law countries, strong versus weak legal enforcement countries, the extent of country-level earnings management, the extent to which IFRS and local GAAP standards differ from each other, and the extent of differences in differential timeliness across countries. Multicollinearity concerns induced by the correlations among these conditioning variables and the extensive breakdowns of earnings pre- and post-IFRS interacted with a relatively large number of conditioning variables preclude us from controlling for all five additional factors simultaneously. Instead, we evaluate each of the conditioning variables separately. Specifically, we condition ROA on the institutional variables, and benchmark the results against the matched U.S. sample. (Insert Table 4 about here)

21

The CODE column of Table 4 conditions earnings on code law versus common law countries. ROA_PRE_COND (ROA_PRE_NO_COND) equals ROA if the fiscal quarter-end is in the pre-IFRS period and the country is characterized as a code (common) law system, and zero otherwise. ROA_POST_COND and

ROA_POST_NO_COND are defined similarly if the fiscal quarter-end is in the postIFRS period. The ROA coefficient is negative and insignificant in pre-IFRS (pvalue=0.17) and in post-IFRS periods (p-value=0.54) for code law countries, and negative and highly significant (p-value<0.01) in pre and post-IFRS periods for common law countries. Tabulated F-tests indicate that the difference in the ROA coefficients between the pre and post periods is not significant both for code (p-value=0.52) and common law (p-value=0.81) countries. Comparing the coefficients on ROA across the two law systems, we find that the coefficient is more negative for common law countries, and the difference is highly significant in the post period (p-value=0.02) and marginally significant in the pre period (p-value=0.14). Overall, these results suggest that the origin of the legal system is related to the credit informativeness of earnings with higher credit informativeness of earnings in common law countries, consistent with the conjecture that the accounting in common law countries prior to IFRS better reflected economic reality in comparison to code law countries. However, the adoption of IFRS appears to be inconsequential; the credit informativeness of earnings in code law countries as well as in common law countries did not change significantly following the adoption of IFRS. These results suggest that the adoption of IFRS did not mitigate differences in the transparency of accounting information across the two regimes.14 The RULE column conditions earnings on the strength of legal enforcement. We partition ROA pre and post-IFRS based on the legal enforcement score; values above (below) the median represent countries with strong (weak) legal enforcement. ROA_PRE_COND (ROA_PRE_NO_COND) equals ROA if the period is pre-IFRS and the observation belongs to a country with strong (weak) legal enforcement. ROA_POST_COND (ROA_POST_NO_COND) is defined similarly if the fiscal quarter
14

For completeness, we also tabulate F-tests in Table 4 comparing U.S. and IFRS ROA coefficients but we do not comment on these in the text.

22

end is in the post-IFRS period. The coefficient on ROA is negative and significant before and after IFRS adoption for countries with strong legal enforcement (p-value<0.01), but not significant for countries with weak legal enforcement (p-value=0.43 in pre-period and p-value=0.76 in post-period). F-tests indicate that the adoption of IFRS had no effect on the credit risk information conveyed by earnings irrespective of the strength of legal enforcement (p-value=0.71 for firms in countries with strong legal enforcement and pvalue=0.72 for firms in countries with weak legal enforcement). These results suggest that earnings convey credit risk relevant information only in countries with strong legal enforcement and that IFRS adoption had no impact on the credit informativeness of earnings both for countries with strong and weak legal enforcement. The EM column of Table 4 conditions earnings on the extent of earnings management. Values above (below) the median represent countries with high (low) earnings management. ROA_PRE_COND (ROA_PRE_NO_COND) equals ROA if the period is pre-IFRS and the observation belongs to a country with high (low) earnings management. ROA_POST_COND (ROA_POST_NO_COND) is defined similarly if the fiscal quarter-end is in the post-IFRS period. ROA is negative and significant before and after IFRS adoption for countries with low earnings management (p-value<0.1), but not for countries with high earnings management (p-value=0.31 in pre-period and pvalue=0.87 in post-period). F-tests indicate that the adoption of IFRS had no significant effect on the credit risk information conveyed by earnings irrespective of whether earnings management is high (p-value=0.27) or low (p-value=0.58). While these results suggest that earnings convey credit risk relevant information only when the level of earnings management is low, they show again that IFRS adoption is not consequential. The DIFF column of Table 4 conditions earnings on the extent of accounting differences between local GAAP and IFRS. Values above (below) the median represent countries with high (low) local GAAP-IFRS differences. ROA_PRE_COND

(ROA_PRE_NO_COND) equals ROA if the period is pre-IFRS and the observation belongs to a country with high (low) accounting differences. ROA_POST_COND (ROA_POST_NO_COND) is defined similarly if the fiscal quarter-end is in the postIFRS period. The ROA coefficients are insignificant in the pre-IFRS (p-value=0.91) and
23

in the post-IFRS (p-value=0.89) periods for countries with large differences across local GAAP and IFRS. The coefficients on ROA are negative and highly significant (pvalue<0.01) in the pre and post-IFRS periods for countries with small differences between local GAAP and IFRS. F-tests show that differences in the pre and post ROA coefficients are not statistically significant whether GAAP differences are large (pvalue=0.83) or small (p-value=0.97). Overall, these results suggest that earnings convey credit risk relevant information for countries with low differences between local GAAP and IFRS both before and after IFRS adoption, but not for countries with high differences between local GAAP and IFRS. Moreover, the adoption of IFRS had no impact on the relation between CDS and profitability irrespective of whether GAAP/IFRS differences are large or small. The DT column of Table 4 conditions earnings on the extent of conditional conservatism measured at the country-year level.15 We rank countries according to the DT measure in the pre and post-IFRS periods, and classify countries with DT values above (below) the median as COND (NO_COND). ROA_PRE_COND

(ROA_PRE_COND) equals ROA if the period is pre-IFRS and the country is characterised as high (low) DT. ROA_POST_COND (ROA_POST_NO_COND) is defined similarly if the fiscal quarter-end is in the post-IFRS period. The coefficients on ROA for the high conservatism countries are negative and highly significant (pvalue<0.01) both in the pre and post periods, but only marginally significant (p-value = 0.10 and 0.11 in the pre and post periods, respectively) for low conservatism countries. However, the differences in the ROA coefficients for the high DT (p-value=0.83) and low DT (p-value=0.80) countries are statistically insignificant both pre and post, suggesting that DT measured using the Basu (1997) specification does not affect the credit informativeness of earnings. More importantly for our study, the coefficients for high and low DT countries did not change significantly following the adoption of IFRS, indicating that IFRS had no impact on the credit informativeness of earnings.

15

In untabulated results, we find that DT decreased significantly after the adoption of IFRS; mean (median) DT in the pre and post-IFRS periods is 0.33 (0.33) and 0.15 (0.19), respectively. This is consistent with the anti-conservative tendency of IFRS.

24

Overall, the evidence from Table 4 suggests that the institutional factors - origin of law, strength of the legal system, level of earnings management, and difference between local GAAP and IFRS, affect the CDS-earnings relation, but IFRS was not consequential for the credit risk informativeness of earnings. These results further validate the results documented in Table 3 that the adoption of IFRS had no discernable impact on the credit informativeness of earnings. 4.3 Asymmetry in CDS-earnings relation Callen, Livnat and Segal (2009) show that credit risk, as measured by CDS premia, is more sensitive to earnings decreases than earnings increases, which is consistent with a non-linear payoff function for debt holders. To explore the possibility that the impact of IFRS is more pronounced for firms with lower profitability, we estimate our regressions partitioning the sample firms according to the level of ROA. Table 5, column 1 shows the regression results for the IFRS matched sample only with pre and post ROA further partitioned based on the median level of earnings by quarter. ROA_PRE_IFRS_HIGH (ROA_PRE_IFRS_LOW) equals ROA if the period is pre IFRS, the observation belongs to the IFRS sample and ROA is above (below) the median ROA. ROA_POST_IFRS_HIGH (ROA_POST_IFRS_LOW) is defined similarly if the fiscal quarter-end is in the post IFRS period. Table 5, Column 1 shows that ROA estimated coefficients are negative and significant before and after IFRS adoption for firms with earnings below (p-values<0.05) and above (p-values<0.01) the median. Consistent with asymmetry of the CDS-earnings relation, the estimated coefficients of below median earnings firms are far more negative than the coefficients of the above median earnings firms. F tests indicate that the differences are highly significant both before (p-value<0.01) and after (p-value<0.01) IFRS adoption. However, F-tests also indicate that the adoption of IFRS had no effect on the credit risk information conveyed by earnings, both for high (p-value=0.94) and low (p-value=0.69) earnings. (Insert Table 5 about here) Column 2 (column 3) shows results benchmarked against the matched U.S. sample (excluding the financial crisis quarters). The results are similar to those just
25

described. Overall, these findings suggest that there is asymmetry in the CDS-earnings relation and that earnings are more credit risk relevant for firms with earnings below the median. However, IFRS adoption is inconsequential to the CDS-earnings relation whether earnings are below or above median. 4.4 Robustness Checks We perform several additional tests to ensure that our results are robust. First, following Cram, Karan and Stuart (2010), we use a dummy variable for each matched pair in the matched sample regressions (1,663 additional dummies). While the coefficient on ROA_PRE_IFRS is negative (-1.767) but insignificant (p-value=0.27), and the ROA_POST_IFRS is negative (-3.140) and significant (p-value=0.03), the difference between these coefficients is not statistically significant, confirming our main results that adoption of IFRS did not have an impact on the CDS-earnings relation. Furthermore, untabulated results related to institutional differences, level of differential timeliness and the level of earnings are similar to those presented in Table 4 and Table 5. Second, we use an alternative research design for the matched sample. Rather than employing a difference in differences design with dummies for matched pairs, we estimate the regressions based on differences between the matched pairs for dependent and independent variables. The results are very similar to those reported. Third, we check the results at country level for three countries in our sample which have at least 20 firms and 50 firm-quarters in the pre and post IFRS periods, namely Australia, France and United Kingdom. We estimate the regressions at the country level (similar to column 2 in Table 3) and with a matched U.S. sample (similar to column 5 in Table 3). While both sets of tests reveal wide variation of the estimated ROA coefficients among these three countries, they indicate that the coefficient on ROA_POST_IFRS is not significantly different from the coefficient on ROA_PRE_IFRS. Fourth, we use the log of book values as an alternative proxy for size. Our main result that IFRS did not have an impact on the CDS-earnings relation remains unchanged for each of these robustness checks.

26

5. Conclusion
The proponents of IFRS argue that one of the major benefits of IFRS is increased financial statement transparency. If so, the adoption of IFRS provides a unique setting for examining how changes in transparency affect the credit risk informativeness of accounting numbers. Our analysis indicates that earnings are a significant determinant of credit risk, but the adoption of IFRS did not change the overall credit informativeness of earnings. While it is difficult to measure transparency at the country level, the extant prior literature documents that institutional factors affect accounting information quality and hence arguably transparency. Therefore, we investigate whether credit risk informativeness is affected by country-level institutional factors, including code law vs. common law, strength of security law enforcement, prevalence of earning management, IFRS-local GAAP differences, and differential timeliness. We find that earnings are informative about credit risk in pre as well as post adoption periods for common law countries, countries with strong legal enforcement, countries with low earnings management, countries with low differences between local GAAP and IFRS, and countries with both high and low differential timeliness. However, similar to the results related to earnings in general, IFRS adoption does not appear to have an impact on the credit informativeness of earnings even when controlling for institutional differences. We also explore the asymmetry in the credit risk informativeness of earnings and find that while earnings of low profitability firms are more credit risk relevant than the earnings of the firms with high profitability, IFRS does not have any impact on the asymmetry.

27

References Ahmed, A.S., M.J. Neel, and D.D. Wang. 2010. Does mandatory adoption of ifrs improve accounting quality? Preliminary evidence. W.P., Texas A&M University. Alford, A., J. Jones, R. Leftwich and M. Zmijewski. 1993. The relative informativeness of accounting disclosures in different countries. Journal of Accounting Research 31(Supp): 183-229. Ali, A. and L. Hwang. 2000. Country-specific factors related to financial reporting and the value relevance of accounting data. Journal of Accounting Research 38 (1): 1-21. Bae, K., H. Tan and M. Welker. 2008. International gaap differences: The impact on foreign analysts. The Accounting Review 83 (3): 593-628. Ball, R. 2006. International financial reporting standards (ifrs): pros and cons for investors. Accounting and Business Research Special Issue: International accounting policy forum: 5-28. Ball, R., A. Robin and J. S. Wu. 2003. Incentives versus standards: Properties of accounting income in four east asian countries. Journal of Accounting and Economics 36 (1-3): 235-270. Ball, R., S. P. Kothari and A. Robin. 2000. The effect of international institutional factors on properties of accounting earnings. Journal of Accounting and Economics 29 (1): 1-51. Barth, M.E., D. Cram, and K. Nelson. 2001. Accruals and the prediction of future cash flows. The Accounting Review 76 (1): 27-58. Bartov, E. and S. R. Goldberg. 2001. The valuation-relevance of earnings and cash flows: An international perspective. Journal of International Financial Management and Accounting 12 (2): 103-132. Basu, S. 1997. The conservatism principle and asymmetric timeliness of earnings. Journal of Accounting and Economics 24: 3-37. Beneish, M.D., P.B. Miller and T.L. Yohn. 2009. The effect of IFRS adoption on crossborder investment in equity and debt markets. W.P. Indiana University. Berndt, A. and A. Ostrovnaya. 2008. Do equity markets favor credit markets news over options market news? W.P. Carnegie-Mellon University. Bertrand, M., E. Duflo and S. Mullainathan. 2004. How much should we trust differences-in-differences estimates? Quarterly Journal of Economics 119: 24975. Blanco, R., S. Brennan, and I. W. Marsh. 2005. An empirical analysis of the dynamic relation between investment-grade bonds and credit default swaps. Journal of Finance 60 (5): 2255-2281.
28

Bushman, R. and J. D. Piotroski. 2006. Financial reporting incentives for conservative accounting: The influence of legal and political institutions. Journal of Accounting and Economics 42(1/2): 107-148. Callen, J.L., J. Livnat and D. Segal. 2009. The impact of earnings on the pricing of credit default swaps. The Accounting Review 84 (5): 1363-1394. Chen, H., Q. Tang, Y. Jiang, and Z. Lin 2010. The role of international financial reporting standards in accounting quality: Evidence from the european union. Journal of International Financial Management & Accounting: forthcoming. Christensen, H, E. Lee, and M. Walker. 2009. Do ifrs reconciliations convey information? The effect of debt contracting. Journal of Accounting Research 47: 1167 1199. Cram, D. P., V. Karan and I. Stuart. 2009. Three threats to validity of choice-based and matched-sample studies in accounting research. Contemporary Accounting Research 26(2): 477-516. Daniels, K. N., and M. S. Jensen. 2005. The effect of credit ratings on credit default swap spreads and credit spreads. Journal of Fixed Income 15 (3): 16-33. Das, S. R., P. Hanouna, and A. Sarin. 2009. Accounting-based versus market-based cross-sectional models of cds spreads. Journal of Banking and Finance 33 (4): 719-730. Daske, H., L. Hail, C. Leuz, and R. Verdi. 2007. Adopting a label: Heterogeneity in the economic consequences of ifrs adoptions. Journal of Accounting Research 46 (5): 10851142. Datta, S. and U. S. Dhillon. 1993. Bond and stock market response to unexpected earnings announcements. Journal of Financial and Quantitative Analysis 28 (4): 565577. Dechow, P. 1994. Accounting earnings and cash flows as measures of firm performance: The role of accounting accruals. Journal of Accounting and Economics 18 (1): 3-42. Dechow, P., S. Kothari, and R. Watts. 1998. The relation between earnings and cash flows. Journal of Accounting and Economics 25 (2): 133-168. Duffie, D. 1999. Credit swap valuation. Financial Analyst Journal 55 (1): 73-87. Duffie, D. and D. Lando. 2001. Term structures of credit spreads with incomplete accounting information. Econometrica 69 (3): 633-644. Easton, P, Monahan S. and Vasvari, F. 2009. Initial evidence on the role of earnings in the bond market, Journal of Accounting Research, 47, 2009, 721-766.

29

Elton, E. J., M. J. Gruber, D. Agrawal, and C. Mann. 2001. Explaining the rate of spread on corporate bonds. Journal of Finance 56 (1): 247-277. Finger, C. 1994. The ability of earnings to predict earnings and cash flows. Journal of Accounting Research 32 (2): 210-223. Florou A. and U. Kossi. 2009. The economic consequences of mandatory ifrs adoption for debt financing. W.P. University of Lancaster and University of Macedonia. Hail, L., C. Leuz, and P. D. Wysocki, 2009. Global accounting convergence and the potential adoption of IFRS by the United States: An analysis of economic and policy factors. Working Paper, University of Pennsylvania, University of Chicago and MIT. Houweling, P., and T. Vorst. 2005. Pricing default swaps: Empirical evidence. Journal of International Money and Finance 24 (8): 1200-1225. Huang, M., and J. Huang. 2002. How much of the corporate-treasury yield spread is due to credit risk? Working paper, Stanford University. Kaufmann, D., A. Kraay, and M. Mastruzzi. 2007. Governance matters vi: Governance indicators for 1996-2006. Research W.P. No. 4280, The World Bank. Kosi, U., P.F. Pope, and A. Florou, A. Credit Relevance and Mandatory IFRS Adoption (August 2010). INTACCT Working Paper Series. Available at SSRN: http://ssrn.com/abstract=1679672. Leuz, C., D, Nanda, and P. D. Wysockic, 2003. Earnings management and investor protection: An international comparison. Journal of Financial Economics 69: 505-527. Li S. 2010. Does mandatory adoption of international financial reporting standards in the european union reduce the cost of equity capital? The Accounting Review 85 (2): 607640. Linsmeier, T., 2011. Financial reporting and financial crises: The case for measuring financial instruments at fair value in the financial statements. Accounting Horizons, 25(2), 409- 417. Longstaff, F. A., S. Mithal, and E. Neis. 2005. Corporate yield spreads: default risk or liquidity? New evidence from the credit default swap market. Journal of Finance 60 (5): 2213-2253. Merton, R. C. 1974. On the pricing of corporate debt: The risk structure of interest rates. Journal of Finance 29 (2): 449-470. Meyer, B. D. 1995. Natural and quasi-experiments in economics. Journal of Business & Economic Statistics 13: 15161.

30

Moodys. Special Comment. October 2004. The impact of international financial reporting standards (ifrs, formerly known as ias) on the credit ratings of european corporates. Nissim, D., and S. Penman. 2001. Ratio analysis and equity valuation: From research to practice. Review of Accounting Studies 6 (1): 109-154. Petersen, M. A. 2009. Estimating standard errors in finance panel data sets: Comparing approaches. Review of Financial Studies 22 (1): 435-480. Schipper, K. 1995. Academic accounting research and the standard setting process, Accounting Horizons (December), 61-73. Soderstrom, Naomi S. and Sun, Kevin Jialin. 2011. IFRS adoption and accounting quality: A review. European Accounting Review, Forthcoming. Available at SSRN: http://ssrn.com/abstract=1008416. Watts, R.L. 2003. Conservatism in accounting part I: Explanations and implications. Accounting Horizons 17: 207-221. Zhu, H. 2006. An empirical comparison of credit spreads between the bond market and the credit default swap market. Journal of Financial Services Research 29 (3): 211-235.

31

Appendix A - Variable Definitions

Variable CDS_PRM CODE COMMON COND Description Log of CDS spread (mid-point between bid and ask) from the daily quotes, 45 days from the fiscal quarter end Indicator value equal to 1 if the firm belongs to a country characterized by the code law system; 0 otherwise Indicator value equal to 1 if the firm belongs to a country characterized by the common law system; 0 otherwise Indicator variable equal to 1 if firm belongs to a country DIFF characterized by code law system; 0 otherwise in country with above median RULE, 0 otherwise in country with above median EM, 0 otherwise in country with above median DIFF, 0 otherwise in country with above median DT, 0 otherwise Bae, Tan & Welker (2008) Datastream Leuz, Nanda & Wysocki (2003) Thomson Financial Source Datastream

Accounting differences between local GAAP and IFRS based on the summary score of 21 key accounting dimensions. Higher value indicates larger differences between local GAAP and IFRS.

DT EM IFRS LEV NO_COND

Accounting conservatism measured using the Basu (1997) differential timeliness (DT) metric at the country level following Ball, Robin and Sadka (2000). Country-wide aggregate earnings management score. Higher value indicates higher extent of earnings management Indicator variable equal to 1 if firm is from country that adopted IFRS; 0 otherwise Short term debt plus long term debt scaled by market value of equity plus total liabilities Indicator variable equal to 1 if firm belongs to a country characterized by common law system; 0 otherwise in country with below median RULE, 0 otherwise in country with below median EM, 0 otherwise in country with below median DIFF, 0 otherwise in country with below median DT, 0 otherwise

POST PRE RATEN ROA RULE SD_RET SIZE SPOT U.S. HIGH LOW

Indicator variable equal to 1 if observation is post 2004; 0 otherwise Indicator variable equal to 1 if observation is pre 2005; 0 otherwise Credit rating, takes numerical values from 1 to 20, 1 being the highest for rating AAA+ and 20 being the lowest for rating CCC. Income before extraordinary items scaled by total assets Strength of legal enforcement. Higher value indicates stronger enforcement Standard Deviation of the most recent 12 monthly returns (with at least 6 data points) Log of market value Annualized 3-month Treasury-Bill rate Indicator variable equal to 1 if U.S. firms; 0 otherwise Indicator variable equal to 1 if ROA is above median Indicator variable equal to 1 if ROA is below median Standard & Poors Thomson Financial Kaufmann, Kraay, and Mastruzzi (2007) Thomson Financial Thomson Financial Datastream Thomson Financial

32

Table 1 Sample Selection and Distribution

Panel A: Sample selection by number of firms and CDS contract quarters
Non-U.S. Firms CDS prices available on datastream Firms with data available on Thomson Financial Firms from countries which adopted IFRS in 2005 Final Full Sample (delete observations with missing financial statement data including credit rating) Final Matched Sample 390 (41countries) 211 (17 countries) 155 (17 countries) 392 392 1,699 308 5,295 CDS Contract Quarters Firms U.S. CDS Contract Quarters Total CDS Firms Contract Quarters 1,382 782 603 463 6,994

155 (17 countries)

1,663

279

1,663

434

3,326

Panel B: Number of firms, firm quarters and CDS contract quarters by country
FIRMS PRE POST AUSTRALIA BELGIUM DENMARK FINLAND FRANCE GERMANY HONGKONG IRELAND ITALY NETHERLANDS NORWAY POLAND PORTUGAL SPAIN SWEDEN SWITZERLAND UNITED KINGDOM IFRS TOTAL UNITED STATES GRAND TOTAL 23 2 2 2 22 1 5 3 9 7 1 2 1 6 12 1 23 122 215 26 3 2 4 26 18 9 3 9 7 4 2 1 7 13 8 26 168 294 FIRM-QUARTERS POST 81 29 22 42 136 159 40 17 98 63 14 21 10 66 137 55 142 1,132 1,132 3,326

PRE 55 6 14 15 85 5 16 11 34 45 3 19 8 45 91 4 75 531 531

Notes Table 1: Panel A describes the sample selection process. Panel B shows the number of firms, the number of firmquarter observations and the number of CDS contracts before and after IFRS adoption by country for the the matched sample.

33

Table 2 - Descriptive Statistics

Panel A: Descriptive Statistics of Main Variables for matched IFRS and U.S. sample (n=3,326)
IFRS SAMPLE N=1,663 U.S. SAMPLE N=1,663 IFRS U.S. Mean p value STD 0.878 0.015 1.225 0.142 3.279 0.035 1.210 Std 1.051 0.017 1.222 0.139 3.044 0.034 1.065 IFRS U.S. Median p value

PRE

CDS_PRM ROA SIZE LEV RATEN SD_RET SPOT

MEAN 3.860 0.016 9.140 0.249 9.429 0.063 2.788 Mean 4.136 0.018 9.683 0.232 8.943 0.065 4.211 POSTPRE p value 0.000 0.034 0.000 0.014 0.003 0.091 0.000

POST

CDS_PRM ROA SIZE LEV RATEN SD_RET SPOT

POSTPRE

CDS_PRM ROA SIZE LEV RATEN SD_RET SPOT

N=531 MEDIAN 3.697 0.012 9.099 0.241 9.000 0.056 2.146 N=N=1,132 Median 4.102 0.013 9.695 0.207 9.000 0.060 4.150 POSTPRE p value 0.000 0.043 0.000 0.001 0.472 0.005 0.000

STD 0.834 0.022 1.138 0.128 3.280 0.028 1.305 Std 1.019 0.023 1.132 0.133 3.106 0.026 1.324

MEAN 3.893 0.013 9.106 0.247 9.593 0.065 2.488 Mean 4.068 0.016 9.596 0.221 9.143 0.067 4.573 POSTPRE p value 0.001 0.001 0.000 0.001 0.006 0.444 0.000

N=531 MEDIAN 3.773 0.010 8.986 0.238 9.000 0.056 2.450 N=1,132 Median 3.932 0.014 9.603 0.199 9.000 0.059 5.030 POSTPRE p value 0.001 0.000 0.000 0.005 0.001 0.138 0.000

0.527 0.012 0.639 0.808 0.416 0.279 0.000

0.021 0.022 0.010 0.427 0.000 0.380 0.000

0.121 0.002 0.078 0.071 0.123 0.386 0.000

0.000 0.467 0.055 0.225 0.002 0.137 0.000

34

Table 2 contd
Panel B: Mean of Main variables by country for the matched sample (n=3,326)
COUNTRY AUSTRALIA BELGIUM DENMARK FINLAND FRANCE GERMANY HONG KONG IRELAND ITALY NETHERLANDS NORWAY POLAND PORTUGAL SPAIN SWEDEN SWITZERLAND UNITED KINGDOM UNITED STATES N CDS_PRM ROA SIZE LEV RATEN SD_RET SPOT

136 35 36 57 221 164 56 28 132 108 17 40 18 111 228 59 217 1,663

3.637 3.509 3.750 4.830 4.234 4.245 4.029 3.305 3.992 4.006 4.967 3.653 3.194 4.014 3.837 3.830 4.422 4.012

0.040 0.013 0.007 0.015 0.016 0.007 0.035 0.005 0.006 0.016 0.006 0.003 0.002 0.013 0.019 0.020 0.027 0.015

9.046 9.871 9.580 8.988 9.938 9.952 9.314 9.356 9.414 9.869 9.023 7.005 9.209 9.909 9.099 10.305 9.584 9.440

0.170 0.176 0.451 0.227 0.220 0.255 0.230 0.362 0.315 0.170 0.303 0.340 0.401 0.302 0.186 0.235 0.213 0.229

8.588 8.229 10.472 11.193 9.665 8.049 10.393 8.000 9.061 9.157 8.765 11.075 8.500 8.910 8.539 8.424 9.378 9.287

0.061 0.050 0.046 0.074 0.066 0.068 0.070 0.059 0.058 0.063 0.083 0.107 0.069 0.057 0.065 0.058 0.066 0.066

6.039 3.709 3.320 3.415 3.245 3.946 5.119 3.173 3.476 3.235 4.811 3.155 3.031 3.200 2.803 2.139 4.840 3.907

Panel C: Institutional variables by country for the IFRS sample

CODE AUSTRALIA BELGIUM DENMARK FINLAND FRANCE GERMANY HONG KONG IRELAND ITALY NETHERLANDS NORWAY POLAND PORTUGAL SPAIN SWEDEN SWITZERLAND UNITED KINGDOM MEDIAN 0 1 1 1 1 1 0 0 1 1 1 1 1 1 1 1 0 RULE 1.81 1.45 2.03 1.95 1.31 1.77 1.45 1.62 0.37 1.75 2.02 NA NA 1.10 1.86 1.96 1.73 1.75 EM -4.8 -19.5 -16.0 -12.0 -13.5 -21.5 -19.5 -5.1 -24.8 -16.5 -5.8 NA -25.1 -18.6 -6.8 -22.0 -7.0 -16.25 DIFF 4 13 11 14 12 11 3 1 12 4 7 12 13 14 10 12 1 11 DT 0.285 0.080 0.236 0.216 0.191 0.269 0.103 0.025 0.457 0.154 0.271 0.190 0.430 0.406 0.279 0.259 0.292 0.259

35

Panel D: Correlations with CDS spread (CDS_PRM) for the matched sample (n=3,326)
IFRS SAMPLE PRE ROA SIZE LEV RATEN SD_RET SPOT (0.190) (0.315) 0.219 0.480 0.433 (0.025) POST (0.082) (0.225) 0.202 0.250 0.367 0.351 PRE (0.445) (0.513) 0.532 0.407 0.445 (0.169) U.S. SAMPLE POST (0.370) (0.418) 0.452 0.339 0.469 (0.323)

Notes Table 2: Panel A presents means, medians, p-values of the difference between means, and p-values of the difference between medians of the main variables for the matched IFRS and U.S. sample by adoption periods. Panel B presents means of the main variables in the analysis for the IFRS and the U.S. matched sample by country. Panel C presents the institutional factors by country, origin of legal system (CODE), strength of enforcement (RULE), level of earnings management (EM), difference between local GAAP and IFRS (DIFF) and differential timeliness (DT). Panel D presents sample correlation of CDS_PRM (log of CDS spread) with the selected variables for the IFRS and U.S. matched sample by adoption periods. Variable definitions are provided in Appendix A.

36

Table 3 - The Determinants of CDS Spreads Before and After IFRS Adoption
VARIABLES ALL (1) ROA_PRE_IFRS ROA_POST_IFRS ROA_PRE_US ROA_POST_US SIZE LEV RATEN SD_RET SPOT POST PRE -0.260*** (0.000) 0.901** (0.020) 0.052*** (0.000) 3.288** (0.020) 0.094* (0.055) 4.963*** (0.000) 5.066*** (0.000) YES YES YES 1,699 0.983 -0.260*** (0.000) 0.948** (0.015) 0.052*** (0.000) 3.450** (0.013) 0.096* (0.055) 4.886*** (0.000) 4.999*** (0.000) YES YES YES 1,663 0.983 -4.248*** (0.002) -4.140*** (0.003) IFRS MATCHED (2) -4.153*** (0.002) -3.990*** (0.004) -7.877*** (0.000) -4.680*** (0.000) -0.204*** (0.000) 2.006*** (0.000) 0.057*** (0.000) 5.581*** (0.000) -0.034 (0.416) 4.530*** (0.000) 4.553*** (0.000) -6.382** (0.014) -6.007*** (0.000) -0.174*** (0.000) 2.256*** (0.000) 0.036*** (0.000) 6.067*** (0.000) -0.084 (0.160) 4.125*** (0.000) 3.965*** (0.000) ALL (3) U.S. IFRS AND U.S. MATCHED MATCHED MATCHED EXCL. CRISIS (4) (5) (6) -3.334*** (0.007) -3.213** (0.018) -7.332*** (0.003) -8.709*** (0.000) -0.218*** (0.000) 1.535*** (0.000) 0.045*** (0.000) 5.318*** (0.000) 0.026 (0.169) 4.742*** (0.000) 4.731*** (0.000) YES YES YES 3,326 0.981 -4.004*** (0.003) -2.444 (0.103) -8.714*** (0.001) -9.656*** (0.000) -0.249*** (0.000) 1.111*** (0.000) 0.050*** (0.000) 6.295*** (0.000) 0.035 (0.227) 4.846*** (0.000) 4.838*** (0.000) YES YES YES 2,358 0.981

COUNTRY DUMMIES INDUSTRY DUMMIES QUARTER DUMMIES OBSERVATIONS R-SQUARED F-TESTS ROA_PRE_IFRS=ROA_POST_IFRS ROA_PRE_US=ROA_POST_US

YES YES 5,295 0.982

YES YES 1,663 0.981

0.005 (0.943)

0.012 (0.912) 5.286 (0.022) 0.019 (0.891)

0.009 (0.923) 0.384 (0.536)

1.364 (0.243) 0.155 (0.694)

Notes to Table 3: Table 3 shows the regression estimates of the CDS spread on its determinants. Column 1 (Column 2) shows the regressions using the full IFRS (matched IFRS) samples. Column 3 (Column 4) shows the regressions using the full U.S. (matched U.S.) samples. Column 5 (Column 6) shows the matched sample results (excluding the crisis quarters). The CDS spread determinants include profitability (ROA), log of market value (SIZE), leverage (LEV), credit rating (RATEN), volatility of returns (SD_RET), the country risk-free interest rate (SPOT). The PRE (POST) indicator variable takes the value of 1 if the fiscal quarter is in the pre-adoption (post-adoption) period and zero otherwise. Variables are defined in Appendix A. All regressions control for industry, fiscal quarter, and country fixed-effects (not reported). Two-tailed pvalues are in parentheses. All regressions are estimated using OLS with robust standard errors corrected for firm and time clustering.

37

Table 4 - The Determinants of CDS Spreads Before and After IFRS Adoption and Five Conditioning Variables
VARIABLES CODE (1) -2.666 (0.165) -1.194 (0.542) -6.195*** (0.000) -6.532*** (0.000) -8.858*** (0.000) -10.479*** (0.000) -0.172*** (0.000) 1.704*** (0.000) 0.050*** (0.000) 5.410*** (0.000) 0.026 (0.159) 4.291*** (0.000) 4.263*** (0.000) YES YES YES 3,326 0.979 RULE (2) -5.158*** (0.000) -4.648*** (0.000) -2.038 (0.427) -0.950 (0.760) -9.102*** (0.000) -10.494*** (0.000) -0.171*** (0.000) 1.675*** (0.000) 0.049*** (0.000) 5.379*** (0.000) 0.015 (0.393) 4.306*** (0.000) 4.291*** (0.000) YES YES YES 3,326 0.979 EM (3) -2.277 (0.309) 0.348 (0.868) -5.518*** (0.000) -6.268*** (0.000) -8.865*** (0.000) -10.222*** (0.000) -0.196*** (0.000) 1.589*** (0.000) 0.048*** (0.000) 5.904*** (0.000) 0.017 (0.356) 4.532*** (0.000) 4.505*** (0.000) YES YES YES 3,286 0.979 DIFF (4) -0.270 (0.912) 0.355 (0.891) -6.006*** (0.000) -5.953*** (0.000) -8.919*** (0.000) -10.501*** (0.000) -0.175*** (0.000) 1.663*** (0.000) 0.049*** (0.000) 5.418*** (0.000) 0.023 (0.201) 4.326*** (0.000) 4.301*** (0.000) YES YES YES 3,326 0.979 DT (5) -5.134*** (0.001) -4.816*** (0.001) -3.500 (0.100) -2.921 (0.112) -9.042*** (0.000) -10.564*** (0.000) -0.169*** (0.000) 1.700*** (0.000) 0.049*** (0.000) 5.396*** (0.000) 0.017 (0.349) 4.285*** (0.000) 4.262*** (0.000) YES YES YES 3,326 0.979

ROA_PRE_COND ROA_POST_COND ROA_PRE_NO_COND ROA_POST_NO_COND ROA_PRE_US ROA_POST_US SIZE LEV RATEN SD_RET SPOT POST PRE

COUNTRY DUMMIES INDUSTRY DUMMIES QUARTER DUMMIES OBSERVATIONS R-SQUARED F-TESTS ROA_PRE_COND=ROA_POST_COND ROA_PRE_NO_COND=ROA_POST_NO_COND ROA_PRE_COND=ROA_PRE_NO_COND ROA_POST_COND=ROA_POST_NO_COND ROA_PRE_US=ROA_POST_US ROA_PRE_US=ROA_PRE_COND ROA_PRE_US=ROA_PRE_NO_COND ROA_POST_US=ROA_POST_COND ROA_POST_US=ROA_POST_NO_COND

0.418 (0.518) 0.056 (0.813) 2.208 (0.138) 5.978 (0.481) 0.497 (0.481) 6.373 (0.012) 1.101 (0.295) 21.92 (0.000) 4.369 (0.037)

0.134 (0.715) 0.137 (0.711) 1.435 (0.232) 1.532 (0.547) 0.364 (0.547) 3.106 (0.079) 5.578 (0.019) 13.35 (0.000) 9.718 (0.002)

1.245 (0.265) 0.313 (0.576) 1.780 (0.183) 9.483 (0.556) 0.347 (0.556) 6.121 (0.014) 1.994 (0.159) 23.55 (0.000) 5.563 (0.019)

0.048 (0.827) 0.002 (0.969) 4.613 (0.032) 5.786 (0.491) 0.475 (0.491) 8.922 (0.003) 1.584 (0.209) 18.46 (0.000) 7.104 (0.008)

0.044 (0.834) 0.061 (0.804) 0.556 (0.456) 0.935 (0.506) 0.442 (0.506) 2.738 (0.099) 4.710 (0.031) 10.35 (0.001) 14.92 (0.000)

38

Notes to Table 4:
Table 4 shows the regressions of the CDS spread on its determinants conditioning for institutional factors by country using the matched IFRS and U.S. samples. Columns 1 to 5 condition (COND) on the origin of legal system (CODE), strength of enforcement (RULE), level of earnings management (EM), difference between local GAAP and IFRS (DIFF) and differential timeliness (DT), respectively and show the regression of the CDS spread on profitability (ROA), log of market value (SIZE), leverage (LEV), credit rating (RATEN), volatility of returns (SD_RET), the country risk-free interest rate (SPOT). The constant PRE and POST indicator variables take the value of 1 if the fiscal quarter is in the pre-adoption and post-adoption period, respectively, and zero otherwise. Pre and Post are trifurcated into three separate coefficients for IFRS countries with COND equal to 1, IFRS countries with NON_COND equal to 1, and the U.S. Variables are defined in Appendix A. All regressions control for industry and fiscal quarter (not reported). Two-tailed p-values are in parentheses. All regressions are estimated using OLS with robust standard errors corrected for firm and time clustering and multiple CDS contracts per firm. The number of observations in the EM column is smaller because we do not have the EM measure for Poland. The ROA variables in each column are defined as follows: The column labeled CODE conditions on Code vs. Common law countries. COND=1 if code law country and zero otherwise ROA_PRE_COND=ROA pre-IFRS adoption if code law and zero otherwise. ROA_POST_COND=ROA post-IFRS adoption if code law and zero otherwise. ROA_PRE_NO_COND=ROA pre-IFRS adoption if common law and zero otherwise. ROA_POST_NO_COND=ROA post-IFRS adoption if common law and zero otherwise. The column labeled RULE conditions on the strength of legal enforcement. Values above (below) the median represent countries with strong (weak) legal enforcement. COND=1 if strong legal enforcement and zero otherwise ROA_PRE_COND=ROA pre-IFRS adoption if strong legal enforcement and zero otherwise. ROA_POST_COND=ROA post-IFRS adoption if strong legal enforcement and zero otherwise. ROA_PRE_NO_COND=ROA pre-IFRS adoption if weak legal enforcement and zero otherwise. ROA_POST_NO_COND=ROA post-IFRS adoption if weak legal enforcement and zero otherwise. The column labeled EM conditions on extent of earnings management prior to the adoption of IFRS. Values above (below) the median represent countries with high (low) earnings management. COND=1 if high earnings management and zero otherwise ROA_PRE_COND=ROA pre-IFRS adoption if high earnings management and zero otherwise. ROA_POST_COND=ROA post-IFRS adoption if high earnings management and zero otherwise. ROA_PRE_NO_COND=ROA pre-IFRS adoption if low earnings management and zero otherwise. ROA_POST_NO_COND=ROA post-IFRS adoption if low earnings management and zero otherwise. The column denoted DIFF conditions on the extent to which there are differences between local GAAP and IFRS. Values above (below) the median represent countries with large (small) IFRS-local GAAP differences. COND=1 if large accounting differences and zero otherwise ROA_PRE_COND=ROA pre-IFRS adoption if large accounting differences and zero otherwise. ROA_POST_COND=ROA post-IFRS adoption if large accounting differences and zero otherwise. ROA_PRE_NO_COND=ROA pre-IFRS adoption if small accounting differences and zero otherwise. ROA_POST_NO_COND=ROA post-IFRS adoption if small accounting differences and zero otherwise. The column denoted by DT conditions on the level of differential timeliness We estimate DT for each country separately in the pre (post) IFRS periods of 2002-2004 (2006-2008) using the Basu (1997) specification for those firms in each country for which earnings, returns and market values are available on Datastream. We classify each country to high (low) DT if the estimated DT is above (below) the median. COND=1 if the country-year is classified as high DT and zero otherwise. ROA_PRE_COND=ROA pre-IFRS adoption if high DT and zero otherwise. ROA_POST_COND=ROA post-IFRS adoption if high DT and zero otherwise. ROA_PRE_NO_COND=ROA pre-IFRS adoption if low DT and zero otherwise. ROA_POST_NO_COND=ROA post-IFRS adoption if low DT and zero otherwise.

39

Table 5 - The Determinants of CDS Spreads Before and After IFRS Adoption for High vs. Low Earnings
VARIABLES IFRS MATCHED (1) -3.318** (0.014) -3.214** (0.023) -13.195*** (0.000) -11.628*** (0.000) MATCHED (2) -2.595** (0.033) -2.450* (0.067) -12.882*** (0.000) -11.400*** (0.000) -6.610*** (0.006) -8.233*** (0.000) -9.204 (0.127) -10.300*** (0.007) -0.219*** (0.000) 1.574*** (0.000) 0.043*** (0.000) 5.259*** (0.000) 0.024 (0.205) 4.763*** (0.000) 4.763*** (0.000) YES YES YES 3,326 0.981 IFRS AND U.S. MATCHED EXCL. CRISIS (3) -3.436** (0.010) -2.269 (0.131) -11.708*** (0.000) -7.494** (0.019) -8.254*** (0.001) -9.157*** (0.000) -10.069 (0.118) -15.471*** (0.000) -0.249*** (0.000) 1.147*** (0.000) 0.049*** (0.000) 6.112*** (0.000) 0.030 (0.297) 4.889*** (0.000) 4.883*** (0.000) YES YES YES 2,358 0.981

ROA_PRE_IFRS_HIGH ROA_POST_IFRS_HIGH ROA_PRE_IFRS_LOW ROA_POST_IFRS_LOW ROA_PRE_US_HIGH ROA_POST_US_HIGH ROA_PRE_US_LOW ROA_POST_US_LOW SIZE LEV RATEN SD_RET SPOT POST PRE

-0.262*** (0.000) 0.992*** (0.009) 0.050*** (0.000) 3.368** (0.011) 0.096* (0.051) 4.940*** (0.000) 5.077*** (0.000) YES YES YES 1,663 0.984

COUNTRY DUMMIES INDUSTRY DUMMIES QUARTER DUMMIES OBSERVATIONS R-SQUARED F-TESTS ROA_PRE_IFRS_HIGH=ROA_POST_IFRS_HIGH ROA_PRE_IFRS_LOW=ROA_POST_IFRS_LOW ROA_PRE_IFRS_HIGH=ROA_PRE_IFRS_LOW ROA_POST_IFRS_HIGH=ROA_POST_IFRS_LOW ROA_PRE_US_HIGH=ROA_POST_US_HIGH ROA_PRE_US_LOW=ROA_POST_US_LOW ROA_PRE_US_HIGH=ROA_PRE_US_LOW ROA_POST_US_HIGH=ROA_POST_US_LOW

0.006 (0.940) 0.160 (0.690) 9.491 (0.002) 7.537 (0.007)

0.014 (0.906) 0.140 (0.708) 9.890 (0.002) 8.269 (0.004) 0.668 (0.414) 0.025 (0.876) 0.202 (0.654) 0.301 (0.584)

0.766 (0.382) 1.249 (0.264) 6.625 (0.010) 2.697 (0.101) 0.173 (0.678) 0.541 (0.462) 0.086 (0.769) 2.389 (0.123)

40

Notes to Table 5: Table 5 shows the regressions of the CDS spread on its determinants conditioning for high vs. low ROA. Column 1 shows the regression coefficients for the matched sample. Column 2 (column 3) shows the regression coefficients for the matched IFRS and U.S. sample (excluding the crisis quarters). The CDS spread determinants include profitability (ROA), log of market value (SIZE), leverage (LEV), credit rating (RATEN), volatility of returns (SD_RET), the country risk-free interest rate (SPOT). The constant PRE and POST indicator variables take the value of 1 if the fiscal quarter is in the pre-adoption and post-adoption period, respectively, and zero otherwise. The HIGH (LOW) variable column is equal to 1 if ROA is above median (below median). Variables are defined in Appendix A. All regressions control for industry and fiscal quarter (not reported). Two-tailed p-values are in parentheses. All regressions are estimated using OLS with robust standard errors corrected for firm and time clustering and multiple CDS contracts per firm.

41