Holger Daske University of Mannheim Luzi Hail The Wharton School, University of Pennsylvania Christian Leuz The University of Chicago Booth School of Business and NBER Rodrigo Verdi Sloan School of Management, MIT Current draft: October 2009 (First draft: February 2007)
Abstract
This paper examines market liquidity and cost of capital effects associated with voluntary IFRS adoptions around the world. In contrast to prior work, we focus on the heterogeneity in the economic consequences, recognizing that firms have considerable discretion in how they implement IFRS. Some firms may simply adopt the label, while for others IFRS adoption may be part of a strategy to increase their commitment to transparency. To illustrate these differences, we classify firms into ‘label’ and ‘serious’ adopters using changes in firms’ underlying reporting incentives and actual reporting behavior, and then analyze whether capital markets respond differently around IFRS adoptions. We find that, on average, voluntary IFRS adoptions are not associated with capital market benefits, especially when compared to other forms of commitment such as cross-listing in the U.S. Consistent with our predictions, we find an increase in market liquidity and a decline in the cost of capital for ‘serious’ adopters. These benefits are likely attributable to broader changes in firms’ commitment to transparency, and not just IFRS. JEL classification: G14, G15, G30, K22, M41, M42 Key Words: International accounting, Reporting incentives, IAS, U.S. GAAP, Disclosure, Cost of equity, Enforcement, IFRS implementation * We appreciate the helpful comments of Ray Ball, Joachim Gassen, Günther Gebhardt, Bob Holthausen, Bjorn Jorgensen, Steve Young, Peter Wysocki, and workshop participants at the 2007 Utah Winter Accounting Conference, 2007 Harvard Business School IMO Conference, 2007 Global Issues in Accounting Conference, 2007 Journal of Accounting, Auditing and Finance Conference, 2008 AAA Annual Meeting, University of Chicago, INSEAD, Lancaster University, University of Lugano, University of Mannheim, Massachusetts Institute of Technology, and New York University. We thank Eric Blouin, Emily Goergen, Wannia Hu, Yuji Maruyama, Anindya Mishra, Michael Sall, Caleb Smith, Richie Wan, and Kai Wright for their excellent research assistance. 1 1. Introduction Over the last number of years the adoption of International Financial Reporting Standards (IFRS) has gained considerable momentum around the world.1 Almost 100 countries require or permit the use of IFRS for financial reporting purposes, and several more have decided to require IFRS in the near future (www.iasplus.com). But even before IFRS became mandatory, many firms around the world have voluntarily adopted or switched to IFRS. Studying the effects around voluntary IFRS adoptions can provide important insights and are the focus of this paper. There are several prior studies analyzing capital-market effects around voluntary IFRS adoptions (which are reviewed in detail in Section 2). These studies typically focus on the average effect around IFRS adoptions with respect to some outcome variable (e.g., earnings quality, liquidity, cost of capital), and the results are often mixed. However, cross-sectional differences in the adoption effects and the reasons for this heterogeneity are rarely analyzed. Moreover, most studies characterize the estimated effects as being attributable to IFRS adoption per se, neglecting the underlying forces that drive the adoption decision. By analyzing the crosssectional variation in the effects around IFRS adoptions, we can shed light on the question of whether the observed effects stem from IFRS reporting per se or, more broadly, from firmspecific differences and changes in reporting incentives. Our main prediction is that the observed economic consequences around IFRS adoptions largely depend on firms’ reporting incentives and the underlying motivations for the accounting regime change. As a result, there should be predictable heterogeneity in the economic 1 See, e.g., Barth (2008) and Hail et al. (2009). International Accounting Standards (IAS) were renamed to IFRS in 2001. In this paper, we use IAS and IFRS interchangeably but we do not presume that earlier IAS and later IFRS adoptions necessarily have the same consequences. In principle, our tests are equipped to capture differences between these adoptions. 2 consequences across firms. IFRS, like any other set of accounting standards, offers firms substantial discretion in applying the standards. Moreover, firms’ reporting incentives are different and the strength of enforcement differs considerably across countries (e.g., Ball et al., 2003; Leuz et al., 2003; Ball and Shivakumar 2005; Lang et al., 2006; Burgstahler et al., 2006). For these reasons, one frequently voiced concern is that some firms may adopt IFRS merely as a label without making material changes in their reporting policies (e.g., Ball, 2001, 2006). In contrast, other firms may experience a major change in their reporting incentives and adopt IFRS as part of a broader strategy to increase their commitment to transparency. Our study is designed to identify and examine such differences across firms. Provided that investors can differentiate between firms that adopt IFRS as a ‘label’ and those that make ‘serious’ changes in their reporting strategy, we should observe differential economic consequences. An increased commitment to transparency is expected to reduce information asymmetry and estimation risk, and hence should be rewarded with higher market liquidity and a lower cost of capital (e.g., Verrecchia, 2001; Lambert et al., 2007). To examine these hypotheses, we analyze a large panel of voluntary IFRS (and IAS) adoptions from 1990 to 2005 across 30 countries in the universe of firms on Worldscope. We identify voluntary IFRS adoptions based on accounting standards classifications in Worldscope, Global Vantage, and an extensive hand-collection of firms’ annual reports and provide a comparison of these three classifications in Appendix A. After combining and cross-checking these sources, we provide descriptive evidence on the frequency of IFRS adoptions as well as on adoption strategies and trends around the world. As expected, the number of firms reporting under IFRS is steadily increasing over the years. In addition, there is substantial variation in the frequency of voluntary adoptions across countries. Recognizing this variation is important when 3 analyzing the effects of mandatory IFRS reporting around the world and, as such, our paper also provides relevant evidence for the literature on mandatory adoptions. We begin by analyzing whether voluntary IFRS reporting is associated with higher market liquidity and a lower cost of capital relative to local GAAP firms, and relative to themselves prior to the adoption of IFRS. We examine three measures of economic outcomes, namely the price impact of trades suggested by Amihud (2002), the percentage bid-ask spread, and the implied cost of capital. Using these variables (and several others in our sensitivity analyses), we find little evidence that IFRS reporting is, on average, associated with higher market liquidity or a lower cost of capital, after controlling for various firm characteristics and industry-, country-, and yearfixed effects. We obtain similar results for voluntary adoptions of U.S. GAAP. In contrast, cross-listings in the U.S., which represent a credible commitment to more transparency, are associated with higher market liquidity and a lower cost of capital (Hail and Leuz, 2009). Furthermore, we show that stronger reporting incentives are significantly associated with higher market liquidity and a lower cost of capital, regardless of firms’ accounting standards choice. Both of these findings make it unlikely that measurement error in the dependent variables or lack of power is responsible for the weak average effects around voluntary IFRS adoptions. Next, we analyze the heterogeneity in the economic consequences across firms. Towards this end, we use two proxies to create variables indicating major changes in firms’ reporting incentives around IFRS adoptions. Our first proxy is input-based and focuses directly on firm characteristics that shape firms’ reporting incentives. Specifically, we expect firms that are larger, more profitable, more international, have larger financing needs, larger growth opportunities, and more dispersed ownership structures to have stronger incentives for transparent financial reporting. We use factor analysis and extract a factor that has consistent loadings for 4 these firm characteristics, and then use the distribution of changes in factor scores around IFRS adoption to split the sample into ‘serious’ and ‘label’ adopters. Our second proxy is output-based and measures changes in reporting behavior around IFRS adoptions. It is built on the idea that actual reporting changes are ultimately driven by changes in the underlying incentives. Following Leuz et al. (2003), we use the magnitude of accruals relative to the cash flow from operations as a simple characterization of earnings quality and use changes in this metric to capture improvements in reporting behavior. As before, we use the distribution of changes around IFRS adoption in this score to split firms into ‘serious’ and ‘label’ adopters. It is important to note that the construction of our split variables and the focus on changes allows for the possibility that firms with strong reporting incentives already provide high-quality reports under local GAAP. Thus, ‘label’ adopters are not necessarily firms with poor reporting. They are simply characterized by changing their accounting standards without material changes in their reporting incentives or behavior. Conversely, ‘serious’ adopters experience material changes and are not just adopting IFRS. Using these two proxies as partitioning variables, we show that ‘serious’ IFRS adopters experience larger increases in market liquidity and larger declines in the cost of equity capital than ‘label’ adopters. The results are statistically and economically significant and very similar across the two partitioning variables, although slightly stronger for the (input-based) reporting incentives factor. Compared to local GAAP firms, serious IFRS adopters experience a positive and economically significant net effect on market liquidity, consistent with the notion that these firms increase their commitment to transparency. The net effects on the cost of capital are less robust and generally close to zero, except in one specification for which we find that serious IFRS adoptions are associated with a significant decrease in the cost of capital. There is also evidence 5 that the net effect on Tobin’s q (used in the sensitivity analyses) is significantly positive for serious IFRS adopters using the reporting incentives factor as partitioning variable. Our paper makes several contributions to the literature. First, to our knowledge, this is the first study to analyze the heterogeneity in the economic consequences around voluntary IFRS adoptions for a large sample of international firms. We show that the effects around IFRS adoptions depend on changes in individual firms’ reporting incentives and actual reporting behavior. These findings imply that one has to exercise caution in attributing effects to (voluntary) IFRS reporting per se. Second, we contribute to the growing literature on the importance of firms’ reporting incentives in determining reporting outcomes (e.g., Ball et al., 2003; Ball and Shivakumar, 2005; Burgstahler et al., 2006). These studies typically provide comparisons across countries with similar reporting standards or draw inferences across private and public firms that are subject to the same accounting standards, but quite disparate otherwise. Our study applies the notion of reporting incentives at the firm level, and shows that changes in these incentives characterize the market effects around IFRS adoptions.2 Using changes and hence, each firm as its own control, also alleviates concerns about cross-sectional comparisons across countries or disparate groups of firms. Third, our study uses a considerable amount of hand-collection to cross-check and complement readily available data on the accounting standards used by firms around the world. Our comparisons highlight that the codings by Worldscope and Global Vantage are problematic for a variety of reasons and have to be used cautiously. Section 2 develops our hypotheses and reviews the literature. Section 3 delineates our research design and describes the data. Section 4 presents our analyses and results. Section 5 2 In a recent paper, Lang et al. (2009) also analyze how firm-level transparency choices, among them IFRS adoption, relate to liquidity and the cost of capital in an international setting. For IFRS adoption, they follow our approach of classifying cases that are likely to be more serious in their commitment to transparency. 6 concludes. The appendix provides a comparison of accounting standards classifications and describes the construction of our implied cost of capital measures. 2. Conceptual Underpinnings and Literature Review 2.1 Hypothesis Development The starting point of our study is the hypothesis that the observed economic consequences around IFRS adoptions largely depend on firms’ reporting incentives and the underlying motivations for the accounting regime change. Some firms may adopt IFRS merely as a label without making material reporting changes, whereas other firms may experience a major change in their reporting incentives and adopt IFRS as part of a broader strategy that credibly commits them to more transparency. As a result, there should be predictable heterogeneity in the economic consequences around voluntary IFRS adoptions related to firms’ reporting incentives. The conceptual underpinnings for this hypothesis are recent studies highlighting the importance of firms’ reporting incentives, rather than accounting standards, as key drivers of observed accounting properties and actual practices.3 This literature calls into question the extent to which the adoption of IFRS alone can provide a credible commitment to transparency. IFRS, like any other set of accounting standards, affords firms with substantial discretion as the application of accounting standards involves judgment and the underlying measurements are often based on private information. The way in which firms use this discretion is likely to depend on their reporting incentives, which are shaped by many factors, including countries’ institutional 3 The literature on the role of reporting incentives versus standards is rapidly growing. Examples are Ball et al. (2000, 2003), Leuz (2003), Ball and Shivakumar (2005), Burgstahler et al. (2006), Lang et al. (2006), and Bradshaw and Miller (2008). 7 frameworks, various market forces and firm characteristics.4 In this paper, we emphasize the role of firm-level reporting incentives. Their existence implies that the effects of (voluntary or mandatory) IFRS adoption are not just an enforcement issue. Even with perfect enforcement, observed reporting behavior will differ as long as the accounting standards offer some discretion (which they do for a good reason) and firm-level reporting incentives differ (Leuz, 2006). These arguments cast doubt on whether we can attribute changes in reporting quality and associated capital-market effects around IFRS adoptions to IFRS itself. Rather than the standards, the effects may reflect the differences in the incentives for credible reporting and the circumstances that accompanied the adoption of IFRS in the first place. The reporting incentives view implies predictable cross-sectional variation in the economic consequences around IFRS adoptions. We expect markets to react more favorably to IFRS adoptions around which firms make material and credible changes to their reporting and disclosure policies than to cases in which firms merely swap their reporting standards. For instance, firms that experience a positive shock to their growth opportunities have larger future financing needs and hence, see more value in improving transparency. Such firms may adopt IFRS as part of a broader strategy that credibly commits a firm to more transparent reporting. To the extent that the overall strategy is difficult to mimic (or more costly) for firms that are not serious about improving transparency, it constitutes a credible commitment.5 In this case, markets likely react favorably, e.g., by lowering the cost of capital, but the reaction reflects the entire commitment strategy, and IFRS adoption is only a proxy for this strategy (Leuz and Verrecchia, 2000). 4 This insight is also at the heart of the accounting choice and earnings management literature. See Watts and Zimmerman (1986), Healy and Wahlen (1999), and Dechow and Skinner (2000). 5 This argument is similar to the bonding hypothesis for U.S. cross-listings (e.g., Coffee, 1999; Stulz, 1999). The bonding hypothesis also presumes that cross-listing in the U.S. is more attractive for firms with large financing needs and growth opportunities, and that it entails higher costs for firms in which insiders consume large private control benefits and engage in expropriating outside investors (Doidge et al., 2004). 8 Thus, we do not claim that ‘serious’ IFRS adopters experience the documented effects because of IFRS adoption or because they comply strictly with IFRS. We classify them as such because they experience a substantial change in their underlying reporting incentives, which likely makes them serious about improving transparency. Put differently, we do not attempt to demonstrate effects from IFRS adoption per se or to identify the marginal effect of IFRS adoption. To the contrary, a key point of our study is to highlight that the estimated coefficient on the IFRS indicator variable cannot be simply attributed to the accounting standards per, but that it likely reflects differences in firms’ underlying motivations for IFRS adoption. Thus, addressing the self-selection inherent in the voluntary choice of an accounting regime is not appropriate in our paper, as we aim to highlight its existence, but it would be critical for studies that interpret their results as showing IFRS effects per se. An alternative hypothesis is that IFRS adoption itself leads to relatively homogenous reporting quality across firms. In this case, observed heterogeneity in the economic consequences mainly stems from prior differences in reporting quality, rather than differences in firms’ motivations for IFRS adoption. This hypothesis predicts that IFRS adoption per se has positive capital-market effects for most firms, assuming that IFRS reporting is more demanding and capital-market oriented compared to most local GAAP regimes. It also predicts that firms with lower levels of reporting quality (or incentives) prior to the switch exhibit more positive effects (holding all else equal) because they experience larger improvements in reporting quality by switching to IFRS. Given this alternative, our research design uses changes in firms’ reporting incentives and actual reporting behavior to partition firms into serious and label adopters. In addition, we estimate to specifications controlling for pre-IFRS levels in the proxies for firms’ reporting incentives and reporting behavior to address this alternative hypothesis. 9 Our main hypothesis for the observed heterogeneity in capital-market effects around IFRS adoptions presupposes that investors can discern between serious and label adoptions, at least imperfectly. However, IFRS adoption itself can be almost costless, in particular in countries with low-quality institutions, and hence uninformative about reporting quality (Ball, 2006). One concern then is that discretion in reporting standards and lack of enforcement make it difficult and very costly for investors to figure out the extent to which firms implement serious changes around IFRS adoption. Consequently, we may observe market reactions that are fairly similar across firms. We realize that, in this case, our inferences are limited because we do not know whether the proposed proxies do not characterize meaningful differences in firms’ IFRS adoptions, or whether investors simply cannot (or do not) differentiate between serious and label adopters. Finally, we note that it is difficult to predict market reactions to label adoptions even if investors are able to discern between serious and label firms. One prediction is that the effect around label adoptions is essentially zero or negligible. Alternatively, the market reaction could be adverse, as a label adoption makes it apparent to investors that a firm is unwilling to commit to more transparency. Moreover, label adoptions may increase investor uncertainty, for example, because the change in standards makes it harder to forecast future earnings. But if markets can unravel label adoptions and react unfavorably, one wonders why firms choose such adoptions in the first place. One potential explanation is prior evidence that managers sometimes engage in seemingly strategic reporting behavior, despite the fact that markets unravel their accounting choices and price them accordingly (e.g., Watts and Zimmerman, 1986; Kothari, 2001). Another explanation is that firms perceive a trend towards IFRS and simply jump on the bandwagon, but adopt IFRS in the least costly way. Allowing for these possibilities, we do not explicitly sign our expectation for the effects of label adoptions relative to local GAAP firms. 10 2.2 Related Studies Despite the large and increasing number of international reporting studies, there is not much research that exploits or analyzes heterogeneity in the economic consequences of voluntary IFRS adoptions. Generally, voluntary IFRS studies examine the effects on capital-market metrics (e.g., liquidity or cost of capital), the properties of reported accounting numbers (or reporting quality), or various market participants (e.g., analysts or institutional investors).6 In the first category, Leuz and Verrecchia (2000) examine German firms that adopt IAS or U.S. GAAP and find that those firms exhibit lower bid-ask spreads and higher turnover compared with German GAAP firms. Cuijpers and Buijink (2005) find that the differences in the implied cost of capital are insignificant across local GAAP and IFRS firms in the European Union. Daske (2006) examines German firms and finds that voluntary IAS adopters exhibit a higher cost of equity capital than local GAAP firms. Comparing IAS and U.S. GAAP firms in Germany’s New Market, Leuz (2003) finds that differences in spreads, turnover, and IPO underpricing are statistically and economically insignificant. For a small sample of international firms, Karamanou and Nishiotis (2009) document positive abnormal returns around the announcement of IAS adoption and a reduction in long-run returns and the implied cost of capital. Focusing on reporting properties, Harris and Muller (1999), Ashbaugh and Olsson (2002), Bartov et al. (2005), and Hung and Subramanyam (2007) analyze the value relevance of IAS accounting numbers relative to local GAAP or U.S. GAAP numbers, and generally find mixed evidence. Barth et al. (2008 and 2009) compute a broad set of earnings quality metrics for firms 6 In addition, there are studies providing evidence on the determinants of voluntary IFRS adoptions (e.g., Leuz and Verrecchia, 2000; Ashbaugh, 2001; Tarca, 2004). More recently, there is an emerging literature examining the determinants and effects of mandatory IFRS adoption in certain countries (e.g., Hope et al., 2006; Armstrong et al., 2009; Christensen et al., 2009), or around the world (e.g., Daske et al., 2008; Florou and Pope, 2009; Gordon et al., 2009; Landsman et al., 2009). 11 using IAS and compare them to those for firms using local GAAP and U.S. GAAP, respectively. The evidence suggests that IAS reports are of higher quality than local GAAP reports, but of lower quality than U.S. GAAP reports. There are also a few studies on the reaction of market participants to voluntary IFRS adoptions. Cuijpers and Buijink (2005) find an increase in analyst following around IFRS, but the effect disappears when they attempt to control for self-selection. Ashbaugh and Pincus (2001) show that analyst forecast errors are positively related to differences in accounting standards between IFRS and various local GAAP, and that the accuracy of these forecasts improves after firms adopt IFRS. Covrig et al. (2007) document that foreign mutual fund ownership is significantly higher for IFRS adopters compared to local GAAP firms, particularly for firms from poorer information environments and with lower visibility. In sum, the literature documents various economic consequences around voluntary IFRS adoptions and, in many cases, firms reporting under IFRS appear to enjoy substantial benefits. However, these results have to be interpreted carefully. While most studies interpret their findings as being attributable to IFRS, IAS or U.S. GAAP, their research design rarely accounts for the fact that firms choose their standards. Thus, it is difficult to attribute the observed effects to the accounting standards per se. It is possible, if not likely, that the effects reflect factors that gave rise to IFRS adoption in the first place. Moreover, studies generally focus on the average effects around IFRS adoption, rather than cross-sectional differences in the effects. Heterogeneity across IFRS adoptions could be one reason for the mixed results in prior studies, but it also highlights the selection issues. Therefore, our study focuses on cross-sectional differences in the adoption effects and the drivers of this heterogeneity. 12 3. Research Design and Data Given our hypotheses, the key variables for our research design are a variable indicating when a firm has voluntarily adopted as well as a variable indicating differences and changes in firms’ reporting incentives to create our proxy for ‘serious’ and ‘label’ adopters. We also need proxies for the economic outcomes and a set of control variables. We combine these variables in the following model: EconCon = β0 + β1 IFRS + β2 Serious IFRS Adopters + Σ βj Controlsj + ε (1) where EconCon stands for three different proxies of economic outcomes (i.e., price impact, bidask spreads, and cost of capital), IFRS is a binary variable coded as ‘1’ for years in which a firm follows IFRS and ‘0’ otherwise, Serious IFRS Adopters denotes a binary classification that identifies firms with substantive changes in reporting incentives around IFRS adoption and estimates differential effects across IFRS adopters, and Controlsj denotes a set of control variables. To estimate this model, we obtain financial data from Worldscope, return, bid-ask spreads and trading volume data from Datastream, and analyst forecasts and share price data for the cost of capital estimation from I/B/E/S. The sample consists of all Worldscope firms from 1990 to 2005 for which we have the necessary data to compute the variables described below.7 3.1 IFRS Reporting and Serious vs. Label IFRS Classifications The coding of our IFRS variables involves three steps. First, we construct a firm-year panel with the binary IFRS indicator variable. Second, we determine the switch year, i.e., the point in 7 Because we study voluntary IFRS adoptions, we exclude firms with fiscal year-ends after December 30, 2005 (December 30, 2003, for Singapore). After this date, IFRS reporting became mandatory in many sample countries (e.g., in all the member states of the European Union). We further require firms to have total assets of 10 US$ million or more, and limit the sample to countries with at least one valid voluntary IFRS firm-year observation. The latter two sample selection choices do not affect the inferences from our analyses. 13 time when IFRS reporting started. Third, we construct proxies measuring changes in reporting incentives around the switch year to partition the IFRS sample into serious and label adopters. The first step, coding a binary IFRS variable, is not a trivial exercise because firms have chosen many different paths, particularly in the early days of IAS. We deliberately use a broad classification in order to capture a wide variety of adoption strategies, including firms that merely create an appearance of IFRS reporting. We begin with information from the “Accounting Standards Followed” field in Worldscope as it offers by far the largest sample. We code firmyear observations as IFRS equal to one if Worldscope indicates that the financials are based on “International Standards”, “IFRS”, or “IASC guidelines.” Appendix A describes this process in more detail and shows that the Worldscope classification has shortcomings, i.e., does not always properly identify firms that claim to follow IFRS. For this reason, we attempt to manually verify the coding of each firm-year observation, for which either Worldscope or Global Vantage indicate IFRS reporting. Towards this end, we download electronic copies of the annual reports from Thomson Research, read the relevant parts in the annual report (e.g., accounting principles’ footnote and auditors’ report), and create a hand-coded classification. In total, we are able to obtain and code 22,213 annual reports, which we use together with the accounting standard information in Global Vantage to triangulate and correct the initial Worldscope coding. This procedure gives rise to an ‘augmented’ Worldscope IFRS classification, which serves as basis for our analyses.8 Table 1 presents the distribution of IFRS and local GAAP firms in the sample by country (Panel A) and year (Panel B). The total sample consists of 70,167 firm-year observations across 8 As Appendix A illustrates, there is a substantial number of cases where the annual report and the Worldscope (or Global Vantage) coding do not coincide. Hence, we conduct sensitivity tests that use the pure Worldscope, Global Vantage, or hand-coded classification instead of the augmented file and report results in Section 4.3. 14 30 countries, of which 4,193 are coded as IFRS. There are several countries with adoption rates exceeding 20% (e.g., Austria, Czech Republic, Hungary). Germany has the largest number of IFRS firm-year observations followed by Switzerland and China. As expected, the number and the percentage of firms reporting under IFRS increases considerably over the sample period. By 2004, almost 9% of the firms in the sample have adopted some form of IFRS reporting. Similarly, the number of U.S. GAAP firms, which we identify via the same procedure as the IFRS firms, increases over time, but the total number remains relatively small. The second step involves identifying the year in which a firm switches to IFRS. As Appendix A illustrates, voluntary IFRS adoption and reporting over time takes on many different patterns. In its simplest form, a firm starts out reporting in accordance to local GAAP and at some point switches to IFRS reporting. In more complicated cases, firms switch back and forth between local GAAP and IFRS reporting multiple times, and also might switch to reporting in accordance with U.S. GAAP, for instance, as a result of a cross-listing on a U.S. exchange. We assign an IFRS switch year only if for two consecutive firm-year observations the first is classified as reporting under local GAAP, and the second as reporting under IFRS.9 Using this procedure, we identify 851 firms with a proper IFRS switch year (Appendix Table A4, Panel A), which excludes 571 firms from the cross-sectional tests using the serious/label partition because we are unable to identify a proper switch year for these firms, even though they reported under IFRS at some point during our sample period (Appendix Table A4, Panel B). 9 Our analyses focus on IFRS adoption and, thus, we do not consider a switch from U.S. GAAP to IFRS reporting as giving rise to a valid switch year. Conceptually, however, our arguments apply to voluntary adoption of U.S. GAAP as well. Thus, we report sensitivity analyses in Section 4.3 combining IFRS and U.S. GAAP reporting into a single global accounting standards variable as well as examining U.S. GAAP reporting on its own. 15 The third and final step involves the creation of partitioning variables to separate IFRS firms into serious and label adopters. This is the key research design innovation of our paper. The underlying idea is that firms’ reporting incentives largely determine the quality of reported numbers, rather than the standards themselves. Based on this logic, we attempt to identify firms that experience changes in their reporting incentives, for which IFRS adoption is more likely to be part of a commitment to transparency. Firms with no or small changes in reporting incentives are not expected to materially change their reporting behavior around IFRS adoption; they likely adopt a label. Since reporting incentives are unobservable, we construct two proxies for the underlying construct, one that is input-based and focuses on the determinants of firms’ incentives, and one that is output-based and uses actual reporting behavior based on the notion that it reflects firms’ incentives. We use stylized, binary partitions based on these two proxies to highlight the heterogeneity across IFRS adoptions and because it facilitates the interpretation of our analyses, but we realize that the differences across firms likely are gradual. Relying on economic theory and prior research, we use observable firm characteristics to create a Reporting Incentives Factor because firms that are larger, more profitable, more international, have greater financing needs, larger growth opportunities, and more dispersed ownership structures likely have stronger incentives to provide accounting reports that are informative to outside investors.10 Specifically, we estimate the first principal factor using firm size (natural logarithm of the market value in US$), financial leverage (total liabilities divided by total assets), profitability (return on assets), growth opportunities (book-to-market ratio), ownership concentration (percentage of closely held shares), and internationalization (percentage 10 These determinants are consistent with the disclosure literature (see Leuz and Wysocki, 2009, for a survey) and also the cross-listing literature (e.g., Lang et al., 2003; Doidge et al., 2004 and 2009) 16 of foreign sales out of total sales). The factor solution retains three factors. For all variables, the first and primary factor exhibits the expected loadings (i.e., is increasing in size, leverage, profitability, growth and foreign sales, and decreasing in ownership concentration), and therefore serves as our Reporting Incentives Factor.11 Our second proxy, the Reporting Behavior Score, relies on a simple characterization of actual reporting behavior, i.e., the magnitude of accruals relative to the cash flow from operations. Sloan (1996) or Bradshaw et al. (2001) show that the decomposition of earnings into accruals and operating cash flow as well as extreme accruals contain important information. Furthermore, this measure produces plausible earnings management rankings for firms around the world (Leuz et al., 2003; Wysocki, 2004). Following Leuz et al. (2003), we compute the Reporting Behavior Score as the ratio of the absolute value of accruals to the absolute value of cash flows (multiplied by -1 so that higher values indicate more transparent reporting). Scaling by the operating cash flow serves as a performance adjustment, which influences accruals absent any earnings management (Kothari et al., 2005). If Worldscope provides cash flow information, we estimate accruals as the difference between net income before extraordinary items and the cash flow from operations (roughly 55% of the sample). For the remaining firms, we estimate accruals using the indirect method as in Dechow et al. (1995).12 To improve the measurement of reporting incentives and to allow for the possibility that incentives change only slowly over time, we compute the Reporting Incentives Factor and the 11 To maximize sample size we replace missing values for ownership concentration and foreign sales with zero. Unreported sensitivity analyses show that our results do not hinge on the composition of the factor score. That is, when we re-run our analyses eliminating one-by-one each individual firm attribute (or replacing market value with total assets and dropping the book-to-market ratio), the results are not materially affected. 12 We recognize that a change in accounting standards likely has mechanical effects on the magnitude of accruals. However, as IFRS tends to be more accruals-based than local standards around the world (e.g., Hung, 2001; Ding et al., 2006), this effect likely works against our expectations. 17 Reporting Behavior Score as a rolling average over the past three years (i.e., t, t-1, t-2). We use the level of both proxies as a separate control variable for reporting incentives in some specifications. When creating the partitions of serious and label adopters, we attempt to capture changes in firms’ reporting incentives around IFRS adoption and hence use changes in the rolling average from t-3 to t+3, excluding the switch year. We then classify firms with above median changes as Serious IFRS Adopters (coded as ‘1’). We assess the construct validity of our reporting incentives measures by computing the Spearman correlation between the two proxies (=0.26 and statistically significant), and by correlating them with other related measures, for which we have only less comprehensive samples at hand. For instance, we find significantly positive correlations for the Reporting Incentives Factor and the Reporting Behavior Score with (1) a disclosure quality index created by business professionals in Germany, Switzerland and Austria (Daske and Gebhardt, 2006), (2) a Big Five auditor, and (3) the level of disclosure (measured by the number of pages of the annual reports, after adjusting for country effects, firm size and performance). Ultimately, partitions based on these two proxies should not produce consistent and sensible results if they do not capture the underlying construct. In this sense, our analyses are also a validation exercise. 3.2 Dependent Variables In studying the economic consequences of IFRS adoptions, we use proxies for market liquidity, information asymmetry, and the cost of capital. Increasing the commitment to transparency should reduce information asymmetries between investors and increase market liquidity (e.g., Diamond and Verrecchia, 1991; Verrecchia, 2001). More precise disclosures should also lower non-diversifiable estimation risk, which in turn reduces the cost of capital (e.g., Coles et al., 1995; Lambert et al., 2007). Thus, proxies for market liquidity, information 18 asymmetry, and the cost of capital should reflect, among other things, the quality of disclosures and financial reports. In addition, market-based proxies should capture differences in reporting quality more broadly, including differences in recognition, measurement and footnote disclosures. The first dependent variable is a measure of illiquidity suggested by Amihud (2002), which in turn is inspired by Kyle’s (1985) lambda. The proxy is intended to capture the price impact of trades, i.e., the ability of an investor to trade in stock without moving its price.13 We measure Price Impact as the median daily price impact over the year and follow Amihud (2002) in computing price impact as the daily absolute price change in percent divided by US$ trading volume. Higher values indicate more illiquid stocks. To avoid the misclassification of days with no or low trading activity (i.e., days potentially yielding a price impact of zero), we omit zeroreturn days from the computation of the yearly medians. The second dependent variable is the Bid-Ask Spread, which is a commonly used proxy for information asymmetry (e.g., Welker, 1995; Healy et al., 1999; Leuz and Verrecchia, 2000; Lang et al., 2009). We obtain the closing bid and ask prices for each day from Datastream and compute the daily quoted spread as the difference between the two prices divided by the mid-point. To obtain a yearly firm-year observation, we compute the median daily spread over the year. The third dependent variable is Cost of Capital. Following Hail and Leuz (2006), we compute estimates of the implied cost of equity capital using four models suggested in the literature (Claus and Thomas, 2001; Gebhardt et al., 2001; Easton, 2004; Ohlson and Juettner- Nauroth, 2005). All four models are consistent with discounted dividend valuation but rely on different earnings-based representations of this model. For each model, we substitute market 13 While market liquidity is an important construct in its own right, there is also evidence that liquidity is priced in expected returns (e.g., Amihud and Mendelson, 1986; Brennan and Subrahmanyam, 1996). 19 price and analyst forecasts from I/B/E/S into the valuation equation and back out the cost of capital as the internal rate of return that equates current stock price and the expected future sequence of residual incomes or abnormal earnings. We average over the four models to obtain a single estimate per firm-year observation. Appendix B describes the models and the cost of capital estimation in more detail.14 All dependent variables are measured as of month +10 after the fiscal year end for which we code the accounting standards. We choose this month to ensure that firms’ annual reports are publicly available and priced at the time of our computations (Hail and Leuz, 2006). For variables that are computed over an entire year, we start the computation as of month -2 through month +10 relative to the firm’s fiscal year end. Table 2, Panel A, presents descriptive statistics for the dependent variables for the sample of IFRS adopters and firms following local GAAP. The mean price impact (mean spread) for local GAAP firms is 2.62 (3.3%) versus 1.54 (1.9%) for IFRS adopters.15 3.3 Control Variables In all regression models we include industry-, country-, and year-fixed effects. Thus, our specifications control for differences in countries’ adoption rates as well as time trends in IFRS adoption. In unreported regressions, we also check that our results are robust when we include country-year-fixed effects to control for country-wide shifts in the adoption rates over time, e.g., 14 We recognize that there is a debate about the empirical validity of implied cost of capital estimates (e.g., Botosan and Plumlee, 2005; Easton and Monahan, 2005). One alternative is to use realized returns as a proxy for expected returns. However, this proxy has many drawbacks as well, especially with short time series (Elton, 1999). We therefore go down a different route and use proxies for liquidity and information asymmetry as these constructs also capture differences in reporting quality (Leuz, 2003; Lang et al., 2009). 15 We note that spread and price impact are right-skewed. As it is common in the literature to estimate microstructure models in a log-linear specification, we use the natural logarithm of these measures, which also addresses the skewness. 20 due to the announcement of mandatory IFRS reporting. Thus, our IFRS effects reflect withincountry differences relative to local GAAP firms and label adopters, respectively. In addition, we introduce binary indicators to control for firms following U.S. GAAP, having U.S. cross-listings, trading on a ‘new market’, and being a member of a major stock index (Daske et al. 2008). The U.S. cross-listing indicator is equal to one if the shares are also traded over-thecounter or listed on an exchange in the U.S. (Hail and Leuz, 2009). We separately code firms that voluntarily report under U.S. GAAP firms and do not have U.S. cross-listings. We set the new market indicator to one if shares are traded on an exchange, which specializes in technology and other high-growth stocks and has listing requirements that mandate or allow IFRS reporting. Index observations represent firms whose shares are constituents of national or international stock market indices as defined in Worldscope. In the price impact and spread regressions, we control for firm size, share turnover, and return variability (Chordia et al., 2000; Leuz and Verrechia, 2000). Firm size is the market value of equity measured as the stock price times the number of shares outstanding (in US$ millions). Share turnover is the accumulated US$ trading volume during the year divided by the market value of outstanding equity. Return variability is computed as annual standard deviation of monthly stock returns. We compute share turnover and return variability beginning in month -2 through month +10 relative to a firm’s fiscal year end, and lag the market variables by one year to mitigate any confounding effects from contemporaneous measurement. For the cost-of-capital specifications, we follow Hail and Leuz (2009) and control for firm size, financial leverage, return variability, forecast bias, and expected inflation. Size is measured as total assets, leverage is the ratio of total liabilities to total assets, and return variability is the annual standard deviation of monthly stock returns computed from month -2 through month +10 relative to the firm’s fiscal year end. We control for analyst forecast errors for two reasons. First, 21 it is possible that the adoption of IFRS impairs analysts’ ability to forecast earnings, at least during a transitional period. Second, any bias in analyst forecasts could mechanically affect our implied cost of capital estimate if markets back out the bias (e.g., Botosan and Plumlee, 2005; Hail and Leuz, 2006). We compute forecast bias as the one-year-ahead I/B/E/S analyst forecast error (mean forecast minus actual) scaled by lagged total assets. Finally, we control for inflation because analyst forecasts are expressed in nominal terms and local currency, which implies that the resulting cost of capital estimates reflect countries’ expected inflation rates. Inflation is measured as the yearly median of one-year-ahead realized monthly changes in the consumer price index in a given country.16 Table 2, Panel B, presents descriptive statistics on the control variables for the sample of IFRS adopters and for firms following local GAAP. Firms adopting IFRS are on average larger, more financially leveraged, have higher analyst forecast bias, share volatility and lower turnover. 4. Results 4.1 The Average Effects of IFRS Reporting We begin our analysis by examining the average effect of IFRS reporting on firms’ stock market liquidity and cost of capital. We use cross-sectional, time-series panel regressions, which benchmark IFRS firms against local GAAP firms and against the local GAAP history of IFRS adopters. Estimating average effects allows the comparison of our findings to prior work. Towards this end, we estimate the empirical specification outlined in equation (1), but without including the serious IFRS adopter variable. In addition, we estimate two regression specifications, one that excludes and one that includes the level of reporting incentives, measured 16 Using countries’ risk-free rates, rather than the inflation rate, yields very similar results and inferences. 22 either by the Reporting Incentives Factor or the Reporting Behavior Score. Based on the theory that incentives are a primary force shaping firms’ commitment to transparency, we expect these proxies to exhibit a negative relation with market outcomes. We report results from OLS regressions with robust standard errors that are clustered by firm.17 Table 3 presents the average effect of IFRS adoptions on the price impact of trades, bid-ask spreads and the cost of capital. In the first three regressions, the coefficients on the IFRS variable are negative but insignificant, suggesting that IFRS firms do not experience a decline in price impact around the switch to IFRS relative to local GAAP firms. The coefficients on the Reporting Incentives Factor and the Reporting Behavior Score are both negative and highly significant. This finding is consistent with the idea that these proxies capture firms’ incentives for transparent reporting, and further validates our proxies. The coefficient on voluntary U.S. GAAP reporting is positive and marginally significant, but only in Model 1. Cross-listings in the U.S., which represent a significant commitment to transparency and, among other things, require reconciliation of local GAAP numbers to U.S. GAAP, are associated with a strong reduction in price impact, consistent with Foerster and Karolyi (1999) and Baruch et al. (2007). This finding makes it unlikely that lack of power is responsible for the weak IFRS results and shows that significant commitments to transparency matter for liquidity. There are also liquidity differentials for firms on new markets (consistent with high growth, high risk firms), and for constituent firms of major stock indices (consistent with large, well established and less risky firms). The coefficients on market value, share turnover, and return variability are all highly significant and 17 In untabulated analyses, we also estimate the average effects model replacing the country- and industry-fixed effects with firm-fixed effects. This essentially controls for time-invariant and potentially unobserved differences between each individual IFRS and local GAAP firm. In this model, the IFRS variable is estimated off firms with reporting changes only. The results using the firm-fixed effects estimation are for the most part similar to those reported in the tables, if not we discuss differences in the text. 23 exhibit the predicted relations. In a firm-fixed effects specification (not tabulated), the coefficient on IFRS becomes significantly negative. This finding suggests that accounting for heterogeneity across firms is important – a message that our subsequent analyses in Section 4.2 underscore. The next three columns in Table 3 present the analyses of the bid-ask spread. Throughout, the results suggest that firms reporting under IFRS have significantly higher spreads, even after controlling for the level of reporting incentives. This result may seem surprising but it should be noted that the specifications in Table 3 include IFRS firms for which we are unable to identify the IFRS switch year (see Table A4 in the appendix). Thus, the IFRS coefficient could also reflect simple cross-sectional differences between IFRS and local GAAP firms, rather than effects around the switch to IFRS. Consistent with this explanation, the IFRS coefficient becomes significantly negative once we estimate a firm-fixed effects specification (not tabulated). This finding also underscores the importance of requiring firms to have a switch year in our subsequent analyses (Table 5). The other variables in the spread model exhibit associations that are similar to those in the price impact model. The final three columns in Table 3 report the cost of capital results. Similar to the spread regressions, the coefficients on the IFRS adoption variable are positive and significant, suggesting that IFRS firms have higher costs of capital than local GAAP firms. But unlike for the spread, this result still holds in the firm-fixed effects regressions (not tabulated). We discuss potential explanations for this result in Section 4.2. The Reporting Incentives Factor and the Reporting Behavior Score exhibit a negative and significant relation, as in the liquidity models, which is reassuring. Except in Model 2, the coefficients on U.S. listing indicate that being cross-listed in 24 the U.S. significantly lowers firms’ costs of capital, consistent with the cross-listing literature.18 All the remaining control variables are highly significant and have the expected signs. Overall, the average economic consequences associated with voluntary IFRS adoptions are decidedly mixed, that is, either insignificant, slightly negative (i.e., indicating a decline in liquidity and increase in cost of capital), or positive (for market liquidity when controlling for firm-fixed effects). One explanation for the mixed results could be substantial heterogeneity across firms, in which case examining the average effect can be quite misleading. We explore this possibility next. But regardless, our findings for the average IFRS effect are inconsistent with the notion that IFRS adoption itself (and hence for all voluntary adoptions) constitutes a commitment to increased disclosure. This is not surprising in light of our argument that firms have substantial flexibility in how they adopt IFRS, and the descriptive evidence that there are major differences in the quality of IFRS financial statements (e.g., Cairns 1999, 2000; Barth et al., 2008). The findings are also consistent with prior evidence on the role of reporting incentives in shaping the quality of firms’ reporting. The significant coefficients on our two reporting incentives proxies further highlight the importance of firms’ reporting incentives. We therefore use these variables in our subsequent analyses to explain cross-sectional variation in the effects around IFRS adoption. 4.2 Heterogeneity in the Economic Consequences across Serious and Label IFRS Adopters Assuming investors are able to evaluate the extent to which firms change their commitment to transparency around IFRS adoptions, we expect our model specified in equation (1) to indicate 18 The significantly positive coefficient on U.S. listing (and total assets) is primarily due to the inclusion of market values in the computation of the Reporting Incentives Factor. As Hail and Leuz (2009) note, if cross listing indeed reduces the cost of capital, contemporaneous market values absorb much of the effect. Consistent with this explanation, we find a significantly negative coefficient on the U.S. listing variable when we replace the market value with total assets in the computation of the Reporting Incentives Factor. 25 that serious adopters have lower price impact and bid-ask spreads, and a smaller cost of capital than label adopters. To the extent that serious IFRS adopters commit to more transparency, we also expect them to have higher market liquidity and lower costs of capital than local GAAP firms. Naturally, this prediction also hinges critically on our partitioning variables. Our results in Table 3 establish that the two reporting incentives proxies are related to market liquidity and the cost of capital in the predicted fashion. As discussed in Section 2, we do not sign our prediction with regard to the difference between label adopters and local GAAP firms. One possibility is that the switch to IFRS changes only the name of the accounting regime and hence, does not result in discernable capital-market effects. Another possibility is that market participants perceive label adoptions as a negative signal, in which case we could even observe a decline in market liquidity and an increase in the cost of capital. We begin the cross-sectional analysis with simple, univariate comparisons, which provide descriptive evidence. In these tests, we include only IFRS firms with identifiable switch years to use each IFRS adopter as its own control. We also include local GAAP firms that do not switch to IFRS to control for time-period effects. Table 4 reports median values and the number of observations for the dependent variables across label adopters, serious adopters, and local GAAP firms. As Panel A shows, serious IFRS adopters have the lowest price impact. That is, serious IFRS adopters have higher stock market liquidity than label adopters, which in turn have higher liquidity than local GAAP firms. For instance, when using the Reporting Incentives Factor, the price impact for serious adopters is 0.046 compared to 0.111 (0.185) for label adopters (local GAAP firms). These differences are all statistically significant. Panel B repeats the analyses for the bid-ask spread and yields similar insights. Spreads are lowest for serious IFRS adopters, followed by label adopters, followed by local GAAP firms. In Panel C, we compare costs of 26 capital. A split by the change in the Reporting Incentives Factor indicates that serious IFRS adopters exhibit lower costs of capital than local GAAP firms. Label adopters exhibit the highest costs of capital among the three groups. All differences are statistically significant. The results using the Reporting Behavior Score are inconclusive. Next, we conduct OLS regressions estimating equation (1) and report results in Table 5. Our main variable of interest is the indicator for serious IFRS adopters, which we predict to take on a negative sign. This variable measures the differential economic consequences between serious and label adopters, after controlling for various firm characteristics and fixed effects. The IFRS variable captures the liquidity and cost of capital effects for the label adopters. The two coefficients combined compare serious IFRS adopters to local GAAP firms, and we again expect the total difference to be negative. In Panel A, we report the results for price impact. We find that the coefficients on the serious IFRS adopter indicator are always negative and statistically significant. This finding indicates that there is substantial heterogeneity in the price impact effects around IFRS adoption, as predicted. More specifically, we find that serious IFRS adopters experience significant declines in price impact relative to label adopters. The net effect on price impact around IFRS adoption is also negative and significant for serious adopters, consistent with the notion that an increased commitment to transparency increases market liquidity. Splitting by the change in the Reporting Incentives Factor, the coefficient on IFRS is significantly positive, indicating that label adopters exhibit lower market liquidity than local GAAP firms. Using the Reporting Behavior Score to partition IFRS firms suggests no change in market liquidity for label adopters. In Model 2, we control for the level of firms’ reporting incentives (measured at t-1 for firms that switch to IFRS). As in Table 3, both proxies for reporting incentives are negatively 27 associated with price impact. More importantly, the magnitude of the coefficient on serious IFRS adoption increases in magnitude and statistical significance and hence sharpens our results. This finding shows that the effects for serious IFRS adopters and the heterogeneity around IFRS adoptions are not driven by prior differences in reporting quality. That is, it is inconsistent with the alternative hypothesis discussed in Section 2 that IFRS adoption leads to convergence of reporting quality among voluntary IFRS adopters. The results using bid-ask spreads are reported in Panel B and closely resemble those for price impact. That is, the coefficient on serious IFRS adopters is always significantly negative and so is the net effect for serious IFRS adopters relative to local GAAP firms. This suggests that serious adopters experience a significant decline in bid-ask spreads compared to label adopters and local GAAP firms, consistent with the link between market liquidity and transparency. For the label adopters, the effect is positive across the four regressions, but only significant when also controlling for the level of the Reporting Incentives Factor. In Panel C, we report the cost of capital results. Across all four models we find that serious IFRS adopters exhibit a significant decrease in the cost of capital relative to label adopters. However, the net effect on the cost of capital around IFRS adoption is significantly negative in only one case (i.e., for the Reporting Incentives Factor in Model 2). In the remaining cases, the combined coefficient is not distinguishable from zero. Thus, there is at best only weak evidence that serious IFRS adopters experience a decline in the cost of capital. Furthermore, the IFRS main effect is positive and significant, suggesting that label adopters experience an increase in their cost of capital. The result implies that the average cost of capital effects shown in Table 3 stem primarily from firms classified as label adopters. As discussed in Section 2, there are several possible explanations for this effect. 28 In sum, the cross-sectional analyses confirm our main hypothesis that there is substantial heterogeneity in economic outcomes around IFRS adoptions, and that these differences can be explained in part by the degree to which firms experience substantial changes in reporting incentives (and are not muted by prior differences in reporting quality). This evidence casts doubt that the capital-market effects around IFRS adoption really stem from the switch in standards per se. In addition, the findings suggest that markets can successfully distinguish between different adoption types and reward only firms that credibly commit to more transparency. 4.3 Sensitivity Analyses In this section, we assess the sensitivity of our tests to various research design choices and report results in Table 6. For brevity, we tabulate only the coefficients and t-statistics of the two IFRS variables but, unless stated otherwise, estimate the regressions using Model 2 in Table 5. First, in Panel A, we extend the analysis using three additional variables to measure economic outcomes around IFRS adoptions (e.g., Daske et al., 2008), namely Total Trading Costs (equal to the roundtrip transaction costs drawn from Lesmond et al., 1999), Zero Returns (equal to the yearly proportion of trading days without stock price movements), and firm value defined as Tobin’s q. For a description of the variables and regression models see the notes to Table 6. The results using total trading costs are very similar to the liquidity results presented in Table 5. Serious IFRS adopters have lower trading costs than label adopters. The results are weaker using zero return days, but point in the same direction. For both variables, the net effect for serious adopters is significantly negative, suggesting that serious adopters have significantly lower trading costs and fewer zero return days than their local GAAP peers. The Tobin’s q results are consistent with the cost of capital effects in Table 5. In particular, serious adopters 29 experience larger effects and a net benefit when using changes in the Reporting Incentives Factor to partition IFRS adopters. The results are weaker for the other partitioning variable. Second, we evaluate the sensitivity of our results to the accounting standards classification used to identify IFRS (and U.S. GAAP) adopters. As explained before and detailed in Appendix A, we use an augmented Worldscope accounting standards classification that incorporates extensive hand-coding based on firms’ annual reports. In Panel B, we re-run our analyses using (1) the original Worldscope classification, (2) the accounting standards classification based on Global Vantage, and (3) the sub-sample of firms for which we are able to classify standards based on firms’ annual reports. We report only the results for price impact as the dependent variable, but the inferences are similar using bid-ask spreads and the costs of capital. Regardless of the accounting standards classification, firms with a substantial change in reporting incentives around IFRS adoption experience an increase in liquidity relative to label adopters and local GAAP firms in all specifications. Label adopters generally exhibit the opposite effect, although it is not always significant. The fact that our results are not sensitive to the accounting standards classification may seem surprising in light of the substantial differences across classifications documented in Appendix A. Note, however, that our partitions into label and serious adopters are independent of the accounting standards classification. Thus, if our partitions work, firms that are misclassified as IFRS reporters do not necessarily hurt our tests. In fact, the partitions likely “correct” for misclassifications by putting firms into the label adopter category. In the remainder of Panel B, we extend our analysis by including voluntary U.S. GAAP adoptions. Combining IFRS and U.S. GAAP adoptions, or alternatively, considering voluntary (non-U.S. listed) U.S. GAAP adopters yields essentially the same results as before. Market liquidity is highest for serious adopters, while the differences between local GAAP firms and 30 label adopters are usually not significant. These results highlight that our prior findings are not specific to voluntary IFRS adoptions. Third, we gauge the effect of sample composition on our results, and re-estimate the models (1) eliminating sample restrictions and using the largest possible sample comprising firms from 50 countries and a maximum of 135,538 firm-year observations, (2) our base sample but excluding the two largest countries, i.e., the U.K. and Canada, (3) a random benchmark sample consisting of up to 150 non-IFRS adopting firms per country and year, and (4) including only firm-years from firms that eventually switch to IFRS during the sample period. Results reported in Panel C indicate that our analyses remain largely unaffected by alternative sample compositions. Even for our smallest sample with just 6,329 firm-year observations (i.e., IFRS firms only), the results are consistent with the earlier findings, albeit at lower significance levels. Finally, when we re-estimate the cross-sectional analyses using firm-fixed effects and hence, eliminate time-invariant unobserved firm attributes as potentially confounding factors, the results continue to hold and the inferences are essentially the same as in Table 5. In sum, the evidence in Table 6 increases the confidence in our results. 5. Conclusion This paper examines the economic consequences of voluntary IFRS adoptions around the world. In contrast to prior work, we focus on the heterogeneity in the consequences, recognizing that firms differ in their motivations to adopt IFRS. Some firms may adopt IFRS merely as a label without making material changes to their reporting policies, while others may adopt IFRS as part of a broader strategy to strengthen their commitment to transparency. To illustrate this heterogeneity around IFRS adoptions, we create two binary partitions that attempt to capture major changes in firms’ reporting incentives and split IFRS firms into ‘label’ 31 and ‘serious’ adopters. The first measure extracts firms’ reporting incentives from various firm attributes using factor analysis. The second proxy is based on a simple characterization of firms’ actual reporting behavior related to earnings quality. Using a large global panel of IFRS adoptions from 1990 to 2005, we document that the economic consequences around IFRS adoptions depend on firms’ reporting incentives and the extent to which these incentives change around IFRS adoption. Specifically, there is little evidence that, on average, IFRS adoptions are associated with an increase in market liquidity or a decline in the cost of capital, after controlling for known determinants of these measures. If anything, the effects seem to go in the opposite direction. This evidence is inconsistent with the notion that IFRS itself, and hence for all voluntary adoptions, constitutes a commitment to increased disclosure. This result is consistent with the notion that firms have substantial discretion in how they adopt and apply IFRS (or any other set of accounting standards). Once we estimate the market effects separately for serious and label adopters, we find substantial differences around IFRS adoptions. We find that serious IFRS adopters experience economically and statistically significant declines in the price impact of trades, bid-ask spreads and costs of capital relative to label adopters, but also relative to local GAAP firms. We also show that these effects do not reflect prior differences in reporting incentives or quality. By documenting that changes in firms’ reporting incentives characterize capital-market effects around IFRS adoptions, we show that one has to exercise caution in attributing these effects to the standards per se. This evidence contributes to the literature on the importance of firms’ incentives for reporting outcomes and their capital-market effects. We extend this literature by highlighting the importance of firm-level incentives. In addition, our study shows that markets can 32 differentiate across IFRS adoptions. This result is important as there is considerable concern that the ongoing global movement towards a single set of accounting standards masks the heterogeneity in actual reporting practices and hence makes it harder for investors to evaluate firms’ reporting quality. Our results do not support this concern. Finally, we caution the reader that our results should not be interpreted as indicating that ‘serious’ IFRS adopters experience the documented effects because of IFRS adoption or because they comply strictly with IFRS. Our classification indicates that these firms experience a substantial change in their reporting incentives, which likely makes them serious about more transparent reporting. For such firms, IFRS adoption is likely part of a broader strategy to increase the commitment to transparency, and the market reaction around IFRS adoption likely reflects the entire commitment strategy. Thus, a key message of our paper is that in order to document the impact of IFRS reporting per se, one has to push harder on the identification of IFRS effects (and their separation from reporting incentive effects). 33 Appendix A: IFRS and Financial Reporting Standards Classifications This appendix describes our coding of firm-year observations from firms following IFRS, U.S. GAAP or local GAAP, and compares accounting standards classifications across different data sources. It also documents firms’ reporting patterns over time, and how we identify the year of voluntary IFRS adoption. We use three primary data sources to construct our panel of firms’ accounting standards: (1) Thomson Financial Worldscope (WS), (2) Compustat Global Vantage (GV), and (3) an extensive manual review of annual reports collected through Thomson Research. WS serves as starting point because its coverage is by far the most comprehensive. In addition, it is directly linked to Datastream thereby reducing the potential for mismatches from combining accounting data with price data. Using the information on accounting standards followed (field 07536), we classify firm-year observations into the three categories IFRS, U.S. GAAP and local GAAP according to Panel A of Table A1. For sensitivity purposes and to assess the quality of commercially available databases, we also classify the same set of observations applying the GV coding scheme in Panel B that is based on the accounting standards information in field “ASTD”.19 To check whether the information contained in WS (and GV) meets our purposes (i.e., identification of firms claiming to report under IFRS), we next conduct an extensive manual data collection and classification. In a first step, we identify all potential voluntary IFRS or U.S. GAAP adopters, i.e., firms with at least one firm-year coded as IFRS or U.S. GAAP in either WS or GV before IFRS adoption became mandatory (35,633 firm-years). We then gather the time- 19 One of the drawbacks of commercial databases is that they attempt to capture many different reporting practices around the world, but often at the expense of consistency through time or across countries. Furthermore, the categories often do not provide clear distinctions between local GAAP and IFRS, as is needed for our study. For instance, category 02 “International standards” in WS comprises not only IFRS observations, but also firms following other non-local, non-U.S. standards (e.g., H.K. GAAP, U.K. GAAP). This problem is even more pronounced in GV as it has only three categories dedicated to international standards. 34 series of annual reports through Thomson Research and were able to collect 22,213 documents in electronic image format. Next, based on a manual review of the accounting principles’ footnote and the auditors’ report, we create our own accounting standards classification (Panel C). The idea behind this classification is to rely on firms’ self-proclaimed financial reporting practices. This leads to six IFRS (U.S. GAAP) reporting categories ranging from the exclusive use of IFRS (U.S. GAAP) for consolidated financial statements to reporting under local GAAP together with a reconciliation of net income or shareholders’ equity to IFRS (U.S. GAAP). Note that we classify firms as local when they only adopt individual IFRS or U.S. GAAP standards (e.g., for leasing or segment reporting). Finally, we complete firms’ time-series by filling in cases with missing individual annual reports, utilizing all data sources at hand (i.e., WS, GV, annual reports). This procedure results in a total hand-coded sample of 27,589 firm-year observations. The main purpose of this massive hand-collection is to assess the suitability of commercially available databases for our research question. To gauge the effect of potential misclassifications, we tabulate firm-years across different classification schemes and report the number of observations and percentages in Table A2. For ease of comparison, we use the base sample from our main analyses (n=70,167).20 In Panel A we compare the classifications across WS and GV. For our comprehensive coding for IFRS, we find that 2.5% of all firm-year observations are classified as IFRS in WS, but as local GAAP in GV (compared to 1.2% the other way round). Thus, the two data sources provide contradictory information on more than every second IFRS firm-year observation. When we limit the WS coding to categories 02 and 23 and the GV coding to “DI” (labeled “Stricter Coding for IFRS” in Panel A), the contradiction rate drops to about 48% of the IFRS firm-year observations. Note that firm coverage in WS is larger by about one 20 In unreported analyses we confirm that the results of the cross-tabulation is very similar when using the entire hand-coded sample instead (n=27,589). 35 fourth compared to GV. Panel B reports results from comparing the hand-coded classification to WS and GV. Neither of the two commercial databases clearly dominates, but both exhibit substantial classification differences compared to our manual review of annual reports, as indicated by large numbers of observations off the diagonals. Our IFRS hand-coding disagrees in 24% (25%) of the cases with WS (GV), computed as the fraction of off-diagonal observations relative to the observations for which we have annual report data and WS (GV) provides a classification. Panel C reports Pearson correlations between the three classifications. As a further validity check we compare the three coding schemes on a country-by-country basis. In Table A3 we tabulate the numbers and proportions of IFRS and U.S. GAAP adopters for all countries with a minimum of 20 observations and total assets available in a given year. The table again highlights the larger firm coverage in WS, which identifies more than twice as many IFRS observations than GV (13,001 IFRS firm-years versus 6,227). The hand-coded sample consists of 8,399 IFRS firm-year observations. It further reveals that the proportion of IFRS adopters at the individual country-level varies substantially. For instance, the percentage of IFRS adopters in Italy is as high as 78.7% according to WS but only 0.2% based on GV. The handcoding, to which we added observations that are uniformly classified as local under WS and GV to make it more comparable in size and to serve as a benchmark, produces a proportion of 25%. As a result of the above validity checks, we conduct our main analyses using an augmented WS accounting standards classification in which our hand-coded data overrides conflicting WS information. This yields 3,183 substitutions. We manually ensure that the splicing of the two data series does not introduce artificial IFRS adoption years. For completeness, we also report results using alternative accounting standards classifications in the sensitivity tests section. Finally, Table A4 illustrates the identification of the IFRS switch years, and indicates the number of unique firms and firm-year observations in each category for our base sample. 36 Appendix B: Implied Cost of Equity Capital Models B.1 Overview and Model-specific Assumptions Claus and Thomas (2001): T CT CT t T CT t T T CT t CT t t t r g r x r bv g r x r bv P bv 1 ˆ 1 1 ˆ + 1 1 + 1 Model-specific assumptions: This is a special case of the residual income valuation model. It uses actual book values per share and forecasted earnings per share up to five years ahead to derive the expected future residual income series. We define residual income as forecasted earnings per share less a cost of capital charge for beginning of fiscal year book value of equity per share. We assume clean surplus, i.e., future book values are imputed from current book values, forecasted earnings and dividends. Dividends are set equal to a constant fraction of forecasted earnings. At time T = 5, it is assumed that (nominal) residual income grows at rate g equal to the expected inflation. As a proxy for g, we use the (annualized) median of country-specific, one-year-ahead realized monthly inflation rates. Note that g sets a lower bound to the cost of capital estimates. Gebhardt, Lee, and Swaminathan (2001): T GLS GLS t T GLS t T T GLS t GLS t t t r r x r bv r P bv x r bv 1 ˆ 1 ˆ 1 + 1 + 1 Model-specific assumptions: This is a special case of the residual income valuation model. It uses actual book values per share and forecasted earnings per share up to three years ahead to impute future expected residual income for an initial three-year period. We assume clean surplus, i.e., future book values are imputed from current book values, forecasted earnings and dividends. Dividends are set equal to a constant fraction of forecasted earnings. After the explicit forecast period of three years, the residual income series is derived by linearly fading the forecasted accounting return on equity to the industry-specific median return. We compute the historic three-year average return on equity in a given country and year based on the industry classification in Campbell (1996). Negative yearly target returns are replaced by countryindustry medians. From T = 12 on residual income is assumed to remain constant. Ohlson and Juettner-Nauroth (2005): Pt xˆ t 1 rOJ gst rOJ dˆ t 1 xˆ t 1 glt rOJ glt Model-specific assumptions: This is a special case of the abnormal earnings growth valuation model developed by Ohlson and Juettner-Nauroth (2005). It uses one-year ahead forecasted earnings and dividends per share as well as forecasts of short-term and long-term abnormal earnings growth. Dividends are set equal to a constant fraction of forecasted earnings. Following Gode and Mohanram (2003), the short-term growth rate gst is estimated as the average between the forecasted percentage change in earnings from year t+1 to t+2 and the five-year growth forecast provided by financial analysts on I/B/E/S. The model requires a positive change in forecasted earnings to yield a numerical solution. The long-term earnings growth rate glt incorporates the assumption that growth in abnormal earnings per share beyond year t+1 equals the expected rate of inflation. We use the (annualized) country-specific median of oneyear- ahead realized monthly inflation rates. Note that glt sets a lower bound to the cost of capital estimates. Modified PEG ratio model by Easton (2004): Pt xˆ t 2 rPEG dˆ t 1 xˆ t 1 rPEG 2 Model-specific assumptions: This is a special case of the abnormal earnings growth valuation model developed by Ohlson and Juettner-Nauroth (2005). It uses one-year ahead and two-year ahead earnings per share forecasts as well as expected dividends per share in period t+1 to derive a measure of abnormal earnings growth. Dividends are set equal to a constant fraction of forecasted earnings. The model embeds the assumption that growth in abnormal earnings persists in perpetuity after the initial period. Note that it requires positive changes in forecasted earnings (including re-invested dividends) to yield a numerical solution. 37 Notes: Pt = Market price of a firm’s stock at date t bvt = Book value per share at the beginning of the fiscal year bvt = Expected future book value per share at date t+ , where bvt bvt 1 xˆt dˆt xˆt = Expected future earnings per share for period (t+–1, t+) using either explicit analyst forecasts or future earnings derived from growth forecasts g, gst, and glt, respectively dˆt = Expected future net dividends per share for period (t+–1, t+), derived from the dividend payout ratio times the earnings per share forecast xˆt g,gst ,glt = Expected (perpetual, short-term or long-term) future growth rate rCT ,rGLS ,rOJ ,rPEG = Implied cost of capital estimates calculated as the internal rate of return solving the above valuation equations, respectively B.2 General Assumptions and Data Requirements The computation of the implied cost of capital estimates is based on several data sources and requires a series of general assumptions. For an observation to be included we require current stock prices (Pt), analyst earnings per share forecasts for two periods ahead ( xˆ t 1 and xˆ t 2), and either forecasted earnings per share for period t+3 ( xˆ t 3) or an estimate of long-term earnings growth (ltg). We obtain this information from the I/B/E/S database. If explicit earnings per share forecasts for the periods t+3 through t+5 are missing, we apply the following relation: xˆ t xˆ t 1 1 ltg. Alternatively, if long-term growth rates are missing, we impute ltg from the percentage change in forecasted earnings per share between periods t+2 and t+3. We only use positive earnings forecasts and growth rates. All estimates are mean analyst consensus forecasts. Stock prices and analyst forecasts are measured as of month +10 after the fiscal year end. This time lag is chosen to ensure that financial data, especially earnings and book values of equity, are publicly available and impounded in firms’ prices at the time we compute the cost of capital. Given this choice, analyst forecasts represent estimates for fiscal years ending in just 2, 14, 26, etc. months. The valuation models, on the other hand, assume discounting for a full year, i.e., they start at the beginning of the fiscal year. For consistent discounting, we first move the 38 month +10 prices (which contain the information available at the time of forecasting) back to the beginning of the fiscal year using the imputed cost of capital and then use full-year discounting. Net dividends ( dˆ t ) are forecasted up to the finite forecast horizon as a constant fraction of expected future earnings per share. We define the dividend payout ratio (kt) as the historic threeyear average for each firm. If kt is missing or outside the range of zero and one, we replace it by the country-year median payout ratio. We use the country-specific median of one-year-ahead realized monthly inflation rates as our proxy for long-run growth (g or glt) in the terminal value computations. Negative values are replaced by the country’s historical inflation rate, estimated as the median of the monthly inflation rates over the 1980 to 2005 period, because deflation cannot persist forever. We obtain all financial data (bvt and kt) from Worldscope. Inflation data are gathered from the Datastream and World Bank databases. Since most of the valuation models do not have a closed form solution, we use an iterative procedure to determine the internal rate of return. This numerical approximation identifies the annual firm-specific discount rate that equates Pt to the right-hand side of the equity valuation model. We stop iterating if the imputed price falls within a 0.001 difference of its actual value. Implied cost of equity capital estimates are restricted to be positive and set to missing otherwise. 39 References Amihud, Y. 2002. Illiquidity and stock returns. Journal of Financial Markets 5: 31–56. Amihud, Y., and H. Mendelson. 1986. Asset pricing and the bid-ask spread. Journal of Financial Economics 17, 223–249. Armstrong, C., M. Barth, A. Jagolinzer, and E. Riedl. 2009. Market reaction to events surrounding the adoption of IFRS in Europe. The Accounting Review: forthcoming. Ashbaugh, H. 2001. Non-US firms’ accounting standard choices. Journal of Accounting and Public Policy 20: 129–153. Ashbaugh, H., and M. Pincus. 2001. Domestic accounting standards, International Accounting Standards, and the predictability of earnings. Journal of Accounting Research 39: 417–434. Ashbaugh, H., and P. Olsson. 2002. An exploratory study of the valuation properties of crosslisted firms’ IAS and U.S. GAAP earnings and book values. The Accounting Review 77: 107– 126. Ball, R. 2001. Infrastructure requirements for an economically efficient system of public financial reporting and disclosure. Brookings-Wharton Papers on Financial Service: 127–169. Ball, R. 2006. IFRS: Pros and cons for investors. Accounting and Business Research, International Accounting Policy Forum: 5–27. Ball, R., A. Robin, and J. Wu. 2003. Incentives versus standards: properties of accounting income in four East Asian countries. Journal of Accounting and Economic 36: 235–270. Ball, R., and L. Shivakumar. 2005. Earnings quality in U.K. private firms. Journal of Accounting and Economics 39: 83–128. Ball, R., S.P. Kothari, and A. Robin. 2000. The effect of international institutional factors of properties of accounting earnings. Journal of Accounting and Economics 29: 1–51. Baruch S., G. A. Karolyi, and M. Lemmon. 2007. Multi-market trading and liquidity: Theory and evidence. The Journal of Finance 62: 2169–2200. Barth, M. 2008. Global financial reporting: Implications for U.S. academics. The Accounting Review 83: 1159–1179. Barth, M., W. Landsman, and M. Lang. 2008. International Accounting Standards and accounting quality. Journal of Accounting Research 46: 467–498. Barth, M., W. Landsman, M. Lang, and C. Williams. 2009. Accounting quality: International Accounting Standards and U.S. GAAP. Working paper, Stanford University and University of North Carolina. Bartov, E., S. Goldberg, and M. Kim. 2005. Comparative value relevance among German, U.S. and International Accounting Standards: A German stock market perspective. Journal of Accounting, Auditing and Finance 20: 95–119. Botosan, C., and M. Plumlee. 2005. Assessing alternative proxies for the expected risk premium. The Accounting Review 80: 21–53. Bradshaw, M., S. Richardson, and R. Sloan. 2001. Do analysts and auditors use information in accruals? Journal of Accounting Research 39: 45–74. 40 Bradshaw, M., and G. Miller. 2008. Will harmonizing accounting standards really harmonize accounting? Evidence from non-U.S. firms adopting U.S. GAAP. Journal of Accounting, Auditing and Finance 23: 233–264. Brennan, M., and A. Subrahmanyam. 1996. Market microstructure and asset pricing: On the compensation for illiquidity in stock returns. Journal of Financial Economics 41: 441–464. Burgstahler, D., L. Hail, and C. Leuz. 2006. The importance of reporting incentives: Earnings management in European private and public firms. The Accounting Review 81: 983–1017. Cairns, D. 1999. International Accounting Standards Survey, LLP Professional Publishing. Cairns, D. 2000. International Accounting Standards Survey. Campbell, J. 1996. Understanding risk and return. Journal of Political Economy 104: 298–345. Christensen, H., E. Lee, and M. Walker. 2009. Do IFRS reconciliations convey information? The effect of debt contracting. Journal of Accounting Research 47: 1167–1248. Chordia, T., R. Roll, and A. Subrahmanyam. 2000. Co-movements in bid-ask spreads and market depth. Financial Analysts Journal 56: 23–27. Claus, J., and J. Thomas. 2001. Equity premia as low as three percent? Evidence from analysts’ earnings forecasts for domestic and international stock markets. The Journal of Finance 56: 1629–1666. Coffee, J. 1999. The future as history: The prospects for global convergence in corporate governance and its implications. Northwestern University Law Review 93: 641–708. Coles, J., U. Loewenstein, and J. Suay. 1995. On equilibrium pricing under parameter uncertainty. Journal of Financial and Quantitative Analysis 30: 347–364. Covrig, V., M. DeFond, and M. Hung. 2007. Home bias, foreign mutual fund holdings, and the voluntary adoption of International Accounting Standards. Journal of Accounting Research 45: 41–70. Cuijpers, R., and W. Buijink. 2005. Voluntary adoption of non-local GAAP in the European Union: A study of determinants and consequences. European Accounting Review 14: 487– 524. Daske, H. 2006. Economic benefits of adopting IFRS or US-GAAP: Have the expected costs of equity capital really decreased? Journal of Business Finance and Accounting 33: 329–373. Daske, H., and G. Gebhardt. 2006. International Financial Reporting Standards and experts’ perceptions of disclosure quality. Abacus 42: 461–498. Daske, H., L. Hail, C. Leuz, and R. Verdi. 2008. Mandatory IFRS reporting around the world: Early evidence on the economic consequences. Journal of Accounting Research 46: 1085– 1142. Dechow, P., R. Sloan, and A. Sweeney. 1995. Detecting earnings management. The Accounting Review 70: 193–225. Dechow, P., and D. Skinner. 2000. Earnings management: Reconciling the views of accounting academics, practitioners, and regulators. Accounting Horizons 14: 235–250. Diamond, D., and R. Verrecchia. 1991. Disclosure, liquidity, and the cost of capital. The Journal of Finance 46: 1325–1359. 41 Ding, Y., O.-K. Hope, T. Jeanjean, and H. Stolowy. 2006. Differences between domestic accounting standards and IAS: Measurement, determinants and implications. Journal of Accounting and Public Policy 26: 1–38. Doidge, C., G. A. Karolyi, and R. Stulz. 2004. Why are foreign firms listed in the U.S. worth more? Journal of Financial Economics 71: 205–238. Doidge, C., G. A. Karolyi, K. Lins, D. Miller, and R. Stulz. 2009. Private benefits of control, ownership, and the cross-listing decision. The Journal of Finance 64: 425–466. Easton, P., and S. Monahan. 2005. An evaluation of accounting-based measures of expected returns. The Accounting Review 80: 501–538. Easton, P. 2004. PE ratios, PEG ratios, and estimating the implied expected rate of return on equity capital. The Accounting Review 79: 79–95. Elton, E. 1999. Expected return, realized return and asset pricing tests. The Journal of Finance 54: 1199–1220. Florou, A., and P. Pope. 2009. Mandatory IFRS adoption and investor asset allocation decisions. Working paper, Lancaster University. Foerster, S., and G. A. Karolyi. 1999. The effects of market segmentation and investor recognition on asset prices: Evidence from foreign stocks listing in the United States. The Journal of Finance 54: 981–1013. Gebhardt, W., C. Lee, and B. Swaminathan. 2001. Toward an implied cost of capital. Journal of Accounting Research 39: 135–176. Gode, D., and P. Mohanram. 2003. Inferring the cost of capital using the Ohlson-Juettner model. Review of Accounting Studies 8: 399–431. Gordon, E., B. Jorgensen, and C. Linthicium. 2009. Could IFRS replace US GAAP? A comparison of earnings attributes and informativeness in the US market. Working paper, Temple University. Hail, L., and C. Leuz. 2006. International differences in the cost of equity capital: Do legal institutions and securities regulation matter? Journal of Accounting Research 44: 485–531. Hail, L., and C. Leuz. 2009. Cost of capital effects and changes in growth expectations around U.S. cross-listings. Journal of Financial Economics 93: 428–454. Hail, L., C. Leuz, and P. 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. Harris, M., and K. Muller III. 1999. The market valuation of IAS versus US-GAAP accounting measures using Form 20-F reconciliations. Journal of Accounting and Economics 26: 285– 312. Healy, P., A. Hutton, and K. Palepu. 1999. Stock performance and intermediation changes surrounding sustained increases in disclosure. Contemporary Accounting Research 16: 485– 520. Healy, P., and J. Wahlen. 1999. A review of the earnings management literature and its implications for standard setting. Accounting Horizons 13: 365–383. Hope, O.-K., J. Jin, and T. Kang. 2006. Empirical evidence on jurisdictions that adopt IFRS. Journal of International Accounting Research 5: 1–20. 42 Hung, M. 2001. Accounting standards and value relevance of financial statements: An international analysis. Journal of Accounting and Economics 30: 401–420. Hung, M., and K. Subramanyam. 2007. Financial statement effects of adopting International Accounting Standards: The case of Germany. Review of Accounting Studies 12: 623–657. Karamanou, I., and G. Nishiotis. 2009. Disclosure and the cost of capital: Evidence from the market’s reaction to firm voluntary adoption of IAS. Journal of Business Finance and Accounting 36: 793–821. Kothari, S.P. 2001. Capital markets research in accounting. Journal of Accounting and Economics 31: 105–231. Kothari, S.P., A. Leone, and C. Wasley. 2005. Performance matched discretionary accrual measures. Journal of Accounting and Economics 39: 163–197. Kyle, A. 1985. Continuous auctions and insider trading. Econometrica 53: 1315–1335. Lambert, R., C. Leuz, and R. Verrecchia. 2007. Accounting information, disclosure, and the cost of capital. Journal of Accounting Research 45: 385–420. Landsman, W., E. Maydew, and J. Thornock. 2009. The information content of annual earnings announcements and mandatory adoption of IFRS. Working paper, University of North Carolina at Chapel Hill. Lang, M., J. Smith Raedy, and W. Wilson. 2006. Earnings management and cross listing: Are reconciled earnings comparable to US earnings? Journal of Accounting and Economics 42: 255–283. Lang, M., K. Lins, and M. Maffett. 2009. Transparency, liquidity, and valuation: International evidence. Working paper, University of North Carolina at Chapel Hill. Lesmond, D., J. Ogden, and C. Trzcinka. 1999. A new estimate of transaction costs. Review of Financial Studies 12: 1113–1141. Leuz, C. 2003. IAS versus US GAAP: Information asymmetry-based evidence from Germany’s New Market. Journal of Accounting Research 41: 445–472. Leuz, C. 2006. Cross listing, bonding and firms’ reporting incentives: A discussion of Lang, Raedy and Wilson. Journal of Accounting and Economics 42 (2006): 285–299. Leuz, C., and R. Verrecchia. 2000. The economic consequences of increased disclosure. Journal of Accounting Research 38: 91–124. Leuz, C., D. Nanda, and P. Wysocki. 2003. Earnings management and investor protection: an international comparison. Journal of Financial Economics, 69: 505–527. Leuz, C., and P. Wysocki. 2008. Economic consequences of financial reporting and disclosure regulation: A review and suggestions for future research. Working paper, University of Chicago. Ohlson, J., and B. Juettner-Nauroth. 2005. Expected EPS and EPS growth as determinants of value. Review of Accounting Studies 10: 349–365. Sloan, R. 1996. Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review 71: 289–315. Stulz, R. 1999. Globalization, corporate finance and the cost of capital. Journal of Applied Corporate Finance 12: 8–25. 43 Tarca, A. 2004. International convergence of accounting practices: Choosing between IAS and US GAAP. Journal of International Financial Management and Accounting 15: 60–91. Verrecchia, R. 2001. Essays on disclosure. Journal of Accounting and Economics 32: 91–180. Watts, R., and J. Zimmermann. 1986. Positive Accounting Theory, London. Welker, M. 1995. Disclosure policy, information asymmetry, and liquidity in equity markets. Contemporary Accounting Research 11: 801–827. Wysocki, P. 2004. Discussion of ultimate ownership, income management, and legal and extralegal institutions. Journal of Accounting Research 42: 463–474. TABLE 1 Sample Composition by Country and Year Panel A: Accounting Standards, Listing Status and Index Membership by Country IFRS U.S. GAAP U.S. Listing New Markets Index Member Country Unique Firms Firm- Years Firm- Years % Firm- Years % Firm- Years % Firm- Years % Firm- Years % Australia 983 5,412 66 1.2 3 0.1 703 13.0 58 1.1 2,892 53.4 Austria 131 974 223 22.9 20 2.1 93 9.5 46 4.7 221 22.7 Belgium 171 1,160 78 6.7 8 0.7 28 2.4 0 0.0 303 26.1 Bermuda 18 66 12 18.2 25 37.9 9 13.6 12 18.2 12 18.2 Canada 1,279 7,598 7 0.1 122 1.6 1,478 19.5 35 0.5 2,837 37.3 China 1,121 3,362 728 21.7 21 0.6 127 3.8 6 0.2 89 2.6 Czech Republic 55 159 42 26.4 0 0.0 8 5.0 0 0.0 21 13.2 Denmark 239 2,129 90 4.2 2 0.1 16 0.8 0 0.0 309 14.5 Finland 152 919 28 3.0 0 0.0 24 2.6 0 0.0 226 24.6 France 988 6,287 136 2.2 44 0.7 210 3.3 0 0.0 2,521 40.1 Germany 724 3,020 781 25.9 372 12.3 122 4.0 738 24.4 610 20.2 Greece 331 2,131 26 1.2 4 0.2 36 1.7 0 0.0 616 28.9 Hong Kong 730 5,029 61 1.2 13 0.3 693 13.8 53 1.1 1,342 26.7 Hungary 35 207 125 60.4 6 2.9 31 15.0 0 0.0 12 5.8 Israel 152 629 18 2.9 158 25.1 69 11.0 26 4.1 190 30.2 Italy 207 1,117 381 34.1 0 0.0 26 2.3 27 2.4 217 19.4 Luxembourg 29 142 32 22.5 17 12.0 23 16.2 4 2.8 67 47.2 The Netherlands 288 2,365 46 1.9 73 3.1 258 10.9 41 1.7 398 16.8 Norway 240 1,624 15 0.9 12 0.7 127 7.8 0 0.0 366 22.5 Pakistan 86 583 44 7.5 0 0.0 0 0.0 0 0.0 449 77.0 Peru 54 282 74 26.2 4 1.4 38 13.5 0 0.0 98 34.8 Poland 104 527 29 5.5 2 0.4 18 3.4 0 0.0 97 18.4 Portugal 100 599 10 1.7 0 0.0 24 4.0 0 0.0 273 45.6 Russian Federation 36 153 48 31.4 14 9.2 87 56.9 0 0.0 6 3.9 Singapore 456 2,421 25 1.0 12 0.5 123 5.1 1,928 79.6 780 32.2 South Africa 494 2,762 88 3.2 1 0.0 370 13.4 0 0.0 851 30.8 Sweden 370 2,415 42 1.7 0 0.0 162 6.7 7 0.3 380 15.7 Switzerland 283 2,081 749 36.0 30 1.4 37 1.8 8 0.4 291 14.0 Turkey 193 1,100 180 16.4 0 0.0 25 2.3 0 0.0 656 59.6 United Kingdom 2,155 12,914 9 0.1 13 0.1 1,148 8.9 747 5.8 5,926 45.9 Total 12,204 70,167 4,193 6.0 976 1.4 6,113 8.7 3,736 5.3 23,056 32.9 (continued) TABLE 1 (continued) Panel B: Accounting Standards, Listing Status and Index Membership by Year IFRS U.S. GAAP U.S. Listing New Markets Index Member Year Firms Firms % Firms % Firms % Firms % Firms % 1990 1,228 26 2.1 4 0.3 113 9.2 38 3.1 478 38.9 1991 1,886 43 2.3 8 0.4 159 8.4 43 2.3 706 37.4 1992 2,385 59 2.5 9 0.4 185 7.8 63 2.6 848 35.6 1993 2,668 65 2.4 11 0.4 210 7.9 76 2.8 929 34.8 1994 3,060 96 3.1 15 0.5 254 8.3 74 2.4 1,039 34.0 1995 3,320 103 3.1 17 0.5 289 8.7 91 2.7 1,130 34.0 1996 3,700 140 3.8 23 0.6 352 9.5 134 3.6 1,264 34.2 1997 4,418 194 4.4 25 0.6 424 9.6 158 3.6 1,446 32.7 1998 4,727 221 4.7 31 0.7 473 10.0 172 3.6 1,620 34.3 1999 5,556 326 5.9 60 1.1 528 9.5 235 4.2 1,887 34.0 2000 5,950 431 7.2 96 1.6 553 9.3 355 6.0 2,021 34.0 2001 6,877 556 8.1 164 2.4 630 9.2 579 8.4 2,224 32.3 2002 7,258 621 8.6 164 2.3 649 8.9 661 9.1 2,358 32.5 2003 7,680 624 8.1 159 2.1 591 7.7 494 6.4 2,228 29.0 2004 7,342 633 8.6 163 2.2 535 7.3 381 5.2 2,088 28.4 2005 2,112 55 2.6 27 1.3 168 8.0 182 8.6 790 37.4 Total 70,167 4,193 6.0 976 1.4 6,113 8.7 3,736 5.3 23,056 32.9 The sample comprises a maximum of 70,167 firm-year observations from 30 countries with fiscal year ends between January 1, 1990, and December 30, 2005, for which we have sufficient Worldscope and Datastream data to estimate our base regressions (see Table 3). We exclude fiscal years when firms are subject to mandatory IFRS reporting. We require firms to have total assets of at least 10 US$ million, and limit the sample to countries with one or more voluntary IFRS firm-year observations. The table reports the number of unique firms, firm-year observations and corresponding percentages by country (Panel A) and year (Panel B) for the following indicator variables (coded one if the definition applies and zero otherwise): IFRS and U.S. GAAP indicate financial reports following the two sets of accounting standards, respectively, based on the “accounting standards followed” field in Worldscope (field 07536), augmented and adjusted for contradictory coding from an extensive manual review of firms’ annual reports (see Appendix A for details). We form a separate category, U.S. Listing, for companies whose shares are traded over-the-counter or listed on a U.S. exchange (see Hail and Leuz, 2009). Note that these observations are not included in the U.S. GAAP indicator. New Market observations stem from firms that are traded on an exchange that specializes in technology shares and other high-growth stocks and that has listing requirements mandating or allowing financial reports according to IFRS (i.e., Alternative Investment Market in the United Kingdom, Expandi market in Italy, Neuer Markt in Germany, Nordic Growth Market in Sweden, and Sesdaq in Singapore). The Index Member variable represents firms whose shares are constituents of national or international stock market indices as defined in Worldscope (field 05661). TABLE 2 Descriptive Statistics for Regression Variables across IFRS and Local GAAP Reporting Firms Panel A: Dependent Variables Variable Accounting Standard N Mean Std. Dev. P1 P25 Median P75 P99 Price Impact Local GAAP 65,974 2.621 9.861 0.000 0.027 0.185 1.169 48.985 IFRS 4,193 1.542 6.554 0.001 0.017 0.096 0.590 26.257 Bid-Ask Spread Local GAAP 45,991 0.033 0.040 0.001 0.009 0.019 0.039 0.204 IFRS 3,485 0.019 0.024 0.001 0.006 0.011 0.024 0.105 Cost of Capital Local GAAP 24,806 11.9% 4.1% 5.5% 9.0% 11.0% 13.9% 25.0% IFRS 1,874 12.0% 4.3% 5.7% 8.9% 11.3% 14.1% 25.8% Panel B: Independent Variables Variable Accounting Standard N Mean Std. Dev. P1 P25 Median P75 P99 Local GAAP 63,949 0.031 0.562 -1.326 -0.349 Reporting Incentives 0.041 0.413 1.255 Factor IFRS 4,127 0.168 0.567 -1.211 -0.203 0.171 0.571 1.330 Reporting Behavior Local GAAP 59,704 -1.654 2.604 -13.604 -1.624 -0.811 -0.466 -0.074 Score IFRS 4,030 -1.709 2.701 -14.452 -1.608 -0.852 -0.519 -0.108 Market Value Local GAAP 65,974 607 1,550 4 38 124 422 8,662 IFRS 4,193 911 1,920 6 73 243 740 10,643 Share Turnover Local GAAP 65,974 0.561 0.898 0.003 0.115 0.315 0.666 4.430 IFRS 4,193 0.525 0.908 0.002 0.098 0.266 0.592 4.310 Return Variability Local GAAP 65,974 0.121 0.080 0.023 0.068 0.099 0.150 0.420 IFRS 4,193 0.132 0.083 0.021 0.075 0.110 0.168 0.417 Total Assets Local GAAP 65,439 1,257 3,876 12 64 194 697 20,507 IFRS 4,144 2,023 4,995 16 121 412 1,438 25,301 Financial Leverage Local GAAP 64,831 0.508 0.241 0.012 0.340 0.521 0.679 0.964 IFRS 4,133 0.530 0.236 0.013 0.367 0.549 0.702 0.969 Forecast Bias Local GAAP 35,460 0.010 0.043 -0.079 -0.004 0.001 0.012 0.212 IFRS 2,642 0.011 0.045 -0.088 -0.004 0.001 0.014 0.214 Inflation – 68,912 2.4% 1.9% 0.0% 1.4% 2.1% 3.0% 9.6% (continued) TABLE 2 (continued) The sample comprises a maximum of 70,167 firm-year observations from 30 countries between 1990 and 2005 with financial data from Worldscope and price/volume data from Datastream (see Table 1). IFRS observations are firm-years with financial reports following IFRS according to our augmented Worldscope accounting standards classification described in Appendix A. The table reports descriptive statistics for the dependent variables (Panel A) and the continuous independent variables (Panel B) across IFRS and local GAAP firm-year observations. We use three dependent variables in the analyses: (1) Price Impact is the yearly median of the Amihud (2002) illiquidity measure (i.e., daily absolute stock return divided by US$ trading volume). (2) The Bid-Ask Spread is the yearly median quoted spread (i.e., difference between the bid and ask price divided by the mid-point and measured at the end of each trading day). (3) Cost of Capital is the average cost of capital estimate implied by the mean I/B/E/S analyst consensus forecasts and stock prices using the Claus and Thomas (2001) model, the Gebhardt, Lee, and Swaminathan (2001) model, the Easton (2004) model, and the Ohlson and Juettner-Nauroth (2005) model. We describe these models in more detail in Appendix B. The independent variables are: using factor analysis, we extract a single factor indicating the strength of firms’ reporting incentives from various firm attributes (i.e., market value of equity, financial leverage, return on assets, book-to-market ratio, percent of closely-held shares, and percent of foreign sales). Higher values denote greater reporting incentives. For each observation in a given year t, we compute the Reporting Incentives Factor as the average of the raw factor scores over the past three years (i.e., t, t-1, t-2). We measure firms’ actual reporting behavior as the absolute value of accruals scaled by the absolute value of cash flows from operations. For each observation in a given year t, we then compute the Reporting Behavior Score as the average of the raw measures over the past three years (i.e., t, t-1, t-2), and multiply this rolling average by -1 so that higher values denote less earnings management and more transparent reporting. Market Value is stock price times the number of shares outstanding (in US$ million). Share Turnover is annual US$ trading volume divided by market value of outstanding equity. We compute Return Variability as annual standard deviation of monthly stock returns. Total Assets are denominated in US$ million. Financial Leverage is computed as the ratio of total liabilities to total assets. Forecast Bias equals the oneyear- ahead I/B/E/S analyst forecast error (mean forecast minus actual) scaled by lagged total assets. Inflation is the yearly median of country-specific, one-year-ahead realized monthly percentage changes in the consumer price index as reported in Datastream. Accounting data and market values are measured as of the fiscal-year end, the dependent variables, forecast bias, return variability and share turnover as of month +10 after the fiscal-year end. Except for variables with natural lower or upper bounds, we truncate all variables at the first and 99th percentile. TABLE 3 Regression Analysis of Liquidity and Cost of Capital Effects of Voluntary IFRS Adoptions Log(Price Impact) Log(Bid-Ask Spread) Cost of Capital Variables Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 IFRS -0.11 -2.37 -1.86 4.58** 3.45* 4.56** 0.31** 0.42*** 0.29** (-0.04) (-0.83) (-0.67) (2.37) (1.80) (2.27) (2.20) (3.29) (1.97) Control Variables: Reporting Incentives Factor – -55.96*** – – -19.68*** – – -3.26*** – (-22.18) (-12.55) (-31.95) Reporting Behavior Score – – -2.56*** – – -1.36*** – – -0.10*** (-10.16) (-9.58) (-5.64) U.S. GAAP 9.60* 5.33 9.35 1.95 -0.71 2.89 -0.22 -0.04 -0.28 (1.69) (0.90) (1.64) (0.56) (-0.20) (0.83) (-0.57) (-0.12) (-0.71) U.S. Listing -26.82*** -25.97*** -26.69*** -2.85* -2.61 -3.05* -0.39*** 0.18* -0.34*** (-10.91) (-10.24) (-10.85) (-1.67) (-1.50) (-1.80) (-3.58) (1.78) (-3.12) New Markets 25.01*** 21.81*** 24.91*** 13.49*** 12.19*** 13.47*** 0.52** 0.41* 0.52** (6.83) (5.90) (6.67) (6.22) (5.62) (6.13) (2.23) (1.86) (2.14) Index Member -58.45*** -56.46*** -60.30*** -18.73*** -18.49*** -19.81*** -1.21*** -0.73*** -1.26*** (-31.46) (-30.07) (-31.28) (-15.33) (-14.72) (-15.93) (-14.54) (-9.24) (-14.46) Log(Market Valuet-1) -98.45*** -82.89*** -98.17*** -31.61*** -25.97*** -31.16*** – – – (-189.01) (-93.95) (-182.43) (-98.40) (-48.51) (-94.18) Log(Share Turnovert-1) -67.47*** -66.28*** -67.43*** -22.03*** -21.53*** -22.18*** – – – (-111.37) (-106.99) (-106.61) (-60.91) (-58.00) (-60.13) Log(Return Variabilityt-1) 46.58*** 45.81*** 44.98*** 33.30*** 33.21*** 32.13*** – – – (38.33) (37.25) (35.94) (42.08) (41.50) (39.50) Log(Total Assets) – – – – – – -0.24*** 0.29*** -0.25*** (-9.24) (9.39) (-8.93) Financial Leverage – – – – – – 3.62*** 3.92*** 3.80*** (18.72) (21.30) (18.90) Return Variability – – – – – – 7.14*** 6.25*** 6.83*** (11.35) (10.05) (10.38) Forecast Bias – – – – – – 18.43*** 19.68*** 18.00*** (18.78) (19.39) (17.75) Inflation – – – – – – 25.44*** 28.06*** 22.81*** (11.79) (13.20) (9.87) Country-, Year-, and Industry-Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes Yes R2 81.39% 81.42% 81.72% 73.24% 73.08% 74.00% 30.06% 35.94% 30.23% # Observations 70,167 68,076 63,734 49,476 47,985 45,863 26,680 25,260 24,522 # Unique Firms 12,204 12,024 11,666 9,875 9,722 9,494 6,088 5,881 5,710 # Countries 30 30 30 24 24 24 29 29 29 (continued) TABLE 3 (continued) The sample comprises a maximum of 70,167 firm-year observations from 30 countries between 1990 and 2005 (see Table 1). The table reports OLS coefficient estimates and (in parentheses) t-statistics based on robust standard errors that are clustered by firm. We use three dependent variables in the analyses: (1) Price Impact is the yearly median of the Amihud (2002) illiquidity measure (i.e., daily absolute stock return divided by US$ trading volume). (2) The Bid-Ask Spread is the yearly median quoted spread (i.e., difference between the bid and ask price divided by the mid-point and measured at the end of each trading day). (3) Cost of Capital is the mean of four estimates for the implied cost of equity capital (see Appendix B). IFRS is a binary variable equal to one for firm-years with financial reports following IFRS according to our augmented Worldscope accounting standards classification described in Appendix A. For a description of the control variables see Table 1 (indicator variables) and Table 2 (continuous variables). We use the natural log of the raw values and lag the variables by one year where indicated. We include industryfixed effects based on the classification in Campbell (1996), country- and year-fixed effects in the regressions, but do not report the coefficients. For expositional purposes we multiply all coefficients by 100. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels (two-tailed), respectively. TABLE 4 Univariate Analysis of Liquidity and Cost of Capital Effects across Label and Serious IFRS Adopters Median Values and Number of Observations Panel A: Price Impact Split by Change in Reporting Incentives Factor Split by Change in Reporting Behavior Score Label Adopters (1) Serious Adopters (2) Difference (2) – (1) Label Adopters (1) Serious Adopters (2) Difference (2) – (1) IFRS 0.111 0.046 -0.065*** 0.106 0.068 -0.038*** (a) 1,060 880 IFRS (a) 975 819 Local GAAP 0.185 0.182 (b) 63,812 Local GAAP (b) 59,732 Difference (a) – (b) -0.074*** -0.139*** Difference (a) – (b) -0.076*** -0.114*** Panel B: Bid-Ask Spread Split by Change in Reporting Incentives Factor Split by Change in Reporting Behavior Score Label Adopters (1) Serious Adopters (2) Difference (2) – (1) Label Adopters (1) Serious Adopters (2) Difference (2) – (1) IFRS 0.012 0.008 -0.004*** 0.011 0.010 -0.001*** (a) 877 717 IFRS (a) 809 681 Local GAAP 0.019 0.019 (b) 44,461 Local GAAP (b) 42,518 Difference (a) – (b) -0.007*** -0.011*** Difference (a) – (b) -0.008*** -0.009*** Panel C: Cost of Capital Split by Change in Reporting Incentives Factor Split by Change in Reporting Behavior Score Label Adopters (1) Serious Adopters (2) Difference (2) – (1) Label Adopters (1) Serious Adopters (2) Difference (2) – (1) IFRS 12.0% 10.7% -1.3%*** 11.6% 11.0% -0.6% (a) 393 437 IFRS (a) 398 460 Local GAAP 11.1% 11.0% (b) 23,423 Local GAAP (b) 22,733 Difference (a) – (b) 0.9%*** -0.4%** Difference (a) – (b) 0.6% 0.0% The sample comprises a maximum of 70,167 firm-year observations from 30 countries between 1990 and 2005 (see Table 1). The table reports median values of the dependent variables and the number of observations across firms reporting under local GAAP, label and serious IFRS adopters. We use three dependent variables in the analyses: (1) Price Impact is the yearly median of the Amihud (2002) illiquidity measure. (2) The Bid-Ask Spread is the yearly median quoted spread. (3) Cost of Capital is the mean of four estimates for the implied cost of equity capital (see Appendix B). IFRS observations are firm-years with financial reports following IFRS according to our augmented Worldscope accounting standards classification described in Appendix A. We partition the IFRS firm-year observations like follows: for each IFRS adopting firm we compute the change in the Reporting Incentives Factor (Reporting Behavior Score) by comparing the rolling threeyear average ending in year t-1 with the rolling three-year average ending in year t+3 (relative to year t of IFRS adoption). See Table 2 for the variable definitions. We then classify firms as Serious Adopters if the change in the respective metric falls above the median change, and as Label Adopters otherwise. We limit the IFRS observations to the initial five firmyears after the adoption. ***, **, and * indicate statistical significance of median differences at the 1%, 5%, and 10% levels (two-tailed), based on a Wilcoxon rank sum test. TABLE 5 Regression Analysis of Liquidity and Cost of Capital Effects across Label and Serious IFRS Adopters Panel A: Log(Price Impact) as Dependent Variable Label versus Serious IFRS Adopter Classification Split by Change in Reporting Incentives Factor Split by Change in Reporting Behavior Score Model 1 Model 2 Model 1 Model 2 IFRS 9.00** 16.95*** 1.11 3.28 (1.97) (3.51) (0.24) (0.70) Serious IFRS Adopters -29.33*** -48.38*** -11.53* -15.41** (-4.57) (-7.04) (-1.73) (-2.31) IFRS + Serious = 0 [p-value] [0.00] [0.00] [0.04] [0.01] Control Variables: Reporting Incentives Factor – -56.48*** – – (-22.47) Reporting Behavior Score – – – -2.57*** (-9.95) U.S. GAAP 9.03 5.81 9.68* 9.57 (1.53) (0.96) (1.66) (1.64) U.S. Listing -25.71*** -24.96*** -25.67*** -26.29*** (-9.92) (-9.58) (-10.08) (-10.39) New Markets 21.44*** 19.43*** 22.84*** 23.06*** (5.38) (4.85) (5.79) (5.86) Index Member -58.92*** -56.76*** -61.00*** -60.88*** (-31.08) (-29.80) (-31.00) (-31.02) Log(Market Valuet-1) -98.49*** -83.01*** -98.78*** -98.29*** (-182.00) (-94.31) (-179.75) (-178.36) Log(Share Turnovert-1) -67.06*** -65.84*** -66.88*** -66.97*** (-107.23) (-103.70) (-102.67) (-103.03) Log(Return Variabilityt-1) 45.84*** 45.61*** 46.18*** 44.65*** (36.89) (36.36) (35.75) (34.50) Country-, Year-, and Industry-Fixed Effects Yes Yes Yes Yes R2 81.2% 81.6% 81.8% 81.9% # Observations 65,752 61,526 # Serious (Label) Observations 880 (1,060) 819 (975) # Unique Firms 11,544 11,180 # Countries 30 30 (continued) TABLE 5 (continued) Panel B: Log(Bid-Ask Spread) as Dependent Variable Label versus Serious IFRS Adopter Classification Split by Change in Reporting Incentives Factor Split by Change in Reporting Behavior Score Model 1 Model 2 Model 1 Model 2 IFRS 3.58 6.17** 1.01 2.20 (1.28) (2.19) (0.32) (0.69) Serious IFRS Adopters -12.51*** -18.33*** -8.14** -10.49** (-3.10) (-4.48) (-1.96) (-2.53) IFRS + Serious = 0 [p-value] [0.00] [0.00] [0.02] [0.01] Control Variables: Reporting Incentives Factor – -18.77*** – – (-11.87) Reporting Behavior Score – – – -1.51*** (-10.50) U.S. GAAP -2.29 -4.39 -0.59 -0.52 (-0.63) (-1.18) (-0.16) (-0.15) U.S. Listing -3.47** -3.51** -3.79** -4.17** (-2.01) (-2.03) (-2.27) (-2.52) New Markets 14.41*** 13.79*** 14.40*** 14.60*** (5.88) (5.69) (6.03) (6.15) Index Member -19.38*** -18.73*** -20.33*** -20.15*** (-15.58) (-14.85) (-15.94) (-15.87) Log(Market Valuet-1) -31.32*** -26.16*** -31.30*** -31.00*** (-93.10) (-49.01) (-92.37) (-91.39) Log(Share Turnovert-1) -21.61*** -21.17*** -21.66*** -21.74*** (-59.25) (-57.48) (-56.85) (-57.25) Log(Return Variabilityt-1) 33.08*** 33.06*** 32.73*** 31.79*** (40.01) (40.23) (39.44) (38.28) Country-, Year-, and Industry-Fixed Effects Yes Yes Yes Yes R2 73.3% 73.5% 74.3% 74.5% # Observations 46,055 44,008 # Serious (Label) Observations 717 (877) 681 (809) # Unique Firms 9,287 9,056 # Countries 24 24 (continued) TABLE 5 (continued) Panel C: Cost of Capital as Dependent Variable Label versus Serious IFRS Adopter Classification Split by Change in Reporting Incentives Factor Split by Change in Reporting Behavior Score Model 1 Model 2 Model 1 Model 2 IFRS 1.14*** 1.34*** 0.84*** 0.89*** (4.11) (5.22) (2.79) (2.95) Serious IFRS Adopters -1.07*** -1.78*** -0.59* -0.69** (-3.11) (-5.59) (-1.71) (-1.99) IFRS + Serious = 0 [p-value] [0.75] [0.04] [0.21] [0.32] Control Variables: Reporting Incentives Factor – -3.15*** – – (-29.50) Reporting Behavior Score – – – -0.10*** (-4.98) U.S. GAAP -0.23 -0.02 -0.26 -0.26 (-0.55) (-0.05) (-0.65) (-0.64) U.S. Listing -0.35*** 0.17 -0.33*** -0.34*** (-3.05) (1.63) (-2.88) (-2.98) New Markets 0.57** 0.48* 0.45* 0.46* (2.08) (1.87) (1.68) (1.72) Index Member -1.21*** -0.76*** -1.28*** -1.28*** (-14.07) (-9.39) (-14.51) (-14.50) Log(Total Assets) -0.23*** 0.27*** -0.26*** -0.25*** (-8.02) (8.45) (-9.07) (-8.77) Financial Leverage 3.68*** 3.95*** 3.85*** 3.80*** (18.00) (20.89) (18.64) (18.54) Return Variability 7.16*** 6.29*** 7.06*** 6.67*** (10.47) (9.73) (10.40) (9.85) Forecast Bias 17.93*** 19.23*** 18.11*** 17.87*** (16.80) (18.34) (17.29) (17.07) Inflation 24.77*** 26.58*** 21.97*** 21.80*** (10.78) (11.94) (9.30) (9.23) Country-, Year-, and Industry-Fixed Effects Yes Yes Yes Yes R2 30.0% 35.6% 30.2% 30.4% # Observations 24,253 23,591 # Serious (Label) Observations 437 (393) 460 (398) # Unique Firms 5,574 5,413 # Countries 29 29 (continued) TABLE 5 (continued) The sample comprises a maximum of 70,167 firm-year observations from 30 countries between 1990 and 2005 (see Table 1). The table reports OLS coefficient estimates and (in parentheses) t-statistics based on robust standard errors that are clustered by firm. It also reports p-values [in brackets] from an F-test indicating joint statistical significance of the coefficients on IFRS and Serious IFRS Adopters. We use three dependent variables in the analyses: (1) Price Impact is the yearly median of the Amihud (2002) illiquidity measure. (2) The Bid-Ask Spread is the yearly median quoted spread. (3) Cost of Capital is the mean of four estimates for the implied cost of equity capital (see Appendix B). IFRS observations are firm-years with financial reports following IFRS according to our augmented Worldscope accounting standards classification described in Appendix A. We partition the IFRS firm-year observations like follows: for each IFRS adopting firm we compute the change in the Reporting Incentives Factor (Reporting Behavior Score) by comparing the rolling threeyear average ending in year t-1 with the rolling three-year average ending in year t+3 (relative to year t of IFRS adoption). We then set the binary Serious IFRS Adopters variable equal to one if a firm’s change in the respective metric falls above the median change, and to zero otherwise (i.e., representing label adopters). We limit the IFRS observations to the initial five firm-years after the adoption. For a description of the control variables see Table 1 (indicator variables) and Table 2 (continuous variables). For IFRS adopting firms, we use the Reporting Incentives Factor or Reporting Behavior Score in year t-1 before IFRS adoption, rather than the rolling average. We use the natural log of the raw values and lag the variables by one year where indicated. We include industry-fixed effects based on the classification in Campbell (1996), country- and year-fixed effects in the regressions, but do not report the coefficients. For expositional purposes we multiply all coefficients by 100. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels (two-tailed), respectively. TABLE 6 Sensitivity Analyses for the Cross-Sectional Liquidity and Cost of Capital Effects Label versus Serious IFRS Adopter Classification Split by Change in Reporting Incentives Factor Split by Change in Reporting Behavior Score IFRS Serious IFRS Adopters IFRS Serious IFRS Adopters Panel A: Alternative Dependent Variables Log(Total Trading Costs) -9.114*** -7.801* -7.439** -11.407*** (-2.93) (-1.87) (-2.46) (-2.61) [0.00] [0.00] Zero Returns -2.486*** -0.222 -2.078*** -1.364 (-4.05) (-0.25) (-3.15) (-1.55) [0.00] [0.00] Tobin’s q -36.024*** 57.173*** -9.260** 7.549 (-12.13) (11.64) (-1.99) (1.17) [0.00] [0.73] Panel B: Alternative Accounting Standards Classifications (Price Impact as Dependent Variable) Worldscope Classification 21.395*** -56.620*** 0.852 -10.222* (5.28) (-9.39) (0.19) (-1.78) [0.00] [0.03] Global Vantage Classification 28.673*** -53.164*** 12.930** -19.253** (5.05) (-5.83) (2.27) (-2.28) [0.00] [0.33] Hand-coded Classification 16.384*** -43.813*** 2.461 -13.014* (3.23) (-5.97) (0.51) (-1.84) [0.00] [0.06] IFRS and U.S. GAAP Combined 16.308*** -48.191*** 4.256 -16.715*** (3.53) (-7.42) (0.90) (-2.68) [0.00] [0.01] U.S. GAAP 11.031 -56.479*** 6.323 -30.817* (0.83) (-2.97) (0.44) (-1.78) [0.00] [0.04] Panel C: Alternative Sample Compositions and Model Specifications (Price Impact as Dependent Variable) All Firms with Available Data 19.090*** -49.254*** 4.626 -17.038*** (4.03) (-7.39) (1.01) (-2.59) [0.00] [0.01] Exclude Firms in the U.K. & Canada 13.543*** -46.643*** -1.604 -12.390* (2.79) (-6.77) (-0.34) (-1.85) [0.00] [0.01] Random Benchmark Sample 18.791*** -46.294*** 4.285 -12.822** (3.98) (-6.94) (0.94) (-1.97) [0.00] [0.09] IFRS Firms only 4.629 -44.694*** -9.098 -9.860 (0.83) (-6.64) (-1.56) (-1.49) [0.00] [0.00] Firm-fixed Effects 15.092** -65.917*** -8.675 -24.481** (2.17) (-6.43) (-1.17) (-2.08) [0.00] [0.00] (continued) TABLE 6 (continued) The base sample comprises a maximum of 70,167 firm-year observations from 30 countries between 1990 and 2005 (see Table 1). The table reports only the coefficients (t-statistics, clustered by firm) of the IFRS and Serious IFRS Adopters variables, but the full set of controls is included (see Model 2 in Table 5). It also reports p-values [in brackets] from an Ftest indicating joint statistical significance of the two variables. We show results for the serious and label IFRS adopters across the change in Reporting Incentives Factor and the change in Reporting Behavior Score partitions. In Panel A, we report results for the following alternative dependent variables: Total Trading Costs is a yearly estimate of total round-trip transaction costs (i.e., bid-ask spreads, commissions, and implicit costs such as short-sale constraints or taxes) inferred from the series of daily security and aggregate market returns, as developed by Lesmond, Ogden, and Trzcinka (1999). Zero Returns is the proportion of trading days with zero daily stock returns out of all potential trading days in a given year. Tobin’s q equals (total assets – book value of equity + market value of equity)/total assets. The total trading costs and zero returns specifications are the same as for price impact. For Tobin’s q we use the log of total assets, financial leverage, asset growth (computed as the one-year percentage change in total assets), and industry q (equal to the yearly median q in a given industry) as continuous control variables. In Panels B and C, the dependent variable is Price Impact. In Panel B, we report results for the classification of IFRS, U.S. GAAP and local GAAP firm-years based on either Worldscope, Global Vantage, or our hand-coding as described in Appendix A. We also create a combined IFRS and U.S. GAAP classification (i.e., all IFRS and U.S. GAAP firm-years taken together), and split the firms that switch from local GAAP to either IFRS or U.S. GAAP into serious and label adopters. Finally, we split the firms switching from local GAAP to U.S. GAAP into serious and label U.S. GAAP adopters while controlling for IFRS reporting in the regression. In Panel C, we report results for alternative sample compositions and model specifications: (1) we include the entire Worldscope universe, i.e., we do not impose a minimum limit on firm size and the number of voluntary IFRS observations per country. (2) We exclude the U.K. and Canada, the two largest sample countries. (3) We limit the benchmark sample consisting of the non-IFRS adopting firms to 150 randomly selected observations per country and year. (4) We only include firms that at some point during the sample period reported under IFRS. (5) We replace the Reporting Incentives Factor (Reporting Behavior Score), the country- and industry-fixed effects with firm-fixed effects. For expositional purposes we multiply all coefficients by 100. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels (two-tailed), respectively. TABLE A1 Definition of Accounting Standards Classifications Panel A: Coding Based on Worldscope (WS) “Accounting Standards Followed” (field 07536) WS Code WS Description Coding for Analyses We code firm-year observations as IFRS if one of the following cases applies: IFRS 02 International standards 06 International standards and some EU guidelines 08 Local standards with EU and IASC guidelines 12 International standards – inconsistency problems 16 International standards and some EU guidelines – inconsistency problems 18 Local standards with some IASC guidelines 19 Local standards with OECD and IASC guidelines 23 IFRS We code firm-year observations as U.S. GAAP if one of the following cases applies: U.S. GAAP 03 U.S. standards (GAAP) 13 US standards – inconsistency problems 20 US GAAP reclassified from local standards We code firm-year observations as local if one of the following cases applies: Local 01 Local standards 05 EU standards 07 Specific standards set by the group 09 Not disclosed 10 Local standards with some EU guidelines 11 Local standards – inconsistency problems 14 Commonwealth standards – inconsistency problems 15 EEC standards – inconsistency problems 17 Local standards with some OECD guidelines 21 Local standards with a certain reclassification for foreign companies 22 Other Panel B: Coding Based on Global Vantage (GV) “Accounting Standard” (field ASTD) GV Code GV Description Coding for Analyses We code firm-year observations as IFRS if one of the following cases applies: IFRS DA Domestic standards generally in accordance with IASC and OECD guidelines DI Domestic standards generally in accordance with IASC guidelines DT Domestic standards in accordance with principles generally accepted in the United States and generally in accordance with IASC and OECD guidelines We code firm-year observations as U.S. GAAP if one of the following cases applies: U.S. GAAP DU Domestic standards in accordance with principles generally accepted in the U.S. MU Modified United States’ standards (Japanese companies’ financial statements translated into English) US United States’ standards We code firm-year observations as local if one of the following cases applies: Local DD Domestic standards for parents and domestic subsidiaries. Native country or United States’ standards for overseas subsidiaries DO Domestic standards generally in accordance with OECD guidelines DR Accounts reclassified to show allowance for doubtful accounts and/or accumulated depreciation as a reduction of assets rather than liabilities DS Domestic standards MI Accounts reclassified by SPCS to combine separate life insurance and non-life insurance accounts LJ Combination DR and MI (continued) TABLE A1 (continued) Panel C: Hand-coded Classification Based on Firms’ Annual Reports Description Coding for Analyses We code firm-year observations as IFRS if one of the following cases applies: (1) Notes to consolidated financial statements refer to IAS/IFRS standards only (2) Annual report has two separate sections with two full sets of consolidated financial statements (balance sheet, income statement, statement of cash flows), one set under local GAAP, and one set under IAS/IFRS (Parallel Reporting) (3) Notes to consolidated financial statements refer to IAS/IFRS in the first place, but also refer to compliance with local GAAP (4) Notes to consolidated financial statements refer to local GAAP in the first place, but also refer to compliance with IAS/IFRS (5) Notes to consolidated financial statements refer to full compliance with local GAAP, but also to application of IAS/IFRS standards if local GAAP is silent about a reporting issue (Dual Reporting) (6) Notes to consolidated financial statements refer to full application local GAAP, but there is also a reconciliation of net income and/or shareholders’ equity to IAS/IFRS in a separate section of the annual report (Reconciliation) IFRS We code firm-year observations as U.S. GAAP if one of the following cases applies: (1) Notes to consolidated financial statements refer to U.S. GAAP standards only (2) Annual report has two separate sections with two full sets of consolidated financial statements (balance sheet, income statement, statement of cash flows), one set under local GAAP, and one set under U.S. GAAP (Parallel Reporting) (3) Notes to consolidated financial statements refer to U.S. GAAP in the first place, but also refer to compliance with local GAAP (4) Notes to consolidated financial statements refer to local GAAP in the first place, but also refer to compliance with U.S. GAAP (5) Notes to consolidated financial statements refer to full compliance with local GAAP, but also to application of U.S. GAAP standards if local GAAP is silent about a reporting issue (Dual Reporting) (6) Notes to consolidated financial statements refer to full application local GAAP, but there is also a reconciliation of net income and/or shareholders’ equity to U.S. GAAP in a separate section of the annual report (Reconciliation) U.S. GAAP We code firm-year observations as local if one of the following cases applies: (1) Notes to consolidated financial statements refer to local GAAP only (2) Notes to consolidated financial statements refer to local GAAP only, but selected individual IAS/IFRS or U.S. GAAP standards are applied on specific reporting issues (e.g. Leasing IAS 17, Segment Reporting SFAS 131) Local The table describes the assignment of firm-year observations to the three reporting categories IFRS, U.S. GAAP or Local GAAP using the accounting standards classification in Worldscope (Panel A), Global Vantage (Panel B), and our own classification applied to a comprehensive set of firms’ annual reports collected through Thomson Research (Panel C).
Tidak ada komentar:
Posting Komentar