Market Reaction to the Adoption of Ifrs in Europe

ABSTRACT: This study examines European stock market reactions to 16 events associated with the adoption of International Financial Reporting Standards (IFRS) in Europe. European IFRS adoption represented a major milestone toward ancial reporting convergence yet spurred controversy reaching the highest levels of government. We nd an incrementally positive reaction for rms with lower quality pre-adoption information, which is more pronounced for banks, and with higher pre-adoption information asymmetry, consistent with investors expecting net information quality benets from IFRS adoption. We nd an incrementally negative reaction for rms domiciled in code law countries, consistent with investors’ concerns over enforcement of IFRS in those countries. Finally, we nd a positive reaction to IFRS adoption events for ms with high-quality pre-adoption information, consistent with investors expecting net convergence benets from IFRS adoption. Keywords: IFRS; IAS 39; convergence; Europe. Data Availability: All data are publicly available from sources indicated in the text. JEL Classications: M41, G15, G38. We thank Robert Bushman, Craig Chapman, Tony Cope, Gilbert Gelard, Ian Gow, Bob Holthausen, Ole-Kristian Hope, Dave Larcker, Jim Leisenring, Dave Maber, Greg Miller, Karl Muller, Joe Piotroski, Thorsten Sellhorn, Dan Taylor, and two anonymous referees for helpful comments and discussions.

We also thank seminar participants at the American Accounting Association Annual Meeting, especially Luzi Hail, discussant, European Accounting Association Annual Congress, Harvard Business School Accounting Seminar Series, Harvard Business School International Seminar Series, The Ohio State University, Southern Methodist University, Stanford Graduate School of Business Accounting Summer Camp, The University of Chicago, The University of Iowa, University of North Carolina Global Issues in Accounting Conference, and University of Toronto. We thank Sarah Eriksen and James Zeitler for data assistance, and Susanna Kim for research assistance.

Editor’s note: Accepted by Steven Kachelmeier, with thanks to Dan Dhaliwal for serving as editor on previous versions. Submitted: August 2007 Accepted: April 2009 Published Online: January 2010 31 32 Armstrong, Barth, Jagolinzer, and Riedl I. INTRODUCTION his study examines European stock market reactions to events associated with the 2005 adoption of International Financial Reporting Standards (IFRS) in Europe. 1 Prior to 2005, most European rms applied domestic accounting standards. Thus, the adoption of IFRS in Europe represented one of the largest ancial reporting changes in recent years and was controversial, generating debate that reached the highest levels of government. The adoption of IFRS as issued by the International Accounting Standards Board (IASB) would result in the application of a common set of nancial reporting standards within Europe, and between Europe and the many other countries that require or permit application of IFRS. Thus, the debate was about not only the benets and costs of IFRS adoption itself, but also the global nancial reporting convergence implications if IFRS were modied as a result of the adoption process. Modifying IFRS would result in European standards differing from those used in other countries, thereby eliminating some potential convergence benets. 3 We refer to the adoption of IFRS as issued by the IASB as the adoption of IFRS—adoption of modied standards is not adoption of IFRS. It is unclear how investors in European rms would react to this anticipated change in nancial reporting. This study examines these reactions. It is possible that investors in European rms would react positively to movement toward IFRS adoption if, for example, investors expected application of IFRS to result in higher quality ancial reporting information, thereby lowering information asymmetry between the rm and investors and information risk and, thus, cost of capital. Investors also might have believed that application of a common set of standards would have convergence benets, such as lowering the costs of comparing rms’ nancial position and performance across countries, and that IFRS adoption would enable European capital markets to become more globally competitive, with consequent increases in liquidity for European rms. Alternatively, it is possible that investors in European rms would react negatively to movement toward IFRS adoption.

This could be the case if investors believed that IFRS would result in lower quality nancial reporting information. For example, IFRS might not adequately reect regional differences in economies that led to differences in domestic accounting standards. Also, investors might have believed that potential variation in the implementation and enforcement of IFRS would lead to an increase in the exercise of opportunistic managerial discretion when applying IFRS. Finally, investors might have believed that the implementation and transition costs associated with IFRS would exceed any benes. To gain insight into investors’ expectations regarding the net cost or benet of IFRS adoption in Europe, we examine three-day market-adjusted returns for rms with equity traded in the European stock market centered on 16 events that we assess as affecting the likelihood of IFRS adoption in Europe. We nd an incrementally positive reaction for rms with lower pre-adoption information quality, which is consistent with investors expecting that IFRS adoption will result in greater informational benets for these rms.

We examine market reactions in all stock markets in Europe. Throughout, we refer to these markets collectively as the European stock market. Also, we examine market reactions for rms trading in the European stock market. Throughout, we refer to these rms as European rms. For example, the U. S. Securities and Exchange Commission (SEC) recently eliminated the requirement for cross-listed rms that prepare their nancial statements using IFRS as issued by the IASB to reconcile net income and shareholders’ equity from those based on IFRS to those based on U. S. standards.

The SEC did not propose to eliminate the requirement for cross-listed rms that use IFRS as modied by any particular jurisdiction, including the European Union. Market reactions studied in this paper may reect perceptions of investors in European rms regarding the global convergence implications of any decision by Europe to modify IFRS before adoption.

This is consistent with investors expecting that IFRS will result in a greater improvement in information quality for these banks, perhaps reecting perceived net benets associated with adoption of the controversial International Accounting Standard (IAS) 39, Financial Instruments: Recognition and Measurement. We also nd an incrementally positive reaction for rms with higher pre-adoption information asymmetry, which is consistent with investors expecting that IFRS adoption will mitigate this asymmetry. Finally, we nd an incrementally negative reaction for ms domiciled in code law countries. This nding is consistent with investors harboring concerns regarding implementation of IFRS in countries that are generally thought to have weaker accounting standards enforcement, although this reaction could be attributable to other factors associated with a rm being domiciled in a code law country. Further analyses reveal a positive reaction to IFRS adoption events for the subset of European rms with the highest quality pre-adoption information. To the extent investors expect little, if any, informational benets from IFRS adoption for these ms, this nding is consistent with investors expecting net benets associated with convergence. Sensitivity analyses reveal that our results are robust to alternative proxies for information quality, measures of standards enforcement environments, and benchmark returns. Section II discusses the background of IFRS adoption in Europe. Section III reviews prior research and provides the basis for interpreting the market reaction to each event. Section IV describes our data and research design. Section V presents our empirical results, and Section VI presents sensitivity analyses.

Background

European Union Adoption Process We focus on the European Union (EU) adoption of IFRS because the EU comprises most countries in Europe. 4 In March 2002, the European Parliament passed a resolution requiring all rms listed on stock exchanges of European member states to apply IFRS when preparing their nancial statements for scal years beginning on or after January 1, 2005. This requirement affected approximately 7,000 rms. 5 The prospects of adopting IFRS represented a substantial shift in nancial reporting for European ms because many requirements in IFRS differ from those in domestic standards of European countries. Also, the adoption of IFRS in Europe reects an EU goal of achieving capital market integration; it is a necessary step toward convergence of nancial reporting not only across Europe, but also between Europe and the rest of the world. Although the resolution requires rms to use IFRS, which are issued by the IASB, a private-sector standard-setter, the European Commission (EC) must endorse the standards before they are required in the EU.

Thus, the EC retains the power to reject any standard, or part of a standard, it believes does not meet its criteria for endorsement. The three primary criteria are: the standard is not contrary to the EU’s true and fair principle; the standard meets the criteria of understandability, 4 5 The exceptions are Iceland, Liechtenstein, and Norway, which are members of the European Economic Area (EEA), and Switzerland. Members of the EEA are committed to following EU Directives, including those relating to IFRS adoption. Switzerland is centrally located in Europe and, thus, has close conomic ties with other European countries. Effective in 2005, Switzerland required all listed rms to use either IFRS or U. S. standards. Presumably its decision was related to that of the EU. Thus, our sample includes rms domiciled in Norway and Switzerland; no rms in Liechtenstein or Iceland met our data requirements. Because Switzerland is not committed to following EU Directives, we conducted all of our tests omitting Swiss rms with no change in our inferences. The required adoption date is January 1, 2007 for rms trading securities in the U. S. and basing their ancial statements on U. S. standards, and rms trading debt securities but not equity securities. The Accounting Review January 2010 American Accounting Association 34 Armstrong, Barth, Jagolinzer, and Riedl relevance, reliability, and comparability; and adopting the standard is in the European public interest. The EC endorsement process, which played a key role in the adoption of IFRS in Europe, is as follows (Brackney and Witmer 2005; KPMG IFRG Limited 2005). The IASB develops IFRS in accordance with due process procedures outlined in its governing constitution (IASB 2006).

This process involves public meetings and extensive input from interested parties around the world. Among these is the European Financial Reporting Advisory Group (EFRAG), a private-sector organization comprised of accounting experts from the EU, which provides advice to the EC regarding technical accounting matters. After the IASB issues a standard, EFRAG reviews it and, after public consultation, EFRAG decides whether to recommend that the EC endorse the standard for use in Europe. Taking EFRAG’s advice into account, the EC drafts proposed regulation. The EC then seeks input from the Accounting Regulatory Committee (ARC).

The ARC, a governmental organization comprised of representatives from each EU member state, reviews the regulation and provides its recommendation about adoption in the EU. The ARC considers the technical merits of the standard as expressed in EFRAG’s recommendation letter, as well as the implications of the standard for the European public interest. If the ARC recommends endorsement, then the EC decides whether to endorse the standard, as written by the IASB or as amended, or to reject it. If endorsed, the standard becomes regulation applicable to rms in the EU.

If the ARC recommends rejection of the standard, then the EC can ask EFRAG to consider it further, or send it to the European Parliament for a decision. 6 The debate surrounding the adoption of IFRS in Europe initially focused on the merits of adopting IFRS, such as whether the benets of the expected increase in capital ows would outweigh the costs of implementation and loss of diversity in domestic accounting standards. The debate later centered on IAS 39, Financial Instruments: Recognition and Measurement, and, to a lesser extent, on IAS 32, Financial Instruments: Disclosure and Presentation (IASB 2004a).

The provisions in these two standards, particularly IAS 39 (IASB 2004b), had the potential to materially affect nancial statement amounts for rms with a large number of nancial instruments, notably banks. The debate regarding IAS 39 ultimately led to the modication of IAS 39 for adoption in Europe. However, modications of IAS 39, or any other IASB standard, undermine the EU’s goal of adopting a set of global standards. Regarding IAS 39, the controversy focused on two types of requirements. The st type relates to use of fair value as the measurement attribute. IAS 39 requires many nancial instruments—notably derivatives—to be recognized at fair value, with changes in fair value recognized in prot or loss. 7 IAS 39 also includes a fair value option that permits rms to designate irrevocably nancial instruments on initial recognition as ones to be measured at 6 7 The process can apply to a single standard or to a group of standards.

For the initial endorsement of IFRS in Europe, the extant set of standards was considered as a group. Specially, the EC considered all standards effective at March 1, 2002, which included IAS 1 through IAS 41, as well as the related Standing Interpretations Committee (SIC) interpretations, i. e. , SIC 1 through SIC 33. IAS 39 classies nancial assets into (1) loans and receivables not held for trading; (2) held-to-maturity investments; (3) nancial instruments held for trading, including derivatives; and (4) available-for-sale nancial assets. Financial assets in (1), (2), (3), and (4) are recognized at, respectively, amortized cost; amortized cost subject to impairment; fair value with changes in fair value recognized in proor loss; and fair value with changes in fair value recognized in other comprehensive income. Most nancial liabilities are recognized at amortized cost, except derivatives and liabilities held for trading, which are recognized at fair value with changes in fair value recognized in prot or loss.

Market Reaction to the Adoption of IFRS in Europe 35 fair value with changes in fair value recognized in prot or loss. The second type of controversial requirements relates to qualifying criteria for hedge accounting—IAS 39’s qualifying criteria are speciand not easy to satisfy. Hedge accounting generally results in gains (losses) on a hedged item and losses (gains) on a designated hedging instrument being recognized in prot or loss at the same time to the extent the gains (losses) on the hedged item result from the hedged risk(s). Thus, hedge accounting reduces volatility in prot or loss resulting from, e. g. , measuring derivatives entered into for hedging purposes at fair value and measuring the hedged item at amortized cost. However, IAS 39 does not permit hedge accounting for many ancial instruments ostensibly entered into for hedging purposes. For example, IAS 39 does not permit hedge accounting for interest rate risk associated with core demand deposits, even though European banks frequently claim to hedge such risk. For many European rms, the fair value and hedging requirements in IAS 39 differ substantially from requirements in their domestic standards. In fact, most European domestic standards do not include standards specifying the nancial reporting for many nancial instruments.

IFRS Adoption Events Although the adoption of IFRS was an involved process with considerable discussion, we identify 16 events between 2002 and 2005 that we assess as affecting the likelihood that IFRS would be adopted in Europe. 8 We identify the events by searching Dow Jones News Retrieval using the terms ‘‘IFRS,’’ ‘‘international nancial reporting standards,’’ ‘‘harmonization,’’ and ‘‘IAS 39,’’ as well as by examining press releases and available listings of documents publicly released by the IASB, the European Parliament, the EC, and EFRAG.

This search provided an initial listing of approximately 40 events. Each author independently veried each event’s timing, content, and likely directional effect on IFRS adoption likelihood. Each author then independently identied the events that likely had the greatest effect on the likelihood of IFRS adoption; events that simply conrmed an earlier event were eliminated. Table 1 lists the resulting 16 events and our assessment as to whether each event increased or decreased adoption likelihood. We assess 13 events as increasing the likelihood of IFRS adoption and three events as decreasing it.

Our assessment of each event’s directional effect on adoption likelihood reects our assessment of how the event likely affected investors’ expectations conditional on prior events and discussions, and enables us to aggregate the market reaction across events. The rst event is March 12, 2002, when the European Parliament passed the resolution requiring all rms listed on stock exchanges in the EU to apply IFRS by 2005. The resolution passed by a vote of 429 for, 5 against, and 29 abstentions, indicating broad support for adoption of IFRS.

Even though convergence toward international standards had been under consideration in Europe prior to 2002, we use the March 12, 2002 resolution as our rst event because passage of the resolution was the rst clear commitment to IFRS 8 Earlier events that we identify generally relate to broad adoption of IFRS in Europe. Later events generally focus on whether Europe might adopt particular standards as written by the IASB. Even though the focus was on particular standards for the later events, the uncertainty remaining to be resolved as the later events unfolded was more general.

In particular, a key dimension of uncertainty was whether Europe would adopt IFRS as written by the IASB or would modify some standards, which could have global convergence implications.

Description European Parliament passes resolution requiring all EU-listed companies to use IFRS by 2005 EFRAG issues draft recommendation to endorse all extant IFRS EFRAG issues al recommendation to endorse all extant IFRS Chirac sends letter to Prodi expressing concerns about IAS 39 and its potential negative effect on Europe Bolkestein sends letter to Tweedie supporting goal of adoption ECOFIN and ARC support adoption of IFRS EC endorses all extant IFRS, except IAS 32 and IAS 39 Bolkestein pledges to postpone endorsement of IAS 32 and IAS 39 until issues are resolved; sets up consultative group to facilitate resolution HSBC announces intentions to implement IAS 39 in full EFRAG issues draft recommendation to endorse IAS 32 and IAS 39 EFRAG issues al recommendation to endorse IAS 32 and IAS 39 ARC recommends endorsement of IAS 39, but recommends provisions relating to the fair value option and portfolio hedging of demand deposits be carved out EC endorses IAS 39 with both carve-out provisions IASB issues revised IAS 39 with new fair value option ARC recommends endorsement of revised fair value option, thereby eliminating one of the carve-outs EC endorses revised fair value option, thereby eliminating one of the carve-outs (continued on next page)

The Accounting Review TABLE 1 (continued) This table presents the 16 events, our assessment of their effect on the likelihood of the European adoption of IFRS as issued by the IASB, and the predicted market reaction to each event. In the last column, IFRSbenets IFRScosts refers to the predicted market reaction if expected benets associated with IFRS adoption exceed expected costs. IFRSbenets IFRScosts refers to the predicted market reaction if expected benets associated with IFRS adoption are lower than expected costs. Key persons / organizations referred to in the event descriptions are deed as follows: ARC (Accounting Regulatory Committee) is a public-sector body that opines on EC proposals regarding international accounting standards, and is comprised of representatives from each member state of the European Union (EU), chaired by the EC. Bolkestein (Frits Bolkestein), a commissioner of the EC, is responsible for internal markets, taxation, and customs union. Chirac (Jacques Chirac) is the President of France.

EC (European Commission) was created to represent the European interest common to all Member States of the EU, and has the right of initiative in the legislative process, i. . , it proposes the legislation on which the European Parliament and the Council decide to enact. ECOFIN (European and Financial Affairs Council) is comprised of the Economics and Finance ministers of the member states, and covers EU policy in several areas, including nancial markets. EFRAG (European Financial Reporting Advisory Group) is a private-sector body created by the accounting profession within Europe, and advises the EC on the technical assessment of IASB-issued nancial reporting standards. IASB (International Accounting Standards Board) is an independent, privately funded ancial reporting standard-setter charged with creating International Financial Reporting Standards. Prodi (Romano Prodi) is the President of the EC. Tweedie (Sir David Tweedie) is the Chairman of the IASB.

Market Reaction to the Adoption of IFRS in Europe January 2010 American Accounting Association 37 38 Armstrong, Barth, Jagolinzer, and Riedl adoption. 9 On May 14, 2002 and June 19, 2002, EFRAG issued its draft and nal recommendations that extant IFRS be endorsed en bloc. The endorsement recommendation letters stated EFRAG’s view that the regulation requiring European ms to adopt IFRS by 2005 is a major achievement in that the common basis for nancial reporting based on highquality global standards provides a platform for efcient cross-border investment both within and beyond the EU. EFRAG further noted that the IASB was reviewing several standards with the objective of making some improvements to them; EFRAG would consider those changes and make its recommendation on them after the IASB issues them. These three events reected clear support for the broad adoption of IFRS. Therefore, we assess these events as increasing the likelihood of IFRS adoption in Europe.

During the remainder of 2002 and into 2003 the EC considered whether to accept EFRAG’s recommendation to endorse extant IFRS. On July 4, 2003, Jacques Chirac, President of France, wrote a letter to Romano Prodi, President of the EC, to express concern that adopting IFRS, particularly IAS 39, would not be in the best interest of Europe. Chirac’s interest in the debate arose at least in part because French banks were among the most critical of IAS 39. Chirac’s involvement showed IFRS-related concern at the highest level of government.

Therefore, we assess this event as decreasing the likelihood of IFRS adoption. On July 9, 2003, Frits Bolkestein, the EC commissioner responsible for internal markets, expressed to Sir David Tweedie, Chairman of the IASB, similar concern about IAS 39, but expressed support for the broader goal of convergence using IFRS. On July 16, 2003, the ARC and the EU’s Economic and Financial Affairs Council (ECOFIN), which is comprised of the Economics and Finance ministers of the EU member states, echoed Bolkestein’s support for adoption of IFRS, despite also echoing his concern about IAS 39.

On September 29, 2003, the EC endorsed all extant IFRS, except IAS 32 and IAS 39. Even though all of these events reect concern about IAS 39, they all expressed clear support for IFRS adoption and the desire to work to resolve in the near term the remaining issues relating to IAS 39. Therefore, we assess these three events as increasing the likelihood of IFRS adoption. After the EC’s endorsement of most extant IFRS, the debate seemingly focused on IAS 32 and IAS 39, although the debate also rected the possibility that the EU would amend IASB-issued standards—and the implications this might have for the adoption of future standards by the EU. Although the IASB revised IAS 39 in December 2003, the revisions did not resolve the controversial issues relating to the fair value option and hedge accounting requirements. Thus, the endorsement of IAS 32 and IAS 39 remained uncertain. On February 3, 2004, Bolkestein indicated his intention to continue postponing endorsement of IAS 32 and IAS 39 until the issues could be resolved. To facilitate resolution, he announced establishment of a high-level consultative group.

On March 30, 2004, HSBC, the largest European bank, expressed its support for adoption by announcing its plans to implement IAS 39 in full, regardless of whether the EC endorsed it. On June 4, 2004, EFRAG issued draft recommendations to endorse IAS 32 and IAS 39. Although the IAS 32 recommendation was unanimous, 6 of 11 EFRAG members voted against the IAS 39 9 Prior research nds little (Comprix et al. 2003) or limited (Pae et al. 2007) evidence of a market reaction to IFRS adoption events before March 12, 2002. These ndings suggest that investors’ IFRS adoption expectations were not ffected by events prior to the March 12, 2002 resolution, the studied events were not those that affected investors’ expectations, or the event windows were too narrow to capture the market reaction to the events. Nonetheless, our ndings may only partially capture the market reaction to EU IFRS adoption to the extent that events prior to March 12, 2002 affected investors’ expectations about the European Parliament’s resolution. The direction of any bias arising from the omission of such events is unclear.

Market Reaction to the Adoption of IFRS in Europe 39 recommendation. A majority negative vote is insufcient for EFRAG to recommend nonendorsement of a standard—the EFRAG constitution requires a two-thirds negative vote. On July 8, 2004, EFRAG issued its nal recommendations to endorse IAS 32 and IAS 39, both based on the same votes as the draft recommendations. Although each of these events continued to reect concern regarding IAS 39, the actions taken by Bolkestein to resolve the conicts, the support of IAS 39 by a major European bank, and EFRAG’s recommended endorsement of IAS 39 all rected events that we assess as having increased the likelihood of IFRS adoption. On October 1, 2004, the ARC added its endorsement recommendation to that of EFRAG. However, the ARC did not recommend endorsement of IAS 39 as issued by the IASB. Rather, the ARC recommended that the EC ‘‘carve out’’ of IAS 39 the two parts of the standard that were the focus of the controversy. Endorsing IAS 39 with this carve-out would mean that IFRS as applied in Europe would differ from IFRS applied elsewhere in the world, thereby thwarting the goal of global convergence described in the 2002 EFRAG endorsement recommendation letters.

On November 19, 2004, the EC followed the ARC recommendation and endorsed a carve-out version of IAS 39. Because these two events indicated that the EC was willing to alter IFRS, we assess these events as decreasing the likelihood of IFRS adoption. The EC indicated its intention that the carve-outs be temporary, only in place until the technical controversies were resolved. On June 16, 2005, the IASB revised the fair value option in IAS 39, and on July 8, 2005 ARC recommended endorsement of it. The EC endorsed the revised fair value option on November 15, 2005, thereby eliminating one of the two carve-outs of IAS 39. 0 Because these three events supported the EC’s intention to eliminate the carve-outs and made IFRS as endorsed in the EU closer to IFRS as issued by the IASB, we assess these events as increasing the likelihood of IFRS adoption.

Prior Research and Expected Market Reaction

Little is known about how investors perceived the possibility of IFRS adoption in Europe. This study infers investor perceptions by examining the equity market reaction to events leading to the adoption. Prior research uses this approach to assess the perceptions of investors in U. S. rms regarding individual standards (e. g. , fair value accounting in Statement of Financial Accounting Standards [SFAS] No. 115 by Beatty et al. [1996] and Cornett et al. [1996]; and stock-based compensation accounting in SFAS No. 123 by Dechow et al. [1996]). However, the setting in our study, which investigates investor perceptions regarding an entire set of accounting standards, is perhaps more analogous to prior research that examines investor perceptions to broad legislation (e. g. , the Sarbanes-Oxley Act by Jain and Rezaee [2006], Zhang [2007], and Li et al. 2008]). It is unclear how investors in European rms would react to movement toward IFRS adoption.

It is possible that investors would react positively to movement toward IFRS adoption if, for example, they expected application of IFRS to result in higher quality nancial reporting relative to application of domestic standards, thereby enhancing nancial reporting transparency, and reducing information asymmetry and information risk and, thus, lowering cost of capital. This prediction is supported by prior research. For example, Barth et al. (2008) ds that application of International Accounting Standards (IAS), which comprise a large portion of extant IFRS, is associated with higher quality accounting amounts than application of non-U. S. domestic standards. Similarly, Karamanou and 10 As of the writing of this manuscript, the second carve-out relating to hedge accounting remains in place. The Accounting Review January 2010 American Accounting Association 40 Armstrong, Barth, Jagolinzer, and Riedl Nishiotis (2005) nds positive abnormal returns for a small set of non-U. S. ?rms announcing voluntary adoption of IAS between 1989 and 1999.

Diamond and Verrecchia (1991), Baiman and Verrecchia (1996), Leuz and Verrecchia (2000), and Barth et al. (2009), among others, nd that higher nancial reporting quality is associated with lower cost of capital. These studies are consistent with Aboody et al. (2004) and Easley and O’Hara (2004), which provide evidence that information risk is priced and, thus, its perceived reduction could result in an empirically detectable market reaction. Investors might also react positively to movement toward IFRS adoption if they expect application of IFRS to have positive cash ow effects. These effects could include reduced contracting costs (e. . , Beatty et al. 1996) or reduced scope for managerial rent extraction associated with greater nancial reporting transparency (e. g. , Hope et al. 2006). It also is possible that investors in European rms would react positively to movement toward IFRS if they believed IFRS would provide convergence benets. For example, Barth et al. (1999) nds that there can be positive market effects associated with convergence. 11 Similarly, Ashbaugh and Pincus (2001) nds that previous convergence efforts relating to IAS resulted in reductions in analyst forecasts errors. Pae et al. (2007) nds that m value reected in Tobin’s Q, increased after the passage of EU regulations intended to converge nancial reporting, particularly for rms with higher agency costs.

Alternatively, it is possible that investors in European rms would react negatively to movement toward IFRS if, for example, they believed IFRS would decrease nancial reporting quality. This could occur if investors believed IFRS would fail to either adequately reect regional differences in economies or accommodate countries’ differing political and economic features that led to existing differences in domestic accounting standards. 2 Investors might also believe that variation in the implementation and enforcement of IFRS could lead to an increase in opportunistic managerial discretion when applying IFRS. Ball (1995, 2006) and Daske et al. (2007), among others, point out that effective nancial reporting convergence requires consistent implementation and enforcement of standards. Unlike the SEC in the U. S. , there was no regulatory counterpart with enforcement authority that spanned the European member states to ensure consistent application of IFRS. Consistent with factors other than standards themselves affecting nancial reporting quality, Ball et al. 2003) reports no detectable information quality difference between East Asian rms with high-quality, i. e. , common law-based, accounting standards and those with lowquality, i. e. , code law-based, accounting standards. There also is evidence that substantial information quality differences within Europe remain even after convergence efforts that preceded the 2005 IFRS adoption EU mandate (e. g. , Tay and Parker 1990; Joos and Lang 1994). Investors also might have believed that any convergence benets arising from adoption of IFRS would be less than the costs to implement and transition to the new set of standards.

Data and Research Design

We infer investor perceptions relating to IFRS adoption by examining European rms’ equity return reactions to our 16 adoption events. We rst provide descriptive evidence on 11 12 Barth et al. (1999) shows that the net market effect of convergence is a function of two effects. The rst is the direct informational effect, i. e. , whether convergence increases or decreases accounting quality. The second is the expertise acquisition effect, i. e. , whether investors become experts in foreign accounting, which depends on how costly it is to develop the expertise.

Therefore, ex ante the net market effect of convergence is uncertain. For example, relative to domestic standards, IFRS generally specify a greater use of fair values, which some believe is more susceptible to opportunistic managerial discretion than are modied historical cost-based amounts.

Market Reaction to the Adoption of IFRS in Europe 41 the overall European market reaction to these events. We then focus our tests on determining whether particular rm characteristics explain cross-sectional variation in ms’ return reactions in a manner consistent with our predictions. Because the adoption of IFRS resulted from a process that evolved over several years, we draw our inferences from analyses of the market reactions associated with all 16 events taken together, rather than with each event separately. Our initial sample comprises all European rms for which event returns are available for all 16 events in Table 1, which yields an initial sample of 3,265 rms. Our crosssectional tests require additional data, which results in a smaller sample size of 1,956 rms. We base our tests on three-day value-weighted market-adjusted returns for these ms centered on each of the 16 event dates, CMARj,e, where j denotes rm and e denotes event. 13 We obtain daily price data between 2002 and 2005 from Datastream through Thomson One Banker Analytic. Table 2 provides a breakdown of the sample by country.

This table presents the sample composition by country. The sample includes all European rms with returns available for all 16 events in 2002–2005 listed in Table 1. 13 We obtain the same inferences if we require sample rms to have a three-day return for only one event. In addition, untabulated statistics provide additional assurance that our sample rms are representative of European rms. In particular, the Pearson correlations between daily non-event date returns for the initial sample portfolio and daily non-event returns for the MSCI European Index, which is a broad-based market index, and the 600 European ms included in the DJ STOXX Global 1800 Index are 0. 993 and 0. 985, respectively. The correlation between the portfolio returns on event and non-event dates for our initial sample and the sample on which we base our cross-sectional analyses is 0. 995. The Accounting Review January 2010 American Accounting Association 42 Armstrong, Barth, Jagolinzer, and Riedl Following the vast event-study literature (e. g. , Campbell et al. 1997), we market-adjust raw event returns to mitigate potentially confounding effects of global news occurring concurrently with our event dates.

Because our events affect a large subset of the global equity market—all of Europe—the appropriate market index is not obvious. We market adjust the returns by subtracting the corresponding three-day return to the Dow Jones STOXX Global 1800 Index excluding the 600 European rms in the index (DJ STOXX 1800 ex Europe, hereafter). 14 We base our inferences on the DJ STOXX 1800 ex Europe index because including returns of European rms in the market adjustment return would remove some of the effect we seek to document. However, untabulated ndings reveal that our inferences are unchanged if we use the full DJ STOXX 1800 Index.

Consistent with the existence of global news that affects returns of European rms and rms in other parts of the world, untabulated statistics reveal a signicantly positive Pearson correlation of 0. 551 between daily non-event returns for our initial sample of European rms and daily non-event returns for the DJ STOXX 1800 ex Europe. 15,16 Our event-study research design relies on a degree of equity market efciency in the sample countries that is sufcient to ensure the information related to each event is reected in equity prices during the event window in an unbiased manner.

In particular, the maintained hypothesis throughout our analysis is that equity prices reect unbiased expectations of the costs and benets of IFRS adoption conditional on available information. Although the size and liquidity of the European equity market suggests this is a tenable assumption, there is likely variation across markets within Europe during our sample period. If a sample country’s equity market is not sufciently efcient to reect event information within the event window, our tests can lack power or be biased (e. g. , Hirshleifer 2001).

Our tests also rely on both the correct identication of information events and there being no confounding news during the event windows. 17 Including non-events likely introduces noise and excluding relevant events likely reduces power; both of which can introduce bias. Our event selection procedures, described in Section II, are intended to minimize the 14 15 16 17 The stated objective of the Dow Jones STOXX Global 1800 Index (hereafter, DJ STOXX 1800) is to ‘‘provide a broad yet investable representation of the world’s developed markets. ’’ The index comprises the largest 600 ms, based on free oat market capitalization, from each of Europe, North and South America, and the Asia / Pacic region. Our sample comprises all European rms that meet our data requirements, including, but not limited to, the 600 largest European rms that are in the DJ STOXX 1800 Index. We do not limit our analyses to the European rms in the index for three reasons. First, IFRS adoption would affect all rms in the European market, not only the largest rms. Thus, limiting our analyses to the largest rms limits the generalizability of our inferences. Second, we predict and d evidence that investors’ reaction to European IFRS adoption events is smaller for larger rms. Thus, limiting our analyses to the largest rms could bias against us nding a signicant market reaction across our event dates. Third, limiting the analyses to the largest rms would reduce cross-sectional variation in our sample on dimensions relevant to our cross-sectional predictions, thereby reducing the power of our tests. It is possible that on our event dates investors in non-European rms revised their expectations of IFRS adoption outside of Europe, which would affect returns that comprise the DJ STOXX 1800 ex Europe index.

If so, then using the DJ STOXX 1800 ex Europe index to market-adjust our event returns could remove some of the effect we seek to document, i. e. , bias our market-adjusted returns toward zero. Nonetheless, we use the index to market adjust our returns because, as for European rms, we expect that any IFRS adoption effect is relatively small for the largest non-European rms, whose returns comprise the index, which mitigates against such bias. See Campbell et al. (1997, Chapter 4) for a detailed discussion of the assumptions and limitations of an eventstudy research design.

January 2010 Market Reaction to the Adoption of IFRS in Europe 43 likelihood of including non-events and excluding relevant events. 18 Regarding potentially confounding news, we search the U. S. and European editions of the Wall Street Journal, including ‘‘World Markets’’ articles, and the European edition of the Financial Times for non-IFRS related news during each event window. We observe other news within nearly all event windows. However, as Table 1 indicates, our predictions are based on one of two patterns of market reactions across event dates, depending on whether investors perceive that the benes of IFRS adoption exceed the costs, or vice versa. Our search for confounding news revealed no discernible pattern of good (bad) news during event windows for events that we assess as increasing (decreasing) the likelihood of IFRS adoption, or vice versa. That is, we do not observe a systematic alternative news pattern that would bias our inferences. Overall European Market Reaction Our descriptive evidence on the overall European market reaction to IFRS adoption in Europe focuses on market-adjusted event returns for a value-weighted portfolio of the initial sample of rms.

We view this evidence as descriptive because of the challenge in identifying the appropriate market index with which to adjust the returns. We construct portfolio event returns by value weighting each rm’s return based on the rm’s equity market value at the end of the most recent quarter prior to the event to obtain a portfolio CMARe for each event. We base our test statistics on portfolio returns because each portfolio return is unaffected by potential cross-sectional correlation (Sefcik and Thompson 1986), and portfolio returns for different events should be uncorrelated.

Because we assess some events as increasing and some as decreasing the likelihood of IFRS adoption, when calculating our test statistics we multiply the event returns for the latter by 1 to enhance interpretation of investors’ response to the aggregated events. We provide three statistics to test the signicance of the portfolio event returns. The rst is a t-test of whether the mean of the 16 event portfolio CMARes differs from zero. This statistic assumes normality of the return distribution and that our market adjustment is the proper benchmark for the expected market return and, thus, the expected marketadjusted return equals zero.

The standard deviation used to calculate the statistic is derived from the distribution of the 16 event portfolio returns. Consistent with Fama and MacBeth (1973), this assumes portfolio returns associated with different events are uncorrelated. The second statistic is a t-test for whether the mean of the 16 event portfolio CMARes differs from the mean of a distribution of similarly constructed non-event portfolio returns. This statistic assumes unequal variances for the event and non-event return distributions. We present this statistic because it admits the possibility that returns of European ms differ systematically from returns of rms in other regions. That is, the statistic does not assume that our market adjustment fully adjusts for the market return.

To the extent returns of European rms do not differ systematically from other rms, this statistic will display diminished power because of estimation noise. To form the distribution of non-event returns, we use returns for European rms for all days in our event years, 2002 to 2005, that do 18 It is possible that the June 19, 2002 and July 8, 2004 events, which nalized EFRAG endorsements, only conmed investors’ prior expectations and did not provide new information. This possibility requires assuming EFRAG’s nal endorsement is perfunctory. If this is the case, they would be non-events and, thus, including them in our tests likely would add noise. Regardless, untabulated ndings reveal that our inferences are unaffected by excluding rms’ returns associated with these two events. Also, some might interpret the February 3, 2004 event as signaling a decreased likelihood of IFRS adoption because Bolkestein delayed endorsement of two standards. Untabulated dings also reveal that our inferences are unchanged if we adopt this alternative interpretation of this event. The Accounting Review January 2010 American Accounting Association 44 Armstrong, Barth, Jagolinzer, and Riedl not overlap with our event windows. We calculate the 300 three-day portfolio returns, ensuring non-overlapping windows, and subtract the corresponding three-day return to the DJ STOXX 1800 ex Europe Index. 19 The third statistic reects the probability that the mean of the 16 event portfolio CMARes exceeds the mean of 16 similarly constructed randomly selected non-event portfolio returns.

This statistic assumes the distribution of these non-event returns is the same as the distribution of event returns, but does not rely on any other distributional assumptions. However, to the extent the randomly selected non-event portfolio returns do not reect the population of portfolio returns, there could be noise or bias in this statistic. To construct this statistic, we randomly select 16 non-event portfolio returns from non-event dates that mimic the year-by-year distribution of our sample events.

That is, we select three, four, six, and three non-event portfolio returns from 2002, 2003, 2004, and 2005, respectively. We then designate one of the non-event portfolio returns from 2003 and two from 2004 as being associated with events that decrease the likelihood of IFRS adoption. We then compare the standardized mean of the non-event portfolio returns to the standardized mean of the 16 event portfolio CMARes. 20 We repeat this procedure 500 times to construct a simulated p-value for the probability that the standardized mean non-event portfolio return is greater than the standardized mean portfolio CMARe. 1 Cross-Sectional Analysis We base our main inferences on tests of whether rm characteristics explain crosssectional variation in the market reaction to IFRS adoption events. This analysis assumes that investors assess the expected costs and benets of IFRS adoption, including those related to accounting information quality, and implementation and transition.

As in our portfolio return statistics, when estimating Equation (1) we multiply by 1 returns associated with events we assess as decreasing the likelihood of IFRS adoption. Our proxy for the rm’s pre-adoption information quality is InfoQualFactor, which is the rst principal component derived from four variables selected to capture information quality. 22 The four variables are ADR, an indicator variable that equals 1 if a rm crosslists in the U. S. using American Depository Receipts (ADR) during the event year, and 0 otherwise; Standards Applied, an indicator variable that equals 1 if the m applies U. S. standards or International Accounting Standards (IAS) during the event year, and 0 if the 19 20 21 22 There are approximately 75 non-event three-day windows in each of the four sample years, which results in 300 non-event return windows. To ensure that our non-event returns do not overlap with each other, we select every fourth trading day as the beginning of the three-day return window. Our inferences are unaffected by altering the starting point for this trading day selection. We standardize means by dividing the mean return by the standard error of the 16 event return distribution.

This procedure effectively calculates how often our event distribution t-statistic is larger than a t-statistic estimated from a similarly constructed distribution of non-event dates. Bootstrap p-values indicate the likelihood of obtaining a similar magnitude statistical rejection of the null hypothesis on non-event dates. See, for example, Hein and Westfall (2004). InfoQualFactor is estimated using varimax orthogonal rotation. The rst and second principal component eigenvalues are 1. 75 and 0. 99, respectively.

January 2010 Market Reaction to the Adoption of IFRS in Europe 45 ?rm applied domestic standards;23 Exchanges, the number of exchanges on which the rm is listed during the event year; and Size, the natural logarithm of the rm’s prior end of year market value of equity. We expect ADR rms to have higher quality pre-adoption information because these rms are subject to U. S. securities regulation and enforcement, are required during our sample period to reconcile domestic standards-based net income and equity book value to those based on U.

S. standards, and are typically larger and more widely followed by analysts. 24 We expect that rms applying U. S. standards or IAS, rms listed on multiple exchanges, and rms that have larger equity market values to have higher quality pre-adoption information. To ease interpretation, we multiply factor scores by 1 so that higher values of InfoQualFactor correspond to lower quality information. If investors perceive the benets to IFRS adoption are higher for rms with lower quality pre-adoption information, then we expect 1 is positive.

To provide insight into whether the event market reactions reect perceptions regarding IFRS generally or IAS 39 in particular, we include in Equation (1) Bank, which is an indicator variable equal to 1 if the rm’s primary two-digit SIC code is 60 or 61, i. e. , depository institutions and non-depository credit institutions, and 0 otherwise. We include Bank because IAS 39 gured prominently in banks’ resistance to EU IFRS adoption. Investors in banks might react more positively than investors in other rms if they perceive informational benes associated with IAS 39, such as those associated with nancial statement recognition of previously unrecognized derivative ?nancial instruments, leading to a positive predicted sign for 2. However, investors in banks might react more negatively if they perceive that banks would incur more IFRS-related transition costs, such as those associated with extensive systems changes to account for large portfolios of nancial instruments and hedging activities or raising the attention of banking regulators if they report more volatile earnings based on IFRS than based on their domestic standards.

This would lead to a negative predicted sign for 2. Thus, we do not predict the sign of 2. Equation (1) also includes the interaction variable InfoQualFactor * Bank, which is intended to capture any incremental market reaction for those banks with lower quality pre-adoption information as reected in InfoQualFactor. For reasons discussed above, we do not predict the sign of 3. Equation (1) also includes three proxies for pre-adoption information asymmetry among investors or between the rm and investors. The rst is Turnover, which is an indicator variable that equals 1 if the m’s ratio of average number of daily shares traded to average total number of shares outstanding for the year is greater than the sample median, and 0 otherwise. The second is CloselyHeld, which is the percentage of shares held by insiders, as provided by Worldscope. The third is Herf, which is the Herndahl index, measured as the sum of squared market shares, i. e. , percentage of total industry sales, for all rms in the rm’s primary two-digit SIC code. Thus, Herf ranges from 0 to 1, with higher values indicating that within-industry sales are concentrated among fewer rms.

We expect that rms with lower turnover, with greater insider ownership, and that have less industry competition have more information asymmetry. If investors expect IFRS adoption to reduce information asymmetry, then they will react more positively to increases in the likelihood 23 24 Belgium, the Czech Republic, Denmark, Finland, Germany, the Netherlands, and Switzerland permitted rms a choice of accounting standards prior to mandatory IFRS adoption. We identify ADR rm years from the Bank of New York Complete Depository Receipt record (http: / / 160. 254. 123. 37 / dr directory. sp), which indicates both the type and date of the ADR listing. We use only Level II or Level III ADRs, not Level I. This is because during our sample period rms with Level II and Level III ADRs were required to provide domestic-to-U. S. standards-based reconciliations and were subject to more stringent requirements than rms with Level I ADRs. The Accounting Review January 2010 American Accounting Association 46 Armstrong, Barth, Jagolinzer, and Riedl of IFRS adoption for rms with greater pre-adoption information asymmetry. This would be consistent with investors perceiving a reduction in the m’s future cost of capital. Thus, we expect 4 is negative, and 5 and 6 are positive.

However, if investors perceive that IFRS adoption will require rms with dominant industry market share to disclose proprietary information, then we expect 6 is negative. Finally, Equation (1) includes two proxies for enforcement and implementation of accounting standards. The rst is Code, which is an indicator variable that equals 1 if a rm is domiciled in a code law country, and 0 otherwise. 25 Investors may expect that nancial reporting standards are less stringently enforced in code law countries (Ball et al. 000, 2003). Therefore, rms in code law countries may retain greater exibility in the application of IFRS. If this is the case, we expect 7 is negative. The second proxy is Big4, which is an indicator variable that equals 1 if the rm’s auditor during the scal year is one of the four largest, as reported by Worldscope, and 0 otherwise. Prior research nds that larger audit rms provide higher quality audits (DeAngelo 1981), and that investors perceive and price quality differences associated with stronger monitors (e. g. , Hogan 1997; Muller and Riedl 2002). Thus, investors may expect larger audit ms to provide more stringent enforcement and have more resources available to facilitate IFRS transition. If this is the case, we expect 8 is positive. To account for potential cross-sectional correlation among the residuals from Equation (1), we calculate two-way clustered standard errors based on two-digit industry and country (Rogers 1993; Gow et al. 2010; Petersen 2009). 26 We cluster along these two dimensions because it is reasonable to assume that nancial reporting practices and changes therein are more homogeneous within industries and countries than across industries and countries.

Results

Overall European Market Reaction Table 3 presents the portfolio event returns statistics. For each of our 16 events, we present the raw return to the portfolio of 3,265 European rms (column labeled ‘‘Raw Return Europe’’); the DJ STOXX 1800 ex Europe index (column labeled ‘‘DJ STOXX 1800 ex Europe Index Return’’); and the difference between them, which is the marketadjusted European return (column labeled ‘‘Market-Adjusted Return Europe’’). Table 3 presents the individual event portfolio returns with their predicted and actual signs; none of the individual event portfolio returns is signiantly different from zero. To compute the mean of the 16 event returns, which we present at the bottom of the table, we multiply by 1 returns associated with events we assess as decreasing the likelihood of IFRS adoption. Thus, we predict that the mean of the event portfolio returns is positive if the benets to IFRS adoption outweigh the costs. We nd that the mean raw return associated with the 16 events for the European portfolio is 0. 0052, versus 0. 0086 for the DJ STOXX 1800 ex Europe index. The difference of 0. 0034 is positive and significantly different from zero (t-statisticvs0 2. 627) and signiantly different from the mean 1. 980). Further, bootstrap estimation difference for the non-event returns (t-statisticvs300 reveals there is less than a 1 percent chance of randomly drawing 16 non-event returns with a standardized mean larger than the mean for our 16 events (p-valuebootstrap 0. 008). 27 25 26 27 All of our sample countries except the U. K. and Ireland are classied as code law countries.

Predicted sign relates to predictions for the sign of Market-Adjusted Return Europe. Mean Return across Events is computed as the mean of the individual event returns, after multiplying by 1 returns from events with a negative predicted sign. t-statisticvs0 assesses whether the mean return differs from 0. t-statisticvs300 assesses whether the mean return differs from the mean return for 300 non-overlapping non-events, chosen across the sample period 2002–2005. p-valuebootstrap is the proportion of 500 draws for which the standardized mean return across 16 randomly selected non-events exceeds the standardized mean event return.

Each draw of randomly selected non-events reects the year-by-year distribution of events. Market Reaction to the Adoption of IFRS in Europe 49 We also conduct these tests separately on the 13 events that we assess as increasing the likelihood of IFRS adoption, and the three events that we assess as decreasing this likelihood. Disaggregating events in this fashion reduces the chance that our event returns reect spurious market reactions to non-IFRS news because such non-IFRS news not only would have to coincide with our event dates, but also would have to be of the same sign that we predict for our event.

Untabulated results tend to support the inference of a general positive (negative) investor reaction to events that increased (decreased) the likelihood of IFRS adoption. In particular, for the 13 events increasing adoption likelihood, untabulated results reveal a mean positive market-adjusted reaction of 0. 0030. t-statisticvs0 (1. 896) and p-valuebootstrap (0. 054) at least marginally support the inference of a positive reaction to these 13 events. t-statisticvs300 (1. 450), however, does not support this inference. For the three events decreasing adoption likelihood, untabulated results reveal a mean market-adjusted reaction of 0. 053 that is negative, and signicantly different from zero based on all three 4. 988; t-statisticvs300 3. 760; p-valuebootstrap 0. 008). statistics (t-statisticvs0 Collectively, the tabulated market-adjusted mean reactions are consistent with investors perceiving that the benets to IFRS adoption outweigh the costs. However, the negative mean raw portfolio return indicates that inferences regarding investors’ overall reaction to European IFRS adoption events are sensitive to whether the raw portfolio returns are market-adjusted, and potentially to the choice of index for market-adjustment.

For this reason, we base our inferences regarding investors’ reaction to IFRS adoption events on the cross-sectional estimation. Cross-Sectional Analysis Table 4, Panel A, presents descriptive statistics for the variables used in Equation (1). We estimate Equation (1) using observations for which data are available for all 16 events and for all 300 non-events (n 31,296 rm-event observations for 1,956 rms). 28 Panel A reveals that 5. 27 percent of the sample rms are banks and an average of 41. 43 percent of rms’ outstanding shares are held by insiders.

Panel A also reveals that within-industry 0. 034), 61. 25 percent of sample rms sales are relatively dispersed across rms (Herf are domiciled in code law countries, and 74. 18 percent of rms are audited by one of the four largest auditing rms. The panel also reveals that almost 11 percent of rms have ADR listings, 24. 76 percent previously applied international or U. S. GAAP, and most rms list shares on only one exchange. Table 4, Panel B, presents Pearson correlations between the variables. Consistent with our expectations, Panel B reveals that CMAR is signiantly positively correlated with InfoQualFactor, InfoQualFactor * Bank, and CloselyHeld, and signicantly negatively correlated with Turnover, Code, and Size. The correlations between CMAR and Bank, Herf, and Big4 are not signicantly different from zero. Table 5 presents regression summary statistics from Equation (1), both excluding and including Size as a control variable. The table also presents two test statistics for each coefcient estimate. The rst (in parentheses) is a t-statistic calculated using two-way clustered standard errors based on industry and country.

The second (in brackets) is a t-statistic associated with a test of whether event date coefcient estimates differ from non-event date coefcient estimates. 29 28 29 Our inferences are the same if we estimate Equation (1) using observations for which data are available for only all 16 events (n 32,848 rm-event observations for 2,053 rms). Specically, coefcient difference t-statistics (in brackets) are estimated for the same set of 1,956 rms using a stacked estimation of Equation (1) for the 16 event dates and for the 300 non-event dates.

The Accounting Review

ADR is an indicator variable that equals 1 if a rm cross-lists in the U. S. using American Depository Receipts during the event year, and 0 otherwise. Standards Applied is an indicator variable equal to 1 if the rm applies U. S. standards or International Accounting Standards during the event year, and 0 if the rm applies domestic (continued on next page)

Size is the log of the rm’s prior end of year market value of equity. Bank is an indicator variable equal to 1 if the rm’s primary two-digit SIC code is 60 or 61, and 0 otherwise. Turnover is an indicator variable equal to 1 if the rm’s mean daily percentage shares traded during the year is above the median for all rms, and 0 otherwise. CloselyHeld is the percentage of the rm’s shares outstanding held by insiders at the end of the scal year. Herf is the Herndahl Index, measured as the sum of each rm’s squared percentage market-share, calculated at the twodigit industry level.

Code is an indicator variable equal to 1 if the rm is domiciled in a country with a codebased legal system (all countries except the U. K. and Ireland), and 0 otherwise. Big4 is an indicator variable equal to 1 if the rm was audited by one of the four largest accounting rms, and 0 otherwise. t-statistic is in parentheses and is the regression coefcient scaled by the coefcient standard error corrected with two-digit SIC code and country double-clustering (Petersen 2009). t-statisticvs300 is in brackets and is derived from comparing the coefcient estimated in a regression of 1,956 ms across the 16 event dates to the coefcient estimated in a regression of 1,956 rms’ 300 non-overlapping non-event dates. Table 5 reveals that the coefcient on InfoQualFactor, 1, is positive and signicantly 0. 0017, t-statistic 5. 81). This indicates different from zero, as predicted (coefcient that market participants reacted more positively to IFRS adoption for rms with lower quality pre-adoption information, consistent with investors in these rms expecting greater improvements in information quality. Although the coefcient on Bank, 2, is not signi. 0002, t-statistic 0. 25), the coefcient on cantly different from zero (coefcient InfoQualFactor * Bank, 3, is positive and signicantly different from zero (coefcient 0. 0017, t-statistic 2. 24).

This is consistent with investors in banks with lower preadoption information quality expecting net benets associated with IFRS adoption. Table 5 also reveals that the coefcient on Turnover, 4, is negative and signicantly 0. 0023, t-statistic 3. 29) and the coefcient on different from zero (coefcient 0. 0031, CloselyHeld, 5, is positive and signiantly different from zero (coefcient 2. 57). These ndings are consistent with predictions that investors in rms t-statistic with higher pre-adoption information asymmetry expected a reduction in information asymmetry from adoption of IFRS. Table 5 also reveals that the coefcient on Code, 7, is 0. 0019, t-statistic 4. 11). 31 negative and signicantly different from zero (coefcient This nding is consistent with predictions that investors in rms domiciled in code law countries expected lower IFRS adoption benets because of enforcement concerns. 32 The coefients on Herf, 6, and Big4, 8, are not signicantly different from zero (coefcients 0. 0051 and 0. 0000, t-statistics 0. 65 and 0. 08).

We obtain identical inferences when we compare event coefcient estimates with nonevent coefcient estimates (t-statisticvs300). This indicates that our inferences are unlikely to reect systematic associations between the rm characteristics and market returns that might be observed on non-event dates. The second column of Table 5 also reveals that our inferences are not affected when Size is included as a control variable. 30 31 32 Untabulated dings from an analysis in which we replaced InfoQualFactor * Bank with Size * Bank reveal a signicantly negative coefcient on the latter interaction variable of –0. 0013 (t-statistic –7. 51), which is consistent with larger banks having an attenuated reaction to IFRS adoption. That is, smaller banks, which are more likely to have lower quality pre-adoption information, have a more positive reaction to IFRS adoption. Inferences relating to the other variables are the same as those obtained from Table 5. Even though code law rms comprise 61. 25 percent of the sample and the coefient on Code is signicantly negative, untabulated statistics reveal that the mean market-adjusted return across the 16 events for code law rms is signicantly positive. This inference may be confounded by other factors associated with code law country domicile, such as differences in institutional structure, e. g. , rm governance. In Section VI, we report results of analyses aimed at assessing the sensitivity of our inference regarding enforcement concerns to alternative enforcement proxies. The Accounting Review January 2010 American Accounting Association 54 Armstrong, Barth, Jagolinzer, and Riedl

Market Expectations Regarding Convergence Benets Evidence thus far indicates that investors reacted positively to IFRS adoption events related to expectations of both improved information quality and reduced information asymmetry. We now provide evidence on whether investor reactions also relate to expectations of convergence benets. To implement this analysis, we estimate Equation (1), but using dichotomous explanatory variables as follows. We dene LowInfoQualFactor as an indicator variable equal to 1 if the rm’s InfoQualFactor is above the sample median (that is, the m has lower quality pre-adoption information); LowTurnover as an indicator variable equal to 1 if the rm has below median share turnover; HighCloselyHeld as an indicator variable equal to 1 if the rm has above median insider share ownership; HighHerf as an indicator variable equal to 1 if the rm has above median industry sales concentration distribution; Code as an indicator variable equal to 1 if the rm is domiciled in a code law country; NonBig4 as an indicator variable equal to 1 if the rm does not have a Big 4 auditor; and SmallFirm as an indicator variable equal to 1 if the rm has below median m size. Each variable equals 0 otherwise.

We dene the variables in this way so that the intercept reects the mean reaction for rms with high-quality information and low information asymmetry prior to IFRS adoption. That is, the intercept reects the mean reaction for those rms with higher quality information, higher turnover, lower percentage of closely held shares, lower industry sales concentration, domiciled in common law countries, using a Big 4 auditor, and that are larger. If investors in these rms expect IFRS adoption to provide little improvement in the information quality of these ms, then any observed reaction to IFRS adoption events more likely reects expected convergence benets. We do not include bank variables in this estimation because we do not have predictions for industry-specic variation in expected convergence benets. Table 6 presents the ndings. Consistent with investors expecting convergence benets, Table 6 reveals a positive and signicant market reaction for rms with higher quality information prior to the adoption of IFRS (intercept 0. 0048, t-statistic 7. 57). That is, investors appear to have anticipated benets associated with convergence from ancial reporting based on several sets of domestic accounting standards to nancial reporting based on a single set of standards, i. e. , IFRS, beyond any benets associated with increased information quality. Inferences relating to the other variables in Table 6 are consistent with those we obtain from Table 5, and are not affected by including size (i. e. , SmallFirm) as a control variable. Alternative Pre-Adoption Information Quality Proxies To assess the robustness of our inferences with respect to InfoQualFactor, we estimate Equation (1) separately for each variable used to derive the factor.

Low Info Qual Factor is a dichotomous variable that equals 1 if InfoQualFactor is at or above the distribution median (indicating low pre-adoption information quality), and equals 0 otherwise. InfoQualFactor is the quality of the rm’s pre-adoption information, measured as the highest eigenvalue factor derived from principal components analysis of the variables ADR, Standards Applied, Exchanges, and Size. InfoQualFactor is multiplied by 1 to ease interpretation, where higher values of InfoQualFactor indicate lower quality pre-adoption information. ADR is an indicator variable that equals 1 if a m cross-lists in the U. S. using American Depository Receipts (ADR) during the event year, and 0 otherwise. Standards Applied is an indicator variable equal to 1 if the rm applies U. S. standards or International Accounting Standards during the event year, and 0 if the rm applies domestic standards. Exchanges is the number of exchanges in which the rm lists during the event year. Size is the log of the rm’s prior end of year market value of equity. LowTurnover is an indicator variable equal to 1 if the rm’s mean daily percentage shares traded during the year is below the median for all ms, and 0 otherwise. HighCloselyHeld is a dichotomous variable that equals 1 if the percentage of the rm’s shares outstanding held by insiders at the end of the scal year is at or above the distribution median, and 0 otherwise. HighHerf is a dichotomous variable that equals 1 if the sum the rm’s squared percentage market-share is at or above the distribution median of the Herndahl index, calculated at the two-digit industry level, and 0 otherwise. Code is an indicator variable equal to 1 if the rm is domiciled in a country with a code-based legal system (all countries except the U.

K. and Ireland), and 0 otherwise. NonBig4 is an indicator variable equal to 1 if the rm was not audited by one of the four largest accounting rms, and 0 otherwise. SmallFirm is a dichotomous variable that equals 1 if Size is below the distribution median, and 0 otherwise. t-statistic is in parentheses and is the regression coefcient scaled by the coefcient standard error corrected with two-digit SIC code and country double-clustering (Petersen 2009). The Accounting Review January 2010 American Accounting Association 56 Armstrong, Barth, Jagolinzer, and Riedl

Evidence in Table 7 highlights the difculty in isolating a single underlying variable that sufciently captures pre-adoption information quality. Creating a factor by using the four variables together to characterize a rm’s information quality, as in the estimation presented in Table 5, appears to improve our ability to capture this construct and, thus, increase the power of our test. Alternative Enforcement Environment Proxies Our primary cross-sectional analyses rely on the variable Code to assess whether the reaction to IFRS adoption differs for ms domiciled in code law and common law countries. We interpret Code as capturing the effect of weaker enforcement on expected benets associated with IFRS adoption because the estimation equation includes controls for other rm characteristics, including the quality of pre-adoption information. Nonetheless, it is possible that Code also captures other rm characteristics, including some related to the rm’s pre-adoption information quality. Thus, to assess the robustness of our inferences related to enforcement, we re-estimate Equation (1) using three alternative enforcement proxies.

The alternative proxies are Rule of Law, Control of Corruption, which we obtain from Kaufmann et al. (2007), and the average of these two variables, Average Enforcement. 34 Rule of Law is an annual country-specic measure of contract enforcement quality, and police and court system quality. Control of Corruption is an annual country-specic measure of the extent to which public power is exercised for private gain and the degree of capture of the state by private interests. We multiply the Kaufmann et al. (2007) variables by 1 so that larger values indicate weaker enforcement environments, analogous to the way we dee Code. Consistent with enforcement being weaker in code law countries, untabulated ndings reveal that the correlation between Average Enforcement and Code is positive, 0. 333, and signicantly different from zero. Table 8 presents the ndings and reveals results consistent with those in Table 5. Specically, the coefcients on all three enforcement variables are negative and marginally 0. 0017, t-statistic 1. 84), signicantly different from zero: Rule of Law (coefcient 0. 0013, t-statistic 1. 70), and Average EnforceControl of Corruption (coefcient 0. 0015, t-statistic 1. 76).

Inferences relating to the other variables ment (coefcient are similar to those obtained from Table 5. Collectively, the ndings in Table 8 support the inference that the market reaction is less positive for rms domiciled in jurisdictions likely to have weaker enforcement.

Conclusion

This study investigates the European equity market reaction to 16 events associated with the adoption of IFRS in Europe. IFRS adoption resulted in a broad cross-section of rms domiciled in European countries with a variety of domestic accounting standards changing to a common set of standards at the same time.

The prospects of IFRS adoption 33 34 The lack of signicance for the ADR * Bank coefcient could result from the small number of observations underlying this coefcient estimation. In particular, only 12 banks in the sample have ADR listings. We use the Kaufmann et al. (2007) variables, rather than those provided by La Porta et al. (1998), because the Kaufmann et al. (2007) variables vary over time and align better in calendar time with our study than do the non-time-varying La Porta et al. (1998) measures.

Standards Applied is an indicator variable equal to 1 if the rm applies U. S. standards or International Accounting Standards during the event year, and 0 if the rm applies domestic standards. Exchanges is the number of exchanges in which the rm lists during the event year. Size is the log of the rm’s prior end of year market value of equity. Bank is an indicator variable equal to 1 if the rm’s primary two-digit SIC code is 60 or 61, and 0 otherwise. Turnover is an indicator variable equal to 1 if the m’s mean daily percentage shares traded during the year is above the median for all rms, and 0 otherwise. CloselyHeld is the percentage of the rm’s shares outstanding held by insiders at the end of the scal year. Herf is the Herndahl Index, measured as the sum of each rm’s squared percentage market-share, calculated at the twodigit industry level. Enf Environ is one of three proxies for the rm’s enforcement environment, derived annually at the country level from Kaufmann et al. (2007). To ease interpretation, the Kaufmann et al. (2007) measures