Mandatory Adoption of IFRS and Stock Price Informativeness

Mandatory Adoption of IFRS and Stock Price Informativeness Christof Beuselinck Tilburg University and CentER Philip Joos Tilburg University and CentER TiasNimbas Business School Fellow Inder Khurana University of Missouri at Columbia Sofie Van der Meulen Tilburg University and CentER First draft: September 2008 Comments appreciated

Mandatory Adoption of IFRS and Stock Price Informativeness Abstract: In this paper, we examine whether mandatory adoption of IFRS ameliorates firm opacity and contributes to stock price informativeness. Using Dasgupta et al. s (2007) framework, we predict a decrease in synchronicity around the time of mandatory adoption of IFRS, and a subsequent increase in synchronicity in the post-ifrs adoption period. Using a sample of 2,173 mandatory IFRS adopters in 14 EU countries, we find stock return synchronicity decreased in 2005 (the year of mandatory IFRS adoption), but subsequently increased in the post-adoption years to levels higher than the pre-adoption period. We also find synchronicity increases more in the post-ifrs adoption period when analyst activity is higher. In contrast, we find synchronicity returned to pre-ifrs adoption levels during the post-ifrs adoption period for firms with higher institutional ownership, which is consistent with a continuing private information advantage enjoyed by institutional investors under the IFRS regime. Overall, our evidence adds to the growing stream of literature on the consequences of mandatory IFRS adoption.

1 Mandatory IFRS Reporting and Stock Price Informativeness 1. Introduction The mandatory adoption of International Financial Reporting Standards (IFRS) by listed enterprises in the European Union (EU) beginning in 2005 represents a significant regulatory reporting change without historical precedent. 1 The change not only increased substantially the number of enterprises that apply IFRS to their consolidated reports but also mandated the use of IFRS by enterprises that vary considerably in size, ownership structure and financial reporting sophistication (Schipper 2005). In this paper, we examine whether mandatory adoption of IFRS ameliorates firm opacity and contributes to stock price informativeness. 2 In part, our inquiry is motivated by recent theory which emphasizes the role of information markets on asset price movements. Stock prices change to reflect new information about a firm, its industry or the market (Roll 1988). The extent to which total stock return volatility is explained by variation in industry-level or market-level return is referred to as systematic volatility, and the remaining volatility is referred to as idiosyncratic volatility. Given this definition, greater idiosyncratic volatility also implies lesser synchronicity (i.e. stock return comovement with market-wide and industry-level returns). Ferreira and Laux (2007, p. 952) note that idiosyncratic volatility is a good candidate for a summary of information flow, especially for private information about firms. 1 In this paper, we use the term IFRS to refer to all standards issued by the International Accounting Standards Committee or the committee s successor, the International Accounting Standards Board, even though the standards issued by the former are typically referred to as International Accounting Standards (IAS). 2 As noted by Tobin (1984), the fundamental economic role of the stock market is to generate prices that serve as public signals for allocating capital to productive uses and thereby contributing to economic growth. To the extent that stock prices impound firm-specific information, they serve to inform investors about firms investment opportunities and assist in directing resources towards firms with superior prospects (Morck et al. 2000; Durnev et al. 2003, Wurgler 2000).

2 Dasgupta et al. (2007) present a simple theoretical model that predicts a decrease in synchronicity at the time when new information is disclosed and impounded into stock prices, due to the fact that more firm-specific information becomes available at the time of adoption of a new information system. However, the model also predicts a subsequent increase in synchronicity in the future as an indication of a more transparent information environment. That is, the new and improved information allows investors to improve their predictions about the occurrence of future firm-specific events. Consequently, stock prices are expected to incorporate the likelihood of the occurrence of these future events, and when these events actually happen in the future, investors react less to such news, and therefore, return synchronicity will be higher. In our context, assuming that mandatory IFRS adoption improves the information or disclosure regime, we predict a decrease in synchronicity around the time of mandatory adoption of IFRS, and a subsequent increase in synchronicity in the post-ifrs adoption period. Next, we investigate the influence of two informed market participants (financial analysts and institutional investors) on the predicted synchronicity patterns around mandatory IFRS adoption. To the extent IFRS adoption improves the comparability of financial reports among firms, financial analysts are likely to increase the amount of industry-level information in prices because of their ability to better interpret and disseminate common information across all firms in the industry (Piotroski and Roulstone 2004; Ramnath 2002). The implication is that greater analyst activity exacerbates the predicted synchronicity patterns around the mandatory IFRS adoption. In contrast, prior research (e.g., Piotroski and Roulstone 2004; Xu and Malkiel 2003) argues that institutions possess an information advantage because of greater monitoring and that they are able to increase the relative flow of firm-specific information through their trading

3 activities. We predict that the level of institutional ownership attenuates the effect of mandatory IFRS adoption on stock return synchronicity. Last, we reanalyze our data by partitioning our sample into those countries with low versus high levels of enforcement of accounting standards to proxy for weak and strong institutions that affect the reporting incentives of firms. We also partition our sample based on the extent of divergence between a country s local GAAP and IFRS. The purpose of this analysis is to examine whether the predicted synchronicity patterns around mandatory IFRS adoption vary across countries with different reporting incentives. We test our predictions on a sample of 2,173 mandatory IFRS adopters in 14 EU countries. 3 Specifically, we use an interrupted time-series design to investigate the synchronicity patterns in the year 2005 when IFRS was mandated in EU for listed enterprises and in the postadoption years (2006-2007) relative to pre-ifrs years (2003-2004). We focus on European firms because Europe offers a unique setting for studying the influence of accounting standards on financial reporting outcomes. First, EU capital markets are well-integrated, capital mobility is high across Europe, and economic harmonization across countries is well established (Adjaouté and Danthine 2005). This feature allows us to treat EU countries as being homogeneous, thereby enhancing our ability to rule out alternative explanations for observed results (Holthausen 2003). Second, IFRS are uniform among EU countries and they did not shift dramatically over time during the years 2005-2007. Third, different ownership structures across EU jurisdictions can allow us to shed light on reporting incentives (Schipper 2005). In short, studying stock price informativeness evolutions in an economically harmonized environment such as the EU can 3 We delete early voluntary IFRS adopters and late IFRS adopters (such as AIM firms on the London Stock Exchange) from our sample. Most prior accounting studies focused on the effects of IFRS adoption on financial reporting quality for voluntary adopters. However, only about 8% of firms voluntarily adopted IFRS before 2005 in the EU, and most voluntary adopters come from Germany and Austria.

4 provide novel insights into the role of institutions in affecting how information from a new disclosure regime (such as IFRS) is impounded into stock prices. Turning to our empirical results, we find stock return synchronicity decreased in 2005 (the year of mandatory IFRS adoption), but subsequently increased in the post-adoption years to levels higher than the pre-adoption period. The first time mandatory adoption of IFRS in 2005 led to a significant increase in disclosed information because firms not only explained transition effects due to the use of IFRS but also disclosed more footnotes about segments, pensions, sharebased payments, and other transactions that were not required to be disclosed under local GAAP. Subsequently, more timely (transparent) information under IFRS led to less surprises in the future and therefore greater synchronicity in the post-ifrs adoption period. These findings are consistent with our prediction of a decrease in synchronicity around the time of mandatory adoption of IFRS, and a subsequent increase in synchronicity in the post-ifrs adoption period. Our results are also similar for subsamples of countries classified on the basis of the level of enforcement of accounting standards and divergence between local GAAP and IFRS, although post-ifrs adoption effects are weaker for countries with weak enforcement and more divergence in local GAAP relative to IFRS. We also find that analyst activity, as measured by analyst revisions, exacerbates the synchronicity effect of mandatory IFRS adoptions in the post-ifrs adoption period. That is, synchronicity increases more in the post-ifrs adoption period when analyst activity is higher. This result holds irrespective of whether the subsample of countries have strong or weak enforcement of accounting standards or have more or less divergence in local GAAP. In contrast, we find an overall negative effect of institutional holdings on synchronicity in all three periods (pre-ifrs, IFRS adoption year, and post-ifrs adoption year) with no statistically

5 significant difference between pre-ifrs and post-ifrs adoption periods. This finding of synchronicity returning to pre-ifrs adoption levels during the post-ifrs adoption period for firms with higher institutional holdings is consistent with a continuing private information advantage enjoyed by institutional investors under the IFRS regime. Again, this result holds in all subsamples except for the subsample of countries with large difference in local GAAP relative to IFRS. Overall, our paper highlights the impact of IFRS on stock price informativeness and helps us understand one factor that can contribute to stock price informativeness, i.e., the financial accounting standards. Morck et al. (2000) show that less respect for private property by government is associated with more synchronous stock price movements. However, when they substitute good government index with an accounting standards index that is a direct measure of the sophistication of each country s accounting standards, they find accounting standards index itself is uniformly statistically insignificant. From this evidence, they conclude that either the effect of accounting standards is unimportant in explaining synchronicity, or that their measure of accounting standards index is flawed. In contrast, our results show that mandatory adoption of IFRS influenced stock price synchronicity in EU countries. Our study also contributes to extant research on the consequences of IFRS. Prior research by Kim and Shi (2008) shows that firms that voluntarily adopted IFRS in the 1998-2004 time period exhibit a decrease in stock return synchronicity and that this decrease in synchronicity is attenuated for firms with high analyst following. We also find a decrease in synchronicity in the year of mandatory IFRS adoption; however this is followed by a substantial increase in synchronicity in the post-adoption period. The effect on synchronicity in the post- IFRS adoption period is exacerbated by analysts and the quality of enforcement of accounting

6 standards. Overall, our results highlight the importance of reporting incentives in making mandated improvements in financial reporting (such as IFRS adoption) effective. The rest of the paper is organized as follows: Section 2 describes the related literature and develops our testable hypotheses. Section 3 discusses the sample and details the empirical methods. In section 4, we present empirical results. Section 5 summarizes and concludes the paper. 2. Related literature and hypothesis development 2.1 Prior research Prior research on IFRS has examined the impact of voluntary IFRS adoption on financial reporting outcomes. Using a sample of 80 firms that voluntarily adopted IFRS before 1993, Ashbaugh and Pincus (2001) find that analysts absolute forecast error decreases on average after switching to IFRS. In a similar vein, Barth et al. (2008) argue that IAS, by limiting allowable alternative accounting practices and providing a more consistent approach to accounting measurement, can produce accounting numbers of higher quality than those determined in accordance with firms home country GAAP. Consistent with their argument, they find 387 public firms applying IFRS over the 1994-2003 time period exhibit less earnings management, more timely loss recognition, and more value relevant accounting numbers than non-ias adopting firms. Thus, the implication is that IFRS can influence the quality of financial reporting. However, as noted by these authors (e.g., Ashbaugh and Pincus 2001; Barth et al. 2008), it is possible that results based on firms that voluntarily adopted IFRS cannot be attributable to a change in financial reporting system; rather these results can be driven by firm characteristics that cause such firms to switch from local GAAP to IFRS.

7 Kim and Shi (2008) investigate whether enhanced disclosures via voluntary adoption of IFRS in 34 countries around the world over the 1998-2004 time period influence stock price synchronicity. They find that IFRS adoption decreases stock price synchronicity and that this effect is attenuated for firms with high analyst following. However, Morck et al. (2000) find that accounting standards index used as a direct measure of the sophistication of a country s accounting standards does not explain synchronous stock price movements in 1995 for a crosssection of firms spanning 37 countries. Similarly, Wang and Yu (2008) find that the adoption of IFRS is not related to stock price synchronicity for a sample of 44 countries around the world over a 10-year period. However, they find some evidence of a negative relation between stock price synchronicity and accounting standards in common-law countries. In short, the evidence on the relation between synchronicity and voluntary IFRS adoptions is mixed. More recent research has begun to focus on mandatory adoption of IFRS by first-time adopters. For example, Daske et al. (2008) examine the economic consequences of mandatory IFRS adoption in 26 countries. They find that market liquidity (as reflected by bid-ask spread) improves around the time of mandatory IFRS adoption. They also find that the capital market benefits occur only in countries where legal enforcement is strong and where firms have incentives to be transparent. Beuselinck et al. (2008) examine timely loss recognition in the EU over the period 1991-2007 and find no significant change around 2005, the year of mandatory IFRS adoption in EU. Christensen et al. (2008) also find no evidence of a change in accounting quality (earnings management and early loss recognition) following mandatory IFRS adoption in Germany. In contrast, Byard et al. (2008) find that mandatory adoption of IFRS in EU countries resulted, on average, in an improvement in analysts forecast accuracy, and that this effect is more pronounced for firms in countries with better law enforcement. In this paper, we examine

8 whether mandatory adoption of IFRS ameliorates firm opacity and contributes to greater stock price informativeness. 2.2. Hypothesis Development Jin and Myers (2006) note that transparency affects the division of risk bearing between managers and investors. For example, in more transparent firms, insiders take on less firmspecific risk, while outsiders bear less market risk. As a result, more firm-specific disclosure results in a firm s stock price reflecting more firm-specific information and less stock price synchronicity. Dasgupta et al. (2007) show that when the information environment surrounding a firm improves and more firm-specific information is disclosed, return synchronicity will first decrease, which is consistent with Jin and Meyers (2006). However, to the extent market participants are able to improve their predictions about the occurrence of future firm-specific events from the disclosure of time-varying firm specific information (or disclosure of timeinvariant information about firm characteristics), return synchronicity will increase subsequently. In our context, mandatory IFRS adoption in 2005 means that during the 2005 calendar year, information mandated by IFRS was disseminated through availability of interim reports, press releases, and documents explaining transition effects of adopting IFRS, and culminated with the issuance of fully IFRS compliant reports in the next calendar year. These disclosures are analogous to Dasgupta et al. s characterization of lumpy new information being released: early disclosures during 2005 followed by the release of fully IFRS compliant reports in 2006 and 2007. Using Dasgupta et al. s framework, the prediction is that stock price synchronicity first decreases in the period leading up to full IFRS-compliant reporting and then increases. The hypotheses can be formally stated as follows:

9 H1: Stock price synchronicity decreases in the period leading up to full IFRS-compliant reporting, ceteris paribus. H2: Stock price synchronicity increases in the period of or following the full IFRScompliant reporting, ceteris paribus. A typical argument in support of the IFRS adoption drive is that it allows financial statement users such as investors to make more sound decisions and hence allow better flow of capital globally. While important, this view downplays the importance of a country s institutions in place and the quality of the local accounting standards in place. For example, Ball et al. (2000) show that the demand for accounting transparency varies internationally as a function of economic and political institutional variables. Similarly, based on evidence of both low timeliness and low asymmetric conservatism in reported earnings for a sample of firms in four East Asian countries, Ball et al. (2003) conclude that reporting incentives are important determinants of international accounting differences, and that information quality is not determined by accounting standards alone. At the extreme, ignoring these forces will result in imposing costs on firms without any associated benefits. We examine the implications of this view by first examining whether mandatory IFRS adoption contributes to more informative stock prices after controlling for firm, industry, and country characteristics. Second, we examine whether a firm s home country characteristics (in terms of the strength of enforcement of accounting standards and of the extent of accounting differences in local GAAP and IFRS) affect the impact of mandatory IFRS adoption on stock price informativeness.

10 3. Sample and Research Design 3.1 Sample Table 1 summarizes the sample selection procedure. We construct our sample from the list of all firms from 14 EU (EU15 minus Luxembourg) member countries that are covered by the Worldscope database. For each firm-year observation, we require that there is sufficient data on Worldscope to compute the financial data items and stock returns used in the empirical tests. To avoid the confounding effects of changes in firm coverage over time (such as the inclusion in later years of younger, less profitable, and more high-tech firms in the database), we restrict our sample to firms with complete data for each year during the 2003-2007 time period. Because the focus of this study is on the effects of mandatory IFRS adoption, we delete firms that voluntarily adopted IFRS prior to 2005 or adopted IFRS after 2007 (such as AIM firms on the London Stock Exchange). We also excluded firms in regulated industries with SIC codes 49 and 62 because equity values of regulated firms are expected to respond similarly to changes in underlying regulations and economic conditions (Piotroski and Roulstone 2004). Our final sample consists of 10,865 firm-year observations corresponding to 2,173 firms. [Insert Table 1] We require each firm to have at least 45 weekly returns available during each of the sample years. In addition, we truncate the financial variables used in our empirical tests at 1% in each tail of the distribution on an annual basis to reduce the effects of outliers. 4 4 The results are robust to inclusion of firm-year observations with financial variables values in the 1% tails of the annual distributions.

11 3.2 Research Design To examine the relation between stock price comovement (and, by implication, the firmspecific informativeness of stock prices) and the mandatory adoption of IFRS, we estimate the following model 5 : SYNCH i,t = α 0 + β 0 ADOPT i + β 1 POST_ADOPT i + γ j IND j + β 2 Log(NREV) i,t + β 3 INSTIT i,t Where: + β 4 Log(MCAP) i,t + β 5 HERF i,t + β 6 NIND i,t + β 7 GDPG i,t + ε i,t (1) SYNCH i,t = synchronicity of firm-level stock returns with market-wide and industry-level returns (based on Morck et al. 2000). ADOPT = dummy variable equal to 1 if a firm observation is from the 2005 calendar year and 0 otherwise. POST_ADOPT = dummy variable equal to 1 if a firm observation is from 2006 or later calendar year and 0 otherwise. IND j = industry fixed effects based on one-digit SIC code. NREV = number of analyst revisions of one-year ahead forecast of annual earnings [Source: IBES detailed files]. INSTIT = the proportion of institutional holdings in a firm in 2006 [Source: Amadeus]. MCAP = the market value of equity of the firm at beginning of the calendar year. HERF = a revenue-based Herfindahl index of industry-level concentration. NIND = the average number of firms used to calculate the weekly industrylevel return index. GDPG = inflation-adjusted country growth in gross domestic product [Source: World Bank development indicators CIA estimates for 2007] Model (1) uses the firm s stock price synchronicity as the dependent variable, and regresses it on the test variables ADOPT (2005 dummy) and POST_ADOPT (2006-07 dummy) and a set of control variables including firm, industry and country characteristics. As discussed 5 Since we are interested in the market-wide impact of IFRS reporting, we measure synchronicity consistently for all firms on a calendar-year basis (regardless of their fiscal-year end). We recognize that this might introduce some asynchronous timing, adding noise to the relation between IFRS reporting and relative firm-specific stock return variation. In unreported analyses, we conduct a robustness check by dropping non-december fiscal-year firms (approximately 27% of our sample). Unreported results for December fiscal year-end firms are qualitatively similar to those reported in the paper using all firms.

12 above, higher values of stock price synchronicity imply lower firm-specific informativeness of stock prices. The dependent and independent variables in model (1) are discussed further below. 3.2.1 Dependent Variable Following prior research (Durnev et al. 2003, Piotroski and Roulstone 2004), we compute the stock price synchronicity for each sample firm and calendar year (variable SYNCH). Specifically, for each firm-year, we regress the firm-level weekly return (RET i,w ) on the current and prior week s value-weighted market-wide return (MARET i,w ) and on the current and prior week s value-weighted industry-level return (INDRET i,w ) based on two-digit SIC codes. 6 Specifically, the model is as follows: RET i,w = α + β 1 MARET i,w + β 2 MARET i,w-1 + β 3 INDRET i,w + β 4 INDRET i,w-1 + ε i,w (2) where, MARET equals the value-weighted market-wide return for the week w and INDRET equals the value-weighted industry-level return for week w using all firms in the market and industry (excluding firm i), respectively, and where industry is defined based on the same two-digit SIC code as firm i. Model (2) is estimated for each sample firm using weekly data over the entire calendar year. 7 Following Morck et al. (2000), we define synchronicity as follows: SYNCH = R2 log 1 R 2 where R 2 is the coefficient of determination obtained from estimating model (2). The log transformation changes the R 2 variable, bound by zero and one, into a continuous variable with a 6 All results relating to the synchronicity variables reported in this paper define MARET and INDRET using country-weighted returns. Unreported results using a EU index and industry indices at the EU level yielded inferences similar to those reported in the paper. 7 In estimating model (2), we follow Piotroski and Roulstone (2004) and require the regression for each firm to have a minimum of 45 weekly observations in each year.

13 more normal distribution. Variable SYNCH is the dependent variable in model (1) discussed previously. By construction, higher values of this variable reflect higher comovement and lower firm-specific informativeness of stock prices. 3.2.2 Test Variables The use of IFRS was made mandatory for all fiscal years commencing on or after January 1, 2005 for all listed-firms in the European Union (Deloitte 2005). In model (1) we introduce two dummy variables, ADOPT and POST_ADOPT. ADOPT reflects the calendar year 2005 or the year in which all EU-listed firms (EC Regulation No. 1606/2002) ultimately needed to adopt IFRS; POST_ADOPT refers to the years after 2005 when full IFRS-compliant reports became public. Both variables are introduced as main effect variables to capture stock market effects of mandatory IFRS adoption relative to the pre-period covering calendar years 2003 and 2004. Recent research (Piotroski and Roulstone 2004, Xu and Malkiel 2003) indicates that the trading and trade generating activities of informed market participant groups such as financial analysts and institutional investors influence the relative amounts of firm-specific, industry-level, and market-wide information impounded in stock prices. For this reason, we also examine alternative specifications of model (1) to explore the interaction effects of ADOPT and POST_ADOPT with informed market participant groups such as financial analysts and institutional investors. 3.2.3 Control Variables Following prior research (Piotroski and Roulstone 2004), model (1) includes several control variables previously shown to be related to stock price comovement. These variables include industry fixed effects (IND) to control for any relation between stock price comovement

14 and the industry to which the firm belongs. Also included as control variables are the market value of firm equity (MCAP) as a proxy for firm size, a proxy for industry-level concentration (HERF), the average number of firms used to calculate the weekly industry-level return index (NIND), and a country-specific measure for macro-economic activity (GDPG). The variable NREV measures the number of analyst revisions during the fiscal year and captures firm-specific analyst activities. Hameed et al. (2005) suggest that firms whose fundamentals are useful in assessing the fundamentals of other firms are covered by more analysts. Further, to the extent that analysts contribute industry-level or market-wide knowledge and expertise to the price formation process (Lys and Sohn, 1990; Park and Stice, 2000), the variable NREV is expected to be positively related to the dependent variable SYNCH. We obtain data on the number of analysts issuing earnings forecasts and subsequent revisions from the IBES detail tape. The variable INSTIT captures the role of institutional investors in the price formation process and is measured as the number of shares held by institutions, as a fraction of the number of shares outstanding at the beginning of the year. Piotroski and Roulstone (2004, p. 1121) note that the type of information conveyed by institutional trading is a complex function of several factors including trade size, pre-trade ownership stake, and the source of information. Xu and Malkiel (2003), however, conjecture and find that institutional trading may have a far greater impact on firm-specific (compared to market-level) return volatility since new information on individual stocks arrives much more frequently than market-wide information. Consistent with Xu and Malkiel (2003), we hypothesize a positive association between INSTIT and SYNCH. Further, we include MCAP to control for a firm s size and its associated effects on stock price synchronicity. To the extent that firm size is positively associated with various dimensions

15 of the firm s information environment not captured by NREV and/or INSTIT, firm size could negatively influence stock return synchronicity. Alternatively, if information acquisition is more costly for small firms, then, in equilibrium, investors may optimally choose to learn less about small firms (Kelly 2005). Thus, firm size could positively influence stock price synchronicity. Given these conflicting arguments, we do not predict the sign for the variable MCAP in the regressions. We also include a proxy for industry-level concentration (HERF) to control for industry concentration. The more concentrated an industry is, the more likely that firm performances are inter-dependent and news about one firm is perceived as value-relevant for the remainder industry peers (Piotroski and Roulstone 2004). Hence, we predict a positive relation between HERF and SYNCH. Next, we include a control variable for the average number of firms (NIND) used in computing weekly industry-level returns and resulting SYNCH measures (using Model (2)). The inclusion of NIND in the multivariate tests may additionally control for any differences in SYNCH arising from differences in sample sizes used for estimation purposes. Finally, also following prior research, we control for variation in economic conditions across countries (i.e., the state of the economy) by including the annual real GDP growth rate (variable GDPG); the higher the real GDP growth rate, the stronger the state of the economy (Kearney and Poti 2008). 3.2.4 Subsample Analyses To examine whether the influence of mandatory IFRS adoption on stock price informativeness varies across countries with different institutional settings, we partition our sample based on two country-level variables; one proxying for the strength of enforcement of accounting standards (ENFORCE) and one proxying for the divergence of local GAAP relative

16 to IFRS (GAAP_DIFF). We examine the enforcement of accounting standards because prior research (e.g., Ball et al. 2003; Hope 2003) argues that accounting is likely to matter most if the rules are properly enforced. Based on Hope (2003), we measure country-level enforcement as the factor-score on a five-factor model, reflecting (1) the level of audit spending in a country, (2) insider trading laws, (3) judicial efficiency, (4) rule of law, and (5) investor protection. 8 A higher score denotes stronger enforcement. Using median value of ENFORCE for all 14 EU countries, we partition our sample into a high and low ENFORCE subsample. We also examine the difference in accounting standards because recent research (e.g., Daske et al. 2008) suggests that IFRS adoption - both mandatory and voluntary- coincides with larger increases in market liquidity when the differences between local GAAP and IFRS are larger. To capture GAAP/IFRS differences, we follow Bae et al. (2008) and create a variable DIFF_GAAP by counting the number of reporting rules on 21 key accounting items in local GAAP that differ from specified IFRS rules. 9 Theoretically, DIFF_GAAP can range between 0 and 21, with higher values reflecting a larger difference between the country s local accounting standards and IFRS. Again, we perform a split based on the median values of DIFF_GAAP for all EU countries, and accordingly partition the sample into a high and low DIFF_GAAP 8 The first variable measures the audit spending in a country as the total fees of a country's ten largest auditing firms as a percentage of GDP for 1990 (Ali and Hwang 2000). The second variable captures the existence and enforcement (i.e., prosecutions) of insider trading laws based on data presented in Bhattacharya et al. (2003). The third variable is judicial efficiency, which measures the "efficiency and integrity of the legal environment as it affects business" (La Porta et al. 1998, 1124). The fourth variable is rule of law, which assesses a country's law and order tradition (La Porta et al. 1998, 1124). The last variable captures the voting rights of minority stockholders, which are proxied by the anti-director rights index calculated by La Porta et al. (1997). This index is based on six specific elements of investor protection namely, (1) the ability to vote by mail, (2) the ability to gain control during the shareholders meeting, (3) the possibility of cumulative voting for directors, (4) the ease of calling an extraordinary shareholders meeting, (5) the availability of mechanisms allowing minority shareholders to make legal claims against directors, and (6) preemptive rights that can be waived only by a shareholders vote. 9 Following Bae et al. (2008), we assign a score of 1 to a country if its rule is different from the rule specified by IFRS and 0 otherwise, for each key accounting variable. The sum of the scores across all the accounting items reflects the overall difference between the country s accounting standards and IAS.

17 subsample where high (low) DIFF_GAAP sample reflects more (less) difference in local GAAP relative to IFRS. 4. Empirical Results 4.1 Descriptive Statistics Panel A of Table 2 reports evolutions in the synchronicity measure across individual EU countries and for the whole EU market over three distinct time periods. Period 1 refers to the calendar years 2003 and 2004 when IFRS adoption was not mandatory. Period 2 is 2005 when mandatory IFRS adoption was effective and firms gradually started to produce IFRS-relevant information. Period 3 refers to the calendar years 2006 and 2007 or the years of full IFRScompliant financial statements. [Insert Table 2] The number of firm-year observations covered within a country range from 45 in Austria to 2,840 in the U.K. For the whole EU market, the dependent variable in our model (1), SYNCH, defined as log[r 2 /(1-R 2 )], where R 2 is the coefficient of determination from model (2) - - exhibits first a decreasing pattern until 2005 (from -1.58 in 2003-04 to -1.76 in 2005) and then an increasing pattern in the 2006-2007 time period (-1.24). On a univariate basis, this result suggests that synchronicity (stock price informativeness) went down (up) in the year when mandatory IFRS adoption became effective, and that synchronicity (stock price informativeness) went up (down) during fiscal years after the first year of mandatory IFRS adoption. This pattern in the evolution of SYNCH prevails for 10 of the 14 EU countries in our sample. The exceptions are Austria, Denmark, Ireland, and the UK. However, even for these 4 countries, the mean values of SYNCH in period 3 (the years after the first year of mandatory IFRS adoption) are higher than those when IFRS adoption was not mandatory.

18 The last two columns in Panel A list the country-level values of two variables, ENFORCE and DIFF_GAAP. There is considerable variation in the two country-level variables. The ENFORCE variable ranges between -3.65 for Spain to 1.16 for UK. An examination of the values of DIFF_GAAP variable reveals that Irish and UK GAAP have the smallest number of differences with IFRS on the 21 accounting items selected by Bae et al. (2008). In contrast, Greek GAAP has the largest difference (15 out of 21) with IFRS. Panel B of Table 2 shows evolutions in the synchronicity measure across different industries based on one-digit SIC codes for the same three time periods identified in Panel A of Table 2. Industries are fairly well represented in the sample. Twenty-one percent (2,325/10,865) of our sample observations belong to SIC code 3 (manufacturing). The least number of observations are from SIC code 8 (services other than entertainment, food and accommodation). More importantly, across all industries except SIC code 1 (agriculture, forestry, and fishing), the SYNCH variable exhibits first a decreasing pattern until 2005 and then an increasing pattern in the 2006-2007 time period. This result suggests that the patterns of SYNCH observed over the three time periods for the entire EU sample are not driven by a specific industry. Synchronicity patterns documented in panel A of Table 2 appear to persist across most industries classified on the basis of one-digit SIC code. 10 To check the validity of these time patterns, we also conduct an out-of-sample comparison test. If time trends in SYNCH reflect a worldwide structural pattern rather than a manifestation of the mandatory IFRS introduction, then we should find SYNCH to evolve in a similar way in a setting such as the US where no such IFRS adoption is mandated. Panel C of Table 2 reports difference-in-differences analyses between our balanced panel data set of 10 Untabulated results indicate that this pattern is consistently found when we classify industries according to twodigit SIC codes.

19 mandatory IFRS adopters in EU over the years 2003-2007 and a sample of US firms for which we require similar data requirements. Time periods are defined in a manner analogous to those defined in Panel A and B. Moreover, the sample composition is kept constant in all periods across all samples. We assess the statistical significance of the difference-in-differences values of SYNCH by comparing means of yearly firm-level changes across EU and US sample firms using t-tests. Results in Panel C of Table 2 show that US stock prices also evolve in a less synchronous way in 2005 (period 2) compared 2003-2004 (period 1), although the reduction is significantly less pronounced in the US (-0.136 versus -0.176; p<0.10). A comparison of SYNCH for US firms in 2006-2007 (period 3) with those in 2005 (period 2) indicates only a marginally significant increase in US synchronicity (0.057; p<0.10). Further, US stock prices in contrast to EU findings do not behave more synchronously in the 2006-2007 period compared to the period 2003-2004 (-0.079 versus +0.344; p<0.01). Taken together, the results in panel C of Table 2 indicate that over the 2003-2007 period, US stock markets follow a rather stable SYNCH pattern which differs from the SYNCH evolution patterns of the EU stock markets. Table 3 presents descriptive statistics about the firm-year observations included in our sample. The mean value of SYNCH is slightly higher than the median value, indicating that the distribution of this variable is a little right-skewed. The mean (median) number of earnings forecast revisions by analyst is 15.25 (2). The mean and median percentage of institutional holdings in a firm is 12.78 and 1.90 percent, respectively. The mean firm size (as measured by the market value of equity or MCAP) is 2,274 billion Euros. The mean value for HERF is 0.011.

20 The average number of firms used in calculating the weekly industry returns is 140. Finally, the mean real annual GDP growth rate over the period of our study (2003-2007) is 2.48 percent. 11 [Insert Table 3] 4.2 Correlations Table 4 presents the Pearson/Spearman pairwise correlations among the dependent and independent variables in model (1). The highest (absolute) pairwise Pearson/Spearman correlation among the control variables is 0.566/0.672 (between variables NREV and MCAP), indicating that larger firms have more revisions (or following) by financial analysts. Moreover, the correlations of INSTIT with MCAP and NREV are positive, suggesting that firms with more institutional following are larger and followed by (more) analysts who revise their forecasts more frequently. As expected, the correlation between HERF and NIND is negative, suggesting that the larger the number of firms in an industry, the lower the concentration ratio in the aggregate industry. All other correlations among the control variables are below 0.10. [Insert Table 4] The correlation between the dependent variable SYNCH and ADOPT is negative and statistically significant at the 0.01 level, suggesting that that there is less stock price synchronicity (i.e., there is more firm-specific information in stock prices) during the first year when IFRS were mandated. In contrast, SYNCH and POST_ADOPT are positively correlated, suggesting that there is more stock price synchronicity (i.e., there is less firm-specific information in stock prices) during the two years after the IFRS are mandated. These correlations are consistent with the evolutions in the synchronicity measure documented in Table 2 for the whole EU market over these two time periods. Moreover, SYNCH has positive and 11 Note that the number of observations is fewer (n=10,700) for GDPG variable because the GDPG estimate in 2007 was unavailable for Greece.

21 statistically significant correlations with NREV, MCAP, and GDPG. Below, we explore the relation between SYNCH and our test variables further in a multivariate context. 4.3 Full Sample Regression Results Table 5 reports regression results for alternative specifications of model (1) for our sample with variable SYNCH (stock price comovement or synchronicity) as the dependent variable. The first specification ignores the role of analysts and institutions and focuses on the main effects of our test variables, ADOPT and POST_ADOPT. The second and third specifications introduce the main and interaction effects of analysts and institutions activities, one at a time. The last specification represents a complete model that considers both the role of financial analysts and institutions in the regression model. As reported in Table 5, the explanatory power of alternative specifications of model (1) range between 0.297 and 0.314. [Insert Table 5] The tests of significance reported in Table 5 are based on robust t-statistics that are adjusted for residual correlation arising from pooling cross-sectional observations across time, i.e., t-statistics are based on White (1980) heteroskedasticity-adjusted robust variance estimates that are adjusted for within-cluster correlation where industry and fiscal year comprise the cluster (Cameron et al. 2006; Petersen 2006; Thompson 2005). Industry fixed effects (dummy variables) are not reported for brevity. Separately, unreported variance inflation factors (VIFs) were low suggesting that collinearity is unlikely to be a problem in interpreting the regression results. 12 In terms of our hypotheses, the coefficients on the main effects of ADOPT and POST_ADOPT are of interest. In Table 5, the coefficient on ADOPT is negative and statistically significant at the 0.01 level, indicating that stock price synchronicity declined with 12 The highest VIF is less than 3 (2.82) for Log(NREV) in the full sample.

22 the mandatory adoption of IFRS. In terms of economic magnitude, a mandatory switch to IFRS reduces stock price synchronicity by approximately 5.7% (the coefficient of -0.229 on the ADOPT variable divided by the coefficient of -4.035 on the intercept in column (1)). Since the firm-specific informativeness of stock prices varies inversely with stock price synchronicity, our findings suggest that the firm-specific information in stock prices increased in the first year of the mandatory adoption of IFRS. In contrast, the coefficient on POST_ADOPT is positive and statistically significant at the 0.01 level, indicating that stock price synchronicity increased (and the firm-specific information in stock prices decreased) in the years after the mandatory adoption of IFRS). This represents approximately 3% increase (computed by dividing the coefficient of 0.125 on the POST_ADOPT variable by the coefficient of -4.035 on the intercept in column (1)) in stock price synchronicity in the POST-IFRS adoption years. The coefficients on variable log(nrev) are positive and statistically significant at the 0.01 level, suggesting that analyst activities improve intra-industry information transfers resulting in greater stock price synchronicity. This result is consistent with Piotroski and Roulstone (2004) who find that analysts in the U.S. increase the amount of industry-level information in prices. Moreover, while the interaction of ADOPT with log(nrev) is not statistically significant at the 0.10 level, the interaction between POST_ADOPT and log(nrev) is statistically significant at the 0.01 level with a positive sign. This positive interaction effect suggests that analysts began to appreciate greater comparability of financial statements after the first year of mandatory IFRS adoption and that contributed to more stock price synchronicity. The insignificant interaction of ADOPT with log(nrev) suggests that early IFRS announcements/documents cause considerable firm return volatility, even in environments with high analyst activity, suggesting that analysts cannot offset the surprise effect.

23 Also, the coefficients on the variable, INSTIT, are negative and statistically significant at the 0.01 level, suggesting that institutional ownership is negatively related to stock price synchronicity. This result is also consistent with Xu and Malkiel (2003) and Roulstone and Piotroski s (2004) finding that institutions possess an information advantage and that they are able to increase the relative flow of firm-specific information through their activities. However, the interaction between POST_ADOPT and INSTIT is not statistically significant the 0.01 level, suggesting that synchronicity returned to pre-ifrs adoption levels during the post-ifrs adoption period for firms with higher institutional holdings. This result is consistent with a continuing private information advantage enjoyed by institutional investors under the IFRS regime. As for the other control variables, we discuss them only briefly. As a proxy for the size of the firm, MCAP is significant at the 0.01 level with a positive sign. By contrast, the variable HERF, which represents industry concentration, is statistically significant at the 0.01 level with the positive sign, which is surprising, given our prediction of a negative relation. Variable NIND (which represents the average number of firms used to create the weekly industry-level return index) appears to impact synchronicity negatively in two of the four alternative specifications, i.e., it is significant with a negative sign at the 0.10 level. Variable GDPG is also significant at the 0.01 level but with a positive sign indicating that stock price comovement increases (and, by implication, the firm-specific informativeness of stock prices decreases) with higher GDP growth rate. 4.4 Subsample Regression Results Table 6 reports OLS regressions results for several subsamples with SYNCH as the dependent variable. Specifically, Panel A reports regression results for subsamples partitioned

24 into two groups, High-ENFORCE and Low-ENFORCE, using the median value of ENFORCE variable for all 14 EU countries. Similarly, Panel B reports regression results for subsamples partitioned into two groups, High-DIFF_GAAP and Low-DIFF_GAAP, using the median value of DIFF_GAAP variable for all 14 EU countries. Within each subsample, we report results of two regression models one that excludes the main and interaction effects of variables proxying for analyst and institutional activities, and the other that includes effects of these variables. [Insert Table 6] Panel A of Table 6 shows that the explanatory power of alternative specifications of model (1) are larger in magnitude for subsample of firms operating in high enforcement countries than those in low enforcement countries (0.392 versus 0.267; 0.397 versus 0.289). Again, our main interest is in examining how IFRS adoption is associated with synchronicity (or stock price informativeness) in the two subsamples. Hence, our focus is on the ADOPT and POST_ADOPT variables. The coefficients on ADOPT are negative and statistically significant at the 0.01 level in both subsamples, suggesting that synchronicity reduces in the IFRS adoption year in both high and low enforcement regimes. However, the coefficient on POST_ADOPT variable is positive and statistically significant at the 0.01 level only for the HIGH enforcement subsample (0.232; p<0.01). This result suggests that synchronicity increases in the years after the first year of mandatory IFRS adoption for the high enforcement subsample but not for the low enforcement subsample. Stated alternatively, the level of enforcement of accounting standards appears to affect the influence of mandatory IFRS on stock price synchronicity. Moreover, for both subsamples of countries with high and low enforcement, synchronicity increases more in the post-ifrs adoption period when analyst activity is higher. However, the relative increase in synchronicity (post- versus pre-adoption period) due to high

25 analyst activity is greater in countries with higher enforcement: the coefficient on log(nrev), as shown in column (2), increases by 45% (0.055/0.121) in the high enforcement sample, and only by 25% (0.048/0.188), as shown in column (4), in the low enforcement sample. This finding is consistent with analyst activity affecting synchronicity more in strong enforcement environments. 13 In contrast, for both subsamples, there is no observable difference between periods (pre-ifrs, IFRS adoption year, and post-ifrs adoption year) in the negative effect of institutional holdings on synchronicity. That is, differences in levels of enforcement of IFRS do not affect firm-specific information access by institutional investors. Panel B of Table 6 shows that the explanatory power of alternative specifications of model (1) is larger in magnitude for subsample of firms operating in countries with smaller differences in their local GAAP than those with larger differences. More importantly, the effects of IFRS adoption on SYNCH are similar in the two subsamples. Specifically, the coefficients on ADOPT are negative and statistically significant at the 0.01 level in both subsamples, and the coefficients on POST_ADOPT variable are positive and statistically significant at the 0.01 level in both subsamples. However, the relative increase in synchronicity (post- versus pre-adoption period) is greater in countries with less IFRS-local GAAP differences: the coefficient on POST_ADOPT relative to the intercept in column (3) is 6% (0.218/3.418) for countries with small differences between local GAAP and IFRS, versus 2.6% (0.078/2.933) for countries with a large difference in local GAAP. This finding is consistent with the level of local GAAP-IFRS 13 In unreported analysis, we find the mean level of analyst revisions in the pre-ifrs period in high enforcement countries is similar to that in low enforcement countries (around 13 revisions per year in 2003 and 14.5 in 2004). The mean number of revisions increases to 17 in low enforcement countries and 22 in high enforcement countries in 2005, i.e., the year of mandatory IFRS adoption. That increase is consistent with lumpy IFRS adoption information being disclosed to the investment community in 2005. Interestingly, the number of revisions decreases to 12.5 for low enforcement and 18 for high enforcement countries in the post-ifrs adoption period. The finding is consistent with more timely IFRS information being disclosed in high enforcement countries, resulting in more analyst revisions in the post-ifrs adoption period. The intra-industry information transfers by financial analysts are therefore more pronounced in high enforcement countries, resulting in a relatively more synchronicity increasing effect.