How do individual investors react to global IFRS adoption? Ulf Brüggemann Humboldt University of Berlin Holger Daske University of Mannheim Carsten Homburg University of Cologne Peter F. Pope Cass Business School August 2012 We appreciate the helpful comments of Kevin Aretz, Hans Christensen, Gary Entwistle, Markus Glaser, JoergMarkus Hitz, Christian Leuz, Steve Lin, Steve Zeff, and seminar participants at HU Berlin, the University of Chicago, Lancaster University, KU Leuven, Erasmus University Rotterdam, WHU Vallendar, the 2009 ASVHB/IAAER meeting in Munich, the 2009 EAA meeting in Tampere, the 2009 AAA meeting in New York, the 2010 FARS Mid-Year meeting in San Diego, the 2010 Accounting Research Conference at Penn State and the INTACCT meetings in Frankfurt, Cyprus, London and Valencia. We are grateful to Markus Glaser for providing us with statistics from his brokerage dataset. Thanks are also due to Paul Rayson for expert advice and Eddie Bell for programming assistance on the collection of Google News archive search results. Ulf Brüggemann, Holger Daske and Peter F. Pope gratefully acknowledge the financial contribution of the European Commission Research Training Network INTACCT (Contract MRTN-CT-2006-035850). Part of this research was carried out while Ulf Brüggemann was visiting the University of Chicago and Holger Daske was visiting London Business School. Correspondence: u.bruggemann@hu-berlin.de. Electronic copy available at: http://ssrn.com/abstract=1458944 How do individual investors react to global IFRS adoption? Abstract We examine the impact of global IFRS adoption on cross-border equity investments by individual investors. Our proxy for cross-border equity investments is trading volume in the Open Market at Frankfurt Stock Exchange. The Open Market is a segment designed for German individual investors to trade a large selection of foreign stocks. Using a sample of 5,637 firms from 31 countries around the world, we find that stocks experience a significant increase in Open Market trading volume following mandatory adoption of IFRS. This effect is more pronounced for attention-grabbing stocks (e.g., stocks experiencing an increase in media coverage following IFRS adoption). Our results suggest that global IFRS adoption enhances cross-border equity investments by individual investors. However, this effect does not materialize equally across stocks due individual investors’ limited attention. JEL classification: Key Words: G14, G38, K22, M41, M48 Individual Investors, International Accounting, IFRS, Open Market, Cross-Border Investments Electronic copy available at: http://ssrn.com/abstract=1458944 1. Introduction Over the last decade, International Financial Reporting Standards (IFRS) have been introduced in over 100 countries around the world (see http://www.iasplus.com/country/ useias.htm). This development is fuelled by the expectation that global adoption of IFRS will, among other benefits, enhance the comparability of financial statements across countries and, thus, reinforce foreign equity investments (e.g., EC Regulation No. 1606/2002).1 The rationale is that global IFRS adoption moves foreign stocks into the choice set of investors by replacing unfamiliar country-specific accounting rules with one single set of standards that investors can familiarize themselves with at lower cost. In this paper, we evaluate this claim by analyzing the impact of global IFRS adoption on cross-border equity investment by individual investors.2 Prior literature on the economic consequences of global IFRS adoption provides evidence that an increase in cross-border equity investments by institutional investors depends on whether compliance with accounting rules is expected to be high and on the degree of change in financial reporting rules as a result of IFRS adoption (e.g., DeFond et al., 2011; Yu, 2010). We hypothesize that individual investors react differently in two ways. First, we expect individual investors to ignore the complex interaction between IFRS adoption and the institutional environment of the stocks, because they tend to be less sophisticated than their institutional peers (e.g., Bhattacharya, 2001; Malmendier and Shantikumar, 2007). Second, we predict that 1 The European Union, Australia, Hong Kong, South Africa and many other countries around the world mandated IFRS or IFRS equivalents for most listed firms from fiscal year 2005 onwards. The treatment group in our empirical analysis is confined to IFRS adopters from these countries (see Table 1, Panel A). Here and in the following, we therefore use the term global IFRS adoption to refer to the mandatory introduction of IFRS in 2005. 2 We use the term individual investors to refer to non-professional investors. Retail investors and private investors are synonymous expressions used in prior studies. It is interesting to note that some of the earliest research examining individual investors’ use and understanding of financial statement information was conducted by Sir David Tweedie, the long-time Chairman of the IASB (e.g., Lee and Tweedie, 1977). 1 Electronic copy available at: http://ssrn.com/abstract=1458944 individual investors focus on those stocks within the IFRS-enhanced choice set that catch their attention, for example, through increased media coverage. This hypothesis is based on the notion that individual investors actively follow a smaller subset of stocks (e.g., Merton, 1987) and that this subset is biased towards attention-grabbing stocks (Barber and Odean, 2008; Engelberg et al., 2012). We test our hypotheses by analyzing trading volume in the Open Market at Frankfurt Stock Exchange (FSE) as a proxy for cross-border equity investments by individual investors. The Open Market is an unofficial trading segment designed for German individual investors to trade foreign (i.e., non-German) stocks that have their main listing at a home market outside Germany. The group of foreign stocks quoted on the Open Market is very large (about one quarter of all firms in the Datastream Universe, see Table 1) and determined by lead brokers who are allowed to include securities in the Open Market at low cost and without the involvement of the issuer. Although lead brokers generally set higher bid-ask spreads than in the home market, the Open Market provides a cost-efficient alternative to the home market when trade sizes are low. German retail banks and brokers pass on high, mostly fixed order fees when local clients choose to trade directly in a stock’s home market, whereas fixed charges for trading at FSE are considerably lower. For small trades executed in the Open Market, lower order processing costs therefore more than compensate for higher bid-ask spreads (see Appendix A for an illustrative example). In short, Open Market lead brokers provide German individual investors with cheaper access to foreign stocks. Open Market trading volume is a valid proxy for increases in cross-border equity investments for two reasons. First, liquidity in the Open Market is low so that lead brokers typically cannot match offsetting orders from individual investors. Instead, lead brokers carry out countertrades in 2 the respective home market when they rebalance inventories. Trading volume in the Open Market therefore largely reflects changes in cross-border equity investments by German individual investors rather than trades between these investors. The second reason is that Open Market investors face no restrictions on the buying side, but they can only sell stock they own and are not able to sell stock short. Consequently, increases in Open Market trading volume are more likely to reflect increases rather than decreases in cross-border equity investments. Appendix B provides statistics from a brokerage dataset supporting this conclusion. Our empirical analyses are based on a proprietary dataset provided by FSE containing stocklevel information on Open Market trading volume for the period January 2002 to June 2008. The Open Market sample comprises 5,637 (43,671) unique firms (semiannual firm-periods) from 31 countries around the world. The mean (median) trade size in the Open Market is about 2,700 (1,700) Euro which corroborates that this segment is used by individual investors. Event study results show that Open Market trading volume increases significantly around annual earnings announcements. This finding confirms insights from official statistics (DAI, 2008; Deutsche Bundesbank, 2005) and recent research (e.g., Bailey et al., 2008; Ernst et al., 2009; Graham et al., 2009) that suggest that individual investors who actively trade foreign stocks are likely to use accounting information for their trading decisions. The empirical analyses proceed in two steps. First, we employ a difference-in-differences design to compare the average impact of global IFRS adoption on Open Market trading volume of IFRS adopters (treatment group) with its impact on firms that apply local GAAP throughout the sample period (control group). The regression results show that Open Market trading volume of mandatory IFRS adopters increases by more than 45% relative to the control group. This effect is statistically significant and robust to the inclusion of variables that control for 3 concurrent changes in market value, changes in home market trading activity and changes in number of German institutional investors, respectively. These control variables are designed to ensure that our results do not reflect already documented effects from global IFRS adoption such as decreases in the cost of equity capital (Li, 2010), liquidity increases in the home markets (Daske et al., 2008) or increases in cross-border equity investments by institutional investors (e.g., DeFond et al., 2011; Yu, 2010). Further tests show that the estimated IFRS effects are not driven by firms that start preparing their financial statements in English or by concurrent reductions in the bid-ask spread difference between the Open Market and the respective home markets. The effect for voluntary IFRS adopters following global IFRS adoption is weaker and not always statistically significant. In additional tests addressing the identification of the IFRS effect, we first repeat the difference-in-differences analysis by artificially changing the starting date of global IFRS adoption. Results confirm that the estimated IFRS effects in the main analysis reflect a structural break in Open Market trading volume rather than the continuation of country-specific time trends. Next, we perform a within-country analysis focusing on the United Kingdom. The results show that mandatory IFRS adopters listed in the Main Market at London Stock Exchange (LSE) experience a significant increase in Open Market trading volume following global IFRS adoption relative to LSE companies that were not required to adopt IFRS before fiscal year 2007. In the second step of the empirical analyses, we examine cross-sectional variation in the estimated IFRS effect. In contrast to studies examining cross-border equity investments by institutional investors (e.g., DeFond et al., 2011; Yu, 2010), we find that the estimated IFRS effect does not vary with institutional variables at the country- or industry-level. While this result is consistent with our hypothesis that individual investors ignore the complex interaction 4 between IFRS adoption and the stocks’ institutional environment, it has to be interpreted with caution due to the specific composition of the Open Market sample. Additional analyses show that the Open Market sample is significantly tilted towards more visible and transparent firms and, thus, may not provide sufficient variation in determinants of institutional quality to enable a powerful test. In contrast, sample selection is less of a concern when testing the effects of limited attention as the Open Market comprises several thousand stocks and, thus, far more than most individual investors have the resources to analyze and retain within their investment choice sets. Our analyses provide strong evidence that the estimated IFRS effect is more pronounced in stocks that catch individual investors’ attention through increased media coverage (measured by the number of articles published on Google News) or positive IFRS effects on net income. This finding supports our hypothesis that individual investors focus on attention-grabbing stocks within the IFRS-enhanced choice set. Taken together, the results of our empirical analyses suggest that global IFRS adoption enhances cross-border equity investments by individual investors. However, this effect does not materialize equally across stocks due individual investors’ limited attention. Our paper contributes to the emerging research on the economic consequences of IFRS. To our knowledge, we provide the first analysis on how individual investors react to global IFRS adoption. Given the important role that individual investors play in financial markets,3 the lack of prior evidence may seem surprising at first. However, a likely explanation is that data on individual investors’ behavior is not publicly available. Prior studies on individual investors have 3 At the end of 2007, domestic individuals owned 14% of the market value of listed stocks in Europe (FESE, 2008). In the United States (U.S.), more than 20% is held directly by individual investors (French, 2008). Anecdotal evidence suggests that individual investors are more likely to pursue long-term objectives than their institutional counterparts. Companies therefore make great efforts to attract individual investors, e.g. via corporate websites and investor relations departments (Vogelheim et al., 2001). The relevance of individual investors is also recognized by the International Accounting Standards Board (IASB) as reflected in its recent efforts to include them in the development of IFRS (IASB, 2010). 5 adopted several strategies to address this challenge, including analysis of small trades using intra-day transactions data (e.g., Lee, 1992), surveys of individual investor opinion (e.g., Elliott et al., 2008) and examination of a proprietary datasets from online brokers (e.g., Odean, 1998; Glaser and Weber, 2009). In contrast to these approaches, the Open Market setting is capable of providing large-sample evidence on the impact of global IFRS adoption on individual investor decisions. The Open Market may therefore also prove useful for other research, because it allows direct observation of aggregate trading activities of a large group of individual investors trading in foreign stocks. 2. Hypotheses Development Regulators expect that global IFRS adoption will, among other benefits, enhance the comparability of financial statements across countries and, thus, reinforce foreign equity investments (e.g., EC Regulation No. 1606/2002). The underlying argument is that global IFRS adoption moves foreign stocks into the choice set of investors by replacing unfamiliar countryspecific accounting rules with one single set of standards that investors can familiarize themselves with at lower cost. Extant evidence on reactions by institutional investors suggests that the regulators’ expectation has indeed been met provided that the stocks’ institutional environment ensures high compliance with accounting rules and allows for a substantial change in financial statements through IFRS adoption. DeFond et al. (2011) find that mandatory IFRS adoption in the EU enhances cross-border equity investments by international mutual funds if the increase in the number of industry peers is sufficiently large and accounting standards are credibly implemented. Yu (2010) provides similar evidence and identifies accounting distance (i.e., the number of differences between IFRS and the domestic accounting standards they replace) as an 6 additional determinant for the increase in cross-border equity investments.4 These findings are consistent with the majority of related literature that shows that capital-market benefits of mandatory IFRS adoption such as higher market liquidity (Daske et al., 2008), lower cost of equity capital (Li, 2010), increased institutional holdings (Florou and Pope, 2012), higher information content of earnings announcements (Landsman et al., 2012) or improved analyst forecast properties (e.g., Byard et al., 2011; Tan et al., 2011) are contingent on the quality of the institutional environment. However, it is not clear whether individual investors react to global IFRS adoption in the same manner. Prior literature suggests that individual investors systematically differ from their institutional counterparts in at least two ways. First, they are more naive about incentives and more likely to use simple heuristics in decision making (see De Bondt, 1998, for a survey of the earlier literature). For example, Bhattacharya (2001) examines abnormal trading reactions to earnings announcements and shows that small traders tend to rely on a seasonal random-walk model when forming earnings expectations. In contrast, large traders incorporate more current information than the previous year’s earnings into their expectations. In a similar vein, Malmendier and Shantikumar (2007) provide evidence that small traders follow analysts’ stock recommendations literally, while large traders display more sophisticated reactions by taking into account that analysts tend to bias these recommendations upward. Second, individual investors have fewer resources and therefore tend to actively follow a smaller subset of stocks (e.g., Merton, 1987). Barber and Odean (2008) find that this subset is biased towards stocks that grab the attention of individual investors (e.g., stocks in the news or stocks experiencing high 4 Khurana and Michas (2011) and Shima and Gordon (2011) focus on institutional investors from the U.S. and also find an increase in cross-border equity investments following mandatory IFRS adoption. The size of this effect is positively associated with country-level factors such as the strength of the enforcement regime and the accounting distance. 7 abnormal trading volume). Engelberg et al. (2012) provide supporting evidence for this attention effect by showing large overnight returns to stock recommendations on the popular television show “Mad Money”. These returns subsequently reverse in the long run. Based on these insights from prior literature, we predict that while global IFRS adoption has the potential to enhance cross-border equity investments by individual investors, this reaction is likely to differ from that of institutional investors in two aspects. First, we expect individual investors to ignore the complex interaction between IFRS adoption and the institutional environment of the stocks, because they tend to be less sophisticated than their institutional peers. Second, we predict that the lack of resources leads individual investors to focus on those stocks in the IFRS-enhanced choice set that grab their attention. Our hypotheses can be summarized as follows: H1: Global IFRS adoption enhances cross-border equity investments by individual investors. H2a: The effect of global IFRS adoption on cross-border equity investments by individual investors is not related to the institutional environment of stocks. H2b: The effect of global IFRS adoption on cross-border equity investments by individual investors is more pronounced for attention-grabbing stocks. In our empirical analyses, we use trading volume in the Open Market at FSE as a proxy for cross-border equity investments by individual investors. The next section provides institutional details on the Open Market. 8 3. The Open Market 3.1 INSTITUTIONAL BACKGROUND The Open Market (“Freiverkehr” in German) is an unofficial trading segment at FSE. In contrast to official stock market segments in Europe (e.g., Prime and General Standard at FSE, Main Market at London Stock Exchange), the Open Market is not subject to regulations and directives of the European Union (EU), but is exclusively governed by stock exchange rules. It covers a variety of financial instruments such as stocks (both from Germany and abroad), bonds, certificates and warrants. The stocks segment is structured into the First Quotation Board and the Second Quotation Board. The First Quotation Board contains companies with a primary listing in the Open Market.5 Companies whose stocks are already listed at another domestic or foreign trading venue (home market) are included in the Second Quotation Board. Since the Open Market is an unofficial trading segment, the EU regulation mandating IFRS is not applicable to companies in the First Quotation Board. In contrast, many companies in the Second Quotation Board are obliged to prepare their financial statements in accordance with IFRS due to regulation in the respective home markets. In this study, we focus on the foreign (i.e., non-German) stocks in the Second Quotation Board. For simplicity, we refer to this sub-segment as the Open Market.6 5 In 2005, FSE introduced the Entry Standard as a sub-segment of the First Quotation Board. Transparency requirements in the Entry Standard are higher than in the rest of the Open Market, but considerably lower than in official FSE stock market segments. While the Entry Standard is open to all companies, it is specifically targeted at small- and mid-caps that seek low cost access to the capital market. The Entry Standard is marketed as an alternative to the Alternative Investment Market at London Stock Exchange (e.g. Sudmeyer et al., 2005; Schlitt and Schäfer, 2006). At the end of 2010, 119 (13) German (foreign) companies were listed in the Entry Standard (FSE, 2009). 6 Official resources and the academic literature provide only little information on the Second Quotation Board of the Open Market. Much of the following description is based on the insights we gained from interviews with FSE staff and brokers. For more general information on the Open Market see e.g. Müller-Michaels and Wecker (2005), Harrer and Müller (2006) or the website of FSE: www.deutsche-boerse.com. 9 Established in 1987, the Open Market has become increasingly popular with German investors in recent years. At the end of 2000, a total of 4,471 foreign stocks were traded in the Open Market. This number more than doubled to 10,095 by the end of 2010. For comparison, the number of domestic stocks traded at FSE increased by merely 17% from 903 to 1,058 during the same period (FSE, 2010). The remarkably high number of foreign stocks available for trading in the Open Market is a consequence of its unique set of rules.7 These rules permit eligible brokerage houses accredited for trading at FSE to include securities in the Open Market on their own initiative (AGB §2.3). The stock issuing company need not be informed, nor need it approve inclusion of its securities in the Open Market.8 For the brokerage house, the inclusion process involves two basic requirements. First, it has to guarantee orderly fulfillment of transactions by acting as a lead broker (AGB §12.1). Second, it has to pay a non-recurring fee of 750 Euro (AGB §25). Follow-up obligations of the lead broker are confined to informing the FSE about essential company news concerning the issuer that can be acquired “by generally accessible information sources in a reasonable way” (AGB §14.2). Lead brokers are authorized to exclude securities from the Open Market at any time “subject to an adequate term” (§ 15.1 AGB).9 In summary, brokerage houses face very few constraints or institutional barriers relating to inclusion or exclusion of securities in the Open Market. 7 We refer to this set of rules as AGB in the following. AGB stands for “Allgemeine Geschäftsbedingungen für den Freiverkehr an der Frankfurter Wertpapierbörse” (General Terms and Conditions for the Regulated Unofficial Market). The specific rules we cite are from the most recent AGB version as of May 2011. The content of these rules is very similar to that of an earlier AGB version as of September 2002. 8 Such involuntary cross-listings are also observed at other stock exchanges. For example, the unsponsored OTC ADR program in the U.S. allows for the inclusion of foreign stocks without legal obligation to notify the firms or obtain their consent (Iliev et al., 2011). In contrast to the Open Market, this program is not specifically designed for individual investors. 9 Order books of Open Market securities can also be terminated by FSE (AGB §15.2). For example, in December 2005, FSE suspended trading in Turkish stocks until further notice because of unanswered questions about a planned tax on Turkish equities (Greil, 2005). 10 Once a security has been included, the lead broker holds the exclusive right to set bid and ask quotes.10 Although officially non-binding, these quotes are de-facto tradable up to a size the lead broker specifies (Freihube et al., 1999). When an investor places an order to trade on the bid (ask) quote, the lead broker buys (delivers) the agreed number of stocks. To entirely eliminate inventory risk the resulting position would then ideally be closed immediately with an offsetting order. However, due to market illiquidity (see below for supporting evidence), perfect offsetting may not be possible for Open Market stocks. The lead broker is then forced to execute inventory rebalancing countertrades in another market, typically the home market where liquidity is usually much higher. Hence, in setting bid and ask quotes the lead broker faces a trade-off. On one hand, she has an incentive to offer low bid-ask spreads to generate trades and earn brokerage fees. On the other hand, there is the risk of losing out on trades if countertrades in home markets are carried out at unfavorable prices. The resulting Open Market quotes are therefore likely to be determined by the home market bid-ask spread as a lower bound plus a premium that reflects the price risks faced by lead brokers in executing inventory rebalancing and due to currency risk exposure during trade execution. Other factors with a potential impact on the bid-ask spread premium in the Open Market include trading volume (i.e., the likelihood that the lead broker is able to match offsetting orders) and competition to the lead broker’s services. Competition may arise from other German exchanges, e.g. in Berlin, Stuttgart or Munich, where similar but much smaller trading segments exist. Within the FSE, trading volume can shift from floor trading where the lead broker operates to the fully electronic platform XETRA where quotes are 10 In case more than one party applies to be the lead broker for a particular stock, the allocation of the order book is decided by lot. Baader Bank AG, mwb fairtrade Wertpapierhandelsbank AG and Wolfgang Steubing AG Wertpapierdienstleister are the leading brokerage houses in the Open Market (Hiller von Gaertringen, 2006), but there are a number of other competitors. Detailed information on the allocation of Open Market order books is not publicly available. 11 automatically determined by an open limit order book. We provide more details on these alternative trading channels in Appendix A. Despite high bid-ask spreads, the Open Market provides a cost-efficient trading venue alternative to home markets under certain circumstances. German retail banks and brokers pass on high, mostly fixed order fees when local clients choose to trade directly abroad, whereas fixed charges for trading at FSE are considerably lower. Hence, for small trade sizes the higher bid-ask spreads (i.e., variable transaction costs that increase with trade size) are outweighed by lower order processing costs (i.e., mostly fixed transaction costs that are independent of trade size) in the Open Market. The combination of low fixed and high variable fees in the Open Market is likely to be particularly attractive for individual investors who trade small sizes of foreign stocks.11 Put differently, the Open Market lead broker provides German individual investors with cheaper access to foreign stocks.12 Appendix B provides an illustrative example on these links. The first part of Appendix C confirms empirically that individual investors of a German online broker trade foreign stocks primarily through the Open Market, particularly when trade sizes are low. 3.2 DATA AND DESCRIPTIVE STATISTICS In this section, we describe the quantitative features of the Open Market. The analysis is based on two main samples of equities: (1) the Datastream (DS) Universe, and (2) the Open Market 11 We refer to Open Market investors as German individual investors for two reasons. First, the economics of the Open Market show that Open Market investors trade through German retail banks and brokers (see Appendix B). Due to institutional barriers most German retail banks and brokers require their clients to be domiciled in Germany. Second, German retail banks and brokers typically target their services at investors who speak German. For example, the web-portal of comdirect bank (www.comdirect.de), the leading online broker in Germany, provides information in German language only. 12 Note that institutional investors typically (1) have preferred and cheaper access to home markets through their lead brokerage houses and (2) trade in volumes well above the break-even where the home market turns into the more cost-efficient trading alternative. 12 sample - a subset of the DS Universe. At the country-level, the DS Universe includes all firms covered by Datastream domiciled in countries other than Germany that either introduced IFRS in 2005 (the treatment group) or mandated domestic accounting standards throughout the sample period (the control group).13 At the firm-level, we restrict the DS Universe to companies that have their primary listing on the main exchange of their country of domicile (home markets)14 and for which sufficient data on trading volume, stock returns and accounting data are available. We only include companies in the treatment group that switched from local GAAP to IFRS in 2005 (mandatory IFRS adopters) or before 2005 (voluntary IFRS adopters).15 The control group consists of companies that used domestic accounting standards throughout the sample period. We retrieve information on accounting standards followed from Worldscope. The Open Market sample covers all firm-years within the DS Universe during which the respective stock could be traded on the FSE. We identify this sample using a proprietary dataset from FSE containing daily trading volume data (both in Euros and in number of shares traded) as well as the number of ticks for every stock traded in the Open Market during the sample period. The FSE trading volume dataset available to us spans the period January 2002 to June 2008. For consistency, we confine capital market data from Datastream (e.g. trading volume for the home markets) to the same period. We partition the dataset into a maximum of 13 semiannual periods 13 We deliberately exclude Germany (our focus is on foreign, that is, non-German, stocks), New Zealand (IFRS introduction in 2007), Singapore (IFRS introduction in 2003) and Switzerland (no mandatory IFRS introduction, many listed firms use IFRS by choice) from the DS Universe. 14 The main exchange is defined as the trading venue with the largest number of companies listed. We consider only one exchange per country except for the United States where firms from both New York Stock Exchange (NYSE) and NASDAQ are included. By focusing on the main exchange(s) in each country, we exclude companies listed at less regulated trading venues (such as the OTC Bulletin Board in the U.S.) and thus ensure a minimum level of transparency among sample firms. 15 Thus, in order to obtain a clean sample, companies from the treatment group countries that did not switch to IFRS during the sample period (e.g., firms that need not prepare consolidated financial statements), did so after 2005 (e.g., firms listed on the Alternative Investment Market at London Stock Exchange) or applied U.S.-GAAP (e.g., due to a cross-listing in the U.S.) are not considered. We include firms from the Alternative Investment Market in an additional analysis in section 4.3.2. 13 per firm covering 2002H1 to 2008H1 with the suffix H1 (H2) indicating the first (second) half of the respective calendar year. Table 1 presents details on the composition of the DS Universe and the Open Market sample. Panel A focuses on the treatment group, that is, countries that introduced IFRS in 2005. The DS Universe includes 266 (6,192) voluntary (mandatory) IFRS adopters from 22 countries. The Open Market sample covers 50% (27%) of all voluntary (mandatory) IFRS adopters in the DS Universe. Panel B shows that the DS Universe includes 15,822 firms from 19 countries within the control group. 24% of these companies are part of the Open Market sample. Open Market coverage differs substantially across countries and firms. For example, while the majority of Austrian and U.S. stocks are tradable in the Open Market, some countries (e.g., Poland, Morocco or South Korea) are not represented at all. At the firm-level, the considerable difference in coverage rates across accounting standards and IFRS adopter types gives a first indication that the lead brokers do not randomly choose the securities they offer in the Open Market. From Panel C, we learn that the number of Open Market firms covered in the DS Universe increases over time, both in absolute as well as in relative terms. The total number of unique Open Market firms (Open Market share) climbs from 2,188 (16%) in 2002H1 to 4,936 (25%) in 2008H1. In total, the Open Market sample (DS Universe) comprises 5,637 (22,280) unique firms and 43,671 (217,772) firm-periods.16 Table 2 shows descriptive statistics on various firm characteristics for the Open Market sample (Panel A) as well as for the rest of the DS Universe (Panel B). Panel A presents trading volume, number of trades, trade size and bid-ask spread statistics from FSE and the respective home markets for the same set of firm-years. Panel B is naturally confined to data from the home 16 Differences in the size of the Open Market sample and the official numbers from the FSE Factbooks stem from the data requirements we impose on the DS Universe. 14 markets, because the covered sample (DS Universe excluding the Open Market sample) is not traded at FSE. In addition to liquidity measures, both panels show statistics on other variables that are independent of the trading venue. Liquidity in stocks from the Open Market sample is low at FSE. During the average firm period, trading occurs on slightly less than 25% of all trading days. Average stock-level daily trading volume is about 18,000 Euros, although the distribution of trading volume is highly positively skewed. In contrast, daily trading volume in home markets averages nearly 30 million Euros. These massive liquidity differences across exchanges are hardly surprising. While trading at FSE is confined to a small subset of individual investors, institutional investors and non-German individual investors will prefer to trade in the respective home markets.17 Average trade size at FSE is about 2,700 Euro which is well below the threshold of 10,000 USD typically used in prior literature to distinguish between trades of individual and institutional investors (e.g., Lee, 1992). Comparison of bid-ask spreads across exchanges indicates that the variable fee the lead broker charges on Open Market transactions is in fact substantial: the median bid-ask spread is 3.13% at FSE compared to only 0.96% in the respective home markets. Taken together, these descriptive statistics confirm that the Open Market at FSE is a trading segment that is specifically designed for individual investors. 3.3 CHARACTERISTICS OF THE OPEN MARKET In this section, we complement the descriptive analysis of the Open Market by examining two important features of the Open Market. First, we provide evidence on the characteristics of firms 17 Untabulated statistics show that trading volume at FSE aggregated over the whole Open Market sample varies between 10 and 20 billion Euros per year. Hence, despite its relative lack of liquidity the Open Market offers substantial income opportunities for its participants. For example, with an average brokerage fee of 0.08% of the order volume (see Appendix A) Open Market lead brokers earn a total of 8 to 16 million Euros per year for their services. 15 included in the Open Market. Second, we discuss how accounting information influences Open Market trading volume. 3.3.1 Determinants of Open Market Inclusion Stocks are tradable in the Open Market if they have been included by the lead broker. This decision depends on a stock’s potential to generate sufficient Open Market trading volume and, thus, brokerage fees. Potential trading volume in the Open Market is ultimately determined by individual investors’ demand for a particular stock. Table 3, Panel A, presents results from probit regressions relating the likelihood of inclusion of a stock in the Open Market to various firm- and country-specific variables. The analysis yields two key findings. First, firms in the Open Market feature higher market values (Market Value), higher trading volume in the home market (Home Trading Volume), are more likely to prepare their financial statements in German or English (German/English Reporting) and attract more German institutional investors (No. of German Inst. Investors) than firms whose stocks are not tradable at FSE.18 These results suggest that Open Market investors prefer stocks of more visible companies that prepare financial statements in a familiar language. The second key finding is that transparent reporting practices are significant determinants of Open Market inclusion, both at the country and at the firm level (EM Measure). The only other significant determinant at the country level is a dummy variable that indicates countries from the eurozone (Euro). In contrast, variables for capital market development (MCAP/GDP) and geographic proximity (Distance Berlin - Capital) do not load significantly.19 18 We transform highly skewed variables using natural logarithms to mitigate the influence of outliers. All variables are described in more detail in Table 2 and 3, respectively. 19 Note that some coefficients lose their statistical significance in multivariate regressions due to high correlations among the determinants. For example, the Pearson correlation coefficient between Log(Market Value) and Log(Home Trading Volume) over the DS Universe is 0.85. 16 Taken together, these results provide strong evidence that the Open Market sample is a nonrandom subset of the DS Universe. Specifically, the Open Market sample is significantly tilted towards more visible and transparent companies. Thus, it seems that lead brokers either act as gatekeepers to the Open Market by proactively screening stocks for inclusion, or they respond to demand from individual investors who prefer more visible and transparent stocks. 3.3.2 Accounting Information and Open Market Trading Volume In this subsection, we present official statistics and event study results to provide insights into how accounting information influences the trading behavior of Open Market investors. Official statistics show that – similar to evidence from other countries – the majority of German individuals do not actively invest in stocks and that those who do exhibit a considerable degree of home bias when selecting individual stocks. For example, in 2004 the number of individual shareholders over 14 years in Germany amounted to 4.6 million or 6.5% of the entire population (DAI, 2008). Despite the well known benefits of international diversification, German individual shareholders invested a total of 145,495 million Euro in domestic, but only 36,873 million Euro in foreign stocks (i.e., 20.2% of all investments in stocks; Deutsche Bundesbank, 2005). These macro-level statistics suggest that the small subset of German individuals that actively trades individual foreign stocks possesses more financial literacy, on average, than those who focus on domestic equity or entirely refrain from stock picking. Consistent with this observation, recent research shows that individual investors are more likely to have internationally diversified portfolios if they are highly educated, wealthy and/or have more trading experience (e.g., Bailey et al., 2008; Graham et al., 2009). Survey evidence by Ernst et al. (2009) reveals that German individual investors use business media and financial statements as central information sources. Ernst et al. (2009) also find that the usage of accounting information 17 increases with the trading experience of individual investors. Taken together, these findings suggest that accounting information affects Open Market trading volume through two nonmutually exclusive channels: either Open Market investors utilize financial statements by themselves and/or they consult other information sources such as the business media and brokers’ analyst reports that in turn reflect information disclosed in financial statements. We complement the insights from official statistics and prior literature with direct evidence on the link between accounting information and Open Market trading by analyzing abnormal trading volume around annual earnings announcements. Table 3, Panel B, compares reactions at FSE with those in the respective home markets for the same set of earnings announcements. The analysis is based on a sample of 18,362 earnings announcement dates from IBES. Abnormal trading volume is the difference between trading volume on the event day and the mean daily volume for that stock over the pre-announcement window (-120, -21), scaled by the mean daily volume (e.g., Bamber et al., 2011). To mitigate the influence of outliers, abnormal trading volume is winsorized at the 99% level by event day. The results show that abnormal trading volume at FSE increases significantly around earnings announcements. While the effect on mean abnormal trading volume is similar to that in the home markets, median abnormal trading volume remains unchanged at -1.0000 throughout the event window because fewer than 50% of Open Market stocks trade each day (see, e.g., the statistics on Trading Days (%) in Table 2, Panel A). These results illustrate that the release of earnings information triggers substantial trading volume reactions in the Open Market if general liquidity is sufficiently high. Hence, the event study provides direct evidence that accounting information influences trading behavior of Open Market investors. 18 4. Global IFRS Adoption and Open Market Trading Volume 4.1 EMPIRICAL STRATEGY We test the hypotheses developed in section 2 by using increases in Open Market trading volume as a proxy for increases in cross-border equity investments by individual investors. This proxy is valid for two reasons. First, liquidity in the Open Market is low so that the lead broker typically cannot match offsetting orders by the individual investors (see section 3.2). Instead, the lead broker has to carry out a countertrade in the respective home market to rebalance her inventory. Trading volume in the Open Market therefore largely reflects changes in cross-border equity investments by German individual investors rather than trades between these investors. The second reason trading volume proxies for cross-border equity investment is that Open Market investors have no restrictions on the buying side, while they can only sell stocks they own. Thus, increases in Open Market trading volume are more likely to reflect increases rather than decreases in cross-border equity investments. Appendix B provides statistics from a proprietary brokerage dataset that support this conclusion. The empirical analyses proceed in two steps. First, we examine the first hypothesis (H1) that global IFRS adoption enhances cross-border investments by individual investors. Section 4.2 comprises the baseline analyses. Since the subsequent tests on cross-sectional variation depend on the IFRS effect being estimated precisely, we devote a separate section 4.3 to alternative identification strategies. Section 4.4 comprises the second step of our analyses where we test the hypotheses on how the IFRS effect relates to the institutional environment of stocks (H2a) and to proxies for individual investors’ attention (H2b). 19 4.2 BASELINE ANALYSES 4.2.1 Research Design In this section, we test the first hypothesis (H1) that global IFRS adoption enhances crossborder equity investments by individual investors. Our proxy for cross-border equity investments is trading volume in the Open Market denoted as FSE Trading Volume.20 The key independent variable is Post-FY2005, a dummy variable that equals one (zero) in periods after (before and including) the fiscal year-end of first-time mandatory adoption. For example, if the fiscal yearend of first-time mandatory adoption is in December 2005, Post-FY2005 has a value of one from 2006H1 onwards.21 To test the impact of global IFRS adoption, we interact Post-FY2005 with binary variables that indicate mandatory IFRS adopters (Mandatory) and voluntary IFRS adopters (Voluntary), respectively. These interaction terms capture the average effect of global IFRS adoption on Open Market trading activity for the respective group of companies relative to the control group of non-IFRS adopters. Combining these variables results in the following basic regression specification: FSE Trading Volume = β0 + β1 Post-FY2005 + β2 Post-FY2005*Mandatory + β3 Post-FY2005*Voluntary + β4 Voluntary*IFRS + Σ βj Controlsj + ε [1] 20 We use trading volume from both the floor and XETRA to calculate FSE Trading Volume. FSE Trading Volume is identical to Trading Volume (Euro) for Frankfurt Stock Exchange in Table 2, Panel A. The results are similar when we use trading volume from the floor only (see Appendix A) or percentage trading volume, that is, the ratio of shares traded and the number of shares outstanding. Since FSE Trading Volume is highly skewed, we use the natural logarithm to mitigate the influence of outliers. To ensure computing the natural logarithm for all firm periods, we replace raw values of zero by a small firm-specific constant. This constant is defined as the average market value divided by the average number of shares outstanding, divided by the number of exchange trading days during the firm period (i.e., we assume that exactly one stock was traded during the firm period). 21 Many listed companies in Germany adopted IFRS well before global IFRS adoption took off, either voluntarily (e.g., Leuz and Verrecchia, 2000; Daske, 2006) or due to exchange regulation of the former New Market (e.g., Leuz, 2003). These early adoptions enabled German investors to familiarize themselves at an early stage with IFRS through their investments in domestic stocks. We therefore assume that Open Market investors react promptly to global IFRS adoption. We test the sensitivity of our results to this assumption in section 4.3.1. 20 where Voluntary*IFRS is an interaction term that equals one (zero) after (before) voluntary IFRS adoption and Controlsj denotes the set of control variables. Consistent with our main hypothesis, we expect the coefficient estimate on Post-FY2005*Mandatory to be positive, that is, β2 > 0. To the extent that the comparability benefits of global IFRS adoption spill over to voluntary adopters, we also expect a positive coefficient estimate on Post-FY2005*Voluntary (β3). Following related IFRS literature (e.g., Daske et al., 2008), we estimate regression specification [1] after including firm fixed-effects. The goal of this difference-in-differences approach is to identify the relationship between a treatment (IFRS introduction) and an endogenous variable (Open Market trading activity) by comparing the treatment’s impact on affected firms (treatment group) with its impact on unaffected firms (control group). To ensure that OLS estimation produces consistent standard errors, we use standard errors clustered by country. Note that in our model the estimated effect of global IFRS adoption is exclusively determined by firms that are part of the Open Market sample both before and after IFRS introduction. Hence, differences in the composition of the Open Market sample pre- versus postIFRS do not directly influence the regression outcomes. Since we estimate a firm fixed effects model, our controls are confined to variables that capture firm-specific changes over time. We include Market Value, Home Trading Volume and No. of German Inst. Investors to control for changes in market values, changes in trading activity in the respective home markets and changes in the number of German institutional investors, respectively. These control variables are designed to ensure that our results do not reflect already documented effects from global IFRS adoption such as decreases in the cost of equity capital (Li, 2010), liquidity increases in the home markets (Daske et al., 2008) or increases in cross-border equity investments by institutional investors (DeFond et al., 2011; Yu, 2010). German/English 21 Reporting is included to ensure that the results are not attributable to firms that start preparing their financial statements in German or English. Finally, Spread Difference controls for changes in trading costs measured by the difference between the bid-ask spread at FSE and the respective home market. We expect the coefficient estimate(s) on Spread Difference (all other control variables) to be negative (positive).22 4.2.2 Empirical Findings Table 4 presents regression results of the trading volume analysis. Regression models 1 (without controls), 2 and 3 (with controls) are based on the full sample and show that both mandatory and voluntary IFRS adopters experience a strong and statistically significant increase in Open Market trading volume following global IFRS adoption. For example, the coefficient estimate on Post-FY2005*Mandatory (Post-FY2005*Voluntary) in model 2 is 0.679 (0.459) which corresponds to an increase in percentage trading volume of 97% (58%) relative to the control group. Untabulated analyses reveal that this coefficient estimate drops to 0.444 (0.200) if U.S. firms, which constitute over 50% of the full sample, are excluded. This result demonstrates that U.S. firms substantially affect the main results. To avoid the possibility that our results are driven by sample firms from a few large countries such as the U.K. and the U.S., we define a subsample that allows a maximum of 100 firms per country. We create this subsample by (1) focusing on firms that are part of the Open Market sample both before and after IFRS introduction,23 (2) sorting these firms within each country by their average market value over the sample period and (3) selecting every N/100th firm if the number of firms per country N is greater than 100. Regression models 4 (without controls), 5 and 22 Again, we use the natural logarithm for those variables that have highly skewed raw values. For details on all control variables, see Table 2, Panel A. 23 Note that firms that do not fulfill this requirement have no direct impact on the estimated IFRS effect due to the inclusion of firm fixed effects (see section 4.2.1). 22 6 (with controls) confirm that the strong IFRS effects for mandatory adopters hold when the subsample is used. Specifically, mandatory IFRS adopters experience an increase in Open Market trading volume of at least 45% (model 6) relative to the control group. This effect corresponds to 8,250 (140) Euro in Open Market trading volume per firm period relative to the full sample’s mean (median) of 18,340 (305) Euro (see Table 2, Panel A). The IFRS effect for voluntary adopters is weaker in the subsample regressions with t-statistics below 2. The coefficient estimates on Voluntary*IFRS are statistically significant throughout all regression specifications and suggest that following voluntary IFRS adoption Open Market trading volume more than doubles relative to the rest of the respective sample. Note, however, that these estimates are determined by a small group of only 27 (21) voluntary adopters in the full sample (subsample) that switched to IFRS in fiscal years 2002, 2003 or 2004, and were tradable in the Open Market both before and after adoption. For the remaining 105 (73) voluntary adopters in the full sample (subsample), the firm fixed effects capture potential effects of voluntary IFRS adoption on Open Market trading volume. In all regressions, the coefficient estimates on the control variables have the expected sign and are statistically significant, except for German/English Reporting24 and No. of German Inst. Investors. Taken together, the regression results provide strong evidence consistent with hypothesis H1 that global IFRS adoption enhances cross-border equity investments by individual investors. 24 There are no switches to German reporting in our dataset. Since our dataset does not comprise financial reports before fiscal year 2003, we only capture switches to English reporting between fiscal year 2004 and 2007. We document 47 (269) switches in fiscal year 2005 (in total). At the country level, most switches are carried out by firms from Hong Kong (71), France (43), Greece (27), Spain (25) and Sweden (23). Where comparable, our statistics are similar to those presented by Jeanjean et al. (2011). 23 4.3 ALTERNATIVE IDENTIFICATION OF THE IFRS EFFECT 4.3.1 Shifting the Switch of the Post-IFRS Dummy The regression results presented in the previous section are based on the assumption that Open Market trading volume reacts to financial statement information immediately after the respective fiscal year-end. We therefore defined that the key independent variable Post-FY2005 switches in the first firm period following the end of fiscal year 2005. In this section, we gauge the impact of this research design choice by defining Post-FY2005 to switch up to two periods earlier or later (see Christensen et al., 2011, for a similar strategy). Table 5, Panel A, reports the results which are based on re-estimations of model 2 and 5 in Table 4, respectively. While the coefficient estimate on Post-FY2005*Mandatory remains largely unchanged for the regressions based on the full sample, the subsample results show that both the coefficient estimate and the t-statistic peak when Post-FY2005 is defined as in the previous section. The untabulated coefficient estimates on Post-FY2005*Voluntary are similar and (not) statistically significant for the full sample (subsample) throughout all regressions. The subsample results suggest that there is a structural break in Open Market trading volume around global IFRS adoption. More importantly, this structural break is confined to stocks of mandatory IFRS adopters. To the extent that the subsample is more powerful than the full sample in disentangling a potential IFRS effect, this result lends support to our first hypothesis H1 of a causal relation between global IFRS adoption and cross-border equity investments by individual investors. 4.3.2 Within-Country Analysis In this section, we exploit institutional peculiarities in the United Kingdom to gain further insights on the existence of an IFRS effect on Open Market trading volume. To this end, we use 24 mandatory IFRS adopters listed in the Main Market at London Stock Exchange (LSE) as the treatment group and companies listed in the Alternative Investment Market (AIM) at LSE as the control group. AIM companies were not required to adopt IFRS before fiscal year 2007 and therefore are excluded from the Open Market sample presented in Table 1. We delete firmperiods of AIM companies after they adopted IFRS. This procedure yields a sample of 2,104 (437) firm periods from 302 (93) mandatory IFRS adopters from the Main Market (AIM companies). Based on hypothesis H1, we expect that following global IFRS adoption firms from the Main Market experience an increase in Open Market trading volume relative to AIM companies. Table 5, Panel B, provides evidence that is consistent with this expectation. For example, the coefficient estimate on Post-FY2005*Mandatory in the regression model with control variables is 0.990 (t-statistic 2.34) which corresponds to a relative increase in percentage trading volume of more than 100%. 4.4 CROSS-SECTIONAL VARIATION IN THE IFRS EFFECT 4.4.1 Research Design In this section, we examine cross-sectional variation in the estimated IFRS effect on Open Market trading volume to test our second set of hypotheses (H2a) and (H2b). Similar to prior literature (e.g., Daske et al., 2008), we interact the key independent variables in regression specification [1] with a binary variable Conditional that partitions the treatment group. This approach translates into the following basic regression specification: FSE Trading Volume = β0 + β1 Post-FY2005 + β2 Post-FY2005*Mandatory + β3 Post-FY2005*Mandatory*Conditional + β4 Post-FY2005*Voluntary + β5 Post-FY2005*Voluntary*Conditional + β6 Voluntary*IFRS + Σ βj Controlsj + ε [2] 25 where all variables are defined as in regression specification [1]. The partitioning variable Conditional explains systematic cross-sectional variation in the estimated IFRS effect for mandatory (voluntary) adopters if the interaction term Post-IFRS*Mandatory*Conditional (PostIFRS*Voluntary*Conditional) is statistically significant. To test hypothesis (H2a) on the relation between the IFRS effect and the institutional environment of the stocks, we use the following country- and industry-level variables to partition the treatment sample (DeFond et al., 2011; Yu, 2010): (1) Long Accounting Distance is based on the Bae et al. (2008) summary score of how a country’s local GAAP differs from IFRS on 21 key accounting dimensions and equals one (zero) for countries with a score greater (equal to or less) than the sample median of 9. (2) Strong Credibility is based on the earnings management score from Leuz et al. (2003) and takes a value of one (zero) for countries with a score of less (equal to or greater) than the sample median of 18.3. (3) Large Δ Uniformity is based on the changes in uniformity measure from DeFond et al. (2011) and equals one (zero) for industrycountry clusters with changes in uniformity greater (equal to or less) than the sample median of 39.67. (4) Strong Credibility * Large Δ Uniformity is the interaction of the second and third variable. It takes a value of one if both variables equal one, and zero otherwise. To test hypothesis (H2b) whether the IFRS effect is more pronounced for attention-grabbing stocks, we use the following firm-level variables to proxy for individual investors’ attention and partition the treatment sample: (1) Increase in Total Media Coverage equals one (zero) if worldwide media coverage is, on average, higher (lower or the same) after fiscal year 2005 than before. Total media coverage is measured as the number of search results in the Google News archive (http://news.google.com/archivesearch) in any language. This number reflects the number of articles that were published on Google News during the relevant period and that 26 contain either the company name (as provided by Worldscope item WC06001) or the firmspecific ISIN code. (2) Increase in German Media Coverage is similar to the first variable except that it only counts Google News archive search results in German language. 25 The media coverage variables are based on the notion that stocks in the news are more likely to draw individual investors’ attention (Barber and Odean, 2008). (3) Large IFRS Restatements equals one (zero) if the percentage difference between the restated net income under IFRS and the originally reported net income under local GAAP for fiscal year 2004 is above (below or equal to) the median percentage difference of 2.17%. Restatement information is from Worldscope (item WC01551R) and only available for a subset of mandatory IFRS adopters. The underlying assumption of this variable is that mandatory IFRS adopters are more likely to trigger individual investors’ attention if the transition from local GAAP to IFRS suggests a substantially positive impact on net income. (4) Strong EA Reactions equals one if the average Open Market trading volume during the three-day window around the earnings announcement is higher than the average Open Market trading volume over the relevant firm period, and zero otherwise. To the extent that earnings announcements are events that trigger attention-driven Open Market trading volume (see section 3.3.2), firm periods where trading volume is clustered around these events are likely to reflect periods of relatively high attention by individual investors. 4.4.2 Empirical Findings Table 6 presents results of the analyses on cross-sectional variation in the estimated IFRS effect on Open Market trading volume. All reported regressions are based on the subsample used in the previous sections and include the same control variables as regression model 5 in Table 4. 25 The number of search results in the Google News archive in any language (German language only) varies between 0 (0) and 1,615,000 (26,800) per firm period with a mean of 4,656 (88) and a median of 82 (2). 27 Panel A provides evidence that the IFRS effect is not related to institutional variables at the country- or industry-level as none of the treatment sample partitions load significantly. Untabulated analyses confirm that other institutional variables such as the Rule of Law or Membership in the European Union (e.g., Daske et al., 2008) neither explain cross-sectional variation in the estimated IFRS effects.26 These results are consistent with hypothesis (H2a) that the effect of global IFRS adoption on cross-border equity investments by individual investors is not related to the institutional environment of the stocks. However, since the Open Market sample is tilted towards more visible and transparent firms (see section 3.3.1), it may not provide sufficient variation in determinants of institutional quality to enable a powerful test of this hypothesis. We therefore abstain from drawing strong conclusions with regard to the relation between the estimated IFRS effect and the institutional environment of the stocks. In contrast, the Open Market sample provides a useful setting to test hypothesis (H2b), because it comprises several thousands of stocks and, thus, far more than a single individual investor is able to draw her attention on. Panel B shows that the estimated IFRS effect is related to proxies for increases in individual investors’ attention. In the regression partitioning the sample based on Increase in Total Media Coverage, the coefficient estimate on IFRS*Mandatory*Conditional is positive (0.258) and statistically significant (t-statistic 2.02) suggesting that the IFRS effect is particularly strong for firms that experience enhanced worldwide media coverage after mandatory IFRS adoption. This pattern is even more pronounced when the increase in media coverage is based only on articles in German language (coefficient estimate: 0.455, t-statistic: 2.78). The third regression illustrates that mandatory 26 For example, the coefficient estimate on Post-FY2005*Mandatory*Conditional is -0.034 with a t-statistic of -0.18 if the regression model in Table 6 is re-estimated with EU Membership as partitioning variable. This result also suggests that other EU regulation such as the Transparency Directive that was implemented towards the end our sample period does not drive the estimated IFRS effect in our analyses (Christensen et al., 2011). 28 adopters that report a substantially higher net income under IFRS than under local GAAP experience a stronger increase in Open Market trading volume following global IFRS adoption. This effect is statistically significant at the 10% level. The final regression of Panel B shows that the IFRS effect is significantly stronger during firm periods where Open Market trading volume is clustered around earnings announcements. Taken together, these findings are consistent with hypothesis (H2b) that the effect of global IFRS adoption on cross-border equity investments by individual investors is more pronounced for attention-grabbing stocks. 5. Conclusions This study examines the impact of global IFRS adoption on cross-border equity investments by individual investors. Our proxy for these cross-border equity investments is trading volume in the Open Market, a segment at FSE designed for German individual investors to trade foreign (i.e. non-German) stocks. The empirical analyses provide evidence that stocks experience a significant increase in Open Market trading volume following mandatory adoption of IFRS. This effect is more pronounced for attention-grabbing stocks (e.g., stocks experiencing an increase in media coverage following IFRS adoption). Our results suggest that global IFRS adoption enhances cross-border equity investments by individual investors. However, this effect does not materialize equally across stocks due to individual investors’ limited attention. Taken at face value, our results support the efforts by the IASB and standard setters around the world to foster a single global set of financial reporting standards. However, we urge caution in interpreting the results in this study. First, our dataset does not allow us to directly observe through which channels global IFRS adoption influences individual investors’ decision making and trading in the Open Market. Despite our extensive efforts to identify the IFRS effect, we can therefore not fully rule out the possibility that the results reflect concurrent changes that are not 29 related to global IFRS adoption. Second, our analyses are based on the Open Market sample. The Open Market sample is a large but selected subset of the universe of global stocks comprising companies that are significantly more visible and transparent. It is an open issue whether our results on individual investor demand also apply to other less-visible and less-transparent stocks. However, this potential selection issue is common in IFRS literature as most studies rely on commercial databases that suffer from biases towards large firms (e.g., Brüggemann et al., 2012). Moreover, the size of the Open Market sample compares well to sample sizes in related studies. For example, while DeFond et al. (2011) examine 1,365 mandatory IFRS adopters from 14 countries, the Open Market sample comprises 1,693 mandatory IFRS adopters from 20 countries (see Table 1, Panel A). Third, we recognize that individual Open Market investors taking active positions in individual foreign stocks are not necessarily representative of the universe of individual investors in the global economy. However, the IASB’s and other standard setters’ efforts are naturally targeted towards investors who use financial statement information in their investment decisions. Individual investors who engage in cross-border investments are an important subset of this group. 30 APPENDIX A Competitors of the FSE Lead Broker The lead brokers at FSE face competition from alternative trading channels within and outside FSE. This appendix provides details on these trading channels. Within FSE, there are two trading platforms that work in parallel: (1) the floor where the lead brokers operate and (2) the fully electronic XETRA where quotes are automatically determined by an open limit order book. Order processing costs are lower in XETRA, but its anonymity induces higher costs arising from adverse selection. Since the adverse selection component becomes more important when trading volume is low, the floor is more attractive for less liquid stocks (Theissen, 2002). Consistent with this evidence, we find that trading activity in the illiquid Open Market usually takes place in the floor, but shifts to XETRA if liquidity is high. Specifically, trading activity in the floor is higher in 97.71% of all firm periods in the Open Market sample. In contrast, the difference in mean trading volume (Trading Volume (Euro)) is less pronounced between both systems (floor: 11,430 Euro, XETRA: 6,910 Euro). For our main analyses, we use trading volume from both the floor and from XETRA. We perform the same set of tests using only trading volume from the floor. The results remain largely unchanged. For example, the coefficient estimate on Post-FY2005*Mandatory is 0.384 with a t-statistic of 2.61 if regression model 5 in Table 4 is re-estimated with floor trading volume as the dependent variable. Outside FSE, trading segments similar to the Open Market exist at regional German exchanges in Berlin, Stuttgart and Munich. Datastream data indicates that these segments are smaller and less liquid than the Open Market in Frankfurt. Applying the selection criteria described in section 3.2, we identify a sample of 13,805 firm periods (6% of the DS Universe) for Berlin, 2,872 firm periods (1%) for Stuttgart and 441 firm periods (0%) for Munich Stock 31 Exchange. Hence, the Open Market sample (43,671 firm periods) is more than twice as large as the combined samples of the other three German exchanges. The mean proportion of non-zero trading volume days is 2.41% in Berlin, 3.79% in Stuttgart and 4.76% in Munich, compared to 24.73% at FSE (see the statistics on Trading Days (%) in Table 2, Panel A). We conclude that trading is a rare phenomenon at the regional German exchanges. To address potential coverage and quality issues with Datastream data, we analyze a further proprietary dataset from FSE that contains monthly trading volume in all Open Market stocks separately for FSE and, if applicable, for other German exchanges.27 Descriptive statistics show that FSE combines 82% of the aggregate trading volume in these stocks. Taken together, these results suggest that the Open Market lead broker at FSE faces little competition from other German exchanges. 27 Note that the FSE dataset is confined to Open Market stocks, i.e. stocks that are tradable at FSE. However, analysis of Datastream data shows that some stocks are tradable in Berlin, Stuttgart and/or Munich that are not included in the Open Market. 32 APPENDIX B Comparison of Transaction Costs: Open Market versus Home Markets German investors have two options when trading foreign stocks that are included in the Open Market. They can either trade at FSE or they can trade abroad, that is, in the respective home market. Both options involve transaction costs that differ considerably in nature. This appendix seeks to illustrate these differences by means of an example. Table A1 shows concurrent quotes for Fiat stock (ISIN IT0001976403) at the home market in Milan and at the Open Market in Frankfurt. We collect this information through the web-portal of comdirect bank, the leading online broker for German individual investors. Table A2 presents order fees at comdirect bank for trading stocks at Milan Stock Exchange (MSE) and FSE, respectively. While MSE offers better prices (i.e., the MSE bid-ask spread is inside the FSE quotes), comdirect bank clients incur lower order fees (both fixed and variable) for trading at FSE. Lower order fees at FSE outweigh the price advantage at MSE if and only if the size of the trade is sufficiently small.28 For example, a buy of 100 Fiat stocks would cost 785.06 Euro at MSE, but only 772.01 Euro at FSE.29 In contrast, buying 1,000 Fiat stocks is cheaper at MSE (7,609.42 Euro versus 7,617.48 Euro at FSE).30 The break-even where MSE turns into the more cost-efficient trading alternative is about 500 units or 3,800 Euro in this particular case. This is also reflected in the different quote sizes at MSE (around 10,000 stocks) and FSE (200 stocks). Taken together, this example confirms our depiction of the Open Market as a platform for German individual investors to trade small sizes of foreign stocks. 28 The difference in transaction costs between FSE and MSE is a monotonic function, because the price advantage of MSE (about 0.5%) is larger than the advantage of FSE in variable order fees (about 0.12%). The advantage of FSE in fixed order fees therefore has less impact on the total transaction costs the higher the trade size. 29 The order value (order processing costs) is (are) 756.00 (29.06) Euro at MSE and 760.00 (12.01) Euro at FSE. 30 The order value (order processing costs) is (are) 7,560.00 (49.42) Euro at MSE and 7,600.00 (17.48) Euro at FSE. 33 Table A1: Concurrent Quotes for Fiat Stocks (ISIN IT0001976403) This table presents concurrent quotes for Fiat stocks (ISIN IT0001976403) at the home market in Milan (German: Mailand) and at the Open Market in Frankfurt. The information was retrieved from the website of comdirect bank (www.comdirect.de) on 22 April 2009. The upper part of the table contains information on the relevant exchange (Börse), the last price (Aktuell), the time the last price was set (Zeit), the percentage difference between the last price and the price of the previous day (Diff. Vortrag) as well as the trading volume in Euro (Tages-Vol.) and in units (Gehandelte Stück). The lower part of the table provides details on the current bid (Geld) and ask quote (Brief), the time these quotes were set (Zeit), the percentage spread (Spread) as well as the size of the current bid (Geld Stk.) and ask quote (Brief Stk.). Table A2: Order Fees at comdirect bank Type of Fee Milan Stock Exchange (MSE) Frankfurt Stock Exchange (FSE) Order Provision 7.90 Euro + 0.25% of order value (min. 12.90 Euro, max. 62.90 Euro) 4.90 Euro + 0.25% of order value (min. 9.90 Euro, max. 59.90 Euro) Brokerage Fee - 0.08% of order value Exchange Fee 0.20% of order value (min. 8.66 Euro) 0.0015% of order value (min. 1.50 Euro) Delivery Fee 7.50 Euro - This table presents order fees at comdirect bank for trading stocks at Milan Stock Exchange (MSE) and Frankfurt Stock Exchange (FSE), respectively. The information was provided by comdirect bank customer support. Order Provision is the fee comdirect bank charges for its services. All other fees are charges by third parties that comdirect bank passes on. Brokerage Fee (Exchange Fee) is a charge for the FSE lead broker (respective exchange). Delivery Fee is a charge for stock clearing. 34 APPENDIX C Statistics from Brokerage Dataset This appendix describes a proprietary dataset from a German online broker which comprises information on trading activities and portfolio positions for a sample of approximately 3,000 German individual investors during the period January 1997 – April 2001.31 We use this dataset to empirically validate the following two arguments. First, we examine where the individual investors covered by the brokerage dataset trade foreign (i.e., non-German) stocks. Table B1 shows that these investors make 93.5% (113,352 / (113,352 + 7,817) of the trades in foreign stocks through the Open Market and that trade sizes in the Open Market are substantially lower than trades that are made through the home markets. Untabulated statistics show that the mean size for trades in the Open Market (home markets) is 4,840 (11,289) Euro. These results confirm that German individual investors trade foreign stocks primarily through the Open Market, especially when trade sizes are low. The second analysis focuses on the relation between changes in Open Market trading volume and changes in portfolio holdings. To ensure consistency with the main analyses, we calculate these changes for each semiannual firm-period and aggregate over all individual investors in the brokerage dataset. Table B2 illustrates that an increase in aggregate Open Market trading volume coincides with an increase in aggregate portfolio holdings in the same foreign stock in 763 out of 1,201 firm periods (64%). Similarly, a decrease in aggregate Open Market trading volume coincides with a decrease in aggregate portfolio holdings in the same stock in 651 out of 1,031 firm periods (63%). The Pearson correlation coefficient (or phi coefficient) between changes in Open Market trading volume and changes in portfolio holdings is 0.27 and statistically 31 The full dataset is described in detail in Glaser and Weber (2009). We are grateful to Markus Glaser for providing us with the statistics presented in this appendix. 35 significant at the 1% level. These results provide evidence that changes in Open Market trading volume are positively correlated with changes in portfolio holdings and, thus, cross-border equity investments. Table B1: Open Market versus Home Markets Number of Trades Trade Size in Euro (converted) from to Open Market Home Markets 500 7,309 6% 178 2% 500 1,000 12,485 11% 232 3% 1,000 2,500 34,185 30% 933 12% 2,500 5,000 28,288 25% 1,586 20% 5,000 10,000 18,920 17% 2,134 27% 10,000 25,000 9,826 9% 2,031 26% 25,000 50,000 1,875 2% 551 7% 464 0% 172 2% 113,352 100% 7,817 100% 50,000 Total This table presents statistics on foreign (i.e., non-German) stock trading by individual investors in the brokerage dataset. The dataset covers the period January 1997 – April 2001. The table distinguishes between trades in the Open Market and trades in the home markets. The markets are identified via the currency of the trade. For example, U.S. stocks traded in Euro (U.S. Dollar) are classified as Open Market (home market) trades. Observations where such a classification cannot be made unambiguously (e.g., Italian stocks traded in Euro) are eliminated. These eliminations are more prevalent after 1999 when many European exchanges started to quote stocks in Euro. The table shows the number of trades and related trade sizes in Euro (converted at the historical exchange rate for non-Euro stocks) across markets. Table B2: Changes in Open Market Trading Volume versus Changes in Portfolio Holdings Change in Aggregate Portfolio Holdings Change in Aggregate Open Market Trading Volume Decrease Total Increase Decrease 651 438 1,089 Increase 380 763 1,143 1,031 1,201 2,232 Total This table shows how changes in aggregate Open Market trading volume relate to changes in aggregate portfolio holdings in the same foreign (i.e., non-German) stock. The analysis is based on the brokerage dataset described in Table B1. Aggregate Open Market trading volume is the total number of stocks of a firm that the individual investors in the brokerage dataset bought or sold during semiannual firm periods. Aggregate portfolio holdings reflect the total number of stocks of a firm held by these investors at the end of the respective semiannual firm period. The resulting sample comprises 2,232 firm periods. 36 REFERENCES BAE, K.-H., H. TAN, and M. WELKER. ‘International GAAP Differences: The Impact on Foreign Analysts.’ The Accounting Review 83 (2008): 593-628. BAILEY, W., A. KUMAR, and D. NG. ‘Foreign Investments of U.S. Individual Investors: Causes and Consequences.’ Management Science 54 (2008): 443-459. BAMBER, L. S., O. E. BARRON, and D. E. STEVENS. ‘Trading Volume Around Earnings Announcements and Other Financial Reports: Theory, Research Design, Empirical Evidence, and Directions for Future Research.’ Contemporary Accounting Research 28 (2011): 431-471. BARBER, B. M., and T. ODEAN. ‘All That Glitters: The Effect of Attention and News on the Buying Behavior of Individual and Institutional Investors.’ The Review of Financial Studies 21 (2008): 785-818. BHATTACHARYA, N. ‘Investors’ Trade Size and Trading Responses around Earnings Announcements: An Empirical Investigation.’ The Accounting Review 76 (2001): 221-244. BRÜGGEMANN, U., J.-M. HITZ, and T. SELLHORN. ‘Intended and Unintended Consequences of Mandatory IFRS Adoption: A Review of Extant Evidence and Suggestions for Future Research.’ European Accounting Review (2012): forthcoming. BYARD, D., Y. LI, and Y. YU. ‘The Effect of Mandatory IFRS Adoption on Financial Analysts’ Information Environment.’ Journal of Accounting Research 49 (2011): 69-96. CHRISTENSEN, H., L. HAIL, and C. LEUZ. ‘Capital-Market Effects of Securities Regulation: Hysteresis, Implementation, and Enforcement.’ Unpublished paper, University of Chicago and University of Pennsylvania, 2011. DASKE, H. ‘Economic Benefits of Adoption IFRS or US-GAAP – Have the Expected Cost of Equity Capital Really Decreased?’ Journal of Business Finance & Accounting 33 (2006): 329373. DASKE, H., L. HAIL, C. LEUZ, and R. S. VERDI. ‘Mandatory IFRS Reporting Around the World: Early Evidence on the Economic Consequences.’ Journal of Accounting Research 46 (2008): 1085-1142. DEFOND, M., X. HU, M. HUNG, and S. LI. ‘The Impact of Mandatory IFRS Adoption on Foreign Mutual Fund Ownership: The Role of Comparability.’ Journal of Accounting and Economics 51 (2011): 240-258. DEUTSCHE BUNDESBANK. Securities Deposits. Special Statistical Publication 9, 2005. DAI. Factbook 2008. Deutsches Aktieninstitut, 2008. 37 ELLIOTT, W. B., F. D. HOIDGE, and K. E. JACKSON. ‘The Association between Nonprofessional Investors’ Information Choices and Their Portfolio Returns: The Importance of Investing Experience.’ Contemporary Accounting Research 25 (2008): 473-498. ENGELBERG, J., C. SASSEVILLE, and J. WILLIAMS. ‘Market Madness? The Case of Mad Money.’ Management Science 58 (2012): 351-364. ERNST, E., J. GASSEN, and B. PELLENS. Verhalten und Präferenzen deutscher Aktionäre. Studien des Deutschten Aktieninstituts 42, 2009. FESE. Share Ownership Structure in Europe. Federation of European Securities Exchanges, 2008. FLOROU, A., and P. F. POPE. ‘Mandatory IFRS Adoption and Institutional Investment Decisions.’ The Accounting Review (2012): forthcoming. FREIHUBE, T., C.-H. KEHR, J. P. KRAHNEN, and E. THEISSEN. ‘Was leisten die Kursmakler: Eine empirische Untersuchung am Beispiel der Frankfurter Wertpapierbörse.‘ Kredit und Kapital 32 (1999): 426-460. FRENCH, K. R. ‘Presidential Address: The Cost of Active Investing.’ Journal of Finance 63 (2008): 1537-1573 FSE. Cash Market: Monthly Statistics December 2010. Deutsche Börse Group, 2010. GLASER, M., and M. WEBER. ‘Which Past Returns Affect Trading Volume?’ Journal of Financial Markets 12 (2009): 1-31. GRAHAM, J. R., C. R. HARVEY, and H. HUANG. ‘Investor Competence, Trading Frequency, and Home Bias.’ Management Science 55 (2009): 1094-1106. GREIL, R. ‘Ab 29. Dezember werden in Frankfurt keine türkischen Aktien mehr gehandelt.‘ Börse Online (2005). HARRER, H., and R. MÜLLER. ‘Die Renaissance des Freiverkehrs – Eine aktuelle Analyse mit internationalem Vergleich.’ Wertpapier-Mitteilungen (2006): 653-696. HILLER VON GARTRINGEN, C. ‘Ausländische Aktien sind wieder gefragt.’ Frankfurter Allgemeine Zeitung (2006): 25 (No. 71). ILIEV, P., D. P. MILLER, and L. ROTH. ‘Uninvited U.S. Investors? Economic Consequences of Involuntary Cross-listings.’ Unpublished paper, Pennsylvania State University, Southern Methodist University and University of Alberta, 2011. 38 IASB. ‘IASC Foundation and IASB Emphasise Greater Investor Participation in the Development of IFRSs.’ International Accounting Standards Board, Press release, April 29, 2010. JEANJEAN, T., H. STOLOWY, and M. ERKENS. ‘The Economic Consequences of Increasing the International Visibility of Financial Reports.’ Unpublished paper, ESSEC Business School, HEC Paris and University of Trier, 2011. KHURANA, I. K., and P. MICHAS. ‘Mandatory IFRS Adoption and the U.S. Home Bias.’ Accounting Horizons 25 (2011): 729-753. LANDSMAN, W. R., E. L. MAYDEW, and J. R. THORNOCK. ‘The Information Content of Annual Earnings Announcements and Mandatory Adoption of IFRS.’ Journal of Accounting and Economics 53 (2012): 34-54. LEE, C. M. C. ‘Earnings News and Small Traders.’ Journal of Accounting and Economics 15 (1992): 265-302. LEE, T. A., and D. P. TWEEDIE. The Private Shareholder & the Corporate Report. The Institute of Chartered Accountants in England and Wales, London, 1977. LEUZ, C. ‘IAS Versus U.S. GAAP: Information Asymmetry-Based Evidence from Germany’s New Market.’ Journal of Accounting Research 41 (2003): 445-472. LEUZ, C., D. NANDA, and P. D. WYSOCKI. ‘Earnings management and investor protection: an international comparison.’ Journal of Financial Economics 69 (2003): 505-527. LEUZ, C., and R. E. VERRECCHIA. ‘The Economic Consequences of Increased Disclosure.’ Journal of Accounting Research 38 (2000): 91-124. LI, S. ‘Does Mandatory Adoption of International Financial Reporting Standards in the European Union Reduce the Cost of Equity Capital?’ The Accounting Review 82 (2010): 607-636. MALMENDIER, U., and D. SHANTIKUMAR. ‘Are small investors naïve about incentives?’ Journal of Financial Economics 85 (2007): 457-489. MERTON, R. C. ‘A Simple Model of Capital Market Equilibrium with Incomplete Information.’ Journal of Finance 42 (1987): 483-510. MÜLLER-MICHAELS, O., and J. WECKER. ‘Freiverkehr: gesetzliche Rahmenbedingungen und Börsenordnungen.’ Finanz Betrieb (2005): 736-743. ODEAN, T. ‘Are Investors Reluctant to Realize Their Losses?’ Journal of Finance 53 (1998): 1775-1798. 39 SCHLITT, M., and S. SCHÄFER. ‘Der neue Entry Standard der Frankfurter Wertpapierbörse.’ Die Aktiengesellschaft (2006): 147-155. SHIMA, K. M., and E. A. GORDON. ‘IFRS and the Regulatory Environment: The Case of U.S. Investor Allocation Choice.’ Journal of Accounting and Public Policy 30 (2011): 481-500. SUDMEYER, J., S. RÜCKERT, and T. KUTHE. ‘Entry Standard – Das neue Börsensegment für den Mittelstand.’ Betriebs-Berater (2006): 2703-2706. TAN, H., S. WANG, and M. WELKER. ‘Analyst Following and Forecast Accuracy after Mandated IFRS Adoptions.’ Journal of Accounting Research 49 (2011): 1307-1357. THEISSEN, E. ‘Floor versus Screen Trading: Evidence from the German Stock Market.’ Journal of Institutional and Theoretical Economics 158 (2002): 32-54. VOGELHEIM, P., D. D. SCHOENBACHLER, G. L. GORDON, and C. C. GORDON. ‘The Importance of Courting the Individual Investor.’ Business Horizons 44 (2001): 69-76. YU, G. ‘Accounting Standards and International Portfolio Holdings: Analysis of Cross-border Holdings Following Mandatory Adoption of IFRS.’ Ph.D. dissertation, University of Michigan, 2010. 40 TABLE 1 Sample Composition by Country and Firm Period Panel A: Treatment Group IFRS Adoption Countries IFRS Voluntary IFRS Mandatory Local GAAP Total Unique Firms Unique Firms Unique Firms Firm-Periods Open Market Australia DS Universe Share Open Market DS Universe Share Open Market DS Universe Share Open Market DS Universe Share 5 14 36% 301 1,455 21% - - - 1,908 15,122 13% Austria 46 53 87% 15 24 63% - - - 437 744 59% Belgium 10 22 45% 28 98 29% - - - 280 1,279 22% Czech Republic 5 14 36% 3 10 30% - - - 43 169 25% Denmark 8 17 47% 22 113 19% - - - 256 1,554 16% Finland 8 10 80% 39 124 31% - - - 469 1,607 29% France 3 13 23% 178 550 32% - - - 1,543 6,462 24% Greece 3 10 30% 29 265 11% - - - 286 2,898 10% Hong Kong 5 15 33% 349 1,038 34% - - - 2,187 10,416 21% Hungary 16 18 89% 2 7 29% - - - 144 268 54% Ireland 0 0 - 18 37 49% - - - 115 396 29% Italy 1 2 50% 97 263 37% - - - 839 2,894 29% Luxembourg 0 4 0% 1 14 7% - - - 5 159 3% Netherlands 2 3 67% 29 65 45% - - - 180 800 23% Norway 2 4 50% 54 172 31% - - - 319 1,748 18% Philippines 0 1 0% 0 166 0% - - - 0 1,645 0% Poland 0 17 0% 0 152 0% - - - 0 1,344 0% Portugal 3 5 60% 14 40 35% - - - 164 502 33% South Africa 10 27 37% 38 291 13% - - - 225 3,275 7% Spain 0 0 - 85 127 67% - - - 693 1,399 50% Sweden 3 7 43% 89 288 31% - - - 757 3,302 23% United Kingdom 2 10 20% 302 893 34% - - - 2,118 9,264 23% 132 266 50% 1,693 6,192 27% - - - 12,968 67,247 19% Total Panel B: Control Group Non-IFRS Adoption Countries IFRS Voluntary IFRS Mandatory Local GAAP Total Unique Firms Unique Firms Unique Firms Firm-Periods Open Market DS Universe Share Open Market DS Universe Share Open Market DS Universe Argentina - - - - - - 1 58 2% 4 485 Brazil - - - - - - 11 353 3% 57 2,362 2% Canada - - - - - - 456 1,380 33% 2,375 11,043 22% Chile - - - - - - 1 190 1% 4 1,671 0% China - - - - - - 30 1,508 2% 183 11,345 2% Colombia - - - - - - 0 22 0% 0 158 0% India - - - - - - 0 946 0% 0 6,095 0% Indonesia - - - - - - 68 262 26% 467 2,370 20% Israel - - - - - - 12 108 11% 50 996 5% Japan - - - - - - 458 2,635 17% 4,441 29,032 15% Malaysia - - - - - - 0 857 0% 0 8,978 0% Mexico - - - - - - 19 127 15% 49 1,117 4% Morocco - - - - - - 0 18 0% 0 147 0% Pakistan - - - - - - 0 52 0% 0 359 Share Open Market DS Universe Share 1% 0% (continued) TABLE 1 (continued) Panel B: Control Group Non-IFRS Adoption Countries IFRS Voluntary IFRS Mandatory Local GAAP Total Unique Firms Unique Firms Unique Firms Firm-Periods Open Market DS Universe Share Open Market DS Universe Share Open Market South Korea - - - - - - 0 Sri Lanka - - - - - - 0 Taiwan - - - - - - Thailand - - - - - United States - - - - Total - - - - DS Universe Share Open Market DS Universe 740 0% 0 8,117 0% 23 0% 0 213 0% 2 790 0% 4 8,888 0% - 0 402 0% 0 3,902 0% - - 2,754 5,351 51% 23,069 53,247 43% - - 3,812 15,822 24% 30,703 150,525 20% Share Panel C: Treatment and Control Group Firm Period Open Market IFRS Voluntary IFRS Mandatory Local GAAP Total Unique Firms Unique Firms Unique Firms Unique Firms DS Universe Share Open Market DS Universe Share Open Market DS Universe Share Open Market DS Universe Share 2002H1 55 181 30% 488 3,996 12% 1,645 9,551 17% 2,188 13,728 16% 2002H2 55 182 30% 516 4,062 13% 1,776 9,595 19% 2,347 13,839 17% 2003H1 54 185 29% 533 4,241 13% 1,834 9,689 19% 2,421 14,115 17% 2003H2 62 196 32% 602 4,410 14% 1,906 9,823 19% 2,570 14,429 18% 2004H1 85 210 40% 680 4,672 15% 2,045 10,834 19% 2,810 15,716 18% 2004H2 87 210 41% 728 4,838 15% 2,149 11,373 19% 2,964 16,421 18% 2005H1 96 225 43% 834 5,035 17% 2,323 11,762 20% 3,253 17,022 19% 2005H2 102 229 45% 934 5,219 18% 2,502 12,030 21% 3,538 17,478 20% 2006H1 100 223 45% 1,048 5,501 19% 2,629 12,800 21% 3,777 18,524 20% 2006H2 105 220 48% 1,150 5,606 21% 2,739 12,949 21% 3,994 18,775 21% 2007H1 106 214 50% 1,265 5,646 22% 2,874 13,219 22% 4,245 19,079 22% 2007H2 108 211 51% 1,473 5,688 26% 3,047 13,360 23% 4,628 19,259 24% 2008H1 112 211 53% 1,590 5,636 28% 3,234 13,540 24% 4,936 19,387 25% Total Firm-Periods 1,127 2,697 42% 11,841 64,550 18% 30,703 150,525 20% 43,671 217,772 20% Total Unique Firms 132 266 50% 1,693 6,192 27% 3,812 15,822 24% 5,637 22,280 25% This table presents the sample composition by country and across time. The Datastream Universe (DS Universe) comprises a total of 217,772 semiannual firm periods from 41 countries between 2002H1 and 2008H1 with sufficient data on trading volume, stock returns and accounting standards followed. H1 (H2) indicates the first (second) half of the respective year. We split the DS Universe into two groups: (1) countries that introduced IFRS in fiscal year 2005 (treatment group), and (2) countries that required domestic accounting standards throughout the sample period (control group). We include only companies in the treatment group that switched from local GAAP to IFRS before (IFRS Voluntary) or in fiscal year 2005 (IFRS Mandatory). The control group consists of companies that used domestic accounting standards (Local GAAP) throughout the sample period. Information on accounting standards followed is from Worldscope. For simplicity, we refer to Hong Kong as a country in our analyses, although it has the status of a Special Administrative Region of the People’s Republic of China. Using proprietary data from Frankfurt Stock Exchange (FSE) we identify all firms within the DS Universe whose stocks are traded in the Open Market. The Open Market sample (Open Market) is a subset of the DS Universe and consists of 43,671 firm periods from 31 countries. Share indicates the proportion of the Open Market sample relative to the DS Universe. Panel A (Panel B) reports the number of unique firms and the number of firm periods by country for the treatment (control) group. Panel C shows the number of unique firms across time for treatment and control group combined. TABLE 2 Descriptive Statistics Panel A: Open Market Sample Variables Firm-Periods Mean Std.Dev. P1 P25 Median P75 P99 Frankfurt Stock Exchange Trading Volume (Euro) 43,671 18,340 133,673 0 46 305 1,889 429,491 Trading Days (%) 43,671 24.73% 31.41% 0.00% 2.36% 9.02% 36.15% 100.00% No. of Trades 43,671 2.40 10.89 0.00 0.05 0.20 0.94 45.71 Trade Size (Euro) 39,815 2,659 10,754 60 845 1,713 3,178 15,358 Bid-Ask Spread (%) 38,913 5.24% 9.19% 0.40% 1.97% 3.13% 4.99% 50.00% Trading Volume (Euro) 43,671 29,678,350 86,143,152 6,756 778,230 5,600,336 24,427,630 394,500,000 Trading Days (%) 43,671 96.55% 5.63% 70.40% 96.85% 97.62% 97.69% 100.00% Bid-Ask Spread 38,913 0.70% 1.56% 0.03% 0.11% 0.28% 0.67% 6.70% Market Value (m Euro) 43,671 4,904 15,128 7 196 875 3,516 65,046 German/English Reporting 38,205 0.83 No. of German Inst. Investors 35,049 0.48 1.77 0.00 0.00 0.00 0.00 9.50 EM Aggregate (Firm Level) 36,822 0.45 0.21 0.07 0.28 0.45 0.60 0.91 Home Markets Other Variables Panel B: Datastream Universe (excluding Open Market Sample) Variables Firm-Periods Mean Std.Dev. P1 P25 Trading Volume (Euro) 174,101 2,369,077 20,538,086 126 28,431 Trading Days (%) 174,101 86.51% 20.50% 7.87% 88.19% Bid-Ask Spread (%) 149,594 2.93% 7.06% 0.06% Market Value (m Euro) 174,101 523 2,555 German/English Reporting 158,467 0.42 No. of German Inst. Investors 137,334 0.05 EM Aggregate (Firm Level) 131,550 0.51 Median P75 P99 190,251 1,153,291 30,613,758 94.62% 97.60% 100.00% 0.41% 0.96% 2.53% 33.33% 2 32 106 342 6,468 0.40 0.00 0.00 0.00 0.00 1.50 0.21 0.08 0.36 0.51 0.67 0.93 Home Markets Other Variables This table reports descriptive statistics of all relevant variables for the Open Market sample (Panel A) and for the rest of the Datastream (DS) Universe (Panel B). The DS Universe comprises a total of 217,772 semiannual firm periods from 41 countries between 2002H1 and 2008H1. The Open Market sample is a subset of the DS Universe and consists of 43,671 firm periods from 31 countries. Each panel is subdivided into groups of variables that depend (Frankfurt Stock Exchange (FSE) and/or Home Markets) or do not depend (Other Variables) on the trading venue. Trading Volume (Euro) is the firm period’s trading volume in Euro divided by the number of exchange trading days during the firm period. Trading Days (%) is the number of exchange trading days with non-zero trading volume divided by the number of exchange trading days during the firm period. No. of Trades is the number of ticks divided by the number of exchange trading days during the firm period. Trade Size (Euro) is the firm period’s trading volume in Euro divided by the number of trades during the firm period. FSE trading volume information is based on a proprietary dataset from Frankfurt Stock Exchange. We use trading volume from both the floor and from XETRA to compute FSE trading volume variables. Trading volume data for the home markets is retrieved from Datastream. No. of Trades and Trade Size (Euro) are available for FSE only, because Datastream does not provide information on number of trades. Bid-Ask Spread is the firm period’s median quoted spread (i.e., the difference between the closing bid and the closing ask price divided by the midpoint). For FSE, we use bid-ask spreads from the floor. For the home markets, information on closing bid and ask prices is obtained from Datastream (CRSP) for non-U.S. (U.S.) exchanges. Market Value (m Euro) is the market value of outstanding equity in Million Euro at the end of the firm period (Datastream). German/English Reporting indicates whether a firm prepares its financial statements in German or English language. We assume that all firms from Austria report in German and all firms from Australia, Canada, Ireland, New Zealand, South Africa, the United Kingdom and the United States report in English. For all other countries, we manually collect information on the reporting language from financial statements available on Thomson ONE Banker. No. of German Inst. Investors is the number of institutional investors (mutual funds, pension funds, insurance companies, TABLE 2 (continued) hedge funds, private equity funds and venture capital funds) domiciled in Germany that hold shares of the firm during the firm period. We obtain this information from the Thomson Financial Ownership quarterly data feed. For details on this database, see Florou and Pope (2012). EM Aggregate (Firm Level) is the average within-country rank of two earnings management (EM) measures: (1) the ratio of the firm level standard deviations of operating earnings and operating cash-flow (both scaled by lagged total assets), and (2) the ratio of the absolute value of accruals and the absolute value of operating cash-flows. Accounting data is obtained from Worldscope. TABLE 3 Characteristics of the Open Market Panel A: Determinants of Open Market Inclusion Dependent Variable: Open Market Sample (0 = No, 1 = Yes) Independent variables Univariate Log(Market Value) Multivariate 0.329 0.015 0.060 (12.67)*** (0.14) (0.66) Log(Home Trading Volume) German/English Reporting Log(No. of German Inst. Investors) EM Measure (Firm Level) 0.280 0.270 0.244 (8.77)*** (3.75)*** (4.19)*** 1.045 0.557 (3.57)*** (2.12)** 0.847 0.177 (5.98)*** (1.35) -0.750 -0.791 (-9.81)*** EM Measure (Country Level) (-4.47)*** -0.035 -0.038 -0.027 (-3.56)*** (-4.41)*** (-3.21)*** 0.416 0.173 0.099 (1.84)* (0.82) (0.52) 0.223 1.040 0.864 Log(MCAP/GDP) Euro Log(Distance Berlin - Capital) Fixed Effects (0.83) (4.74)*** (3.83)*** -0.123 0.002 0.081 (-1.56) (0.03) (1.00) Year Year 0.31 0.36 217,772 133,072 - Pseudo R-squared Observations Panel B: Abnormal Trading Volume of Open Market Stocks around Earnings Announcements Day Frankfurt Stock Exchange Mean Median -5 0.0010 -1.0000 -4 -0.0989 -1.0000 -3 -0.0503 -2 -1 Home Markets Mean Median -0.30 -0.0068 -0.1961 -0.33 -0.70 0.0017 -0.1850 -0.11 -1.0000 -0.44 -0.0227 -0.2016 -0.47 -0.0272 -1.0000 -0.30 0.0309 -0.1702 0.10 0.0147 -1.0000 0.58 0.1321 -0.0693 1.45 0 0.5795 -1.0000 4.91 *** 0.9632 0.4005 5.82 *** 1 1.1099 -1.0000 6.63 *** 1.1195 0.4890 6.29 *** 2 0.3569 -1.0000 3.34 *** 0.4165 0.1266 3.53 *** 3 0.1903 -1.0000 1.54 0.2714 0.0254 2.55 ** 4 0.1512 -1.0000 1.17 0.1850 -0.0393 1.79 * 5 0.0585 -1.0000 0.20 0.1531 -0.0536 1.58 Observations 18,362 Rank Statistic 18,362 Rank Statistic TABLE 3 (continued) This table reports descriptive evidence on the determinants of Open Market inclusion (Panel A) and on trading volume reactions around earnings announcements in the Open Market versus the respective home markets (Panel B). In Panel A, we relate the likelihood of inclusion into the Open Market to various country- and firm-specific variables. The DS Universe comprises a total of 217,772 semiannual firm periods from 41 countries between 2002H1 and 2008H1. The Open Market sample is a subset of the DS Universe and consists of 43,671 firm periods from 31 countries. The dependent variable in all regressions is a binary variable that equals one (zero) for all firm periods in the Open Market sample (the rest of the DS Universe). EM Measure (Country Level) is an updated country-specific earnings management score based on Leuz et al. (2003). To construct this score we use all firms in the DS Universe with sufficient accounting data in Worldscope for fiscal years 2001-2009. The score is the average countryspecific rank of three earnings management measures: (1) the country’s median ratio of the firm level standard deviations of operating earnings and cash flow from operations (both scaled by lagged total assets), (2) the country’s Spearman correlation between the change in accruals and the change in cash flow from operations (both scaled by lagged total assets), and (3) the country’s median ratio of the absolute value of accruals and the absolute value of the cash flow from operations. MCAP/GDP is the ratio of a country’s stock market capitalization to its Gross Domestic Product (GDP). Yearly ratios are obtained from the World Bank (www.fsdi.org). We compute this variable as the mean ratio over the period 2001-2004. Euro distinguishes between countries that have adopted the Euro as their national currency (variable equals one) and countries that have not (variable equals zero). We collect this information from the European Central Bank (www.ecb.int). Distance Berlin – Capital is the distance between Berlin, capital of Germany, and the capital of the respective country in kilometers. Data source is the French research centre in international economics (CEPII). For a description of the remaining variables see Table 2. Panel A reports probit coefficient estimates, and (in parentheses) t-statistics. The t-statistics are based on standard errors that are clustered by country. We use the natural logarithm of the raw values where indicated. ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels (two-tailed) respectively. In Panel B, we focus on Open Market stocks and compare abnormal trading volume around earnings announcements at Frankfurt Stock Exchange (FSE) versus the respective home markets. The analysis is based on a sample of 18,362 annual earnings announcement dates retrieved from IBES. Abnormal trading volume is the difference between trading volume on the event day and the mean daily volume for that stock over the pre-announcement window (-120, -21), scaled by the mean daily volume. Abnormal trading volume is winsorized, by event day, at the 99% level. Since the average Open Market stock is traded on less than half of the exchange trading days (see descriptive statistics on Trading Days (%) in Table 2, Panel A), the median abnormal trading volume at FSE remains at -1.000 throughout the event window. To test for significance we use Corrado’s (1989) non-parametric rank test. ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels (two-tailed) respectively. TABLE 4 The Effect of Global IFRS Adoption on Open Market Trading Volume Dependent Variable: Log(FSE Trading Volume) Independent Variables Full Sample (1) Subsample (max. 100 Firms per Country) (2) (3) 0.052 -0.359 -0.315 (0.48) (-6.82)*** 0.912 0.679 (6.59)*** (5.23)*** (4) (5) (6) 0.361 -0.203 -0.279 (-11.20)*** (1.67) (-1.46) (-3.67)*** 0.511 0.647 0.457 0.374 (5.42)*** (2.83)*** (2.96)*** (2.91)*** Test Variables Post-FY2005 Post-FY2005 * Mandatory Post-FY2005 * Voluntary Voluntary * IFRS 0.841 0.459 0.163 0.523 0.206 0.001 (3.16)*** (2.17)** (0.97) (1.61) (0.97) (0.00) 1.247 0.737 0.847 1.269 0.759 0.987 (4.16)*** (2.85)*** (2.82)*** (4.32)*** (3.56)*** (3.38)*** Control Variables Log(Market Value) Log(Home Trading Volume) 0.304 0.245 0.672 0.715 (1.33) (1.05) (4.90)*** (4.79)*** 0.864 0.894 0.723 0.733 (12.68)*** (12.44)*** (10.44)*** (9.76)*** German/English Reporting Log(No. of German Inst. Investors) Log(Spread Difference) 0.008 0.174 (0.03) (0.49) 0.040 0.181 (0.32) (1.13) -0.380 -0.322 (-11.30)*** (-5.86)*** Fixed Effects Firm Firm Firm Firm Firm Firm R-squared 0.72 0.76 0.79 0.74 0.78 0.81 43,671 43,671 27,305 12,063 12,063 7,867 Observations This table presents results from regressions that relate Open Market trading volume at Frankfurt Stock Exchange (FSE) to IFRS adoption. The Open Market sample comprises a total of 43,671 semiannual firm periods from 31 countries between 2002H1 and 2008H1. Regression model 1-3 (4-6) is based on the full sample (a subsample that allows a maximum of 100 firms per country). To create the subsample we focus on firms that are part of the Open Market sample both before and after IFRS introduction, sort these firms within each country by their average market value over the sample period and then select every N/100th firm if the number of firms per country N is greater than 100. Post-FY2005 equals one (zero) for fiscal year 2005 and later (2004 and earlier). We define that Post-FY2005 switches in the first firm period following the end of fiscal year 2005, that is, if the fiscal year ends in December 2005 Post-FY2005 switches in 2006H1. Voluntary (Mandatory) is a firm-level dummy variable and indicates companies that switched from local GAAP to IFRS before (in) fiscal year 2005. IFRS is a binary variable measured at the firm period level that indicates whether IFRS is applied or not. Spread Difference is the difference between Bid-Ask Spread at FSE and Bid-Ask Spread in the respective home market. For a description of the remaining variables see Table 2. The table reports OLS coefficient estimates and (in parentheses) t-statistics. The t-statistics are based on standard errors that are clustered by country. We use the natural logarithm of the raw values (plus a small constant when the raw value is zero) where indicated. ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels (two-tailed) respectively. TABLE 5 Alternative Identification of the IFRS Effect Panel A: Shifting the Switch of Post-FY2005 Dummy Dependent Variable: Log(FSE Trading Volume) Full Sample Subsample (max. 100 Firms per Country) Re-estimation of Table 4, Model (2) Re-estimation of Table 4, Model (5) Post-FY2005 * Mandatory Post-FY2005 * Mandatory Coefficient t-stat Coefficient t-stat Post-FY2005 Dummy switches... ... 2 Periods earlier 0.667 (5.00)*** 0.338 (1.65) ... 1 Period earlier 0.672 (4.80)*** 0.408 (2.30)** ... in the same Period (= Table 4) 0.680 (5.22)*** 0.454 (2.93)*** ... 1 Period later 0.597 (5.52)*** 0.463 (2.85)*** ... 2 Periods later 0.569 (4.58)*** 0.455 (2.67)** Panel B: Within-Country Analysis Independent Variables Dependent Variable: Log(FSE Trading Volume) United Kingdom: Main Market versus AIM Test Variables Post-FY2005 Post-FY2005 * Mandatory -0.587 -0.482 (-1.38) (-1.22) 1.386 0.990 (3.03)*** (2.34)** Control Variables Log(Market Value) 0.318 (1.34) Log(Home Trading Volume) 0.885 (6.12)*** Fixed Effects R-squared Observations Firm Firm 0.78 0.80 2,541 2,541 This table shows alternative analyses to identify the IFRS effect on Open Market trading volume. The Open Market sample comprises a total of 43,671 semiannual firm periods from 31 countries between 2002H1 and 2008H1. In Panel A, we re-estimate regression model 2 and 5 (Table 4), respectively, by defining Post-FY2005 to switch up to two periods earlier or later. For brevity, we only report coefficient estimates and t-statistics for the interaction term Post-FY2005 * Mandatory. In Panel B, we perform within-country analysis focusing on the United Kingdom. We use mandatory IFRS adopters listed in the Main Market at London Stock Exchange (LSE) as the treatment group and companies listed in the Alternative Investment Market (AIM) at LSE as the control group. AIM companies were not required to adopt IFRS before fiscal year 2007 and therefore not included in the Open Market sample presented in Table 1. All variables are defined as in Table 4. We delete firm periods of AIM companies after these adopted IFRS. This procedure yields a sample of 2,104 (437) firm periods from 302 (93) mandatory IFRS adopters (AIM companies). Panel B reports OLS coefficient estimates and (in parentheses) t-statistics. The t-statistics are based on standard errors that are clustered by firm. We use the natural logarithm of the raw values (plus a small constant when the raw value is zero) where indicated. ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels (two-tailed) respectively. TABLE 6 Cross-Sectional Variation in the IFRS Effect Panel A: IFRS Effect Conditional on the Institutional Environment Dependent Variable: Log(FSE Trading Volume) Independent Variables Post-FY2005 Post-FY2005 * Mandatory Post-FY2005 * Mandatory * Conditional Post-FY2005 * Voluntary Post-FY2005 * Voluntary * Conditional Voluntary * IFRS Control Variables Fixed Effects R-squared Observations Strong Credibility Large Δ Uniformity -0.204 -0.204 -0.194 -0.194 (-1.46) (-1.46) (-1.33) (-1.33) 0.513 0.452 0.492 0.478 (2.53)** (3.17)*** (2.95)*** (2.93)*** -0.099 0.006 -0.090 -0.032 (-0.61) (0.04) (-0.41) (-0.14) 0.187 0.423 -0.072 0.393 (0.44) (2.22)** (-0.22) (1.81)* Long Accounting Distance Strong Credibility * Large Δ Uniformity 0.022 -0.394 0.422 -0.666 (0.05) (-1.37) (1.07) (-1.52) 0.759 0.766 0.805 0.850 (3.56)*** (3.46)*** (4.33)*** (4.08)*** Included Included Included Included Firm Firm Firm Firm 0.78 0.78 0.78 0.78 12,063 11,889 9,589 9,589 Panel B: IFRS Effect Conditional on Attention Proxies Dependent Variable: Log(FSE Trading Volume) Independent Variables Post-FY2005 Post-FY2005 * Mandatory Post-FY2005 * Mandatory * Conditional Post-FY2005 * Voluntary Post-FY2005 * Voluntary * Conditional Increase in Total Media Coverage Increase in German Media Coverage Large IFRS Restatements Strong EA Reactions -0.200 -0.200 -0.190 -0.202 (-1.44) (-1.44) (-1.31) (-1.46) 0.254 0.166 0.356 0.427 (1.69) (0.96) (2.07)** (2.79)*** 0.258 0.455 0.206 0.272 (2.02)* (2.78)** (1.81)* (2.88)*** 0.262 0.094 0.220 0.195 (1.03) (0.37) (1.02) (0.91) -0.063 0.166 0.113 (-0.29) (0.82) (0.87) 0.762 0.751 0.778 0.759 (3.57)*** (3.48)*** (3.56)*** (3.56)*** Included Included Included Included Fixed Effects Firm Firm Firm Firm R-squared 0.96 0.91 0.79 0.78 12,063 12,063 10,811 12,063 Voluntary * IFRS Control Variables Observations TABLE 6 (continued) This table presents results from regressions that relate Open Market trading volume at Frankfurt Stock Exchange (FSE) to IFRS adoption conditional on variables related to the institutional environment (Panel A) and to individual investors’ attention (Panel B). The Open Market sample comprises a total of 43,671 semiannual firm periods from 31 countries between 2002H1 and 2008H1. All regressions are based on model 5 in Table 4 using Conditional as an additional independent variable to partition the treatment group. The conditional variables in Panel A are defined as follows: Long Accounting Distance is based on the Bae et al. (2008) summary score of how a country’s local GAAP differs from IFRS on 21 key accounting dimensions and equals one (zero) for countries with a score greater (equal to or less) than the sample median of 9. Strong Credibility is based on the earnings management score from Leuz et al. (2003) and takes a value of one (zero) for countries with a score of less (equal to or greater) than the sample median of 18.3. Large Δ Uniformity is based on the changes in uniformity measure from DeFond et al. (2011) and equals one (zero) for industry-country clusters with changes in uniformity greater (equal to or less) than the sample median of 39.67. The conditional variables in Panel B are defined as follows: Total Media Coverage (German Media Coverage) is based on the number of search results in the Google News archive (http://news.google.com/archivesearch) in any language (German only). This number reflects the number of articles that were published on Google News during the relevant period and that contain either the company name (as provided by Worldscope item WC06001) or the firm-specific ISIN code. The related conditional variables equal one (zero) if media coverage of treatment group firm is, on average, higher (lower or the same) after fiscal year 2005 than before. Large IFRS Restatements equals one (zero) if the percentage difference between the restated net income under IFRS and the originally reported net income under local GAAP for fiscal year 2004 is above (below or equal to) the median percentage difference of 2,17%. Restatement information is from Worldscope (item WC01551R) and only available for a subset of mandatory IFRS adopters. Strong EA Reactions equals one if the average Open Market trading volume during the three-day window around the earnings announcement is higher than the average Open Market trading volume over the relevant firm period, and zero otherwise. The table reports OLS coefficient estimates and (in parentheses) t-statistics. The t-statistics are based on standard errors that are clustered by country. We use the natural logarithm of the raw values (plus a small constant when the raw value is zero) where indicated. ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels (two-tailed) respectively.