When does the Adoption and Use of IFRS increase Foreign Investment?

When does the Adoption and Use of IFRS increase Foreign Investment? Bowe Hansen Virginia Tech University Mihail Miletkov University of New Hampshire M. Babajide Wintoki University of Kansas Current Draft: April 5, 2013 ABSTRACT Extant research suggests that the use of IFRS provides firms an opportunity to reduce information asymmetry and make themselves more attractive to foreign investors. However, Daske et al., (2013) suggest that any benefits of IFRS adoption will accrue only to firms with incentives to provide transparent financial reporting. We thus predict that foreign ownership will be higher in IFRS firms with strong reporting incentives than in IFRS firms with weaker reporting incentives. Using a sample of over 54,000 firm-years from 72 countries during the period 2001 to 2011, we find evidence supporting this hypothesis. Further, a one quartile increase in transparency is associated with a 0.74% higher level of foreign ownership, which is economically meaningful compared to a mean level of foreign ownership of 8.14% across our sample. Additional tests show that our results are driven by countries with weak investor protection and by investments made by institutional investors. Key Words: IFRS, Home Bias, International Accounting, Reporting Incentives We are grateful for the research support of Virginia Tech University, the University of New Hampshire, and the University of Kansas. We appreciate comments from Valaria Vendrzyk (the AAA discussant), Bryan Cloyd, Mitch Oler, and workshop participants at Virginia Tech University and the 2012 AAA Annual Meeting. Contact information B. Hansen: bowe.hansen@vt.edu; M. Miletkov: mihail.miletkov@unh.edu; M. B. Wintoki: jwintoki@ku.edu. 1

I. INTRODUCTION The home bias literature, documents that observed levels of foreign investment are below what would appear to be economically efficient (French and Poterba, 1991). One factor that has been proposed to contribute to home bias is information asymmetry between domestic and foreign investors (Kang and Stulz, 1997). Recent research by DeFond et al. (2011) and Yu (2010) has used the adoption of International Financial Reporting Standards (IFRS) by over 80 countries in the last decade as a natural experiment in which to test this hypothesis. In general, these studies have found a positive relation between IFRS adoption and foreign shareholding. However, neither of these studies considers heterogeneity in the application of IFRS at the firm level, and how this could relate to foreign investment. The goal of the current study is to address this gap in the literature by investigating whether firm reporting incentives and observed application of IFRS are associated with the level of foreign ownership. We view this as a significant contribution given the long-standing debate regarding the relative importance of accounting standards versus reporting incentives in determining accounting quality (Ball et al., 2003; Burgstahler et al., 2006). In particular, the findings in Daske et al., (2013) indicate that variation in the reporting incentives and application of IFRS by voluntary adopters leads to systematic variation in capital market effects such that only firms with strong reporting incentives experience decreases in their cost of capital and increases in liquidity subsequent to their adoption of IFRS. 1 In line with their findings, we argue that foreign investors will be able to discriminate between those firms that use the adoption of IFRS as part of an overall effort to provide high quality financial reporting, and those that simply adopt the IFRS name, but apply it in a manner which does not result in transparent accounting. Firms that 1 Daske et al. (2013) do not propose that the adoption of IFRS causes these capital market benefits, rather that the economic benefits result from an underlying motivation to improve financial reporting quality, which may be related to the decision to adopt IFRS. 2

combine IFRS adoption with more transparent financial accounting, rather than merely adopting the IFRS label, will experience lower levels of information asymmetry between foreign and domestic investors, making such firms more attractive to foreign investors. This leads us to hypothesize that the level of foreign ownership will be higher in IFRS firms with strong reporting incentives than in IFRS firms with weaker reporting incentives. We test this hypothesis using a broad sample of over 54,000 firm-years from 72 countries during the period 2001 to 2011. We identify foreign ownership using the OSIRIS database compiled by Bureau van Dijk Electronic Publishing, and measure firm level reporting incentives based on a metric proposed by Daske et al. (2013). Specifically, our measure of observed firmlevel reporting, Transparency, is the magnitude of accruals relative to cash flows from operations averaged over the prior three years. We classify as IFRS all firms that had adopted IFRS either voluntarily or as a result of a country level adoption. We do not separate mandatory and voluntary adopters because we argue that the impact of IFRS usage will vary as a function of firm reporting incentives. If our hypothesis is correct, this result will hold regardless of whether the firm adopted IFRS voluntarily or adopted IFRS when it was mandated in its country. 2 We find evidence that supports our hypothesis. Specifically, IFRS firms with higher transparency have higher foreign ownership than IFRS firms with lower transparency, and this result is robust to the inclusion of firm-fixed effects. Moreover, the higher level of foreign ownership for IFRS firms with greater transparency is economically significant, with a one quartile increase in our Transparency variable being associated with a 74 basis point increase in the level of foreign ownership, relative to a mean foreign ownership level of 8.14%. This 2 The research design choice to split firms based on whether they voluntarily or mandatorily adopt IFRS is often used as proxy for reporting incentives (e.g., Christensen et al., 2008). However, our Transparency measure allows us to measure reporting incentives directly, thus allowing for the possibility that some voluntarily adopters likely adopted IFRS in name only, and some mandatory adopters have incentives to provide high quality accounting information. 3

relationship appears to be causal; we find that IFRS firms with higher transparency experienced a significantly larger increase in foreign ownership following IFRS adoption than IFRS firms with lower transparency. We extend this primary finding by performing three additional analyses. In the first of these, we investigate the impact of country-level investor protection on the relationship between firm-level reporting incentives and foreign ownership. Existing research has documented that variation in country-level institutions has a systematic impact on the application of accounting standards. For example, Leuz et al. (2003) document that country-level investor protection plays an important role in determining the level of earnings management in which firms engage, and Ball et al. (2000) find that accounting in civil law countries is less timely than in common law countries. While some prior evidence 3 suggests that only firms from countries with strong investor protection would experience higher levels of foreign ownership as a result of the conversion to IFRS, ex ante, it is unclear whether country-level and firm-level reporting incentives act as complements or substitutes with respect to foreign ownership following IFRS adoption. If they act as complements, then we expect to find the highest levels of foreign ownership in firms with both strong firm- and country- level reporting incentives. On the other hand, if they act as substitutes, then we expect to find the greatest benefit for firms with either strong firm-level, or strong country-level incentives, but not necessarily both. We find that higher levels of transparency for IFRS users in weak investor protection countries are associated with greater foreign ownership. However, the same is not true in countries with strong investor protection. Based on these results, we infer that firm-level and country-level reporting incentives act primarily as substitutes. Specifically, investors are 3 See for example, DeFond et al. (2011) 4

primarily concerned with firm level reporting incentives when they are concerned that country level enforcement is lax. In our second additional analysis, we note that the OSIRIS database allows us to distinguish among different investor types. Specifically, we are able to determine whether the foreign owner is an institution, an industrial firm, an individual/family, or a government. We repeat our primary analysis separately for institutional investors and the group comprised of all other investors. Consistent with prior evidence we view institutional investors as sophisticated investors (Badrinath et al., 1995; Nagel, 2005; Field and Lowry, 2009). Therefore, we expect to find that higher levels of transparency are associated with greater foreign investment in IFRS firms by institutional investors. Individual investors are more likely than institutional investors to be subject to behavioral biases and to misinterpret financial accounting information (Odean, 1998, Cohen et al., 2002). Relatedly, cross-border investing by industrial firms and governments are more likely to be related to strategic and political motivations. Therefore, we do not expect to find an association between cross-border investing and financial reporting transparency among these groups. Our results support these expectations. Consistent with the joint hypotheses that our measure, Transparency, reflects the information environment of the firm, and that institutional investors are sophisticated, we find a significantly higher level of foreign institutional ownership in IFRS firms with higher transparency, and find no association between transparency and foreign ownership of IFRS firms by the other types of investors. In our final additional analysis, we rerun our primary tests using two alternative proxies measuring the firms incentives to provide transparent financial disclosures also drawn from Daske et al. (2013). First, we use the primary factor from a factor analysis of the following six firm attributes which prior literature indicates are associated with reporting incentives: firm size, 5

leverage, profitability, growth opportunities, ownership concentration, and internationalization. Second, we use the natural log of the number of analysts following the firm (plus one) as a proxy for the capital market s demand for information from the firm. For both of these alternative proxies, we find that IFRS firms with higher reporting incentives have higher levels of foreign ownership than IFRS firms with lower reporting incentives. Our findings contribute to the literature in several ways. We contribute to the literature that investigates the role of accounting information, and in particular, the adoption of IFRS, in mitigating home bias. Covrig et al. (2007), present evidence that voluntary adoption of international accounting standards reduces home bias. In more recent research, Yu (2010) finds that cross-border holdings by international mutual funds increase following mandatory adoption of IFRS, and argues that one mechanism driving this increase is a decrease in information acquisition costs that results from IFRS closing the gap between investors and investees accounting standards. DeFond et al. (2011) present evidence supporting the argument that the adoption of IFRS in many countries makes cross-border investment less costly because it allows for a more efficient analysis of alternative investments. Our paper differs from, and extends, Yu (2010) and DeFond et al. (2011) in two ways. First, both of these papers propose a benefit of IFRS adoption that results from its role as a common accounting language. In contrast, we consider heterogeneity in the application of IFRS at the firm level and assert that the adoption of IFRS, combined with strong reporting incentives, leads to more transparent financial statements. These transparent financial statements in turn lead to lower levels of information asymmetry between domestic and foreign investors, making a firm more attractive to foreign investors. 6

Second, Yu (2010) and DeFond et al. (2011) use foreign mutual fund holdings as a proxy for foreign ownership. We use a much broader dataset which contains data on all types of investors (e.g., institutions, individuals, governments). This allows us to investigate the impact of IFRS adoption and reporting incentives on the investment decisions of different types of investors which adds richness to our study and serves to validate our proxy for reporting transparency. In addition, we contribute to the stream of literature contrasting the role of accounting standards and reporting incentives in determining accounting quality. Prior literature has documented that country level institutions impact the quality of reported accounting numbers (Ball et al., 2003 and Lang et al., 2006). In addition, Ball and Shivakumar (2005) and Burgstahler et al. (2006) document that differences in the reporting incentives of public and private firms lead to different earnings management and earnings quality measures between these two groups of firms. Applying the reporting incentives argument to IFRS adoption, Daske et al. (2013) find that firms with strong reporting incentives experience a reduction in their cost of capital and an increase in their liquidity around the time of IFRS adoption. We extend this literature by investigating how reporting incentives interact with IFRS adoption and usage to impact the ability of firms to attract foreign investors. The rest of this paper proceeds as follows. In the next section, we review the related literature and develop our hypothesis. Section three provides data definitions, presents our research methodology, and provides descriptive statistics. We present and discuss the results of our primary analyses in the fourth section. In the fifth section, we describe our robustness tests. The final section concludes. 7

II. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT Home bias, the tendency for investors to dedicate a higher percentage of their portfolio to domestic investments than is considered to be optimal under contemporary finance theory, is a much investigated phenomenon in the finance, accounting, and economics literature (see, for example, French and Proterba, 1991; Cooper and Kaplanis, 1994; Karolyi and Stulz, 2002, among others). Institutional factors such as restrictions on ownership and taxation concerns are likely to be a contributing factor in home bias (Black, 1974). However, literature in the area to date indicates that information asymmetries between foreign investors and domestic investors are likely to be the primary cause of home bias. Kang and Stulz (1997) find that foreign ownership of Japanese firms is related to attributes that reflect information asymmetry such as size, foreign sales, and the presence of an American Depository Receipt (ADR). In addition, Edison and Warnock (2004) present evidence that US investors do not exhibit home bias against firms from emerging markets that are cross-listed on a US exchange. Further, Ahearne et al. (2004) find that the portion of a country s firms cross-listed in the U.S. is positively associated with that country s weight in US investors portfolios. The accounting choices of the firm are likely to have a direct effect on the level of information asymmetry between foreign and domestic investors due to accounting s role as a primary information dissemination channel (Gehrig, 1993; Young and Guenther, 2003). Specifically, the adoption of IFRS could lead to improved financial reporting in terms of both the quantity and quality of information disclosed, which would reduce the level of information asymmetry, and thereby make the firm more attractive to foreign investors. Consistent with this argument, Ding et al. (2007), and Bae et al. (2008) provide evidence that IFRS has more comprehensive disclosure requirements than most domestic GAAP. Further, Ashbaugh and 8

Pincus (2001) find that non-u.s. firms that adopt international accounting standards experience a decrease in analyst forecast errors. Building on the above evidence, Covrig et al. (2007) find that foreign mutual funds are more likely to invest in firms that voluntarily adopted IFRS. However, their findings do not clarify whether the higher level of ownership results from increased accounting quality, increased comparability, and/or reduced information acquisition costs. More recent research has capitalized on the widespread adoption of IFRS in the last decade to provide further insight into the relationship between IFRS and foreign ownership. The results of research that investigates the impact of mandatory IFRS adoption on foreign ownership are mixed. Amiram (2009) investigates the impact of country-level conversion to IFRS on aggregate foreign equity portfolio investments and finds that countries that convert to IFRS have a significantly higher level of foreign investment. Yu (2010) finds that foreign mutual funds increase their investments in firms that adopt IFRS relative to non-adopters in the same country. However, Beneish et al. (2010) find that countries that adopt IFRS do not experience an increase in aggregate foreign equity investment, but do experience an increase in foreign debt investment. One potential reason for these inconsistent results may be that the effect of IFRS adoption varies across firms. Along these lines, DeFond et al. (2011) argue that the mandatory adoption of IFRS is beneficial only when it increases the comparability of foreign firms to other firms in the same industry. They find that IFRS firms experience an increase in foreign mutual fund ownership if they have strong increases in comparability. In a related vein, Yu (2010) presents evidence that a primary benefit of the adoption of IFRS is that it reduces the differences between the accounting used by the firm and the accounting used in the investors home markets, thereby reducing information asymmetries between foreign and domestic investors. 9

However, DeFond et al. (2011) and Yu (2010) do not investigate the impact that variation in the application of IFRS has on foreign investors. 4 Prior research has shown that variation in reporting incentives, both across countries and within countries, have a significant impact on accounting outcomes (Ball et al. 2003; Ball and Shivakumar, 2005; Burgstahler et al., 2006). Firm-level incentives are likely to have a particularly strong effect under IFRS as it is a principles-based accounting system which allows for a substantial amount of flexibility in its application (Henry, 2008). Supporting this claim, Christensen et al. (2008) present evidence that German firms that voluntarily adopt IFRS (i.e., those German firms with incentives to do so) display greater levels of timely loss recognition, and lower levels of earnings management after adoption, while mandatory adopters do not display such changes. Further, Byard et al. (2011) find that the impact of mandatory IFRS adoption on analysts forecast errors and forecast dispersion is a product of both country-level and firm-level reporting incentives. Finally, Daske et al. (2013) use a sample of voluntary IFRS adopters from 30 countries over the time period 1990 2005, and find an increase in liquidity and a decrease in the cost of capital only for those firms with strong incentives to report transparent financial information. Based on the above studies we expect that the quality of financial accounting information will vary systematically for IFRS firms in a manner that reflects their reporting incentives. As a result of this variation, it is likely that the use of IFRS will decrease information asymmetry between domestic and foreign investors only for those firms which apply IFRS in a manner that leads to more transparent financial statements. This leads us to hypothesize that: Hypothesis 1: The use of IFRS will be associated with higher levels of foreign ownership for firms with stronger incentives to provide transparent financial reporting disclosures than for firms with weaker reporting incentives. 4 Both Yu (2010) and DeFond et al. (2011) do control for country-level enforcement, but unlike our paper, neither paper looks at firm level reporting incentives. 10

In addition, it is likely that country-level reporting incentives will impact the application of IFRS, and therefore its ability to reduce information asymmetries between foreign and domestic investors. However, the impact of country-level incentives and, in particular, their interaction with firm specific incentives is not clear. Country- and firm-level incentives could act as complements, or as substitutes. Evidence that firm- and country- level incentives act as complements can be seen in DeFond et al. (2011) who find that foreign mutual fund ownership increases for IFRS adopters as a result of increased comparability only in countries with high levels of average earnings quality. If country- and firm-level incentives are complementary, then we would expect IFRS firms with greater reporting transparency that are located in countries with strong enforcement mechanisms to have a higher level of foreign ownership than IFRS firms with lower reporting transparency. On the other hand, Byard et al. (2011) find that analyst forecast errors and forecast dispersion decrease more for firms with strong reporting incentives in countries with weak legal enforcement than for firms with strong reporting incentives in countries with strong legal enforcement. This indicates that at least to some degree firm- and country- level incentives may act as substitutes. In this case we would expect a higher level of foreign ownership in IFRS firms with either strong country-level or strong firm-level reporting incentives. Given these competing arguments, and prior evidence supporting them, we do not hypothesize regarding the effect of the interaction between firm- and country- level reporting incentives on foreign ownership. However, we explore this relationship in additional tests by dividing the sample based on country-level enforcement mechanisms and performing our analysis separately in higher enforcement and lower enforcement countries. 11

In addition, the importance of a rigorous application of IFRS may differ across investor types. The OSIRIS database provides information on the type of foreign investor (institutions, individuals/families, industrial firms, and governments/sovereign wealth funds) for each firm. We are able to leverage this data to investigate the extent to which different investor types react differently to variations in transparency. We expect that institutional investors are more likely to evaluate the quality of accounting numbers when making their investment choice, and therefore, are more likely to invest in foreign IFRS firms with higher levels of transparency than are other types of investors. Prior evidence has demonstrated that institutional investors are sophisticated and are able to earn abnormal returns on their investments as a result of a superior ability to interpret publicly available information. Cohen et al. (2002) provide evidence that institutional investors better understand the impact of cash flow news on future performance than do individual investors. Further, Barber and Odean (2008) argue that the greater resources available to institutional investors allow them to evaluate a larger pool of stocks for potential investment than individual investors, who are more likely to invest in firms that are attention-grabbing. Finally, Field and Lowry (2009) find that institutions are able to earn higher returns around IPOs than other investor groups, and that these higher returns are primarily driven by a better ability to interpret publicly available information. In contrast, individual investors are more likely to be subject to behavioral biases and information processing constraints (Cohen et al., 2002; Field and Lowry, 2009). Further, investments by industrial firms are more likely to be driven by strategic motivations, such as: tax implications, diversification, or, most likely in this case, an effort to gain access to a foreign market. Similarly, government investment is likely to be influenced by strategic political goals, 12

and sovereign wealth fund investments may also be influenced by political motivations, as well as regulations designed to limit the riskiness of their investment portfolios. Therefore, to the extent that our measure, Transparency, accurately reflects the information environment of the firm, we expect institutional, but not other types of investors to have higher levels of foreign investment in IFRS firms with higher Transparency. III. DATA, METHODOLOGY, AND SAMPLE DESCRIPTION Descriptions of Key Variables Foreign Ownership We measure foreign ownership for all firms covered in the OSIRIS database for the years 2001 2011. 5 The dataset is unique compared to other studies investigating the impact of IFRS adoption, or usage, on foreign ownership in that it includes all types of foreign owners (i.e., institutions, individuals, governments, etc.). The vast majority of previous literature in the area has used holdings by foreign mutual funds collected from Thomson Financial Services as a proxy for total foreign holdings. We measure our dependent variable, Foreign Ownership, as the percentage of the firm s total shares outstanding owned by foreign investors as of year t. Transparency We measure firms observable financial reporting transparency by adapting the reporting behavior score measure from Daske et al. (2013). The measure is based on attributes of the firm s reporting observable to investors prior to their investment decision. Specifically, we first measure the absolute value of accruals divided by the absolute value of cash flows multiplied by negative 1. Multiplying by negative 1 increases the intuitiveness of the result as 5 The OSIRIS database is administered by Bureau van Dijk Electronic Publishing, and aims to provide coverage of all publicly listed companies worldwide. The database includes financial, ownership, stock data, and other information on over 45,000 firms from more than 140 countries. The OSIRIS database and its European counterpart Amadeus (also a product of Bureau van Dijk Electronic Publishing) have been recently used by Li et al. (2006), John et al. (2008) and Faccio et al. (2011) among others. 13

larger (less negative) numbers can be interpreted as indicative of stronger reporting incentives. Consistent with Daske et al. (2013), we calculate this measure over the three-year time period: t- 3, t-2, and t-1, and use the average over the three year period to reduce the impact of noise in individual year measures. 6 We then modify the Daske et al. (2013) measure by ranking each firm on this measure and dividing the firms into 4 equally sized quartiles to form our measure of reporting incentives, Transparency. We use the quartiles of the measure because the raw measure is truncated at zero by definition and has a significant number of outliers in the left tail, resulting in a highly skewed distribution. Ranking the observations removes the skewness and increases the interpretability of our results. 7 The use of this measure as a proxy for transparency is based on evidence that accruals and cash flows provide different information (Sloan, 1996), and the argument that large accruals could indicate that managers are using their discretion to provide misleading financial statements (Leuz et al., 2003). Consistent with this argument, Wysocki (2004) shows that this measure is highly associated with several other earnings management proxies used in international settings. An implicit goal of earnings management is to conceal information from financial statement users; therefore, evidence of less earnings management indicates more transparent financial information. Identification of IFRS Firms We identify firms accounting standards using the Worldscope database. We collect this data for the years 1998-2011, and classify a firm as using IFRS if, and only if, it used IFRS in the 6 Daske et al. (2013) use the change in this measure from the period before IFRS adoption to the period after adoption as their independent variable. We use the level in the prior three years because: (1) we use a levels design in our primary analysis, and (2) using the level in the prior three years allows our proxy to reflect only information observable by investors at the time of the investment decision. 7 We repeat all of our analysis using the raw data after separately truncating, or winsorizing, the bottom 1% of the data and obtain qualitatively similar results. 14

three prior years, t-3, t-2, and t-1. The three-year requirement increases the reliability of Transparency since it reduces the possibility that the magnitude of accruals relative to cash flows for a firm we classify as an IFRS firm is impacted by a change in accounting standards. 8 The primary drawback of this requirement is that it results in us classifying firms that use IFRS as non-ifrs firms during the first two years of IFRS adoption. This design choice primarily acts to limit the power of our tests. If the market can determine which firms will faithfully apply IFRS as of the date of adoption, then those firms should experience significantly higher levels of foreign ownership in the first three years of adoption, and we would be less likely to find a result when classifying those years as non-ifrs years. The primary strength of this research design choice is that it allows us to base our measure of transparency on information that is not impacted by a change in accounting standards, and that is available to foreign investors before they make the investment decision. Therefore, we do not have to rely on rational expectations, or other forward looking decision theories to explain investor behavior. 9 Methodology We first test our hypothesis using a research design that estimates the effect of IFRS use, the level of reporting incentives, and their interaction, on the level of foreign ownership, as follows: 8 Daske et al. (2013) document several inconsistencies in the Worldscope database which bring into question the accuracy of Worldscope s accounting standards designations. The requirement that a firm use IFRS for three years prior to being classified as an IFRS firm serves to reduce the likelihood of classification error resulting from the documented inconsistencies in the Worldscope database However, to further address this concern, we reran our analysis using accounting standard designations from the Osiris database and found statistically consistent results. 9 As a robustness test, we reran our primary analyses classifying firms as IFRS users in the first year they are noted as IFRS users in the Worldscope database. Consistent with our claim that our classification method weakens our tests, the results of these analyses were quantitatively stronger with regard to our primary variable of interest. However, we retain our classification method for our analysis to ensure that our Transparency variable is not influenced by a firm changing GAAP during its measurement period. We also reran our analysis excluding firms in their first three years of IFRS use from the sample and found statistically and economically consistent results. 15

Foreign Ownership i = β 1 IFRS i X Transparency i + β 2 Transparency i + β 3 IFRS i + (1) Σ βj Controls t + ε t, where: Foreign Ownership Transparency IFRS Controls equals the percentage of the firms shares outstanding owned by foreign shareholders as of year t, and equals the quartile rank (1 4) for firm i in year t of the absolute value of accruals divided by the absolute value of cash flows multiplied by negative 1, averaged over the years t-3, t-2, and t-1, and equals 1 if the firm used IFRS in the years t-3, t-2, and t-1, and zero otherwise, and are variables found in prior literature to be significantly related to foreign ownership. We include controls for firm visibility, performance, credibility, and risk/growth characteristics which prior literature (DeFond et al., 2011, Yu, 2010, among others) has found to be related to cross-border investment. We also include country-level controls measuring the size of the economy and the development of the financial markets in the firm s home country. The control variables are defined in Appendix A. We perform the analysis with and without firm fixed effects. If our first hypothesis is correct, and the use of IFRS is associated with a higher level of foreign ownership for firms with stronger incentives for financial transparency, then the coefficient on the interaction variable should be positive. A significant coefficient on the IFRS variable would indicate that either the adoption of IFRS is associated with foreign ownership regardless of reporting incentives, or that our proxy for reporting transparency fails to completely capture the underlying construct. Further, a significant coefficient on Transparency would indicate that reporting quality is associated with the level of foreign ownership regardless of whether a firm uses IFRS or not. 16

Next, we modify our analysis to investigate the impact of IFRS adoption and firm transparency on changes in foreign ownership around IFRS adoption. We test for the effect around IFRS adoption using the following regression: ΔForeign Ownership i = β 1 Adopt i X Transparency i + β 2 Transparency i + β 3 Adopt i + (2) Σ βj ΔControls t + ε t, Where: ΔForeign Ownership equals the change in the percentage of the firms shares outstanding owned by foreign shareholders from year t-1 to year t+1 10, and Transparency equals the quartile rank (1 4) for firm i in year t of the absolute value of accruals divided by the absolute value of cash flows multiplied by negative 1, averaged over the years t-3, t-2, and t-1, and Adopt ΔControls equals 1 if the firm adopted IFRS in year t, and zero otherwise, and are the changes in variables found in prior literature to be significantly related to foreign ownership from year t-1 to year t+1. We use a two year change in foreign ownership to allow time for financial statements prepared using IFRS to be published prior to the investment decision. In this analysis, Transparency is based on accounting numbers prepared using non-ifrs accounting standards and therefore its effectiveness as a proxy for reporting incentives will be dependent on the ability of the firm to signal its commitment to accounting quality based on its application of non-ifrs accounting standards. As in our primary analysis, a positive and significant coefficient on the interaction term provides support for our first hypothesis. The control variables are the same as those included in our foreign ownership levels tests. Sample Description Our sample consists of all firms for which we are able to collect ownership data, for which we are able to identify the accounting standards they used in the three prior years, and for which we are able to obtain data necessary for the firm-level control variables. We collect firm- 10 For example, if a firm adopts IFRS in 2004, we measure the change in foreign ownership as: foreign ownership in 2005 minus foreign ownership in 2003. 17

level data for the years 2001 through 2011 from the OSIRIS and WorldScope databases, and country level data from the World Bank s World Development Indicators database. The intersection of these datasets provides us with 54,552 firm-year observations from 72 countries. In table 1, we present the number of firm-years from each country, and report the extent of IFRS usage and foreign ownership within each country. 11 Both the use of IFRS and the level of foreign ownership vary substantially across countries. The largest number of firm-years in any one country using IFRS is 2,774, in Australia, which is 61.20% of the Australian firm-years in the sample. Nine countries have 100% of the firms using IFRS, though combined these countries represent less than 0.2% of our sample. South Africa has 73.32% of the firm-years using IFRS which is the highest among countries which make up at least 1% of the sample. The largest number of firm-years with foreign ownership is Japan, with 6,820, 77.47%. Similar to the use of IFRS, a number of countries with very few observations achieve 100% of firm-years with some level of foreign ownership. Further, at least 70% of the firms in many of the countries which are well represented in our sample have some level of foreign ownership. [Insert Table 1 about Here] We present the mean and median, as well as the standard deviation and 25 th and 75 th percentiles, of foreign ownership, Transparency, and the control variables for the full sample in Panel A, of Table 2. The unconditional means indicate that IFRS firm-years represent approximately 27% of the sample. For the full sample the mean of foreign ownership is 8.14%, while the median is only 1.35% indicating that home bias is likely present in our sample. In Panel B, of Table 2 we present the mean and standard deviations for foreign ownership, Transparency, and the control variables separately for IFRS and non-ifrs firm- 11 We exclude the United States from our sample because firms in the United States are not allowed to use IFRS, and the large number of observations that would be available from the United States would likely have a significant impact on the efficacy of the non-ifrs control group. 18

years. Consistent with our argument that accounting quality is a function of both standards and incentives, we find no significant difference in Transparency between the IFRS and non-ifrs firms. The univariate tests do indicate a significantly higher level of foreign ownership for IFRS firms relative to non-ifrs firms. However, we also find significant differences for the majority of the control variables indicating that inferences should be reserved for the multivariate tests. [Insert Table 2 about Here] Table 3 presents Pearson and Spearman correlations among the variables used to test our hypothesis. The strongest correlation between any two variables is 0.437 which is observed for the Pearson correlation between the natural log of the firm s market capitalization and inclusion in a market index. The vast majority of the remaining correlations are below 0.2, indicating that multicollinearity is unlikely to have a significant effect on our results. The variables having the strongest correlations (both Pearson and Spearman) with Foreign Ownership are IFRS usage, inclusion in a market index, having a big-n auditor, and size (measured as the natural log of the firm s market capitalization). All of these variables are positively correlated with Foreign Ownership. [Insert Table 3 about Here] IV. RESULTS We first investigate the impact of IFRS usage and reporting incentives on the level of foreign ownership by estimating equation (1), with, and without firm fixed effects (OLS and fixed effects regressions, respectively). For comparative purposes we also present the results excluding the interaction of IFRS and Transparency. The results are presented in table 4. Tests of significance for all specifications are performed using robust standard errors clustered by firm. 19

Supporting our hypothesis, we find a significant and positive result for the interaction between IFRS and Transparency for both the OLS and fixed effects specifications. Further, the economic impact is significant in both specifications. For example, the coefficient on the interaction in the fixed effects specification of 0.7438 translates to a difference of 2.23% in the level of foreign ownership between an IFRS firm in the highest quartile of Transparency and an IFRS firm in the lowest quartile of Transparency. This is a substantial difference given that the mean level of foreign ownership for all firms in our sample is 8.14%, and the median is only 1.35%. [Insert Table 4 about Here] We find little evidence of a direct effect on the level of foreign ownership for either IFRS or Transparency. The coefficient on the IFRS indicator variable is significant in all specifications, but its sign is unstable, and the coefficient on Transparency is significant only when the interaction term is excluded from the model. The coefficients on both IFRS and Transparency are positive and significant in both the fixed effects and OLS models excluding the interaction term. This indicates the importance of including the interaction of these two variables in tests regarding the efficacy of IFRS adoption to draw correct inferences. In summary, these results indicate that IFRS firms with higher Transparency exhibit a significantly higher level of foreign ownership than IFRS firms that have lower levels of Transparency. In addition, we find no evidence that higher Transparency is associated with higher levels of foreign ownership for non-ifrs firms, and weak evidence (in the full model including fixed effects) that the usage of IFRS by firms with weak reporting incentives actually leads to a lower level of foreign ownership. Based on these results we infer that both incentives 20

and standards matter. A higher level of foreign ownership is associated with the presence of both IFRS usage and the quality of its application. We next investigate the relationship between reporting incentives and changes in foreign ownership around the adoption of IFRS. We use a change specification looking at the change in ownership from the year before adoption to the year after adoption (i.e., if a firm adopted IFRS in 2004 our dependent variable is the change in foreign ownership from 2003 to 2005). We use the two year change because it increases the probability that investors will be able to observe a set of financial statements prepared under IFRS before making the investment decision. This test is performed using equation (2). The results are presented in Table 5. [Insert Table 5 about Here] Providing further support for our hypothesis, we find that firms with higher Transparency experience greater increases in foreign ownership around IFRS adoption than firms with lower levels of Transparency. The economic impact of the change is similar to that found in the levels design. We find no evidence of a relationship between IFRS adoption and foreign ownership for firms with lower Transparency. We do find evidence that firms with higher Transparency experience increases in foreign ownership even when they do not adopt IFRS. Specifically, the coefficient on Transparency is positive and significant at the 5% level. However, it has substantially less economic significance than the interaction of Adopt and Transparency. V. ADDITIONAL ANALYSIS AND ROBUSTNESS TESTS Self-selection and the adoption of IFRS Since the decision to adopt IFRS is voluntary for some firms in our sample there is a potential for self-selection bias. Firms committed to a higher level of reporting transparency may be more likely to voluntarily adopt IFRS. Any change in foreign ownership observed around 21

adoption may therefore be due to this commitment to higher disclosure without being associated with IFRS adoption at all. Therefore, we also test the impact of reporting incentives and IFRS adoption on foreign ownership using a matched pairs design. We first run a logit regression on the entire sample with the dependent variable being an indicator equal to 1 if the firm adopted IFRS in that year and using all control variables included in our primary analysis as predictors. We then use the coefficients from that regression to calculate propensity scores for all firm-years in the sample. Next, we match each IFRS adopter with its nearest propensity neighbor in the same year that is a non-ifrs adopter. Finally, we compare the mean 2-year change in foreign ownership for the adopting firms, to the mean 2-year change in foreign ownership for their non-adopting counterparts. This calculation is made for all adopters and counterparts, and separately for the low reporting transparency and high transparency firms (split at the median). A significantly larger increase in foreign ownership for the high reporting incentive firms relative to their counterparts than for the low reporting incentive firms relative to their counterparts (i.e., a significant difference-in-difference) would support our hypothesis. The results of this analysis are shown in Table 6. The results indicate that IFRS adopters with low Transparency do not experience significantly different changes in foreign ownership than their matched counterparts. In contrast, IFRS adopters with high Transparency experience an increase in ownership significantly greater than their non-adopting matched firms. Finally, the difference-in-difference indicates that the increase for high Transparency firms is significantly greater, at the 10% level, than the difference found for low Transparency firms. [Insert Table 6 about Here] 22

Country-level investor protection, IFRS adoption and change in foreign ownership We next investigate the interaction between our proxy for firm-level reporting transparency and country-level reporting incentives. We divide the sample based on the level of investor protection in the firm s home country, measured using the Legal Structure and Security of Property Rights Index from Economic Freedom of the World (Gwartney and Lawson, 2006). 12 We then perform the regressions separately in countries with strong (weak) investor protection with strong investor protection being defined as firm-years in the top quartile, and weak investor protection being defined as firm-years in the bottom three quartiles. This analysis provides insight into whether firm-level and country-level incentives act as complements or substitutes in the eyes of foreign investors. 13 If we find a significant result for the interaction of Transparency and IFRS for the weak (strong) investor protection group, this would indicate that investors view the two sets of incentives as substitutes (compliments). If we find significant results for both, or neither, subsets, then we would be unable to draw an inference in this regard. The results of these additional analyses are presented in table 7. [Insert Table 7 about Here] The results for these two sets of firms are significantly different. For the firms from weak investor protection countries we find a positive, and significant, association between the interaction of IFRS and Transparency. Further, the point estimate for this sample is more than twice the estimate found in the combined analysis indicating that the economic significance is substantially greater for firms from weaker investor protection regimes. In contrast, for firms in countries with strong investor protection we find a positive association between the level of 12 The Economic Freedom of the World index and its individual components have been used by La Porta, Lopez-de Silanes, and Shleifer (2002) Heckelman and Knack (2008), and Fung (2009), among others. 13 The separate regressions allow for a non-linear (interactive) relationship to exist between firm-specific and country-specific reporting incentives. 23

foreign ownership and the interaction between IFRS and Transparency only when firm fixed effects are excluded from the model. Finally, we find that the positive relationship between foreign ownership and the interaction of IFRS and Transparency in low investor protection countries is significantly greater than the relationship found in high investor protection countries. These results indicate that country-level and firm-level incentives act as substitutes in the eyes of non-domestic investors. We find little evidence of a direct relationship between IFRS usage and foreign ownership. The coefficient on the IFRS indicator variable is positive and significant in only one of our specifications. The coefficient on Transparency is unstable across the regression being positive and significant in one specification, and negative and significant in two specifications. Therefore, while the evidence seems to lean toward a negative relationship between foreign ownership and Transparency, particularly in the low investor protection countries, inferences should be drawn with caution. The results for the control variables are generally similar in high and low investor protection countries. Though the evidence does indicate that inclusion in a market index and the presence of a controlling shareholder are particularly important in the low investor protection environment and not in the high investor protection environment. The latter possibly resulting from foreign owners being more likely to acquire a controlling interest themselves when they have less confidence in their ability to rely on institutional mechanisms to protect their investment. IFRS adoption and change in foreign ownership across different types of investors Next, we divide foreign investors by investor type and perform our analysis separately for foreign institutional investors and for all other foreign investors. Throughout the paper we argue 24

that IFRS firms with a higher Transparency will experience higher levels of foreign ownership as a result of an improved information environment. Therefore, we expect investors who use accounting information as a primary input in their investment decisions to drive this result. We proxy for these sophisticated investors using institutional investors and expect that we will find the strongest result for that investor group. Therefore, we rerun our primary test after measuring foreign ownership separately for institutional investors and the group of other investors. The results of this analysis are shown in Table 8. [Insert Table 8 about Here] We find a positive and significant relationship between foreign ownership and the interaction of IFRS usage and Transparency only for institutional investors. To the extent that institutional investors are more likely to rely on, and better able to interpret accounting information than other investor groups, this supports our argument that Transparency is a proxy for the information environment of the firm. Consistent with our main analysis (including fixed effects), the coefficient on IFRS is negative and significant in the regression using foreign institutional investment as the dependent variable. The coefficient on Transparency is insignificant for both investor groups indicating that non-ifrs firms with greater transparency do not experience significantly higher levels of foreign ownership in either investor group. In additional untabulated analyses, we split the other investors group into three subgroups: foreign industrial firms, foreign individuals and families, and foreign governments and sovereign wealth funds. The coefficient on the interaction of IFRS and Transparency is positive and significant only for the specification using investment by foreign governments and sovereign wealth funds as the dependent variable. However, the coefficient is an order of magnitude smaller for this group than for the foreign institutional investor group indicating that a 25