IFRS and the Complexity Hurdle

IFRS and the Complexity Hurdle Nicolas Schrödl 1 Christian Klein* Chair of Accounting and Finance, University of Hohenheim, 70593 Stuttgart, Germany Abstract Regulators expect that the introduction of International Financial Reporting Standards (IFRS) improves firms transparency and consequently benefits the capital market. According to the literature and the enforcement panels reports, the standards introduction is associated with various operational hurdles, which are due to difficulties in implementing and understanding the IFRS due to their complexity. Consequently, the following question arises: Does the standards complexity partially impede the expected benefits? We investigate this issue by examining the changes in market liquidity. Higher transparency and lower information risk should increase market liquidity. We assume that the more the regulations deviate from the former Generally Accepted Accounting Principles (GAAP), the greater the degree of complexity that countries will experience. The analysis of a worldwide sample finds that, owing to the introduction of IFRS, firms from countries with little deviation have significantly higher market liquidity. Keywords: IFRS, complexity, transparency, market liquidity, information risk * E-mail address: cklein@uni-hohenheim.de. Phone: +49 711 45922657. Fax: +49 711 45922721. 1

IFRS and the Complexity Hurdle Abstract Regulators expect that the introduction of International Financial Reporting Standards (IFRS) improves firms transparency and consequently benefits the capital market. According to the literature and the enforcement panels reports, the standards introduction is associated with various operational hurdles, which are due to difficulties in implementing and understanding the IFRS due to their complexity. Consequently, the following question arises: Does the standards complexity partially impede the expected benefits? We investigate this issue by examining the changes in market liquidity. Higher transparency and lower information risk should increase market liquidity. We assume that the more the regulations deviate from the former Generally Accepted Accounting Principles (GAAP), the greater the degree of complexity that countries will experience. The analysis of a worldwide sample finds that, owing to the introduction of IFRS, firms from countries with little deviation have significantly higher market liquidity. Keywords: IFRS, complexity, transparency, market liquidity, information risk 2

1. Introduction Recently, the introduction of IFRS 2 has become mandatory in different countries around the world in order to provide financial statement users with information that is useful for making economic decisions (e.g., IASB Framework; EC Regulation No. 1606/2002). Consequently, IFRS should lower information risks, optimize capital allocation, and increase market liquidity. However, a closer look at the related literature casts doubt on the expected capital market benefits. Early surveys and reports from official institutions document the risks of and dissatisfaction with the increasing complexity within international reporting standards. Overall, the studies underline the difficulties firms and investors face when implementing the standards and analyzing the financial statements, respectively (e.g., Fearnley and Hines, 2007; Palmrose, 2009; CESR, 2007; FREP, 2009). While there is extensive literature on complexity concerns, we are not aware of any studies that provide evidence on IFRS complexity observing capital market reactions. We therefore concentrate on different variables for market liquidity and examine whether complexity within the standards represents a hurdle for the proper IFRS adoption, which consequently reduces the expected capital market benefits. The variables are Price Impact, Zero Returns, and Bid-Ask Spreads. We divided our sample into voluntary and mandatory IFRS adopters to control for selection effects. Voluntary adopters applied IFRS before the official mandatory adoption date, which was 1 January 2005 for our treatment firms. The analysis faces various challenges. Since the IFRS mandate usually forces all firms from one country to adopt IFRS at the same point in time, there are no firms with which 2 The International Financial Reporting Standards (IFRS), formerly called the International Accounting Standards (IAS), are issued by the International Accounting Standards Board (IASB). The IAS were issued by the IASB s predecessor: the International Accounting Standards Committee (IASC). As the IASB has adopted all standards issued by IASC, we will refer to these standards as the IFRS. 3

to compare the findings. Consequently, the analysis demands a precise benchmark definition. According to the finance literature, countries institutional features influence the IFRS introduction s effects (e.g., Daske et. al., 2008). We therefore chose treatment and benchmark countries with similar enforcement regimes and similar reporting incentives to be transparent. The selected benchmark countries did not mandate IFRS introduction in 2005. We kept the worldwide sample constant between 2003 and 2008. This sample selection method assures the comparability between the treatment and the benchmark countries, guaranteeing for similar institutional characteristics. Another challenge is measuring complexity. You and Zhang (2009) examine the US GAAP s complexity and use a word-count as a proxy for the complexity level. Li (2008) and Miller (2008) also concentrate on the length of 10-K filings. Li (2008) additionally uses the Fog Index. Filzen and Peterson (2010) observe the disclosure length. Extant IFRS studies perform analyses based on surveys (e.g., Fearnley and Hines, 2007). We decided to use a different approach that is more market orientated. We therefore separated our treatment sample into two groups. One group consists of countries with strong differences between the local GAAP and IFRS and the other one of countries with weak differences. Compared to the weak differences group, we believe that firms (preparers) from countries with strong differences are subject to greater expenses to understand the standards and convert their accounting processes. As a result, they experience a higher susceptibility to commit mistakes. Investors (users) from countries with strong differences have to spend more in order to absorb the additional information and to understand the changed accounting numbers as well as (possibly) complexly transferred management assessments. They also face a 4

higher risk of receiving financial statements with IFRS application errors. As a result, they experience a high level of uncertainty and information risk. 3 We conclude that countries with strong differences between the former local GAAP and IFRS perceive the complexity in adopting IFRS to a higher degree than countries with weaker differences. Consequently, we suppose that stronger capital market benefits through IFRS introduction will materialize in countries with weaker differences between the former local GAAP and IFRS. Consistent with our expectation, we show that the complexity in the IFRS indeed represents a hurdle and consequently reduces the expected capital market benefits. A difference-in-differences as well as various regression analyses all document a better development in market liquidity for the group with weak accounting discrepancies. We find that the percentaged increase in market liquidity from the year before the IFRS adoption to the adoption year was generally higher for the weak differences group. For instance, mandatory adopters Bid-Ask Spreads indicate a percentaged increase in market liquidity that is 14.92% higher than the strong differences group s increase. The improvement in Zero Returns is 19.19% stronger for voluntary and 9.56% stronger for mandatory adopters. The results from our regression analyses are also remarkable. The coefficients indicate that the voluntary adopters Bid-Ask Spreads increased by 500 basis points between 2003 and 2008 for the group with strong accounting discrepancies while the group with weak accounting discrepancies showed a decrease of 76 basis points, compared to a benchmark that is unaffected by the IFRS adoption. This equates to an augmentation of 3 See Section 2. 5

169.9% and a reduction of 25.9% for the group with strong and weak accounting discrepancies, respectively. The coefficients for Price Impact and Zero Returns underlined the better development in market liquidity for the group with weak accounting discrepancies during this period. Consistent with these results, both treatment samples market liquidity coefficients, compared from the time before the IFRS adoption to the coefficients after the adoption, underline the weak differences group s advantages over the entire period. This evidence stems from all our liquidity variables and remains so after various sensitivity analyses and robustness checks. The results are, in general, statistically significant at the 1% level. This study contributes to the literature examining international financial reporting standards complexity (e.g., Fearnley and Hines, 2007; You and Zhang, 2009) in that it is the first to examine complexity during the certain period using a worldwide sample of IFRS adopter firms. Moreover, it extends previous IFRS complexity studies since we base our investigation on capital market data. Furthermore, we open a new methodology to examine the topic by observing different levels of perceived complexity. The results also contribute to studies examining the introduction of IFRS (e.g., Daske et al., 2008; Holthausen, 2009; Armstrong et al., 2010) as it provides additional information on prerequisites for positive capital market effects. The sample size enhances the generalizability of our results that the complexity in the IFRS represents a hurdle and consequently reduces the capital market benefits expected from the IFRS introduction. We believe that this study is the first to provide comprehensive evidence of the IFRS complexity s capital market effects. Our findings are therefore of special interest to the IASB, policy-makers, and countries planning to implement IFRS. 6

In Section 2 of this paper, we review the literature. We develop the hypotheses in Section 3 and present the data selection in Section 4. The research design is described in Section 5, while the empirical results are presented in Section 6. The conclusion with suggestions for future research follows in Section 7. 2. Literature Review Our review summarizes essential findings from studies on IFRS introduction. Subsequently, we present exemplary studies to define the term complexity as examined in the literature. Literature on the IFRS Introduction Various studies prove the context of reporting transparency and capital market benefits (e.g., Welker, 1995; Botosan, 1997; La Porta et al., 1998; Healy et al., 1999; Lang and Lundholm, 2000; Botosan and Plumlee, 2002). 4 Hence, if IFRS introduction achieves higher transparency, positive capital market effects will materialize. Daske et al. (2008) present an overview of the capital market effects through IFRS introduction. These - and other - authors identified specific company and country characteristics as prerequisites for positive capital market effects. Armstrong et al. (2010), for example, supplement this overview. They examine equity return reactions to 16 events associated with IFRS introduction in Europe. They noticed that firms with lower-quality pre-adoption information and higher pre-adoption information asymmetry showed positive effects. Conversely, firms from code law countries (associated with weak enforcement regimes) showed negative reactions. Holthausen (2009) argues that benefits will not fully realize unless the underlying institutional and economic factors also develop and become more similar. Furthermore, 4 For a comprehensive overview of the transparency literature see Barth and Schipper (2008). 7

other studies have found evidence that the global adoption of a single set of accounting standards does not have the potential to increase accounting information s comparability across countries that differ economically, politically, and culturally (e.g., Ball et al., 2003; Lang et al., 2006). Burgstahler et al. (2006) as well as Christensen et al. (2007) underline incentives importance, while Ball and Shivakumar (2005) emphasize the key role of regulation, and Leuz et al. (2003), Christensen et al. (2011), as well as Burgstahler et al. (2006) underline the importance of enforcement. In short, the literature demonstrated that the main characteristics for positive capital market effects are an institutional environment that provides firms with strong transparency incentives and a strict enforcement regime (e.g., Daske et al., 2008). Literature on Complexity The IFRS introduction s objective relies on the principles understandability, relevance, reliability, and comparability (IASB Framework). Simultaneously, these principles document the requirement to transfer the standards contents in a low-complex and informative manner. Nevertheless, the framework targets financial statement users with reasonable knowledge and advises that relevant information should not be excluded solely because it may be too complex or difficult for some users to understand" (IASB, 2006). Authors have questioned the interpretation of reasonable knowledge and the difficulty of making complex underlying economics understandable (Barth and Schipper, 2008). Early surveys and reports from official institutions document the difficulties faced when introducing IFRS. On the one hand, these difficulties arise for companies (preparers) when implementing IFRS. According to Larson and Street s (2004) survey, companies perceive complexity 8

as a key challenge when implementing IFRS. Jermakowicz and Gornik-Tomaszewski s (2006) survey confirms this result and identifies the lack of implementation guidance and uniform interpretation as challenges. Fearnley and Hines (2007) also found evidence from UK surveys that the IFRS are overly complex. Dunne et al. (2008) have conducted surveys, which largely lead to the same conclusion. Palmrose (2009) claims widespread discontent with complexity and deep disenchantment with the current state of accounting affairs. The Financial Reporting Council (FRC) warns that IFRS are becoming increasingly complex (FRC, 2009). Official institutions like the Financial Reporting Enforcement Panel (FREP), the Financial Reporting Review Panel (FRRP), and the Committee of European Securities Regulators (CESR) detected various areas where companies had not complied with the requirements of the relevant standard or legislation (e.g., FREP, 2009; FRRP, 2009; CESR, 2007). In their annual report, the FREP (2009) still presented an error rate of 27%. On the other hand, these difficulties arise for investors (users) when analyzing the financial statements. Various studies generally confirm that analysts fail to access or ignore certain complex information which could result in an incomplete use of available information (e.g., Hirst and Hopkins, 1998; McEwen and Hunton, 1999; Hirshleifer, 2001; Bergstresser et al., 2006; Picconi, 2006).5 Brav and Heaton (2002) reveal that investors uncertainty about information structures can lead to a pattern of underreaction that varies with the level of uncertainty. Hong and Stein (1999) also find evidence of a relation between underreaction and complexity. You and Zhang (2009) study the 5 We also mention studies on the US GAAP, as they provide insights on the general topic of analysts reaction to complexity. Moreover, it is worth it to consider concurring investigations and developments in view of the IASB s and FASB s conjoint improvement projects and their path of convergence to international accounting standards. 9

immediate and delayed market response to SEC EDGAR 10-K filings. They use a word-count as a proxy for the complexity level and find that investors underreaction is stronger for firms with more complex 10-K filings. Li (2008) and Miller (2008) confirm that longer 10-K filings increase the difficulty to understand the financial information. Daske (2005) finds lower accuracy and higher dispersion among analysts forecasts for German firms which adopted IAS between 1993 and 2002. Filzen and Peterson (2010) find out that complex accounting is exploited by managers to meet or beat analysts expectations when expectations are close to actual earnings. In conclusion, we have noticed a general concern regarding a high level of complexity within the reporting standards. Our understanding of accounting complexity, in general, is consistent with the SEC s (2008) view. Accordingly, it affects both preparers and investors and can impede effective communication between a company and its stakeholders, create inefficiencies in the marketplace and suboptimal allocation of capital. 3. Hypothesis Development Regulators expect that the introduction of IFRS benefits the capital market as the standards are supposed to reveal more information and consequently improve firms transparency which will result in more efficient economic decision-making (e.g., EFRAG, 2011). Nevertheless, the high information requirements also increase the complexity within the standards. The literature review demonstrates that both 10

companies fail to implement IFRS properly 6 due to the standards complexity and analysts fail to absorb complex information. We derive that complexity increases uncertainty for financial statement users in several ways: Financial statement preparers mistakes in applying IFRS can transfer information that impairs economic decisions. Furthermore, complexity creates information asymmetry between managers and investors, as managers can communicate the economic substance of a transaction in a way that is difficult to understand for financial statement users (see also Filzen and Peterson, 2010; SEC, 2008). Overall, it is a sophisticated task to absorb the transferred information. All issues increase information risks. Thus, our first hypothesis is: H1: The complexity within the standards reduces the expected capital market benefits from IFRS since complexity increases uncertainty for investors. 7 We observe two groups to examine the capital market effects. Compared to countries with weak differences, we believe that firms (preparers) from countries with strong differences between the former local GAAP and IFRS are subject to greater expenses to understand the standards, convert their accounting processes, and come to an agreement with auditors on sophisticated principles-based IFRS interpretations. As a result, they experience a higher susceptibility to commit mistakes. Investors (users) from countries with strong differences have higher expenses to absorb the additional information provided in the disclosures and to understand the accounting numbers and (possibly) complexly transferred management assessments. They also face a higher risk of 6 The enforcement panels (as presented in Section 2) annual reports note that IFRS have not been implemented properly in almost all countries. Unfortunately, the enforcement panels measures and the presentations of the results differ between the countries. Consequently, we are not able to conclude how well the standards are implemented in each country. 7 There is, of course, a high extent of uncertainty in the year of the adoption. We cannot separate between the influences of first year s (natural) adoption uncertainty and general complexity-uncertainty. Nevertheless, we keep these influences in mind when we interpret the results in Section 6. 11

receiving financial statements with IFRS application errors. As a result, they experience a high level of uncertainty and information risk. We conclude that countries with strong differences between the former local GAAP and IFRS perceive the complexity in adopting IFRS to a higher extent than countries with weaker differences. Consequently, our second hypothesis is: H2: Stronger capital market benefits through IFRS introduction will materialize in countries with weaker differences between the former local GAAP and IFRS. 8 To measure the two clusters capital market effects, we focus on market liquidity. Market liquidity comprehensively represents change due to IFRS introduction, since higher transparency and lower information risk should optimize capital allocation and increase trading and liquidity. We also found proof in the international finance literature that market liquidity is an appropriate measure to capture both clusters reactions (e.g., Daske et al., 2008; Hail and Leuz, 2007). 4. Data Selection We define two samples, a treatment and a benchmark sample, in order to observe the capital market effects. As the countries institutional features influence IFRS introduction s effects, we need treatment and benchmark sample countries whose features are similar and can be kept constant over the sample period. Primarily, these features are legal enforcement and transparency incentives for companies. 9 As a result, we define the following criteria: 8 We empirically test the second hypothesis and make a conclusion from its results on the first hypothesis. 9 See Section 2. 12

Criteria for the Sample Definition (1) Enforcement: we chose the Rule of Law (Kaufmann et al., 2009) as criterion for our sample definition. 10 Higher values represent countries with stricter enforcement regimes. We divided our possible sample countries into two groups: countries with relatively weak and countries with relatively strong enforcement regimes. The cut-off point is the median score. (2) Reporting incentives to be transparent: we chose Bellver and Kaufmann s (2005) institutional transparency index and the average transparency scores for the years 2003 to 2008 (http://www.transparency.org/policy_research/ surveys_indices/cpi/2009/cpi_2009_table). We also chose the second variable to assure the index reliability. 11 Higher values indicate stronger reporting incentives for firms. We divided possible sample countries into two groups: countries with strong incentives and countries with weak incentives. The cut-off point is the median score. 12 (3) IFRS convergence: countries converge towards IFRS differently, for example, by adapting the local GAAP to IFRS. We therefore ignore all countries with an official convergence strategy, with local regulators officially announcing a gradual move towards IFRS over a predetermined time-frame. Daske et al. (2008) provide an overview of countries with an IFRS convergence strategy. 10 There are also other scores for accounting enforcement, but they did not have sufficient data available (e.g. Brown et al., 2009). 11 Bellver and Kaufmann s (2005) study was a draft; it was not published. Nevertheless, in the light of the authors reputations, we see no material risk concerning data reliability. The second variable simply serves to support Bellver and Kaufmann s (2005) index. All countries chosen for our sample according to the transparency index also belong to the strong incentives group according to the second variable. Both variables are closely related, as low corruption levels represent strong governance infrastructure regarding transparency incentives. 12 Other variables, such as Leuz et al. s (2003) earnings management score cannot be used, because data are not available for all the sample countries. 13

We divided all possible treatment and benchmark sample countries into two groups according to the enforcement and transparency criteria. The cut-off point is the median value of all countries variables which is 0.82, 1.05, and 5.03 for Rule of Law (Enforcement), Institutional Transparency, and CPI Index (both Transparency Incentives), respectively. We excluded countries with values below the medians and used the groups with strong institutional structures. Ultimately, the IFRS adoption countries whose enforcement regimes are relatively strict and whose institutional environments provide strong incentives for company transparency are: Denmark, Finland, Germany, Ireland, Netherlands, Norway, Portugal, Sweden, Switzerland, and the UK. The benchmark countries with strong institutional features are Canada, Chile, Israel, Japan, New Zealand, and the U.S. This sample selection method ensures the comparability between treatment countries and benchmark countries, guaranteeing similar institutional characteristics. We believe that this approach is more precise compared to a comprehensive global benchmark. 13 We chose listed firms from these treatment countries and benchmark countries, provided that their market capitalization was EUR 10 million or more. The market capitalization criterion is important because some countries did not have data for smaller firms. We selected a randomly drawn sample of 350 firms from each country if more data were available. This approach disallows strong effects from any particular country that might be due to country-specific regulatory changes. The examined sample consists of 49% treatment and 51% benchmark observations. Furthermore, we allocated firms from our treatment countries (mandatory adoption) that had not yet adopted IFRS to the benchmark sample. 14 Voluntary IFRS adopters from the benchmark countries 13 Especially owing to Daske et al. s (2008) results, we could define the specific benchmark group. Nevertheless, we also exceeded the benchmark sample as part of the sensitivity analysis in Section 4.2. 14 Including company year observations from the treatment countries to the benchmark sample brings the risk that observations might be concentrated in specific countries during certain periods. We omit these observations when conducting the sensitivity analysis (Section 4.2.). 14

were dropped as well as country years if IFRS was mandated at a later point in time. 15 We illustrate the sample selection process and the descriptive statistics for IFRS adopter countries and benchmark countries in Table 1. 16 Since we have selected treatment (IFRS adopters) and benchmark (non-ifrs adopters) countries with similar institutional features, we can focus on one variable within the treatment countries: the complexity they experienced with the IFRS application. As concluded from Section 3, countries with stronger differences between IFRS and the local GAAP experience complexity to a higher degree when applying IFRS. To measure the differences between the local GAAP and the IFRS, we use Bae et al. s (2008) summary score, which is based on 21 key accounting dimensions. Higher positive scores represent larger differences with the local GAAP. 17 In addition, we utilize Daske et al. s (2008) modified score. They orthogonalized Bae et al. s (2008) score for accounting discrepancies with respect to fundamental country characteristics such as countries legal origins and their gross domestic product (GDP) per capita as a response to the concern that the variables are highly correlated and that they are outcomes of more fundamental qualities of countries institutional frameworks. Applying their score on our sample led to the same results. Furthermore, we tested Bae et al. s (2008) variable using the mean of Ding et al. s (2009) absence and divergence results per country. We divided the treatment group at 15 Israel partially mandated IFRS in 2008 and New Zealand in 2007. See http://www.iasplus.com. 16 Alternatively, we tried to form a benchmark based on company level matching and assembled a propensity score. However, there were insufficient data for a robust score. We therefore omitted this approach. 17 An alternative method would be to choose only countries with an official convergence strategy as one can assume that these countries have low differences between IFRS and local GAAP. However, this method does not account for the GAAP differences at the beginning of the convergence strategy. For instance, Daske et al. (2008) identify a country with an official convergence strategy and strong differences between IFRS and the local GAAP (see Table 6). 15

their median into groups with weak and strong differences and, again, obtained the same results. Bae et al. s (2008) score is available for the period before the IFRS adoption and not for each year of our sample period. Thus, we face the risk of a change in the GAAP differences not recorded in our investigation. However, through our choice of sample criteria, we have addressed this risk. Important changes in the GAAP differences score are not expected, as we have excluded all countries with an IFRS convergence strategy. Based on the treatment countries median (10.5), we form two clusters: 1) Strong differences (over the median): Germany, Denmark, Finland, Portugal, and Switzerland. 2) Weak differences (below the median): Ireland, The Netherlands, Norway, Sweden, and the UK. We use an ANOVA to test the significance of the groups differences. The results indicate that the group with strong differences and that of weak differences differ significantly (the p-value is always below 1.0%). <insert table 1 about here> One concern about the data is that the selected treatment countries represent all four legal origins (La Porta et al., 1998) and, hence, might contain cross-border heterogeneity. La Porta et al. (1998) generally classify a country based on the origin of the initial laws it adopted. Consequently, the observations cannot account for current revisions or changes in the laws. At the time La Porta et al. (1998) examined the legal rules 1993-1994, the European Community was already attempting to harmonize European laws (e.g., Andenas and Kenyon-Slade, 1993; Werlauff, 1993). Eight of our ten selected treatment countries have joined the European Union (former: European 16

Community) largely before 2003, the start of our investigation period. Therefore, we expect the selected countries to be legally harmonized in accordance with our examination s requirements. The two remaining treatment countries that have not joined the European Union are Switzerland and Norway. We therefore conduct different sensitivity analyses (as described in Section 6) and exclude these countries data from the investigations. 5. Research Design To measure the two clusters capital market effects, we focused on market liquidity. 18 Market liquidity represents change due to IFRS introduction, as trading is expected to increase due to lower information risks. We also found proof in the international finance literature that market liquidity is an appropriate measure to capture both clusters reactions (e.g., Daske et al.,2008; Hail and Leuz, 2007). We examine market liquidity according to the following variables: Zero Returns is the proportion of trading days with zero daily stock returns of all the potential trading days in a given year. Illiquidity or Price Impact is a variation of the Amihud (2002) illiquidity measure, i.e. the annual average of daily absolute stock returns divided by the trading volume. This measure provides the price impact of each EUR traded on the stock price. As verified by Amihud s (2002) study, the price impact or the return increases with illiquidity. 19 Bid-Ask Spreads are the annual average of daily quoted spreads measured at the end of each trading day by calculating the difference between the bid price and the asking price divided by the mid-point. 18 See Section 3. 19 See also, for example, Amihud and Mendelson (1986). 17

As our investigation starts in 2003 and covers the entire period until 2008. We start in 2003 to ensure that we have sufficient data before the IFRS adoption to compare the IFRS adopters variables with. We furthermore want to capture voluntary IFRS adopters to control for self-selection effects. When we intended to extend the sample period back to 2001 we experienced that most of the early IFRS adopters data stems from the strong differences group. Our sample covers four years of mandatory IFRS adoption and consequently truncates distorting adoption effects in the first year of the mandate. It ends in 2008 due to data availability at the time of our analysis. The measurement period is defined from month -5 to month +7 relative to the firm s fiscal year-end (e.g., Hail and Leuz, 2007). Consequently, we ensure that information from interim reports and annual reports are priced in our data. We obtained the financial, price, and trading volume data from Bloomberg. In the event that fiscal year-end or reporting standard data were not available, we compared the company information from Datastream and Reuters. To define the control variables, we followed the literature (e.g., Chordia et al., 2000; Leuz and Verrecchia, 2000) and controlled for firm size, share turnover, and return variability. 20 We logarithmized the control variables and lagged them by one year (e.g., Stoll, 1978; Glosten and Milgrom, 1985). Table 2 illustrates descriptive statistics on the variables. Another control variable is the market benchmark. It is computed as the dependent variable s annual mean from the benchmark sample (Daske et al., 2008). 21 We therefore controlled for unobserved time-invariant company characteristics. Furthermore, we included an indicator variable for every year, country, and industry to 20 The variables are explained in Table 2. 21 Daske et al. (2008) used US GAAP reporting, US Listing, and New Markets Listing as additional control variables. These variables had little empirical validity as they generally did not lead to statistically significant results. To concentrate on the main determinants, we omitted these control variables. 18

deal with industry, year, and country fixed effects. We followed Campbell (1996) categorizing the firms into the following industries: Petroleum, Finance/Real Estate, Consumer durables, Basic, Food/Tobacco, Construction, Capital Goods, Transportation, Utilities, Textiles/Trade, Service, and Leisure. We excluded values outside the 1% and 99% percentile, except for variables with natural lower and upper limits. Throughout the tests and analyses, we ensured that the data fulfilled the necessary statistical premises. 22 <insert table 2 about here> We conducted univariate and multiple regression analyses. In the univariate analysis, we calculated the dependent variables mean values in the preadoption year and in the IFRS adoption year, holding the sample constant over the two years. The examined groups are voluntary and mandatory IFRS adopters with weak and strong accounting discrepancies between IFRS and the local GAAP. We differentiated between mandatory and voluntary IFRS adopters to examine whether the capital market reactions were distorted by self-selection effects. It is possible that voluntary adopters capital market effects cannot be attributed to IFRS alone, as they might adopt the new standards in advance to signal superior company characteristics. In that case, the results were only attributable to IFRS after the first adoption years. However, the early adoption can also be part of a new commitment to transparency. These companies are not forced to adopt IFRS and are therefore probably more committed to overcome complexity and implement the standards properly. 22 We excluded the year control variables from our calculation for Price Impact in respect of the statistical premises. 19

Voluntary adopter companies first adopted IFRS before the adoption became mandatory in their country, which was in 2005 23 for our treatment sample. 24 Mandatory adopters applied IFRS for the first time at fiscal year-ends on or after 31 December 2005. We compared the means and examined two-sided t-tests to assess statistical significance. In our regression analysis, we calculated the ordinary least squares (OLS) coefficient estimates throughout the firm-years. Indicator variables separated our IFRS adopters into voluntary and mandatory adopters with weak and strong accounting discrepancies between IFRS and the local GAAP. By using indicator variables for IFRS adopters, we could also distinguish different periods to present the firms development over the years. We assume that the influence of the natural complexity of the accounting change (implementation costs) should diminish over the years. 25 Another indicator variable labeled companies from our treatment countries between 2003 and 2004 that did not adopt IFRS during this period. We therefore could compare the liquidity values before and after the IFRS introduction for the two groups having strong and weak accounting discrepancies between the local GAAP and IFRS. The variables are presented in the following regression model: 23 See Daske et al. (2008), Table 6. 24 Some firms belonged to an index that prescribed IFRS application prior to 2005. These firms comply with the voluntary adopters criteria as the mentioned stock segments presume superior firm characteristics and commitment to innovation and transparency. We therefore did not add these firms to the mandatory adopter groups. 25 See Section 3, hypothesis H1. 20

DepVar it VoluntarySD(2005 2006) (2005 2006) WD (2007 2008) Before IFRS AdoptionSD it VoluntaryWD (2003 2004) 8 (2007 2008) 0 it 1 it Controls j it Mandatory SD(2007 2008) 6 ijt it it it VoluntarySD(2007 2008) VoluntarySD(2003 2004) it VoluntaryWD (2005 2006) MandatoryWD (2005 2006) 11 4 9 2 it it MandatorySD Before IFRS AdoptionWD 7 12 5 it Voluntary 10 Mandatory WD it it 3 DepVar it represents the firms (i) dependent variables for each year (t): Zero Returns, Price Impact, and Bid-Ask Spreads. SD and WD indicate our sample groups with strong and weak accounting discrepancies, respectively. Dummy variables that take the value of one or zero represent different IFRS adopter types. Controls ij comprises the control variables and fixed effects for every firm, country, and year. We examined two-sided t- tests to assess the statistical significance and undertook various sensitivity analyses and robustness checks. 6. Empirical Results 6.1 Results Based on Univariate Analysis Table 3 illustrates the results from our univariate analysis. Statistical significance is indicated at the 1%, 5%, and 10% level with ***, **, and *, respectively. The absolute change in market liquidity in the year of the IFRS adoption (column (b)-(a)) is equated. For three of the six variables, the improvement in market liquidity is higher for the group with weak accounting discrepancies, compared to the group with strong accounting discrepancies. These variables are voluntary adopters Bid-Ask Spreads as well as mandatory adopters Bid-Ask Spreads and Price Impact. For instance, mandatory adopters with weak accounting discrepancies decreased Bid-Ask Spreads by 57 basis points. Mandatory adopters with strong accounting discrepancies decreased Bid-Ask Spreads by 17 basis points. This is an improvement of 22.18% for the group 21

with weak and 7.26% for the group with strong discrepancies. With regard to the relative change in market liquidity (column (b)-(a) in %), five of the six variables show a stronger improvement in market liquidity for the group with weak accounting discrepancies. The one exception is the Price Impact variable for voluntary adopters. Regarding the difference-in-differences results for the relative changes that are statistically significant, all three variables show a stronger improvement for the group with weak discrepancies. For example, the advantage in Zero Returns is 19.19% for voluntary adopters and 9.56% for mandatory adopters. The Bid-Ask Spreads difference for mandatory adopters is 14.92%. For both, absolute and relative changes, only half of the results are statistically significant. Nevertheless, these findings are overall supportive of advantages in market liquidity for the group with weak accounting discrepancies. <insert table 3 about here> 6.2 Results Based on Regression Analysis We present the results of the ordinary least squares coefficient estimates in Table 4. The t-statistics in parentheses indicate statistical significance. IFRS Adopters Development over the Years The observation from 2003 to 2008 demonstrates that all the liquidity coefficients developed better for the group with weak accounting discrepancies. Their market liquidity either increased more or decreased less than the strong differences group s coefficients compared to the benchmark group, which is unaffected by the IFRS adoption. For instance, voluntary adopters Bid-Ask Spreads increased by 500 basis points between 2003 and 2008 for the group with strong accounting discrepancies, compared 22

to the benchmark s mean of 2.94%. This equates to an augmentation of 169.9%. 26 At the same time, the variables for the group with weak accounting discrepancies decreased by 76 basis points which equates a reduction of 25.9%. The mandatory adopter groups with strong and weak accounting discrepancies increased by 526 (187.7%) and 131 (44.6 %) basis points, respectively, and, hence, underlined the better development in market liquidity for the group with weak accounting discrepancies. The coefficients for Price Impact and Zero Returns show the same trends. 27 Furthermore, they are generally statistically significant at the 1% level. The conclusion stays unchanged when we observe the examined periods 2003 to 2004, to 2006, and to 2008, separately. IFRS Adopters Development compared to the Time before IFRS In addition, we compare the treatment samples market liquidity coefficients before the IFRS adoption (see Table 4, (a)) to the coefficients after the adoption. The results confirm the former conclusions. For example, the strong differences group s Bid-Ask Spreads before IFRS were 3.66%. 28 The Bid-Ask Spreads for the voluntary IFRS adopters developed from 3.31% to 4.0% and 7.94% over the measured periods 2003 to 26 ln( 0.0294) 0.119 0.189 0.685 We tabulate the Bid-Ask Spreads variables as e =0.0794, 0.0794-0.0294=0.0500 for voluntary adopters with strong accounting discrepancies, and as ln( 0.0294) 0.200 0.230 0.130 e =0.0218, 0.0218 0.0294=-0.0076 for voluntary adopters with weak accounting discrepancies. The value 0.02942 represents the benchmark s mean for the Bid-Ask Spread variable. The other values are the Bid-Ask Spreads coefficients presented in Table 4. The mandatory ln( 0.0294) 0.255 0.770 adopters with strong and weak accounting discrepancies are calculated as e =0.0820; ln( 0.0294) 0.047 0.416 0.0820-0.0294=0.0526 and e =0.0425; 0.0425-0.0294=0.0131, respectively. 27 The benchmark s means for Price Impact and the Proportion of Zero Returns to tabulate the variables are 1.26 and 7.67%, respectively. 28 0.0294) 0.218 Calculated as ln( e =0.0366. 23

2004, 2005 to 2006, and 2007 to 2008, respectively. 29 This increase of 428 basis points is a deterioration of 117.1% compared to the coefficient before IFRS. The weak differences group s Bid-Ask Spreads before IFRS were 2.64%. The Bid-Ask Spreads for the voluntary IFRS adopters developed from 2.41% to 1.91% and 2.18% over the measured periods 2003 to 2004, to 2006, and to 2008, respectively. This decrease of 46 basis points is an improvement of 17.5% compared to the coefficient before IFRS. Mandatory adopters with weak differences also developed better than the strong differences group. Over the entire period, the weak differences group s Bid-Ask Spreads increased by 161 basis points (deterioration of 61.1%) while the strong differences group s increased by 454 basis points (deterioration of 124.1%). The Price Impact coefficients for the weak differences group s voluntary and mandatory adopters also showed wide advantages over the entire period. The weak (strong) differences group s Proportion of Zero Returns showed an improvement of 3.5% (2.4%) for voluntary adopters. 30 Mandatory adopters Proportion of Zero Returns decreased from 7.45% to 7.34% (improvement of 11 basis points or 1.5%) for the group with weak differences and increased from 7.56% to 7.60% (deterioration of 4 basis points or 0.5%) for the group with strong differences. <insert table 4 about here> In Section 3, we developed the hypothesis that the expected positive capital market effects through the IFRS introduction s higher transparency cannot prevail if analysts fail to interpret financial statements correctly and perceive uncertainty and information risks. We identified the IFRS complexity as a possible cause of uncertainty. A variable 29 0.0294) 0.119 Calculated as ln( ln( 0.0294) 0.119 0.189 ln( 0.0294) 0.119 0.189 0.685 e =0.0331, e =0.0400, e =0.0794. 30 From 7.45% to 7.19% and from 7.56% to 7.38%, respectively. 7.45% (7.56%) is the Proportion of Zero Returns before IFRS for the weak (strong) differences group. 24

that accounts for discrepancies between the local GAAP and the IFRS measures the degree of complexity that IFRS adopters and financial statement users experience. The outcomes from our univariate and regression analyses clearly demonstrate the higher market liquidity of firms from countries with weaker differences between the local GAAP and IFRS and support our hypothesis of the complexity hurdle. Furthermore, the effects did not diminish over the years as one could expect, given that there is an additional natural uncertainty in the year of the adoption. 31 No significant deviations appeared between the voluntary and mandatory IFRS adopter groups. Both groups experience the complexity hurdle. The coefficient estimates as well as the control variables are mainly significant at the 1% level. We then examined our results by performing various sensitivity analyses. We excluded observations from specific countries to control for other country characteristics that could affect firm liquidity. When we remove Portugal, 32 we find that the average enforcement and transparency scores are higher for the strong differences group. Nevertheless previous papers results about IFRS being more beneficial for firms in countries with strong enforcement and transparency, the advantages for the weak differences group persist. This result articulately strengthens the relation between market liquidity and the country of origin. We find the same results when we also exclude Sweden. 33 In addition, we varied the benchmark definitions, tested our regressions robustness by applying a random effect, and omitted the observations for the voluntary adopter years 2003 and 2004 to start the investigation at the time of the IFRS mandate. The main 31 See Section 3, hypothesis H1. 32 We chose Portugal since it belongs to the strong differences group and demonstrates relatively weak transparency and enforcement values. 33 We chose Sweden since it belongs to the weak differences group and demonstrates relatively strong transparency and enforcement values. 25

assertion - that IFRS adopters with weak accounting differences between the local GAAP and IFRS experience a better development in market liquidity than IFRS adopters with strong accounting differences - was usually confirmed. However, statistical significance varies. Furthermore, we excluded the benchmark observations and calculated a regression only between the strong and weak differences treatment groups. We therefore demonstrated that the differences in the coefficients are statistically significant, not only against the benchmark countries but also between the two adopter groups. We illustrate some of the sensitivity analyses in Table 5. <insert table 5 about here> In contrast to our results, Daske et al. (2008) concluded that firms from countries with strong accounting discrepancies between IFRS and the local GAAP profit more from the IFRS adoption compared to firms with weak discrepancies. Their cross-sectional analysis ends in 2005. During the first years of our regression analysis, comparisons in market liquidity between the time before and after IFRS adoption also partially resulted in advantages for the group with strong accounting differences. However, over the entire period, the advantages clearly prevailed for the group with weak accounting discrepancies. We assume Daske et al. s (2008) different conclusions are due to the different observation period. Although our study s focus is on complexity, the coefficients from the regression analysis present an interesting development, which is worth it to be mentioned. Compared to the benchmark, which is unaffected by the IFRS introduction, the weak differences group s coefficients for Bid-Ask Spreads and Price Impact generally show higher market liquidity as a consequence of the IFRS introduction only until 2006. IFRS adopters progress decreased between 2006 and 2008, resulting in even lower liquidity 26

values than those of the benchmark firms. This result was not expected at all, as adoption and analysing difficulties were supposed to be reduced after the early IFRS adoption years. 34 We deduce that companies and investors difficulties in implementing IFRS and analyzing the financial statements, respectively, increased over the years. 35 Consequently, the information risk due to complexity increased. This evidence questions IFRS long-term benefits in general as well as previous studies early results of capital market benefits. The latter were possibly influenced by introduction effects rather than the IFRS adoption itself. These effects can emerge through misinterpretations or investors expectations and enthusiasm. This field is open to future research. 7. Conclusion In this paper, we examined the IFRS introduction s effects in the context of the standards complexity. We formed two groups of countries that experienced the complexity differently. The outcomes from our univariate and regression analyses clearly demonstrate that the complexity in IFRS represents a hurdle and consequently reduces the capital market benefits expected from the IFRS introduction. Countries that experienced the complexity to a lesser extent, showed a stronger percentaged increase in market liquidity in the year of the adoption and a better development in market liquidity over the years calculated against an unaffected benchmark or against the liquidity coefficients before the adoption. 34 Performing our sensitivity analysis, we also find various scenarios for a decrease of the Zero Returns coefficients between 2006 and 2008. Nevertheless, the weak differences group always maintains its advance against the benchmark. 35 This interpretation is consistent with FREP s (2009) report of an increase in the adoption error rate from 26% to 27% and various studies view that IFRS are becoming increasingly complex (see Section 2). 27