Mandatory IFRS and properties of analysts forecasts: How much does enforcement matter?

Transcription

1 Mandatory IFRS and properties of analysts forecasts: How much does enforcement matter? John Preiato, University of Western Australia. Philip Brown, University of New South Wales and University of Western Australia. Ann Tarca, Business School, University of Western Australia, 35 Stirling Highway, Crawley, Western Australia Telephone Fax Ann.Tarca@uwa.edu.au We thank the UWA Business School for research funds. Helpful comments from Mike Bradbury, Julie Cotter, Izan, Marvin Wee and seminar participants at the University of Sydney, the Accounting and Finance Association of Australia and New Zealand (AFAANZ) annual meeting Adelaide 2009 and the American Accounting Association (AAA) annual meeting New York 2009 are gratefully acknowledged.

2 Mandatory IFRS and properties of analysts forecasts: How much does enforcement matter? Abstract (250 words) We examine (1) whether analysts forecast accuracy and the extent to which they disagree have declined since the mandatory adoption of International Financial Reporting Standards (IFRS) in the European Union (EU) in 2005, and (2) whether differences between countries in their enforcement of accounting standards are a contributing factor. Based on a sample of 40,123 firm-month observations from 13 European countries during , we find analysts forecast errors and dispersion are indeed significantly smaller in the post-ifrs period, implying an increase in overall quality of financial reporting following widespread mandatory adoption of IFRS. We show enforcement proxies that focus on accounting enforcement (we use a selfconstructed proxy that distinguishes between countries based on legal aspects, statutory audit and independent oversight of financial reporting) and public and private enforcement in securities markets (La Porta et al. 2006) are more closely associated with properties of analysts forecasts than more general regulation proxies such as Kaufmann et al. (2007). The accounting enforcement proxies are statistically significant variables in our models, indicating analysts forecasts are more accurate and less disperse when enforcement is more developed. However, overall we find that enforcement proxies do not strongly influence the size of the forecast error or the extent to which analysts disagree, suggesting that country institutional differences may be less important than is commonly believed. Key words: Mandatory IFRS adoption; analyst forecast accuracy and dispersion; enforcement of accounting standards. JEL codes: M41, M42, M48, K20. 2

3 1. Introduction The widespread adoption of International Financial Reporting Standards (IFRS), promulgated by the International Accounting Standards Board (IASB), has been described as a watershed in financial reporting (Cairns 2003). From 2005 onwards publicly traded companies in more than 100 countries have been required prepare consolidated financial statements under IFRS and it is expected this number will exceed 150 by 2011 (IASB 2008). Throughout the European Union (EU), IFRS are now used by listed entities in their consolidated accounts. Support for adoption of IFRS is based on the view that a single set of high-quality accounting standards will promote transparency and comparability, thus improving overall quality of financial reporting and efficiency of capital markets (EC Regulation No. 1606/2002). Realization of the anticipated benefits of mandatory IFRS adoption relies on the change from national GAAP to IFRS actually improving the quality of financial reporting. Supporters of IFRS adoption argue that benefits will flow from expanded financial statement disclosures, improved measurement and recognition practices and the narrowing of differences in company reporting which arise when a variety of national GAAP are used. However, a wide range of factors affect the quality of financial reporting, many of which are embedded in a country s institutional framework (SEC 2000; Ball 2006; Ruland, Shon and Zhou 2007). Adoption of a set of high quality standards may be viewed as one of several elements required to achieve high quality financial reporting, others being independent external auditors, public enforcement bodies and courts. Studies suggest that while there has been considerable activity in developing international auditing standards and promoting high quality auditing, there remain many differences between countries in the extent to which they have active enforcement bodies and systems to promote comparable application of IFRS (IFAC 2006; CESR 2007; Zeff 2007). An interesting question is whether these differences between countries are sufficiently important to limit the benefits to be gained from IFRS adoption. In the first stage of this paper we investigate the consequences of mandatory IFRS adoption in the EU for financial analysts, who are among the most important users of financial information. Analysts earnings forecasts provide a way to evaluate use of IFRS because their forecasts reflect firms information environments and the overall quality of financial statement information available to them. In addition, forecasts proxy for investor beliefs, which cannot be directly observed (Arbarbanell, Lanen and Verrecchia 3

4 1995). The accuracy of analysts earnings forecasts has also been linked to information asymmetry and the cost of capital (Gebhardt, Lee and Swaminathan 2001). Thus, we also provide indirect evidence as to whether mandatory IFRS adoption has resulted in reduced information asymmetry and a lower cost of capital, as expected by supporters of IFRS adoption. In the second stage of the paper we explore whether countries institutional frameworks influence the outcomes of mandatory IFRS adoption. Researchers argue that the level of enforcement of accounting standards will influence the extent to which IFRS adoption is beneficial (Schipper 2005; Ball 2006; Holthausen 2009). Given that managerial incentives influence both accounting policy choice and disclosure (Fields, Lys and Vincent 2001; Healy and Palepu 2001), independent enforcement of IFRS is likely to be necessary to ensure high quality comparable reporting. We investigate the relationship between the level of enforcement, in 13 IFRS adopting countries in the EU, and analysts forecast accuracy and dispersion to shed light on the question of whether institutional differences, which are pervasive and difficult to remove, have any real impact on forecasts by financial analysts. Mandatory adoption of IFRS began from 1 January 2005 for consolidated accounts of listed companies from EU member states. Based on a sample of 40,123 firm-month observations over the period for firms which first adopted IFRS from 2005, we find a significant reduction in analysts forecast errors and dispersion following mandatory IFRS adoption. Contrary to suggestions that IFRS earnings are more difficult to forecast (Aubert and Dumontier 2005) and the benefits take time to be realised (Cuijpers and Buijink 2005; Horton, Serafeim and Serafeim 2008) we provide evidence of a strong positive effect for analysts in the immediate post-adoption period Studies document considerable international variation in regulation and enforcement in securities markets (Frost, Gordon and Hayes 2006; Gadinis and Jackson 2007) be that via stock exchanges, government regulators or other bodies. We employ four proxies to capture country difference in institutional setting: a self-constructed proxy, which aims to distinguish between countries based on legal aspects, statutory audit and independent oversight of financial reporting; an amalgam of the La Porta, Lopez-de-Silanes and Shleifer (2006) measures of public and private enforcement; Hope s (2003) composite enforcement proxy; and Kaufmann, Kraay and Mastruzzi s (2007) rule of law measure. We find that the proxies which are more 4

5 recent and focus more on accounting enforcement (the self-constructed proxy and La Porta et al. 2006) have stronger association with properties of analysts forecasts than the older proxy (Hope 2003) and the more general regulation proxy of Kaufmann et al. (2007). Consistent with Hope s (2003) earlier work on national GAAP users, our results suggest that, for IFRS adopting firms, analysts error and dispersion are lower when enforcement is more developed, although the benefits of enforcement are less than many might expect. We add to the literature in the following ways. First, we extend studies which have explored the impact of IFRS on analysts issuing forecasts in a voluntary setting, where its effects appear to be mixed. For example, Ashbaugh and Pincus (2001) find lower forecast errors among voluntary adopters (primarily from Switzerland, France and Canada). Contrary evidence is provided by Daske (2005) who finds lower accuracy and higher dispersion among analysts forecasts for German firms which adopted international accounting standards (IAS) between 1993 and Cuipers and Buijink (2005) find higher dispersion among EU firms using non-national standards (US GAAP and IAS). Hodgson, Rasoul, Harless and Adhikari (2008) conclude that compliance with IFRS disclosure requirements enhances analysts forecast accuracy (based predominantly on firms from Switzerland and Germany). Since the mandatory IFRS setting represents a fundamentally different situation, it cannot be assumed that findings of studies of voluntary adoption are applicable post We extend the literature as we include a large sample of first-time mandatory adopters from 13 countries, thus providing deeper and more generalizable evidence about IFRS effects on analysts in the mandatory setting. 1 To date, only a few studies evaluate the effects of IFRS adoption on a large scale (notable examples include Daske, Hail, Leuz and Verdi 2008 and Horton et al. 2008). There is a need for empirical evidence that assists market participants and regulators to evaluate the benefits of widespread mandatory IFRS adoption and our evidence is relevant to scholars and practitioners reviewing IFRS effects. Our study also offers insights to regulators and governments as they make policy decisions about optimal institutional arrangements in their national capital markets. Although a relationship of enforcement to accounting quality is presumed to exist, few studies of IFRS effects on analysts have considered the role of country differences in enforcement. In an international study of national GAAP users, Hope (2003) finds proxies for enforcement (based on measures of audit spending, audit firm type, stock exchange listing, insider trading, judicial efficiency, rule of law and anti-director rights) are positively related to forecast accuracy. Byard, Li and Yu (2008) report improved forecast accuracy in EU 5

6 countries is associated with stronger enforcement, as proxied by Kaufman et al. s (2007) rule of law variable. Our enforcement index (hereafter PBT index) is based on material provided by IFAC (2006) and the Committee of European Securities Regulators (CESR 2007) in relation to enforcement mechanisms in EU countries in post The PBT index focuses on cross-country variation in factors that may encourage managers to comply with accounting standards. Other widely used enforcement proxies (e.g. Kaufmann et al. 2007) often employ variables expected to correlate with the enforcement of accounting standards, but may not actually reflect the institutional features that influence high quality reporting. Prior studies have raised questions about the influence of institutional setting on the application of IFRS (Ball, Robin and Wu 2003; Hail and Leuz 2007). Our study is the first to use a comprehensive measure of enforcement of accounting standards to capture key aspects of country differences in institutional setting in the post-ifrs period. We expect our index could be used in future research to proxy for differences between countries in enforcement of accounting standards in the post-ifrs period. An important caveat is that none of the enforcement proxies we use has particularly strong explanatory power and there will be a need for more refined measures of enforcement as more information becomes available. The remainder of the paper is organized as follows. Section 2 provides background and develops testable hypotheses. Section 3 describes our data and method and Section 4 presents the results. Section 5 concludes the paper. 2. Background and hypotheses Mandatory IFRS adoption and analysts forecast accuracy IFRS are promoted as high quality accounting standards which will improve the comparability and transparency of financial statements. For many European firms the mandatory adoption of IFRS should lead to more extensive and informative financial statement disclosures. In a voluntary adoption setting, Daske and Gebhardt (2006) confirm that the quantity and quality of corporate disclosures improved with the adoption of IFRS in Austria, Germany and Switzerland. As increased financial statement disclosures potentially allow analysts to better understand a firm s strategies, prospects and accounting practices, we might expect their forecasts to be more accurate under IFRS. This would be consistent with results of prior studies investigating 6

7 financial disclosure and analysts forecast accuracy and dispersion. Basu, Hwang and Jan (1998), Chang, Khanna and Palepu (2000) and Hope (2003) report greater financial statement disclosure and improved analysts forecast accuracy in cross-country studies. It is also argued that adopting IFRS leads to a more restricted set of accounting measurement methods and, with fewer measurement rules to deal with, analysts can more easily master the existing set (Ashbaugh and Pincus 2001). Hope (2004) examines the relationship between the extent of choice among accounting methods and analysts forecast accuracy for a sample of 18 countries, including 12 European countries, and confirms that a more restricted set of accounting methods leads to greater forecast accuracy. Another argument in support of improved forecast accuracy under IFRS relates to diversity in national GAAP. Bae, Tan and Welker (2008) study 6,888 foreign analysts following 6,169 firms from 49 countries over and show cross-country differences in GAAP are costly for financial analysts. They find a greater difference in GAAP of two countries reduces the forecast accuracy of foreign financial analysts. On this basis, the adoption of IFRS, which increases the harmony of accounting practices across European countries, should lead to a reduction in analysts forecast errors. Thus, the first hypothesis is: H1: Analysts forecasts are more accurate in the post-ifrs mandatory adoption period. Mandatory IFRS adoption and analysts forecast dispersion Many of the arguments related to the proposition that mandatory IFRS adoption reduces analysts forecast errors also apply to forecast dispersion. For example, a decrease in the number of inconsistencies in GAAP across countries should reduce the extent of disagreement among financial analysts as well as analysts forecast errors. However, the effect of additional financial statement disclosures on analysts forecast dispersion is not unambiguous. Lang and Lundholm (1996) indicate the effect of increased disclosure on analyst dispersion depends on whether differences in the earnings forecasts of individual analysts are attributable to differences in private information or differences in forecasting models. If earnings forecasts differ because analysts possess different private information, then an increase in the availability of public information will encourage them to place less weight on private information, which would lead to a decrease in forecast dispersion. However, if 7

8 analysts forecasts differ because they interpret annual report disclosures in different ways, an increase in public information would lead to an increase in forecast dispersion. Consequently, ex ante we cannot be confident of the effect of mandatory adoption of IFRS on forecast dispersion. However, given that mandatory IFRS adoption was intended to improve the overall quality of financial information and thereby reduce both information asymmetry and investor uncertainty in securities markets, the second hypothesis is: 2 H2: Analysts forecasts are less disperse in the post-ifrs adoption period. We do note other reasons why H1 and H2 might not be supported. Although the adoption of IFRS was expected to be beneficial to users of financial information, it may be that IFRS adoption actually impairs analysts ability to accurately predict future earnings. It may be difficult for analysts to generate accurate earnings forecasts around IFRS adoption because of their unfamiliarity with the new accounting standards. Under UK GAAP, Acker, Horton and Tonks (2002) show that the introduction of UK Financial Reporting Standard 3 reduced analysts forecast accuracy in the first year of mandatory adoption. In relation to EU firms in 2005, Aubert and Dumontier (2005) conclude that forecast error is associated with the extent of impact of IFRS adoption on earnings and that generally analysts were unable to predict the impact of IFRS adoption on earnings. Both Cuijpers and Buijink (2005) and Horton et al. (2008) suggest the benefits of IFRS adoption take time to materialize. Ashbaugh and Pincus (2001) suggest more volatile earnings in the post-ifrs period might lead to lower forecast accuracy. Managers may have less ability to smooth earnings under IFRS because of the more restricted set of measurement methods. Ball (2006) argues that the focus of IFRS on fair value accounting may lead to increased earnings volatility and, consequently, less accurate earnings forecasts. Barth, Landsman and Lang (2008) confirm greater volatility in earnings after the voluntary adoption of IFRS. The authors find that a sample of voluntary adopters from 21 countries (56% of their sample are from EU countries) exhibit lower levels of earnings smoothing than a matched sample of firms using local GAAP. A uniform set of accounting standards may not improve the quality of financial reporting internationally due to marked economic, political and cultural differences between countries (Zeff 2007; Ball 2006). The argument is that diversity among countries may mean a uniform set of accounting standards does not provide relevant financial information in some instances. In this vein, Horton et al. (2008) find the benefits of IFRS 8

9 for analysts are not observed in all situations. They report improved accuracy and less dispersion for mandatory adopters in EU countries, but find the improvement in the information environment (proxied by forecast accuracy, dispersion, analyst following and volatility of revisions) is greater for firms that had voluntarily adopted IFRS earlier and only for non-financial firms. The role of enforcement Prior research suggests a country s institutional setting affects accounting outcomes and that the introduction of higher-quality accounting standards is no guarantee of higher quality financial reporting (Ball, Kothari and Robin 2000; Ball et al. 2003; Leuz, Nanda and Wysocki 2003). That is, adopting higher quality accounting standards (such as IFRS) does not confer benefits upon users of financial information unless managers comply with the new standards, which in turn may require a strong system of enforcement. Regulators in the EU and elsewhere have long recognized the need for enforcement in promoting compliance with accounting standards (SEC 2000; CESR 2003). Tweedie and Seidenstein (2005, 590) state that a sound financial reporting infrastructure must be founded upon an enforcement or oversight mechanism that ensures that the principles as laid out by the accounting and auditing standards are followed. 3 Given that enforcement may well play an important role in determining the effectiveness of accounting standards, the potential benefits of mandatory IFRS adoption in the EU may vary among countries reflecting differences in their enforcement regimes. Hail and Leuz (2007) suggest that, despite a greater emphasis on strengthening the enforcement of accounting standards in recent times, such variation does exist. Indeed, Ball (2006) suggests political and economic factors mean IFRS enforcement will be uneven around the world, including within Europe. Although national standard setters and the IASB can encourage and promote the adoption of IFRS, they cannot require that managers be incentivized to comply with accounting standards nor can they directly influence the audit function. Healy and Palepu (2001) consider that differences in auditing standards, the enforcement of these standards, the legal framework governing the audit profession and differences in professional training requirements will influence the credibility of audit reports and financial statements across countries. Francis, Khurana and Pereira (2003) confirm financial reporting and disclosure quality are 9

10 higher in countries with developed auditing infrastructure and greater auditing enforcement. The development of international auditing standards by IFAC is an important step for promoting consistency in auditing practices between countries. However, country differences in the institutional framework for setting or endorsing auditing standards and promoting compliance with them remain (IFAC 2006). IFRS adoption is subject to the risk that application of the standards will vary among countries in response to firm- and country-level incentives. Ernst & Young (2006) concludes that national patterns in use of IFRS are emerging. Jermakowicz and Gornik-Tomaszewski (2006) report survey evidence that EU preparers considered differences in interpretation of standards would be an obstacle to achieving accounting convergence. Armstrong, Barth, Jagolinzer and Riedl (2008) also present evidence that country-level differences will be relevant. They find the market reaction to announcements about adoption of IFRS varied among European countries and conjecture the reaction is less positive in jurisdictions expected to have weaker enforcement. Daske et al. (2008) provide some early evidence about capital market effects of IFRS. Subject to the caveat that they consider only the immediate adoption period, they suggest that the benefits of IFRS (favourable effects on market liquidity and firm valuation) exist only in countries with strict enforcement regimes and institutional environments that provide strong incentives for high quality reporting. Jeanjean and Stolowy (2008) investigate accounting quality under IFRS by examining earnings management pre- and post-ifrs in the UK, France and Australia (countries which did not permit early adoption of IFRS and have enforcement mechanisms in place). They conclude that the pervasiveness of earnings management did not decline and that future initiatives should focus on harmonising legal enforcement systems and their effectiveness rather than accounting standards. Using an international sample of national GAAP firms from 22 countries, Hope (2003) supports the proposition that the level of enforcement of accounting standards is positively related to the accuracy of analysts earnings forecasts. He suggests strong enforcement is associated with higher forecast accuracy because enforcement encourages managers to follow accounting rules and thus reduces analysts uncertainty about future earnings. In addition, Hope (2003) finds enforcement is important when standards allow a choice of methods, which remains the situation with many IFRS (Nobes and Parker 2006). Barniv, Myring and 10

11 Thomas (2005) also find country differences have economic consequences. Using a sample of national GAAP firms from 12 common law and 21 code law countries, they show differences in the legal and financial reporting environment are associated with analysts forecast accuracy and that the characteristics of analyst behaviour can be linked to countries legal origins. To explore the influence of institutional setting and IFRS, Hodgson, Tondkar, Harless and Adhikari (2008) focus on compliance with standards. They consider voluntary adopters of international standards (predominantly from Switzerland and Germany) and conclude compliance with disclosure requirements enhances analysts forecast accuracy. In the post-2005 mandatory environment, Byard et al. (2008) report lower forecast error in for a sample of 1,163 EU firms, predominantly from the UK (N=400), France (172) and Italy (96). Using a constant analyst-firm sample, they show the decrease in error is greater for firms domiciled in countries where differences to national GAAP are greater and enforcement (Kaufmann et al. s (2007) rule of law proxy) is stronger. These findings are consistent with the proposition that strong enforcement encourages managers to comply with accounting standards, which, in turn, reduces analysts uncertainty about managers disclosures and improves forecast accuracy. In the light of the foregoing literature, the third hypothesis is: H3: Analysts forecasts are more accurate when enforcement is greater. In this area of literature, studies focus more on forecast accuracy than forecast dispersion, which arguably measures a different property. Aspects of IFRS adoption mentioned previously, such as unfamiliarity with standards and increased volatility in earnings under IFRS, may affect dispersion differently than accuracy. Nevertheless, Horton et al. (2008) find reduced dispersion for both early IFRS adopters and later mandatory adopters. Consistent with IFRS effects providing an improved information environment, the fourth hypothesis is: H4: Analysts forecasts are less disperse when enforcement is greater. 3. Data and method Sample and period 11

12 Table 1 summarizes the sample selection process. The initial sample of median consensus forecasts available on I/B/E/S consists of 217,226 observations for 18 EU member states and Norway (member of the European Economic Area). The pre-ifrs adoption period includes fiscal years commencing from 2002 because it is the year immediately preceding the announcement that the EU would adopt IFRS for consolidated accounts of listed firms. The post-ifrs period includes fiscal years commencing during , the three years immediately following IFRS adoption. We determine actual financial year ends (i.e. 31 December, 30 June or otherwise) to ensure forecasts are correctly classified into pre-adoption and postadoption years. The sample was reduced by the following: actual earnings, stock price and other accounting data not available (61,053); consensus forecasts announced more than one year before the earnings announcement date (8,637); number of days between the end of the fiscal year and the earnings announcement date exceeded 365 (8); less than three analysts included in the consensus forecast (34,306); absolute analyst forecast error in the corresponding month of the previous year cannot be calculated due to missing consensus forecast or actual earnings data (38,204); type of accounting standards used was not available (2,012); and firms not switching from national GAAP to IFRS on 1 January 2005 (18,795). Therefore we restrict the sample to firms that mandatorily switched to IFRS for fiscal years beginning on or after 1 January In addition, we exclude countries when the number of firms was below ten (Austria n = 4, Hungary n = 2, Czech Republic n = 1 and Luxembourg n = 1) and they are not covered by the La Porta et al. (2006) or Hope (2003) indexes (Poland and Greece), leaving 13 countries. We censor the top and bottom 1% of observations on forecast errors and dispersion of forecasts to give a final sample of 40,123 observations. Method The following regression models (firm, month and year subscripts omitted for convenience) are used to investigate the relationship of analysts forecast errors and dispersion before and after IFRS adoption. ERROR = β 0 + β 1 IFRS + β 2 LOSS + β 3 HORIZON + β 4 DISAGREEMENT + β 5 FOLLOWING + β 6 LOG_SIZE + β 7 PREV_AFE + ε (1) DISAGREEMENT = β 0 + β 1 IFRS + β 2 LOSS + β 3 HORIZON + β 4 FOLLOWING + β 5 LOG_SIZE + β 6 PREV_AFE + ε (2) 12

13 where ERROR DISAGREEMENT IFRS LOSS HORIZON FOLLOWING LOG_SIZE PREV_EPS The absolute difference between actual EPS and the monthly median consensus forecast scaled by stock price at the middle of the month. The inter-analyst standard deviation of analysts earnings forecasts scaled by stock price at the middle of the month. An indicator variable equal to 1 where actual EPS is reported under IFRS, zero otherwise. An indicator variable equal to 1 if the current year s EPS is negative, zero otherwise. The number of days between the announcement of the median consensus forecast and the earnings announcement date. The number of analyst earnings forecasts included in the median consensus forecast. Natural log of the market value of equity in USD million at the middle of the month. The absolute value of last year s forecast error scaled by price, measured at the corresponding month in the previous year. We expect the coefficient on IFRS in models 1 and 2 to be negative, consistent with a reduction in analysts forecast errors (H1) and forecast dispersion (H2) in the post-ifrs adoption period. We add ENF_PROXY to models 3 and 4 to explore the role of enforcement on attributes of analysts forecasts. We expect a negative coefficient ENF_PROXY indicating that lower forecast errors (H3) and dispersion (H4) are associated with greater enforcement. ERROR = β 0 + β 1 IFRS + β 2 ENF_PROXY + β 3 LOSS + β 4 HORIZON + β 5 DISAGREEMENT + β 6 FOLLOWING + β 7 LOG_SIZE + β 8 PREV + ε (3) DISAGREEMENT = β 0 + β 1 IFRS + β 2 ENF_PROXY + β 3 LOSS + β 4 HORIZON + β 5 FOLLOWING + β 6 LOG_SIZE + β 7 PREV + ε (4) Measuring enforcement Hope (2003) notes there is no straightforward and uncontroversial way to measure the enforcement of accounting standards because of the various forms that enforcement mechanisms take in different countries and the subjectivity involved in evaluating the effectiveness of these mechanisms. He builds a proxy based on measures of audit spending, audit firm type, stock exchange listing, insider trading, judicial efficiency, rule of law and anti-director rights. The last three items are sourced from widely cited work by La Porta, Lopez-de- Silanes, Shleifer and Vishny (1998), which provides useful proxies for capturing country differences. 4 However, the La Porta et al. (1998) measures are limited in that they are based on data from and thus do not capture subsequent changes in institutional setting. Further, proxies that measure rules may be deficient because they do not capture how the law is actually used (Jackson and Roe 2009; Mahoney 2009). 13

14 La Porta et al. (2006) provide more recent measures (based on interview data) that capture the level of public and private enforcement in securities markets in 49 countries. The proxy for public enforcement summarizes information about country-specific supervisor characteristics, rule-making power, investigative powers, orders and criminal sanctions. The proxy for private enforcement summarizes information relating to the liability standard for companies and managers, for distributors (underwriters) and accountants. It also captures variation in mandatory disclosures such as the prospectus, compensation, shareholders, inside ownership, irregular contracts and transactions. La Porta et al. s (2006) measures, while compiled some years ago, may still be relevant in the IFRS setting because security market regulation and laws may change slowly. On the other hand, because the measures relates primarily to securities laws relating to initial public offerings (rather than laws that promote compliance with accounting standards) and does not account for cross-country variation in the audit function it may not have explanatory power in the IFRS setting. We create an index based (hereafter LAPORTA) by taking the average of the public and private enforcement scores for each country. Researchers also use measures provided by Kaufmann et al. (2007) (Daske et al. 2008; Byard et al. 2008; Armstrong et al. 2008). These indexes have the advantage of covering several aspects of institutional setting (including proxies for rule of law and regulatory effectiveness ). The rule of law variable measures perceptions of the extent to which market participants have confidence in and comply with the laws of society, and is an aggregate indicator based on a range of specific and disaggregated individual variables measuring various dimensions of governance. For example, it captures the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence. The primary advantage of this proxy is that the rule of law variable is available for each year of the sample period ( ) so that improvements in the institutional setting over time are taken into account. However, Kaufman et al. s (2007) measure is subject to a number of limitations for accounting research. First, the rule of law variable does not directly capture differences in the effectiveness of the statutory audit function or the institutional oversight system, which have traditionally been important enforcement mechanisms in European countries (FEE, 2001). Second, the rule of law captures variation in factors that may be entirely unrelated to the enforcement of accounting standards, such as the kidnapping of foreigners and the prevalence of violent 14

15 and organized crime. To address the limitations of existing proxies and to provide a measure which focuses on enforcement of accounting standards post-ifrs, we create our own proxy. Self-constructed enforcement proxy FEE (2001) defines enforcement as the substantive procedures that exist in a country to ensure the proper application of accounting principles and standards. Enforcement of accounting standards in the EU may involve any of the following: (1) Self-enforcement in the preparation of financial statements; (2) Statutory audit of financial statements; (3) Approval of financial statements; (4) Institutional oversight system; (5) Court sanctions; and (6) Public and press sanctions. FEE (2001) concludes that mechanisms such as the routine approval of financial statements (not considered an enforcement mechanism) as well as statutory audit existed in all European countries pre Entities such as stock exchanges, regulators of stock exchanges, governments departments, and private sector bodies also play a role in some countries. Prior to 2005 not all EU countries had an independent enforcement body and the type of body used varied between countries (Brown and Tarca 2005). To promote comparable use of IFRS post-2005, CESR (2003) issued a standard requiring that all EU member countries have an independent enforcement body with particular characteristics in place from 1 January 2005 (see appendix). The self-constructed measure (PBT index) focuses on each country s institutional setting relating to legal environment and institutional oversight. Our approach is similar to La Porta et al. (2006) and Al-Shammari, Brown and Tarca (2008). That is, relevant features of the institutional framework are identified and countries are assigned one point when a feature is present and zero when it is absent. Scores are then summed to provide a country-level measure of enforcement. In Table 2 Legal environment we gave one point to countries that permitted the early adoption of IFRS (EARLY) because early adoption may provide useful exposure to IFRS. We gave one point to countries that give shareholders the right to bring a cause of action against managers for non-compliance with accounting standards (PROSEC). The relatively low weightings for the legal environment reflect the fact that complainants rarely consider going to the courts to obtain revision of (alleged) misstatements in the financial reports (FEE, 2001). Thus, a country environment that promotes compliance with IFRS solely though legal provisions may be less effective than one that uses other more proactive institutional oversight systems. 15

16 In Table 2 Statutory Audit we gave two (one) points to countries which have fully (partially) adopted national auditing standards (NAT. STDS) and two points when auditors are required to complete a university degree (DEGREE). We gave two points when auditors are required to sit an examination prior to admission to the profession (EXAM) and two points when the professional accounting body or bodies supply additional training to ensure auditors meet a minimum level of competence (TRAINING). Two was assigned to countries in which the rotation of auditors is compulsory (ROTATION). Data was sourced from FEE (2001) and updated for 2005 from IFAC (2007) which contains data about national requirements provided by bodies representing the auditing profession. The Statutory audit measures (maximum 10 points) are given a higher weighting than Legal environment (maximum 2 points) because auditors play a more active role in ensuring that firms comply with accounting standards than does the judiciary (FEE, 2001). Table 2 Institutional oversight measures the effectiveness of the institutional oversight systems that were in place in The measures are derived from CESR (2007) which summarizes the compliance of EU countries with CESR Standard No 1 on Financial Information (CESR, 2003). The standard stipulates 21 principles by which, according to CESR, convergence of the institutional oversight systems existing within European countries may be achieved. First, we gave a score of three to countries that are fully compliant with Principle 3 to Principle 8, which relate to the existence of a public enforcer. Second, we gave a score of three to countries fully compliant with Principles 11 to 15, which outline the methods to be employed by the public enforcer. Finally, we gave a score of three if countries met the requirements of Principles 16 to 19, which relate to actions taken by the public enforcer to achieve an appropriate level of disclosure or, where applicable, a public restatement of the financial reports (maximum nine points). We gave a score of two to countries in partial compliance with the above sets of enforcement principles. The final enforcement score (maximum 21) assigned to each country is equal to the sum of the scores across the three dimensions (legal framework, statutory audit and institutional oversight). Dependent Variables Analysts forecast error is defined in month t (ERROR) as the absolute value of actual earnings per share less the median consensus forecast of EPS in month t, scaled by price that month. Analysts forecast error is calculated for each month in the year preceding the earnings announcement date. This controls for the effect that consensus earnings forecasts become more accurate as the earnings announcement date approaches 16

17 (Clement 1999; Brown, Taylor and Walter 1999). To facilitate comparison of forecast errors across firms, we divide the absolute value of analysts forecast error by the stock price. We use the median forecast rather than the mean as the measure of the consensus forecast because the median is less sensitive to outliers. Following Lang and Lundholm (1996), analysts forecast dispersion in month t is measured by the inter-analyst standard deviation of forecasts included in the consensus forecast of EPS in month t, again divided by price that month. Control Variables We include a set of variables to control for other factors influencing analysts forecast error and dispersion. Larger firms are more likely to provide additional disclosures than smaller firms, which resolve uncertainty about a firm s future prospects and thus reduce forecast errors and dispersion (Lang and Lundholm 1996). However, larger firms are also more likely to have complex earnings, which may increase forecast errors and dispersion. We control for firm size (LOG_SIZE) using the market capitalization in the month of the forecast. We control for the extent of analyst following (FOLLOWING) because more intense competition between analysts provides them with a greater incentive to accurately forecast future earnings. Consistent with the literature (Hodgdon et al. 2008), we predict a negative relationship between analyst following and analysts forecast errors. Although the expected relationship between analyst following and forecast dispersion is less clear (as there is a greater chance of a dissenting earnings forecast when more analysts are following a firm) we predict a negative relationship, reasoning that pressure for greater accuracy will also result in lower dispersion. Similar to Herrmann, Hope and Thomas (2007), we measure the extent of analyst following as the number of forecasts included in the median consensus forecast. We control for the absolute value of the forecast error in the corresponding month of the previous year scaled by price (PREV_AFE) because we expect the current year s forecast error and dispersion to be positively related to analysts forecast error in the previous year (Brown et al. 1999). We include the dummy variable LOSS as an ex-post measure that takes into account the fact that analysts are more likely to have issued optimistic forecasts in loss periods (Herrmann et al. 2007). We expect to observe larger forecast errors and greater dispersion in forecasts for loss-making firms. We include the number of days between the announcement of the consensus forecast and the announcement of actual earnings (HORIZON) to control for the fact that earnings forecasts issued tend to become more accurate as the earnings announcement date approaches (Clement 1999; Brown et al. 1999). Consistent with Brown et al. (1999), we include 17

18 DISAGREEMENT as a control variable in models with analysts forecast error as the dependent variable because the extent of analyst disagreement can also reflect the difficulty in making a forecast. All analyst forecast data is collected from I/B/E/S and information about accounting standards is sourced from Datastream (Worldscope). 5 We normalise all explanatory variables (subtract the mean and divide by the standard deviation for each continuous variable, and subtract the mean for each indicator variable) to assist in interpreting the regression coefficients. We include ENF_PROXY to capture differences between countries and therefore do not include country fixed effects. 6 We use pooled OLS with White (1980) standard errors to allow for heteroscedasticity, and undertake a detailed Monte Carlo analysis to verify repeated sampling of forecasts for the same company-year does not unduly distort the significance levels in the regressions. 4. Results Descriptive statistics Panel A of table 3 shows that mean (median) ERROR is (0.008) with a maximum value of and standard deviation of DISAGREEMENT shows less variation: the mean (median) value is (0.006) with a maximum value of 0.08 and standard deviation of PREV_ERROR, which captures error in the same month of the previous year, has a mean (median) of (0.009) and standard deviation of LOG_SIZE shows sample firms range in size (2.439 to ), with mean (median) natural log market capitalization of (7.510). The number of forecasts (FOLLOWING) included in the consensus or median forecast also varies, with an average of 11 (standard deviation = 7). The mean (median) timing of the forecasts (HORIZON) is 183 (182) days from the date of the consensus median forecast to the earnings announcement date. Panel B shows significant differences between the pre- and post-ifrs periods. ERROR, DISAGREEMENT and PREV_ERROR are significantly lower post-ifrs. LOG_SIZE and HORIZON are greater in the post-ifrs period, the latter at the 1% level. There is no significant difference in FOLLOWING pre-and post-ifrs. There are 1,727 loss firms overall (panel C) and significantly fewer loss firms in the post-ifrs period (717 compared to 1,010). Untabulated tests of correlations show significant positive product-moment correlations (p<0.01) between ERROR and both DISAGREEMENT (0.40) and PREV_ERROR (0.36). As expected, ERROR is negatively correlated with SIZE (-0.20) and FOLLOWING (-0.16) and positively correlated with 18

19 HORIZON (0.12). Similarly, DISAGREEMENT is negatively correlated with SIZE (-0.16) and FOLLOWING (-0.11) and positively correlated with HORIZON (0.01). Table 3 panel A reports descriptive statistics for the four enforcement proxies for the 13 countries included in the statistical analysis. These are the countries covered in all indexes; i.e., the PBT index as well as proxies from Kaufmann et al. (2007), La Porta et al. (2006) and Hope (2003). The PBT index has a mean (median) of 15.6 (16), a minimum of 9 and maximum of 21, and standard deviation of The KAUFMANN index covers the period The mean (median) score is 1.56 (1.72) and the standard deviation is The LAPORTA proxy ranges from to with a mean (median) of (0.583). The HOPE proxy has a mean (median) score of (-0.190) and a standard deviation of Due to the diverse ways in which the proxies are constructed, none of these numerical measures is directly comparable. However, HOPE is positively correlated with KAUMANN (0.81) and LAPORTA (0.54); LAPORTA is positively correlated with PBT (0.15) and KAUFMANN (0.09); and PBT is negatively correlation with KAUFMANN (-0.36) and HOPE (-0.31). The diversity in the correlations suggests they measure different underlying aspects of a country s institutional setting. Table 4 shows that the PBT index is highest in Denmark, Germany and France and weakest in Luxembourg, Sweden, Hungary, the Netherlands and Poland. Our rankings reflect changes to enforcement mechanisms made around For example, Germany ranks second highest, reflecting regulatory changes made in Germany to ensure full compliance with the enforcement principles of CESR s Standard No. 1. In contrast, La Porta et al. s (2006) index (which is based on earlier data and includes different factors) shows Germany as having the second-least effective enforcement system. In some other countries, the rankings of PBT and La Porta et al. (2006) are similar. Austria has the second-lowest and France the second-highest enforcement score in both indexes. The KAUFMANN proxy ranks Norway and Denmark as the countries with the most effective enforcement and Italy and Greece as the countries with the least effective regulatory regimes. This differs from the PBT index, which assigns relatively low scores to Scandinavian countries such as Sweden and Norway, but comparatively higher scores to the Italy and Spain. These differences reflect the data provided by the IFAC (2007) and CESR (2007) surveys and highlight the fact that a country might rank highly in terms of rule of 19

20 law, but it does not necessarily follow that it has a strong audit function or effective public enforcement. Nevertheless, the usefulness of the PBT index may be compromised if the data on which it is based was incomplete or inaccurate (for example, because a country changed its enforcements mechanisms after the data was collected). In addition, the PBT index is based largely on input measures which may not effectively capture how the institutional setting affects the practices of auditors and preparers. Multivariate Results Analysts forecast error Table 5 panel A reports the results of OLS regressions of analysts forecast error (ERROR) (models 1, 2 and 3) on an indicator variable for the pre- and post-ifrs adoption periods (IFRS), variables representing the institutional setting under IFRS (PBT) and regulation in general (KAUFMANN), and several control variables. Consistent with our predictions and descriptive evidence reported in table 3, model 1 shows forecast errors are on average per cent smaller (i.e., greater accuracy) in the post-ifrs adoption period. 7 The coefficient on the IFRS indicator variable is negative and significant (p<0.01). The result complements that of Horton et al. (2008) who report improved forecast accuracy for some EU firms under IFRS. It extends Byard et al. (2008) who consider a smaller sample of firms and shorter time period. The control variables are significant in the expected direction. ERROR is positively associated with the age of the forecast, being a loss firm, having greater analyst dispersion and having higher forecast error in the previous year (HORIZON, LOSS, DISAGREEMENT and PREV_AFE are positive and significant p<0.01). ERROR is lower for firms which are larger and have a greater number of analyst forecasts (LOG_SIZE and FOLLOWING are negative and significant at p<0.01). While model 1 shows that ERROR is less post-ifrs, it does not offer insights into why this may be so. In model 2 we include the self-constructed enforcement measure (PBT) to proxy for the changed institutional setting under IFRS. The coefficient of PBT is negative and significant (p<0.01), providing support for our prediction that greater enforcement is associated with lower forecast error in the post-ifrs period. However, in reality a one standard deviation increase in PBT has relatively little (about 2%) effect on average error. The explanatory power of models 1 and 2 is comparable (adjusted R 2 is 32% in both models) and signs, coefficients, t-statistics and significance levels for control variables in each model are similar. 20

21 In model 3 we include the PBT proxy, which focuses on accounting enforcement, as well as the KAUFMANN proxy to capture more general regulatory differences. PBT is positive and significant (p<0.01) but KAUFMANN does not have the predicted negative sign nor is it significant. The explanatory power of model 3 is similar to that of model 2, which suggests the rule of law proxy does not add to the explanatory power provided by the PBT proxy. The implication is that the more specific proxy (PBT) is statistically more helpful in explaining variation in forecast error than the more general proxy (KAUFMANN) although its effect on the size of error is relatively small. We leave further discussion of the KAUFMANN proxy until later in this section. Analysts forecast dispersion Table 6 panel B reports the results of OLS regressions of analysts forecast dispersion (DISAGREEMENT) (models 4, 5 and 6) on an indicator variable for the pre- and post-ifrs adoption period (IFRS), variables representing the institutional setting under IFRS (PBT) and regulation in general (KAUFMANN), and several control variables. In model 4 we observe that the coefficient for IFRS is negative and significant (p<0.01) supporting our prediction. Forecast dispersion is on average per cent lower under IFRS. This result is consistent with Horton et al. (2008) who report lower dispersion for first time mandatory IFRS adopters in The control variables are mainly significant in the expected direction. DISAGREEMENT is greater when the forecast horizon is longer, when firms made a loss in the current year and when there was higher forecast error in the previous year (HORIZON, LOSS and PREV_AFE are positive and significant at p<0.01). DISAGREEMENT is lower for larger firms (LOG_SIZE is negative and significant at p<0.01). Contrary to our prediction and the sign on the variable in model 1, the coefficient of FOLLOWING is positive indicating that disagreement is greater when more forecasts are included in the measure of dispersion and after controlling for firm size. We add the PBT proxy to model 5. Again, contrary to our prediction, the coefficient for PBT is positive and significant (0.0001, p<0.10), suggesting dispersion is higher for firms subject to greater enforcement. We expected more enforcement to be associated with lower error and lower dispersion; however we find evidence only for the former. Thus firms from countries where enforcement is greater appear to have lower error but higher dispersion under IFRS. A possible explanation is that higher enforcement is associated with greater 21