Abstract. A Comparative Analysis of the Distribution of Earnings Relative to Targets in the European Union

Transcription

1 Abstract A Comparative Analysis of the Distribution of Earnings Relative to Targets in the European Union by HOLGER DASKE, GÜNTHER GEBHARDT AND STUART MCLEAY This paper provides evidence on the distribution of reported earnings relative to targets across and within 14 countries of the European Union (EU). We find for a large sample of over 6, firm-years between 1986 and 21 that more than expected firms (1) report small positive earnings, (2) report small positive earnings changes and (3) have zero or small positive forecast errors. These discontinuities are much more pronounced in our EU sample as compared to the (former) U.S. or pure U.K. evidence. In addition, the functional distributions of reported earnings and earnings changes are less spread and more clustered around zero in Europe, indicating a high(er) extend of income smoothing. Further, the phenomenon is compared between the EU countries: Analyzing all countries individually as well as (stylized) groupings of accounting regimes, we find first that the avoidance of losses and of decreasing earnings is most heavily practiced in Continental European countries, particularly in German accounting origin countries, and more pronounced than in the benchmarking British origin countries. Second, pronounced discontinuities are also detected for Continental European firms using the Anglo-American reporting standards IAS/IFRS or US-GAAP. These results hold both for the application of different deflators as well as after controlling for differences in sample characteristics. Finally, we analyze whether European firms really utilize their managerial discretion in order to beat benchmarks: We suggest (long-term) income smoothing as one plausible reason why the discontinuities are so pronounced in the EU. In the case of (short-term) accruals management to achieve targets, our results are inconclusive: Although the distributions of earnings, earnings changes and forecast errors before discretionary accruals show less small positive surpluses and are more spread overall, we cannot detect unusually high positive discretionary accruals in the areas slightly above the targets. In the case of analyst forecasts, however, firms seem to engage in expectations management by walking down consensus estimates. We provide evidence of a switch from upwardbiased to unbiased forecasts of annual earnings and, in particular, an increase in the ratio of small positive to small negative surprises as the announcement date approaches over the forecast horizon. Keywords: earnings targets, properties of multinational accounting earnings, earnings management, expectations management. Data availability: Data are available from the public or commercial sources identified in the text (Thomson Financial Worldscope & I/B/E/S) Program Codes: SAS Programs are available for the referee on request 1

2 A Comparative Analysis of the Distribution of Earnings Relative to Targets in the European Union HOLGER DASKE, GÜNTHER GEBHARDT AND STUART MCLEAY * May 23 * The authors are respectively from J.W. Goethe-Universtiät Frankfurt am Main, Chair of Accounting and Auditing (authors 1 and 2) and from the University of Wales, Bangor. This paper forms part of the HARMONIA program on research into accounting harmonization and standardization in Europe. The main research was conducted while Holger Daske was visiting researcher at Lancaster University and at the University of Wales, Bangor. We gratefully acknowledge the financial contribution of the European Commission through the Human Potential Program, Contract HPRN-CT-2-62, the Arthur Andersen Foundation and Ernst & Young Deutschland. We are grateful to Thomson Financial I/B/E/S for providing analyst earnings forecast data as part of a broad academic program to encourage earnings expectations research. We thank workshop participants at the EAA and AAA Doctoral Consortia 22, the EAA Annual Meeting 22 Copenhagen and 23 Seville, the Lancaster University Friday Seminar, and the HARMONIA group for helpful comments. Adress for correspondence: Holger Daske, Professur für Betriebswirtschaftslehre, insb. Wirtschaftsprüfung (Accounting and Auditing), J.W. Goethe-Universität Frankfurt am Main, Mertonstr , D-6325 Frankfurt/M daske@wiwi.uni-frankfurt.de 2

3 1. INTRODUCTION This paper provides first detailed multinational evidence on the distribution of reported earnings relative to targets and its relation to managerial discretion across and within 14 countries of the European Union (EU). Both academic accounting research (e.g. Pope and Walker 1999; Ball, Kothari and Robin 2) and standard setters (EU, FASB, IASB) have shown recently a rising interest in the impact of different economic environments and GAAP regimes on the properties of accounting earnings. In particular, existing research has shown that differences in accounting principles and standards alone are not sufficient to explain differences in properties of earnings internationally as the application and enforcement of accounting rules in different institutional settings might set uneven incentives for prepares to select among accounting alternatives inherent in any accounting regime. Therefore, empirical research in a multinational context on earnings management (Leuz, Nanda and Wysocki 21; Land and Lang 22) and expectations guidance (Brown and Higgins 21) has tried to shed light on the impact of managerial flexibility on the properties of accounting earnings. This need for comparative research on earnings management under different institutional and accounting regimes has been underlined in recent times by accounting manipulations in various countries 1 and the global discussion of whether a similar set of global accounting standards, such as the International Financial Reporting Standards (IAS/IFRS), will lead to similar properties of reported earnings, despite the potentially remaining different incentives for earnings management across countries 2. In this context, the interesting finding of systematic patterns in the distribution of reported earnings under US-GAAP around specific targets are consistent with the view that managers have several benchmarks they seek to beat, such as to report (1) positive earnings, (2) an increase in earnings relative to the last reporting period and (3) consensus analyst earnings forecasts which proxy for the stock markets expectation (Burgstahler and Dichev 1997; Degeorge, Patel and Zeckhauser 1999). Although those discontinuities are commonly interpreted as evidence of earnings management and are used to compare its magnitude between subgroups, only further concurrent research has shown on top that executives really exercise their discretion by using accruals management or expectations guidance in the case of analyst forecasts to achieve such earnings targets (Gore, Pope and Singh 21; Beaver, McNichols and Nelson 2). In sum, this doublehurdle approach of detecting managerial discretion which relies on (at least) 2 independent methodologies can be considered the current state-of-the-art approach of researching earnings management (McNichols 2, Dechow, Richardson and Tuna 23). Earnings management in general and benchmark beating in particular has been researched extensively only for the US capital market; comparable empirical evidence for other central markets 3

4 such as in Asia or Continental Europe is still scarce in the international literature (see Ronen and Sadan 1981, Schipper 1989, Dechow and Skinner 2, Fields, Lys and Vincent 21 for excellent surveys). Notable exceptions are the contemporary comparative studies by Leuz, Nanda and Wysocki (23), Brown and Higgins (21) and Land and Lang (22) which study properties of earnings in relation to earnings management using global data-bases and covering all large countries across the world. The work by Leuz, Nanda and Wysocki (23) is the closest to our study: They present evidence using various measures for earnings management, including the positive earnings target as one among those measures. However, they focus more on differences in investor protection across countries and its relation to the properties of earnings, rather than on a detailed description of the benchmark beating phenomenon itself. Brown and Higgins (21) present evidence on the distribution of forecast errors relative to targets globally and set their findings in relation to the U.S. discussion on the importance of meeting or beating analyst expectations in equity markets. Finally, Land and Lang (2) analyze the different quality of earnings worldwide and show a trend of convergence over the nineties. A potential limitation of the research design of all those studies is that they cover as many countries as possible globally, and thus mix countries which are very different in their state of economic development and, by definition of their sample selection, do not cover smaller countries. Also, give their sample size, they cannot control for differences in the characteristics of the companies they cover across country subsamples. Methodologically, those studies rely on the discussion of properties of earnings alone in their final judgment on the (relative) magnitude of earnings management, without further demonstrating actual managerial intend around the targets on top as suggested by the double-hurdle approach. In Europe, recent contributions by Peasnell, Pope and Young (21) and Gore, Pope and Singh (21) not only show that earnings targets are also important in the U.K context, but further demonstrate that working capital discretionary accruals are used to achieve such targets. However, the U.K. is considered the most similar European country to the U.S. in its institutional setting, as key features of the two countries are quite comparable (e.g. common law legal system, equity markets focus, dual enforcement, Anglo-Saxon, etc.; see Ball, Kothari and Robin 2). Thus, those studies have not researched the influence of further diverging institutional settings from the U.S. system on earnings management in general and benchmark beating in particular (see Leuz, Nanda and Wysocki 21). Our study presents new evidence by examining the benchmark beating phenomenon and its relation to the management of earnings and expectations into detail for a broad European sample, covering all 14 large EU member states. The context of the EU is particular interesting because the European Commission has tried to achieve a certain level of accounting harmonization through the 4th and 7th directive, although various other institutional features such as the legal system, 4

5 enforcement, corporate financing or ownership structures remain particularly heterogeneous across countries, despite their similarities in economic and social conditions (see McLeay 2). Our specific design thus allows to analyze earnings management not only on a global EU level which is comparable in sample size to the (former) U.S. studies, but also on a comparative level among individual countries or particular groupings of such with different accounting origins or accounting standards in place. These findings do not only demonstrate the (comparative) extent of benchmark beating, but can also be interpreted as a measure for the level of comparability achieved so far in the European Union though harmonization. In addition, we try to really document that managerial discretion and executive interference in the financial reporting process at least partially is an explanation for the discontinuities potentially to be detected in Europe as suggested by the double-hurdle approach. We therefore analyze the management of earnings, both as the long-term smoothing of earnings and the short-term management of accruals, as well as the management of expectations in the case of analyst consensus forecasts. In light of the recent EU decision to further move to standardization in Europe by requiring listed companies in the EU to apply International Financial Reporting Standards (IAS/IFRS) beginning 25/27, we finally analyze an interesting subgroup of Continental European firms which have already (pre)adopted the high transparency Anglo-American accounting standards IAS/IFRS or US- GAAP, but which continue to operate under an institutional environment with supposingly low transparency, low outside investor protection and therefore potentially high incentives for earnings management (Leuz, Nanda and Wysocki 23). Those findings can be particularly interesting on the background of the discussion on whether a similar application of IFRS can be achieved in the future across Europe and elsewhere without further shaping the regulatory oversight. 3 Similar to prior research in the U.S. and the U.K., we find that more than expected firms (1) report small positive earnings, (2) report small positive earnings increases and (3) have zero or small positive forecast errors in our total EU sample. Those results are presented both graphically in the form of histograms of the pooled cross-sectional distributions of scaled earnings (changes, forecast errors) as well as with popular test statistics. Interestingly, these discontinuities are much more pronounced in our European sample as compared to the U.S. or pure U.K. evidence. When comparing the phenomenon within the EU by analyzing different (stylized) accounting origin regimes as well as individual countries, we do find that its magnitude differs substantially: The avoidance of losses or decreasing earnings is most heavily practiced Continental European countries, in particular in the German accounting origin countries, and much more pronounced than in the British accounting origin countries. Those results are shown to be robust for the application of 5

6 different deflators when scaling the earnings variables as well as for using matched firm subsamples. Our results thus confirm and extend prior research which uses different research methods and smaller or limited EU (sub)samples (Langendijk and van Praag 1996; Leuz, Nanda and Wysocki 21). The substantial difference in the distribution of earnings and earnings changes also underlines the still remaining differences in the properties of earnings within the EU up to date despite the EU efforts to harmonize accounting. This can therefore be considered an argument to move from harmonization to standardization if comparability of accounting figures across EU countries is a (political) goal to be achieved. With respect to the issue whether managers really utilize their discretion in the reporting process to achieve the earnings targets analyzed, we study the practice of (1) income smoothing, (2) accruals management, and (3) the management of expectations. We argue (1) that there is supposingly a positive relationship between the level of smoothing and the magnitude of benchmark beating. We interpret the high degree of income smoothing over time which results in enough slack in the firms books (and thus earnings flexibility in bad times) as one reason why (particularly Continental) European companies are so successful in achieving important earnings targets. We further (2) discover that our estimates for discretionary accruals have both the effect of significantly increasing the frequency of firms achieving earnings targets as well as to smooth income. The distributions of earnings, earnings changes and forecast errors before discretionary accruals show less small positive surpluses and are more spread overall. However, we fail to detect unusually high positive discretionary accruals in the areas slightly above the target areas. (3) In the case of analyst forecasts, firms seem to engage in expectations management by walking down consensus estimates: We provide evidence of a switch from upward-biased to nearly unbiased forecasts of annual earnings and, in particular, a (sharp) increase in the ratio of small positive to small negative surprises as the announcement date approaches over the forecast horizon. Finally, for our mostly Continental European subsample of firms which already apply IAS/IFRS or US-GAAP, we find the same severe discontinuities as under our local GAAP groups. To the extend that such discontinuity really proxy for earnings management, this would suggest that even under a more strict and transparent, Anglo-American accounting regime, EU firms would practice earnings management to the same (high) extend as under the more discrete and flexible old local GAAP before. Therefore, those initial results would suggest that the incentives for earnings management in a given environment are important factors in determining the properties of earnings and have to be considered along with the accounting standards to be applied. A further strengthening of the regulatory oversight and enforcement in the EU would therefore be necessary in order to achieve comparability across the EU. 6

7 In sum, this paper contributes to the growing international accounting literature by providing detailed multinational evidence on the distribution of reported earning relative to targets across and within 14 countries of the European Union (EU) by addressing the questions of (1) whether the phenomenon of discontinuities in the distributions is also found in the EU context, (2) how the magnitude of difference (and thus level of harmonization achieved) compares among EU accounting origins, member countries or accounting standards applied and (3) how managers of EU listed firms use their discretion to accomplish earnings targets. The remainder of the paper is organized as follows: Section II reviews the related literature and develops the empirical predictions and hypothesis; Section III describes the sample and our test methodologies; Section IV summarizes the results and our conclusions are in section V. 7

8 2. THE DISTRIBUTION OF EARNINGS AND MANAGERIAL DISCRETION Earnings are a key summary measure of firm performance and the output of a specific accounting process which gives prepares some flexibility in reporting their numbers. Under the assumption that financial statement users are (at least partially) unable to completely detect and reverse the effects of such flexibility, managers can exploit their discretion to manage earnings in a desired direction. Incentives for such managerial intervention in the financial reporting process might result from multiple reasons, typically categorized into political, contractual, and valuation motivations (see e.g. Dechow and Skinner 2). Such incentives to exercise discretion in the reporting process might also apply around specific earnings benchmarks which can be important in contracting or firm valuation. Targets based on accounting numbers in general and earnings (or earnings components) in particular play an important role in debt contracts and covenants as well as in managerial compensation plans (e.g. Matsungana and Park 21). Furthermore, recent research in the U.S. has documented not only that the failure to meet or beat consensus earnings forecasts may result in a (large) decline of the reporting firm s stock price (e.g. Burgstahler, Kinney and Martin 21; Bartov, Givoly and Hayn 21), but also that the U.S. equity market assigns a valuation premium on firms with a constant and positively growing earnings trend over time (Barth, Eliott and Finn 1999). Such findings are not surprising as investment practitioners have for decades stressed the importance of positive and growing earnings for a firm s valuation. The large U.S. investment intermediary Standard & Poor s for example considers an increase in earnings relative to last year s earnings, labeled earnings stability, a key criteria in its stock Quality Rank Model which is used for equity investment purposes. 4 Empirical research in the U.S. has provided evidence about systematic patterns in the distribution of earnings relative to specific earnings targets, consistent with the view that managers have several benchmarks they seek to beat such as (1) reporting positive earnings, (2) reporting an increase in earnings relative to the last period and (3) meeting or beating consensus analyst earnings forecasts which proxy for the capital market s earnings expectations (Hayn 1995; Burgstahler and Dichev 1997; Burgstahler and Eames 1999; Brown 1998; Degeorge, Patel and Zeckhauser 1999). Although there are a number of possible other earnings targets potentially relevant for a manager choice when deciding on the firm s accounting policy such as e.g. an industry average or a competitor s income, the literature has mainly focused up to date on these three benchmarks. Also, popular press reports in the print media or the television about a firm s reported financial numbers almost uniformingly comment on whether a firm has earned a profit or loss in the reporting period (usually even in the article s headline, positive earnings target ), has increased or decreased its numbers relative to last 8

9 period ( increase in earnings target ) and has met or beaten the stock market s expectations ( meet or beat consensus forecasts ). Benchmark beating has only been researched extensively for the U.S. and the U.K. capital market so far. Further evidence for other central markets such as in Asia or Continental Europe is still scarce: Leuz, Nanda and Wysocki (23) present first multinational results on the positive earnings target by comparing the ratio of small positive to small negative earnings. However, their focus is more on explaining institutional factors which lead to earnings management in general rather than on exploring the benchmark beating phenomenon into more detail themselves. Brown and Higgins (21) concentrate on the forecast error target and the large European countries only. There is no further evidence on the earnings changes target. The European Union is an ideal laboratory for a comparative analysis on earnings targets for two reasons: First, it has a broad common political, social and economic base, consists of fully developed countries with major capital markets and overall across its countries is large enough to compare in sample size with the existing U.S. studies. Second, the substantial variety in individual capital markets, institutional settings or legal origins of the member states makes it particularly interesting to analyze earnings targets within the EU as well. Last, the trend of listed companies reporting their results under IAS/IFRS or US-GAAP emerging in various Continental European countries such as Austria and Germany allows analyzing the effect of an application of high-transparency, Anglo- American accounting standards outside the U.S. financial system. (I) Given the fact that (particularly Continental) European firms are generally regarded in textbooks of comparative international accounting as being discrete and engaging in earnings management even to a higher extend than their U.S. counterparts, we expect to find discontinuities around the three earnings targets under investigation in the EU similar to those reported for the U.S. or the U.K. alone. Accordingly, we test whether we find the same discontinuities in the EU and compare our results to the magnitude of the discontinuity for a benchmarking U.S. sample. Therefore, we test whether the frequency of small negative earnings (earnings changes, forecast errors) is lower than expected and the frequency of small positive earnings (earnings changes, forecast errors) is higher than expected under a smooth distribution. Although the European Commission has tried to achieve a certain level of accounting harmonization within the European Union (EU) through their 4 th and 7 th directives, various institutional features in the financial reporting regimes such as the legal system, enforcement, ownership structures or the relative importance of equity vs. debt markets remain particularly heterogeneous among member countries (Leuz, Nanda and Wysocki 23). In addition, the underlying various accounting origins and traditions in Europe have resulted in still diverging de 9

10 facto national accounting practices in the EU despite the binding global EU accounting regulations (see McLeay 2). Thus, this set-up allows analyzing benchmark beating not only on a total EU level, but also on a comparative level among counties within the EU with different accounting origins and institutional backgrounds. As a result, one can assess whether (or not) significant differences in the properties of accounting earnings still exist among various European accounting origins and thus measure the level of comparability achieved so far in the EU through the harmonization strategy. The literature generally broadly distinguishes between the British and the Continental European accounting origin traditions. Within the Continental European group, the French, Scandinavian and German origin groups are usually separated (see e.g. Nobes and Parker 1998). The British and German origin countries are often described as forming the boundaries of financial reporting traditions (Joos and Lang 1994). With respect to the hypothesis of whether one accounting regime (or country) vs. another will ceteris paribus engage (relatively) more into benchmark beating and/or earnings management depends on two different aspects: (1) the relative incentives and (2) the relative opportunities and constraints for prepares in the financial reporting system to beat a particular earnings target. Although many potential incentives for achieving e.g. positive earnings or earnings changes have been put forward, it is still an open issue to judge their validity absolutely in an absence of a complete understanding of the benchmark beating phenomenon itself up to date (Beaver, Nelson and McNichols 23). Even more troublesome seems to be to weight the explanations put forward relatively across countries and regimes as various contracting and/or valuation incentives might contribute differently multinationally (e.g. valuation vs. taxation motives). The same applies to the opportunities and constraints side for benchmark beating, given the various alternatives and judgments involved in any accounting regime. Accordingly, we leave the question of the relative magnitude of benchmark beating as an empirical issue and analyze the 4 accounting origin subgroups as well as the 14 EU countries individually without formulation a priori hypothesis. (II) Furthermore, the contemporary literature has indicated that the discontinuities around targets alone are not sufficient conditions to prove managerial discretion and intervention in the financial reporting process (Gore, Pope and Singh 21; Beaver, McNichols and Nelson 2; Dechow, Richardson and Tuna 23). Consequently, we investigate around the discontinuities the use of discretionary tools managers are given within the financial reporting process to beat earnings targets: Such tools include (1) the (long-term) strategic management of earnings through smoothing of earnings over various reporting periods which results in enough slack in the books and increases the probability of achieving earnings targets over time, (2) the (short-term) management of earnings 1

11 through accruals management for a single reporting period and (3) the management of expectations through analyst guidance in the case of market earnings expectations. (1) In the case of strategic, long-term earnings management in the form of income smoothing, managers save earnings through accruals management for the future when the performance for a financial reporting period is particularly good. This strategy increases the slack in the books and thus increases the overall flexibility preparers have for a given reporting period. Applied to a firm which would fall from their performance slightly below an earnings target, a manager who has saved earnings from the years before will have a higher probability to achieve earnings above the target after accruals management. Therefore, we expect to find a positive relationship between the level of smoothing and the ratio of firms which are slightly above relative to slightly below the target across reporting regimes. This reasoning can be considered in the same line of thought as Beaver, McNichols and Nelson (2) who have been the first to study earnings management using the distribution of earnings not only for the analysis around a specific target, but also to exploit the information in the upper and lower tails of the distribution to analyze for smoothing. For the property-casualty insurance industry, they show that loss reserve accruals are used not only to report positive earnings, but also to smooth income over the entire distribution. (2) In the case of short-term earnings management, single period accruals management, DeFond and Park (1999), Pope, Gore and Singh (21) and Dechow, Richardson and Tuna (23) use a discretionary accruals approach to test for earnings management around earnings targets. The first two studies compare the distributions and test-statistics of reported earnings (changes, forecast errors) relative to non-discretionary earnings (changes, forecast errors). They show that earnings management as proxied by the discretionary accruals can explain part of the phenomenon, as, in fact, the irregularities around the specific earnings targets analyzed are not present in the nondiscretionary earnings, earnings changes or forecast error sub samples. More recently, however, Dechow, Richardson and Tuna (23) cannot detect an unusually high level of positive discretionary accruals in the areas slightly above the benchmark. Overall in this line of research, both the total discretionary accruals models (Jones 1991; Dechow, Richardson and Tuna 2) as well as working capital versions in the cross-sectional form (Peasnell, Pope and Young 21; Pope, Gore and Singh 21) have been employed. Wiedmann and Marquardt (21) and Moehrle (22) also present evidence for the use of specific earnings components such as special & extraordinary items or specific accruals, respectively. (3) In the case of consensus analyst earnings forecasts which proxy for the stock market s earnings expectations, the target itself is not a fixed number, but it is potentially possible to influence 11

12 not only the actual earnings figure (by earnings management) relative to a given target, but also the target itself by managerial intervention through a specific investor s communication policy ( expectations management ). In fact, the number and size of investor relations (IR) departments and related investor communications has expended sharply both in the U.S. and oversees. Such management of expectations, also labeled earnings guidance, can take various forms such as, e.g. individual and conference calls with (key) analysts, press releases or public announcements (Burgstahler and Eames 2) or by effective disclosure of bad news in advance of the formal earnings announcement date (Soffler et. al. 1997; Skinner 1997). The logic behind the surprising idea that company executives would like analysts to be more pessimistic about the firm s future income prospects over the horizon of the reporting year is the assumption that a walk-down of forecasts closer to the announcement period plus the meeting or beating of expectations at the reporting date would have a superior effect on the firm s stock price rather than a strategy which leaves overoptimistic forecasts over the forecast horizon and fails to meet or beat the markets expectations when the annual or interim figures are disclosed. Theoretically, the reward and therefore incentives of such a strategy is highly questionable in an fully information efficient market, however, evidence so far in the U.S. shows that managers do walk analyst forecasts down over the forecast horizon (Richardson, Teoh and Wysocki 21; Mastsumoto 22) and such a strategy is even appreciated by a market premium on the companies stock value (Soffer, Thiagarajan and Walther 2). In order to test for expectations management, several potential approaches have been used in the U.S.: Cotter, Tuna and Wysocki 22 document guidance directly by showing the link between management earnings projections and security analyst s reactions to this specific investor relations communication mechanism. On the other hand, Matsumoto (22) and Richardson, Teoh and Wysocki (21) provide indirect evidence that firms appear to be guiding analysts toward beatable earnings forecasts. By examining the monthly path of forecast errors over the forecast horizon, they document a shift from systematic analyst optimism at the beginning of the fiscal year which turns into a systematic pessimism shortly before the earnings announcement. Even more, Burgstahler and Eames (1999) document that the particularly in this respect important ratio of small positive relative to small negative forecast errors increases over the forecast horizon. The advantage of this indirect approach is that the net effect of all IR activities on analyst forecasts can be studied, including nonpublic or non-available IR mechanisms such as telephone conferences or informal tasks. In Europe, large sample size research on IR in general and earnings guidance in special is missing, largely because databases on such activities are virtually non-existing. Research undertaken in some European countries using (small sample size) interview, survey or experimental techniques indicate that analyst guidance does exist. Even more, since the mid 9s the capital market focus of European companies has sharply increased. Such a focus has resulted in the creation of new stock market 12

13 segments in Europe (e.g. the German Neuer Markt or the Pan-European EASDAQ ), an increase in the level of disclosure or the possibility for stock corporations to report their consolidated results using international accounting standards such as IAS or US-GAAP. At the individual company level, this development has also resulted in a constant increase in IR-related activities, along with a rising importance of equity capital for the listed European companies. In summary, this provides managers in Europe with both the tools and incentives to exercise expectations management. We therefore test whether the optimistic bias in analyst forecasts at beginning of the year turns into pessimistic bias around the announcement date and whether the percentage of firm-years with small positive forecast errors relative to small negative forecast errors increases as the forecast horizon approaches the announcement date. (III) Last, as the number of EU firms which report their consolidated financial results using IAS/IFRS or US-GAAP has increased sharply over the late 9s especially in the Continental European countries Austria and Germany which legalized such behavior after 1998, we can analyze the effect of an application of such high-transparency Anglo-American accounting standards outside the U.S. financial system. This setting is particularly interesting on the basis of the discussion whether accounting standards alone, without shaping the institutional background, would lead to similar properties of earnings globally: The Anglo-American standards taken alone would predict high-transparency and low earnings management (and thus benchmark beating), whereas the codelaw, low outside investor protection and low enforcement financial system in Austria and Germany would predict a high level of earnings management (and benchmark beating). Again, we investigate this issue only empirically. 3. DATA AND METHODOLOGY (i) Data and Definition of Variables We use the full universe of quoted companies in the EU, covering incorporation in the following 14 EU member states: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, Spain, Sweden, and the U.K. We excluded small countries for which there is not sufficient data as well as non-eu European countries as they were not (fully) effected by the accounting harmonization process of the 4th and 7th directives. We employ the maximum data available for our firms of interest from two sources: Annual financial statement data is obtained from the Thomson Financial Analytics - Worldscope database, analyst forecasts data comes from the Thomson Financial - I/B/E/S Summary Tape 3/23. We cover data from 1986 to 21 over a period of 16 years thus cover data across a range of different economic cycles. From the same sources comes the data on our U.S. sample, which serves as a natural reference benchmark for our results. 13

14 In absence of a theory it is unclear which earnings measure managers try to manipulate. 5 For the positive earnings variable (E), we report results on net income, as earnings after extraordinary items, in this paper. 6 This is consistent with Burgstahler and Dichev (1997), but contrast with other studies which examine earnings before extraordinary items (Degeorge, Patel and Zeckhauser 1999; Gore, Pope and Singh 21). We focus on net income (WS #1651) as the bottom-line number attracts a wide interest both from a firm s stakeholders and the investment community. In addition, the definition and differentiation of special and/or extraordinary items not only differ among EU member states, but has also changed in some countries during our sample period. 7 We further use change in net income (WS #1651 t WS #1651 t-1 ) for our earnings change variable ( E). For the forecast error variable (FE), we use the most recent consensus annual earnings forecast before the announcement month ( Forecast Period End Date ) as taken from the Institutional Brokers Estimate System (I/B/E/S). In order to ensure the consistency between forecasted and actual number, we take the actual earnings figure from the same source. 8 The forecast error reported here is defined as actual earnings minus mean forecast. We normalize all variables by deflation since we pool data across time and firms which vary widely in size and share price. We primarily use opening total assets (TA; WS #2999) as deflator of all variables for consistency with the deflators commonly used in the discretionary accruals models described later. 9 However, we also report results using opening year market value of equity (MV; WS YrEndMarketCap ) and previous year sales (SA; WS #11) as alternative deflators. We winsorize all variables, per country, by the top and bottom 1% of the distribution in order to cancel out data errors. 1 When comparing results across various reporting regimes, we use the following schema: We first compare the total EU sample to the benchmarking U.S. Then, we divide the EU sample into various subsamples: We start by comparing the Anglo-Saxon British accounting origin countries which are most comparable to the U.S. accounting regime to the remaining Continental European countries. We next use a classification schema to further group our Continental European countries into 3 stylized accounting origin sub-samples, labeled French, German and Scandinavian accounting origin. 11 Although there is no common agreement on a schema to correctly cluster such countries (depending on the relative weighting of various institutional characteristics in each particular country, see McLeay 2), we rely on the existing comparative accounting literature (Nobes and Parker 2; Choi, Frost and Meek 1999) and are inspired by the latest classification schema by Nobes 1998 to assign our countries of analysis into accounting origin groups: (1) British Accounting Origin (Ireland, Netherlands, the U.K.) (2) French Accounting Origin (Belgium, France, Greece, Portugal, Spain) 14

15 (3) German Accounting Origin (Austria, Germany, Italy) (4) Scandinavian Accounting Origin (Denmark, Finland, Sweden) In addition, since we want to analyze differences across countries given their domestic accounting standards, we exclude from the country analysis European firms which report their results under IAS or US-GAAP. These firms form a subgroup labeled Anglo-American reporting. In order to control for potentially interfering differences in firm characteristics across our subgroups of analysis, we further match companies for our key groups of analysis (U.S., British and German origin) according to industry (SIC-code), size (average sales) and age (time series of variables in the database). Those matched sample results are reported next. Last, we split the EU sample into the 14 countries individually. Finally, table 1 presents the descriptive statistics for our (scaled) variables E (Panel A), E (Panel B) and FE (Panel C) for each subgroup described earlier. (ii) Methodology (a) Detecting Discontinuities around Targets Our hypotheses predict earnings management driven discontinuities in distributions of our earnings variables. We use two approaches to test our hypothesis: (1) Histograms: We graphically plot the univariate distribution of (scaled) E, E and FE. We focus on the on the area around the target and compare the discontinuity in the distribution using the graphical approach to detect differences in the distribution among countries (Heiler and Michels 1994, p. 8). (2) Test-Statistics: We perform formal test-statistics developed by Burgstahler and Dichev (1997) to examine the significance of any discontinuities around the target area (BD-Test Statistics). When comparing the relative magnitude of this test-statistic, one has to take into account differing sample sizes of subgroups. Since, ceteris paribus, the standardized difference of the BD statistics increases approximately linearly with the square root of sample size (see Burgstahler and Eames 1999), we adjust for this difference for country or accounting origin subgroup i by an adjustment factor relative to our base sample country of sqrt(n i )/sqrt(n base ). The U.K., as the biggest sample size country, forms the base for this adjustment. An important issue in univariate empirical distributions and the test statistics is the choice of class width for the variable. A common suggestion in standard statistics textbooks calls for a bin width of 2(IQR)n -1/3, where IQR is the interquartile range of the variable and n is the number of observations 15

16 (Heiler and Michels 1994). We use this rule of the thumb and round the class width calculated to the next logical even number. We also compare our bin size to those used by other studies in this field. 12 (b) Detecting Managerial Intervention in Financial Reporting to Achieve Targets Even though a discontinuity around targets might indicate the practice of managerial intervention in the financial reporting process, it is no sufficient condition to prove earnings- and/or expectations management. Therefore, we use a double-hurdle approach and analyze the practice of accruals management around the target area on top of the discontinuity analysis. Such an approach, relying on (at least) two different methodologies, has been used most recently as a response to the problem of adequately detecting earnings management (Beaver et. al. 21; Dechow et. al. 23) and can be considered the current state-of-the-art approach of researching earnings management (McNichols 21). In section (1) to (3), we explain the methodologies which try to detect one- or multiperiod earnings management employed in our study. (1) Detecting (multi-period) Earnings Management - Income Smoothing Studies on earnings management in a multiperiod context have concentrated on detecting income smoothing over time. While there has been an ongoing discussion for the past 3 years on how to correctly quantify an income smoothing strategy (e.g. Ronen and Sadan 1981), we use a measure put forward by Eckel (1981) and modified by Albrecht/Richardson (199) which classifies companies into two binary groups of smoothers and non-smoothers. This measure has been used in most studies on comparative income smoothing (see Ashari, Tan and Wong, 1994, for two Asian or Langendijk and van Praag, 1996, for 4 selected European countries). The index, which basically compares the variability of sales (which measure the underlying real economic degree of smoothness in a firm s activities) to the variability of reported income over time, is considered a firm approximation of the extend of income smoothing in various countries. A firm i is classified as income smoother, if: CV( E i ) < CV( Sales i ) (1) where CV(x) = coefficient of variation (x) = Std(x)/Mean(x), E = annual change in earnings, sales = annual change in sales/net revenues. Since smoothing is a multiperiod concept, a crucial issue in any income smoothing study is the number of successive periods to be studied. The literature uses different periods: We use 3 consecutive periods to classify income smoothers according to the modified Eckel (1981)-Index. We calculate the index for each firm individually and then calculate the overall magnitude of smoothing in Europe and its relative scale across subgroups

17 More recent measures for the relative magnitude of smoothing include the correlation-approach used by Myers and Skinner 1999 and exploited in comparative international research by Leuz et. al. 23. We use the contemporaneous correlation of the change in cash flow from operations to the change in total accruals (as defined in equation (2)) to compare the relative magnitude of smoothing across countries: Managers can use their discretion to balance real economic shocks to a firm s operating cash flows by adjusting accruals in the opposite direction. 14 ρ( Acc, CFO) (2) Finally, an approach we suggest here is to exploit the information content in the overall distribution of the earnings and earnings changes samples in order to compare the spread of those variables. If smoothing exists in one country or accounting origin vs. another to a larger extend, the overall distribution of both earnings and earnings changes should be less spread around their means. This analysis can be verified both graphically and by using (simple) statistical measures such as the interquartile range or the density distribution which summarizes the spread information. We thus report the interquatile range (IQR) of the (scaled) earnings and earnings changes variables for the various subgroups. All the presented methods on detecting income smoothing have one disadvantage in common they do not show directly (as opposed to the discretionary accruals models) how managerial intention and intervention in financial reporting leads to smoothing earnings over time: The accruals process as the heart of accounting itself buffers (through the matching principle) fluctuations in cash flows over time and thus smoothes reported earnings even without actual earnings management (see Dechow 1994). If thus accounting systems smooth income to a varying degree through their regular accruals process (e.g. completed contract vs. percentage-of-completion method), those methods to detect earnings management in a multiperiod context will show a differing degree of smoothing across borders even in the absence of earnings management. We try to balance this limitation by the application of a number independent approaches described above. (2) Detecting (one-period) Earnings Management Existing studies on earnings management use numerous methodologies to detect the influence of accounting discretion on earnings for a given reporting period. In relation to earnings targets, researchers have focused on investigating specific individual accruals (Beaver et. al. 2, Moehrle 21, Marquardt and Wiedman 21) or discretionary accruals models, which try to evaluate the total effect of managerial intervention on reported earnings (Peasnell et. al. 21; Gore et. al. 21; Dechow et. al. 23). Specific accruals may represent only a small portion of the discretion component of earnings and therefore may fail to reflect earnings management and also to reveal its 17

18 total extent. In addition, individual accruals might be used differently among countries (e.g. discretionary reserves). Therefore, we use the more comprehensive discretionary accruals models, which try to estimate the total effect of accruals management. Building on the original Jones (1991) model, there have been a number of proposals for such models which assume for normal accruals (or non-discretionary, NDACC) a linear relationship to specific factors or drivers. The components of reported accruals that cannot be explained by such drivers are denoted as discretionary or abnormal (DACC). We use the cross-sectional versions of such models as they have outperformed their time-series counterparts in recent evaluations both in the U.S. (Bartov, Gul and Tsui 21) and the U.K. (Peasnell et. al. 2). Even though discretionary models generally lack power, they appear to be well specified (Peasnell et. al. 2). We use both a working capital accruals version [see equations (1), (3)] and a total accruals version [(2), (4)]. Both models have unique attractive features: On the one hand, Young (1999) argues that working capital manipulations are less visible and therefore more attractive for earnings management. On the other hand, using only working capital accruals would neglect other central accruals components in the Continental European context, in particular the important changes in discretionary, non-equity reserves. First, due to a lack of comparable data on cash flow statements, we use the balance sheet approach to calculate total accruals (TACC) and total working capital accruals (WCACC) per year t. We approximate accruals as in the former U.S. literature (Dechow 1994, Dechow, Hutton and Sloan 1995), but correctly adjusted for changes in provisions in the European context 15 : WCACC t = ( CA t - Cash t ) ( CL t STD t TP t ), (1) TACC t = ( CA t - Cash t ) ( CL t STD t TP t ) DEP t Provisions t, (2) where CA = change in current assets (WS # 221) Cash = change in cash and cash equivalents (WS # 21) CL = change in current liabilities (WS # 311) STD = change in short-term debt included in current liabilities (WS # 351) TP = change in tax payable, if available (WS # 363) DEP = depreciation, depletion & amortization expense (WS # 1151) Provisions = change in provisions for risks & charges (WS # 326). Second, we estimate cross-sectional OLS regressions for each two-digit SIC industry-year, using all firm-years with a minimum of 6 observations per regression. 16 We use pooled European data in order to estimate the parameters of regressions (3) and (4) even though we are aware of the fact that differences in the accruals process internationally might cause interference. However, the estimation per country in addition with the requirements of 6 observations per two-digit firm-year would make 18