Journal of Contemporary Accounting & Economics

Transcription

1 Journal of Contemporary Accounting & Economics 7 (2011) 1 17 Contents lists available at ScienceDirect Journal of Contemporary Accounting & Economics journal homepage: The impact of IFRS adoption on the value relevance of book value and earnings Peter Clarkson a,b,, J. Douglas Hanna c, Gordon D. Richardson d, Rex Thompson c a University of Queensland, Brisbane 4072, Australia b Business Administration, Simon Fraser University, Canada V5A 1S6 c Edwin L. Cox School of Business, Southern Methodist University, Dallas, TX 75275, USA d Joseph L. Rotman School of Management, University of Toronto, Toronto, Ontario, Canada M5S 3E6 article info abstract Article history: Received 20 August 2009 Revised 28 September 2010 Accepted 29 September 2010 Available online 1 April 2011 Keywords: IFRS Value Relevance Nonlinear Pricing Model In this study, we investigate the impact of IFRS adoption in Europe and Australia on the relevance of book value and earnings for equity valuation. Using a sample of 3488 firms that initially adopted International Financial Reporting Standards (IFRS) in 2005, we are able to compare the figures originally reported for the 2004 fiscal years to the IFRS figures that were provided in 2005 as the 2004 IFRS comparative figures. As part of the inquiry, we introduce a cross-product term, equal to the product of EPS and BVPS, into the traditional linear pricing models. The estimated coefficient on the cross-product term is statistically significant and negative, as theory suggests in the presence of important nonlinearities. Further, there is increased non-linearity in the data subsequent to IFRS adoption, with the increase being most pronounced for firms in Common Law countries. With non-linear effects controlled for, there is no observed change in price relevance for firms in either Code Law or Common Law countries, contradicting the results from the linear pricing models. The results also suggest that the distribution of measurement errors becomes more similar across Code Law and Common Law countries after the adoption of IFRS, removing one difference between these groups. Thus, IFRS enhances comparability, an inference that would not be possible had we confined the analysis only to linear pricing models. Ó 2011 Elsevier Ltd. All rights reserved. 1. Introduction In this study, we investigate the impact of International Financial Reporting Standards (IFRS) adoption on the relevance of book value and earnings for equity valuation. The 2005 switchover to IFRS in Europe and Australia provides a natural quasiexperimental setting. The 2005 filings of European and Australian public companies contain restated 2004 GAAP numbers in accordance with IFRS which can be compared to the former country level GAAP numbers reported in the original 2004 filings. This allows the measurement of performance for the 2004 fiscal year using both the original local GAAP and the new IAS figures. Each firm serves as its own control and the time period studied is the same for both sets of figures. 1 Hung and Subramanyam (2007) refer to this same firm year research design as a powerful one since it controls for cross sectional and time series differences in the sample firms. The problem of identifying the effects of changes in accounting standards per se has plagued the IFRS mandatory adoption literature because control groups are difficult to find for a country wide change for an extensive discussion, see Daske et al. (2008) and Hail et al. (2009). This identification problem is impor- Corresponding author at: University of Queensland, Brisbane 4072, Australia. addresses: p.clarkson@business.uq.edu.au (P. Clarkson), dhanna@cox.smu.edu (J.D. Hanna), gordon.richardson@rotman.utoronto.ca (G.D. Richardson), rex@cox.smu.edu (R. Thompson). 1 For some firms, depending upon the firm s fiscal year-end and the effective date for the standard in the firm s home country, the switchover alternatively occurred in /$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi: /j.jcae

2 2 P. Clarkson et al. / Journal of Contemporary Accounting & Economics 7 (2011) 1 17 tant since, as Hail et al. point out, financial reporting quality depends not only on accounting standards but also on a firm s reporting incentives and its country s institutional framework. Our same firm year design holds incentives and institutional frameworks constant and turns one dial only, namely, standards. Thus, we solve the identification problem as we can isolate the capital market effects due to changes to IFRS accounting standards. The value relevance of aggregate book value and earnings is a natural place to look for the impact of IFRS adoption on financial reporting quality given the paramount role of equity valuation in the IFRS conceptual framework. We view book value and earnings as summary descriptors of value and explore how value relevance changes subsequent to the adoption of IFRS. Our value relevance perspective is not the only one for examining the capital market benefits of IFRS adoption and concurrent research explores other benefits of IFRS adoption (for example, Daske et al. (2008) examine changes in cost of capital). Our research complements the cost of capital evidence of Daske et al. in the following way. Greater value relevance is one dimension of accrual quality (Francis et al., 2004) and higher accrual quality can translate into a lower cost of capital, especially when firms raise equity financing in distant countries where investors are not insiders in the sense described by Ball et al. (2000). Put simply, a greater value relevance of book value and earnings means investors need to rely less on other information (Ohlson, 1995). Hence, firm information risk declines. From a methodological perspective, we begin with the traditional linear valuation model to evaluate the improvement in fit for equity valuation as the GAAP regime changes. These are our benchmark models, since they represent current operating procedure for analyses in the spirit of Ball and Brown (1968). We then supplement the linear model with the addition of a cross-product term, the product of book value and earnings. While the true valuation model relating assets and cash flow to equity value is proportional, book values and earnings measure assets and cash flow with some degree of measurement error. One important source of measurement error is that contributed by a specific set of accounting standards. Consider, for example, the impact of Code Law conservatism prior to IFRS adoption. This downward bias in book and earnings is a source of measurement error induced by accounting standards (see Beatty et al., 1999). As a second example, consider the move within IFRS towards more fair value accounting. As pointed out by Ohlson (2009), the ultimate use of fair value accounting is perfect mark-to-market accounting. In such a regime, earnings measure permanent (i.e., expected ) earnings with considerable measurement error and while earnings perfectly explain returns, earnings do a poor job in explaining the level of price. This is a source of measurement error introduced by IFRS accounting standards. Making reasonable assumptions about the nature of this measurement error, there is support for the use of a non-linear pricing model that includes a cross-product term of some form. Moreover, as will be seen, adding the cross-product term changes the inferences reported in the paper. Our sample data, obtained from Worldscope, consist of 3488 firms from 14 EU countries and Australia (three Common Law and 12 Code Law countries). Our results suggest the following. First, the adoption of IFRS has had a greater impact on the financial statements of Code versus Common Law countries (the mean percentage changes in both earnings per share and book value per share are larger for firms from Code Law countries). Also, for Code Law countries, IFRS increases book value and earnings, consistent with the consensus view expressed in the literature that Code Law GAAP was conservative in nature (for example, see Jermakowicz and Gornik-Tomaszewski, 2006, p. 186). Second, using same firm year valuation tests and linear pricing models, the adoption of IFRS increased absolute pricing errors for Common Law countries, on average, relative to Code Law countries. This result is counterintuitive given that IFRS resulted in a greater change in accounting statements for Code Law countries. Using the non-linear pricing model (termed the Product Model), however, we find no evidence of a reduction in relative price relevance for Common Law countries. This suggests that the deterioration in the linear model s explanatory power for Common Law countries is caused by an increase in non-linearity of the relation between stock prices and accounting information. Changes in the significance of the crossproduct term reveal this to be the case. Aggregating across the 15 countries in our sample, we conclude that, after controlling for nonlinearities, there is no change in the association between book value, earnings and price. Thus, using association as a market-based attribute, IFRS adoption benefits (see Francis et al., 2004; Schipper, 2009) would appear to be limited. However, the Product Model reveals capital market benefits of IFRS adoption when we examine an alternative market-based attribute, namely, the importance of the product term which we view as capturing measurement error. After IFRS adoption, there is no difference in measurement error between Code and Common Law countries, whereas before IFRS there was a difference, which we attribute to prior cross-sectional variation in Code Law conservatism. If one interprets measurement error as one dimension of financial reporting quality, this implies the same financial reporting quality after IFRS for two groups of countries that had different financial reporting quality before IFRS. Thus, IFRS enhances comparability, an inference we can draw from the data while holding enforcement and incentives constant to the analysis. This result points to a benefit of IFRS adoption, an inference that would not have been possible had we confined the analysis to linear models and traditional goodness-of-fit metrics. In terms of contribution, our analyses would seem retrospective in that all European countries and Australia have already switched to IFRS and the historical relevance of former home county GAAP may be less interesting. Nevertheless, there are many countries which have not as yet adopted IFRS, including Canada (which will adopt IFRS in 2011) and the United States which is contemplating IFRS adoption. Our results will be of interest to these countries, and to standard setters and accounting professionals around the globe who are interested in the impact of IFRS conversion. The remainder of the document is organized as follows. In the next section, we present a review of the relevant literature. Section 3 then presents a conceptual case for an alternative pricing model which adds an additional explanatory variable the cross-product term to the traditional OLS regression model. The methodology is discussed in Section 4, followed by the sample data in Section 5. Section 6 then presents the empirical results and Section 7 concludes.

3 P. Clarkson et al. / Journal of Contemporary Accounting & Economics 7 (2011) Table 1 Divergence in GAAP between IFRS and Local GAAP. Common Law gaapdiff1 Code Law gaapdiff1 Australia 4 Belgium 13 Ireland 1 Denmark 11 UK 1 Finland 15 France 12 Germany 11 Greece 17 Holland 4 Italy 12 Norway 7 Portugal 13 Spain 16 Sweden 10 Average 2.00 Average Adapted from Bae et al. (2008). Bae et al. identify 21 reconciling items between IFRS and local GAAP. The measure presented (gaapdiff1) is the number of these differences on a country basis. 2. Literature review Prior to the voluntary adoption of IFRS, international accounting researchers examined the value relevance of book value and earnings using cross sectional designs. For example, Arce and Mora (2002) explore the valuation relevance of book value and earnings in a levels valuation model across eight European countries using data from 1990 to Basing their inferences on model R 2, they conclude that earnings are more relevant than book value in Common Law countries and vice versa for Code Law countries. Using data from 21 countries during , Hung (2001) explores the association between earnings and returns using a perfect foresight returns approach and concludes that the value relevance of financial statements is weaker in Code compared to Common Law countries. 2 The Code versus Common Law partition, which is popular in the literature, is based on the expectation that the valuation role of earnings should be more important in shareholder as opposed to stakeholder economies. As explained by Soderstrom and Sun (2007), legal systems influence accounting standards. In Common Law countries, accounting standards are set by private sector bodies and the purpose of standard setting is to satisfy the information needs of investors. In Code Law countries, standards are influenced by governments and accounting serves as a measure to divide profits between stakeholder groups (see Ball et al., 2000 for an elaboration). Thus, legal system suggests a natural partition for our enquiry regarding IFRS adoption impacts on value relevance. Bae et al. (2008) report a greater divergence between local GAAP and IFRS in Code versus Common Law countries, presumably because Code Law countries adopt an insider economy orientation (Ball et al., 2000). Table 1 reports the Bae et al. divergence findings: for 21 IAS reconciling items making up their initial GAAP differences measure ( gaapdiff1 ), Bae et al. observe an average of two GAAP differences for the three Common Law countries in our sample and 12 GAAP differences for the 12 Code Law countries in our sample. Some studies have examined the valuation relevance of book value and earnings for voluntary (early) adopters of International Accounting Standards (IAS, a precursor to IFRS). The samples have tended to be small and the results mixed. For example, Bartov et al. (2005) examine the effects of IAS versus German GAAP using a linear pricing model. For their sample of 37 IAS adopters, they employ a pre-post design and find an increase in the value relevance of earnings when firms switch to IAS from German GAAP. These findings differ from those of Hung and Subramanyam (2007). Using the same firm year design (as discussed in the Introduction) and a linear price on book value and earnings valuation model, Hung and Subramanyam explore the valuation relevance of restatement differences for 80 voluntary (early) IAS adopters in Germany. They find that the combined value relevance of book value and earnings decreases after the switch to IAS. Similar results are observed by Stergios et al. (2007) for 40 voluntary IAS adopters in Greece. Barth et al. (2008) also focus on voluntary IAS adoption, employing a sample of 327 firms that voluntarily adopted IAS between 1994 and Using a pre-post design and a linear price on book value and earnings valuation model as one of their metrics, they find that the R 2 for the price level model increases from 28% to 40% in the adoption year, relative to the pre adoption year, for IAS adopters. They conclude that accounting quality is higher following IAS adoption. Again, this conclusion is based on voluntary (early) IAS adopters. By focusing on all 2005 adopters in a particular country with data on Worldscope, we employ a much larger sample. Perhaps most importantly, we base inferences on our same firm-year approach and a variety of econometric models. One country at a time mandatory IFRS adoption studies which focus on value relevance are beginning to appear. These include Ahmed and Goodwin (2007) and Goodwin et al. (2008) Australia; Christensen et al. (2008) and Horton and Serafeim (2009) UK; and Gjerde et al. (2008) Norway. In addition, multiple country mandatory IFRS adoption studies focusing 2 Ball et al. (2000) also find that the accounting numbers from common law countries are more highly correlated with prices than are the numbers from code law countries.

4 4 P. Clarkson et al. / Journal of Contemporary Accounting & Economics 7 (2011) 1 17 on value relevance are also starting to appear. These include Capkun et al. (2008) for seven EU countries; and Wang (2008) for 14 EU countries and Australia (the same 15 countries we examine). For Australia, Ahmed and Goodwin (2007) and Goodwin et al. (2008) conclude that aggregate differences between IFRS and local GAAP have no incremental information for price in their samples. For the UK, Christensen et al. (2008) and Horton and Serafeim (2009) report that earnings reconciliations have incremental price relevance over local UK GAAP numbers. For the seven countries they examine (two Common Law and five Code Law), Capkun et al. (2008) report incremental tests which again point to IFRS financial statements conveying incremental value relevant information relative to local GAAP. Finally, Wang (2008) observes that IFRS net income reconciliation differences are positively associated with returns, a finding that is once again consistent with incremental value relevance for IFRS EPS. All of the mandatory adoption studies described above use incremental value relevance approaches. Relative value relevance tests address the question of which GAAP numbers fit price better, whereas incremental value relevance tests ask whether, given knowledge of local GAAP numbers, IFRS numbers have incremental explanatory power for price. It is well established in the literature (see, for example, Biddle et al., 1995) that two GAAP regimes (e.g., local and IFRS) could each have incremental value relevance for price given the other, but one of these regimes could have greater value relevance for price compared to the other. In our application, the relative tests seem more meaningful because investors will have one set of numbers or the other, but not both, when countries switch to IFRS. 3 Finally, Daske et al. (2008) examine the impact of mandatory IFRS adoption on cost of capital around the world, including EU countries. They report that first time mandatory adopters experience a modest decrease in the cost of capital (28 basis points). They observe that the capital market benefits of IFRS adoption are most pronounced for countries with strict enforcement regimes. Their effects are stronger for countries where local GAAP differs more from IFRS. A challenge facing mandatory adoption researchers is the problem of early adoption and any residual self-selection issues involving late adopters. As discussed by Capkun et al. (2008), of the 15 countries in our sample, a number of countries allowed IFRS adoption prior to 2005 and the greatest incidence of early adoption was in Germany. Our search of Worldscope reveals that 326 public companies on Worldscope indicated as residing in Germany made the switch prior to 2005, compared to the 185 mandatory adopters in our sample. Other than Germany, only a minor percentage of companies in our 15 countries opted for early adoption. Our search of Worldscope reveals that, for the 14 remaining countries in our sample, 392 public companies switched prior to 2005 compared to the 3303 mandatory adopters in our sample. Jermakowicz and Gornik- Tomaszewski (2006) survey the early adoption literature and identify the following firm characteristics of early adopters: lower debt-equity ratio, higher percentage foreign sales, higher number of foreign listings, more diffuse ownership and larger capitalization. Thus, mandatory adopters, especially in Germany, are not a random subset of the population. Christensen et al. (2008) show that mandatory adopters are smaller, less dependent on foreign capital markets, more dependent on local bank financing, and have more concentrated ownership. While it is difficult to ascertain the impact of this self selection on our tests, especially for German firms, we believe this constitutes a conservative bias for our tests for the following reason. For the Hung and Subramanyam (2007) sample of 80 German early adopters, available firm-year data (as reported in their Table 3) indicates that the book value of equity increased approximately 50% for IAS GAAP over local German GAAP. As we report in our Table 1, we observe a corresponding percentage increase in the book value of equity of only 11.9%. Thus, it appears that early German adopters faced larger local versus IAS impacts. As such, our tests of IFRS impact would appear to yield a lower bound on the valuation impact of IFRS for German firms. 3. A suggestion concerning the model form linking stock value to accounting information 3.1. Basic intuition With a few exceptions, the literature surrounding price level regressions, wherein either market capitalization or price per share are modeled in terms of accounting-based explanatory variables, has tended to focus on a linear functional form. 4 In this section, we propose extending the traditional linear pricing equation to include an additional explanatory term, the product of earnings and book value. This term is motivated by assuming that accounting earnings and book value are measurements of permanent (i.e., expected) income and net assets, along with associated measurement errors. 5 Formal models developed in Riffe and Thompson (1998) and Beatty et al. (2002) contain similar product terms within closed-form non-linear model specifications. The existence of closed-form solutions to these non-linear models depends on specific assumptions about the distributions of errors. The central driver behind the models, however, is easy to visualize and would seem to transcend the exact distributional properties of the variables. 3 As a robustness check, we also conducted (but do not tabulate) incremental value relevance tests. For Code and Common Law countries, our results (available upon request) are broadly consistent with IFRS financial statements conveying modest incremental value relevant information relative to local GAAP. The average adjusted R 2 improvement when reconciliation differences are added to a local only OLS model are 3.97% and 2.82%, respectively, for Common and Code Law countries. These findings are not inconsistent with the relative tests which we tabulate for linear models, which suggest that a modestly better (worse) fit, overall, for IFRS over local GAAP for Code (Common Law) countries. 4 Within the linear form, Christie (1987), Landsman and Magliolo (1988), Barth and Kallapur (1996), Beatty et al. (1996), Easton (1998), and Barth and Clinch (2009) study the question of deflation and adjustments for heteroscedasticity. 5 Accounting standards have evolved to serve the needs of many users of financial statements; and in bridging across these uses can be thought of as creating a signal about the underlying or true economic value of the firm with an associated error.

5 P. Clarkson et al. / Journal of Contemporary Accounting & Economics 7 (2011) Here is the core intuition. Imagine that accounting earnings is a measurement of the economic cash flow driving price, although it contains a deviation largely uncorrelated with this cash flow. Likewise, imagine accounting book value as a measure of the economic net assets that drive price but containing an error also largely uncorrelated with price. In addition, assume that the error variances are heteroscedastic such that higher-valued firms have higher absolute variance in the deviation between price and accounting information. It follows that a product term in a regression of price on earnings and book will have explanatory power in helping to identify the size of these measurement errors. Because of heteroscedasticity in measurement errors, where earnings and book give similar value predictions, Baye s Rule can be used to infer that these measures come from a firm with relatively low error variance. Conversely, where one value measure is high and the other low, it is more likely that there is higher error variance. This inference about error variance influences the ultimate prediction of price (because we earlier assumed that higher error variances are associated with more highly-valued firms). Moreover, because the geometric mean falls below the arithmetic mean as an increasing function of variance, the cross-product term should have a negative sign in so far as the influence of heteroscedasticity is concerned. To illustrate, assume there are two firms which can only be of three valuation types, high (H), medium (M), or low (L). 6 Each firm reports earnings and book value which also can only be one of three signals: H, M or L. Investors combine the signals on earnings and book value to get a prediction on overall firm type. Further, assume that we know that there is one type M firm and one type H firm. When the two firms report their accounting numbers for the period, firm A reports earnings of M and book value of M, while firm B reports earnings of L and book value of H. Which firm is more likely to be the high value (H) firm? Both firm A (M, M) and firm B (L, H) have average signals of M in this case but the difference between them is the variance in the errors of the earnings and book value signals. Because our model assumes that higher-valued firms have more error variance in their signals, it is more likely that firm B s underlying true value is H when compared to firm A. In a regression of price on earnings and book value with a cross-product term included, if M is half way between H and L in this example, the cross product of H and L will be below M squared. Thus, a negative coefficient on the cross-product term predicts a higher value for a firm with signals H and L than for a firm with two M signals. Imagining a regression incorporating this cross-product term with a negative coefficient, what does the pricing model actually look like? If the data are cut holding either book or earnings constant while varying the other, the model is conditionally linear. If the data are cut so that the spread between earnings and book is increasing, the model will contain a convexity. Alternatively if the data are cut so that the spread is decreasing, a concavity follows. An appendix, available from the authors, fleshes out this intuition within the context of a formal model. It shows that the cross-product term is a complex function of error variances and cross-correlations so that changes in any of these influence the degree of non-linearity of the pricing function Simulation results supporting a cross-product term To substantiate the importance of a cross-product term in the presence of heteroscedastic measurement errors, we simulated (results available upon request) data under varying assumptions about the true linkage between price and accounting variables. If, for example, price is treated as the dependent variable where the independent variables of earnings and book value are multiplied by coefficients and an error term added to the right hand side, no cross-product term emerges in the data, regardless of whether or not heteroscedasticity is introduced. This result should be obvious. The data do not reveal a cross-product term because the pricing model does not contain one. On the other hand, let price be generated with both earnings (times a multiplier) and book created as measures of price, and both with measurement errors. Further, let the error variances be linked back to the level of price. In this case, a negatively signed cross-product term emerges consistently in simulated data. Ultimately, of course, the proof is revealed only in the application to actual data. We find significant cross-product terms in the data used in this study (see Table 5). So, from a practical point of view, that in itself is evidence of non-linearity and motivation to include the cross-product term in the pricing model. 4. Empirical strategy and econometric models As discussed in the Introduction, we employ a same firm year research design (Hung and Subramanyam, 2007) wherein we compare the relevance of local GAAP accounting measures as originally reported for the pre-ifrs adoption year with the relevance of the restated accounting measures for the same year, as presented in the comparative financial statements for the IFRS adoption year. This design varies GAAP while holding the sample composition and time period constant. For all analyses, we partition our sample countries in accordance with the Common Law versus Code Law dichotomy following the argument that the impact of IFRS adoption for any given country should depend upon the quality of the original (local) GAAP earnings figures in that country. 6 This example is an oversimplification given that it unrealistically restricts the signals that firms of type H or L may provide. The intuition regarding firms signaling type M still provides a useful illustration. 7 The formal model of Riffe and Thompson (1998) also implies a convexity in the pricing function when either earnings or book is held constant while the other varies. But this is of second order importance except at extreme values.

6 6 P. Clarkson et al. / Journal of Contemporary Accounting & Economics 7 (2011) 1 17 In conducting the analysis, we begin by considering the linear valuation model P ¼ d 0 þ d 1 BVPS þ d 2 EPS þ t where P = price per share as at the Balance Sheet date; BVPS = per share book value of common equity; EPS = earnings per share from continuing operations; and t = OLS-appropriate error term. We estimate this basic model first as an OLS regression and then as a weighted OLS regression using 1/(sqrt( BVPS + 0.1)) as the weight. Following these analyses, we then consider an alternative non-linear model which incorporates the cross-product of book value and earnings into the model. This so-called product model adopts the following form: P ¼ c 0 þ c 1 BVPS þ c 2 EPS þ c 3 BVPS EPS þ s ð2þ with all measures as previously defined (and the error, s, is determined accordingly). As discussed in Section 3 above, the cross-product term has the potential to shed light on the source of a change in overall model fit. ð1þ 5. Sample data The primary source of data used in this study is Worldscope which contains financial information on public companies worldwide. We identified the countries where mandatory IFRS adoption was required with effect from fiscal year 2005 onwards. For these countries, we downloaded the accounting and fiscal year end stock price data from Worldscope, and the monthly stock returns necessary to complete our analysis from Datastream. The data were extracted using the July 2007 version of the Worldscope files. For the year prior to the mandatory adoption, we required that each firm have local GAAP and restated IFRS data available. Firms for which prior year restated financials did not exist on Worldscope were deleted. We only included firms that switched from their local GAAP to IFRS; those which used US GAAP instead of local GAAP prior to the switch to IFRS were deleted. We further included only those firms which restated the prior year s data due to change in GAAP in our sample. If part of the restatement was identified by Worldscope as due to other reasons, we deleted the observation. We additionally deleted firms with a share price equivalent to $0.10 US or less 8 and finally, we deleted those with price, earnings, or book values are in the top or bottom 1% of the respective country distributions (with a minimum deletion of both the maximum and minimum for each country). As reported in Table 2, our final sample consists of 3488 firms from 14 European Union (EU) countries and Australia. The countries with the largest number of observations are Australia (895), the United Kingdom (UK, 699), and France (484), while the smallest number of observations are for Portugal (32), Ireland (30), and Belgium (72). 9 Overall, the data comprise three Common Law countries (Australia, Ireland and the United Kingdom) and 12 Code Law countries. Table 2 also presents descriptive statistics for the impact of IFRS adoption on book value per share (BVPS) and earnings per share before extraordinary items (EPS) by country, and aggregated by the Common Law versus Code Law partitions. The statistics presented are the mean and standard deviation of BVPS and EPS for the year prior to mandatory IFRS adoption based on both local GAAP (as originally reported) and their counterparts restated to IFRS. In conjunction, we also present the percentage change in these figures using price as the scalar and the percentage of firms for which the restatement to IFRS resulted in an increase or a decrease to BVPS and EPS. 10 As revealed at the end of Panel A, summing across the three Common Law countries, BVPS increased for 46.4% of firms (753 firms) whereas it decreased for 53.6% (870 firms), with the remainder experiencing no change. Alternatively, EPS increased for 49.3% of firms (800 firms) and decreased for 34.0% (552 firms). In contrast, as revealed at the end of Panel B, the adoption of IFRS had a much greater impact on the financial statements of Code Law countries. Here, BVPS figures increased for 67.7% of firms (1262 firms) whereas only 32.2% (601 firms) experienced a decrease. Similarly, 64.4% of firms (1201 firms) experienced an increase in EPS while only 29.6% (551 firms) experienced a decrease. Thus, the patterns for Code Law countries appear quite different than those Common Law countries and point to conservatism in Code Law countries prior to the adoption of IFRS. This conclusion is substantively supported by mean percentage change figures. 11 For the three Common Law countries, the overall mean percentage decline in BVPS is 2.5% while the overall mean percentage decrease in EPS is 0.2%. Thus, for Common Law countries, the restatement appears to have had a negative impact on book value and earnings. Alternatively, for the Code Law subsample, there was a mean overall increase in BVPS of 3.6%. The mean percentage increases in BVPS for the largest Code Law countries (in terms of sample size) are 11.9% (Germany, N = 185), 5.9% (France, N = 484), 5.4% (Italy, N = 207), and 3.2% (Sweden, N = 217). The mean overall percentage increase in EPS for firms from Code Law countries is 0.7%. 8 This sampling constraint had the greatest impact on the Australian sample which was reduced by 342 firms. The next greatest impact was on the UK sample which was reduced by only seven firms. For the 13 remaining countries, the impact was negligible. 9 The UK sample only includes firms listed on the main exchange of the London Stock Exchange. Firms listed on the Alternative Investment Market (AIM), which comprises approximately one-half of listed UK firms, are not included because their IFRS adoption date was for fiscal years starting January 1, Given the relative heterogeneity of firms both intra-country and most certainly inter-country, as well as currency differences, we focus our discussion primarily on the percentage of firms for which the adoption of IFRS resulted in either an increase or a decrease in BVPS and/or EPS. For the same reason, we do not present standard deviation figures for the aggregate Common Law and Code Law samples. 11 The percentage change figure is calculated as the IFRS restated amount minus the local GAAP counterpart scaled by price.

7 P. Clarkson et al. / Journal of Contemporary Accounting & Economics 7 (2011) Table 2 Descriptive Statistics for the Impact of IFRS Adoption. Country Local GAAP IFRS Restated % Change % Observations Increase Decrease Panel A: Common Law (CML) (English Legal Origin) Countries Australia (AUD) N = 895 BVPS Mean Std dev EPS Mean Std dev Ireland (Euro) N = 30 BVPS Mean Std dev EPS Mean Std dev United Kingdom (GBP) N = 669 BVPS Mean Std dev EPS Mean Std dev Common Law Aggregate N = 1624 BVPS Mean EPS Mean Panel B: Code Law Countries Belgium (Euro) N = 72 BVPS Mean Std dev EPS Mean Std dev Denmark (DKK) N = 97 BVPS Mean Std dev EPS Mean Std dev Finland (Euro) N = 101 BVPS Mean Std dev EPS Mean Std dev France (Euro) N = 484 BVPS Mean Std dev EPS Mean Std dev Germany (Euro) N = 185 BVPS Mean Std dev EPS Mean Std dev Greece (Euro) N = 144 BVPS Mean Std dev EPS Mean Std dev Holland (Euro) N = 106 BVPS Mean Std dev EPS Mean Std dev Italy (Euro) N = 207 BVPS Mean Std dev EPS Mean Std dev (continued on next page)

8 8 P. Clarkson et al. / Journal of Contemporary Accounting & Economics 7 (2011) 1 17 Table 2 (continued) Country Local GAAP IFRS Restated % Change % Observations Increase Decrease Norway (NOK) N = 110 BVPS Mean Std dev EPS Mean Std dev Portugal (Euro) N = 32 BVPS Mean Std dev EPS Mean Std dev Spain (Euro) N = 109 BVPS Mean Std dev EPS Mean Std dev Sweden (SEK) N = 217 BVPS Mean Std dev EPS Mean Std dev Code Law Aggregate N = 1864 BVPS Mean EPS Mean The table presents the number of observations for each country (N), the country s currency, in parenthesis, and descriptive statistics for book value per share (BVPS) and earnings per share (EPS). The first three columns present the mean value and standard deviation of: (1) the reported values based on local GAAP for the year preceding the adopting of IFRS; (2) the restated amount using IFRS for the same year (i.e., the year preceding) as presented for comparative purposes in the statements for the year of adoption; and (3) the signed percentage change in the figures measured as the difference between the IFRS restated and the local GAAP amounts, scaled by price. The last two columns present the percentage of firms for which the restated figure exceeds the initially reported local GAAP figure (increase) or for which the restated figure is lower (decrease). Taken together, these figures suggest that the adoption of IFRS had a relatively limited impact on the financial statements of Common Law countries. In contrast, they also point to balance sheet conservatism and conservative income measurement in Code Law countries prior to the adoption of IFRS. These differences observed in the initial application of IFRS accounting standards to Common and Code Law countries is not unexpected given the local versus IAS GAAP studies by Bae et al. (2008) and Ding et al. (2007). Both groups of authors based their analyses on GAAP 2001: A Survey of National Accounting Rules Benchmarked Against International Accounting Standards edited by Nobes (2001). An example of these differences is when local GAAP allows the recognition of liability reserves, a common practice for Code Law countries, but IAS prohibits such behavior. The Bae et al. results, discussed in Section 2, point to greater divergence in local-ifrs GAAP for Code Law countries. Similarly, Ding et al. observe greater absence (no local GAAP rule matching an IAS rule) and divergence (local GAAP differs from IAS) for Code Law countries relative to Common Law countries. The IFRS rules used in 2005 evolved from the IAS rules that existed in 2001 and, as such, the findings of these studies complement ours. Finally, with the exception being Ireland, 14 of our 15 countries experienced an increase in EPS volatility under IFRS. This is consistent with the fair value orientation of IFRS, which results in reduced opportunities for income smoothing. Barth et al. (2008) observe similar patterns for voluntary IAS adopters. 6. Empirical results In this section, we discuss the impact of IFRS adoption on the valuation relevance of book value and earnings across the 15 countries considered in this study. We use tests that compare relative value relevance across GAAP regimes. All reported results are for stock price observed 6 months following the fiscal year of adoption, which is in effect 18 months after the fiscal year for which we compare the value relevance of local and IFRS GAAP (see Capkun et al. (2008) and Hung and Subramanyam (2007, p. 648) for a similar approach). These are in the spirit of future price-based tests as advocated by Aboody et al. (2002). The advantage of this approach is that we can be sure investors have access to the IFRS numbers from the following year s financial statements. While at first glance the disadvantage of this approach is that future price

9 P. Clarkson et al. / Journal of Contemporary Accounting & Economics 7 (2011) Table 3 Regression Results across all 15 Countries. Country Local GAAP results Restated IFRS Results PRICE= b 0 +b 1 BVPS +b 2 EPS +b 3 (BVPS EPS) ADJ R 2 b 0 +b 1 BVPS +b 2 EPS +b 3 (BVPS EPS) ADJ R 2 Panel A: Common Law Countries Australia OLS N = 895 (6.270) (9.834) (18.680) (6.885) (7.881) (19.726) WLS (8.499) (11.645) (15.383) (8.918) (10.352) (15.627) PROD (6.072) (9.011) (13.639) (0.296) (6.988) (6.665) (14.506) (1.402) Ireland OLS N = 30 (1.589) (2.192) (2.897) (2.346) ( 0.837) (5.359) WLS (1.587) (3.927) (2.165) (2.537) ( 0.293) (4.337) PROD (1.826) (0.608) (2.004) (0.912) (1.327) ( 0.200) (4.618) ( 1.047) UK OLS N = 699 (10.069) (14.584) (18.986) (10.473) (10.051) (16.145) WLS (11.424) (12.956) (18.559) (11.273) (8.959) (17.380) PROD (7.971) (13.153) (14.991) ( 2.373) (5.945) (12.818) (17.305) ( 7.489) Panel B: Code Law Countries Belgium OLS N =72 ( 1.150) (6.262) (9.985) (0.254) (7.408) (0.108) WLS (0.041) (6.679) (6.930) (0.471) (7.013) (0.684) PROD (2.431) ( 4.102) (11.359) (7.200) ( 2.120) (7.033) ( 0.466) ( 4.037) Denmark OLS N = 97 (0.617) (9.644) ( 0.282) (0.681) (9.887) ( 0.550) WLS (2.099) (8.681) (0.797) (2.151) (8.691) (0.621) PROD (1.729) (8.563) ( 1.901) (2.264) (1.745) (8.728) ( 1.997) (2.192) Finland OLS N = 101 (3.223) (5.793) (5.215) (3.188) (4.522) (5.994) WLS (3.298) (6.065) (5.364) (3.243) (4.828) (6.204) PROD (2.148) (4.544) (3.569) ( 0.715) (1.705) (4.214) (5.139) ( 1.381) France OLS N = 484 (3.865) (12.971) (3.973) (4.104) (21.087) (6.653) WLS (5.596) (15.891) (4.268) (7.168) (18.594) (7.913) PROD (2.252) (12.248) (5.009) ( 3.012) (1.487) (18.187) (8.653) ( 5.309) Germany OLS N = 185 (1.069) (7.366) (10.839) (0.769) (13.520) (10.135) WLS (1.700) (9.287) (7.177) (1.921) (11.346) (8.349) PROD (0.866) (6.230) (7.998) ( 0.416) (0.025) (14.162) (8.107) ( 3.086) Greece OLS N = 144 (2.441) (2.573) (5.713) (1.805) (3.926) (3.218) WLS (2.753) (2.041) (6.190) (2.147) (3.697) (3.780) PROD (1.335) (3.260) (4.839) ( 2.069) (0.224) (5.232) (4.808) ( 3.692) Holland OLS N = 106 (9.765) (5.799) (6.964) (8.967) (7.787) (5.176) WLS (9.170) (5.670) (6.415) (8.313) (7.155) (4.822) PROD (7.756) (4.466) (6.025) ( 0.822) (7.148) (6.538) (4.640) ( 1.315) (continued on next page)

10 10 P. Clarkson et al. / Journal of Contemporary Accounting & Economics 7 (2011) 1 17 Table 3 (continued) Country Local GAAP results Restated IFRS Results PRICE= b 0 +b 1 BVPS +b 2 EPS +b 3 (BVPS EPS) ADJ R 2 b 0 +b 1 BVPS +b 2 EPS +b 3 (BVPS EPS) ADJ R 2 Italy OLS N = 207 (4.670) (19.871) (4.686) (4.314) (18.718) (4.637) WLS (4.857) (16.913) (4.214) (4.458) (16.392) (4.388) PROD (4.663) (18.650) (2.835) ( 0.271) (4.304) (18.572) (3.855) ( 1.483) Norway OLS N = 110 (5.346) (2.788) (1.366) (5.425) (0.919) (3.330) WLS (4.944) (4.658) (1.064) (5.475) (2.759) (2.493) PROD (2.573) (6.254) (3.248) ( 5.532) (3.808) (1.994) (3.878) ( 2.074) Portugal OLS N = 32 (3.936) (0.589) (4.273) (1.366) (3.434) (4.891) WLS (3.300) (1.598) (3.821) (1.591) (3.334) (4.836) PROD (2.237) (1.284) (4.227) ( 1.180) (1.028) (3.611) (3.231) ( 1.084) Spain OLS N = 109 (4.044) (1.865) (10.693) (4.651) (3.678) (12.146) WLS (4.072) (1.893) (10.657) (4.758) (2.501) (11.693) PROD (2.076) (2.495) (9.625) ( 1.736) (2.181) (4.248) (10.672) ( 2.194) Sweden OLS N = 217 (6.794) (9.021) (2.577) (7.384) (7.726) (2.593) WLS (6.578) (9.144) (3.791) (6.890) (7.988) (3.713) PROD (4.034) (10.204) (5.020) ( 4.490) (4.230) (9.658) (5.569) ( 5.154) Variable definitions: PRICE is the firm s market price measured 18 months after the fiscal year end balance sheet date; BVPS is the firm s accounting book value deflated by the number of common shares outstanding; and EPS is the firm s earnings per share for the fiscal year. PRICE will be the same number for both regressions reported above. The only variables changing are the measures of BVPS and EPS. OLS indicates ordinary least squares regression estimates. WLS indicates the estimation is done using weighted least squares, where the weighting factor is 1/(sqrt(BVPS) ). PROD indicates the estimates are the result of using the product model described in the text (Eq. (2)). The product model adds the product of BVPS and EPS to the OLS estimation model. The additional variable is intended to reflect measurement error in the underlying accounting variables. t-statistics presented in parentheses. reflects future information, Aboody et al. argue that future information adds uncorrelated measurement error to the dependent variable. 12 For the relative tests, we consider goodness-of-fit statistics (adjusted R 2 ) for the various valuation models we employ. Again, the three regression analyses that we report are the simple linear model (Eq. (1)) estimated both using OLS and WLS, and the product model (Eq. (2)) estimated using OLS. These results are presented in Table 3. In Table 4 we present summary model error measures based on the difference in absolute percentage pricing error between accounting regimes, partitioned on Common Law versus Code Law country origin. Table 5 contains an analysis of the impact of IFRS adoption on the cross-product term of the product model Relative value relevance results This section examines the ability of accounting numbers to explain price variation. We consider the Common Law and Code Law countries separately, and compare the change in this explanatory power upon the switch from Local GAAP to IFRS Common law countries Perhaps as expected, given the descriptive statistics relating to the impact of IFRS adoption on BVPS and EPS, the results for the Common Law countries suggest that adoption only modestly impacted the valuation relevance of book value or 12 Results (untabulated) and conclusions are qualitatively similar when the valuation models are based on the price at the comparative year balance sheet date (P 0 ) or price 3 months (P +3 ) after the comparative year balance sheet date instead of the price 18 months (P +18 ) after the comparative year balance sheet date.