Compliance with IFRS 3- and IAS 36-required disclosures across 17 European countries: company- and country-level determinants

Abstract

In this study, we analyse compliance for a large sample of European companies mandatorily applying International Financial Reporting Standards (IFRS). Focusing on disclosures required by IFRS 3 Business Combinations and International Accounting Standard 36 Impairment of Assets, we find substantial non-compliance. Compliance levels are determined jointly by company- and country-level variables, indicating that accounting traditions and other country-specific factors continue to play a role despite the use of common reporting standards under IFRS. At the company level, we identify the importance of goodwill positions, prior experience with IFRS, type of auditor, the existence of audit committees, the issuance of equity shares or bonds in the reporting period or in the subsequent period, ownership structure and the financial services industry as influential factors. At the country level, the strength of the enforcement system and the size of the national stock market are associated with compliance. Both factors not only directly influence compliance but also moderate and mediate some company-level factors. Finally, national culture in the form of the strength of national traditions (‘conservation’) also influences compliance, in combination with company-level factors.

Keywords:

• IFRS
• compliance
• business combinations
• cross-national analysis
• goodwill
• impairment testing

Acknowledgements

We are grateful to the editor and two anonymous reviewers as well as to Andreas Barckow, Holger Himmel, Allan Hodgson, Dennis Jullens, Andreas Mackenstedt and Bill Rees. We also appreciate comments on earlier versions of this paper from participants at the EAA Conference in Rotterdam as well as from workshop participants at the Vienna University of Economics and Business, the University of Glasgow and at the Financial Reporting Review Panel in London. Peter Schmidt gratefully acknowledges support from the Higher School of Economics (HSE), Basic Research Programme (International Laboratory for Sociocultural Research), Moscow.

Notes

From 1 January 2011 onwards, CESR has been replaced by the European Securities and Markets Authority; for details, see www.esma.europa.eu.

IFRS 36, paragraph BC205.

See Beattie et al. 2008 and Ernst and Young (2006) for practice-based surveys of smaller samples.

To ensure that our findings regarding non-compliance in 2005 financial statements do not merely reflect transitory implementation problems pertaining only to the first year of IFRS application, we also collect data for a select group of companies for 2007. The findings for 2007 are very similar to those for 2005, indicating that non-compliance is not a temporary phenomenon.

During phase two, the IASB considered implementation issues arising from application of the purchase method. The phase was completed with publication of a revised IFRS 3 in January 2008.

In general, goodwill is the excess of the cost of an acquisition over the sum of the fair values of the assets acquired less the liabilities assumed, taking into account deferred taxes.

For instance, there is a substantial established literature on earnings management (e.g. Healy and Wahlen 1999, Ronen and Yaari 2008). However, earnings management is usually defined as comprising practices within the limits of accounting standards or laws.

Studies on companies that disclose weaknesses in internal control following section 404 of the Sarbanes-Oxley act are also confined to the US environment. These studies find that firms with serious control problems are mostly relatively young and small (e.g. Doyle et al. 2007). In contrast, our sample comprises large European blue chip companies.

Cascino and Gassen (2011) also find evidence suggesting that IFRS compliance levels may differ systematically within countries. More precisely, they find that companies located in the South of Italy exhibit compliance levels significantly lower than those of companies domiciled in Northern Italy.

Some sample companies are audited by two firms (for example, this is mandatory in France). For a company audited jointly by one of the Big 4 and by one of the non-Big 4, we assume that the audit procedures of the large firm dominate and code AUDITOR as 1. Thus, when AUDITOR is coded as 0, the audit is done only by non-Big 4 firms, even if the audit is conducted jointly.

The 8th EU directive on auditing was changed in May 2006. The directive mandates EU member states to implement legislation that generally requires stock-listed companies to establish audit committees. For details, see Article 41 of EU Directive 2006/43/EC of 17 May 2006.

As a robustness check, we also estimate our models with an industry classification based on first-digit SIC codes. Our findings remain essentially the same.

We expect compliance with disclosure requirements regarding business combinations and impairment testing to be related to the overall differences between IFRS and national GAAP. This expectation is justified if one of the following holds: (1) the overall measure DIFFER proxies for differences between IFRS and national disclosure requirements in our specific areas of interest (i.e. business combinations and impairment testing) and (2) the quality of the footnote disclosures in our area of interest is affected by the overall differences between IFRS and national reporting standards and thus the challenges faced by companies when adopting the new international standards. Furthermore, it is not feasible to construct a meaningful measure for differences between IFRS and national GAAP with regard to business combinations and impairment testing disclosures because for our sample companies most national GAAPs did not require specific, detailed disclosures of this nature prior to 2005.

The data underlying the original La Porta et al. (1998) index relates to the 1980s and early 1990s and thus should not be used to assess enforcement in 2005.

Measurement of the effectiveness of the enforcement of financial reporting standards in national capital markets is controversial. Like the earlier La Porta et al. (1998) index, the index developed by Djankov et al. (2008) measures the enforcement of financial reporting standards only indirectly; it was actually designed to measure the legal protection of minority shareholders against expropriation by corporate insiders (“anti-self dealing index”). Preiato et al. (2012) construct alternative enforcement indices for different years based on cross-country data related to the regulation of auditors and to characteristics of national financial reporting enforcement bodies. As a robustness test we also estimate our models with their index for the year 2005 and find that the results are qualitatively very similar to those we obtain when using the enforcement measure of Djankov et al. (2008). We also get similar results when, in a further test, we employ GOVERN, a measure that is based on the World Bank's Worldwide Governance Indicators (Kaufmann et al. 2008) and indicates the effectiveness of countries' governments, their regulatory quality and their ‘rule of law’.

For details of the European Social Survey, see http://www.europeansocialsurvey.org/index.php.

Conceptually, the ‘conservation’ dimension of the Schwartz Value Survey combines elements from Hofstede's dimensions ‘uncertainty avoidance and ‘individualism’. As a robustness check, we estimate our model with data for these two Hofstede dimensions. Our findings are qualitatively very similar. For a comparison of the Hofstede's culture dimensions and the Schwartz Value Survey, see Hofstede and McCrae (2004).

For years ending 31 December 2005 and later, European listed companies are required to prepare consolidated accounts based on IFRS. Thus companies with year-ends earlier than 31 December tended to postpone IFRS adoption until 2006. When referring to 2005, we are referencing the first year of mandatory IFRS adoption. For example, for a 31 March 2006 financial year-end company the 2006 report is used in our analysis.

Where possible, we include dual-listed companies in the country where they are domiciled and eliminate them from other sub-samples. In a very few cases, however, companies are listed on exchanges outside their home country without being included in their home country index. For instance, one FTSE 100 company is legally domiciled in Switzerland and is not included in the SMI index. Such companies voluntarily submit themselves under the regulatory framework of the country of listing. Therefore, we include these companies in the primary country where they are listed (in our example, the company is included in the UK sub-sample).

A copy of the disclosure checklist is available on request from the authors.

For goodwill impairment testing, companies are required to disclose information on estimates used to measure recoverable amounts of CGUs containing goodwill and describe how they estimate ‘value in use’ or ‘fair value less costs to sell’ of CGUs containing significant goodwill. Some companies with material goodwill balances did not present information concerning the basis on which the CGUs recoverable amount was determined. In these instances, we estimate how many disclosure items should have been presented. This is challenging since a different number of disclosures apply depending on whether companies determine recoverable amounts on the basis of ‘value in use’ or ‘fair value less costs to sell’. Of the sample companies presenting the required information, approximately 95% used value in use; the other 5% used fair value less costs to sell. Thus, we use a weighted average of the number of required disclosures under these two alternatives to gauge non-compliance.

As suggested by the descriptive statistics, companies fully complying with the disclosure requirements mostly come from Anglo-Saxon and Northern European countries. These companies also have in general undertaken fewer acquisitions than the other sample companies. On the other side of the spectrum, a total of 16 companies provide less than 40% of the required disclosures. Most of these companies are domiciled in Eastern or Southern Europe, or in Austria, and a relatively high number of them are in the financial sector. Furthermore, companies with below average disclosure levels tend to have lower goodwill positions than average and higher proportions of closely held equity share capital. Regarding other company characteristics, we do not identify clear univariate patterns.

We measure compliance with IFRS disclosure requirements pertaining to acquisitions and impairment testing. Since not all companies undertake acquisitions every year and since the transactions themselves are not identical, disclosure compliance levels are not expected to be completely identical over time.

The data for our national culture variable CONSERV is from the European Social Survey 2004. Italy is not covered in the European Social Survey 2004. We therefore lose our 27 Italian observations in the analysis of country-level determinants of compliance.

Additionally, some companies issue different classes of equity shares, and CLOSELY_HELD can differ across classes. For companies with multiple share classes, we use the ownership data for the most important class of voting shares; this may produce measurement error. However, information is not always available for minor classes, thereby precluding computation of weighted averages. In most countries, multiple share classes are rare or non-existent, but they are more frequent in some countries (e.g. Sweden).

In Table 5, Models 1 and 2 are estimated with the 332 observations for which complete data is available in order to ensure that the results are comparable to the results for our fully specified model 3 (with CLOSELY_HELD). If we estimate models 1 and 2 with the full sample of 357 observations (i.e. without the variable CLOSELY_HELD in Model 1), the estimation results for all other variables are very similar. 

In a robustness check, we also estimate our model with censored regression (using MPLUS, Version 6) rather than OLS to take into consideration that our dependent variable COMPLIANCE is defined only for values between 0 and 1. The results are qualitatively the same as those from the OLS estimations. 

The linear term CLOSELY_HELD is not significant in the fully specified Model 3. Thus, we can assume that the maximum of the quadratic term is at CLOSELY_HELD = 0. Since we mean-centered CLOSELY_HELD, this implies that, ceteris paribus, the maximum level of compliance is given when strategic investors hold about 27% of equity.

Caution should be exercised in interpreting the result for Luxembourg as it is based on only four companies. 

Eleven companies reported more than 10 business combinations in 2005; the maximum number of transactions reported was 41. In order to check whether outliers bias our results, we estimate two alternative model variants. In the first variant, we wincorise COMBINATIONS at the value of 10, in the second equation we recode COMBINATIONS as ln (1 + number of acquisitions). However, COMBINATION remains insignificant in both model variants, and all other results are qualitatively unaffected.

SEASONED is significant at the 10% level in Model 1, i.e. the model without country indicator variables. Recent research on the economic consequences of the introduction of IFRS has found it helpful to distinguish between three groups of companies, (i) early adopters, i.e. SEASONED adopters in our terminology, (ii) late adopters, i.e. companies that reside in countries where early adoption was allowed but decided to wait until IFRS became mandatory (‘resisters’), and (iii) mandatory adopters, i.e. companies domiciled in countries not allowing early adoption (e.g. Christensen et al. 2008, Daske et al. 2008, 2011, Capkun et al. 2012). Based on this research, in an additional investigation we replace our country indicator variables with an indicator variable that is 1 for all companies domiciled in countries that allowed early adoption and 0 for all companies domiciled in countries that did not allow early adoption (Capkun et al. 2012). When we estimate the model, the indicator variable for countries allowing early adoption is not significant. The other results are similar to those for Model 1, i.e. the model that only includes company-level variables (see Table 5). The exception is the result for the coefficient of SEASONED. While this coefficient is significant at the 10% level in Model 1, it now loses its significance. Furthermore, we do not find a significant difference between the average levels of compliance of early adopters, late adopters, and mandatory adopters (ANOVA: F = 0.627, p = 0.535).

Central European countries (Czech Republic, Hungary and Poland) are not addressed by La Porta et al. (1998). Given their common history as former Soviet-bloc transition countries, we group them into one country class.

It should be noted that the number of observations for the Central European sub-sample is relatively small (Czech Republic: 6, Hungary: 6 and Poland: 13).

To further address the possibility of multicollinearity, especially between our four country variables, we again inspect the VIFs. All VIFs are low (the highest being 2.249 for DIFFER), indicating that multicollinearity does not appear to be a problem. In addition, we run our model four more times, each time including only one of the four country variables. The results for the model variants that include ENFORCE, S-MARKET and CONSERV confirm the results for the full model, i.e. the coefficients for the three variables have the same signs as in the full models, and they are significant. When we include DIFFER as the only country-level variable, in contrast to the full model, the estimator turns out significantly negative (t = −2.639, p < 0.01), suggesting DIFFER now picks up some of the variance that in the full model is explained by the other three country variables. In accordance with our expectations, the negative coefficient indicates that companies that reside in countries with large differences between traditional local GAAP and IFRS tend to have lower levels of compliance than companies that reside in countries where differences between traditional local GAAP and IFRS are smaller. 

By construction, the interaction terms are highly correlated with the single variables from which they are formed. To preclude multicollinearity, we orthogonalise the interaction terms. That is, we regress the interaction terms on the single variables from which the interaction terms are constructed and use the residuals as ‘pure’ interaction effects in our regression analysis. Inspection of the VIFs of the estimation of our model with all interaction terms reveals that most VIFs are very moderate, and that all are lower than the critical value of 10 (Gujarati 1995, p. 328).

For example, the estimated coefficients for the four country-level variables in the OLS regression (see Table 7, Panel A) and in the multilevel regression are as follows: DIFFER: 0.000 and 0.001; ENFORCE: 0.046 and 0.052; S-MARKET: 0.009 and 0.008; and CONSERVATION: −0.097 and −0.105.

The exception is the coefficient of the country-level variable ENFORCE. This variable is significant at the 10% level in the OLS regression (see Table 7, Panel A), but misses the 10% significance level in the multilevel regression (p = 0.120).

We exclude AUDITOR from the equation because of the small number of observations for clients of non-Big 4 firms. Thus, our multigroup SEM tests are based on the following model: COMPLIANCE = α + β ₁GOODWILL + β ₂SEASONED + β ₃AUDIT_COM + β ₄CAPITAL + β ₅CLOSELY_HELD2 + β ₆FINANCIAL + ϵ.