Voluntary adoption of non-local GAAP in the European Union: A study of determinants and consequences

Abstract

This study examines the determinants and consequences of voluntary adoption of non-local accounting principles (non-local GAAP) by firms listed and domiciled in the European Union (EU). We restrict ourselves to the two predominant internationally accepted sets of accounting standards: International Accounting Standards (IAS) and United States generally accepted accounting principles (US GAAP). We have used various sources to identify EU firms that use non-local GAAP. We examine the 1999 annual reports of all these firms, because accounting standard choices in more recent years may be affected by the announcement of the proposal by the European Commission in February 2001 to mandate IAS usage from 2005 on. The maintained hypothesis is that firms that voluntarily adopt IAS or US GAAP expect to experience net benefits from adoption. The finding that 133 non-financial firms in the EU voluntarily used non-local GAAP in 1999 suggests that the majority of listed EU firms does not expect to benefit from non-local GAAP adoption. By studying the characteristics of non-local GAAP adopters this study provides insight into the determinants of non-local GAAP adoption. We find that firms voluntarily using non-local GAAP are more likely to be listed on a US exchange, the EASDAQ exchange in Brussels, and have more geographically dispersed operations. Furthermore, they are more likely to be domiciled in a country with lower quality financial reporting and where IAS is explicitly allowed as an alternative to local GAAP. We also study whether non-local GAAP adopters have lower levels of information asymmetry, a much cited benefit of using more transparent financial reporting, than non-adopters. We examine three proxies for information asymmetry: analyst following, cost of equity capital, and uncertainty among analysts and investors (forecast dispersion and stock return volatility). We document a positive effect of non-local GAAP adoption on analyst following, but fail to find evidence of a lower cost of capital for non-local GAAP adopters. Contrary to expectations, uncertainty among analysts and investors appears to be higher for firms using IAS or US GAAP than for firms using local GAAP. However, by comparing ‘early’ and ‘late’ adopters, we find some evidence that suggests that benefits take some time to fully materialise.

Acknowledgements

We gratefully acknowledge helpful comments and suggestions from the editor and two anonymous reviewers of EAR, Steven Maijoor, Frank Moers, Erik Peek, and workshop participants at Maastricht University, the 25th Annual Congress of the European Accounting Association in Copenhagen, and the EIASM Workshop on Capital Market Research in Accounting in Frankfurt. Furthermore, we gratefully acknowledge the contribution of Thomson Financial for providing earnings per share forecast data, available through the I/B/E/S – Institutional Brokers Estimate System. This data has been provided as part of a broad academic programme to encourage earnings expectations research.

Notes

¹ For an overview of listing requirements of EU stock exchanges, see Haller (2002).

² Private communication with Stefano Zambon confirmed this. His research shows that only three Italian firms comply fully with IAS.

³ The authors would like to thank Paul Pacter of the IASC for providing the list that was no longer available from the IASC website.

⁴ To be included, the auditor's report and summary of accounting policies, or footnotes must state that the financial statements comply with IAS without qualification (IASC, 2001).

⁵ Not necessarily IAS, however, no more specific data was available.

⁶ Item number GF66: Accounting standard – Note (ASTD). Again, no more specific data was available.

⁷ There was considerable overlap between the different data sources.

⁸ We also excluded a small number of firms classified as SIC code 9.

⁹ The inclusion of firms domiciled in the UK in our regression analyses (we do not have data on the quality of Irish accounting standards) does not materially influence the results.

¹⁰ See Maddala (2001, section 8.3) or, for a discussion of methodological issues in using in an accounting context, see Zmijewski (1984), Palepu (1986) and Maddala (1991).

¹¹ Mutual recognition in this context means that EU firms listing on a stock exchange in the EU can use their local GAAP to draw up their financial statements.

¹² The measure developed by Ashbaugh (2001) is not available for Austria, Greece and Italy, whereas the CIFAR score is only unavailable for Luxembourg.

¹³ Although this score was constructed based on 1990 annual reports, we still expect the measure to reflect differences in accounting disclosures across countries, because regulation and practice only change gradually.

¹⁴ EU countries without any institutional oversight systems for financial reporting standards are: Austria, Finland, Germany, Ireland, Luxembourg and the Netherlands (Ireland and Luxembourg are not included in our analyses). The report does not analyse the enforcement mechanisms in Greece, which forces us to exclude Greek firms from this sensitivity analysis.

¹⁵ The coefficient on the indicator variable is statistically significant (1% level, one-sided). Inclusion of this variable leaves the results reported earlier qualitatively unchanged.

¹⁶ We estimate the logistic regression equation (1) including only firms using IAS or local GAAP (i.e. excluding US GAAP users), with IAS as the dependent variable (an indicator variable that takes the value 1 if a firm uses IAS and 0 otherwise) and without the dummy variable indicating a US listing (US_EX).

¹⁷ In contrast to Auer (1998) and Ashbaugh and Pincus (2001) we employ a cross-sectional design to study the consequences of non-local GAAP adoption, mainly because of data availability constraints. After contacting the 114 firms using non-local GAAP examined in Section 3, we were able to obtain the date of first usage for 92 firms. However, many of these firms did not switch to non-local GAAP but were using (or started using) IAS or US GAAP at the time of the IPO. Consequently, for many firms we do not have pre-adoption data available. Furthermore, Leuz and Verrecchia (2000) point out that, based on an event study (around the switch to non-local GAAP), it is hard to separate the effect on information asymmetry from the news effect (the revision in capital market participants' expectations about future firm performance). Therefore, Leuz and Verrecchia perform time-series analysis only as a consistency check for their cross-sectional findings.

¹⁸ We use earnings forecasts outstanding six months after fiscal yearend to allow the publication of (possible first-time) non-local GAAP annual reports to affect our information asymmetry proxies. When forecasts for the next four years are not available, we use the forecasted earnings growth rate from I/B/E/S or the growth rate implicit in forecasts that are available to calculate the missing forecasts. We exclude observations that have negative cumulative four year earnings forecasts.

¹⁹ We estimate beta by using daily return observations (from Datastream, at least 150 observations required) over the firm's fiscal year for which we analyse the accounting standard choice.

²⁰ Specifically, we minimise the Euclidean distance between local and non-local GAAP users per country based on standardised measures of beta, market capitalisation and probability of adoption (the fitted value per firm from regression equation (1)).

²¹ We thank one of the anonymous reviewers for suggesting this possibility.

²² We contacted the 114 firms using non-local GAAP included in the determinants analysis and asked when they started using IAS or US GAAP. In this way we were able to obtain the date of first usage of non-local GAAP for 92 firms. Data availability constraints reduce the number of observations in the various consequences analyses even further.

²³ An alternative would be to directly examine changes in a firm's investor base. We did not perform this analysis because, based on our data, it is hard to come up with a proxy for investor base.

²⁴ Data is available at the statistics section of the website of the World Federation of Exchanges (http://www.world-exchanges.org).

²⁵ At least 150 daily return observations are required.

²⁶ Wilcoxon two-sample tests for differences between the groups of non-local and local GAAP users for the independent variables appearing in equation (7) reveal differences comparable to those in Table 3, suggesting that the reduction in observations has not changed the composition of the sample significantly. The level of stock market development (STMRKTDEV) is significantly lower (1% level, two-sided) for non-local GAAP than for local GAAP users. Furthermore, there is no significant difference in abnormal stock return volatility (VOLATILITY) between the groups of non-local and local GAAP users.

²⁷ The mean natural logarithm of one plus analyst following (FOLLOWING) for the group of non-local (local) GAAP users is 2.92 (2.02), the standard deviation is 1.09 (1.21) and the median is 3.11 (2.08). For the full sample (non-local and local GAAP users), the mean is 2.08, the standard deviation is 1.22 and the median is 2.08.

²⁸ Wilcoxon two-sample tests for differences between the groups of non-local and local GAAP users for the independent variables appearing in equations (8) and (9) confirm earlier findings (also see Table 3 and n. 26).

²⁹ The mean forecast dispersion (DISPERSION) for the group of non-local (local) GAAP users is 0.63 (0.39), the standard deviation is 1.87 (0.78) and the median is 0.26 (0.25). For the full sample (non-local and local GAAP users), the mean is 0.41, the standard deviation is 0.91 and the median is 0.25. The mean stock return volatility (VOLATILITY) for the group of non-local (local) GAAP users is 0.03 (0.03), the standard deviation is 0.01 (0.01) and the median is 0.02 (0.02). For the full sample (non-local and local GAAP users), the mean is 0.03, the standard deviation is 0.01 and the median is 0.02.

³⁰ In the analysis of cost of capital differences between non-local and local GAAP users, self-selection bias is less of an issue, since we match local GAAP users to non-local GAAP users on the estimated benefit of adoption, i.e. the fitted values from equation (1), among others (also see n. 20).

³¹ The results (not tabulated) from estimating the probit model are comparable to the results from the logistic regression model (see Table 5, model a).

³² Specifically, we find that the coefficient on GAAP remains positive and statistically significant in equation (7). In equation (8), the effect of non-local GAAP adoption on forecast dispersion is no longer statistically significant. Controlling for possible self-selection bias does not influence the results obtained from estimating equation (9), i.e. we do not find an effect of non-local GAAP usage on stock return volatility. Furthermore, the coefficients on the control variables included in equations (7), (8) and (9) remain qualitatively similar. In addition to the two-stage model described in the text, we also use an instrumental variables approach (as described in Leuz and Verrecchia, 2000, fn. 23), with similar results. Note that we do not control for possible self-selection in the equations where we separately measure the effect of early and late adoption of non-local GAAP on our information asymmetry proxies (see Table 7, panel B and Table 8, panel B), since this would require explicit modelling of the early versus late adoption choice to avoid self-selection bias. We consider this to be outside of the scope of the paper.