1. THE VALUE RELEVANCE OF IFRS IN THE EUROPEAN BANKING INDUSTRYby
M. Agostino, D. Drago and D. B. Silipo
Università della Calabria (Italy)
This version, October 2008
PAPER PREPARED FOR
The European Financial Management Symposium Risk Management in Financial Institutions April 23-25, Audencia School of Management-Nantes, France

2. OUTLINE
We investigate the market valuation of accounting information in the European banking industry before and after the adoption of IFRS, compulsory since 2005.
To this aim, we specify a multiplicative interaction model in which the partial effects of earnings and book value on share prices are conditional on the adoption of IFRS. In addition, we use panel methods to control for time-invariant unobserved heterogeneity.
Our main finding is that the impact of earnings on share prices increased after IFRS became compulsory in Europe, while book value did not significantly affect market valuation.

3. BACKGROUND
Since 1 January 2005 all listed companies in the European Union have been required to publish their consolidated financial statements in accordance with International Accounting Standards, known as IAS/IFRS rather than national requirements (local GAAP).
The assumption is that financial reporting under IAS/IFRS results in better information quality than local GAAP, with beneficial effects on the cost of capital.

4. An accounting number is value-relevant if it has a predicted significant relation to share prices. This is the case when the item conveys information that investors use in valuing the firm and when it is measured reliably enough to be reflected in share prices.
This paper takes market value relevance as a proxy for overall information quality and investigates whether the value relevance of account numbers increased after the adoption of IAS/IFRS by listed banks in Europe.

5. A number of studies compare the value relevance of IAS, US-GAAP and local GAAP in other countries, with contrasting results.
Barth et al. (2005) found that IAS produce better accounting quality than local GAAP outside the US, with beneficial effects on the cost of capital.
By contrast, Eccher and Healy (2000), Wu et al. (2005) and Hu (2002) find that the accounting amounts based on IAS are no more value-relevant than those based on Chinese GAAP. Similar results are obtained by Dumontier and Labelle (1998), comparing IAS and French accounting principles.

6. More interestingly, Ball et al
More interestingly, Ball et al. (2003) studying the quality of financial reporting in Hong Kong, Malaysia, Singapore and Thailand find that the application of accounting standards also depends significantly on the regulatory, enforcement, and attestation environment.
The evidence concerning the effects of IAS on accounting quality, then, is mixed, both in Europe and elsewhere. The qualitative results for the banking sector are similar.

7. However, the above results may suffer from sample selection bias, as the adoption of IAS was still voluntary. The European Union now offers a unique opportunity to address the issue of whether the earlier findings can be extended to situations in which the adoption of IAS is compulsory.
Preliminary results indicate significant positive market reactions to events that increased the likelihood of the adoption of IFRS (see Armstrong et al., 2006, Horton and Serafeim, 2007, Morais and Curto, 2007)

8. EMPIRICAL QUESTIONS
Using data from 2000 to 2006, we investigate whether replacing local GAAP with IAS improves the quality of accounting amounts in Europe.
We test this prediction on European listed banks, for which IAS became mandatory in 2005.
We investigate value-relevance by estimating a panel valuation model to see whether the value-relevance of accounting information changed.

9. THE MODEL
Building on the Ohlson (1995) framework, we estimate the following model:
P = stock price six months after fiscal year end (robustness: three months after);
BVPS = book value per share;
EPS= earnings per share;
postIAS = dummy coded 1 when IAS become mandatory (years 2005 and 2006), and 0 otherwise;
T = trend;
ε= composite error summarizing unobserved time-invariant bank characteristics + idiosyncratic shocks to market value.

10. Estimation Method
A poolability test (Breusch - Pagan test) on our sample shows that bank-specific fixed effects are statistically relevant.
In addition, the Hausman test leads to the adoption of a fixed effects estimator.
Although bank-fixed effects subsume all other time-invariant effects, the market value of a bank is likely to be sensitive to a variety of factors reflecting differences in national economic environments. So, some of these confounding effects may remain and the error terms may be correlated for banks belonging to same country over time. Accordingly, it is reasonable to cluster the observations at country level.

11. Turning to the parameters of interest:
We specify a multiplicative interaction model: the partial effects of book value and earnings are dependent on IAS adoption.
The marginal effects of earnings and book value depend on the postIAS dummy.
when postIAS= 0 the marginal effects coincide with the estimated coefficients α1 and α2, and significance is given by the t-statistic
when postIAS = 1, we compute the appropriate marginal effects α1+α4 and α2+α5 (sum of the regressor and postIAS coefficients), and the relative standard errors.
So the marginal effects may change sign and gain or lose significance depending on IAS adoption.

12. SAMPLES
The panel includes the listed banks in EU-15 in the period :
1.1 The unbalanced panel (all available observations) is a sample of 1201 annual observations for 221 European listed banks.
1.2 The balanced panel (only the banks with data available for every year) falls from 1201 to 567 observations and from 221 to 81 banks.
We can compare the effects of IAS adoption on the same set of banks
but data points are drastically reduced, so that the unbalanced sample may involve a gain in efficiency.

13. Results: Unbalanced panel estimates
Before IAS adoption
book value marginal effect: positive and insignificant in itself, but tends to be significant jointly with its interaction term (see the F-test)
earnings marginal effect : positive and significant
After IAS adoption
book value marginal effect : negative but not always significant (see march price estimates).
earnings marginal effect : positive and significant, its magnitude increases.

15. Results: Balanced panel estimates
Before IAS adoption
book value marginal effect : positive but not always significant, even jointly with its interaction term (see the F-test)
earnings marginal effect : positive and significant
After IAS adoption
book value marginal effect : positive (June price) and negative (March price) but always not significant
earnings marginal effect : positive and significant, its magnitude increases

17. Country-by-country analysis
Some heterogeneity of market reactions but a common pattern emerges:
The earnings interaction term is always positive, and from 2005 on the partial effect of earnings is positive and statistically significant;
the book value interaction term is mostly negative and the marginal effect of book value tends to become insignificant in the years following IFRS adoption (except in Denmark).
In short, there is evidence that the impact of earnings data increased after IFRS were made compulsory. The largest additional effect of earnings data was in Germany and Italy, the smallest in Britain.

18. What drive these results?
We take separate sub-samples of:
High and low-capitalization banks
(On the ground that problems of information asymmetry are more severe for the smaller institutions)
Cooperatives and public limited companies
(As the former are expected to be more opaque)
Rated and non-rated banks.
(As the latter are expected to be more opaque)

19. High and low-capitalization banks
pattern comparable to that for the entire sample: earnings tend to have a greater impact, book value to turn negative and statistically insignificant.
High-capitalization group:
the coefficients of both earnings and book value are positive and their marginal effects increase after IAS adoption

20. Cooperatives and public limited companies
earnings tend to have a significant greater impact, book value to turn negative and statistically significant.
Public limited companies:
the increase in the impact of earnings and the decrease in that of book value again emerge, but both variables exert a positive marginal effect subsequent to IFRS introduction

21. Rated and non-rated banks
The results of the rated banks (105 banks and 406 obs) are similar to those of the capitalized group (46 banks, 295 obs) even though the sub-samples are different.
Non-rated banks confirm the results of the whole sample: the (positive) earning coefficient increases and the book value coefficient becomes lower and not statistically significant.

22. Robustness Alternative way of coping with extreme values
Excluding the banks that voluntarily adopted IFRS before 2005.

23. Concluding remarks
Did the mandatory application of IAS/IFRS increase the value relevance of accounting information of the European banks?
As we expected, the marginal effect (value relevance) of earnings on share prices after the IFRS were made compulsory increased for the entire sample. This is a robust result. The largest incremental effect was in Germany and Italy, the smallest in the United Kingdom.
For equity book value, our results are less clear-cut. The marginal effect of this variable became negative (for the unbalanced panel) or insignificant (the balanced panel).
The last result is driven by small and more opaque banks.

24. The segmentations’ results suggest that data for more opaque banks explain the negative effect of the book value Restricting the field to the biggest or more transparent banks the value-relevance of both earning and book value increases with the switch-over. By contrast, the introduction of the new accounting standards does not appear to have enhanced the quality of accounting information for the smaller and more opaque banks.
However, the results for the entire sample do not support the thesis that IAS produced a significant increase in the value relevance of book value.
This calls for future research connected with possible effects of two recent developments in the banking industry: the diversification of some banks (money centre//financial marketplace banks) into trading more liquid securities and the enormous growth of securitization and financial crash.