Mandatory IFRS adoption and the cost of debt in Italy and UK

Abstract

This paper analyses the effect of the mandatory adoption of International Financial Reporting Standards (IFRS) within the EU on the cost of corporate debt. In order to avoid the imprecision involved in a large-scale cross-country study, we examine the impact of IFRS in two very clearly different institutional settings, the UK and Italy. The UK is a common-law country characterised by strong enforcement and national generally accepted accounting principles (GAAP) which are equivalent to IFRS. Italy is a typical European code-law country, characterised by a weak outside investor protection system, and national GAAP significantly different from the IFRS model. No IFRS effect is observed in the UK, consistent with it having standards which are close to IFRS. During the post-IFRS period, in Italy more weight is placed on the accounting numbers to assess the cost of debt. We also find that accruals quality improves in Italy, thus suggesting that public financial reporting data are enhanced relative to privately held information about borrowers' credit ratings.

Keywords:

• cost of debt
• IAS/IFRS
• mandatory adoption
• standard setting

Acknowledgements

We are grateful to Gunther Gebhardt, Stuart McLeay, the participants of the Workshop on Accounting and Regulation (Siena, 2010), the Eufin Annual Workshop (Stirling, September, 2010), the European Accounting Association Annual Congress (Rome, 2011) and the Loughborough Symposium in Financial Reporting (Loughborough, 2012) for useful comments and suggestions. We would also like to thank the anonymous reviewer whose many invaluable comments considerably sharpened the exposition and improved the substance of the paper.

Notes

1. Francis et al. (2005) and Liu and Wysocki (2007) scale the standard deviation of NIBE by assets. We do not follow this procedure but instead take logs. There is no debt theoretic justification for scaling by assets; rather the purpose seems to be to reduce the effect of the outermost observations. However, since the standard deviation is likely to be small in relation to assets, the scaling procedure may, in fact, obscure any effect of the variation in NIBE on the cost of debt. Therefore, we use the log of the standard deviation of NIBE. When, for comparison with these studies, we scale by assets, the conclusions are unchanged and are available from the authors.

2. It is worthwhile to stress that it is possible to capture this aspect by using ROA and Leverage ratios. When Leverage is divided by ROA, it gives debt/profit which is the value of debt as a multiple of profit, reflecting the indebtedness of the firm in relation to its performance. However, we do not use this approach, since it reflects the capital aspects of debt rather than the more relevant income aspects captured by the Interest Coverage ratio. Of course the Interest Coverage ratio is negative when operating income is negative. We interpret this as meaning that the more negative the ratio, the less likely is the return to profit as suggested by Joos and Plesko (2005), thus increasing the company-specific risk.

3. The variable Log_NIBE, a measure of variability, is not thought to be influenced by the adoption of IFRS since it is measured over a rolling prior five-year period. Therefore, it is not given an interactive term.

4. This follows common practice, since the formal Kolmogorov–Smirnov and Shapiro–Wilk tests for normality are unreliable in large samples. Variables are likely to fail these tests even though the deviation from normality is insufficient to make any real difference (Kline 2005, p. 63).

5. This result does not necessarily imply that the cost of debt born by the Italian companies is lower than in the UK. Indeed, it is generally understood that the average price of basic banking services in Italy is amongst the highest in Europe (BBA 2006). However, this is largely due to the non-interest components of cost, such as advice and internet banking, that are not captured by our empirical analysis (Drummond et al. 2007).

6. In addition to the pairwise correlation analysis, a variance inflation factor (VIF) test has also been developed to control for multicollinearity. The VIF indices are always far below the critical value of 10, so permitting the assertion that multicollinearity is not a troublesome problem (Hair et al. 1995). For the sake of brevity, the results of the VIF test are not tabled, but they are available from the authors.

7. Sánchez-Ballesta and García-Meca (2011, Table 6) report R²s of around 0.28, whilst Moir and Sudarsanam (2007, Table 5) and Liu and Wysocki (2007, Table 6) find much lower values of 0.06 and 0.005, respectively.

8. As with Table 2, the residuals show no signs of non-normality since the skewness and kurtosis values are below the value of 2 (see Garson 2012, Lomax and Hahs-Vaughn 2012).