Value-Driving Activities in Euro-Zone Banks

Abstract

We develop and test accounting-based valuation models for commercial banks. We extend Begley et al.'s framework (2006) and propose a valuation model where goodwill is generated by virtually all commercial and investment banking activities. Key features of our model are: the development of a relation between future cash flows from fee income and the bank value that depends on lending, borrowing and off-balance sheet business; and the inclusion of proprietary investment and trading as value-driving activities. Empirical tests on a sample of Euro-zone banks from 1998 to 2006 provide the following evidence. Unrealised expected cash flows from fee income are the most important source of unrecorded goodwill. This is consistent with the increasing importance of revenue from the sale of financial services to banks' income. The contribution of fee income to goodwill is higher for banks with large deposits and new loans. Equity securities are a source of unrecorded goodwill, but the introduction of fair value accounting, with the adoption of the International Financial Reporting Standards (IFRS), reduces their valuation role. Yet equity securities remain positively associated with unrecorded goodwill after IFRS adoption, suggesting that the fair value standards do not fully capture market expectations about future cash flows of investment assets.

Notes

For details on the derivation of the models see Begley et al. (2006).

Legal acts such as the Second Banking Directive (1989), Financial Services Action Plan (1999) and the Economic and Monetary Union (1990–99) permitted European banks to operate as universal banks and to harmonise activities.

Banks engage actively in off-balance sheet instruments to enhance profitability (by increasing income without increasing net assets), to enlarge activity without the need to increase cash asset reserves (in the EU off-balance sheet business does not require more cash asset reserves – Casu et al., 2006, p. 229), and to boost their relationships with customers (Nissim and Penman, 2007).

See Appendix A for details.

The European Central Bank reports that income from trading in securities in EU banks has more than doubled between 1994 and 2003 (ECB, 2005). For details on the importance of trading and investment income in the sample, see Table 2.

IAS 39 defines fair value as ‘the price at which an asset or liability could be exchanged in a current transaction between knowledgeable, unrelated willing parties’. This definition implies that fair value does not depend on the existence of active markets. Further, even if active markets exist, the fair value of the asset need not equal its market price. (See Landsman (2006) for a detailed discussion of implementation issues of fair value accounting.)

Available-for-sale investments can represent a substantial part of European banks' financial instruments. Although IAS 39 conceptualised available-for-sale as a residual category many European banks often choose that category to avoid the reclassification restrictions imposed by ‘financial assets at fair value through profit or loss’.

Research on income reporting presents evidence supportive of both investors understanding comprehensive income (Chambers et al., 2007) and investors failing to capture items reported outside net income (Lee et al., 2006; Hunton et al., 2006).

An example is a marked-to-market asset that the bank has the intention and ability to hold with a market price of 60 and a present value of expected cash flows of 100. IAS 39 permits the reclassification of the asset at cost or amortised cost, and remeasures it at 100. Note, however, that the financial markets turmoil and the amendments to IAS 39 occurred at the end of 2008 and thus these events are not likely to affect our analysis.

Note that the model refers to proprietary investment and trading that distinguish between similar services that banks provide to customers, namely, asset management. Hence investment and trading assets do not generate fee income. See Appendix B for details.

The book value of equity is a commonly used scaling factor and it was used in Begley et al. (2006). We acknowledge Barth and Clinch's suggestion (2009) that generally the ‘number of shares outstanding’ effectively mitigates the scale effects but practical reasons prevent us from using this deflator. Our main source of data (Compustat Global Financial Services) does not provide data on shares outstanding and manual collection from Worldscope results in missing data for many sample banks. Also, Barth and Clinch find that the performance of deflators varies with the type of scale effects that suggests that number of shares outstanding may not be the best deflator in all circumstances. We experimented with other scaling factors for which data was available, namely, total assets and market value and concluded that book value of equity deals more effectively with scale problems.

The relatively high level of holdings in debt securities in Euro-zone banks is explained by the European Central Bank (ECB, 2004): (1) European firms and individuals rely strongly on their banks to manage their savings and investments, and (2) they have a preference for long-term fixed-income investments. As a result, European banks end up with a large surplus of resources that is mostly reinvested in debt securities.

In line with the theoretical models, we do not include a constant. We re-run all regressions with: (1) a standard constant, (2) a constant equal to 1/bvce (i.e. constant of 1 scaled by book value of equity) and (3) a time-varying constant equal to 1/bvce * year dummies (similar to Begley et al., 2006). In all cases, the results remain materially the same.

Belsley et al. (1980) suggest a cut-off value of |DFITS| > 2 * sqrt(p/n), where DFITS represents the observation height and residual, n is the number of observations and p the number of regression parameters.

Estimating the models with country and year fixed effects (i.e. including indicator variables) generally yields the same results but we opt for clustering because this is more likely to produce unbiased standard errors (Peterson, 2009). However, clustering by country and year becomes problematic when we work with two subsamples because a small number of clusters and observations per cluster impair the efficiency of clustered standard errors. For this reason we report t-statistics in Tables 6 and 7 based on standard errors clustered by the most frequent dimension (country) and more observations by cluster as suggested by Peterson (2009). Clustering by two dimensions does not materially change the results. We also repeat the main results in Table 5 eliminating observations for the three countries that have only one bank (Belgium, Luxembourg and the Netherlands) and find no major differences.

The negative effect is not attributed to the correlation between new and lagged deposits because when we estimate the model without the interaction term, * , it yields the same result.

A good illustration of the significant impact of IFRS on banks' financial statements is the strong opposition of the European banking sector to IAS 39 implementation. The antagonism towards the application of the standard reached its peak when French President Jacques Chirac wrote to Frits Bolkenstein, the EU commissioner, declaring that the adoption of the rules would be ‘a disaster’ for French banking.

For each bank and year we identify the standards used in the preparation of financial statements and accordingly divided the sample into IFRS/non-IFRS financial reports. The IFRS subsample includes not only financial reports presented after 1 January 2005 when the standards became mandatory in Europe but also those that voluntarily adopted the standards before that date.

We used other size partition criteria, specifically above and below mean total assets, and quintiles 1 and 2 vs. quintiles 3 and 4 (quintiles based on total assets). Results do not materially change.

An alternative approach is to collect all the terms that interact with fee _t directly under this variable. However, this results in a more complex and less intuitive solution.

Note that dep _t−1 (deposits at the beginning of period t) is equivalent to ldep _t .

Additional information

Notes on contributors

Helena Isidro

Paper accepted by Salvador Carmona.

David Grilo

Paper accepted by Salvador Carmona.