M&A Operations and Performance in Banking

Supplementary results associated with this article can be found in Beccalli and Frantz (2008).

As noted in Goddard et al. (2007), the legislation impacting on the EU banking and financial sectors includes among the others: Consolidated Banking Directive (2000), Financial Conglomerates Directive (2002), New EU Takeover Directive (2004).

In order to test for robustness, we also used the pre-merger performance of the hypothetical consolidated bank, obtained as aggregate net income divided by aggregate total equity (acquirer+target), when performance is measured by ROE. Our results were unaffected by this change.

In order to test for robustness, we also used industry medians for the industry adjustment figure when performance is measured by ROE. Our results were unaffected by this change.

This potential problem though is common in the investment literature when both a portfolio’s α and β are jointly estimated from the same regression.

We are grateful to Robert DeYoung for this suggestion.

We also tested for other variables on bank-level characteristics: quality of loan portfolio (proxied either by total problematic loans or by loan loss provision) and equity capital. However these bank-level variables are not proven to be statistically significant either for cost or for profit efficiency changes.

As stated by Palepu et al. (2007), “In computing ROE, one can either use the beginning equity, ending equity, or an average of the two. However, for most companies, this computational choice makes little difference as long as the analyst is consistent. This comment applies to all ratios [...] where one of the items in the ratio is a flow variable and the other one is a stock variable”. In this manuscript, we use beginning balances of equity for both ROE and CFR.

Although off balance sheet items are not earning assets, they do represent an increasing source of income for all types of banks and are therefore included in order to avoid understating total output (Jagtiani and Khanthavit 1996).

\(z_i = \mu _i \left( {\ln Q_i + w_i } \right)\), where μ _i and w _i are scaling factors, limiting the periodic sine and cosine trigonometric functions within one period length 2π (see for a discussion: Gallant 1982; for an application: Mitchell and Onvural 1996).

We have also tested for a larger set of control variables (real growth in GDP, population, and concentration index). Between these frontiers (including either one control variable or the set of control variables), we found large, positive, and significant rank order correlation. Thus we have chosen to minimise the number of control variables and have kept the specification with a country dummy only.

The latter result is however not robust to alternative specifications of ROE.

The industry-adjusted values show that banks involved in M&A operations are more (cost and profit) efficient than banks not involved in M&A (control sample) on average in any year (1991–2005) and in most of the countries under investigation. Further details can be found in Beccalli and Frantz (2008).

Empirical evidence on robustness is available from the authors on request.

An alternative base period (consisting of the 3years prior to the deal) was considered throughout the analysis. All the results are confirmed.

As in much of the literature (Berger 2007), the terms “developed” and “developing” countries are based on International Monetary Fund (IMF) classifications: developed countries are “high-income” countries, developing countries are “middle income” or “low income” countries.

The authors would like to thank Allen Berger for suggesting such an interpretation.

Moreover, we investigate the correlation between accounting performance and size as measured by total assets in the population of banks not involved in mergers and acquisitions. Correlation is shown to be very low. For instance, when performance is measured by returns on equity, the Pearson correlation coefficient is equal to 2.4% whilst the Spearman correlation coefficient is equal to 13.7%.

Accordingly to the analysis made for ROE and CFR, in order to make sure that our results are not driven by failure to control for size, we test the correlation between the (profit and cost) efficiency and the (profit and cost) efficiency measure normalised by equity. We find large, positive, and significant rank order correlation for each year before and after the deal. This correlation suggests that the efficiency ordering generated by the unadjusted estimates is consistent with the efficiency ordering generated by the normalised estimates. However the magnitude of the correlation coefficient for profit efficiency (about 0.75 significant at 1% level) is lower than the magnitude of the correlation coefficient for cost efficiency (about 0.90 significant at 1% level). Therefore we further investigate any scale bias effect on (profit and cost) efficiency by testing the correlation between efficiency and size (proxied by ln total assets) for the control sample. The magnitude of this correlation is very low (about 5% significant at 1% level), which suggests that there is no failure to adjust for size the (profit and cost) efficiency estimates. Furthermore, we use the OLS regressions outlined in Eq. 2). The value of the intercept α becomes positive and statistically significant at 1% level, but the coefficient γ is found to be negative and statistically significant. This suggests that the larger the size of the acquirer, the lower the profit efficiency associated with the M&A (Table 6 [Panel G).

A table with the precise definition of each variable (and the relevant statistics) is provided in Beccalli and Frantz (2008).