Are the financial characteristics of acquired banks similar across the EU? Evidence from the principal markets

Abstract

We use a sample of acquired and non-acquired commercial banks from the principal EU markets and logistic regression analysis to investigate the relationship between bank characteristics and the likelihood to be acquired. The results indicate the existence of differences across countries either in terms of the significance of the variables or the sign of their coefficients.

Notes

¹ Acquisitions were identified in Bankscope database, BANKERSalmanac and Zephyr. To be included in the sample acquired banks should have available financial data in Bankscope for 2 years prior to the acquisition. The control sample of the 373 banks includes all non-acquired commercial banks, as of end 2002, with available data in Bankscope for the period 1995–2002.

² Other studies that have used a case–control approach are the ones of Meric et al. (1991), Hadlock et al. (1999), Pasiouras and Gaganis (2007) among others.

³ For example, in the case of Italy, acquired banks in 1998 (3) represent 8.57% of total acquired banks in Italy (35). Hence, from a total of 47 non-acquired banks that were available for Italy, 8.57% are assigned to 1998, that is four banks.

⁴ Obviously, there are numerous other variables that could be considered. For example, Hadlock et al. (1999), analyse the effect of variables related to management incentives. Owing to data availability, our study is restricted to financial variables. We hope that future research will improve upon this.

⁵ Considering that bank acquisition can take some time to complete, when estimating our models we assume, as in previous studies, that acquisitions completed during year t, reflect bank's and market characteristics during year t − 1.

⁶ Prior to the logit estimations we weighted the data to compensate for differences in the proportion of acquired and non-acquired banks in the sample. Furthermore, we smoothed all financial variables by replacing all observations above the 95th percentile and below the 5th percentile with these values.

Table 3. Logistic regression results