ABSTRACT
Empirical studies analysing the determinants of bank failures ignore the role of bank holding company affiliation in these failures. In this article, we propose a new approach of estimating affiliated banks’ failures that incorporates holding companies’ role in failures. Our logit regression results show that the holding company’s features, especially internal flows, mattered on failures more than those of the banks during the sub-prime mortgage crisis.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 Regulators in the United States use a summary measure to evaluate banks, called the CAMELS rating, based on six factors: capital adequacy, asset quality, management quality, earnings, liquidity and sensitivity to market risk.
2 Like CW, we also control for loan portfolio indicators (results are available upon request).
3 Bankers would respond to changes in their environment in making their daily management decisions, including changes in their likelihood of default. For this reason, using contemporaneous variables in the estimations may cause an endogeneity problem. Authors such as CW, Antoniades (Citation2015) and Berger and Bouwman (Citation2013) address this potential issue by using pre-crisis values of the banks’ CAMELS indicators in estimating bank failures. (For details of this discussion, see Antoniades (Citation2015).) We follow their approach in this paper and use pre-crisis average values of our independent variables in Equations (1)–(3).
4 Although it would have been interesting to extend our results to examine multinational banks, limitations of our data prevent us from doing so. Our sample includes only 49 multinational banks for which data are available for both the bank itself and its parent institution. Due to the large number of independent variables, which includes the loan portfolio variables that are not reported, we are not able to analyse these multinational banks separately and further compare their results with banks that are not multinational (which we call domestic banks). However, we found that once we exclude the multinational banks from our sample, our results for domestic banks are very similar to the results reported previously. These results for domestic banks are available upon request.