Determinants of profitability of community banks in the USA: a cost-frontier-based decomposition approach

An exception is the normalized quadratic functional form, in which case global curvature can be imposed using Cholesky factorization (see, for example, Feng and Serletis 2008).

Following Diewert (2014), we choose the mixed vectors \(\left( { {\varvec{q}} }_{it}{, {\varvec{w}} }_{ht}\right) \) and \(\left( { {\varvec{q}} }_{ht},{ {\varvec{w}} }_{it}\right) \) in order to have cost decompositions into explanatory factors that are multiplicatively complete.

See Footnote 23 of Diewert (2014) for more details.

Strictly speaking, \(\mu _{i}\) in (10) is the sum of the fixed effects and the constant term of the log-linear function.

An alternative approach is the user cost approach (e.g., Diewert et al. 2012). This approach is theoretically more appealing. However, to keep consistency with previous studies in the literature, we use the commonly accepted intermediation approach in this study.

This can also be seen from the density plots of the cost efficiency for both the community banks and large banks in Figure 1. From this figure, we see that on average community banks were moderately less cost efficient than large banks. In addition, the distribution for large banks is more dispersed than that for community banks.

Note that we do not report the price index for nontraditional income and the price index for fixed assets, because the former is unity and the expenditure on fixed assets only represented a very small fraction of the total costs.

We would like to thank a referee for pointing out this issue.

For example, Hugh and Mester (2013) consider banks with assets greater than $100 billion as being TBTF. In contrast, Davies and Tracey (2014) argue that “size is not the only determinant of TBTF; some smaller banks received government assistance during the crisis.” This latter study instead considers a bank to be TBTF if its ratings uplift by ratings agencies is greater than zero. In addition, there is no consensus about how to quantify the implicit TBTF subsidies. For example, Ueda and Weder di Mauro (2012) quantify a structural subsidy based on information contained in credit ratings; Noss and Sowerbutts (2012) employ a contingent-claims–based approach as a measure of an implicit subsidy, while Davies and Tracey (2014) assume that ratings-sensitive liabilities are funded at the support credit rating, but in the absence of any implicit government support banks would need to fund themselves at the stand-alone credit rating.