Are the less productive banks catching up to the more productive ones and, if so, how quickly and by what means? The objective of this study is to answer these questions by looking for convergence in productivity among bank holding companies (BHCs) in the US Past research has identified two major factors governing productivity in the banking sector – scale economies and X-efficiency. If the gains from scale economies decline with firm size and if the only difference between BHCs lies in their initial size, then the initially smaller BHCs should eventually catch up to the initially larger ones because the former tend to grow more quickly. However, the findings from this study do not support this hypothesis of “absolute convergence”. Indeed, the findings show strong evidence for “conditional convergence”, which means that the steady-state productivity to which a BHC is converging is conditional on the BHCs own level of X-efficiency. Conditional convergence implies that initial differences in X-efficiency among BHCs can, between them, create permanent differences in steady-state productivity.

Labor statistics show that the average labor hours per dollar of banking output fell by more than 30% between 1992 and 2002. The proliferation of labor-saving technologies was widely believed to be the major reason. While the first-round effect of a labor-saving technology with a given level of output is a reduction in required labor per unit of output, the second-round effect is a reduction in wage costs that will increase output. Analytically, a given type of labor-saving technology is more likely to have a positive effect on employment if the elasticity of substitution between capital and labor, the price elasticity of demand, and the cost-reducing impact of the new technology are sufficiently large. The main empirical findings of this study are that labor-saving technologies, and the spillovers of these technologies, are associated with higher firm-level employment. These results seem robust to a wide range of specifications and controls.