This paper compares the productivity and efficiency of large banks and community banks in the United States over the period 1997–2006. This comparison is performed by estimating a true random effects stochastic distance frontier model—a model that is capable of disentangling unobserved heterogeneity from inefficiency—within a Bayesian framework. We find that failure to consider unobserved heterogeneity results in a misleading ranking of banks and mismeasured technical efficiency, productivity growth, and returns to scale. Our results show that, compared with community banks, large banks have experienced much higher productivity growth and higher levels of returns to scale. Our estimates of total factor productivity growth show a clear downward trend for both large and community banks, and our decomposition of the output-distance-function-based Divisia productivity index indicates that technical change is the driving force behind this trend.

This paper investigates the returns to scale of large banks in the US over the period 1997–2010. This investigation is performed by estimating a random coefficient stochastic distance frontier model in the spirit of Tsionas (2002) and Greene (2005, 2008). The primary advantage of this model is that its coefficients can vary across banks, thereby allowing for unobserved technology heterogeneity among large banks in the US We find that failure to consider unobserved technology heterogeneity results in a misleading ranking of banks and mismeasured returns to scale. Our results show that the majority of large banks in the US exhibit constant returns to scale. In addition, our results suggest that banks of the same size can have different levels of returns to scale and there is no clear pattern among large banks in the US concerning the relationship between asset size and returns to scale, due to the presence of technology heterogeneity.