I compare returns to scale in the U.S. and Canadian banking system from 1996 to 2015. I estimate a parametric trans-log cost function and, for robustness, an inputoriented distance function. I do this in a way that is commensurate with the limitations of these models. Among the ten largest...