This paper covers one phase of a regression analysis of 117 bank stocks for the period 1945-52, which was conducted primarily to shed light on some current problems of banks in the capital markets. Since doubts have arisen concerning the reliability of regression analysis in such an application, this paper considers evidence of heterogeneity in the assumed underlying bank stock population and of auto-correlation in the residuals. Finally, it experiments with models designed in the hope of eliminating the worst effects of heterogeneity and auto-correlation.