Previous studies of the mortgage lending process, which fail to adequately account for inter-lender differences, suffer from a potentially serious omitted-variables bias. Using a sample of 23,094 home purchase mortgage applications from Pinellas County (St. Petersburg) Florida, over the 1993–1995 interval, I demonstrate that the inclusion of lender characteristics enhances the predictive power of accept/reject decision models of mortgage origination. In addition, the inclusion of lender attributes is shown to substantially influence the coefficient estimates of additional model parameters of interest, such as the applicant race and neighborhood racial composition variables. Finally, this study provides unique and supportive evidence regarding the existence and persistence of information externalities in home purchase mortgage markets.