Statistical models were used to control for—., remove—the impact of socioeconomic factors that might account for the correlation between race/ethnicity and credit scores. The inclusion of such controls slightly weakened, but by no means eliminated (or accounted for) the association between minority status and credit scores. Among all such control variables, race/ethnicity proved to be the most robust single predictor of credit scores; in most instances it had a significantly greater impact than education, marital status, income and housing values. It was also the only variable for which a consistent correlation was found across all companies
