We sought to replicate prior findings that admissions tests' underprediction of female college performance was driven in part by the omission of Big 5 personality factors from the predictive model, using 5,400 college students. We investigated gender differences in an elaborated model subdividing the Big 5 into ten aspects. We found differences at the aspect level that were not found at the factor level, and some aspects had unique relationships with academic outcomes. The findings demonstrated the effect of omitted variables on predictive bias.