Despite being criticized as unrepresentative and misleading, retention and graduation rates are an important part of college-search web sites and accountability systems, and they frequently have been used as indicators of institutional quality and effectiveness in educational research. Retention and graduation rates are often compared over time and across institutions. However, such comparisons can be confounded by differences in entering student cohorts and differences among the institutions being compared. This research examined the effects of institutional and cohort characteristics on one-year retention rates using random-effect and fixed-effect regression models for panel data. The use of a fixed-effect model allowed the researchers to account for omitted variables (unobserved heterogeneity) in the analyses. Results indicated that unobserved heterogeneity was a significant issue in the study, and that traditional regression methods may overstate the effects of institutional characteristics on retention rates. Results also indicated that the effects of institutional and cohort characteristics were essentially stable over time and across cohorts.