There is considerable disagreement about whether cohabitors are more or less likely to be educationally homogamous than married couples. Using data from the National Longitudinal Survey of Youth, I reconcile many of the disparate findings of previous research by conducting a "stock and flow" analysis of assortative cohabitation and marriage. I find that cohabitors are less likely to be educationally homogamous than married couples overall, but these differences are not apparent when cohabiting and marital unions begin. Instead, the results suggest that differences in educational homogamy by union type are driven by selective exits from marriage and cohabitation rather than by differences in partner choice. Marriages that cross educational boundaries are particularly likely to end. The findings suggest that although cohabitors place greater emphasis on egalitarianism than married couples, this does not translate into greater educational homogamy. The findings are also consistent with a large body of research on cohabitation and divorce questioning the effectiveness of cohabitation as a trial marriage.