Social scientists have devoted substantial research to economic basis for matching of men and women in marriage. A common feature of existing studies is their reliance on samples of married couples. The principal shortcoming of spouse data is that spouses' earnings correlations are contaminated by the partners' behaviors and other events that occur after marriage and affect their earnings. This study addresses that problem by exploiting a longitudinal data file from the Swedish population. By selecting a sample of married couples in a given year, we retreat through the file to years before the marriage. Using data from the spouses' single years, we apply the correlation methodology to their earnings. Evidence from the model supports positive assortative mating. © 2004 by the Board of Regents of the University of Wisconsin System.