On the mutual causalities between age at marriage and divorce

Abstract

The existence of an inverse relationship between age at marriage and probability of divorce has been repeatedly documented by a variety of empirical studies in the economics, demography, and sociology literatures [e.g., Becker et al., "An Economic Analysis of Marital Instability," JPE, 1977:85; Weed, "Age at Marriage as a Factor in State Divorce Rate Differentials," Demography, 1974:11; Lee, "Age at Marriage and Marital Satisfaction: A Multivariate Analysis with Implications for Marital Stability," Journal of Marriage and the Family, August 1977]. Although the causality from age at marriage to divorce is theoretically well determined [Becker, "A Theory of Marriage. Parts I and II," JPE, 1973:81, 82], not much has been said on the causation from divorce to age at marriage. A simple search model, however, demonstrates that optimal stopping for marriage (age at marriage) depends on the expected duration of marriage. If one assumes that the actual divorce rate shapes the expectations of the individuals on the duration of their potential marriages, one obtains the causality running from divorce rate to the age at marriage [Mocan, "Love, Marriage, and Dissolution: An Eclectic Approach," Mimeo, Graduate Center of the City University of New York, 1987]. Since the cross sectional studies, which focus on the estimation of divorce equations neglect the reverse causation from divorce to age at marriage, and include age at marriage among the set of exogenous variables, a potential bias arises in these estimations. This note provides the evidence on the existence of mutual causalities between divorce and age at marriage. The monthly observations between 1963-82 are obtained from The Vital Statistics of the U.S. The divorce rate (DIVORCE) is measured as the number of divorces per married women who are 18 years of age and over. YOUNGMAR measures the proportion of brides between ages 16 to 24 in their first marriages to the number of women in the same age group and reflects the tendency to marry early. To ensure stationarity, the variables are expressed in logarithms, and linear and quadratic trend variables are included. The system also contains 11 dummy variables to account for monthly variation. To test the mutual dependence of the variables in the Granger-Causality sense, the following system with 12 lags is estimated: