This paper explores the role of marriage when markets are incomplete so that individuals cannot diversify their idiosyncratic labor income risk. Ceteris paribus, an individual would prefer to marry a "hedge" (i.e. a spouse whose income is negatively correlated with her own) as it raises her expected utility. The presence of love, however, complicates the picture. If love is very persistent, for example, and the resolution of uncertainty to agents' income is early, then those who in fact married hedges are the ones most likely to be caught short with too little love in order to save a marriage in the event of an adverse shock. Consequently, under these conditions individuals who are good hedges for one another are more likely to marry one another, although once married, they will be more likely to divorce. In contrast, if love is fleeting and the resolution of uncertainty to agents' income is predominantly later, then those who in fact marry hedges will in fact be less likely to subsequently divorce. Evidence is provided to distinguish which of these alternative scenarios is in support of these aspects of the decision to stay married. Additional hypotheses regarding the effect of differences in the expected means and volatilities of partners' incomes are also derived from the theory and tested.