In contrast to many studies of the general population, all but one of the empirical studies (10 out of 11) uses longitudinal panel data that includes measures of religious denomination and church attendance. Notably, the study relying on cross-sectional data uses couple fixed effects models to minimize selection bias. All of the quantitative studies control for an extensive set of demographic and economic characteristics associated with marriage. The studies that use the Fragile Families and Child Well-Being (FFCW) data (described subsequently) to study marriage formation and dissolution also control for relationship quality and attitudes toward marriage.
The majority of the studies (8 out of 11) draw on the FFCW Study, a unique research project that focuses on unmarried parents who recently had a child together. These families are known in the research and policy literature as fragile families. The FFCW Study measures several dimensions of relationship quality and attitudes, includes a comparison group of married parents, and contains some information about whether each partner has other children with the same or different partners (known in the literature as multipartnered fertility). At baseline, approximately two-thirds of the sample had incomes below 200 percent of the poverty line (McLanahan et al., 2003).
In addition to the FFCW study, other data sources used in the marriage studies include the National Longitudinal Survey of Youth (NLSY); Welfare, Children, and Families (WCF); and the Marital and Relationship Survey (MARS). All of these studies include a low-income subsample drawn from a national data set or a sample drawn from low-income communities.