Relation of divorce to years since first marriage and religion

Data from the General Social Survey, 1972 through 2024 (GSS) was used to look the degree to which ever having been divorced was related to years since first marriage and religious affiliation: No religion, Catholic, Protestant, and Other religion. Total sample size was 22,014.

A binary logistic regression examined whether ever having divorced was associated with religious affiliation and years since first marriage, using 22,014 cases. The model showed significant differences by religion, Wald χ²(3) = 141.27, p < .001. It also showed a nonlinear duration association: both the linear and quadratic duration terms were significant (ps < .001). The religion-by-duration interactions were not significant (p = .061 for the linear interaction; p = .549 for the quadratic interaction), providing little evidence that the duration pattern differed across religious groups.

Bonferroni-adjusted pairwise comparisons found significant differences between every pair of groups. Model-estimated proportions ever divorced were 26.1% for respondents reporting no religion (n = 1,197), 18.7% for Protestants (n = 14,540), 14.8% for the “Other” group (n = 849), and 11.0% for Catholics (n = 5,428). These are model-adjusted estimates evaluated at the model’s covariate means, not raw group percentages at a specific marriage duration.

Summary

• The analysis included 22,014 respondents from the General Social Survey, conducted from 1972 to 2024. The survey is designed to represent the U.S. adult population.
• The percentage of respondents who had ever divorced generally rises with the number of years since their first marriage for about 20 years, then levels off and declines.
• Having divorced is associated with religious affiliation. The percentage is highest among respondents with no religious affiliation, followed by Protestants, respondents reporting another religion, and Catholics.
• We found no clear evidence that this pattern differs across religious groups.
• These results compare people and marriage cohorts at one point in time; they do not track the same people as their marriages progress. The decline at longer durations could reflect differences between marriage cohorts, as well as years since first marriage.