We have a paper forthcoming in the Journal of Marriage and Family about cohabitation and marital dissolution (Rosenfeld and Roesler 2018). You can find the prepublication version of our paper here or linked from Rosenfeld’s website here. We used the National Survey of Family Growth (NSFG), and we found, as many other researchers have found, that premarital cohabitation is associated with a higher risk of marital dissolution at ten years’ marital duration. We also discovered that in the first year of marriages, couples who had cohabited before marriage had a lower rate of marital dissolution, presumably because the cohabiters had the benefit of experience at living together.

Kuperberg’s blog post argues for a specific operationalization of age of relationship’s initiation. For married couples who did not cohabit before marriage, she defines the age at marriage as the beginning of the relationship. For couples who cohabited before marriage, she defines the age at cohabitation before marriage as the age of relationship initiation. Since cohabitation precedes marriage (for those who cohabited before marriage), Kuperberg defines the cohabiters as having an earlier relationship start. The earlier relationship start, in her view, explains away the difference in later divorce rates (and see also Kuperberg 2014 where a similar argument is made). Kuperberg’s way of operationalizing the age of relationship initiation is not standard in the scholarly literature on cohabitation and divorce, but it is interesting and worth examining.

The earlier people marry, the less time they have to gather information about potential partners, and the less time potential partners have to gather information about them (Oppenheimer 1988). People who marry later are more established in their careers and presumably more mature at the time of marriage. Earlier age at marriage is associated with higher subsequent rates of divorce.

Since premarital cohabitation typically precedes marriage by a year or two, the age at first cohabitation is earlier than the age at first marriage. Nonmarital unions have much higher breakup rates than marital unions (Cherlin 2009; Rosenfeld 2014). Expectations of sexual exclusivity are different before and after marriage. In the NSFG data it is unknown when relationships started, when couples made the promise to marry, or when (if ever) couples first made the promise of sexual exclusivity. It is safe to assume that most couples who married but did not cohabit before marriage had a period of engagement before marriage, but the length of that engagement is unknown in NSFG. The age at first marriage (for couples who did not cohabit before marriage) is therefore not functionally the same point in the relationship, in terms of commitment and search and exclusivity, as the age at first cohabitation (for couples who cohabited before marriage).

For couples who did not cohabit before marriage in Kuperberg (2014), their cohabitation began when the marriage began. Treating the beginning of the cohabitation as the age when the relationship started would be consistent (for cohabiters and for non-cohabiters) if the outcome variable included breakups of cohabiting couples as well as breakups of marriages. In Kuperberg (2014), the outcome is divorce, meaning only married couples were at risk for breakup in Kuperberg’s analyses. Cohabiting couples started their relationships earlier but were not at risk of divorce until they were married. Using age at first cohabitation as the age the relationship started in combination with an outcome variable that ignores cohabitation breakups potentially biases the effect of cohabitation to be less associated with marital dissolutions. This is why Kuperberg’s analyses show no effect of premarital cohabitation on later risk of divorce.

If the NSFG included a similar commitment age for the non-cohabiters, perhaps the age at which the couple were engaged to be married, then the age of commitment to marry could be consistently controlled for across the subsamples of cohabiters and non-cohabiters. Unfortunately, NSFG did not include consistent measures of age at engagement for all women who subsequently married.

We agree with Kuperberg that controlling for age of first cohabitation or age at first marriage (whichever came first) appears to lower the effect of premarital cohabitation on later risk of divorce. Depending on what other controls are included, the relationship between premarital cohabitation and later divorce can be reduced to zero. We don’t agree that Kuperberg’s approach is the correct or the only reasonable approach.

Table 1: Replacing age at marriage with age at first cohabitation or marriage

Source: NSFG data on first marriages for women age 44 and under, NSFG wave 1988 and later.

Note: Additional controls not reported above: marital duration  (1df), year of marriage (1df) marital duration first calendar year dummy variable (1df), premarital cohabitation interacted with first year of marriage (1df), age at marriage for M1 (for M2 and M3, age at first cohabitation with spouse) (categorical, 3df), presence of children under 18 (1df), calendar decade (5df), educational attainment (3df), race (2df), stable family of origin (1df), mother’s education (3df), NSFG wave (5 df). N of couple-years is 216,455. N of couples is 24,888.

*** P<0.001; ** P<0.01; * P<0.05; + P<0.10, two tailed tests.

Table 1 starts with the association between premarital cohabitation and later odds of marital dissolution from Rosenfeld and Roesler’s (2018) Table 2, Model 5. The log odds ratio coefficient was 0.29, and highly significant, indicating couples who cohabited before marriage had an odds ratio of marital dissolution of e^0.29=1.33 times higher, after the first year of marriage, compared to couples who never cohabited before marriage. If we replace age at marriage with age at marriage or age at first cohabitation (whichever came first), then the coefficient for premarital cohabitation’s association with marital dissolution drops from 0.29 to 0.16, but remains statistically significant.

Table 2: Replacing age at marriage with age at first cohabitation or marriage, while also controlling for nonmarital cohabitations

Source: NSFG data on first marriages for women age 44 and under, NSFG wave 1995 and later.

Note: Additional controls not reported above: marital duration (1df), year of marriage (1df) marital duration first year dummy variable (1df), premarital cohabitation interacted with first year of marriage (1df), age at marriage or age at first cohabitation (categorical, 3df), presence of children under 18 (1df), calendar decade (4df), educational attainment (3df), race (2df), stable family of origin (1df), mother’s education (3df), NSFG wave (4 df), nonmarital cohabitation (2df). Premarital cohabitation is cohabitation with the woman’s first husband before marriage. Nonmarital cohabitation is prior cohabitation with men other than the man who would later become the woman’s first husband. N of couple-years is 167,723. N of couples is 19,777.

*** P<0.001; ** P<0.01; * P<0.05; + P<0.10, two tailed tests.

Table 2 adds additional controls for nonmarital cohabitation. If we control for nonmarital cohabitations, which are much more strongly associated with marital dissolution than are premarital cohabitations (Rosenfeld and Roesler 2018; Teachman 2003), the association between premarital cohabitation and marital dissolution is reduced to a log odds ratio coefficient of 0.13. The smaller association between premarital cohabitation and marital dissolution in Table 2 (compared to Table 1) allows us to replicate Kuperberg’s central finding. In Model 2 of Table 2, controlling for age at first cohabitation reduces the association between premarital cohabitation and marital dissolution to zero.

Conclusion:

While we agree that using the age of first cohabitation instead of the age of first marriage can reduce the apparent association between premarital cohabitation and marital dissolution to zero, we do not agree that this is the best or the only reasonable approach. As we are examining marital dissolution (rather than relationship dissolution), the beginning of the marriage is a start time that has consistent meaning for all couples whether they cohabited before marriage or not. The beginning of cohabitation (for couples who cohabited before marriage) is not the same relationship decision point as the beginning of the first marriage (for couples who did not cohabit before marriage). Whether one chooses to control for the age at marriage or the age at first cohabitation depends on one’s substantive interpretation of the meaning of relationship milestones. We appreciate Kuperberg’s thoughtful and innovative work. We disagree with her conclusions, because we do not think that age at first cohabitation is a consistent relationship milestone for couples who did and who did not cohabit before marriage.

References:

Cherlin, Andrew J. 2009. The Marriage-Go-Round: The State of Marriage and the Family in America Today. New York: Knopf.

Kuperberg, Arielle. 2014. “Age at Coresidence, Premarital Cohabitation, and Marriage Dissolution: 1985-2009.” Journal of Marriage and Family 76:352-369.

Oppenheimer, Valerie Kinckaid. 1988. “A Theory of Marriage Timing.” American Journal of Sociology 94:563-591.

Rosenfeld, Michael J. 2014. “Couple Longevity in the Era of Same-Sex Marriage in the US.” Journal of Marriage and Family 76:905-918.

Rosenfeld, Michael J., and Katharina Roesler. 2018. “Cohabitation Experience and Cohabitation’s Association with Marital Dissolution.” Journal of Marriage and Family.

Teachman, Jay. 2003. “Premarital Sex, Premarital Cohabitation, and the Risk of Subsequent Marital Dissolution among Women.” Journal of Marriage and Family 65 (2):444-455.