A public policy that promoted entry into marriage but decreased its stability would be deemed a failure. And yet in “Determinants of Long-Term Unions: Who Survives the ‘Seven Year Itch’?” Audrey Light and Yoshiaki Omori make the case that unions entered via cohabitation contribute to women’s chances of experiencing a union that lasts 12 years or more precisely because “the high probability of entering a cohabiting union more than offsets the relatively low probability of maintaining it for the long-term.” In other words, the high volume of cohabitation means that even with a short average duration, there will still be enough outliers that a significant fraction of long-term unions are cohabitations or began with cohabitation.

Outliers are easy to imagine. All I have to do is think about my grandmother who was married, divorced, and remarried before she graduated from high school. Later she divorced again, and was then married to husband number three for over 30 years before he died. Third marriages aren’t known for their stability, but some of them make it. Unions that start in cohabitation are similar, only there are more of them, so more of them make it (some last as cohabitations and more last after becoming marriages).

In other words, it is their quantity and not their quality that has made them a sizable fraction of long-term unions. Cohabitation is less stable than marriage, and marriages preceded by cohabitation are also more likely to end in divorce. Thus the proportion of women whose unions last for at least 12 years is far greater among women whose unions started with marriage than among those whose unions started with cohabitation (Tables 3-7). This fact alone obviates the generally accepted wisdom that it is better to drive before you buy.

Cohabitation is less stable than marriage, and marriages preceded by cohabitation are more likely to end in divorce.

I think that most reading “Determinants of Long-Term Unions: Who Survives the ‘Seven Year Itch’?” would conclude that although there are strong social forces determining which unions will make it for the long haul, public policy is not among them. On the one hand, this conclusion is fair: Light and Omori show that race, income, and teen childbearing all matter far more than “pro-marriage” environments. But their evidence falls far short of what would be necessary to substantiate that policy is of little relevance.

Most importantly, three of the four “environmental factors that can potentially be manipulated by public policy” that they consider have income-specific relevance: the maximum levels of welfare benefits, sex ratios (which vary the most where incarceration rates are high), and marriage penalties in state tax codes (which disproportionately affect those with higher incomes). Yet they estimate the impact on the average woman.  Only divorce laws seem likely to affect the average woman, and yet they conclude that the likelihood of long-term unions is largely invariant to changes in environmental factors.

None of the activities allowed by the Healthy Marriage Initiative are tested, so we still don’t know their impact on long-term unions. Plus there are a lot of policy options that aren’t marriage policy per se, but that could still matter for marriage through pathways that Light and Omori identify—like reducing teen childbearing. Even building walkable communities could be construed as marriage policy through reducing social isolation and shortening commutes to make more time available for family life. In short, the definition of a pro-marriage environment that they test is far too limited to conclude that environmental factors don’t matter much.

The evidence in this study falls far short of what would be necessary to show that policy is of little relevance.

Their focus on long-term unions rather than marriages per se also helps downplay the potential relevance of policy environments. According to their data, pro-marriage environments significantly increased women’s likelihood both of getting married and of staying married, but because of their negative impact on cohabitation, the overall impact on long-term unions was more modest than the impact on marriages.

Light and Omori make no claims about the association between long-term unions and healthy unions—in fact, they explicitly state that their study is silent on this question. But with an abundance of literature showing marriage premiums in domains ranging from infant health to men’s earnings to sexual satisfaction, we need to be careful not to assume that all long-term unions are equal.