A short paper on the decline of all things marriage in the U.S.

I took a short break from my other deadlines and wrote a short paper on the decline of all things marriage in the U.S. “All things” is the important part of that phrase. The decline of marriage is well-trodden territory, but the decline of divorce, and also — oddly — widowhood is what makes this curious to me.

Is there such a thing as the “marriage metabolism” of a society, and if so, what does it mean? People have used the concept in the past, but they only included rates of marriage and divorce. Digging into the history of this research and its etiology (especially the “retreat from marriage” — a term I think demographers never should have used), I think the metabolism concept was meant to capture the choices people make about marriage, so they excluded death and widowhood. That’s a mistake, because if you’re interested in the properties of the system, why would you leave out the main way people exit from it? It doesn’t matter whether a person choses widowhood, what matters is their widowhood shrank the body of married people and increased the pool of unmarried people. So I made a measure that includes all the “marital events” in the American Community Survey: marriage, divorce, and widowhood.

The details are below, but the gist of it is I used a life table approach. I used the ACS data to calculate the proportion of people at each age who got married, divorced, or widowhood. Then I took the U.S. life table, which tells you how many people out of a hypothetical cohort of 100,000 births are still alive at every age. By multiplying those proportions times the number of survivors, at each age (the Lx column, formally), I can see how many people out of the original 100,000 experience each event. If you add up all the ages, you get the lifetime number of events per person. Fun!

The details are in the paper. The gist of it is the Total Rate of Marital Events fell about 24% in 12 years. That seems like a lot to me. Also, the total number of marriages per person fell below 1.0 during the time, which may be something of a milestone. Conclusion: “Thus, one partial explanation for falling marriage, divorce, and widowhood rates may be greater selectivity into marriage, with fewer people achieving a more desirable status – and as a result exiting that status less often. A higher status marriage system is a smaller, slower, and more stable marriage system.”

Anyway, the Stata code, IPUMS data extract, and Excel file (sorry!) are in the supplemental materials, here: https://osf.io/at64y/. After I get back to my other work for a little while, and see if there’s any feedback on this, I’ll fix it all up to be a little more presentable (and unless you convince me I’m wrong, submit it somewhere).

I updated this post on 30 Nov 2022, if you want to be sure you’re seeing the latest version, you can get it on SocArXiv here. The abstract and key figure are below.

---

Rethinking marriage metabolism: The declining frequency of marital events in the United States

Abstract

Previous research has employed an inadequate measure of marriage metabolism, but the concept may be useful for understanding the system of marriage. This paper addresses changes in the incidence of marital events in the United States from 2008 to 2020. I offer a measure, the Total Rate of Marital Events (TRME), that captures the lifetime experience of marital transitions (marriage, divorce, and widowhood) for a life table cohort. I find that the TRME declined steeply over this relatively short period: 25 percent for men and 23 percent for women. All three components declined in every age group below 90. I suggest that the slowing churn of the marriage system reflects the diminished social presence of marriage in daily life – if not its declining importance – which coincides with the increasingly selective status of married life. A higher status marriage system is a smaller, slower, and more stable marriage system.

Figure 3. Cumulative rates of marriage, divorce, and widowhood by age for men (A) and women (B), showing 2008 (solid lines) and 2020 (dotted lines). Rates from single-year American Community Survey data applied to life table person-years lived to produce lifetime events per person. See text.