First, some chit chat, then some data analysis.
I’m not good at converting demographic fact patterns into personal advice. However, choosing an age to get married to minimize your odds of divorce seems like a bad idea. Age is a factor in the aggregate, but at the individual level I would have to guess that marrying the right person at the right time for you matters more. And, in case it’s not obvious, the best way to minimize the odds of divorce is not marrying.
Like a needy would-be grandparent who badgers you to get married already (not that there’s anything wrong with that), some pundits promoting marriage will try to convince you that sooner is better. They are especially keen to advocate for this option because they know they’ve probably already failed to convince you to abstain from premarital sex. They might help you get into Heaven, but their earthly advice is mostly just virtue signaling to their followers — not really to make your life better. That report is from the National Marriage Project and its corrupt director, Brad Wilcox, whose religious conservative scholars have been turning out crap propaganda research for a decades. I won’t get into what’s wrong with the report, but I do like their happy-couple stock photos.
The stock photos give us a sociological glimpse into the inner lives of the happy couples. For example, that first head-butting pair may be Calm and Cool, but I think they’re only really happy because they’re Corrected. Maybe the secret to a happy marriage is actually doing carbon laser facials together?
On the other hand, the horsing around couple is afflicted by both heart disease and joint problems. Fortunately for them they’ve been married since they were 20 — they can get through anything!
Anyway, even though I like disagreeing with conservative family pundits, I am quite skeptical that the model I’m about to present is a causal one: that is, that the age at which women marry is a strong causal determinant of divorce — even if it is a reliable predictor. I think it’s likely that people who marry at different ages are selecting into better or worse marriages (or those more or less likely to break up.) Women who marry in their teens are in general probably not well situated to make their best match, or their best decisions (or even their own decisions). However, if everything happens to be lined up right, there’s no reason an early marriage can’t work. So, it would be wrong to turn this into advice that says, “Delay your marriage to that dream partner you’re already living with who would lay down their life for you forever and ever — because that will decrease your odds of divorce.”
So, here goes.
Age at marriage and divorce
1. What is Person X’s age? We’ll just take the people who are ages 15 to 59, but that’s optional.
2. What is this person’s marital status? Surprisingly, we don’t want to know if they’re divorced, just if they’re currently married (I include people are are separated and those who live apart from their spouses for other reasons).
3. In the past 12 months, did this person get divorced? The number of people who got divorced in the last year is the numerator in the divorce rate. For an analysis of divorce odds, I’m going to mix all the currently-married and just-divorced people together, and then treat the divorces as an event, asking: which people just got divorced?
4. In what year did this person last get married? This is crucial for estimating divorce rates according to marriage duration. When you subtract this from the current year, that’s how long they are (or were) married. When you subtract the marriage duration from age, you get the age at marriage. (For example, a person who is 40 years old in 2020, who last got married in 2010, has a marriage duration of 10 years, and an age at marriage of 30.)
5. How many times has this person been married? I limit this to people who have only been married once.
I also restrict this analysis to women, which is just a sexist convention for simplifying things (since men and women do things at different ages). Note, however, that these women may be married to or divorced from spouses of any gender.
I estimate a logistic regression model, with divorce as the dependent variable, and age at marriage, marital duration, education, race, Hispanic origin, and nativity as predictors. Note I can’t do anything with children, because it’s just a household survey, and doesn’t tell me anything about children who don’t currently live with the woman. I have no information about class background. Also, I don’t have any demographic information about the former spouses. (The results table, and link to the other materials, is at the end of the post.)
Here are the predicted probabilities of divorce by age at marriage, for women who married for the first time between the ages of 15 and 59:
The odds of divorce decrease the older the woman was when she first got married. The decline is steepest before age 30, then flat till age 45, when it decreases some more.
If we spin out the model a little more, we can get divorce odds for the whole matrix of age-by-duration conditions. So, the heat map below shows women who got married from their teens to their fifties, and their annual odds of marriage after 1 to 39 years of marriage (from the same model, but with an interaction between age at marriage and marital duration). The figure shows that the highest odds of divorce are in years 4-8 of marriages that started when the woman was younger than 25 — and especially under 20. The lowest rates are at the oldest ages and the longest-lasting marriages. Close inspection also reveals that the first year or two of late marriages are also a relatively risky moment for divorce.
So, if you want to avoid divorce, marry the right person and have a great marriage (I told you I’m not good at giving advice). If you want to know about divorce rates in the aggregate, know this: Divorce odds are lowest for women who marry later, and for people who’ve been married a long time.
I put the IPUMS extraction codebook, the Stata code that makes the file and runs the regressions, and the Excel file that made the heatmap, in an Open Science Framework project, here: osf.io/xpsba/. Help yourself (if you want different genders, race/ethnicities, etc., especially).
Here is the logistic regression table: