"Half of marriages end in divorce" was never a statistic

In the 1970s, divorce in the United States rose steeply. No-fault divorce laws spread state by state, women's participation in the paid workforce climbed, and a large backlog of unhappy marriages that had previously been difficult to exit was released in a relatively short window.

Demographers extended the trend, as demographers do. If divorce continued climbing at that rate, roughly half of the marriages formed in that period would eventually end in divorce. That was a reasonable projection given what was known at the time.

The rate peaked around 1980 and has been falling since[1]. The projection was never revised in public memory, because a corrected statistic is far less quotable than an alarming one.

The clearest way to measure divorce is the number of divorces per 1,000 married women, a rate that accounts for how many people are married in the first place. The common alternative, dividing this year's divorces by this year's weddings, compares two completely different populations of people and tells you very little.

On the better measure, divorce in the United States has been declining for four decades. The National Center for Family and Marriage Research at Bowling Green State University tracks this and publishes the underlying figures[2]. The Centers for Disease Control publishes the national marriage and divorce counts the estimates are built from[3].

The decline is not evenly distributed. It is driven substantially by younger cohorts: people marrying in the 2000s and 2010s divorce at noticeably lower rates than people who married in the 1970s and 1980s did at the same point in their marriages. Pew Research has documented the parallel and less-discussed trend of divorce rising among adults over 50[4].

Partly because it is easy to say. "Fifty percent" is memorable in a way that "the refined divorce rate has declined from roughly 23 per 1,000 married women to around 14" is not.

Partly because it is useful. If you are anxious about your own marriage, a statistic saying that half of everyone fails is oddly comforting. If you are arguing against marriage as an institution, it is ammunition. If you are selling a book, a course, or a service, a crisis is a better opening than a slow improvement.

And partly because almost nobody who repeats it has any reason to check. It sounds like the kind of thing that was measured. It has the shape of a fact.

Here is the part that matters most, and it is the part the argument about the number usually obscures.

A population-level divorce rate tells you almost nothing about your own marriage. It is an average across every marriage in a country: first marriages and fourth ones, couples who married at nineteen and couples who married at forty, marriages entered under financial desperation and marriages entered with a decade of shared history behind them. Your risk is not the average of those.

The variables that actually predict outcomes for an individual couple vary enormously and are far more informative than any headline rate. Age at marriage, financial stability, education, whether either partner has divorced before, and how the two of you handle conflict all matter more. Some of them you cannot change. Some of them you can.

A statistic about strangers is not a forecast about you, in either direction. It does not condemn your marriage and it does not protect it.