Half of all marriages end in divorce. It is one of the most repeated statistics in American life — in sermons, in wedding toasts, in arguments about the state of the family, and in the private calculations of people deciding whether to get married at all.
It is possible to reconstruct exactly where it came from, and the answer is not a study. It is a division problem, performed on two numbers from the National Center for Health Statistics, in a single year.
The arithmetic
In 1981, the United States recorded 2,422,000 marriages. In the same year it recorded 1,213,000 divorces. Divide the second by the first and you get 50.1%.
1981 is the only year in American history where that ratio lands on approximately one half. It was also the all-time peak of the American divorce rate.
The slogan was struck from the highest point on the curve, by a method that was invalid to begin with, and then carried unchanged for forty-five years.
Why the method is invalid
The problem is that the numerator and denominator describe different populations.
The divorces granted in 1981 were overwhelmingly granted to people who had married in the 1960s and early 1970s. The marriages contracted in 1981 were a different group of people entirely. The two sets barely overlap.
Worse, the denominator is a flow — one year’s weddings — while the population actually at risk of divorce is a stock: everyone already married. Using the NCHS refined rate for 1981 of 22.6 divorces per 1,000 married women, the true at-risk population was roughly 53.7 million married women. The ratio used a denominator about 22 times too small, drawn from the wrong people.
This is not a mildly imprecise estimate of the lifetime probability of divorce. It is not an estimate of it at all.
What actually happened to the rate
The crude divorce rate has fallen from 5.3 per 1,000 population at its 1979 and 1981 peaks to 2.4 in 2023 — a decline of roughly 55%. The refined rate, which uses the correct denominator of married women, has fallen from around 22.8 in 1979 to about 14.6 in 2022.
The Census Bureau’s direct measure is the cleanest answer to the question people are actually asking. As of 2016, 34% of ever-married women and 33% of ever-married men had ever been divorced. The figure peaks at around 43% for those aged 55 to 64 — the cohort that married during the peak-divorce era, and therefore an upper bound for anyone marrying now.
The data problem nobody mentions
Before celebrating, there is a complication that cuts against the good news.
NCHS suspended detailed divorce data collection in January 1996. Current national figures are compiled from provisional counts supplied by state health departments, and several states — including California, the largest — have not reported divorces for decades. The 2023 rate covers 45 states and the District of Columbia. Any “national” divorce rate from this source is not national.
Sheela Kennedy and Steven Ruggles took this seriously in a 2014 paper in Demography. Vital statistics show the refined rate falling from 22.6 in 1980 to 17.8 in 2008. Using American Community Survey marital-history data instead, they put 2008 at 22.1 — essentially unchanged from 1980. Age-standardised, they found divorce rose substantially between 1990 and 2008, with rates roughly doubling among people over 35, while falling among the youngest couples.
Philip Cohen, writing in Socius in 2019, found a 21% drop in the refined rate between 2008 and 2017 — but only 10% once adjusted for demographic composition.
These two findings are partly reconcilable, since Cohen’s window begins where Kennedy and Ruggles’s ends. They also reflect a real, unresolved dispute about how much of the apparent decline is compositional rather than behavioural. Neither should be presented as settled.
The finding that actually matters
The more consequential fact is that there is no longer a single American divorce trend. There are at least three, running in different directions.
Steven Martin’s 2006 analysis in Demographic Research built life tables of first marriages by the wife’s education, comparing the 1975–79 marriage cohort with the 1990–94 cohort.
For women with a four-year degree, the ten-year dissolution rate fell from 29.0% to 16.5%. For women without a high school diploma, it rose from 38.3% to 46.3%. The gap between them widened from 9.3 percentage points to 29.8.
Meanwhile, Susan Brown and I-Fen Lin documented what they called the gray divorce revolution: among people aged 50 and over, the divorce rate per 1,000 married persons rose from 4.9 in 1990 to 10.1 in 2010, a 106% increase, against a falling national rate. By 2010, roughly one in four divorces involved someone over 50.
What is replacing the old statistic
Cohen’s analysis identified what changed between 2008 and 2017 among newly married women: mean age at marriage rose from 32.1 to 33.4, the share with a bachelor’s degree or higher rose from 30% to 39%, and the share entering a first marriage rose from 69% to 74%.
That is not evidence that marriages became happier. It is evidence that marriage became more selective. Fewer bad marriages are being formed, which is a different phenomenon from fewer marriages going bad. The crude marriage rate fell from 10.6 per 1,000 in 1970 and 1981 to 6.1 in 2023.
Cohen’s own summary is the most economical description of where things stand: the United States is progressing toward a system in which marriage is rarer and more stable than it was.
So the slogan is wrong twice
It is wrong about the level — the rate is roughly half what it was when the number was coined, and only about a third of ever-married Americans have ever divorced.
And it is wrong in what it implies: that marriage carries one uniform risk for everybody. It does not, and the spread between groups is now larger than the difference between the old number and the new one.
The 50% figure is a fossil of the exact moment the old system broke. It is still being used to describe the one that replaced it.
Sources
- National Center for Health Statistics. Marriage Rates in the United States, 1900–2018; FastStats: Marriage and Divorce; Advance Report of Final Divorce Statistics, 1987.
- Martin SP. Trends in Marital Dissolution by Women’s Education in the United States. Demographic Research 2006;15(20):537-560.
- Brown SL, Lin I-F. The Gray Divorce Revolution. Journals of Gerontology Series B 2012;67(6):731-741.
- Kennedy S, Ruggles S. Breaking Up Is Hard to Count. Demography 2014;51(2):587-598.
- Cohen PN. The Coming Divorce Decline. Socius 2019;5.
- US Census Bureau. Number, Timing, and Duration of Marriages and Divorces: 2016, P70-167.
