A Stanford study found that women accounted for 69% of heterosexual marital breakups in its 2009 to 2015 U.S. panel. The result came from 92 marital breakups and a question about who wanted the breakup more. It was not a count of court filings, and it does not show who caused a marriage to end.
See what the U.S. divorce rate of 2.4 actually measures→
On this page
What does the 69% divorce statistic actually count?
The study asked one person from each former couple: between you and your partner, who wanted the divorce, separation, or breakup more? The choices were the respondent, the partner, or both equally. Court filing was outside this measure. Wanting a divorce, saying so first, hiring a lawyer, and filing a petition are four different actions.
For the published estimate, Rosenfeld scored a breakup as 1 when the woman wanted it more, 0 when the man wanted it more, and 0.5 when both wanted it equally. Survey weights then produced 69%. In the 92 unweighted marital responses, 56 said the wife wanted the breakup more, 18 said the husband wanted it more, and 18 said both wanted it equally. The raw wife-only share was therefore 56 of 92, or about 61%, while the published scoring gives half weight to mutual decisions.
The accurate wording is that women accounted for 69% of the scored marital breakups. The study measured who wanted the breakup more. The author notes that the person who starts the formal process can differ from the person who most wanted the relationship to end.
How precise is 69%?
The estimate rests on 92 heterosexual marital breakups reported between the follow-up waves. The paper gives a 95% confidence interval of 61% to 78%. In plain terms, 69% is the study's best estimate, and the interval shows the uncertainty around it.
The analysis sample began in 2009 with 2,538 adults who had a partner of a different gender. Follow-ups took place in 2010, 2011, 2013, and early 2015. Across the analysis, 371 heterosexual breakups had a usable answer to the who-wanted-it-more question. Only the 92 marital breakups produced the divorce figure.
Follow-up participation also fell over time, from 85% in the second wave to 46% in the fifth. Rosenfeld repeated one key Table 2 model with weights adjusted for people leaving the panel and obtained the same result for that model. The 69% figure remains a panel estimate from relationships observed more than a decade ago. Its time basis is the 2009 to 2015 panel, and the study provides no annual series.
How did married and unmarried breakups compare?
The strongest comparison inside the study is marriage against nonmarital relationships. Women accounted for 69% of the scored marital breakups and 56% of breakups among unmarried couples who had lived together. That 56% estimate had a confidence interval of 47% to 65%, which includes an even 50% split.
Among unmarried couples who had never lived together, the estimate was 53.4%, with a confidence interval of 47.9% to 58.9%. This interval also includes 50%. In the weighted results, mutual decisions made up 32% of former cohabiter breakups and 35% of breakups among couples who had never cohabited, compared with 19% among married couples.
The panel supports a marriage-specific pattern. The two unmarried groups were statistically consistent with an even split between women and men.
Who wanted the breakup more in the Stanford panel
The table may be wider than the screen. Focus it and use the arrow keys to reach every column.
| Relationship at breakup | Breakups | Both wanted it equally | Women's weighted share | 95% confidence interval |
|---|---|---|---|---|
| Married | 92 | 19% | 69% | 61% to 78% |
| Unmarried, had lived together | 76 | 32% | 56% | 47% to 65% |
| Unmarried, never lived together | 203 | 35% | 53.4% | 47.9% to 58.9% |
How the scoring works Woman wanted it more = 1; both equally = 0.5; man wanted it more = 0. Survey weights produce the published shares.
How to read it: Source: Rosenfeld, Table 1. The married result is statistically different from 50%; the two unmarried results are not.
Download the table (CSV) ↓Did the study find that college-educated women initiate 90% of divorces?
No. Rosenfeld added a direct note to the paper because the 90% claim kept spreading. He states that the figure is absent from the data and paper and is too high to be feasible in his view.
This study cannot support a page that combines “69% overall” with “90% for college-educated women.” The paper tested education differences as part of its statistical models and published no 90% result. Treat that number as unsupported unless a page supplies a different primary dataset and a clear calculation.
Why were wives more likely to want divorce?
The study identifies no single proven cause. It tested several possible explanations, including relationship quality, income differences, education differences, age, children, and evangelical Christian identity for both partners. None explained the marriage-specific difference in who wanted the breakup.
Rosenfeld argues that the pattern is consistent with heterosexual marriage operating as a gendered institution. The paper discusses unequal expectations for housework, care, names, earnings, and authority as possible context. This is a contextual interpretation. Causation remains unresolved because the survey observed too little inside each marriage to determine why a specific spouse wanted to leave.
The number also leaves blame unresolved. A spouse can file after the other person withdrew, after both agreed to separate, or because one person had to complete the paperwork. The study's question reduces one source of confusion and cannot settle fault.
How should you use the statistic?
Use 69% as a study estimate for who wanted heterosexual marital breakups more in one U.S. panel from 2009 to 2015. Keep the sample of 92 and the 61% to 78% interval beside it. Its scope stops short of all current divorces, all women, all relationships, court filings, and fault.
For another divorce number that is often misread, the U.S. divorce rate of 2.4 per 1,000 people measures yearly population events. A lifetime share needs different data. For timing, average length of marriage separates current marriages from marriages that ended. For couple-type comparisons, lesbian divorce rate explains the limits of overseas registers for a current U.S. rate.
To evaluate a claim, match it to its measure. Ask who got counted, what action counted as initiation, which years the data cover, and whether mutual decisions were kept. Those four checks turn a viral percentage into a result you can evaluate.
