Where does the claim that marriage harms women come from, and does the evidence on happiness and mental health by marital status and sex support it?

Answer in brief

The belief that marriage harms women rests on a comparison error: the founding studies showed married women were identified as mentally ill more often than married men, and this was read as marriage harming women. But those same studies showed married women were better off than unmarried women — the benefit was simply smaller for women than for men. Every subsequent study confirms the direction: married women report less depression and higher life satisfaction than unmarried women. The open questions are about how large the benefit is and whether marriage causes it, not about whether married women are worse off.

What the evidence shows

The origin: a comparison that answered the wrong question. The belief entered academic literature through Gove 1972, who reviewed seventeen postwar studies and found married women were identified as mentally ill more often than married men. But Gove's tables also showed that unmarried people of both sexes had higher rates of mental illness than married people. The single-to-married ratio was about 3.1 for men and 1.7 for women — meaning marriage was associated with a large benefit for both sexes, just a smaller one for women. The data never showed marriage harming women; they showed it helping women less than it helped men.

The belief entered public discourse through Jessie Bernard's 1972 book The Future of Marriage, which called marriage an "emotional health hazard" for women. As Williams 2003 documents, Bernard's empirical evidence "was certainly limited," and the claim "has rarely been questioned" because it fit what sociologists expected about gender. Both Bernard and Gove compared married women to married men and treated that as evidence about whether marriage itself was the problem.

The companion paper's female excess was produced by its definition. Gove 1973 argued that women had higher rates of mental illness overall. But this held only after excluding personality disorders — the category where men predominated. The one large community survey that included personality disorders found no sex difference and was excluded on definitional grounds.

Later studies find no harm to women. No paper in this set finds married women worse off than unmarried women on any measure of happiness or mental health:

• Stutzer 2006, following roughly 15,000 Germans over 17 years, found life satisfaction rose about a third of a point (on a 0–10 scale) after entering a partnership, with no evidence the gain was smaller for women.
• Grover 2017, using a British panel, found the marriage–satisfaction association was slightly larger for women than for men (0.16 points on a 1–7 scale). The causal claims do not hold up, but the descriptive finding is consistent with the others.
• Simon 2002, using a US national panel, found married people of both sexes reported less depression. The paper claimed equal benefits, but this rested on failing to detect a gender difference in subgroups as small as 70 people — not proof of equality. For alcohol, unmarried men had substantially more problems than unmarried women relative to their married counterparts, meaning men benefited more on that measure. Every longitudinal drinking finding is also compromised by a measurement mismatch between survey waves.
• Williams 2003, also using a US panel, found no detectable gender difference in the marriage–depression link but lacked the statistical power to detect one.

Marital quality matters, but the sex difference is uncertain. Robles 2014, a meta-analysis of 126 studies, found a small but consistent correlation between marital quality and physical health (0.07 to 0.21). There was a hint the link was slightly larger for women, but this rested on only six studies and showed signs of publication bias. Kiecolt-Glaser 2001 documented that hostile behavior during lab arguments co-occurred with larger stress-hormone and immune changes in women than in men. That physiological pattern is real, but no study connected it to any later disease outcome.

What is claimed but not shown

That marriage causes better mental health. The association between marriage and lower depression is consistent, but every study is observational. Happier, healthier people are more likely to marry and stay married; unhappier ones are more likely to divorce. Stutzer 2006 showed that happier singles were more likely to marry later, and that future divorcers were already less happy before they married. Within-person analyses offer the best evidence of a real effect but cannot fully separate causation from selection.

That marriage benefits men and women equally. Simon 2002 and Williams 2003 both claim this, but both rest on non-detection in underpowered subgroups. Gove's original data and subsequent work suggest the benefit is likely larger for men, at least for some outcomes. Whether the difference is small or substantial remains unsettled.

That childhood emotional socialization explains gendered symptom patterns. Simon 2002 proposed this — women express distress as depression, men as drinking — but collected no data on childhood experiences. Biological sex differences and gendered reporting norms would produce the same pattern.

That bad marriages cause disease through physiological pathways. Kiecolt-Glaser 2001 showed real physiological responses during lab arguments, but the leap from a half-hour argument to years of disease has never been demonstrated.

What would settle it

• A large pre-registered panel study with validated mental health measures at both time points, enough power to detect moderate gender differences, and controls for baseline health and personality.
• Within-person analyses following the same people through marriage, divorce, and widowhood, reporting results separately by sex with adequate sample sizes.
• Studies measuring marital quality with external observation rather than self-report, to separate "bad marriage" from "depressed person rates everything negatively."

Limitations of this evidence base

The founding studies (Gove 1972, 1973) drew on treatment records and surveys from the 1950s and 1960s, when most married women were housewives and mental-health treatment was heavily gendered. The later studies (Simon 2002, Williams 2003) used data from the late 1980s and early 1990s, now three decades old. Marriage patterns, women's employment, and diagnostic practices have all changed substantially.

The evidence clusters around the US, UK, and Germany. Grover 2017 found that marriage was associated with lower life satisfaction in Latin America and Sub-Saharan Africa, suggesting Western findings may not generalize.

Every study relies on self-reported well-being or treatment records. None uses clinical diagnosis by blinded assessors. Robles 2014 found signs of publication bias in the mortality and self-rated health pools.

Missing from this set: any randomized or natural-experiment evidence, any study from the past decade, and any study that directly measured the mechanisms the various theories propose.