Revisiting the Relationships among Gender, Marital Status, and Mental Health

The study uses a national probability sample and panel data with a baseline measure of distress — genuine improvements over the cross-sectional work it critiques. Depression is measured with the same well-validated scale at both waves.

But the alcohol measure changes between waves in a way that undermines every longitudinal finding about drinking. At the first interview, respondents answered a single yes-or-no question: "Do you have a drinking problem?" At the second, they reported how many days per month they had five or more drinks — a continuous scale from 0 to 30. Every model that claims to measure change in alcohol problems is comparing answers to two fundamentally different questions. The baseline control absorbs almost none of the variance in the outcome, so these "change" models are poorly controlled.

26% of first-wave respondents dropped out by the second wave, and the dropouts were disproportionately unmarried, male, and depressed — exactly the people whose data matter most. Income and parental status are measured after marital transitions, meaning they may be consequences of marriage rather than independent controls. The paper's central theoretical contribution — that gendered symptom patterns reflect childhood emotional socialization — is never tested with any data on childhood or socialization. Multiple alternative explanations (biological sex differences, structural inequality, gendered reporting norms) would produce the same observed pattern, and the study cannot distinguish among them.

The overall sample is large (8,161), but the paper's claims depend on far smaller subgroups — 70 widowed men, 231 separated or divorced men. No power analysis is reported, and several core conclusions rest on the study's failure to detect gender differences in these small cells. Failing to detect a difference is not the same as showing one doesn't exist, but the paper treats it that way. This affects at least five claims, including the headline finding that marriage benefits men and women equally.

The mismatched alcohol measures create a structural problem for every longitudinal model. The yes-or-no baseline cannot properly control for prior drinking when predicting a 0-to-30 continuous outcome, making the change estimates for alcohol unreliable.

Twelve separate gender interaction terms are tested without any correction for multiple comparisons. The single one that reaches the conventional threshold — remarriage reducing men's drinking more than women's — is within the range expected by chance alone. Yet this finding anchors one of the paper's core claims about gendered symptom responses.

The descriptive findings (women report more depression, men report more drinking, married people report less of both) are well-supported by the sample sizes and standard methods.

The paper claims to measure "the emotional benefits of marriage" but operationalizes this as just two outcomes: depression and heavy drinking. The results are mixed. For depression, there is no detectable gender difference in the married-versus-unmarried gap. For alcohol, there is a difference: unmarried men drink substantially more relative to married men than unmarried women do relative to married women. Whether this supports "equal benefits" depends on how you weight the two outcomes, and the paper offers no reason for privileging depression over alcohol.

The claim that men and women respond to marital stress with "different types" of problems is only partially supported. Women do show more depression after divorce, but men do not show a correspondingly larger increase in drinking after marital loss. The one finding that fits the pattern — remarriage reducing men's drinking — is a single result out of twelve tests.

The paper's central theoretical construct, emotional socialization, is entirely unmeasured. The supporting evidence (that gendered symptom patterns appear in adolescence) comes from other studies and is ambiguous because puberty — which intensifies biological sex differentiation — happens on the same timeline.

The paper frames mixed results as a cleaner story than the data support. The depression finding (no detectable gender difference) becomes "equal benefits," while the alcohol finding (men benefit more) becomes evidence for a different theory. There is no stated reason for treating one outcome as the measure of "benefit" and the other as evidence of something else. A reader encountering only the abstract gets a materially incomplete picture.

The study treats its inability to detect a difference as proof that no difference exists. Several of the relevant subgroups are too small to reliably detect moderate-sized effects. The paper never discusses whether it had the precision to support its equality claims.

The emotional socialization interpretation is unfalsifiable with these data. Any gendered pattern in symptoms can be attributed to socialization. Results favoring equality are taken as confirmation, while results favoring men are reclassified as a different phenomenon. The measurement problem in the alcohol data — which could explain why some alcohol findings failed to emerge — is never disclosed.

The paper discloses its data source (a publicly available national survey) and funding (government grants with grant numbers). It acknowledges differential attrition and warns that some findings should be interpreted cautiously. It is honest about the ambiguity of the selection findings for marital loss.

But it does not flag the alcohol measurement mismatch as a limitation, despite drawing conclusions from models built on mismatched measures. It does not address the multiple-comparisons problem. Both omissions bear directly on the paper's core conclusions.

The literature review is genuinely balanced, presenting studies that found men benefit more from marriage, studies that found women benefit more, and studies that found no difference. The treatment of Gove's original theory is fair and acknowledges it was intended as historically specific.

The theoretical literature on emotional socialization (Gordon 1981, Hochschild 1979, and others) describes cultural emotion norms in general terms but does not demonstrate that socialization causes gender differences in psychopathology — these citations establish less than the causal claim they are asked to support. Directly relevant complicating evidence — that the gender-marital status-mental health pattern differs between Black and white Americans (Williams, Takeuchi, and Adair 1992), and that biological differences could produce identical patterns — appears only in footnotes. One critical citation (Waite and Gallagher 2000) is a popular-audience book rather than a peer-reviewed study.

This paper has no financial conflicts of interest. It is funded by government agencies, uses publicly available data, and operates in a field without significant industry distortion. The interpretive choices consistently favor the author's own prior theoretical framework (emotional socialization, developed in Simon 1998 and Simon and Marcussen 1999), but this is theoretical investment, not financial, and typical of normal academic work.