Marriage & well-being · General Social Survey, 1972–2024

Does the Ring Still Matter?

For fifty years, married Americans have reported being happier than everyone else. The gap has barely budged — and most of it survives even after you account for who tends to marry.

An analysis of 70,869 General Social Survey respondents · weighted estimates · association, not causation

Marriage in America became, over half a century, a later and more selective thing. People marry older, marry richer, and increasingly only after they feel settled. If marriage were turning into a luxury good — something the already-advantaged opt into — you might expect the happiness gap between the married and everyone else to balloon, as the people walking down the aisle grew ever more sorted toward contentment.

That is not quite what the data show. In the General Social Survey, the share of married people who call themselves “very happy” led the unmarried by about 21 points in the mid-1970s. Fifty years later, in 2021–24, it led by about 17 points. The premium is remarkably durable — neither the runaway widening the luxury-good story predicts, nor the collapse the “marriage doesn’t matter anymore” story predicts. It just persists.

The harder question is what the gap is made of. Married people are, on average, older, healthier, richer, and more educated than the unmarried — and all of those things track happiness on their own. So how much of the marriage premium is the marriage, and how much is simply who marries whom? The chart below puts both numbers on the same axis.

Net of age, income, health, children and education, the married still lead by 0.26 of a point on the 1–3 scale — about seven-eighths of the raw 0.30-point gap.

How much is selection?

Pool the whole half-century into one weighted regression and the raw married-minus-unmarried gap on the 1–3 happiness scale is about 0.30 points. Add the obvious confounders — age and its square, constant-dollar family income, self-rated health, number of children, and highest degree — and the gap shrinks to about 0.26. That is the shadow line in the chart: only about 13% of the raw gap is accounted for by those observable differences. The rest — the overwhelming majority — is something the controls do not see.

This is not the clean answer either camp wants. Thirteen percent is real attenuation: married people genuinely are richer and healthier, and that buys some of their reported happiness. But it is a long way from explaining the premium away. And note the shadow is not constant. In the earliest bins the raw and adjusted lines nearly touch — the controls barely move the 1970s gap. In the most recent bins they pull apart: the 2013–18 gap of 0.32 on the analytic sample falls to 0.26 once adjusted, and the 2021–22 gap of 0.33 falls to 0.25. As marriage grew more selective, a larger slice of the premium became explainable by sorting — exactly the luxury-good mechanism, visible not in the headline gap but in its composition.

Even so, the honest verdict is restraint. The controls are crude; income is a bracket midpoint, health is self-rated, and none of them capture temperament, the very thing that makes some people both happier and more marriageable. Stutzer & Frey, following German panel data, argued that happier people are more likely to marry in the first place — reverse causation our cross-section cannot rule out. The adjusted line narrows the gap; it cannot purify it. Read the band as “at least this much is who-marries-whom,” never as a clean causal estimate.

Not all unmarried are alike

Folding everyone who isn’t married into one “unmarried” lump hides a lot. The never-married, the divorced, the separated and the widowed are very different groups, and they sit at different heights. The married line runs along the top of the next chart in every era; the separated tend to sit lowest. The whole panel drops in 2021–24 — partly a real post-pandemic dip, partly the artifact of the GSS’s 2021 switch to web interviewing, which shifted measured happiness downward across the board.

Does cohabitation now rival marriage?

If the premium is partly a story about commitment, the natural follow-up is whether living together without the certificate delivers the same thing. Since 2012 the GSS has asked cohabitors to rate their relationship (hapcohab) on the same scale it has long asked the married to rate their marriages (hapmar). Across the seven years both questions exist, the answer is striking: cohabitors rate their relationships almost exactly as highly as married people rate their marriages.

And the “parenthood penalty”?

One more contested cut. A long line of research — Glass, Simon & Andersson on the cross-national happiness gap between parents and non-parents — warns that raising children can cost well-being, especially where support for parents is thin. In the raw GSS, though, parents come out slightly ahead: about 33% of people with children call themselves very happy, against 26% of the childless, a gap of roughly 0.08 on the 1–3 scale.

Treat that with the same caution as the marriage premium. Parents are disproportionately married, older and settled — the very traits that carry happiness — so this raw edge is selection-soaked, not evidence that children make you happier. The honest reading is a small, persistent association that runs the opposite direction from the “penalty” headline, and that the controls in our marriage model would shrink. The gap also widens in 2021–24, but so does everything in that mode-shifted bin.

What this does and doesn’t show

It shows a durable association and nothing more. The married report more happiness than the unmarried, by a margin that has held for fifty years and that observable advantages explain only a small slice of. That is consistent with the case Waite & Gallagher made in The Case for Marriage — but it is equally consistent with the selection story they were arguing against. Happier, healthier, steadier people are more likely to marry and to stay married; our regression narrows that channel but cannot close it, and it cannot touch reverse causation or the parts of temperament no survey records.

The numbers are also modest in absolute terms, and we report them plainly: a few percentage points here, a few hundredths of a scale point there, on a three-point item. The 2021 switch to web interviewing lowered measured happiness across every group at once, so the recent dip is partly a ruler change, not only a mood change — which is why we annotate it rather than narrate it. And the cohabitation near-parity is real but survivorship-bound: we see the unions that lasted, not the ones that didn’t. The ring, on this evidence, still travels with more reported happiness. Whether it causes any of it is a question fifty years of cross-sections cannot settle.