The marriage-mortality evidence

The headline statistic — married adults live longer — survives scrutiny. But the number is smaller than popular memory says, unevenly distributed across groups, and structurally entangled with a caveat: the benefit travels with relationship quality, not with the license. This page lays out the actual numbers, where they come from, and how far they can honestly be pushed.

The Meta-Analytic Core

The literature is old, large, and consistent in direction. Among older adults, Manzoli and colleagues pooled dozens of studies and found widowed people carrying roughly 11–16% higher all-cause mortality than married peers (Social Science & Medicine, 2007). For marital dissolution specifically, Shor and colleagues (Social Science & Medicine, 2012) meta-analyzed studies of divorce and separation and estimated a mortality rate ratio near 1.3 for the divorced versus the married — with the excess consistently larger for men than for women. A parallel meta-analysis of never-married, divorced, and widowed adults (Roelfs et al., American Journal of Epidemiology, 2011) found the singles' relative risk drifting upward across cohorts — a widening gap, not a shrinking one, over the decades the data cover. And the pattern reaches cardiovascular outcomes specifically: a systematic review of more than two million participants found unmarried status associated with elevated risk of coronary heart disease and cardiovascular death (Heart, 2018). Note what the same literature does not say: not all unmarried statuses are equal. Never-married young adults show little penalty in several cohorts, while divorced and widowed adults carry the largest ratios — a pattern that matters for the selection argument below.

The Numbers in Human Units

Rate ratios need translating. In a large US cohort, Kaplan and Kronick (Journal of Epidemiology & Community Health, 2006) found married men and women outliving their unmarried peers across essentially every adult age band, and the association persisted after adjusting for income and other measured confounders — narrowing it, but not removing it. Panel data tell the transition story more sharply: in the Panel Study of Income Dynamics, divorce raised subsequent mortality risk, and remarriage partially restored it (Lillard & Waite, American Journal of Sociology, 1995). Dupre and colleagues (American Journal of Epidemiology, 2009) grouped people by marital trajectory rather than snapshot status and found the lowest mortality among the continuously married, with risk concentrating in the years immediately around divorce or widowhood. The honest magnitude: at the cohort level the association is measured in years, not months — but the confidence intervals are wide, the effect size is modest in absolute terms (comparable to a small-to-moderate behavioral risk, not to smoking), and none of it is randomized.

The Selection Problem

Every number above shares one structural weakness: people are not assigned to marriage. Healthier, more agreeable, more economically stable people marry more and stay married longer, while the divorced include people whose risk profiles — drinking, smoking, temperament, chronic illness — predated the divorce. The pattern in the data is consistent with selection doing real work: large ratios for dissolution and widowhood (events that select on prior characteristics), small ratios for the never-married young (a group that self-selects into singlehood for many reasons). No experiment can randomize marriage, so the causal share of the effect is permanently unknown — the parent topic walks this debate with the protection-versus-selection studies (Rendall et al., 2011). The honest translation for a longevity site: treat the marriage-mortality association as real, modest, and only partly causal — and read its practical value not as an argument to marry, but as a catalog of mechanisms you can replicate without a license, which is what the next two sections do.

What the Numbers Hide: The Mechanisms

• 🩺 Health monitoring: the partner who notices the changing mole, books the checkup, enforces the follow-up. Men's larger marriage benefit is thought to run heavily through this channel — men historically outsource health surveillance to a spouse (the parent topic's health-behavior section owns this finding).
• 💰 Economic pooling: two incomes smooth shocks, share housing and insurance, and buffer job loss. When Kaplan and Kronick adjusted for income the association shrank but did not disappear — money is part of the story, not all of it.
• 🧘 Stress buffering: a built-in confidant dampens the stress response to life events — the physiology the biology of connection topic maps in detail.
• 🍽️ Behavior regulation: drinking, smoking, diet, and sleep all improve with a cohabiting partner who shares the routine and notices the drift — the mundane channel that carries much of the effect.
• 📅 Routine stability: regular meals, shared sleep schedules, and standing appointments are protective infrastructure, and they are easier to run with two people.

The Quality Caveat, Inside the Data

The most important sentence in this literature is not about marriage at all. The same datasets that produce the protective number show the benefit concentrating in high-quality marriages — and distressed, hostile marriages erasing it, leaving their occupants with health patterns worse than the single comparison group. The relationship-quality page owns that physiology: blood pressure, wound healing, and heart-rate variability all respond to the emotional climate of the marriage, not its existence. Symmetrically, unmarried adults with strong networks show no penalty — the solo-by-choice page covers that side of the ledger. Read together, the two findings reframe the whole topic: the license is a proxy, and the actual variable is whether your closest relationship is a source of safety or a source of stress.

The Marital-Status Ledger

Plotted against the married reference, the pooled estimates line up in a consistent order — dissolution worst, widowhood close behind, never-married mildest. The chart summarizes the direction; the table below lists what moves the numbers.

Questions, Answered Briefly

• ❓ Is the marriage effect causal? Partly, but the share is unknown. Selection is real, no experiment can randomize marriage, and the strongest reading is "association with mechanisms that are partly replicable."
• ⏳ How big is it in years? Years, not months, in the largest cohorts — but with wide confidence intervals, a smaller effect for women, and a modest absolute size once you adjust for the obvious confounders.
• 💍 Does remarriage restore the benefit? Partially, in panel data (Lillard & Waite, 1995) — the transition out of dissolution matters more than the new status itself.
• 🧾 Should I marry for my health? No. The license is not the mechanism — support, monitoring, and routine are, and all three are available unmarried, as the solo-by-choice page shows.

Related Topics