Summary
Stevenson and Wolfers survey 150 years of U.S. marriage and divorce trends using vital statistics, census data, and cross-national comparisons to characterize the current state of marriage and identify its driving forces. The paper's central theoretical contribution is that the declining returns to household specialization have shifted the basis of marriage from production complementarities (Becker 1981) to consumption complementarities and insurance — a framework better suited to explaining rising assortative mating, the education "divorce gap," and the persistence of marriage despite declining fertility rates and expanded female labor force participation.
Key Claims
- Divorce peaked in 1981 at 22.8 per 1,000 married couples; by 2005 the rate had fallen to 16.7 — the lowest since the early 1970s. The 1950s–60s were anomalously low-divorce; the 1970s overshot a long-run rising trend dating from 1860; post-1981 decline is a reversion toward the pre-WWII trend.
- The "half of all marriages end in divorce" claim applies specifically to the 1970s marriage cohort (48% dissolution within 25 years). 1980s and 1990s cohorts show lower cumulative dissolution rates — fewer than half likely to end in divorce.
- Marriage rates are at recorded-history lows, but this reflects later timing (median age at first marriage: men 23→27, women 22→26 between mid-1950s and 2004) and rising cohabitation as a precursor (59% of early-2000s first marriages preceded by cohabitation with the future spouse).
- The Black–White marriage gap is almost entirely an entry gap (lower marriage rates for Black Americans), not a divorce rate gap. Conditional on marrying, divorce rates are similar by race. The gap opened in the 1940s–1960s and is explained by labor market and incarceration effects on male "marriageability," not differential attitudes toward marriage.
- The education "divorce gap" is real and large: ~10 percentage points (pp) lower first-marriage dissolution for college graduates vs. non-college. College-educated women's marriage rates have converged upward: 29% never-married (1960 census, women in their 60s) → 8% (2000 census). The historical claim that educated women are less likely to marry is now reversed.
- Cohabitation was nearly absent before 1970; 50% of women 15–44 had ever cohabited by 2002, and 75% of remarriages were preceded by cohabitation. Pre-marital cohabitation was historically associated with higher subsequent divorce (selection, not causation).
- Americans are international outliers: highest marriage rates, highest divorce rates, highest remarriage rates among comparable nations. Only 10.1% of Americans view marriage as outdated (vs. 18–36% in Europe). Mean age at first birth is 27.3 (US) vs. 29–30.3 elsewhere.
- Driving forces — (1) Birth control pill: FDA approved 1960, generalized to unmarried women via Eisenstadt v. Baird (1972); reduced cost of waiting to marry, allowing more informative courtship → fewer bad matches → lower divorce, later marriage (Goldin and Katz 2002: state-cohort pill access predicts all three outcomes). (2) Shotgun marriage decline: Akerlof, Yellen, and Katz (1996) show the pill undermined women's ability to demand marriage on pregnancy → rising non-marital births despite better contraception. (3) Household technology: Labor-saving appliances reduced skill and time cost of home production → declining returns to within-household specialization → rising female labor force participation (FLFP) and lower marriage-based surplus. (4) Wage structure: Rising male wage inequality raises the option value of continuing to search for a high-wage partner → lower marriage rates (Gould and Paserman 2003: explains up to 1/3 of marriage-rate decline for women aged 21–30); narrowing gender wage gap reduces female comparative advantage in home production. (5) Unilateral divorce law: Wolfers (2006) — divorce spiked 2 years post-reform then reverted; explains little of the long-run trend. But Stevenson and Wolfers (2006) — female suicide and domestic violence fell in unilateral-divorce states, consistent with a bargaining-power shift toward women.
- Theory shift: Becker's (1981) production-complementarity model of marriage is increasingly inapplicable. By 2005, only 41% of married people lived with own children (vs. 75% in 1880); rising FLFP eliminates specialization; appliances commoditize home production; longevity extends marriage far past childrearing. The authors argue modern marriage is better understood through consumption complementarities — joint leisure, shared public goods, insurance against health and income shocks. Signature: rising positive assortative mating by education and income (consistent with consumption complementarity, contradicting Becker's specialization-based prediction of negative assortative mating).
- Welfare: Marriage–happiness correlation reflects selection, not causation. British Household Panel Survey: the happiest singles are most likely to marry, and those who married experienced similar happiness gains to those who didn't. The marginal marriage/divorce does not conform to the average.
Concepts Introduced or Extended
Entities Mentioned
Quotes
"The shift in the nature of marriage away from a focus on production complementarities toward a focus on consumption complementarities suggests that the returns to marriage are changing." (p. 47)
"The claim that 'half of all marriages end in divorce' is most descriptively accurate for people who married in the late 1970s." (p. 29)
My Take
A model survey paper — encyclopedic on facts, honest on mechanism. The production-to-consumption-complementarity reframing is compelling and explains several otherwise puzzling patterns (rising assortative mating, education gradient, persistence of marriage among the highly educated despite FLFP). The weakness is that "consumption complementarities" is a residual category rather than a testable structural model; the paper identifies what marriage is not grounded in more precisely than what it is grounded in. The international comparison section usefully demolishes the idea that the U.S. trend is simply global convergence — Americans remain structural outliers in both direction and degree.