The marriage squeeze: How the college gender gap is reshaping who marries whom

Today, 1.6 million more women than men are enrolled in four-year colleges and universities in the US. An emerging narrative holds that, as women increasingly outnumber men among college graduates, college-educated women will struggle to find suitable partners. In some high-income countries with more traditional gender norms, marriage rates for highly educated women have indeed fallen as women’s education has outpaced men’s (Bertrand et al. 2016). Yet in the US, marriage rates among college-educated women have remained relatively stable. Instead, the decline in marriage has been concentrated among women and men without college degrees (Kearney 2022).

The results in our new paper suggest that college women have sustained relatively stable marriage rates in part by increasingly marrying non-college men (Chambers et al 2026). This shift, together with the weakening labour market position of non-college men, has coincided with especially large declines in marriage among non-college women. Over the past 50 years, the share of all women who are college-educated and married to a man without a four-year degree has quadrupled. The growth in these marriages has been concentrated among higher-earning non-college men, while real earnings have fallen for other non-college men. As a result, the share of non-college men who earn above the national median for their birth cohort and are not married to a college woman has fallen by more than half. These patterns point to an important question for policy: whether improving outcomes for non-college men could also improve marriage prospects for non-college Americans.

How college women adjusted

A simple comparison of whom college women marry over time can be misleading because group sizes have changed. Even a stable share of college women marrying college men could signal stronger sorting by education as women compete for a smaller group of similarly educated partners. We therefore adapt a standard marriage-market model (Choo and Siow 2006) that separates changes in group sizes from changes in ‘match surplus’ – the value of each education pairing relative to remaining single. We predict later marriage rates while holding match surplus at its 1930 level and allowing group sizes to change. The gap between this prediction and observed marriage rates shows how much of the change cannot be explained mechanically by shifting group sizes.

Holding the match surplus for each education pairing at its 1930 level, the model predicts that marriage rates for college women would have fallen by 16.7 percentage points. This decline reflects both the relative scarcity of college men and the smaller overall number of non-college men. Instead, marriage rates for college women fell by only 6.7 percentage points. Much of this 10-point gap reflects college women marrying non-college men more often than changing group sizes alone would predict.

The shift is also clear in the raw data. Over the past 50 years, the share of all women who are college-educated and married to a non-college man has quadrupled, from 2.3% to 9.6% – well above the model’s prediction holding match surplus at its 1930 level. With the gender gap in college enrolment continuing to widen (Reeves 2024a), college women’s ability to sustain marriage rates in this way may not continue indefinitely.

A shrinking pool for non-college women

While marriage rates for college women have remained higher than the model predicts, marriage rates for women without a four-year degree have fallen by 26.3 percentage points, from 78.7% for the 1930 cohort to 52.4% for the 1980 cohort. This decline mainly reflects fewer marriages to non-college men, historically their most common partners.

Two shifts have changed this part of the marriage market. First, real earnings for non-college men fell from $56,400 for the 1930 cohort to $50,000 for the 1980 cohort. This decline is part of a longer-run weakening of labour market opportunities for less-educated men (Rao et al. 2016). It occurred even as the broader economy grew and earnings for non-college women rose sharply.

Second, college women are not only marrying more non-college men; the non-college men they marry increasingly come from the upper end of the earnings distribution. Real earnings rose by roughly $7,000 among non-college men married to college women, while they fell by roughly $10,000 among all other non-college men.

Figure 1 Real earnings for non-college men, separately for those married to college women and all others

Notes: Earnings measured in thousands of 2024 dollars. |
Source: Chambers et al. (2026).

To gauge how these shifts may relate to non-college women’s marriage prospects, we construct a rough measure of ‘marriageable’ non-college men: those earning above the national median for their birth cohort and not married to a college woman. The share of non-college men who meet this criterion fell from 72.9% to 35.3% over the past 50 years—a decline of more than half.

Figure 2 Share of non-college men earning above the national median for their birth cohort and not married to a college-educated woman

Source: Chambers et al. (2026).

Together, the economic decline of non-college men and the growing number of marriages between higher-earning non-college men and college women have reduced this measure of the pool of economically stable partners for non-college women, even as non-college women’s own earnings have risen.

Similar patterns across places

The historical pattern has a geographic counterpart in the modern US. Figure 3 plots women’s marriage rates against the employment-to-population ratio for non-college men. The gap in marriage rates between college and non-college women is more than 50% smaller in areas with the highest employment rates for non-college men than in areas with the lowest rates.

Figure 3 Binned scatter plot using 2017–2021 American Community Survey data at the Public Use Microdata Area (PUMA) level, for individuals aged 40–49

Note: Employment rates for non-college men are standardised across PUMAs. Both series are residualised against the area’s overall college attendance rate for women.
Source: Chambers et al. (2026).

Similar associations appear across racial groups and for measures of hardship beyond employment. Areas where non-college men have higher rates of incarceration, high school dropout, or mortality also have larger education gaps in women’s marriage rates. The consistency of these relationships suggests that the pattern is not specific to employment, but is part of a broader link between men’s economic and social outcomes and women’s marriage prospects.

Because these comparisons are across places, they do not by themselves show that improving men’s outcomes would cause marriage rates to rise. They do, however, motivate further study of whether policies that improve outcomes for less-educated men could also narrow class gaps in marriage and family formation (Reeves 2022). Improving those outcomes remains important in its own right.

Source: cepr.org