BERKELEY, Calif. — A new working paper from researchers at the University of California, Berkeley's Goldman School of Public Policy offers the most comprehensive look to date at how political differences shape the durability of romantic relationships in America. Analyzing voter registration records for approximately 60 million couples between 2014 and 2024, the study finds that couples with different party affiliations are over 14 percent less likely to remain together from one election cycle to the next. These mismatched couples are also more likely to split up in response to politically salient events, such as the criminalization of abortion in 13 states following the Supreme Court's 2022 Dobbs decision.

"A House Divided: Partisan Mismatch, Abortion Bans, and Relationship Duration," by Democracy Policy Lab researchers Rachel Funk Fordham and Ernesto Tiburcio, represents the first study to use longitudinal voter file data to track couple-level separation rates over time and to connect a specific policy shock to differential breakup rates among politically mismatched partners.

Drawing on voter file snapshots provided by the data firm L2, the researchers identified couples of registered voters living at the same address and followed them across election cycles from 2014 through 2024. Couples were classified as "politically mismatched" if their members were registered to different parties, a group that has consistently represented about 26 percent of all identifiable couples throughout the study period.

Even after controlling for a range of factors including age, marital status, income, years together, and ethnic differences, mismatched couples remained roughly 0.61 percentage points less likely to stay together over a two-year period. While that may sound small in absolute terms, it amounts to about 14 percent of the mean biennial separation rate observed in the data.

The researchers also found that the gap grows during politically charged moments. In counties where voter turnout was above the local median, already-mismatched couples were significantly more likely to separate than their matched counterparts, while the separation rates of politically aligned couples showed no such response to heightened political activity.

"What strikes us most is the consistency of this pattern across different parts of the country and across different types of couples," said Funk Fordham. "We've long known that people tend to seek out partners who share their values, but this data lets us observe how political differences within established couples shape relationship outcomes in an increasingly politically divisive era."

The Dobbs Effect on Relationships

To move beyond a purely descriptive finding, the researchers took advantage of a natural experiment: the rapid criminalization of abortion following the Dobbs v. Jackson Women's Health Organization ruling in June 2022. By the time of the 2022 midterm elections, abortion had been effectively banned or severely restricted in 13 states that record party registration. The researchers compared the trajectory of newly formed mismatched couples in those states against newly formed mismatched couples in states that maintained or expanded abortion protections.

Relative to 2020, the election cycle immediately before the bans took effect, newly formed mismatched couples in abortion-ban states were 1.47 percentage points less likely to remain together by 2022. Given that the average separation rate for mismatched couples that year was roughly 13 percent, the effect represents about an 11 percent increase over the baseline.

Politically matched couples showed no comparable change in separation rates in the same states and the same period, a finding the researchers say helps rule out the possibility that some broader social or economic shock was driving the results. Similarly, mismatched couples in states that only imposed abortion bans after 2022 showed no elevated separation rates in 2022, but did show a similar effect in 2024, after their own bans went into effect.

"The pattern follows the timing of the bans in a way that is hard to explain by anything other than the policy itself," said Tiburcio. "We tested for alternative explanations, including economic conditions, political trends, demographic shifts. The finding holds across all of them. The bans appear to have forced political differences to the surface in relationships where those differences had perhaps been manageable before."

The effect was concentrated among couples composed of one Democrat and one Republican, which aligns with the researchers' expectation that abortion preferences largely track party lines. Couples in which one partner was registered as Independent or unaffiliated did not show a statistically significant response to the bans.

Scale and Methodology

The study's dataset covers roughly 60 million unique couples across 35 states that record party registration. Because voter files do not include a couple-level identifier, the researchers developed an algorithm to reconstruct couples within households based on gender, age, marital status, and last names, then tracked those couples forward in time. The resulting estimates of marriage and separation rates correlate strongly with figures from the American Community Survey, lending confidence to the approach.

The researchers also acknowledge limitations. The data captures only registered voters, and mismatched couples were found to be modestly more likely to drop out of voter rolls than matched ones, which could introduce some underestimation of the true separation gap. The study also covers only heterosexual couples, reflecting constraints in how L2 models household relationships.

Broader Context

The findings contribute to a growing literature on how partisan polarization extends beyond the ballot box and into everyday social life, touching where people choose to live, work, and spend their time. Prior surveys had suggested that political disagreement strains relationships, but evidence tying that strain to actual behavior, tracked at population scale over time, has been limited.

The paper is currently circulating as a working paper and has not yet undergone peer review. The authors have posted it as a preliminary version dated August 2026.

For more information contact:
Rachel Funk Fordham
Ernesto Tiburcio