Politically mismatched couples are 14% less likely to still be together by the next election cycle than couples who share a party, according to a working paper from UC Berkeley’s Goldman School of Public Policy. The researchers didn’t survey anyone. They tracked 60 million real couples through voter registration records across six election cycles, and found that the 2022 fall of Roe v. Wade turned a quiet risk factor into a measurable one.
The paper, titled “A House Divided: Partisan Mismatch, Abortion Bans, and Relationship Duration,” comes from Goldman School researchers Rachel Funk Fordham and Ernesto Tiburcio, working out of the school’s Democracy Policy Lab. Instead of relying on self-reported survey data, the usual method in relationship research, they matched L2 voter file snapshots from 35 states between 2014 and 2024, using shared addresses, last names, and ages to identify romantic partners and track their party registration over time.
About 26% of the couples they could identify were politically mismatched, meaning one partner was registered with a different party than the other. On its own, that mismatch translates to a modest but real gap: mismatched couples were 0.61 percentage points less likely to still be together after two years than politically matched couples, against a baseline separation rate of roughly 13% a year for cross-party pairs. Small per year, but it compounds across a relationship, which is how the researchers arrive at the 14% figure over a full decade of tracking.
The real story is what happened after June 2022. In the 13 states that moved quickly to restrict or ban abortion following the Dobbs decision, newly formed politically mismatched couples saw a 1.47 percentage point jump in separation rates, more than double the baseline mismatch effect and concentrated almost entirely among Democrat-Republican pairings. Politically matched couples in those same states showed no comparable shift. Neither did mismatched couples in states where abortion access stayed protected.
That state-by-state contrast is what makes the finding more than a coincidence. If Dobbs simply made everyone more stressed or more political, you’d expect separation rates to move for matched and mismatched couples alike, and in states where nothing legally changed too. Instead, the effect shows up specifically where the law changed and specifically in couples where partners already disagreed politically, evidence that the bans forced a private disagreement into a household-level decision that couples could no longer avoid.
Independent reporting from the Daily Californian, UC Berkeley’s student newspaper, confirmed the same headline figures after reviewing the study. The researchers themselves are careful not to overstate the finding as deterministic: being politically mismatched doesn’t guarantee a relationship fails, plenty of cross-party couples stay together long-term, and some of the gap may reflect other unmeasured differences between people who end up in mismatched pairings in the first place rather than the mismatch itself causing the split.
It also fits a longer trend demographers have been tracking for years: Americans increasingly marry and partner within their own party to begin with. The 26% mismatch rate Berkeley found among identifiable couples is already lower than you’d get from random pairing in a roughly 50-50 electorate, which points to a pattern researchers call political assortative mating, people sorting into relationships along party lines well before a wedding or even a first date, whether consciously or not. That makes the couples in this study who did end up mismatched something of a self-selected group already more tolerant of political difference than average, which is part of why the abortion-ban effect stands out: it broke through in exactly the couples who had already proven they could handle disagreement.
Still, the mechanism the researchers describe is specific and plausible: abortion law is one of the rare political issues that stops being abstract the moment it affects your own body, your own pregnancy, your own medical options. A couple that has spent years agreeing to disagree about tax policy or foreign affairs suddenly has to agree, in practice, not in theory, about what happens if she gets pregnant in a state that has criminalized the outcome one of them might choose. The Berkeley data suggests that’s exactly the kind of disagreement relationships don’t absorb quietly.
The paper is a working paper, posted in August 2026 and not yet peer-reviewed, so its conclusions should be read as strong preliminary evidence rather than a settled fact. But the scale of the dataset, 60 million couples, a full decade, real administrative records instead of memory-dependent surveys, makes it one of the more rigorous looks yet at a question couples increasingly can’t avoid asking themselves: how much political difference can a relationship actually carry, and does that answer change when the politics stop staying outside the house.
