If you’ve ever used an online dating website, you’ve answered a litany of questions about yourself and your ideal match, and maybe even about your political affiliation. Although this information is sometimes hidden from view, new research shows that time and time again men and women choose to pursue romantic connections with people from their own political party and with similar political beliefs.

This was one of the key findings that emerged after Neil Malhotra, Stanford Graduate School of Business associate professor, and Gregory Huber, political science professor at Yale University, analyzed thousands of interactions from an online dating website. Their findings, presented in a research paper titled “Political Sorting in Social Relationships,” show that political affiliation rivals education level as one of the most important factors in identifying potential mates.

“We underestimate how much politics affects our daily lives,” says Malhotra. “After an election is over, we don’t think about it, but in fact our political affiliations strongly affect other aspects of our lives, such as our romantic choices.” And that has important implications beyond the households that politically similar individuals may form, he says.

“At the highest levels within our political system, we increasingly see that people are unwilling to work and communicate with each other,” he observes. “Simply put, our society has become more and more polarized, and we wanted to explore if political preferences in romantic relationships could begin to explain part of the divide in America.”

As Malhotra explains, when people pair with individuals of similar political beliefs, their households can become echo chambers that transmit extreme views to the children. In fact, research shows that children are more moderate if their parents have differing political viewpoints. There is a genetic story at play, as well. Studies of twins demonstrate a genetic predisposition for certain political beliefs, which suggests that offspring of like-minded individuals may be predisposed to more extreme beliefs.

With this context in mind, Malhotra and Huber embarked on a laboratory experiment in which they presented participants with online dating profiles. Participants evaluated profiles more positively (e.g. had greater interest in dating the targeted individual) when the target had their same political ideology and level of interest in politics. Study participants even found online candidate profiles more physically attractive if they shared similar political beliefs.

To validate these results, the researchers then formed a partnership with an online dating website, which provided the team a unique window to observe people’s beliefs and preferences before they meet and interact in a marriage market. It also provided a wealth of data since, according to a Pew Research study, 74% of single Americans seeking partners have used an online dating site.

The team developed a set of seven new questions that users were asked when signing up for the online dating service. The questions measured three different political characteristics: political identity, including party affiliation; issue positions; and political participation. Most users opted to keep their answers to these questions private, meaning that other users could not proactively search for potential mates using these criteria.

Nevertheless, after assessing how men and women interacted via the site’s messaging function, Malhotra and Huber found that — in line with the results from the lab study — shared political characteristics increased the messaging rates in statistically significant ways above a baseline rate. Shared partisanship increased messaging rates by 9.5%, shared levels of political interest increased messaging rates by 10.7%, and shared ideas about how to balance the budget increased messaging rates by 10.8%.

These are similar to the messaging boosts found from shared educational background (10.6%) and height (9.8%); slightly lower than race (16.6%); and lower than religion (50%). However, since political characteristics were not disclosed — unlike these other publicly disclosed characteristics — it shows “how strong the political effect is, and how easy it is for people to pick up on cues about political beliefs,” says Malhotra.

Unlike other research in the field, which demonstrates that shared political preferences develop after a relationship is formed, Malhotra and Huber have, in fact, uncovered that shared political preferences actually precede the formation of romantic relationships. This kind of “sorting” reduces political disagreement within the household, which, in turn, risks the creation of homogenous political enclaves. The bottom line? According to Malhotra, “Partisan polarization could get much worse.”

When asked what might help combat some of these effects, he says that interaction between people of differing viewpoints, backgrounds, and identities could help mitigate political polarization. It’s been suggested, for example, that racial integration in the workplace has resulted in greater tolerance. If schools and colleges give more attention to political discourse, it might have the same positive effects, he observes. On a larger scale, says Malhotra, our systems of political districting create extreme enclaves of like-minded people — a problem that needs to be addressed if we want to have any hope of changing the national political conversation.