Eitan Hersh | June 28, 2016
Evidence abounds that Democrats and Republicans really do not like each other. Researchers have found that they avoid dating one another, desire not to live near one another and disapprove of the idea that their offspring would marry someone outside their party (see here, here, here). Sure, most people are not very political, but among those who are, partisanship seems to be affecting nonpolitical realms of their lives.
That phenomenon motivated a colleague and me to gather data about mixed-partisan marriages. We were curious: How many Americans are married to someone of the other party? Who are these people? Are they old or young? Where do they live? Do they vote?
To answer these questions, I teamed up with Yair Ghitza, chief scientist at Catalist, a prominent political data firm that sells data to left-of-center campaigns and interest groups, and also to academics like me who use the data for scholarly research. Catalist maintains a continuously updated database containing records of personal, political and commercial data for nearly all American adults.
We focused on registered voters in the 30 states that track voters’ party affiliation. For simplicity, we mostly focused on male-female partners who live at the same address, share a last name, are within 15 years of age (sorry, Donald and Melania Trump), and are the oldest such pair in the household.
We also cut the data in other ways, such as incorporating same-sex couples as well as couples who do not share a last name. In our research paper, we try out 32 different ways to define marriage in the data. Without getting too deep into the details, there’s a trade-off in how we define marriage here. For instance, if we include same-sex pairs and pairs with different last names, we are both more likely to count nonmarried people as married (e.g., 20-something platonic, same-sex roommates — not our population of interest) and also more likely to count as married those in less “traditional” marriages, who are in the population we care about.
How we define marriage affects the overall partisan composition of married couples (i.e., when we include less traditional couples, the population appears more Democratic), but the definitions do not much affect the key findings below.
What are those key findings? Here are the five most important ones.
First, 30 percent of married households contain a mismatched partisan pair. A third of those are Democrats married to Republicans. The others are partisans married to independents. Perhaps unsurprisingly, there are twice as many Democratic-Republican pairs in which the male partner, rather than the female partner, is the Republican.
Second, 55 percent of married couples are Democratic-only or Republican-only, which raises a question: Is that a big number or a small number? In other words, is there more or less partisan intermarriage than we should expect? Here are two ways we try to answer that. We can compare interparty marriages to interracial marriages. Using voter registration data, we can do this in three states, Florida, Louisiana and North Carolina, where public voter files list everyone by their party affiliation and their racial identity. In those states, 11 percent of married couples are in Democratic-Republican households. In comparison, only 6 percent of married couples are in any kind of interracial household. At least in these states, there’s about twice as much interparty marriage as interracial marriage.
We also evaluate the degree of sorting through another exercise: Suppose that all of us choose partners from the pool of people who share our age and geographic location. So, a 30-year-old New Yorker seeks a spouse from the pool of other 30-year-old New Yorkers. In the graph above, the bottom of the gray band indicates the percentage of couples that would be Democratic only (left panel) or Republican only (right panel) if people were pairing off randomly with regard to partisanship. The top of the gray band reflects the percentage of couples that would be Democratic-only or Republican-only if partisans exclusively married people who share their party. And the blue and red lines show the actual percentage of same-party marriage for each age group in New York. The upshot — which is the same for every city we have explored — is that the red and blue lines fall close to the middle of the gray band. People sort into relationships with co-partisans, but not that much.
Third, there is a much higher rate of mixed-partisan couples among younger pairs than older pairs.
The main reason for the dramatic relationship with age is that younger voters are more inclined to register as independents than older voters are. This was true 50 years ago, and it is true today. As the chart shows, while the proportion of Democratic-independent and Republican-independent pairs shrinks from the youngest couples to the oldest couples, the proportion of Democratic-Republican pairs actually doubles — i.e., the purple band becomes bigger.
Fourth, we looked at the neighborhoods where couples live. In this graph, we situate each voter in his or her neighborhood, and we look at the percentage of the vote in that neighborhood that went to President Obama in 2012.
If you have an image in your head of a “battleground neighborhood” that is fiercely divided between Democrats and Republicans, you might imagine tension between neighbors. But the truth is that in these neighborhoods, half of the married couples living under the same roof are not one-party pairs. In fact, except in overwhelmingly Democratic neighborhoods (which tend to be African-American neighborhoods), close to half of households are not Democratic-only or Republican-only. This is likely to contribute to a more tempered political climate in battleground areas than we might first expect.
Finally, we looked at voter participation. Accounting for a voter’s state, age, gender, race and party, we see huge effects of household composition on voter turnout. Partisans married to like-partisans voted at much higher rates than partisans married to independents or to members of the opposite party.
|D-D COUPLES TURNOUT VS.||R-R COUPLES TURNOUT VS.|
Estimates show marginal turnout change at maximum part of logit curve. Model controls for state, race, gender and age.
SOURCE: HERSH AND GHITZA
In the 2012 and 2014 general elections, a Republican married to a Republican was about 10 percentage points more likely to vote than the same kind of Republican (e.g., same age, gender, race, state) married to a Democrat or independent. That effect is about twice as big as for a Democrat married to a Democrat.
The effect is even bigger in primaries, especially in closed primaries where independent voters are not eligible to vote. In closed primaries, the partisans who are married to independents have especially low turnout compared with the same kind of partisans who are married within their party. In closed primaries in 2012 and 2014, Democrats and Republicans were 17 to 18 percentage points less likely to vote if they were married to an independent, which is enormous considering that overall turnout in these elections is only 30 to 40 percent among registered partisans.
Why is there such a big effect on turnout? From this data alone, it is hard to say for sure. But it is likely a combination of two factors. First, voters who are not particularly interested in voting are probably more willing to be in mixed-partisan relationships. So their low engagement is not so much an effect of their mixed marriage as a contributing cause of that marriage. Second, living with an independent or opposite-partisan probably also directly affects one’s behavior. If your spouse is not going to vote in a primary because he or she is ineligible or does not care, you are probably more likely to skip voting too rather than walk to the polling place alone.
In addition to what this analysis can tell us about marriages and partisanship, there’s also an important lesson here for any political data junkie or journalist. Almost all data about politics that you encounter comes from polls and surveys of individuals or else from analysis of geographic units such as precincts, counties and states. Individual data and geographic data do not capture the essential networks in which we all live — households and friendships and communities. But other and newer kinds of data — such as voter files that connect individuals to their households or network data that capture online connections — revolutionize how we understand politics. By the end of this election cycle, expect to see many more discoveries about the social groupings that define our lives.
Eitan Hersh is an assistant professor of political science at Yale University. @eitanhersh