The subject of candidate electability is naturally in the news, as Democratic activists and potential primary voters sort through the various presidential candidates to decide who has the best chance of defeating Donald Trump next year. Jonathan Bernstein and Julia Azari helpfully examined some of the problems with the concept this week. I want to focus here on some of the built-in biases surrounding electability, especially when it comes to race and gender.
We have quite a bit of evidence that ideology is relevant for electability. Moderate candidates simply tend to do better in general elections than ideologically extreme ones do. Of course, much of the evidence we have on moderation and electability comes from House elections, where we have a) lots of data, and b) pretty reliable measures of candidate ideology based on donations and/or roll call votes. We obviously have far fewer cases at the presidential level. Some recent studies suggest that the role of ideology may be overstated there – partisans close ranks almost no matter who is the nominee. Nonetheless, the idea that moderates do better is a pretty well-substantiated finding.
What is less well-substantiated is the relationship between identity and electability. That is, some party leaders and voters are convinced that white male candidates will do better in elections than women or people of color. The idea is that Democrats can do something daring, or they can play it safe with a white man.
But the evidence, for the most part, doesn’t back that up. Women run as strongly as men in elections. It’s possible that African American candidates do worse among white voters, but their presence may also boost turnout among African American voters. (Hillary Clinton could have become president with just a bit more support among white voters or slightly higher black turnout in a few key states.)
Yet here’s where it gets tricky: ideology and identity aren’t unrelated, at least in voters’ minds. Many voters are convinced that African American and female candidates are more liberal and more out of step with their districts.
To examine this idea, I conducted a simple analysis of Democratic members of the 115th House of Representatives (2017-19). A recent survey by the Cooperative Congressional Election Study asked voters in the members’ districts to place their member ideologically on a scale from -5 (the most liberal position) to +5 (the most conservative position). (My thanks to Kevin Reuning at Data for Progress for the data.) Meanwhile, Voteview uses members’ roll call votes to derive estimated ideal points for each member. By comparing these two measures, we can see how close constituents are to estimating the ideology of their members as revealed in their congressional voting. The correlation between the two measures is a reasonably-strong .55; voters have a pretty good idea of how liberal or conservative their members’ voting behavior is.
But a regression analysis, using the respondents' estimations of their member of Congress' ideology as the dependent variable, reveals some interesting differences. Constituents generally perceive both African American and female members of Congress as disproportionately liberal relative to their roll call votes. This holds up even controlling for district partisanship and the percent of the district that is African American. (You can see the Stata output below, which I am pasting here because independent blog.)
The differences are not enormous. Translating the coefficients above, African Americans and women are seen as about eight and six percentage points more liberal, respectively, than other House members, controlling for their roll call behavior. But these results are statistically significant and quite possibly enough for party activists to care about when considering a nominee. In other words, this evidence suggests that voters see African Americans and women as somewhat out of step with their districts just by virtue of their demographic characteristics.
This finding is consistent with studies by Brian Schaffner and Seth Stephens-Davidowitz suggesting that Barack Obama lost several percentage points of the vote in 2008 and 2012 by virtue of his race. It’s also consistent with findings by Neil Visalvanich and Matthew Jacobsmeier that African American congressional candidates underperform because white voters see them as ideologically extreme.
But in many ways, this shows how the discussion of electability as a desirable quality in a candidate is, in a sense, a toxic one. Voters and party insiders aren’t crazy to consider electability when evaluating candidates. And perceptions of the candidates’ ideological positions are certainly fair game in that evaluation. But evaluating ideology is not an exact science; it’s subject to distortions and preconceptions.
If party insiders think a candidate is too liberal simply because of her demographic profile, they may dismiss that candidate and under-support her for the nomination. Even if they understand that these distortions exist, they may dismiss the candidate because they think primary voters will be subject to those same prejudices.
This dynamic is probably at work in the current presidential nomination process, and quite possibly affecting the candidates' chances. That doesn't mean that if people had perfect knowledge about the candidates' ideologies Democrats would end up with a different nominee -- a lot of different forces will affect the outcome. But electability is a major factor in this contest, and it's got some issues.