Lynn Vavreck
Lynn Vavreck

Lynn Vavreck is the Marvin Hoffenberg Professor of American Politics and Public Policy at UCLA, and is a co-author of the coming “Identity Crisis: The 2016 Presidential Campaign and the Battle for the Meaning of America.” This column appeared in the New York Times.

Donald J. Trump’s rise to the G.O.P. nomination and eventual presidential victory gave heft to the idea that extreme candidates can beat moderates by galvanizing their party’s base. That premise is shaping competitive races in both parties.

But an analysis of more than 30 years of House general elections suggests the opposite.

It seems that when parties nominate an extreme candidate — in a district where a more moderate candidate might have had a chance to win the primary — the extreme candidate does worse in the general election.

Extreme candidates fire up voters, but not always in the way the candidates would like.

A party that nominates an extreme candidate when it could have nominated a more moderate one may lose as much as seven points of vote share in the general election, according to Andrew Hall, a political scientist at Stanford University, and Daniel Thompson, a doctoral student there. Seven points would have been enough to swing dozens of House races in 2016 alone.

Extreme candidates, the researchers say, may mobilize their party’s base — but they tend to activate their opponent’s base even more than their own, resulting in a net loss on turnout. They also seem to lose habitual voters — the people who almost always turn out to vote — to the opposition party, while moderate nominees hold on to those voters. In this way, extreme nominees lose votes in two ways: mobilization and persuasion.

The findings initially seem counterintuitive because Congress has become more separated by ideology and party than ever. So it seems as if extreme candidates must be winning general elections more often than moderates.

But as Mr. Hall points out, both patterns can be true. Extreme candidates may be less likely to win general elections than more moderate ones, but if the supply of candidates over all is becoming more ideologically extreme, the average ideological divide in Congress could be increasing.

Mr. Hall and Mr. Thompson had to overcome significant measurement and research challenges to illustrate these patterns. They had to find a way to characterize the ideologies of candidates who were running in primaries (most of whom had no political records before running). And they had to find a way to hold everything in a district constant except whether a party nominated a moderate or extreme candidate in a place where both were equally likely to win the primary.

To overcome the first challenge, the researchers developed a donation-based method of characterizing candidates’ ideologies (similar to work done by Adam Bonica, also from Stanford). The measure assumes that candidates who receive contributions from the same donors and groups are ideologically similar.

To overcome the second hurdle, they focused on close primary races between a more moderate and an extreme candidate between 2006 and 2014. Because the researchers focus on close outcomes, they know that both candidates were viable choices for that party in that district. In these races, the final count was so close as to be nearly decided by luck, as in a coin flip. Between 2006 and 2014, about 125 races fell into this category.

This technique is important to the researcher’s approach. It’s as if the victory by the extreme candidate over the moderate one were a randomized experiment. If there are differences in the vote share or turnout patterns of the general election races in these districts — which are nearly the same in every way other than which candidate won the party primary — they are attributed to the nominees’ characteristics, including their ideology.

Essentially, these quasi-natural experiments allow the researchers to isolate the effect of the candidate’s ideology and compare the general election outcomes across districts that got one type of candidate (nearly by chance) with districts that got the other.

Using a variety of methods, Mr. Hall and Mr. Thompson estimate that in these 125 close elections, a party that chooses the more extreme nominee in the primary sees roughly a six-point drop in the number of its registered voters who show up at the polls in November. It’s a drop-off that does not exist when the party nominates a more moderate candidate.

The notion that turnout drives midterm elections is not new, nor is the idea that ideologies of members of Congress can affect vote share, nor that competition may have a moderating influence on candidates. A rich body of work in political science examines these ideas from different angles.

But few have explicitly tried to highlight the role candidates’ ideologies play in affecting turnout. It’s an important niche, especially in today’s polarized climate as party leaders may be tempted to give primary voters “what they want,” ignoring what general-election voters may prefer.

It is worth noting that many other things affect turnout in elections, including the other races on the ballot and electioneering efforts by all the candidates and groups with a stake in an outcome. But the takeaway for parties is that even if they nominate the more extreme candidate and he or she wins the general election in November — it is likely the party could have done even better in that race by nominating a viable candidate who was more moderate.

Voters may be more emotionally wedded to their party than ever before, but party strategists would be foolish to think this means voters will always show up on Election Day or that the share of swing voters in the electorate don’t care how extreme a nominee may be.

Mr. Hall and Mr. Thompson have shown political parties one way to avoid losing votes: Keep an eye on how likely your nominee is to motivate the other side to show up and nominate accordingly.