The oldest debate in electoral politics is between moderation and idealism. Some say that parties need to nominate ideologically-aligned candidates who can fire up the base; in theory, they claim, this would engage more low-propensity voters and drive up turnout, overwhelming the opposition through a surge of new voters. Others claim that the nominee must appeal to swing voters and practice moderation; the party base, they insist, will largely show up regardless of the nominee, and so the way to win is instead in persuading as many swing and crossover voters as possible.
The argument itself is central to electoral calculus and can influence everything from legislative strategy to primary elections, in which party officials often choose the candidate to back according to the school of thought they subscribe to.
Research suggests that the path of electoral moderation is generally more influential, and comparing ideology against electoral performance shows that there is a significant and quantifiable penalty for ideological extremism. Political Scientist Dr. Chris Warshaw found that the penalty for ideological extremism generally ranges between 1.1 and 2.3 points in vote share for House incumbents; while the penalty for this has declined over time, it remains sizable, especially in close races (and especially considering that 39 seats were within 5 points of the national environment, on margin).
Our research at Split Ticket, which uses GovTrack’s measurement of incumbent ideology, finds something similar. Within a certain range, there isn’t really a clear ideological penalty; members aren’t likely to be penalized for being on one side or the other of the partisan median in the House, so long as this deviation is not overly significant*. Representatives were, on the whole, not exceptionally likely to pay a significant penalty for being close to to the median ideological positions of their caucus.
But ideologically extreme members were significantly more likely to have severe underperformances than more centrist members were. Of the 60 most centrist members in the House in the 116th Congress, only 1 underperformed our 2020 modeled expectations (controlling for incumbency, demographics, presidential lean, and fundraising) by 7 or more points above expected, on margin. But among the 60 incumbents that were most ideologically extreme, 7 of them underperformed by this much.
A similar thing is observed among candidates likely to overperform as well. Of the 60 most centrist members in the House in contested races, 4 of them overperformed modeled expectations by 7 or more points above expected. But of the 60 most extreme incumbents, none of them had a significant overperformance after controlling for spending, district demographics, and presidential lean.
|60 Most Centrist Members||60 Most Extreme Members|
|7+ point Underperformances||1 (0 Dem, 1 Rep)||7 (5 Dem, 2 Rep)|
|7+ point Overperformances||4 (2 Dem, 2 Rep)||0|
So it’s somewhat clear that moderation is generally correlated with improved electoral performance among swing voters (or at least, in helping to avoid significant underperfomances). But there’s even evidence that it helps with turnout differentials, and that the so-called “galvanizing” of the primary base through a more ideologically in-tune candidate may be fools’ gold.
Prior studies show that ideological extremism, in general, tends to actually hurt turnout rather than helping it. A 2018 study by Andrew Hall and Daniel Thompson found that ideologically extreme candidates tend to give the opposite party a higher turnout boost than anything and actually get lower relative turnout than more moderate candidates. This is backed by a 2014 study by Jon Rogowski, which finds that increased levels of ideological conflict actually drove down turnout. All of this serves to counter the notion that attuning one’s stances to primary electorates helps margins in the general election through firing up base voters.
With that said, there are caveats to this argument. If someone is in a district on the ends of the extreme spectrum, matching their district’s ideology is probably the smarter route for them. General election overperformance is not too important in a safe district, but a continuous insistence on centrism and moderation might eventually cost these candidates dearly in primary elections, as Dan Lipinski and Joe Lieberman will readily attest to. The nature of America’s electoral system does demand some degree of primary pandering.
Moreover, it’s important to acknowledge that there are things that matter beyond inflating electoral margins. No one would expect Pramila Jayapal to moderate to try and get 2 percent more of the vote in her safe D district at the cost of sacrificing the progressive ideals she believes in — the safety of their districts gives members the freedom to advocate for what they want ideologically, and it is a valuable freedom.
Lastly, it’s certainly possible to be a more base-attuned candidate and overperform; Barbara Lee was the most liberal incumbent and still outperformed expectations by 2 points on margin. There are other reasons beyond ideology for performance deviations, and so this is not meant to say that centrism should be the ideology of all candidates, or that centrism is the only thing that wins elections.
But this is meant to push back on the somewhat far-fetched notion that ideological extremism will win elections through tapping into a latent pool of voters who stay home. Many on the left tend to claim that progressives are more electable because they activate a legion of nonvoters who are drawn to the energy and the platform of the candidate. And several on the right claim that not nominating a candidate pandering to the far-right wings of the party will lead to millions of disaffected conservatives staying home.
These are all notions that make a lot of logical sense and fit our priors nicely. The only problem is that they’re just untrue, and there is a legion of contradicting evidence available for this. To the degree that ideology appears to help with electoral margins, moderation is generally the path to follow.
*Significant movement is quantified as moving roughly one standard deviation away from the ideological mean of the member’s party; applying this a window of approximately this length in either direction filters all but 120 members in the House out — 60 centrists and 60 more extreme members.
A special thanks to Eric DeBellis and Chris Warshaw for their help with this piece.
I’m a software engineer and a computer scientist (UC Berkeley class of 2019 BA, class of 2020 MS) who has an interest in machine learning, politics, and electoral data. I’m a partner at Split Ticket, handle our Senate races, and make many kinds of electoral models.