What it cost to be extreme, by election

Vote-share points lost per unit of candidate |CFscore|, net of district presidential lean and incumbency, each cycle 1980–2022. Filled points are statistically significant (p<0.05); hollow points are not. Shaded band is the 95% interval.

Takeaway: the penalty sits well below zero for three decades, then climbs to the zero line and stays there from 2018 on.

Each point is the coefficient on candidate extremity in a cycle-specific regression of general-election vote share on extremity, own-party district favorability, and incumbency (HC1 robust standard errors), among major-party House and Senate candidates in contested races. 2024 is omitted: general-election outcomes are not yet coded in this DIME release. Source: Stanford DIME v4.0.

In the closest districts, moderates once dominated — less so now

Win rate of moderate vs. extreme candidates (own-party extremity split at the median) in competitive districts (district presidential vote 45–55%), by party, early (1990–2008) vs. recent (2010–2024).

moderate half extreme half

Takeaway: the moderate edge in toss-up seats was enormous early (a moderate Democrat won 71% of the time, an extreme one 28%) and has narrowed since.

Raw win rates, not adjusted for incumbency — part of the persistent moderate edge is that moderates are disproportionately incumbents; the incumbency-adjusted penalty is the cycle line above, and it is what vanishes. Source: DIME v4.0.

Safe seats do not breed extremists

Mean candidate extremity (|CFscore|) by district safety, by party.

Extremism is not lucrative

Correlation of candidate extremity with the size of the donor base, per cycle.