Skip to main content
ABC News
The ‘Enthusiasm Gap’ Could Turn A Democratic Wave Into A Tsunami

It’s a busy time of year, so we’ll keep this relatively short. But I wanted to echo and underscore a point made by FiveThirtyEight contributor Nathaniel Rakich in his article on Tuesday’s special election in Pennsylvania’s 18th Congressional District. One reason that the results are especially scary for Republicans — Democrat Conor Lamb is the apparent winner1 in a district that President Trump won by 20 percentage points — is because it came on reasonably high turnout, the sort of turnout one might expect in this year’s midterms.

As of early Wednesday morning, about 228,000 votes had been counted in the special election. That equals 62 percent of the 370,000 votes cast for the presidential candidates in the district in November 2016, according to data from Daily Kos Elections. That’s right in line with midterm turnout, which is typically 60 percent to 70 percent as high as turnout in presidential years.

Turnout was similarly high in December’s special Senate election in Alabama — 64 percent of presidential-year turnout in the state. That was also an impressive win for Democrats, albeit one with more mitigating circumstances.

The high-turnout wins in Pennsylvania and Alabama ought to reassure Democrats — and worry Republicans — because there had previously seemed to be a pattern in which Democratic results were most impressive in low-turnout special elections. For instance, Democrat Archie Parnell surprisingly came within 3 percentage points of defeating Republican Ralph Norman in the special election in South Carolina’s 5th Congressional District, which represented a 16-point blue swing relative to the district’s partisan lean.2 But that result came on minuscule turnout — only about 88,000 votes, or 29 percent of the district’s 2016 presidential turnout. By contrast, Democrat Jon Ossoff’s underwhelming performance in the Georgia 6 runoff that same evening — which represented only a 6-point Democratic swing from the district’s partisan baseline — had come on exceptionally high turnout, equalling 79 percent of the district’s presidential vote.

After nine federal special elections that pitted the two parties against one another, however — including both the open primary3 and the runoff in Georgia 6 — there’s no overall pattern between turnout levels and how Democrats perform:4

Democrats are performing strongly in both high-turnout and low-turnout special elections

Since the 2016 presidential election

Special Elections
2016 pres. vote count Vote count as SHARE of pres. votes Dem. Swing
Georgia 6th runoff 331,246 260,316 79% +6
Montana At-Large 501,822 381,416 76 +16
Alabama U.S. Senate 2,123,372 1,348,720 64 +31
Pennsylvania 18th 370,497 228,378 62 +22
Georgia 6th open primary* 331,246 192,569 58 +7
Utah 3rd 289,923 147,741 51 +3
Kansas 4th 274,547 122,594 45 +23
South Carolina 5th 306,200 88,316 29 +16
California 34th open primary* 184,601 42,308 23 +18

Democratic swing is the difference between the special election result and the district’s partisan lean. Partisan lean is the average difference between how the constituency voted and how the country voted overall in the last two presidential elections, with 2016 weighted 75 percent and 2012 weighted 25 percent.

* Primary potentially included multiple candidates from each party, and results show vote share for all candidates from each party combined.

Sources: State websites, Daily Kos, New York Times, dave leip’s atlas of U.S. presidential Elections

Thus, Republicans have one less excuse for their string of really awful special election performances. It’s true that other measures aren’t as bad for Republicans as these special elections — for instance, they trail Democrats by “only” 8 or 9 percentage points on the generic congressional ballot, which suggests a close race for control of the House this year that only narrowly favors Democrats. By contrast, the 16- or 17-point5 average Democratic overperformance in special elections so far suggests a Democratic mega-tsunami.

But those special election results consist of actual people voting, whereas generic ballot polls are mostly conducted among registered voters — or sometimes all adults. (Very few pollsters will apply their likely voter models until later this year.) In midterm years, polls of likely voters sometimes show a substantial gap from those of registered voters — there was about a 6-point enthusiasm gap favoring Republicans in 2010, for instance, which took that year from being mildly problematic for Democrats into a massive Republican wave that saw them pick up 63 House seats.

And there were signs of an enthusiasm gap even within Pennsylvania 18 on Tuesday night. According to the Cook Political Report’s David Wasserman, turnout in Democratic-leaning Allegheny County equaled 67 percent of presidential-year turnout, but voters turned out at only 60 percent of presidential levels in Republican-leaning Westmoreland County. That sort of turnout gap suggests that registered-voter polls could be underrating Democrats in this year’s midterms — and could turn a challenging year for Republicans into a catastrophic one.


  1. This analysis won’t really change if Republican Rick Saccone wins by a few tenths of a point instead of Lamb, however.

  2. Partisan lean is the average difference between how the constituency voted and how the country voted overall in the last two presidential elections, with 2016 weighted 75 percent and 2012 weighted 25 percent.

  3. For the open primaries in Georgia 6 and California 34, I aggregate the votes for all Republican candidates and all Democratic candidates in calculating the partisan swing. I don’t include the runoff in California 34, which was between two Democrats.

  4. The correlation between turnout (relative to 2016) and Democratic overperformance (relative to the district’s partisan lean) is -0.13, or essentially nothing.

  5. Depending on whether you count both the primary and runoff in Georgia 6 or just the runoff.

Nate Silver founded and was the editor in chief of FiveThirtyEight.