This article has been updated.
Lamb’s apparent victory over Republican Rick Saccone the morning after Election Day runs in the hundreds of votes in a contest where more than 220,000 people cast ballots. In the 2016 presidential contest (according to Daily Kos’s invaluable tallies of presidential votes by legislative districts), more than 370,000 votes were cast in the district, with Trump receiving 215,000 of them. Lamb got about 78 percent of the votes that Hillary Clinton received in the district — but Saccone received about 52 percent of the votes Trump received. The result? A tie. A shift of 20 points toward the Democrats.
That shift is not uncommon in special elections since the 2016 presidential race. In 86 special elections we examined since the presidential race, the average shift has been 13.4 points to the Democrats. In 34 races, the Democratic candidate has done at least 20 points better in the district than Clinton did. While the results of those races have been mixed — the Republican candidate has won about half of the special elections — the direction of the shift in the results was in the Democrats’ favor three-quarters of the time. (Remember this bit of information; we’ll come back to it.)
Let’s break those results out by party. Democratic candidates have consistently overperformed Clinton in special elections.
Republican candidates have consistently underperformed Trump.
Those two shifts are linked, of course. We’re looking at percentages here, so the Democrat generally can’t gain relative to Clinton — that is, see an increase in the percentage of the vote earned — without the Republican faring worse. It’s zero-sum or close to it.
But then there’s the question of turnout. Sure, lots of places have seen big swings to the Democrats, but, in part, that’s because turnout is low and a few thousand more Democrats turned out to the polls. We looked at this in the context of special elections in Missouri last month. Yes, the Democrats picked up a seat, but the narrower results overall were also a function of Democrats seeing slightly less terrible declines in turnout. This graph from that article tells the whole story.
So let’s overlay the shift in turnout on our party graphs above. On the charts below, the size of the bubble is scaled to how many Democratic or Republican votes were cast in the special election relative to those cast for Clinton or Trump. Bigger bubbles means that the special-election candidate received almost as many votes as the presidential candidate.
There are a lot more big bubbles on that Democratic graph than on the Republican one. Which is to say: In special elections, Democrats came closer to matching 2016 presidential turnout than did Republicans.
Overlaying those two graphs on one another:
Notice the two circles representing Pennsylvania’s 18th District. A bigger blue circle indicating that Lamb got more votes relative to Clinton than Saccone got relative to Trump. Lamb’s circle to the left of the diagonal line, indicating that he outperformed Clinton in vote share.
In nearly every special election, votes for the Democrat were a greater percentage of Clinton’s total in the district than votes for the Republican were relative to Trump’s. (That was the case in 67 of the 86 races we looked at.)
We’re using the term “turnout,” but that’s misleading. It’s really a measure of vote totals for the candidates of each party in each district, which doesn’t overlap perfectly with party support. For example: Conor Lamb was, for a variety of reasons, a more appealing candidate than Hillary Clinton in the 18th District. There were people who voted for Trump and then voted for Lamb, meaning that the shifts above aren’t entirely partisan turnout shifts. But it’s still an instructive proxy.
Let’s compare two metrics: The change in support for the Democratic candidate in the special election in a district relative to how Clinton fared vs. the change in turnout for the Democrat relative to Clinton. In other words, comparing the 20-point improvement Lamb saw in the election results vs. Clinton (she lost by 20; he tied) with the 27-point margin by which Lamb’s turnout relative to Clinton bested Saccone’s (remember that Lamb got 78 percent as many votes as Clinton and Saccone 52 percent of Trump’s).
The correlation is strong.
Which you would expect! If the Democrat gets more votes relative to Clinton than the Republican got relative to Trump, you’d expect the Democrat to do better relative to the Republican than Clinton did to Trump.
What the graph above suggests, though, is that the results of special elections aren’t necessarily related to those turnout shifts. Most of the special elections have taken place in districts that Trump won; shifts in the Democrat-to-Republican margin in those places may or may not lead to victory depending on how big a deficit the Democrats faced there in 2016.
If we compare the shift in turnout to the actual results, that’s what we see. In most cases, the Democrats earned more votes relative to Clinton than the Republicans did relative to Trump — but in only about half the cases did the Democrats then win.
Since the shift in turnout correlates to the shift in the margin, this maps to what we noted earlier in that bit of data you were supposed to remember. (Did you remember it?)
There is a theory that’s floated around the special elections that in higher turnout elections, the Democrats would fare less well because the effects of a flood of enthusiastic voters would be muted. The case in point is Georgia’s 6th Congressional District, where Democrats did no better than they did in 2016 in a House race that attracted national attention.
The difference there was that the Republican, Karen Handel, did about as well in turnout (by our definition here) as did the Democrat, Jon Ossoff. The number of votes she received was 84 percent of the votes received by Trump in the district; Ossoff got 81 percent of Clinton’s votes. Handel’s relative vote total was one of the largest of any Republican in special elections since 2016.
In other words, that race was the outlier for special elections. Lamb’s isn’t.
The question for November is which of these races will be closer to capturing what’s happening nationally. Will Republicans run even with Democrats relative to 2016 as they did in Georgia or will Democrats overwhelm them as they did in Pennsylvania?
Special election results suggest that if Democrats can match 2016 votes for Clinton by 10 points more than Republicans match Trump’s 2016 votes, they will gain 21 points relative to how they did during the presidential contest. If Republican candidates get about as many votes relative to Trump as the Democrats do to Clinton, the Democrats have a slight advantage.
That’s the question pollsters are trying to answer: What will that enthusiasm gap be? Once again, we’re left waiting until Election Day.
Update: A tweet from Nate Silver inspired me to think about another way of looking at this data.
“Democrats are now winning **high-turnout** special elections in very red districts,” Silver wrote. “That’s really bad news for the GOP.”
So let’s look at the data above using turnout in special elections overall relative to turnout in 2016.
The special elections in the 6th District in Georgia and the at-large seat in Montana were both high-turnout, as measured as a percentage of 2016 turnout (which, as a presidential election, was about as high as you’ll generally see). Republicans won both of those. But then there is that Alabama Senate seat and the 18th in Pennsylvania — both won by the Democrats.
Both of those wins happened after November of last year. That certainly does not, as Silver says, seem like good news for the Republicans.