(CNN)Democrats have been winning elections in places they aren't supposed to since Donald Trump became President.
Special elections are good for Democrats, but beware the margin of error
On Tuesday night, Democrat Linda Belcher won by 37 percentage points in a Kentucky special House election. Trump carried the same district by 49 points.
Any one race can be an outlier, but this Kentucky race -- along with the more than 90 special elections that have been held since Trump became President -- tell a clear story: Democrats are outperforming their baselines by a significant enough margin in special elections so far as to be favored to take back the House in November.
Special elections are far from perfect in predicting future results, however, so we need to be humble about this projection.
Democrats are doing about 13 percentage points better in 92 state and federal special elections held since January 2018 than the presidential baseline (based off the prior two presidential election results with 2016 weighted more heavily) would indicate they would do in a neutral political environment. All told, they're doing better than their baseline in 74% of the special state and federal elections so far this cycle.
Using special federal election data in midterm cycles dating back to 1994, there's a clear relationship between the average special election overperformance and the national House popular vote. Democrats would be projected to win the national House race by about 13 percentage points, given special election results so far. Now, it should be emphasized that this is based off data in special elections held so far, so it could shift around as more special elections occur this year.
The far bigger issue for Democrats, though, is the margin of error on this projection: +/- 10 percentage points with 95% confidence. An example of this error in action is that Democrats outperformed their presidential baseline by 15 points in the average special election in the 2006 cycle and only won the national House vote by Democrats 8 percentage points in November 2006. If the Democrats current overperformance holds through to November, the Democrats would probably win the national House vote by anywhere from 3 to 23 percentage points. A win of less than 7 percentage points makes them no better than a 50-50 shot to win the House of Representatives.
This projection though may be too optimistic for Democrats, however. Daily Kos Elections' Daniel Donner has compared special election results since 1990 to the baseline of how previous candidates running for a specific seat did in both state and federal elections (e.g. looking at previous House candidates for a House district as opposed to looking at the presidential baseline as we have in this analysis so far). Like with the presidential baseline, these numbers also suggest Democrats are vastly outperforming what we'd expect in a neutral year.
Donner's data forecasts that Democrats would be expected to win the House popular vote in November by about 10 percentage points. Again, that would be enough to win back a majority of seats.
This popular vote projection comes, however, with a true margin of error of about +/- 9 percentage points. So any national House result between a Democratic win of 1 and 19 percentage points is plausible. Republicans would almost assuredly be able hold onto the House if they lose the House popular vote by only a point. Obviously, something closer to the 10 point margin is more likely to happen than something on the extremes. Though, as 2016 taught us, outliers sometimes happen.
Democrat's aren't always overperforming, which means there's plenty of variance from special election to special election. This month, for example, they're outperforming the baseline by about 20 percentage points on average, though they did only 6 percentage points better on average in November 2017 races. (Note: I pulled these two months to illustrate some inconsistency in the data, not to illustrate that there's a trend in the Democrats' direction over time.) The variation demonstrates why it's important not to read too much into one election outcome and why forecasting any individual race based off the average is foolish.
Still, holding all things steady, most races should fall at least someone close to the average. And if the current trends continue, they'll have a decent shot of winning in the deeply red Pennsylvania 18th congressional district in next month's special election.
That's good news for Democrats heading into the November House elections. In that month, all 435 House races will be up for grabs, so even if Democrats underperform in a few, history indicates that they should vastly outperform their baseline in most of them.