Help me keep The Crosstab ad free. Subscribe to premium content and join the community on Patreon for just $2.
Become a Patron now!

Congressional Polling Versus Special Elections: Are Democrats Really Heading for a Wave?

Congressional Polling Versus Special Elections: Are Democrats Really Heading for a Wave?

Note: this post was released early for the site’s supporters on Patreon. If you’d like early access to posts and other content, follow the link above.



Democrats are up seven points in my average of generic ballot polls today, but they’ve overperformed by an average of 17 points in special elections to Federal office since 2016. What’s with this discrepancy? Which measure should we believe, and what consequences do the differences have for the 2018 midterm elections to the U.S. House?

This post is a quick dive into this question, first exploring why the difference between the two measures exists, and then discussing why it matters.

First, most generic polls are polls of registered voters, but special election voters aren’t the same as all the registered voters nationwide; according to the current numbers, they’re likely about 10 points more Democratic-leaning. Congressional polls that poll likely voters, or those who are enthusiastic about voting, are much more Democratic. So a polling average that doesn’t include them will underestimate Democrat’s advantage in the national political environment..

Here’s an example: a NBC/Wall Street Journal poll from mid-April found that the Democrats have a 7-point lead among registered voters, but a 21-point advantage among those who said they have a high interest in voting. This could approximate the difference between

It’s worth noting before I move on that the disconnect between the average Democratic margin in national generic ballot polling also shows a disconnect from state-level generic ballot polling: a poll in Missouri showed a ~15 point swing from Trump’s 2016 margin in the state. Congressional polls conducted at the Congressional district-level show a 13 point average swing in Democratic performance relative to a House seat’s partisan lean.

Second, special elections are contests that necessarily take place in open seats — they are uncontested by incumbents — these open seats usually see larger swings in midterm cycles than races with incumbents.

So the Democratic over-performance in special elections is pronounced, higher than it ought to be in November. Historically, this has meant a 2-3-point smaller swing in November than in the special elections. So the current 17-point Democratic overperformance in special elections is likely to be closer to 14 points in Nov, based on this data alone.

However, both of these methods alone don’t suffice. We should combine both congressional election polling and special elections to congress, and state legislative elections.

When I combine both of these in a model predicting the Democratic vote margin in November, the most likely outcome is Democrats winning by 10 to 11 percent, with a 3% margin of error. So they could win by as little as 7 points or as many as 13. And this is still no guarantee that Democrats will pick up enough seats to win the majority in the U.S. House of Representatives — though a national environment in which they are up by 11 is certainly one in which Democrats favored (maybe odds of two-in-three) to do so.

The graph above charts the average shift from Democrats’ 2016 vote margin in a seat to their margin in the 2018 special election there




In short, though the individual data may slightly disagree, there is a clear story being told in aggregate: Democrats are favored to flip control of the House of Representatives come November (my model currently gives them a 55% chance). Though things can change between now and the fall midterms (there are, after all, margins of error around all the predictions I make), it is wise to expect that the Speaker’s gavel will no longer be wielded by a Republican


G. Elliott Morris avatar
About G. Elliott Morris
Elliott is a ("big") data journalist and data scientist who specializes in American politics, public opinion, and predictive analytics. He is a government, history, and computer science student at The University of Texas at Austin and worked previously for the Pew Research Center and Decision Desk HQ.
comments powered by Disqus