Who Is Voting Early?
By Kevin Reuning (@KevinReuning)
Election Day is next Tuesday but already more than 22 million Americans have already cast their ballots. Early voting records have been set in several states including Ohio and Texas. There is a lot of speculation about what this means for Democrats. I dove into the polling data from the New York Times Upshot to try to get a better picture on early voting. What I find is that those most likely to have already voted are Democrats and those voting for Democrats. The motivation is real.
Before continuing, I do want to stress, that there is one thing that this data can definitely not answer: who will vote on Election Day. It is entirely possible that Democrats just like voting early, Republicans like voting on Election Day. That said, the extent of early voting can still be informative.
In order to investigate who is voting early I created a model predicting who would report that they have “Already Voted” in the New York Times Upshot surveys. I included only districts where at least 10 people had reported that they already voted (people sometimes report that they have voted already even when they haven’t), which narrowed that analysis down to the 29 surveys that they did (N= Not voted: 20,129 , Voted: 2,470 ). This is of course not a representative sample, but it can still provide guidance. To account for the variation across races I used a random effects model including random effects for the race (this will help reduce the chances that any findings across demographics are the result certain races attracting a lot of early votes in districts that are heavily tilted towards a single demographic group).
First let’s explore who is voting early based on race, gender, and party affiliation. In this figure I show the probability that someone reports that they were voting early based on these characteristics. In general Democrats are the most likely to report that they voted early, while Republicans are substantially less likely. This difference is most extreme among white men and women which make up a sizable chunk of the sample of voters here (the percentages below each category show size of the group in the sample).
Another important note is that we see that Democratic identifying Latina women are very likely to have already voted.
We can also look at the differential likelihood to have already voted based on who they say they will vote for. To do this I estimated the same model again but included the reported vote choice. You might be surprised that there is a positive probability for people to be undecided and to have already voted. This combination does appear in the data and reflects either voters that are uncomfortable saying who they voted for or are confused about the question on when they will vote.
There are perhaps two shocking points from these results. We see that again white women and men are highly motivated to have voted already but only if they report that they are voting for Democratic candidates. Second, among non-whites there is a lot of motivation to vote no matter who they are voting for. It is important to note here that the likelihood of non-white voters voting for Republican candidates is relatively small. Given this all of this should be taken as evidence that there is a lot of motivation among Democratic identifiers and Democratic supporters to vote early.
Finally, one less happy note for a Democrats, which is persistent age gaps in early voting. Older Democrats and Independents report having already voted at higher rates than younger groups. In addition the partisan gap in who is voting early diminishes among younger votes.
It is obviously not too late for young voters to turnout at higher rates. Over the next week more and more people will vote early. Hopefully the results here provide some guidance in how we interpret these early votes.
Kevin Reuning (@KevinReuning) is an assistant professor of political science at Miami University.
All the R Scripts to run the above models are available here on Github.