Now that the numbers are starting to come in, we can start observing some voting patterns at the precinct level. One thing that is for certain (though we don’t have all the numbers yet), is that is around .9% of Florida voters cast a write-in vote for the presidential race. However, if you look at the Division of Elections’s site, they show only 153 votes cast for write-in candidates. The reason for this is that the state (and some counties) only count write-in votes if they actually wrote in the name of an official write-in candidate. Still, most counties do not do this until the official results are given. Therefore, during the “unofficial results” phase, counties will still give the number of voters who actually selected a write-in option at first, and then will adjust later. So in this initial phase, we get a lot of information that could be lost quite soon.
One of the possibilities discussed before the election was if voters who were upset with either Hillary Clinton or Donald Trump might write in a candidate instead of selecting a candidate on the ballot. With .9% being write-in votes, it is not enough to sway the election. Still, did they break on Election Day to favor a candidate?
To see, I just did a regression analysis. Simply, I asked if a precinct that had a higher amount of write-in votes also had a higher amount of votes for Clinton (so, basically, a #BernieOrBust vote). For those technical people out there, yes, I did do a cluster analysis because I know that OLS assumes that the observations are independent (just saying that now). Anyway, when doing it there was some directional evidence that the higher the percent for Clinton, the higher the amount of write-in votes. Here is a scatter plot graph of those results, with 2,572 precincts being observed:
When looking at the graph, it looks as if the number of write-ins go up, so does the Clinton vote. However, when we look at the R-Squared of the model (which is, using a basic definition, how much this model can explain the relationship between the variables. Click here for a video), it is .009, which means that this model only explains .9% of the relationship.
What does that mean? It means there is no relationship. Therefore, the #BernieOrBust theory does not hold. In fact, we really cannot get a good idea of how the write-in vote impacted the election. If anything, looking at the scatter plot might give us more of an indication. We see that the mass of observations are between the 30% and 50% range, which means it actually could have hurt Trump as well.
Yes, nearly 1% of the voters decided to write-in a candidate. But looking at the numbers, Mickey Mouse might have received just as many votes as any other write-in candidate.