Here we are, finally, the last day of early voting. As in other states, early voting and vote-by-mail turnout has been massive. With 4.85 million Floridians casting ballots, the difference between registered Democrats and Republicans voting is only 32,489. If yesterday’s voting pattern continues, then it is possible that more Democrats will have voted in this election than Republicans.

Turnout Summary: So far, 4,850,897 people have voted, which is a turnout rate of 36.53%. The turnout rate among Republicans is 42.32%, while for Democrats it is 39.41%, and for NPA/minor party voters it is 25.22%, a decent increase from their totals in 2014. The Enthusiasm Gap is down to -2.65%, favoring Republicans but a .64% swing to the Democrats, the largest swing we have seen during the entire election. And also, I might want to reconsider, and call this the “Registration Gap”, as that is truly what it is.

Bottom Line: I will keep the original method map up on Tableau, but from this point forward I will only be using the conflated model for comparison because the pre-Election Day electorate is not resembling past electorates when it comes to voting methods. As I said yesterday, the conflated model is more parsimonious and cleaner.

With that, the model shows that a generic Democratic candidate would receive 49.01% of the vote while a generic Republican candidate would receive 48.09% of the vote, which is a .92% gap. This puts the race outside of recount territory, and is a .83% shift to the Democrats, the largest one-day swing that we have seen throughout the entire election.

Details: What is causing this shift. That is simple, South Florida. In 2014, South Florida was 23.18% of the overall vote. In this election, South Florida is 25.09%. Additionally, it seems as if South Florida will have a stronger Democratic performance than in other recent midterm or presidential elections. Combine these two factors and it is a winning combination for Democrats. Also, the influence of Lee County is shrinking. Remember, Lee County was inflating the vote to favor Republicans early on. But now we are seeing some balance.

What to watch for: From this point on, the NPA/minor party vote is the key to the election. And so far, the numbers show that NPA/minor party voters are turning out in counties where Hillary Clinton did well in 2016. If you look at the graph to the left, you see that, in most cases, as Hillary’s vote goes up, so does the number of NPA/minor party voters in 2018. There are some outliers, but the model has a adjusted R-squared of .4149 and shows a coefficient favoring Democrats, and is statistically significant (P>|t| = .000). This means that NPAs are probably going to vote more Democratic based on their past performance. This, along with everything else, is good news for Democrats.

However, Election Day can change everything. We have two days to wait and wonder. (By the way, I will be doing live voter model updates on here and Twitter @danumbersguy on Election Day. with a final prediction made when the first polls close).

Caveat: So, I am adding this hear because there is something I want to address. Can this vote model be wrong? Absolutely. However, because it is based on previous voting behavior, the only way that the model would be drastically incorrect is if there has been a change in voting behavior, particularly among NPA/minor party voters. Usually, I would not include this caveat because I would feel comfortable in the model. But something seems to be quite different with this election, as I wouldn’t be surprised if NPA voters deviated drastically from previous voting trends. If that is the case, then back to the drawing board, but also time for a new analysis of the electorate.