On Election Night, I wanted to make a prediction, but it was hard. What made it hard was that we weren’t getting Election Day numbers from some of the largest counties in Florida, like Miami-Dade and Palm Beach. Because of that, I had to predict the Election Day turnout in those counties, and predicting turnout is impossible. However, with that, I made this less-than-confident call:

Last map tonight. Over the last hour, the race has been getting away from the Democrats. Even a high turnout in Miami-Dade favoring Democrats might not help. So I will predict, with incomplete data*, @FloridaGOP favored. #flapolhttps://t.co/LtUdzAibbN

— Dave Trotter (@DaNumbersGuy) November 6, 2018

And there it was. On paper, the final margin (based on some turnout predictions, which is why I wasn’t confident) showed an advantage of 1.07% to the Republicans. Yes, the model was skewed a little to the GOP, but overall it was correct. And when the actual Election Day party turnout numbers come out, we may see that the model was spot on.

With that, we are starting to see some numbers come in from Florida. Again, we can’t get a complete picture of the entire state yet because we are still waiting for counties to report their returns by voting method. Regardless, there are some major trends and theories that we can look at to understand the election on Tuesday. Here are just a few of those observations.

1. The Scott-Gillum split didn’t happen: Before the election, Democrats were making the argument that voters who had voted for Scott at the top of the ticket might change to Gillum in the governor’s race because DeSantis was “too extreme”. However, as we can see by the closeness of the election, that didn’t happen. Even on a county-by-county level, we see no real change between the two races. However, because of the margins, there were probably more Nelson-DeSantis voters. Still, that number would be very small.

2. Race didn’t have an impact: Over the next week, I expect we are going to see the Florida Democratic Party establish and consultants try to deflect blame and create narrative that Gillum lost the election because of race, which was beyond their control. However, there is no evidence to back up this claim. First, the closeness of the Senate and Governor’s race shows that this isn’t really possible. However, some might argue that racists who voted against Gillum could have brought down Nelson. Again, the evidence does not show this. Doing logistic regression tests, using region as the binary dependent variable (North Florida = 1, Rest of Florida = 0, with North, Central, South, Southwest, Panhandle), there was no strong relationship that showed that racism existed. If anything, race had an impact of around 1% in the Panhandle and some North Florida counties, and that is about it.

3. Democrats likely lost on Election Day: We are still waiting for all the numbers to come in, but everything indicates that Democrats, yet again, probably had a lead before Election Day, and blew it on Election Day. Blame Jim Messina. He has created a culture in Florida Democratic politics that pre-election polls mean everything, and that Election Day is “just another day”. No it isn’t. It will now be responsible for yet another statewide Democratic loss.

4. Trump election remains a baseline for voting behavior: As I mentioned previously, this model was based solely on the 2016 presidential election results (not turnout, that is something totally different). My argument is that we are in a new era of politics, where Trumpism defines the Republican voter, and that we wouldn’t see much of a shift. Basically, if you used my model and the voter turnout composition was exactly the same, the result of the model would be exactly like the 2016 presidential election. If we had used the 2016 Senate race, or even averaged the races, the model would be drastically off (which is why having theory behind a model matters!!!). While we did see some shifts in voting, they were not extreme. I will be writing about these in greater detail once we get all the numbers in.

5. Registration reflects vote choice: Building on #4, the model assumed the likelihood of a voter supporting a certain party based on the performance of a county and the registration composition of voter turnout. Those numbers remain the same between 2016 and 2018. However, we did see an increase in support for Democrats in most rural Florida, possibly by an average of 2%. This means that a few Democrats who voted for Trump came back to the Democratic side in rural. But at 2%, it isn’t much to write home about.

6. Possible shift in NPA support in Southwest Florida: These shifts seem to be minor, but they will have an impact. Both Collier and Lake Counties saw a shift (compared to the model) of around 2% in the Democrats’ favor in Vote by Mail, and Collier possibly seeing a 2.5% swing in early vote (still need to finish a new model to determine the exact number). These are also small percentages, but over five election cycles, that can be 10%, which would make Lee County a tossup county.