Florida Vote Projection Model – Day 13: Big Momentum for Democrats with Vote by Mail ballots.

Over the last few days, there has been a momentous swing toward the Democrats in the latest projection model. Many might think that this is because the Democrats have performed well in early voting. But that really isn’t the case. Yes, Democrats are outperforming the Republicans by 2.28% on the early vote model, but Democrats have closed the gap when it comes to VBM ballots. Yesterday, it seemed as if the Democrats hit a wall when it came to VBM, with the model gap staying at 2.94%. Today, that gap is now down to 1.19%, still favoring the Republicans. Part of the reason for this is that the highly Republican counties are starting to see VBM ballots trickle in, while the Big Three in South Florida still see more VBM ballots coming in, usually favoring Democrats.

To give you an indication how things are trending for Democrats in the VBM, in Miami-Dade County, Democrats have now submitted more VBM ballots than Republicans. In 2014, Republicans actually submitted more ballots. If this trend continues, Democrats chances of winning will improve drastically. This also may indicate that Election Day turnout might be lower than usual, but stay tuned on that one.

Turnout Summary: With the book closing numbers finally published by the Division of Elections, we can finally get an accurate turnout rate. As of this morning, 13.23%, or ~1,756,111 of voters have submitted their ballots. The turnout rate for Republicans is 16.01%, while for Democrats it is 14.11%. The Enthusiasm Gap is now down to -4.92%, still favoring Republicans.

The Bottom Line: The gap between Democrats and Republicans has drastically dropped over the last few days. The gap between a generic Democrat and generic Republican is now down to .41%, swinging to the Democrats by 2.53% over the last four days. In a hypothetical generic race, a Republican candidate would win 48.6% of the vote while a Democrat would win 48.19%. The Enthusiasm Gap continues to somewhat mirror the overall swing gap at 2.58, indicating that it is still a base race. But it being a little bit off indicates that some NPAs are starting to favor Democratic candidates.

Details: We already mentioned Miami-Dade, but the margin in Broward County continues to climb. While it has been argued by many in Democratic party circles that Democrats shouldn’t “rely on the Big Three”, it looks like they might be the key to pulling off an election win for the Democrats especially with other Democratic counties, like Orange and Osceola, under-performing for Democrats..

On the flip side, early voting is heavily favoring Republicans in Sarasota County (as it has in the past). At the rate in which early voting is influencing the electorate, Sarasota could flip to the Republican side by the end of the day.

Seminole and St. Lucie Counties continue to be slightly Republican.

What to watch: We, as observers of Florida politics, have to ask ourselves is if Hillsborough County can be considered a “swing county” anymore? The margin for Democrats continues to increase during the vote projection, with Democrats never seeing a real threat by the Republicans in Hillsborough County. Yes, the races can be close, but the vote continues to trend Democratic.

Also, Alachua County might be challenging Broward County and Gadsden County as the “most Democratic county” in the state. Right now, Alachua has a better Democratic performance than Gadsden County, and is only slightly off of Broward’s performance. Additionally, the Enthusiasm Gap in Alachua County is at 11.27, strongly favoring Democrats.

Here is a link to the latest model.


  1. I’m surprised you think the #s look good for Democrats the last few days. I disagree.

    In 2016 on this day, the net Rs vs D ballots (in person + mail in) was only net +6,975 to Rs and declining (mail in + in person) and Rs still won. This year its currently +53,160 R votes and holding steady even with the most conservative area of the state (panhandle) votes down 66% from 2016 levels with most of the rest of the state nearly the same as 2016. And it’s likely the Panhandle will catch up on election day.


    This looks like an R blowout in Florida IMO


    1. NPA/Minor party voters vote as well. 17.58% of the current electorate. You don’t account for that in your numbers. Also, this is a projection of votes currently cast, not overall projection. Yes, Election Day matters. This model does not predict Nov 6th vote, but, as of now, Oct 25th vote.


      1. In 2016, 60% of the vote came from early vote and the trends were pretty consistent – Mail in favored R, early in person favored Ds (Ds had 100k more early votes cast than Rs in 2016 on election day), and in person voting on election day favored Rs. This was true in 2012, 2014 and 2016. That was how I knew FLorida was likely going Trump (barely) in 2016 and NC was going Trump as well. The trend was remarkably stable in 2016 and has been so far this year as well. Right now the mail in is favoring Rs like 2016 but early in person voting has not been D favorable like in prior elections. Net, Rs are doing much better vs 2016 even with the Panhandle mostly off line. Unless Independents go 2:1 or better Ds, which I highly doubt, this will be an R blowout unless something changes soon.


    2. I agree with Rob, Dave how are you calculating your NPA vote? at what percentage do you have at skewed toward Democrat candidates? and how are you coming up with that skew, through polling?


      1. NPA split is based off of projected split from 2016 (since that voting behavior will more than likely reflect 2018 voting behavior since we are in a Post-Trump electoral landscape). So, in a county like Okaloosa, NPAs will strongly favor Republicans, while they strongly favor Democrats in Broward.

        The problem with Rob’s methodology is that he only uses voter registration, and assumes a Republican or Democratic vote equates to a vote choice for the same party. Using Rob’s methodology, this would mean that Democratic candidates are winning in Calhoun, Liberty, Jackson, Taylor, Lafayette Counties simply because more registered Democrats have voted in those counties. My model shows that Republicans are winning those by 70%+. His model doesn’t account for any theory regarding vote choice.

        I’m pretty confident with my methodology 🙂


      2. Dave – Early voters tend to be the most partisan, eager to vote and very rarely cross party lines. Even on election day, it’s rare that more than 10% ever cross over to vote in states like FL with candidates like this and that is usually canceled out by the other party having the same amount vote the other party. The only real question is independents, which do swing from election to election. The methodology worked extremely well in 2016. Even if more Rs cross-over to Ds than in 2016 (I doubt, with T at the top of the ticket in 2016 would probably have led to the most cross votes in any recent election), the fact that the early vote greatly depresses the Rs margins because of the Panhandle would more than cover that.


      3. But that is all theory. What methodology are you using the test the theory?

        Also, how do you account for the voting patterns of the counties listed above?


      4. Methodology is fairly simple – Assume 90-92% of Rs vote R, 90-92% of Ds vote Ds (which study after study shows this to be true – even in 2016,when many Rs thought about voting D or 3rd party b/c of Trump), and Indies are 50% +-10% (only wild card).

        Then you compare the # of votes at this time statewide today for each of those groups compared to the same time in prior elections (as people are creatures of habit and parties GOTV efforts in each state are fairly similar election to election).

        At a starting point of Indies are neutral, at this point in 2016, the Rs led by less than 10k votes. Today, it’s over 50k even with the most Republican area voting down 2/3 from 2016 with all other areas combined roughly in line with last time.

        If you look at 2012, 2014 and 2016, you’ll see the D votes led early in person voting by a considerable margin (34k votes at this point in 2016). In 2018, it’s less than 3k votes and yesterday Rs actually added more than Ds did (+2500 net pickup for Rs). Early mail in voting is +20k more votes to Rs net than Ds at the same as 2016 even with the Panhandle at depressed levels. Independent votes at this point are slightly smaller than in 2016, making up about 20% of the total vote so far. But again, unless Independents go 2:1 D (doubtful), or trend drastically changes, once the Panhandle votes get going this could be a bloodbath (btw this methodology worked in IA, FL, NC, OH and other states in 2016 almost perfectly assuming 50/50 with Independents)



      5. The methodology is based on too many assumptions and lacks parsimony.

        1. The methods are base on assumptions. For example, assume 90-92% of Ds vote D and Rs vote R. That does not account for variance in county-level results where a high number of registered Democrats vote in a county that went 70% for Trump. The models cannot explain variance, thus, even at face value, presents challenges.

        2. When you mentioned at “this time” compared to previous elections, you are again assuming that the voting patterns are going to remain the same without any theoretical basis for that assumption. Voting patterns aren’t linear or static, but can be erratic. Also, variables can change, such as movement in early voting dates, increase in early voting booths and locations, increase in VBM ballots requested, whether the date you are comparing falls on a weekend or a weekday (in comparison to where the date lands this year), “if there is a hurricane” (Katherine Harris’s favorite). and various other factors. To simply say that “on October 25th in 2016, it looked this way” doesn’t account for any movement in independent variables. Instead, the model requires the numbers to be static to be accurate. And if the model doesn’t allow for movement in the variables, and thus only replicates previous elections, how can you reject the null hypothesis? Basically, it only compares numbers.

        3. You treat the State of Florida in a monolithic manner. What I mean by this is that your model assume that a Democratic voter in Lafayette County is going to behave the same way as a Democratic voter does in Broward County. However, there is a 77% chance that a Democratic voter in Lafayette County will vote Republican. Back channel this, and you will see my methodology. Additionally, if Democratic and Republican performance isn’t figured out, then one cannot begin to understand NPA/minor party voting. Again, it would only be based on assumption.

        4. I still cannot tell by your description if you have a forecasting model or a projection. It seems as if you are trying to use projection (even if static) to forecast. Still, the forecast is based on assumption (exp: once the panhandle comes in there will be a bloodbath).

        Simply comparing the mean of previous years to current trends is not methodologically sound. It is also not theoretically based. I mean, that isn’t a real problem, but I figured you would want you model based on some theory in order to come up with a methodology. Instead, it seems like a model looking for a theory. And comparing previous elections to current election trends provides a number of threats to external validity. Even with internal validity, the inference is based on assumptions and not necessarily causation. So maybe you do have a theory, but county-level variance throws it out of sync.

        Good luck with your model. It is good to see other people trying to figure this election out.


    3. Of course u r correct
      If missing Panhandle were included, Mail-in total margin would likely be close to 80k
      Worse for D’s is the early vote. Almost tied without 32 Counties (strong GOP) have not begun early vote till today – some not till Sat


      1. But EV and VBM is not static by county.


  2. Arthur Madden · · Reply

    Is the Panhandle back to reporting as normal?


  3. Yes, but I am interested in seeing how Escambia performs.


    1. Escambia on both Absentee & early is approx 1.6/1 in favor R


  4. Obviously 2018 is a series of local elections.
    However this date in 16 Dems were +7 in early vote. In 18 Dems -4


  5. As of 7am 19/27
    Absentee +60k GOP – up 7k from yesterday
    Early voting +1k GOP – up 2k from yesterday
    21 mostly small Districts (20 pro GOP) not voting yet
    If trend continues GOP wins across the board


    1. Correction meant 10/27


    2. But today is the 26th? 🙂


      1. My error on dates
        But facts correct
        Chk “Statistics” on FL Election site

        If history (last 4 elections) plays out GOP wins Election Day & wins 4-6% across the board

        My sense (not validated) is that Broward running at same great % rate for Dems BUT #’s underperforming. GOP Counties mostly overperforming on #’s


      2. Again, as was mentioned previously, assumptions. Look at my threats to internal validity listed above in #2.


  6. Dems are in trouble unless souls for polls hits big on Sunday. The only other way things could be going good for them is if Indies are tilting a lot more Dem this year than 2016.


    1. Actually, when I wrote “unless something changes, something I will address on November 5th” in today’s article, that is basically what I am going to say on November 5th 🙂 Independents could shift.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: