Florida Prediction edition
The Florida early voting page has gone viral, and spawned a bunch of people speculating on what it all means. Is 480K too insurmountable of a lead for Republicans to overcome? Should Democrats be worried about the in-person early voting (IPEV) Republican performance? Who should be worried, anyway?
This post attempts to answer that question. Richard Baris would call this useless tea-leaf reading, but tea-leaf reading is the only thing we have at this point.
Florida isn’t like other states
The premise of “JoeIsDone” is that Democrats are not turning out the VBM numbers they need to win the election. That applies to a state like NC, but not FL so much. Why?
Florida already has a strong history of VBM and early voting, so it’s harder to infer how COVID-19 fears will impact Florida specifically.
Here are the raw vote numbers from 2016 broken by party and method:
| Voting Method | Democrats | Republicans | Unaffliated / NPA | Total (includes third parties) |
|---|---|---|---|---|
| In-person early vote (IPEV) | 1,580,003 (43.27%) | 1,425,309 (38.35%) | 779,627 (39.32%) | 3,874,929 (41.13%) |
| Vote by mail (VBM) | 1,049,809 (28.75%) | 1,108,053 (29.81%) | 504,895 (25.47%) | 2,732,075 (29.00%) |
| Election Day votes | 1,021,740 (27.98%) | 1,183,162 (31.84%) | 698,034 (35.21%) | 2,813,035 (29.86%) |
| Total votes | 3,651,552 | 3,716,524 | 1,982,556 | 9,420,039 |
| IPEV + VBM combined | 2,629,812 (72.02%) | 2,533,362 (68.16%) | 1,284,522 (64.79%) | 6,607,004 (70.14%) |
Note: I pulled the totals off this website, so the numbers are not exact.
Put simply, in 2016, a full 70% of the electorate had already early voted. Trump came into the election with a 96,450 point deficit, which was then overcame with Republicans compromising 42% of the Election Day electorate.
Current state (21-Oct)
This is the current state as of writing (excluding today’s Saratosa and Miami-Dade data):
| Voting Method | Democrats | Republicans | Unaffliated / NPA | Total (includes third parties) |
|---|---|---|---|---|
| In-person early vote (IPEV) | 378,540 | 465,313 | 161,975 | 1,017,620 |
| Vote by mail (VBM) | 1,516,008 | 968,224 | 633,935 | 3,155,740 |
| IPEV + VBM combined | 1,894,548 | 1,433,537 | 795,910 | 4,173,360 |
Breaking out by their respective 2016 metrics, we see the following:
| Voting Method | Democrats | Republicans | Unaffliated / NPA | Total (includes third parties) |
|---|---|---|---|---|
| In-person early vote (IPEV) | 23.96% of 2016 Dem IPEV | 32.65% of 2016 Rep IPEV | 20.78% of 2016 NPA IPEV | 26.26% of 2016 IPEV |
| Vote by mail (VBM) | 144.41% of 2016 Dem VBM | 87.38% of 2016 Rep VBM | 125.56% of 2016 NPA VBM | 115.5% of 2016 VBM |
| IPEV + VBM combined | 72.04% of 2016 Dem EV | 56.69% of 2016 Rep EV | 61.96% of 2016 NPA EV | 63.17% of 2016 EV |
What do the polls say?
Currently, the best poll we have on how people will vote is probably the University of North Florida poll that was released yesterday.
Within the crosstabs, we find the following vote preference by party registration:
| Voting Method | Democrat | Republican | NPA |
|---|---|---|---|
| IPEV | 32% | 24% | 33% |
| 55% | 37% | 37% | |
| Election Day | 12% | 37% | 29% |
| Don’t know | 2% | 3% | 1% |
Some oddities should jump out. For example, if the poll is to be believed, we should see a larger proportion of Republicans turn out for Election Day than in 2016. That doesn’t seem right, but Gallup says that early voting has become a partisan issue.
Reading the tea leaves…
Let’s read the tea leaves in the turnout. What if we cast 2016’s turnout to the poll numbers? We arrive at the following:
| Voting Method | Democrats | Republicans | Unaffliated / NPA |
|---|---|---|---|
| IPEV | 1,168,494 (32%) | 891,966 (24%) | 654,243 (33%) |
| VBM | 2,008,353 (55%) | 1,375,114 (37%) | 733,545 (37%) |
| Election Day | 438,186 (12%) | 1,375,114 (37%) | 574,941 (29%) |
| 2016 total votes | 3,651,552 | 3,716,524 | 1,982,556 |
| IPEV + VBM combined | 3,176,850 (87%) | 2,267,079 (61%) | 1,387,789 (70%) |
(We will ignore Don’t Knows.)
Obviously, this doesn’t align with our data. Republicans are lagging way behind on VBM, independents are way overperforming on VBM, Republicans are overperforming with early voting. .
But it’s the best we have.
But by that model:
- Republicans are 63.23% to their EV target (1,433,537 / 2,267,079).
- Democrats are 59.64% to their EV target (1,894,548 / 3,651,552).
- NPAs are 57.35% to their EV target (795,910 / 1,387,789).
Before you go out Tweeting that Republicans are actually kicking butt and taking names, there are many reasons to not to take this model as gospel.
- It doesn’t match up with the real world data.
- We don’t know how the ‘don’t knows’ allocate.
- It is doubtful that Republicans will have higher turnout on Election Day 2020 than they did in 2016.
- The Democrats appear to have requested more VBM than their target would imply (2,631,697).
And most importantly:
It’s been only three days since IPEV started.
The return rates and IPEV have been rather volatile.
The “magic numbers”
There’s been extensive discussion of the “magic numbers” - aka the gap between Democrat and Republican VBM and IPEV to secure a victory.
From the raw model above, these would be 633,239 for VBM (which closely matches the one that’s been tossed around) and 276,528 for IPEV for a combined magic number of 909,767 (!).
But, again, there are many unknowns. Conservatively assigning all of the 2% “Unknown” to the Republican EV (-74,330) and 3% to the Democrat Election Day (-109,465) brings that number down to 725,972 (combined IPEV and VBM lead).
And finally, just because I don’t trust that 37% of Republicans will turn out on Election Day, I’d further conservatively knock off a 6% (-222,990).
That brings the “magic number” to 502,982. Higher turnout will push the number up and lower turnout will push the number down.
Overall, Florida isn’t going to be a cakewalk. But should be doable for Trump with a few more days like today’s.
All this article really is, is just tea leaf entertainment reading in the end.
What’s a good “advantage” for Trump?
One of the more discussed things about the Florida early voting is the “Advantage.”
At first, I measured the advantage as the R-D gap compared to the Trump-Hillary gap. I noticed this greatly favored blue counties at the expense of deep red counties, so I sought a fairer comparison.
So I based the advantage off the 2016 early voting R-D gap. But, again, since Democrats have declared a preference to early vote, this wasn’t completely fair either.
In 2016, 39.80% of early voters were Democrat, and 38.34% of early voters were Republican (39.8%-38.34%).
According to the above poll, and adding in the 3% unknowns plus 6% Election Day bias to the Republican part to bring their early vote targets to 2,601,566, the gap becomes 44%D-36%R.
That’s a 7 point swing, so I will keep the statement on the website that anything within 5% is good news.