Florida Early Vote, a retrospective

It is time for one last big data piece on Florida 2016.

For about 18 hours a day over 2+ weeks, I found myself living and breathing early voting data. So now that all the data has been reported from counties, I wanted to take a look back at some assumptions, and compare them to the actual voting data.

Before I begin, there are five things to keep in mind:

1. Every time I talk in percentages, those percentages are relative to the two-party, ie Trump v Clinton numbers. I have no use or interest in playing the "what if" questions around third-party votes, so the data in here is just the two party vote. For what it is worth, this is pretty standard for my blogs.

2. I compiled this data over the entire month of November, often by pestering counties to provide data they don't have on their website. Some of the data came before the final, final certified versions, so there might be exceptionally slight variances - like tens of votes in a county - from the state final counts. However, there is nothing that happened so significant to change any findings.

3. When I talk about early voting, that is both in-person and vote by mail combined, unless I specify otherwise.

4. For the sake of interpreting the data, everything that wasn't an in-person or traditional vote by mail ballot was allocated to Election Day. So this means that there are likely provisionals from in-person early, and vbm, as well as late military ballots in Election Day. I don't think the impact of this is significant, but I'm flagging it regardless.

5. We know how people voted on Election day, but we do not know yet who voted on Election Day. In terms of firm lessons and takeaways, some of that will have to wait.

And since I was wrong about the final outcome, before we get started, here were some of my macro-assumptions going into Election Day.

When early voting started, I thought Presidential turnout would fall about 9.2 million votes. Because of early vote turnout, and based on who was left to vote on Election Day - namely voters who voted on Election Day in 2012, I modified that projection to 9.5 million late during the second week of early voting, and assuming that 3% of those would vote for someone else, this meant slightly over 9.2 million would vote for either Trump or Clinton.

I was assuming going into Election Day, we were at about 67-68% of our total turnout, and while the Democrats had a 96,000 lead among registered voters heading into Election Day, I was operating from a place that her lead was between 3-4%, largely due to the overwhelmingly diverse nature of the NPA vote, which would put her raw vote lead between 180-250K votes.

This meant Trump had to win Election Day, on the low-end by about 5.8% to upper end of 8%, just to break even. Both of these numbers are above Romney's Election Day win in 2012 (I can't remember McCain, but I suspect it is above McCain as well).

Here are two other things baked into my assumptions: Republicans had about 100,000 more "certain" voters left to vote, though when you looked at just 2012 voters, the number was about 40K.

So worse-case scenario, Democratic turnout struggled and only the certain voters turn-out. the R versus D lands about even for the entire election, and the early vote strength combined with a a more diverse NPA vote would carry the day. I think my final memo pegged her winning Florida by about 1.5%, which was about 130K votes, meaning on the more optimistic view of Clinton's early vote lead, Trump could still win Election Day by more than Romney, and she'd still win.

Since Trump is a golfer, I described his challenge on Election Day in golf terms: a 250 yard shot over water.

So here are the toplines:

9.42 million Floridians cast a ballot for President. For what it is worth, 9.58 million Floridians cast a ballot, though it was only 9.3 million in the Senate race.

9,122,861 Floridians voted for either Donald Trump or Hillary Clinton in 2016.

Trump's margin was about 113K votes, or roughly 1.2% out of the two-party voters.

69.3% of the vote was cast before Election Day.

Of the VBM/early vote, Clinton won by just over 247K votes -- roughly a 4 point edge (she won both VBM and early vote)

On election day, Trump won by 360K, or a roughly 13 point margin over Clinton.

Toplines versus basic assumptions:

Turnout on Election Day was slightly lower than I expected, by about 80-100K votes. Given that my projection was based largely on the number of 2012 voters who had yet to vote, it was almost certainly lower because some share of 2012 Election Day Democrats didn't show up, and, more than likely, another share voted for Trump. This is the big question I will be looking at when the state updates the final 2016 voter file.

Clinton's nearly 250K vote lead was actually at the upper-end of my projections. Honestly, this surprised me. I suspected some of my optimism in the numbers leading up to the election was misplaced, and honestly thought as I put numbers into Excel, that we'd see she had gone into Election Day with a narrower lead. However, almost everything was landing right on target for her to win. As I get more into this, and look at some of the benchmarks I tracked throughout, you can see the pattern for my optimism going into Election Day.

However, Trump just crushed Election Day. There is no other way to look at it. And as I discussed in the first look back at the numbers, it really happened in just a handful of places: namely the Tampa and Orlando media markets. For example, his two-party vote share was 8.39% higher on Election Day (56.44) than Early Vote. (48.05), but in Tampa it was up 8.92% (51.5% EV, 60.42 ED), and Orlando was up 9.08% (48.8% EV, 57.88% ED). Less than 3 million voted for Bush or Clinton on Election Day, yet he won the day by 360K votes.

How big is that? Bush won Florida in 2004 by landslide for Florida proportions: 380K votes -- out of 7.6 million cast. Trump's Election Day margin almost matched it.

Benchmarks

For most of early voting, I tracked a variety of benchmarks, namely Hillsborough (the only county that voted for Bush and Obama both times), the I-4 corridor counties, South Florida and #Duuuval county. So for the sake of this exercise, let's start there:

Hillsborough:

Clinton went into Election Day with about a 29K partisan advantage among early voters, or a partisan lead of about 6.8%.

When the votes were cast, she carried the early voting period almost 44,000 votes, or almost 11% of the two-party vote. Trump won election day by just under 2 points, or right at 3,000 votes, so when all was done, Clinton carried the county by 41,000 votes. The final percentage margin, 6.8% was almost the same as Obama, and her raw vote win was about 5,000 votes larger.

The county was a little below where it should have been for turnout. Hillsborough is typically about 6.% f the statewide vote, but it landed at 6.3%, largely because its Election Day share was down -- only 29% of Hillsborough votes came on Election Day.

Long and short of it, Hillsborough could have been a little better, but that number is right at what a win for Democrats looks like.

I-4 Corridor

Hillary Clinton won the I-4 counties by almost 162K votes, but here the Trump surge on Election Day is very evident. She won these counties by almost 200,000 votes in the early/vbm phase, yet Trump won Election Day by almost 35,000 votes. Overall, Clinton won the early phase with 56.3% of the two-party vote, though only won 47.3% of the Elecion Day vote -- a surge which exceeded his statewide average.

When you look at the Volusia and Polk numbers, you can see the seeds of how Trump won on Election Day. Compared to the state, both saw their Election Day turnout levels exceed Early Vote -- with 34% of the Volusia vote coming on Election Day, and over 40% for Polk. Once fairly Democratic Volusia has been the canary in the coal mine for a free cycles - there is a reason I've highlighted it in blogs for years. If I was going to do qualitative research into 2016, I'd start with focus groups in Volusia.

Pinellas is a slightly different kind of animal, but his Election Day performance is probably indicative of late deciders breaking almost exclusively for Trump. Had the FBI Director not chosen to insert himself into the campaign with a week to ago, I suspect Clinton would have carried Pinellas (albeit very narrowly).

In total, 24.1% of the statewide vote came from these counties, of which 70.6% of the vote came before Election Day. Another way to look at it: while only 29.4% of the total vote from these counties came in on Election Day, 33.4% of Trumps' vote total from these counties came in on Election Day. I suspect when Election Day voter data comes out, we will see a cratering of minority participation.

Volusia (Daytona)
Final early vote party spread: 39.6 R, 37.1 D, 23.3 NPA R + 4,302
Actual early vote spread: Trump +8.88% (+14,754 votes)
Actual Election Day spread: Trump +22.28% (+19,162 votes)
Final results: Trump +33,916 (54.3-41.4%). In 12, Romney was +2700 (+1.15%)

Seminole – suburban Orlando
Final early vote party spread: 41.0 R, 35.0D, 24.0 NPA R +10,316
Actual Early Vote spread: Clinton +1.84% (+2,989 votes)
Actual Election Day spread: Trump +12.36% (+6,518 votes)
Final results: Trump +3,529 votes (48.1-46.5%). In 12, Romney was +13,500 (+6.5%)

Orange (Orlando)
Final early party spread: 45.8 D, 29.5 R, 24.7 NPA D +67,155
Actual Early Vote spread: Clinton +29.71% (+116,949 votes)
Actual Election Day spread: Clinton +13.49% (+17.729 votes)
Final spread: Clinton +134,678 votes (59.7%-35.4%). In 2012, Obama was +85,000 (+18.2%)

Osceola – heavy Hispanic suburban Orlando.
Final early vote party spread : 47.1 D, 26.2 R, 26.7 NPA D + 22,625
Actual Early Vote spread: Clinton +29.71% (+30,645 votes)
Actual Election Day spread: Clinton+13.98% (+4,512 votes)
Final results: Clinton: +35,157 votes (60.4-30.6%). In 2012, Obama was roughly +27K (+24.4%)

Imperial Polk – between Tampa/Orlando
Final Early Vote Party Spread: 39.6 R, 39 D, 21.4 NPA R +1,085
Actual Early Vote Spread: Trump +7.55% (+12,424 votes)
Actual Election Day spread: Trump +25.01% (+27,573 votes)
Final results: Trump +13.94% (+39,997 votes). In 2012, Romney was +19K votes (+6.8%)

Hillsborough (See Above)

Pinellas (Clearwater/St. Pete)
Final early vote party spread: 38.5 R, 38.2 D, 23.3 NPA D +752
Actual early vote spread: Clinton +4.58% (+14,460 votes)
Actual Election Day spread: Trump +12.72% (+19,960 votes)
Final results: Trump +1.1% (+5,500 votes). In 2012, Obama won by about 26K votes (+5.5%)

South Florida

Going into Election Day, there was almost nothing that I didn't feel good about in South Florida, and here is why: 87.7% of the entire 2012 election turnout voted early in Dade. In Broward, it was a respectable 81%. In fact, 11.9% of all early votes came in from Dade (should be 10.3%), and Broward was at 9.65% (should have been 8.75%).

And then Election Day happened. The issue here was different than I-4. Trump's share of the two-party vote in Broward and Dade went from 32% to 38.7%, a growth of 6.7%, which while significant, is lower than his statewide average increase of 8.4%. What happened on Election Day is people didn't vote. Statewide, 30.7% of the vote came on Election Day -- in Broward and Dade, it was 23.2%. Another way of looking ai: these two counties made up 21.5% of early vote, and only 14.7 of Election Day

That being said, these two counties both exceeded their projected share of the statewide vote, as well as set records for vote margins. Democrats can not blame losing on Broward and Dade not doing their jobs.

On the flip side, I was concerned about Palm Beach County the entire early vote period. Even in my last memo, I called Palm Beach a "red flag" largely due to lagging turnout. While the Democratic margins were good, Palm Beach was only 5.9% of the statewide early vote, and it should have been 7%. Well it turned out on Election Day -- 41.1% of the total Palm Beach County vote came in on Election Day, making up 9.5% of the total statewide vote, the biggest single jump in the state. And it was a Trump vote that showed up: after running up a 95K vote lead in the early vote, Clinton won Election Day by just over 7K.

When it boils down to it, Clinton won the county by about the same vote margin as Obama in 2012 (which was down from 08), but her vote share was down. Frankly going forward, Palm Beach is a place where Democrats need to up their game.

Palm Beach
Final early vote party spread: 47.3 D, 28.4 R, 24.3 NPA D +74,728
Actual early vote spread: Clinton +24.9% (+94,888 votes)
Actual Election Day spread: Clinton +2.78 (+7,383 votes)
Final results: Clinton +15.1% (+102,271 votes). In 2012, Obama won by just over 102K (+17%).

Broward
Final early vote party spread: 55.4 D, 21.7 R, 22.9 NPA D +212,077
Actual early vote spread: Clinton +41.7% (+254,391 votes)
Actual Election Day spread: Clinton +18.6 (+37,978 votes)
Final results: Clinton +34.9% (+292,369 votes). In 2012, Obama won by 264K votes (+34.9%)

Miami-Dade
Final early vote party spread: 43.9 D, 29.2 R, 26.9 NPA D +114,767
Actual early vote spread: Clinton +34.4% (+234,758 votes)
Actual Election Day spread: Clinton +26.7% (+55,389 votes)
Final results: Clinton +29.4% (+290,147 votes). In 2012, Obama won by 208.5K votes (+23.6%)

#DUUUUVAL

Clinton had one job in Duval, keep it manageable. If you had given the Clinton campaign the option of spotting Trump a 20,000 vote win in Duval in exchange for both campaigns walking away, I would have urged them to take it. After all, this is a county where Bush in 04 won by 61,000 votes, and given that Trump exceeded the Bush 04 margins in most counties, running up a big number here was a real possibility.

But she did her job here, plus some. In keeping Trump's Duval margins under 6,000 votes, she had the best showing in Duval for a Presidential Democratic candidate since Jimmy Carter, and she held Trump well below the Rubio numbers, who won the county by 70,000 votes. If #NeverTrump succeeded anywhere, it was in Duval.

Final early vote party spread: 42.5D, 41.1 R, 16.4 NPA D +4,279
Actual early vote spread: Clinton +1.9% (+5.439 votes)
Actual Election Day spread: Trump +8.9% (+11,407 votes)
Final results: Trump +1.4% (+5,968 votes). In 2012, Romney won by 15K votes (+3.6%)

Final Thoughts

There isn't much more to say -- Clinton had the race where it needed to be, and Trump won it on Election Day.

First, one quick note on the votes before Election Day. Democrats had about 1.5% edge in the voters who had voted either in-person early or a vote by mail ballot, yet she won the early voting period by almost 4%. This was likely do to her over-performing with NPAs, given that non-white voters made up 37% of NPA voters (comapred to 33% of partisans).

I suspect what we will see when the Election Day voter data comes out that white NPA participation was quite high, balancing out the racial make-up of the NPA voter to look more like the electorate at-large.

So where did Trump really win it? The data from the early vote/Election Day totals confirms my first glance: This was a win primarily in suburban/exurban I-4.

Here's why.

Start with my favorite analogy, Florida as a scale. The GOP media market buckets (Pensacola, Panama City, Jacksonville and Fort Myers) and the Dem buckets (Tallahassee, Gainesville, West Palm and Miami) largely balance themselves out, and I-4 tilts it one way or the other. This year, in their core markets, Republicans did much better on Election Day than the Democrats, winning them by 188K votes, compared to the Democrats only winning theirs by 70K, carrying a margin of roughly 120K votes out of their core markets.

However, Democrats went into Election Day with a bigger margin, having crushed the Republicans in early vote, by almost 260K votes. In fact, Clinton's 141K final margin over Trump in the core partisan markets was a few thousand votes higher than Obama in 2012.

Then we get to I-4, and this time, we look at it not as just as the counties on I-4, but every county in the two media markets. Going into Election Day, I-4 was pretty balanced, with Trump holding a 11K vote lead. But on Election Day, Trump won by 242K votes. In other words, 95.5% of Trump's total margin in the Tampa and Orlando media markets came on Election Day. In total, Trump won 59% of the two-party vote in the Tampa and Orlando media markets on Election Day.

And of those 242K votes, 200K of that margin came from the non-urban counties in the media market, in other words. Just on Election Day.

And while it is true that Republicans always do better on Election Day, his Election Day "improvement" particularly in the Tampa media markets far exceeded Romney.

For example, in Pasco, his vote share was 7.69% higher on Election Day than in Early Vote, where as Romney was 2.59% higher, or 5.1% greater than Romney. In Polk, he was also 5.1% higher, Seminole 5.1%, Sarasota 5.4%, and Pinellas 7.2%. We saw similar things in the outlying counties in the Palm Beach market, where in St. Lucie, his vote share was 11.1% higher on Election Day, a 5.2% increase on Romney, and in Martin County, where his Election Day improvement was 6.3% higher than Romney.

I could keep writing on this, but until we get actual voter data from Election Day back, there isn't much else to add. I will do a piece on my thoughts on where the Democrats should go from here sometime in the next few weeks, but as I mentioned in my last piece, the Trump loss, at least regionally, looks a lot like the Bush win in 04 -- and there is a road map for how to reverse it (see Obama).

And again, I don't think it is as simple as Republicans had more voters left to vote, because best case scenario, that number was only about 100,000 more voters. No, this almost surely a cratering of Democratic turnout, all Election Day deciders going to Trump, and an Election Day surge contributing to the comeback.

The combination of two unliked candidates, Trump's success at driving the narrative into the ground, and all of the late-breaking issues going to Trump, it ended up being the perfect storm on November 8th, or in Trump's case, the perfect 3-wood over water to that green 250 yards away.

And I lied in the first sentence -- I'll be back once we have the full voter file with Election Day voters. Until then, Happy Holidays, unless you are a Jags fan,,because we will surely all get a Gus Bradley extension for Christmas.