In my earliest days on the Obama campaign in 2008, one of our first statewide polls showed a weakness with Black voters, at least compared to other states. It wasn't necessarily that McCain was doing better than elsewhere, just that there were more voters on the sidelines. It didn't take long to figure out the initial weakness was among Caribbean voters, which over time, we were able to address.

A couple of days ago, an old Obamaland friend who was a big part of those 2008 Caribbean conversations, texted me a quick question about the Haitian vote in Florida, and specifically if there was any truth to the chatter, and/or anecdotal evidence that Clinton under performed among Haitians. I had sensed some of the same, but honestly hadn't taken a look at the data yet.

Before starting, it is important to consider there are three significant challenges when thinking about the Haitian, and in a larger sense, Caribbean Black vote in Florida.

First, unlike the vast majority of other states, the Black vote in Florida is not monolithically African American. Here, a significant share is either Caribbean and/or Hispanic. The same challenge exists when analyzing the Hispanic vote. On other battleground states, Hispanics ten to be nearly universally Mexican, while here in Florida,both Hispanic and Black voters come from a large mosaic of nationalities.

Secondly, along these same lines, Florida's voter registration data is woefully overly-generic about the population. When it comes to Caribbean and African American voters, the voter registration form provides really just three options: Black, Multi-racial, or Other. Therefore, it is impossible to solely pull out voters of Caribbean descent. There are some analytic tools, but that is generally built on a model, and as such, isn't exact (nor available to the public at large).

Third, and finally, the census data, isn't a ton better. The generic census form does not drill down for information on "Black or African American" residents (it does with certain Hispanics, and Asian populations). There are census tools that dig into nation of origin, but again are sampled and not individual specific.

So in answering my friend's query, I came up with what was a granted, an inexact performance model, yet one I think provides some insight -- and in this case, caution for Democrats -- or at least cause for more research. The model: Florida House District 108, the home of "Little Haiti." The question -- how did Clinton/Trump play both in this district and specifically in the Little Haiti precincts, versus Obama/Romney? For sake of adding more data, I also looked at Rick Scott in 2010 and 2014. Understanding the limitations I laid out above, here is what the data says.

Obama won the district in 2012 by 90-10, and Clinton won it 87-11 (interestingly - this shift matches the 2 point margin shift from Obama to Clinton). Also, voter turnout in the seat at large was about the same, at least among Black voters (70% in 2012, 70.5% in 2016). On the surface, these are not insignificant changes, but in no way are the kind of massive shifts we saw in places like Pasco County, north of Tampa, where the shift among Republican support was almost 10 points.

But looking deeper, there is more than the story. First, there were actually 6,000 fewer registered voters in the district in 16 than 12, which a combination of two things: purges of "inactive voters" and at a certain level, some voters not being interested enough to care to keep registration up to date. As a result, Clinton got 6,000 fewer votes than Obama in the district -- while Trump got about the same as Romney. In other words, Clinton carried the district by 6,000 fewer votes than Obama's 2012 margin. The total shift in the vote margin statewide was roughly 180K votes -- so just over 3% of the total shift from Obama to Trump
happened just in this one state house seat - a seat that by comparison only made up 0.6% of the entire statewide vote in the Presidential election.

Secondly, it gets even more interesting in just the Little Haiti precincts.

So inside House District 108, during the Obama re-election, voters in the Little Haiti precincts made up just over 17% of registered voters, and in the election, just over 16% of the actual 2012 voters. Looking at it another way, turnout among all Black voters in the district was roughly 70% in 2012, but within the Little Haiti precincts, was about 63%.

My guy won Little Haiti by 92% (96-4). Clinton won it by 85% (91-6%). Honestly, this data point actually surprised me. My hunch going in was Trump might have done better in these precincts than he did districtwide (10%). But here is where the huge red flag shows up. Little Haiti residents in 2016 actually made up a bigger share of registered voters than 2016 -- almost 19%, but saw their share of the district's actual vote drop to 16%. Why? Black turnout was right at 71% in the district in 2016, but inside Little Haiti, it fell to 58%.

As a result, Clinton carried these 10 precincts by 1,300 votes less than Obama did, or roughly 0.7% of the total shift from Obama to Trump -- ten precincts that by the way, make up less than one-tenth of one percent of the 2016 statewide vote. Why? Simply, Little Haiti voter participation was 13% lower than Black turnout districtwide. While Trump got better margins than Romney did four years earlier, but it had almost nothing to do with more support for him, and almost everything to do with lower participation from people who in 2012 voted for Barack Obama.

It is interesting when comparing Democratic performance in Little Haiti between 2010 and 2014, Crist did better than Sink, both in terms of turnout and performance. But I suspect, just as we saw overall Black turnout prove to be robust in 14, a lot of that was a factor of voters showing up to protect President Obama. Interestingly enough, Scott put a lot more emphasis on Caribbean voters in 2014 than 2010, so it would be useful to look outside of this one neighborhood to see if the 2014 results hold up elsewhere. Moreover, Crist's 2014 strength in Little Haiti doesn't mean, as 2016 shows, that one can expect 2018 to be the same without work.

Granted, there are lots of reasons to be cautious about reading much of anything into a ten precinct sample of one State House seat in a state like Florida, but I do think there is enough to take a longer look at this, overlaying census data with precinct maps throughout South Florida, and comparing the Presidential election in precincts with significant Caribbean population. My hunch is we would see a lot of the same.