The Washington PostDemocracy Dies in Darkness

Allowing felons to vote likely would have changed the result in Florida’s Senate race

Following Democratic House gains and Republican Senate gains, The Fix's Aaron Blake analyzes the winners and losers from the 2018 midterm elections. (Video: JM Rieger/The Washington Post)

On Tuesday evening, voters in Florida made two decisions. The first was a split: Incumbent Sen. Bill Nelson (D) trails Gov. Rick Scott (R) by less than a point in the contest for Nelson’s seat. The other was a strong majority, over the required 60 percent threshold, in fact: Florida voted to restore the right to vote to felons who had completed their sentences. Prior to the passage of Amendment 4, there was a complicated process required before a former felon could vote. That will no longer be the case.

It raises an interesting question. Had those individuals been able to vote before Tuesday, would it have affected the results in the Nelson-Scott race?

To answer that question, we need to be able to answer a number of other questions. How many felons are affected? How many would have voted? How would they have voted? What follows is our attempt to answer those questions — and the one about who might have been headed to Washington.

As of this writing — Thursday morning — Scott has a lead of about 30,000 votes, according to the Associated Press.

We can get a breakdown of the Senate vote by various demographics, thanks to polls of voters conducted in all 50 states on Tuesday by NORC at the University of Chicago for the AP and Fox News.

Let’s look at how the Scott-Nelson vote broke down by race. White voters preferred Scott by about 18 points, but black voters supported Nelson by 75 points. Hispanic Floridians and those in other groups gave Nelson an edge.

If we break out that first graph by race, it’s clear that much of Scott’s support came from white Floridians, while Nelson’s was a broader mix.

So that’s the baseline. Now let’s assess the number of people who might have been able to vote.

The Brennan Center for Justice at New York University School of Law estimates that there are about 1.4 million people living and working in the state who would gain the right to vote. The specific demographic composition of that population is harder to determine. A staffer for the organization pointed us to center analysis that shows the racial composition of those incarcerated, probably a reasonable proxy for the population on the whole. A separate analysis published by the Prison Policy Initiative has generally similar — but not identical — numbers.

In each case, we’ll note, the density of black people in the population far exceeds the population in the state on the whole. That’s important for our political calculus (and is why we isolated race in the voter poll data above.)

So how many of those felons, if they’d had the right, would actually have voted? Let’s assume that they turn out at the same rate as other Floridians, a perhaps incorrect assumption. How many would have voted on Tuesday?

We can compare two sets of data: The U.S. Elections Project’s estimates of turnout in Florida 2014, 2016 and 2018 with Census Bureau estimates of turnout by race and state in the first two elections. McDonald of the Elections Project would be quick to note that the Census Bureau data has flaws, but broadly we can get a sense for how each group likely voted this year. McDonald’s data has 2018 turnout much higher than 2014 but not quite at 2016 levels. If we use that same ratio for each racial group, we get turnout estimates in the 40 to 55 percent range.

Those Census Bureau figures are a percentage of the citizen population in the state. So let’s combine three things: The size of the population, the estimates of turnout and the likely support for Nelson or Scott indicated in the VoteCast voter polls. That gives us rough estimates of the number of votes each candidate might have received, depending on which estimate of the racial composition of the population we use, PPI’s or Brennan’s.

If we add those to each candidates' existing totals, the race shifts. Nelson now leads by between 1.2 and 1.5 percentage points, or by at least 100,000 votes.

There are innumerable asterisks that apply here, including that a rough application of the VoteCast numbers by race might not be the best estimate. Using the data for white voters without a college degree, perhaps a better population in this case, drops Nelson’s total by 30,000 votes in the Brennan distribution example.

The result, though, shouldn’t be surprising. A shift to the Democrats should be expected, given the density of the black population in Florida’s prison population. In a race as close as the Scott-Nelson contest, even the addition of small groups of voters could have a dramatic effect.

Update: After this article was published, a colleague noted research into the partisan lean and turnout rates of ex-felons in Florida published by Vox earlier this month.

A key point is this one: “One thing that limits the electoral impact of restoring ex-felons’ voting rights is that they turn out at particularly low rates. ... We find that just 16 percent of black and 12 percent of nonblack ex-felons voted. (We defined nonblack as white, Hispanic, Asian, and other due to data limitations.)”

How does that affect the numbers above? If all other figures stay the same, Nelson wins by 4,266 using the PPI racial distribution of the population -- and Scott wins by 166 votes using the Brennan Center distribution.

Only 94,000 to 97,000 new votes are added.