Brian Amos, Daniel A. Smith, and Casey Ste. Claire
University of Florida
Despite the expansion of convenience voting across the American states, millions of voters continue to cast ballots at their local precincts on Election Day. We argue that those registered voters who are reassigned to a different Election Day polling place prior to an election are less likely to turn out to vote than those assigned to vote at the same precinct location, as a new precinct location incurs both search and transportation costs on reassigned voters. Utilizing voter file data and precinct shape files from Manatee County, Florida, from before and after the 2014 General Election, we demonstrate that the redrawing of precinct boundaries and the designation of Election Day polling places is not a purely technical matter for local election administrators, but may affect voter turnout of some registered voters more than others. Controlling for a host of demographic, partisan, vote history, and geospatial factors, we find significantly lower turnout among registered voters who were reassigned to a new Election Day precinct compared to those who were not, an effect not equally offset by those voters turning to other available modes of voting (either early in-person or absentee). All else equal, we find that registered Hispanic voters were significantly more likely to abstain from voting as a result of being reassigned than any other racial group.
Here’s the link to the pre-publication draft.
What should be the denominator when calculating voter turnout? This is an ongoing discussion, usually with respect to Voting Age Population or Voting Eligible Population, the latter construct developed by my UF colleague, Mike McDonald.
But what should be the denominator when drawing on official voter files?
The Florida Division of Elections has provided two different measures for the 2016 Presidential Preference Primary.
The following is from the FDOS:
- Turnout for PPP Election Only (This information is also provided on the homepage of the Florida Election Watch website)
As of 3/27/2016 12:00 pm, the voter turnout percentage for the subset of active registered voters statewide eligible to vote in the Presidential Preference Primary race was 46.23%. The subset percentage consists of 4,089,516 votes cast out of the total number of active registered Republicans and Democrats (8,845,892) as of the registration deadline.
This sounds reasonable. The total votes cast by Republicans and Democrats divided by the total number of ACTIVE registered voters registered with the two major parties.
But it’s not really the total turnout in the March 15, 2016 election. Many counties and municipalities had local elections with nonpartisan contests or ballot issues; thousands of registered voters who belong to a third party, or who don’t belong to any party, turned out to vote.
So, FDOS also has offered this turnout figure:
- Total turnout on PPP Election Day (including municipal/county level races)
The total voter turnout percentage, including nonpartisan municipal/county level races, was 34.53% (4,164,001) compared against the total number of active registered voters in the state (12,060,748) as of the registration deadline.
That’s a huuuuge difference in turnout.
And turnout rates would be even lower if the denominator included all of Florida’s registrants, including those who have INACTIVE status. FDOS (and the county SOEs) don’t include these legally registered voters in their denominators, even though thousands of them turned out in the election.
I’m not sure what’s the correct way of determining turnout. But it matters with respect to our perception of the level of enthusiasm in the Florida electorate.
I’ve had a very productive conversation with the Florida Division of Elections regarding my earlier post, below.
Statutorily (Section 98.0981, F.S.), Florida counties are not required to compile and upload their complete voter history for the March 31, 2016 PPP until April 28, 2016. As such, the complete PPP individual-level voter histories that the counties provide to the FVRS will not be reflected until the April 30, 2016 statewide voter extract. Of course, most of the counties did voluntarily report their PPP data to the FVRS in time to be captured in the March 31, 2016 voter history file. A handful of counties, though, voluntarily uploaded only partial voter histories for the PPP, which I document below.
The month-long delay by some counties (even though permitted statutorily) in uploading their voter histories seems to be somewhat problematic. It can lead to an asynchronized archiving of official voting state data.
For example, in counties where SOEs immediately report individuals’ vote histories in their counties, a contemporaneous snapshot of turnout is possible. On the other hand, in counties where SOEs legally delay their reporting, it is possible that the vote histories of registered voters who turned out might be mis-assigned. If a registered voter who cast a ballot in one county in the PPP moves to another county and re-registers in that county, the PPP vote history that is uploaded to the FVRS by the voter’s previous county will appear to have been cast in the voter’s new county, not the actual county in which the PPP was cast.
I’ll have more to say on the issue later.
But I want to reemphasize that my post last night was intended to be a cautionary tale to those who utilize these data, not an indictment of the FDOS of the SOEs or Secretary Detzner.
There are some problems with the March 31, 2016 statewide Voter History File not syncing with the official vote totals. Nothing serious, I hope. But nothing new, either…
I was going to start cranking out a simple analysis of who turned out to vote in the March 15, 2016 Presidential Preference Primary, but it looks like some county Supervisors of Elections did not successfully upload their official results, which were due to the Division no later than noon on Sunday, March 27, 2016.
According to an announcement issued by the Florida Division of Elections, voter turnout in the 2016 Presidential Preference Primary (as of 3/27/2016) was 46.23%. The Division reports that 4,089,516 votes were cast (out of the total number of active registered Republicans and Democrats (8,845,892) as of the registration deadline).
This official vote is slightly different from the FDOE’s “2016 Presidential Preference Primary March 15, 2016, Official Election Results” website, which reports that the total votes cast was 4,164,001. There’s no explanation from FDOE why the totals are not the same, but I can certainly understand why they may differ. Really, why quibble over a couple (OK, several) thousand votes cast in Florida?
What is more disconcerting, however, is that these two totals are far greater than what is recorded in the March 31, 2016 Voter History File. The Voter History File reports the method of voting of all those who who cast ballots in the election. A quick sum finds only 3,408,945 ballots appear to have been cast by individual voters.
What accounts for the 755,056 missing from in Voter History File (if we compare it to the “Official Election Results” website)?
A quick analysis reveals that several counties have not properly uploaded their individual vote histories to the statewide Florida Voter History File, maintained by the Bureau of Voter Registration Services (BVRS).
Here’s a quick and dirty scatter-plot of the total votes cast in the PPP 2016, with the Voter History File on Y-Axis and the “Official Election Results” on the X-Axis.
As is pretty clear, 7 counties have failed to synchronize their own individual-level records with the BVRS’s statewide Voter History File: Collier, Sarasota, Polk, Brevard, Hillsborough, Palm Beach, and Broward.The discrepancies are not small. Brevard, Broward, Hillsborough, and Polk counties report precisely zero (0) votes were cast on Election Day. Other counties, too, failed to sync their own records with the BVRS’s Vote History File, but they’re too small to see on the scatter-plot. These additional counties with zero (0) votes cast include Desoto, Franklin, Glades, Hardee, Hernando, Highlands, Layfayette, and Taylor. Other counties, too, have obvious problems: Collier reports one (1) Election Day vote was cast; Palm Beach reports 126 votes were cast on Election Day; Sarasota, 48 cast on Election Day.
Clearly, there’s a syncing issue with these counties.
Data uploaded by the counties and synced with the BVRS database by many other counties looks pretty clean. In Miami-Dade County, for example, the totals only differ by 5,674 votes (compared to the 134,045 missing individual-level votes in neighboring Broward). Orange county’s total is only off by 2,338 votes cast. Hamilton County’s total votes cast are spot-on: 2,451 and 2,451. Pasco County, whose SOE Brian Corley heads the Florida Association of Supervisors of Elections (FSASE), recorded 103,987 votes cast in the Voter History File, and 103,722 on the Official Election Results webpage, a difference of only 265 votes cast.
To be sure, this is not the first time I’ve discovered these kinds of discrepancies with the official vote tally and what is reported in the BVRS’s voter files. For an article (see footnote 8) on early voting patterns in Florida that I wrote with Michael Herron (as well as several other articles), we found major discrepancies in the December 31, 2012 Vote History File. After discussing the issue with the counties that had syncing problems, it was finally corrected in the March 31, 2013 file.
I’m sure FDOE and the county SOEs will get to the bottom of this in due time. But until then, data-hounds should be very weary about using the statewide Vote History File for any analyses.
Which is good, as this data-hound has some papers to revise and final papers to grade!
Again, the math doesn’t add up. CNN’s exit poll of Florida voters reports that 16% self-reported Republicans who voted in the Florida PPP as being “latino” [sic].
We know that prior to Election Day, of the more than 1.2m registered Republicans who had already voted, more than 86% self-identified as “white” when they registered to vote. Only 10.3% marked on their voter registration cards that they were “Hispanic.”
It’s stretches the imagination, then, that one in five of the 1.16m Republicans who voted on Election Day (some 254k) were Hispanic voters. Sure, some 192k Republican Hispanics didn’t vote early in Miami-Dade, but chances are, a few of them also stayed home on Election Day. Indeed, there were less than 400k Republican Hispanics statewide who had yet to vote on Election Day. Nearly every one of them would have had to have voted on Election Day in order for the CNN exit poll figure for Hispanic turnout to map out.
With such dubious baseline figures, I’d throw caution to the wind for anyone digging any deeper into the CNN exit poll crosstabs in Florida. One wonders how far off the exit polls are in the other states that have had primaries?
I will be digging into this some more as time permits. Although I can’t find a link to the exit poll methodology or how weighting was done, I’m assuming CNN drew its sample of 1907 Republicans not only from Election Day voters, but sampled early in-person and called absentee voters who cast ballots ahead of the March 15, 2016 PPP in Florida.
But some quick observations of the marginals…
First, and most glaringly, I am hard-pressed to believe that only 39% of Republican respondents were 60+. Sure, Election Day voters tend to be younger than convenience voters (early in-person and absentee mail), but my analysis of the statewide voter file and absentee and early in-person voting indicates that of the nearly 1.2m Republicans who voted prior to Election Day, 63.4% were 60+.
So, some simple arithmetic: A total of nearly 2.36m votes were cast by Republicans in the PPP; less than half of the total (roughly 1.16m) were cast on Election Day. If CNN’s exit poll is accurate, that 39% of Republican voters were 60+, it would mean that 919.9k of the 2.36m Republican voters were in this group of older voters. But roughly 756.9k Republicans 60+ had already voted early (in-person and absentee), which leaves only 163k Republicans over 60 to show up on Election Day. That would mean that only 1 in 8 of the 1.64m Republican voters who showed up on Election Day were 60+.
That’s just not credible.
This is not the first time I’ve found problems with CNN’s exit polls; its 2014 General Election exit polling breakdown for the age of voters was also way off.
But we’ll know for sure about the accuracy of CNN’s 2016 PPP exit polls in Florida next month.
Floridians wanting to participate in the March 15, 2016 presidential preference primary had until February 16 to either register to vote or change their party registration to Democrat or Republican in order to vote in either closed primary.
Between February 1 and February 16, more than 36k Floridians became newly registered voters during the final run-up to the registration deadline. Roughly one-third of them cast ballots prior to Election Day. Some 5.4k newly registered Democratic voted (3.4k cast early in-person ballots and 2.0k mailed in their absentee ballots), and 6.1k newly registered Republicans voted (3.7k early in-person and 2.4k absentee ballots). A smattering of newly registered NPAs and 3rd party registrants also voted before Election Day.
We won’t know until next month if the 24k other newly registered voters waited to cast ballots on Election Day. We’ll never know (because the state of Florida doesn’t track this) if those who voted, or those who waited until Election Day or didn’t vote at all, opted-in to register at DMV offices or other state or federal agencies that are required to ask voters if they’d like to register. My Ph.D. graduate student, Lia Merivaki, is looking into questions like these in her dissertation. It will be interesting to see if the ongoing implementation of online voter registration across many states (and in Florida, in 2017), or automatic registration (which is already in effect in Oregon and will be in California), will lead to these new registrants voting, or deciding not to exercise their franchise.
If it’s the later, turnout rates will likely take a tip dip due to an inflated denominator of registered voters who didn’t opt when applying for their driver’s license or other services.
As of this morning, some 1.269m absentee ballots have been counted in the Florida election. The number will continue to trickle up in the coming days as county canvassing boards examine the overseas mail ballots coming in over the next week.
Republicans cast 719.1k absentee ballots. Thus far, .56% of them were rejected by canvassing boards due to voter error (likely a mismatched or invalid signature); another .27% were rejected because they lacked an accompanying signature on the return envelope. So, nearly 99.2% of GOP absentee ballots were valid.
On the Democratic side, nearly 524.8k were received by county SOEs. Slightly less than 99% were processed as valid (98.97%, to be exact). Of the rejected Democratic mail ballots, .61% had a voter error and .42% lacked a signature on the envelope.
Only 21.7k NPAs cast ballots, and roughly 10k more by 3rd party registrants.
More analysis, as time permits, on the partisan/racial/ethnic/age breakdown of absentee ballots as time permits…
Here are the total ballots cast by Republican Hispanics and the percentages of registered Republican Hispanic voters who cast ballots, as of the start of Election Day voting this morning. I’ve sorted by the top 10 counties in terms of percentages. Not surprisingly, Miami-Dade (DAD), has the highest overall vote total and percentage of Republican Hispanic turnout, perhaps an indication of Marco Rubio’s pull in his home county. We’ll see later tonight what the tally is for late arriving absentee ballots as well as Election Day voting.
|County||Hispanic Republican votes cast||Tot Registered Hispanic Republicans||% Republican Hispanic Turnout|
When precincts opened their doors this morning at 7am, more than 2.09m Floridians had already cast their ballots.
Nearly 1.17m out of the 4.56m active and inactive registered Republicans in the state had cast early in-person or absentee ballots ahead of Election Day, or 25.6% of all registered Republicans.
Nearly 865k out of the 5.04m active and inactive registered Democrats in the state had cast early in-person or absentee ballots ahead of Election Day, or 17.2% of all registered Democrats.
An additional 52.8k NPAs and 3rd party registrants cast ballots (but not for the GOP or Democratic nominees), as did another 3.6k (out of the 28.9k) who registered to vote between February 1 and February 16, the last day to register to vote in the presidential preference primary election.
Older voters in both parties came out in force. 739.3k (63%) of all Republican ballots cast prior to Election Day were cast by voters 60 and older. 514.5k (59%) of all Democratic ballots cast prior to Election Day were cast by voters 60 and older.Only 50.3k Republican and 59.8k Democrats under the age of 30 voted during the early voting period or by absentee ballot.
Sliced differently, only 8.7% of all registered Republicans under 30 and only 7% of all registered Democrats under 30 have voted prior to Election Day, whereas nearly 40% of all registered Republicans over 60 and 29% of all registered Democrats over 60 cast ballots ahead of today, March 15, 2016.
What about the partisan breakdown along racial/ethnic lines?
Slightly more than 1m white Republicans have cast ballots, nearly ten times the 119.5k Hispanic Republicans who voted early in-person and absentee.
Among Democratic early and absentee voters, 547k are white,86.9k are Hispanic, and 191.5k are black.