Did Ballot-Design Oust a US Senator? A Study of the 2018 Election in Florida

• In Event: Evidence about Voter Suppression, Voter Fraud, and Administrative Discretion

Abstract

In 2018, Florida’s two-term Republican Governor, Rick Scott, defeated Bill Nelson, a two-term Democratic incumbent United States Senator, by a mere 10,033 votes out of over 8,100,000 votes cast statewide. Reminiscent of the 2000 Presidential Election in Florida (Wand et al. APSR 2001) and the 2006 contest in Florida's 13th Congressional District (Frisina et al. ELJ 2008), one county’s ballot design may have changed the outcome of the Scott versus Nelson race by leading a sufficient number of voters to undervote in it. In heavily Democratic Broward County, and unlike in other Florida counties, the U.S. Senate and U.S. House races appeared at the bottom of the left column of the county ballot, under a long set of instructions printed in three languages. The location of these races on the ballot apparently caused many voters in Broward to miss them. Indeed, 24,975 more votes were cast in Broward County for Florida governor than for U.S. Senator, raising the question of whether the Broward ballot design and the undervotes it generated changed the outcome of the Senate election.



In this paper, we will offer two methodologies to address this possibility. In the first, we will collect county-level (or precinct-level data, if available) from all Florida counties, and using regression models, we will determine how anomalous the undervote rate in the 2018 gubernatorial and Senate elections was in Broward County (or its precincts) and how many additional votes Nelson could have expected in this county, given votes received by other statewide candidates and the demographic (age, race, sex, and party registration) characteristics of voters in those counties (or precincts).



In the second approach, we will obtain ballot-images from Broward County (directly from the county or through another source) that will allow us to estimate how many additional votes Nelson and Scott should have each expected to receive in Broward County among voters who undervoted in that race, conditional on individual-level votes for the other offices on the ballot.



Thus, this paper will address the possibility that a ballot design unintentionally affected the outcome of a heavily-contested U.S. Senate election, and through its analysis will demonstrate the importance of administrative aspects of elections. We will also show whether and how closely different methodologies converge in their estimates of that effect.

Authors

• Michael C. Herron, Dartmouth College

• Michael D. Martinez, University of Florida

• Michael P. McDonald, University of Florida