Introduction

In September 2022 two months before the November 2022 midterm elections, Hurricane Ian ravaged Southwest Florida leaving a path of destruction in the region which affected millions of residents. In response to the hurricane, Governor Ron DeSantis issued Executive Order 22-234, which extended early voting in Lee, Charlotte, and Sarasota counties and called for the availability of funds to those counties if requested by the county supervisor of elections and the creation of vote centers where on election day voters could go and cast their ballot in lieu of going to their assigned precinct. However, bordering counties were not included in this albeit the fact that they are likely to have experienced similar hardships as the county line is just an artificial line and oftentimes residents do not know where the county ends and begins. Using a difference in difference model I examine voter turnout in the precincts that were under the purview of the Governor’s executive order and see if they voted at rates comparable to their neighbors in bordering precincts which were not under the executive order and if voters would alter their voting method. This study seeks to build on the work done by Zelin and Smith who examined voter turnout following Hurricane Michael in 2018, finding that turnout was lower in affected counties subject to an executive order by Governor Rick Scott but the privileges granted to the affected counties helped mitigate the hurricane’s effects (Zelin and Smith, 2022). Building on the work Zelin and Smith did, I examine turnout between precincts in these Ian counties that are on the county line of neighboring precincts which did not receive the special treatment. By examining the bordering precints theres a better chance that there will be more similarity between voter since the precinct line is just an artificial line and most of the times people do not know they live on either side of the line but they are likley to face the similar hardships due to the storm however, one gets this new chance to participate in the process because hes covered by the executive order and the other does not. Past work is mixed on whether vote centers like those implemented in the affected countries actually contribute to voter turnout and what other effects they may have mainly due to the dichotomy in the implementation methods and there is also muuch debate as to what gets Voters to turnout or not turnout. Cortino and Rottinghaus examined the impact of a switch to voting centers from precincts in Texas on the voting behavior of ethnic minorities, particularly Latinos in Texas, across election cycles. They argue that an imbalance in voter turnout by race, class, and ethnicity can lead to a reduction in representation, and that changes in voting arrangements may have unintended consequences (Cortina and Rottinghaus, 2023). The results overall show that the displacement of traditional precinct level voting and the increase in distance between polling locations takes a greater toll on voter turnout for voters in rural counties and Latinos, especially in counties that have increased in Latino population (Cortina and Rottinghaus, 2023). The use of voting centers is not always convenient for everyone involved; sometimes they may actually end up being farther than a person’s regular precinct which increases the cost of voting for them. When Governor DeSantis issued the executive order calling for the use of these super voting centers on election day it caused some voters to have a polling place that is farther away than their normal precinct and in that case it made it harder for them to vote making the Governors executive order counter intuitive to some. When looking at how geography can affect turnout, some scholars contend that turnout is affected by the geography and distance voters have to travel to vote, Dyck and Gimpel argued that the political geography of a voters residence affects not only the likelihood that they will vote, but also what modality they will use to vote, such as casting an absentee ballot, mail ballot, or going to an early-voting site as they looked at turnout in Clark County, Nevada while similarly Haspel and Knotts found that that even small changes in the distance from voters home to the precincts can make a big difference in whether voters vote or not it also found that if a polling place is moved, it can change whether people decide to vote as they used GIS to map voters residences in Atlanta to polling places (Dyck and Gimpel, 2005 and Haspel, et al 2005). Furthermore, there is always the potential impact of precipitation which can effect voters going to the polls, albeit the fact that it may not have be as severe as a hurricane. In a study done by Gomez, Hanford and Krause they found that rain significantly reduced voter participation by just less than 1% per inch, while an inch of snowfall decreases turnout by almost 0.5% (Gomez, et al 2007). This decrease in turnout mainly helped Republicans as the study examined weather conditions in 14 US Presidential elections with a noticeable effect in the 1960 and 2000 elections. Fraga and Hersh also found that rain mainly impacted competitive elections however they found that the campaigns not necessarily election officials find ways to assuage the effects of precipitations such as in 2008 when Senator Hilary Clinton’s presidential campaign bought shovels for every precinct and offered to shovel driveways for supporters in Iowa during a bout of heavy snow (Fraga and Hersch, 2010). With campaigns and candidates and parties addressing these barriers instead of the actual election administrators it raises questions about the role and responsibilities of election administrators Moreover, Harder and Krosnick posited that the likelihood of voting can be influenced by three factors: the motivation to vote, ability to vote and the difficulty of voting (Harder, et al,. 2008). This can also include actions by state actors interfering and making the process more difficult to register and vote as more and more states tighten their election and voting laws.

Methods

Utilizing a difference and difference approach I test whether voters registered in precints that were under the purview of the Governors executive order voted at rates comparable to bordering precints that were not given the same privileges but are likely to have experienced similar hardships. In order to build off prior work such as (Zelin and Smith 2020) who examined counties I take an additional step in examining precincts on the county border. Lee, Charlotte, and Sarasota were the counties under the purview of the Governors executive order and the counties that border these three are Collier, Hendry, Glades, Highlands, Desoto and Manatee. They will be known as Ian and Non-Ian counties for the purpose of this study.

Joining the Data

In order to determine who registered to vote on Election Day, Re Cap files for the 2018 and 2022 November elections in Florida for the three counties covered by the order Lee, Charlotte, and Sarasota along with the counties that border the affected counties Collier, Hendry, Glades, Highlands, Desoto and Manatee will be used this file similar to a regular voter file but it has the method of voting used and additional information. I also collected precinct information for the counties in questions, identifying the precincts that border neighboring precincts which were under the Governors executive order.

mergedIan <- inner_join(Recap2018_Ian, Recap2022_Ian, by = c("VoterID")) # suppressWarnings()

I use an inner join in order to merge the two voter files (Recap2018_Ian and Recap2022_Ian) based on the “VoterID” column creating a new dataframe called mergedIan.

precinctcomplete <- precinctcomplete %>% 
    mutate(Precinct = as.numeric(Precinct))

mergedIan_precinct <- mergedIan %>%
  left_join(precinctcomplete, by = c("CountyCode", "Precinct"))

mergedIan_precinct <- mergedIan_precinct %>% 
  select(IanCounty, PrecinctIdentifier, CountyCode, VoterID, ResidenceCity.x, Gender.x, Race.x, RegistrationDate.x, PartyAffiliation.x, Precinct, Precinct.x, VoterStatus.x, Military.x, MilitaryDependent.x, Overseas.x, VoteHistoryCode.x, VoteHistoryCode.y, Agecat2.x, Age.x, VotingAssistance.x, reg.year, CountyCode.x)

Here I bring in the precinct level data which I collected that identify the precints that border the Ian precinsts which were under the purview of the Governors executive order and join them with the merged voter file also with some data cleaning to reduce the number of colimns in the voter file by just bringing in the ones that are necessary for analysis

### County Ian nonIan analysis of turnout in 2018 and 2022

turnout2018Ian <- mergedIan_precinct %>%
  filter(IanCounty == 1) %>% 
  count(IanCounty, VoteHistoryCode.x) %>% 
  mutate(Percentage = n / sum(n) * 100) 

turnout2018nonIan <- mergedIan_precinct %>%
  filter(IanCounty == 0) %>% 
  count(IanCounty, VoteHistoryCode.x) %>% 
  mutate(Percentage = n / sum(n) * 100) 

turnout2022Ian <- mergedIan_precinct %>%
  filter(IanCounty == 1) %>% 
  count(IanCounty, VoteHistoryCode.y) %>% 
  mutate(Percentage = n / sum(n) * 100) 

turnout2022nonIan <- mergedIan_precinct %>%
  filter(IanCounty == 0) %>% 
  count(IanCounty, VoteHistoryCode.y) %>% 
  mutate(Percentage = n / sum(n) * 100) 

# wow, turnout in non Ian in 2018 was slightly higher than in Ian counties; much higher in 2022 after the hurricane.
# Did the vote centers in the 3 Ian Counties fail to assuage the damage of the hurricane? maybe not.


# narrowing down the analysis to look only at adjacent precincts in Ian and non Ian

In this code the voter turnout data from both Ian and non-Ian counties are brought in for the elections in 2018 and 2022 along with the precinct and county identifiers in order to calculate the voter turnout among the counties that were under the purview of the governors executive order known as Ian Counties and those bordering counties which were not under the order are called non-Ian counties.

#context of who lives along the borders. also, why look at bording precincts?
table(mergedIan_precinct$PrecinctIdentifier, mergedIan_precinct$CountyCode)
##    
##        CHA    CLL    DES    GLA    HEN    HIG    LEE    MAN    SAR
##   0  77325 159451   6041   2566   7058  17674 340794 173696 226155
##   1  38961  32387   7910   3196  11621  33154  58119  39335  50142

This table helpos give context of why to look at the bordering precincts with 0 denoting the precints that are not on the border and 1 denoting the precisnts that are on the border, looking at the table it shows that there are large swaths of the population that live on these bordering precints, more people living in DeSoto County are on bordering precints which means a large portion of the population is being affected by these election administration decisions.

This table gives an overview of the voter turnout in the counties that received the special treatment and the ones that did not in the 2018 and 2022 elections. The results show that voter turnout in the counties that had the special rules in place which allowed the use of super voting centers which allowed voters to cast ballots on election day at any of these sites was actually lower when compared to the counties that did not have these super voting centers in both the 2018 and 2022 elections.

precinctturnout2018Ian <- mergedIan_precinct %>%
  filter(IanCounty == 1) %>% 
  filter(PrecinctIdentifier == 1) %>% 
  count(IanCounty, VoteHistoryCode.x) %>% 
  mutate(Percentage = n / sum(n) * 100) 

precicntturnout2018nonIan <- mergedIan_precinct %>%
  filter(PrecinctIdentifier == 1) %>% 
  filter(IanCounty == 0) %>% 
  count(IanCounty, VoteHistoryCode.x) %>% 
  mutate(Percentage = n / sum(n) * 100) 


precinctturnout2022Ian <- mergedIan_precinct %>%
  filter(IanCounty == 1) %>% 
  filter(PrecinctIdentifier == 1) %>% 
  count(IanCounty, VoteHistoryCode.y) %>% 
  mutate(Percentage = n / sum(n) * 100) 

precinctturnout2022nonIan <- mergedIan_precinct %>%
  filter(IanCounty == 0) %>% 
  filter(PrecinctIdentifier == 1) %>% 
  count(IanCounty, VoteHistoryCode.y) %>% 
  mutate(Percentage = n / sum(n) * 100)

In this series of code the turnout among Ian precincts and non-Ian precincts is being calculated from the data set where I joined the precinct identifier data I collected with the vote history code in the Recap vote history files. The percentages are calculated within each unique combination of identifiers with Ian County identifiers a 0 denotes counties that were not under the governors executive order and did not have super voting centers and 1 denotes the counties that were under the executive order. Likewise for precint identifers a 0 denotes precincts that are not on the county border and a 1 denotes precincts that are on the county border

###now race

# Black Turnout 2018 Ian Precints 
blackprecinctturnout2018Ian <- mergedIan_precinct %>%
  filter(IanCounty == 1) %>% 
  filter(PrecinctIdentifier == 1) %>% 
  filter(Race.x == "Black") %>% 
  count(IanCounty, VoteHistoryCode.x) %>% 
  mutate(Percentage = n / sum(n) * 100) 

# Black Turnout 2018 non Ian Precints 
blackprecinctturnout2018nonIan <- mergedIan_precinct %>%
  filter(IanCounty == 0) %>% 
  filter(PrecinctIdentifier == 1) %>% 
  filter(Race.x == "Black") %>% 
  count(IanCounty, VoteHistoryCode.x) %>% 
  mutate(Percentage = n / sum(n) * 100) 

# Black turnout 2022 Ian Precints 

blackprecinctturnout2022Ian <- mergedIan_precinct %>%
  filter(IanCounty == 1) %>% 
  filter(PrecinctIdentifier == 1) %>% 
  filter(Race.x == "Black") %>% 
  count(IanCounty, VoteHistoryCode.y) %>% 
  mutate(Percentage = n / sum(n) * 100) 

# Black turnout 2022 non Ian Precints 
blackprecinctturnout2022nonIan <- mergedIan_precinct %>%
  filter(IanCounty == 0) %>% 
  filter(PrecinctIdentifier == 1) %>% 
  filter(Race.x == "Black") %>% 
  count(IanCounty, VoteHistoryCode.y) %>% 
  mutate(Percentage = n / sum(n) * 100)

Further research can be done by delving into turnout among African Americans and see if the executive order helped mitigate the effects of the hurricane for them. In this series of code I filter by race to measure the turnout by African Americans in 2018 and 2022 in both Ian counties and non-Ian counties

- The table shows that African American Turnout in the Ian counties was lower than in the non-Ian counties. It also shows that in both Ian and Non-Ian counties African American turnout fell significantly in 2022 which could mean that actions taken by state officials did not do enough to assist these African American voters

Visualize the Data

- Turnout in 2022 among these Ian precincts was 56% with less than 10% voting on election day. Turnout in 2022 among non-Ian precincts was 59.6% and 20% voting on election day. Turnout was higher in non-Ian precincts compared to the Ian precincts
Turnout in 2018 among Ian precincts was 65% and 20% voting on election day. Turnout in 2018 among non-Ian precincts was 69% with 25% voting on election day. Turnout in 2018 was higher in these non-Ian precints compared to the Ian precints which were under the Governors executive order in 2022 The data from the line graph shows that that turnout was lower for both 2018 and 2022 in the Ian precincts which had these super voting centers and under the order of the Governors executive order

Analysis and Conclusion

The findings show that turnout in 2022 was lower among those precincts that received the special treatment by the Governors order compared to the bordering ones that did not however the turnout was consistent with the previous 2018 midterm election with non-Ian counties maintaining a 4-5 point edge in turnout compared to Ian Counties. Hurricane Ian caused catastrophic destruction in the region and affected the electoral process notably in person election day voting decreased in Ian counties and was as low as 9% while absentee and early voting increased which can show that the relief provided by the state helped assuage the effects of the hurricane. Among African Americans turnout dipped to low levels compared to the 2018 election where in the Ian and non Ian counties they turned out at rates upwards of 60%. This turnout could be attributed to the presence and popularity of Democratic nominee for Governor Andrew Gillum or the amendment issues on the ballot to restore felons voting rights in Florida. Overall these results are consistent with Zelin and Smith 2022 who found that turnout in counties affected by Hurricane Michael was lower than those counties that were not but the effects were mitigated by these actions by the state.

Citations

Cortina, Jeronimo, and Brandon Rottinghaus. ““The quiet revolution”: convenience voting, vote centers, and turnout in Texas elections.” Politics, Groups, and Identities 11, no. 2 (2023): 246-266.

Dyck, J.J. and Gimpel, J.G. (2005), Distance, Turnout, and the Convenience of Voting. Social Science Quarterly, 86: 531-548. https://doi.org/10.1111/j.0038-4941.2005.00316.x

Fraga, Bernard L., and Hersh, Eitan. 2010. “Voting Costs and Voter Turnout in Competitive Elections.” APSA 2010 Annual Meeting Paper. Available at SSRN

Gomez, Brad T., Thomas G. Hansford, and George A. Krause. 2007. “The Republicans Should Pray for Rain: Weather, Turnout, and Voting in U.S. Presidential Elections.” The Journal of Politics 69, no. 3. Cambridge University Press.

Harder, Joshua, and Jon A. Krosnick. 2008. “Why Do People Vote? A Psychological Analysis of the Causes of Voter Turnout.” Journal of Social Issues 64, no. 3. Wiley-Blackwell. DOI: 10.1111/j.1540-4560.2008.00576.x12.

Haspel, Moshe, and H. Gibbs Knotts. 2005. “Location, Location, Location: Precinct Placement and the Costs of Voting.” The Journal of Politics 67, no. 2: 560-573.

Zelin, W. A., & Smith, D. A. (2023). Weather to Vote: How Natural Disasters Shape Turnout Decisions. Political Research Quarterly, 76(2), 553-564. https://doi.org/10.1177/10659129221093386