Transfer of Partisanship on Voting Behavior
Maddie Wagner
12/6/2019
Motivation and Background
The election landscape has shifted rapidly over the past few election cycles. One element of this is the increasing role that the Hispanic population plays in American politics. According to the Pew Research center, Hispanics have become “the largest racial or ethnic minority group in the electorate,” for the 2020 presidential election, putting them ahead of African and Asian Americans (Cilluffo and Fry, 2019). This has large implications for the future of the US political climate, with Hispanics yielding a substantial amount of power over the election cycle (Bell, 2016). It is therefore important to know how Hispanics tend to vote, and whether or not the voting behavior expressed by one generation of Hispanic voters is the same as their children and grandchildren. For this project, I would like to analyze the voter registration files of Miami-Dade county, Florida. I have chosen Miami-Dade county as it has a large Hispanic population (65%), the largest subset of which is Cubans (34.3%), (Census Bureau, 2010). Cubans are an interesting case, as unlike other Hispanic ethnic groups, they tend to register as Republican. First generation Cuban immigrants who came to the US in the 50s and 60s may have been fleeing Fidel Castro’s rise to power and the subsequent communist takeover in their county, which has influenced their right-wing voter choices (Bishin and Klofstad, 2011). As Cubans make up such a large portion of the Miami-Dade voter file, their turnout can have an impact on the outcome of election cycles. What would be interesting then to analyze is how strong the transfer of partisanship is between different generations of Cuban Americans. Young people as a whole are registering in increasing numbers an independents or democrats. If this trend holds up in the Cuban community as well, this can lead to major implications in political atmosphere of Miami-Dade county
Objectives
- To determine if there is a relationship between country of origin and voting behavior for Cuban Americans, and if so, is this partisanship transferable to newer generations who have grown up in the United States
- Determine if this transfer of partisanship is stronger for Cubans than other Hispanic Ethnic groups (for the purposes of this project, I will be comparing them to Puerto Ricans)
- Determine if this transfer of partisanship is stronger for Cuban households than the rest of Miami-Dade County
- Determine if one political party has a higher transfer of partisanship than others, comparing these rates with country of origin
Hypothesis
- Cuban households will have an overall stronger rate of partisan transfer than the rest of Miami-Dade (higher homogeneity)
- The Latino households I will be studying, Cuban, Puerto Rican (PR), and Cuban-Puerto Rican (PRCuba), will have a stronger transfer of partisanship than the rest of Miami Dade (“Other”)
- For Cuban households, the rate of transfer for the Repubican Party may be lower than expected due to younger voters moving towards Non-Party Affliation and the Democratic Party
Data Source
- Miami-Dade County Voter File
- Total voters: 114,836
- Total households: 35,471
- Cuban: 8,884
- Puerto Rican: 777
- Cuban-Puerto Rican: 172
- Others: 25,638
- Note: The entire Miami-Dade voter file consists of roughly 1.5 million voters. For this project, I will only be looking at 3-7 person households in which everyone in the household has the same last name. This leads to the number of voters you see above.
Methodology
- Isolate the addresses with the same last names – ensures familial relations
- Separate Cuban, PR, and PRCuban households from the others
- Code each household according to different categorizations (tests). There are 3 I will be using for this project:
- Homogeneous vs. Heterogeneous (0,1)
- Full vs. Partial vs. No Partisan Transfer (1,2,3)
- Party/Age by Household (1-15)
- Visualize on maps
Test 1 - Homogeneous vs. Heterogeneous Households
To begin, open Tidyverse and read in my files from excel (there are 4 total)
library(tidyverse)## -- Attaching packages -------------------------------------------------------------------------------- tidyverse 1.2.1 --## v ggplot2 3.2.1 v purrr 0.3.2
## v tibble 2.1.3 v dplyr 0.8.3
## v tidyr 0.8.3 v stringr 1.4.0
## v readr 1.3.1 v forcats 0.4.0## -- Conflicts ----------------------------------------------------------------------------------- tidyverse_conflicts() --
## x dplyr::filter() masks stats::filter()
## x dplyr::lag() masks stats::lag()library(readxl)
DAD_Cuba <- read_excel("DAD_Other.xlsx")## Warning in read_fun(path = enc2native(normalizePath(path)), sheet_i =
## sheet, : Expecting numeric in R19485 / R19485C18: got 'k'library(readxl)
DAD_PR <- read_excel("DAD_PR.xlsx")
library(readxl)
DAD_PRCuba <- read_excel("DAD_PRCuba.xlsx")
library(readxl)
DAD_Other <- read_excel("DAD_Other.xlsx")## Warning in read_fun(path = enc2native(normalizePath(path)), sheet_i =
## sheet, : Expecting numeric in R19485 / R19485C18: got 'k'Next, I used the “count” to get a count of the total 0’s and 1’s I’ve assigned to each household for each country of origin. * 0’s are homogeneous households, meaning everyone in the household is registered to the same party * 1’s are households in which at least one member is not registered to the same party * This test is used to determine the level of homogeneity among the households of each country of origin for comparison purposes
count(DAD_Cuba, Cat1)## # A tibble: 3 x 2
## Cat1 n
## <dbl> <int>
## 1 0 11692
## 2 1 13946
## 3 NA 57566count(DAD_Other, Cat1)## # A tibble: 3 x 2
## Cat1 n
## <dbl> <int>
## 1 0 11692
## 2 1 13946
## 3 NA 57566count(DAD_PR, Cat1)## # A tibble: 3 x 2
## Cat1 n
## <dbl> <int>
## 1 0 296
## 2 1 481
## 3 NA 1743count(DAD_PRCuba, Cat1)## # A tibble: 3 x 2
## Cat1 n
## <dbl> <int>
## 1 0 77
## 2 1 95
## 3 NA 384Stray Note: I assigned these numbers in excel. Since I only assigned one number per household, the NA’s come from the other two or more members that didn’t have a number assigned to them. I’m going by household, not by person. So, for a three person homogeneous household, they would have a 0, followed by two NA’s.
Continuing on - I calculated the proportions of 0’s and 1’s and wrote them in excel before reading the file into R.
library(readxl)
DAD_Cat1 <- read_excel("DAD_Cat1.xlsx")Here’s the table:
*
Graph:
ggplot(data=DAD_Cat1, aes(x = Country_of_Origin, y = Proportion, fill = Cat1)) + geom_col(position = "dodge")
Takeaways
- All households (Other, Cuban, Puerto Rican, Cuban-Puerto Rican) have a higher proportion of heterogeneous households than homogeneous
- Puerto Rican households have the highest difference in proportions – 23.8% more heterogeneous households than homogeneous
- The others have the lowest difference in proportions – 8.8%
- Cuban and Cuban-Puerto Rican proportions are very similar – 11.4% and 10.4%, respectively
- Conclusion: Cuban households are more homogeneous than Puerto Rican households, but less so for Cuban-Puerto Rican households and others
Chi-Squared Test
cat1cs <- matrix(c(11691,13947, 3935,4948, 296,481, 77,95), byrow=F, ncol=4)
colnames(cat1cs) <- c("Other","Cuba","PR","PRCuba")
rownames(cat1cs) <- c("Homogeneous", "Heterogeneous")
cat1cs## Other Cuba PR PRCuba
## Homogeneous 11691 3935 296 77
## Heterogeneous 13947 4948 481 95chisq.test(cat1cs)##
## Pearson's Chi-squared test
##
## data: cat1cs
## X-squared = 20.303, df = 3, p-value = 0.0001469The P-value suggests that there is an extremely low probability that the relationship between country of origin and level of homogeneity is by chance
Test 2 - Full vs. Partial vs. No Partisan Transfer
- Note: Due to time limitations, this test only includes 3 person households
- Adjusted Household Breakdown:
- Cuba: 7,594
- PR: 641
- PRCuba: 165
- Other: 21,178
- Note 2: Similar to my last test, I assigned the numbers in excel and imported the dataset into R. 0’s are Full Transfer Households, 1’s are Partial Transfer Households, and 2’s are No Transfer Households
- Unlike my last test, I am looking specifically at the relationship between the parent and child, as opposed to simple levels of homogeneity
- Disclaimers:
- I don’t specifically mean “parent” and “child” as I do a transfer between different generations. In this case, I defined generation as 18 years, which is a narrow definition. Some would define it as being larger – 20-25 years, but again, I am not looking at parent to child specifically.
- Many of these households are single parent households. I have coded single parent households the same as I would households that have two parents of the same party.
- I did not differentiate between older children and younger children. I coded a household with a 90 year old and a 60 year old the same as a household with a 50 year old and a 20 year old. This was again due to time limitations and can certainly be explored in future projects.
Table:
Graph
Takeaways:
- All households (Other, Cuban, Puerto Rican, Cuban-Puerto Rican) have a higher proportion of full transfer than partial or no transfer
- Cuban-Puerto Rican households have the highest proportion of no transfers, at 37.6%
- The others have the highest proportion of full transfers at 55.2%, followed by Cuba at 51.8%
- The others have the highest discrepancy between categories –there are 32.7% more full transfers than partial transfers, and 32.9% more full transfers than no transfers
Chi-Squared Test
cat3cs <- matrix(c(11691,4765,4723,3934,1488,2172,296,157,187,78,25,62), byrow=F, ncol=4)
colnames(cat3cs) <- c("Other","Cuba","PR","PRCuba")
rownames(cat3cs) <- c("Full_Transfer","Partial_Transfer","No_Transfer")
cat3cs## Other Cuba PR PRCuba
## Full_Transfer 11691 3934 296 78
## Partial_Transfer 4765 1488 157 25
## No_Transfer 4723 2172 187 62chisq.test(cat3cs)##
## Pearson's Chi-squared test
##
## data: cat3cs
## X-squared = 159.53, df = 6, p-value < 2.2e-16The P-value suggests that there is an extremely low probability that the relationship between country of origin and level of partisan transfer is by chance
Test 3 - Partisan Transfer by Party
To start, I looked at Miami-Dade county by country of origin and by age to get an overview of the demographic breakdown of the parties (Note: these graphs aren’t representative of the entirety of Miami-Dade, only the dataset I am working with - meaning this is Miami-Dade when only looking at 3-7 person households consisting of people with the same last name).
Graph: Miami-Dade by Country of Origin
Graph: Miami-Dade by Age
I also broke down country of origin by age:
Graph: Cuba by Age
Graph: Puerto Rico by Age
Graph: Cuba-Puerto Rico by Age
Graph: Others by Age
Takeaways:
- Unsurprisingly, Cuban households have a large amount of Republicans, though it seems to be much more prominent among older voters. Cuban-Puerto Ricans are showing similar results. Puerto Ricans seem to be more spread apart. Others lean heavily democratic. For all graphs, NPA is much more prominent among younger voters.
- Note: There was a 5th age category which covered people under 18 and over 105. I took that out as there were very few people in this category. Also, I took out everyone who identified with a party that was not DEM, REP, or NPA. I categorized “IND” the same as NPA.
Household- Party Combinations
- All Disclaimers I have stated in test two above apply here. This is only considering three person households.
- Code Key:
- All REP household
- All DEM household
- All NPA household
- REP parents, at least one DEM child
- REP parents, at least one NPA child
- REP parents, children are DEM and NPA
- DEM parents, at least one REP child
- DEM parents, at least one NPA child
- DEM parents, children are REP and NPA
- NPA parents, at least one REP child
- NPA parents, at least one DEM child
- NPA parents, children are DEM and REP
- Parents differ, all children are the party of at least one of their parents
- Parents differ, at least one child is of a different party
- Un-categorized
Graph: Cuban Household Combinations
Graph: Puerto Rican Household Combinations
Graph: Cuban-Puerto Rican Household Combinations
Graph: Others Household Combinations
Stray Notes:
- I adjusted the y-axis of the Puerto Rican Households to match the other three.
- Cuban Households most common combinations: 1 and 5
- Puerto Rican Households most common combinations: 2 and 10
- Cuban- Puerto Rican Households most common combinations: 1 and 5
- Others Households most common combinations: 1 and 2
Proportion of Party Transfer
- Divided homogeneous Party A households by all households with Party A parents
- Repeat for each country of origin with all parties (DEM, REP, and NPA)
- See the strength of partisan transfer from parents to kids on a party by party basis between countries
Table:
Take Cuban Republicans as an example. Of all of the Cuban Households where the parents are identified as Repubican, roughly 61.8% of these households have children that also identify as Republican.
Graph: Proportion of Party Transfer
Stray Notes:
- This test only accounts for Full Transfer households.
- Cubans have the highest rate of Republican transfer. Others have the highest rate of democratic transfer. Puerto Ricans have the highest rate of NPA transfer.
- Interestingly, DEM, REP, and NPA transfer in Cuban households seems to be pretty close in proportion, indicating that REP households aren’t as dominant as one would think.
Conclusion
- My first test sought to analyze the level of homogeneity of Cuban households, and compare them to Puerto Rican households and everyone else. As I have mentioned, Cuban households are more heterogeneous than everyone else except for Puerto Ricans. I certainly wouldn’t have predicted this. Based on a paper by Escaleras et. al, (2019), they propose the idea of linked fate, which ties into group identity. They found that as a group, Latinos will vote together to satisfy their individual self interests. In this case, their fate is “linked”. Because of this, I would have expected that Cuban households would be the most homogeneous, but this was not this case and needs to be looked at more in depth (which is what I did in my second and third tests).
- My second test expanded on my first, adding in age as a factor. In this case, I looked at specific parent-child interaction in partisanship. In this case, over half of Cuban households are full transfer households. When including partial transfer, Cubans come out at 77.7%. This is tied for highest, along with the “others”.
- My third test broke it down even further, adding in party as a factor. I wanted to see if one party had a higher transfer rate from parent to child than another party, and if this differs based on households with different countries of origin. Cuba, unsurprisingly, has the highest proportion of Republican transfer. What is surprising, however, is how close this proportion is to DEM and NPA. Given that Cubans are so dominantly Republican (especially the older ones), I can theorize that this closeness may be due to the growing number of DEM and NPA in Cuban youth as they move away from their parents’ and grandparents’ parties. Therefore, more and more households with Republican parents may be seeing their children leave their party, which then leads to a lower proportion of Republican transfer than one may have guessed based just on party demographics without accounting for age.
Works Cited
Bell, Aaron. “The Role of the Latino Vote in the 2016 Elections.” SSRN Electronic Journal, vol. 13, May 2016.
Bishin, Benjamin G., and Casey A. Klofstad. “The Political Incorporation of Cuban Americans.” Political Research Quarterly, vol. 65, no. 3, 2011, pp. 586–599.
Cilluffo, Anthony, and Richard Fry. “An Early Look at the 2020 Electorate.” Pew Research Center’s Social & Demographic Trends Project, 30 Jan. 2019, https://www.pewsocialtrends.org/essay/an-early-look-at-the-2020-electorate/.
Data Access and Dissemination Systems (DADS). “American FactFinder - Results.” United States Census Bureau, 5 Oct. 2010, https://factfinder.census.gov/faces/tableservices/jsf/pages/productview.xhtml?src=CF.
Escaleras, Monica, et al. “You Are Who You Think You Are: Linked Fate and Vote Choices among Latino Voters.” Wiley Online Library, John Wiley & Sons, Ltd (10.1111), 30 Aug. 2019, https://onlinelibrary.wiley.com/doi/abs/10.1111/polp.12329?af=R.
Data Collected by Daniel A. Smith