This project investigates the data surrounding California’s Proposition 61 which appeared on the 2016 Ballot. Proposition 61 would prohibit the state from buying prescription drugs above the lowest price offered for the Department of Veteran Affairs. This particular proposition is interesting to investigate because pharmaceutical companies spent a whopping $109 million dollars on advertising against the measure. The Proposition was not passed, by a 54% vote for No. Did the advertising from Big Pharma target veteran and rural populations, or are there more subtle factors that influenced the outcome of this proposition? To begin, we will briefly compare and contrast differences in census statistics across counties that voted yes and no.



To create these scatter plots, specific census data was pulled and filtered by county. For the Veteran Proportion, the number of veterans was devided by county population. Below is an example code 

variables <- c(
  veterans = "B21001_002E",  
  total_pop = "B01001_001E"  
)

county_data <- get_acs(
  geography = "county",
  variables = variables,
  state = "CA",
  year = 2016,
  survey = "acs5",
  output = "wide"
)

filtered_data <- county_data[county_data$NAME %in% paste(counties, "County, California"), ]

filtered_data$Veteran_Proportion <- filtered_data$veterans / filtered_data$total_pop


ggplot(filtered_data, aes(x = reorder(NAME, -Veteran_Proportion), y = Veteran_Proportion)) +
  geom_bar(stat = "identity", fill = "darkorange") +
  coord_flip() +
  labs(
    title = "Veteran Population by County that voted No for Prop 61",
    x = "County",
    y = "Proportion of Veterans"
  ) +
  scale_y_continuous(labels = scales::percent) +
  theme_minimal() +
  theme(
    axis.text.x = element_text(angle = 45, hjust = 1),
    plot.title = element_text(hjust = 0.5, size = 16)

Veterans as Proportion of Total Population: No on Prop 61

Veterans as Proportion of Total Population: Yes on Prop 61



Proposition 61 was significant to veteran populations and was a large focus point pharmaceutical companies advertised. Theoretically, if Medi-Cal was getting the same low prices as Veteran Affairs, nothing would stop pharmaceutical companies from raising the prices for everyone. Therefore, it was relevant to analyze veteran status with a vote for Proposition 61. These graphs indicate a difference. Counties that voted no, on average, had a much greater veteran proportion of the county population. However, some counties had a similar veteran population. The following graphs will compare counties with similar veteran populations by differentiating factors.

Median Age: Yes on Prop 61



These graphs show similar average ages. Age is an important factor to consider because older individuals will qualify for Medi-Cal, and the proposition would then be advantageous. However, in counties with similar veteran populations, there was not a significant difference in median age.

Median Income: Yes on 61



This data shows a sharper contrast. Counties that voted Yes had a significantly higher average income. This is counter intuitive, as you would expect counties with lower come, with more people who qualify for Medi-Cal, to support Proposition 61. However, high income most likely indicates that these individuals are more likely to be privately insured. This disconnect could have allowed voters to be less persuaded by the advertisements surrounding Proposition 61.

Median Household Size: Yes on Prop 61

Median Household Size: No on Prop 61



Counties that voted Yes on Prop 61 had smaller households. This could be correlated with a number of factors. Larger households using a wider variety of healthcare services, differences in ages requiring a greater range of coverage, and eligibility.

Alameda County



We will examine Alameda County to preform precinct level analysis. Alameda County voted yes and has a lower veteran population. Therefore, will the most obvious factor for a no vote excluded, we can examine more closely how income, age, and household size correlate with a Yes or No vote.



To approach this code, I first overlayed the precinct with a Yes Vote percentage with data from the statewide data base. I mereged these colums by their precinct identifier.

join <- left_join(shapefile_data, vote61, join_by(PRECINCT == svprec))
## Reading layer `mprec_001_g16_v01' from data source 
##   `/private/var/folders/c6/wd3l8r_d0pv0df44zx1m3fq40000gn/T/RtmpWlfVRw/mprec_001_g16_v01.shp' 
##   using driver `ESRI Shapefile'
## Simple feature collection with 1664 features and 1 field
## Geometry type: MULTIPOLYGON
## Dimension:     XY
## Bounding box:  xmin: -122.3738 ymin: 37.45419 xmax: -121.4692 ymax: 37.90669
## CRS:           NA



This is the precincts overlapped with their percentage of a Yes vote. Green indicates a precinct that voted over 50% Yes and red indicates a precinct that voted over 50% No. This is our baseline for comparison

I then preformed a spatial merge between the precinct shapefile and a cnesus block shapefile. I encountered challanged here, attempting to make the block fit the shapefile because the boundaries are different. I overcame this by using st_transform.

tracts_with_data <- tracts %>%
  left_join(census_data, by = "GEOID")

st_crs(shapefile_data) <- invisible(st_crs(tracts))
shapefile_data <- invisible(st_transform(shapefile_data, crs = st_crs(tracts)))

invisible(st_is_valid(shapefile_data))
invisible(st_is_valid(tracts))

precincts_with_data <- suppressMessages(st_join(shapefile_data, tracts_with_data, left = TRUE))

precinct_summary <- precincts_with_data %>%
  group_by(PRECINCT) %>%  
  summarize(
    avg_income = mean(estimate, na.rm = TRUE)
  )


ggplot(data = precinct_summary) +
  geom_sf(aes(fill = avg_income)) +
  scale_fill_gradient(
    low = "red",    
    high = "green",  
  ) +
  theme_minimal() +
  labs(
    title = "Average Household Income by Precinct in 2016 Alameda County",
    fill = "Income"
  )



Precinct Level Median Income



Here we see a positive correlation. Low income is associated with a No vote. This suggests that Advertising was effective in spreading the message that this Proposition could impact drug prices overall, which would harm low income precincts much more than high income precincts. However, this differentiates when we reach the analyze the precincts with the lowest income, which is correlated with a Yes vote. This is visible in the upper left region of the map. Could this region be differentiated by other census factors?

Precinct Level Median Age



If you look at the dark red cluster in the upper left region of the map, this region was correlated with a Yes vote. The area had a much higher average age, indicating a significant amount of individuals on Medi-Cal. However, the other, larger, dark red region of the map is correlated with a mixed vote. Younger individuals along the line of small precincts also voted Yes on average. These results confirm what the county level analysis that age is not a strong correlating factor

Precinct Level Median Household Size



With this analysis, we see more clarity surrounding the upper left larger precincts. This area has a smaller household size correlated with a yes vote

Summary of Visualization

Significant Precincts Upper Left Region (Larger Precincts): This area is correlated with a Yes vote. The larger precincts on the most left side consist of predominately low-income, middle-aged, individuals with small households. Left Cluster (Significant Elderly Population): The area is correlated with a Yes Vote: The precincts consist of high-income, older individuals with a small average household size. Left Region (Smaller Precincts Resembling a Diagonal Line):This area is more mixed in regards to a yes or no vote. These precincts consist of predominately low-income, younger individuals with a range of household sizes. Right Cluster of Smaller Precincts: This region is correlated with a No vote.These precincts consist of low income, middle-aged, individuals in smaller households.

Conclusion

This project explored the voting patterns of California’s 2016 Proposition 61, which aimed to regulate prescription drug pricing by aligning state purchases with the lowest prices offered to the Department of Veteran Affairs. Using county-level and precinct-level data, the analysis examined census data, like veteran status, income, age, and household size to determine their relationship with a Vote for Proposition 61.

At the county level, a clear correlation emerged between veteran populations and voting preferences. Counties with a higher proportion of veterans were more likely to vote “No,” possibly due to concerns that lowering drug prices for Medi-Cal would affect veterans’ access to medications. Conversely, counties with higher average incomes tended to vote Yes, suggesting that higher-income voters, possibly with private insurance, were less influenced by pharmaceutical advertisements.

Further analysis at the precinct level in Alameda County provided more nuanced insights. While low-income areas generally showed a No vote, there was an exception in some of the poorest precincts, which voted Yes. This was evident in the larger precincts in the upper left region of the map, which were predominantly low-income, middle-aged, and had small households. This region’s support for Proposition 61 may have been influenced by the significant number of older individuals, presumably on Medi-Cal, who stood to benefit from lower drug prices.

The map visualizations also revealed patterns in household size, with smaller households correlating with a Yes vote in certain precincts. Additionally, precincts with a higher average age, possibly indicating more Medi-Cal recipients, also leaned towards a Yes vote. These findings suggest that household size and age, in conjunction with income levels, played a significant role in determining support for Proposition 61.

In conclusion, the data analysis and visualizations illustrated the complex relationship between demographic factors and voting behavior. While income and household size were key predictors, age and veteran status also contributed to the variability in support for the proposition. The mixed results across precincts underscore the challenges pharmaceutical companies faced in swaying public opinion through advertising.

Future studies could focus on low-income areas where misinformation about the proposition may have had a larger influence. Additionally, a more granular examination of precincts with mixed voting patterns could help identify other demographic or social factors that were not accounted for in this analysis. Expanding this analysis to include a broader range of propositions related to healthcare policy could provide further insights into the intersection of healthcare, income inequality, and voting behavior.