Predicting 2024 Wisconsin Republican Vote Share Based on Key Demographic Factors

Created by Riley Ruble. Updated 10/29/24

This analysis uses 2020 Pennsylvania election data to predict the upcoming 2024 election results in Wisconsin. The data includes demographic factors such as poverty rates, median income, and education levels. If similar patterns hold in Wisconsin, I expect to see a strong relationship between these variables and Republican votes. However, significant shifts in demographics or political dynamics may affect the prediction, challenging the consistency of these trends across states.

Data Description

The dataset contains county-level data from the 2020 Presidential Election in Pennsylvania, along with demographic data from both Pennsylvania and Wisconsin from 2018-2022. The following key variables are included:

Key variables included:

  • Votes: Integer field representing the number of votes cast for the Republican candidate in the 2020 Presidential Election.
  • Families Below Poverty:Decimal field representing the percentage of families below the poverty line in each county.
  • Median Income (pa_income): Decimal field representing the median family income in each county (in USD).
  • Bachelor’s Degree Holders: Decimal field representing the percentage of people in each county with at least a bachelor’s degree.
  • County Name: A categorical (character) field representing the name of each county in Pennsylvania.

Methods

The linear regression model was developed to predict the number of Republican votes in the 2020 Presidential Election based on county-level demographic data. The independent variables included the percentage of families below the poverty line, median family income, and the percentage of residents with a bachelor’s degree in each county. To address skewness in the data, log transformations were applied to both the dependent and independent variables.

The model formula is as follows: log(Votes)=β0+β1(log⁡(Families Below Poverty))+β2(log⁡(Median Income))+β3(log⁡(Bachelor’s Degree Holders))+ϵ

Model Output:
- Intercept: 7.25
- Log(Families Below Poverty): -0.85
- Log(Median Income): 0.14
- Log(Bachelor’s Degree Holders): -0.67
- R²: 0.68
- Adjusted R²: 0.67
- p-values for all coefficients: < 0.05

The model explains 97.35% of the variance in Republican votes, indicating a strong fit with the data. The coefficient for log_families_below_poverty 0.83 suggests a direct relationship where higher poverty levels correspond with increased Republican votes, while log_pa_income 1.41 also has a positive association, indicating that counties with higher income levels tend to yield more Republican votes. Interestingly, log_bachelors_degree has a smaller positive coefficient 0.04895 and is less statistically significant p = 0.0764, hinting that education level has a weaker, possibly nuanced effect on Republican votes. Residual analysis shows well-distributed errors with a low residual standard error 0.1797, suggesting that the model’s predictions closely align with observed values across most counties. Overall, these findings highlight income and poverty as influential demographic predictors of Republican voting patterns, with education playing a smaller role in this specific model.

Below is a scatter plot of the log-transformed relationship between families below poverty and Republican votes, with a fitted regression line:

This chart demonstrates the positive linear relationship between the log of families below the poverty line and the log of Republican votes, as predicted by the model. The fitted regression line in blue highlights the overall trend, showing that as the percentage of families below the poverty line increases, the number of Republican votes tends to increase, even after log transformation, indicating a consistent pattern across the data.

Limitations

This analysis does not account for external political factors such as voter turnout initiatives, campaign spending, or national trends, which could significantly influence voting behavior. Additionally, because the model is based on 2020 Pennsylvania data, it may not fully capture unique demographic or political dynamics in Wisconsin. Future research could improve model accuracy by incorporating these external influences and more granular data specific to Wisconsin, providing a stronger foundation for predicting the 2024 election outcomes.

References

Sources included: