View on GitHub


Polls and Probabilistic Models

Oct 9, 2020

Throughout previous blogs, I have relied on linear regression models to derive election results. Although such has helped me analyze the relationships between variables like average poll support and popular vote shares, the outcome of a linear regression model can be any value in a continuous range; thus, some predictions (i.e., predicted popular vote shares) could be outside the 0-100% range. This clearly reduces the practicality of some of the linear regression results. Therefore this week, I will explore 2020 predictions through probabilistic models, like a binomial logistic regression, limiting election outcomes to a finite draw of voters from a given population (i.e., the voting-eligible population). I will base such probabilistic models on the average poll support for each of the major party candidates in 2020, first considering the entire country and then focusing on the key battleground states in the 2020 election.

Win Margins for the United States in 2020

Biden’s Simulated Win Margins in 2020

This map demonstrates the 2020 win margins for the Democratic candidate, Joe Biden, relative to the Republican candidate Donald Trump. The win margin is derived from simulating the predicted distribution of draws (10,000 draws) for each candidate from the population (i.e., the voting-eligible population) in each state. The draw probability for Biden and Trump is based on their average support. The main takeaways are:

Win Margins for the Battleground States in 2020

This section will consider Biden’s 2020 win margins in the 5 battleground states highlighted by the New York Times. I will again use probabilistic models and rely on average poll support data in each state for the two major-party candidates. By focusing on these battleground states, I will determine which candidate has a higher probability of winning in the pivotal “swing states” and thus the 2020 election.

Biden’s Win Margin in Florida Biden’s Win Margin in Ohio
Biden’s Win Margin in Georgia Biden’s Win Margin in North Carolina
Biden’s Win Margin in Iowa

The breakdown for Biden’s win margins in the above battleground states are as follows:

Based on my probabilistic model and resulting win margins, Biden has a clear advantage in the afromentioned battleground states. In fact, my projected win margins indicate Biden could carry as much as 60 of the total 85 electoral votes in the battleground states. Such is significant, as a decisive win in the battleground states gives Biden a significantly higher chance of winning the overall election. This stands in stark contrast to 2016, where, as my first blog demonstrated, Trump took the majority of the electoral votes from the swing states, ultimately leading him to victory.

Final Takeaways

In this blog, I used probabilistic models to evaluate the electoral prospects for the 2020 candidates. By calculating the Democratic candidate’s win margin in various states, based on their average poll support, I projected that Biden would take most of the battleground states’ electoral votes and thus stand a greater chance of winning the 2020 election. This analysis affirms previous model predictions based on linear regressions of average poll support. Of course, polls are not the only variable in predicting elections; future probabilistic models could consider incorporating other variables, like incumbency status and 2nd quarter GDP.