By Sam Thau
This blog is a log of the assignments for the Harvard College class Government 1347: Election Analytics taught by Dr. Ryan Enos. The blog will be updated every Saturday throughout the fall semester of 2020.
This is my prediction for the 2020 election. I build a probabilistic model that includes correlation between states, based on a host of economic, demographic, and polling data.
The postmortem of my model, assessing what went right, what went wrong, and what I should have done to correct my mistakes.
A look at the urban-rural divide, and if it increased during the 2020 election. Discussion of the urban-rural divide itself, and how it relates to other factors about the election.
An introduction to the blog. Covers different ways to think about election results. Includes a discussion of “tipping point states.”
Models using the economy to predict election outcomes. National and state by state regressions of various forms.
Incorporating polls into linear regressions, following similarly to the pervious post. Shows how in certain models, including polls increases model accuracy at the state level.
Discussions of incumbency in the 2020 election, along with an introduction to probabilistic models.
Examination of a binomial model, and thinking about some alternate models that may be useful in the coming weeks.
Examination of a beta-binomial model, estimated using hierarchical empirical Bayes.
An examination of the predictive power of different two sided models over time. In addition, I take yet another crack at a probabilistic model, this time with some success in introducing variance.
A running log of content that does not quite fit into the blog. If updated to go along with a post, a note will be made in the corresponding blog post. Last updated for 7. Parties and Uncertainty.