Campaign Narrative Reflection
District Background For my campaign reflection, I decided to focus on the election in New York’s 17th District. NY-17 covers parts of the Lower Hudson River Valley. It was previously represented by Mondaire Jones and had been represented by a Democrat since 1983. During the 2020 redistricting cycle, the district’s boundaries were redrawn, making it somewhat more Republican. Prior to redistricting, the Cook Political Report had NY-17 as a D+7 district, but after redistricting, it became a D+3 district.
Model Reflection
In this blog post, I’ll reflect on the performance of my model using actual results from the 2022 midterms. As a reminder, my final model found that Republicans were strongly favored to take the House. My model predicted that Democrats would win 200 seats and Republicans would win 235 seats. Moreover, I gave Democrats an 8.14% chance of winning the House. In 80% of simulations of my model, Democrats won between 185 and 216 seats, and Republicans won between 219 and 250 seats.
Final Prediction Model
National Seat Forecast Republicans are strongly favored to take the House. My model predicts that Democrats will win 200 seats and Republicans will win 235 seats. In 80% of simulations, Democrats win between 185 and 216 seats, and Republicans win between 219 and 250 seats. District-Level Forecasts Zoom and hover over this district map to get predicted vote share, win probability, and prediction intervals for every district. Note: the template for this cartogram comes from Daily Kos Elections
Blog 7: Building a Pooled District-Level Model
This blog is an ongoing assignment for Gov 1347: Election Analytics, a course at Harvard College taught by Professor Ryan Enos. It will be updated weekly and culminate in a predictive model of the 2022 midterm elections. For my last blog post until my final prediction, I’m going to construct a pooled district-level model to predict Democratic vote share based on the demographics of the districts. It became clear that a pooled model — rather than 435 distinct models for each district — would be necessary to predict district-level results.
Week 6: The Ground Game and Probabilistic Models
This blog is an ongoing assignment for Gov 1347: Election Analytics, a course at Harvard College taught by Professor Ryan Enos. It will be updated weekly and culminate in a predictive model of the 2022 midterm elections. The main focus of my blog this week will be transferring my linear model to a probabilistic model. I’ll thus complete blog extension three as I create a binomial logistic regression model for every congressional district in the country.
Week 5: Advertising
This blog is an ongoing assignment for Gov 1347: Election Analytics, a course at Harvard College taught by Professor Ryan Enos. It will be updated weekly and culminate in a predictive model of the 2022 midterm elections. This week, I’ll be trying to answer one central question: does advertising in midterm elections predict the results of congressional elections? As we don’t yet have access to 2022 advertising data, I won’t be able to make a prediction for 2022 districts.
Blog 4: Expert Predictions and Incumbency
This blog is an ongoing assignment for Gov 1347: Election Analytics, a course at Harvard College taught by Professor Ryan Enos. It will be updated weekly and culminate in a predictive model of the 2022 midterm elections. This week, I’ll focus on the value of expert predictions and incumbency in building my model for the 2022 midterms. First, I’ll analyze the accuracy of experts in predicting the 2018 midterm elections, completing blog extensions 1 and 2.
Blog 3: Polling
This blog is an ongoing assignment for Gov 1347: Election Analytics, a course at Harvard College taught by Professor Ryan Enos. It will be updated weekly and culminate in a predictive model of the 2022 midterm elections. In this week’s blog, I will be incorporating polling into my predictive modeling. I will create predictions using polling in two ways. First, I’ll update a version of my economic fundamentals model from last week with polling data.
Blog 2: The Economy
This blog is an ongoing assignment for Gov 1347: Election Analytics, a course at Harvard College taught by Professor Ryan Enos. It will be updated weekly and culminate in a predictive model of the 2022 midterm elections. “It’s the economy, stupid,” became a slogan of Bill Clinton’s 1992 presidential campaign. With the nation facing recession under incumbent President George H.W. Bush, political strategist James Carville coined the phrase to keep the campaign focused on the big picture.
Blog Post 1: Exploring House Election Data
This blog is an ongoing assignment for Gov 1347: Election Analytics, a course at Harvard College taught by Professor Ryan Enos. It will be updated weekly and culminate in a predictive model of the 2022 midterm elections. The objective of this first blog post is to explore and visualize historical House election data in order to gain a better understanding of the electoral context underpinning the 2022 midterms. I will complete blog extensions 2 and 3, as well as parts of extension 1.