Stanford Seminar - How Can We Understand and Evaluate Election Forecasts, Given That N=15 or Less

Stanford Seminar - How Can We Understand and Evaluate Election Forecasts, Given That N=15 or Less

Explore the complexities of election forecasting in this Stanford seminar featuring Andrew Gelman from Columbia University. Delve into the challenges of understanding and evaluating election predictions with limited data points. Learn about preelection polls, differential nonresponse, external calibration, and political polarization. Examine the intricacies of interpreting forecasts, including the significance of polls above the 95% interval. Analyze the shortcomings of popular forecasting models like 538 and SimMap, and understand the concept of National Swing in electoral predictions.