Panelists discuss interpreting election polls, models and prediction markets

11 articles · Updated · ssrs.com · May 6

At a panel moderated by Doug Usher, Joe Lenski, Andrew Gelman, John Aristotle Phillips and Kabir Khanna examined battleground states, uncertainty and forecast communication.
Gelman stressed polling limits in close races, while Phillips said prediction markets quickly aggregate expectations but do less to explain voter behaviour.
The discussion emphasised that polls, models and markets are complementary tools, and that probabilities can be widely misunderstood depending on context.

When polls and prediction markets give conflicting forecasts, which signal is more reliable for predicting election outcomes?

As prediction markets grow, how can we prevent them from being used to influence elections, not just forecast them?

If high campaign spending creates a statistical draw, how much does forecasting accuracy in close races truly matter?