Accuracy of Political Event Contracts

It's interesting to compare the implied probabilities on Kalshi for political outcomes versus traditional polling. Often, the market seems to lead the polls. Any thoughts on why this might be, or specific examples where you found Kalshi to be more accurate?