If we could run hundreds of presidential elections, we would be able to get a pretty good idea of whether the 538 model or the Economist model is better calibrated. But we only get one every four years, and the models get refit between those elections anyway.
So suppose Biden wins, as both models currently say is quite likely. What evidence would we look for, ex post, to help make the call that one performed better than the other? To put it another way, what evidence (conditional on a Biden win) would convince you that 538 had actually done a better job?
bbbsssjjj681 karma
If we could run hundreds of presidential elections, we would be able to get a pretty good idea of whether the 538 model or the Economist model is better calibrated. But we only get one every four years, and the models get refit between those elections anyway.
So suppose Biden wins, as both models currently say is quite likely. What evidence would we look for, ex post, to help make the call that one performed better than the other? To put it another way, what evidence (conditional on a Biden win) would convince you that 538 had actually done a better job?
Keep up the awesome work.
View HistoryShare Link