https://goodreason.substack.com/p/nate-silvers-finest-hour-part-1-of
And then there’s 538, which gives Trump a 29% chance of winning. This is actually down from his peak a day or two earlier, when they gave him a 35% chance.
Silver gives some thoughtful explanations for why they’re way more bullish on Trump than others, most clearly laid out here. I don’t want to jump the gun, but I also feel obligated to draw attention to one reason he gives for 538’s relative bullishness on Trump. So, pre-election Nate, why might Trump have a chance?
State outcomes are highly correlated with one another, so polling errors in one state are likely to be replicated in other, similar states.
… Basically, this means that you shouldn’t count on states to behave independently of one another, especially if they’re demographically similar. If Clinton loses Pennsylvania despite having a big lead in the polls there, for instance, she might also have problems in Michigan, North Carolina and other swing states. What seems like an impregnable firewall in the Electoral College may begin to collapse.
Silver wrote this 2 weeks before the election, and - spoilers for the election outcome - it is so spot-on I had to double-check that it wasn’t written after the election. Clinton lost almost entirely because a polling error with one demographic group - non-college-educated white people - meant that she lost at least three states with a lot of them.
Whatever Silver’s reasons, the Twitterati are, at best, uncomfortable with his conclusions. The pundit response is to gently cast doubt on 538’s model.