,
3 tweets,
1 min read
Read on Twitter
For people interested in Nov. 2020: The rule is look at macro indicators early (it's *very* early now) and micro indicators (i.e. state by state polls) late. Right now, a relatively minor increase in the risk of recession is a lot more material than a new poll in WI or whatever.
tbh I think people recall 538's highly detailed analysis of individual polls or the Electoral College math and sort of mimic the style, but they forget that there's a fairly particular time and place when we do it: Within a few months of the election and *not* a ton before that.
Doesn't mean you can't indulge in a little micro once in a while this far out from the election, it's fun to click on those Electoral College maps. But from a forecasting accuracy standpoint your view at this point should be 40% macro, 10% micro, 50% "we don't know anything yet".
