Thread by @gelliottmorris

FYI there is some evidence that polarization reduces the size of convention bounces, and polls in recent fall campaigns have been much stabler than in, say, the 1980s

General election polling in 2020 has been very stable (as we predicted). That _could_ change before November, yes, but there are at least two good reasons to think it won't: the low variance from Jan to July, & that election polls now are like half as volatile as they used to be.

Yes, candidates can get convention bounces, but #ackshually polls in the first half of the campaign have been more volatile than polls in the second half, historically speaking. This is the trend for the variance over the whole electio nyear:

Anyway, there seems to be some disagreement among prognosticators about how volatile (or not) modern contests are. I'm pretty compelled by these charts and the research into political polarization, but reasonable people can disagree & there is uncertainty around our assumptions.

(PS that uncertainty about variance is in our model; if the election suddenly gets more volatile, the variance in the forecasting component is allowed to increase a bit. But the mean estimate we provide constrains outcomes; & as we've discussed it might be like 0.1 % pts to low)

Oh, I should also say that our average expectation IS for the election to get tighter before November (a change in the mean of the distribution of outcomes) but that's not the same as expecting more volatility later (a change in the variance of the distribution). /end