I beg everyone please do not write takes based off the detailed internals of unweighted exit polls https://twitter.com/CharlesMBlow/status/1323982596980826113

The way this works is that they have too few older white people with no college degree in the sample.

So to match a poll that has too fewer members of Trumpy demographics in it to the actual result, you end up inflating Trump's numbers with non-Trumpy demographics.

Some day we will get better data!

But remember the polls understated Trump's support.

The exit polls can't do that (they are matched to the vote count) so they have to twist a bad sample into the right shape. It's very misleading.