Party Identification

 When looking at polls that show Biden and Trump tied, it's helpful to look at how pollsters are sampling the population. For instance a poll that has Biden losing Michigan, but samples the Black population at 7% when they make up 14% of the population of Michigan is a questionable poll. 

One thing that's especially tricky these days is simply getting people to answer the phone. Young people, in particular, do not answer the phone from people they don't know. This also makes sampling tricky, because you have to extrapolate out from the 15 Black people under 65 who answered the phone to cover all of Michigan's Black population.

What bears watching, however, is partisan breakdown in the polls. The new CNN poll that shows a dead heat between Biden and Trump has partisan samples as 32% Democrat, 32% Republican and 35% Independent. The number of "Democrats," "Republicans," and "Independents" depends on self-reporting. I don't have CNN's self-reporting data on that, but I have Gallup's. Their most recent sample has this breakdown: D-26; R-25, I-47. Looking over the past two years, we can see that Independents typically come in around 40%, give or take, with Republicans and Democrats largely split in the mid-20s to 30. 

Keep this in mind, when we see polling that shows Trump tied or winning for a few reasons. One is that Independents, unlike Republicans, are legitimately concerned about January 6th and Trump's criminal cases. After January 6th, we saw roughly Republican self-identification drop 5%, from around 30 to around 25, for obvious reasons. We are expected to believe that subsequent events - including Dobbs and Trump's cavalcade of indictments has led that 5% to return to the GOP. I have my doubts.

As we have seen in every actual election since Dobbs, Democrats have done about 5% better than the polls suggest. The implication to me is that either polls are just missing younger voters who are showing up to protect reproductive freedom or they are oversampling Republicans in order to avoid the polling failures of 2016 and 2020. They clearly missed the Blue Wave of 2018 and misread a Red Wave of 2022 (which wouldn't have even happened if it wasn't for the ineptitude of the NY Democratic Party). 

So leaving aside the fact that Biden's numbers are roughly where Obama, Dubya and Clinton's were a year before they were re-elected, leaving aside the fact that Republican voters are older and therefore more likely to have died since 2020, and leaving aside that incumbents usually get re-elected and Trump is uniquely unpopular, there could be a sampling problem if pollsters assume that there are the same number of Republicans now as there used to be.

UPDATE: Check out this tweet from Liz Cheney.