2020 election: Exit ballot accuracy isn’t superb

Earlier than the votes from the 2020 presidential election have been absolutely tabulated, pundits all throughout the land are venturing forth to supply evaluation based mostly on the demographic breakdowns supplied within the exit polls.

If we’re seeking to reiterate our personal prior convictions, the information is serviceable. But when we would like precise details about voting habits, exit polls usually are not very helpful — and the early exit polls that haven’t even been weighted to the ultimate vote rely are even worse.

To start out with, as all of us continue learning over and over, it’s troublesome to conduct correct surveys of voters’ opinion. Nothing about conducting an correct exit ballot is any simpler than conducting an correct pre-election ballot. If something, it’s tougher. Some individuals vote on Election Day, and different individuals vote early by completely different means, so particularly this 12 months, with an enormous shift in how individuals vote, the trendy “exit” polls want to mix a number of waves of survey knowledge.

Final however in no way least, one benefit exit pollsters do have is that they know the ultimate final result of the election. So that you by no means publish an exit ballot saying Biden gained the election by eight factors when he actually gained by 4. As a substitute, the exits are weighted to match the precise final result.

B.A. Van Sise/NurPhoto by way of Getty Photos

However we don’t but know what the precise final result is, so the “outcomes” of the exit polls hold altering as extra votes are counted. CNN’s exit ballot write-up has a disclaimer that reads “exit ballot knowledge for 2020 will proceed to replace and can mechanically mirror within the charts under,” that means that not solely is the information you’re working with inaccurate the very web page you’re citing will say one thing completely different sooner or later.

Sadly, factoids derived from exit polls — like “53 p.c of white ladies voted for Trump” — are inclined to develop into hardened typical knowledge fairly rapidly. Extra correct data solely turns into obtainable later (a Pew evaluation based mostly on administrative knowledge about who truly voted suggests the share was round 47 p.c), at which level the reality comes out, however no person cares anymore. It’s tempting to grab on this type of data as an instance broader factors we wish to make about social or cultural traits, however realistically it’s simply not potential to reply sure questions with the extent of precision implied by exit ballot outcomes.

Trump’s mysterious overperformance with white faculty graduates

Exit polls, public opinion surveys, and county-level demographic knowledge all clearly present that Black and Hispanic voters overwhelmingly backed Biden (regardless of some Trump features with these communities) whereas whites with no faculty diploma love Trump and college-educated whites are extra skeptical.

However in accordance with the exit polls, Trump truly did a lot better with white faculty grads than pre-election polls recommended. That’s spurred some robust takes, together with Andrew Sullivan’s suggestion that white faculty graduates lied to pollsters en masse and a data-rich article by political scientist Erik Kaufman arguing that college-educated professionals grew to become Shy Trumpers due to issues about political correctness within the office.

It’s actually potential that one thing like that is true. However the extra boring (and in addition extra doubtless) rationalization is that the exit polls are overstating Trump’s help amongst white faculty graduates.

In keeping with the exit polls, for instance, non-college whites had been 34 p.c of the citizens in 2020. That’s precisely the identical because the 34 p.c share that exit pollsters present in 2016. However when Pew went again and examined the citizens based mostly on validated voter knowledge, they discovered that non-college whites had been 44 p.c of the 2016 citizens. Catalist, a Democratic knowledge agency that works with voter file knowledge discovered that non-college whites had been 48 p.c of the 2016 citizens.

One risk is that the non-college white share of the citizens plummeted for mysterious causes. A second, extra believable idea is that the exit polls are undercounting non-college voters once more. So what occurs whenever you undercount non-college voters in a ballot? In a traditional ballot, you’ll simply find yourself underestimating Trump’s help. However in an exit ballot, you know what number of votes Trump bought. So with too few working-class whites in your pattern and too many white professionals, to make the numbers add up it’s good to exaggerate Trump’s stage of help with each subgroups.

In actuality, Trump is much less fashionable with working-class whites than the exit polls say and in addition much less fashionable with white faculty graduates. It’s simply that the white inhabitants is extra working-class than the exit polls say.

The 53 p.c delusion

Again in 2016, the parable of majority help for Trump amongst white ladies was based mostly on an analogous error.

As Molly Ball wrote in a little-read 2018 debunking, Pew’s evaluation means that the exit polls had been merely undercounting the variety of white voters. By doing that, the polls ended up exaggerating the share of the white vote that Trump gained. On this specific case the distinction between 53 p.c and 47 p.c isn’t giant, however it occurs to cross the psychologically vital 50 p.c threshold. (Remember the fact that all polls are topic to error, and Pew’s methodology doubtless will get nearer to the right share however can also be not precise.)

The reality took a very long time to come back out, nonetheless, as a result of it merely takes a very long time for complete voter file or census knowledge to be obtainable.

If you wish to be extra dependable together with your sizzling takes, take a look at maps and census details about county-level demographics. Trump, for instance, did worse general in 2020 however higher in Miami and within the Rio Grande Valley in Texas and within the handful of Massachusetts cities with giant Hispanic populations and within the precincts of Milwaukee with giant Hispanic populations. It’s theoretically potential that that is all some form of bizarre coincidence, however it actually seems to be like some broad features for Trump with Hispanic voters. To know the way broad precisely, we have to look ahead to full tallies from Arizona, Nevada, and the notoriously gradual counters in California.

You can too simply see together with your eyes that Trump collapsed within the suburbs of Atlanta, in Fort Price, and in lots of different jurisdictions which might be stuffed with economically comfy faculty graduates. Catalist says it should have extra data in a month or two and we are able to examine again then.

It’s potential there was some form of secret anti-anti-Trump backlash lurking someplace on the market amongst faculty grads, however it’s exhausting to see precisely the place it could be.