Wednesday, August 16, 2017

Turnout Troofing

Yesterday the Times Upshot ran a piece by Nate Cohn analyzing the 2016 presidential election in terms of what looks to me like another take in the Legendary White Working Class family of takes, identifying the crucial factor in Trump's victory as that particular set of white-no-college voters who voted for Obama in 2008 and/or 2012 but for Trump in 2016, and who are said to have made this remarkable switch mostly out of racial resentment (I actually don't think that's as bizarre as it sounds, but the obvious question it raises, of why a white person who voted for Obama one year would turn around against Clinton out of racial resentment the next time, is one Cohn doesn't even discuss) and then out of disappointment with Obama and then lastly because they agree with Trump's policy prescriptions as they understand them.

Which Cohn does not take to mean that Democrats need to appeal more to racists, even though that's what his data makes it sound like, but that we should take positions more like those of imaginary Trump, in favor of lots of infrastructure spending, and trade protectionism, and relatively relaxed sexual views. The great Zandar of Kentucky, though, hears Cohn thinking it, and he doesn't like it:

What that means is that Cohn is strongly suggesting that in order to be competitive, Democrats have to make a sea change to attract voters that harbor no small amount of racial resentment. Trump was able to leverage that resentment into massive distrust of the Obama administration and Democrats in general.
The problem is that this will come at a cost, and the cost will be borne by black, Latinx, and Asian voters and candidates [and female candidates too, I'd add].  I've said before that this path is suicidal for the Dems and so far Trump is making it incredibly easy to make the Democrats be the party of inclusiveness in comparison by simple dint of Trump's overwhelmingly awful racism, if not open support of white supremacists.

Nor do I.

Both for the range of moral-emotional reasons that make me revolt against the thought of moving the party back into that ugly territory of accepting little homeopathic doses of racism once again, and for the obvious political-science reason that we can't win without the full-hearted support of black and brown people, and abandoning them (or allowing them to feel abandoned) in pursuit of these dubious Trump voters looking for Mr. Good-Dem is just really bad tactics, and probably bad strategy. It's not only wrong, it's dumb to think there are that many of those guys waiting to get picked up.

Which is where I want to go here, just looking at those numbers. Especially, how does Cohn know how many Obama-to-Trump switch voters there are? Before the party starts chasing them, how important a group is it, in fact?

Now, I have my own current theory of the 2016 election, which is that Sam Wang was more or less right, and Clinton had a 95% chance of winning, and it was just that one out of 20 times when enough things go wrong that the outside chance prevails. This doesn't mean I'm claiming that it was a well-run campaign, because I don't think it should have been anywhere near that close in the first place, and I especially can't see why Democrats didn't take the Senate in Wisconsin and Pennsylvania, but it does mean I'm suspicious of anybody who insists that there's one overwhelming factor, whether it's Russians or all these sour-ass white guys, I just don't believe that's going to work.

I also have my own theory of American politics in general, which is that non-voters and unlikely voters play a decisive role that never gets enough attention from the pandits and apostles. Nearly half the population stays home in a national election in the US, and if they all came out it would certainly change things (this was Bernie's theory—he was just wrong in thinking he would bring them out). And one of the annoying things is the data is never presented in a way that makes it easy for me to figure out what the actual role of the nonvoter in a given election is going to be.

So I'm looking at some of the data Cohn is working with, as examined at Sabato's Crystal Ball at the University of Virginia, which are of very large-scale surveys of eligible voter (as opposed to registered or likely voters or people who definitely did vote) from the American National Election Study, which found that 13% of Obama's voters in 2012 went to Trump in 2016 (6% of total vote)

and the Cooperative Congressional Election Study, which found that it was more like 11% (5% of the total).

which  comes down to somewhere between 6.7 million and 9.2 million Obama-to-Trumpers over all.

So how do they know? The poll respondents told them so, and there is a problem with that, not necessarily a big problem (in the world of social science data self-report data is on the whole a lot more accurate than you might imagine), but something that should always be addressed with voting behavior in particular: Sabato's writes,
self-reported past voting behavior has limited value. Voters do not always accurately recall the past. Academic research has found that a number of different factors can explain these discrepancies. Bias toward current political preferences can affect responses, as can simple forgetfulness — as much as it may surprise hardcore politicos, many Americans don’t necessarily remember their voter information with precision.
And indeed, with the ANES data,
Overall, 58% of respondents to the 2016 ANES who said they voted in 2012 (including those who didn’t vote in 2016) claimed to have voted for Obama, seven points more than the 51% Obama actually won in 2012. Meanwhile, Romney only won 40% of the vote of those who said they cast a ballot in 2012, seven points fewer than the 47% Romney really won in 2012. Although there may have been a sizable number of voters who cast ballots in 2012 but didn’t vote in 2016 and vice versa, it would not be surprising if Obama’s percentage were somewhat inflated and Romney’s a bit diminished.
That's kind of a lot. And indeed there were a large number of voters who voted in one election and not the other and lied about it:
Overall, 59% of respondents in the 2016 ANES reported voting in both 2012 and 2016, with 68% claiming to have voted in 2012 while 74% voted in 2016.
But in reality just 55% of the eligible voters turned out in 2012, and 55.5% in 2016 (which was a not-bad turnout by US standards). That's a 13-point difference in those who claimed to have voted but hadn't in Obama's last election, and an 18.5-point difference in those who claimed falsely to have voted in Trump's. Self-report of turnout is really terrible!

And here's the thing: that sounds gross, but the overall numbers for the 2016 election are pretty good: even though people lied about whether they had voted or not, the difference between the reported Trump vs. Clinton vote is quite close to the actual vote (it gave Clinton a very slightly wider lead). That's one of the lovely things about social science, which is that in the broad picture, unsystematic errors don't mess it up much. It's only the errors of bias that lead to wrong conclusions.

What I'd love to know but don't is what the corresponding figures for Obama vs. Romney are, how what people claimed in 2016 they had done in 2012 compares to what they actually did in 2012 (if anybody can figure out how to work the numbers please let me know), but this is still very suggestive. I'd like to see the rememberers of 2016 giving Obama a ridiculously large margin in 2012, in, precisely, one of those errors of bias that make things happen

What I'd propose as an alternative to the zombie WWC theory is that there were no significant numbers of Obama-to-Trump voters, but rather two different sets of nonvoters: the ones who decided to come out for Obama in 2012 are different from the ones who turned out for Trump in 2016, but they're all likelier not to vote than the rest of us. And guess what: white people with no more than a high school degree don't vote all that much, and that's who's most likely to have produced the fictional reports (because you feel guilty when you admit to not voting).

And in these surveys, people in particular who really voted for Trump in 2016 and stayed home in 2012 were especially likely to claim they had voted for Obama, either because they thought it sounded good ("I'm not a racist, I voted for Obama") or really believed they had because he won ("I had so much hope but he disappointed me"). While people who didn't vote at all in 2016 but wished they had voted for Clinton claimed in many cases truthfully that they had voted for Obama in 2012.

But if I'm right, we don't need to change the demographic character of the party, is what I really want to say. We need to convince more less-educated people of all races including whites obviously to vote at all (and stop voter suppression for poor people of all races any way we can, of course). The issue isn't the WWC at all but hope, you know, the sense that voting's going to make a difference, hope and change.

Cross-posted at The Rectification of Names.

No comments: