Data Nerds

Via Sierra Madre Tattler.

From his R&R location, Steve is wondering:

Am I the only one who thinks the data nerds are overcompensating for 2016 by portraying Trump as nearly unbeatable in 2020?

— Steve M. (@nomoremister) July 21, 2019

And I figured that's something he might be writing about if he were here.

He's talking among other things about the New York Times Upshot department and Nate Cohn, who reported over the weekend that

President Trump’s approval ratings are under water in national polls. His position for re-election, on the other hand, might not be quite so bleak.

His advantage in the Electoral College, relative to the national popular vote, may be even larger than it was in 2016, according to an Upshot analysis of election results and polling data.

Based on that, I'd say yes, it's overcompensating. The basic hypothesis seems to be about those key states that Trump unexpectedly won in 2016, Pennsylvania, Wisconsin, and Michigan, and significance of the broad Democratic swings they saw in the 2018 midterms; Cohn calls attention to Trump's approval numbers in November 2018 and suggests that he was a lot less unpopular than the election results might make you think—and less unpopular than he was in 2016, when he won the states, and especially certain population centers, and especially in Milwaukee, and if Democrats lose Milwaukee in 2020 they'll lose Wisconsin. And they may not be able to replace Wisconsin with some other state they narrowly lost in 2016, like Arizona, North Carolina, Iowa, or Florida, because he's less unpopular in those states than you might think, and especially in certain population centers, and especially Miami-Dade County.


If you follow me, and if you do, you may be noticing that this is assigning a vast significance to a tiny set of possibilities as if it were guaranteed that nothing else is happening at all—like, to be obvious, a much livelier turnout in Madison, or Milwaukee's black community, or whatever parts of North Carolina Cohn hasn't considered. Or a change in the president's personal popularity;

All of this is based on the president’s approval rating — well ahead of the election. Most presidents manage to improve their approval rating between this point and the election, particularly with a strong economy. But unforeseen events could also hurt his approval rating; it is even imaginable that the president could go too far on immigration for some of his more moderate supporters.

Uh and maybe the increasing perception that the president is a criminal, which is more than a little higher than it was when Cohn got his numbers last November, and might "even imaginably" have some kind of effect on turnout (which is so far expected to be enormous by US standards, up to 70%). To say nothing of a change in the economic outlook as the recovery that started in 2009 has already broken the records for durability and seems more and more endangered by the president's loony trade policy.

And the possibility that the Democrats' presidential candidate could have an effect on the race (you sometimes feel these guys would be more comfortable making predictions if the Democrats decided not to nominate anybody at all), which Cohn doesn't completely ignore:

Democrats could nominate a candidate who tries to win the presidency by mobilizing a new, diverse coalition with relative strength in Sun Belt states, while making little or no effort to secure the support of the white working-class voters with reservations about the president.

The Democrats could certainly win in the Sun Belt states, even in Texas. Perhaps this kind of Democrat could generate such a favorable turnout that it helps the party even in relatively white states.

(As ever, I have to protest against that glib use of the "white working class" as a constituency; Democrats will certainly make an effort to get votes from all working class voters, and if it's the right nominee even the New York Times and CNN could be forced to notice it. Health care and student debt are serious issues for real people, sharpened by the Trump administration's failure to do anything about them, and even those guys who eat breakfast in a diner care about Social Security payments, which any good candidate will be offering to boost through a raised cap on the income subject to payroll tax.)

Evidence that Trump is more likely to win Wisconsin in 2020 than he was in 2016?

What I think the analysis shows is how bad the tools we have are for talking about the Electoral College, in this new world in which profound unpopularity is not necessarily a bar to victory.

It's understandable that they should be shaken up by Trump's crazy win, but in the first place they should recognize that they weren't all that wrong (people often fail to understand that the mainstream polling predicting near certainty of a Clinton victory was a prediction of the popular vote, and reasonably accurate). And that what happened in 2016 is likely to be something they weren't measuring at all.

Cross-posted at No More Mister Nice Blog.