Steven Ratner argues that Trump is in great shape heading into 2020:
The economy invariably ranks among the top issues on the minds of voters in presidential elections. At the moment, it appears to offer President Trump a meaningful tailwind.
But how big is that tailwind? Fortunately, economists have worked hard to develop models for predicting election outcomes, and according to one of the best of these, it should be quite large.
One of the first — and perhaps still the best — of these models was created by Ray Fair, a professor at Yale. He found that the growth rates of gross domestic product and inflation have been the two most important economic predictors — but he also found that incumbency was also an important determinant of presidential election outcomes.
One should be very suspicious of models built on small number of variables based on a small sample of elections. It’s not hard to come up with simple models that can explain most outcomes retroactively, but they’re prone to falling apart when applied going forward.
But, hey, if Fair’s model did well in 2016, it did well, so let’s see…
When the 2016 election rolled round, a surprising result emerged. According to the model, Donald Trump should have received 54.1 percent of the vote; in actuality he received 48.8 percent.
So, this model overestimated Trump’s share of the top-two-party popular vote by more than ten points, predicting Trump to beat Clinton by more than 8 points when in fact he was came in at -2. I am just a mere political scientist, but this strikes me as…a massive failure! But don’t worry, those hands are getting ready to wave:
I’m quite confident that the gap was a function of the generally unfavorable rankings on Mr. Trump’s personal qualities. In other words, a more “normal” Republican would likely have won the popular vote by a substantial margin (instead of losing it by three million votes).
So if you just throw in a factor — in this case, nothing but unempirical conventional wisdom from summer 2016 — the model ignores, actually the model that failed miserably worked! This is just junk science:
The ones that use economic data as a prior, and blend it with polling data, do OK and seem to add some value. But the ones based on economy/”fundamentals” alone range from mediocre for the best ones to bordering on junk science for the worst ones.https://t.co/uzhRzj9s2F— Nate Silver (@NateSilver538) May 27, 2019
And, in addition to all that, if we peg Trump’s value as a candidate at at least -10 points, that seems like a problem for his re-election, since he will still be on the ballot in 2020. I’m not sure I can stand much more of this analytical rigor from the One True Social Science.