Moody's Analytics' Very Accurate Election Model -- Which Has Only Been Wrong Once in 40 Years -- Points to Trump Victory, Possibly a Crushing One

Before getting to the current model, we should look at the one time the model broke down.

Which was... 2016. As the prediction stood on November 1, 2016:

Low gas prices and President Barack Obama’s high approval ratings are key factors that favor Democrat Hillary Clinton winning the White Hou:se in next week’s election, according to a model from Moody's Analytics that has accurately predicted the last nine U.S. presidential contests.

...

The Reuters-Ipsos States of the Nation project also predicts a Clinton win, with a 95 percent probability of her winning at least 278 electoral votes.

So it got that one wrong.

Moody's projection is largely based on underlying economic factors -- pocketbook, stock market, unemployment rate -- with a comparatively small dependence on polling information.

So, assuming the economy stays healthy, the current forecast is for a Trump victory even bigger than the one erroneously predicted for Hillary.

Moody's Analytics released the results of its 2020 prediction model on Tuesday showing Trump winning with 332 electoral votes, an increase over his 2016 win of 306, if voter turnout remains relatively close to the historical average.

...

"If the economy a year from now is the same as it is today, or roughly so, then the power of incumbency is strong and Trump's election odds are very good, particularly if Democrats aren't enthusiastic and don't get out to vote," the authors said. "It's about turnout."

I don't know if that proviso if voter turnout remains relatively close to the historical average is likely to be true. Seems to me 2020 will be abnormally passionate and have an abnormally high turnout.

But that doesn't necessarily cut against Trump too badly, as the forecast goes on to say.

Moody's forecast is here.

They explain that their erroneous prediction of a Clinton win was based on a 2-year graph of low gas prices, but, had they instead used a one-year frame....

Beginning in 2014, gasoline prices experienced their largest two-year decline leading up to a presidential election. Historically, two-year declines in gasoline prices have a strong statistical relationship with incumbent parties maintaining control of the White House. Therefore, we used the two-year decline in gasoline prices as an independent variable in the 2016 election model, and it was enough to offset many other explanatory variables that were working against Clinton at the time. However, if we had shortened the time frame for the decline in gasoline prices from two years to one year, the 2016 model would have instead predicted a Trump win. This owed, at least in part, to the timing of the decline in gasoline prices. Though the two-year drop was the largest leading up to an election, most of the decline occurred in 2014 and early 2015 (see Chart 2). This meant that the price decline in the 12 months before the 2016 election was barely noticeable, providing little boost to the then-incumbent Democratic Party.

Gas prices dropped a lot two years before 2016, but almost not at all since 2015. So the public was not feeling great about gas prices.

It's very easy to adjust a "forecasting" model post hoc and change your variables to retroactively "predict" past events you already know happened. Still, that's their explanation.

As to turnout: in an election with turnout near the historical average, Trump wins with 322 votes.

But even assuming maximum turnout seen in past elections (that is, the highest turnout one could reasonably expect), the Democrats are predicted to win -- but only barely. That projection suggests that Democrats would just barely cross the 270 threshold with 279, and Trump would lose with 259.

Even with maximum historical turnout, the Democrats can't afford to have any states wind up more Trump-leaning than average, or else they'll lose their slight edge.

Meanwhile, in a low turnout election (something I'm pretty sure will not happen), Trump wins with 380 electoral votes.

This is probably premature, but we get so much bad polling news, I thought it's maybe worth pointing out that a model based on economic modeling projects a likely Trump hold.

posted by Ace of Spades at 11:55 AM

| Access Comments