Search This Blog

De Omnibus Dubitandum - Lux Veritas

Monday, October 21, 2019

Economics and Election Predictions

October 18, 2019 by Dan Mitchell @ International Liberty

There’s an entire field of economics called “public choice” that analyzes the (largely perverse) incentive structures of politicians and bureaucrats.

But is economic analysis also helpful to understand voting and elections?

In the past, I’ve suggested that political betting markets are a useful place to start since “you are seeing estimates based on people defending their views with cold, hard cash.”

In his Bloomberg column, Professor Tyler Cowen takes a more rigorous look at the potential insights of political betting markets.
Prediction markets…are a quick way to get an overview of the state of the campaign. President Donald Trump is currently at about 0.40 to be re-elected… Under normal assumptions about the uncertainty of future economic growth, the markets rate Trump’s chances of winning at 40%. …it is a useful corrective to the argument that Trump is toast — or, alternatively, that he is a shoo-in.   
The market incorporates the relevant uncertainties in both directions. (Interestingly, Trump’s re-election odds have stayed pretty steady over the last week or so of negative news.) In many cases, prediction markets…“see through” the day-to-day volatility that may buffet the polls but not affect the final outcome. …Prediction markets…also made me think that a possible Hillary Clinton candidacy…is perhaps an undercovered story. …It is not a valid criticism of prediction markets to say that they didn’t predict Trump, say, or Brexit. The purpose of prediction markets is not to foresee particular upsets.  
They can, however, tell you in advance what would be an upset — much like probability theory can tell you that getting three heads in a row is unlikely but is of no help in predicting exactly when it will happen.
There are also people who build models that predict elections based largely on economic factors.

The Washington Post just published a very interesting review of how three of these models show Trump comfortably winning.
President Trump is on a fast track to an easy reelection. That’s the conclusion reached by economic forecasters… Moody’s Analytics projects the president will win handily next year if the economy doesn’t badly stumble — and in fact, rack up a greater margin in the electoral college than the 304-to-227 victory he secured against Hillary Clinton in 2016. … 
The finding jibes with those of other forecasting models that rely on measures of the economy’s strength to predict which major party’s candidate will win the White House next. Oxford Economics sees Trump winning 55 percent of the popular vote next year barring a “significant downturn” in the economy. … 
by the reckoning of the firm’s model, three key economic indicators — unemployment, inflation and real disposable income growth — all favor Trump’s reelection. They outweigh a “negative exhaustion factor” with Trump that dents his support in the projection. … 
Another model, assembled by Trend Macrolytics, accurately predicts every presidential victor back to 1952 by focusing on the effects of the economy and incumbency on the electoral college, according to Donald Luskin, the firm’s chief investment officer. It projects Trump will win reelection next year with 354 electoral votes — a margin that seems staggering on its face. Here’s the Moody’s electoral map, which doubtlessly will cause sleepless nights for the anti-Trump crowd.

Wow, not only do they show Trump winning every state he won in 2016, but they show him picking up New Hampshire, Virginia, and Minnesota.

So which approach is more accurate, betting markets of election models?

Given my inaccurate 2016 predictions, I’m probably not the right person to ask.

I’ll simply observe that both approaches have erred in the past.

And if you believe in guilt by association, some of the people who put together political prediction models also put together deeply flawed Keynesian economic models.

No comments:

Post a Comment