Bet the Election

Bet the Election
Townhall.com ^ | October 21, 2020 | John Stossel

Posted on 10/21/2020 3:26:13 AM PDT by Kaslin

Donald Trump will probably lose the election.

As I write, The Economist says he has only an 8 percent chance of winning.

Nate Silver's FiveThirtyEight, which came closest to predicting Trump's win in 2016 and has the best track record among modelers, gives Trump just a 12 percent chance.

But people who "put money where their mouths are" give Trump a better chance: 37 percent.

That's according to roughly that often., the website I created with Maxim Lott. It tracks multiple betting sites around the world.

Though 61 to 37 percent seems like a giant lead for Joe Biden, 37 percent means Trump is likely to win one-third of the time.

Four years ago, most bettors were wrong about Trump and Brexit. I assume they learned from that and adjusted their 2020 bets.

But since bettors were wrong in 2016, why trust betting odds now?

Because betting is a better predictor than polls, pundits, statistical models and everything else.

ElectionBettingOdds.com has tracked hundreds of races. When bettors think a candidate has a 37 percent chance -- they really do win roughly that often.

A research scientist at Amazon concluded that in the last presidential election, ElectionBettingOdds.com beat all other existing public prediction models except for Nate Silver's polls-plus model.

Silver says: "Betting markets are populated by people with a sophomoric knowledge of politics... Traders are emotionally invested in political outcomes." Also, "Markets (are) not super liquid... way different than sports where you have a much more sophisticated player base and more liquidity."

But our site takes odds from betting sites in Europe, the U.S. and a cryptocurrency-based exchange. More than $200 million has been bet.

As Silver says in his excellent book, "The Signal and the Noise," "A lot of smart people have failed miserably when they thought they could beat the market."

Overall, bettors have the best track record. Last election, The New York Times' "expert model" had Hillary Clinton ahead 85 percent to 15 percent. The Princeton Election Consortium gave Clinton a 99 percent chance. (Now they give Biden 98.2 percent.)

Daily Kos had Clinton at 92 percent. Huffington Post had 98 percent. Those two stopped operating after that embarrassment.

Silver is one modeler who's often beaten the market. In 2016, he gave Trump the highest odds, and in 2018, he was the most confident that Democrats would win the House.

On the other hand, his FiveThirtyEight model was confident Democrats would win Florida's and Indiana's Senate races, making Democrats 70% favorites in both states. But Republicans won. Bettors were closer to predicting the actual results.

Bettors do well because they consider many things not easily captured by polls and statistical models.

How many mail-in ballots do not get counted? In the New York state primary this year, 20 percent were disqualified for irregularities.

FiveThirtyEight "built in an extra layer of uncertainty this year because of the possibility that the pandemic will disrupt usual turnout patterns." But bettors believe it's not enough.

Bettors also consider the possibility that polls are wrong in some new way.

In 2016, polls showed Clinton well ahead in Wisconsin, Michigan and Pennsylvania, but pollsters hadn't questioned enough voters without college degrees. Who knows what mistakes pollsters are making now?