9

For a couple of weeks now, the betting market site PredictIt has pretty consistently shown Trump as having a low-double-digit chance of winning the election. However, I was under the impression that:

  • Vote counting is effectively finished
  • Although many states are close, their margins are large enough such that possible recounts would not affect the outcome
  • Although Republicans have initialized many legal cases, many of them have been dismissed, and there has not been sufficient evidence of voter fraud presented, nor could sufficient ballots be thrown out for the overall result to be overturned
  • Although electors are rarely faithless, electors are chosen by the party, and so are quite unlikely to vote against their party's (and likely their own) interests, especially when it has a chance of changing the outcome

It seems to me essentially certain that Biden has won. However, everyone participating in PredictIt has a strong financial incentive to make predictions that are as accurate as possible, and the crowd has been fluctuating on odds of Trump winning between 10% and 15%, rather than the 2% or so that I'd expect.

What am I missing? Some people appear to be seeing a path to victory somewhere, and have put significant amounts of money on it, but I don't see where it might lie. What, if any, are the unlikely but not utterly outlandish ways that Trump can still win the Electoral College?

3
  • 5
    Banks and credit rating agencies also have strong financial incentives to make accurate predictions, that's even their job. Yet they sometimes make costly mistakes. – Erwan Nov 28 '20 at 1:36
  • 5
    I think we should reopen because it is based on facts and does not seem like a clear attempt to discredit Trump. And if is theoretically possible. – Number File Nov 28 '20 at 17:34
  • Perhaps those who "know" that the election was rigged by illuminati reptiloids make up an unusually large part of the "wisdom" of the crowd – Hagen von Eitzen Nov 28 '20 at 22:40
15

The betting markets did a good job of predicting the state-by-state results of the 2020 election. In most cases they did better than polls. For example, Predictit strongly favored Trump to win Florida and Texas, while polls were showing that Biden was slightly ahead in Florida and had a chance in Texas.

However, everyone participating in PredictIt has a strong financial incentive to make predictions that are as accurate as possible[...]

But this is not really true. Prediction markets in general have a well documented favored-longshot bias, in which the odds of a low-probability event are systematically overestimated. Furthermore, these markets are distorted by fees, illiquidity, and regulation. Regarding fees, once a contract on Predictit goes as high in price as 0.95, it is no longer possible to deposit fresh money on the site, buy the contract, win, and take the money back out at a profit. Even for a contract priced as low as .87, by the time you work out all the fees and the income tax, it really isn't a good use of your time to put fresh money on the site and buy a dollar for 87 cents.

There is also a maximum number of bettors on each contract, and a limit imposed by the US government of $850 on the amount of money any individual bettor can bet on any contract. This is currently excluding fresh money from coming in to many of the bets. You can see this, for example, if you go to the Pennsylvania market, click on "read the full rules," and look at the bottom where it says that the contracts for this state have "reached the limit for the allowable number of traders."

There are also many stupid and/or crazy people on Predictit, as well as many people who are day trading rather than buying and holding a contract because they actually believe that the outcome will be a certain way.

What, if any, are the unlikely but not utterly outlandish ways that Trump can still win the Electoral College?

There are no such ways.

1

The betting markets have been poor predictors of results in this cycle, far worse than the polls. People just don't believe that Trump can lose, and while this belief persiste, there will be money bet on Trump. The odds there just reflect how we have come to distrust conventional political analysis.

There are substantial minorities that believe one of two things:

  • That the election was rigged, Trump actually won by a clear majority. The media reports are fake. And it is just a matter of time before the courts realise this and step in to prevent the illegal takeover by Joe Biden.
  • Or that Trump will use improper, and illegal, steps to prevent the President-Elect from succeeding him. He has appointed court justices that will do his bidding and despite the obvious fact that Biden won a clear majority of the state's electors, the puppet Justices on the supreme court will disqualify all of Biden's electors and a tame Congress will re-instate Trump.

Both ideas seem far fetched, but there are people who genuinely believe a version of one or other. It is perhaps because the election of Trump seemed (back in the Spring of 2016) to be such an unlikely outcome. He was a joke candidate, like K.West. We've got used to the impossible in politics. Why then is a rigged election or Presidential coup so unlikely(?)

3
  • 7
    "The betting markets have been poor predictors of results in this cycle, far worse than the polls." This is simply not true. Prediction markets correctly said that the 2020 election was going to be fairly close, which was more accurate than poll-based predictions of a blue wave. What the prediction markets have failed at, ludicrously, is "predicting" the result of the presidential election after it was over. – Ben Crowell Nov 28 '20 at 2:57
  • @BenCrowell Well, they haven't actually failed until January 20, 2021. But I agree it looks very likely they will. I think they may still think Trump could win because gamblers have a good understanding of the public's opinions, but a poor understanding of the law. – Ryan_L Nov 28 '20 at 16:25
  • 1
    Yes they did. They predicted 290 electoral votes for Biden and he got 306. – Number File Nov 28 '20 at 17:35

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .