Polls say Biden. BetFair says Trump. Which source do you trust?

Last time both polls and betting markets failed to predict Trump. Which source will get it right this time?
Author:
Publish date:

The website 538, which aggregates leading polls, show Biden being favored in the 2020 election by somewhere between 4-7 points in recent polls. Some of these polls even show Biden taking Texas. Based on these polling results, some commentators believe that Trump has no chance of winning in November.

On the other hand, betting markets are showing a Trump victory. The website electionbettingodds.org, which aggregates bets made at FTX.com, a cryptocurrency betting site, and BetFair, a British betting site, show Trump with a 48.9% chance of winning and Biden at only a 42% chance of winning (Clinton at 2.8%, Cuomo at 1.5%, and others make up the remainder). More than $38 million have been bet on the presidential election. Should we trust these betting sites more than we trust the polls?

Since George Gallup created the Gallup polling organization in 1935, and successfully predicted FDR's victory in the 1936 election, most Americans have looked to polling data for insight into who is likely to win elections. Gallup failed to predict the president correctly in 1948, 1976, 2004 and 2016, the popular vote in 2000, so by no means has their track record been perfect.

In recent decades, several economists have made the case that prediction markets or betting markets are likely to be more accurate than are polls, both because bettors have "skin in the game" and because people may not be honest when asked by a pollster. In an article summarizing the failure of polls to predict Trump's 2016 victory, Pew Research notes,

"Across the board, polls underestimated Trump’s level of support. With few exceptions, the final round of public polling showed Clinton with a lead of 1 to 7 percentage points in the national popular vote."

Pew goes on to explain the failure to predict accurately as likely due to "nonresponse bias," suggesting that Trump supporters were less likely to respond to a poll, and "shy Trumpers," poll respondents who did not admit to their preference for Trump.

If these explanations are accurate, then it is possible that polls are once again underestimating support for Trump.

The comparison between polls and prediction markets or betting markets is complicated by the fact that both major prediction markets, such as predictit.org, as well as the leading betting market, BetFair, also failed to predict a Trump victory in 2016. Indeed, BetFair was showing heavy odds favoring Clinton even as election results began to come in.

Meanwhile, Predictit.org, an academic prediction market limiting the amount of event bets to $850, is currently predicting that Democrats will win the 2020 election, 53% to 49%. On the other hand, BetFair allows bets of any amount provided that the payout does not exceed 1 million pounds. Insofar as a large, liquid market with no limits on betting is likely to be more accurate, we should regard BetFair as more likely to be accurate than is Predictit.org.

Finally, HyperMind, which claims to aggregate the opinions of hundreds of "champion forecasters," predicts the probability of a Democratic victory at 52.9% and a Republican one at 46.1%. Do we trust the aggregated opinions of these "champion forecasters" over the bettors at BetFair?

Whether you believe that Biden is more likely to win due to polls, Predictit.org, or HyperMind, there is an opportunity to monetize your conviction over at BetFair.