bettingmarket
2020 US Election: Political Betting Markets Massively Overreact To Early Counting
Political betting markets (aka prediction markets) are supposedly predictive because of such widely embraced notions as "skin in the game" and "the wisdom of crowds." When the money is down and the crowd comes out to play, so the logic goes, these markets outperform professional forecasters and polls. That is the theory.
Contrary to the popular narrative, prediction markets do not incorporate new information in a rational bayesian manner - infact, they behave like in-running horse racing betting markets being prone to massive overreaction to short term news, displaying excessive volatility and bizarre swings away from the correct probability distribution. Those that trade in these markets are prone to all of the normal biases, and succumb often to ideological motivated reasoning.
In the 2020 US Presidential Election the political prediction markets ascribed to Trump an implied probability of 80% of winning the election on the back of the early counting, even though it was widely known that the late counted postal votes in the states that were actually going to decide the election would significantly favour Joe Biden - and in fact in every instance the outstanding mail ballots in these states did lean heavily towards the Democrats, supporting the notion that Democrats were more likely to vote by mail. In Philadelphia, for example, the mail ballots in some counties were reported to be going to Biden by more than a 9 to 1 ratio.
When the news broke that Trump was leading in Florida - a wave of fear engulfed the political betting market. Panic was the order of the day - had the polls got it wrong again. Positions were unwound, people hedged, some closed out - arbitrageurs arbitraged - vultures swooped. There was nothing rational about it - nothing textbook - this was chaos - political betting market style. There was no predicting going on - people were running to save themselves - fear eats the soul. The simple playing out of miscalibrated priors, unanchored anchors and inadequate models. Those that were following the data new better - there was nothing in the Florida numbers to suggest that Trump was going to win the election as was made clear by FiveThirtyEight’s election results simulator. (Plenty of wishful thinking, and reality denial at work alonside many examples of the the Dunning-Kruger effect and the familiarity bias: "after Florida it was 2016 all over again ..it wasn't.. but it serves as a great piece of post rationalisation for those that either stormed into backing Trump or cashed out their Biden positions.).)
The excessive volatility, and dramatic belief changes displayed in the political betting markets once the votes start to trickle through from the counts are essentially a product of people having skin in the game; they panic, close out opened positions; change their minds; herd; open new positions; arbitrage etc.. Prior to the commencement of the counting the betting/prediction markets had not actually bought into the notion that Biden was heading for a landslide - the implied probability of a Trump win was 39% on Betfair compared with 10% with Nate Silver. Following Trump's early victories in Florida, Iowa, Texas and Ohio panic set in and many people rushed to close down their positions on Biden. It was only when reality set in and people realised that the early results had done nothing to undermine Biden's chances in Pennsylvania, Michigan, Wisconsin and Arizona that the ship steadied and the betting started to move back in favour of Biden.
The notion of a tight race between Biden and Trump was a delusion brought about by the order in which the votes were counted - nonetheless many bought into it. The human needle was however unmoved.
The behaviour of the Betfair Political Betting Market on the 2020 US Presidential Election, on the first day of vote counting, supports the notion that when the outcome of a future event becomes more unclear, traders will overreact to any new information that further raises the spectre that they have miscalibrated their priors, in a manner that will induce excess volatility.
@FiveThirtyEight's last forecast said that there was a 10% chance of Trump winning and that even a 2016-style polling error wouldn’t be enough for him to do so.."As of right now, Biden’s lead is large enough that Trump’s chances of winning are 10 percent, considerably lower than the 35 percent chance he had at this point in 2016." The Betfair Betting Exchange Political Betting Market at the same time had the implied probability of Trump winning at 39%.
| Outcome | Implied Probability Trump Winds |
|---|---|
| FiveThirtyEight | 10% |
| Betfair Exchange | 39% |
We will leave the last word to Nate Silver.
The author Tim Harford in his recent book How To Make The World Add Up stated that "naïve realism can lead us badly astray when we confuse our personal perspective on the world with some universal truth." The 2020 US Presidential Election represented another failure for the skin in the game and wisdom of crowds merchants.
Show me a man who claims he is objective and I'll show you a man with illusions. Henry R. Luce
To cite this article: Niall O'Connor 2020 US Election: Political Betting Markets Massively Overreact To Early Counting.
(Published on Bettingmarket.com 05/11/2020 . Cognitive Flexibity = Greater Plasticity. All Rights Reserved.)