top of page

The use of prediction markets to make better forecasts

Updated: Jan 24, 2022

Revealed Preferences and the 2020 US election

Revealed preferences are a way to use "The wisdom of the crowds" to make high-quality forecasting decisions. This approach will work, especially when the individual participants can make well-informed and independent decisions. It is also important the participants have a uniform and forecast aligned decision motivation.

Prediction markets are a great method to reveal preferences. Prediction markets allow individuals to buy and sell shares of some commodity, based on their belief of price at some designated future point. We will use the 2020 U.S. presidential elections as an example.

First, a word about revealed preferences. Revealed preference theory, pioneered by economist Paul Samuelson, is a method of analyzing choices made by individuals, mostly used for comparing the influence of policies on consumer behavior. Revealed preference models assume that the preferences of consumers can be revealed by their purchasing habits.

Those that have read my other articles, especially on risk, know that I take exception to some of Dr. Samuelson’s methodology. In this case, though, he got it right. Think of revealed preferences as people “voting with their feet” or “voting with their pocketbook.” In other words, according to advocates of revealed preference theory "It is not what you say, it is what you do that reveals what you want." An example of a securities market whose goal is to reveal political preferences is called the Iowa Electronic Markets (IEM).

I first started tracking the IEM in 1992, the election Bill Clinton won. At that time, it was called the Iowa Political Stockmarket and was used as a laboratory to study revealed preferences. I was an economics graduate student and the IEM was an economics laboratory. Back then, the students were the subjects to help the professors research revealed preferences. I have fond memories of trading shares of Clinton, Bush, and Perot. Today, the market is larger and more liquid. The IEM is still focused on revealed preferences. Based on the earlier rule of thumb, the IEM enables:

* Individual participants can make well-informed decisions - Participants have access to all public election news sources.

* Individual participants can make independent decisions - Participants are generally physically separate and trade anonymously.

* Individual participants have a uniform and forecast aligned decision motivation - All participants are attempting to maximize return. They wish to buy low and sell high. They have predefined forecast time frames, just prior to election day.

Pre-election observations

I write this on the night before the 2020 US Election Day. By the end of the day tomorrow we should have a good sense of who the next president will be, along with all members of the US House of Representatives and ⅓ of the US Senators.

So what does the IEM indicate about election outcomes the night before the election? The aggregated revealed preferences of the traders suggest a Biden victory, with about a 55% popular vote. Below you can see historical closing prices, through the market close last night. In this “voter share” market, traders get paid out on the actual percentage of the official popular vote. So this market should settle on prices based upon traders' “at-risk” popular vote beliefs. As such, payout incentives are aligned with the popular vote percentage.

Also, below is the “winner take all” market. This is different from the popular vote market as a single share will payout 100% to the holder of the victor’s shares and 0% to the holder of the loser’s shares. I interpret this market as 1) a form of statistical error measurement relevant to the popular vote market, and 2) resolving for the differences between the popular vote and the electoral college. As such, payout incentives are aligned with the election winner.

Therefore and based on the IEM, there is a 90% chance of a Biden victory, given Biden’s popular vote victory.

We will see! Also, it is worth looking at the other markets for congressional outcomes. See the IEM site.

Two weeks after the election

At this point, Biden has been declared the winner, having enough electoral votes to clinch. The popular vote was closer than the pre-election IEM close predicted. While there are still a few votes to be counted, Biden ended up with 51.7%, compared to the 55% predicted from the pre-election close.

The winner-take-all market got it right, given the Biden victory. I infer the difference in the popular vote relates to the error as suggested for the winner-take-all market. That is, winner-take-all would have been priced Biden at 100% if there was no predicted error possible in the voter share market. Though, this doesn’t account for the electoral college reconciliation.


The Iowa Electronic Market is a great example of revealed preference theory in action. The IEM successfully forecasted the 2020 U.S. election outcome. Markets like this may be useful to help forecast other outcomes, as well as provide a decision-making method for complex organizational decision-making.

13 views0 comments


bottom of page