How manipulation-resistant are Prediction Markets?
Our Undertaking in Empirical Cryptoeconomics
It often is after-the-fact that we realize how much harm results from a decision made by key organizations such as nations, firms, cities, or clubs. Before such a decision was made, there were certainly people who had a deep understanding about its consequences and therefore good reasons to disapprove it. However, these relevant experts were not enticed enough to share their knowledge with properly-motivated decision makers, nor were non-experts induced to learn that these decisions are inferior. Most importantly, decision-makers ultimately are not held accountable if the decision turns out to not have the consequences they promised.
The Failure of our Information Institutions
An important contribution to the implementation of inefficient decisions is that our information institutions, i.e. public relations teams, organized interest groups, news media, conversation forums, think tanks, universities, journals, elite committees, or state agencies fail to induce people to acquire and share relevant information.
Robin Hanson, the inventor of Futarchy, father of modern prediction markets, and currently Professor of Economics at George Mason University, therefore argues that we should consider augmenting our (political) information institutions with speculative market institutions. Since speculative markets excel at inducing people to acquire information, share it via trades, and aggregate that information into consensus prices that convince wider audiences, they seem to be ideal information institutions.
Futarchy as the new Form of Decision-making
Inspired by the information-aggregation power of speculative markets, Hanson developed the concept of Futarchy — a form of governance where only those policies become law for which speculative markets have clearly estimated that they would increase national welfare.
The prices of speculative markets could estimate a national welfare metric (such as GDP) conditional on a proposed policy being adopted, and on that policy not being adopted. This would be possible through called-off trades (i.e. which are made null and void if the proposed policy is being adopted (or not)) in assets that pay in proportion to the measured national welfare. Finally, the policy would only be adopted if the market expects that policy to increase national welfare relative to the status quo.
This also means that with Futarchy, decision-makers (i.e. the market participants) are held accountable if the predicted consequence (the national welfare metric estimate f.ex.) would not come true: they would lose their money. To impact a speculative market, you literally have to put your money where your mouth is.
Futarchy can be applied as a form of decision-making to many kinds of organizations. The organization would simply need to define a publicly verifiable metric, and their decision-making would then depend on the outcome of a pair of prediction markets on that metric.
As Vitalik Buterin described in a blog post back in 2014, the first market would be denominated in a currency which pays out $1 if a company makes decision A and $0 otherwise. The second market would be denominated in a currency which pays out $1 if a company takes decision B and $0 otherwise.
Therefore, the first market only has a value if its condition is met, i.e. if decision A is made. Similarly, the second market only has a value if decision B is made. If governed by Futarchy, only the decision for which the market has estimated a higher price will be made.
To illustrate the application of Futarchy in a corporate environment, let’s assume that a company sets up two prediction markets on its revenues for the next 5 years conditional on hiring a given CEO A or a given CEO B by the end of the month.
At decision time (f.ex. at the end of the month), the expected value of the second market B (if CEO B is hired) is clearly higher than the expected value of the first market A (if CEO A is hired). Thus, CEO B is hired.
Whether the winning outcome is chosen by the integral over the entire market length, the integral over the past day, or a weighted integral over the past week doesn’t matter as long as the decision metric is specified in the smart contract beforehand.
CEO B is the winner of the conditional prediction market. Market participants who bet on CEO A get their investment refunded: Since CEO A has not been hired, CEO A tokens do not contain any value.
Those who bet on CEO B, however, are still participating in the prediction market which is just no longer a conditional market — the hiring decision has been made.
Who wins and who loses at the end of the prediction market, at time X? Those market participants who bet on the losing decision (CEO A) make neither profit nor loss. Those who bet on the winning decision (CEO B) and bought CEO B tokens at a price higher than the market price at time X turn a loss, while those who bought CEO B tokens at a lower price than the market price at time X make a profit. This is at least true for the long tokens that predict a higher revenue. To every long bet there needs to be a counterpart—a short bet—for which the payout structure is just the opposite.
How manipulation-resistant are Prediction Markets?
If decisions are made based on prediction market outcomes, we need to be absolutely sure that our markets resist attempts to influence a decision via distorted participation.
Speculative markets have shown to resist manipulation particularly well. This is due to the way they handle noise trading, i.e. trading for reasons other than information about common asset values. Manipulators trading in order to distort prices are in fact noise traders since they trade for reasons other than asset value information.
Hanson explains that traders, when they can’t predict noise trading exactly, compensate for its expected average by an opposite average trade, and then compensate for its expected variation by trading more, and by trying harder to find the relevant information. In fact, Theory states that, given that trader risk-aversion is mild and more effort results in more information, increased noise trading actually increases price accuracy.
5 Cryptoeconomic Experiments
We would like to empirically prove that adding manipulators (or noisy traders) to a prediction market built on the Ethereum infrastructure does not reduce average price accuracy, and that prediction markets will therefore be among the most manipulation-resistant.
With support from the Ethereum Foundation, we’ll be running the five experiments proposed by Vitalik Buterin to test the viability of Futarchy.
A detailed overview of the different experiments will follow soon, so stay tuned! Meanwhile, you can start thinking about manipulation strategies and join the conversation :)
Will you be able to beat the market?
