Prediction Markets and the Donald
For the 2012 Presidential election the Intrade prediction market correctly forecasted outcome of almost every state (Flordia was expected to vote Romney) [link]. However just before the election started, a Trump victory was priced at $0.22 on Predictit.com and $0.258 on the Iowa Electronic Markets. While the fact that an event with a proablity of ~1/4, actually occurred is not surprising, this is the second major political vote that prediction markets failed to get right. The other voting being the Brexit vote where a vote to leave was price at $0.26 on Predictit.com the day before the vote.
Unfortunately like all markets, prediction markets also fall prey to information cascades. As each bettor in a prediction market set’s their price based off of the information that they have (Their private signals) and the going prediction of the market (the public declaration of other bettors), thus setting up a system ripe for an information cascade to occur. Further compounding the problem has been the recent tenancy for “all sorts of experts have assured us capture the wisdom of crowds,” which in turn has encouraged people treat the market price as the best estimate the correct probabilities [2]. Which in turn has lead to “Traders … not updating enough based on outside information.” [2] In short a full blow information cascade has occurred as the primary source of information has become the market itself.
Unfortunately it doesn’t appear much can be done to correct this. Prediction market rely on low risk bets to reduce barriers to entry, which has the side effect of the “total payouts from prediction markets are too low to create a strong incentive for participants to work really hard to become substantially better-informed” [1]. While this barrier could easily be raised at present the problem appears to also results from “key players aren’t [being] represented” (ie Rural White Rupublicans). Thus while a higher payoff would increase the liklyhood that bettors would activly seek out new information, it may also eliminate group of bettors who ultimatly will decided via thier votes the election results.
[1] https://www.bloomberg.com/view/articles/2016-11-15/prediction-markets-didn-t-call-trump-s-win-either