Thursday, July 10, 2008

Polls Vs Prediction Markets

With elections around the corner in India, both political parties and candidates are naturally eager to assess their respective electoral prospects. The media feed this appetite by indulging in a frenzy of polling as the election day draws closer. News papers and news channels compete with each other in publishing opinion polls and also claiming greater credibility based on selective interpretation of thier previous poll results and the final election verdict.

These opinion polls survey the preferences of a representative enough sample (and this is most often a subject of debate and controversy) of voters and a reasonably large enough number of voters. The voters make their choices based on their present perceptions of the respective parties or candidates. This in turn is generally based on their assessment of the past performance or actions of the political parties and candidates. Opinion polls are therefore largely a reflection of the past performance of the candidates.

Justin Wolfers argues that Prediction Markets are a better indicator of the actual result than opinion polls because, unlike the latter, the former makes its assessments based on the expectations of the future performance. He writes that prediction market traders even take cue from the opinion poll results and change their predictions assuming that the contestants respond to their poll results by changing strategies.

Prof Wolfers quotes the works of Robert Erikson and Christopher Wlezien, who find that opinion poll results tend to be skewed by time inconsistency problems, and therefore needs to be discounted for. The initial advantages or leads generally tend to get dissipated over time, and the voting public tends to be more strongly anti-incumbent three-and-a-half years into an administration than they are on Election Day. Further, polls are slow to reflect economic conditions.

Prediction markets are "information markets" or "speculative markets created for the purpose of making predictions" where the participants trade in contracts whose pay-offs depend on outcomes of unknown or uncertain future events. It is based on one of the basic maxims of classical economics - in a truly efficient market the market price will be the best predictor of an unknown future event. James Suroweicki had claimed that prediction markets are a more reflective statement of the reality as it represents the "wisdom of the crowds".

In the most basic prediction markets, the "winner-takes-all" market, the contract costs, say Rs 5, and pays off, say Rs 25, if and only if a specific event occurs. The price on a winner take all market represents the markets expectation of the probability that an event will occur.

In an "index contract", the payoff varies in a continuous way based on a number that rises or falls, like the percentage of vote received by a candidate. The price of such a contract represents the mean value that the market assigns to the outcome. In "spread betting", trades differentiate themselves by bidding on the cutoff that determines whether an event occurs or not, like a candidate receives more than a certain percentage of the popular vote.

The Iowa Electronic Markets, run by the University of Iowa Tippie College of Business, is one of the most successful prediction market in the world and has predicted many US Presidential elections. Our own informal and illegal "satta" markets are another example of a crude prediction market, though these are more one-off markets than continuing markets.

Given the coming elections and the amount of time available to make meaningful trades, it may be a good idea for some university or newspaper to run a prediction market on the election results. Besides, it will also throw up interesting statistics and information, which can be analyzed to better understand the preferences and motivations of the Indian voter.

No comments: