The U.S. presidential election campaign is now in the home stretch. Yet I remain confused by the welter of polls and their many interpretations. I read a quote in a recent Newseek magazine: Another day, another 37 new statewide polls.1 Do the polls predictably measure voting behavior? Do we really care about national polls when national elections are resolved state by state? What adjustments must be made to polling figures to remove their bias? Can landline telephone polling reach the critical Millennium Generation, many of whom use only cell phones? Its a bit overwhelming for me. I think Ill take the maverick path: buy a six-pack, watch a hockey game - and look to the Iowa Election Markets (IEM) for guidance.2
Citing the excellent book The Wisdom of Crowds, by James Surowiecki, the OpenBI Forum mentioned the IEM as an illustration of prediction markets in Part 1 of the BI Ensemble series.3 The IEM, established by the Tippie School of Business at the University of Iowa, allows paying participants to buy and sell future contracts on candidates to either predict the winners of elections or the vote share per candidate - and has proven remarkably prescient in forecasting national elections. Alas, the current IEM presidential market does not bode well for Senator McCain. As of October 22, Senator Obama leads on both the election winner and vote-share markets, and the disparity has increased post Republican convention. Indeed, the bid prices on the winner take all presidential market are $0.868 for Obama and $0.127 for McCain, which translate to confidence in Obamas prospects in excess of 6 to 1.
The insightful article Discovering and Managing New Product Blockbusters: The Magic and Science of Prediction Markets from the Haas Business School at the University of California, Berkeley, sheds light on the workings of prediction markets. The authors, one a Haas professor and the other an HP analyst, address the problem of failed product launches, noting the inadequacies of traditional approaches to demand forecast, such as target customer surveys, pooled expert opinion and company meetings. As an inexpensive but reliable replacement, they propose prediction markets to harness the wisdom of crowds.4 Key to the success of such markets is a learning platform that compensates for individual prediction limitations by aggregating information from the active participation of large numbers of public, diverse and independent thinkers. The authors detail their prediction market framework, cleverly named I4C, which rewards solely information using price as an indicator that aggregates opinions of all participants. The market encourages individuals to improve their knowledge, learning from others and subsequently making the markets smarter. The markets ultimately assure success by pooling independent information sources from large crowds to promote liquidity.
The Harvard Business School article Prediction Markets at Google elaborates on the Haas paper by chronicling the development of prediction markets at Google. Google of course has a history of exploiting crowd wisdom with its industry-defining PageRank algorithm for search. And so with corridors of very smart employees and a rule that promotes 20 percent time allocation to innovation projects, its hardly surprising that Google has become a leader in prediction markets. With Google Prediction Markets (GPM), participants buy and sell decision securities using an allocation of Goobles, which are replenished quarterly. Individual markets that forecast product success, product launch date or competitor activities generally remain active for a single quarter only, at which time cash and other prizes are awarded. As might be expected, wily Google traders have written bots to take advantage of arbitrage opportunities in the evolving markets. While overall internal reception to prediction markets is somewhat mixed, the evaluation of market results is quite gratifying. Price appears to be an excellent proxy for the probability that a given decision will win in a market. In addition, the most expensive stock decisions overwhelmingly dominate as winners, and that domination increases as the markets progress in time to the end. 5
Prediction markets have become visible in hallowed business publications as well. The Wall Street Journal narrated the successful prediction market experience of Best Buy in a recent article. TagTrade, the Best Buy program, trades imaginary stocks based on answers to company manager questions involving forecasts and the success of ongoing or prospective initiatives. The company has focused attention on the organizational side of prediction markets, going to great lengths to assure that their experiments are non-threatening and encourage substantial, diverse and independent participation. Best Buy CEO, Brad Anderson, thinks narrowing the gap between management and workers has helped make the company more nimble and responsive, elevating sales and profits. Researchers from the University of Chicago have been enlisted to help evaluate their prediction market program, comparing market forecasts with official company predictions.6
Not to be outdone, the October 2008 Harvard Business Review opines on the wisdom of crowds for the advancement of philanthropy. With billions of dollars now inefficiently allocated across 1.5 million organizations, author Steven Goldberg conjectures that prediction markets could shed light on which non-profits are providing the highest social ROI, creating virtual stocks out of hypothetical questions related to organization performance, such as how many new charter schools would be built and running in a given time period.7 Ideally, the markets would create consensus judgments on the relative success of organizations competing for scarce funding dollars. One conjectured benefit of prediction markets in this space is agency obsession with evaluating and reporting performance data that can educate prospective philanthropists.