Prediction: Political-futures markets clean up
.jpg)
Alan Boyle, a former colleague at msnbc.com, and creator and shepherd of the celebrated Cosmic Blog, wrote a column on Tuesday based on findings from the Iowa Electronic Markets (IEM), an experimental online-futures project of the University of Iowa’s Henry B. Tippie School of Business, something that’s mainly used for purposes of academic research and instruction.
To go by the IEM’s conclusion, it’s all over: “Political prediction markets suggest that the outcome of next week's congressional elections is settled, with Republicans taking control of the House and Democrats slightly favored to hold onto a majority of sorts in the Senate,” Boyle writes.
“That forecast follows the current conventional wisdom. The interesting part is how the markets arrive at that conclusion: Rather than relying on polls of registered or likely voters, the markets render their judgment based on real-money bets from hundreds of traders.”
.jpg){kind=small}
“The IEM has special dispensation from the Securities and Exchange Commission to run this kind of operation because it's regarded as a University of Iowa research project with the purpose of studying how real-money behavior plays out in non-traditional markets. ...”
Boyle also breaks down where the shares have been trading lately:
But while the IEM and other proposition futures markets are entirely legal, their conclusions aren’t any more substantive, no more illuminating about the mood of the country, than a simple garden-variety bet on who’ll win based on who’s winning (or who seems to be winning) at a given moment. In that respect, it’s not much more than entertainment; they should make it a game at the casino down the interstate from where you live.
◊ ◊ ◊
That’s partly because of a wide universe of traders, investors — you and me and everyone we pass on the street. Real-money bets from traders around the country may be a benefit in terms of generating the funding sources needed to make such predictive modeling possible, certainly diverting, and maybe even lucrative. But without drilling down into the identity of those traders, such wide-open political futures have to yield questionable forecasts. It’s a given of regression analysis — the cornerstone of modern statistics — that the sample should be representative of a wider population for any meaningful prediction. “A basic condition in almost all inferential statistics is that a set of data constitutes a random sample from a given homogeneous population,” says Hossein Arsham, a leading professor in business statistics, decision science and systems simulation at the University of Baltimore.
By that yardstick, any measure of the day-by-day trend of one proposition over another would seem to be compromised from the start. There’s no way, apparently, of knowing or caring if IEM traders are registered or likely voters, Democratic or Republican voters, women or undecideds, independents or Tea Partiers; without that basic information, such an election forecast amounts to a wish list of undifferentiated millions, a measure that may or may not have any relation to reality as a barometer of expectation.
◊ ◊ ◊
Without some idea of who the universe of traders is composed of, or what percentage of those traders who say they intend to vote, what‘s the prediction based on in the first place? If a nonvoter (or hundreds of them) buys lots of shares in the proposition that the GOP takes the House, other traders will be buying into that proposition based, in part, of the feelings of that nonvoter — someone whose interest in the election outcome is purely financial, someone whose ability to affect the election doesn't exist.
It’s an impure, incomplete sample of the public, like polls that rely on hardline telephone contacts with respondents but fail to account for a growing universe of cell-phone users. It fails to make a necessary distinction between nonvoters and voters — between those betting on the outcome and those who’re investing in the outcome.
◊ ◊ ◊
“In the political sphere,” Boyle writes, “the IEM has done at least as well as the more traditional opinion surveys at predicting election outcomes. The reason has to do not only with the oft-cited "wisdom of crowds," but also with the fact that the potential payoff entices knowledgeable traders to swoop in on what they perceive as a good deal. That produces a market that quickly reflects the latest line on how the contest will turn out. Like the Vegas line on next weekend's big game, the result is authoritative but not foolproof.”
At the halftime of Super Bowl XLIV, the smart money wouldn’t have bet on the New Orleans Saints opening the second half with an onside kick that thoroughly flummoxed the heavily-favored Indianapolis Colts and set the stage for the Saints winning the NFL Championship.
At the start of the 2010 football season, the smart money wouldn’t have bet on the Dallas Cowboys being 1-5 approaching the halfway point.
At the start of last week, the smart money didn’t have Brett Favre getting a hairline fracture of his left ankle, possibly sidelining him for weeks, and hobbling the rest of the season for the Minnesota Vikings.
Never mind authoritative; there’s a reason why you actually play the game. Shit happens. The wisdom of crowds is often indecisive and easily seduced. You have to wonder what the IEM and other online futures operations will make of the new variable, the latest Associated Press-GfK poll, finding that 1 in 3 Americans has not yet decided how they intend to vote.
◊ ◊ ◊
But of course that poll doesn’t matter if you’ve already voted; how do predictive markets take into account the millions of Americans who’ve cast their ballots by mail, early, days if not weeks ago? How can political futures markets be accurate when they don’t allow for — and can’t allow for, based on the business model — what’s already happened? How reliable is the information they’re responding to?
As represented by the IEM illustration above, the political futures market (like all markets) is a dynamic phenomenon, it’s serial, it’s ongoing, it’s responsive, it’s based on people buying proposition shares in something close to real time. If millions of absentee ballots have already been cast weeks or months ago, and more are being mailed in while proposition shares are being purchased based on more immediate events (or hunches), the markets’ predictive process is at odds with the reality it hopes to
Data on historical voter behavior won’t help; the X factor that could prove or disprove the proposition is literally in the mail, or safely locked away until Election Night. It’s information that’s not historical enough to be accessible. Stock traders are exchanging the certainty of favorable reaction to a proposition for a game-changing unknown: the ballots of early voters whose decision may have already derailed the proposition.
◊ ◊ ◊
The IEM itself is an intriguing natural application of science, statistics, economics and politics, and deserves attention as a smart diversion. But still. Is it just me or do you get a slightly queasy feeling when considering how, how much and how long the national political mood has been the substance of a stock ticker, fully commoditized like orange juice and pork bellies?
Research is research, though. As a snapshot of the mood of the nation, the IEM may well have its greatest value in being what it truly is: a snapshot of this nation, one of any number of possibly meaningful metrics that tell us who and what and where we are.
There may be a more informative and more telling gauge, another tale of the economy, a kind of meta-metric: One that distinguishes between those who have the money to place such wagers in the first place ... and those who don’t.
Image credits: Muhammad Ali: Source unknown. Koch Industries logo: © 2010 Koch Industries Inc. All other images: Iowa Electronic Markets, Henry B. Tippie School of Business, University of Iowa.
.jpg)
Comments
Post a Comment