Thursday, January 17, 2008

previous entry | main | next entry | TrackBack (0)

Old-time prediction markets

In my latest Marketplace commentary, I pointed out that the accuracy of prediction markets would improve as they went more mainstream. Essentially, as markets widen and deepen, their informational efficiency should improve.

I had assumed that we would need to wait for the future for this to happen. However, Paul Rhode and Koleman Strumpf provide some fascinating evidence from the past in thieir paper, "Historical Presidential Betting Markets." The highlights:

This paper analyzes the large and often well organized markets for betting on presidential elections that operated between 1868 and 1940. Over $165 million (in 2002 dollars) was wagered in one election, and betting activity at times dominated transactions in the stock exchanges on Wall Street.

Drawing on an investigation of several thousand newspaper articles, we develop and analyze data on betting volumes and prices to address four main points. First, we show that the market did a remarkable job forecasting elections in an era before scientific polling. In only one case did the candidate clearly favored in the betting a month before Election Day lose, and even state-specific forecasts were quite accurate. This performance compares favorably with that of the Iowa Electronic Market (currently the only legal venue for election betting in the U.S.). Second, the market was fairly efficient, despite the limited information of participants and attempts to manipulate the odds by political parties and newspapers. Third, we argue political betting markets disappeared largely because of the rise of scientific polls and the increasing availability of other forms of gambling. Finally, we discuss lessons this experience provides for the present....

The extent of activity in the presidential betting markets of this time was astonishingly large. For brief periods, betting on political outcomes at the Curb Exchange in New York would exceed trading in stocks and bonds. Crowds formed in the financial district – on the Curb or in the lobby of the New York Stock Exchange— and brokers would call out bid and ask odds as if trading securities. In presidential races such as 1896, 1900, 1904, 1916, and 1924, the New York Times, Sun, and World provided nearly daily quotes from early October until Election Day....

In the 15 elections between 1884 and 1940, the mid-October betting favorite won 11 times (73 percent) and the underdog won only once (when in 1916 Wilson upset Hughes on the West Coast). In the remaining three contests (1884-92), the odds were essentially even throughout and the races very close. The capacity of the betting markets to aggregate information is all the more remarkable given the absence of scientific polls before the mid-1930s. The betting odds possessed much better predictive power than other generally available information. Moreover, the betting market was not succeeding by just picking one party or by picking incumbents. Over this period, Republicans won eight of the elections in the Electoral College and Democrats seven; the party in power won eight, the opposition seven.

Hat tip: The Monkey Cage's John Sides.

posted by Dan on 01.17.08 at 05:58 PM


and yet, and yet... even perfectly efficient, rational prediction markets can only process the information that is actually out there. The most efficient markets out there can't reduce the inherent uncertainties of life; they are not soothsayers.

We've witnessed a great example of this over the last couple of weeks, in the reactions of the political prediction markets to the recent events in Democratic presidential primaries. I was watching them closely, to see whether they could add any real predictive value, and the answer was a resounding "no." Before Iowa, Hillary was the favorite; after Iowa, Obama shot up to become the favorite; after New Hampshire, Hillary shot back into the lead. The graphs of the candidates look like an EKG, and with good reason: the people betting had the same information as everybody else, were reading the same polls and analyses, etc. The markets, in short, behaved like a second-rate pundit on CNN-- they expressed whatever the conventional wisdom of the moment happened to be, with a lag time of a few hours.

Sure, whatever the outcome of the race eventually is, the markets will have "predicted" it at some point beforehand, but so will Jeff Greenstein--along with every other outcome that seemed likely on a given day but never panned out.


posted by: lamont cranston on 01.17.08 at 05:58 PM [permalink]

This is the political equivalent of the Heisenberg Uncertainty Principle. If nobody has noticed this and coined a political version of the term, then I will lay claim to it now and call it the Sixpack Uncertainty Principle.

Heisenberg Uncertainty Principle: locating a particle in a small region makes the momentum of the particle uncertain, and conversely, measuring the momentum of a particle precisely makes the position uncertain.

Sixpack Uncertainty Principle: measuring the political support for a candidate and publishing the poll results makes the candidate's momentum uncertain (due to effect of the poll on public opinion), and conversely, measuring the momentum of a candidate makes determination of the candidate's political support uncertain due to claims of momentum upon the opinion polls.

posted by: Joseph Sixpack on 01.17.08 at 05:58 PM [permalink]

I saw on you BHTV talking (or not talking) about Jonah Goldberg's book "Liberal Fascism." You pointed out that not having read the book, you should reserver judgement. You're a political scientist and an honest person, not a hack like Goldberg or Byron York or Ann Coulter, so I would actually trust your impressions of the book. I hope you take Henry up on his offer to read the book and return to BHTV to have an extended conversation about it. I'd love to hear what you have to say.

Cheers and regards

posted by: Frank on 01.17.08 at 05:58 PM [permalink]

Prediction Markets? Just another way of saying "ad hoc bookmakers." Considering the Rhode and Strumpf paper--I think it would have benefitted from further incorporation of mathematical factors. Sometimes, even in economics, basic math must be used to really see what is happening. There are too many variables unexplored or unaccounted for in their four points.

Frank--Jonah Goldberg a "hack?" Really, that's pretty inaccurate. Question: Have YOU read Goldberg's book? Why is it important to you what Dr. Drezner thinks about it? Read it and think for yourself.

posted by: Useless Sam Grant on 01.17.08 at 05:58 PM [permalink]

Post a Comment:


Email Address:



Remember your info?