An “influential forecast” is one in which knowledge of the prediction changes the uncertainty in the thing being predicted. Consider a (respected) political opinion poll which announces Candidate B has an 85% chance of winning an election. Some people supporting Candidate A might hear this and figure the margin is so large that they needn’t bother voting. The result of the vote will edge in the direction of Candidate B.
This isn’t far-fetched. Exit polls for presidential races used to be reported from states in the east while voting continued in the west, but it was felt these polls influenced behavior and so are now withheld from general public knowledge (journalists, better equipped to deal with reality, still know their dark secrets).
Polls are not only error-prone measures of opinion, but they can always be viewed in a predictive sense, as a guess of what the vote percentages will be.
Also of interest are those forecasts produced by prediction markets, such as Intrade, Betfair, and the Iowa Electronics Market (IEM). The price on the presidential race on Intrade is often quoted in the press and taken to be indicative of the status of the race.
Some people are betting the predictive markets like Intrade are not just accurate but influential. That is, some folks have pumped money into the system hoping to drive the price up for a candidate and therefore make that candidate appear viable to the press, which will then dutifully report what it sees. According to this Daily Caller article, market manipulation has happened several times in the past on Intrade, particularly in presidential races.
If Intrade races were fiddled, then their forecast accuracy will have suffered. What of nominee races? In a University of Nottingham dissertation Ian Saxon1 found that Intrade predictions for the 2004 and 2008 Democrate nominee contests were good. He compared Intrade against an average of polls (he also tracked the individual polls, which were not much different than the average). Recall that polls are not forecasting the winner, but the vote percentage.
Intrade predicted a solid Kerry victory, with probabilities (prices normalized to 100) near 100% early on in January 2004 (pic). In the Clinton et al.-Obama matchup, Intrade’s prediction also showed an Obama victory, but perhaps not a quickly as in 2004 considering the price did not near the top until the close of the race.
Intrade performed their own study of the 2004 Presidential, Bush-Kerry race, and produced this picture:
The events 1-7 are given by Intrade as:
- April 03: Coalition troops overtake Baghdad. Major fighting expected to be over
- December 03: Saddam Hussein captured in Tikrit
- April/May 04: Abu Ghraib prison scandal photos released
- June 04: Coalition troops hand over power to Iraqi interim government
- July 04: Democratic National Convention
- Early September 04: Republican National Convention; Swift Boat Veterans’ ads
- September/October 04: John Kerry “wins” Presidential debates
But the Daily Caller claims these undulations might have due to manipulation:
In 2004, recalls Eric Zitzewitz, an associate professor of economics at Dartmouth College who has co-written several papers on prediction markets, “somebody, in the middle of the night, drove the price of the [George W.] Bush re-elect contract — which was probably trading about 60 at the time — down to 15.”
Similar fluctuations were present in the McCain-Obama election in McCain’s favor. According to the Daily Caller, “The now-late founder of Intrade, John Delaney, concluded that the fluctuations were not the result of someone trying to rig the system.” Here’s the picture:
If manipulation is present, it’s pretty subtle, particularly near the end of the race.
Here’s the latest (as of yesterday) Intrade Chart for Romeny (red)-Obama (blue). Any manipulations? Or is this a good prediction?
As of this writing, Intrade has Obama at 57.9%.
Whether or not there is manipulation, Intrade weeks and months before the election is certainly not an infallible guide to who will win.
1Intrade Prediction Market Accuracy and Efficiency: An Analysis of the 2004 and 2008 Democratic Presidential Nomination Contests: pdf).