Predicting Palin
By Ocean on Sep 2, 2008 in Articles of Interest, Current Events, Featured, Prediction Markets
Those who read our previous post regarding prediction markets know that we were following the VP candidate announcements with interest. They may have also noticed the the prediction markets were incorrect.
Once again, Joseph Biden was the top Democratic VP pick for the markets, while the Republican pick varied between Mitt Romney and Tim Pawlenty. As we know, Sarah Palin received the nomination, while prediction markets had her likelihood in the single digits.
A comprehensive post on Concurring Opinions addresses this and what may have gone wrong. Nate Oman proposes three options:
First, inTrade may simply be bunk. Second, Palin’s choice may simply have been a very improbable event that happened to come to pass. Third, it may be that in the period running up to the choice of Palin there just wasn’t all that much relevant information available about the relative probabilities of who was being chosen
We agree with the second and third statements (judging that they are not mutually exclusive). The reason we do not agree with the first is that Prediction Markets are not Crystal Balls. They can be very useful in aggregating public information and determining certain outcomes, such as with elections and business world occurrences. As we mentioned in our previous post, they cannot enter a politician’s head to see what he (or his advisers behind closed doors) will choose. The only way to do this would be through insider information which, if it were legal, would eliminate the use for markets in the first place.
What seems to have happened here is that there was very little, if any, information regarding VP picks, but what little information there was was used and thereby blown out of proportion. It is important to remember the limits of prediction markets when using them. Always check what information sources there are, as well as how many. And, as always, use your head.
(6 votes, average: 8.83 out of 10)
Loading ...
Yo, Ocean– what is your take on the latest Prediction Market folks suggesting that Palin will resign before too much longer. Are they likewise working on too-little info, or how would they arrive at such a prediction? Just curious. — bjg
Uncle Bear | Sep 2, 2008 | Reply
Ah, thank you Uncle Bear. I actually think this is an interesting example of the same problem:
The fact that the question has been propping up everywhere suggests it is a real possibility.
On the other hand, all of the markets (at least the ones I’ve seen) put the chances of her resignation at less than 20%.
However, the volume for many of these questions is extremely low, so a few answers could skew the average greatly.
I think this makes for a good question because it sounds scandalous, however it is not optimal for prediction markets since there aren’t a multitude of sources of input (there’s just one). I believe a majority of questions like these are likely to be predicted correctly by prediction markets, but it is impossible to say for any one specific question.
Personally, if a bet had to be made I’d bet with the majority (ie. she’s staying on), but I’d make many more bets of the same kind so I could be sure they’d average out in my favor.
Ocean | Sep 2, 2008 | Reply