Old friends and prediction markets

There can be no pretence when meeting up with an old friend, sure you may wear a quality cut suit nowadays and you may well keep your room clean and tidy; but they still see you as the spotty and badly dressed kid you were when you first met.

So it was last night, catching up with Matt, with whom I went to school.

He is working on a neat little project launching next month which has got me thinking, it is folksonomic, combined with a healthy dose of buying low and selling high.

It is all about predicting the future

Matt is working on a prediction market for New Zealand, a joint venture between VUW and ISCR, aptly named iPredict.

Prediction markets work by offering a platform for people to buy and sell one dollar securities on the outcome of future events; they are big in the states for predicting political outcomes and currently being talked about in relation to predictions for the results of the primaries.

They differ from a political poll because instead of asking people which outcome they prefer, i.e. which presidential candidate they would vote for, they encourage people to guess which outcome will be favoured i.e. which presidential candidate do they think is going to win.

This is a subtle and profound difference.

Predicting politics

It is no coincidence Matt and the team are launching iPredict in New Zealand in 2008, the year of the general election. Come April it will be possible for us all to buy a stake in the political party we think will win the election, and cash in when we are proved right.

As an example of how it will work: if you think that National will win, you will want to “buy” some securities in a National victory outcome.

Let’s say the current market price is 65c, and this represents an overall market consensus that the National Party has a 65% probability of winning the next election; this will mean for every 65c that you invest, if National do win: you will receive one dollar; but if they lose you get nothing. The price on the market shifts as events and announcements unfold, so a Labour Party announcement on tax cuts may make participants in the market think a Labour victory more likely, and so the price of the National victory outcome will drop, reflecting the consensus a National Party victory is now less likely.

The end result is a system that more accurately predicts certain outcomes than any other method, including polling.

Politics schmolitics

Being a-political, I am not that excited in the application of prediction markets for politics, but there are also other, far more interesting applications.

Imagine using the same technology for a company that is working on a number of new products – they don’t have the resources to bring all of them to market, so they need to select the one that is mostly likely to be a commercial success. Internally the company could run a predictive market through their intranet and get everyone in the company to make a prediction on the product most likely to succeed commercially; and as such gain insight on the direction for resources to be applied. There could also be applications in large government departments, wanting to get some insight on the areas of the business which with focus are more likely to achieve strategic goals.

Google use predictive markets internally - and they seem to do pretty well predicting what to do next on the interwebs.

I was also interested to come across the local site Leaguist.com yesterday. There you can pick winners of sporting events as well as see percentages of what others have picked which is an element of a predictive market.

For the economists: iPredict are using a system of automated market makers to ensure liquidity but as I have almost know idea what this means – I will leave it to others to discuss.