Some of the challenges of launching a new, useful and innovative internet based product are apparent in the iPredict.co.nz business.
Officially launched in 2008, just before that year’s general election, iPredict quantifies what are otherwise qualitative measures. Traders are betting on different predictions, allowing people to buy and sell stocks in future political and economic events. They are literally putting their money where their mouth is.
Put another way, the current market price (seen on iPredict) can be interpreted as predictions of a probability of an event. (There’s a Wikipedia explanation of prediction markets in general here).
This, from a psychological and accuracy point of view, makes the prediction of future events much more accurate than polling.
When iPredict asks binary questions (yes/no) it has a much more accurate estimation of what the actual outcome will be.
“Because people are using real dollars, it is more resilient to bias and manipulation,” says iPredict chief executive, Matt Burgess. “It makes that sort of behaviour costly.”
It isn’t a platform for betting on sports, ‘Dancing with the Stars’, or the Oscars. It is about measurable future events however, though not too far distant.
Burgess has an economics background, and successfully launched (and recently sold) a start-up web-based company called mytradesman.co.nz. What became Burgess’s main job on joining iPredict in 2006 was getting the Viclink/ISCR joint venture project past the rules of the Financial Markets Authority (FMA). The idea itself is traceable back to a 1991 idea by Lew Evans and Bryce Wilkinson, though the legislation then and now wasn’t written with prediction markets in mind.
The FMA eventually allowed iPredict to burst into life, without the need to issue a prospectus – as is usually the case when asking for someone else’s money for a business project.
The company’s trading fees are one third of a cent for every share traded, and a 2% withdrawal fee if traders take money out of their account (though only if they’ve made money). iPredict also sells the license to use its facility as a software-as-a-service model.
The site has 5500 traders, with about 600 being active in any one month and the current election doubling active participants.
Most peoples’ trading float is about $80, and there’s about $500,000 of deposited funds says Burgess.
“But we’ve never had a report of problem gambling through our site,” he says. “Prediction markets are not about those you’d find at night playing the pokie machines. It attracts students and professionals, and is less about entertainment and more about testing your opinion and the quality of your ideas.”
One of iPredict’s features is the almost instant reaction it can provide for changes in events. “It is real time, accurate information,” Burgess says.
The beauty of prediction markets is that manipulation is in a sense, self-correcting. Because each bet on one outcome, requires someone else to bet on the equal and opposite outcome, the odds should always add up to one. And they do over time.
But, as happened recently, someone bet heavily ($500) for Labour to win. The price on this lifted sharply. Almost instantly, keen and observant traders were in, taking the opposite bet – figuring such an event is of such little likelihood that they were sure to make some money.
“A manipulator, by definition, is willing to pay too much for a trade,” says Burgess. “It increases the accuracy of the market, since where you have people willing to pay too much you drag more people, more dollars in. Manipulation can’t be understood as a rational trading strategy. What they do do though is create opportunities for everyone else, creates an arbitrage position [where someone backing both sides of a bet at the same time can’t lose money]. Manipulation doesn’t permanently affect the market trade price at all.”
As a business model, iPredict has its challenges, though “we’ve got really good software that we can license,” says Burgess. “It’s battle-hardened, and has had three years of operating using real money. Every error has been caught.”
The platform is also scalable as well as being robust, and the next targets for growth are Australia and Asia. The USA is currently off the target, its Securities Commission not allowing the concept. Indeed, a disclaimer suggests that Americans can’t use what is, at this stage, NZ-centric political and economic event prediction stocks.
Burgess is somewhat surprised, and not for want of trying and selling, that iPredict hasn’t been used more by mainstream media during the election run-up.
“It’s more accurate, proven, than polls,” he says. “When something happens, the market reacts instantly. It should be front and centre of major organisation’s election coverage strategy.”
This is one of the reasons iPredict set up its own election oriented TV channel (on Stratos, Sky/Freeview 89). electionresults.co.nz has been fronted by Burgess himself, using iPredict charts as the basis of discussion.
Burgess says that the platform is actually complementary rather than a competitor to traditional polling.
“Where there are no polls, iPredict may be less accurate,” he says. “In last year’s mayoral elections, traders got it right in Auckland, Christchurch and Invercargill. They got it wrong in Wellington, Hamilton, and Dunedin. There were few polls in these places, and where there’s no polls, there is no raw material for traders to predict with, a less clear view of what is happening.”
Burgess points out that iPredict also asks a different question than polls.
“One of the few places there’s an overlap between polls and prediction markets is forecast party vote share, but polls cannot answer questions about likelihood, for example how likely a certain party is to win the election, or how likely a coalition deal between X and Y will be struck, and so on. And prediction markets cannot answer questions about how certain demographics will vote or which policies they will be most responsive to. So there is actually strong complementarity between polls and markets. Each source is rich in its own way.”
iPredict has proven that the economic theory of a prediction market works on a web platform, there’s robust software behind peoples’ trading and accounts, and that it is a clever product.
The next challenge is to take it international – but that in itself may be at least as hard as proving its worth in the New Zealand market.
Between yarning to Matt Burgess and posting the story, the election’s come and gone. Yet again iPredict beat all the poll people – who had been predicting National with over a low 50s to mid 50s majority.
IPredict’s verdict was 48% – which it was.
The site is currently running over 340 contracts on which politicians will fill which particular portfolios.