Victoria University's iPredict is claimed by its managers to be more accurate than most polls. Adam Bennett reports
Whether it's as a highly accurate forecasting tool, gambling in drag for political junkies, or a skirmishing ground for fringe party supporters, iPredict is a small but increasingly influential feature of New Zealand's political landscape.
Though betting on the outcome of elections in New Zealand is banned under the Gambling Act, a dispensation from the Financial Markets Regulator means iPredict - which describes itself as a "prediction market" - effectively offers odds on political events.
It also offers odds on a range of financial and economic indicators, scientific and technological developments and current affairs.
Examples include Fonterra's payout, the sharemarket's closing level, crude oil prices, when the Higgs boson or "god particle" will be first observed by physicists, and the release date of the iPad 3.
To protect the TAB's monopoly, iPredict cannot offer odds on sporting events.
Currently, for 93c, iPredict is offering punters a "contract" which pays a dollar if National wins the election. Put another way, it's paying just over $1.07 for a National win. The $1 contract for Labour to win costs 7c, therefore paying the equivalent of $14.28.
Compare that with Australian betting company Centrebet which is offering odds on the New Zealand election and will pay $1.08 on National and $7.40 on Labour.
But iPredict, owned by Victoria University, is not just a de facto political gambling platform. The "market-based political and economic forecasting system" is touted by its managers as more accurate than most polls and is being promoted to the media as such.
Chief executive Matt Burgess told the Herald analysis following the 2008 election showed that in the months before, the site "quite substantially" outperformed polls, proving more accurate than 15 of the 19 conducted over that period.
He puts that down to several factors such as polls' susceptibility to different forms of "selection bias" including the fact that landline phones, the preferred means of contacting the population being sampled, are less ubiquitous than they used to be, arguably among particular socio-economic groups.
"The composition of our population doesn't really affect the forecast", says Mr Burgess, "but the composition of the sampling population for polls does."
He says iPredict's accuracy aided by the "wisdom of crowds" or the fact that it aggregates information and judgments from large group in a way that produces a more accurate conclusion than any single member of that crowd might arrive at.
Mr Burgess says the members of that crowd on iPredict are reaching their conclusions using a wide array of information.
"Any trader on iPredict can take the result of the latest poll but also combine that with what they read in the paper, their experience in politics if they have any, or the fact their best friend's mate is the prime minister or whatever.
"All of that gets combined to produce a market price."
Political commentator Matthew Hooton, whose firm Exeltium does public relations work for iPredict puts it another way.
"This is not asking people how they will vote, it's asking how they think their neighbours or the community will vote."
Otago University politics lecturer Bryce Edwards says he has a lot of faith in the market.
"It is quite sophisticated, it's not a blunt mechanism like a poll. It really does take into account lots of different elements."
Labour MP Trevor Mallard said iPredict was "clearly not gospel but it's still an interesting thing because it shows where people are prepared to put some money".
But among the 5200 strong crowd applying their wisdom on iPredict - about 600 of whom are active in any given month - are political insiders like Mr Mallard or blogger and pollster for the National Party David Farrar.
As it goes, one of the key features of iPredict's special dispensation from the Financial Markets Authority is the freedom given to insider trading - a criminal offence on the financial markets the authority oversees.
"The fact that we predict so well is only a product of the fact people come to our market with information", says Mr Burgess. But for most iPredict contracts he says, "there can't be an insider trading problem".
"For example - with who's going to win the election, there really are no insiders on a question like that, the answer's in the community, it's just there's no one person has access to all the information at any one time."
On the other hand Mr Burgess says there have been instances were insider trading does appear to explain movements in price.
"We can never be sure but occasionally we've been given advance warning through the market of some event before it's been made public through media or other sources."
In March this year iPredict punters who were previously putting the chances of a Labour list MP resigning before the election at 15 per cent suddenly lifted that to 70 per cent within the space of a few days as it became known around Parliament - but not the wider public - that Darren Hughes was under investigation by the police.
Mr Mallard said though there were no official restrictions on doing so, Labour would be "appalled" if any of its MPs or staff used inside information to make money on the site.
He hasn't traded much on iPredict for some time, but does take advantage of what he considers relatively good deals occasionally - buying and selling according to "what I consider my superior judgement rather than on the basis of insider information. I'm privy to our polling and I would never use that for an iPredict gamble."
Mr Burgess says iPredict has had no discussions with regulators about insider trading and has never taken any action against a trader.
"It's not a serious or an important concern to most traders. Everybody knows exactly what the deal is on iPredict. I just think it's a 'buyer beware' issue."
He suggests insider trading on iPredict is good thing for the political process. "iPredict is a way for information to be made public and it's valuable that its out in public."
Otago University's Mr Edwards agrees. As well as contributing to the accuracy of the market's predictions, "insider trading, even by owners of political polling companies for example, can be of benefit to the public and the democratic process.
"Voters need information to inform them on their votes and this is a way of democratising that information. If we don't have opinion polls being published, if we don't have iPredict, it means a lot of information about politics is pretty much being held by an elite."
But it's arguable whether the wisdom of crowds and of political insiders actually gives iPredict an edge over other forms of forecasting. Though hindsight shows iPredict outperformed most individual polls in the run up to the 2008 election, it was not as accurate as the so called poll of polls, of which the Herald's version proved the most accurate.
One reason iPredict doesn't beat aggregated polls may lie in the fact that as a small market - its entire value including current contracts and traders' uncommitted account balances is currently $320,000 - it is fairly affordable to manipulate. That is especially true in the contracts for election results for smaller parties.
"You can see for example that Act is always over-priced," says Mr Mallard. "Often double or treble what they're polling. That to me is an indication of someone trying to make them look better than they are."
Mr Edwards agrees that Act's contracts often seem "extraordinarily high" suggesting they are being pushed up by supporters. That was not necessarily market manipulation but "people with more money that want to back their own side or perhaps who have an optimistic or naive forecast for their own party".
Such activity, he says, should be welcomed as a money-making opportunity by o rs. "That's why I personally go in and short-sell on Act."
But whether the pricing for each political party is correct depends on the eye of the beholder.
Mr Hooton doesn't believe Act is priced too high and thinks pricing is about right across all the political parties "with the exception of United Future which is priced too high".
The sums involved to push the price of a stock around on iPredict are relatively small. Daily turnover across the entire site is currently somewhere between $4000 and $6000 but Mr Burgess expects to see that increase as the election approaches. In fact he expects trading to spike enormously heading into the election.
It has been growing anyway; its total value including funds held in traders' accounts stands at $325,000, up from $220,000 a year ago.
Clearly the site and its backers have big plans and it is already profitable, although Mr Burgess refuses to give numbers.
But iPredict's greatest value may lie in the fact is gives armchair experts the chance to put their money where their mouth is. "There could be an argument that this makes politics more real and interesting for people and perhaps it's a way of encouraging their involvement."