I understand the theory behind the claim that election markets might be more useful tools for predicting election outcomes than opinion polls:
- Pollsters ask about current opinion ('If an election were held today, …'), and election markets focus attention on what really matters: what actually happens on election day.
- Poll respondents have essentially no incentive to tell the truth, while election market participants have a direct financial interest in revealing their subjective beliefs about election outcomes.
These are not-insubstantial arguments. So when the people behind the UBC election stock market started their 1997 federal election market, I joined in.
I was soon disabused of my preconceived notions. For one thing, the people participating didn't seem to understand how the market worked. Payoffs were allocated for the share of seats won in the House of Commons, but prices tracked national poll numbers almost exactly. Now, anyone who knows anything about Canadian politics knows that vote shares do not correspond to seat shares. But the election market didn't.
Then there was the day when I was trying to translate prices into seat forecasts, and couldn't understand the numbers I was getting. It eventually dawned on me that the reason the math didn't work out was that the bid and ask prices didn't satisfy the no-pure-arbitrage condition. Worse, this happened pretty much every day.
I ended up doing very well cashing in on those arbitrage opportunities (hey, somebody had to do it) and by betting on the Bloc Québécois (their national vote share underestimated their share of seats). But the experience left me somewhat jaded about the merits of election markets – in Canada, anyway.
There are at least two reasons for thinking that election markets should not be taken seriously as indicators for outcomes:
- Participants regard the market as a casino, and treat it as such. Betting $20 on your party's victory is essentially the same thing as betting $20 on your football team's victory next weekend. When that money is put in the market, the
investorbettor doesn't really expect to see it again. - The markets are way too thin – at least in Canada. It's so easy to generate headline-inducing movements that it's hard to believe that no-one wouldn't try.
So it was with some dismay when I saw yesterday that Andrew Coyne saw fit to bring attention to this election market as evidence that the federal Conservatives might be likely to win the election that may or may not be held this fall. The trouble was – as I noted at the time – it would have cost about $30 to send the implied probability of a Conservative victory to 100%. Since then, no trades have occurred, but blocs of bids and offers (in nice, round quantities – no hint of co-ordinated activity here!) have been placed as tripwires around the last closing price.
It would appear that both sides have deployed the resources necessary ($164 between them) to make sure that we don't see any more media stories with the theme 'The Election Market Speaks'.
I'm hoping that they succeed. But I'm not betting on it.
Update: Felix Salmon notes that US election markets don't seem to work any better (h/t to Mike Moffat).
Worthwhile Canadian Initiative 
Yet somehow those betting $20 on their team end up creating spreads that are remarkably close to the final scores in games. I’ve never heard of the UBC election market, perhaps there aren’t enough people trading to make it accurate. After all someone has to take the other side of the trade. But if you were able to make money, then you should have kept exploiting every opportunity until the market reflected the true odds. That itself would have made the market accurate.
“I ended up doing very well cashing in on those arbitrage opportunities (hey, somebody had to do it) and by betting on the Bloc Québécois (their national vote share underestimated their share of seats).”
LOL.. I did the exact same thing and made out like a bandit. I also made a bunch of money on the NDP, on the theory that the other people in the market were disproportionately right-of-centre (and their purchases reflected their biases). I was right.
“The trouble was – as I noted at the time – it would have cost about $30 to send the implied probability of a Conservative victory to 100%. Since then, no trades have occurred, but blocs of bids and offers (in nice, round quantities – no hint of co-ordinated activity here!) have been placed as tripwires around the last closing price.”
If you’re so sure why don’t you just sell the contract and make a nice profit?
I suppose I could, but that misses the point: the informational content of the market price would then consist of whatever I thought it should be.
I disagree. I am a fairly heavy ESM trader. I do agree that the market is thin. What does that mean? That means that pricing anomolies (such as the impact of a partisan big money trader) do happen. In thick markets, these anomolies disappear instantaneously. In thinner markets like the UBC ESM, these anomolies just take a little longer to work their way out.
For example if a partisan NDPer comes in and pushes the NDP popular vote contract up to 50 cents with lots of buy orders, then someone who happened to take a look at that particular moment would think that the market is not a very good predictor of anything. However, all it takes is one non-partisan, deep pocketed arbitrage minded trader to come in undo the anomoly. I’m happy to fulfill that role. If I logged in and saw that 50Cent NDP contract I would sell NDP until it got back down to the ‘right’ range. So long as my pockets are deep enough, I am happy to take those trades. But, I don’t log in every day. So, if I were the only non-partisan trader (and I am surely not) then perhaps the anomoly might stick around for awhile. But, it would soon enough be undone by me or by someone else.
I suppose it would, eventually. But I’m much less confident that it would occur before some journalist would see the anomaly and say “Hey! Here’s a big spike in the election market! I should bring attention to it!”