Granite Geek: Wanna Bet That Predictive Markets Are More Accurate Than Polls?

Oct 7, 2014

You know the election season is gearing up when pollsters and survey researchers start calling residents, trying to gauge where the electorate stands on the issues and the candidates of 2014.

Politicians are, of course, quick to remind us that the only poll that matters is the one on election day, but there’s some evidence that another method, used regularly in the UK, may provide a clearer picture of where a campaign is headed than a traditional poll.

They’re known as predictive markets, and to explain them, we turn to David Brooks, who writes the weekly Granite Geek science column for the Nashua Telegraph and GraniteGeek.org.

How do these systems work?

We all know about them, because they’re basically what the stock market is - people placing monetary bets on their predictions on how companies are going to do in the future. And so the idea is that if you get people to place monetary bets on how various other things are going to do in the future, including political races, you can get a sort of crowdsourcing effect.

A tradition poll asks you, “who are you going to vote for?” It asks you about your own personal behavior and what you know about your own personal behavior. These polls ask you, “who do you think is going to win?” They’re asking you your knowledge about everybody else’s behavior. If you get enough people combining their partial knowledge, the idea is you get the crowdsourced wisdom, the wisdom of crowds. And the net result, in certain circumstances, can be more effective than traditional polling.

Putting money in – if you’re really confident that a particular candidate or party is going to prevail, that’s more serious than just sort of backing one candidate over another candidate.

I think that’s key even more than the crowdsourcing question part of it. It forces you to think about how confident you are. Some predictive markets out there online that don’t require payments are asking about – I saw one that’s asking about Brown and Shaheen, the US Senate race. But anybody can just put what they think in without having to pony up any dollars. You know if all of the campaigns find it they’ll just go game the system and people will do their wishful thinking – “I think Vermin Supreme’s going to win, so I’ll just click on that and say that!” If you don’t have to put money in it’s a little less effective.

How far can you take a predictive market? Can you put money up on virtually anything if someone’s willing to take your bet?

Well, you can’t in the US, because we have laws against gambling to a large extent. That’s why these markets are much more prevalent in Britain, where they bet on everything all the time. The High Street betting parlor – you can place bets on the equivalent of congressional races. In fact, you can place bets on who’s going to be the leader of particular political parties and really obscure stuff, from my point of view.

What do traditional pollsters make of all of this?

The evidence that it’s better, more predictive – there are some questions. I talked to Andy Smith, director of the UNH Survey Center, who, of course, has been polling forever and ever. He talked about how there’s some evidence that the crowdsourcing approach – without money involved, at least – can perhaps do better on saying who will win but does not do better on what the margin of victory will be. It’s early days yet.

Having said that, they do some of these questions. They’re doing them in the Maine governor’s race at the moment, I believe, for the Portland Press-Herald. When they ask the traditional polling question, “Who are you going to vote for?” the current governor, Paul LePage, is marginally behind. But when they ask, “Who do you think is going to win?” the current governor is marginally ahead. It’s an interesting difference, and it will be interesting to see how that runs out.

It’s likely, then, that 10-15 years from now, pollsters will still be calling around election time, whether or not the betting markets open up in this state.

I assume so – although I don’t see why the New Hampshire Lottery doesn’t get in on this somehow. You’d think they’d be able to make some money off this. Instead of buying a scratch ticket, you’d buy a ticket and say who you think is going to win, and then you can get some sort of return. I think that could be a real money maker for Concord in that.