How people elect parliaments
Since the 2000 election year, one of the more intriguing predictors of US political outcomes has been the Iowa Electronic Market.
Run out of the University of Iowa’s College of Business, the IEM was an early adopter in the field of prediction markets.
The principle behind the market is simple. A large pool of people – many thousands are said to participate – sign up to the markets. They involve real money, although the stakes are not high – they are essentially $1 bets. The number people can take out is regulated, as the project does not aim to become an exercise merely of online gambling.
Participants buys contracts for political outcomes, such as “the Democrats will win the presidency in 2016”. The day after the election, if their prediction was correct, the IEM pays out $1. If they’re wrong, the participant gets nothing. Participants can then buy and sell their contracts to other participants in an open market, so when you first buy one unit of contract, it costs you whatever the market is selling for that day.
What happens day-to-day is that the prices of all IEM’s contracts (they have a range, including predictions of legislative majorities and other election events) fluctuate at values between zero and $1. The result resembles an opinion poll, except that instead of being a measure of what those who are polled want to happen, it is a measure of what they actually expect to happen.
The logic behind this idea is psychological; because people have a stake in the outcome, they focus on true probabilities, not their own desires. What’s more, people can be participants in markets where they think they can predict the result, not just because they personally care about who wins.
Market speculation is also possible, and to an extent is a healthy contribution to price setting. But IEM tries to regulate contract sales to prevent irrational speculation surges from affecting the exercise.
At any time, IEM can generate from it’s daily market prices data and graphs of price movements, such as the following:
The graph of prediction prices for unnamed Democratic and the Republican candidates winning the US presidential election to be held on 8 November 2016 (source: IEM accessed on 18 February 2016)
Shown above is perhaps the most important of IEM’s markets – the prediction of the coming presidential election of 2016.
Notice that IEM opened this market (that is it opened contracts on this outcome for sale) back in November 2014.
Notice also how the prices for the two parties’ prospects broadly mirror each other. This graph is saying that buyers have always thought the Democratic candidate had the best chance; the current odds are about 63% for a Democratic win and 39% for a Republican win.
Note carefully that the two prices don’t need to add to exactly $1.00, or 100% probabilities. At any given moment the buyers and sellers in each market can be over- or under-confident. Price surges up or down can occur due to public events.
For example, note the spike in the Democratic price in May 2015, coinciding with the announcement that Hillary Clinton was launching her campaign. There was wild purchasing for a few days, before cooler heads prevailed and settled into a more balanced valuation of the probability of her party now winning.
And note the matching smaller spikes in November 2015. (I’m not sure what this was about, but it was probably a market reaction to something Donald Trump did.) Again the price spikes only lasted a day or so before buyers and setters settled on more rational prices.
IEM has a good track record as a real predictor. It made its name successfully predicting the very close election of the year 2000.
Overall, studies have shown that prediction markets consistently achieve levels of prediction accuracy at or around the very best of opinion polling exercises. They have never proved to be dramatically wrong.
Prediction markets have been around in most countries for over 10 years now. Some are online gambling services pure and simple. IEM, created as a tool to teach university students about the psychology of markets, avoids any hint of gambling hype and retains it’s non-commercial, academic focus. Of course, it still is real money that participants are punting!
The example graph shown above is still a generic party candidate’s chance of winning the US presidency 9 months from now. What about the primary races?
Well, IEM is selling prediction contracts on those as well. Here are the graphs for the past four weeks:
The Democratic Party presidential nominee prediction price graph as at 18 February 2016; “CLIN_NOM” (the blue line) means ‘Clinton wins nomination’, similarly “SAND_NOM” (yellow) is Bernie Sanders’ price, and “DROF_NOM” means ‘Democratic rest-of-field’.
IEM’s punter community has always been sure Hillary Clinton would win the Democratic contest, and the events of the first few weeks of primary elections haven’t changed their mind. Prices for the $1 Clinton contract have consistently been above 70 cents, and haven’t dipped in recent weeks at all – in fact they’ve gone up slightly.
On 14 February Bernie Sanders stock was abruptly dumped for a day – it’s not entirely clear why – but it has recovered to it’s tracking price of around 27 cents.
For some reason people still think a bet on the Democratic party ‘rest of the field’ is worth 4 cents. This is the price prediction of both Sanders and Clinton tumbling out of the race somehow and the party unexpectedly picking someone else (after all this trouble!).
The Republican Party presidential nominee prediction price graph as at 18 February 2016
The Republican candidate market has always been more mixed. Donald Trump (the green line in the graph above) has generally had the best prices for the past month, but was only briefly above 50 cents just for a day or two. Marco Rubio (grey) surged for a week up to 62 cents, then got dumped to 22 cents, and has recovered to 39 cents, about equal today with Trump. Notice the significant price volatility in this market, as the participants struggle to work out the true value of each of the predictions.
Anyway, keep an eye on these markets in the months ahead – they are a compliment to commercial opinion polling, and can sometimes pick up trends faster and more accurately than polling does.