Trading politics: University runs market to predict provincial election results

By , on March 15, 2017


Antweiler teaches at the University of British Columbia's Sauder School of Business and runs its online election market, where people use cash to buy contracts or “shares” representing political parties ahead of provincial or federal elections. (Photo: University of British Columbia)
Antweiler teaches at the University of British Columbia’s Sauder School of Business and runs its online election market, where people use cash to buy contracts or “shares” representing political parties ahead of provincial or federal elections. (Photo: University of British Columbia)

VANCOUVER –After 20 years of running election stock markets, Prof. Werner Antweiler believes the buying and selling of shares in political parties is a more accurate way of predicting election results than public opinion polling.

Antweiler teaches at the University of British Columbia’s Sauder School of Business and runs its online election market, where people use cash to buy contracts or “shares” representing political parties ahead of provincial or federal elections.

Traders can take long positions, which means buying shares in parties they think will grow in value, or short positions where they buy shares they think are overvalued. Outcomes are all tied to real election results and traders are allowed to spend between $25 and $1,000

Election markets raise the stakes compared with traditional public opinion polls, which can be influenced by personal bias on the outcome, Antweiler said.

“But if people have to put money where their mouth is, then they may be more eager to reveal their true belief about how an election is going, even if it’s not their preferred outcome.”

Election markets are better than polls at absorbing new information and anticipating trends because they take into account not only what’s happening today, but what people expect to happen in the future, Antweiler said.

“The idea is that futures markets are forward looking whereas polls are kind of backwards looking in the sense that they aggregate information from the past few days,” he said.

Kara Mitchelmore, CEO of the Marketing Institute and Intelligence Association, said there are significant differences between polls and election markets.

“I don’t think it’s fair to pit one tool against another,” she said, adding that investors in election markets often base their decisions on data from polls.

The reason polls work when they’re done well is that they’re based on valid science, Mitchelmore said.

Election markets can be off in their predictions, Antweiler said.

A market operated by Iowa University in last year’s U.S. presidential race showed it was likely that Hillary Clinton would win, but Antweiler said the margins were much narrower than in public opinion polls.

“Even though the probabilities were still in Hillary Clinton’s favour, they allowed, quite clearly, for another outcome,” he said.

The university’s market is the only one of its kind in Canada. The school is running three markets tied to the B.C. provincial election this spring, including a market gauging the popular vote, another on the share of seats each party wins in the legislature, and a third on which party will form a majority government.

Election markets are particularly good at predicting seat shares and who may or may not form a majority government because traders have to gather a lot of information to be comfortable participating, Antweiler explained.

“People who are more educated, politically, may feel more comfortable taking positions, especially aggressive positions, in the market. Whereas newcomers now have a strong incentive to inform themselves now because they have money at stake,” he said.

On Wednesday, the governing Liberals in B.C. had slight leads over the Opposition NDP in every market. But Antweiler cautioned there are several issues that could cause the markets to fluctuate before election day on May 9, including housing prices, infrastructure, public transportation and campaign financing.