Prediction Markets Are Good, Actually

Will Alberta vote to separate from Canada? Will Katy Perry and Justin Trudeau get hitched this year? Soon, you might be able to legally bet on real-world events like these—on Wealthsimple. Earlier this year, news broke that the Toronto-based fintech company had received regulatory approval to offer prediction trading in Canada. It’s already a booming industry elsewhere: Polymarket and Kalshi, the two biggest companies offering this kind of trading, have handled over US$60 billion in trades this year alone.  

To understand why this is such a big deal, you first need to understand how prediction markets work. At its core, a prediction market allows traders to bet on whether a specific event will occur. The event could be, for example, “The Bank of Canada will change interest rates on June 10” or “Katy Perry and Justin Trudeau will be engaged by the end of 2026.” Unlike the long-established bookmaker model—where the house calculates the odds, takes the other side of every bet and earns the spread—platforms don’t set prices. Instead, prediction markets operate as exchanges, and match “yes” and “no” traders against each other in a transparent order book. In function and spirit, they’re more like a stock exchange. 

Traders take positions by buying contracts, which, by default, pay out a dollar each if the event turns out in their favour. The contracts’ price reflects the market’s estimate of the event’s probability. If the Perry-Trudeau engagement contract trades at 30 cents, for example, the market is saying that the event’s likelihood is roughly 30 per cent. At a fixed point in the future—which is pre-determined but sometimes arbitrary, like the last day of the year—the contract is set to expire. By then, if the event has occurred, the owners of “yes” contracts receive a payout, and the owners of “no” contracts receive nothing. If the event did not occur, the reverse happens. Yes, the returns can be high. If a trader buys a contract for five cents and it pays $1, the gross payoff is 20 times the stake and the net return is 1,900 per cent. But if they lose, their contract is worthless. Since traders can end up unable to recover any of what they paid, regulators often compare prediction trading to gambling, in the sense that you’re either right or wrong—and if you’re wrong, you lose everything. 

But prediction trading is more than that. For one, these contracts can serve as financial insurance. Suppose a tradesperson wants to finance a new project, but the loan they’re offered has a floating interest rate tied to the Bank of Canada’s rate. If rates rise, that loan will become more expensive and the tradesperson will face a loss. So they can buy a contract that pays when rates go up, and thereby hedge part of that interest rate risk. It’s just like car insurance: it costs you something up front, but if things go wrong, you get a payout.

More importantly, when making decisions, many people and firms seek out useful information about future events. A firm considering an investment in Alberta needs to account for a possible split from Canada. A prospective student wants to know if a specific degree will lead to an AI-proof job. A 30-year-old who inherited money from their grandma is considering a condo investment and wants a forecast for the rental market. 

For many of these questions, airtight information is difficult to obtain. Often, all we have is politically motivated punditry. But there is so much more information out there. People in Alberta talk to their neighbours and friends, so they know something about local political sentiment. HR departments plan future hiring and have a sense of what kinds of graduates they will want. Realtors collectively speak to thousands of clients and have insight into housing demand. When people with relevant on-the-ground knowledge buy prediction market contracts, they’re making informed bets—putting their money where their mouths are. In doing so, they aggregate decentralized information faster and better than most public data sources. Unlike a bookmaker, who sets the odds behind a curtain, prices emerge in the open on a prediction market. 

How accurate is this information? The Canadian Investment Regulatory Organization has barred political prediction trading, but in American elections, the evidence has been impressive. Faculty at the University of Iowa launched the first modern prediction market, called the Iowa Electronic Markets, in 1988. That year, it forecast the results of the Bush-Dukakis presidential election far better than major polls. It estimated that Bush was likely to win 53.2 per cent of the popular vote; he emerged with a tidy 53.4 per cent. In the following decades, the IEM continued to outperform polls 74 per cent of the time. Meanwhile, Stanford researchers Hall and Paschal developed an online app called Bellwether, which tracks success probabilities for political markets on Kalshi and Polymarket. For major races, accuracy is high, at 86 per cent. When these markets work well, they are a serious forecasting tool, making them valuable to anyone whose fortunes depend on political or economic events.

But there are a few common objections. I believe that prediction markets are economically useful, but many argue these markets will be used primarily for sports betting and other irrelevant celebrity engagements—and that betting on these serves no social purpose. It’s also possible that, when financial institutions offer prediction contracts, people may confuse them with investments. Investing means handing money to a firm or fund that uses the capital productively and, if things go well, returns the principal plus profit. Meanwhile, a prediction contract is a short-term bet with an all-or-nothing payoff. The two aren’t the same. But I don’t think people need special education to understand that a losing prediction contract is worth zero. What are the odds that someone playing the lottery does not know that a losing ticket is worthless? A prominently placed, two-sentence explainer can clarify the consequences. The problem is not comprehension in that narrow sense. Instead, it’s human vulnerability. Gambling is seriously addictive for some people, and for those people, prediction markets may trigger the same psychological response. 

Meanwhile, although the information-production case is strong in principle, Bellwether reports that prices are meaningful in less than one per cent of markets. That’s because there is often very little trading volume on economically relevant markets, and a market requires sufficient activity to offer real information. When a market aggregates a multitude of opinions, it’s more likely to be right; a market with thin trading doesn’t offer reliable information about what people believe, because it relies too heavily on the opinions of a few people. In other words, the wisdom of the crowds can’t unfold without a crowd. Low-volume markets are even vulnerable to manipulation, because individual trades can shift prices very quickly.

Moreover, there are serious insider-trading concerns. A real estate developer will know more than a layperson about the new construction market—that’s broad personal knowledge, and its use is acceptable (and desirable). But some people have direct, exclusive information about a specific event: for instance, government officials know what policy decisions may be coming up. Earlier this year, for example, a U.S. Special Forces member allegedly traded on advance knowledge of the American operation against Venezuelan president Nicolás Maduro. In some cases, insiders’ actions can even affect the outcome directly. That’s why referees, coaches and players are restricted from sports betting.  

These markets will reach their potential for information aggregation only if participants trust them to be fair. If ordinary traders fear that they are systematically trading against someone who already knows the outcome, they will stay away. Maintaining market integrity is therefore essential. In my view, this is best done in a regulated financial-market environment. In the U.S., the Commodity Futures Trading Commission has asserted jurisdiction over prediction markets, and in Canada, Weathsimple is a regulated financial institution, so we are moving in the right direction.

But what should the destination look like? Prohibition is not a serious option—anyone in Canada who wants to trade on Polymarket can already do so with a VPN. Realistically, we must choose between regulated onshore prediction markets and unregulated offshore ones. In an ideal scenario, Canada should apply securities regulations, not gambling regulations, to prediction markets. That’s because securities regulations come with position limits, transparent order books, market-maker obligations, suitability requirements, trade surveillance and laws against insider trading and manipulation. They’re also backed by a century of accumulated infrastructure: Canada’s first securities regulations were introduced to Manitoba in 1912. Gambling regulation brings none of these benefits. 

The country also needs to implement insider trading rules with teeth. Securities regulators have spent a century working out who can trade on what information and when. Those rules translate naturally to economically relevant prediction markets. Government officials, central bankers, and people with operational knowledge of military, regulatory, or corporate events should be presumptively restricted from trading on their own information—and platforms should be required to surveil for it, the way stock exchanges already do.

We must also establish clear rules on what can and what cannot be listed. Some contracts serve a genuine economic purpose when they aggregate information, such as interest rates, employment data and election outcomes. Any contracts that create financial incentives to harm specific people or groups must be illegal. 

Finally, traders need tools to manage any susceptibility to gambling addiction: cooling-off periods, spending caps, mandatory self-exclusion, and the power to opt out from commercials. Regulators have a lot of power over companies, so they must ensure that exploitative engagement-mining patterns—which have made sports-betting apps so dangerous—don’t find their way into the financial industry. Some betting apps hire armies of psychologists to increase engagement and spending, so there’s a case to be made that vulnerable users may be better protected when engaging with regulated financial institutions. In the world of finance, prediction market contracts are functionally options, which are subject to suitability tests. In other words, financial institutions cannot make these products available to everyone. 

Throughout history, many novelties have promised quick riches—railroads, the stock market, dot-com firms and crypto trading. Some people get rich off of them, while others join late and are left holding the bag when the bubble bursts or the excitement eventually fades, as it always does. What receives less attention is that these bubbles often leave behind technology and tools that serve a broader purpose. Developers overbuilt railroads, then went bankrupt—but society benefited from that overbuilding. Many people lost fortunes with stock options in the 1990s, but options are now well-established risk-management products that people use to offset, for instance, exchange rate risks in cross-border trade. Many dot-com firms went bust and lost investors’ money, but they also gave us the modern internet. And crypto came with plenty of scams, but also introduced blockchain-based infrastructure to the finance industry.

My hope is that prediction markets will become an information-aggregation tool that helps investors build more resilient portfolios and enables entrepreneurs to make better capital-allocation decisions. That would advance both society and the economy. And honestly, is it really so bad to bet a dollar or two on meaningless celebrity frivolities?

Andreas Park is a professor of finance at the University of Toronto.

What did you think of this story? Send a comment for publication to letters@macleans.ca, and supply your name and address. Comments should be fewer than 200 words, and may be edited for length and clarity.