The first time I heard about prediction markets was when my one of my favourite sports team became the first U.S. pro team to partner with the second biggest market, Kalshi. I saw a lot of people complaining about the deal and how unethical it was, and given how pervasive gambling already is in sports, I knew it had to be something interesting. I did nearly get Kalshi to see how it worked, but I quickly found out that it’s banned in Canada, which is probably for the best. 99% of bets on Kalshi are currently traditional sports bets, but I wanted to look further into the growing political side of prediction markets instead. Prediction markets like Kalshi are regularly cited to positively improve predictive accuracy and increase civic engagement, creating a more efficient democracy. Initially, I was pretty skeptical of this, so I decided to do a bit more reading into the theory behind prediction market forecasts and how they might shape governments and corporations alike as they continue to grow.
If you haven’t heard of a prediction market, it’s pretty much a cross between a sportsbook and the stock market. I’ll mostly be focusing on Kalshi in this article. While Polymarket is bigger by volume, it’s only just returned to the United States a month ago due to various legal issues, and it’s entirely cryptocurrency-based, which is an entirely separate can of worms. When you make a bet on Kalshi, you’re buying a “contract” between 1-99 cents, where a winning contract will settle at $1, while a losing contract settles at $0. For example, if the bet was “It will rain in Kingston tomorrow”, and the contract for “Yes” was 70 cents, I could buy 10 contracts for $7. If it did rain, my payout would be $10 with $3 profit, and if it didn’t, I’d lose my $7. This has parallels with sports betting, but the big difference is the ability to trade in and out of positions like in the stock market. If a few hours from now, the probability of it raining increased, the value of my contracts would also increase, so I could cash out early at a slight profit, like you could with a stock. Because these markets are two-sided, in that you’re betting against a different user, there’s no “house”, though Kalshi will take a fee on top of the price of contracts.
A Forbes article recently claimed that prediction markets are beating traditional forecasting methods like polling in both accuracy and speed. Betting money on an outcome in complete privacy would probably incentivise someone to tell the truth more than a poll where they have to share their answer with a stranger. Also, people are probably more likely to want to do their research before putting their money on a contract, while there’s no immediate consequence to answering a poll or voting without doing research.
It’s generally true that people who bet money on sports tend to be more involved in the finer details of statistics, so it might make sense that that translates to civic discussion as well. I could see a future where people on the whole are more educated in politics simply because it’s socially expected for them to put money on the outcome, as insane as that sounds now. However, there are some factors that I think should be considered before policymakers look to the prediction market to find out what a collective thinks about something. The demographics of prediction markets is nowhere similar to the demographics of the voting populace, especially in age. I would also assume that they might skew towards higher income, though I didn’t find any available statistics on user income. Taken together, using Kalshi market odds to predict the larger population can have some flaws.
Also, I think there’s some massive conflation between wanting something to happen (i.e., voting for a candidate) and believing that it’s likely to happen (i.e., betting that it will happen), whether or not they want it to. For instance, Kalshi made a post during the 2024 election showing that women on the platform bet on Trump winning the election at higher proportions than men. While they did disclaim that this was a market, not a poll, it was still clear at least a portion of their audience interpreted the data to mean women across America were more favorable towards Trump, which is nowhere close to what the data really showed.
With both of these issues, an election is unlikely to be directly affected. However, what if a prediction market could manipulate voting patterns? Election odds on Kalshi are calculated by the total money bet on a candidate, not the number of distinctive betters. If one individual put tens of millions of dollars on Trump winning the election, making it seem that he was much more favored, would that impact the actual result of votes? While attempts have been made to manipulate elections in the past, there hasn’t really been enough success for us to know for certain the effect it’d have. However, Andrew Hall, a professor of Political Science at Stanford, wrote of both the bandwagon effect and complacency effect. While he believes the effect would be pretty small relative to the amount of effort it would take, particularly in low-liquidity markets, the increasing usage of election markets might mean it’s something worth looking out for and creating safeguards against. An investigation was carried out after the 2024 U.S. election after an anonymous French trader received a payout of $85 million after placing a massive bet on Trump across four separate accounts, but no market manipulation was found.
I’m not much a gambler, but I find myself even pretty tempted by the idea of being able to bet on just about anything. While there are so many headlines of people quitting their jobs to day-trade on Kalshi, I think it’s worth remembering why it’s headline-worthy every time someone actually makes a profit. It doesn’t happen often, and all it takes is a scroll through online forums to see just how many people are losing. I’ve seen a lot of super interesting hypotheticals of the future of prediction markets, and while I don’t think I necessarily want them to be the future, I think it’s always worth researching to understand it fully, just in case.