Prediction markets are incentive machines · ↗ www.forbes.com

Robinhood CEO Vlad Tenev calls prediction markets “truth machines”. This isn’t quite right, though. Prediction markets are incentive machines.

One incentive is to bet using inside information—or, as Polymarket CEO Shayne Coplan put it: “What’s cool about Polymarket is that it creates this financial incentive for people to go and divulge the information to the market.”

Another incentive is to create the desired outcome, such as by manipulating temperature sensors or throwing sex toys at female athletes. In the same vein, Kalshi reportedly paused plans to offer bets on flight cancellations, fearing that bettors might deliberately disrupt airports.

Luckily, this lesson apparently applies narrowly to airport cancellations and not, say, clinical trials and FDA drug approvals, which Kalshi recently started offering bets on.

This is an objectively terrible idea because, as with sports gambling, it shrouds the whole endeavour in uncertainty and suspicion. Kalshi says it will only open markets after trial enrolment has closed, which at least prevents betting odds from discouraging patients from signing up. But bettors still have incentives to convince enrolled patients to report or exaggerate negative effects. The public may also become more skeptical of regulatory decisions now that insiders have another way to profit from confidential information and outsiders have an incentive to undermine the process.

The good thing is that Kalshi has limited bets to Phase III trials and FDA drug approvals, though it is open to expanding the program later. Phase III trials are often double-blind—neither the patient nor the investigator is told which treatment the patient receives—and generally involve hundreds or thousands of patients. Both factors make the results more difficult to manipulate. Tampering is less useful if bettors cannot tell who received the treatment rather than the placebo/control, and the scale of Phase III trials means that many more patients would have to be turned to produce a meaningful effect. But blinding is not always successful or feasible, depending on the treatment under investigation.

If betting expands to Phase I trials, manipulation becomes much easier. Phase I trials are conducted mainly to establish the safety and tolerability of a new treatment, may involve only a few dozen patients, and are commonly unblinded and uncontrolled. At that scale, a few well-placed adverse events can derail the entire clinical pipeline.