Election prediction is an interesting one. The first digital prediction market on the US was election only, and it has a $500 cap. It was run by a university and had a limited scope, but it still be a surprising amount.
It did a better job of predicting election outcomes than the polls. For markets without a cap, companies have used them to have against a politician that would pass policy that's can for their business.
This doesn't mean we should have them, but there is utility. It's just not worth the trade-offs
Prediction markets are a superior method for determining public sentiment, and understanding public sentiment is useful for many reasons. It can guide economic decisions, picking policy, etc.
Basically, prediction markets are for public sentiment what free markets are for supply and demand. They create an emergent system that balances/optimizes variables across a large domain.
It did a better job of predicting election outcomes than the polls. For markets without a cap, companies have used them to have against a politician that would pass policy that's can for their business.
This doesn't mean we should have them, but there is utility. It's just not worth the trade-offs