Use Prediction Markets To Fund Investigative Reporting


Hindenburg Research has a great business model:

1. Investigate companies
2. ...until they find one that is committing fraud
3. Short the fraudulent company
4. Publicly reveal the fraud
5. Company's stock goes down
6. Profit!

I've been thinking about them recently because of the debate around funding investigative reporting. It goes something like: investigative reporting is a public good. Everyone benefits from knowing about Watergate. But it's hard for investigative reporters to capture the value they produce. Very few of the people who cared about Watergate bought subscriptions to the Washington Post. There's no reason to - you can let the Washington Post uncover Watergate at no cost to you, then hear about it for free on the nightly news. The traditional solution is bundled media. Newspapers have their profitable bread-and-butter in the form of easy things like commentary and sports, then do some unprofitable investigative reporting on the side to gain prestige.

This traditional solution is failing, because the Internet unbundles media. If you want commentary, you can get it here on Substack; if you want to know who won the big game, you can go to or just Google it.

I feel like this is mostly good. It lets customers pay for things they want and not things they don't want. It diversifies media, preventing a few big gatekeepers from spinning everything to fit their political agendas. And it lets journalists capture the full value of their own labor instead of surrendering most of it to big corporations. But obviously gatekeepers and big corporations are sad about it. And the argument they've come up with for why it's bad is that it threatens investigative reporting. This is honestly a pretty compelling argument.

But: if there were good prediction markets, you could fund investigative reporting the same way Hindenburg Research funds its investigative reporting.

Right now there's about $25 million on Predictit saying that Eric Adams will win this November's election for NYC mayor. Suppose some kind of media property uncovered evidence that Adams took bribes or committed some other career-ending indiscretion. In a world where prediction markets were as liquid as stock markets, they could short Adams, reveal his dirty laundry, crash his election prospects, and collect some portion of $25 million.

Even if their dirt isn't quite that good - even if it only makes Adams seem a little bit worse than people thought - it would probably affect his chances by 1 or 2 percent. And that's still $250K to $500K - more than enough to hire a couple of good investigative reporters.

What if Adams had such monolithic support that nothing a dedicated investigative reporter turns up could possibly shake his chance of winning by 1% or even 0.01%? Maybe then we should just accept that the market is correctly telling people not to bother investigating Adams. If we've gotten so polarized that no conceivable fact could change who we vote for, then investigative reporting should have the same status as golf - a weird activity that rich people like to talk about, but which has no ability to change the world for the better. We wouldn't be trying to think up clever mechanisms to publicly fund golf, so there's no reason to be sad that this particular clever mechanism fails at funding investigative reporting.

But what about important stories that can't easily be translated into prediction markets? I'm looking at the front page of today's Washington Post, which has something about how there's a culture of sexism in some important military school. This could charitably be termed investigative reporting - we can imagine the journalist painstakingly tracked down women at that military school, heard their stories, hunted down documents proving that this school discriminated against women in some way, et cetera. But even assuming it's true, how would you translate this into a prediction market? Maybe some important person is more likely to fire the head of this military school now that they know he allowed a culture of misogyny to continue there? But number one, this would require there to be prediction markets in the careers of every military school principal. And number two, this relies on some pretty strong assumptions about how likely it is that some head honcho really wants to fire anyone who perpetuates a culture of sexism. There are potential ways to get around the second problem - you can imagine that, this having been revealed, the school will take actions to change it, and the self-reported life satisfaction of women in the school will go up - a measurable result. But just brings us back to problem one again - you have to have a pre-existing prediction market on the life satisfaction of every demographic group in every school in the country!

I think I bite these bullets. First of all, of course there should be fire-the-head-of-a-military-school prediction markets! Robin Hanson has long said that if he had a million dollars, the first thing he would do is subsidize fire-company-CEO markets. Why not give him a billion instead, and have him extend it to military school principals?

And second, yeah, I even think there should be life-satisfaction-of-female-students-at-this-particular-school prediction markets. If current-day schools care enough about gender equality to hire a chief diversity officer (which they definitely do), future schools should care enough about gender equality to subsidize prediction markets in the self-reported life satisfaction of female students. It'll be cheaper than the CDO and more effective. I want a world where protesters march into colleges and ask them why, if they claim to care about female students, they don't have a subsidized female-student-life-satisfaction prediction market, so that they can credibly set targets for gender equality (see this post for more on how that would work).

I realize this system is a weird proposed funding model for a newspaper. I'm sure people will think up lots of perverse incentives it might cause, and some them might even be true. But it's not like our current system doesn't cause a lot of perverse incentives. Prediction-market-moving newspapers would get right a lot of things that our current model gets wrong. For example, our current model incentivizes fake news - if something is exciting and outrageous enough, it'll sell a lot of papers, and even if a few spoilsport researchers come out a few years later with definitive proof it was wrong, by that time nobody will care. In a prediction market, once you're wrong a couple of times, traders will stop updating on your reports and you'll lose most of your power to move the market. You'll have no particular reason to care whether what you write is outrageous or exciting, only that it's true and relevant to something that people care about.

In the far future, the news might diverge into an "infotainment" industry where people try to get clicks by accusing each other of being racist pedophiles, and a prediction-market-movement industry where people try to generate true information on important issues. Then everyone could just ignore the first thing and go about their lives.