McDonalds is a famous food brand. Not everyone likes what they sell, but many do, and under this brand they can reliably find a kind of food they like at a predictable price which is below their value.
Imagine you were hungry and came across a Capitalist Food joint. You’ve never heard of them before, but their pitch is that you should like them because they are capitalist, and all the best food comes from capitalists. E.g., McDonalds. Which is in fact true.
But this should not persuade you much to buy from them. Yes, if you could choose only between a generic capitalist food place and generic non=profit food place, you’d probably do better with the capitalist one. But the reason the best food comes from capitalists is that they usually develop much narrower food brands. Like McDonalds.
Imagine that you knew how to make a better yet cheaper burger. Most people who like burgers and who tried your burger for a half dozen times would conclude that yours are better. But by itself knowing how to make your better burger would not let you profit from selling them. Because you’d also need to create (or merge with) an acceptable brand to go with them.
For example, if your burger was branded with disliked and low status associations, people might avoid it even if your burgers were better. Such as being associated with the Russian side of their war with Ukraine when your customers live in Europe, or with Trump when your customers live in Seattle. Or being associated with insects, such as if your burger meat was made out of them.
Now consider prediction markets. We have good reasons to think that speculative markets are a great way to generate parameter estimates and decision advice. And many good people are now trying to sell this as a truth brand, that is, as a generic way to find truth. They set up a website where such markets can exist, put in a few sample claims, invite folks to suggest more claims, and step back. Somewhat like Capitalist Food as a brand.
But the thing that I’ve long been struggling to explain to these good folks is: that is too wide a brand to work well. Few people want truth in general. Yes decision theory says that people want truth near their decisions, and want it more the biggest their decision. But there are many kinds of truths that they positively do not want, and many more truths where their generally positive value for truth is below its cost of production.
In fact, most of the claims on most of these prediction market sites are actually of this sort: general world events, politics, and celebrity gossip topics. Topics where people care a bit about truth, all else equal, but aren’t much willing to pay to improve on the level of truth that results from the usual news, gossip, and punditry on such topics. A few people are willing to pay to gamble on these topics just for fun, and that can support a few small businesses that serve them. But that leaves the huge social potential of prediction markets unrealized.
A related failure happens when other good people see lamentably low levels of truth in public conversations, and decide that their fix is to just think honestly and carefully and tell the truth as they see it. The problem is that their audiences cannot reliably distinguish sources that are actually more accurate due to being truly honest and careful, from the many other sources that look just like those, yet merely like to tell themselves that they are being honest and careful, but are actually motivated and sloppy.
That is, these good people have failed to create a brand to distinguish their superior truth product. Most individually honest and careful people don’t live long enough or have consistent enough reliability to enable most audiences to distinguish them via personal topic-specific brands. So we mainly distinguish them via larger existing truth brands, e.g., via academic or news media brands. But to gain such brand approval, they must make the many usual compromises re honesty and truth that such brands demand.
A solution here I think is: application-specific prediction market brands. For example, a brand that specializes in estimating the chance of making project deadlines, sold to orgs that actually want to know if they will make their deadlines. Or a brand that specializes in estimating the two-year-later employee evaluation that each new hire candidate would have if hired, sold to orgs that actually want to evaluate new hires.
Such brands would invest in early trials, first to learn the many details of how to do these applications well, and then second to collect a track record proving such knowledge. And they would also do what it takes to acquire and maintain whatever prestige associations their customers demand, and to avoid disliked associations that put off customers. Which yes could be a lot more work than just putting up a betting website with a few sample questions on current events.
But this is the work that needs to happen to create narrow-enough truth brands to be useful. Don’t try to sell Capitalist Food, but instead create your version of McDonalds in the truth space. Find the particular kinds of truths whose value of use is plausibly more than its cost of production, learn how to increase value and lower costs in that particular area, and then prove your learning to potential customers via a statistically-validated track record.
That's a pretty niche market, which may well be served eventually. But first movers in a new market area should focus on the largest market niches to fill.
Request: can you try to figure out how a prediction market for NFL general managers would work for the draft?
That is, we know empirically that the consensus draft wisdom is more likely to be correct than the work of any single general manager or team. This is a field RIPE for use of a prediction market.
The problem--well, one among a few--is that if general managers outsource their power to these markets, at least publicly, they lose status. Moreover, they have a high degree of interest in gaining a competitive advantage over other teams, and thereby would want to keep their findings confidential.
My best guess would be to set up an NFL-specific prediction market (and branded as such!) anonymously but publicly, with questions like "Will Dan Sarmold be selected to the Pro Bowl within three years of being drafted?" and so on. Then the GM could base their decisions on the market without anyone being the wiser.