Robin Hanson 0:00
So the fundamental policy problem is how can we collect the information we all have about the consequences of our choices so that we can all pull it together into a form that we can share and use. And if you look at the various kinds of institutions we’ve ever used in the world to collect information and share it together into a consistent form that we can use together, the institution that seems to consistently work the best is speculative markets, betting markets, stock markets, currency, markets, commodity markets.
These markets have consistently take in giving people the incentive to find things out, bring the information that have to bear to this center, which is the market include their information in the current market price, and then allow that current market price to be information spread everywhere that tells us all what we all know. So we’ve had many tests where we have have a betting market or speculative market price compared at the same time to other markets. tutions that we often use like committees, or expert surveys, or all sorts of other mechanisms. And basically, we have many dozens of these comparisons.
And consistently, what we find is that either they’re both about the same in terms of the accuracy of their forecasts, or the betting market is substantially better, more accurate. That’s the consistent data we have. Now we have some good theoretical reasons for understanding why but I don’t want to make the pitch here to be based on that theory, it seems safer to make the pitch based on the empirics consistently, these betting market prices beat now you can understand in the sense that what they do is they offer a challenge to anybody, they say, look at my price. If you think this is wrong, come fix it and you will make money.
And that’s not only true with respect to sort of base things you might know about a factory burning down or something else. It’s about the patterns like do prices go up on Monday. There’s too much do they go down when it rains too much, etc. And any pattern you can find where you can see a mistake in the market, you can make money by betting on that pattern and making it go away. And that’s the powerful incentive that speculative markets offer.
And in general, we’ve had specular markets just on the random topics that have Rosen, like the price of gold or IBM. But we can make a betting market on purpose on a topic you care about, say even COVID COVID infections, and then get the information on the question you care about. So for example, if you have a project with a deadline, you want to know will I make the deadline? Typically, you ask people on the project will we make the deadline when you get back as a politically correct responsive whatever the leader wants you to think which is typically Yes, we’ll make the deadline.
Robin Hanson 2:51
And if you say something else, you might get punished for you know by the person running it for violating their their dogma, but if you have betting markets with people Bet anonymously, then you tend to get the accurate forecast. So we consistently see that people get accurate forecasts, and more accurate forecasts when we have betting markets on things.
And there’s an enormous potential for this to be applied across society and institutions, especially through what I call conditional betting markets or decision markets. I’ve given the name a few Turkey to sort of governance that way. And the key idea is that you can bet on the outcome of a decision conditional on which decision is made and betting markets on that can basically tell you which decision to make. So my favorite example is firing the CEO. So you have a company, you have a CEO, should you keep the CEO should you fire them? Well, usually a board makes that decision, but that’s relatively political.
And it seems like we make a lot of mistakes there. So the idea is you have a stock market where people trade on the stock and then you have a stock market. We let them trade on the stock conditional on whether or not we keep the CEO We have two prices, the price of the stock if we keep the CEO, the price of the stock if we don’t, and the difference of those prices tells us Should we keep the CEO. And that would be a straightforward way to give us advice on a key decision. And we could do that everywhere.
And so the bottom line is we have a lot of data on prediction markets in many different contexts, showing that the they basically work, we have a lot of innovations and better ways to make them work better and new contacts. We have a lot of tests of those. But what we don’t have is a lot of adoption. And the key problem is that they are very politically disruptive. So for example, if you think about this person with a deadline, and wanting to know if he’ll make the deadline, he can make a prediction market on the project and that will tell them that they will make the deadline.
But the person writing the project cares more about having a good excuse. If he fails, then lowering the chance of failing. And so typically a person with a project deadline, their best excuse if they fail will be To say, well, things were going along fine. And then at the last minute, something came out of left field, no one could see him coming. noctus flat, it’s so rare and strange, it’ll never happen again. So there’s no need to hold anyone responsible here. Let’s just continue on. And the person running the project wants that to be the excuse, and their boss.
And even boss’s boss wants that to be the excuse, because they all would look bad if they were in charge of a project that failed for no reason that could have been prevented. So if you make a prediction market on this project, you could even do conditional versions, which will tell you how to make the project more likely to make a deadline. But the key problem is, if the market says you won’t make the deadline that will tell you long in advance, consistently and then you can’t have this excuse that at the last minute, we will not buy something at a left field.
Everybody can say look, the market told you you were going to fail. You didn’t you didn’t prevent the failure, you’re at fault. And that’s an example of how in our world political disruption of prediction markets tends to make People not want to adopt them. That is most people running projects and running organizations care more about defending themselves from blame than they do about making the overall organization worthwhile.
Anthony Pompliano 6:12
Yeah, this reminds me a lot of the book super forecasters. And so would it be fair to say that you think the stock market is a prediction market as it exists in today’s form? Or does it not fit that definition?
Robin Hanson 6:25
It does, but the thing it’s predicting is not very easy to interpret. You know, it’s like if aliens had a betting market, and you could see the prices in the betting market, but you didn’t understand what the words were the best, then you wouldn’t be able to make much use of it, right. So the stock market is a bet on the value of that company. But that’s a complicated thing. And so it’s hard to apply that insight and information to other things you care about. So the more that we can make betting in speculative markets more directly on the things we care about, including especially the decisions we make The more useful we can make
Anthony Pompliano 7:03
it, how does this get applied to government? Obviously, that is like the ultimate decision making machine that probably has the widest impact on a citizens life. How do we incorporate the ideas of a prediction market into that bureaucratic process?
Robin Hanson 7:20
Well, mechanically, it wouldn’t be that hard to do. The harder problem is do people want to. So what I recommend is that we start with small scale experiments and work our way up. So instead of trying initially, to apply prediction markets to government, we apply them to small clubs, small companies, small organizations, and slowly show through a long, you know, detailed track record that this works and in each case, try to entice neighboring organizations and slightly bigger versions to copy what it seems to have worked at a smaller scale.
And then eventually we could get government so as you may know, government isn’t usually the initiator of a lot of innovations in society, but it tends to belatedly copy them when something works well enough elsewhere. And so that’s probably the best route to produce innovation as government is to produce in a similar innovation in the private sector, and then shame the government into copying.
But in order to inspire people about what should be eventually possible, I do sometimes describe government applications to give you a sense of just how far this can go. So for example, the simple simplest thing might be just there are many government projects with budgets that lie. That is, there’s an official budget, but then later on, then I come back and ask for some more money. And they didn’t admit ahead of time. They wanted that extra money, but they kind of knew.
So you can have a betting market on the budget for a project that says, well, eventually, how much will this cost? And you can even have that conditional if we adopt this project, how much will it cost eventually, and now we can get a more honest budget, on how much things will cost and of course, people Who are pushing these projects now don’t want that to happen because they tend to want to convince us to adopt their project by believing it won’t cost as much as it really well. freeways cost more, and all sorts of NASA projects cost more than they initially announced. So that’s just one simple example of how you can use it. But these conditional versions of the market would let us make big key decisions. Like you might justify building a stadium by saying, Oh, look at all the revenue we’ll have if we have a stadium.
But we might say, well, if we build the city, how much revenue will we really have over these coming years? And then that could be more disciplined our choice of do we build a stadium? We could even do this at the, you know, presidential election level, we could have betting markets and say there’s the two you know, Republican and Democrat. You know, what are the outcomes you care about? pandemic deaths, war deaths, GDP, you know, stock market, whatever it is even international respect. We can have betting markets in each of these things condition along with which candidate is elected?
Actually, today there are betting markets for which nominee should each party produce if they want to win the election. So there are betting markets on who will win the election and also on who will be nominated. And you just have to divide those two numbers to get the probability of being elected if nominated. So, those betting markets could be much thicker and much more accurate, and then they would give a much stronger message to each party of who to nominate. Of course, the parties are pretending that they really want to win. But often the insiders care more about who’s nominated and who wins so they may not welcome such things.