Prediction markets are an interesting tool for determining the most likely answer or result for almost any question. Some economists hope that we could use prediction markets to make difficult political decisions. Many companies already use them internally.
Sadly there's only one prediction market that's accessible from the US, and it's an academic project with a 10% fee on profits and a 5% fee on all withdrawals. It's also limited in what you can predict or bet about. Augur or some other decentralized solution is probably our best bet for developing prediction markets further.
I really think these markets are the key to overcoming our failure to make good political decisions. It is a wild idea, but I don't know of any other solutions that directly incentivize truth-seeking.
Vitalik Buterin actually wrote about these concepts in more length
I participate in Good Judgment, but I think it will be great to compare Augur's results on certain questions to GJP's results (whenever there is a question that overlaps at least).
> Augur will probably be used for betting but really it's a prediction market.
A prediction market is an attempt to leverage a hypothesized predictive power of aggregated
betting, so being used for being is essential to being prediction market.
> Some economists hope that we could use prediction markets to make difficult political decisions.
Political decisions aren't verfiable fact questions, so they aren't really subject to prediction markets. You could use them as a tool to get possible answers to fact questions supporting policy decisions, but there is no real compelling case for them being better than the best available other methods in most domains for that, and in any case getting answers on fact questions related to policy decisions isn't even the hard part of getting good policy decisions.
> > Some economists hope that we could use prediction markets to make difficult political decisions.
> Political decisions aren't verfiable fact questions, so they aren't really subject to prediction markets. You could use them as a tool to get possible answers to fact questions supporting policy decisions, but there is no real compelling case for them being better than the best available other methods in most domains for that, and in any case getting answers on fact questions related to policy decisions isn't even the hard part of getting good policy decisions.
i wish this comment/POV had gotten more traction in this thread. i am increasingly of the opinion that building consensus on values and goals is more of a problem in politics than is building consensus on facts.
In direct democracy, you vote for what you want to happen. In a prediction market you vote for what you think will happen.
The whole point is that a sufficiently large and diverse groupe of people is better at predicting the future using those system than an individual. Being able to correctly predicting the future can lead to better decisions.
Yes, it works very well but not for the 'wisdom of the crowds' effect. It works because the people with the most information and best models sharpen up the price. Over time, the sharpest bettors grow their bankrolls or capital exponentially and place larger bets and have a disproportionate weight on the markets and their opinions matter more, making markets more efficient.
In theory you're right - but this isn't how it works in practice. The people who earn the most money are the best traders, not necessarily the people with the "most information and best models." The overall forecasts tend to be accurate, but you cannot look to individual participants, see who has the most money, and assume they know the most about the subject at hand.
Both right: on an individual level the people with the most money might or might not be the smartest, but in aggregate the money moves towards the smarter bettors, weighting their votes higher in future rounds, improving the aggregate intelligence of the market.
The stock markets are basically prediction markets. Over the short term they are wildly inaccurate and are determined mostly by fear, greed, and crowd psychology. But averaged over the long term they are very accurate.
but are they accurate in a predictive sense, or only in their reaction to events? If the latter, which I think is more likely, it seems kind of a useless way to try and answer questions...
The idea is that they create an incentive to discover all possible information that may effect the price of the stock in the future, and so the eventual market price ends up reflecting all that information. That in turn lets people use that price as a signal for how to allocate capital and make decisions in the future.
It's not that they're totally accurate, it's that they're as accurate as it is possible to be. Why? Because if you knew of a better way to determine the truth, you would do so, trade on that information, bump up (or down, if shorting) the price to reflect that information, and take your profits so you could do so again. Conversely, if you're just trading randomly, somebody with better information will take all your money and you won't be able to continue trading.
The price that's reflected in a market could be considered residual uncertainty that nobody knows, i.e. if the prediction market for "Will Donald Trump win the 2020 election ?" sells at $0.60 on the $1.00, it means that after all the market participants have crunched their models, aggregated their polls, gone out and canvassed neighborhoods, ran their economic analyses, spoken to campaign insiders, and whatever else they can do, the equilibrium price assumes he has a 60% chance of winning. You don't have to actually perform all those information-gathering tasks to figure this out, you can assume that other people have been incentivized to do so and their results are now reflected in the price.
Plenty of market participants are not actually doing this. Day traders, high speed traders, even index funds come to mind. And even "real" investors are influenced by psychology that has little to do with an impartial analysis of all possible information. (It's not what you know, but what
"everybody knows that everybody knows" [1])
Sure, but markets ensure that traders who don't effectively do this lose money and get flushed out. That's the case with most day-traders, who are losing money they would otherwise put into consumption and using it to get an emotional high off price movements. Gambling, basically.
Quant high-speed traders (and arbitragers) are using information in a different way. Typically, the way these strategies work is that they discover correlations between prices. For example, when China enacts tariffs on soybeans in response to Trump's tariffs on steel & aluminum, that's going to depress demand (and hence prices) on soybeans. However, it will also have a lot of knock-on effects on other industries: shipping companies transporting soybeans will lose money, railroads connecting soybean farms to the Great Lakes will lose money, railroads connecting soybean farms to domestic demand will gain money, ranchers and chicken farms and other domestic food producers who use soybeans as feed will gain money, commodity traders could go either way, house prices in wealthy Chicago neighborhoods might increase if commodity traders are making a killing, etc. Say that the inside information entering the market is "China is enacting tariffs on soybeans". A bunch of people with that information will short soybean futures, and then it's the job of an HFT to spread that information through all the other markets that it may affect, updating the relevant prices on their securities before the profits actually trickle down through the economy.
Index funds are the freeriders of the markets. The philosophy there is that all of the work that people do is going to produce some value, and the shareholders will capture some of that value as returns to equity. And if the rest of the financial system is functioning effectively, all of this will be reflected in prices. So rather than worry about what the price will be, just buy & hold securities in proportion to their value, let the price mechanism do its job, and profit in proportion to the initial amount of capital you put in.
What does being "predictively inaccurate" mean in the context of the stock market? I assume that any time a stock jumps or falls is a symptom of inaccuracy--investors were surprised by some previously unknown--and unpriced--event.
So the low volatility of the stock market would indicate that it is actually predictively accurate, and that it is getting more so.
Then again, volatility is kind of a silly measure, because we also care about the frequency domain. The market could be very good at predicting the very short term or the very long term or both.
I wonder what a Fourier transform of the market would show?
The Signal and the Noise by Nate Silver talks about this a lot. From predicting politics to economics to sports to climate change, group predictions are almost universally better than individual expert predictions.
In theory. But from my experience, it's the opposite, and you can tell by the comments. I made a lot of money betting on Trump to win PA, FL, and POTUS on predictit. The only one I was unsure of was POTUS, but it was good odds so I took a gamble. The fact that FL, and to a lesser extent PA, had underdog odds meant people weren't betting logically. And the comments were akin to youtube comments.
Is there any well established study which outlines this? It feels very counter intuitive that masses are better at prediction than specialists of that domain.
Well, a simple average is a pretty reasonable ways to make predictions when you have no better information... like a regression model with 0 predictors.
But really the idea is more about a feedback mechanism in which successful predictors those who are successful at making predictions end up with more money to invest back into the market making more predictions. So its a weighted average of the masses, in which those who are successful predictors get weighted more heavily in virtue of their having more financial capital with which to invest. For that to work well, you really need a reasonable level of participation. Also, if the inputs from new participants continually outweighs those from experienced quality participants, you are not going to get good predictions. Also, you have to have a rock-solid mechanism to decide the market, or it will be gamed.
You should visit the good judgment project website. They've studied this for a while and have documentation outlining and explaining their claims.
The idea is not that an untrained mass of people are necessarily better than experts. Instead, it's that some people, with training, can learn how to make good predictions about many different fields. These predictions exceed the predictions of domain experts.
Sorry, no citation handy, and this might be semi-tangential (more "wisdom of he crowds" per se), but I recall a study that reviewed those "guess how many jellybeans are in this jar" -type contests, in which the guesses' median value converged with the correct answer...
not sure how they are comparable to begin with, since one is people guessing an outcome, and another is people voting for an outcome. in the first, you are trying to guess how everyone will vote, in the second you say what you want. sorry if thats not what you are actually asking.
theres definitely more than one prediction market available in the us. for instance, https://www.predictit.org/ (assuming you were referring to IEM in your comment)
The iCash team is working to solve the governance issues in blockchain. Their work towards creating an automated & self-sufficient economy has really impressed me.
Especially being their work is in the prediction market space and covers a lot of the current limitation issues we face.
There is a much larger global prediction market that comes with its own set of nuances and quirks, and that of course is the stock market. You can even get involved with no fee at all through Robinhood, but you will have to fill out some annoying and potentially costly IRS forms.
If I could paraphrase Charlie Munger, in the stock market, the only people who are successful long term are those who bet massive sums, very rarely, when they are sure that the price of a security is severely and unproperly valued.
If Tesla is currently trading at $355, and you have evidence for yourself that it should be worth something like $0 or $3,000, then this would be an example of the current market prediction being extremely wrong.
A stock's price is really just a prediction of the company's future cash flow in the event of profits, losses, sale of assets, so on.
Actually, there are also people [0, 1, 2, 3, 4, 5] who are very successful betting small sums (relative to their total investable capital), very frequently, on securities they believe are slightly over or under valued.
He was daring a human-managed fund to outperform S&P500. I don't think Warren is an authority on market's microstructure and quant trading, which these companies exploit.
Basically on very quick trades machines outperform humans. For long term investment a prediction market (stock market index) outperforms humans and machines.
It's called survivorship bias. With hundreds of millions of people making investments in the stock market, almost any strategy has many examples of people who "won" with it.
These firms make many thousands of trades per day, year after year. They have multi-year track records of exceptional returns.
If their edge was truly no better than random then they would have been bankrupted years ago. You can do some back of the envelope math to easily convince yourself of this.
nobody said it's totally innefective (or random), just that it is vastly less effective (and you're exposed to terminal risk more often) than throwing lots of money on an index.
prediction markets are not well regarded academically , outside of the cristall ball economics.
the academic consensus is that as soon as you create a second incentive (the prediction market itself) you lose all the insight you could gain from external factors. At that point they are no better than simply looking at the stocks (the predicted market). hence prediction markets are by definition, useless.
the cristall ball people still trying to make it happen are now trying to see how much they can limit the prediction market itself for it to not affect itself (hence the limitations you don't like on the one you mention)
The stock market is already a prediction market. Using a prediction market to predict a prediction market is a silly idea. But that doesn't make prediction markets bad at other things, like predicting the weather.
> What about a market about the winner of the next US presidential election?
IIRC, those exist, were the focus of attention a few cycles back (because cherry picked just right they were arguably better than polling in the cycle preceding the attention), but as soon as media started paying attention to them, people started diverting money to them to manipulate the results, because there are plenty of people that care more about influencing the result of that question (and it's one where perceived likelihood influences the actual behavior that controls the result) more than predicting it.
That kind of thing is actually a significant problem in many of the areas where prediction markets have been hyped.
> What's the evidence for people putting in large amounts of money to influence betting on elections?
Going back and reviewing things it seems my memory was wrong: the risk of coordinated efforts of this type if the prediction markets became seen as strong bellwethers of electoral success was discussed in the wake of some isolated, apparently single-trader incidents that momentarily moved the prediction markets in 2004 and again in 2008, but there doesn't appear to have been any detected widespread manipulation.
> Isn't it making a bit assumption that people vote in a certain way because they see things in prediction markets?
No, there's a pretty strong well-known effect that media coverage of electoral momentum reinforces that momentum; it's not doing much changing cited as impacting enthusiasm and turnout.
> Regular people are surely not checking odds on who is likely to win an election.
People—especially the people who are reasonably likely to vote—to consume political news, including news relating to whatever the media sees as strong horserace indicators. For a few cycles startibg around 2000 that had started including prediction markets, though it seems to have faded the last couple cycles.
The thing is it is supposed to solicit insider information exactly like this. Potentially it incentivises people to change outcomes if the have the resources available, but at least we will all know in advance.
To quote a prior Senate Minority Leader Tom Daschle on the floor of the Senate:
"How long would it be before you saw traders investing in a way that would bring about the desired result?"
As in, if one found a way to bet against your car remaining in tact and realized sabotaging your car was cost effective to manipulate their bet, then they reap the market pay out. Is this good?
> As in, if one found a way to bet against your car remaining in tact and realized sabotaging your car was cost effective to manipulate their bet, then they reap the market pay out.
You just described insurance fraud. And just like in insurance, the fix is to be very careful about how you describe the "win" condition—not just "will the car become damaged" but "will the car become damaged through no fault of the owner" (to begin with; the actual terms would be much more complex).
We already have a much more liquid terrorism futures market, i.e. the regular stock market. Worrying about incentives for a large attack from a prediction market is missing the ocean for a hypothetical raindrop.
Incentives in smaller markets (e.g. local elections, or personal car crashes as you say) are super interesting though.
Interesting...so when enough "life insurance" has been taken out on someone, it becomes more profitable to protect them or to kill them, depending on what the relative odds are, and which side the potential assassins have placed their money?
The thing I don't like about prediction markets is the capped liquidity. I want to put my money where my mouth is but then find out there is only 1 share to buy in the market.
What I like about financial/forecasting crypto projects like Pareto Network is that there is no capped upside, and you can still determine the conviction some people have about an outcome
A close runner up is Numerai, given that you can tell the 'confidence' someone has, but the capped upside is arbitrarily capped, and the purpose of the project has nothing to do with censorship resistance and is beholden to the whims of the Numerai hedge fund.
I'm confused. With Augur, "complete sets" can, aiui, be created by anyone, so I don't see why there would be a problem with only one share being available?
Unless I guess if you want to make sure the trade goes through immediately?
the problem is the limited amount of money one can make off of a random market.
makes it less appealing to some (or many) and less efficient in determining an outcome
yes assassin markets and popular political events can now be traded in the US even if the Augur company gets sanctions, but for everything else this is an inefficient system.
Well, the money has to come from somewhere. In any case, someone has to be willing to take the other side of your bet, whether it is other participants or the people running the market/betting thing.
I'm not sure what alternative you are considering to not have that problem, or what it would look like for something to not have that problem.
> determine the conviction some people have about an outcome
But in order to do that, wouldn't you also have to know a lot about those people?
If you bet $1 and I bet $50 that could just mean I have a lot more money than you. Or for a million other reasons, I'm pretty likely to have not exactly 50x your conviction. I may even have less conviction than you.
And even this is assuming good faith, because I might have reasons other than "winning" for placing my bet.
https://en.wikipedia.org/wiki/Prediction_market
Prediction markets are an interesting tool for determining the most likely answer or result for almost any question. Some economists hope that we could use prediction markets to make difficult political decisions. Many companies already use them internally.
Sadly there's only one prediction market that's accessible from the US, and it's an academic project with a 10% fee on profits and a 5% fee on all withdrawals. It's also limited in what you can predict or bet about. Augur or some other decentralized solution is probably our best bet for developing prediction markets further.
I really think these markets are the key to overcoming our failure to make good political decisions. It is a wild idea, but I don't know of any other solutions that directly incentivize truth-seeking.
Vitalik Buterin actually wrote about these concepts in more length
https://blog.ethereum.org/2014/08/21/introduction-futarchy/