July 7, 2012

Failure of prediction markets

David Leonhardt writes:
IN the days leading up to the Supreme Court’s health care decision, rumors began to circulate in Washington that the justices had decided to uphold the law. Liberals around town who might have reason to know the outcome seemed happy, according to the gossip, and a couple of conservative justices had seemed angry when the court met three days before the announcement. On the eve of the ruling, a few respected court watchers went so far as to predict publicly that the law would be upheld. 
It is impossible to know how much of the gossip sprang from actual information. Several dozen people — the justices, their clerks, other members of the court staff — did indeed know the outcome in advance. Although they have a record of discretion exceeding even that of some parts of the national security apparatus, they are human. They have friends and relatives, and they have emotions. 
With the rumors swirling, I began to check the odds at Intrade, the online prediction market where people can bet on real-world events, several times a day. The odds had barely budged. They continued to show about a 75 percent chance that the law’s so-called mandate would be ruled unconstitutional, right up until the morning it was ruled constitutional. 
The market — the wisdom of crowds — turned out to be wrong.

As I pointed out in a Taki column about the pretend Hollywood Stock Exchange for predicting box office revenue, markets have several advantages over individuals taking SWAGs about upcoming events: random noise and individual bias are reduced. You thought Hunger Games was kind of stupid? Well, Mr. Smartypants, a lot of people thought it was pretty smart. 

But, a huge part of the value comes if inside information plays a sizable role in setting prices. For example, HSX.com encourages insider trading. It's just pretend money, so nothing is hurt (except outsider players' pride) when somebody who has seen the studio's marketing research uses it in playing HSX.

Why didn't InTrade reflect inside information on the health care decision? I don't know. Perhaps the ratio of rubes to insiders among the players has swollen dramatically. Perhaps the insiders didn't play the game fearing that they could go to jail if found to have profited from this hot info. 

My larger point about prediction markets is that most of the stuff we get excited about in the short run wondering about what will happen isn't all that important in the long run. For example, as I pointed out on the Forth of July, the constitutional history of the U.S. and Canada is quite different, but, in the very long run, famous events haven't played that big of a difference in why daily life in Vancouver is more like daily life in Seattle than daily life in San Diego is like daily life in Mexico. But there's not much of a betting market for that kind of wisdom, which most people find boring and depressing.

In fact, a few years ago I came up with the dullest breakthrough insight in the history of the philosophy of science. (Or, more realistically, I'm only the latest in a long line of individuals who have figured this out, but nobody else was interested enough to make the names of my predecessors familiar.) Namely, the reason the social sciences are universally believed to have a poor track record of making predictions is that the kind of things that people are most interested in predictions about are those things are most likely to go either way. The closer something is to being a coin-flip toss-up, the more exciting it is to speculate about what will happen. 

In contrast, the more reliable the evidence there is for a prediction -- e.g., one I've been making for 40 years: San Diego will tend to be a nicer place to live than Tijuana -- the less people are interested in it. 

People always tell you that physics is a true science because astronomers can predict an eclipse 10 years from now, but the social sciences aren't true sciences because they can't predict accurately whether the stock market will go up or down tomorrow. 

Okay, but isn't it obvious that the price of the stock market today is set by a huge number of people all trying to guess where it's going to go, which drives the stock market in the direction of the consensus? In contrast, nothing anybody does has any influence on the orbit of the moon.

Consider the Supreme Court's ruling on the health bill? Why was it of so much (temporary) interest? Because it was hard to tell which way it would go. It was an anomalous case and thus hard to predict. The activities of lawyers and judges are voluminous, but most is of trivial interest to the public because it is relatively predictable. In fact, Oliver Wendell Holmes Jr.'s view of the law said that the chief aim of the law should be predictability rather than justice. This was something of an anticipation of Coase Theorem: if people can predict how the law will be applied, they will adjust their affairs accordingly.

So, my view is that it makes sense to study the implications of what the social sciences are good at predicting. If San Diego continues to be, as I've been predicting for 40 years, nicer than Tijuana, don't avert your eyes, but ask why is that? What implications does that have for policies, such as immigration?

26 comments:

DCThrowback said...

A 9-10% drop (from, say 80 to 70% the day of) is still quite significant. In fact, one might say it was similar to the Pacquiao/Bradley odds, which dropped from Pacquiao being -600 (laying $600 to win $100) to being -400 (laying $400 to win $100) the day before the fight. One could argue, reasonably, that "somebody knew something" and that may have led to the "wise guy steam" that moved the line. I know when I found out that the upheld on intrade dropped from 80% to 70%, I had a sneaking suspicion that a Pacquiao/Bradley-like upset was on the horizon. The line drop (sharp money) tells you all you need to know (most of the time).

Anonymous said...

intrade is thin, you typically can't make real money on inside trading.

nazgulnarsil said...

How is this a failure? People can doubt the veracity of insider info.

AL said...

It's ancillary to the thrust of Steve's post, but Leonhardt is incorrect in concluding that "the market... turned out to be wrong." If I were about to roll two dice, the market would give me less than a 3% chance of throwing a twelve. If the roll in fact turned out to be twelve, it wouldn't mean the market was wrong. Likewise, if a stock traded at $20 yesterday, but today closed at $25 after strong earnings were revealed, that doesn't mean that yesterday's price was wrong. Yesterday's price incorporated the information available yesterday, and now we have more information.

Under Leonhardt's system, a market would be "wrong" if it priced the likelihood of the eventually-realized outcome at anything other than 100%!

Anonymous said...

In fact, Oliver Wendell Holmes Jr.'s view of the law said that the chief aim of the law should be predictability rather than justice.

Technical point that probably no one but a lawyer will care about: The above is a mischaracterization of Holmes's view and the so-called "prediction theory of law." The prediction theory is not about predictability. On the contrary, it's about the inescapable "judge-iness" of law. It's based on the insight (or assertion) that law doesn't exist until a judge makes it by deciding a case. Thus, in Holmes's aphorism, "the law" is just "a prediction of what some judge will do."

I should add that when Holmes wrote that, most law was still judge-made law, as opposed to law made by legislatures and bureaucrats. Now, about a century later, the opposite is the case.

- A Solid Citizen

Henry Canaday said...

Economists get a bad rap because they are often bad at predicting close calls that involve many imperfectly understood relationships and variables, such as exactly how the economy is going to behave over the next quarter or two. But economists could provide, if not predictions, very useful perspectives on such larger and important subjects as the reason for Tijuana-San Diego difference. But this would require economists to integrate their models with research from other fields, and these other fields involve people, which can be a very sensitive subject.

Consider this contrast: Yale economist Robert Nordhaus has for a long time been building increasingly ambitious models that integrate long-run macroeconomic performance, a tough challenge in itself, with the highly complex effects of fuel use, carbon emissions, climate change and the economic effects of climate change. His work is highly useful and widely respected. But how many economic models have you heard of that connect macroeconomic performance with: a) the demographic composition of the population; and b) the decline in two-parent families, even at birth, and the effects of this decline on adult behavior?

Clutch cargo cult said...

Re: AL's comment, leonhardts statement about the market being wrong is just another plug for government intervention not about the veracity of the bet.

Daybreaker said...

People and things that you are tacitly but effectively forbidden to talk about in the wrong way are problematic for making good predictions. Not only do you miss out on the predictions covered by forbidden utterances, but people who can make things forbidden and keep them forbidden tend to be powerful and driving events.

If you the great seer are performing in Rome during the reign of the Emperor Little-boots, all predictions based on the emperor being violently insane are right out. Predictions that involve the Praetorian Guard, what it may do soon, and the long term effects of praetorianism are if anything even more forbidden.

If you were performing in London in the court of the Virgin Queen, you would be badly advised to lay out an accurate history of England up to the Glorious Revolution. The consequences of the ruler's failure to produce an heir and a working dynasty (and her unwillingness to let people talk this through and come up with alternative arrangements) were so big that you might be reduced to saying that you knew the future and (except for the parts you couldn't talk about) it was pretty dull.

It's worse if you are a social scientist rather than a performer with a crystal ball and a flashy silk robe, because then you have to "show your working," still without any reference to forbidden things and forbidden people. Your an analysis has to start by omitting key disruptive factors, and follow through with unswerving consistency, and be correct, a tough trick.

Anne said...

Slightly OT, it would make more sense from a Canadian or European viewpoint to just hold a referendum on healthcare rather than leave it to nine individuals parsing the US Constitution in a Talmudic way.

(I am Canadian).

sunbeam said...

The Supreme Court decision didn't surprise me. At least not Roberts' decision.

The rest of the eight justices were pretty predictable honestly.

As far as Roberts go, I wasn't surprised, because I believe he is some kind of ... I don't know what... corporatist maybe?

I can't really see him as a Randian, nor a Libertarian for sure.

He will vote conservative (whatever that means exactly) a fair percentage of the time, because "conservative" correlates with his core beliefs to a fair percentage right now.

But at the end of the day, this decision was for the best for the insurance companies. And I believe that is why he voted the way he did.

And that is the sum total of his principles. Not for insurance companies specifically, but for companies in general.

I really thinks he puts the friendly in fascism, and yes I'm not just throwing that out there. I think Giuliani would be his ideal President.

He is a lot more like Mitt Romney than the rest of the court I think. If things had been different he might have been a character in "Wall Street."

BTW, I have the world's worst gaydar, and I know about his dancing kid (Adopted, from Ireland. You can take the kid our of Riverdance, but you can't take the Riverdance...) at his swearing in. But I do think he is gay.

BTW I am curious Steve. Did you ever read any of the "Masked Political Consultant's" three posts on the late Joe Bageant's website? Opinions differ, but he pretty much nailed what I believe about current American Politics.

Of course old Joe was on the left side of the fence, but damn those were some insightfult posts over there. I guess I need to google and save them, I think that website is caput now.

At least I still have Fred Reed.

Silver said...

In fact, a few years ago I came up with the dullest breakthrough insight in the history of the philosophy of science. Namely, the reason the social sciences are universally believed to have a poor track record of making predictions is that the kind of things that people are most interested in predictions about are those things are most likely to go either way.

By "a poor track record of making predictions," do you mean

(a) the track record of the predictions they have made is poor, ie they are too often wrong?

or

(b) that they have a poor track record of making predictions, ie of going to the trouble to make a prediction, ie they don't make enough predictions, ie they could presumably be making many more?

or

(c): (a) and (b)?

Whatever the case, I fail to see how the reason/explanation for any of (a), (b) or (c) is that "the kind of things that people are most interested in predictions about are those things are most likely to go either way."

So I don't understand what the actual "insight" here is.

The closer something is to being a coin-flip toss-up, the more exciting it is to speculate about what will happen.

True, but the possibility that the probability of longshot events occurring could be greater than is commonly believed also generates excitement. Eg Could Rocky Balboa really have a chance against Apollo Creed?

Aaron in Israel said...

Your dullest insight seems wrong. "Likely" and "probability" must be subjective here, if the statement means anything. So "likely to go either way" means according to somebody's (or many somebodies') subjective opinion. But social sciences get the most scorn when they fail to predict things that were not seen as likely to go either way: the fall of the Soviet Union, economic disasters, etc. Of course, I'm not talking about predictions like, "Something bad is going to happen sooner or later."

It's forgivable to miss close calls like presidential elections (which social science is pretty good at anyway). That's just a matter of accuracy. People wouldn't mind if social science were just inaccurate. Physics was a lot less accurate two centuries ago, but no less scientific. And as far as predicting that San Diego will be nicer than Tijuana, that's something that can be predicted without social science, so it seems unimportant, just as it's unimportant that physics predicts that an apple will fall down, not up.

Orthodox said...

The insider with the information was Justice Roberts because he switched his vote very late, and I don't think he or his staff play InTrade.

Anonymous said...

http://www.davekehr.com/?p=1382#comments

Anonymous said...

Well the point is that any reasonably informed observer can make shrewd guesses about the likelihood of certain trends in the world coming to pass - no different to betting on horses really.
The breakthroughs in the real, hard sciences are distinguished by the fact that they were never obvious to our ancient forbears, and the ramifications from those discoveries were truly earth shattering - the breakthroughs were the fruit of painstaking observation, invsetigation and theorizing - the idea was to get to the underlying 'truth' of natural phenomena.
One small example think of the development of current electricity in the 19th century. Oersted observed that there was a linkage between electricity and magnetism, a few decades later James Clerk Maxwell published his work on electro magnetic waves - hence we got the theoretical speed of light and radio and television as a by product.
All the economists have beem able to do is to stink up the world economy and turn millions of worthy US citizens into paupers.

Anonymous said...

My own theory of why academic and government appointed economists fail, fail and fail again is that most of what they believe in and ractice is just simply falsehood and wrong, except they can't see it, being so blinded by their psuedo academicisim and unwarranted self belief.
Macro economics is, in general, a matter of feel, intuition, feeling and instinct - nothing to do with theory but being shrewd and informed enough to spot trends, expect trouble coming and knowing your own strengths and weaknesses, it's not a science at all in fact, but something a shrewd and cunning horse dealer or horse gambler would excell at.
Unforyunately, the pompous know-it-alls would scoff at such a notion (ie the best apprenticeship for the US chief economist position is 30 years as a used car dealer), and thus we get the dummies in charge who have stunk up and will continue to stink up the US economy.

Anonymous said...

Intrade market on Obamacare was probably skewed by

1. a lot of rubes who don't understand how the supreme court works

2. a lot of rubes who really really wanted it to be overturned and made a weird kind of "i got money on this now, its gotta happen" voodoo bet

Steve Sailer said...

I had a friend in high school whose father was a prominent football bookie. About once a decade he'd go to prison for awhile, but was a popular member of the local community. My friend gave me a subscription to his dad's "Gold Sheet" one year and I tried picking out a few college football games a week where I disagreed with his point spread. My recollection is I beat his picks by a good margin over the course of the year, but I never tried it again and never for real money. It was kind of fun to be a dispassionate fan but it was a lot of work and dispassion isn't as much fun as passion.

kaganovitch said...

Steve, Do you really mean Coase's theorem? Or did you mean "rational expectations" associated most prominently with Muth and Lucas?

John A. MacDonald said...

Slightly OT, it would make more sense from a Canadian or European viewpoint to just hold a referendum on healthcare rather than leave it to nine individuals parsing the US Constitution in a Talmudic way.

(I am Canadian).


I don't see how this is a Canadian or European viewpoint at all. Canadians routinely sneer at Americans for all their referenda questions (how gauche to let the rubes decide! next thing you know, they'll be voting on who should be dog catcher).

As for Europe, perhaps I'm mistaken, but the EU has a pretty poor record of success when the transfer of national sovereignty has ever been put to the electorate in a referendum, and these days, they regularly bully anybody (say, Ireland or Greece) who has the effrontery to let the views of the electorate stand in the way of the "European project" being directed by the unelected geniuses in Brussels.

Anonymous said...

"BTW I am curious Steve. Did you ever read any of the "Masked Political Consultant's" three posts on the late Joe Bageant's website? Opinions differ, but he pretty much nailed what I believe about current American Politics."

Sunbeam, based on your mentioning them I found them still posted on Mr. Bageant's site.

Very worth reading without a doubt.

Thanks for tipping me and other readers off.

Tony said...

It makes me happy that InTrade could be wrong. They're predicting an obama victory.

beowulf said...

"Keynes himself made a sharp distinction between calculable risk and events which are inherently unpredictable. For instance, life expectancy is a calculable entity, while a terror attack or the performance of the stock market is not.
For Keynes, this absolute uncertainty was perhaps the most important element in economics. It replaced the completely rational actor with an actor who was often rational if presented with computable odds, but also often had to rely upon others and upon convention for guidance in how to act."
http://www.perspy.com/?p=401

from Plame to Flame said...

"a record of discretion exceeding even that of some parts of the national security apparatus"--sounds impressive!

TGGP said...

Coase wrote about the effect of a legal decision assigning property rights (his background was in law, not economics), so I don't think Steve was confused. Rational expectations more often features agents who stymie the efforts of policy-makers by acting in response to what they expect policy to be.

kaganovitch said...

TGGP I think I see your point, but Steve's restatement of Coase "if people can predict how the law will be applied, they will adjust their affairs accordingly" seems to have little to do with Coase's insight that the assignment of property rights is irrelevant to the final outcome absent transaction costs- or am I missing something here?