Russ Roberts

Hanson on Health

EconTalk Episode with Robin Hanson
Hosted by Russ Roberts
PRINT
Vernon Smith on Markets and Ex... Shlaes on the Great Depression...

Robin Hanson, of George Mason University, argues that health care is different, but not in the usual ways people claim. He describes a set of paradoxical empirical findings in the study of health care and tries to explain these paradoxes in a unified way. One of his arguments is that the human brain evolved in ways that make it hard for us to be rational about health care. He also discusses using prediction markets as a way of designing health care policy.

Size: 33.1 MB
Right-click or Option-click, and select "Save Link/Target As MP3.

Readings and Links related to this podcast

Podcast Readings
HIDE READINGS
About this week's guest: About ideas and people mentioned in this podcast:

Highlights

Time
Podcast Highlights
HIDE HIGHLIGHTS
0:38Intro. Is health care different? Does government have to regulate it because we "need" it? Bigger choices have bigger consequences, but that doesn't mean by itself that it should be regulated or subsidized. It's different because people feel it is, but what actually is different is not obvious. Historically, medical spending was 1-2% of the economy at most. Recently up to 16%. Historically had plenty of doctors, but still people didn't spend much. When you have an accident or condition, it's argued that that's when you'd spend the money--even everything you've got. But even then people don't spend everything, don't get every test or treatment. People do value health insurance but at the moment of choice, like in the emergency room, they often choose less than the largest choice possible. Opposite of usual intuition. Even in a situation of crisis, the fact that you'd be willing to pay a lot for it doesn't mean that therefore markets don't work. People really like food and would pay a lot for it, but they don't have to because competitive supply of food keeps the price down. Same could be true of medical care if competitive supply of medical care were allowed. Health care is also claimed to be different because it is hard to evaluate the quality of health care. Because you don't know how your own body works as well as someone else (a doctor) might, it's hard to evaluate if a doctor or treatment is good. But you can get on the Internet; and just because you don't know as much as an expert doesn't mean you can't be trusted to consume it. People vastly overestimate how much they know about most things they buy. We make do. If you are ignorant you do have to delegate to someone, but why government rather than private?
6:44RAND Health Insurance experiment. In late 1970s, U.S. Government created a randomized experiment. Previously people had looked at correlations with health--exercise, sleep, wealth, diet, pollution--but you don't see a correlation with medicine! Puzzle. True in both developed and undeveloped countries. Can be criticized for not controlling for enough things. RAND experiment was to test overall. 7000 random people in many cities, let half the people get free medicine and half had to pay full price. Those who had to pay didn't take as much--so far as expected, demand curves slope down. Health care demand is relatively inelastic--that is, relatively unresponsive to price. Dropping full price to 0 resulted in a 40% increase in quantity. Did the people who got more medicine get any healthier? You'd think it would make a difference. But maybe people who got it free mostly just used that free aspect for trivial things like the sniffles or a twisted ankle. Turned out that comparing the extra care people got when it was free to the care bought at full price, doctor evaluations were the same. Looked afterwards at the case and evaluated appropriateness (failed 1/4 of the time), severity of diagnosis--same percentages in both halves of the population. That is, both groups went in for medical care for the same mix of things, same severity levels, and got the same care. Bottom line: quality of health care wasn't any different by any standard. One qualifier was eye glasses. People in the "free" group got free dental and eye care as well. Those who got free glasses could see better--no surprise. If you count that as medicine, then more medicine is correlated with improved health. But if you take glasses out of the sample, there was no effect. How did they measure effect? Previous studies looked at death rates. RAND study looked at 25 measures such as blood pressure, other overall measures. Treated children and the elderly separately. No tangible measure of health showed that amount of medical care had any impact. Inconsistent with previous correlation experiments that had apparently shown that countries or states that spent more on health care had healthier citizens.
14:56Could argue that the RAND experiment still didn't do the experiment right--maybe have to do it with more people or more recently. Or, could say maybe it's true: Medical expenditure may not have the bang for the buck that we might think it would. Distinguish between health, health care, and health insurance. We want health, not health care, and not health insurance. Not intuitive. How could all the other studies be wrong? Studies are constantly published showing that specific medicines or treatments are effective. And, why are people buying all that health care and health insurance if it's not very useful? This has been going on for centuries. Why were people going to the doctor historically? Just because we can look back on it now and say the procedures were not good for your health--saws, leeches, etc.--doesn't mean people don't do it. Maybe they have other reasons. Tolstoy, in War and Peace, has a description of a lovelorn patient whose family spends a lot on her health trying to show that they cared for her. But maybe it's not either. Maybe medicine is strange. As economists, let's take all the strange things and look at them at once instead of one thing at a time. What's strange? 1. Maybe health care doesn't affect your health. 2. But then why do people keep going to the doctor? 3. When you offer people information about health care or medicine, they are remarkably uninterested. Late 1980s experiment, published mortality rates at different hospitals for different surgeons. A 4% difference in mortality should be worth a lot to you! You'd think people would want to look at this information and be willing to pay for it to select a hospital. But people were only willing to pay $50 for it. Most hadn't read the study. Influenced the decisions of hospital choices for only a few people. But maybe people are sophisticated and knew that people know that different mortality rates are not just due to different hospitals. 4. Puzzling patterns of people liking community medicine. They like to get medicine from family members. 5., 6., etc. People with children get more medicine, women get more medicine. People like to regulate things that might have health risks. We want to regulate medicine but we don't want to regulate other big things like job choice, city choice.
25:33Don't try to explain each of these things individually. Instead try to explain them all with one explanation. Here's one possibility: Maybe we are just terrified of thinking about death. Would imply that we just don't want to think about not trusting our doctor. Want to push decision off on someone else. By buying health insurance, the vision is that someone else is responsible. This explains some of the puzzles, but not all. Explains apparent excess medical amount. If you were terrified of not getting enough food, you'd go only to all-you-can-eat restaurants. Wasteful-appearing care is not wasteful because it consoles you. Supporters of socialized medicine can reasonably argue that the government is supplying that assurance. Don't have to feel guilty about not getting enough. But you do need the part where you are sure your government is picking the right amount. Maybe you should switch countries. Canada, rationing. Explains not wanting to think about quality, explains wanting to regulate to make sure things are safe so you don't have to think about it or blame yourself. Close to the unsavory argument that people are stupid, though.
30:34Second unified answer: Humans evolved various ways to signal their loyalty to each other. Who's on my side? Coalition formation is important for human evolution. Ancestors did things that were small cost to signal loyalty—small talk, grooming hair. Also did big things—go to war for you, throw a feast. If you were about to betray someone the time to do it was just before one of these things. If someone spends a lot of cost on you just when you are weak is an especially potent signal of loyalty, since it's just at the time you'd most appropriately dump them. Medicine fits this category. How valuable it is isn't as important as how much you need to spend, in order to demonstrate loyalty. Valentine's Day boxes. Ancestors had signals like showing how witty they are, displaying sport prowess, showing wealth by display--health, wealth, and intelligence. Health care is the sort of thing our ancestors did and we have a deeply embedded sympathy for. Quality doesn't matter—amount of expenditure is what matters. Public, not private signal, matters. A hospital did something wrong, and one article in one paper on that had greater effect than the study reporting the hospital's standings. Can also explain aspects of the status of health correlation and why we want to regulate health a lot.
38:01Overcoming Bias, blog, looks at ways we are hard-wired or struggle to make rational decisions. Long literature on biases in our reasoning, but little on how to overcome them. "I'm shocked! I'm terrified of saying my brain can't be trusted! It's going to tell me all sorts of lies! How do I make this brain trustworthy?" How can we form social institutions that overcome bias? Want to overcome inefficiency of health care puzzle. Alternative explanation: Common view is that you get what you pay for. View is that expensive stuff is better than cheaper stuff. Often true. A Lexis is more comfortable, has more features than a Hyundai. Expensive fountain pen writes better than a $20 fountain pen. The general view is a good enough guide. But it runs into problems in areas like health care. Take RAND study as true, that people who consumed more health care didn't get better health care. Maybe it's just that health care is not always consistent with the idea that you get what you pay for because it is not a competitive market. In situations where you are spending other people's money, you shouldn't expect to get what you pay for. Education, Rick Hanushek, found that schools that spend more per student don't get better-educated students. Either the money is being spent on things not related to schooling, or it's wasted. Spending other people's money distorts incentives. How is health care now any different? The level. Why do people tolerate it? People are eager to have a low co-pay. Bryan Caplan would argue people just don't realize it. But why then do people not choose all-you-want plan for gasoline? Medicine and education are distinct: people tolerate crappy institutions that produce crappy incentives. Not just that they don't understand it because they seem to understand it better in other walks of life. We spend $2 trillion on health care per year in the U.S., a lot of money. What percent is pharmaceutical? Less than 15%, say 10%. Suppose we made all pharmaceutical payments 0, simultaneously compensating the pharmaceutical companies as a pure subsidy so that their incentives for future innovation are not affected. That would dramatically lower the cost! You'd think people would want this, but people are against it because it's the only area left where choice is still possible.
50:03Medicine and education have been weird for longer than government has been big. Can't just blame government for the weirdness. Suppose this hypothesis is true. Yet, Russ and Robin, with same genetic heritage are aware of it. What might we encourage our neighbors to do? We need to socialize medicine, but everyone will be really happy because of the signal involved. But Canadians seem to be becoming somewhat alarmed about the efficacy of their system. Maybe this hypothesis is not a strong basis for making health care recommendations, don't understand enough about it yet. Not that confident in hypothesis, how to apply it. If signaling theory is correct, better to tax a signal than to subsidize it. Hard to overcome a signaling problem. Eager to create social institutions that don't lie as much. Another explanation for RAND experiment: Are you telling me I should never go to the doctor? No. RAND experiment says: Ask yourself if you were going to go to the doctor anyway, paying out of your own pocket. If answer is no, don't bother to go. Experiment doesn't tell you about the stuff you were going to be willing to pay for. Marginal benefit. Consistent with a world where there are many things in medicine that are valuable, but the ones that you would buy if it were really cheap, those are probably not good uses of your time either because they're ineffective or because the bad things that could happen cancel out the good things. Yes, if successful, surgery can help you. But there is a risk that surgery won't help you because sometimes bad things happen. Lots of possibilities for things going wrong.
57:31Prediction markets. Talked about this recently in podcast with Cass Sunstein. If you call someone on the phone asking for opinion, speaker doesn't have an incentive to say he doesn't know anything, and no incentive to say the truth. Betting market has two wonderful features: If you don't know, you have incentive to shut up; and it gives people the incentive to pause and really think, because they might lose money. Betting markets for who will be next president, and even on complex conditional estimates, which can be thought of as advice. If a particular health care reform passes, what is life expectancy ten years later, and if it doesn't pass, what is life expectancy ten years later? Russ: Although I'm a skeptic on RAND experiment, when you put it as a betting market I'd guess there won't be any difference in life expectancy. Also worried about intervening events. Phoenix Suns playing basketball; those who betted probably didn't anticipate that two players would be out because of a brawl. It's not currently legal to have betting markets on policy issues. Legal in Las Vegas to have sports betting but not betting on Presidential elections. Don't want scandal. Even if it were legal, do people want to hear the answer? Robin's advocacy of prediction markets, brouhaha. Research project to look at betting markets to advise on military policy. Were going to set it up, but some Senators found some of the bets in a corner of the website and argued that it would allow making money off of terrorism. Once the accusation was made, Secretary of Defense declared the project over, DARPA person out of town, John Poindexter resignation. All happened in one day. Tetlock, Expert Political Judgment. Not supposed to make certain trade-offs, not even supposed to think about them. Are there other areas where these markets might be used? Private corporations. Less interested than you might think about getting accurate forecasts. Management may be more about theater. May even like the fact that the software managers' forecasts to be wrong because when they are behind schedule they'll get more done. Competition tends to produce firms that can use bets well. Entrepreneurs' overconfidence in ability to produce. Adam Smith quote.

COMMENTS (30 to date)
Bruce G Charlton writes:

This is a terrific interview - in some ways my favourite EconTalk so far, because of the 'no holds barred' interaction between two colleagues. Its a pleasure to listen to a thorough conversation between such smart and thoughtful people.

RR has developed into an ideal interviewer, such that he seems able to bring-out and clarify things which the interviewee didn't quite know they knew.

I certainly found the interview provided a very valuable background for understanding and integrating Robin Hanson's ideas - many of which I had already encountered, but I lacked the perspective to put them together.

Robin Hanson writes:

Thanks Bruce - out of the 170+ interviews I have done so far, I think this one is my favorite. At least it is the most fun one I can remember. :)

muirgeo writes:

Great interview. As a physician I was glad to see late interview clarification of no marginal benefit between the two groups in the RAND study as opposed to no benefit at all for health care.

The experiment didn't have a control group that received no health care. Unfortunately our country does have a group 45 million plus that have no insurance and many more that are under insured. Medical illness is one of the top causes for bankruptcy as I understand.

First to give my 2 cents on why we think of health care as we do. It might be somewhat hardwired. I suspect through the ages all members of the tribe received care from their Shaman. The tribe depended on its members being healthy so free care has a mutual benefit. That and they didn't have any expensive MRI scanners.

Then consider the free market alternative. Imagine you are one of 4 doctor/owners of a small rural hospital. You're on call and drug addicted ,homeless women with no insurance comes to your emergency room about to deliver a 28 week premature baby. What do you do?

As far as betting on health care policy. I would put my money on a single payer plan to both improve quality of care and markedly decrease cost. This would be based on all the other developed countries that do have such plans yet have better outcomes and sometimes at half the cost. For those who are insured in our country it will be hard to improve on the quality they get but ideally they would save some money. But the large pool of under-insured and un-insured suddenly having access will bring up our mortality and morbidity stats.

One last thing in general I understand that 15 of ever 100 health care dollars goes to medicine. 10 go to docs, another 15 to nurses and ancillary health care workers. Some 15 to 25 goes to paper pushers and stock dividends.

http://content.nejm.org/cgi/content/abstract/349/8/768


http://www.pnhp.org/images/cartoons/hightower_cartoon.gif

mz writes:

Prof Hanson,

Does the RAND study control for possible changes to behavior of the two groups? My thinking is that if a group people has free access to medicine that members of this group might engage in more risky behaviors than they otherwise might have. The net result of the increased unhealthy behavior may wash out the effects increased medical care. This is similar to observation that seat belt laws have minimal effect on safety because of an equally large negative effect on safety due to increased risky driving.

Also this is one my favorite podcasts. Great discussion.

Robin Hanson writes:

Mz, many kinds of behavior were tracked, and no changes were seen.

Muirgeo, the group in the RAND experiment that faced full prices at the moment of service in essence had no insurance, yet was no less healthy; being without insurance is not at all the same as being without medicine. In modern hunter-gatherer tribes, not all members receive all possible help from shamans. And I would love for you to be able to put your money where your mouth is on single payer effectiveness.

Brad Hutchings writes:

Well, Robin's quip about "maybe I'll listen to be polite" had me laughing out loud at the gym tonight, then having to explain to my buddy that it was a podcast on health care policy (among other topics). That was just like dinner with my crowd, except we don't often discuss prediction markets.

My bet on health care policy is that it will get nasty. Muirgeo launches the nice paternalistic "underinsured" salvo. As a single 36-year old guy, I keep a high deductible policy. It's $1800 per year out the window. Frankly, I don't blame any 20-something who has a history of being healthy, doesn't have a family, and opts out of the insurance system. It's a bad bet for many young people. Of course, what Massachusetts figured out in their universal coverage system was that you couldn't do it unless you brought all of these healthy people in and basically overinsured them. If our Governator gets his way, we'll have the same thing in California. If doctors would just offer free WiFi, I might go get my money's worth.

Pascal Bernhard writes:

Sir,

in fact in Germany there is a sort of betting market on politics although not with real money. Since the election of 1998, the newspaper "Die Zeit" has instituted an online "election stock market", in which everyone signing up gets allotted a certain amount of virtual money with which he or she can buy stocks of specific political parties in order to build up a political portfolio. The price of these "stocks" are determined by the factors of supply and demand (depending on how the campaign is going). Here in Germany we don't have a first-past-the-pole election system, but a proprortional system, with the effect that there are several parties, since all votes are somehow represented in the German "Congress".
The people betting on parties do not buy stocks according to their personal preferences but according to their judgment about the voting intentions of all voters. The incentive is that the people whose portfolio reflects best the final election outcome win some prize.
This stockmarket has proved to be far more accurate than "casual" polls (in the 2005 election the error margin was only 0.2% on average. Besides as a result of the proportional vote, there are (sadly) far-right extremist parties and people do not want to admit to pollsters that they have voted for them, so in the official result these parties' score is in the end always higher than predicted.

paul writes:

the money management industry suffers from the same "bias."

Robin Hanson writes:

Pascal, there are and have been many betting markets on who will win various elections. There have not been markets on the consequences of who we elect.

Paul, yes, and the medicine and investment advisor puzzles may have some similar explanations.

William Walker writes:

A brief comment on Prof. Hanson's point about hard-wired social signaling, but with reference to an earlier podcast on price gouging.

I have tried to explain the economic problems with price gouging to my state government colleagues. They have a hard time following me, not because they are stupid, nor even because I can't explain, but because they are repulsed by gouging itself.

I suspect that a similar social signaling issue is at work there as in health care. We see gouging as taking advantage of someone in distress, and therefore of breaking the tribal bond that is essential to our nature.

Note that our usual explanation as economists for the public's ignorance of economic principles is either stupidity or poor education. I would enjoy hearing a podcast on social signaling interpretations of economic "ignorance."

As always, these podcasts are supremely informative, entertaining, and thought provoking. Thanks for brightening my week.

Russ Roberts writes:

William Walker,

You make an excellent point. I think of it this way. When there aren't enough cookies to go around, what parent sells the cookies to the kid willing to pay the most? In a small town where everyone knows everyone else, using prices to allocate goods in an emergency would be frowned on.

In both of those situations, the informational role of prices is relatively or totally unnecessary. Surely, a parent wouldn't want use a high price for a cookie as a way of figuring out which kid wouldn't mind a piece of fruit instead of the cookie.

But in a city the size of Washington, DC, failing to use prices is going to cause suffering. Hayek writes about the tension between these different situations in The Fatal Conceit:

"Part of our present difficulty is that we must constantly adjust our lives, our thoughts and our emotions, in order to live simultaneously within the different kinds of orders according to different rules. If we were to apply the unmodified, uncurbed, rules of the micro-cosmos (i.e. of the small band or troop, or of, say, our families) to the macro-cosmos (our wider civilisation), as our instincts and sentimental yearnings often make us wish to do, we would destroy it. Yet if we were always to apply the rules of the extended order to our more intimate groupings, we would crush them. So we must learn to live in two sorts of world at once."

But we're not good at living in two worlds at once. So retailers who raise prices in large cities are treated with the same disgust as a parent deciding which kid's little league game to attend based on an auction.

I write about this at length in my next book, The Price of Everything. I use the bank run scence from "It's a Wonderful Life" as an example of the small town example. The guy who wants all of his money is treated as a thug.

If I remember correctly, the issue does come up in the podcast on gouging with Mike Munger. It might be worth another podcast down the road to focus on this tension explicitly.

Steve writes:

Russ (and Robin),

I agree that this is one of my favorite podcasts so far. Great give and take on a very interesting topic.

Steve

Chris Fulmer writes:

I terrificly enjoyed this podcast and wanted to offer an alternative explanation to the conclusions of the RAND study: substitutes. There are many non-healthcare things people can do to increase their general health. For example, I could manage my cholesterol through my diet by eating the right foods or through the healthcare system by seeing my doctor and getting cholesterol medicine. If the healthcare is free, I might just eat the foods I enjoy and get the drugs to treat it. But, if I have to pay for that healthcare, well, eating fish instead of steak might be a good tradeoff.

Also, recognize that just because it's medicine doesn't mean it's "healthcare." Maybe by measuring "health," RAND was measuring the wrong thing. I suggest that a person is equally healthy with or with out the viagara prescription, tummy tuck, breast implants or hair transplant. That person may not be as happy, though.

Robin Hanson writes:

Chris, I don't think there was any evidence that diets were different in the RAND experiment for people who had free medicine. They used a wide range of health measures, so it is hard to believe they missed the "real" health measures.

Russ Roberts writes:

Robin,

I think Chris's point speaks more to the implications of the Rand experiment rather than to the experiment itself. When you make something free, people will consumer a lot more of it for all kinds of reasons. And when that free stuff is subsidized, people will step forward to create all kinds of variations that people might want. So the Rand experiment didn't have Viagra or Lipitor or Botox or other innovations because they weren't available yet. But when health care is widely subsidized, those options will get created and people will consume them. They'll also choose, as Chris points out, to eat tastier but more cholesterol-heavy food if they know there's an alternative subsidized way (Lipitor) to keep their cholesterol down.

So while the Rand experiment may have looked at a wide variety of health care measures, it probably did not catch every margin of choice people can respond on when they know health care is free. But even if the experiment was perfectly designed and caught everything, it's good to remember the creativity people use (both consumers and suppliers) to respond to a world of the free lunch.

Robin Hanson writes:

Russ, yes, an experiment of the effect of price on a few thousand people does not capture how price changes in an entire society change the products available in that society.

Jake Barna writes:

This was a very entertaining Podcast.

Can you give us some further reading on Prediction Markets and how managers should use them?

Thanks.

enronal writes:

Just adding my three cheers for this terrific interview. Good economic reasoning in the give and take, and lots of delightful and counterintuitive nuggets of fact, starting with the Rand study. Robin Hanson is on my radar as of now. Next on my list is to find his 14 wild ideas...

Andy writes:

I agree that prediction markets are conceptually a great way to get people's honest opinions. Talk is cheap, but putting your money where your mouth is can be a different story. I think, however, there is a dangerous side effect. A prediction market creates cheerleaders and detractors based on where people have placed their bets. There is now a financial incentive to influence the outcome of the event you are are attempting to predict. Much like in professional sports, you need to regulate the system such that those with influence are not allowed to participate in the betting, which can be difficult in practice. Such regulation might actually remove the potentially most interesting players from the game.

Mark G writes:

Wondering if you could comment on a slightly different take on the Rand study by Malcolm Gladwell not that long ago in New Yorker. I haven't parsed your interpretation versus Gladwells too closely, but I don't recall you differentiating between preventive and acute care in the podcast as Gladwell does here. (Excerpt below)

*************

For that matter, when you have to pay for your own health care, does your consumption really become more efficient? In the late nineteen-seventies, the rand Corporation did an extensive study on the question, randomly assigning families to health plans with co-payment levels at zero per cent, twenty-five per cent, fifty per cent, or ninety-five per cent, up to six thousand dollars. As you might expect, the more that people were asked to chip in for their health care the less care they used. The problem was that they cut back equally on both frivolous care and useful care. Poor people in the high-deductible group with hypertension, for instance, didn't do nearly as good a job of controlling their blood pressure as those in other groups, resulting in a ten-per-cent increase in the likelihood of death. As a recent Commonwealth Fund study concluded, cost sharing is "a blunt instrument." Of course it is: how should the average consumer be expected to know beforehand what care is frivolous and what care is useful? I just went to the dermatologist to get moles checked for skin cancer. If I had had to pay a hundred per cent, or even fifty per cent, of the cost of the visit, I might not have gone. Would that have been a wise decision? I have no idea. But if one of those moles really is cancerous, that simple, inexpensive visit could save the health-care system tens of thousands of dollars (not to mention saving me a great deal of heartbreak). The focus on moral hazard suggests that the changes we make in our behavior when we have insurance are nearly always wasteful. Yet, when it comes to health care, many of the things we do only because we have insurance—like getting our moles checked, or getting our teeth cleaned regularly, or getting a mammogram or engaging in other routine preventive care—are anything but wasteful and inefficient. In fact, they are behaviors that could end up saving the health-care system a good deal of money.

[The complete article from The New Yorker is available online: "The Moral Hazard Myth".--Econlib Ed.]

Robin Hanson writes:

Mark, Gladwell is the victim of data-mining. The overall results were no net effect on health. If you go look at thirty specific results, you should expect to see a result significant at the 3% level, even if there are no real effects. To claim a net effect from that sort of result is to misunderstand statistics at a fundamental level.

Salaam Yitbarek writes:

I hope you don't mind that I comment on your Michael Lewis interview here - I've only just listened to it, I feel compelled to comment, but comments are disabled there.

[Hi, Salaam. I've moved your comment to the Michael Lewis thread and reopened comments there--at least till they again start to attract spam. Thanks for your interest!--Econlib Ed.]

Luke O writes:

Interesting interview. Just wanted to comment on the prediction markets section of the show. It's surprising to me that you have had Nassim Taleb on your show, and I am assuming you have read his book, and yet you continue to tout the greatness of prediction markets in all areas.

Prediction markets can be very good but in my opinion they must be used very carefully. You have to really think about the question you are asking people to predict before you start giving them any weight. You run the risk of having too much faith in "fake" knowledge.

If the question exists in a normally distributed world then your market will be able to yield you a good result. But if you ask a question on a prediction market like what when will the next terrorist attack be? This exists in the "extremistan" world, as Taleb calls it. There is too much chaos and unpredictability to have any faith in the market. The question is of course whether the only person with true knowledge of the event, the terrorist, would be dumb enough to bet on himself.

Russ Roberts writes:

Luke O,


Good point that Taleb is relevant for the efficacy of prediction markets. It's an empirical question. For some questions, they work fairly well.

By the way, I doubt the terrorist has true knowledge. Life is uncertain even for terrorists.

But I don't mean to "tout" them, just explore them. I'm interested in them. I don't know as much about them as I'd like.

Robin Hanson writes:

Luke, some problems are surely easier than other problems, but the question is the relative performance of different institutions on the same problem. Yes it is hard to predict rare events, but that doesn't mean prediction markets couldn't do as well as the alternatives.

Luke O writes:

Russ and Robin,

I'm also interested in prediction markets and I'm reading through Cass Sunstein's book now.

The mathematics of many opinions being better than any individual, even "experts", and imposing consequences on those that are wrong makes sense for questions that adhere to a normal distribution. It is often the presumption of the models that a question like how many beans are in a jar?, is in anyway the same thing as asking, What will happen to the economy of Saudi Arabia be like 5 years from now? The former question adheres to a mild randomness, while the latter is subject to extreme randomness in which we do not know the all the variables that will influence the question and some of them are likely to have an extreme impact. The error in the idea of using prediction markets on the wrong question is the false belief that we can know all the variables, and that the impact of any individual being wrong or right is equal in the in a non bell curve question.

Robin, you say that the important thing is the relative performance of different institutions on the same problem. I would agree. My contention is that there are some questions in which both a prediction market and alternative method will both fail to predict equally badly, because we can't possibly predict what's going to happen. While prediction markets will be just as good or better at answering a question like what is the weather going to be tomorrow as group of experts, if you asked what will the weather be 100 years from today, neither group has much predictive knowledge.

Remember also that many things in history can only be played out once, and some things occur for simply random reasons. We have to have an idea of error rate, before any prediction can be taken seriously. I'm reminded of a gambling example in which everyone in the world goes out and plays Russian roulette and agrees to play for 60 days. So the first day everyone goes out spins their six shooter and fires, 1/6 dies. We expect that everyday 1/6 of the population will be eliminated each day. After about 30 days of this there will be around 1 million people left people left and after 60 days there will only be about 1000 left. Why did these 1000 survive? Simple luck, but many of them would think they had some method for surviving that long.

As a professional poker player, I am always thinking about whether I am a Russian roulette survivor or if I have actually made good decisions. It is really difficult, and most of them take place in a normally distributed world! As far as your question of life expectancies ten years from now based on whether socialized medicine is introduced, it would be difficult to know, even if the life expectancy goes up or down, whether it was in fact caused by the change in medicine. You have to be aware of how often and what types of questions the markets are bad at predicting, and watch out for times in which you are only randomly right, not predictively right.

Thank you both for your time

Richard Sprague writes:

Has anybody tried to confirm the RAND study by looking at the health of employees at large companies? My employer, Microsoft, provides essentially free healthcare -- no copay on drugs, hospital visits, etc. -- to 70,000 employees. I wonder if the health of those employees (adjusted for age, etc.) is any higher than at companies without such a generous plan?

Tom writes:

I agree with this completely, thanks for the post.

Floccina writes:

Your articles on healthcare remind me of the Softram story. In short, the Softram product claimed to increase effective memory available in windows be compressing data in RAM. The funny thing is that a competitor came along "Quarterdeck" and made a product that actually worked. Healthcare started with useless leeches but that showed that there were big bucks to be made in medicine.

So perhaps since the time of the rand study doctors have actually made more healthcare that really works.

Comments for this podcast episode have been closed
Return to top