
Robin Hanson on Health 5/28/2007
What this covers
This EconTalk conversation pairs Russ Roberts with Robin Hanson of George Mason University. Hanson collects a set of empirical oddities about medicine and tries to explain them with one account instead of treating each separately. The evidence comes first. Before the RAND experiment, cross-sectional studies linked health to exercise, sleep, wealth, diet and air pollution, but not to medical care. The RAND experiment of the late 1970s then randomly assigned thousands of people to free care or cost-sharing. The free-care group used roughly a third more care, rated no less appropriate by physician reviewers, and showed no measurable health gain apart from better vision from free eyeglasses. Hanson reads this as a claim about marginal value: the treatments people buy at zero price are probably not good uses of their time.
The range then widens. Roberts raises the puzzle of why people kept visiting doctors for centuries when the visits were probably harmful. He also cites a late-1980s Pennsylvania study in which very few surgery patients looked at hospital mortality data. Hanson offers several explanations. Trial results may not carry over to real patients with other conditions, drug interactions and infections. Patients may be sophisticated rather than irrational about mortality statistics. And care may serve as a costly signal of loyalty and caring, rooted in how human brains evolved for tribal coalitions.
Some of the more contested turns follow. Hanson compares health to education, where dozens of studies find that higher spending does not improve outcomes. He argues that regulation worsens these sectors but does not fundamentally cause their problems. He also describes the DARPA project's cancellation as an instance of a taboo against weighing trade-offs involving human life. On policy he is cautious. He declines to make detailed recommendations, prefers decentralized decisions that use dispersed patient information, and champions prediction markets because they reward honesty and make people stop to ask what they actually believe.
Hanson argues that medicine's weak measurable health payoff, as shown by the RAND experiment, is better explained by medicine's role as a costly signal of caring and loyalty than by its health production, and that prediction markets could expose the institutional distortions that hide this.
- The RAND experiment found 30-40% more medical use with free care and no measurable health gain beyond eyeglasses
- Costly signaling of loyalty and caring explains spending better than health production, and people respond to public signals over private quality data
- Prediction markets give people incentives to speak honestly and to reconsider their views, unlike surveys or reporter calls
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
The DARPA project's termination was driven by an emotional taboo against even considering trade-offs involving human life, and the outcry was so fast because thinking about such trade-offs is viewed as wrong.
“you're not supposed to make sacred trade offs and you're not even supposed to think about them”
Treatments that work in randomized trials may fail in practice because trial patients are the best candidates, careful doctors run the trial, and real-world patients have other conditions, medications, and random accidents such as infections and wrong medicines.
“they picked the very best patients the patients most likely to benefit from something”
People treat some trade-offs as sacred: even considering cutting hospital costs at risk to patients is judged wrong, so a hospital administrator who thinks about it for a week is seen as worse than one who cuts immediately, which made the DARPA project politically untenable.
“you're not supposed to make sacred trade offs and you're not even supposed to think about them”
The RAND free-care group showed no measurable improvement in health: none of about 25 physical measures or the overall health measures changed, and no subgroup (rich/poor, initially sick/healthy) showed an effect, except that free eyeglasses improved vision.
“no tangible objective measure of health showed that that had any impact”
People are strikingly uninterested in information about hospital quality: in a late-1980s Pennsylvania study, only about 8% of surgery patients said they had looked at mortality data, and only about 5 of 500 said it influenced their choice, even though a 4-point mortality difference is worth tens of thousands of dollars.
“only 8% of the people said that if you ask them did you see the information”
Medical historians say that going to a doctor a century or more ago was probably harmful, since doctors did not wash their hands and used crude tools, yet people kept going, which poses a puzzle about why people buy ineffective medicine.
“medical historians usually actually tell you that if you know go back a century or more it was probably a bad idea to go to the doctor”
The free-medicine group consumed about 32 to 40 percent more medical care over three to five years, and the extra care was not more appropriate or more serious by doctor evaluation.
“it turns out when we looked at the extra care people got because of the free medicine and compared it to the common care both groups get and you had sort of doctor evaluations after the fact of the medical appropriateness it was just the same”
A DARPA-funded project to set up prediction markets on Middle East stability was cancelled within a day after senators claimed it was a market to bet on terrorism, even though bets were capped near $10 and no individual attacks were to be traded.
“once this accusation was made on a Monday morning by Tuesday morning the Secretary of Defense declared the project over”
The RAND Health Insurance Experiment in the late 1970s randomly assigned about 5,000-7,000 people in six US cities to free care or to cost-sharing; the free-care group used roughly 30-40% more medical care over three to five years but had no measurable improvement in health.
“we took five thousand seven thousand people in six US cities and we randomly assign some of them to get free medicine”
In a late-1980s Pennsylvania study, when hospital mortality rates were published, only about 8% of heart-surgery patients said they looked at the data, only about 5 of 500 said it influenced their choice, and most would not pay $50 for it, even though a 1% versus 5% mortality difference is worth tens of thousands of dollars by ordinary risk-taking.
“only 8% of the people said that if you ask them did you see the information a lot of fair number of people said if you ask them for the actual number they couldn't remember what it was”
The get-what-you-pay-for principle works in ordinary markets, where competition drives out poor sellers, but fails in health care and education because third parties spend other people's money, so people do not get what they pay for.
“in other areas where people are explicitly spending other people's money rather than their own we really shouldn't expect that you get what you pay for”
Health and education differ from other sectors because people do not adopt institutions with strong monitoring and incentives there, even though they accept low-copay plans that encourage overuse; people reject all-you-can-eat grocery or gasoline plans because they understand the bad incentives.
“people tolerate crappy institutions that produce crappy incentives”
Medicine and education have been strange for longer than government has been big and remain strange under very different regulatory regimes, so regulation probably exacerbates but does not fundamentally cause their problems.
“medicine and education have been weird for longer than government has been big”
Human brains evolved large because of social complexity, so a central evolved function of many behaviors is to signal loyalty and ability to allies, since knowing who is on your side was critical in tribal coalitions.
“human brains became large it seems because there were social brains they had to figure out each other and one of the most important things they had to figure out was who's on my side”
Prediction markets are valuable because they give people who lack knowledge a strong incentive to stay quiet, and they make people pause to ask what they really think, since they lose money if wrong, unlike phone surveys or peer review.
“a betting market has two wonderful features one it gives people who don't know a strong incentive to shut up”
The RAND results do not show that medicine is useless; the extra care may be useful, but many interventions do no good because bad outcomes such as hospital infections, wrong medicines, and side effects offset the benefit, and randomized trial results often fail to generalize.
“because unfortunately sometimes bad things happen ... you should pick up an infection in the hospital that wasn't what you came in with”
A betting market has two key features: it gives people who don't know much a strong incentive to stay quiet rather than offer opinions, and it makes people pause to ask what they really think, because being wrong costs money.
“a betting market has two wonderful features one it gives people who don't know a strong incentive to shut up which other institutions don't do”
The RAND results are consistent with many medical treatments being effective, but the treatments people would buy at zero price are probably not good uses of their time either, because they are ineffective or their risks cancel their benefits.
“the ones that you would buy if it were really cheap those are probably not good uses of your time either”
Managers may prefer over-optimistic forecasts from software staff, because managers who promise more than they can deliver signal expertise and work harder when behind schedule.
“managers like to have software managers who are who have inaccurate forecasts of how long it will take them to do a project”
Employer-bundled health plans with low copays encourage overuse, and people do not understand that the low copay is paid for through premiums, so they accept bad incentives they would reject elsewhere.
“the low copay encourages overuse that they're really paying for it in the premium that's coming out of their cell”
People behave differently ex ante and at the moment of choice: they will forgo enormous salary to get generous health insurance, but in the emergency room they don't spend as much as the 'spend anything' view predicts, so they still spend large but finite amounts.
“people actually do seem to value health insurance enormously in terms of the benefit on a job they're willing to forego enormous amounts of salary to get generous health care health insurance packages but at the moment of choice they aren't actually willing to spend quite as much”
In the RAND experiment, the only clear health benefit of the extra care was improved vision from free eyeglasses; the other extra care, including free doctors and dentists, showed essentially no effect on health.
“people with free eyeglasses can see better that was one of the strong results of the experiment”
The RAND study found that extra care was similarly appropriate and similarly severe in both groups, with physician reviewers rating the extra care as no more or less appropriate than common care, so the extra spending was not on trivial conditions.
“it turns out when we looked at the extra care people got because of the free medicine and compared it to the common care both groups get it was just the same”
'You get what you pay for' fails in health care because most spending is by third parties, not patients, so bad sellers are not driven out; the same is true in education, where schools spending more do not produce better students.
“in other areas where people are explicitly spending other people's money rather than their own we really shouldn't expect that you get what you pay for”
Roughly a third of medical studies are inaccurate, and even well-run randomized trials select the best patients, so treatments that look helpful in trials can fail in everyday practice.
“one third of medical studies were found to be inaccurate”
Dozens of studies find that schools spending more per student do not produce better educated students, and the null results persist even though publication incentives favor finding effects, suggesting spending is wasted or diverted to non-educational uses like gyms and buildings.
“schools that spend more per student don't get better educated students”
Before the RAND experiment, decades of cross-sectional correlational studies across states, counties, and countries found correlations between health and exercise, sleep, wealth, diet, and air pollution, but not between health and medical care.
“you see correlations with exercise sleep wealth diet air pollution but you don't see a correlation with medicine”
Hanson believes the mechanism behind the RAND results is not wasteful spending aimed at reassurance alone, but that medical care often has zero net benefit because of accidents and interacting risks, so the effect on average health is small.
“because the bad things that can happen cancel out the good things”
The extra care from free medicine was judged by doctors to be as medically appropriate and as serious in diagnosis as the care the paying group received, so the extra care was not trivial.
“the extra care people got because of the free medicine and compared it to the common care both groups get and you had sort of doctor evaluations after the fact of the medical appropriateness it was just the same”
Patients may be sophisticated rather than irrational about hospital mortality: a 5% versus 1% gap could reflect patient mix rather than hospital quality, and people rely on trusted doctors' recommendations instead.
“people are sophisticated and they understood that a 5% mortality rate versus a 1 percent mortality rate could be due to lots factors other than the quality of the hospital”
Software managers often prefer overoptimistic project estimates, because promising more than can be delivered signals expertise and motivates people to work harder when behind schedule.
“managers of software managers like to have software managers who have inaccurate forecasts of how long it will take them to do a project”
Medicine's marginal care, the care people would buy if it were cheap, is probably not a good use of time either, because it is ineffective or its harms cancel its benefits.
“the ones that you would buy if it were really cheap those are probably not good uses of your time either”
Health care bureaucrats engage in extensive social engineering that people are unaware of, which destroys performance incentives because the designers assume they understand what treatments are for and what they do.
“social engineering that people are totally unaware of that goes on in healthcare that destroys the incentives for performance”
Much of medicine is ineffective, not because it is a sham, but because treatments that help in trials can fail in practice through infections, wrong medicines, interactions with other drugs, and trial patients being unrepresentative; these are the reasons that buying medicine may still be a poor use of money.
“because the bad things that can happen cancel out the good things right so yes if you have a certain health care issue surgery if it's successful we'll help you but there's a risk”
Publication incentives push researchers to report positive effects, yet many studies still find no effect, which implies education money is either spent on non-learning items like gyms and buildings or is simply wasted.
“one of two things either the money is being spent on things that are not related to schooling to true education like a really nice gym”
Before RAND, decades of correlational studies across states, counties, and countries found health correlated with exercise, sleep, wealth, diet, and air pollution, but not with medicine.
“when they did those correlation studies turns out you see lots of things a correlative health you know exercise sleep wealth diet air pollution but you don't see a correlation with medicine”
Free medical care increased total medical spending and utilization by about 30-40% over the three-to-five-year RAND period (about 32% more quantity at zero price), which is elastic but not extreme demand.
“the price to zero price gave you roughly a 40 percent increase in quantity so that's relatively elastic but still it's junk”
Medical care functions as a costly signal of loyalty: because our ancestors showed loyalty through expensive acts at moments of weakness, humans evolved to treat expensive care for the sick or injured as a sign of caring, even when it does not improve their health.
“taking care of somebody who is sicker injured was an especially potent signal of loyalty”
Given how little is known about health policy and how confused the speaker is, detailed policy recommendations are unwarranted, and a strong preference should be given to decentralized, bottom-up decisions where dispersed patient information is used.
“I think I'll just recommend taking your damn hands off it and stepping back”
About one third of medical studies are found to be inaccurate when retested years later, so the field's published findings are unreliable.
“one third of medical studies were found to be inaccurate”
If medical spending is a signal, people should be uninterested in private quality information but very responsive to public signals; in the Pennsylvania study, a single newspaper article about one bad hospital event had more effect on choices than all the statistics combined.
“one article in one paper about one event like that had more effect than all the rest of this those statistics combined”
Medical care evolved as a costly signal of loyalty: in human tribes, expensive help such as nursing or caring for someone who is sick or injured is a potent signal of commitment, because a betrayer would have to pay the cost at the moment they were most vulnerable.
“an even more potent a signal of loyalty is when somebody spends a lot of costs just at the moment you're really weak and it would be just the opportune time to dump you right and that's when you're sick or injured”
People respond to public signals of quality rather than private signals: a single newspaper story about a bad hospital event had more effect on choices in the Pennsylvania study than all the statistical data combined, because others could see it.
“one article in one paper about one event like that had more effect than all the rest of this those statistics combined in that study because it was in the newspaper where everybody could see it”
Pharmaceuticals account for roughly 10 percent of the roughly $2 trillion US health bill, so making drugs free would cut only about 200 billion dollars a year, yet that is the one area where people still face out-of-pocket choices.
“10 percent so 10 percent of two trillion is 200 billion so suppose we made all pharmaceutical payments zero”
The proximate explanation for weak incentives is that people do not adopt institutions with strong incentives and monitoring, but the deeper question is why people tolerate those institutions in health and education and not in areas like groceries or gasoline.
“the question is why do people tolerate those sorts of institutions and make those choices in this context versus other ones”
Costly signals of loyalty are the most potent: a sacrifice such as a revenge killing, building a hut, or hosting a feast is a strong reassurance, and taking care of someone when they are sick or injured is the most potent signal because it is expensive at a moment when betrayal would be easy.
“taking care of somebody who is sicker injured was an especially potent signal of loyalty”
Corporations such as Microsoft, Yahoo, Google, Pfizer, and Eli Lilly use internal prediction markets, but managers are less interested in accuracy than expected because much management is theater and corporate decisions are political rather than driven by new information.
“most a lot of management is more about theater and a lot of political a lot of corporate decisions”
Because medicine signals caring, its cost matters more than its quality: like a Valentine's box of chocolates, people choose the expensive option rather than the most useful one, so signals of caring prioritize expense over effectiveness.
“the more important thing is how much it costs you so when you're if you're buying a box of chocolates for your Valentine you don't ask how hungry is she”
Firms such as Microsoft, Yahoo, Google, Pfizer, and Eli Lilly use internal prediction markets, but managers are less interested in more accurate forecasts than expected, because much management is theater and corporate decisions are political rather than driven by last-minute information.
“most a lot of management is more about theater and a lot of political a lot of corporate decisions are in a sense political decisions”
Conditional betting markets can forecast policy by comparing outcomes under alternative choices, such as life expectancy ten years after a health reform passes versus after it fails; Hanson would bet the two prices are nearly equal, because he expects little difference from such policy.
“if their health care reform passes then what is the life expectancy in the United States ten years late and if it doesn't pass what's life expectancy”
What matters in a costly signal is how much it costs the signaler, not how valuable it is to the recipient, so people buy gaudy expensive boxes of chocolates rather than high-quality bargains because price shows caring.
“how much it costs you so when you're buying a box of chocolates for your Valentine you don't ask how hungry is she”
A DARPA-funded project to set up markets forecasting Middle East stability was shut down within a day after senators publicized a sample page about terrorism, because politicians treated any betting on terrorism as scandalous even though bets were limited to about $10 and did not target individual attacks.
“once this accusation was made on a Monday morning by Tuesday morning the Secretary of Defense declared the project over”
People are terrified of thinking about death, so they hand medical decisions to doctors, treating them like shamans, and buy health insurance and blank-check care to avoid having to make life-or-death choices themselves.
“we're just terrified of thinking about death”
People refuse all-you-can-eat grocery or gasoline plans because they understand the bad incentives, but accept similar bad incentives in medicine and education, suggesting these fields are distinct in ways that allow crappy institutions to persist.
“people tolerate crappy institutions that produce crappy incentives and it's not just that they don't understand”
Medicine and education have been strange for longer than government has been large, and they are weird across very different regulatory regimes, so bad regulation probably amplifies but does not fundamentally cause their problems.
“medicine and education have been weird for longer than government has been big okay and they're weirder all across the world under very different regimes”
People regulate health risks far more intensely than other risks: they accept a crappy-tasting apple with tiny health risks but want that risk regulated, and regulation seems to track the salience of health rather than the probability of harm.
“people don't mind you taking a chance with an apple that tastes crappy but the Apple might have a health risk however tiny well then they want that regulated”
Hanson argues that if the signaling theory is right, medicine is largely wasteful, and though taxing the signal would be better than subsidizing it, a global competition over the appearance of medical care makes a signaling problem hard to fix by tax alone.
“if my signaling theory is correct then overall it's a waste and you know often it's better to tax the signal than to subsidize it”
Hanson's core argument is that medicine is not the only thing that seems strange: if healthcare does not affect health, then many other puzzles follow, such as why people keep going to doctors and why they ignore quality information, so one should look for a single holistic explanation rather than ad hoc explanations for each puzzle.
“try to collect a whole bunch of strange things lay them out on the table and stare at them for a while and see if you can find some simple explanation”
Medicine's strange pattern (no health effect, yet persistent spending) should be explained by collecting many strange facts at once and seeking one simple unifying explanation, not by ad hoc explanations for each puzzle.
“I think the right way to do armchair economics since I'm sitting here at an armchair is to not just look at one strange thing but try to collect a whole bunch of strange things lay them out on the table”
People value health insurance enormously when choosing jobs, forgoing large amounts of salary for generous coverage, yet at the moment of medical choice they are not willing to spend nearly as much, which is the opposite of the usual intuition.
“they're willing to forego enormous amounts of salary to get generous health care health insurance packages but at the moment of choice they aren't actually willing to spend quite as much”
If the signaling theory is correct, medical spending is largely wasteful and taxing the signal would be better than subsidizing it, but this is hard to do at the national level because nations compare medical spending with each other.
“it's better to tax the signal than to subsidize it but then of course when we there's a national competition”
Betting markets make people reconsider their views: Hanson, a skeptic of the RAND result who still sees some value in marginal care, says he would bet that a radical health reform would barely change life expectancy, but he would not say this in a casual interview.
“when you ask me how different is life expectancy gonna be after a 10 year run of this radical new policy I'd say oh that's not gonna change much there you go and that's the beauty of a betting market”
Most people are poor at overcoming their own biases; although there is a long literature on cognitive biases, people seem surprisingly uninterested in seriously overcoming them.
“people seem surprisingly uninterested in actually seriously overcoming those biases”
Medical care is structured in ways that are weakly connected to quality because bureaucrats and engineers make treatment decisions, and their engineering destroys incentives for performance, while those decisions rest on overconfident beliefs about what treatments do.
“an unbelievable amount of engineering social engineering that people are totally unaware of that goes on in healthcare that destroys the incentives for performance”
Explaining a set of strange facts about health care requires gathering many anomalies together and seeking a single simple explanation; ad hoc explanations for each anomaly are undisciplined.
“try to collect a whole bunch of strange things lay them out on the table and stare at them for a while”
The puzzle has two parts: medicine's effect on health looks null even though people keep buying it, and medical journals keep publishing studies that show benefits, so the appropriate response is to collect many such anomalies and look for a single explanation rather than ad hoc fixes for each.
“the right way to do armchair economics since I'm sitting here at an armchair is to not just look at one strange thing but try to collect a whole bunch of strange things lay them out on the table”
Hanson's method is 'armchair economics': collect many strange facts about a domain, lay them side by side, and look for a single simple explanation that accounts for many of them, since ad hoc explanations for each oddity are too easy and undisciplined.
“the right way to do armchair economics since I'm sitting here at an armchair is to not just look at one strange thing but try to collect a whole bunch of strange things lay them out on the table and stare at them”
People buy blank-check medical coverage like an all-you-can-eat restaurant so they never worry they have missed something, and waste on wasteful care may be felt as reassurance that one did everything possible.
“the blank check is what optional stuff is it's awesome because that way I'll have it all”
A second explanation for medicine's puzzles is fear of death: people avoid thinking about death, so they hand health decisions to doctors and insurers, buy 'blank check' coverage, and avoid thinking about medical choices.
“we're just terrified of thinking about death no doubt about that and we all sort of know that”
Betting markets produce firm numerical estimates where white papers would hedge, because people who put money down reveal actual beliefs.
“if you asked people to write a white paper to predict this analysis they would have all sorts of qualifiers in hem and haw and it'd be hard to actually get them to make a forecast”
Conditional prediction markets can forecast policy effects: for example, a market could estimate US life expectancy ten years after a Democratic health reform passes versus if it fails, and Hanson predicts the two prices would be nearly identical, implying little health effect.
“if their health care reform passes then what is the life expectancy in the United States ten years later”
People vastly overestimate how much they know about most things they buy, so the argument that patients cannot evaluate medical care applies to many other purchases too, and delegating evaluation to an agency raises the question of whether government does it better.
“people vastly overestimate how much they know about everything else they buy”
Health care and health insurance are not as important for producing health as people think, implying that what matters is health itself rather than the health care or insurance people obsess over.
“neither health care nor health insurance are as important in producing health as we might have thought”
Consumers cannot easily judge medical quality, but people overestimate how well they understand most of what they buy, so the difficulty of evaluating medicine does not uniquely justify restricting consumer choice.
“people vastly overestimate how much they know about everything else they buy that's really pretty ignorant about most things we buy we make do”
Regulators and society should regulate higher-risk activities more intensely than lower-risk ones, and people want regulation of small health risks such as a crappy-tasting apple with a tiny risk, partly because wealthier people want to consume less risk, an income effect.
“some of that as I assume what we call in economics an income effect as we get wealthier we tend to want less risk”
Most people overestimate how well they know what they buy, and most of what consumers buy they make do with limited knowledge, so the difficulty of evaluating medical quality is not in itself a special reason to restrict consumer choice.
“people vastly overestimate how much they know about everything else they buy that's really pretty ignorant about most things we buy we make do”
Many puzzling patterns in health behavior exist: people want medicine from government or employers, value family involvement in care, married people and parents get more medicine, women get more medicine, and people demand regulation of small health risks more than other big life choices.
“married people get have more medicine than unmarried people especially people with children”
Patients in the 1970s and 1860s had good reason to be skeptical of doctors, as the best medical devices were saws and doctors often spread disease, yet people kept going to doctors anyway.
“going to the doctor in in 1860 where his main medical device was a saw oral is oral each is really not a good thing”
Puzzling patterns in health behavior include people wanting medicine from government or employers, married people and parents getting more medicine, women getting more care, and strong demand for regulation of small health risks such as an apple's tiny risk.
“married people get have more medicine than unmarried people especially people with children”
People want medicine from family and communities, and married people and parents receive more medical care, which suggests health care serves emotional and relational functions beyond physical benefit.
“married people get have more medicine than unmarried people especially people with children you have more medicine than people without children”
The RAND experiment measured health through about 25 physical measures and overall health indicators such as feelings and functioning; no overall measure showed an effect, including when broken down by income, initial health, or the four groups formed by those splits.
“they had 25 different measures like blood pressure and then they had some overall health measures”
The RAND Health Insurance Experiment, begun in the late 1970s, randomly assigned about 5,000-7,000 people in six US cities to free ('blank check') medical care or to cost-sharing plans.
“we take we took five thousand seven thousand people in six US cities and we randomly assign some of them to get free medicine”
Entrepreneurs are overconfident about how long projects will take and how much money they can raise, a bias similar to managers' over-optimism about project timelines.
“entrepreneurs overconfidence about how long it will take them to produce something right or raise the money they need”
Socialized medicine with rationing can give the same reassurance at lower cost, because people feel they did not get enough care but are reassured that the government decided the care was not worth it, though many find this less comfortable.
“people who are supporters of socialized medicine have an argument when they say well when we have the government decide how much medicine to get”
People do not spend everything on health care even in emergencies; in the emergency room they refuse some tests and treatments that are not worth the cost or are too risky, and they act differently when paying out of pocket.
“people don't write a check for everything they don't spend everything”
Food shows that people can value something enormously without paying high prices, because competition among suppliers pushes prices down; the same could be true of health care if more competition were allowed.
“there's competition among food suppliers brings the price down and that competition protects us”
Bryan Caplan's argument is that people do not realize low copays encourage overuse and that they pay for this through higher premiums, which is why they prefer employer bundles with low copays.
“people just don't realize it they don't realize that that low copay encourages overuse”
Most people's concern is health insurance, but what people really care about is health care and, ultimately, health itself; neither health care nor health insurance seems as important for health as people assume.
“it's not only not health insurance that matters it's not only health care that matters it's just health and unfortunately neither health care nor health insurance are as important in producing health”
US law restricts prediction markets: only Las Vegas has a special legal exemption, and that exemption allows only sports betting, which is why policy and political betting markets are hard to create legally.
“Las Vegas has that legal exemption actually Las Vegas can only bet on sports”
Competition among firms should eventually push them toward accurate forecasting, although reputational and political reasons can delay this, and entrepreneurs' overconfidence about how long projects take shows a similar bias.
“the competition among firms of course is going to push firms toward more accuracy”
Betting on policy outcomes is hard in practice because unexpected intervening events complicate resolution, as when a basketball bet is affected by injuries or brawls, though Hanson argues that is a complication of most real-world bets, not a fatal flaw.
“the problem is before I put that money down I'd be worried about inter unexpected intervening events”
Entrepreneurs' overconfidence about how long projects take and how fast they can raise money is the same kind of bias as managers' over-optimism about schedules.
“entrepreneurs overconfidence about how long it will take them to produce something”
Willingness to pay a lot for something like health does not mean markets fail, because competition among suppliers brings prices down, as with food, and the same protection could apply to health care if more competition were allowed.
“the fact you'd be willing to pay a lot for it does not mean that therefore markets don't work because it's true food so for example people really like food”
The main reason health care lacks bang for the buck in America is that people spend others' money: when care is nearly free at the point of use, people use it anyway, so the cost to the patient disconnects use from value.
“if it's almost free to me I'm just gonna have it anyway is there any reason to think that that's not a major problem in America today”
Competition among firms should eventually push organizations toward using accurate forecasts, though reputational and political motives can delay that for a long time.
“the competition among firms of course is going to push firms toward more accuracy”
Hanson's main two possible responses to the RAND null result are that the study was too short (three to five years) and too small with too few older people, or that the result is real and medicine's failure to help needs an explanation.
“it was only three to five years you needed more old people yes you need more and that was 30 years ago”
The argument that people will spend whatever it takes on health care, so markets fail, is false historically: medical spending was about 1-2% of the economy through most of history and has only recently risen to about 16%.
“through most of history a medical spending was a you know one or two percent of the economy at most recently it's up to 16 percent”
The overall lesson is that the world is weirder than people appreciate; the first step is to acknowledge and list the weirdness, and only then build coherent explanations for many puzzles at once.
“my overall theme here is the world is weird in ways that we don't appreciate and we don't like to acknowledge and the first step is to acknowledge it's weird”
Government regulation based on consumers delegating evaluation to an agency requires showing that the agency will do better than other institutions, which is hard to establish.
“in order to justify some sort of regulation there you'd have to make that which is sure hard to do right”
Medical care is not fundamentally different from other big purchases like houses, jobs, or careers; bigger consequences justify more thought but not by themselves more regulation or subsidy.
“by itself that doesn't mean they should be more regulated or more subsidized”
People regard their own decisions in a market as informed and disciplined, yet they accept reliance on expert institutions for medicine even though they cannot verify the experts' claims.
“if you're ignorant about this you'll have to delegate to some institution a consumer reports and evaluation committee or something”
Health care is not obviously different from other goods; people make big high-stakes purchases like houses, jobs, and careers without justifying heavy regulation, so scale of consequences alone does not justify government intervention in medical markets.
“obviously bigger choices have bigger consequences and we should think more about them but by itself that doesn't mean they should be more regulated or more subsidized”
Dating shows that people rely on ancient costly signals like sportiness, wit, and conspicuous expenditure to show health, wealth, and intelligence, even though modern signals such as bank statements or IQ scores are more efficient, because ancient signals are deeply embedded.
“those ancient signals of health wealth and intelligence still dominate the dating scene”
Making decisions through political romance may be preferable to hard-nosed estimates: people prefer a romantic politics that is comforting, and a betting market would take some romance out of it, which Hanson sees as a benefit but acknowledges most people would not want.
“a lot of people would not like that world they prefer the more romantic world”
People regulate high-probability health risks more intensely than low-probability ones, while ignoring large life-shaping choices such as marriage, children, and moving, which are not regulated despite their consequences.
“people say choose to get married or choose to have children or choose to move cities or jobs those are big important decisions they have a lot of consequences”
Medicine behaves like a status display in dating: people signal health, wealth, and intelligence using ancient signals (sports prowess, expensive clothing, wit) rather than modern, more efficient ones (medical records, bank statements, IQ scores).
“those ancient signals of health wealth and intelligence still dominate the dating scene”
Medicine and social institutions often lie to us about how effective care is, so Hanson is eager to build institutions that do not lie, such as prediction markets, which he sees as the practical response to the signaling problem.
“once you realize how little we know and how much our social institutions lie to us about these various things I'm eager to create institute social institutions that don't lie as well”
Regulation of product safety, such as car safety standards, may also be driven by a desire to avoid blame for children's deaths, not just by actual safety benefits.
“I don't have to worry about my kid buying an unsafe car because the government makes sure all the cars are safe”
Under the Canadian rationed system, people feel guilty about not getting enough care, and the existence of the option to move countries undermines the reassurance that rationing gives.
“the so the more that becomes an option the less reassuring that”
The fear-of-death explanation also predicts that people accept regulation and reassurance, such as government car-safety rules, because they don't want to be blamed for a loved one's harm.
“I don't have to worry about my kid buying an unsafe car because the government makes sure all the cars are safe and I don't have to think about my kid dying in a way that I would be to blame”
Medicine's apparent failure to deliver value is partly explained by people's willingness to accept seemingly useless care as evidence of effort, and by a sense of blamelessness: when care fails, people feel blameless because authorities said it was not worth it.
“you feel guilty about not having gotten enough because everybody around you does the same thing and you you did what you were supposed to do”
Several limits of the RAND study remain: it lasted only three to five years, enrolled too few older people, and covered a single 1970s experiment, so it may not generalize to current medicine.
“it was only three to five years you needed more old people yes you need more and that was 30 years ago everything's changed”
US health spending was about 1-2% of the economy for most of history and has risen to about 16% recently, a shift that frames the question of whether more spending means better health.
“through most of history a medical spending was a you know one or two percent of the economy at most recently it's up to 16 percent”
Prediction markets on US politics are restricted because Las Vegas has a legal exemption that allows only sports betting, so a presidential market is effectively illegal, and gambling operators avoid political markets to limit scandal risk.
“Las Vegas has that legal exemption actually Las Vegas can only bet on sports”
Medical spending was about one or two percent of the economy through most of history and has risen to about 16 percent recently, so the idea that people will spend whatever it takes on health care is historically false.
“through most of history a medical spending was a you know one or two percent of the economy at most recently it's up to 16 percent”
The uncertainty-first approach implies that a policy adviser who does not understand health care should avoid detailed recommendations and instead recommend stepping back.
“I'll just recommend taking your damn hands off it and stepping back and realizing you don't understand”
The world is weird in ways we do not appreciate, so the first step is to acknowledge and list the peculiarities before trying to construct coherent explanations.
“the world is weird in ways that we don't appreciate and we don't like to acknowledge and the first step is to acknowledge it”
The 'health' concept is expandable: once a good is labeled medicine, it is hard to limit, as with mental health parity, dental parity, eyeglass parity, and even big-screen televisions for counseling sessions.
“the whole concept of medicine is pretty expandable if you think about it too”
Politics is romantic, and people may not want hard-nosed probability estimates for reform because such estimates would remove the romance from politics, while a figurehead like a queen can absorb the romance and leave real decisions to others.
“a lot of people would not like that world they prefer the more romantic world”
The RAND null result could be due to study limits: it lasted only three to five years, had too few elderly and too few deaths to measure mortality, and used one experiment of thousands rather than millions of people.
“it was only three to five years you needed more old people yes you need more and that was 30 years ago”
A person may avoid a hospital after reading about it in the paper so that a loved one sees they care, illustrating how public signals drive medical choices more than private quality information.
“I should take her somewhere else to show that I care”
Health care is also used to show social status and loyalty, which helps explain why low-status people are especially focused on regulating health and why status correlates with health.
“you can also explain some of the aspects of the social status and health correlation and why we want to regulate health a lot especially for low status people”
Patients' disinterest in hospital mortality data may be rational: people trust a doctor friend's referral to a surgeon over a ranking, so they may already be relying on trusted social information rather than on public statistics.
“I have a good friend who is a doctor ... I go along probably with my friends recommendation”
The death-anxiety explanation accounts for only a modest fraction of the puzzles, so it is not an impressive unifying theory, even though it explains some of them.
“it explains some of them not an impressive fraction but you know get you a certain way”
Roberts counters that the signaling argument is interesting but unconvincing, and suggests that social status and health correlations, and the desire to regulate health especially for low-status people, can also be explained by the same general approach.
“I don't like the argument Robin but it's really interesting”
Two responses to the RAND null result exist: the study was too short (three to five years), too small, and too dominated by younger people, or the result is real and medicine's value is overstated; Roberts argued the second view is more chilling and thought-provoking.
“one is now they still didn't do it right it was only three to five years”
Roberts argues that the death-avoidance theory explains only a small fraction of the puzzles, and that it is hard to show evidence for it beyond its plausibility.
“what's the evidence that that's true other than it's a thought-provoking idea”
Given how little we understand health care, Hanson is skeptical of detailed policy recommendations and prefers decentralized, bottom-up decisions by patients who spend their own money, because dispersed information and incentives are better there than with bureaucrats.
“I have a strong preference toward decentralized decision making and a bottom-up set of decision making in a world where the information that's dispersed among all of us”
Eliminating all pharmaceutical spending, around 10% of a two-trillion-dollar US health bill, by having government pay drug companies directly would still have a large effect on future health through innovation and spending incentives.
“so that would lower our health care bill from two trillion to one point eight trillion”
The signaling explanation for medicine is hard to accept because it implies people are not much interested in quality; Roberts agrees it is interesting but finds it uncomfortable, and suggests status effects also explain why low-status people demand more regulation.
“I don't like the argument Robin but it's really interesting”
Hanson presents his 14 wild ideas list, of which he says about one third are true, as an example of the unusual and often mistaken views he publishes.
“Robin has a page what we'll post called I think it's called 14 wild ideas and I think is one third of which a third of which are true”
Roberts' rebuttal that patients are sophisticated and that a 1% versus 5% mortality gap could reflect case mix does not fully explain the behavior, since patients who ignore quality data still fail to act even when a trusted doctor-friend's recommendation is available.
“I'm gonna argue that people are sophisticated and they understood that a 5% mortality rate versus a 1 percent mortality rate could be due to lots factors”
Roberts doubts the signaling argument and suggests that social status and health correlations, and the tendency to regulate health more for low-status people, can also be explained by the same general approach.
“you can also Blayne some of the aspects of the social status and health correlation”
Pharmaceutical purchases are only about 10-15% of US health spending, roughly $200 billion of the $2 trillion total, so making drugs free would cut spending only modestly and would be a small share of the bill, but it would delay one year's cost growth.
“10 percent of two trillion is 200 billion so suppose we made all pharmaceutical payments zero as a government program”
Hanson's ideas are often wild, and about one third of his 14 'wild ideas' are true.
“I think is one third of which a third of which are true which is shocking”