I reported yesterday on the RAND health insurance experiment: Thousands of people randomly given free medicine in the late 1970s consumed 30-40% more medical services, paid one more "restricted activity day" per year to deal with the medical system, but were not noticeably healthier. So unless the marginal value of medicine has changed in the last thirty years, if you would not pay for medicine out of your own pocket, then don’t bother to go when others offer to pay.
If people knew that they weren't going to get medical care, could they not have been more careful, resulting in better health? Is this accounted for in the study?
"the marginal value of medicine is zero" means that the value of more medicine than you are willing to pay for is zero. Whether this corresponds with as much as you need is an issue of contention.
Obviously for any non-free medicine a person (however rich or poor they are) will have to sacrifice whatever else they would have done with the money, so there will be an effect on their other consumption (school, whatever). Subsidised medicine will undeniably relieve this. However it will also compel them to use a lot more medicine, the benefit of which appears to be zero (at a large cost). Perhaps it would be better to just give them the money in welfare to cover the potential effects on schooling etc. and not pay for all the useless medicine. A usual argument for not doing this (people will spend it all on other things and still have no healthcare) is irrelevant if these costs of not having money for healthcare are borne by their lifestyle (where they are presumed to have spent the money).
As usual, I'm having some trouble understanding Robin Hanson, who, while the thematics of his posts are interesting, has trouble expressing himself quite as clearly as other contributors to this blog.
I wonder, for instance, about the value of Robin's finding that "the marginal value of medicine is zero". If I can paraphrase this correctly in language that is slightly more plain, the meaning should be: "once you have had the medicine you need, the value of additional medicine is zero".
Well, duh.
Pretty much the same could be said about restaurant meals. Once you have eaten your fill, the marginal value of an additional meal is zero. That is, until you are hungry enough to want to eat again.
Similarly, the same could be said about transportation. Once you have used a mode of transportation to go where you want to be, the marginal value of more transportation is zero. That is, until you want to go somewhere else again.
In the same vein, medicine obviously has value when you need it. I once got bitten deeply by a cat, and if I couldn't get antibiotics, I would be typing this now single-handed. A friend in his thirties got cancer, and if he didn't get immediate chemotherapy, friends would now be visiting him at the graveyard instead of at his place.
Given how important their own health is to most people, including poor people, one would expect that the value of free health care cannot be measured by looking at how healthy poor people are. Obviously, if they have any kind of health-threatening condition, and they're not on a free health care plan, they will scrap up the money they have and give it to the doctors to get help. The difference between a free plan and pay-for-yourself will not be observed in the poor people's health; it will be observed in their living standards, in their childrens' education, and so on.
I'm not saying this to argue that we should be giving free health care to everyone. There are other reasons, including matters of principle, to consider when toying with that idea. For example, do we give free health care to monkeys? If not, then why should we be giving free health care to borderline people? And if we give free health care to borderline people - then I say, we should also give it to monkeys.
We need to have an altruism cutoff at some point. If we have literally no altruism cutoff, then we need to go as far as to provide free health care to bacteria. A person with no altruism cutoff cannot live, because we depend on consuming other organisms to function.
Most people draw the altruism cutoff line at the edge of the human species, regardless of the quality of the recipient human specimen. I think such people should speak for themselves, and help others however much they want, but not force the same altruism cutoff on other people, because drawing it at the edge of the human species, including substandard specimens, is arbitrary and is not at all self-evident.
That would be a principled argument about whether or not there should be universal health care. On the other hand, saying that we studied X poor people, observed that their health did not improve dramatically under free health care, and concluding that the value of free health care must be zero, is misleading, because it ignores differences in other things that were not under consideration in the study - such as, say, whether the poor people's kids are still going to school.
An interesting question is whether the added medical spending was mostly at the patients' request, or recommended by the doctor based at least partly on the existence of good insurance coverage. It seems like we could be seeing a pattern of patients wanting more medical care because of lower cost, or a pattern of doctors recommending more medical care because of higher price.
Stuart, I'm not sure which "the summary" you mean,
The summary on the link you gave to the Rand website.
many people went fishing for results, and reported "significant" results that didn't correct for their data mining.
Fair enough. But the Rand report "The Effect of Coinsurance on the Health of Adults - Results from the RAND Health Insurance Experiment" from 1984 (the earliest one I could find) claim:
Poor people at elevated risk apparently benefited from receiving free care...However, they don't make clear whether this is connected with the spectacles or the hypertension measures. I'll just categorise it as 'unresolved' and move on (after all, a health plan could be easily tweaked to provide extra benefits to the sick poor, if needed).
I'm not familiar with the literature, so I can't suggest the best among the many, but the report I cited from your last article seems to be relevant, even if it isn't perfect.
I think it is sufficient to say something like, "Based on the RAND study, it appears that there are diminishing returns on how much medical care someone recieved in the 1970's. Comparing the outcomes of no insurance to some insurance was not studied, but as coverage increases, it reaches a point where there is no apparent benefit of any kind, or at least on aggregate the benefit equals the risk."
Do you see the difference between that and "Thousands of people randomly given free medicine in the late 1970s consumed 30-40% more medical services ... but were not noticeably healthier."
If I were to stick with your formulation, I would change it to read, "Thousands of people randomly given free medicine in the late 1970s consumed 30-40% more medical services than others on plans with varying amounts of shared cost ... but were not noticeably healthier."
Chuck, I mean to say everyone had at least catastrophic insurance with a high deductible. We must draw some general conclusion about the marginal value of medicine. What better source could you suggest than this experiment?
Stuart, I'm not sure which "the summary" you mean, but because the experiment didn't find any results according to the statistical tests they originally planned, many people went fishing for results, and reported "significant" results that didn't correct for their data mining.
"David, everyone had catastrophic insurance with a high deductible, but that can still leave them paying full price for most of their medicine. They looked separately at just the poor people in their sample and saw no effect on that group."
It isn't true that everyone had a high deductable. From the report, there were multiple plans in 3 categories 1) free, 2) individual deductable, 3) intermediate coinsurance, and 4) income-related catasrophic plans. That is from page v of the report. Did I read it wrong?
Robin Hanson:
"I'm not sure why you are describing me as a "proponent of a `market based' system to `fix' our healthcare system."
I apologize if you feel like I put words in your mouth. It seems to me when someone concludes that "free" healthcare is more expensive and doesn't help, that they would be in favor of a different approach, especially in the context of a fairly libertarian blog. In fairness, you've not proposed the approach I described in these posts.
Bottom line is that I think you are over-generalizing. You are comparing the relative benefit of medicine between two sub-groups of insured people and concluding that in every case (including the un-insured) lowering the cost of healthcare won't benefit someone, which is clearly not true.
They looked separately at just the poor people in their sample and saw no effect on that group.
Ehr... I just read the summary and they claim
Serious symptoms[2] were less prevalent for poorer people on the free plan.
They defined serious symptons as:
...chest pain when exercising, bleeding (other than nosebleed or menstrual period) not caused by accident or injury, loss of consciousness, shortness of breath with light exercise of work, and weight loss of more than ten pounds (except when dieting).
And later on they added
The study suggested that cost sharing should be minimal or nonexistent for the poor, especially those with chronic disease.
David, everyone had catastrophic insurance with a high deductible, but that can still leave them paying full price for most of their medicine. They looked separately at just the poor people in their sample and saw no effect on that group.
Robin, I don't understand what you mean. As I understand it from the comments above, everyone in the experiment had insurance, just some people had better insurance than others, and that was the basis of the comparison. No one had anything even close to no insurance. Isn't that right? Also, in my limited understanding of survey methodology, you "oversample" a group if the group is small and you are afraid that your sample in that group will be too small to be representative. But then you reweight those people so that they are back to their actual population share. So oversampling poor people does not mean that it is a study of disproportionately poor people, and even if it were, I doubt would be as poor-heavy as is the population of uninsured people.
But I haven't read the paper so you tell me if I got any of that wrong.
James, the question is how much weight to put on a statistical study compared to your prior expectations. There are always reasons to doubt any study, but the question is how they compare to the reasons you must have to doubt your prior expectations.
I suspect you are being too hard on Jed. Statistics are often misused and can be very misleading. If you can't independently determine the quality of some statistical study but the results of this study contradict what you believe to be true then you should be suspicious of the study's results.
David, the experiment lowered the copay from 95% to 0%, just as insurance does. The experiment oversampled poor people, and looked specifically at the five general health measures for poor people broken down by initial health status of good versus bad.
If people knew that they weren't going to get medical care, could they not have been more careful, resulting in better health? Is this accounted for in the study?
Denis:
"the marginal value of medicine is zero" means that the value of more medicine than you are willing to pay for is zero. Whether this corresponds with as much as you need is an issue of contention.
Obviously for any non-free medicine a person (however rich or poor they are) will have to sacrifice whatever else they would have done with the money, so there will be an effect on their other consumption (school, whatever). Subsidised medicine will undeniably relieve this. However it will also compel them to use a lot more medicine, the benefit of which appears to be zero (at a large cost). Perhaps it would be better to just give them the money in welfare to cover the potential effects on schooling etc. and not pay for all the useless medicine. A usual argument for not doing this (people will spend it all on other things and still have no healthcare) is irrelevant if these costs of not having money for healthcare are borne by their lifestyle (where they are presumed to have spent the money).
As usual, I'm having some trouble understanding Robin Hanson, who, while the thematics of his posts are interesting, has trouble expressing himself quite as clearly as other contributors to this blog.
I wonder, for instance, about the value of Robin's finding that "the marginal value of medicine is zero". If I can paraphrase this correctly in language that is slightly more plain, the meaning should be: "once you have had the medicine you need, the value of additional medicine is zero".
Well, duh.
Pretty much the same could be said about restaurant meals. Once you have eaten your fill, the marginal value of an additional meal is zero. That is, until you are hungry enough to want to eat again.
Similarly, the same could be said about transportation. Once you have used a mode of transportation to go where you want to be, the marginal value of more transportation is zero. That is, until you want to go somewhere else again.
In the same vein, medicine obviously has value when you need it. I once got bitten deeply by a cat, and if I couldn't get antibiotics, I would be typing this now single-handed. A friend in his thirties got cancer, and if he didn't get immediate chemotherapy, friends would now be visiting him at the graveyard instead of at his place.
Given how important their own health is to most people, including poor people, one would expect that the value of free health care cannot be measured by looking at how healthy poor people are. Obviously, if they have any kind of health-threatening condition, and they're not on a free health care plan, they will scrap up the money they have and give it to the doctors to get help. The difference between a free plan and pay-for-yourself will not be observed in the poor people's health; it will be observed in their living standards, in their childrens' education, and so on.
I'm not saying this to argue that we should be giving free health care to everyone. There are other reasons, including matters of principle, to consider when toying with that idea. For example, do we give free health care to monkeys? If not, then why should we be giving free health care to borderline people? And if we give free health care to borderline people - then I say, we should also give it to monkeys.
We need to have an altruism cutoff at some point. If we have literally no altruism cutoff, then we need to go as far as to provide free health care to bacteria. A person with no altruism cutoff cannot live, because we depend on consuming other organisms to function.
Most people draw the altruism cutoff line at the edge of the human species, regardless of the quality of the recipient human specimen. I think such people should speak for themselves, and help others however much they want, but not force the same altruism cutoff on other people, because drawing it at the edge of the human species, including substandard specimens, is arbitrary and is not at all self-evident.
That would be a principled argument about whether or not there should be universal health care. On the other hand, saying that we studied X poor people, observed that their health did not improve dramatically under free health care, and concluding that the value of free health care must be zero, is misleading, because it ignores differences in other things that were not under consideration in the study - such as, say, whether the poor people's kids are still going to school.
Should I Get LASIK?
Hansonian doubts aside, I've been thinking about getting LASIK (laser corrective eye surgery) for a couple of years. By making...
Should I Get LASIK?
Hansonian doubts aside, I've been thinking about getting LASIK (laser corrective eye surgery) for a couple of years. By making...
An interesting question is whether the added medical spending was mostly at the patients' request, or recommended by the doctor based at least partly on the existence of good insurance coverage. It seems like we could be seeing a pattern of patients wanting more medical care because of lower cost, or a pattern of doctors recommending more medical care because of higher price.
Stuart, I'm not sure which "the summary" you mean,
The summary on the link you gave to the Rand website.
many people went fishing for results, and reported "significant" results that didn't correct for their data mining.
Fair enough. But the Rand report "The Effect of Coinsurance on the Health of Adults - Results from the RAND Health Insurance Experiment" from 1984 (the earliest one I could find) claim:
Poor people at elevated risk apparently benefited from receiving free care...However, they don't make clear whether this is connected with the spectacles or the hypertension measures. I'll just categorise it as 'unresolved' and move on (after all, a health plan could be easily tweaked to provide extra benefits to the sick poor, if needed).
I'm not familiar with the literature, so I can't suggest the best among the many, but the report I cited from your last article seems to be relevant, even if it isn't perfect.
I think it is sufficient to say something like, "Based on the RAND study, it appears that there are diminishing returns on how much medical care someone recieved in the 1970's. Comparing the outcomes of no insurance to some insurance was not studied, but as coverage increases, it reaches a point where there is no apparent benefit of any kind, or at least on aggregate the benefit equals the risk."
Do you see the difference between that and "Thousands of people randomly given free medicine in the late 1970s consumed 30-40% more medical services ... but were not noticeably healthier."
If I were to stick with your formulation, I would change it to read, "Thousands of people randomly given free medicine in the late 1970s consumed 30-40% more medical services than others on plans with varying amounts of shared cost ... but were not noticeably healthier."
Chuck, I mean to say everyone had at least catastrophic insurance with a high deductible. We must draw some general conclusion about the marginal value of medicine. What better source could you suggest than this experiment?
Stuart, I'm not sure which "the summary" you mean, but because the experiment didn't find any results according to the statistical tests they originally planned, many people went fishing for results, and reported "significant" results that didn't correct for their data mining.
Robin Hanson wrote:
"David, everyone had catastrophic insurance with a high deductible, but that can still leave them paying full price for most of their medicine. They looked separately at just the poor people in their sample and saw no effect on that group."
It isn't true that everyone had a high deductable. From the report, there were multiple plans in 3 categories 1) free, 2) individual deductable, 3) intermediate coinsurance, and 4) income-related catasrophic plans. That is from page v of the report. Did I read it wrong?
Robin Hanson:
"I'm not sure why you are describing me as a "proponent of a `market based' system to `fix' our healthcare system."
I apologize if you feel like I put words in your mouth. It seems to me when someone concludes that "free" healthcare is more expensive and doesn't help, that they would be in favor of a different approach, especially in the context of a fairly libertarian blog. In fairness, you've not proposed the approach I described in these posts.
Bottom line is that I think you are over-generalizing. You are comparing the relative benefit of medicine between two sub-groups of insured people and concluding that in every case (including the un-insured) lowering the cost of healthcare won't benefit someone, which is clearly not true.
They looked separately at just the poor people in their sample and saw no effect on that group.
Ehr... I just read the summary and they claim
Serious symptoms[2] were less prevalent for poorer people on the free plan.
They defined serious symptons as:
...chest pain when exercising, bleeding (other than nosebleed or menstrual period) not caused by accident or injury, loss of consciousness, shortness of breath with light exercise of work, and weight loss of more than ten pounds (except when dieting).
And later on they added
The study suggested that cost sharing should be minimal or nonexistent for the poor, especially those with chronic disease.
David, everyone had catastrophic insurance with a high deductible, but that can still leave them paying full price for most of their medicine. They looked separately at just the poor people in their sample and saw no effect on that group.
Robin, I don't understand what you mean. As I understand it from the comments above, everyone in the experiment had insurance, just some people had better insurance than others, and that was the basis of the comparison. No one had anything even close to no insurance. Isn't that right? Also, in my limited understanding of survey methodology, you "oversample" a group if the group is small and you are afraid that your sample in that group will be too small to be representative. But then you reweight those people so that they are back to their actual population share. So oversampling poor people does not mean that it is a study of disproportionately poor people, and even if it were, I doubt would be as poor-heavy as is the population of uninsured people.
But I haven't read the paper so you tell me if I got any of that wrong.
James, the question is how much weight to put on a statistical study compared to your prior expectations. There are always reasons to doubt any study, but the question is how they compare to the reasons you must have to doubt your prior expectations.
Dear Eliezer Yudkowsky,
I suspect you are being too hard on Jed. Statistics are often misused and can be very misleading. If you can't independently determine the quality of some statistical study but the results of this study contradict what you believe to be true then you should be suspicious of the study's results.
David, the experiment lowered the copay from 95% to 0%, just as insurance does. The experiment oversampled poor people, and looked specifically at the five general health measures for poor people broken down by initial health status of good versus bad.