So: onto adverse selection.
Adverse selection is the idea that information asymmetries in markets can lead to sub-optimal outcomes. Say 30% of all used cars are lemons that will cost you a fortune to repair. A new car is worth $10,000 while a good used car is worth $5,000, but a lemon is only worth $1,000, because of the repairs.
The problem is, you can't tell which cars are lemons. So you won't be willing to pay $5,000. You might be willing to pay more like $3,500, which compensates you for the risk. But the owners of the good cars will not wish to sell them at such a steep discount. The owners of the crap cars, however, will leap at the chance to unload them for such a great price. The percentage of lemons in the market goes up to 50%. You demand a bigger discount to compensate you for the higher risk; now you'll only pay $2,500. More owners of high-quality used cars decide to keep them rather than buy a new one. The percentage of lemons rises to 70% . . .
You can see where this is going. And, presumably, how it might apply to insurance markets: you end up with a pool of very sick people who cost a lot to cover. This (along with a lavish schedule of mandatory benefits) is arguably why health insurance in New York State costs so much.
Adverse selection is a favorite explanation of why markets for health care can't work, and health care therefore needs to be provided by the government. Often, the proponents of this theory add a wrinkle: insurance companies spend huge amounts of money on trying to keep from treating people.
Alex Tabarrok points out a hole in this theory:
If insurance companies do avoid covering people who are "likely to need care," this suggests that the uninsured are unhealthy. But 60% of the uninsured are in excellent health (Table 10) (In fact, overall the uninsured are only slightly less healthy than the insured).
Henry Farrell voices the objection that immediately occurred to me.
But the statement '[Insurance companies] try to avoid covering people who are actually likely to need care' very obviously does not imply the statement 'All uninsured people are being refused coverage because they have expensive conditions.' The logical connection between the two implied by the 'contra Paul' bit is not, to put it mildly, clear to me. As a result, I am not sure what Alex's actual point is. Is he suggesting that the incentive problems that Paul identifies are in some sense unimportant?
But on reflection, I think Henry and I were wrong. I was very surprised to follow through to the table Alex Tabarrok links (from the Kaiser foundation, hardly a right-wing advocacy group) and find out that the percentage of the uninsured* who are in "fair" or "poor" health really isn't much larger than the percentage of the insured in those categories: 10.3% of the uninsured, versus 8.4% of the insured.
That is surprising because we would expect the uninsured to be sicker than the general population, even if the insurance companies were doing nothing to weed out the sick. Being uninsured is correlated with other things that are strongly correlated with poor health: being born in another (poorer) country; being too sick to work full time; little education; low SES.
Of course, it's also true that the population of the uninsured is correlated with something that's also correlated with good health: being young. But then, this sort of undercuts the adverse selection argument, and also the moral imperative of giving them health insurance. If you could reasonably afford health insurance by dropping down to a lower-priced cell phone plan and cutting back on your bar tab, you are not a national emergency.
So for the nonce, let's ignore all those confounding factors, and assume that without virulent machinations by the insurance companies, the population of the uninsured ought to mirror the population of the insured in health status. How many people are being pushed out of the insurance market because of their poor health status? It can't be many; there are only 4.6 million people without health insurance who report poor or fair health status. And not all of those people are very sick. I'd be hard pressed to call my health better than fair because of my asthma and my Hashimoto's Thyroiditis. But neither of those diseases costs my insurance company, or me, much.
If there were no evil insurance companies involved, what would those percentages look like? A slightly higher percentage of the insured market would report fair health, and a slightly lower percentage of the uninsured market. But the difference would be less than 1 million people: 0.3% of the US non-elderly population, and 0.2% of the total population.
What this means is that even if this is a problem, it is not a big problem. You don't gut rehab the US health system because of a "market failure" affecting a small fraction of one percent of the population.
Of course, as Alex notes, that doesn't mean that uninsured sick people aren't a problem. But as both he and Tyler have pointed out, they seem more likely to be a distributional problem than a market failure. In other words, the problem is not that the market cannot provide them insurance; the problem is that they don't have the money to buy it.
This seems meaningless, a distinction without a difference, to most people. But in fact, distributional problems and market failures call for very different kinds of solutions. You fix distributional problems with cash or vouchers. You fix market failures with regulation or some other form of intervention.
Now, arguably, there could be a market failure that is being fixed by existing government interventions: regulations preventing insurance companies from ripping off customers; government programs to cover the indigent, which take care of those who have become too sick to work. But I have to agree with Alex; the empirical data does not seem to back up the notion that there is a large and persistent selection problem in current markets.
How many of the "people who can't afford insurance" are no longer uninsured because they're covered by Medicaid/Medicare?
Does including this group skew the numbers for your argument?
Because if you're uninsured and in poor health and either poor or aging, you make it onto the public roles eventually.
That's sort of not relevant to the discussion of building new programs.
If people have to spend themselves down to the poverty level or go without treatment until they are unable to work, those are deadweight losses that are relevant to building new programs.
It's relevant because it bears on whether it's actually true that the uninsured aren't much less healthy than the insured. The "Medicare" part isn't relevant to this data, because the data is for the "nonelderly uninsured" -- anyone on Medicare presumably isn't part of this table. But the "Medicaid" part means that a lot of people who are sick-because-poor are actually in the "insured" column, not the "uninsured" column. You need to exclude the people covered by Medicaid so you can instead compare those covered by private insurers against those lacking coverage.
Acutally, the Medicare part does matter if people in their late 50's and early 60's put off getting medical care, waiting for that magic 65.
Oh, wait, I think I see what you're saying. You're saying it doesn't matter because if they are poor and sick, Medicaid already picks them up, so why do we need a new program?
Exactly.
And yet the mantra is that because we can't control costs of Medicare/Medicaid, we should keep the gov. out?
Yet because the poor/ill/elderly are already covered (or will be covered once they get poor and sick enough), we don't really have a problem with private insurance?
Circular.
I'm not sure the high number of healthy people without insurance really undercuts the adverse selection model. These people are playing the role of the owners of quality used cars, who do not want to enter the market because they will pay a premium because of the lemons (people in poor health).
--
One more quibble I have is the mocking mention of "evil insurance comapanies." I don't like demonizations, and I am sure most people working for insurance comapanies are fine folks who got into their businesses with the best of intentions.
That doesn't mean that the current system doesn't have some incentives that lead insurance companies to behave in ways that are contrary to the common good. And one need not assume that insurance companies are run by moustache-twirling villains to note that they will respond to those incentives, and it should be our goal to limit those.
As it is, it does seem that there are incentives for insurance companies to avoid convering people in poor health, and if they respond to that incentive, that will cause some problems. Noting this is not the same as demonizing.
JohnMcG beat me to one point that I was going to make.
Also, one way that the adverse selection problem is sort-of addressed in the U.S. is by making the majority of health-insurance coverage linked to group plans based on employment. This creates a quasi-pooling equilibrium that allows people who are in bad health (but good enough health to work) to get insurance. But it seems that this is a sub-optimal solution for a number of reasons, both from an efficiency (it creates labor market frictions) and an equity (it makes it harder for the unemployed, who presumably need the help, to get insurance) standpoint.
This doesn't automatically make a government plan better, but I'm not convinced the the adverse selection problem isn't real.
This is exactly what happened at a contracting company I worked for. The company didn't pay anything for employee health insurance. There was an employee plan, but every year it got more and more expensive as the healthier employees dropped it and went with a retail plan. Eventually even people who were in poor health but weren't, you know, waiting for a new liver could do better on the open market.
It's not surprising to me the uninsured aren't much sicker than the insured. People who have problems but aren't yet too sick to work make sure to get a job with good benefits.
I think that the market for individual coverage is less well developed than group coverage. For a short time (2 years) I needed individual coverage but was unable to purchase it at any price. I was quite healthy but didn't meet the underwriting requirements. I wasn't charged more to compensate - just denied. I was easily able to acquire group insurance for a small group (2 people). They didn't even raise the rates after medical underwriting.*
* When purchasing group insurance you will typically get a quote based upon limited information (age, gender) of the group. If you decide to purchase each person will fill out a detailed application with medical history. At this point the insurance company may decide that you carry an additional risk and will charge you a higher premium to compensate. You of, of course, don't have to accept the higher rate and can purchase insurance elsewhere.
Another fact that keeps quite a few chronically ill people insured: spousal benefits. Even if one is too sick to work, one's spouse probably is not. And with domestic partner benefits becoming increasingly common I'm betting there's probably a non-trivial number of such partnerships contracted for the sake of health insurance. Of course marriages too may exist for similar reason: thrity years ago my step-mother married my father in a classic marriage of convenience because, among several reasons, she needed health insurance for herself and her kids (she was widowed; so was my father-- he needed a mother for me; neither one was deceived about the reasons for their marriage. Before anyone jumps on this, I should also mention that they were good friends too, and I would rather have a marriage or partnership based on stable friendship than on some grand romantic passion.)
And as someone else has pointed out, anyone who is really, really sick and uninsured will end up on one of the public programs eventually. Which to some small extent makes the private companies free riders on them. The opposite complaint, that Medicare and Medicaid free ride on the private insurers because of tehir low reimbursement rates, has some truth to it too, but the reverse is that the private plans can offload some of their worst risks on the public plans too.
I think this line of argument, as far as I understand it, is more than a little deceiving (e.g. uninsured are just as "healthy" as insured) and missing a huge countervailing force. Others have made allusions to it, but I think I can flesh out the mechanism a little bit.
Anyone who has any interface with the health care setting on an inpatient basis (e.g. is "sick") will-- if they qualify-- walk out of the hospital with some kind of insurance via social workers. I can't tell you how many patients I have cared for who have had social workers fill out their paperwork during their admission. Hospitals aren't dummies; truly uninsured patients are much less likely to pay. A lot of times the question of insurance for truly poor people (or the elderly) is simply a question of filling out the paperwork. I bet this has a skewing effect and masks any "adverse selection." Your percentages, for example, do not include the type of insurance (e.g. public vs. private).
One of the major benefits that I can see to a mandatory national health care system is that it forces everyone to fill out the paperwork. Not so great from a libertarian perspective, but I am not a libertarian. To a still young physician who is tired of bankrupting the working poor (contractors, small businessmen, etc), seems like a reasonable governmental intrusion. Someone else wrote this (Ezra or someone, can't remember) but every time I see one of those "collect money for young uninsured patient x with horrendous cancer/trauma/autoimmune disease, it is really depressing.
I go back and forth as to whether Megan just doesn't argue in good faith or isn't very smart, because every time I visit her blog I run across something like this.
Anyway, a few points from the numbers that weren't made in the comments above. First, this is junky data, although you'd have to be pretty deep into health care wonkery to know that. The CPS is not a health care survey, even though it's frequently used because it's easy access and big enough to give state results. The question asking for self-reported "health status -- excellent/good-fair/poor" is not a precise indicator of health status. (The insurance questions on the CPS are also way too crude and give answers that are significantly off from almost every specialized health care survey).
But anyway, taking the data at face value, just look at the actual numbers in the chart. Almost 70 percent of the total population says they are in excellent/very good health. So taking yourself out of the top "excellent/very good" category is an indicator that you are in poor health -- in the bottom 30 percent of the non-elderly population in terms of health status. These are the people who cost money (the healthiest 50-60 percent of the nonelderly population has basically no costs).
This bottom 30 percent are a lot, not a little, more likely to be uninsured -- specifically, the uninsurance rate among the bottom 30 percent in self-identified health status is 18.3 percent, while it's 13 percent among the top 70 percent. That means that people in the bottom two categories for health status are a whopping 40 percent more likely to be uninsured than people in the most healthy 70 percent! Not a small difference.
Megan obscures this by just glancing at the fair/poor category, where the populations are so small that the absolute percentage point differences don't look so great. The true percentage differences are a little smaller when you look at only this category, but not that much smaller -- the "fair/poor" people have an uninsurance rate of 17.4 percent, which is still fully one-third higher than the healthy folks. (The slight difference may be because of William's point if you're actively sick, which this category probably is, you get signed up for Medicaid at the hospital).
Then add on to that that these differences have to be adjusted for the large chunk of people who work at big/prosperous or public sector firms and therefore can get covered by good insurance even when unhealthy. My guess is that the gap would be much greater among those working for smaller firms or the self-employed. You also have to adjust for the flaws in using a 1/0 indicator of insurance, which is all that the CPS gives you. What is needed is a measure of real coverage, not insurance status. The individual insurance market is notorious for selling policies that count as "insurance" but don't really cover you when you get sick, because of pre-existing condition clauses, annual or treatment limits, or a refusal to renew when you get sick. This is well known in health policy circles.
Anyway, all of this is conflated about nine different ways because income and health status are so deeply connected, and health status also gives you a greater incentive to sign up for insurance. So you have biases going both ways. But there is basically no evidence in this simple tabulation that Krugman is wrong.
This is circular. You've provided no evidence that citing your health status as "good" means that you're uninsurable; only that you're in the bottom 30%. I'm in the bottom 30% for a number of individually minor reasons, but I'm insurable.
Moreover, you're taking percentages of small numbers and claiming large effects. As a percentage of the overall population, the number of people who are getting forced out of the insurance market has to be small, particularly after you control for the 25% of the uninsured population that is foreign born and therefore usually unhealthier (and has less access to health care), and the even higher percentage who are low SES.
Bringing in Medicare, Medicaid, and private insurance is irrelevant. You're proving that in some other market, there might be a large selection problem, a point I myself brought up in the post. But in the market we actually have, we do not seem to have a big selection problem, because the combination of employment pooling and the Medicaid safety net pick up most of the people who might be selected out.
@ Megan
I would argue that this is a 'market failure' because I have never met a person who has had a serious health problem that hasn't been frustrated to tears with whatever insurance company they used put them through.
Can someone please name a good health insurance company? Seriously. I own my own business and am shopping for some health insurance, which is the best one.
Suppose fire insurance companies were actually finding ways to deny coverage to people whose houses were likely to burn down. And then suppose you decided to find out whether this was happening by comparing the percentages of people with and without fire insurance whose houses had burned down. And, unsurprisingly, you found that the percentages weren't really that different, and thus declared that there was little evidence of adverse selection in the fire insurance market. And that even if it were happening, it wasn't worth a major overhaul of the insurance system (like mandating group rating) because it only affects less than 1% of homeowners.
The flaw in reasoning here is that the percentages will never be very different, because having your house burn down is a very high-cost but very rare event. It only happens to less than 1% of homeowners. Those are the people whom fire insurance is supposed to serve.
The allegation is that health insurance companies seek to exclude from coverage the people who are likely to be extremely expensive to them. The absolute numbers of such people, the ones who wind up getting very sick and costing huge amounts of money, may be relatively small. And so a system of health insurance that excludes many or even all of them might be said to be only affecting a trivial portion of the population. The problem is that you, Citizen, do not know whether you will end up in that trivial portion of the population. This is the rationale for having insurance in the first place. With an insurance system that excludes a whole class of people likely to contain many of the sickest among us, it's not surprising that the excluded class shows only slightly higher percentages of sick people. But that slightly higher percentage probably includes a large chunk of the people whom health insurance exists for. To present the reductio ad absurdum: if the entirety of the sickest 0.2% of Americans fell into that 5% of people denied insurance because they were likely to become sick, it's not just 0.2% of the country or even 5% of the country that would lack insurance due to adverse selection. Everyone in the country would lack insurance, because falling into that 0.2% was what they thought they were insuring themselves against.
Obviously the insurance companies aren't going to be discerning enough to exclude everyone who gets severely, expensively sick. And if it were so clear to the insurance companies that you wouldn't get very sick, it'd probably be clear to you too, in which case you would be well served by getting a cheaper plan and not having to subsidize those poor high-risk suckers. But here's the thing: health insurance is an iterative process. People don't usually get sick overnight, and people don't sign a single health insurance plan at birth that covers them for the rest of their life. (Not in the US, anyway.) The signs that you are going to get sick come along gradually through your lifetime. And through your lifetime, you are likely to end up having to buy health insurance a number of times, with each event presenting the insurance company with an opportunity to check whether you now fall into a high-risk category. You insured yourself five years ago when you didn't fall into the 10% likely to contain the sickest 0.2%. But then you developed high blood pressure, and then you switched jobs or got divorced, and now it turns out you do fall into the 10% likely to contain the sickest 0.2%. And so your new premiums are twice as high. And if you end up falling into the riskiest 5%, and then change jobs again, you won't be able to get insurance at all. And so all the premiums you paid back when you were in the healthiest 90%, against the risk you might someday fall into the sickest 0.2%, were wasted money. As were the premiums of most of the other people in that healthiest 90%. They thought they were insuring themselves, but as it turned out, if they ever really got sick, they would be gradually pushed out of their insurance. It's not just those few who actually get sick that lack insurance. It's everyone.
This is the allegation. What's the best way to test it? Not looking at relative percentages of sickness among the uninsured, because those may not be much higher. Health costs among the sickest of the uninsured vs. the sickest of the insured? But these will be distorted -- the poor uninsured may get Medicaid when they get sick, which will spend less on them than a private insurer might and will technically take them off the "uninsured" rolls anyway, and so forth. How about looking at whether the sickest are likely to lack private insurance? This is obvious: the answer is yes, as commenter MQ above notes from the same data set Tabarrok uses. But that either proves the entire case, or doesn't prove anything, depending on your ideological inclinations.
How about something simpler: investigate whether private health insurers employ lots of highly paid people to try to deny coverage to people who are likely to get sick, or who already are getting sick? And the answer is: yes they do. And presumably they don't think they're wasting their money.
The pools are not the same. On the one hand, they're younger, and thus healthier. On the other hand, they're poorer, and disproportionately foreign born.
Moreover, I'm not arguing that this never happens; only that there's no evidence it's a sizeable problems. As I said in the post, you do not gut rehab the US health system to solve the problems of a tiny fraction of one percent of the population; you find a way to deal with the problems of those people.
Sorry, this is too concise for me to follow. Which pool isn't the same as which?
Because what I was saying was that the evidence you're using to show that it's not a sizable problem isn't the right kind of evidence to show that. What we're talking about is the problem of private insurers shunting the actual sick people out of their pool. That doesn't need to be a sizable fraction of all the people in order to eviscerate the health insurance system. It just needs to be a sizable fraction of the *sick* people. As I said, reductio ad absurdum: if only 0.01% of the country gets leukemia, but the insurers have managed to reliably shift that entire 0.01% out of their pool, then it's not just 0.01% of the population that has no leukemia insurance. 100% of the country has no leukemia insurance.
You need, as Oliver Rivers suggests, to look at the sick people, and see if they have a harder time getting insurance. Not the other way around.
That was a response to Olivier. My point is that this data is very crude, but you can't simply look at p(uninsured/sick) v. p(uninsured/healthy) because there are confounding distributional issues:
1) Sick people may lose their jobs
2) Sick people may be poorer to start with
3) Sick people may be from poorer countries where poor prenatal and early child nutrition make for unhealthier later lives
All of these things are correlated with both poorer health, and uninsured status. But the poor health is not causing the uninsured status.
Aren't we getting our prior probabilities confused?
What you should be interested in is the quesion what are the chances of being uninsured, given that I'm in poor health, vs the chances of being uninsured, given that I'm in good health?
The table doesn't provide enough detail to give a robust answer to that question, but the last column, "Uninsured rate for nonelderly", gives a pretty good indication; 15% of the uninsured are in excellent health, 23% in good health, and 21% in fair/poor health - which is exactly what you'd expect to see if P(being uninsured, given poor heath) > P(being uninsured, given good health).
I'd take that last column as good evidence of market failure, rather than a distribution problem.
Newsflash: Many people with preexisting conditions are very healthy. Check the insurance rates for people with Type I diabetes, hemophilia, childhood leukemia. Your looking at around $1500/month, with out of pocket exposure adding another 15-20K, and possible exclusion for anything that can be related back to your preexisting condition. What percentage of your "60% of the uninsured who are healthy" do these folks comprise? I don't know, but I suspect a vast majority of people who do not have employer based health coverage but do have a preexisting condition like one of these, does not have health insurance. I certainly know several.
We need to be more like Wayne Gretzky, and see were the puck is going instead of where it is. We have some minor problems with adverse selection now. However, we have a number of other issues we are seeking to solve, so we will be changing things dramatically. This may very well create the conditions for a significant increase in adverse selection.
One other major question has been "Where is the competition in health insurance?". The answer may be that the risks of adverse selection have prevented it.
This is a hugely illogical argument. You're arguing about a ratio that is completely unecessary to the argument at hand.
So long as insurance in linked to employment, you'd first need to prove that poor health leads to unemployment in order for your adverse selection model to make sense. It's such a basic mistake it's kind of shocking.
To whit:
In both insured and uninsured populations, you'll have a mix of healthy and unhealthy people.
But the idea of adverse selection doesn't necessitate having more unhealthy people uninsured. It just requires that the unhealthy uninsured have disproportiantely negative outcomes based on market forces.
In this case, the ratios don't matter. What matters is that the unhealthy insured are pooled on group plans that drives their individual costs down, since health insurance is inextricably linked to employment in this country.
Counter to this, you have healthy uninsured people who chose not to pursue health insurance because it's not necessary or cost effective for them, leaving the unhealthy uninsured in need of health insurance but unable to attain it. An extreme adverse selection.
Which is currently the case. You argue that there should be more unhealthy people that are uninsured, but under the employer/insurance plan, this would only occur if being unhealthy was a predisposition to being unemployed - and we have equal protection laws that specifically prevent this. Prove a simple chain of causality.
This is what happens when complex problems become political - we stop seeing the forrest because of all those damn trees. To wit:
Everyone's discussing insurance as if it's one-dimmensional (companies paying claims to patients), when it's really three-dimmensional (actuaries determining premiums, patients paying premiums to the companies, companies paying claims to patients).
If the cost to insure you is higher than average, it doesn't necessarily follow that an insurance company won't want to insure you. It just means that they prefer to charge you a higher premium rate (which isn't always possible because of employer-based pools). Also, if you're likely to get sick several years from now, an insurance company is happy to collect premiums for some period of time and then pay the claims later (again, assuming a favorable premium rate).
Also, on the topic of healthy, uninsured people: given the rising costs of healthcare generally, it is entirely plausible for a young, healthy person to "self-insure" (read: choose to pay out of pocket if & when they get sick, rather than pay premiums "just in case"). This gets into the whole "personal decision vs. government mandate" argument, which I'll leave to the politicos. Point is - not everyone who doesn't have health insurance is a victim (financial, adverse selection, or otherwise).
I'm not sure the data Alex Tabarrok cites support the conclusion he reaches.
If insurance companies do avoid covering people who are "likely to need care," this suggests that the uninsured are unhealthy. But 60% of the uninsured are in excellent health (Table 10) (In fact, overall the uninsured are only slightly less healthy than the insured).
Health status of the non-elderly (from Table 10)
Health status of the non-elderly uninsured (from Table 10)
The first problem is that the non-elderly includes both insured and uninsured individuals. It makes more sense to compare the elderly insured with the elderly uninsured.
Health status of non-elderly insured individuals (computed from Table 10)
Whether you consider the uninsured "only slightly less healthy", based on this data, is a matter of opinion. But there is a far more serious problem with this comparison. Adverse selection and refusal to insure the sick does not occur in company health plans or Medicaid; it only occurs with private individual health insurance. So we should look at the health status of non-elderly individuals with private individual health insurance.
Health status of non-elderly individuals with private individual health insurance (computed from Table 1)
Compared to the non-elderly uninsured, we see a smaller percentage of people in excellent or very good health (51.7% versus 60.2%), probably because of adverse selection. And we see a far smaller percentage of people in fair or poor health (3.8% versus 10.3%), probably because insurers refuse to insure them.
You point out that "there are only 4.6 million people without health insurance who report poor or fair health status". I think this misses the point. There are 45 million non-elderly Americans, or 17.2% of the non-elderly population, who lack health insurance. Any one of these individuals could become catastrophically ill tomorrow. Most trapeze artists are protected by a net even though they do not expect to need it.
Megan, I think you've misunderstood the adverse selection that is expected to go on in insurance markets.
There is asymmetric information, but the presumption is that the purchaser of the insurance, not the insurer, is the one with the better information. Healthy people cannot buy insurance at actuarially fair prices because if insurance companies were to make such an offer, unhealthy people would jump on board and make things unprofitable. But if an insurance company offers something that is only actuarially fair to unhealthy people (expensive insurance) then it is not as if healthy people can crash the party.
In practice, you cannot derive much meaningful information from a survey of the characteristics of the uninsured for two reasons:
One, the majority of people are insured through their employer (and most others through means-tested or age-tested programs such as Medicaid and Medicare), and within these domains, adverse selection is not a powerful force. People have many other things to consider when choosing a job beside the health benefits, and so adverse selection is very much mitigated when insurance is provided through this channel. Two, insurance companies have worked hard to overcome the problem of adverse selection: for example, suppose I offer two types of insurance, a smaller insurance plan at rates which are actuarially fair to healthy people, and a broader, more encompassing plan at rates that are only actuarially fair to unhealthy people. Risk-averse unhealthy people, despite paying a higher cost per dollar of coverage, would prefer the larger coverage. In this manner, insurance companies can eke out little areas of stability where risk aversion is just strong enough to let insurance companies offer actuarially unfair rates and partial coverage and make profit in the face of adverse selection
Really, what you may be revealing with the observation of the demographics of who is insured and who is uninsured is not the revealed hand of adverse selection (which in pure terms would do the opposite of what you are saying: insure the worst at extremely high rates and leave the rest to hang) but really is more of an artefact of the relative risk aversion and risk neutrality of these demographics.
But on reflection, I think Henry and I were wrong. I was very surprised to follow through to the table Alex Tabarrok links (from the Kaiser foundation, hardly a right-wing advocacy group) and find out that the percentage of the uninsured* who are in "fair" or "poor" health really isn't much larger than the percentage of the insured in those categories: 10.3% of the uninsured, versus 8.4% of the insured.
Don't know if this has been pointed out yet but this seems to me a logical consequence of our system which insures people essentially at random. Have a good job and you have insurance. Have a bad job or just got laid off or in business for yourself and you probably don't have insurance (or if you do you won't for very long). Of course the uninsured population isn't going to be much sicker than those with insurance. The problem is that with a market they should be! Don't people with horrible driving records, numerous DWIs have a harder time getting auto insurance than the guy who went 30 years with no accidents? Don't people who rent apartments above crack houses have a harder time getting renters insurance? The lack of more sick people among the uninsured is a sign the market is NOT working, not that it is.
If the idealized libertarian reform happened and people were buying their own insurance with maybe a gov't voucher/tax credit but no employer tax incentive, I predict the market would produce the adverse selection result where people who are known to be sick have a hard or impossible time finding coverage.