Some progressives claim that they have an easy solution, one that proceeds from their belief that more government it is often the answer: Medicaid or Medicare for all. What is the easy solution of classical liberals? There are two sets of reforms: one on the demand side and one on the supply side. On the demand side are a surprisingly simple combination of out-of-pocket payments, a new type of event-based health insurance, traditional care-based health insurance for some, and, perhaps, judicious subsidies. A later article will deal with reforms on the supply side.
This is from Charles L. Hooper and David R. Henderson, “A Cure for Our Health Care Ills,” Econlib Featured Article, June 4, 2018.
One of my favorite lines is the last line in this paragraph, written by co-author Charley:
A fascinating 2008 experiment in Oregon punctured this double-barreled myth. Oregon’s government conducted a lottery to enroll a limited number of low-income adults in Medicaid. The results? According to an article in the New England Journal of Medicine, those in the Medicaid group spent about 35 percent per person more than those in the control group. But, although the increased spending did lead to some improvement in mental health, it “generated no significant improvements in measured physical health outcomes.” If Medicaid were a new drug, the Food and Drug Administration would reject it.
Also, although we will deal with supply-side reforms in a future article, we note the likely effect on the supply side of our proposed demand-side reforms:
Although, as noted at the outset, we have purposely avoided discussing supply-side reforms of health care, we should note that the demand side reforms we offer would have a salutary effect on the supply side. Because the reforms would cause consumers to be much more cost-conscious, their awareness of costs would drive positive changes in supply. One of the best illustrations of this is the evolution in eye surgery over the last few decades–a corner of the health care market that is largely free of government and commercial third-party interference. The original surgical procedure, called radial keratotomy (RK), relied on the skill of the surgeon to make large cuts in the cornea, required a six-week recovery, and originally cost about $8,000 ($18,600 in 2018 dollars). Today, LASIK has largely replaced RK for those who are eligible. According to George Mason University economist Alex Tabarrok, in 1998 the average price of LASIK laser eye surgery was approximately $4,400. Just six years later, the price had fallen to $2,700, a 38-percent reduction. Adjusted for inflation, the price had fallen by over half, a result we are used to seeing in computers but rarely in medical procedures.
Read the whole thing.
READER COMMENTS
JFA
Jun 4 2018 at 2:07pm
These aren’t really reforms (aside from the subsidies) so much as features you would like to see. I wish you would discuss why these things don’t already exist. Some current cheap insurance policies are about as skimpy as catastrophic care and the evidence on coinsurance and copayments is mixed but those two things seem to reduce spending for those who are budget constrained. Why is there not more event-based pricing already? You have hip-replacement and a few other procedures under these schemes. Why not more already?
I don’t think that the differential tax treatment on employer contributions to health insurance explains why your preferred features aren’t more wide-spread. I imagine it has to do with supply-side more than demand-side.
Airman Spry Shark
Jun 4 2018 at 2:13pm
From the “Event-Based Health Insurance” section of the Econlib article:
This makes sense to me & I like it. I wonder if you would endorse this corollary:
Michael Byrnes
Jun 5 2018 at 7:09am
This, to me, is highly misleading. There’s a lot of confirmation bias in the response of Medicaid critics to these results.
Simply put, this study was not adequately powered (ie, did not enroll enough patients) to provide a definitive answer (ie, in a statistical sense, as would be the case in a drug trial) to the measured physical health outcomes studied in the experiment.
This is not surprising, in the sense that the health economists who ran this study were taking advantage of an opportunity (Oregon was going to conduct a lottery to assign Medicaid coverage to a subset of eligible patients), not designing a study from scratch and ensuring that it was adequately powered to detect clinically significant changes on health.
Another limitation of this study was that it compared “lottery winners” to “lottery losers”, not “people covered by Medicaid” to “people not covered by Medicaid”. Not all winners accepted coverage, some losers later became eligible and received coverage.
One great source of information on this study is this great EconTalk interview with health economist Austin Frakt. (See also, this EconTalk with Jim Manzi, who has a differing view.)
JFA
Jun 5 2018 at 8:47am
I think an important point about the Medicaid study is that while it confirms many other studies on the lack of relationship between health and insurance, one of the strongest findings is that those who received Medicaid had much better financial health than those who did not. This should be expected since health insurance is a financial product.
robc
Jun 5 2018 at 10:43am
JFA,
I wonder if drivers with collision insurance have better financial health than drivers with just liability insurance, all other things being equal?
Mark Z
Jun 5 2018 at 12:13pm
Michael Byrnes,
The ‘lack of power’ criticism could likely be applied just the same to the majority of empirical studies, including most clinical trials, that find negative results: they don’t ‘definitively’ prove the null hypothesis correct. That doesn’t mean they don’t provide evidence in favor of the null hypothesis. There’s not some magic number at which you have ‘enough’ statistical power. A given sample size provides a measure of confidence that you will accurately reject the null hypothesis of the effect size is bigger than some value and/or variance is smaller than some value. The bigger n, the larger the variance and the smaller the effect size can be while still being detectable. The Oregon study therefore can reduce our estimate for the impact of Medicaid. Maybe there is a nonzero effect, but it simply
isn’t big enough to be detectable reliably with sample size n. That’s not a negligible result.
JFA
Jun 5 2018 at 12:32pm
robc,
On a purely observational analysis, you would probably see the reverse relationship, most likely due to income effects and wealthier people being more able to afford collision insurance.
But I imagine if you had a large enough sample (given the rarity of car accidents, especially those with repairs that cost more than the deductible) stratified by income and you randomized the type of car insurance within those strata, you probably would see differences in average financial health (especially at the low end of the income distribution where people are more budget constrained).
Alan Goldhammer
Jun 5 2018 at 2:12pm
@David Henderson – I did read the whole thing and I’m sorry to say that the paper you and Mr. Hooper wrote fails on several counts. Others have already pointed to the lack of statistical power of the Oregon Medicaid ‘experiment.’ Contrast this against the Framingham heart study which was among the first long term studies to look at health outcomes. It’s now into its third generation of population study. Two years is too short a period of time to measure health outcomes other than looking at whether some lab values change or don’t. The Oregon study did establish that diabetes could be diagnosed and appropriately treated which is no small matter.
The comment in the paper about the ACA is misleading and ultimately wrong. The ACA was never designed to solve the multitude of problems that affect our healthcare system (one is better off reading TR Reid’s wonderful book “The Health of America” that surveys not only the US system but a number of foreign systems that Reid and his family accessed when he was a Washington Post foreign correspondent. Those countries provide equivalent care at a much lower price than in the US.). You write, “…so health plans are designed to dissuade these types of customers through restricted access to specialists and expensive drugs…” This is NOT unique to ACA plans as most all health plans have a variety of restrictions on specialist visits, ER usage, drug formularies (that include mandatory use of a low cost therapy before the expensive, mainly newer biologic therapy, will be approved by the insurer), etc.
You write, “…. insurance premiums for many people have approximately doubled in just the four years that Obamacare has been in force…” and this means what? The premiums of my Medigap insurance that is provided by my former employer has doubled as well! I’m sure that anyone out there who is on a health plan from an employer has seen the premiums go up along with co-payments and deductibles. This is not unique to the ACA.
Reference 5 in the paper regarding the myth of preventative care is just a short op-ed and not original research. The op-ed has several flaws in terms of overlooking a variety of studies regarding prevention and outcomes. Elmendorf’s comment regarding increased cost has nothing to do with health outcomes which are measured on an entirely different scale. I read Elmendorg’s article when he submitted it to Congress and the conclusion is far more nuanced than you make it out to be.
In your Myth #2 you error in drawing conclusions about the impact of the ACA on increased access to health insurance. Again, as with the Oregon Medicaid experiment you can not draw any such conclusion unless there is longitudinal data that looks at both the enrollees in the ACA and their health outcomes. Neither sufficient time has passed nor is there the kind of patient specific data that would allow such an inference. Some individuals will be in the ACA for a short period of time depending on their employment status (many individual contractors use the ACA for insurance as they do not work for an employer directly and don’t qualify for employer paid insurance. This was the case with my two daughters who were doing contract work for a couple of years but are now getting employer insurance. You have to be able to discern how many are in this position.
The section on ‘Out of Pocket Payments’ is disjointed and makes little sense. You state earlier that insurance is actuarial in nature and it also has a cost control component. What better way to control this than by formularies and prior-authorization. Why is this wrong? I guess the standard Libertarian answer is that it curtails freedom of choice; so does the many state requirements for car insurance and mortgage companies for property casualty insurance.
In the ‘Event Based Health Insurance’ section you reference how automobile insurance works in the case of a fender bender. As one who went through this recently with my car, I can tell you that this is not how my insurer does business. You file a claim and they provide you with the repair shop appointment from a list of shops that they have negotiated a pay structure with (hmmmm, this sounds kind of like health insurance).
You point to the lowering of the cost of LASIK surgery as a good example of how things ought to work. Other than improved access to cheaper generic drugs following patent expiry is there any other aspect of the health care system that has come down in price?
I would conclude by noting that the esteemed vice-chairman of Berkshire-Hathaway, Charley Munger, who is certainly no raging liberal stated during press events at the recent annual meeting that we likely destined to move to some form of national health insurance.
Sorry to be so frank but the paper is not convincing.
Michael Byrnes
Jun 5 2018 at 5:17pm
Mark Z wrote:
No. This isn’t what is meant by ‘lack of power’. Sometimes the null hypothesis is correct. Statistical power refers to the probability that a given experiment will disprove the null hypothesis if, in fact, the null hypothesis is false.
Calculation of statistical power depends on 3 parameters: the average effect size, the variance around that average, and the number of participants in the trial. The way this works in practice in most clinical trials is that the trial designers have estimates of the effect size and the variance based on prior work, and they then figure out how many patients are necessary to give them a high (usually 85% to 90%) chance of success if the null is false. In the wolrd of drug-company funded clinical research, this type of approach is basically statnard operating procdure, with some exceptions. (The big exception being orphan diseases, where recruiting enough patients is a major challenge).
The Oregon Medicaid Study had none of this.
The sample size in this study was was limited by Oregon’s available budget for this Medicaid expansion – only so many people could be covered based on the funds allotted. If this number led to an underpowered study (meaning a study with a low probability of detecting a treatment benefit if there was one), so be it – that is all Oregon was going to fund.
The study was also limited, as I recall, by a relatively high rate of lottery “winners” who ended up not receiving coverage, either due to changes in eligibility status, acquisition of other coverage, or because they elected not to enroll (in his EconTalk interview, Manzi makes a big deal of this latter point, arguing that if people who can get Medicaid coverage choose not to do so, that says something about Medicaid coverage).
But, anyway, what this means in practice is that the researchers can’t directly do the comparison of interest: people randomly assigned to receive Medicaid vs those randomly assigned not to receive Medicaid. The actual random assignment was lottery winners (some but not all of whom enrolled in Medicaid) vs lottery losers (who didn’t get to enroll in Medicaid, except – as I’ll get to below – some actually did get into Medicaid). There are statistical techniques that can be used to address these things, but having to resort to such things mean a lower ratio of signal to noise. No pharmaceutical company would run a large trial where a huge chink of the treatment group never got the drug!
The other “treatment” wrinkle in the Medicaid study was that Oregon had a set of Medicaid eligibility criteria that was in effect long before this experiment. The Medicaid lottery that led to the experiment was a sort of small scale Medicaid expansion. People who met one set of criteria were simply eligible, as they had been; those who met a wider set of criteria were eligible to participate in the lottery. But what this meant is that some of the “losers” who did not receive coverage in this expansion of Medicaid ended up becoming eligible for Medicaid by standard criteria and receiving it anyway – adding more noise.
Despite all of that, the study did detect some significant benefits in the Medicaid group: improved mental health and improved financial security. However, there are statistical reasons to not necessarily be overly excited about thise results, too. When underpowered studies do find statistically significant results, the measured effect sizes are often overestimates of the real-world effect.
Charley Hooper
Jun 6 2018 at 1:28pm
Michael Byrnes,
I listened to both of the informative EconTalk interviews with Russ Roberts. I can see how one would argue that the study was underpowered, but I can also see how one would argue that it was adequately powered and the benefits just weren’t there.
My question is one of presumption.
If a drug company runs an underpowered study that shows no benefits to a new drug, the FDA will not approve that drug. The presumption is one of guilt until proven innocence, and the drug will be rejected.
If the government spends money on health insurance systems that, as far as I know, no study has ever justified in terms of cost per increase in measured health outcomes, why shouldn’t an analogous government agency “reject” those health insurance systems, at least until, and if, some positive studies are finally run and the results clearly shown?
Why are drugs guilty until proven innocent but Medicaid is innocent until proven guilty?
Charley Hooper
Jun 6 2018 at 2:15pm
Alan Goldhammer,
You have a number of criticisms and remain unconvinced. Perhaps your insurance company told you which repair shop to use. Perhaps the Oregon Medicaid study was underwhelming.
But you never addressed our central thesis, that health insurance has substantial advantages if it is focused on the health event instead of the health repair.
Alan Goldhammer
Jun 6 2018 at 4:10pm
@Charley Hooper – there is a simple solution that is in keeping with what you propose and that is Medicare for All!!! Medicare of which I am a recipient now focuses on payments for health events under Part B (I’ll leave aside the very real problems with Part A that have been known for a long time and the recent statement from the actuaries on Part A should surprise nobody).
Medicare Part B has defined payments for health events and the lowest administrative costs around. I think you and David ended up writing a long paper whose outcome ought to have surprised you.
As to my own leanings, a universal payer system for the US is the only logical outcome. When it will happen is of course uncertain. I’m also ambivalent whether it is a Medicare model or a regulated insurance model. Anything else is just tinkering around the margins (this includes the ACA which was just a kludge to addressed those who were not eligible for employer sponsored insurance).
You also write in responding to Michael Byrnes,
The exact same quote can be applied with just some minor word changes to employer sponsored insurance plans. Again, I would urge you to read the TR Reid book that I refer to in my original post where these issues are covered concisely.
WalterB
Jun 7 2018 at 12:03am
Putting more power with consumers and less with government is an excellent plan. However, politicians and bureaucrats don’t want that to happen, and too few consumers understand why it’s important. I will post a solution here when one comes to me!
robc
Jun 7 2018 at 10:28am
Alan,
I have had insurance companies tell me which repair shop I needed to get my estimate from, but they have never paid for the repair directly or told me I had to use that shop.
Their shop gives me the estimate, then a check is cut, then I can use it wherever I want (or not at all).
I did the last after my car took a few thousand dollars worth of hail damage. I cashed the check and drove a hail damaged car for a few more years.
When I sold it, there was a deduction vs book value due to the hail damage, but it was a net win for me, even after deductible.
Comments are closed.