In Which I Pick a Fight with the Cato Institute
It’s not my purpose here to talk at length about The Cato Institute.
Some people say that it does important work to push towards a free-market ideology in this country to the benefit of all. Others will say that it is a non-scientific political organization controlled by a small group of GOP donors. You can read about it yourself and decide how credible you think they are. I’m just looking at one piece of research they put out; don’t going extrapolating it too far.
Michael F. Cannon is the Director of Health Policy Studies at Cato. As I mentioned, we’ve had a number of interactions on Twitter. I’ve honestly been pleased at his willingness to engage a semi-amateur. He brings up good points of which I hadn’t previously thought. I’ve learned from him and from sources he has referred me to. However, in this exchange we had a fundamental disagreement. I claimed that Medicare/Medicaid (which I’ll jointly refer to as the “Meds” provide health coverage for less than private insurers. He responded that this was “go-home-you’re-drunk false.” I shared this report from Kaiser Family Foundation which summarized the literature comparing spending per enrollee for government and private health insurance programs. He responded, cautioning that spending and costs are not the same thing. To illustrate this, he sent me a report he had written in 2009 on the very topic. Does it show what he claims – that government health insurance is not actually less expensive than private? Let’s find out.
The crux of the paper is that, especially for health insurance programs, “spending” is not the same as “total cost.” I agree. By spending, I mean (and I think he agrees) the amounts that are actually disbursed. It includes the amount paid to providers and overhead; for private programs, it should include profits as well. Beneficiaries’ out-of-pocket costs should also be added. If two systems have the same “internal” spending, but one system has an additional $1,000 per person in co-pays and deductibles, then we can agree that health spending in said system is $1,000 higher.
Health insurance systems all have costs beyond spending. For example, if the government subsidizes a specific insurance system, this must be included when comparing different systems. If these subsidies are direct, they are generally easy to quantify; as they become less direct, the result becomes more uncertain. Insurance programs may also increase costs to society as a whole; these are called externalities. Quantifying externalities can be challenging, but they should be considered and included where reasonably identifiable.
Before I look at his arguments, I think it’s important to define the context and a notion of “efficiency.” This discussion occurred in the context of my claim that we could, generally, save on health costs (without sacrificing quality) by moving people from private to public insurance. In this context, I’ll define efficiency of a health insurance system as the incremental total cost to ensure an average enrollee. The context of Cato’s paper is the proposed “public option,” which would allow people a choice to buy government versus private health insurance. I think my definition of efficiency is relevant here also. As described above, we should consider the “costs,” which will be broader than direct spending.
(Note: the article in question predated our discussion by eight years, his paper is obviously not a direct response to my views. I will note places where its content does not address my questions; this is in no way a value judgment. I have offered Mr. Cannon the chance to respond before this paper is published, but have not yet received a response.)
Let’s begin with the direct spending. Which spends less? Government or private health insurance? You can go look at the Kaiser report yourself, but here is a quick summary:
A study covering the period 1996-1999 found that total spending for the average Medicaid enrollee would be 18% higher if they were moved to private insurance. The study noted that this difference is not due to lower service use by Medicaid enrollees.
Another study, conducted from 2005 data, found private insurance spending to be 26% higher for adults and 37% higher for children. It also noted that out-of-pocket spending specifically is 500-600% higher for private coverage.
A third study used the same 2005 data, but split spending by type. Medicaid compared favorably for every type of health spending. (No web link; see Kaiser for citation).
A fourth study, looking at the period 2003-2009, found private insurance spending to be “over 25 percent higher,” after adjusting for the population in Medicaid. This 25% higher does not even include out-of-pocket spending, which would be at least three times higher for those using private insurance.
There are some more studies, but you get the point. I’m confident to state that direct spending per Medicaid enrollee is around 25% lower than for a comparable private insurance enrollee. Referring to the Cato paper, I do not see any figures or studies to dispute this conclusion. Given the topic of the paper, silence looks a lot like assent in this case. If private companies are going to end up more efficient, they make up significant ground outside of direct spending.
Those who claim government health programs are more efficient often note their lower overhead. Cannon’s piece disagrees with this approach, saying that private insurance’s higher administrative costs may actually make it more efficient. Specifically, he notes that profits (which represent 3% of premiums, according to the CBO) “could lead a government program to be less efficient.” He also notes that the Meds save administrative costs by not reviewing claims as aggressively. This is a plausible claim – if every $1 spent on administrative costs eliminated $2 of wasteful spending, then increased administrative costs could reduce total costs. As evidence, Cannon notes that Medicare reported making $10.4 billion in improper payments in 2008.
That $10 billion sounds big, but remember that it is only around 2% of the $469 billion in Medicare spending that year. In any case, it’s already included in direct spending. Adding it again would be double-counting. When all payments – proper and improper – are included, the government program is 25% cheaper. And, we are making a comparison here: Medicare does make some improper payments, but certainly private insurance companies do too. What is the comparable number for them? Cannon doesn’t say, and I’m not aware of a comparable study.
Now, it is possible that if Medicare increased its administrative costs, this would be offset by larger reductions in spending. And Medicare should of course look for ways to become even more efficient than private insurance. It is, however, a bit disconcerting to see the “small government” Cato Institute suggesting that the problem with the Meds is that they don’t have enough bureaucrats, paperwork and red tape. But this is all moot: any case for private efficiency will have to be made outside of direct spending.
The first area of indirect costs mentioned in the paper is the Meds’ use of other government resources. Cannon is correct, some of these costs should be allocated. He quotes several areas, specifically “parts of salaries for legislators, staff and others working on Medicare, building costs, marketing costs, collection of premiums and taxes, accounting including auditing and fraud issues, etc.” Being able to use the IRS to collect its “premiums” is a big benefit for the government insurers, no question. Legislative and staff salaries are a bit more remote – the number of Congresspersons would be the same if Medicare had more enrollees. But sure, apply some of this too. My understanding is that building and marketing costs are paid by Center for Medicare and Medicaid Services (CMS). Auditing and fraud studies are also often done by CMS, such as the one mentioned above studying improper payments. The costs of running CMS are already included in the overhead of these programs, i.e. we’ve already counted it. That being said, I’m sure there are various other ways CMS uses various resources here and there.
But let’s put this in perspective – total annual spending on the IRS is around $11 billion and Congress is an additional $5 billion. If you assign the entire cost of both to the government health programs, it would add around 1.5% to their total cost. And both Congress and the IRS do things other than work on government health programs. When you allocate only a portion of the costs, shared government resources would increase costs by less than 1%. Real, but barely moves the needle.
Next up is taxes. Existing Medicare and Medicaid programs are funded largely by taxes. Cannon says that “Economists estimate that it can easily cost society $1.30 to raise just $1 in tax revenue, and it may sometimes cost as much as $2.” I take this to mean that he thinks taxes have a fiscal multiplier in the range 1.3-2.0. We could get into a debate on fiscal multipliers (the CBO says that taxes have fiscal multipliers between 0 and 1.5, specific to 1-2 year cuts; a contemporaneous 2008 Moody’s report said permanent taxes have multipliers in the range of 0.29-0.48). But I don’t know why it is in a paper about the public option. In the public option, individuals would just have a choice to purchase government rather than private insurance, always using the same source of funds. There would be no increase in taxes, hence no drag.
After taxes, Cannon moves on to quality of care. Maybe the government programs are cheaper, but the quality is far worse. Cannon says, “For example, if a government program refused to pay for lifesaving medical procedures, it would incur considerable non-monetary costs.” Note the “if” – it gives the game away. If you assume this hypothetical, then yes, this would be a cost to society. However, the paper does not offer any evidence that government programs are in fact paying for fewer “lifesaving medical procedures” than private insurance. The reference cited says that Medicare, during 1965-1975, did not reduce mortality during that period. I’d be more impressed by a direct comparison of benefits, or by a study of Medicare less than 40 years old.
The paper then begins a long discussion comparing quality of outcomes for Medicaid against private insurance. He cites several compelling sources saying that private insurance has better health outcomes than Medicaid. I don’t deny these studies. He also includes a scary piece of information: 100,000 Americans die each year from medical errors. Unfortunately, he does not link this figure to the type of insurance these unfortunate people held. They could have been all privately insured. I can’t think of a reason why Medicare patients would be more prone to medical errors. Is the theory that doctors are “not careful” when somebody brings in their Medicare card? This doesn’t seem plausible.
Measuring relative outcomes in Medicaid is tricky; how do you control for the fact that Medicaid specifically covers the old, young, sick, poor and disabled? These people all have worse health statuses than the general population. If you compare the Medicaid population to the general one, your study will be highly skewed. The studies Cannon quoted regarding Medicaid outcomes do correct for this. But I can also quote studies saying that Medicaid does pretty well compared to private insurers. The literature is split, as they say. Maybe, private insurance probably has slightly better outcomes than Medicaid, but nowhere near enough to justify the additional cost. It is almost impossible to make such a comparison for Medicare, since there really isn’t any private insurance in the 65+ market.
The paper then discusses various subsidies that exist in government insurance markets. He has a list of both direct and indirect subsidies which, according to him, should be included in any accounting of the costs of the Meds. I won’t go through all of them, but here are the highlights.
First, because they are funded by tax revenue, Cannon says that the Meds “crowd out private insurance among individuals who could otherwise obtain coverage on their own.” Let’s take a look at this. Medicaid beneficiaries earn below 133% of the Federal Poverty Level, (around $32,000 for a family of four). Under the CBO’s review of the AHCA (aka “Trumpcare”), health insurance for said family would cost around $20,000 per year. Similarly, in the GOP plan, insurance for a 64-year-old would cost $19,000 on average. It doesn’t even consider premiums for those old enough for general Medicare eligibility, but they would clearly be even higher. How many families earning $32,000 per year can afford $20,000 for their health insurance? How many retired couples can afford $40,000 per year for their health insurance? People on Medicaid and Medicare, generally, have no way to “obtain coverage on their own.” Hence there are no plans with actual buyers being crowded out and the point is moot.
He then notes that the proposed new public option would have been allocated to $10 billion of start-up costs. This should certainly be allocated as a “cost” of the new government program. Currently, the government can borrow money for thirty years for around 3%. This $10 billion will therefore cost $300 million per year to finance. Let’s be conservative and say 3 million people sign up for the public option. The cost would be $100 per enrollee per year. So we should add 1-2% to spending for start-up costs.
Cannon also discusses the “risk corridors.” I won’t go through the risk corridors at length; the GOP Congress killed the ACA’s risk corridor program when it illegally decided not to pay insurers what they were contractually owed. This caused many insurers to go bankrupt, a major cause of the ACA exchange competition problems in some states. In any case, Cannon says that risk corridors “could easily become a tool for taxing private insurers to stabilize the government plan.” He also says that government programs might be “more attractive to high-cost patients or [do] a poorer job of controlling unnecessary expenditures.” Note the “could” and “might” – there is no explanation given why this would be expected. We already know that government health programs do a better job than private of controlling spending. The risk corridor spending is therefore more likely to go the other way: payments to the private companies. And hey, if you really don’t like it, we can just exclude the government plan in the risk corridor program.
After this he moves on to indirect subsidies. First up is the new program’s ability to utilize the negotiating leverage of the existing programs, as well as some other advantages inherent to government. I agree completely – this is the point of the public option in the first place! Moving people into systems that have more advantages should be the goal! Any attempt to control health care spending will involve more price controls. This logic seems to say that government programs should have a cost applied because they are better organizationally. This doesn’t make sense.
His next point may be more valid. More patients on Medicare means more will pay the lower prices that Medicare has negotiated. He says that these decreases “will” increase the costs of private insurers. I believe he means spending of private insurers – he made the distinction, I to be careful to follow it. I’m unfortunately not able to locate some of his sources, but the CBO includes a discussion in a 2008 paper (pg. 113) of so-called “cost shifting.” To summarize: the CBO has no real idea of the how much cost shifting should be expected. It also notes that, to the extent that the people moving to government-sponsored insurance reduces uncompensated care, the effect could go the other way.
Cannon also considers the possibility of cost shifting in prescription drugs, where government also pays a lower rate than private. Here, he is on even less firm ground. For most providers (i.e. doctors, hospitals), the market is very local; your primary care physician is probably located within ten miles of your home. Therefore, said provider has two sources of payments: government and local private insurance. It is at least plausible that if the low was reduced that the latter would increase.
But prescription drugs are a global market. Pfizer sells drugs in basically every country in the world. Currently, the United States pays far higher rates than other developed countries for prescription drugs. If a public option caused cost shifting in drug prices, a lot of the shifting would be to patients in other countries. You all know how much I love Canada, but it is a wealthy country, there is no reason why their drugs should cost a fraction of what they do in the U.S. Giving the government the ability to negotiate drug prices will help lower this disparity. This is a good thing.
Cannon also notes that the public option would carry a Federal Government guarantee should it lose money. Completely true. He thinks a cost should be applied due to this. He notes that similar “implicit guarantee[s] saved Fannie Mae and Freddie Mac an estimated $6 billion per year.” But, there is a nuance. Just as spending and costs aren’t the same thing, the benefit of this guarantee to Fannie and Freddie is not the same thing as the cost to the government. We know this because the Federal Government had to make good on these guarantees. I don’t love bailouts any more than the next guy, but taxpayers have done well. We’ve already gotten back our entire investment back – along with an additional $78 billion. We can expect to continue receiving $10 billion per year in dividends; or the government could sell its stakes for a big lump sum. Hard to see how this “cost” taxpayers.
He is correct when noting that moving people to government health insurance foregoes the tax revenue on private insurer profits. Now, I could go back to his point on tax drag on the economy, but we agreed to ignore this factor. So taxes foregone should be included in the government program’s costs; they represent around 1% of premiums. Add that to the tab.
Cannon also notes that the government could disadvantage private insurance companies in other ways. I was actually unaware that if you turn down Medicare Part A (i.e., the free part), you also lose your Social Security. I really don’t understand why somebody would turn down Medicare Part A – again, it’s free. But whatever – no reason to think this would apply to any public option. The fact that “Congress might, one day, pass a law that advantages the government program” is not an advantage to the government program.
There are a few other, smaller points, but this is probably enough. What to make of all of it?
First, going back, Cannon did not challenge that the spending of government insurance programs is significantly lower than private. Until I see something to the contrary, I will say declaratively: direct spending on comparable health benefits via government programs is 20-35% cheaper than via private insurers.
Michael Cannon has certainly identified some subsides that should rightfully be included as incremental costs on government programs. But, the “other costs” identified are less than 5% of premiums, probably closer to 2 or 3%. The tax drag might be a real thing, and we can debate the effect of taxes on aggregate demand if we want to. But for the “buy-in” program being discussed, it isn’t relevant.
More importantly, Cannon did not consider “other costs” of private insurers. Private insurers create their own negative externalities; if we are considering “total costs”, then we need to include them. In the pre-ACA era (recall this paper is from 2008), most states permitted insurers to reject applicants for medical reasons. These people were either pushed into government programs or used significant uncompensated care. Either way, this was an external cost that should be included “against” the private insurers. All those people working in medical billing? Most of them deal with the private insurers (as you can guess from the volume of paperwork you receive from them for each doctor’s visit). Providers spend huge amounts on billing. Some of this is not being borne by private insurers. The NIH theorizes that $350 billion per year could be saved by simplifying medical billing (split between insurer overhead and provider-side).
As I said at the beginning, I’ve appreciated discussions with Mr. Cannon, including this one. I’m also more than happy to look at further data to refine my view of total costs of public versus private health insurance. But, the basis that direct spending by government programs is much lower is not challenged here. And I don’t see anything showing that “other costs” significantly moves the needle towards our private insurance system.