How high are Medicare overhead costs?

Yuval Levin makes a few points of relevance:

…many of Medicare’s most significant administrative costs are just covered by other federal agencies, and so don’t appear on Medicare’s particular budget, but are still huge costs of the program. The IRS collects the taxes that fund the program; Social Security collects many of the premiums paid by beneficiaries; HHS pays for a great deal of what you would think of as basic overhead, but doesn’t put it on the Medicare program’s budget. Obviously private insurers have to pay for such things themselves. Medicare’s administration is also exempt from taxes, while insurers pay an excise tax on premiums (which is counted as overhead). And private insurers also spend a great deal of money fighting fraud, while Medicare doesn’t. That might reduce the program’s administrative costs, but it greatly increases its overall costs. Some administrative costs save money, after all: The GAO has estimated that a $1 investment in pre-payment review of claims, for instance, would save $21 in improper Medicare payments.

My original source is this Reihan Salam post.  I also would add the deadweight cost of taxation.  Arguably that does not count as a cost of “overhead,” but very often it runs 20% or more and still it is a cost.


So much absurdity, but let's just take this one -
'The IRS collects the taxes that fund the program'
And the banks and collection agencies (and let us not forget employers who transfer the money directly to health insurance companies) responsible for transferring payments to insurance companies/doctors/hospitals aren't included in 'overhead' either.

At some point, I expected the author to point out that the roads which are used by Medicare providers are also taxpayer funded and thus not included in Medicare's 'overhead', but luickily, it wasn't even necessary to read beyond the blogpost - ' I also would add the deadweight cost of taxation.'

Also, what is the marginal collection cost of collecting an extra percentage point of any given personal tax?

Come on people, the optimal amount to spend on tax enforcement, collection infrastructure, etc. surely depends heavily on revenue and is also easy enough to take any private insurance expenditure and call it MC = 0 for a given infrastructure...

'surely depends heavily on revenue and spending'
Or merely depends on having the infrastructure already in place. Take this as a baseline -
'The current FICA rate is 7.65% (6.2% for social security tax and 1.45% for medicare tax) for employers and 5.65% (4.2% for social security tax and 1.45% for medicare tax) for employees.' ( ) Changing the rates .1% in either direction is utterly meaningless beyond the fairly trivial (and already expected) maintenance of accurate percentages in the payroll software (note that the link is to a company that provides exactly these services, though with a bit of specialization).

It is not completely wrong to note that Medicare collects its revenue through a mechanism which is required by the government. It is just utterly uninformed to think that Medicare is funded through a process which is any different than that through which a typical company pays its employee health insurance premiums - it is all just handled by payroll software for any company of normal size. And it is simply wrong to say that when a company uses a service like that provided by a company like PCS to pay disability insurance, workmen's comp, IRA/401K plans, or even something like employee health insurance premiums, this is in any way, shape, or form different than FICA taxes being handled in exactly the same way.

Obviously, there are individual cases where details are different (such as the fact that Social Security payments into the system are capped) - but really, anyone currently paying into Medicare (a group that just happens to include every single tenured professor at GMU, I believe) is also paying into Social Security - and just as transparently and easily as they pay into TIAA-CREF (in the past, at least). And yet, one would not argue that TIAA-CREF has additional overhead compared to Social Security merely because the government requires Social Security to be collected as a tax, while TIAA-CREF is part of a non-tax system. The same software is used to handle the transactions, both those mandated by the government, and those merely mandated by a contract of employment.

And from this perspective, the deadweight cost is merely that of having paid employees, and needing to handle various financial transactions on a regular basis, generally using software to make the task straightforward and efficient. However, as seen here at marginal revolution, it is possible to extol programs where work is performed for an 'employer' without any of the costs normally associated with actual employment - such as transferring funds from the employer to various accounts to cover the costs associated with employment.


The IRS and the government rely on the banking system too. Insurers don't collect via taxes...oh wait, they do now!!!

' Insurers don’t collect via taxes'
Of course they do - if Prof. Cowen is insured through Kaiser-Permanente (a standard option at GMU in the past), taxes pay essentially all of his health insurance premiums. This entire discussion is just spinning off into absurdity. And somehow, I doubt that Prof. Cowen would be willing have his wages (or health insurance fees) changed to reflect the 20% DWL of how his salary at GMU is collected before he is paid.

Unless there is a still unknown private market for people with a legal claim to lifetime employment (along with health care as a part of his contract of employment) which pays at least the same amount as he currently earns as a professor. Anyone know where that market can be found?


You are just incoherent. Maybe try again with a focus on clarity.

I think the larger point is that Medicare just skims off its statutory tax rates where private insurers have to set rates based on the covered group, its claims costs, expenses, etc. Private insurers also have to deal with a competitive bidding process as often as annually. The private insurers have a much more difficult problem to solve.

Levin swept away Krugman's claims.

The payroll in small companies is complicated by he existence of the SS and medicare taxes.

True Tyler, MC != 0 for tax collection for Medicare. But fact is economies of scale exist. Medicare taxes are a small portion of payroll and income taxes which the infrastructure already has to exist to collect. You can argue that increasing taxes for any reason also increases collection costs because you'll get more people who evade the taxes or simply won't pay them, but your still talking about pretty marginal costs for collection.

Likewise opportunity cost should also be factored in here. What are the alternatives? Consider a state-by-state system like Medicaid where each state decides to fund it by whatever measure it pleases (sales taxes, payroll, etc.). Since that would require multiple, small, and different collection systems the total cost would likely be higher. Or consider a Republican type tax credit system where individuals get credits and deductions to make it easy to buy insurance directly. This will require tax rates to be high enough to make it appealing to buy insurance to lower them....which means the IRS tax infrastructure has to be there as well as infrastructure for the IRS to verify that people claiming the credits and deductions are being honest about it.

Finally to make an apples to apples comparison, you should consider the infrastructure private insurance companies tap from the government. For example, consider health insurance provided to workers by corporations. This provides insurance companies with dramatically lower selling and collection costs. One sales rep can negotiate with the company's benefits department and the insurance company can send one bill every month to the company rather than sending out thousands of individual bills and worrying about collecting from individuals. To the degree this happens because the company must pay taxes on its payroll and receives credit for providing health coverage, the private insurance company is effectively tapping the IRS/Social Security/Payroll tax infrastructure just as Medicare is.

I wish Yuval had just stricken the tax collection bit, and Tyler didn't subsequently defend it. It was unessential. Yes both government and private insurers take advantage of infrastructure provided by the other. The fact remains that HHS does keep medicare overhead off of medicare's books. It's that simple, so the popular accounting that is used to bludgeon private insurers is wrong.

The article was really strong in pointing out that the "negotiating power" supposed is just price controls with a new tagline. The IRS bit let everyone miss those points in favor of having the millionth "you didn't build that" rant exchange today.

The fact remains that HHS does keep medicare overhead off of medicare’s books. It’s that simple, so the popular accounting that is used to bludgeon private insurers is wrong.

It's perfectly fair to say that we should calculate 'off the books' estimates of Medicare's overhead. But then that should be done for private insurers as well. At that point you start descending into hair splitting debates like the cost of the roads that people who work at the insurance company (and Medicare)use to drive to work or the defense department that keeps terrorists from crashing planes into the HMO's skyscrapper. As has been pointed out, if you really try to do a fair estimate of Medicare's fair share of this 'hidden overhead' it probably still comes out looking good compared to private insurance.

The 'books' are imperfect but pretty objective. The IRS does not issue a special bill to HMO's for the service of maintaining and supplying everyone social security numbers nor does it issue a bill to Medicare.

"Probably"? Well, that's good enough for government work, I guess...

Indeed, but then private business won't even try to make such a calculation either. If income exceeds expenses by $50M, then the business is profitable to the tune of $50M period. One doesn't say 'wait a minute, we referred twenty cases to the prosecutor for fraud last year, that cost the gov't $10M to work on....we would have to spend $30M more in data verification if we didn't use people's social security numbers provided by the gov't...we'd have to spend $500M more on an army of sales agents and many more billing people if we sold policies directly to individuals rather than to corporate benefit departments due to the tax code being set up as it is so in reality Mr. CEO we are costing society more than we are bringing in therefore we should shut down immediately despite our accounting profits!'

While we are at it, just about every bank and private insurance company I'm aware of collects your social security number. Since the only reason we have social security numbers is because the gov't pays the IRS and SSA to keep track of assigning and maintaining numbers for everyone, we can say the gov't is picking up some of the overhead for all insurance!

OK, then quantify it. "How high are Medicare overhead costs?" Neither you nor the article
you link to answers that question. It seems extremely unlikely to me, for example, that the marginal cost to the IRS of collecting medicare taxes is material, but who knows - maybe not.
Let's see the numbers.

DWL of taxation is often estimated at 20%, right there the entire supposed cost advantage of government goes away. And indeed that number is offered in my post.

NOONE has quantified it.

But the other side claimed that no evidence equated to zero numbers in their slipshod rush to grease job the voters.

I am always amazed that there are these huge, complex and compelling issues, with very little serious (dedicated) coverage from economists. There's Amy Finkelstein at MIT. Who else is there?

Isn't using a generic DWL estimate oversimplifying the matter here? After all, when we imagine the negative impacts of taxation, it must depend on what services the government is providing with those taxes.

It seems to me there's a pretty strong argument that there are DWL when the government taxes us to build roads, because in the absence of government action private citizens would spend significantly less money building and maintaining that public good. So in that case it makes sense to estimate the DWL of taxation against the social benefit of a well-maintained system of roads and decide whether it's worth it.

In the case of medicare, private citizens will not likely purchase significantly more or less healthcare in the absence of government provision. The effects of taxation here are primarily redistributive, so I think the argument for a large DWL is much weaker. And when you add in evidence that medicare is controlling costs better than the private sector would, it seems to me that this arrangement is extremely efficient. You could even argue that market forces have driven governments around the world to centralize healthcare markets. Perhaps free-market economists place too little value on the clear price signals that tell us single-payer healthcare is more efficient. A case of confirmation bias?

"In the case of medicare, private citizens will not likely purchase significantly more or less healthcare in the absence of government provision."

So in your opinion making something "free" for consumers will not at all affect how much they consume? Seriously?

Yes, seriously, price won't affect consumption much in the case of healthcare. In the absence of government-provided care, some people would buy insurance anyway. This group of people will consume the same services as if the government provided insurance.

There would be a more noticeable impact on the people who choose not to buy insurance, but it still wouldn't be significant. This is easy to see by comparing the data between the insured and uninsured populations today. We spend most of our healthcare dollars on a very small slice of the population, the people with the worst health issues. In the absence of medicare, medicaid, or private insurance, those people show up in emergency rooms regardless. Either they pay their own costs, or they freeload off taxpayers or the insured population. Every serious study of the many failures of the healthcare market shows these kinds of problems.

Assuming your primary care physician will prescribe a colonoscopy on demand once a month, how many months in a row will you ask him to order a colonoscopy??

Of course, neither is covered for "free" under Medicare Part A or Part B, and supplement is private insurance run by the same insurers you pay premiums to if you are an individual, or that provides your employer group benefit.


That is absurd. Surely you are aware of the drastic differences in utilization between insured patients in various parts of the country without any differences in outcome? Or the differences in utilization between various countries? Surely you know that emergency room non-payers account for a very small percentage of total costs? You are saying health care is the only market in existence where price doesn't matter and no one will shop around or get a second opinion or investigate their options to reduce costs?

My understanding is that the uninsured have a lower rate of utilization on routine services for less serious conditions, and a higher rate of utilization for services relating to expensive and preventable conditions. The net effect of this is modestly lower healthcare spending for uninsured patients, at the cost of worsened health outcomes and decreased productivity.

Just to be clear, I don't claim that price has no effect on consumer decisions -- just that the various problems with the healthcare market make price signals much less effective than in the market for, say, gasoline. Healthcare consumers in general have very little knowledge about the products they are purchasing and can't evaluate the effectiveness of treatments they receive relative to their costs. They mainly rely on doctors for advice, who often do not have the same incentives as the consumer to weigh cost vs benefit. It is also difficult for consumers to place an economic value on the potential for their own death or disability, which is of course necessary for a well-functioning market. The healthcare market is fairly unique in its disfunction, and I think a lot of pragmatic conservatives admit this and concede that there is a much stronger case for regulation of healthcare than for regulation of most other markets.

@Jon Rodney a good place to start:

After all, when we imagine the negative impacts of taxation, it must depend on what services the government is providing with those taxes. ... In the case of medicare, private citizens will not likely purchase significantly more or less healthcare in the absence of government provision.

Your logic does not follow.
Say that you are considering hiring a house cleaner, perhaps a student who wants to work part-time and rely less on his parents, call the would-be cleaner A. Without taxes, you pay $x and A gets $x. With taxes, you pay $X+T*(b) and A gets $X-T*(1-b), where b measures how much of the tax incidence falls on you. With the tax in place, the gap between the value you get from A's labour and the effort A must put in is too large, so you don't hire A and instead do your own cleaning. That's the DWL.
I don't see how publicly funded healthcare payments would necessarily affect that DWL. If anything, publicly-funded healthcare should lower A's reservation wage, as A's healthcare would be provided for regardless of how many hours A works (you mention Medicare specifically, which I presume A is not eligible for.) I suppose if buying healthcare for yourself was more expensive than buying it publicly-funded, you'd have less money left over to hire housecleaners, but on the other hand what extra you spend on healthcare would be paying people working in the healthcare sector, or showing up in shareholders' dividends. The point of DWL is that it's value that's not captured by anyone, it's not just a transfer.

'DWL of taxation is often estimated at 20%'
Which needs to be explained in a way that distinguishes between payroll software paying funds into a TIAA-CREF account and Social Security in a way that convincingly shows where that 20% loss comes from. And it isn't from the collection mechanism, since the same software is used (essentially always) in both cases. The example of the IRS being a free billing service for Medicare is just silly - and shows an utter lack of awareness of how actual Medicare taxes are collected in comparison to how other payroll items, such as health insurance premiums, IRAs, workmen's comp, etc. are handled.

DWL comes from that taxes drive a wedge between what a good or service costs someone, and the value they get from it.

It has nothing to do with payroll software in particular (though taxes on payroll software would create a DWL).

There is only a DWL if people would spend those taxes on something other than medical care. The same DWL would apply to private insurance as people have no say over its provision or would not apply to either since this is something they would buy anyway. There can be no DWL without consideration of a LWG.

"There is only a DWL if people would spend those taxes on something other than medical care."

This is simply false. Go grab a 101 textbook.

How do you figure that 20 percent is accurate?

Also, in a more general sense, why hasn't there been any attempt to calculate the overall cost of administration?

I wonder what the deadweight loss of taxation is in comparison to the dead weight loss associated with market concentration in healthcare. I imaging Dr. Cowen would have to admit the DWL from concentration in healthcare is bigger than zero, not sure if he would admit it is a large part of 20%?

An attempt to quantify these advantages to create a fair comparison would be helpful, though beyond the scope of any blogger on Labor Day.
Nonetheless, if these structural advantages exist, we'd be Damn fools not to take advantage of them.

Here is a report from an n=2 (my wife and myself) who have just gone onto Medicare this year. As those of us who are in Medicare know, premiums are means tested based on ones AGI from the most recent full year tax return. For retirees going into the plan this year, the 2010 filing is used. Now one would think that it would be easy enough for two government Agencies to share the appropriate information, particularly since they are all stored electronically. LOL!!! My wife was first to go into Medicare and received a premium letter one month prior to officially entering the program that outlined the AGI adjusted premiums. Now the base amount that everyone has to pay is $99 for Part B; Part A is free and we did not need the drug benefit (Part D) as we stay on my former employer's plan as retirees. Because we were aggregate high earners (though not in the 1% club) in 2010, our means tested surcharge is $199.80 on top of the $99. What happened to both of us is that we got an initial bill for the surcharge only. About 60 days later my wife received the bill for the base amount that was not included in the original invoice. I'm still waiting for my second invoice as I just went into Medicare in August. Now maybe this inefficiency is really a very small amount and only applies to those who are not yet taking Social Security (if one goes on SS right away on turning 65, Medicare premiums are taken directly out of the SS account) but it has to add up to something unnecessary.

I cannot comment on any reimbursement experience as neither of us has visited a physician since going on Medicare. I suspect what is being compared here is the relative efficiency of claims adjudication between Medicare and private insurance plans. I can tell you that from our historical experience with three different health insurers over our working careers, there have been multiple problems that required letters/calls/emails to our carriers trying (and mostly succeeding) to get reimbursed from covered medical procedures. Because I'm anally compulsive about tracking such things, we did have 7 instances where the insurer reported never receiving the claim and we suspected that some of these were 'intentionally' lost.

My experience dealing with insurers has quite bad as well. I am not sure this is the reason their costs are high or low but they put the onus on the provider and patient to do their work for them. This does likely lower fraud but also represents a huge shift of administrative costs on to providers and patients that can only be shifted back to them in the form of premiums in the former case, so insurer administrative costs are understated as well.

Another expense of Medicare that is reflected on another agency's budget is the legal representation provided by the Department of Justice.

If a private insurance company uncovers criminal fraud, it will go to to the prosecutor. What portion of the costs of Federal and state courts as well as prisons and jails be allocated to insurance companies as 'hidden overhead'?

The thing you have missed, and missed above, too (as well as some of the other commenters) is who is paying for the government that both Medicare and private insurance companies utilize. Things are not identical in this regard. This is why it really isn't a matter of hair splitting when accounting for intragovernmental services provided to Medicare. Medicare isn't being billed for them, but private insurance companies are being billed for them, so they aren't "off the book".

I don't think it's a meaningful miss. If a private insurance company uncovers fraud and alerts police and prosecutors, it does not get a bill for the costs of the arrest, prosecution and sentencing of the criminal. Yes the private insurance company pays for this with taxes but it's taxes are not based on how much 'service' it uses. The business which never has to call in the police gets no reduction in its taxes.

I suppose if you want to be predantic, you can pretend Medicare is some type of company and make it pay 'taxes' to the IRS. That would increase its costs but then it would also create an equal benefit to the IRS's books.


You and others were trying to claim that the intragovernmental services provided to Medicare are a matter of hair splitting since private insurers also use governmental services. I am just pointing out that private insurers pay for their services through the taxes they pay, even if, as you say, they don't get specific bills for it. In addition, their taxes pay for part of Medicare's intragovernmental services, too. It is deeply dishonest to compare Medicare's "costs" with private insurance costs without accounting for these hidden services.

As was pointed out the IRS's budget is something like $12.5B a year. If you want to go ahead and assign a piece of that to Medicare. You're not going to alter the numbers very much. Now if you want to keep going take the Department of Justice and assign a portion of its budget to Medicare for the fraud cases it refers to them each year. Then take the Department of Corrections and assign them a portion of the costs of locking up convicted Medicare cheats. Do all this and I don't think you're going to inflate Medicare's administrative costs so dramatically as to explain why they seem less than private insurance. And to really be fair you must do this to private insurance too. Whenever Blue Cross unocovers fraud and has the gov't arrest, prosecute and punish the criminal they are securing a free service from the DOJ and DOC.

And we forget what conservatives love to tell us about private companies and taxes....corporations don't pay taxes people do! A company is either going to pass its tax bill onto customers in the form of higher prices, employees in the form of lower pay or shareholders in the form of lower profits. So in a real sense a corporation hasn't paid for the gov't anymore than Medicare has paid for it.

And private insurers also spend a great deal of money fighting fraud, while Medicare doesn’t. That might reduce the program’s administrative costs, but it greatly increases its overall costs

Perhaps but that would show up as a higher cost per person. Why did Medicare Advantage cost more per person than regular Medicare if the private insurance companies under Medicare Advantage were keeping close watch on fraud while regular Medicare was allowing con men to bill willy nilly?

As anyone who has filed a claim with an insurance company can attest, the "fraud fighting" they perform includes denying legitimate claims. Giving customers a hard time is part of their business model.

While I don't think this *worse* for health insurance than other insurance markets, the cost of dealing with private insurers denying legitimately insured claims is not trivial.

It used to be routine for insurers to deny all claims to provide more justification for them or as an opening for a lower reimbursement. Once providers caught on, they learned to resubmit them at higher cost but these higher provider costs are not considered insurer overhead.

Again fraud doesn't factor into this comparison here. Say you have an insurance group (Medicare or a private company) spend more money on fraud prevention and detection. Yes it's administrative costs would go up but if it spent the money wisely the actual spending per patient covered would go down. On the other hand if an organization decided to cut back on fraud prevention you'd see administrative savings but an increase in per cost patient (assuming that fraudsters took advantage of the opening).

Spending money on fraud prevention, then, is pretty simple. Ask yourself if you can spend $1 more on fraud prevention to achieve more than $1 in savings. If yes, then spend $1 more then ask yourself that question again. Repeat until you get to a point where the additional $1 yields less than $1 in prevented fraud.

In deciding which entity is doing a better job (Medicare, HMO, other group etc.) all you need to do is look at total cost per patient and then look at the net funds spent on patient care (total cost - administrative). If the secret to Medicare's success was just letting fraudsters bill it like crazy then the net funds spent on patient care would look great but the total cost per patient would be higher.

Tyler, this blog has descended into parody. It used to be my single favourite blog on the internet, but when I have to wade through nonsense like this every couple of posts, I am seriously thinking of taking it off my blogroll. What exactly do you hope to achieve? I am not sure if you don't know better (I am beginning to suspect as much as you usually just resort to some high-level hand-waving -- "great stagnation!" "deadweight loss!"). There are other blogs for serious policy analysis and discussions.

Is there much of a deadweight loss?

Certainly many people would voluntarily pay for insurance now to guarantee coverage in their elder years.

Second, many of those people who would not voluntarily pay would gladly impose costs on Good Samaritan Hospital at 68. So the tax might reduce a market failure.

I agree the administrative costs are glossed over in a dedicated effort to make Medicare seem cheaper than it really is. As some point out, the additional cost per recipient is marginal, but pennies add up to dollars with 300 million beneficiaries. The marginal overhead costs from adding an extra layer of taxes for another dedicated purpose is not trivial. Average fixed costs do decline with output and there are economies of scope, though.

Those who say Medicare is an inexpensive add-on to an existing tax infrastructure aren't considering whether that existing tax infrastructure and the taxes it administers is necessary or efficient.

Those who say Medicare is an inexpensive add-on to an existing tax infrastructure aren’t considering whether that existing tax infrastructure and the taxes it administers is necessary or efficient.

Errr, true but the existing infrastructure is there now and that doesn' t change the fact that it makes Medicare pretty cheap in terms of overhead. If you want to propose a radical overhaul of the tax system, you should take into account the cost of trying to keep Medicare's funding mechanism into your cost estimates then.

Those are good points.

Dead weight loss is the loss of value from transactions that don't happen because of the tax. Deadweight losses are linked to the total tax burden (although if taxes are more administratively complex, that also increases the dead weight loss).

Basically, say you're happy to employ someone if you pay them no more $10 an hour, and they're happy to work if they earn at least $9 an hour (maybe they're a student looking for part-time work). So you'd hire them at a wage somewhere between $9 and $10. In that case, a $2 an hour tax means that you won't hire them, as it drives a wedge between their value to you, and the earnings they take home. It's called a DWL, as no one captures that value, not the employer, not the employee, not the government (as the transaction doesn't take place because of the tax).

The DWL is, at a first order, independent of the benefits of whatever is paid for by the taxes. (One can make up stories about how various government provided-services make people more willing to work, eg healthcare means they're in better shape to work, or less willing to work, eg they don't need to work to obtain healthcare).

This calculation, though, must also be done for the private sector and applied to private health care.

Student is willing to work for at least $8 an hour. Employer doesn't want to pay more than $10. In the absense of everything else the kid gets hired at $9. But there's taxes of $3 so either the employer has to pay $11 to hire the kid for $8 or the kid has to take only $7 to work for $10 of cost to the those cases the transaction won't happen.

But consider health care as compensation isn't now the employer may be able to spend $10 for the kid and give him $7 plus health benefits that cost $3 with taxes. The kid only values the health benefits at $1 per hour despite the $3 per hour cost to the employer. As a result, the kid suffers a $1 per hour loss as he would have taken $9 an hour in a world without complication. Also the employer would have paid only $9 an hour instead of $10 in a simple world so he too has a $1 loss. That $2 deadweight loss, then, should likewise be added to the 'hidden overhead' of private insurance you are comparing Medicare too.

Only if the employer is obliged to pay health benefits (be that by law, or by an inflexible HR department).

Can anyone explain Tyler's invocation of deadweight loss here in comparing private vs. public provision of health insurance? I can deadweight loss's relevance to public provision of health insurance vs. no insurance.

Simple. Take the current cost of medicare for a given population and compare it to what amount of money would be spent in a completely private system. I suspect the difference would be immense.

Really? Give me a break. Others have addressed many of the weaknesses of Yuval and Tyler's arguments, but I want to point out an issue with the assertion about fraud. It is completely incorrect the Medicare doesn't work to fight and prevent fraud. In fact the ACA includes $340 million in additional funds to fight fraud. Despite the fact that this is one of the administrative costs that the author claims is beneficial, I suppose Yuval would turn this wise CMS expenditure on its head with a "Look, look another hidden administrative cost!!"

Sorry, this is a meme that sounds good to rabid anti-government types, but does not add up.

Let's take one of the first items he lists: the IRS, surely spends lots and lots of money collecting Medicare taxes, but those devious Federal bureaucrats cleverly hide that when talking about Medicare's efficient administration.


The _entire_ IRS budget is currently about $12.5 bn pa (from the IRS website). To be sure, the cost of collecting and transferring payroll taxes to the Medicare administration is a rather small part of that cost, especially viewed as a marginal cost. Medicare payments are not complex to calculate, there's no deductions, exemptions, extra forms, etc., and the payments flow in highly regular and predictable ways. We'd be safe in saying that the IRS's administrative costs for collecting Medicare taxes add very little to IRS total costs -- but let's be wildly generous and call it 5% of IRS total costs. That's $625 million

HI and SMI Medicare revenue in 2011 were $228.9 bn and $301 bn, respectively, or a total of about $530 bn. (from the Trustees' report).

So the IRS spent (probably much less than) $625 mn to bring in revenue of $530 bn.

And what proportion of $530 bn is $625 mn? I get about .0012, or about 0.12% of the amount collected (and remember, that's assuming that Medicare collection costs are 5% of the IRS budget, which is a wildly generous estimate). Or to put it per capita, Kaiser estimates that there are 47m Medicare beneficiaries, currently. That makes the IRS's collection cost per beneficiary about $13 pa. Darn those slippery Federal bureaucrats hiding such a breathtakingly high cost!

This is what Dr. Cowen would have us consider a component of what he's claiming are "significant administrative costs"? My calendar does not say it's April 1, but surely he's jesting.

This is why arguments and claims are better with numbers. If Tyler had calculated that the collection overhead was 0.1% he would have realised it wasn't worth mentioning. If he had calculated a higher value, we could argue details.

Not being careful with numbers just got Martin Feldstein in trouble as well!

I'm going to guess what Tyler would say in response. Yes the actual IRS spending is pretty small (about $15.5B) hence even if Medicare's administrative books were charged for its 'services' it will be a pretty small change in its cost accounting.

But the $12.5B the IRS spends on its computers, agents, auditors, printing instruction books and so on hides the larger cost of the entire tax system. For example, every employer has to carefully keep track of how much they pay employees and report that to the IRS every quarter, as well as withhold taxes and remit that.

That would be a very fair point, however I think it leads us into a thicket of conceptual problems:

1. Private enterprises also benefit from this structure. When you get a loan, for example, banks find it helpful to see a copy of your tax return. If there was no IRS and no income taxes, they would have to turn to other ways of verifying your income. As I pointed out before, private health insurance companies benefit greatly from the fact that the IRS tax structure makes it easier to sell their products to big employers and collect premiums from them. To do a fair apples to apples comparision you have to construct a plauisble alternative universe under some other tax regime and try to see what types of administrative burdens/benefits that would be to private insurance companies.

2. Tyler is, I think, confusing Medicare with Medicare's funding mechanism. It's pretty easy to imagine, say, a VAT being instituted to replace the payroll tax. Or a series of import duties. Or a cap-n-trade carbon tax system. And so on. Even if he is correct that paying for Medicare via a payroll tax has non-trivial hidden administrative costs, it is a bit besides the point. You can keep Medicare's output exactly the same while drastically changing its payment mechanism. And the point of this inquiry is the assertion that the cost of Medicare's output (providing coverage) is less than the cost of private insurance company's output.

JTA that if taxes weren't collected via payroll, they would be collected some other way that could well be more expensive or less efficient.

If the IRS / SSA etc. didn't do any of what they currently do for Medicare how much lower would their budgets be?

I'm going to guess maybe a fraction of a percentage less.

if administrative costs for managing two patients are both $100 but one patient has $200 of health expenses in a year and the other has $2,000 of health expenses, the insurer that covers the first patient will have a far higher administrative-cost ratio, even though both have the same administrative costs.

Are per-capita administrative costs really invariant? Sounds like a terrible assumption.

$2000 worth of billing is going to take far more labor to aggregate, calculate etc. Agreed there are fixed costs but the situation isn't what he portrays.

Actually, it won't. The amount of work necessary to process reimbursement is pretty much the same whether the charge is $200 or $2000. Mark Litow of Milliman, a private actuarial firm, finds that the costs of administration are actually 5.2% for Medicare, and 8.9% for private insurers. Further, since the medical costs for a Medicare beneficiary are substantially higher on average ($6600 versus $2700), much of the apparent saving is illusory – it simply reflects the fact that the denominator is different.

Are you assuming they are both a single claim?

Sounds like yet another argument over the "facts" that has little to do with the facts. Even if we knew the true economic cost of Medicare to some absurd level of over-precision....we would still disagree on what to do with that information. So why not focus on some core disagreements like the government can't be efficient/innovative or the private sector won't provide social insurance? Sure factual details matter and need to be considered, but they can also distract from the real issues.

"if administrative costs for managing two patients are both $100 but one patient has $200 of health expenses in a year and the other has $2,000 of health expenses, the insurer that covers the first patient will have a far higher administrative-cost ratio"

This I don't agree with, and may shoot much of the argument down.

Folks with low medical risks, like the young and healthy are of course much cheaper to insure, and require much less attention, administration and management at the insurer level. Then compare to the senior who is already on 6-10 pills and has 3-6 diseases. There is much more to be done at the insurer level as this patient will require more care, more medical encounters, more medications, more hospital days and more technologically advanced testing, treatment and interventions. So the insurer has a huge job looking after all that!


What is your explanation for the very large differences between the US and other state funded health systems in the OECD?

It seems that they are much cheaper. For example it is 10% of UK GDP (State run), 11% of Canadian GDP (State Funded) and 18% of US GDP.

The mean health results of the state funded OECD states are generally better than those in the US. Surveys of satisfaction with health care provision compares favorably.

One explanation is that state funded health care is more efficient than the US private health care + medicare/aid system.

I am receptive to other explanations. The above one is the most obvious to me.

Do you think there is a better explanation?

Out of curiousity, what's your source for the statement that state-funded health care is generally more efficient?
A brief google got me this <a href=""OECD report, which stated that no particular financing system was more efficient. Some of the countries that had private healthcare insurance for basic healthcare were Germany, the Netherlands, Slovakia and Switzerland (page 7). And 3 of those countries had relatively equal outcomes, Slovakia being the exception (page 8).

This post reminds me of the person who says your shoes are untied so you look down and miss the big tree in front of you.

The post assumes administrative costs of collecting and billing are the issue, and that is how one would compare private insurance with social insurance.. but, it isn't.

If you didn't have social insurance (all covered over a certain age, regardless of pre-existing condition or age) you would have a private system with 1) underwriting costs and underwriters to determine each individuals premium or class rate; 2) medical exams for coverage; 3) adverse selection; 4) marketing costs, particularly when carriers would be seeking healthy risks.

Do you think the issue really is the color of each systems shoe laces (collecting fees and administering claims) or did you notice that while you looked down and examined that issue you missed the obvious issues that differentiate private versus social insurance, and explain why social insurance for the elderly is universal among developed nations.

I just finished reading Levin's piece, and it is pretty clear that most of the commenters ignored most of its substance.

But Levin's response is not entirely correct either. He exaggerates some of what he terms "administrative" costs. The vast majority of Medicare recipients are on SS and their premiums are paid directly out of their SS account. This is just a simple electronic transfer of payment and it's linked directly because surprise - your Medicare ID number is your SS# with a letter after it (at least mine is). Levin assumes in his example that administrative costs are fixed per patient and while this may be true for 'part' of it, most of the administrative costs are linked to actual usage; this is no different from what the private insurance industry experiences where each individual claim represents a new administrative expense. Thus, it is a little disingenuous to make the argument about two different patients, one incurring a 10 fold increase in expenses without knowing the # of medical claims. With respect to the 'Harvard' study that Levin sites, his exact interpretation has already been refuted by the study's lead author, David Cutler so it's Levin who is not being straight forward on this point. Others have already made the point about Medicare fighting fraud and there have already been a lot of people put into jail for this.

There are some other minor points as well but you know it's really not worth getting into. Trying to guess what any kind of a system will look like in 10 years is pure fantasy conjecture and quite frankly a waste of time. Neither side will win this battle since as long as Medicare is kept separate and we do not have a capital budget of health care that covers all people the problems will continue. At the end of the day we will either have a socialized government run system or a comprehensive approach to providing private coverage where a core of benefits are available to everyone. My particular bias is to do this with a dedicated VAT and eliminate all the other stuff (Medicare, Medicaid, employer paid for insurance) and allow individuals to purchase extra coverage should they so desire. This at least levels the playing field and gets everyone covered. Anything else is a kludge and you can only put so many fingers in the dike before it floods.

The discussion of Deadweight Loss (DWL) in the comments has been frustrating. While collection costs can be considered such a loss, that’s not at what Tyler is talking about.

Once individuals become eligible for Medicare, the size of their benefit is unaffected by the amount of taxes they pay towards financing the program. Consequently, when thinking about the spending power one gets out of an additional dollar of income, the taxes used to finance Medicare benefits are a pure personal loss. (The fact that the individual might buy private insurance in Medicare’s absence is not relevant because the amount of insurance the individual could purchase would rise with additional dollars of income while the size of the Medicare benefit does not; think Marginally.) The tax payments thus distort classic choice margins including hours of work, work effort, and consumption of things that are deductible (e.g., mortgage interest) and excludable (e.g., fringe benefits) from taxable income.

Let’s put some numbers to the calculation using the standard DWL formula one would learn early in an undergraduate public finance course. The formula states:

DWL = ½ x (marginal tax rate squared)x (elasticity of taxable income) x (size of the tax base).

To keep things simple, I’ll think of the tax base as being all wages and salaries (about $6.7T in 2011) and compare the average marginal tax rates required in worlds with and without Medicare (about $550B in 2011). The average marginal tax rate in Medicare’s absence is not a trivial number to calculate, as it must incorporate state and local taxation as well as federal; I use 0.35 (for a time series on historical average marginal tax rates, see Figure 1 here: To a first approximation (see caveat below) Medicare adds 8.2 percentage points to the required tax rate ($550B/$6.7T), for an average marginal tax rate of 0.432. For the elasticity of taxable labor income, I’ll go with 0.3. (Martin Feldstein tends to use a number like 0.5 while relatively liberal economists might nudge the number down to 0.2.)

In the absence of Medicare we have: DWL = ½ x 0.35^2 x 0.3 x $6.7T = $123 Billion

With Medicare we have DWL = ½ x 0.432^2 x 0.3 x $6.7T = $188 Billion

By this calculation, there’s an incremental DWL of $65B ($188 - $123) on Medicare expenditure of $550, for a “marginal cost of funds” of about 12%. This number is somewhat understated for the following reason. The response of the tax base to the rise in tax rates will increase the size of the rate increase needed to generate the $550B. A hasty calculation suggests that this would increase the estimate of the marginal cost of funds to about 15%, but I won’t stake my life on having done this last piece perfectly. The marginal cost of funds will also rise as the rate increases used to finance the program are made to be increasingly progressive.

i don't know what public finance course you took, but can see you dont understand the difference between social insurance and private insurance. You might want to go to Shillers lectures on social insurance v private insurance at his finance Econ course at

In private insurance you have adverse selection and various underwriting cost, those who do not purchase, thereby changing the risk profile of the pool, eyc.

I also have a problem with your analysis which treats prior contributions as a fixed element thereby making other elements marginal.

If you want to do a cost study, you can take a history lesson from the data on private insurance for the elderly as it existed before Medicare. That is the natural experiment anyone can use, and no one does, for obvious reasons.

By the way, in your calculations you also are assuming that those who did not purchase medical insurance, would not be covered in some other way by government expenditures if they were poor or unable to purchase insurance because of prior conditions. Your comments fail to posit the costs of the but for world. For the but for world, go to the world prior to Medicare and take the Shiller finance course.

Agreed that my calculation involves a Medicare world vs. a world with no Medicare or other public substitute. An extreme version of your argument in this second point is "if we got rid of Medicare, we would ultimately decide to cover the elderly through Medicare 2." This doesn't seem quite like the right way to think through the cost side of the public/private comparison.

Medicare two is not an extreme example. Prior to medicare, what you had were public hospitals and county run retirement homes, and the costs would be picked up by the local property taxpayer or by the hospital which provided indigent care. At least with Medicare, poor people pay something into the system over their lifetime.

Yes, it is an unjust comparison to make. The counterfactual is an unknown, but certainly not the one presented in the Clemens calculation. So the the key point of the comment, that DWL is $65 billion, is a false conclusion.

The benefits of universal social insurance are certainly important. They should be estimated and put on the benefit side of the ledger, and in many cases they may overwhelm a comprehensive accounting of the costs. We still have to do a full accounting of the costs, however, and those costs include the deadweight loss from paying for the program using income and payroll taxes.

P.S. I'm not sure why you're so excited by these Shiller lectures, but I'll be teaching a course on Redistribution and Social Insurance this spring and I think that you'd find my lecture notes to be quite similar to Shiller's, Raj Chetty's, etc.

I thought about this early this morning when I woke up and realized that what is missing from this argument and only tangentially touched on by Bill is that Medicare is a not for profit program and private insurance is not. Of course this one of the reasons why private insurers want to either exclude very ill patients from their program or charge risk adjusted premiums. We know that under either approach the elderly will not be covered. While I appreciate Professor Clemens calculations above, any discussion in the absence of a consideration of what an appropriate ROI/ROE for a private insurer might be is by definition deficient. Once you add this cost in the argument about DWL is really not all that important.

Unlike life and property casualty insurers, the float for medical insurance underwriting is probably not as significant (otherwise you would see Bershire Hathaway entering into this market big time).

One final note - the largest single purchaser of private health care in the country is the Federal government and they are exempt from taxation as an employer.

Thank you Jeff Clemens!!!

I'm reading through these post and thinking "JC....none of these people know what a deadweight loss is!!"

I was going to post a simplified definition to try to move the debate forward, but your post is better than anything I would have come up with.

It is reasonable to point out that dead weight loss is a fascinating topic - personally, I find this German wiki perspective fairly good -

'Unter einem Wohlfahrtsverlust (auch Nettowohlfahrtsverlust, Allokationsverlust, Zusatzlast der Steuer, Steuerkeil, dead-weight loss, excess burden, Harberger Dreieck) versteht man in diesem Zusammenhang den durch eine Marktstörung im Vergleich zur Situation vollkommener Konkurrenz verursachten Verlust an Konsumenten- und Produzentenrente. Die Ursache für den Verlust an Wohlfahrt ist jeweils, dass die gehandelte (= produzierte) Menge von der pareto-optimalen Menge abweicht, die sich auf einem vollkommenen Konkurrenzmarkt im Gleichgewicht einstellt.

Da man den Wohlfahrtsverlust (zumindest theoretisch) wertmäßig berechnen kann, können die Kosten von Markteingriffen wie Steuern, Höchstpreisen, Zöllen oder Marktversagen (z. B. aufgrund von monopolistischen Strukturen oder externen Effekten) berechnet werden.'

It would seem that a lot of people are getting seriously confused about the idea (for example) that Medicare benefits from having its own government funded collection agency, and that somehow, the cost of taxation in and of itself is 20%.

I will merely note that in Germany, my health insurance is completely private, a transfer of a portion of my wages to a private company (most insurance companies in Germany are, in the sense of not being a government agency). And yet, just like other line items, the deduction is made in an identical fashion to that of my income taxes. The (essentially) single government health insurer AOK uses exactly the same mechanism, in exactly the same fashion.

But who knows? Since the incredibly expensive German health care system (only 1/3 cheaper than the U-S.'s, for essentially universal coverage) does not use a goverment taxing facility, maybe that decrease in German costs is due to not using the Finanzbehörden as health insurance premium collectors.

I'm sure that someone in America is capable of making such an argument.

(As a side note to the German text, which does not ignore a moral dimension when talking about the subject, as the very term 'Wohlfahrtverlust' indicates - 'loss of welfare' meaning that the loss should be seen in the same light as is actually written in the Preamble 'promote the general Welfare.' We are talking about health insurance, a 'good' which most industrialized societies feel should be something all citizens are provided, as promoting the general welfare. Only the U.S., oddly enough in a Gandhian sense, doesn't understand the very message its system was founded under.)

Wall Street would laud the "synergism" of having related programs share costs, functions and data. I look forward to reading a like assessment of the indirect costs of government operations related to financial firms.

Comments for this post are closed