The American Enterprise Institute (AEI) was founded in 1938, partly by executives from corporations such as Eli Lilly, General Mills, Bristol-Myers, Chemical Bank, and Chrysler who were disgruntled with the effects that the New Deal was having on society. Its board is today made up almost entirely of executives from major corporations, and it’s staffed mainly by intellectuals and former government officials. Growing rapidly between 1970 and 1980, when its revenue expanded from less than $1 million with a staff of ten to about $8 million with a staff of 125, it played an important role in ushering in the current era of Reaganist, supply side, neoliberal economics.³ It is currently ranked sixth on a list of America’s most influential think tanks, the second highest of "partisan" think tanks, just behind the Heritage Foundation.⁴

The Heritage Foundation was founded in 1973 with help from beer magnate John Coors. It was instrumental in creating the Contract with America that helped Republicans win a 1994 majority in Congress. In partnership with the Wall Street Journal, it publishes the annual Index of Economic Freedom. The Index notes that this year, "Regrettably, populist attacks on the free market, fueled by the economic slowdown and the political temptation of quick interventionist remedies, have gained momentum." It is currently ranked fifth on a list of America’s most influential think tanks,  the highest ranked of those that are considered "partisan."^5,6

Falsehood #1. The US does not spend an excessive proportion of its GDP per capita on health care.⁷

This statement is so clearly untrue that it doesn’t seem to be quoted much even by other right wing sources, but it’s still important because it demonstrates a total lack of intellectual integrity by the AEI authors who made the claim.

To understand how they rationalize this, let’s start with another analysis of US health spending by Princeton economist Uwe Reinhardt published on the New York Times Economix blog.⁸ Here’s the graph from that website.

Reinhardt explains:

You’ll notice that there is enormous variation in health spending per capita in different countries within the O.E.C.D. But the graph also indicates that there exists a very strong relationship between the G.D.P. per capita of these countries (roughly a measure of ability to pay) and per-capita health spending. The dark line in the graph is a so-called regression equation (whose precise mathematical form is shown in the upper left corner).

That line tells us something important about the relationship between a country’s wealth and its health care spending.

An additional insight from the graph, however, is that even after adjustment for differences in G.D.P. per capita, the United States in 2006 spent $1,895 more on health care than would have been predicted after such an adjustment. If G.D.P. per capita were the only factor driving the difference between United States health spending and that of other nations, the United States would be expected to have spent an average of only $4,819 per capita on health care rather than the $6,714 it actually spent.

Now let’s compare the American Enterprise Institute source. After listing a similar graph to Reinhardt’s, they then explain that it is irrelevant because the U.S. "residual" is sensitive to "model specification." In other words, you can just do this:

But as other bloggers have noted,⁹ they give no justification for why you would do that. They do state the US is an extreme value, because it’s richer than almost all other countries, but that difference pales in comparison between differences between many of the other countries in the model (in other words, US GDP isn’t an "outlier" in a technical statistical sense). And as Reinhardt states, the more intuitive model is a very good fit:

Just knowing the G.D.P. per capita of nations helps us explain about 86 percent of the variation in how much different countries pay for health care for the average person.

Finally, there’s also the small matter that if we extend the above AEI model out another twenty-five years and assume a reasonable rate of economic growth, it eats up literally the entire US economy. But doing the slightest sanity checking is evidently uninteresting to these authors.

Falsehood #2. If you adjust US life expectancy for violent deaths, it becomes #1 in the world.

This is simply a lie. Unlike the falsehood above, though, it’s not completely obvious that it’s wrong, and so it’s been used extremely successfully to muddy the debate.

What AEI did was create a model that predicted life expectancy from two factors: GDP per capita and violent deaths. But despite that the US ranks #1 in this model if you remove sources of violent death, that doesn’t mean anything because the other factor, GDP per capita, totally fails to correctly predict US life expectancy. Putting it simply: the model’s wrong, so any conclusions drawn from it are also.¹⁰

The OECD explains it thus:

It has been claimed (Ohsfeldt and Scheider, 2006) that adjusting for the higher death rate from accident or injury in the United States over 1980-99 than the OECD average would increase US life expectancy at birth from 18th out of 28 OECD countries to the highest. In fact, what the panel regression estimated by these authors shows is that predicted life expectancy at birth based on US GDP per capita and OECD average death rates from these causees is the highest in the OECD. The adjustment for the gap in injury death rates between the United States and the OECD average alone only increases life expectancy at birth marginally, from 19th among 28 countries on average over 1980-99 to 17th. Hence, the high ranking of adjusted life expectancy at birth mainly reflects high US GDP per capita, not the effects of unusually high death rates from accident or injury.

It took about a year for this study to be refuted after it was published, finally prompting the authors to divulge to the Wall Street Journal that they were "not trying to say that these are the precisely correct life-expectancy estimates. We’re just trying to show that there are other factors that affect life-expectancy-at-birth estimates that people quote all the time."¹¹

What is really more interesting, though, than the study itself is the vast extent to which this keeps being repeated in the (supposedly more democratic) blogosphere because people are just unaware of the falsity. For example, when Betsy McCaughey cited these statistics on the Daily Show, most people even on Democratic blogs like Daily Kos and Firedoglake did not realize that this study had already been definitively refuted, though of course they were very skeptical of McCaughey in general. On Daily Kos not one post mentioned it at the time, though it is possible that people were just distracted by McCaughey’s claims of "death panels." But on Firedoglake we still have people wondering about this in blog posts—and no one pointing out it’s already been refuted—as of a few weeks ago.¹² Meanwhile a July post on the "slightly left of center" blog Angry Bear cited this study as "casting serious doubt" on the significance of the life expectancy difference, with no refutations in the first few pages of comments.¹³

Senators like John Ensign have also cited these statistics in Congress. Meanwhile Amanda Terkel, a Deputy Research Director of the Center for American Progress, doesn’t realize in her Think Progress blog response to Ensign that it’s all a complete lie despite being skeptical in general.¹⁴ And don’t even mention the right wing blogs!

What seems to be happening here is that the traditional media has mostly ignored the fact that these figures are false, leaving amateur bloggers to wander about hopelessly in a swamp of misinformation. If the media was more decent, they would call these people out vigorously as liars, not write, say, one Wall Street Journal blog post and then forget about it. (The post cited above concludes ambiguously, "What do you think? Should certain deaths be excluded from life expectancy? Is it a solid basis for comparing health systems? Please let me know in the comments.")  

Falsehood #3. Medicare’s administrative overhead is higher, or at least not much lower, than that of private insurers.

This is a very oft-cited claim and has been promoted aggressively by Robert Book at the Heritage Foundation. Book claims that the traditionally cited overhead of Medicare of around 3% is vastly understated because Medicare administration depends on the number of beneficiaries, not the amount of claims that are actually paid out. If this were true, then Medicare would have understated percentage costs simply because its elderly beneficiaries require many more claims than average.¹⁵

But in the first place, Medicare apparently does not have higher per person costs at all: Book’s estimate is based on a highly problematic study by Benjamin Zycher at the Manhattan Institute, another neoliberal think tank. This study states that Medicare’s per claim administrative costs are not 3% but rather 6% because of factors like part of the cost of running the criminal justice system, of which Medicare is assumed to be partly responsible. Many of these factors, though not all, seem highly dubious.^16,17 And in contrast to Book, the Kaiser Family Foundation finds using the official government figures that:¹⁸

For plans operating a PPO on an administrative services-only basis—essentially the way Blue Cross operates under FEHBP—mean administrative costs per member year were $271 in 2002 (Sherlock Company, 2002). Medicare’s costs in 2002, $4.8 billion for about 36 million fee-for service beneficiaries, were about $133 per beneficiary, or about one-half as large.

The second and more fundamental problem with Book’s argument is that there is no evidence that Medicare’s administration actually depends more on the number of beneficiaries than on other factors. In fact, Book does not even bother to offer any compelling reason why this would be true, instead simply alleging that the total proportion of Medicare administration that is directed at processing claims is 4%, so therefore Medicare administration depends on the number of beneficiaries. There are two direct problems with this:

1. Even if this proportion is true, it doesn’t take into account constant and ambiguous expenses.

2. From what I can make out from the Medicare budget, the proportion probably isn’t true.

To understand this, note that there are at least three kinds of expenses in Medicare administration.

1. Constant expenses, such as deciding which medical procedures to cover, that are incurred regardless of the number of beneficiaries or the amount of claims paid out.

2. Per beneficiary expenses, such as enrollment, member education, IT support for these, and so on that do not depend on the number of claims paid out.

3. Per claim expenses, including provider reimbursements, appeals, IT support for these, and so on.

Given the complexity of Medicare’s administration, we might also add a fourth kind.

4. Ambiguous or composite expenses, those which are not obviously dependent on either the number of claims or the number of beneficiaries, but could be, and could also depend on other factors.

The first problem with Book’s assertion is that there are seemingly many expenses in the last category and perhaps the first category.

Given that he acknowledges the first category exists, it seems odd that he then completely ignores it, because if the amount of constant expense is very large then it could dominate all other factors. I have no idea whether this is true, but Book ought to have at least examined it if he is asserting that per claim expenses are as small as 4%.

The fourth category is also problematic, which in turn leads us to the second problem with Book’s assertion, namely that his 4% proportion seems flawed. To see why, consider Medicare’s administration budget.¹⁹

CMS is committed to administering its programs as efficiently as possible. In FY2009, benefit costs are expected to total $703.9 billion. Non-benefit costs, most of which are administrative costs such as Program Management, Medicaid State and local administration, non-CMS administrative costs, costs associated with the health care fraud and abuse control account (HCFAC), the Quality Improvement Organizations (QIO), the Clinical Laboratory Improvement Amendments program (CLIA), and the Medicare Advantage user fees, among others, are estimated at $18.6 billion or 2.6 percent of total benefits under current services. CMS’ non-benefit costs are minute when compared to Medicare benefits and the Federal share of Medicaid and SCHIP benefits. Program Management costs are only one half of one percent of these benefits.

I decided to examine the very first cost listed, Program Management, since Book apparently did too.

The Program Management account provides the funding needed to administer CMS’ programs, including Medicare, Medicaid, SCHIP, CLIA, QIO, State Grants and Demonstrations, and HCFAC. There are four line items in the Program Management account—Medicare Operations, Federal Administration, Survey and Certification, and Research–each one with a distinct purpose. Medicare Operations primarily funds the Medicare contractors that process fee-for-service claims as well as the IT infrastructure, operational support and oversight needed to run the fee-for-service program and the new Medicare Advantage and Prescription Drug programs. In addition, it funds legislative mandates (e.g., HIGLAS, HIPAA, contracting reform, competitive bidding) which improve and enhance CMS’ programs. Federal Administration pays for the salaries of CMS employees and for the overhead (rent, building services, equipment, supplies, etc.) associated with running a large organization. The Survey and Certification account pays State surveyors to inspect health care facilities, both when they enter the program and on a regular basis thereafter, to ensure that they meet Federal standards for health, safety, and quality. The Research line item supports a variety of research projects, demonstrations, and evaluations designed to improve the quality of healthcare furnished to Medicare beneficiaries and slow the cost of health care spending.

The Program Management account is $3266.7 million, accounting for 17.5% of estimated overhead costs. Of that, Medicare Operations is by the far the biggest part, making up $2237.8 million. Of that, the biggest single part is FI/Carrier/MAC Ongoing Operations, and this is the page that Book cites in his report. But there are other parts of Medicare Operations that plausibly lend support to claims processing, most obviously including Contracting Reform at $189.6 million, HIGLAS at $153.7 million, and probably some of the miscellaneous IT Investments category which is valued at $166.9 million. And that’s just a look at a few of the biggest expenses inside Medicare Operations. The second biggest part of Program Management is called Federal Administration, most of which is composed of salaries valued at $508.9 million—has Book factored in the cost of paying the workers that handle the claims? And outside of Program Management, what about QIOs (Quality Improvement Organizations), which cost $1,128,400,000 and help decide if reimbursements are being well spent? Even not considering the large majority of estimated overhead, there is a lot open to question here.

The Medicare Advantage Comparison

But in fact, all of this figuring is not strictly necessary, because there is a surefire way of testing the claim that treating a sicker, elderly population leads to lower administrative overhead per dollar spent, as Book alleges. This way is to just compare Medicare with Medicare Advantage, the privatized version of Medicare serving about a fifth of all enrollees, which treats a similar population.²⁰ The simple and irrefutable truth is that insurance companies under Medicare Advantage have almost exactly the same administrative costs as for private insurance generally: anywhere from 11% to 16% for Medicare Advantage plans,^21,22,23 versus about 12% on average.

Administrative Costs. Private health plans that participate in Medicare have higher administrative costs per enrollee than the traditional Medicare program because of their smaller scale of operations and their costs associated with marketing, utilization management, network development and retention, and reinsurance. Administrative costs and return on investment account for about 11 percent of private plans’ costs of delivering Medicare benefits, whereas the administrative costs of the fee-for-service Medicare program (as reported by CMS) account for less than 2 percent of its expenditures.17 There is some disagreement among analysts about the size of that difference in costs, however. Some analysts contend that the latter estimate understates the administrative costs of the Medicare program because it excludes some costs that should be allocated to Medicare, such as portions of the salaries of legislators, legislative staff, and CMS administrators; depreciation on certain federal office buildings; and the costs of collecting Medicare premiums and payroll taxes.18 Following that reasoning, however, a portion of those additional costs should also be allocated to the Medicare Advantage program.

Of course, Book is not about to give up: he has claimed in reply that "costs of administration incurred by providers" (his emphasis) are picked up to an unusual extent by HMO plans. This might or might not be true, but he then draws what appears to me to be an unjustifiable conclusion: "An administrative cost figure for Medicare FFS that would be comparable to figures reported for Medicare Advantage would be the portion of the payments to providers that are related to their administrative expenses plus the program-level administrative expenses reflected in the CMS budget (3.1 percent for 2005). This would be approximately 19.4 percent for 2005." Here Book claims that the only way to fairly compare administrative expenses between HMOs (of which Medicare Advantage is primary made up) and Medicare is to lump all the provider-level administrative costs of Medicare in with its official overhead. For this to be a fair comparison, it would apparently imply that HMOs incur all provider-level insurance related overhead as part of their own overhead. That is clearly impossible.  

The falsehood of high Medicare overhead is complicated by the fact that it probably contains a grain of truth: estimates for the percentage of legal costs in the justice system attributable to Medicare, though  too high in the Zycher study, are clearly not zero. Still, such hidden overhead has been exaggerated, and enormously so by Heritage. Such inflated estimates are then used by insurance companies to promote bad public policies,²⁴ as well as repeated by industry blogs like The Health Care Blog and many right wing sources.²⁵

Author’s notes:

1. I don’t have any background in accounting or related fields, and couldn’t find a very specific breakdown of where most of Medicare’s overhead comes from, though I feel I found enough to draw the conclusions above.  If anybody can find a document that thoroughly tallies up the sources of where 100% (or even most) of it is from, I’d be grateful. My e-mail is khin2718 [at] gmail [dot] com.