A remarkably important and persuasive paper that calls into question the need for “reforming” Medicare has not gotten the attention it warrants. “An Examination of Health-Spending Growth In The United States: Past Trends And Future Prospects” (hat tip nathan) by Glenn Follette and Louise Sheiner looks at the model used by the Congressional Budgetary Office to estimate long term health care cost increases. Bear in mind that this model is THE driver of virtually all forecasts of future budget deficits.
This paper, although written in typically anodyne economese, is devastating in the range and nature of its criticisms. And the reason this assessment should be taken seriously, independent of the importance of the issues it raises, is that the authors are uniquely qualified to make this critique. Follette is chief of the Fed’s fiscal analysis section. Sheiner, a fellow member of that group, has worked for both the Treasury and the Council of Economic Advisers previously. In other words, the sort of analysis they have made here is the core of what they do on a daily basis.
The argument made by the opponents of the plans to cut Social Security and Medicare generally take this form: concerns about Social Security are greatly exaggerated. They are based on long-term forecasts, which are notoriously inaccurate in outlying years. The most commonly cited, by the Trustees of the Social Security system, projects the exahustion of the famous trust fund in 2033. As several analysts have observed, if Social Security really has a problem, we’ll know it in plenty of time; there’s no need to do anything immediately.
By contrast, conventional wisdom is that Medicare does have a long term cost predicament, but the problem is not demographic, but that of the steep rise of health care costs in general.
The fundamental beef of Follette and Sheiner with the CBO model is that it naively assumes past growth in health care spending as the basis for its long-term projections. The result is that it shows that trees will grow to the sky. One of the things anyone who has build forecasting models will tell you is you come up with assumptions that look reasonable and then sanity check the output (for instance, does your model say in year 10 that your revenues will be 3x what you can produce given your forecast level in plant and investment? If so, you need to make some revisions). The Fed economists point out numerous ways that the model output flies in the face of what amounts to common sense in the world of long term budget forecasting. From the opening of the paper (emphasis ours):
Long-run projections of the U.S. federal budget have played a prominent role in discussions about fiscal policy and the design of major transfer programs for several decades. The projections typically show large fiscal imbalances owing to ramping up of retirement and health care costs relative to GDP. Health care costs are the key factor in these projections for two reasons. First, in current projections they are the prime source of growth of spending as a share of GDP. Second, they are the most uncertain part of the forecast. For example, the Congressional Budget Office’s most recent long run outlook shows spending on Medicare and Medicaid, the governments health programs for the old and poor, respectively, rising from 4.1 per cent of GDP in 2007 to 19.1 per cent of GDP in 2082.1 By contrast, Social Security benefits (the government’s main old-age pension program) increase only 2 percentage points, from 4.3 per cent of GDP in 2007 to 6.4 per cent in 2082. Another analysis by CBO suggests that an 80 per cent confidence band around the Social Security projection would be from 51⁄2 to 91⁄2 per cent of GDP.2 CBO did not present similar calculations for health spending; instead, they projected health spending under three different assumptions about the rate of growth of age-adjusted health care spending in excess of per capita income. Their projections show health spending ranging from 7 to nearly 40 per cent of GDP by 2082.
By comparison, defense spending as a percent of GDP peaked at 42% of GDP in World War II. A model that presents as a possible outcome that the US will devote nearly 40% of GDP to health care spending a long-term, sustained outcome, is ludicrous on its face. The CBO assuming public health care spending will sustain its growth rate of the last 50 years for as long as they do (see further discussion below) with no policy changes is like budget analysts in 1946 assuming that military spending will grow at the same rate it did during World War II without any policy changes. Yet they further assume that, having reached this crushing level, Medicare costs in 2082 will still be growing faster than GDP!
The underlying issue is that nothing that is a large portion of GDP can exceed the growth rate of GDP forever, or even for all that long; that’s how we’ve gotten in the insane position of having health care reach 16% of GDP. The term of art is “excess health care spending growth” which as noted above, they define in relationship to per capita incomes. The Fed economists make the following observations in their paper:
1. Given the concern about health care cost escalation, and the pressure being exerted now by employers to contain these costs, as well as the fact that other advanced economies have shown declines in excess costs, the growth assumptions look very aggressive. Moreover, the excess growth took place during a period when government and private health care insurance expanded greatly. In 1960, out-of-pocket spending was 52% of medical spending; by 2006, it had fallen to only 13% by 2006.
Follette and Shiener also looked at the composition of cost drivers and argued that certain key ones will moderate:
Accordingly, 1.1 percentage points of the [historical] 2.2 percentage points of age-adjusted excess growth over the period resulted from technological change and other factors and 1.1 percentage points reflected the effects increased insurance coverage and administrative costs.16 The historical data therefore support excess growth of 1.1 per cent per year if we assume that out-of-pocket costs do not decline further and that administrative costs (as a share of expenditures) do not rise further. However, demand should fall short of this as consumers respond to rising health bills.
The Fed economists separately find that excess growth has averaged 2% over the last 40 years but has been slowing and argue that 1% excess growth is a likely upper bound for the long term average. They posit that excess growth of 2% can be maintained for only a few years at most because consumption as a percentage of GDP is anticipated to fall (this is in the CBO’s own models; it’s mainly the result of the trade deficit falling). By contrast, the CBO assumes 2.4% excess growth for Medicare and 2.2% for Medicaid over the next decade, falling monotonically to 1.1% for Medicare and 0% for Medicare by 2082. If you’ve ever run financial models, you’ll know that goosing the growth rates in the early years has an impressive impact on the final result.
2. The CBO model produces the peculiar result that government funded plans will show faster cost growth than private plans. From the article:
Our chief concern with their projection is their assumption that per capita spending by Medicare grows much more rapidly than that of the private sector. The projected divergence seems inconsistent with the underlying assumption that policies are unchanged because Medicare and private sector insurance plans have had similar payment rates historically.
3. The Fed budget experts note that the CBO analysis violates the requirements for CBO budget projections. The CBO is tasked to forecast assuming no policy changes. But in simply relying on historical trends, which as noted above include considerable expansion of government-funded health coverage, they have effectively incorporated the hidden assumption of continued expansion. Put it another way, the forecasts should have explicitly backed out the impact of historical increases in health insurance coverage to arrive at a true baseline growth rate. Per Follette and Sheiner:
The lack of distinction between policy and other factors is a particular problem because CBO uses different excess cost growth assumptions for Medicare, Medicaid, and other health spending, and the CBO projection is supposed to be under the assumption of no policy changes. Past Medicare spending growth includes factors that are not assumed to continue in the future. For example, the Medicare Part B premium was not previously indexed to Medicare spending; thus Medicare spending growth grew faster than overall health spending, as Medicare picked up a higher share of spending. Similarly, Medicare policies changed to include renal dialysis, HMOs, coverage of the SSI population, and more broadened coverage of home health care. As CBO is not assuming further expansion of Medicare, it does not seem reasonable to forecast future growth based on historical growth rates that include such expansions. Similar issues arise with Medicaid.
4. Follette and Sheiner find that, contrary to conventional wisdom, that more realistic health care cost assumptions allow for some improvements in coverage to low income groups, provided government coverage to the better off is not increased further:
We find that a narrow expansion of assistance to the lowest quintile would have only minor consequences to government finances but more broad-based programs would have materially deleterious effects.
* * *
The CBO’s performance on this front looks like malpractice. The Fed economists note telling irregularities, such as the substitution of scenarios, as opposed to the use of confidence band analysis, as the CBO employed in its Social Security forecasts. And this would not the first time that CBO has apparently allowed political considerations to interfere with its pretense of objectivity. First we have the case of CBO analyst Lan Pham, who was fired for attempting to incorporate the impact of foreclosures and chain of title issues on home price and property tax forecasts. Second, we have the instance of Tom Ferguson and Rob Johnson of alerting the CBO to a significant omission in their deficit analysis, that of failing to include financial assets in their debt-to-GDP ratio calculation. CBO staffers have not disputed the accuracy of the Ferguson/Johnson research but nevertheless will not change their projections. Now we have what is demonstrably an overly aggressive set of assumptions driving health policy debate, with two Federal Reserve analysts sufficiently taken aback by the model as to publish a serious takedown of it.
The CBO’s independence is, like its output, treated as above question. It’s time to subject both to harsh scrutiny.