It has been a while since I last had the opportunity to analyze the slowdown in health spending and the extent to which it represents a lasting bend in the cost curve, as opposed to lingering effects of the “Great Recession or other temporary changes.” (See Note 1)
Distinguishing Health Care Cost Curves
When we discuss bending the health care cost curve, two questions arise: “Which curve?” and “Short run or long run?” In this post, I focus on the curve represented by the growth rate in national health expenditures (NHE) pre- and post-recession. Other curves of interest include “excess growth” (health spending growth in excess of gross domestic product [GDP] growth) and the closely related health spending share of GDP. For analysis of all three curves over the very long run, including a provocative “big bang” theory about the origins of excess growth, see Tom Getzen’s blog. A fourth curve that has gotten my attention, through the work of Gene Steuerle, is the health spending share of the growth in real per capita GDP. (See Note 2)
I now turn to the present topic, the record low growth in NHE that began in 2009 (the year in which the recession ended) and continued through 2013 (the most recent year for which we have official data). There has been extensive discussion about whether these low rates are the result of temporary cyclical factors, such as the recession, or more permanent structural factors. As detailed below, I conclude that, to a surprisingly large extent, the answer is neither: the bulk of the decline in the health care spending growth rate resulted from lower economy-wide price inflation and some temporary factors not tied to the recession.
What Drove The Record Low Growth In Health Spending From 2009 Through 2013?
Using data from the National Health Expenditure Accounts (NHEA), Exhibit 1 shows that the average growth rate in NHE dropped by 2.6 percentage points, from 6.5 percent to 3.9 percent, comparing the 3 years prior to the recession (2005–2007) with the 5 years following the recession (2009–2013) (See Note 3). It also shows that average economy-wide price inflation dropped by 1.6 percentage points between these two periods and thus accounts for about 60 percent of the drop in the health spending growth rate (See Note 4). The “real” inflation-adjusted growth rate fell by only 1 percentage point.
Which components of NHE contributed to this percentage point decline? Are there insights into whether the slowdown was due primarily to the recession, other temporary factors, or more permanent structural changes?
Exhibit 1. Growth In NHE: Pre- And Post-Recession
From a detailed, line-by-line examination of the NHEAs and changes in growth rates pre-and post-recession, I developed a set of all-inclusive categories to analyze the slowdown. These categories are shown in Exhibit 2, with their shares of total NHE in 2011, the middle of the post-recession period. The two largest categories are health care services, those paid by Medicare, and those paid by some other source. Spending on dental care services is broken out separately and excluded from these two categories to avoid double counting.
The exhibit shows that health care services paid by non-Medicare sources account for more than half of NHE, while the Medicare-paid services add another 16 percent. These large shares are not surprising as this spending includes hospital and physician services. The next largest category is prescription drugs at 10 percent of total health spending. Smaller categories that are broken out for special attention include dental care, public health, and investment in structures. The residual, labeled “Other NHE,” accounts for 15 percent.
Exhibit 2. Shares Of NHE: Middle Of Post-Recession Period (2011)
Exhibit 3 shows, for each category, how much the real growth rate changed between the pre- and post-recession periods. The vertical dashed line represents the change in the real growth rate for total health spending (-1 percent), highlighting those categories whose growth rates fell more rapidly than the overall average.
The motivation for including the smaller categories now becomes obvious, as each showed a much larger slowdown than overall health spending. The real growth rate dropped for spending on structures by 10 percentage points (from 9 percent to -1 percent), for public health by about 3.5 percentage points (from 3.1 percent to -0.4 percent), for prescription drugs by 3 percentage points (from 3.7 percent to 0.7 percent), and for dental by 2.6 percentage points (2.7 percent to 0.1 percent). As will be shown below, these smaller categories combine to account for a disproportionate share of the slowdown.
Results for the health care services categories are enlightening. The growth rate in Medicare spending on health care services (net of dental) fell by a bit more than overall health spending. This is particularly impressive since the rate of growth in Medicare enrollment grew nearly a percentage point faster in the post-recession period than in the pre-recession period. Spending on health care services (net of dental) from other payment sources, on the other hand, actually grew a bit faster in the post-recession period.
Exhibit 3. Decline In Real Growth Rates: Pre-Recession To Post-Recession
Exhibit 4 displays similar data for selected components of health care services, broken out by whether the source of payment was Medicare or other payers. Much of Medicare’s slowdown in spending for health care services can be attributed to a large reduction for home health and nursing home care. (Medicare accounts for about 44 percent of all home health payments and 22 percent of nursing home payments.)
Exhibit 4. Decline In Real Growth Rates: Medicare Versus Other Payers For Selected Services
The impact of each category on the slowdown in overall health spending depends both on its individual change in real growth just shown and on its share of total health spending, both previously shown (See Note 5). The combined results are shown in Exhibit 5.
The largest contributor, accounting for nearly 30 percent of the 1-percentage point slowdown, was prescription drugs. By implication, if the real growth rate of spending on prescription drugs had not fallen, the overall growth in real health spending would have declined by about 0.7 percentage points. The combination of structures, public health, and dental care explain 36 percent of the slowdown. Medicare spending on health care services explains about 15 percent.
Exhibit 5. Percent Contribution To The Slowdown In Real Health Spending: Pre- To Post-Recession
Tying It All Together
The post-recession period exhibited a 2.6 percentage point slowdown in the rate of growth in health spending compared to the pre-recession years, resulting in 5 years of record low growth rates. There has been much debate about how much of the slowdown should be attributed to the recession, which could be reversed with more robust economic growth; and how much is due to structural changes, which would be more permanent. I would add two more factors: economy-wide price inflation and occurrences that cause a temporary change in the growth rate but are not cyclical. Exhibit 6 summarizes the sources of the slowdown and serves as a guide for subsequent discussion.
Exhibit 6. Sources Of The Post-recession Slowdown In Health Spending Growth
Slower economy-wide inflation accounts for more than 60 percent of the slowdown, explaining 1.6 percentage points out of the 2.6-percentage point decline. This leaves 1 percentage point to allocate among the recession, temporary effects that are not cyclical, and structural changes.
The next largest explainer, accounting for about 0.4 percentage points, is the slowdown in the real growth rate in spending on structures, public health, and dental care. These smaller categories can be a bit noisy (temporary noncyclical changes) but should also be somewhat sensitive to the recession. For example, about 85 percent of public health spending is funded by state and local governments who are mostly required to balance their budgets even during recessions, when tax revenues drop substantially.
Slower growth in real spending on prescription drugs accounts for 0.3 of the 1-percentage point post-recession slowdown. Nearly half of this prescription drug slowdown was caused by two factors that are temporary but not cyclical. First, the 2006 introduction of Medicare Part D caused a temporary jump in the growth rate for prescription drugs during the pre-recession period (See Note 6). Second, a series of patent cliffs that occurred in the post-recession period caused a temporary decline in the growth rates.
I estimate that Part D added more than a percentage point to the average annual growth in prescription drug spending during the 2005–2007 period. I suspect that the patent cliff shaved off close to 1 percentage point from the average growth rate during the 2009–2013 period. That would leave a scant 0.2 percentage points of the prescription drug impact to distribute between the recession and structural changes.
Medicare spending on health care services contributed 0.2 percentage points to the slowdown. About three-fourths of this effect is due to lower growth in home health and nursing home spending (See Note 7). I would divide this effect between the categories of structural and temporary noncyclical. If you wish to make your own estimate, I recommend a paper from the Kaiser Family Foundation that provides estimates of the sources of the Medicare spending slowdown.
The residual category of “Other NHE” explains another 0.2 percentage points. But most noteworthy is that spending on health care services by payers other than Medicare, which makes up more than 50 percent of NHE, actually showed a small increased real growth rate in the post-recession period and subtracted about 0.1 percentage points from the slowdown. This is quite surprising because of expectations that the recession would slow such spending, both because of individuals losing insurance coverage but also because of evidence that the recession caused reductions in spending even by those who kept their employer-sponsored insurance.
This all suggests that the vast majority of the post-recession slowdown is attributable to lower economy-wide price inflation along with some temporary noncyclical factors. Of the surprisingly small amount left to be explained by the recession and structural changes, I would attribute the greatest share to the recession.
Thanks to the Robert Wood Johnson Foundation for supporting the continuation of our Center’s work in this area. In addition to blogs, this grant supports our new monthly Health Sector Trend reports, as well as our annual Symposium on Sustainable Health Spending (scheduled for July 21 this year and featuring another A-list lineup).
I previously demonstrated that this share was remarkably stable between 1980 and 2007 (after smoothing for business cycles) at about 25 percent and that a mathematical consequence was a steady decline in the long-run excess growth curve from GDP+4 (percentage points) in 1980 to GDP+1.3 in 2007. In a future post, I will argue that 2007 is likely to mark an upward bend in this particular curve, even as the health expenditure curve I am about to discuss bends downward.
To be clear, these are average compound growth rates describing the increase in spending from 2004 to 2007 and from 2008 to 2013. They are similar to averaging the annual growth rates for the years 2005 through 2007 and 2009 through 2013.
See the previously referenced blog by Getzen for evidence of the impact of economy-wide price inflation on health spending.
The semi-precise formula, which I used in my calculations, is as follows: The contribution of any given category to the change in the rate of growth in overall health spending is equal to the change in the category’s rate of growth times the category’s initial share of total health spending plus the initial category growth rate times the change in the category’s share of health spending. This would be precise if we were comparing single years.
The growth rates in prescription drug spending were 6.4 percent, 9.3 percent, and 5.2 percent for the years 2005, 2006, and 2007, respectively.
I derived this figure from detailed tabulations not shown here.