Tuesday, July 16, 2013

This blog was inspired by a post and follow-up data from Gene Steuerle, and informed by discussions with Tom Getzen.  The three of us will be elaborating on this general topic at our Robert Wood Johnson Foundation-funded sustainable health spending symposium on July 30th in Washington DC.  Hope to see you there!

The historical growth in health spending as a share of GDP is well-documented and increasingly well-known, having increased from about 6 percent in 1965 to 18 percent in 2012.  We know this share will eventually level off – but where?  I begin by reviewing the historical trend.

As shown in the first chart, the historical trend since 1965 is roughly on a linear path, increasing by 4 percentage points each 15 years (see, for example, 1975 to 1990 and 1990 to 2005).  At this rate, the share would reach 20 percent in 2020, 28 percent in 2050, and 40 percent by 2095.  And yes, health spending would reach 100 percent of GDP in about 310 years!  OK – we obviously don’t believe this trend will continue for the next 300 years but the chart provides no clue about a leveling-off.  What macro variable might we examine for clues as to the longer term path of this line and its possible leveling-off point?

Graph of national health expenditure share of GDP

The Health Spending Share of the Growth in GDP

Gene Steuerle has long been a proponent of looking at the per capita growth in health spending as a share of the per capita growth in GDP (a marginal approach).  His blog presents a variety of historical statistics including the chart below.  The health share of GDP growth rose during the first three decades beginning in 1960, dropped during the managed care decade of the 1990s, and jumped dramatically during the most recent decade.  Between 2000 and 2010, health spending accounted for a whopping 43 percent of total and 87 percent of per capita GDP growth!

  Graph of health spending as a share of total and per capita GDP growth

Reprinted with permission from Gene Steuerle’s May 2013 blog.

It is difficult to discern any long-term pattern in these shares, particularly with the sharp jump from 2000 to 2010.  As Gene notes, this jump is influenced by the recession which affected GDP growth much more than health spending growth.1  Perhaps we could learn more about the long-term pattern if we eliminated the noise introduced by business cycles, especially the 2007-2009 recession.

Removing the Effects of Business Cycles

The next chart plots the health spending share of real per capita GDP growth for various time periods between 1965 and 2011, after removing the effects of business cycles (the Appendix Chart presents annual data).2  I refer to the health spending share of real per capita GDP growth as the “marginal” share.  The chart shows that this marginal share increased steadily between 1965 and 1990, from 13 percent for 1965-70 to 29 percent for 1985-90.  The 1990-2005 period encompasses the managed care era of very slow health spending growth (roughly 1994-2000), and three years of backlash featuring very high spending growth.  Over the total 15-year period, the marginal share is 27 percent.  From 2005 through 2011, the marginal share is 31 percent.

Graph of health spending share of GDP growth

Source:  Altarum Center for Sustainable Health Spending

Once business cycle effects have been removed, and managed care and its backlash are averaged together, we see a very interesting pattern.  The marginal share grows steadily between 1965 and 1990 and then levels off at about 30 percent (a mathematical asymptote).  While this may not represent the equivalent of Plank’s constant, it is astounding to find a key macro-level health spending statistic that has been relatively stable for over 20 years!  Obviously, there is no guarantee that it will persist long-term, but it is certainly mathematically possible.3  It seems unlikely to decrease in the next couple of decades, given our rapidly aging population.  But it also seems unlikely to increase, given the unrelenting pressures to control spending.4  Thus, 30 percent appears to be a reasonable baseline assumption for longer-term forecasts of health spending growth around which alternative assumptions could be investigated.

What If the 30 Percent Marginal Share Persists? 

Let us assume that health spending continues to consume 30 percent of the growth in real per capita GDP.  Intuition should tell you that, over time, this will cause the health spending share of GDP to grow from its current 18 percent toward 30 percent.  The faster is the growth in real per capita GDP, the faster the growth in the health spending share of GDP toward its upper limit of 30 percent.5

The next chart reproduces the historical health spending share of GDP and its linear trend line through 2012 (shown above), and then projects forward to the year 2100 under three alternative growth rates in real per capita GDP.  The projections confirm that greater economic growth accelerates the growth in the health spending share of GDP.  The middle rate of 1.5 percent is similar to what the Congressional Budget Office (CBO) projected for the latter half of their 10-year forecast.  If this rate persists, the health spending share of GDP would reach 20 percent in 2025 and 25 percent in 2070.  At the 2 percent growth rate, these milestones would be met sooner (2022 and 2056); while at 1 percent growth they would be met later (2031 and 2100).  The projections illustrate how the roughly linear historical trend eventually begins to bend downward toward a long-run upper limit of 30 percent.

   Graph of projected health spending share of GDP growth at 30%

Source:  Altarum Center for Sustainable Health Spending

Income Elasticity and Excess Growth 

The “income elasticity” of health spending and rate of “excess growth” in health spending are statistics of great interest to health economists.  The income elasticity refers to the percent increase in real per capita health spending associated with a one percent increase in real per capita GDP, and excess spending refers to health spending growth minus GDP growth.  Here are the formulas (via Steuerle) linking these concepts to the marginal share:

  1. Income Elasticity = marginal share / cumulative share
  2. Excess Growth = (Income Elasticity – 1) * GDP growth rate (real per capita)

Under our assumption of a constant 30 percent marginal share, the income elasticity declines steadily over time due to growth in the cumulative share.  Consider the 1.5 percent growth projection line from the chart above.  In 2013 the cumulative share is 18 percent so the elasticity is 1.7 and excess growth is 1 percent, i.e., we are at GDP+1.6  By 2025 the elasticity has fallen to 1.6 and excess growth to 0.8 percent and by 2050 these figures are 1.3 and 0.5 percent respectively.  In the long run, the elasticity approaches 1.0 and excess growth approaches 0.0.

As shown in formula (2), excess growth depends upon both the income elasticity and economic growth.7  Over the past 25 years of a stable marginal share and growing cumulative share, the income elasticity has declined, and this explains part of the decline in excess growth observed over that period.  The underlying trend in real per capita GDP growth began dropping in about 2004, resulting in additional reductions to excess growth in recent years and contributing to the pre-recession slowdown.8


Given the focus of the Center for Sustainable Health Spending, I should address the sustainability of a constant 30 percent marginal share.  It is certainly sustainable in the sense that it could, mathematically, go on forever.  But the more interesting question is whether it is sustainable in terms of federal government financing requirements.9  In previous work, we have used 2035 as a target year for a balanced federal budget.  Between now and 2035, if the marginal share remains at 30 percent and real per capita GDP growth averages 1.5 percent, excess growth will slow from 1.0 percent to 0.6 percent, averaging about 0.8 percent.  According to our Triangle of Painful Choices, tax revenues would have to increase to more than 22 percent of GDP for a balanced budget even if defense and other non-health federal spending (other than social security) were kept to historically low shares of GDP. 

If you have managed to follow all of this, you might be thinking we should hope for slower growth in GDP in order to reduce excess growth in health spending.  Of course this turns out to be something we should not hope for – but there’s no space to address that here.  Come to our symposium where we can delve more deeply into this and other related issues!


Graph of health spending, real per capita without business cycles

Source:  Altarum Center for Sustainable Health Spending


1. Gene finds a much more stable pattern in recent years by averaging over longer periods, thereby diluting the business cycle effects (see the last chart in his blog).

2. I eliminated business cycle effects from GDP using “potential” GDP, and from health spending using a variant of the model described here.  More specifically, to purge GDP of business cycles, I began with real potential GDP (PGDP) estimates from the February 2013 Congressional Budget Office (CBO) report.  This represents CBO’s estimate of what real GDP would have been each year if the nation was at full employment.  Even this series has some significant business cycle-related noise so I used an 11 year centered moving average to provide further smoothing.  For health spending, I regressed real per capita health spending on smoothed real per capita PGDP along with a set of business cycle variables (current year and five lagged years) representing the difference between GDP growth and smoothed PGDP growth.  I used this model to estimate, for each year, the impact of business cycles on real per capita health spending growth.  I then created a health spending series in which these business cycle effects were eliminated.  Details are available upon request.

3.This mathematical possibility contrasts with the commonly-used statistic – growth in excess of GDP – which cannot remain positive forever.

4. Expanded coverage under the Affordable Care Act will almost certainly drive the marginal share well above 30 percent during the years of implementation.  But this is a very short-term effect with a minimal impact on the longer-term trend.

5. The overall share is a mixture of the historical share (18 percent now) and the accumulated future share (30%), with the weights depending upon the cumulative increase in real per capita GDP.  Note that if there was no growth in real per capita GDP, the health spending share of GDP would remain constant.

6. The elasticity is .30/.18 = 1.7.  Excess growth is (1.7 – 1)*1.5 percent = 1 percent.

7. We ignore the short-term effects of business cycles; think of real per capita GDP growing along its long term path in this analysis.

8. The trend in smoothed real per capita potential GDP averaged 1.9 percent from 1985 through 2004 and 1.3 percent from 2005 through 2011.

9. This is the definition of sustainability that our Center has developed.

All postings to the Health Policy Forum (whether from employees or those outside the Institute) represent the views of the individual authors and/or organizations and do not necessarily represent the position, interests, strategy, or opinions of Altarum Institute. Altarum is a nonprofit, nonpartisan organization. No posting should be considered an endorsement by Altarum of individual candidates, political parties, opinions or policy positions.