May 7th, 2013
Much has been made of the slowdown in health spending growth and the role played by the economy. I have to confess that my first take, after studying plots of business cycles and health spending, was that health spending “had a mind of its own” and paid no attention to business cycles. Consider the two most recent recessions depicted in the chart below. During the recession of 2001, health spending growth actually shot up at the same time that the growth in gross domestic product (GDP) was dropping, and continued to rise even after the recession officially ended.
During the Great Recession, spanning December 2007 through June 2009, the growth in health spending dropped by about 2 percentage points and then leveled off while GDP growth dropped by nearly 10 percentage points and then quickly rebounded to a more normal long run rate of growth (though not sufficient to make a large dent in unemployment). I hope you can see why I was skeptical of a predictable relationship.
Recently, I have had the good fortune to work with Larry Levitt and Gary Claxton of the Kaiser Family Foundation, and Temple University’s Tom Getzen (Note 1) on a statistical model that relates health spending to business cycles. So much for my initial impressions; we found convincing statistical evidence that business cycles do impact health spending (something Tom already knew well). However, the effects are incremental and spread over a period of years (six in our model), so they are not easily discerned visually.
In the remainder of this blog, I will provide some intuition behind estimates of the recession’s impact and then relate the findings to two papers in the May 2013 issue of Health Affairs.
Some Intuition on the Statistical Findings
It’s always a good idea to check whether statistical findings make intuitive sense – particularly when the relationships are complex and difficult to “see” directly from the data. Toward that end, I examine the recent recession. My second chart plots GDP and potential GDP (PGDP) in constant 2005 dollars from January 2006 through January 2013. (Note 2) PGDP can be thought of as the long run growth path of GDP if the nation was always at full employment.
Between mid-2008 and mid-2009, the recession knocked GDP about 8 percent below its full employment level. Since then, GDP has returned to a growth path that is slightly steeper than the long-term full employment growth path represented by PGDP. The limited narrowing of this gap represents the painfully slow recovery from the recession. (Note 3)
If health spending adjusted immediately and fully to the recession, by the middle of 2009 we would expect it to have fallen (at least) 8 percent below the path it was on at the beginning of 2008. (Note 4) This clearly did not happen. Our model suggests that the adjustment takes place fairly smoothly over a six year time frame. (Note 5) An 8 percent adjustment, spread over six years, amounts to 1.3 percent per year. Thus, the pure effect of the recession would be to reduce health spending by (at least) 1.3 percent per year for a period of six years. (Note 6)
This all seems plausible to me. Health spending is insulated from the recession by insurance but, in the absence of a recovery, it would eventually adjust to fit the lower “new normal” level of GDP.
Interpreting the Record Low Growth Rates in Health Spending Since 2009
In June 2009, when the recession officially ended, the health spending share of GDP had risen to 18 percent, and it has hovered around that figure ever since. Thus, the rate of growth in health spending in the almost four years since mid-2009 has been GDP+0! Furthermore, over this period, GDP has grown at roughly the same rate as (smoothed) PGDP, which means it is also at (smoothed) PGPD+0. (Note 7) While even this low growth is not sufficient to solve the U.S. budget crisis, it would be a celebrated achievement if it could be maintained long-term. So are we on a long-run GDP+0 spending path?
The preceding analysis suggests that the pure recession effect would be pushing health spending down by at least 1.3 percent per year from 2009 – 2014. This suggests that in the absence of the recession, instead of GDP+0 we would be at GDP+1.3 or higher. But this is an oversimplification because health spending in 2009 and 2010 is also being influenced by fairly high GDP growth rates from 2004 through 2006 (years within the six-year time frame), which push spending up. That is, while the Great Recession impact is likely to be the dominant economic factor affecting health spending in 2009 – 2012, there are other economic effects as well.
This next chart, taken from our study, presents estimates of how much health spending growth would exceed long-term economic growth if health spending is adjusted to remove the effects of business cycles and inflation. (Note 8) Let’s focus on the most recent four years (in green) in which health spending has been at long term GDP+0. When the effects of the business cycles and inflation are removed, we estimate that the underlying path of health spending averages GDP+1.3. (Note 9) Put another way, the roughly 4 percent rate of growth in health spending observed in 2009-12 would be well over 5 percent if not for the weak economy.
Two papers in the newly released May issue of Health Affairs (Ryu, Chernew et al., and Cutler and Sahni) provide fresh perspectives on the recent low growth rates in health spending. Both papers present evidence that the slowdown has been greater than what can be explained by the recent recession and, therefore, that a strengthening economy will not mean a return to previous high rates of health spending growth.
Cutler and Sahni estimate that the 2007-09 recession pushed health spending growth rates down by about one percentage point and that this accounted for 37 percent of the slowdown in health spending between 2003 and 2012. (Note 10) They attribute the rest of the slowdown to a decline in private insurance coverage, cuts in certain Medicare payment rates, and “a host of fundamental changes — including less rapid development of imaging technology and new pharmaceuticals, increased patient cost sharing, and greater provider efficiency”. Presuming that most of these changes were not driven by the recession, they would likely persist during an economic recovery.
Ryu et al. focus on health spending patterns for individuals who were continuously covered by employer-sponsored insurance from 2007 through 2011. One would expect these individuals to be fairly insulated from the effects of the recession and, therefore, a slowdown in their spending could be evidence of more permanent structural changes. They do find a significant slowdown, even after holding benefit generosity constant, and express “cautious optimism that the slowdown in the growth of health spending may persist”.
Comments and Conclusions
Despite the emphasis on part of the slowdown not attributable to the recession, I don’t think either of these studies suggests that spending growth is likely to remain at the 4 percent levels seen over the past four years. (Note 11) Some portion of the slowdown is permanent but some will be given back during a recovery. There is likely some disagreement with our work as to the exact portions, but it is not possible to be more precise than that since we all deal in different definitions and time periods.
When discussing the slowdown, it is important to track spending growth relative to long term economic growth rather than simply on its own. In the long run, this is what governs the share of GDP going to health spending and it is this share that drives the long term budget woes of the federal government. From 2001 to 2003, the underlying real growth rate in the economy was about 3 percent, while in 2009-2012 it had fallen to about 2 percent and it is projected to remain there. Thus, over the past decade, health spending growth has been chasing a moving economic growth target, needing to fall by one percent just to keep pace.
With this background, let’s take another look at the third chart above. After eliminating the noise introduced by business cycles and inflation, health spending growth, relative to long-run GDP, looks similar between 2009-2012 and 2004-2007. Each period averages a bit more than GDP+1. (Note 12) It does not appear that the curve has bent any further since the pre-recession bend. (Note 13) There is also no indication that excess growth is moving back to the GDP+3 levels experienced in 1965-1994 and 2001-2003. (Note 14)
Our paper forecasted how spending growth would look under two scenarios: GDP+0 and GDP+1.6. This brackets the current GDP+1 trend. Interestingly, we have been near GDP+1 from 1994 to present with the exception of the three “managed-care-backlash” years of 2001-2003. Perhaps we should only claim one bend in the curve, occurring in 1994, with a three year interruption.
The forecasts we presented in our study were hypothetical in that they assume historical relationships between the economy and health spending persist into the future. They did not include the impact of ACA policies including the one-time bump in spending growth due to expanded coverage and the provisions (such as smaller provider payment increases) to constrain the growth in Medicare spending. They also do not take into account the unprecedented pressures to control health spending in the coming decade due to current levels of federal debt and the aging of the population into eligibility for Social Security and Medicare. In other words, even with the economic relationships we uncovered that would predict renewed excess health care cost growth once robust economic growth resumed, there will be unprecedented pressures, especially on public expenditures, to constrain excess growth.
Note 1. Tom chairs our Center’s National Advisory Committee and has been studying the impact of the economy on health spending for many years. For an example of his work see: Forecasting health expenditures: short, medium, and long (long) term. Journal of Health Care Finance. 200; 26: 56-72.
Note 2. This is a first approximation since there is some impact of business cycles on PGDP, and I have begun using smoothed PGDP to represent the long term trend in economic growth. This chart shows actual PGDP as estimated by the Congressional Budget Office (CBO).
Note 3. The average gap in 2012 was about 6 percent and recent projections by CBO suggest it will narrow only to 5 percent by 2014.
Note 4. A one percent change in the long-term real growth rate of GDP is expected to cause at least a one percent change in real health spending (in economic terms, a “superior” good). This is consistent with health spending rising faster than GDP over the long-run. There is a question about whether the excess growth in health spending is a fixed amount (GDP+X) or is proportional to GDP (GDP*[1+X]). This is a critical subject that I hope to address in a future post. The statistical results in our model favor the proportional formulation.
Note 5. Interestingly, the largest estimated impact occurs in the fourth year following the recession.
Note 6. I say “pure” because I don’t want to confuse the impact of the recession with the impact of the recovery (or lack thereof). The pure recession impact refers to a case where there is a six year gap before there is any recovery. In the chart above, this would be the same as holding the gap to 8 percent through 2015.
Note 7. I use an eleven year moving average of real PGDP to provide a smoother and, to me, more accurate measure of the longer term underlying trend in real economic growth.
Note 8. The chart presents estimates for the period 1965 – 2012. While the official CMS health spending data for 2012 won’t be available until next January, we include it in our analysis using the estimate from the Altarum Health Sector Economic IndicatorsSM.
Note 9. The match between this GDP+1.3 and that in the previous paragraph is purely coincidental. I’m guessing the pure recession effect is bigger than 1.3 and other factors lowered our modeled effect to 1.3.
Note 10. I suspect this figure will be contrasted with our finding that 77 percent of the slowdown between the period 2001-2003 and 2008-2012 was explained by patterns of economic growth and inflation. This comparison is not apples-to-apples for two main reasons. First, the definitions of the “slowdown” are very different, as we compare the current rate with the historical rate and Cutler and Sahni examine how the current rate differs from a previous government projection. Second, our study includes the impact of business cycles and inflation throughout the period, while Cutler and Sahni’s estimate is limited to the impact of the recent recession.
Note 11. The Cutler paper estimates that the recession has subtracted a percent or so from the growth rate and presumably this will be given back as the economy recovers. And at least some of the drop in spending growth seen for those with continuous private insurance coverage could be attributed to the pressure that a weak economy places on employers and insurers to control spending.
Note 12. I’m using GDP here to represent the underlying long term growth rate rather than the actual GDP growth rate.
Note 13. In a 2012 article, we identified 2005 as the year of the pre-recession bend. That was based upon comparing health spending growth to PGDP growth without accounting for the economic effects described here.
Note 14. The GDP+2 estimate for 2012 (shown in the last chart) is preliminary and will undergo adjustments in July and then again in January.Email This Post Print This Post
Don't miss the insightful policy recommendations and thought-provoking research findings published in Health Affairs.