The Case for Tracking Health Spending as a Share of "Potential" GDP

In my last Altarum Health Policy blog, I discussed how to recognize a bent health care cost curve. This curve was defined in terms of health spending as a share of gross domestic product (GDP), a measure that is closely watched by health economists and policymakers who are concerned with the share of national income needed to fund health care.

I noted the difficulty in judging the underlying trend in the health spending share of GDP from the observed trend because of the turbulence caused by business cycles (the pattern of recurring recessions and recoveries). I ended by proposing a rough rule of thumb for inferring a bent cost curve from data on the health spending share of GDP. But the primary message was that growth in the health spending share of GDP during a recession does not necessarily indicate that health spending is out of control. Conversely, a period of zero growth in the health spending share of GDP during a recovery does not necessarily indicate a bent cost curve.

In this blog, I propose a new statistic that filters out the misleading signals arising from transitory business cycles and provides clearer information on the extent to which health spending is on a sustainable path. The idea is to develop time series estimates of what the health spending share of GDP would be if the nation were to remain at a fixed point in the business cycle. By definition, this eliminates business cycle effects from the statistic and provides a clearer picture of where the health spending share of GDP is headed in the longer term.

A convenient point in the business cycle to use for this statistic is non-inflationary full employment. The level of GDP at this point in the business cycle is referred to as “potential” GDP (PGDP), and there are both historical estimates and forecasts of PGDP available. Actual GDP sometimes falls below PGDP (recessions) or exceeds it (inflationary booms), but these deviations are normally temporary (government countercyclical stabilization policies play a role here). The usual business cycle pattern is for GDP to move about PGDP, but continually return to it.

Since health spending is largely unaffected by the business cycle, we can use observed health spending as our rough estimate of full-employment health spending. Thus, the full-employment health spending share of GDP can be approximated by the ratio of health spending to PGDP.

In the exhibits that follow, our source for GDP is the Bureau of Economic Analysis, while estimates of PGDP come from the Congressional Budget Office. Health spending estimates are derived from our monthly Health Sector Economic Indicators series and are consistent with the annual estimates from the National Health Expenditure Accounts. Our data are quarterly and in constant dollars. Recessions are shown as shaded vertical stripes.

Before comparing trends in health spending as a share of GDP and PGDP, let’s examine historical growth rates of health spending, GDP, and PGDP. These are displayed in the exhibit below. Health spending growth, shown in the solid red line, is quite volatile and seems unaffected by recessions. It falls during the first recession, rises during the second, and shows little change during the third.

PGDP growth rates, shown in the solid purple line, change slowly over time, starting at about 3 percent in 1989 and ending at just below 2 percent in 2010. This primarily reflects trends in labor supply and productivity.

GDP growth rates, shown in the dotted blue line, fall precipitously during recessions and then move above PGDP growth rates during expansions as the economy accelerates back to full employment.

Now let’s look at the health spending shares of GDP and PGDP over time. In the exhibit that follows, the health spending share of PGDP is shown in the solid line while the share of GDP is shown in the dotted line. Thus the solid line represents the “full-employment” trend, while the dotted line represents the same trend but accompanied by the transitory effects of the business cycle. The difference between the two lines isolates the effect of the business cycle.

Consider first health spending as a share of GDP, the statistic that is most commonly tracked and reported (the dotted line). It stairsteps upward with each of the three recessions and shows little or no growth between recessions. Do the upward sloping portions represent excessive growth in health spending or just the impact of the recessions?  Do the periods of leveling off that occurred in 1993, 2003, and 2009 represent a bending of the cost curve or just the impact of economic recoveries?

Now look at health spending as a share of PGDP (the solid line). The rapid increase between 1988 and 1993 cannot be attributed to the recession since business cycle effects are adjusted out. Thus, it must be a period of high growth in health spending (relative to PGDP). There is a clear bend in the curve starting around 1994 and continuing through 2000. This was the managed care era, and I believe it would be appropriate to say that the “cost curve” was (temporarily) bent during this time. The upward slope between 2000 and 2004 represents another period of high spending growth, since it cannot be attributed to the concurrent recession.

Another bend in the curve appears in 2004, and the slow to moderate growth continues through the fourth quarter of 2010. Notice how this contrasts with the health spending share of GDP (dotted line) which jumps by about 2 percentage points during the most recent recession. In the absence of our underlying trend measure (solid line), we would not have known that this jump was due to the recession alone and did not represent excess growth in health spending.

In addition to providing a clearer interpretation of historical underlying trends, this measure of health spending as a share of PGDP also provides a mechanism for setting and monitoring future targets in the underlying trend. This will be a subject of future work in our Center for Sustainable Health Spending that I will cover in a future blog.


Columnist Charles Roehrig is a health economist and director of the Altarum Center for Sustainable Health Spending. His column for the Health Policy Forum considers health economics issues and health spending. Columnists are regular contributors to the Health Policy Forum who pose their own opinions and policy positions in the realm of health care and health policy. As a leading nonprofit health care research and consulting institute dedicated to improving human health, Altarum encourages open discussion and debate about the many challenges in health care today. 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.


Compelling graphs. USe of PGDP is a nice way of dealing with some of the variable lag problems and making a better indicator that gives fewer false signals.
Tom Getzen


Thanks for taking the time to comment! Of course I thought of you when I wrote that health spending is “largely” unaffected by business cycles. In an extended version of this analysis, it would be useful to add detail regarding the lagged effects that you have identified in your research.


Great to see you pick up on the use of Potential GDP. I’m not sure if I shared my thesis work when we first met. I looked at health spending v. potential GDP in the US, Australia, Canada, Finland, France and Germany. You confirm what I saw which is that using potential GDP allows you to isolate the growth of health spending separate from the business cycles. Also all countries grow right through recessions/inflations unless it is a large shock as seen in Finland. I have not published it in a journal article. Let me know if I could get you a copy of my thesis (reference below). Would love to talk more on this topic.

Frogner B.K. (2008) “Long run modeling of health care expenditure growth in industrialized countries,” Doctoral Thesis. Johns Hopkins University: Baltimore, MD.

Best regards,
Bianca Frogner

Thanks for the note Bianca and for the reminder about your thesis work. I’ll have to go back and review it as it has been awhile. In your research, did you encounter any previous work looking at health spending and potential GDP? We will be adding this statistic to our monthly spending brief and I will want to mention other applications.

Hope you are staying active in this area of research.



Add new comment

Click to Print