Healthcare cost growth analysis (1)

As explained in this prior post, I am working on a project to identify fertile opportunity spaces resulting from runaway healthcare costs. Here is the first installment of my investigation, my first steps to understanding the healthcare cost curves. (For details of data sources see the references at the bottom.)

Typical media depiction of healthcare costs

Here is the typical “scary graph” we have all seen in the media. The basic data comes from the US government (1). It shows “runaway cost growth” and more significantly, it shows healthcare costs approaching 20% of GDP. Yikes!

Healthcare costs: scary plot
Real, per-capita Healthcare costs more informative

First of all, it seems clear that if the number of people in the country grow, healthcare costs and GDP will both grow and to the extent that growth comes from more people, that is not relevant to what we want to understand. Second, there is some overall inflation rate, and so $1 in1960 is not at all the same as $1 in 2010. That shows up in the graph above as part of the “growth”. It is easy to back out inflation (2) and put all the numbers into 2010 dollars. And by looking at costs and GDP per capita, we eliminate the impact of the growth in population. Oh, and one more thing. When dealing with these types of quasi-exponential phenomena, semi log graphs are much more useful.

So here is what we get when we make all these adjustments to the basic data to hone in on what we care about. “Real”(ie after inflation) GDP per capita is in red, and “Real” National Health Expenditure per capita is in blue. The vertical axis is in $2010.

Per capita Healthcare costs (real (after inflation)

Healthcare grows faster than GDP over the long term

  1. These two curves lie fairly well on a pair of straight lines (the black “best fit” lines are shown). Because we are using a semi-log plot, this means these data sets fit pretty well (to first order) to two exponential functions. In other words, we can describe GDP and Healthcare per capita as each growing on average at x%/y% per year.
  2. Certainly there is some wiggling around the black fitted lines. But basically, the trends are astonishingly stable over the last 50 years.
  3. The basic problem is that the blue curve (healthcare costs per capita) is steeper (faster growth rate) than the red curve (GDP per capita). Now this did not matter much when healthcare per capita was a small fraction of GDP per capita. But now, and soon even more so, it will start crowding out other things we would like to spend money on. One can debate whether 20% of GDP is “too much”. But 100% of GDP clearly is “too much”.
  4. Of course, you cannot really extrapolate from curves like this to predict the future. But given the amazing stability of these two growth trends, you would have to say that something pretty major is going to have to change if we are to avoid these trends continuing. Small tweaks around the edge like buying drugs from Canada are just not going to do it. The system is clearly programmed to keep growing at pretty much constant rates.

What about all these reports that healthcare growth is slowing / speeding up?

There is certainly a lot of attention paid to the latest annual growth rate of healthcare costs, and whether it is bigger or smaller than the year before. Basically this is all about looking at the wiggles in the overall trend. I think you have to be careful of reading too much into short term wiggles. Often it is “noise”. BUT, if you eyeball the curves above carefully, you can see that the blue curve is actually a bit banana shaped, suggesting somewhat faster growth in the earlier years on the graph. Lets dissect this a bit more by looking more closely at the last 30 years. Here is the same graph, but over a shorter time scale, with the curve of best fit fitted only to data after 1980.

Recent healthcare costs corrected
To me, this looks like a pretty stable exponential growth rate for both Real GDP per Capita and for Real Healthcare Expenditure per Capita. You can see the fitting formulas, which show average growth rates of 3.5% per annum for Real per Capita Healthcare Expenditures, compared to 1.5% per annum growth rates for Real per Capita GDP.

Latest decade is not better

Here is the same set of numbers, but looking only at the most recent decade (and the fit of the data to that decade only).

NHECP last decade
For this latest decade, growth rates are slower for both Real GDP per Capita (0.7% per annum) and Real NHE per Capita (3% per annum). I have seen a lot of articles recently about how healthcare costs were growing slower than in prior decades. Often those articles use that claim as a basis to suggest that somehow we have mastered runaway healthcare costs, or that things are not as bad as we thought, or that recent policy changes have made a positive impact on the growth of healthcare costs.

But if you look at these graphs, it is very clear that the real problem is not just the rate of growth of healthcare costs, but the fact that those costs are growing faster than GDP, thus leading in due course to crowding out of other things we would like to spend our money.

The graph above shows that things are definitely not better in the last decade. It is certainly correct that the rate of growth of Real NHE per capita has dropped from its 30 year average of 3.5% per annum to about 3% per annum. BUT, the rate of growth of Real GDP per Capita has dropped too. And the differential growth rate remains about 2-3%.


  1. The basic problem is this growth rate differential of 2-3% of healthcare costs compared to GDP! (3)
  2. We keep hearing in the news about “bending the curve” of healthcare costs growth. Maybe you can see a bit of that the last few years. But it is accompanied by a similar “bending of the curve of GDP”. I would put some money on there being some common causes. To me, this graph suggests meaningful, long term, curve bending remains in the future.
  3. When pundits talk about healthcare cost growth rates without backing out the effects of inflation, we need to take the conclusions with extreme caution, because changes in inflation rates may mask the underlying changes in real healthcare cost growth.
  4. It seems to me that to be useful, the discussion really needs to be about the differential growth rate between GDP and healthcare costs.

Seems like our “mission” as a society is to find a way to reduce the “real” (after inflation) growth rate of National Healthcare Expenditures Per Capita (NHEPC) by a couple of percent per year. [Assuming that increasing GDP growth substantially (ie a couple of percent beyond its long term trends) is not an option.]

Next installment of this work

So, what exactly is NHEPC, and what parts of it are growing and which are shrinking and where is the apparent low hanging fruit? Read the next installment in this series of posts.

And if you want to get involved or get more details about this project, or get access to later portions of this work that might not make it into this blog:
Sign up to be part of the discussion/information loop (no charges).

Or feel free to add your comments or questions below.





  1. US National Health Expenditure Accounts. Published by the US Department of Health and Human Services. Overview and description available at Centers for Medicaid & Medicare Services. Data downloads available here.
  2. To correct for inflation we used the CPI index (available at Bureau of Labor Statistics). This is a simple way to get a general idea of how inflation correction changes the results. There are more sophisticated approaches to this using more specialized inflation indexes (discussed in the National Health Expenditure Accounts Methodology Paper, 2010 eg. pp35-37 [PDF warning]). If our later work requires that we use exactly the right measure of inflation, we may need to revise this a bit. However for the initial purposes of this post, we think this is a good start.
  3. Whether the differential growth rate is exactly 2% or some slightly different number could certainly be refined. Getting the correct metric of inflation is one example that would change this number somewhat. We will return to this if getting the number exact turns out to matter for the project we are focused on here.


  1. Bob Wilcox says:

    Just started into your series, and I look forward to this useful analysis. I would only recommend adding context to the analysis, in two ways. The healthcare product that we buy today is substantively different than the product we bought in 1960. That difference can justify part of the increased portion of GDP spending. In addition, the % of GDP spent on three other major life purchases – education, housing, transportation- have changed. Comparing the qualitative changes in all four arenas as a function of their increased share of GDP would be revealing.
    An additional footnote: the voters have chosen to avoid the single-payor model. That consumer choice could be separated from the other cost data, since it is unrelated to healthcare and is actually a lifestyle purchasing decision.

    • Hi Bob:

      I agree with this point. The book called “The Cost Disease” by William Baumol goes into more detail about these points and is well worth a read if you have not yet discovered it.

      Thanks for reading my stuff and sharing your thoughts.

  2. I really got into this article. I found it to be interesting as well as packed with unique
    points of view. I wish to read material which makes me think.
    Thanks a lot for writing this great content.

  3. Ronald Baltrunas says:

    Your graph showing GDP and healthcare growth leaves out a very painful reality. GDP and national income are becoming extremely concentrated with the richest sector. If you had a graph of income for the lowest incomes like 50%, 60% you would see healthcare costs growing way ahead. Real income for the middle and lower classes hasn’t grown much at all in the last 20 years while healthcare has climbed considerably.


  1. […] the prior post in this series, I concluded that controlling healthcare cost growth was about reducing the […]

  2. […] fertile opportunity spaces resulting from runaway healthcare costs, see the introduction; the first installment; and second installment of the […]

  3. […] again into the National Healthcare Expenditures, and the subset thereof called “Personal Healthcare Expenditures” which is the relevant […]

  4. […] from the numbers in the NHEA (National Health Expenditure Accounts) analysis I have used in prior posts. This difference is analyzed in detail by Bernard et al (PDF here). The key details are as […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: