• The Cost Disease: Chapters 7-8

    Chapter 7 of The Cost Disease can be summed up with the fact that the elasticity of health care with respect to wealth is greater than one, the hallmark of a luxury good. Translation: we tend to spend a greater proportion of our wealth on health as our wealth increases. This gives rise to charts like the one Baumol shows in the chapter for nations at a point in time:

    We have seen something similar in the U.S. over time:

    However, the U.S. is still an outlier in the sense that we spend more than one would expect considering wealth alone:

    It should be noted that all of the above pertains to spending not cost. The former is price times quantity (p x q), summed over the various services and products in the health sector; it’s an economic measure of the output of health care consumed. The later is a measure of the resources required to produce those services and products, i.e., it’s an economic measure of the inputs to the production of health care. Cost and spending are not the same thing because inputs and outputs are not the same thing.

    Let’s turn to Chapter 8, about which I have little to say. It makes the point that sectors have different rates of change in productivity, due to their mix of inputs. Some inputs are inherently stagnant, like human ingenuity, while others exhibit rapid increases in productivity, like computer resources. Baumol provides several examples of mixtures of stagnant and progressive inputs, including R&D.

    Being a researcher, I can attest to some of the changes in productivity over the last 20 years or so. One thing I’ve observed is that as computational resources have become cheaper, it has been possible to do more work with fewer person-hours input. For instance, I require fewer hours of clerical and programmer support to do the same type of work. This is a replacement of stagnant inputs for progressive ones, a change in the mixture of inputs. To be sure, there may be a limit to the extent the mix can change, but until the limit is reached the cost disease may be ameliorated.

    All posts about the book are under the Cost Disease tag.


    • log (expenditures vs.GDP)/capita data would be far more informative regarding efficiency if it were normalized by some measure of market-basket costs. Does US expend the same approximate # of health care “units”/capita, but each US health care unit has higher cost? This distinction fundamentally informs the issue: inefficient system or intrinsically high cost? Where’s the best bang for the buck: increase efficiency or lower costs? (no doubt both…).

    • But the last graph is linear-linear, while the first two are log-log. This is inconsistent.

      Robert L. Ohsfeldt and John R. Schneider in their book, The Business of Health:The Role of Competition, Markets, and Regulation (Washington, DC:The AEI Press, 2006), p. 8, showed a log-log graph and the U.S. was no longer an outlier.

      I was not able to get to the source document for your first graph, but I suspect that the U.S. is one of the 178 countries in that scattergram. Indeed, I’d almost bet that it’s the dot at the top right end of the line, which means at least one, and likely three, countries are spending more than the U.S. according to the log-log model.

      • The only thing log-log-ing the last graph would do is make the (valid) point more obscure. There’s no law that every chart has to be transformed identically. So, I’m not sure what your point is.

        • They don’t have to be transformed identically but if one specifies a better relationship, then I think that’s the one to use.

          Unfortunately, I misrepresented Ohsfeldt & Schneider, who specified a semilog regression, not log-log, comparing these international relationships for 2000. I just quickly did a scatter for 2009, using 13 developed countries, which I’ve posted at http://tinyurl.com/cskwyt3.

          (When I uploaded the Excel file to Google Docs, the latter could not convert the log charts, so I had to upload it as a picture, but if the print mode is selected it’s legible).

          The lin-lin relationship (chart 2) looks very much like yours in your last chart, although I used only fully developed countries, not Poland or Portugal, or Czech Republic. However, in the lin-log relationship (chart 2) and the log-log (chart 3) the U.S. almost (not quite) falls into line.

          • I’m not taking a position on which relationship is better. I don’t care how close the US dot looks to the line. What’s the estimate of overspending, controlling for wealth? (You may be aware there are complexities in transforming E(log(x)) to E(x) …)

            • But if we’re not prepared to venture to question whether “an increase of $X in GDP per capita explains an increase of $Y in health spending” versus “an increase in $X of GDP explains an increase of Y% of health spending” or “an increase of X% of GDP explains an increase of Y% in health spending” I don’t see how we can begin to proceed to decide whether the U.S. is out of whack at all.

              Discussions many magnitudes more technically complex than ours cannot settle on the the model (e.g. Baltagi & Moscone at .

              In any other context I’d not be so nit-picky, but I am quite sure than in both your graph and your comment, when you write “wealth”, “income” is the correct word.

            • Yes. Is it not clear that none of these graphs are mine?

    • I’m sorry, I thought the first one was from Prof. Baumol’s book and the latter two your own. I look forward to reading the book, as your review convinces me it is very interesting.