A new set of projections released last week by Medicare's actuaries has drawn much attention, in part because it suggests the deceleration in the growth of health costs we've seen over the past few years is ephemeral. The actuaries attribute the slowdown to the "lingering effects of the economic downturn and sluggish recovery" and to increases in cost sharing.
Both of these explanations have serious shortcomings -- and that, in turn, suggests something larger is in fact at work.
The assertion that economic sluggishness is playing a dominant role is based on an econometric analysis that links past slow growth in gross domestic product to current aggregate health spending. However, such a macroeconomic explanation is difficult to reconcile with the slowdown in Medicare, because there is no reason to expect Medicare -- a subsidy that goes mainly to retired people -- to be much affected by the economy.
What's more, the econometric methodology itself, which I have long disliked, was debunked in another paper released last week. Amitabh Chandra and Jonathan Holmes of Harvard University and Jonathan Skinner of Dartmouth College found that when they made even modest changes in this type of model, the results changed drastically. They concluded, "We are reluctant to make much of the time-series evidence between GDP growth and healthcare spending."
A simple rule of thumb is that econometric results are untrustworthy if small specification changes -- such as shifting the time period by a couple of years -- substantially alter the results. (The same critique applies to an earlier analysis of this ilk, also much cited in the news media, from the Kaiser Family Foundation. The report is unfortunately less scientific than it pretended to be.)