One does not normally expect to find a political scientist in the midst of a debate within the pages of the prestigious British medical journal “The Lancet”:http://www.thelancet.com/. Yet that is exactly where “Scott Gehlbach”:http://users.polisci.wisc.edu/gehlbach/research.html of UW-Madison and co-author “John Earle”:http://www.upjohninst.org/staff/earle.html of the Upjohn Institute find themselves at the moment. Here’s the background (from “materials posted at the Upjohn Institute”:http://www.upjohninst.org/mortality/index.html):
bq. Was mass privatization a “crucial determinant” of the increased mortality in postcommunist societies during the 1990s? This claim appears in a recent article [by “David Stuckler”:http://users.ox.ac.uk/~chri3110/, “Lawrence King”:http://www.sociology.cam.ac.uk/contacts/staff/profiles/lking.html, and “Martin McKee”:http://www.lshtm.ac.uk/people/mckee.martin] in the British medical journal Lancet. The article shows a positive correlation between the extent of enterprise privatization and the adult male mortality rate using country-level data for 15 economies of the former Soviet Union.
The results of this study were widely reported in the mass media (see for example “here”:http://www.nytimes.com/2009/01/16/world/europe/16europe.html, “here”:http://www.rferl.org/content/Mass_Privatization_Linked_To_Higher_Death_Rate_In_Postcommunist_Transition/1371049.html, and “here”:http://www.economist.com/opinion/displayStory.cfm?story_id=12972677&source=login_payBarrier), including articles with titles such as “Privatization Killed a Million People in Eastern Europe”:http://news.softpedia.com/news/Privatization-Killed-a-Million-People-in-Eastern-Europe-102173.shtml.
If true, the claims made by the article are profoundly important, both for assigning blame for suffering in the past but also as a guide to potential policy-making in the future. With that in mind, Gehlbach and Earle set out to examine the robustness of the findings in the original article. Their “conclusion”:http://www.upjohninst.org/mortality/index.html:
bq. [Our] analysis shows that the estimated correlation of privatization and mortality in country-level data is not robust to recomputing the mass-privatization measure, to assuming a short lag for economic policies to affect mortality, and to controlling for country-specific mortality trends. Further, in an analysis of the determinants of mortality in Russian regions, the analysis finds no evidence that privatization increased mortality during the early 1990s. Finally, reanalysis of the relationship between privatization and unemployment in postcommunist countries shows that there is little support for the Lancet article’s proposed mechanism by which privatization might have increased mortality.
The full letter to the editor by Gehlbach and Earlecan be found “here”:http://www.upjohninst.org/mortality/mp_summary.pdf. You can get the response of the authors to the Gehlbach and Earle criticque on the “Lancet website”:http://www.thelancet.com/journals/lancet/article/PIIS0140-6736%2810%2960160-2/fulltext (you have to register for access first, but it is free). You can also find the full paper by Gehlbach and Earle “here”:http://www.upjohninst.org/mortality/mp_paper.pdf; more information, including a “press release”:http://www.upjohninst.org/mortality/press_release_1-29-10.html is available “here”:http://www.upjohninst.org/mortality/index.html.
A few quick comments in response. It seems clear that the claims in the original Lancet article are not robust to the re-specifications proposed by Gehlbach and Earle. Thus the debate shifts to the question of whether these are “legitimate” robustness tests, and here there seem to be two big questions in play (and I would welcome comments from readers of the Monkey Cage on both of these points) leaving aside some coding disagreements between the two sets of authors. First, Stuckler, King, and McKee argue that in “in a situation where mortality rates were undergoing fluctuations that were unprecedented in a peacetime era” it is inappropriate to add control variables that attempt to pick up other aspects of mortality time trends across the different countries. Gehlbach and Earle argue that the original specification assumes these trends are constant across all countries, and then present data to show that this was not in fact the case. Second, Stuckler, King, and McKee argue that the best specification for testing a relationship between privatization and unemployment is to use measures from the same period of time, justifying this on the grounds that “given evidence that workers’ stress rose in anticipation of privatisation, adverse causal effects could have occurred in the period before privatisation”. Thus, they argue, the findings do not need to be robust to re-estimation with one and two year lags in the effect of mass privatization.
[In the interest of full disclosure, Gehlbach is a friend of mine and I had the opportunity to see his research presented while in progress at an academic conference this summer.]
*****
Update: I originally posted this without the link to the full Gehlbach and Earle paper. This has now been corrected in the text above, and is also available “here”:http://www.upjohninst.org/mortality/mp_paper.pdf.