Quick, answer this without thinking: which is better for your health, living under a communist regime or a capitalist system?
If you said “communist”, may I wager you’re an academic? Now read this, the first part of an opening sentence to a peer-reviewed paper in The Lancet:
The transition from communism to capitalism in Europe and central Asia during the early to mid-1990s has had devastating consequences for health:
The paper is “Mass privatisation and the post-communist mortality crisis: a cross-national analysis” by David Stuckler, Lawrence King, and Martin McKee (DOI:10.1016/S0140-6736(09)60005-2). Note carefully the word “crisis.” The The University of Oxford’s PR department summarizes the work:
As many as one million working-age men died due to the economic shock of mass privatisation policies followed by post-communist countries in the 1990s,
The gist is that the “shock treatment” of switching from a system where government controlled everything and where “everybody” was “employed”, to one of (more or less) freedom caused the (indirect) slaughter of a whole bunch of folks.
Stuckler and co. relied partly on a data source about which other authors say1 “This series is largely free from the distortions introduced to the published data during the Soviet period to disguise mortality from cholera, plague, suicide, homicide, and work accidents.” “Largely free” is not free; therefore any subsequent analysis which uses this data without considering the error and uncertainty inherent in the data, will itself be too sure of itself. Stuckler did not account for this uncertainty.
In order to create the paper’s stunningly counter-intuitive findings, besides the iffy data, the authors relied on a complex statistical model, which was composed of certain assumptions. One was inclusion of the “European Bank for Reconstruction and Development (EBRD) small-scale privatisation index”, a scale from 1 to 4.3, where e.g. 3 meant “Comprehensive programme almost ready for implementation” and 4 was “Complete privatisation of small companies with tradeable ownership rights.”
They also used the “EBRD large-scale privatisation index”, which is also subjectively defined. Then their own “Mass privatisation” indicator, which was positive when a “Country implemented a programme that transferred the ownership of at least 25% of large state-owned enterprises to the private sector through vouchers and give-aways to firm insiders.”
Amusingly, not only do they assume the data sacrosanct, they treat the “Mass privatisation” indicator and other scales as being unambiguous, definitions which every human being would agree to and which are measured without error in each of several countries2. Their model was this:
AMR is “logged adult male standardised mortality rates” and where, for example, “DEM is the democratisation index, WAR is a dummy for military conflict, EDUC is the percentage of population with tertiary education, URBAN is the percentage of the population living in urban settings, DEP is the population dependency ratio”3. Statisticians call this kind of model “the kitchen sink”, whereby everything a researcher can think of is thrown in with the hope that something clogs the drain.
Things stuck in this one, in the sense Stuckler saw wee p-values pop out of his equation for PRIV, the privatization measure (the measure is plural; the authors created more than one, but they never adjusted for multiple testing). He also found that unemployment increased with the collapse of communism, and that this unemployment contributed to corpses. This finding depends on the definition of “employment.” Under tyranny, everybody had their name on the roles, near 100% employment; under freedom, people found jobs instead, with unemployment now a non-negative number.
All results are purely statistical. The authors admit “This study has not examined how privatisation and unemployment led to illness.” Well, maybe they don’t.
Yours Truly is not the only skeptic of this curious paper. John Earle and Scott Gehlbach of Upjohn attempted to replicate the study but could not. In a letter to Lancet (same issue), the pair wrote, “We attempted to replicate [Stuckler’s] results and found that the relationship is not robust.” They ran three robustness checks, changing the original paper’s assumptions along plausible lines. They “demonstrate that any one of these changes substantially weakens the positive correlation between privatisation and mortality reported by Stuckler and colleagues, and a combination of any two changes eliminates it entirely.” They also ask, “does privatisation in fact lead to substantial job loss? My coauthors and I have found that the answer is a clear ‘no’.”
Even accepting the data, it is not unlikely the enormous societal upheavals which took place from pre- to post-tyranny could cause premature deaths. (The authors speculate booze and pills took their tolls.) If that’s so, then all this paper proves is that we can chalk up a few more hundred thousand to communism’s already astonishing body count.
Addendum Their conclusion is that “Great caution should be taken when macroeconomic policies seek radically to overhaul the economy without considering potential effects on the population’s health.” I’ll make another bet with you: the trio who penned these words were in favor of Obamacare. Irony, irony everywhere!
1Leon DA, Chenet L, Shkolnikov VM, et al. Huge variation in Russian mortality rates 1984â€“94: artefact, alcohol, or what? Lancet 1997; 350: 383â€“88.
2Don’t laugh. This practice is so common with economists and sociologists that nobody questions it anymore.
3All of which falsely supposed to be measured without error, and similarly across each country (giggle).
Thanks to Sam Schulman (@Sam_Schulman) for suggesting this topic.