Excess Famine Deaths: Once Again Into the Breach!
As obsessive readers of this blog know, roughly once a year someone publishes a professional article estimating excess famine deaths and I provide some commentary. About half the time the author(s) then write me angry emails protesting that I have slandered their work. Thus far, no death threats. (Think of it, we could have a new category: excess deaths due to blogging. “We estimate that in 2015, one adult American male suffered an untimely death due to blogging…”).
Anyway, Saebom Jeon, Seong Eun Kim, and Yousung Park have published a paper, “A Study on the Population Structure of Democratic People’s Republic of Korea,” in Communications for Statistical Applications and Methods, the official journal of the Korean Statistical Society and Korean International Statistical Society. Their paper has been published together with commentaries by Daniel Goodkind, Keewhan Kim, Myung Jin Hwang, and a rejoinder by the authors.
The paper is similar to previously published work by Goodkind, West, and Johnson, and by Spoorenberg and Schwekendiek in which the 1993 and 2008 population censuses are used to infer excess deaths during the famine period. Like the aforementioned studies, a critical issue is how to allocate deaths during the post-famine period (alternatively, what to assume about the rapidity of the increase in life expectancy following the famine). The studies also have to contend with the “missing male” problem of the 1993 census due to the apparent exclusion of the KPA from the figures, as well as the implications of ignoring emigration.
These papers have generated estimates of famine deaths that are lower than contemporaneous estimates, or the “second wave” of studies that came out before the 2008 census was released. Goodkind et al. generated an estimate of 500,000-600,000 deaths; Spoorenberg and Schwekendiek produced an even lower range of 240,000-420,000, but this estimate has been criticized (including by Goodkind and Jeon, Kim, and Park, for implying an implausible three year drop in life expectancy after 2000).
Jeon, Kim, and Park, using somewhat different assumptions and slightly different demographic models than Goodkind et al., estimate that there were 489,972 to 574,306 excess deaths during the famine period. They also generate the unusual result that female death rates were significantly higher than death rates for males. This finding cuts against the presumption of a female “demographic advantage” that has been observed in past famines. Jeon et al. argue that this result is possibly due to males in the KPA being relatively protected from the ravages of the famine while older females were uniquely exposed to the collapse of the state-supported safety net. These arguments don’t strike me as implausible. Yet these studies don’t account for emigration, and if emigrants were disproportionately female, then some of the disparity may be explained by emigration wrongly attributed to death. (The existence of emigration means that on their own terms, excess deaths in all of these studies have been overestimated by an unknown degree.) Or it may be the case that with so many assumptions piled on to shaky data, the result may be spurious. My guess is that we’ll have to wait for the opening of the official DPRK archive to ever know for sure. I’m not holding my breath.
In their rejoinder, Jeon et al. express annoyance with Dr. Patrick Gerland of the United Nations Population Bureau for apparently having the temerity to express skepticism about reliability of the data. We’ve been through this argument what seems like a thousand times so I’ll be short and sweet: many of us who come to this work via interest in North Korea find the professional demographers excessively credulous when it comes to DPRK official data. Let me cite one example, I could cite a dozen: during the famine period it was necessary to have a population estimate to calculate the grain balance—and hence how much aid North Korea needed. In 1997 the government was claiming that the mid-year 1998 population would be 23.5 million. (By point of comparison, South Korea estimated that it would be 21.9 million.) By June 1999, however, famine mortality was of such a magnitude that it was germane to calculate overall demand and when pressed by the WFP, North Korea provided a population estimate of 22.55 million for August 1999—amounting to a downward revision of nearly one million people from the previous operating assumption. This magnitude of adjustment puts a different cast on the sorts of smoothing parameter disputes highlighted in some of the commentaries.
Which brings me to Suk Lee. Given the degree of uncertainty we face it is regrettable that only one of the commentaries (the one by Keewhan Kim) cites Lee’s 2011 paper, and even then doesn’t mention the key issue. In his analysis, Lee focuses on women, arguing that the KPA “missing male” distortion in the 1993 census is likely to be less severe for women, generating cleaner estimates of excess deaths. He calculates excess deaths 1993-2008 among the female population who were over 30 years old in 1993, and finds that deaths were 196,307 higher than expected, or 3.8 percent. If one extrapolates this percentage to the entire population one generates a baseline estimate of 815,000 excess deaths. Lee produces other variants on this calculation which produce estimates ranging from 506,000 to 1,125,000. Of course, if Jeon et al. are correct, and contrary to past famines, females were disproportionately impacted in the North Korean case, then Lee’s projection onto the entire population would overestimate famine deaths.
We are unlikely to ever know how many people died in the famine. Probably the sanest conclusion is one that I reached years ago, and Dan Goodkind reprises in his comments: suppose Jeon et al. are correct. Their estimates imply that roughly 3 percent of the population perished. That would be equivalent to nearly 10 million excess deaths if a similar tragedy were to befall the United States today--no small thing. Puts hysteria over Ebola in perspective, doesn’t it?