Body
A couple of weeks ago, during a presentation at the Korea Society, Homer Williams posed a question during the discussion period regarding the 2008 North Korean census. In his public comment as well as our private discussion afterward, he expressed skepticism about the veracity of the census, in particular with regard to the age structure of the population compared to the 1993 census. Nick Eberstadt has raised similar concerns about possible inconsistencies in the age and sex patterns in the two censuses.
This exchange reminded me that I had been meaning to write a post on a recent paper, “A Reassessment of Mortality in North Korea, 1993-2008,” by Dan Goodkind, Loraine West, and Peter Johnson, three demographers from the U.S. Census Bureau. Goodkind and West are notable for producing one of the first rigorous analyses of the 1990s famine, concluding that it probably resulted in something like 600,000-1,000,000 excess deaths. Suk Lee, now at KDI, working independently of Goodkind and West, and indeed unaware of their work, using a different methodology and different data, came to a similar numerical conclusion in his dissertation.
Although I am quite sure the authors would disagree with this characterization, the current paper could be interpreted as an exercise along the lines of “are there a plausible set of assumptions that would make the 1993 and 2008 censuses consistent?” and, “if so, how many excess deaths does the new census imply?”. Broadly speaking the answers appear to be “yes” and "600,000-1,000,000” though one has to jump through a number of hoops-including extending the period of "excess deaths" to 2008—to reach these apparently jejune conclusions.
The basic issue is that the authors take the veracity data that is highly contested as given. A quick (and incomplete) run-though:
- The authors mention in a footnote that an analysis by Eberstadt and Banister had concluded that the 1993 census implied a 7 percent undercount of the population; based on their analysis the authors conclude that the undercount was less than half that large. This matters, because if the 1993 population base is assumed to be higher, then there will be more excess deaths to explain if the 2008 census figures are taken as accurate. But more broadly, the authors downplay the sheer degree of uncertainty surrounding these figures. For example, during the famine, on the basis of its interactions with the North Korean government, in 1995 the FAO/WFP assumed that the population was just over 22 million for the purposes of doing their grain balance calculation. In the November 1997 food balance calculation the FAO/WFP assumed that mid-year 1998 population would be 23.5 million. As a point of comparison, the South Korean government’s estimate was 21.9 million. By June 1999, however, famine mortality was of such a magnitude that it was germane to calculate overall demand. Food balances were subsequently recalculated based on an official North Korean government population estimate of 22.55 million for August 1999—amounting to a downward revision of nearly one million people from the FAO/WFP’s previous assumption.
- The authors use North Korean child nutritional surveys to calibrate morbidity estimates. But again, these surveys are highly vulnerable to challenge. Steph Haggard and I examine in some detail these surveys in our book Famine in North Korea (pages 195-208). In terms of the work at hand, the authors simply dismiss surveys done during the famine that generated very high estimates of stunting among North Korean children and accept at face value two later surveys which yielded much lower estimates (and would hence generate lower estimates of morbidity). It is beyond the scope of this post to examine this issue in full detail (interested readers should take a look at the book), but sufficed to say that the surveys used in this paper excluded roughly 20 percent of the population and at the time were frankly disbelieved by some of the specialists from the UN agencies.
- In a similar vein, the authors note, the age 1-4 child mortality rates reported in the two censuses are “surprisingly low compared to other Asian societies.” Indeed. The reported North Korean child mortality rate is less that of China, Indonesia, the Philippines, Vietnam, and other Asian countries. Presumably how one assesses this claim is a function of prior experience. That the North Koreans have put out “national” figures on infant mortality and low birth weights that were lower than those of the US or UK, and later revealed to have been taken from a single hospital serving the elite in Pyongyang, might give one pause.
- The authors use Court Robinson’s low-ball estimate of 40,000 net outmigration. I have enormous respect for Court, but this estimate is by definition a floor, since due to Chinese policy North Koreans are forced to on-migrate to third countries to make asylum claims and hence exit his survey catchment area in the border region. FWIW, the South Korean government claims that there are currently 100,000 North Koreans in China. Admittedly, this is rounding error from the standpoint of estimating famine deaths, but in this case, applying a larger estimate for net migration would imply fewer famine deaths since the population shrinkage would be explained by migration, not death.
Finally, one minor point: the authors claim that earlier, separate, and higher estimates of famine deaths, up to 3 million, by Andrew Natsios and Hwang Jang-yop, were based on extrapolating estimates by Court Robinson and colleagues who had estimated mortality rates for some North Korean counties based on refugee interviews. True for Natsios, but not true for Hwang. Whatever one thinks of Hwang he was quite clear in attributing his claims to internal Party discussions prior to his defection, including estimates of the number of Korean Workers Party members who had died. Say what you like about Hwang (I was not a big fan) but you cannot write him off as an extrapolator—I sincerely doubt that he was even aware of the Robinson et al. work.
So why does it matter? Perhaps this paper does not quite measure up to Nick Eberstadt’s initial characterization “oops, I guess you and Haggard will have to re-write that book on the famine…oops, I guess you’ll have to re-write that refugee book…” But if you took some of the figures in this paper seriously, including one estimate of famine deaths of 330,000, the analysis would suggest that collectively we way overestimated the magnitude of the North Korean famine. This reconsideration would quite naturally inform our views of the severity of the current situation--if we were that wrong about the past, then could we be once again making the same mistake?
