Excess mortality estimates

Definition and method

Excess mortality is the difference between the deaths actually recorded in a given period and the counterfactual deaths implied by the prevailing death rate. The demographer Patrick Heuveline defined the death toll of the Pol Pot regime as the difference between deaths estimated under prevailing mortality and under counterfactual mortality, that is, as 'excess' deaths; he confined the figure strictly to the period the regime held power and estimated violent deaths such as executions separately as one subset of that total. Excess mortality is therefore the broader category that contains violent deaths.

Distinction from violent deaths

Mass grave data in Cambodia are partial evidence of violent deaths, not a total of excess deaths. By 2009 the Documentation Center of Cambodia (DC-Cam) had mapped 23,745 mass graves containing approximately 1.3 million suspected victims of execution. Kiernan estimates executions at about 31 percent of all deaths and Sliwinski at about 40 percent, while other accounts put direct execution at up to 60 percent and attribute the remainder to starvation, exhaustion and disease. The share itself is not settled, so the figures should be reported side by side rather than reconciled.

The spread of estimates for Cambodia, 1975-1979

Democratic Kampuchea ran from 17 April 1975 to 7 January 1979, about three years, eight months and twenty days. In the demographic balancing equation Heuveline set t-x as the onset of the regime and t as its fall. According to his survey, seven existing violent-death estimates range from about 100,000 to 1.1 million, the three lowest at roughly 100,000 to 300,000, while nine excess-mortality estimates run from 741,000 to 3,315,000. The highest, based on a direct tally of excess-death reports, is a government estimate usually dismissed as propaganda because it was not adjusted for multiple counting or reporting probabilities, while the lowest consists of only total accounting, not disaggregated by age and sex. Setting aside these two exceptional estimates, the remaining estimates fall between 1 million and 2.1 million, with the one-to-two-million interval the most frequently reported range.

The boundaries of the count are themselves disputed. This discussion counts excess deaths reasonably interpreted as resulting from Khmer Rouge policy; deaths in the civil war that brought the regime to power are excluded. The start of the period is clear, 17 April 1975, but the end point is difficult to fix: after the 'official' overthrow of January 1979 the Khmer Rouge still controlled a substantial population, and the famine of late 1979 owed much to their policies, so in practice the period is treated as extending to early 1979.

Uncertainty and dispute

Heuveline's probabilistic population reconstruction produced a 95 percent simulation interval of 1.2 million to 2.8 million excess deaths, with 1.5 million to 2.25 million covering 69 percent of simulations and a median of 1.9 million, or 21 percent of the at-risk population; he stressed substantial uncertainty about the exact scale. The reasons for the divergence are also documented. Individual point estimates each carry an unstated inherent uncertainty that cannot be recovered after the fact, so they cannot be statistically interpreted. Sliwinski's sample-based interval of 1,843,000 to 1,869,000 certainly understates the uncertainty inherent in that approach. Sample surveys can be biased by the concentration of deaths within families targeted along kin lines, by geographic clustering, and by uncertainty about both the total population at risk and baseline mortality, while population reconstruction depends on inputs such as the 1962 and 1998 censuses and on assumptions about about twenty years of births, deaths and migration after 1979.

The debate over the lowest figure is likewise unresolved. Vickery derived about 740,800 excess deaths by total accounting from the 1962 census of 5.74 million and a 1981 population estimate. Kiernan, attributing the difference chiefly to discrepant 1975 population estimates, criticised him for choosing a much lower CIA 'guess' instead of the pre-war calculation of Cambodia's leading demographer, and put his own estimate at 1,671,000. The gap between the two is not resolved by a single number.