Cambodia Section Home Mekong Network Home
About this site...

Counting Hell

by Bruce Sharp

In India, there is a story about a group of blind men. One day, they encounter an elephant. The first man feels the side of the elephant, and announces that it is a wall. The second man feels the trunk, and declares that it is a snake. The third man feels the leg and insists that it is a tree. The fourth man feels the tail, and says that it is a rope. (1)

We describe things on the basis of what we know... but what we know is often incomplete. Grappling in the dark, we try to determine the size, the shape, the nature of the unknown.

This is the problem of describing Pol Pot's Cambodia. We know the broad outlines of what happened during Khmer Rouge regime; it was, to borrow a phrase from one refugee, "Hell on the Earth." (2) These words could serve as a broader description of Cambodia during the entire decade of the 1970s. War, genocide, and starvation devastated the country to a degree that has few parallels in modern times. The details of these events, however, are often confusing and sometimes contradictory. A wall, or a tree? A snake, or a rope?

How can we place Cambodia within the context of revolutions in general? It is impossible to quantify the suffering of the Cambodian people in the 1970s. How do you measure hell? There is, however, one element common to all revolutions: the death toll.


Definitions and Assumptions

For the purposes of this article, any discussion of the death toll refers to excess mortality. Deaths attributed to the Khmer Rouge regime are, in short, those deaths which can reasonably be interpreted as having been the result of Khmer Rouge policies. Deaths from the civil war which brought the Khmer Rouge to power should not be considered as a part of the regime's death toll; after all, death is expected in war. The beginning of the period, then, can be clearly determined as April 17, 1975. The end point, however, is more difficult to fix. The Khmer Rouge continued to control a significant part of the population beyond the "official" overthrow of the regime in January 1979. Famine affected parts of the country later in 1979. As the famine was largely brought about by Khmer Rouge policies, one could argue that these deaths, too, should be considered within the toll. However, because the data is imprecise, shifting the end point by a few months in one direction or another makes little difference; for practical purposes, it is sufficient to say that the period in question extends to early 1979.

In any discussion of the death toll, it is important to stress that these figures are still only estimates. These estimates vary widely; yet the fact that they are all estimates does not mean that they are all equally valid. Careful examination of the available evidence reveals inconsistencies that indicate whether or not the number in question is worthy of serious consideration.


A Range of Estimates

As of this writing (April 2005), the most commonly cited estimates of the death toll fall in the range of 1.5 to 1.7 million. The frequency with which a statistic is repeated, of course, is not an indication of its accuracy. Estimates often gain superficial credibility through repetition: a figure is put forth by one source, and is repeated without citation in another article; another commentator, searching for an accurate estimate, sees the same figure cited in two separate articles, and assumes that these two accounts have corroborated each other, and can therefore be considered reliable.

Dubious statistics about Cambodia abound, often surfacing in books and articles that are otherwise very carefully researched.

A recent article by Carol Pucci in the Seattle Times serves as a good example: "An invasion by the Viet Cong ended the Pol Pot regime but threw the country into political turmoil and economic isolation until the United Nations brokered elections in 1993," Pucci writes. "The population, once 30 million, now stands at just 13 million; the average family of five earns about $375 per year." (3) Surely it should have been obvious that whatever the nature of the tragedy in Cambodia, the country did not lose 17 million people.

Statistics often exhibit a snowball effect: arbitrary figures become widely repeated, and soon become part of the conventional wisdom. In his book Damned Lies and Statistics Joel Best notes that a number "takes on a life of its own... through repetition, it comes to be treated as a straightforward fact -- accurate and authoritative." (4)

Samantha Power's A Problem from Hell serves as an example. Overall, Power's book is excellent; yet in the chapter on Cambodia, she states that "The toll of the civil war on Cambodia's civilians had been immense. Some 1 million Cambodians had been killed." (5) As her source, she cites Elizabeth Becker's When the War was Over. Becker does indeed write that "Officially half a million Cambodians died on the Lon Nol side of the war; another 600,000 were said to have died in Khmer Rouge zones." (6) She does not, however, explain the origin of the statistic... and the statistic is, in fact, wrong.

It might seem unfair to criticize Power, Becker, and Pucci for statistical errors in books and articles which are not primarily concerned with statistics. But it is precisely because these works are thoughtful and articulate that they serve as good examples. Numbers are unforgiving: small errors can become big errors, but even faulty statistics are still imbued with authority by virtue of their source.

How can we separate good statistics from the bad? The first step is to approach all numbers with some skepticism. For every figure, we need to ask: Does this make sense in light of other evidence?

This article attempts to evaluate some of the prevailing estimates of the death toll in Cambodia. Are the methods and assumptions used to construct these estimates sound? Are the conclusions truly supported by the evidence? Are any of these estimates subject to "single points of failure," where uncorroborated data affects the final outcome?


The Estimates

The actions of the Khmer Rouge in the aftermath of their victory immediately raised concerns of a human rights catastrophe; in particular, the brutal evacuation of Phnom Penh, and testimony of refugees led many to fear that the country might be in the throes of a bloodbath. Francois Ponchaud was among the first to suggest that the death toll might exceed one million. Writing in mid-1977 (before some of the worst massacres), Ponchaud stated that "nobody can suggest a reliable figure but it certainly exceeds one million" (7). This could hardly be described as an estimate; it was more of an intuitive guess. As it turned out, however, it was a reasonably good guess.

After the overthrow of the regime, scholars began the grim task of assessing the damage. One of the first efforts to evaluate the scale of the disaster was conducted by the Central Intelligence Agency. In a report entitled "Kampuchea: A Demographic Catastrophe," the CIA made a methodical attempt to estimate deaths from various causes, among various groups, from 1975 through January 1979. The conclusion: net population loss was probably around 1.4 million people. Under normal conditions (a birth rate of 2.7%, and a death rate of 18 per thousand) there would have been around 500,000 deaths during this period; thus, excess mortality would be on the order of 900,000. (8)

The methods employed were straightforward: beginning with an estimate of the 1975 population, the authors applied estimates for birth and death rates for various social groups in different phases of the revolution, and also estimated the number of executions and population loss due to migration. None of these factors, however, was quantified with any certainty; the number of executions, for example, was apparently calculated by estimating the numbers of individuals who, based on membership in various groups (soldiers, intellectuals, and so on) would likely have been targeted for execution; a proportion of these individuals were assumed to have been executed, presumably on the assumption that other members of these groups might have escaped detection.

As the situation in Cambodia began to slowly stabilize, journalists, scholars, and aid workers gained greater access to the survivors; and as they did, the magnitude of the disaster became undeniable. An emerging consensus suggested that the toll was somewhere between one and two million. (9)

A few researchers attempted to determine whether or not the these estimates were feasible. Their conclusions, however, varied widely. To understand why, it's first necessary to examine some of these estimates in detail. This article will focus primarily on estimates by four different researchers. Each of these analysts - Michael Vickery, Ben Kiernan, Craig Etcheson, and Patrick Heuveline - approaches the problem differently. In each case, there is what we might call a core assumption: a central piece of evidence that serves both as a starting point, and a standard for evaluating the legitimacy of other data.


1. Michael Vickery

In 1984, Michael Vickery addressed the question of the death toll in his book Cambodia 1975-1982. Vickery argued that the toll could not possibly have been as high as some observers suggested. "It is simply impossible to take the generally accepted population figure for April 1975, the population alive today, demographically acceptable birth rates, and project an extermination figure of 1-2,000,000." (10)

Vickery's analysis was based on a reconstruction from census data. Total mortality was estimated by calculating the difference between the population at the beginning and end of the regime; subtracting the number of deaths which would be expected under normal conditions from the total mortality would then yield the excess mortality.

The first step, then, was to establish the size of the population in 1975:

"The only nationwide Cambodian census ever taken was in 1962, and it produced a total population figure of 5,740,000. Official statistics for subsequent years were nothing more than the 1962 figure compounded annually by a growth rate of 2.2 percent, giving about 6.8 million for 1970 and 6.9 million for 1971.

"More sophisticated tinkering with the data gave, for 1970, 6,993,000 or 7,143,000; and the most elaborate manipulation of the base figures produced seven different possible projections, ranging for 1970, from 7,029,000 to 7,524,000... The author of these extrapolations seemed to prefer 7,363,000 for 1970." (11)

Continuing, Vickery cites the CIA's "Demographic Catastrophe" report:

"In their report on Cambodian demography, the CIA used a figure of just over 7 million for 1970, which is as good a guess as any for our purposes, and 7.3 million by April 17, 1975, which means they accepted both a decrease in the rate of growth and a war loss of over 500,000. Of that 7.3 million, there were about 200,000 Vietnamese who were immediately repatriated to Vietnam, leaving 7.1 million Cambodians (including Chinese and Chams) for the starting DK [Democratic Kampuchea] population." (12)

For the surviving population, Vickery noted that the new regime, the People's Republic of Kampuchea (PRK), had projected a mid-1981 population within Cambodia of 6.8 million. PRK statistics also tallied 6,353,690 people registered in krom samakki ("solidarity groups") in the countryside. "Since there were possibly 3-400,000 in the larger towns (Phnom Penh, Battambang) unregistered in solidarity groups and some unregistered in the countryside, a total population figure of 6.8 million is not unreasonable, and, including people who had fled the country since 1979, a figure of 7.1 living Cambodians could be postulated. Assuming they had increased at 2.2 percent per annum since early 1979, the number of DK survivors at that date would have again been over 6.7 million; and if the rate of increase in 1979 was less, which seems likely, the total for early 1979 would have been even higher." (13)

Making allowances for decreased fertility, Vickery concluded that the Khmer Rouge regime led to approximately 740,800 excess deaths. (14)

There are, however, a number of problems with Vickery's analysis. By relying exclusively on census data, an error in any of the census figures results in a corresponding error in the estimate. Other analysts -- in particular, Ben Kiernan -- have raised this objection. Kiernan argues that the difference between his own estimate and Vickery's is primarily the result of the discrepancy between their estimates of the 1975 population. Vickery, Kiernan writes, "disregards the prewar population calculations of the leading demographer of Cambodia in favor of a much lower 'guess' by the U.S. Central Intelligence Agency. His position is unsustainable." (15)

Vickery gives no indication of why he chooses one figure instead of the other. He does not help his case with the cavalier remark that the figure cited by the CIA is "as good a guess as any for our purposes." His use of the CIA's figure particularly difficult to justify, given that Vickery himself claims that the goal of the CIA's report was to minimize the crimes of the Khmer Rouge. (16) If he believes his own argument, why rely on the CIA's estimate of the 1975 population? If the CIA's goal was "abetting the Pol Pot clique in their worst crimes," (17) (to use Vickery's own phrase), one must recognize that a simple way to minimize the death toll would be to undercount the population at the start of the regime... and thereby undercount the number of victims who perished.

It should also be noted that the estimate of the 1975 population was predicated on the assumption that the civil war had resulted in around 500,000 excess deaths. This, as we shall see, is doubtful.

The figure Vickery cites for the 1979 population is also highly questionable. The population projections were compiled by administrators who had an incentive to overestimate the number of persons under their jurisdiction, and most analysts believe that the figures were inflated. (18) Vickery's calculations suffer from another oversight, as well: Vickery does not account for immigration and repatriation. He assumes that (except for those born after 1979) all of the 1981 population should be counted as survivors of the DK regime. His estimate of the starting population assumed that 200,000 ethnic Vietnamese were exiled to Vietnam immediately after the Khmer Rouge came to power. (19) How many of these 200,000 returned after the regime was overthrown? Vickery's estimate assumes that none of them did; but it is highly likely that many, if not most, did return, and also quite possible that other Vietnamese who had not previously lived in Cambodia migrated to the country during this period. A constant complaint of those opposed to the PRK regime from 1979 onward centered on claims that the country was being overrun by ethnic Vietnamese. There is no question that these criticisms were blown out of proportion; nonetheless, there is also no question that a certain amount of migration did occur.

Finally, it is generally accepted that the end of a mortality crisis is frequently accompanied by a drastic increase in the birth rate, often beyond the normal rate; thus, projecting backward from a later population estimate using the normal peacetime birth rate would result in an overestimate of the surviving population. (20) On top of this, the 2.2% growth rate that Vickery assumes for 1980-1981 is substantially lower than the 2.7% growth rate used to arrive at the estimate he favors for 1970.

The method of estimating deaths by comparing census figures is prone to underestimation for another reason. Demographers recognize that a census will almost always be an undercount. If we assume a similar undercount in each census, we need remember that our death toll is also, in effect, an undercount; in order to be accurate, it needs to be adjusted upward by the margin of error in the census. At first glance, this might seem counterintuitive; after all, if our population estimate at the start of the crisis is an undercount, this would be offset by the fact that we will also undercount the number of survivors. Closer consideration, however, reveals that this is not correct: a proportion of the dead will be unaccounted for, just as surely as a proportion of the living have been unaccounted for. This can best be illustrated by example. Let's assume that our starting population was roughly 7.9 million, and our surviving population was 6.2 million; this would suggest a net loss of 1.7 million. However, if each figure was actually a 3% undercount, our 1975 population figure reflects an actual population of 8.137 million, and our 1979 estimate reflects an actual population of 6.386 million. The actual net loss - 1.751 million - is 51,000 higher than our original calculation.

Cambodia's only full census prior to the crisis was 1962; thus, the discrepancy in this case would be even greater, since the initial underestimate would have been compounded by the annual growth rate for roughly 13 years.

Some might argue that Vickery's conclusions reflected the limited evidence available at the time. Ten years later, however, Vickery was still defending his position. In a letter to Z magazine, in 1994, Vickery insisted that "I still believe in the relative accuracy of those figures, and in recent years have come to the conclusion that if they are in error, it is on the high side." However, he noted, "on the weakness of all Cambodian population statistics," any estimate could be off by 2-300,000. "I would not argue about a figure of one million or so, even though my preferred total is less," Vickery wrote. "I do insist that the 2-3 million estimates are demographically impossible." (21)

Because Vickery does not consider any evidence other than the population estimates, his estimates of the death toll are entirely dependent upon the accuracy of the figures he cites... and as we shall see, there is substantial evidence to indicate that these figures cannot possibly be correct.


2. Ben Kiernan

In the early 1980s, Ben Kiernan conducted a series of interviews with survivors of the Pol Pot regime. Whereas Vickery's approach relied primarily on estimating the population at different times, Kiernan attempted to determine the toll by randomly sampling the survivors. By calculating death rates on the basis of some 500 interviews, Kiernan estimated the death toll to be around 1.5 million.(22)

In his book The Pol Pot Regime, Kiernan notes that interviews conducted by Milton Osborne led to similar conclusions. In 1980, Osborne interviewed 100 refugees in camps on the Thai-Cambodian border. The families of these refugees included about 500 persons, 128 of whom died during the Khmer Rouge regime. This would suggest a death toll of roughly 25%. Kiernan points out, however, that Osborne's sample did not represent an accurate breakdown of the urban and rural population. Of the 100 refugees, 42 were from rural areas; another 17 were low-level urban workers. "If Osborne's sample were representative," Kiernan writes, "the 25% death rate would suggest a national toll of 2 million." Kiernan argues, however, that the urban population composed only about 20% of Cambodia's total population, and not the 58% seen in Osborne's sample. Since individuals from urban areas died in greater numbers than the rural population, extrapolations from this sample would tend to be too high. On the other hand, Kiernan notes, some of Osborne's surveys were conducted in camps controlled by the Khmer Rouge, and therefore also overrepresented Khmer Rouge supporters, who suffered less than the rest of the population. Considering these factors, Kiernan argues that Osborne's data "therefore point[s] to a possible toll of around 1.5 million." (23)

Extensive interviews were also conducted by Stephen Heder in 1980 and 1981. Heder interviewed about 1,500 refugees on the Thai-Cambodian border. Based on Heder's sample, Kiernan estimates that 33% of the urban "new people" died, along with roughly 15% of the rural "base people." In both of these groups, roughly half the deaths were from executions, and half from starvation and disease. "These data also point to a national toll of over 1.5 million, about 20% of the population." (24) In fact, Kiernan notes, "All known surveys of the toll in individual communities also suggest a death rate well above 20 percent (a figure suggesting a national toll of 1.6 million)." (25)

In order to calculate the total number of victims based on the death rates, it is necessary to estimate the 1975 population. Citing an estimate by demographer Jacques Migozzi, Kiernan believes that the population at the start of the 1970 civil war as 7.363 million; and by mid-1974 the U.N. estimated the population to be about 7.89 million. Kiernan suggests that the April 1975 population was probably around 7.894 million. (26)

This underscores the questionable nature of the estimates Vickery relies upon. Vickery's figure, from the CIA demographic report, was based on extrapolations from earlier UN estimates. Since the UN itself suggested a significantly higher figure for 1974, there would seem to be no good reason for using the older, lower estimates.

Kiernan also cites other evidence that Vickery's 1975 figure is improbable. A former teacher who worked as a statistician for the Khmer Rouge (Democratic Kampuchea) government told Kiernan in 1980 that "'In 1975 there were about 8 million Khmers, according to what I learnt while doing statistics.'" Moreover, Kiernan writes, "In March 1976, Democratic Kampuchea aired its own reduced count of the population, 7,735,239. By August, its estimate was 7,333,000. The progressively smaller figures are obviously due to the death rate after April 1975. Both remain substantially higher than Vickery's April 1975 figure of only 7.1 million, ruling it out as impossible even before considering the known intervening death toll." (27)

For the surviving population, Kiernan notes a range of estimates, from a low of 6 million, to a high of 6.7 million. (28) The high end of this range is preferred by Vickery; but Kiernan's analysis of the death rates suggests that this figure is an overestimate, and that the actual number was closer to the lower end of the range.

Assuming that Kiernan's estimate of the 1975 population is accurate, his figures for the surviving population would mean that the number of deaths would fall within the range of 1.19 and 1.89 million.

Ultimately, however, the census estimates are not the core assumption for Kiernan's estimate; the driving force behind Kiernan's insistence on a higher toll is the death rates suggested by the survivor interviews. Analyzing the interviews on the basis of the respondent's social groups, Kiernan breaks down his estimates for deaths during the regime as shown in Figure 1. (29)

Figure 1: Death Rates in Cambodia, 1975 - 1979

GROUP 1975 Pop. Number of Deaths Death Rate (%)
New People:
Urban Khmer 2,000,000 500,000 25
Rural Khmer 600,000 150,000 25
Chinese (all urban) 430,000 215,000 50
Vietnamese (urban) 10,000 10,000 100
Lao (rural) 10,000 4,000 40
Total new people 3,050,000 879,000 29
Base People:
Rural Khmer 4,500,000 675,000 15
Cham (all rural) 250,000 90,000 36
Vietnamese (rural) 10,000 10,000 100
Thai (rural) 20,000 8,000 40
upland minorities 60,000 9,000 15
Total base people 4,848,000 792,000 16
Total Cambodia 7,890,000 1,671,000 21
Source: Ben Kiernan, The Pol Pot Regime, p. 458

These death rates would leave 6.219 million survivors, a total that falls within the generally accepted range of estimates for the 1979 population. In a subsequent article, evaluating the toll on the basis of population estimates, Kiernan notes that the toll could be somewhat higher than his initial calculations would suggest: "We may safely conclude, from known pre- and post-genocide population figures and from professional demographic calculations, that the 1975-79 death toll was between 1.671 and 1.871 million people, 21 to 24 percent of Cambodia's 1975 population." (30)

Those who have dealt with Cambodian refugees are likely to find Kiernan's estimates more in line with their expectations than Vickery's. There are, however, some aspects of Kiernan's calculations which begin to seem questionable in light of other evidence. A few of these issues become apparent when we consider an alternative analysis, put forth by Craig Etcheson.


3. Craig Etcheson

In 1994, the Documentation Center of Cambodia began to address the question of how many victims died during Khmer Rouge regime. In 1999, summarizing the previous five years' research Craig Etcheson wrote a lengthy article entitled "The Number: Quantifying Crimes Against Humanity in Cambodia." In this highly provocative analysis, Etcheson does not actually offer an estimate of the toll; he does, however, outline some of the evidence that suggests that Vickery's and Kiernan's estimates are too low.

"Many Cambodians believe, almost as an article of faith, that the Khmer Rouge killed more than three million people during the Democratic Kampuchea regime. When this estimate of the Khmer Rouge death toll was first publicized in the early 1980s, commentators in the West almost universally dismissed it as a product of 'Vietnamese propaganda,' an invented figure designed strictly for political purposes.

"Yet, the three million figure was not a complete invention. In the early 1980s, the authorities of the People's Republic of Kampuchea carried out what amounted to a national household survey, aiming to interview every head of household in the entire country about what had happened to their families during the Pol Pot regime. On July 25, 1983, the 'Research Committee on Pol Pot's Genocidal Regime' issued its final report, including detailed province by province data. Among other things, their data showed that 3,314,768 people lost their lives in the 'Pol Pot time.' But that report was quickly forgotten inside Cambodia, and it never became known outside of Cambodia -- until 1995.

"More than a decade after the PRK report was published, researchers at the Documentation Center of Cambodia discovered many of the records from this remarkable research project. Those records allowed DC-Cam researchers to reconstruct the methodology employed by the PRK Research Committee, and some flaws were detected in the research design, flaws, which would tend to lead to an overestimation of the total casualty figure. The Research Committee interviewers of the early 1980's had gone from house to house, and from village to village, collecting information regarding death during the Khmer Rouge regime. It appears, however, that they did not adequately account for the fact that extended families are usually spread out across more than one household or village, and therefore double counting of some victims could occur based on reports from different households belonging to the same extended family." (31)

Noting other methodological errors, Etcheson ultimately concluded that might have been overestimated by as much as 50%, putting the toll in the vicinity of two million. (32)

Meanwhile, the Documentation Center also embarked on a detailed project to chart the location of the mass graves scattered throughout the country. Between 1995 and 1999, DC-Cam teams surveyed some 20,000 graves. "According to the data," Etcheson writes, "these mass graves contain the remains of 1,112,829 victims of execution." Moreover, this number represented only a partial accounting. The final toll, Etcheson predicted, "could eventually reach substantially higher, perhaps as high as 1.5 million." (33)

If the remains in the mass graves give us a reasonable approximation of the number of executions, what does that tell us about the overall death toll? Etcheson explains that, according to Kiernan, Osborne's data suggests that executions amounted to 31% of the death toll; meanwhile, "The demographer Marek Sliwinski estimates that about 40% of the death toll resulted from execution, 36% from starvation, 13% from disease, and the remainder from either combat or natural causes. Other work carried out by a political scientist, Steve Heder, suggested that different proportions of the total death toll could be attributed to execution for urban versus rural dwellers, about 33% among 'new people' and 50% among 'base people.' Thus the various estimates of the proportion of deaths resulting from execution range from a low of about 30% for the overall population to a high of 50% among base people." (34)

Considering these figures, Etcheson continues:

"The implications of these figures are enormous. If these calculations of the proportion of deaths due to causes other than execution are accurate, then we begin to approach an astonishing conclusion. It begins to look possible that the original Cambodian estimate of 3.3 million deaths during the Khmer Rouge regime might be very nearly correct.

"If as little as 31% of the death toll was the result of executions, then a total of 3.3 million deaths would imply slightly more than one million executions, and the Documentation Center data suggest they have already found more victims of execution than that. If we apply Heder's top estimate of 50% for base people to the entire population, and find upon the completion of the mass grave surveys that the number of suspected victims of execution is around 1.5 million, then we again end up with a figure in the vicinity of three million total dead in the Pol Pot time. In either case, we would be driven to the conclusion that not one million, not two million, but rather three million Cambodians died untimely deaths during the Khmer Rouge regime." (35)

A casual reading of Etcheson's report suggests that his own estimate of the death toll, then, would be approximately three million. According to Etcheson, however, this was not his intention. Personally, Etcheson suggests that the death toll was between 2 and 2.5 million, with a "most likely" figure of 2.2 million. (36)

What factors would we need to consider before asserting a higher toll on the basis of the forensic data? Etcheson outlines a few of the problems inherent in relying on the data from mass graves. He cites a passage from Vickery's Cambodia 1975-1982, where Vickery writes:

"Some of the burial pits discovered provide the evidence that mass executions occurred, but there is as yet no way to count the number of executions separately from death due to other causes. Yathay pointed out that in Pursat in 1976-77 mass graves were for those who died of hunger and illness, while executions took place in isolation in the forest. Moreover, some of the 500,000 war victims are buried in mass graves, and without forensic tests it is probably impossible to determine whether death occurred before or after 1975." (37)

Indeed, Etcheson notes, "Documentation Center mapping teams have located a number of sites over the years where the local informants say that the mass graves were in fact not from Khmer Rouge executions, but rather from the bombing, the 1970-1975 war, from victims of mass starvation and even one or two associated with the Vietnamese invasion of 1979." The overwhelming majority of the graves, however, were unquestionably from the Khmer Rouge period. "On two occasions only, if memory serves, have Documentation Center mapping teams discovered mass graves, which were attributed to the victims of bombing during the 1970-1975 war." Moreover, Etcheson notes, "It bears repeating that the virtually all of these mass graves sites are located at, or quite near -- usually within a kilometer or so -- of the Khmer Rouge security centers. It turns out that local informants, in most cases, recall the names of the cadre who were in charge of these Khmer Rouge prisons or killing centers. As often as not, it is subsequently possible to locate this former Khmer Rouge security cadre for follow-up interviews." (38)

Nonetheless, it is critical to clarify one point regarding the forensic data: the number of remains is still only an estimate. Etcheson argues, however, that the experience of the teams has enabled them to achieve good results overall, and that the methodology used in compiling the estimates contains, in general, a conservative bias.

Kiernan, meanwhile, disputes suggestions that the mapping data suggests a higher toll than his own estimate of roughly 1.7 million deaths. He points out that Etcheson's remark that the mass graves contain victims of execution does not preclude the possibility that many of these same graves might also contain victims who died from other causes. "[Etcheson] overlooks the evidence of Pin Yathay that in Pursat in 1976-77, for instance, graves were dug for the victims of hunger and disease, not execution," Kiernan writes. (39) This is precisely the same reference (down to the very page) invoked by Vickery. (40)

It is unclear, however, whether or not Yathay's remark really supports the interpretation that the two scholars suggest. Kiernan's phrasing implies that Yathay says that victims of hunger and disease were buried, but that victims of execution were not; yet Yathay said no such thing. Moreover, it is not at all clear that what Yathay describes in Pursat should be called mass graves. The burials took place in what were, in effect, makeshift cemeteries, bearing little resemblance to the mass graves unearthed by the mapping project. (Click here to see an English translation of the excerpt.)

Etcheson's assumption in the mapping report is that the mass graves contain victims of execution, and not victims of starvation and disease. If Vickery and Kiernan's interpretation of Yathay were correct, and victims of execution were killed in isolation in the forest, their remains would probably not have wound up in the mass graves. In such a scenario, the number of remains from the mass graves still represents only a partial toll. Since the surveys suggest that executions accounted for between 30 and 50% of the total death toll, the implications are minimal. In fact, at 50%, it would make no difference whatsoever; either way, half the victims in Pursat would be uncounted.

Kiernan's objections, however, go deeper. He argues that Etcheson's extrapolation "ignores all the Cambodian demographic data." "Such sloppiness," Kiernan writes, "is unworthy of the Documentation Center of Cambodia. Exaggerating a horrific death toll, it contributes to the ethnic auctioneering of genocide research." (41)

The charge of "ethnic auctioneering" is gratuitous; it is difficult to see how arguing that existing estimates of the Cambodian genocide are wrong would have any bearing on other, unrelated tragedies.

The more important issue, however, is the question of whether or not Etcheson's article ignores the demographic data. Is Kiernan's charge correct? Yes and no. In the mapping article, Etcheson summarized the implications of the number of remains with the comment that "we would be driven to the conclusion that not one million, not two million, but rather three million Cambodians died untimely deaths during the Khmer Rouge regime." There is exactly one word that makes this sentence inaccurate: the word untimely.

What Etcheson seems to have overlooked in this particular remark is how the difficulty of separating natural and unnatural deaths would affect calculations of the ratios of deaths from execution, and deaths from other causes. It is very doubtful that those respondents in the surveys conducted by Heder, Osborne, and Kiernan would be able to discern which deaths should be considered normal mortality, and which should be considered excess mortality; thus, the ratios Etcheson refers to are more valuable for calculating all deaths, and not only untimely deaths.

Prewar death rates would suggest that roughly 500,000 deaths would be represent normal mortality for the duration of the Khmer Rouge regime. (42) Three million deaths, then, would mean roughly 2.5 million excess deaths; and if the proportion of deaths by execution was higher than 30%, the toll would be somewhat less.

How does this fit with the demographic evidence? As it turns out, the demographer Kiernan cites in his article -- Patrick Heuveline -- suggests that Kiernan's own estimates are low. Heuveline's estimates, in fact, are quite close to Etcheson's.


4. Patrick Heuveline

Vickery's early estimates of the toll had relied on census data; Kiernan's relied primarily on survey samples, and Etcheson's 1999 article had dealt mainly with forensic data. Approaching the question from a new perspective, Heuveline analyzed the sizes of various population groups, based on age and sex, before and after the 1970s. By studying the sizes of these birth cohort groups, and projecting trends forward and backward from census totals, Heuveline calculated excess mortality for the entire period between 1970 and 1980.

Heuveline stresses that "any estimate, including mine, is subject to high uncertainty." (43) Nonetheless, it does seem that the toll might be higher than previously believed. "As best as can now be estimated," Heuveline writes, "over two million Cambodians died during the 1970s because of the political events of the decade, the vast majority of them during the mere four years of the 'Khmer Rouge' regime." (44)

Heuveline's analysis is based on a review of trends reflected in population growth rates; thus, the periods of relative normalcy, i.e., before 1970 and after 1980, provide a standard against which the effects of crises can be measured.

Figure 2: Mortality Rates in Cambodia. Source: Patrick Heuveline. Reprinted with permission from Forced Migration and Mortality, 2001, by the National Academy of Sciences,
courtesy of the National Academies Press, Washington, D.C.Under normal circumstances, mortality follows a J-shaped curve. Moving from left to right along the J, we see a relatively high incidence of mortality immediately after birth (i.e., infant mortality). The number of deaths will decline as the individuals pass through childhood, then will increase again after adolescence and into the later years of adulthood, as shown in Figure 2. (45)

Given the relatively consistent nature of mortality patterns under normal circumstances (the red lines on the graph), census data from long after a mortality crisis can be used to estimate what proportion of a given demographic group survived the crisis. This allows demographers to determine which groups suffered disproportionate deaths during the crisis. In the case of Cambodia, the data indicates a severe loss of life during the crisis, particularly among young adult males.

The "crisis," in the context of Heuveline's analysis, consists of the entire decade of the 1970s. Like Kiernan, Heuveline believes that the population estimates relied on by Vickery are significantly too low. Heuveline examined records from the 1962 census, and a subsequent analysis of those records prepared by demographer George Siampos. Concurring with Siampos' contention that the birth rate had been underestimated, Heuveline estimated the 1970 population at roughly 7.7 million. (46) Statistics from the 1998 census data implied a decrease in the birth rate from 1970 on, and a severe decline from 1975 on. ("One-half to three-fourths of antebellum levels, implying a deficit in these four years in the order of 300,000 to 600,000 births.") (47)

Figure 3: Demographic Groups, Cambodia and Thailand. Source: U.S. Census Bureau, International Data BaseComparing the contemporary population structure of Cambodia with that of Thailand, (see Figure 3) one can easily see lingering effects of the turmoil of the Seventies. A population pyramid shows the severe deficit among certain age and sex groups. The group born between 1976 and 1980 is particularly small, and virtually every age group born prior to 1976 shows a significant disparity in the ratio of women to men. (48)

Still, the uncertainty of the original data renders the estimates themselves uncertain. As Heuveline puts it: "The logic of indirect estimation using the balancing equation of population change is similar to assembling a jigsaw puzzle with a single missing piece, whose size and shape can be revealed by fitting together the remaining pieces. But most often, the other pieces can only be estimated more or less precisely, and the size and shape of the missing piece only appears fuzzy." (49)

Heuveline's calculations suggested that the death toll might have been higher than generally believed. Several of the parameters in Heuveline's formulas cannot be fixed with precision, but they can be approximated. By varying each parameter from its minimum to maximum plausible value, the calculations suggested a theoretically plausible range of 1.17 million to 3.42 million excess deaths for the entire decade. (50) Comparing forward projections from pre-1970 trends, and backward projections from post-1980 trends, Heuveline arrived at a central estimate of 2.52 million excess deaths for the decade. Adjusting this figure to account for war deaths, changes in fertility and infant mortality rates, and the impact of the 1979 famine, Heuveline concluded that the number of excess deaths during the Khmer Rouge regime was probably around 2 million. (51)


Evaluating the Estimates

Four different methods of evaluating the death toll yield four very different results:

1. Reconstruction from the census data: 740,000 (Vickery)
2. Random sampling from surveys: 1,670,000 (Kiernan)
3. House-to-house survey and forensic estimations: 3,000,000 (PRK/Mapping project)
4. Demographic trends/cohort analysis: 2,000,000 (Heuveline)

First, let's consider Vickery's estimate. Is it plausible? It is certainly not consistent with the data from DC-Cam's mapping project. Etcheson's report on the mapping project was based on data gathered as of 1999. The project is still ongoing, and its most recent summary revises the earlier figures. The current estimate suggests that the mass graves contain 1,386,734 victims. (52)

Thus, the number of victims in the mass graves - presumed to be primarily victims of execution - exceeds the toll that Vickery suggests for deaths from all causes during the Khmer Rouge years. In fact, it even exceeds Vickery's estimate for the civil war and the Khmer Rouge period combined. It seems likely that combat casualties were generally not buried in mass graves; and certainly deaths owing to the tangential effects of the war (increased disease, malnutrition, and so on) would not have generally been buried in mass graves.

We can exclude the lowest of the four estimates: but what about the highest? If the mapping project data suggests that the number of executions may have been as high as 1.3 million, could the PRK's estimate of 3.3 million deaths possibly be correct?

As Etcheson points out, the PRK's survey contained a number of methodological errors. The death toll derived from the house-to-house survey was combined with the number of remains exhumed from mass graves; yet there was no way to determine whether or not any of the victims from the graves were also included in the deaths reported by family members. On top of this, the researchers were unable to properly account for instances in which the same death was reported by more than one family; for example, one man's death might be reported by both his own family and his in-laws. It is also possible that some of those who were believed to have died might have resettled abroad. Nonetheless, there is also the possibility that some families were killed entirely; with no survivors left, their deaths would be unreported in the survey. This is a critical point, and its implications will be discussed shortly. In spite of this, however, there is no other evidence to suggest that the actual death toll could possibly be on the order of three million.

This brings us to what is probably the most widely accepted estimate: Kiernan's figure of 1.67 million excess deaths. As with all other estimates, there are a number of variables that are open to question; but Kiernan's estimate, derived primarily from the estimates of death rates, seems to be roughly in line with his estimates of the 1975 and 1979 population.

Kiernan's figure also falls within the plausible range suggested by Heuveline's analysis. Yet it is significantly lower than Heuveline's "most likely" figure, and also significantly lower than what we would expect the based on the forensic evidence.

One possible factor is that Kiernan's estimates of "old people" and "new people" may not be accurate. Kiernan suggests a figure of 3.05 million "new people," including an urban population of 2.53 million. He bases this on a figure of 1.8 million in Phnom Penh, including refugees. William Shawcross, however, cites a figure of over 2 million in Phnom Penh, and adds that, of the rural population, "almost half fled their fields for the sanctuary of the towns." (53) George Hildebrand and Gareth Porter, meanwhile, suggested that the population of Phnom Penh may have been around 3 million. (54) Even when we add in the population of other cities, this is still lower than the CIA's assumption in the Demographic Catastrophe report, which suggested an urban population of 4 million in 1975. (55)

Kiernan calculates the death rate among urban Khmer at 25%, versus 15% for rural Khmer. Accepting the higher estimates of the displaced population (i.e., 4 million "new people," instead of 3 million) would mean that the death toll would be around 1.77 million. However, while the number of new people might have been higher than Kiernan suggests, there is no compelling evidence one way or the other; all we have are two estimates, either of which is plausible.

However, there is another factor which may significantly affect Kiernan's estimate. The survey methodology Kiernan relies on is marred by a subtle flaw: Clustering of mortality will typically cause random surveys to underestimate the death toll.

Initially, the survey method seems like a reasonable approach: by interviewing survivors who can count deaths within their own family, we can determine approximate death rates; and by applying these mortality rates to the general population we can calculate an overall estimate. And, as Heuveline points out, the survey method "is often the only technique that can be implemented by a lone investigator and a few collaborators very shortly after the mortality crisis. It thus provides invaluable preliminary estimates when other techniques are much harder and costlier to implement." (56) And, when there are relatively few deaths, distributed fairly evenly, the survey method works reasonably well. But what happens when deaths are concentrated disproportionably within certain groups?

Let's consider deaths within families during the Khmer Rouge regime. The first problem with the survey method is fairly obvious: families who are killed entirely cannot be represented in the survey. There is no one left to speak for them. Although this fact alone will affect estimates of the death rates, it is still not the main problem. More significant is the fact that post-crisis surveys are likely to over-represent families with more survivors; and because of this, overall death rates derived from these surveys will tend to be too low.

Why? To illustrate, let's consider a hypothetical example.

Suppose we want to estimate how many people died in a village of 3,000 people. We'll do this by asking 50 survivors how many deaths there were in each family.

What happens if the fatalities in this village are not distributed evenly? Suppose that we have a total of 300 families, each with 10 members. Of those 300 families, 150 of those families suffered 3 deaths per family. The other 150 families, meanwhile, suffered 1 death per family. The total death toll, then, is 600.

Initially, it might seem that sampling survivors will work fine: half of the families had three fatalities, and half had one; so on the average, there were two deaths per family, which matches the actual death toll precisely.

So why is there a problem? Remember, we will not be polling all of the survivors. We're going to poll only 50 people, on the assumption that a random sample will accurately reflect the entire group. Is that true of our hypothetical village? Of the 2,400 survivors, there are 1,350 from families where only one member died... but only 1,050 from families where three members died. We're not likely to get an equal number of survivors from each of our two groups: we will probably get more from the group where only 1 family member died. Imagine a raffle where one person buys 1,350 tickets, and another buys 1,050. Who has a better chance of winning? The same principle applies to the random sample: because survivors from families with a lower death rate outnumber the survivors from families with a higher death rate, the former are more likely to be selected for inclusion in the survey.

(There is a simulation of this effect, along with a more detailed description, at

This problem becomes much more pronounced as the number of deaths increases, and is particularly severe when deaths are heavily concentrated within individual families. Consider an extreme example: what if all 600 fatalities in the village occurred within a group of 75 families? Of our 2,400 survivors, there are now 2,250 who will tell us that no one in their family died, and only 150 who will tell us that eight members of their family died. With survivors from the untouched families outnumbering the others by a ratio of 15 to 1, it is highly unlikely that a random survey will give us an accurate estimate of the death toll.

In effect, a random survey conducted after a crisis reflects the composition of the survivors... and the survivors may be a very different group than the population in general at the onset of the crisis. The normal method of addressing this problem is to attempt to select respondents in a way that will accurately reflect the target population.

One could argue that by calculating mortality separately for different social groups, Kiernan has addressed this issue. Yet this is not correct; within each social group, the same dynamic exists: families with larger numbers of survivors will have a greater likelihood of inclusion.

The difficulty in obtaining a truly random sample may, in part, account for the fact that Kiernan's extrapolations from the survey are lower than the "direct" estimates when we calculate the toll based on the census data that Kiernan himself favors: while the survey suggested a toll of around 1.67 million, the direct estimates indicated a toll of around 1.797 million. (57)

Considering the fact that the death rates suggested by the surveys are already staggeringly high, it is probably difficult for interviewers to accept the idea that the people who are recounting such mind-numbing horror are, in a statistical sense, the lucky ones. The mind rebels against the idea that those who lost one-quarter of their families were more fortunate than most. Yet this may indeed be true, and it may account, at least in part, for the fact that Kiernan's estimate is lower than the figures advanced by Etcheson and Heuveline.

With regard to Heuveline's estimate, however, we might ask ourselves a question: if it is possible to alter the parameters of Heuveline's estimate to be consistent with Kiernan's analysis, is there a good reason to assume that it is Heuveline who is correct? I can find no compelling reason to dispute Heuveline's assumptions, nor his conclusions. Kiernan's estimate, however, is more problematic: both the forensic evidence and the laws of probability suggest that it is low. Moreover, the house-to-house survey conducted by the PRK also suggests that Kiernan's estimate is low, a point discussed in greater detail below.


A Compound Estimate

Is it possible to reconcile estimates derived from different methods? And are there other factors which have not been considered in any of the prevailing estimates? Should we re-evaluate any of the common assumptions? Possibly. If we consider the strengths and weaknesses of the existing estimates, and consider the implications of some additional data, we might arrive at a figure which fits reasonably well within the plausible parameters of each data point.

First, let's consider the question of the starting population. Are the prevailing assumptions about the growth rate reasonable? The 1970 estimate favored by Vickery reflects a growth rate of just under 2.7%; the higher estimates favored by Kiernan represent a growth rate of around 3%, approximately equal to the growth rate in Thailand for the same period. Laos, meanwhile, had a growth rate of around 2.4%. (58) Either estimate, then, seems plausible in light of trends elsewhere in the region. As noted previously, however, the lower estimates favored by Vickery cannot easily be reconciled with other data.

Kiernan's calculations are based on a starting toll of 7.894 million. Heuveline believes that it was around 7.952 million, but points out that it is probably more worthwhile to think in terms of a plausible range, given the uncertainty of the data. (59)

Which figure is more accurate? I would argue that Heuveline's figure is probably closer. Migozzi (whom Kiernan describes as "the leading demographer of Cambodia" (60) had, in 1972, predicted a 1975 population of 8.5 million; and as this estimate is from 1972, it seems likely that Migozzi was taking the war into consideration. W.J. Sampson, an economist and statistician working in Phnom Penh until March 1975, also believed that the mid-1974 population was around 7.89 million. (61) Demographers Bannister and Johnson, meanwhile, also suggested that the number of children aged 6 - 15 in the PRK's 1980 administrative count indicated that the war did not impact birth rates as drastically as was widely believed. (62)

Aside from the question of growth rates, estimating the 1975 population also requires making assumptions about the number of deaths during the 1970-1975 civil war. Heuveline and Kiernan rely on an estimate of approximately 300,000 deaths during the civil war. This figure, it should be noted, is already lower than the frequently cited estimates of 500,000 - 600,000 deaths. (63) Are these figures plausible?

There is substantial anecdotal evidence which suggests that the war toll is lower than what is commonly believed. In the many memoirs written by Cambodians in the aftermath of the Pol Pot time, it is surprisingly difficult to find firsthand accounts of deaths during the 1970-75 war. In Chanrithy Him's When Broken Glass Floats, Him describes fleeing her family's home in Takeo province, and describes finding the house destroyed upon their return. In spite of this, no one in the family is killed. She describes two deaths in the family, "children not touched by bombs but who might have survived if there had been access to hospitals and medical care." During the Khmer Rouge years, by contrast, Him lost 28 members of her extended family. (64) Haing Ngor, too, mentions the destruction of his father's house, but also does not mention any deaths in his family.(65) In Leaving the House of Ghosts, Sam and Sokhary You describes the bombing of their family's villages in Kompong Speu province; most villagers left immediately, but You's family remained for another six months, until another bombing raid destroyed their house; again, however, there is no mention of casualties. (66) Someth May and Thida Mam both describe seeing rocket and terrorist attacks on Phnom Penh, but do not mention any deaths in their families. (67) Vann Nath describes the death of a friend who was killed in an attack by the Khmer Rouge (68), but, again, does not mention any deaths in his own family. Memoirs by Loung Ung, Sophal Leng Stagg, Paul Thai and Molyda Szymusiak similarly do not discuss family deaths during the war years.(69)

It is important to remember, of course, that these authors are not a representative sampling of Cambodian society. Publishers look for compelling stories, and the harsh experiences of the "new people" are compelling indeed. Moreover, the well-educated urban elite would be those most likely to commit their experiences to paper. Additionally, there are factors which could tend to discourage those who suffered most during the war from relating their experiences. It is widely accepted that the destruction brought by the American bombing led many peasants to side with the Khmer Rouge. Consequently, it seems reasonable to assume that those confronted by the greatest hardship would be the most likely to have become Khmer Rouge. Considering the horrors subsequently committed by the Khmer Rouge during the DK years, however, these people are certainly not likely to be writing memoirs; oppressors do not make good protagonists.

Other evidence also suggests that the war toll should be re-evaluated. Anthropologist May Ebihara, who conducted fieldwork in a village in Kandal province in 1959-1960, returned to the village in 1990. Of the 159 people she had known in 1960, she found that by 1975, 16 persons had died from old age or illness, and 4 had died during the war. Of the 139 remaining people, half of them -- 69 people -- died during the Khmer Rouge regime. (70) Discussing Ebihara's research, Kiernan notes that "Eighteen new families had formed in the hamlet after 1960; but from 1975 to 1979, 26 of the 36 spouses and 29 of their children also perished." (71)

Ebihara's data highlights the disparity in the death ratios between the civil war and the Pol Pot regime. The number of deaths in 1975-1979 was roughly seventeen times the number of deaths during the war.

To demonstrate the implications of these ratios, let's return to Vickery's original estimates of 500,000 war dead, and 740,000 deaths during the Khmer Rouge regime. If this is accurate, one would expect that interviews with survivors would reveal a substantial death toll for both periods: there would have been roughly 2 deaths during the war for every 3 deaths during the Khmer Rouge regime. If we accept Kiernan's estimates, meanwhile, we would expect that for each family member who died during the war, five or six would have died under the Khmer Rouge.

However, even this ratio may still be too high. Steve Heder concluded from his interviews that the death toll among peasants was roughly seven times higher than the toll from the war. (72) We should also bear in mind that Heder referred to peasants, who in general were more affected by the war than those in urban areas, and slightly less affected by mortality in the Khmer Rouge years.

The 50% death rate in Ebihara's village was clearly higher than normal during the Khmer Rouge years; yet even if we applied a more typical death rate of around 25%, we would be still be left with a ratio of 1 to 8.5.

While the exceptionally high ratios from this village might not be typical, they were also certainly not unique. My own conversations with refugees also suggests very high ratios. When questioned about the Pol Pot years, most Cambodians will immediately begin listing names: "the terse tally of the dead," as author Minfong Ho once put it. (73) Yet when asked about deaths during the war, the list of names is almost invariably short: perhaps a cousin who was a soldier, an uncle whose fate was never entirely clear, and so on.

It is important to stress, however, that the subjects I have interviewed would not represent an broad cross-section of the overall population, or even the overall refugee population. One would expect that, in general, refugees represent the people who have suffered greatly under the regime they have fled; after all, it takes a great deal of hardship to motivate an individual to leave her or his own country. Additionally, the majority of the refugees I have interviewed were predominantly from either Battambang province, or Phnom Penh. Both of these locations were less affected by the war than other regions.

Another significant consideration is the likelihood that respondents may not immediately recognize some deaths as war-related. Establishing cause-and-effect is not always easy. As Craig Etcheson remarked, "Your buffalo gets blasted, so you can't plow your field, so your crop fails, so your kids get hungry, and then they get sick and die from beriberi or something." (74)

Clustering of mortality during the civil war might account for the difficulty of finding survivors who lost family members between 1970 and 1975. Casualties from bombing are likely to be heavily clustered; a bomb that falls on a house may kill an entire family, while leaving every other household in the area unscathed. The repression of the ethnic Vietnamese during the Lon Nol regime, too, would have resulted in a clustering of mortality among specific families. (75)

Another critical consideration is that those most likely to have died during the war -- that is, residents of the areas close to the Vietnam-Cambodia border -- were also among the most likely to die during the Khmer Rouge regime. The bloodiest purges of the Khmer Rouge reign targeted peasants in the Eastern zone. Substantially higher mortality among these families during the war might have affected our interpretations of the death rates in each period.

A final consideration is the possibility that the cultural emphasis on not offending others is relevant here: would Cambodians be hesitant to discuss deaths caused by Americans with an American interviewer?

Nonetheless, while one could theorize about what other evidence might have existed, it is not clear that these factors would account for the apparent discrepancy that we see when we compare the ratio of war deaths to deaths during the Khmer Rouge regime. The clustering of mortality in during the Khmer Rouge regime would likely have been even more pronounced than the clustering during the war. The repression and murder of ethnic minorities was even more severe. Lon Nol's wrath focused primarily on ethnic Vietnamese. Under the Khmer Rouge, however, ethnic Chinese, Chams, and Vietnamese all died in numbers in excess of the general population. Additionally, the paranoia of the Khmer Rouge meant that entire families were often murdered in the drive to exterminate "enemies." In many instances purges went beyond immediate family, and reached into extended families as well.

There are also other open questions regarding the war toll. What are we to make of Etcheson's assertion that only two mass graves contained victims of the bombing campaign? Of those who died in the bombing, how many were North Vietnamese, and not Cambodian? Consider the assertions by Noam Chomsky and Edward Herman, that contemporary press accounts of the war failed to quote victims of the bombing: was this because the press was unsympathetic, or was it because there simply were not as many victims as was widely believed? Compared to Vietnam -- where Communist guerrillas could strike anywhere at any time, and Americans and South Vietnamese could rely on rapid mobility to immediately retaliate -- it seems likely that the front lines in Cambodia were more predictable, making it easier to evacuate before battles took place.

Is a figure of 300,000 more reasonable than the estimates of 5-600,000? Using casualty rates from Vietnam as a guide, let's consider how 300,000 deaths would be constituted. In Vietnam, there were about 1,380,000 military deaths, and around 2,000,000 civilian deaths(76); this suggests a ratio of around 1.45 civilian deaths for each military death. If we assume a similar ratio in Cambodia, we would have around 122,488 military deaths, and 177,512 civilian deaths. It's exceptionally difficult to calculate the civilian toll, and (as noted previously) it is often difficult to determine precisely what constitutes a war-related death. But let's consider the number of military deaths. Again relying on statistics from Vietnam as a guide, let's consider the number of total casualties implied by roughly 122,500 deaths. The ratio of dead-to-wounded for American forces in Vietnam was approximately 1:3. For the Vietnamese communists, however, the lack of access to adequate medical care meant that wounds were much more likely to be fatal: the ratio of dead to wounded was roughly 2:1.(77) There are few statistics available for Cambodia, but government at the end of 1970 the government reported their losses 3,888 killed and 7,895 wounded. (78) It seems likely that this 1:2 casualty ratio probably deteriorated over time, as their battlefield situation worsened and limited medical resources were depleted. Khmer Rouge forces, meanwhile, probably suffered the same rate of dead-to-wounded as the Vietnamese communists. If the overall ratio was around 1:1, the number of wounded combatants would have been somewhere between 100,000 and 125,000, yielding a total military casualty figure of around 240,000. This seems exceptionally high when we consider the number of soldiers fielded by both sides. As of 1973, Khmer Rouge troop strength was estimated to be around 200,000. Lon Nol's forces, including paramilitary groups, probably peaked at around 290,000. (79) If we assume that all of the dead and wounded were replaced by fresh recruits as the war progressed, we would have a total of around 730,000 fighters. This would mean that nearly one out of every three combatants was either killed or wounded, an incredibly high rate in comparison with other conflicts.

The fact that this rate is exceptionally high does not mean that it cannot possibly be correct; after all, the death rates during the Khmer Rouge years were also exceptionally high in comparison to other regimes. In the sad history of human conflict, there will be one war that holds the distinction of being bloodier and more destructive than any other. However, there is no compelling evidence to suggest that the Cambodian civil war was markedly more destructive than other conflicts.

Bannister and Johnson estimated the death toll from the war to be around 275,000. (80) Sampson, too, believed that the toll from the war was overestimated. He suggested that civilian deaths "could be numbered in tens of thousands, but not more," and also noted that military attachés estimated the size of each army to be between 100,000 and 150,000. (81) If correct, these figures would mean that even a figure of 300,000 deaths is far too high.

We should also remember what the number of deaths implies in terms of the number of wounded. If we were to accept a figure of 500,000 dead, we would be expect to see least another half-million wounded; this would mean that about 1 out of every 16 people among the 1975 population would have been wounded during the war. Again, interviews with survivors suggests that this is cannot possibly be accurate. I would consider any figure between 150,000 and 300,000 as plausible, and would regard 250,000 as the most likely figure. (82)


Finding A Convergence

How does this affect estimates of deaths during the Khmer Rouge regime? Since Heuveline's estimate of the 1975 population (7.952 million) was presumably based on an assumption of 300,000 war deaths, I would suggest a figure of 8.002 million; the highest plausible starting population would probably be around 8.102 million.

Kiernan, meanwhile, suggests a starting population of 7.894 million. One point of contention, however, concerns the number of Vietnamese who were repatriated in 1975. Kiernan suggests 150,000; Vickery estimates 200,000. On the assumption that Vickery might be correct, we could adjust Kiernan's figure downward slightly; the resulting figure - 7.844 million - could be regarded as the lowest likely starting population.

As difficult as it is to determine the 1975 population, it is even more difficult to accurately estimate the 1979 population. Bannister and Johnson cite an administrative count, compiled by the PRK in 1980, and estimate the population at that point to have been 6,589,954.(83) However, they note Ea Meng-Try's contention that "local and provincial officials had financial reasons to exaggerate the numbers of people in their jurisdictions." (84) Kiernan suggests that the population in January 1979 was somewhere between 6 and 6.7 million, and his own calculations of the death rates would suggest a figure of 6.219 million. Other sources, however, suggest that the number of survivors may have been lower. A 1990 UNICEF publication, charting population trends in Cambodia, cites statistics from the PRK's Ministries of Health and Agriculture indicating a 1981 population, inside Cambodia, of 6,684,000. (85) The UNICEF analysis also cites an estimate of 6.7 million for 1980; (86) as this is higher than the 1981 figure, it seems likely that this included some 300,000 refugees along the Thai border. (87) These figures are probably derived from the same survey cited by Bannister and Johnson, and Vickery.

Estimating the "expected" 1979 population (that is, what the population would have been in the absence of excess mortality) requires estimating the growth rate, and the effects of immigration.

Kiernan suggests that the growth rate during the KR years was around 1% per annum;(88) he does not, however, cite any evidence for this assertion. Anecdotal evidence from survivors clearly indicates that fertility declined. But how drastic was that decline? It seems likely that Kiernan's estimate is indirectly derived from the survey results. Working backward from the presumed toll of roughly 1.7 million, one must accept a very low birth rate in order to reconcile the estimates of the 1975 and 1979 populations.

There is a far better way to determine fertility: cohort data. Naturally, there were no statistics on births compiled during the Khmer Rouge years; but it is possible to estimate the number of births based on the age structure of the population, as reflected in later census data. Based on the age structure of the population in the PRK's 1980 count, Bannister and Johnson concluded that fertility declined by about 34%. (89) Similarly, Heuveline and Bunnak Poch analyzed data from the Mekong Island Population Laboratory, and concluded that birth rates during the Khmer Rouge regime had declined by about one-third. (90) Lacking precise data on infant mortality during the crisis, these estimates will be imperfect. Drastically increased infant mortality would mean that the cohort data would cause us to undercount births; but in spite of this, the demographic data still suggests birth rates which are higher than the assumptions underlying Kiernan's estimate of 1% population growth.

It is important to remember that neither Kiernan nor Heuveline relies on population figures for a direct estimate. Examining their estimates of the beginning and ending populations, however, allows us to compare their assumptions more easily.

First, although both men suggest a similar figure for the 1975 population, Heuveline's analysis of the original census data seems to suggest a somewhat higher growth rate, which affects our interpretation of "normal" mortality. (91)

Second, Heuveline believes that the estimates for the 1979 population were inflated in an attempt to secure additional food aid. He suggests a figure of just under 6 million is probably more accurate. Other analysts, too, including Robert Muscat and Jonathan Stromseth, also suggested that the post-KR counts were inflated. (92)

Although there are other subtle differences in Heuveline and Kiernan's assumptions, it is probably fair to say that these two factors are primary.

How do different population estimates fit with the forensic evidence? Once again, let's consider the implications of DCCam's mapping project, using their current estimate for the mass graves: 1,386,734 remains. If these were exclusively victims of execution, and if executions accounted for half of the deaths during the regime (the figure suggested by Heuveline) (93), we would be left with some 2,773,468 deaths. Subtracting some 500,000 to account for "normal" mortality, we would conclude that the excess death toll was roughly 2.27 million. If executions were 40% (the figure cited by Sliwinski) (94), we would have 2,966,835 excess deaths.

For Kiernan's estimate of 1.67 million deaths to be accurate, we could make several different assumptions... but none of them seems particularly likely. We could assume that the mapping project's teams have significantly overestimated the number of remains, but there is no particular evidence to suggest that this is the case. We could assume that many of these mass graves are from the civil war, or from the Vietnamese invasion; but as Etcheson noted, there is significant evidence that they are not. Alternatively, we could assume that the overwhelming majority of those who died during the Pol Pot years were buried in mass graves, regardless of the cause of death. Again, however, this seems unlikely: refugee testimonies indicate that many of those who died from illness definitely were not buried in mass graves; and there were many others, such as those who died trying to escape, who would not have been buried in mass graves... if indeed they were buried at all.

We might also consider another bit of evidence, discussed only briefly in DCCam's mapping project report: the PRK's house-to-house survey. Etcheson noted that the PRK's 3.3 million figure was derived by combining the number of remains from mass graves, and the number of deaths reported as a result of the survey. He pointed out that combining these figures would result in the double-counting of many victims, and that the survey was also marred by other methodological errors, such as a failure to account for instances where the same deaths were reported by two families.

The decision to combine the totals from the exhumations and the survey was clearly incorrect, but what about the survey itself? The summary of the PRK's report separated the totals: 2,746,105 from the survey, and 568,663 from the exhumations. (95) If there is no reason to combine these figures, correcting that error is simple enough: we can simply disregard the figure from the exhumations, and focus our attention on the survey itself.

Because there appears to be no clear record detailing the methodology of the survey, we can only speculate on the statistical flaws. We can, however, make some broad guesses about what mistakes would be likely to occur, and from this we can infer an absolute minimum toll.

The most frequent error would be probably be when a single victim was included in tallies from both sides of the family tree: for example, both the paternal and maternal grandparents might report the death of a grandchild. This would be highly probable where the two sides of the family tree were in different villages. There might also occasionally be instances where deaths in widely dispersed families would be incorrectly recorded; to take an extreme example, the death of a father with four children, living in four different towns, might be recorded four times. We might also consider the possibility that the fate of many of those who fled to Thailand might be unknown to their families in Cambodia; individuals alive and safe in the refugee camps or resettled in other countries might have been listed as dead by relatives in Cambodia.

Since it seems likely that the most common error would be the double-counts, this should be our first focus when we attempt to estimate a margin of error. How often would these incorrect counts occur?

Generally, for a death to be double-counted, there would need to be survivors on both sides of the family tree. With death rates for some groups approaching 30%, this would often not be the case. And, while it might not be surprising to find duplicates where the families were widely dispersed, it is likely that Cambodian families generally remain fairly close to their relatives. Moreover, there is no reason to assume that the PRK's researchers were completely lacking in common sense. Suppose family A says that their grandchildren Sophal, Haing, and Chhea died in 1977; and family B, two houses down, says that their grandchildren Sophal, Haing, and Chhea died in 1977. It's difficult to imagine that the researchers would fail to recognize these as the same three victims. In larger communities, however, the task of identifying duplicate names could have been quite difficult. (Consider the magnitude of the task: compare the difficulty of finding 100 duplicates on a list of 5,000 items, versus the difficulty of finding 10 duplicates on a list of 500 items.) To correct for duplicates, then, let's make a very broad assertion: let's suppose that double-counts were avoided in the tallies from rural areas, but that every death reported by the urban population was counted twice. Bannister and Johnson suggest that about 12.6% of the population was located in the major cities in 1980, and the 1998 census describes 17.7% of the population as urban. (96) On the assumption that in general families from urban areas suffered more deaths than those from rural areas, we should adjust this upward significantly. Let's assume that 20% of the reported deaths were duplicates; in that case, we would subtract half of that number from our total, leaving us with 2,471,495.

Next, we could adjust for individuals counted as dead, who were in fact alive in refugee camps or other countries. It seems very likely that in most cases, the extended family would know of the escape in advance, or would realize subsequently that the others had fled; or, those in the camps would find a way to contact their relatives in Cambodia. In the interest of establishing a conservative estimate, however, let's assume that roughly one-third of those in the camps were mistakenly counted as dead. (97) Thus, we could subtract another 100,000, giving us a total of 2,371,495.

Our final adjustment would be to subtract from this figure the number of deaths that could be expected in normal conditions. The generally accepted figures for the pre-war death rate would suggest around 500,000 deaths. This would leave us with some 1.87 million excess deaths.

Coincidentally, this figure matches almost exactly the high end of the range proposed by Kiernan. In our ad hoc extrapolation, however, 1.87 million probably represents an absolute minimum toll, since we have corrected only for factors which would lead to an overestimate. There are other factors which would contribute to an undercount victims. At least three groups were unrepresented in the survey: families who were killed entirely, and families whose surviving members had all fled abroad, and families whose only surviving members remained in areas inaccessible to the census takers. Our assumptions here are predicated on an estimate of some 300,000 people in the refugee camps; and we know with certainty that the number of families killed entirely was substantial. Considering these factors, the survey strongly suggests a death toll in excess of 2 million.

Still, with no good data on the survey's methodology, we should consider some other possibilities as well: for example, did the researchers already attempt to adjust the results to account for underrepresented groups? Or, is it possible that the results simply falsified to demonize the previous regime?

Of course, if the survey results reflect the PRK's ideological bias, or simply gross incompetence, then we should regard other statistics from the PRK -- such as the 1980 population estimates -- with equal skepticism.

It seems likely that the survey was conducted to the best of the researchers' ability. Virtually every family in Cambodia bore the scars of the Pol Pot time. In the aftermath, every Cambodian wanted to know what really happened.


Tentative Conclusions

What does all of this tell us? There are still no precise answers. Existing evidence, however, strongly suggests a toll in excess of two million. I would suggest a "most likely" figure of around 2.18 million. Current evidence also suggests that the highest of Kiernan's estimates - 1.87 million - is probably the lowest plausible figure. The highest plausible estimate would be around 2.495 million.

These figures are calculated on the basis of the population and growth estimates, but any such calculations must be subject to certain constraints. I would argue that any estimate must fulfill at least five conditions:

A 1975 population of roughly 8 million seems highly likely. As noted, most demographers believe that a census will generally be an undercount, particularly in a underdeveloped country. Bannister and Johnson note that "A 1964 survey in Cambodia indicated that the 1962 census may have missed about four percent of the population." (98) Migozzi, too, believed that the census was an undercount; (99) and, as noted previously, Siampos and Heuveline believed that birth rates were underestimated. Meanwhile, prevailing estimates of the war toll were probably overestimated; Bannister and Johnson described their estimate of 275,000 excess deaths as "the highest mortality we can justify."(100)

Demographic evidence also suggests that the birth rate for the Pol Pot time did not decline quite as drastically as previously believed. Assuming a normal growth rate of 3%, and a normal crude death rate of 19 per thousand, the normal crude birth rate would have been approximately 49 per thousand. Heuveline and Poch estimated a 34% decrease in total fertility; and over a relatively short period of time, a decrease in fertility would have been accompanied by an approximately equal decline in the crude birth rate. (101) This would mean that the crude birth rate would have been around 32 births per thousand, which in turn would mean that the annual growth rate (crude births per thousand, minus crude deaths per thousand) would have been around 1.3%. A starting population of 8.002 million, compounded by an annual growth rate of 1.3%, would have meant an expected population of roughly 8.39 million in 1979. (102) I would regard this as the "most likely" figure for the expected population. Given the imprecision of the data, it is probably more reasonable to specify a plausible range, rather than a single figure. The high end of the range, assuming these same growth rates, but with a starting population of 8.102 million, would yield an expected 1979 population of 8.495 million. The low end of the plausible range, meanwhile, would be fairly close to Kiernan's estimate of an expected 1979 population of 8.215 million. On the assumption that Vickery's higher estimate of the numbers exiled to Vietnam might be correct, we could lower this figure by about 50,000, yielding an expected population of 8.165 million for 1979.

How many survivors were there? Bannister and Johnson suggest about 6.36 million inside Cambodia at the end of 1978. They also suggest a population loss of 218,000 due to migration. The majority of these however, were the 150,000 ethnic Vietnamese who were previously subtracted from the starting population; the remaining 58,000 fled to Thailand, Laos, and Vietnam before 1979. (103) The number of survivors, then, would be 6.418 million.

But was the 1980 count accurate? Heuveline believes that the surviving population was just under 6 million. (104) Bannister and Johnson acknowledge that the administrators compiling the 1980 count had an incentive to overestimate the number of persons within their jurisdiction, but also note that a census will generally be an undercount. Presumably because they viewed these two factors as offsetting one another, they did not adjust the figure. However, I would argue that if, as Bannister and Johnson suggest, we should treat the 1980 figure as an undercount typical of a census, we should also make a corresponding adjustment to our starting population to compensate for that probable undercount... and, as noted previously, such an adjustment would result in an even higher estimate of excess deaths. I'm also inclined to believe that many of those who fled to Vietnam during the civil war probably returned prior to the 1980 count.(105)

I am unaware of any evidence to suggest that Bannister and Johnson's figure would be an underestimate; thus, we could affix 6.418 million at the upper end of the range of plausible estimates. The lower end of the range would probably be roughly 6 million. In the absence of any precise data, I would regard the median - 6.209 million - as the most likely figure.

Subtracting this figure from the most likely expected population - 8.39 million - suggests roughly 2.18 million excess deaths. The range based on the figures above extends from a minimum of 1.747 million, to a maximum of 2.495 million. Although figures at the low end of the range seem unlikely on the basis of the PRK survey, the midpoint of the range - 2.12 million - falls fairly close to the most likely figure.


Matters for Further Study

All of these figures, of course, are still tenuous. Although it is nearly impossible that the toll will ever be calculated precisely, additional research can allow us to refine and improve our estimates.

I do not regard the analysis presented here as definitive. At best, I hope that it is catalytic: that is, that some of the arguments outlined here might provide ideas that would enable other researchers to make more accurate calculations.

In compiling these estimates, I make a number of assumptions in the absence of evidence to the contrary. This is a critical point, and it requires clarification: I am not claiming that other evidence does not exist. I am merely noting that I have not seen such evidence. Scholars with access to other resources will hopefully be able to elaborate on the points discussed here, and hopefully will be able to provide a more accurate analysis.

Additional research might clarify several issues. Among these:

Confronted with incomplete and inconclusive evidence, it is tempting to say that we will never really see the full picture of what happened in Cambodia. It is tempting to say that thirty years have passed, and it is time to move on.

Historians are accustomed to the idea of genocide. It is easy to accept that Cambodia was not the first occurrence of genocide; it is harder to accept that it was not the last. There are newer crimes; do we still need to worry about the old ones? Why we should bother with numbers? One and a half million, two and a half million... does it matter?

There was once a time when they were not merely numbers. They had names, and that is why it matters.




A number of people gave generously of their time and knowledge in the preparation of this article. They bear no responsibility for the conclusions, confusions, and outright errors in this article; those are mine and mine alone. For their kind assistance, I would like to thank the following people: Craig Etcheson, Patrick Heuveline, Steve Heder, Judy Ledgerwood, Damien de Walque, Ben Kiernan, Ed Moise, Carl Haub, Donald Kirk, John May, Suzanne Rubio, Jim Hauser, Maureen Sharp, Srey Sharp, Ra Thach.



(1)   A good retelling of this story is at

(2)   Narath Tan interview, 1991. See The Art of Survival, at

(3)   Pucci, Carol: "Touching the soul of Cambodia," Seattle Times, February 22, 2005.

(4)   Best, Joel: Damned Lies and Statistics, p. 35.

(5)   Power, Samantha: A Problem from Hell, p. 100.

(6)   Becker, Elizabeth: When the War was Over, p. 170.

(7)   Ponchaud, Francois: Cambodia: Year Zero, (p. 71.

(8)   Central Intelligence Agency (National Foreign Assessment Center): "Kampuchea: A Demographic Catastrophe," 1980. This report is available online at Two clarifications are necessary here. First, the report outlines three ranges of estimates: the High, Medium, and Low series. The figures cited here reflect the assumptions of the Medium series. Second, it is unclear to me whether or not the figure cited for population loss from April 1975 to January 1979 includes the loss from refugees who fled to Thailand; I believe it does not. The report estimates only about 55,000 refugees for this period, however, so the impact of this is not terribly great.

(9)   Elizabeth Becker, for example, cites a figure of "as many as two million." (When the War was Over, p.1) David Chandler refers to "over a million," The Tragedy of Cambodian History, p. 236.

(10)   Vickery, Michael: Cambodia 1975 - 1982, p. 188.

(11)   Vickery, p. 185.

(12)   Vickery, p. 185.

(13)   Vickery, p. 186.

(14)   Vickery, p. 187.

(15)   Kiernan, Ben: The Pol Pot Regime, p. 457.

(16)   Vickery, Michael: letter, posted to bit.listserv.seasia-l mailing list, 2/17/1997. Online at

(17)   ibid.

(18)   Bannister, Judith, and Johnson, Paige: p.91;, Patrick Heuveline personal communication, 4/6/2005.

(19)   Vickery, p. 185.

(20)   Heuveline, Patrick, and Poch, Bunnak, "Mortality and fertility changes: Cambodia before, during, and after the Khmers Rouges," p. 7.

(21)   Vickery, Michael: Letter to Z Magazine, May 9, 1994; retrieved from, March 2005.

(22)   Kiernan, The Pol Pot Regime, p.456.

(23)   ibid., p. 456.

(24)   ibid., p. 457.

(25)   ibid., p. 459.

(26)   ibid., pp. 456-457.

(27)   ibid., p.457.

(28)   ibid., p. 457.

(29)   ibid., p. 458.

(30)   Kiernan, Ben: "The Demography of Genocide in Southeast Asia," online at

(31)   Etcheson, Craig: "The Number: Quantifying Crimes Against Humanity in Cambodia," online at

(32)   ibid.

(33)   ibid.

(34)   ibid.

(35)   ibid.

(36)   Etcheson, Craig: personal communication, 5/6/2003.

(37)   Etcheson, "The Number," citing Vickery, Cambodia 1975 - 1982, pp. 187-188.

(38)   Etcheson, ibid.

(39)   Kiernan, Ben: "Demography of Genocide."

(40)   Vickery, Cambodia 1975 - 1982, pp.187-188. The original source Vickery and Kiernan are citing is Pin Yathay, L'Utopie Meurtrière, p. 149.

(41)   Kiernan, Ben: "Demography of Genocide."

(42)   Munson, Frederick P., et al: Area Handbook for Cambodia, p. 35; CIA, "Demographic Catastrophe."

(43)   Heuveline, Patrick: Personal communication, 5/20/2005.

(44)   Heuveline, Patrick: "The Demographic Analysis of Mortality Crises: The Case of Cambodia, 1970-1979."

(45)   ibid.

(46)   Heuveline, Patrick: Personal communication, 4/6/2005.

(46)   Birth rate underestimated: Heuveline, Patrick: Personal communication, 4/6/2005; Estimate of 1970 population: Heuveline, "Demographic Analysis," p. 121.

(47)   ibid., p. 106.

(48)   U.S. Census Bureau, International Data Base: Population Pyramid Summary for Cambodia, at; Population Pyramid Summary for Thailand, at

(49)   Heuveline, "Demographic Analysis," p. 112.

(50)   ibid., p. 123.

(51)   Heuveline, Patrick: Personal communication, 5/20/2005. In the "Demographic Analysis" article (p. 124), Heuveline suggests a figure of 2.2 million, but subsequently indicated that had inadvertently neglected to subtract the toll of the war. He stresses, however, that "any POINT estimate is by nature uncertain and that in the case of Cambodia, the uncertainty is high." (personal communication, 5/10/2005).

(52)   Documentation Center of Cambodia, "Master Genocide Site Data." The number of mass graves has been revised downward slightly, presumably because some sites were double-counted. The overall estimate of the number of victims contained in the graves, however, has been revised upward.

(53)   Shawcross, William: The Quality of Mercy, p. 50.

(54)   Hildebrand, George, and Porter, Gareth: Cambodia: Starvation and Revolution, p. 42.

(55)   Central Intelligence Agency, "Demographic Catastrophe."

(56)   Heuveline, "Demographic Analysis," p. 108.

(57)   Kiernan, "Demography of Genocide," p. 586. Kiernan cites a figure of 1,854,528 when calculating survivors based on the population estimates. However, I believe that this is slightly off. Bannister and Johnson suggest that in addition to the 6.36 million inside Cambodia, there were 218,000 refugees in Vietnam and Thailand. The majority of these (150,000) were the ethnic Vietnamese who were exiled when the Khmer Rouge came to power. Kiernan has already subtracted these from his starting population. However, this still leaves another 58,000 refugees, who should be added to the 6.36 million survivors. In this case, Kiernan's highest estimate (1.855 million) should actually be adjusted downward slightly, to about 1.797 million.

(58)   Thailand's growth rate: see Growth rate for Laos: see

(59)   Heuveline, Patrick: personal communication, 4/6/2005.

(60)   Kiernan, Ben: The Pol Pot Regime, p. 457.

(61)   Sampson, W.J.: Letter to The Economist, March 26, 1977.

(62)   Bannister and Johnson, p. 85. I'm uncertain how much credence I give this conclusion, however: I believe that Bannister and Johnson's population estimates are significantly low for the 1975 population, and their conclusion that birth rates had not been substantially affected could in theory be the result of this. A decreased population among the very young might appear roughly normal if the size of the entire population was underestimated. This, however, would be the case only if the very young were not undercounted along with all other demographic groups.

(63)   Vickery and the CIA, for example, accepted these estimates. Francois Ponchaud (Cambodia Year Zero, p. 71) notes that Sihanouk had cited a figure of 600,000. Elizabeth Becker's claim of more than one million deaths may have been based on this same statistic: Becker probably misinterpreted claims of 5-600,000 for both sides to mean 5-600,000 on each side.

(64)   Him, Chanrithy: When Broken Glass Floats, pp. 27-44, p. 330

(65)   Ngor, Haing, and Warner, Roger: A Cambodian Odyssey. See Chapters 3 and 4, and p. 129.

(66)   Streed, Sarah: Leaving the House of Ghosts, pp. 72-73. Although this is attributed to a B-52, that seems unlikely; the attacking plane was sighted "flying low in the distance." The B-52 is a high-altitude bomber.

(67)   May, Someth: Cambodian Witness, pp. 91-99; Criddle, JoAn: To Destroy You is No Loss, pp. 3-12.

(68)   Nath, Vann: A Cambodian Prison Portrait, pp. 3-4.

(69)   See Loung Ung's First They Killed My Father, Sophal Leng Stagg's Hear Me Now, Sharon Sloan Fiffer's Imagining America (about Paul Thai) and Molyda Szymusiak's The Stones Cry Out.

(70)   Ebihara, May: "A Cambodian Village under the Khmer Rouge, 1975-1979," in Genocide and Democracy, p. 57.

(71)   Kiernan, The Pol Pot Regime, p. 459.

(72)   Heder, Stephen, personal communication 3/8/2005.

(73)   I highly recommend Ho's novel, The Clay Marble, for young readers. A review is at

(74)   Etcheson, Craig: personal communication, 5/8/2003.

(75)   A clarification may be in order here. I've argued that estimates of the death toll from the Khmer Rouge regime are low, in part because of clustering of mortality. Would clustering of mortality could cause estimates of the war toll to be low, also? No. Underestimates are not caused by clustering; they are caused by survey methodologies that do not account for clustering. There is no evidence that the estimate of 500,000 war deaths was based on surveying survivors. In fact, there seems to be no evidence that it was based on anything at all.

(76)   Associated Press: "Vietnam Says 1.1 Million Died Fighting For North," April 4, 1995, retrieved from, retrieved 3/1/2005.

(77)   "Summary of Vietnam Casualty Statistics, Courtesy of VVA Chapter 172," online at, viewed 1/8/2005. This source indicates 47,378 American deaths due to hostile action, and 304,704 wounded in action; but of the wounded, only about half (153,329) required hospitalization. ARVN casualties are 223,748 KIA, and 1,169,763 WIA. NVA/VC losses are estimated as 1,100,000 KIA and 600,000 WIA. This very high ratio presumably reflects the lack of adequate medical care for Communist soldiers. If we discard the numbers of American wounded who did not require hospitalization, and calculate on the basis of the combined figures for US, South Vietnamese, and North Vietnamese/VC forces, we see 1,371,126 killed, and 1,923,092 wounded.

(78)   Clodfelter, Michael Vietnam in Military Statistics p. 273.

(79)   Kiernan, in Peasants and Politics in Kampuchea 1942-1981 suggests that the Khmer Rouge may have numbered 200,000 by 1973. He cites as his source 1973 hearings in the House of Representatives. (p. 219) Clodfelter's Vietnam in Military Statistics indicates that the Lon Nol army expanded from 40,000 to 100,000 in 1970. (p. 273) By late 1972, the on-paper strength of the Republic's forces was 223,000, but as many as 100,000 may have been non-existent "phantom" soldiers, and of the remainder, only about 80,000 were truly combat-ready. (p. 274) Late 1974 figures indicated 200,000 in the army, 11,000 in the Navy, an Air Force of 9,500, and another 150,000 paramilitary forces. Again, however, of these numbers, 80,000 or more were believed to be phantoms. (p. 277.) Clodfelter's figures also suggest that the Khmer Rouge forces were smaller than Kiernan indicates; he puts their 1972 strength at 35,000 (p. 274), and indicates 60,000 fighters in 175 battalions in 1975. (p. 277)

(80)   Bannister and Johnson, p.66.

(81)   Sampson, letter to The Economist, March 26, 1977.

(82)   One other factor that complicates precise calculations of the death toll centers on the fact that there is almost inevitably excess mortality in the wake of war. However, it is not always the case: consider the implications of a brutal regime, overthrown by war. The Vietnamese invasion of Cambodia, in fact, is a good example of an instance when war led to a decrease in mortality.

(83)   Bannister and Johnson, p.79.

(84)   ibid., p. 91.

(85)   UNICEF, Cambodia: The Situation of Children and Women, p. 8.

(86)   ibid., p. 9.

(87)   ibid., p. 9.

(88)   Kiernan, "Demography of Genocide," p. 586.

(89)   Bannister and Johnson, p.86. They suggest a pre-crisis fertility rate of 7 births per woman, and a rate of 4.6 per woman for the Khmer Rouge years.; this represents a decline of 34.3%.

(90)   Heuveline, Patrick, and Poch, Bunnak, "Mortality and fertility changes: Cambodia before, during, and after the Khmers Rouges," p. 2.

(91)   I should stress that the growth rates I discuss here are not figures supplied by Heuveline; however, I believe that they do accurately reflect the trends he describes.

(92)   Huguet, Jerrold: "The Demographic Situation in Cambodia."

(93)   Heuveline, "Demographic Analysis," p. 125.

(94)   cited in Etcheson, "The Number."

(95)   Etcheson, personal communication, 2/23/2005.

(96)   Bannister and Johnson, p. 68 - 12.6% in major cities in 1980. For comparison, a 1966 estimate had cited a figure of 11% (Munson et al, Area Handbook for Cambodia, Oct. 1968, p. 32). The 1998 census, indicating 17.7%, is available online at

(97)   The UNICEF publication had suggested around 300,000 refugees in the camps in the early 80s. (Cambodia: The Situation of Children and Women, p. 9). The assumption that one-third of these people were mistakenly reported dead is, in my opinion, probably a substantial overestimate. It supposes that there were 100,000 people, on the very border of their home country, who failed to contact any of their surviving relatives inside Cambodia.)

(98)   Bannister and Johnson, p. 91.

(99)   Kiernan, "Demography of Genocide," p. 586.

(100)   Bannister and Johnson, p. 87.

(101)   Heuveline, Patrick: personal communication, 4/6/2005; Haub, Carl (Population Reference Bureau,, personal communication, 4/4/2005.

(102)   This figure is derived by compounding the 1975 estimate by .86% to reflect the remaining eight months of 1975, and compounding this figure by 1.3% for 1976, 1977, and 1978.

(103)   Bannister and Johnson, p.90, pp. 121-122.

(104)   Hueveline, Patrick: personal communication, 4/6/2005.

(105)   There is no question that there was a substantial amount of immigration from Vietnam, but it is difficult to determine numbers, and difficult to determine dates. My wife, who lived just outside Phnom Penh after the fall of the Khmer Rouge regime, believes that there was a great deal of immigration at this time. Another friend who lived in the Kampuchea Krom section of Vietnam at the time also indicated that many Vietnamese returned (or moved to) Cambodia shortly after Pol Pot's fall.



Bannister, Judith, and Johnson, E. Paige: "After the Nightmare: The Population of Cambodia," in Kiernan, Genocide and Democracy in Cambodia, 1993.

Becker, Elizabeth: When the War Was Over, Simon and Schuster, 1986.

Best, Joel: Damned Lies and Statistics, University of California Press, Berkeley and Los Angeles, 2001.

Chandler, David: The Tragedy of Cambodian History, New Haven, 1991.

Clodfelter, Michael: Vietnam in Military Statistics, McFarland, North Carolina, 1995.

Criddle, JoAn: To Destroy You is No Loss, East/West Bridge, 1987.

de Walque, Damien: The Long-Term Legacy of the Khmer Rouge Period in Cambodia, World Bank Policy Research Working Paper 3446, November 2004, online at

Documentation Center of Cambodia, "Master Genocide Site Data," May 26, 2003.

Hildebrand, George, and Porter, Gareth: Cambodia: Starvation and Revolution. Monthly Review Press, New York, 1976.

Him, Chanrithy: When Broken Glass Floats W.W. Norton, 2000.

Heuveline, Patrick: "The Demographic Analysis of Mortality Crises: The Case of Cambodia, 1970-1979," in Forced Migration and Mortality, National Academies Press, 2001.

Heuveline, Patrick, and Poch, Bunnak, "Mortality and fertility changes: Cambodia before, during, and after the Khmers Rouges," Population Research Center, NORC and The University of Chicago, 2002(?)

Huguet, Jerrold: "The Demographic Situation in Cambodia," Asia-Pacific Population Journal, Vol. 6, No. 4, 1991.

Kiernan, Ben (ed.): Genocide and Democracy in Cambodia, Yale University Southeast Asia Studies, New Haven, CT, 1993.

Kiernan, Ben: "The Demography Of Genocide In Southeast Asia - The Death Tolls in Cambodia, 1975-79, and East Timor, 1975-80."

Kiernan, Ben: Peasants and Politics in Kampuchea, 1942-1981. London, Zed Press, 1982.

Kiernan, Ben: The Pol Pot Regime, Yale University Press, New Haven, 1996.

Ledgerwood, Judy: Personal communication, 2/25/2005.

May, Someth: Cambodian Witness, Random House, New York, 1986.

Munson, Frederick P., and Martindale, Kenneth W.; McMorris, David S.; Parachini, Kathryn E.; Raiford, William N.; and Townsend, Charles: Area Handbook for Cambodia, U.S. Government Printing Office, 1968.

Nath, Vann: A Cambodian Prison Portrait, White Lotus, Bangkok, 1998.

Ponchaud, Francois: Cambodia Year Zero, Holt, Rinehart and Winston, 1978.

Pucci, Carol: "Touching the soul of Cambodia," Seattle Times, February 22, 2005. Online at

Sampson, W.J.: letter, The Economist, March 26, 1977.

Shawcross, William: The Quality of Mercy, Simon and Schuster, 1984.

Sloan Fiffer, Sharon: Imagining America, Paragon House, New York, 1991.

Streed, Sarah: Leaving the House of Ghosts, McFarland, Jefferson, NC, 2002.

UNICEF: Cambodia: The Situation of Children and Women, Phnom Penh, 1990.

Vickery, Michael: Cambodia 1975-1982, South End Press, Boston, 1984.

Vickery, Michael: letter, Z Magazine, May 9, 1994.

Yathay, Pin: L'Utopie Meurtrière, Paris, 1980.

W3C Validation