Noam Chomsky, Crooked Timber and crooked numbers


Posted to on July 4, 2006


Over on the Crooked Timber group blog, which occupies a kind of no-man's-land between academia and the normal world, there's been a heated discussion of Noam Chomsky--how could it be otherwise?


These periodic flare-ups always involve charges that Chomsky is a David Irving-like "revisionist," who minimizes civilian deaths in Srebrenica. Among the Chomsky bashers is one Oliver Kamm, an oleaginous bond trader from London who has written a book titled "Anti -Totalitarianism: The Left-wing Case for a Neoconservative Foreign Policy" that ranks #73,123 in sales at By contrast, Chomsky's "Failed States" ranks #352 on So I guess the average person is not that much into neoconservative foreign policies. Can't say you can blame them with reports of GI's raping and murdering Iraqi civilians appearing with depressing regularity nowadays.


People like Kamm have a tough job. It is really an uphill battle to get anybody with a brain to make common cause with the Donald Rumsfelds and Paul Wolfowitz's of the world, unless your brain is fairly rotted out with alcohol like Christopher Hitchens'. Lately some of these pro-war leftists have given up on the whole project, most notably David Rieff who wrote the following in his new book "Point of the Gun":


"At the time of the Kosovo war, I had written that, if I had to make the choice, I would choose imperialism over barbarism. In retrospect, though, I did not realize the extent to which imperialism is or at least can always become barbarism."


Berkeley professor Brad Delong, another well-known Chomsky-basher, has entered the fray as well. He has gotten all worked up over Chomsky's statement that "such journals as the Far Eastern Economic Review, the London Economist, the Melbourne Journal of Politics, and others elsewhere, have provided analyses by highly qualified specialists who have studied the full range of evidence available, and who concluded that executions have numbered at most in the thousands." Evidently, this puts him in the same category as holocaust deniers who are always trying to minimize Jewish deaths.


Delong once wrote an article on his blog titled "My Very, Very Allergic Reaction to Noam Chomsky: Khmer Rouge, Faurisson, Milosevic" ( that described him as an intellectual totalitarian and an apologist for Stalin's crimes. Like a piece of dirt in an oyster, this article serves a higher purpose, namely to allow Edward Herman to dismantle it:


Since Delong was Deputy Assistant Secretary for Economic Policy in the Clinton administration, it should come as no big surprise that he like to misuse statistics to support questionable political objectives. The Treasury Department, under Republicans and Democrats alike, has always seen fit to exaggerate economic growth and minimize unemployment. When they hired Brad Delong, they obviously understood that he knew how to juggle the books like a seasoned performer.


For all the venom directed at Chomsky's alleged misuse of statistics, Delong has no problem counting on Rudy Rummel as an expert when it comes to determining how to rank mass murderers in the 20th century. Rummel is a retired U. of Hawaii professor who has written a number of books that exaggerate the number of deaths under Communism and minimize those that the U.S.A is responsible for. On his website, there's a table that lists the USSR as first with nearly 62 million dead. Remarkably, the U.S.A. is not even listed. This imbalance, bordering on American holocaust denial, is shrugged off by Delong who believes "that there is really no other place to start assessing the history of the twentieth century" when it comes to the numbers that Rummel has produced.


Not everybody is persuaded by Rummel's statistics, least of all a shrewd and canny person named A.J. Philbin who got sick and tired of trolls crossposting Rummel's nonsense to the newsgroup alt.politics.socialism.trotsky. This is what he (or she) thought of Rummel's work (


Would it surprise you to learn that I have downloaded and printed the online version of Dr. Rummel's tome, Lethal Politics? I don't know if it is the same as the printed version, but I assume, since it is linked to his academic website, that old Doc Rummel wouldn't have allowed it to go up if it didn't reflect the substance of his printed work. I haven't read the whole thing thoroughly, but I've read enough to come to the following preliminary conclusions, based on my admittedly cursory reading (subject to revision if and when I have time to give it a more thorough read):


1) Rummel is clearly not a statistician, and the sloppiness of his methods show that he didn't consult one, at least not routinely or in depth. This is obvious from both the method he uses and his own inadvertent statements in the text. His method consists of taking widely (I would say wildly) varying estimates from sources he regards as "authoritative" and averaging them. I saw nothing that indicates how he decide how much credence to give these vastly differing estimates. He also uses categories that clearly are greatly overlapping, but makes no attempt to compensate for the overlap. (I'm no statistician, but the errors are so glaring that I am forced to wonder how Rummel could have missed them).


2) Rummel desperately wishes to prove his basic thesis that government sponsored internal repression has killed more people than war. This is his a priori bias, and it colors his analysis throughout, and renders him susceptible to making (or using) huge estimates as components of his totals.


3) Since Rummel provides much of the data he used, and makes statements that inadvertently show how his bias affected his conclusions, I cannot conclude that he is dishonest. The only reason I suspect any possible deliberate dishonesty (as opposed to credulousness and self-delusion), is the glaring nature of some of the errors. I am a layman, and yet I spotted them without much trouble. Nonetheless, his own inadvertent statements (as well as his apparent inability to grasp their full import) tend to exculpate Rummel of deliberate fakery, while convicting him of credulousness and apriori bias in the first degree. I will now demonstrate what I have stated above.


On the third page of the chapter entitled 61,911,000 Victims: Utopianism Empowered, second full paragraph down, Rummel makes the following statement:


"In sum, the Soviets have committed a democide of 61,911,000 people, 7,142,000 of them foreigners. This staggering total is beyond belief. But, as shown in Figure 1.1, it is only the prudent, most probable tally, in a range from an highly unlikely, low figure of 28,326,000 (4,263,000 foreigners); and an equally highly unlikely figure of 126,891,000 (including 12,134,000 foreigners). This is arange of uncertainty in our democide estimates -- an error range -- of 97,808,000 human beings."


Incredible. This statement demonstrates, concisely and clearly, everything I stated above. Let's start with a priori bias. Rummel characterizes the low estimate (28,326,000) and the high estimate (126,891,000), as being "equally unlikely." Killing 28 million people requires a stupendous, if horrific, effort of political will and organization. yet a figure more than four times that amount is only "equally unlikely?" Let us leave aside the fact that this high figure is greater than the 1910 population of the Russian Empire (120 million), the present day population of European Russia (120 million), and more than half the present day (1992) population of the former USSR (223 million). Only an extremely biased scholar could claim these two figures are "equally unlikely," and not instead conclude that his component estimates are virtually useless. But that isn't the worst, even in this one paragraph. The last sentence demonstrates the worthlessness of his chosen method. The error range (97,808,000) is larger than his "estimate" (61,911,000). I don't have to be a statistician to know that when your estimate is smaller than your error range, your method is about as good as a ouija board -- wait, I take that back: the ouija board has the advantage of being less laborious and more fun. So does a divining rod.


But Rummel goes on, sealing both his conviction on grounds of credulousness and his acquittal on charges of fakery. In the very next paragraph, he exclaims:


"Just consider the error range in Soviet democide, as shown in Figure 1.1. It is larger than the population of 96 percent of the world's nations and countries. Actually, if France, Belgium, the Netherlands, Norway, Sweden, Denmark, and Switzerland were blasted clean of all human life in a nuclear war, the human toll would be less than just this range in the Soviet's probable democide -- the range, and not even the total murdered."


Exactly what are we supposed to conclude from this? Why, those Soviet commies must have been unbelievable monsters -- look at the error range they caused in my estimates!! They must have truly been genocidal (excuse me, democidal) butchers to cause me to find such wide ranging death toll estimates! Just imagine how many people they REALLY must have killed in order to create such a wide disparity in available statistics!


But let us turn to Table 1.1. On the horizontal axis, Rummel lists the following democide components: terror, deportation, camps, and famine. Rummel makes a good case for separating deportation and camp deaths in the text, and mentions that he includes transportation deaths in both categories. But I could find nowhere an explanation of what he meant by "terror," or why "terror" is separated from camps and deportations. While famine might be clearly a separate category, but in the case of the USSR, given the claims that famine was an instrument and result of "terror," this is not at all clear. Isn't transportation and existence in the GULAG considered part of "terror?" It would certainly terrify me! "Terror" is not a method of killing -- it is a description of a general policy or period of repression (which certainly includes killing). Indeed, on the vertical side of the table, the "Great Terror" of 1936 is listed as a period in which all the horizontal categories occured. We are told that, during the Great Terror of 1936, that 3,280,000 people died in the camps (or en route), 65,000 from deportation, and 1,000,000 from "terror." Yet most other periods mentioned on the vertical side show more deaths from "terror" than the Great Terror of 1936. Clearly, "terror" is a category that is partially overlapped by famine, and perhaps almost totally by camps and deportations. "Terror" is supposed to account for 8,298,000 deaths overall. This is assuming the figures are correct in the first place, which I am not, and have dealt with above. There are other examples, this being merely the first one I found. I will relate them if you ask.


Essentially, what Rummel did was to compile hearsay estimates, do sloppy statistical work, and plow ahead to a conclusion that the data couldn't support. The only legitimate coclusion that he could have come to was that no conclusion was possible regarding probable death tolls in the USSR, based on the figures he had. But who wants to publish a book with such an anti-climactic ending?


BTW, the figure of up to 10,000,000 killed, given by Otto Pohl in this thread, is the most common figure I heard when I was associated with the Trotskyite movement. I don't know what these figures were based on, but clearly they were closer to the mark, and formulated with greater caution than either Rummel's estimate, or the estimates he based his work on.


A.J. Philbin