Arguing from More Errors: Reply to Stephen Wheatcroft (with a postscript at the end)
Mark B. Tauger, Department of
This note replies to Stephen Wheatcroft’s recent response (Wheatcroft 2007, pp. 845-866; all subsequent references to page numbers alone refer to this article) to my article “Arguing from Errors” (Tauger 2006) regarding his and Robert Davies misrepresentation of my data and analysis regarding the Soviet famine of 1931-1933. I regret having to write again to defend my work, so my reply will be concise.
Statistics reliable and unreliable
Wheatcroft begins with the statement that “[I]t is impossible to be precise about the absolute reliability of grain statistics in this period” (p. 845), but I did not claim “absolute reliability” for my data, only that they are the correct category of data to estimate the harvests and therefore the causes of the famine. Wheatcroft’s assertion seems inconsistent, however, with his emphasis (p. 854) on a small, statistically insignificant difference he calculated, using these same data as a basis, between his new estimate of the harvest yield and my earlier figures. Wheatcroft writes that he made a “transcription error” in an effort to prove wrong my estimate of the 1932 harvest, offers an apology for this, which I appreciate, and presents a new calculation of that harvest much closer to the conclusion I published. In this new calculation Wheatcroft attempts to distinguish between sown areas of MTS and non-MTS collective farms and weight the data more toward the latter, which according to the annual report data on average had higher yields than MTS farms. This recalculation is ostensibly interesting but it has serious conceptual problems that Wheatcroft does not address and does not seem to be aware of.
First, it involves even more approximation and substitution of surrogate
data than my original calculation. Since the annual report data are not
available for all parts of the USSR, in my original article in 1991 I
interpolated official harvest figures for those data. Since those data
are preharvest forecasts, they are virtually always substantially higher yields
than the actual harvest yields, so such an interpolation inherently must mean
that the resultant average is inflated compare to what the real average harvest
would be if only final harvest data were used, that is, if we had all the
annual report data rather than part of it. I was aware of this problem
acutely in writing that article and I emphasized this point and noted that the
average harvest that resulted from this calculation must be inflated. From his 2007
article Wheatcroft appears totally unaware of this fundamental problem in the
data. This discrepancy between the two types of harvest data involved in
his calculation undermines his assertion that his data imply a larger average
harvest than mine because it means that the slightly higher figure he cites has
to be reduce because it is partly based on preharvest official data.
Second, since so many of his sown area estimates differ from the figures published in Sel’skoe Khoziaistvo SSSR in 1936, and not always in the same direction (p. 863, compare columns AE and AF), he needs to address these differences. He needs to at least explain why his planted area numbers differ so much from the official figures and ideally present some evidence that would justify these differences, that would show that the statisticians in the USSR would not have and did not perform the same calculations. In other words, his "new" calculation raises more questions than it answers, and because it avoids the central issues of different categories of data, it only obscures the real issues .
In Table A2 Wheatcroft attempts to approximate a comparison between the 1931 and 1932 harvests with even more interpolated and approximate figures. In this table, also (pp. 865-866), at least part of the figures in the columns headed “grain prodn in th tons” have their decimal point shifted to the left: the harvest in Ukraine in 1931, for example, was of the order of magnitude of 14 million tons, not 1.4 million tons, and the total Soviet harvest was of the order of magnitude of 59 million tons, not 5.9 million tons. Since errors here may not be uniform, and the data are incomplete, the table is too inconsistent to be meaningful or usable.
Wheatcroft’s new calculation for the 1932 kolkhoz grain yield results in of 5.8 centners per hectare instead of the 5.4 centners figure calculated by the Commissariat of Agriculture and 5.65 centners that I calculated using official sown areas, differences of approximately 9 percent and 4 percent respectively. All of these figures are within each other’s margin of error. and as noted above, Wheatcroft completely avoids discussing the fact that his figure has to be reduced because it is based partly on preharvest official yield data. It appears that when he addresses the data and calculations I published, Wheatcroft emphasizes “it is impossible to be precise,” but when he discusses his calculations, he emphasizes minor differences that are not significant if the data are as uncertain as he maintains.
Since “it is impossible to be precise” with these data, and since Wheatcroft’s calculations include so much interpolated data (such as sown areas and some yield data from official figures which are strictly speaking not compatible with the annual yield statistics) and differ so little from mine, his evidence does not support his assertion that my use of evidence is “oversimplified”, my “methodology faulty” and my “conclusions overall are wrong” (p. 846). Wheatcroft’s condemnation of my work as “overall wrong” (p. 846) when our estimates differ by less than five percent, and when he fails to consider the other data that I use, seems problematic in an article in which he condemns scholars who “in an attempt to make a name for themselves are prepared to distort and ignore the achievements of other serious scholars” (p. 858).
There are two other problems with Wheatcroft's points here. First, I seriously doubt his description of his use of the 30 million hectare figure for the sown area in Siberia instead of the 4.5 million actual area as a "transcription error." The difference between 30 million and 4 million seems too large to be such an error. I am suspicious that he attempted to slip false data into his calculations to discredit me surreptitiously. Second, however, his "apology" is only limited to that error; he does not "apologize" or even mention his much more serious error that he ignored the fact that in my first article I had already performed and included evidence of the calculation he claimed I did not do - a weighted average of the harvest data. I pointed this out in my EAS article in 2003, I even included in that article my table from my 1991 article, and Wheatcroft still avoided mentioning it. His attempt to inpugn my methods again by his slightly altered calculation (which as I will point out below is spurious), represents an attempt to avoid the issue that he either completely overlooked or forgot an entire substantial portion of my work (which would imply less than competent work on his part) or that he intentionally denied that had done this calculation, which would appear to mean that he lied about my work in a conscious effort to discredit me. I hope that readers can understand how frustrating and unjust it is to have one's work attacked in this false and unfair way in a journal in which I am not allowed to reply.
Wheatcroft then criticizes my use of the annual report data in two inconsistent passages. First he writes that “[W]e need to consider the assessments of grain harvests in each year in terms of the relative assessments for other years. Tauger’s failure to do this and to act as though his figures have some absolute significance is the main problem with his approach” (p. 845-46). Does he mean my “failure … to act” as though these figures have some absolute significance or my acting as though they do have significance? In any case, he then writes that “[C]omparing the low uncorrected harvest evaluation of 5.4 ts/ha for kolkhoz grain yields in 1932 with the high corrected figure for 1931 or the official figure for 1931 or 1932 is misleading. It might tell us something about the disjuncture between the government’s view of agriculture and reality, but it tells us very little about how agricultural production had been changing over time.” (p. 848) According to Wheatcroft, therefore, I was wrong because I did not compare the 1932 data with assessments from other years, but I was also wrong because I did compare that data with assessments from other years!
In fact, of course, in all of my publications on this topic I have clearly discussed the 1932 harvest in the context of harvests in the 1930s, with new data on the 1930 and 1931 harvest (Tauger 1991, 76; Tauger 2001b, 40-45). And a comparison between the kolkhoz annual report data, which are final harvest data that Wheatcroft concedes represent “reality,” and official figures for 1932, which are pre-harvest projections or “biological yields,” does documents the gap between an official view of agriculture and “reality” and can be used to tell us a great deal about how agriculture was changing.
Wheatcroft also attempts to minimize the significance of the annual report data by attempting to assimilate the annual report data to some of the 1920s disputes over Soviet grain harvest estimates, even though these are two completely different categories of data ( pp. 847, 849). In fact, as I have shown, the 1932 annual report data are final measured harvest data, the data in the 1920s (with extremely rare exceptions) are pre-harvest projections, speculations, or just guesses, in addition to their distortion by political pressures as Wheatcroft noted, and Soviet peasants and government officials were aware of this distinction (Tauger 2001b pp. 53-60). The annual report data therefore cannot be assimilated to or categorized with any of the 1920s figures as Wheatcroft attempts to do.
One particularly egregious example of this is Wheatcroft’s assertion that the kolkhoz yields were below the average yield for all sectors, and on that basis raises his estimate of the 1932 harvest yield to 6.1 ts/ha (p. 854). This “average yield for all sectors” is pre-harvest projections. We have little or no final harvest data for sovkhozy or non-collectivized peasants, but what we do have casts doubt on his assumption about larger yields in these other sectors. I published data showing that sovkhozy in the North Caucasus had abysmal yields in 1932, and sovkhozy in Ukraine could not fulfill a small procurement quote despite having by official figures (that is, pre-harvest projections) a sufficient harvest, which suggests that their actual harvest was much smaller than these pre-harvest estimates (or guesses). Also, while non-collectivized peasants appeared by (even more uncertain) official statistics to have a slightly higher harvest yields, their dramatically low fulfillment of procurement quotas at least suggests that those data may overstate their actual production (Tauger 1991, 83-84). Wheatcroft’s upward correction thus is clearly based on data that are unreliable and inconsistent with the annual report data. It is not statistically valid to alter final harvest data on the basis of speculative and uncertain pre-harvest estimates and call the result “corrected.”
After these attempts to discredit my estimate of the 1932 harvest as too low, Wheatcroft then asserts that he and Davies have argued since the 1970s that “the 1932 harvest was much smaller than the official data showed,” citing a series of articles (p. 846). This again appears contradictory: my “conclusions overall are wrong” (p. 846) when I prove on the basis of data from tens of thousands of farms that the 1932 harvest was much lower than the official data, but they were right to speculate that this was the case. A careful perusal of their publications shows that they have not argued consistently that the harvest was much smaller than the official figure. Not all of the items that Wheatcroft lists to support this point actually discuss the 1932 harvest directly or provide statistical estimates of its size. Wheatcroft’s first discussion of the 1932 harvest, in a SIPS paper, used a yield estimate of 6.7 centners (compared to the official figure of 6.8), so in that work he did not argue that the harvest was much smaller than the official figure (Wheatcroft 1977, pp. 18, 19). In the 1984 article he cited (which he mistakenly listed as from 1983) he provided a high estimate of 68 million tons, which is essentially the official estimate, and a low one of 62 million tons, less than ten percent lower (Wheatcroft 1984, 42). In a 1994 publication their estimate for the 1932 harvest was 56 million tons plus or minus ten percent, which at the high end (61-62 million tons) is only about ten percent less than the official harvest estimate (Davies, Wheatcroft, and Harrison 1994, p. 286).
Wheatcroft attempts to conclude this issue by writing that “[T]he important question is how did the 1932 harvest compare with those of 1931 and earlier years? And Tauger tells us nothing new in this important area” (pp. 858-859). The annual report harvest data I published proved beyond doubt that the 1932 harvest was actually much smaller than the official figures and than any harvest in the period, which undermine the “man-made famine” view of that crisis so prevalent at that time and today. These data outweigh all the conjectural estimates of and speculative arguments for harvests slightly smaller than official estimates that Davies, Wheatcroft, and others published earlier. Wheatcroft attempts to dismiss my work as “nothing new” when Slavic Review accepted my 1991 article rapidly with extremely positive referee reports emphasizing that the article presented new and valuable information. Both Robert Conquest and Ukrainian nationalist scholars repeatedly attacked the article because of the challenge it posed to their interpretation (in other words for presenting a new perspective with new data) without, however, refuting it (see Conquest 1992 and 1994, Kuzio 2002 and Tauger 2002). Wheatcroft’s attacks, with their inaccurate data, illegitimate statistical practices, contradictory criticisms, and weak appeals to insignificant calculation differences, implicitly recognize that my work has presented something new.
Misunderstanding the Famine
Wheatcroft yet again misrepresents my work in his discussion of the causes of the small 1932 harvest. He asserts that he and Davies viewed the famine as caused by a complex of interacting factors, and then asserts “all of these factors worked dialectically, instead of in the isolated manner implied by Tauger.” (Wheatcroft 2007, p. 857) In fact, in “Natural Disaster and Human Actions in the Soviet Famine of 1931-1933” my discussion of the famine is at least as complex and dialectical as Davies’ and Wheatcroft’s. I take into consideration not only the problems of farm organization, farm labor and draft forces, and state policies, but also a broader set of environmental factors than Davies and Wheatcroft consider (Tauger 2001a).
Wheatcroft errs in his discussion of the rust infestation of 1932. He writes: “Losses from rust may well have been slightly larger than normal, but my understanding is that rust by itself is unlikely to have been a major cause of the losses. Rust attacks the grain in its flowering stages. When the grain is harvested losses from rust cease” (p. 857). In fact, as I documented from numerous highly-respected American, European, and Soviet scientific sources written at the time of the famine and after (including Soviet sources published long before and long after Lysenko gained and then lost control over Soviet agricultural genetics studies), rust affects a plant over a much longer period, in complex ways, and with quite devastating effects (Tauger 2001a, 12-14). Soviet studies of the rust infestation in 1932 estimated that losses from rust alone approximated seven million tons of various grains, which (contra Wheatcroft’s “understanding”) constituted significant losses. These are also the only clearly documented losses to the 1932 Soviet grain harvest attributed to a specific environmental cause recorded in any sources so far available (Tauger 2001a, 17). Of course when the plant is harvested, rust no longer stops growth because the plant has been killed by harvesting, but Wheatcroft again fails to acknowledge the scale of damage that the infestation had done while the harvest matured.
More generally, Wheatcroft’s discussion regarding statistics, especially his doubt that the 1932 harvest was the smallest harvest of the period, ignores the central issue in understanding the famine which I raised: the relationship between the official data of grain availability and the fact of an enormous famine that peaked specifically in 1933. As I documented, the grain procurements from the 1932 harvest were substantially smaller than those from any other harvest in the 1930s. Since the government procurements in 1932 were lower than other years, and the relief provided was greater than in other years, by the official data there should have been more left in the villages in early 1933 than in other years (Tauger 1991, pp. 73-74). This was clearly not the case: it is well documented that the villages were much more short of food in early 1933 than in early 1932 or any other year in the 1930s and as a result several million people died, and the situation was similar in most urban areas, which also had dramatically elevated death rates.
The only explanation compatible with these facts is that the 1932 harvest was extremely small, much smaller than any other harvest in the early 1930s, and the procurements, reduced though they were, took away a larger share of the available food from the villages in fall 1932 than in any other year without fully satisfying the food requirements of the urban population. Wheatcroft and Davies were at least partly aware of this, because the last section of their conclusion in The Years of Hunger is a discussion of why the 1932 harvest was so small (Davies and Wheatcroft 2004, 434ff). Yet neither in that work nor in this 2007 article does Wheatcroft address the basic inconsistency between smaller procurements from the 1932 harvest but much greater mortality afterwards. The evidence that the 1932 harvest was much smaller than that of 1931 (and other years as well) provides the only reasonable explanation of this discrepancy, and therefore of the causes of the famine (Tauger 1991, pp. 84-89).
Wheatcroft concludes his article with two points. First, he attempts to justify his attack on me as doing a “disservice to the field” when I described the “political interpretation” of the famine as the dominant one (p. 858) by dismissing some of the evidence I presented, and by claiming that my disagreement with him is a matter of interpretation. If this evaluation is only a matter of interpretation, then my statement cannot be considered a “disservice to the field.” In fact, of course, and as I documented, the view of the famine as a genocide is widely accepted among historians, the public, and many governments around the world. That a few specialized scholars hold a different view does not make their view is “dominant,” even if they are right.
Second, Wheatcroft concludes by referring to my arguments about the low 1932 harvest and the environmental factors that caused it as a “discovery” (i. e., with quotation marks, which appears to imply sarcasm or mockery; p. 859). In fact, no Western scholar had ever published the annual report data before my 1991 article (the Soviet scholar V. P. Danilov did so incompletely and inaccurately; Tauger 1991, pp. 80-81, fn. 29, and p. 85) or incorporated them into a discussion of the famine. These data provided clear proof of conclusions that Davies and Wheatcroft and certain other scholars had asserted but never documented, and also brought to light factors overlooked in all earlier scholarship. It seems to me that honest, open-minded scholars would welcome new evidence and arguments that supports their earlier conclusions, rather than viewing them with resentment and casting about for false data, minor or unjustified criticisms, and veiled ad hominem attacks in order to attack them.
1. By admitting that he used false data in his web table, Wheatcroft has admitted that the assertions that he and Davies made about my conclusions in The Years of Hunger, p. 444, were incorrect. This fact, which Wheatcroft evades in his reply but which is incontestable logically, has several implications.
In that passage they listed my work as different from other views, in particular describing my estimates as “at the other extreme,” which means that my views are in fact new and add something to the field, and Wheatcroft’s admission in this article that he used false data now means that his criticism of my conclusions are no longer valid, and therefore he cannot dismiss my views as invalid for failing to take into consideration the variations in the data. In other words, the wholesale condemnation of my work in the book is now proven, and I emphasize proven, to based on errors, and therefore his condemnation of my work is also based on errors. So his claim in this recent article that my methods are wrong, my conclusions are wrong, etc., is simply an attempt to resuscitate an attack on me that his own admission of errors simply discredits completely.
2. The fact that Wheatcroft does not address these implications of his admission of error, and the fact that he evades discussing the other errors he made such as his failure to acknowledge that I had already calculated a weighted average in my original article in 1991, means that he is again evading admission of his errors and his responsibility for these errors. In other words, Wheatcroft is guilty here of several offenses against acceptable scholarly practice.
One of these offenses involves a violation of integrity that is essential and basic for historical research. The following passage from the Statement on Standards of Professional Responsibility of the AHA contains the key points:
Professional integrity in the practice of history requires awareness of one’s own biases and a readiness to follow sound method and analysis wherever they may lead. Historians should document their findings and be prepared to make available their sources, evidence, and data, including any documentation they develop through interviews. Historians should not misrepresent their sources. They should report their findings as accurately as possible and not omit evidence that runs counter to their own interpretation. They should not commit plagiarism. They should oppose false or erroneous use of evidence, along with any efforts to ignore or conceal such false or erroneous use.
Wheatcroft did not commit all of these violations but he did commit some of them. He misrepresented his sources, particularly his discussions of my work; he did not report his findings accurately; he omitted evidence that ran counter to his own “interpretations,” and he used false evidence and used evidence erroneously.
But he violated basic principles on another level as well. Wheatcroft committed “bad faith.” Bad faith is a legal term which refers to actions to take advantage of or manipulate the law on the basis of malicious intentions that are not the intent of the law. Applied here, Wheatcroft took advantage of the openness and freedom of scholarly publication to attempt to discredit my work and my capabilities as a scholar in order to advance his own reputation. He acted in bad faith in The Years of Hunger, and he did so again in this latest article.
I realize that few people are likely to read this item and some may misunderstand it, and Wheatcroft may try to attack it in his usual inept and evasive style. I present this page as a supplement to my publications, and all I can hope is that in the long run, enough people will find out about his incompetence and his attempts at character assassination to realize that his work is not the best source on the topics he studies.
Conquest 1992: Robert Conquest, Letter to the Editor, Slavic Review 51 no. 1, Spring 1992, pp. 192-194.
Conquest 1994: Robert Conquest, Letter to the Editor, Slavic Review 53 no. 1, Spring 1994, pp. 318-319
Davies and Wheatcroft 2004: R. W. Davies, S. G. Wheatcroft, The Years of Hunger, New York 2004.
Davies, Wheatcroft, and Harrison
1994: R. W. Davies, Mark Harrison, and S. G. Wheatcroft, The Economic
Transformation of the Soviet Union 1913-1945,
Kuzio 2002: Taras Kuzio, “Denial of Famine-Terror Continues Unabated,” RFE/RL Poland, Belarus, and Ukraine Report, Prague, Czech Republic, Vol. 4, no. 24, 12 June 2002.
Tauger 1991: Mark B. Tauger, “The 1932 Harvest and the Famine of 1933,” Slavic Review, 50 no. 1 (Spring 1991)
Tauger 2001a: Mark B. Tauger,
Natural Disaster and Human Actions in the Soviet Famine of 1931-1933,
Carl Beck Papers no. 1506,
Tauger 2001b: Mark B. Tauger,
Statistical Falsification in the Soviet Union, The Donald W.
Tauger 2002: Mark B. Tauger,
“What Caused Famine in
Wheatcroft 1977: S. G.
Wheatcroft, Soviet Grain Production Statistics for the 1920s and 1930s, CREES
Discussion Papers, SIPS no. 13, University of
Wheatcroft 1984: S. G.
Wheatcroft, “A Re-evaluation of Soviet Agricultural Production in the 1920s and
1930s,” in Stuart, R. C. (ed.) (1983) The Soviet Rural Economy (
Wheatcroft also writes that “[I]n several places in his original article Tauger had referred to this figure of 5.4 tsentners per hectare as if it was an official NKZ evaluation. This is incorrect and I attempted to point this out in terms of the data set alone…” (p. 853). Unfortunately Wheatcroft does not provide any citation for this claim. I used this figure (which came from a NKZ archival source, so I am certainly justified in describing it as “NKZ data”) to estimate the 1932 grain harvest, but I made it clear that this was an approximation (Tauger 1991, pp. 82-83). Wheatcroft’s claim that I use this number “as if it were an official NKZ evaluation” misrepresents the ways I described and used it.
He also does not address the crucial point I made in relation to the memoir accounts that the harvest “appeared” good. Rust, which can kill plants, usually allows the plants to grow and appear normal, but the grain heads will not “fill,” so that the harvest will consist of smaller or fewer grains, with more husks and fibrous materials. In one Soviet study, a 100 percent infestation reduced the weight of 1,000 grains of wheat more than 60 percent. (Tauger 2001a, 13) Thus a grain field infested with rust can look normal to the non-specialist, and it will still produce a harvest, but the harvest will be small because the grains will not have “filled.” This helps explain why the famine surrogates in peasant markets were so full of “chaff”, and why the harvest seemed to many peasants to be unusually “light” (Tauger 2001a, p. 15).
Wheatcroft attempts to support his doubt about the 1932 harvest with “consumption data,” but these are extremely incomplete, speculative, and circular because they are based on assumed norms of consumption rather than on actual conditions (p. 850). These “consumption data” do not take into account the thousands of villages that were left with practically no food after the 1932 procurements, despite smaller procurements that year.