How many died? There will never be a satisfactory answer to that question, if only because in the midst of the great famine so few reliable statistics were kept.
So far, every noteworthy estimate has been based on the official figures on population size and on birth and death rates for 1950–82, published for the first time in the 1984 Statistical Yearbook by the National Statistical Bureau, or on the official figures of the 1953, 1964 and 1982 censuses. Immediately following the publication of the Statistical Yearbook, Basil Ashton used the official evidence to propose a figure of 30 million premature deaths during the 1958–62 period, when the overall population stood at roughly 650 million.1 Judith Banister, a professional demographer, also looked at the population statistics and concluded that an estimated 30 million excess deaths appeared during 1958–61.2 Since the data present a whole range of problems, from lack of internal consistency to the under-registration of births and deaths and the exclusion of the armed forces, different authors have tinkered with this or that variable either to lower or to heighten the number. Peng Xizhi, an expert in population studies, proposed an estimate of 23 million in 1987, while Jung Chang, in her book on Mao, reached a figure of 38 million.3 More recently, retired journalist Yang Jisheng suggested a figure of about 36 million – also based on published statistics.4
New evidence was produced in 2005 when Cao Shuji, an historical demographer from Shanghai, systematically worked his way through more than a thousand gazetteers – official local histories published after 1979 by county or city party committees. While acknowledging that this widely diverse set of data, too, ultimately rests on figures made public by the party, it introduced a much more fine-tuned analysis of regional differences. Cao’s estimate was 32.5 million premature deaths.5
How reliable are official numbers? In the Soviet Union the Central Office of Statistics produced two sets of demographic statistics, one for internal use and one for publication. But as we have already seen in the case of grain procurements, the party archives in China have widely different sets of statistics at every level, from the commune, the county and the province up to the centre. Some were compiled at the height of the collectivisation craze and were intended to convey political zeal. Others were assembled by investigation teams sent to the countryside to oversee the removal of abusive party officials. In other words, debates about whether the released figures are doctored or not miss a very basic point. There is no need for anybody to falsify figures, it is merely a matter of compiling a set of statistics which appears to be the least politically damaging. Or, to put it slightly differently, the fact that public data in a one-party state are not falsified does not necessarily make them reliable.
At least three different sets of unpublished data exist in the archives, namely those compiled by the provincial Public Security Bureau, those from the local party committee and those from the local Statistical Bureau. Nobody has ever gained access to all three sets. But after 1979, as the new leadership wanted to find out more about the Maoist era, a team of 200 was instructed by Zhao Ziyang to go around every province to examine internal party documents. The erstwhile secretary of Guangdong who had pioneered an anti-hiding campaign in 1959 was now premier, and he asked the team to draw a picture of rural China. The team’s report was never published, but one of its members, a senior party official called Chen Yizi, fled to America in the wake of the Tiananmen massacre in 1989. In exile, he claimed that the team had arrived at a death toll of 43 to 46 million people for the famine.6 Only one person who has investigated the famine has taken Chen Yizi’s claim seriously – namely Jasper Becker, who interviewed him for his book Hungry Ghosts published in 1996. The archival evidence presented for the first time below vindicates Chen Yizi’s findings and conservatively puts the number of premature deaths at a minimum of 45 million for the great famine of 1958–62.
Even Chen Yizi and his team would have encountered difficulties in carrying out their research. Archives, in a one-party state, are not public. They belong to the party and are controlled by the party. Except for those under the remit of the Public Security Bureaus, most of them are located in a building inside the party headquarters. Even a high-powered delegation from Beijing could have been fobbed off or deliberately misled by experienced archivists, all the more so since a catalogue would not have existed for every collection. But, most of all, some sets were simply missing. In Hubei, for instance, the file from the party committee which should have contained the figures for all excess deaths during the famine is incomplete. Inside the brown folder is a handwritten note appended by an archivist, dated June 1979, which regrets that the item is ‘missing’.7 As to the Public Security Bureau, in Hubei it offered no more than a vague estimate, speculating that the death rate in 1961 was two to three times lower than the preceding year. The report wonders about the total death toll but provides no answers.8
In any event, all three organisations – the provincial public security, the provincial party committee and the provincial Statistical Bureau – would have had to rely on units lower down the party hierarchy to compile their reports. And obstruction from below was rife. In Gansu province, the provincial party committee sent out a request for estimates in 1962 for excess deaths during the famine. The project foundered, as only a handful of counties ever replied.9
But even when numbers were sent in by county authorities there were problems. First among these was the distinction between ‘normal’ and ‘abnormal’ deaths. Demographers distinguish between ‘natural’ and ‘unnatural’ deaths to tease out a rough estimate of how many people died prematurely as a consequence of famine. But in China the distinction was political. Industrial accidents, suicides, fatal epidemics or deaths from starvation were all a matter of great concern to the authorities. They stood as indices of social and political health, and they were diligently monitored by the party’s regulators. Even a single case of suicide could signal that something was amiss, warranting a political investigation from above. In the middle of mass death in Fuyang, one of the sites of horror in Anhui where up to 70 per cent of some villages were wiped out, the region reported 10,890 deaths for the first quarter of 1961 – of which a mere 524 were described as ‘abnormal’, including 103 deaths due to ‘emaciation’ and ‘oedema’.10 In Rongxian, Sichuan, the county head Xu Wenzheng simply dictated that in the official statistics two rules had to be followed: birth rates had to exceed death rates, and the death rate could not be higher than 2 per cent. In Fuling, also in Sichuan, two sets of statistics were kept. For 1960 local cadres managed to count a total population of 594,324 people but reported 697,590, a difference of more than 100,000.11
Even when cadres were willing to confront the harsh reality of famine, who could have kept track of an avalanche of death? In Jiangjin and Jiangbei counties in Sichuan, up to 250 people died each day in December 1960: the last thing on the minds of local officials would have been to do the rounds each day to produce a neat list of mortality figures, even if they were specifically asked by their superiors to do so.12 When local cadres or police officers did try to report the full extent of death they were generally labelled rightists. Zhao Jian, head of the Public Security Bureau in Wenjiang county, Sichuan, systematically compiled statistics for 1959 and discovered that 27,000 people, or 16 per cent, were missing compared to the previous year. He was taken to task by his superior at the provincial level but refused to modify his findings, which led directly to his political demise.13
To make matters even more complicated, obfuscation went all the way to the top. Provincial boss Liu Zihou – like so many others – dutifully reported 4,700 ‘abnormal deaths’ to Chairman Mao for all of Hebei in 1960, even though his own team of inspectors had discovered that in one county alone some 18,000 had died of hunger since 1958.14 The irony is that he chastised county leaders for covering up the extent of the famine, all the while keeping the incriminating figures from his own superiors in Beijing.15 At every level party officials badgered their subordinates for the truth but were deceitful with their own superiors, contributing to a maze of self-deception. To say that knowledge is power is a truism, and one that does not go very far in explaining why the more absolute power was, the less truth it managed to produce.
But death on such a scale could hardly be hidden all the time. Sometimes local leaders took a chance, sending in hard-hitting reports further up the hierarchy, occasionally directly to Zhou Enlai or Mao Zedong himself. Extraordinarily detailed reports compiled by the investigation teams that fanned out over the countryside after October 1960 led to the removal of a whole series of leaders who had presided over mass death. And sometimes retrospective investigations were carried out in the years following the famine, as the party tried to make sense of what had happened. The result is not so much a neatly arranged set of statistics revealing some absolute truth in a few telling numbers, but rather a mass of uneven and at times messy documentation compiled in different ways, at different times, for different reasons, by different entities, with different degrees of reliability. So assigning a team of 200 people to sift through the evidence would have been a good idea.
The very best of these documents were compiled by a powerful Security Bureau and covered an entire province. As we have seen, this did not happen in Hubei, but it did in Sichuan – by far the most devastated province in all of China. The head of the provincial Security Bureau authorised an investigation into the statistics from 1954 to 1961. The results undermined many of the reported totals, which underestimated the death toll by several per cent in 1960 alone. The corrected death rate for 1954 to 1957 was an average of 1 per cent. This increased to 2.5 per cent in 1958, to 4.7 per cent in 1959, to 5.4 per cent in 1960 and to 2.9 per cent in 1961. It added up to 10.6 million deaths from 1958 to 1961, of which 7.9 million were above 1 per cent and can thus be considered ‘excess deaths’.16 But in Sichuan, unlike the rest of the country, famine did not vanish in 1962. There are countless reports about continuing starvation from a range of counties until the end of 1962. The Public Security Bureau compiled figures which determined that 1.5 per cent died that year, meaning that another 300,000 perished prematurely, bringing the total to 8.2 million.17 Yet even this figure is no doubt too low by at least 10 or 20 per cent, if only because in Sichuan – unlike in other provinces such as Gansu – the party boss Li Jingquan remained firmly in power despite his responsibility for the deaths of many millions of people. Even in 1962 few county leaders in Sichuan would have been prepared to report the full extent of the disaster.
No other similar documents are available – so far. But we do have data collected by regional statistical bureaus. In the case of Yunnan, where the famine started in 1958, the death rate recorded for the year was 2.2 per cent, double the national average for 1957: this alone would have amounted to 430,000 excess deaths, when most historians using official statistics mention only about 800,000 deaths for the entire 1958 to 1961 period.18
The best available evidence comes from carefully compiled reports at the village, commune and county level. Since the work of the historical demographer Cao Shuji, who used published party gazetteers to estimate death rates on a county basis, is in agreement with other population specialists who propose a death toll of roughly 32 million, it provides a very helpful baseline. Common sense indicates that local party committees had every incentive to underestimate published death rates, and in that sense Cao Shuji’s estimate should be considered conservative. The purpose of what follows is to test his figures and provide a rough idea of how they should be adjusted. Not only is a focus on smaller entities such as counties much more accurate than larger aggregations at the national level, but it also allows us to eliminate so many of the variables that have confused demographers working with censuses, from internal migration to the size of the army between 1958 and 1962.
However, an average death rate is required in order to calculate ‘extra’ death figures. What would be reasonable? Here is what Liu Shaoqi, the head of state, thought in 1961 when discussing the famine in his home town of Huaminglou, where hundreds died every month: ‘What are normal deaths? What are abnormal deaths? If you hit a man once and he dies of his injuries, or if somebody jumps into a river, it qualifies as an abnormal death. You can take the figures of the last two years to calculate a normal death rate . . . A normal death rate is below 1 per cent, in general 0.8 per cent, a normal birth rate is 2 per cent, any deaths above 0.8 per cent are abnormal.’19 To err on the safe side, given the wide variations across the country, 1 per cent should be taken as a normal death rate.
In the case of Hebei, we have some very detailed reports for 1960, compiled after provincial boss Liu Zihou gave the green light by asking for investigations into abnormal deaths ‘down to the level of the household’. Hu Kaiming, an outspoken party official in charge of Zhangjiakou who later incurred the wrath of Mao Zedong for proposing greater freedom for farmers to determine their own prices, reported that 1.9 per cent of the population had died in 1960, amounting to 59,000 people. In Weixian county, adjacent to Zhangjiakou, the death rate was 3.4 per cent in 1960, as 18,000 people died.20 That amounted to some 40,000 excessive deaths in one year. Using official documentation, Cao Shuji’s figure for excessive deaths in Zhangjiakou and Weixian is 15,000 for the entire three years of famine.21 In Tianjin and the surrounding countryside – hardly the most deprived part of the country – 30,000 people died within three months at the end of 1960. A normal rate of attrition would have claimed less than half of these lives. The figure provided by Cao Shuji, again based on official rather than archival sources, is 30,000 excess deaths for three years.22 Another example comes from Shijiazhuang, the seat of a region covering some fifteen counties. By reading the official data critically, Cao arrives at 15,000 deaths for the entire region over three years. But in the city of Shijiazhuang alone, close to 4,000 people died in a mere ten days in January 1961, when counting the victims of starvation was no longer politically taboo.23
Tianjin, Zhangjiakou and Shijiazhuang were cities nominally isolated from starvation in the countryside. A very different example comes from Gansu, where the demotion of Zhang Zhongliang in November 1960 was followed by several months of local investigations, bringing to light the extent of the famine. In Longxi county, 16,000 people died in 1959, or 7.5 per cent, followed by 23,000 in 1960, or 11 per cent of the population. So in those two years alone the excess deaths stood at 35,000. But for three years of famine Cao Shuji reaches 24,000 deaths.24 Party archives give 32,000 deaths for Jingning county, about 7 per cent each year in 1959 and 1960. This contrasts with a figure of 19,000 excess deaths for a period of three years arrived at by Cao Shuji.25 In Zhangye, out of a population of roughly 280,000, some 5,000 people died in November, followed by 6,000 in December 1960. Even if we double the normal rate of attrition to 2 per cent, that would still represent over 10,000 excess deaths in less than a quarter of a year. Cao Shuji arrives at 17,000 excess deaths – not for one county in two months, but for four counties over a period of three years.26 In the spring of 1960 some 20,000 people died in Wuwei county alone. Cao Shuji suggests 50,000 premature deaths for a region comprising four counties, over a period of three years.27
In Guizhou the provincial party committee reckoned by 1961 that some 10 per cent of the workforce was missing when compared with 1957 – meaning half a million workers, not counting children and the elderly.28 Not all of these had died, of course, as many had migrated out of the province, but the death rates were high throughout Guizhou, in particular in regions such as Chishui and Meitan. In Chishui some 22,000 people died in half a year – or 10 per cent of the population.29 Cao Shuji, using the official record of the county, proposes 46,000 over a period of three years, which seems reasonable enough. But in the case of Meitan, 45,000 people died in half a year. Cao Shuji suggests 105,000 for four counties over a period of three years, which must be too low.30 Even more interesting is that in his extremely conscientious compilation of the official data for all counties, some places are missing: Yanhe, part of the Tongren region, is not mentioned, although some 40,000–50,000 people died of hunger in that county alone.31
In Shandong the discrepancies are of a similar magnitude – even if few of the relevant archives can be accessed by historians. In Pingyuan county, to take an example from the north-western part of the province, a high-ranking investigation noted that out of a population of 452,000 in 1957, over 46,000 people had died by 1961. Despite 24,000 births, the total population dropped to 371,000, as tens of thousands took to the roads to escape from famine – many to die elsewhere, their deaths being excluded from these figures. Cao Shuji’s examination of the official annals proposes 19,000 premature deaths for Pingyuan county. Even if we take into account a normal death rate of 1 per cent per year over a period of four years, the total of excess deaths reported at the time would be equivalent to 28,000, or 50 per cent higher.32 A similar observation can be made about Qihe, which lost a fifth of its population, or 100,000, between 1957 and 1961. If we deduct a normal death rate of 1 per cent for four years and accept that roughly half of the vanished probably migrated to other areas (the document is not clear on this issue), we are still left with a figure comparable to Pingyuan, or roughly 30,000, although Cao Shuji ventures no more than 19,000, or a third less.33 For the entire Laizhou region, consisting of Qingdao and thirteen counties, Cao Shuji’s estimate is 164,000 premature deaths over four years. But the archives show that in Jimo county alone, according to incomplete statistics, some 47,000 people died (excluding 51,000 farmers who took to the roads) over a period of two years. Even if we deduct 15,000 normal deaths for a population of approximately 750,000 people, it still leaves the county with 32,000 premature deaths – far above Cao Shuji’s estimate.34
In some cases the archival data and the published material are in line. In Xinxing county, Guangdong, 1.5 per cent of the population died in 1959, followed by 2.88 per cent in 1960. This would have amounted to roughly 5,000 deaths, while Cao Shuji arrives at a total of 8,000 deaths for three years.35 For the much larger region of Jiangmen, also in Guangdong, encompassing several counties, the death rate given to the provincial party committee was 2 per cent in 1960 (or 120,000 deaths, half of which would count as ‘premature’). This is difficult to compare with Cao’s reconstruction of the official data, as the administrative borders of the region were extensively redrawn after 1961, but they do seem roughly to fit his estimate of 112,000 excess deaths for three years.36 In the case of Sichuan, as noted above, political pressure under Li Jingquan meant that few if any counties reported high death rates, and none match those found in the official documentation published decades later and consulted by Cao Shuji.
None of this is intended as a criticism of Cao Shuji’s work: on the contrary, his painstaking reconstruction of what happened at the county level, on the basis of well over a thousand local gazetteers, has established a baseline which is very much in accord with figures derived by demographers from more abstract sets of population statistics. A systematic comparison of these figures with archival data compiled at the time or in the immediate aftermath of the famine would not be possible without his work. And when we confront the official data with archival evidence we find a pattern of underestimation, sometimes by 30 to 50 per cent, sometimes by as much as a factor of three or four.
Perhaps some of the reports exaggerated the death rates, but it is very difficult to see why. There was no political advantage to be gained from declaring extra deaths. The death toll was not a major consideration in the purge of party members after October 1960. The manner of death mattered, as local cadres were classified according to different levels of abuse. In fact there was every advantage in inflating the overall population. When a team went to investigate the statistics in Hunan in 1964 it found that the overall population was systematically inflated by more than 1 per cent, in some counties by up to 2 or 3 per cent. The difference for 1963 was half a million people in Hunan who existed on paper alone: ‘through thorough testing we found that in the past the population figures were routinely and severely inflated’.37 When the Ministry of Public Security undertook a more widespread check on population statistics in 1963, it discovered a similar pattern of inflation across the country, sometimes as high as 2.2 per cent in the case of Gansu, for instance. ‘Of a population of 681 million today, we estimate that about 1 to 1.5 per cent of those counted are fake. Many local cadres, in order to obtain greater cloth rations and other goods, intentionally increase the population figures.’38 A year later, during the official 1964 census, the Central Census Office confirmed that ‘the problem of population inflation is far worse than we thought’, as at least a million was added for Hebei and Henan each, and no fewer than 700,000 for Shandong, three of the provinces that had been closely investigated: there was very little that could be done about the issue.39
Even if we ignore some of the most glaring disparities between archival data and official figures, the gap is in the order of 50 to 100 per cent. It is very difficult to venture an alternative death toll, all the more since so many of the key sets of archival statistics remain prudently under lock and key, far removed from the eyes of prying historians. But there is enough archival evidence, from a sufficiently large diversity of party units, to confirm that the figure of 43 to 46 million premature deaths proposed by Chen Yizi, who was a senior member of a large working group that sifted through internal party documents around 1980, is in all likelihood a reliable estimate. The death toll thus stands at a minimum of 45 million excess deaths.
It could be even worse than that. Some historians speculate that the true figure stands as high as 50 to 60 million people. It is unlikely that we will know the full extent of the disaster until the archives are completely opened. But these are the figures informally discussed by a number of party historians. And these are also, according to Chen Yizi, the figures cited at internal meetings of senior party members under Zhao Ziyang.40 Yu Xiguang, an independent researcher with a great deal of experience, puts the figure at 55 million excess deaths.41