China's agricultural crisis and famine of 1959-1961: a survey and comparison to Soviet famines.
The Great Leap Forward (GLF) campaign of 1958-1961, which aimed at quickly transforming China into a powerful industrial state, involved dramatic turns of events, but ended as a national catastrophe. In the 1958 New Year's editorial of the People's Dally, the Chinese Communist Party proclaimed the GLF goal of surpassing the United Kingdom in industrial production in 15 years and the United States in 20-30 years. The nation was soon elevated to a state of euphoria, as news of exaggerated production miracles spread throughout the country like wildfire. However, dreadful reality quickly set in. In 1959, China's grain output suddenly declined by 15%, and in the following 2 years, food supply plunged further, to 70% of the 1958 level. During the same period, a widespread famine raged across China. Years later, based on population census and fertility survey data, demographers were able to estimate that the total excess mortality during the GLF crisis ranged from 16.5 (Coale, 1981) to 30 million people (Banister, 1987). This monumental scale of the famine makes it arguably the largest in recorded history.
However, during the two decades after the famine ended, there was no public recognition of its existence outside of China. Even within China, despite awareness of severe hardship during the crisis years, the magnitude of the famine was largely unknown because the Chinese government prohibited scholarly inquiries into the subject. The public in China and the West began to realise the full severity of the Chinese famine only in 1980, when demographers and Sinologists released systematic research findings. (1) Since the publication of a complete time series of crude mortality rates in the 1983 Statistical Yearbook of China, and the subsequent release of population census and fertility surveys, this catastrophe has received increasing attention among social scientists. Initially, academic research was focused on reliable estimation of excess deaths and delayed births associated with the famine. Then, following the seminal work of Lin (1990) on agricultural collectivisation and performance in China, economists intensified research effort on the causation of the collapse in production as well as the concurrent famine. Research findings include a 1993 special issue of the Journal of Comparative Economics and a 1998 symposium issue of China Economic Review. Thanks to recent progress made possible by retrospective surveys and the collection of archived data, researchers are able to conduct rigorous econometric studies assessing the joint significance of various factors contributing to the GLF crisis. Much has been learned about what caused the precipitous decline in grain output, why mortality climbed to the colossal scale, and why the catastrophe lingered so long.
This article offers an overview of the growing literature on the economic analysis of China's GLF. (2) To establish a reliable factual basis for subsequent discussion, the next section documents the estimates of existing studies on the scale of excess mortality and the magnitude of decline in grain production during the crisis. Relying on Sen's entitlement approach to famine (Sen, 1981), we clarify several conceptual issues that are important for the analysis of China's GLF, particularly with regard to the role of central planning and the relationship between aggregate food supply and the causation of famine. The subsequent section examines the determinants of the decline in output, while the penultimate section analyses the determinants of extra mortality. In each section, brief expositions of economic hypotheses are followed by discussions about data issues and econometric strategies for estimation. We also report up-to-date empirical findings.
Subject to qualifications, the existing research supports the hypothesis that a sequence of failures in central planning was mainly responsible for the sharp declines in China's grain output between 1959 and 1961. At the inception of the GLF, wishfully hoping for a jump in agricultural productivity from collectivisation, the Chinese government implemented an infeasible industrialisation timetable. The diversion of massive amounts of agricultural resources to industry and excessive grain procurements from the peasants, which led to malnutrition and decimation of labour productivity, were both responsible for the collapse of grain supply. The time lag between heavy grain procurement in 1 year and the nutritional deficiency in rural workers in the subsequent year contributed in part to the prolonged decline in production. Bad weather also exacerbated the mistakes of economic planning. The associated famine has multiple determinants. In addition to grain availability decline, during normal years as well as periods of national food emergency, China's food distribution system under planning gave only urban residents legally protected rights for acquiring a certain amount of food. The lack of food entitlement to the rural population due to the urban-biased allocation rule, together with political radicalism and grain exports during the crisis, contributed to the enormous death toll.
Agricultural crisis and famines have long occupied the attention of scholars. The traditional approach to famine analysis, which dates back to the writings of Adam Smith and Thomas Malthus, proposes that famines are primarily caused by a sudden decline in food availability (FAD). From the late 1970s onward, a new literature on famines emerged. The intellectual foundation of this new literature is Sen's entitlement approach, which goes beyond FAD explanations and emphasises a broader set of causal factors that influence hunger and welfare, including food distribution, ownership patterns, relative prices, and famine policies (Sen, 1981; Dreze and Sen, 1989).
However, despite substantial progress in developing a richer conceptual framework and improving empirical knowledge for the new literature, the understanding about famines in socialist economies had long remained rudimentary (see Ravallion, 1997; O'Grada, 2007). This void in research was due in part to the fact that socialist institutional arrangements differed from standard market settings of the entitlement approach, in which private wealth and relative prices could explain a large part of hunger and starvation. In contrast, governments under central planning set commodity prices and controlled food production and distribution. Hence, research findings from the Chinese GLF crisis contribute to our understanding about famine causes under alternative economic systems. A key insight is that central planning embodied a systemic risk. If erroneous policies were made at the top and implemented with dogmatic confidence, policy failures would have nationwide repercussions, resulting in a monumental disaster and an enormous death toll (Li and Yang, 2005). Perhaps this finding provides a partial answer to a perplexing contrast between two large developing countries that was raised by Sen (1983) more than two decades ago: the inability of China to avoid large-scale famines, while achieving a much better record of eradicating regular malnutrition and hunger than that of India. The research findings on China also have implications for analysing the Soviet famines under planning in which overambitious production plans, excessive grain collections, and biases in food distribution were important causal factors. While the major focus of this article is on Chinese government policies as they relate to the GLF, in the last section of the paper, significant parallels will be drawn to the Soviet famines of 1931-1933 and 1947. These correlations corroborate and bring into relief the research findings on China.
FACTUAL AND CONCEPTUAL ISSUES
Existing studies generally agree with the severity of production shortfalls during the Great Leap crisis. The national grain output figures reported in Table 1 indicate yearly grain productions of 195, 200, 170, 143, 148, 160, and 170 million metric tons for the period 1957-1963. The 15% and 16% sudden declines in grain output in the two consecutive years 1959 and 1960, when China was barely self-sufficient in food supply, were catastrophic. These official figures of grain output are broadly consistent with other output estimates based on independent data sources (see Ashton et al., 1984), (3) despite the serious disruptions to the operation of government agencies, including the State Statistical Bureau, during the GLF.
The extent of the demographic catastrophe was not known to Chinese and Western scholars for almost two decades after the event. For years after the GLF, some Western scholars actually praised the ability of the Chinese government to avoid a famine in spite of a sharp production shortfall during the GLF (eg, Perkins, 1966). However, based on 1964 and 1982 Chinese population census and other fertility surveys released since the early 1980s, Western demographers have discovered that the total premature deaths between 1959 and 1961 were astonishing. Among this wide range of estimates, Coale (1981) provides the lower bound of 16.5 million, while Banister (1987) gives the higher bound of 30 million. Because of differences in methods of estimation and data sources, other estimates are reported at 18.48 (Yao, 1999), 23 (Arid, 1982; Peng, 1987), and 29.5 million (Ashton et al., 1984). Total lost or delayed births are estimated at 33 million by Ashton et al. (1984) and 30.79 million by Yao (1999). As Table 2 shows, 1960 registered the highest death rate: 25.4 per thousand.
These mortality estimates identify China's Great Leap crisis as the worst famine in recorded history, measured in absolute number of lives lost. The enormous number of deaths dwarfs the other major historical famine in China between 1876 and 1879, which, due to drought, claimed the lives of between 9.5 and 13 million people (Bohr, 1972), the previously highest death toll. Even by the most conservative estimate, the GLF famine is more destructive than other well-known famines in the world. For instance, the great Irish famine of 1845-1851 destroyed 1 million lives; the Bengal famine of 1943, 2 million; and the Soviet famines of 1921-1922, 5-9 million, 1932-1933, 5-11 million, and 1946-1947, 1.2-5 million. The Ethiopian famine of 1984-1985 resulted in 0.6-1 million premature deaths (see Ravallion, 1997; O'Grada, 2007). (4)
Decline in grain production and massive starvation are two critical aspects of the GLF. Researchers face the challenge of determining the causal factors of each catastrophe despite their concurrent occurrence and the possible interrelationship between the two events.
Economists have long studied the causation of production crises in agriculture. While natural calamities and war are often blamed as the leading causes of crop failures, Rosen (1999) shows that bad human judgements can also be fatal. For the Irish famine that Rosen studies, erroneous expectation of the productivity of seed potatoes provoked 'oversaving', which delayed possible substitution of other crops and led to a sharp reduction in the next year's food supply. (5) Acknowledging the relevance of weather and intertemporal decisions, there is little doubt that China's collapse in grain output was also influenced by the dramatic changes in institutions and policies that are commensurate with the central planned economy of the GLF period.
Severe crop failures are neither a necessary nor a sufficient condition for famines to occur. This is the view of modern economic analysis associated with Sen's entitlement approach (see Osmani (1995) for elaborations). For China's GLF, there was a dramatic decline in grain production and a concurrent famine; the causes of each event are worth careful investigation. One important point is to determine the importance of the decline in food availability among other possible causes of the famine. If the decline in grain output did not contribute significantly to excess mortality, as in Sen's analysis of several well-known historical famines, the causes of the production crisis, such as bad weather, would be irrelevant factors in explaining the famine. Failures in entitlement relationships, institutions, and policies would be the main causes. However, if the decline in grain output was a significant factor responsible for the excess deaths, all causes of the production shortfalls would be indirect contributors to the famine, that is, through their negative effects on food availability.
THE COLLAPSE IN GRAIN PRODUCTION
Economists have analysed multiple factors that were responsible for the precipitous declines in grain output between 1959 and 1961. We briefly discuss these causes and the empirical strategy suitable for testing the proposed hypotheses. Existing evidence suggests that the main culprit of the collapse in production was a series of mistakes in central planning.
Hypotheses on casual factors
The postmortem official explanation for the drop in grain output puts the blame mainly on bad weather (CCP, 1981) and refers to the period 1959-1961 as the 'three years of natural calamities' (Hypothesis 1). According to official weather records, the average percentage of sown areas affected by natural calamities during that period was 14.53%, which compares with an average of 7.63% disaster areas in the 3 years before the crisis (see column (9) of Table 1). Using meteorological data collected independently, Kueh (1995) confirms that bad weather was indeed a contributing factor to the poor harvests (see also Peng, 1987). Although inferior climate could severely damage agricultural production, researchers are cautious about the magnitude of its negative impact. As Kueh notes, bad weather of similar magnitudes did occur in the past, but they did not result in such serious reduction in aggregate grain output as observed in 1959-1961. A closer examination of the official statistics yields a similar conclusion. For instance, the proportion of disaster-hit areas in 1962 and 1963 were 13.1%, comparable to the crisis period, but the negative impact on grain supply was minimal (see Table 1). Another reason not to trust the official measure of meteorological conditions is that the disaster areas are defined as the sown acreage that is hit by flood, drought, and other natural calamities, and that has 30% or more reduction in yield compared to normal yield. By this definition, and given the party line explanation of the GLF disaster, it is plausible that the crop failures caused by other factors, such as the GLF policies, may have been attributed to bad weather. Despite these caveats, scholars generally agree that natural calamities were responsible for a fraction of the decline in grain production during the GLF, although other factors must have played important roles.
Incentive problems due to the unwieldy size of the communes could result in lower production efficiency (Hypothesis 2; see Eckstein, 1966; Perkins and Yusuf, 1984). When 'people's communes' were established in 1958, the average size of the basic production unit in rural China increased dramatically. Since the work-point system practiced in rural collectives rewarded individual workers based on average team performance, the rise in the scale of the production unit would reduce the marginal returns to individual work effort, leading to a decline in peasants' work incentives. To assess this negative effect of collectivisation on agricultural labour productivity, one could approximate the incentive effort by using the size of the basic production unit that had independent accounting. Based on information from retrospective surveys, the provincial average production unit grew from a size of 22 households in 1954 to 176 households in 1957 and then jumped to 2,675 households in 1958. The size of the organisation declined sharply to 41 households in 1962 (Li and Yang, 2005).
Lin (1990) develops a game theoretical explanation, arguing that the main cause of the agricultural collapse was the deprivation of the peasants' right to withdraw from people's communes (Hypothesis 3), which were established in the fall of 1958. Prior to 1958, with voluntary participation in rural cooperatives, the peasants had a self-enforcing agreement to work hard because otherwise, hard-working members would exit the cooperatives. As Lin argues, the switch in organisation to compulsory participation changed the incentive structure so that peasants were prone to shirk within the communes. The total factor productivity (TFP) index estimated by other researchers (eg Wen, 1993) indicates a sudden decline in TFP in 1959, and the index remained at a low level for the entire collectivisation period 1958-1978, giving support to Lin's exit right hypothesis. Since productivity is an important determinant of output, the removal of exit rights may indeed have contributed to the sharp decline in total output. (6)
A number of factors clearly contributed to the sudden decline in food production during the GLF (see Johnson, 1998; Riskin, 1998). But, as Li and Yang (2005) argue, the centerpiece of the puzzle is a systematic failure in central planning (Hypothesis 4). To analyse the relationship between several key variables and obtain a better understanding about the sequencing of GLF events, Li and Yang have formulated a dynamic model of central planning that consists of an agricultural sector and an industrial sector. The objective of the planner is to maximise a discounted flow of industrial output into the future, subject to the constraint that there must be enough grain to feed the industrial work force and function as an intermediate input for industrial production in each year.
In the model of these researchers, two parameters play a crucial role in determining the equilibrium allocation of production inputs between the industrial and agricultural sectors. One is the government's time discount factor, which reflects its preference for speed in achieving industrialisation for China. Another parameter is agricultural productivity, which generally depends on available technology and the form of farming organisation. If the government assigns high priority to rapid industrialisation, it would extract more resources from agriculture to support the industry, but if agriculture has high productivity, less inputs are needed in order to produce sufficient agricultural output to feed the labour force and provide material inputs for the industry. In addition, as a key feature of the model, the physical capacity of agricultural labourers depends on their nutritional status, which increases with food consumption.
Li and Yang argue that on the eve of the GLF movement, two factors acted as the catalysts for a series of policy mistakes that were responsible for the subsequent catastrophe: (a) the central planner became impatient with the slow pace of China's industrialisation and (b) the government formed an erroneous expectation that agricultural productivity would jump to a very high level with the formation of the people's communes. The model predicts that these two changes, which reflect on the central planner's time discount and agricultural productivity parameter, would induce the government to divert large amounts of labour and other resources from agriculture to industry and sharply raise grain procurement from the rural population. Resource diversion reduces agricultural output directly. Excessive procurement, when combined with an actual reduction in productivity due to commune formation, significantly reduces food available for consumption in rural areas, which would lead to a severe nutritional deficiency among rural workers. The resulting reduction in physiological capacity to carry out manual labour would in turn reduce the quality of labour input in growing next year's crops, leading to an additional decline in production. Numerical simulations of the dynamic model suggest that grain output would decline sharply at the onset of these policies, and nutritional deficiency, which realises its effects through intertemporal linkages, could suppress grain production for multiple years. Grain production would not begin to recover until resource allocation and procurement policies are reversed.
Table 1 reveals the extent of resource diversions, heavy procurement, and sharp reduction in food availability in rural areas during the GLF crisis. As shown in column (4), the agricultural labour force was reduced by 38 million between 1957 and 1958, when a large number of workers were mobilised to participate in industrial production. (7) Since the labourers who left agriculture were usually the best workers, there was a decline in both the quantity and the quality of the agricultural labour force.
Under the illusion that the collectivisation drive had solved China's food problem, the government advocated a 'three-three system' of agricultural land use under which only a third of the arable land would be needed for grain cultivation, a third would be allocated for growing cash crops and planting trees, and a third would lie fallow (Walker, 1984; Peng, 1987; Yao, 1999). Another official policy was 'sow less, harvest more'. Table 1 shows that the area sown with grain was reduced by more than 13% between 1957 and 1959.
Influenced by the 'wind of exaggeration' in production forecasts, the government increased the procurement of grain at the onset of the GLE Table 1 shows that state grain procurement increased from 46 million metric tons in 1957 to 64 million in 1959, despite a sharp decline in grain output in 1959. Consequently, grain retained in rural areas declined sharply from 295 kilograms (kg) per person in 1957 to 223 kg in 1959, and further down to 212 kg in 1960. Since grain was the primary source of food energy in China at the time, the decline in per capita food availability would sharply reduce the physical capacity of carrying out manual work, therefore adversely affecting the labour productivity of rural workers.
Regional innovations in radicalism, as exemplified by the establishment of communal kitchens, may have wasted a substantial amount of food and hence compounded the nutritional effects of excess procurement on peasants' work capacity (Hypothesis 5). In the fall of 1958, local governments established more than 2.65 million communal kitchens (Chang and Wen, 1997). By the end of 1959, the participation rate of peasants in communal mess halls reached an average of 64.7% across the Chinese provinces, with a range from 16.7% to 97.8% (see Table 3). In addition to food waste, Yang (1996) argues that the extent of communal dining reflected in large part the degree of radicalism of local political leaders. The more radical the leaders, the more it was likely that local governments would engage in GLF-related non-agricultural activities, such as the construction of mass irrigation and land reclamation projects. These radical initiatives were likely to interfere with grain production in the short run; their negative effects can also be assessed empirically.
Although there has been general awareness about the causes of the production shortfalls analysed in the previous section, for a long time, scholars of the Chinese economy only provided descriptive evidence in support of their individual arguments. Descriptive analyses are helpful in identifying the causal factors and informative of related institutions, but they do not offer quantitative assessments on the effects of the causal factors on output, thus leaving a significant gap in our understanding of the decline in food supply. The paucity of rigorous empirical research is due to several reasons, including the absence of data for several key variables, the need for a unified strategy of testing alternative theories, and the lack of a consistent framework for analysing GLF policies.
In a recent article, Li and Yang (2005) propose a unified framework to assess the relative contributions of the above-mentioned factors to the production shortfalls during the GLF. Their empirical framework adopts a production function approach, in which variations in agricultural output in the Chinese provinces are determined by variations in the quantity and quality of inputs in those provinces. More specifically, the production analysis first takes into account the quantities of conventional inputs--such as land, capital, labour, fertiliser, irrigation, and weather conditions--as determinants of grain output. Then, the framework also takes into account the quality dimensions of the inputs, which depend on factors such as the nutritional status of the rural workers, size of basic production unit, and whether peasants had the right to exit from rural collectivisation. This production function framework is capable of testing not only the hypothesis that the GLF policy package--diversion of agricultural resources and excessive procurement--was responsible for a significant fraction of the collapse in grain output but also other complementary hypotheses on the role of bad weather, labour incentive problems, exit right, and GLF radicalism in causing the production shortfall.
Given the proposed hypotheses, there is still the challenging task of constructing key variables before econometric testing. Li and Yang use a mix of their own constructions as well as innovative measurements suggested by other researchers. The yearly incremental steel and iron output for the period 1954-1964 is used as a proxy for the unobserved diversion of rural labour to nonagricultural GLF projects, as the upheaval associated with the proliferation of backyard steel mills during the crisis absorbed many agricultural labour inputs in rural China. Although yearly food consumption of the peasants was not recorded directly, the retained grain (total grain output minus total grain procurement of the previous year) per person in a specific region is used as a proxy for the nutritional status of local peasants. Collection of historical weather data provides a five-level climate quality index, which is independent of the potentially problematic official weather records. Moreover, the communal dining participation rate at the province level measures the degree of GLF radicalism in local regions (Yang, 1996; Chang and Wen, 1997).
It is also worth noting the unit of observation for empirical analysis. To assess the determinants of grain production, regression analysis requires sufficient variations in agricultural input-output data, either from cross-section units or time series records. This data requirement presents difficulties for studying historical agricultural crises in many parts of the world because detailed data were not available. A unique research opportunity exists for analysing the GLF crisis because China has 25 provinces from which there is systematic historical information on agricultural production. Li and Yang (2005) have compiled a province-level panel data set for the period 1952-1977 consisting of information from published sources and a retrospective survey, which contains additional statistics from local data archives and agricultural experts. The data are rich for estimating econometric models.
The empirical analysis of Li and Yang pays particular attention to two GLF policy variables: food consumption of rural workers and incremental output of steel. Having grain output specified as the explained variable in the regression, the coefficient on the explanatory variable food consumption--a measure of nutritional intake that affects the physical capacity of peasants--enables one to assess the effect of heavy state grain procurement on the collapse in production. The coefficient on the other explanatory variable steel output--a measure of diversion of labour away from farming--enables one to assess the effect of inefficient activities on grain output during the GLF. The estimation results suggest that, ceteris paribus, a 10% reduction in retained grain from the previous year would substantially lower the work capacity of peasants, causing a 2.67% decline in grain output in the current year. Similarly, a 10% increase in the change of steel output, ceteris paribus, would reduce the quantity and quality of agricultural labour and hence result in a 0.99% decline in grain output. Therefore, given the severe decline in food availability in rural areas (see Table 1) and sharp increases in steel and iron output during the GLF, (8) these two factors were important causes of the collapse in gain output.
Through a decomposition analysis, Li and Yang quantitatively assess the contributions of the identified factors to changes in grain output during the GLF crisis beginning in 1958 and ending in 1961. Their findings show that the diversion of agricultural resources to industry was the most important causal factor, responsible for 33% of the observed grain output decline. The intertemporal effect of excessive procurement and nutrition was the second largest contributor to the decline, accounting for 28.3% of the production shortfall. Adverse weather conditions also played a significant role, reducing food supplies by 12.9%. However, despite noticeable effects of other causal factors on output in specific years, none of them had a major impact on the shortfall in grain production for the entire crisis period.
Sen's entitlement approach provides the insight that decline in food availability is only one possible cause of famine. To understand massive starvation, analysts often need to explore factors beyond a shortage in food supply, which is the focus of the last section. Moreover, since the entitlement approach is built primarily on a market system in which commodity prices and rights to private properties are central ingredients of analysis, its application to the Chinese famine must take into account factors unique to China. First, China had a central planning system in which the government set commodity prices and controlled the procurement and distribution of food. Second, the Chinese famine occurred in the chaotic GLF period when radical policies other than food distribution may have aggravated hunger and starvation. Third, unlike many famines examined by Sen in which aggregate food availability did not decline, China indeed experienced severe shortages in food supplies at the national level. The challenge to researchers is to assess the importance of food availability decline, failures in entitlement arrangements, and other institutional and policy factors to the enormous loss in human lives.
Hypotheses on causal factors
Applying Sen's entitlement approach, Lin and Yang (2000) present a framework that considers FAD and urban bias in China's food distribution system as joint determinants of the GLF famine (Hypothesis 1). Under central planning, China had an effective, urban-biased rationing system in which city residents were given protected rights to acquire certain amounts of food. On the other hand, compulsory quotas of grain procurement were imposed on the peasants. As a result, peasants were entitled only to the residual grain output. In years of poor harvests, particularly during the GLF period when there were sharp declines in food production in many rural regions, delivery of production quotas set in the beginning of a year would leave insufficient food supplies for farmers in the following year, causing the caloric intake of some peasants to fall below the minimum threshold required for survival. Hence, urban bias in food procurement and distribution may directly cause excess mortality in the presence of negative shocks on grain production.
Admittedly, dramatic declines in food supply may directly cause starvation and hunger. During the GLF, national grain output declined abruptly by 15% in 1959 and aggregate food availability reached only 70% of the 1958 level in 1960-1961. Lin and Yang argue that the sudden collapse in grain supply, along with urban-biased food allocation, were both likely responsible for the massive death toll. The empirical question they attempt to answer is the relative importance of the two factors in causing the variations in death rates across Chinese provinces.
Factors other than FAD and legal rights to food may also contribute to the severity of the famine. Yang (1996) and Chang and Wen (1997) argue that wasteful preparation and consumption of food associated with communal dining, which led to the depletion of grain before next year's harvest, were the primary causes of the Great Leap famine (Hypothesis 2). Yang (1996) considers wasteful food consumption in communal kitchens as a tragedy of the commons: when meals were freely supplied in communal mess halls, the pursuit of individual gains led to excessive food consumption, a result that was detrimental to all commune members. Yang further argues that agrarian radicalism, of which commune dining is an example, is the more fundamental cause of the catastrophe. He proposes a 'loyalty compensation' hypothesis: that provinces with lower ratios of communist party members would implement radical policies of the GLF more enthusiastically than the provinces with higher ratios of party members. This is because those who wanted to join the party would try hard to gain their party membership by showing unshakable loyalty to the centre, a classic case of overcompensation for the sake of a cause. Therefore, the provinces that had lower ratios of party members were her mess hall participation rates, and consequently, higher levels of mortality.
In a related study, Chang and Wen (1997) single out the communal dining system as the primary cause of the famine. While admitting that multiple factors contributed to the catastrophe, they placed emphasis on one piece of evidence: the national death rate started to rise in 1958, when there were good harvests and abundant food supplies. Since many communal dining halls were established in the second half of 1958 and there was over-consumption and waste of food in the kitchens, they argue that communal dining must be the culprit that first started, and then greatly aggravated, the famine. The paper presents much anecdotal evidence and draws inference from the sequencing of events. Their conclusion is also supported by a positive correlation between 1960 excess death rates and the 1959 dining hall participation rates across the Chinese provinces.
The study by Chang and Wen gives the impression that famine started in 1958. However, on the basis of China's provincial mortality statistics, Lin and Yang (1998) point out that widespread famine did not occur until 1959. Table 2 shows that China's death rate increased to 12.0 per thousand in 1958, from an average of 11.1 per thousand in 1956-1957. However, this increase was driven by dramatic increases in mortality in three provinces: Sichuan's death rate increased from 11.3 to 25.2 per thousand, Yunnan from 15.8 to 21.6, and Gansu from 11.1 to 21.1. In fact, using the 1956-1957 average as a reference, the death rates in 1958 actually declined in 16 out of 28 provinces; therefore, rising mortality was not a widespread phenomenon across the Chinese provinces in 1958. In 1959, the national average mortality jumped to 14.6 per thousand and higher mortality rates are found in 27 out of 28 provinces, marking the beginning of widespread starvation. Since most provinces started communal dining halls in 1958, but rising mortality rates were concentrated in three provinces, these facts do not support the claim that communal dining initiated the famine in that year. (9)
The timing of the 1958 harvests and the creation of communal kitchens also suggests that public dining was an unlikely cause of the excess mortality in 1958. Communal mess halls emerged after the formation of the communes. The first commune appeared in August 1958; most others were established in October or November 1958. Therefore, many communal kitchens did not start until the end of that year. It is well known that in most places in China, crops were not harvested until October or November. For areas with single-crop production, the harvested grain should provide sufficient consumption for 12 months, while in double-crop areas the supply should last for 6 months. Based on the above information, it is highly unlikely that people who participated in communal kitchens consumed all the food in one or two months and went on to hunger and starvation in 1958. The rise in mortality rates in the few provinces was likely due to factors other than food waste and overconsumption.
There is little doubt that overconsumption and waste of food in communal mess halls during the GLF could have reduced food availability in local regions and magnified the severity of famine. The relevant empirical questions are: how does one reliably measure the quantity of food waste resulting from the public dining arrangements, (10) and how does one assess its impact on excess deaths relative to other causes? The information available to approximate the extent of food waste associated with communal dining is the province-level mess hall participation rate (MHPR) at the end of 1959 (see Table 3). Chang and Wen (1997) use MHPR as a proxy for consumption inefficiency; they find a positive correlation with the I960 provincial excess death rates. This result is taken as empirical support for the hypothesis that consumption irrationality led to famine. However, since only 24 sample points are used in the regression and the specification does not control for other possible causes of famine, the correlation can hardly represent reliable evidence on causality.
The political attitude of provincial leaders may have strongly influenced local policies and economic activities during the GLF, which could in turn have contributed to the excess mortality (Hypothesis 3). As Yang (1996, p. 58) observes, mess hall participation rates varied from over 90% for Henan, Hunan, Sichuan, Yunnan, Guizhou, and Anhui to under 50% for provinces such as Liaoning and Heilongjiang (see Table 3). He also notes that the seven provinces with the highest mess-hall participation rates (Shanghai as a municipal city is not included in the table) either were under the influence of the most zealous leaders of the time or had just gone through a political purge. These observations corroborate the view that radical leaders were more responsive to the calls from the party centre in Beijing for achieving high levels of participation in communal kitchens. Since communal dining caused overconsumption and waste of food, the radical provincial leaders and their policies are likely partly responsible for the subsequent death toll.
Another mechanism through which agrarian radicalism may contribute to excess mortality is the exporting of grain from provinces to the central government by zealous provincial leaders who were most enthusiastic to show their loyal support for the GLF policy (eg, Lin and Yang, 1998). For instance, provincial leaders in Sichuan and Hunan, who were among the most cooperative and obedient to the centre, managed to deliver large quantities of grain to the state despite their own shortages (Walker, 1984). In 1959-1960, each of the provinces submitted 2.24 and 0.44 million tons of grain to the centre, while starvation prevailed in their own provinces. As early as 1958, a procurement slogan was propagated in Sichuan: 'First the center, than the locality; first external (commitments), then internal (commitments)'. The province organised 5 million people to transport grain for export, and the procurement reached the highest historical level of 2.595 million tons. Because of heavy extractions, food availability in Sichuan and Hunan fell to dangerously low levels, ultimately causing high levels of mortality. In contrast, provinces such as Guangdong and Jilin only experienced mild increases in excess deaths during the crisis, partly because these provinces successfully reduced their grain export burdens (Walker, 1984).
Finally, provinces with zealous leaders were prone to deeply involve themselves in the radical GLF campaigns, such as backyard steel production, massive irrigation work, land reclamation tasks, and other labour-intensive projects. These energy-consuming projects in nonagricultural activities often increased the demand for calories among participating labourers, leading to faster exhaustion of food supplies before the next harvest and thus malnutrition and starvation (Johnson, 1998; Kung and Lin, 2003). Moreover, since these physically demanding projects may be detrimental to health, and the zealous devotion to these campaigns may also have led to the neglect of health care, both situations could lead to higher mortality (Lin and Yang, 1998). This health-related factor, plus the large quantity of grain outflow from the provinces, could be important factors in explaining the sharp rise in mortality rates in Sichuan, Yunnan, and Gansu in 1958.
With the agreement that GLF radicalism could aggravate the famine, researchers have used two different proxies to measure the degree of radicalism for empirical analysis: (a) the density of party membership (DPM) in individual provinces, as measured by the proportion of the rural population who were communist party members (11); and (b) the revolutionary history of a province, as measured by the time order of liberation for individual provinces (see Table 3). Measure (a) is based on the idea of Yang (1996) that low-level cadres not yet joining the party were usually the most zealous because they were eager to show strong loyalty to the party centre. Therefore, as he argues, a lower density of party membership in a province implies a higher degree of agrarian radicalism. However, because of noticeable limitations of the party density measure, (12) researchers also use (b) to measure radicalism, which is an index known as the time of liberation (TOL) according to the actual month and year when a province was declared 'liberated' by the government (Yang, 1996; Kung and Lin, 2003). The rationale behind this measure is based on the historical fact that newly liberated provinces were more likely to be appointed with zealous leaders, who would in turn pursue more radical GLF policies, leading to higher mortality in those provinces during the GLF.
It is ironic that China's net grain export reached a historical record of 4.2 million tons in 1959 and remained at 2.7 million tons in 1960 when the nation suffered from the horrific death rate of 25.4 per thousand. The bulk of those grain exports functioned as payments to the USSR in exchange for machinery and equipment for the GLF, which peaked in 1959 (Riskin, 1987, p. 132; Lardy, 1987). Since the quantities of China's total grain output were 170 and 143.5 million tons in those 2 years, net grain exports accounted for 2.47% and 1.88% of the total production. Hundreds and thousands of lives could have been saved without those grain exports (Hypothesis 4; eg, Johnson, 1998). Moreover, the combined effects of closing down rural commodity markets and the decline in government coordinated interregional grain trade during the GLF may have contributed to the severity of the famine (Hypothesis 5; Lardy, 1987). The reduction in food trade across provinces was caused in part by the government initiative of establishing local self-sufficiency with the communes. It also reflected the political incentives of local cadres who often concealed food shortages within their jurisdiction and the need for outside help because of their earlier exaggerated production forecasts. There is evidence that mortality was particularly high in rural areas that specialised in non-grain production, where the starvation would have been less severe if there were effective markets and regional redistribution. Lastly, as Sen (1983) emphasises, the lack of news distribution systems and pressure groups within China may have obstructed information flows to provide necessary famine relief or international aid (Hypothesis 6), which could have significantly reduced famine deaths.
In famine analysis, researchers often rely on case studies to investigate the cause of specific famine events, emphasising either food availability or entitlement as a crucial factor. The lack of econometric analysis reflects in part the difficulty of measuring spatial and/or time series variations in entitlement arrangements and food availability decline. In the Chinese famine of 1959-1961, food availability and entitlements to food were both important factors. Moreover, data are available that reveal variations in mortality rates, food availability, entitlement arrangements, communal dining, and the degree of radicalism in the Chinese provinces in specific years and over time. These rich data sources permit researchers to test Hypotheses 1-3 discussed in the last section. Applying Sen's entitlement approach, Lin and Yang (2000) develop a framework that is amenable to empirical testing and that simultaneously considers per capita food supply and the right to food as determinants of famine. The paper is the first econometric study to directly assess the importance of famine causes using the entitlement approach.
Lin and Yang use a panel data set of 28 Chinese provinces for the period 1954-1966 for empirical analysis. They use the percentage of rural population and per capita grain output in a province as, respectively, proxies for the degree of urban bias and the extent of food availability in that province, and assess the contributions of these factors to observed cross-province differences in death rates. Their basic estimation function controls for year and provincial fixed effects. The main hypothesis to be tested is whether during a famine in China, the death rate in a province is positively related to the proportion of rural population in that province (urban bias hypothesis) and negatively related to per capita grain output in that province (food availability hypothesis). (13)
Lin and Yang's empirical results show that in normal years, the cross-province differences in the urban bias and food availability variables did not result in cross-province differences in death rates. However, in the famine period of 1959-1961, both variables contributed significantly to the observed inter-provincial differences in mortality rates. In fact, the Chinese food entitlement system, which was dominated by urban-biased distribution, explains a greater part of the inter-provincial variation in mortality rates than does food availability, providing support to Sen's entitlement approach.
More specifically, estimation results indicate that a 10% increase in the proportion of rural population in a province would result in a 7.19% increase in the provincial death rate. By contrast, a 10% decline in per capita food availability would result in a 4.17% increase in mortality rate. Therefore, the effect on excess death rate from a given percentage change in entitlement is about 72% bigger than from the same percentage change in food availability. Another method of evaluating the relative importance of the two famine causes lies in their power to explain variations in provincial death rates. Using measures of partial correlation coefficients, Lin and Yang show that, among the variations of inter-provincial death rates explained by the two famine causes, 69.5% of the total was attributable to urban-biased food entitlements and the remaining 30.5% was attributable to grain availability. The findings from both methods confirm that urban bias in food allocation, as well as shortages in food supply, played central roles in causing the famine.
In a recent study, Kung and Lin (2003) expand the above framework to also examine the effects of communal dining and the Leap's radical policies on death rates across provinces. They use mess hall participation rate (MHPR) as a proxy for food waste and two other variables, the density of party membership (DPM) and time of liberation (TOL), as alternative proxies for GLF radicalism. This specification improves upon the earlier empirical analysis by Chang and Wen (1997), who examine the correlation between provincial death rates and dining participation rates, without controlling for other key causes of the famine. The findings confirm the importance of food availability decline and urban bias in food distribution in causing the famine. Political radicalism was also an important factor that helps account for variations in the death rates across provinces.
However, empirical evidence fails to support the hypothesis that food waste and overconsumption in communal dining was an important cause of the GLF famine. When either DPM or TOL is included in the regression, the coefficient estimate on MHPR is no longer statistically significant, suggesting that food waste per se did not contribute to provincial variations in death rates, although other aspects of radicalism raised the death toll. To explain why communal dining fails to affect death rate negatively, Kung and Lin present additional evidence. Based on information collected from provincial newspaper archives, they report that food was generally rationed to peasants in the mess halls, a fact that is inconsistent with the popular perception of mess hall operations, as in the political slogan 'opening up your stomach and eating as much as you can'. Moreover, official sources also reveal that communal dining became unpopular in many rural regions as early as the spring of 1959. Local adaptive policies also permitted peasants to cook at home or to open communal kitchens on a seasonal basis. These corrective actions must have mitigated some potential damage of the mess halls in the people's communes.
REMARKS ON THE CHINESE AND SOVIET FAMINES
China's agricultural crisis and the associated famine of 1959-1961 were among the worst catastrophes in human history, but for a long time they were also the least studied and understood of human calamities. This paper reviews a growing economic literature on this historical crisis. Existing evidence suggests that the collapse in grain production was largely attributable to two fallacious GLF policies. Encouraged by expectations of a great leap in agricultural productivity from collectivisation, the government diverted massive amounts of agricultural resources to industry and sharply raised grain procurement from the peasants, eventually leading to malnutrition among peasants and decimation of their labour productivity in growing next year's crops. The consecutive years of bad weather also aggravated the fatal economic policies. The decline in food availability was indeed a cause of the GLF famine. But other institutional factors, including urban bias in China's food distribution system, radical local policies, and grain exports, were also major contributors of the excess mortality. By and large, the GLF catastrophe was the result of a series of failures in central planning.
The literature on the Great Leap crisis not only provides insights into the relationship between economic system and economic performance in China but also suggests a framework to help understand the cause of famine in other centrally planned economies. Under collective agriculture, the USSR experienced two major famines: one in 1931-1933, with 4.5-8.0 millions of excess mortality (eg Davies and Wheatcroft, 2004, pp. 402-403), and the other in 1947, with 1.0-1.5 millions of famine deaths (Ellman, 2000). (14) Lively scholarly debates on the cause of these famines, especially the 1931-1933 disaster, have continued for several decades. In his book The Harvest of Sorrow, Conquest (1986) advocates that the 1931-1933 famine was 'manmade' and 'deliberate', orchestrated by Stalin, who imposed the famine to suppress the nationalist aspirations of the Ukraine and peasant resistance to agricultural collectivisation. Alternatively, Tauger (2001)--who investigates in detail the weather conditions and agricultural performance in the crisis years--concludes that the famine was 'the result of the largest in a series of natural disasters'. Between these polar views, researchers have compiled and analysed a large body of historical evidence on the development of events surrounding the two disasters. (15) A comparison with the Chinese experience reveals that policy failures of central planning were also critical for causing the production catastrophe and the famines in the USSR.
Resembling the Chinese GLF campaign was the USSR's first five-year plan in 1928, a superindustrialisation drive that aimed to achieve lofty production and accumulation goals. Soviet capital stock was to double in 5 years, and even light industry was expected to expand by 70% (Gregory and Stuart, 2001). To support the big push, industry needed an expanding labour force, ample supplies of raw materials, and agricultural export earnings to exchange for machinery and equipment imports for all sectors. There was also an urgent requirement for securing grain from the countryside because of forecasted increasing demand for food during the plan. Such a requirement was also influenced by the earlier grain procurement crisis of 1927-1928, during which Stalin believed that peasants intentionally withheld grain from markets, which resulted in food shortages in cities. Agricultural collectivisation was the proposed solution to these concerns and problems: as the government believed, collectivisation would enable the state to acquire grain more easily and cheaply, and large-scale mechanised farming would quickly transform traditional peasant cultivation into modern and productive agriculture. Based on the false premise of a leap in agricultural productivity, the Soviet government forcefully carried out rural collectivisation in a sweeping fashion. In 1929, only 3.9% of peasant households participated in collective farming; by 1932, 61.5% of peasant households worked under central planning (Volin, 1970, p. 211).
Wishfully expecting a surge in agricultural productivity from technological and organisational innovations with collectivisation, the Soviet government proposed to raise grain collection dramatically in the first 5-year plan. While actual acquisitions encountered fierce resistance from the peasants, total grain collection still increased sharply, from 10.8 million tons in 1928 to 16.1, 22.1, 22.8, 18.5, and 22.6 million tons between 1929 and 1933. Since total grain outputs were at 73.3, 71.7, 83.5, 69.5, 69.6, and 68.4 million tons in those individual years (Clarke and Matko, 1983, p. 149), both production and retained food grain in rural areas declined to very low levels in 1931, 1932, and 1933. (16)
These figures and related data suggest that malnutrition among peasants due to excessive grain procurements and diversion of agricultural resources to industry--the two most important causal factors of the Chinese agricultural crisis--were also critical to the decline in grain production in the USSR. Under the superindustrialisation drive, there were large flows of labour from rural to urban areas: Soviet urban population increased by 12.7 million between 1929 and 1933, while the rural population declined from 126.7 to 125.4 million (Clarke and Matko, 1983, p. 2). Within agriculture, resources were also diverted away from grain production; emphasis was placed on industrial crops instead. For instance, home-produced cotton consumption jumped from 168,400 tons in 1926-1927 to the same as above: 394,800 tons in 1933 (Wheatcroft et al., 1986). Grain production was further undermined by a disastrous decline in the number of draft animals. In 1933, the total heads of cattle declined to 54% of the 1928 level, and the number of work horses declined from 23.4 million in the spring of 1929 to 12.8 million in July 1934. The slaughter of animals on the eve of collectivisation, and the reduction in fodder due to heavy grain procurements during the crisis years, were largely responsible for the destruction of much-needed draft power in agricultural production. The reduction in resources for grain production and the destruction of peasants' labour capacity due to insufficient food consumption must have contributed to the production shortfalls, although their quantitative effects on output are not yet estimated for Soviet agriculture.
As with the Chinese experience, consecutive years of bad weather also contributed to the Soviet production crisis. Following the adverse climate conditions that resulted in the loss of half of the entire winter-sown crop in 1927-1928 and one-third of the winter crop of 1928-1929 in the Ukraine, severe dry weather spells occurred over a large part of the major grain-production regions in the USSR in the early 1930s, especially in 1931 (Wheatcroft et al., 1986). During certain periods in 1932, extremely wet and humid weather hit the USSR, giving rise to severe plant disease infections, smut and rut, in particular, which resulted in the loss of 7.1 million tons of grain, equivalent to 13% of the official harvest figure (Tauger, 2001). Although the quantitative effect of unfavourable weather on output is not yet evaluated statistically, scholars generally agree that natural calamities were important determinants of these production shortfalls.
Taking as given the decline in food supply, policies and institutions of the Soviet planning system were also important causal factors that aggravated hunger and starvation. Closely resembling China's food ration system, employment in industry and other nonagricultural sectors in the USSR generally had legally protected rights for certain amounts of food, while collective farmers were paid out of what was left over after compulsory deliveries. In effect, the state and the urban population were the primary claimants of food grain through the use of compulsory state acquisitions, while the rural population was essentially a residual claimant. Therefore, during the periods of severe food shortages, peasants were the most vulnerable to mortality risks. In fact, when starvation became widespread in 1932 and 1933, the Politburo of the USSR made the decision that state-collected grain must be supplied to the hungry towns; no allocations could be made available to the countryside for seed, food, or fodder (Davies and Wheatcroft, 2004, p. 440). Like China, the USSR also exported grain during the crisis years. Insisting on the priority of industrialisation, the Soviet government continued to export large quantities of grain--5.83, 4.79, 1.61, and 2.32 million tons in each of the individual years between 1930-1931 and 1933-1934--in exchange with other European countries for machinery and equipment to achieve the intended goals of capital accumulation, even as famine devastated the country (Davies and Wheatcroft, 2004, p. 471). There is little doubt that countless lives would have been saved if the Soviet government had stopped exporting grain and used this grain for famine relief.
History repeated itself during the 1947 Soviet famine, which was triggered by long periods of bad weather, especially a serious drought in 1946 that led to a sharp decline in grain production. Given the output shortages, urban bias in food rationing and exports of grain during the crisis emerged again as major causes of massive starvation. As Ellman (2000) documents, the total quantities of grain exports were 1, 0.3, and 2.4 million tons in the three years from 1946 to 1948. The bulk of grain exports went to Poland and East Germany as the Soviet government strove to consolidate the New Soviet bloc (Hanson, 2003, p. 38). The government also started building a state grain reserve after the harvest of 1947, the purpose of which was to deal with future food emergencies and strengthen national food security (Ellman, 2000). Corresponding heavy grain procurements suppressed food availability in rural areas. As a result, policy mistakes of the central planner became critical factors in causing a large number of the rural population who did not have entitlement to food under the Soviet ration system to become famine victims.
These famines in China and the USSR were the worst catastrophes in the loss of human lives in the last century. These famines, each of which lasted for several years, were accompanied by precipitous declines in grain production. The research findings reported in this paper suggest that policy errors and institutional failures of central planning are significant factors in explaining the origin and mechanism of these production and demographic disasters. The blind pursuit of rapid industrialisation at the expense of agriculture sowed the seeds of tragedies for China's GLF campaign of 1959-1961 and the Soviet superindustrialisation drive of 1928-1932. Through a chain of interrelated human actions consisting of agricultural collectivisation, overoptimistic expectations on productivity change, the diversion of agricultural resources to industry, excessive grain procurements, and malnutrition among peasants, both economies experienced sudden declines in aggregate grain production. Compounding the shortages in food supplies, urban bias in the food distribution systems and grain exports during the crisis both contributed significantly to the enormous death tolls. Therefore, central planning was the main culprit of the catastrophes, as the governments acted on false premises, imposed aggregate risks through the implementation of nationwide economic plans, and insisted on deleterious policies with dogmatic certainty. The amplifying effects of central planning on natural calamities can explain much about the long duration and the monumental scale of the catastrophes. Hence, the research reviewed in this article provides additional evidence on the weaknesses of central planning through the perspectives of agricultural crises and famines as economic transition continues to progress in China and the former USSR.
I thank Meg Gottemoeller, James Kung, Ryan Monarch, James Wen, and two anonymous referees for making constructive suggestions and comments on earlier versions of this paper. I am also grateful for the financial support from the Center for China in the World Economy (CCWE) at Tsinghua University, where I carried out much of the research of this paper while serving as Senior Research Fellow at the CCWE.
Aird, JS. 1982: Population studies and population policies in China. Population and Development Review 8: 85-97.
Ashton, B, Hill, K, Piazza, A and Zeitz, R. 1984: Famine in China, 1958-61. Population and Development Review 10: 613-645.
Banister, J. 1987: China's changing population. Stanford University Press: Stanford.
Becker. J. 1996: Hungry ghosts: Mao's secret famine. Henry Holt and Company: New York.
Bernstein, TP. 1984: Stalinism, famine, and Chinese peasants. Theory and Society 13:339-377.
Bohr, PR. 1972: Famine in China and the missionary: Timothy Richard as relief administrator and advocate of national reform, 1876-1884. Harvard East Asian Center: Cambridge.
Bourke, A. 1993: The visitation of god? The potato and the great Irish famine. Lilliput Press: Dublin.
CCP, Central Committee of Chinese Communist Party., 1981: Decisions on several historical issues of the Communist Party of China since the founding of the Republic. People's Press: Beijing.
Chang, GH and Wen, GJ. 1997: Communal dining and the Chinese famine of 1958-1961. Economic Development and Cultural Change 46: 1-34.
Clarke, RA and Matko, DJI. 1983: Soviet economic facts 1917-81. Macmillan Press: London.
Coale, JA. 1981: Population trends, population policy, and population studies in China. Population and Development Review 7: 261-297.
Conquest, R. 1986: The harvest of sorrow: Soviet collectivization and the terror--famine. Oxford University Press: New York.
Davies, RW and Wheatcroft, SG. 2004: The years of hunger: Soviet agriculture, 1931-1933. Palgrave Macmillan: London.
Dreze, J and Sen, AK. 1989: Hunger and public action. Clarendon Press: Oxford.
Eckstein, A. 1966: Communist China's economic growth and foreign trade: Implications for US policy. McGraw-Hill: New York.
Ellman, M. 2000: The 1947 Soviet famine and the entitlement approach to famines. Cambridge Journal of Economics 24: 603-630.
Gregory, PR and Stuart, RC. 2001: Russian and Soviet economic performance and structure, 7th edn. Addison Wesley Longman: Boston.
Hanson, P. 2003: The rise and fall of the Soviet economy: Art economic history of the USSR from 1945. Pearson Education Limited: London.
Johnson, DG. 1998: China's great famine: Introductory remarks. China Economic Review 9: 103-109.
Kueh, YY. 1995: Agricultural instability in China, 1931-1991. Clarendon Press: Oxford.
Kung, J and Lin, JY. 2003: The causes of China's Great Leap famine, 1959 1961. Economic Development and Cultural Change 52: 51-73.
Kung, J and Putterman, L. 1997: China's collectivization puzzle: A new resolution. Journal of Development Studies 33: 741-763.
Lardy, N. 1987: The Chinese economy under stress, 1958-1965 In: MacFarquhar, R and Fairbank, JK (eds) The Cambridge History of China. Cambridge University Press: New York.
Li, W and Yang, DT. 2005: The great leap forward: Anatomy of a central planning disaster. Journal of Political Economy 113(4): 840-877.
Lin, JY. 1990: Collectivization and China's agricultural crisis in 1959-61. Journal of Political Economy 98: 1228-1252.
Lin, JY and Yang, DT. 1998: On the causes of China's agricultural crisis and the Great Leap Famine. China Economic Review 9: 125-140.
Lin, JY and Yang, DT. 2000: Food availability, entitlements, and the Chinese famine of 1959-61. Economic Journal 110: 136-158.
Liu, Z. 1980: The recent selection and development of China's population. Paper presented at Beijing International Round Table Conference on Demography.
MOA, Ministry of Agriculture. 1984: Nongye Jingji Ziliao, 1949-83 (Agricultural economic data, 1943-83). Ministry of Agriculture: Beijing.
Mokyr, J. 1983: Why Ireland starved: A quantitative and analytical history of the Irish economy, 1800-1850. Allen and Unwin: London.
NBS, National Bureau of Statistics. 1983: Statistical yearbook of China. China Statistical Press: Beijing.
NBS, National Bureau of Statistics. 1990: A compilation of historical statistical data of provinces, autonomous regions, and municipalities, 1949-1989. China Statistical Press: Beijing.
O'Grada, C. 1989: The great Irish famine. Macmillan: Houndmills, UK.
O'Grada, C. 2007: Making famine history. Journal of Economic Literature XLV: 3-36.
Osmani, S. 1995: The entitlement approach to famine: An assessment In: Basu, K, Pattanaik, P and Suzumura, K (eds) Choice, Welfare, and Development: Essays in Honor of Amartya Sen. Clarendon Press: Oxford.
Peng, XZ. 1987: Demographic consequences of the Great Leap forward in China's provinces. Population and Development Review 13: 639-670.
Perkins, DH. 1966: Market control and planning in communist China. Harvard University Press: Cambridge.
Perkins, DH and Yusuf, S. 1984: Rural development in China. The John's Hopkins University Press: Baltimore and London.
Potter, S and Potter, JM. 1990: China's peasants--Anthropology of a revolution. Cambridge University Press: Cambridge.
Ravallion, M. 1997: Famines and economics. Journal of Economic Literature 35: 1205-1242.
Riskin, C. 1987: China's political economy: The quest for development since 1949. Oxford University Press: Oxford.
Riskin, C. 1998: Seven questions about the Chinese famine of 1959-61. China Economic Review 9: 111-124.
Rosen, S. 1999: Potato paradoxes. Journal of Political Economy 107: s294-s313.
Sen, AK. 1981: Poverty and famine. Clarendon Press: Oxford.
Sen, AK. 1983: Development: Which way now? Economic Journal 93: 745-762.
Sterba, J. 1981: China believed to have understated scale of drought and flood. The New York Times, 25 April.
Sun, YF. 1981: Consolidate statistics work and reform the statistics system, Jingji Guangli (Economic Management) 2: 3-5.
Tauger, M. 2001: Natural disasters and human actions in the Soviet famine of 1931-1933. The Carl Beck Papers, No. 1506, University of Pittsburgh.
Volin, L. 1970: A century of Russian agriculture. Harvard University Press: Cambridge.
Walker, KR. 1984: Food grain procurement and consumption in China. Cambridge University Press: Cambridge.
Wen, GJ. 1993: Total factor productivity change in China's farming sector: 1952-1989. Economic Development and Cultural Change 42: 1-41.
Wheatcroft, SG, Davies, RW and Cooper, JM. 1986: Soviet industrialization reconsidered: Some preliminary conclusions about economic development between 1926 and 1941. Economic History Review (2nd set.) 39(2): 264-294.
Yang, DL. 1996: Calamity and reform in China: State, rural society, and institutional change since the Great Leap Famine. Stanford University Press: Stanford.
Yao, SJ. 1999: A note on the causal factors of China's famine in 1959-1961. Journal of Political Economy 107: 1365-1369.
Zhu, ZZ. 1980: China's population problem at present and its development trend. Jingji Kexu (Economic Science) 3: 54-58.
(1) The Chinese government did not publish mortality statistics for the years 1959-1961 for two decades. Isolated, yet important, mortality data were published in 1980 by two Chinese demographers, Zhu (1980) and Liu (1980), who inferred the total famine deaths of 14-16.5 million people in China in the crisis years. A Chinese economist, Sun (1981), provided an estimate of 15 million extra deaths for the Great Leap famine. Parallel to these studies, Western demographic researchers also suggested the occurrence of a large-scale famine (Coale, 1981; Aird, 1982), which was mentioned in popular media as well (eg, Sterba, 1981). Soon after, scholars of China carried out systematic and detailed analysis of the disaster (Bernstein, 1984; Walker, 1984).
(2) See Yang (1996) and the references therein for the analysis of the calamity from the perspective of political science. Becker (1996) presents a historical account of the tragedy based on hundreds of interviews and years of painstaking detective work that uncovered facts and personal experiences related to the famine.
(3) The output figure for 1958 is perhaps most unreliable because it has been revised several times. In 1958, the People's Daily (10 September 1958) forecasted a total grain output of 525 million metric tons for that year, or more than 2.6 times the grain output in 1957 (see Table 1). The 'actual' output was estimated to be 375 million metric tons, before it was revised downward to 250 million metric tons in 1959. The last revision of 200 million tons was published in the 1980 Statistical Yearbook of China.
(4) While the total excess mortality of the GLF crisis was the highest among all recorded famines, its excess death rate of less than 20 per thousand is modest compared to other worst famines in history, such as those of 120 per thousand in Ireland in the second half of the 1840s and 60 per thousand during the 1921-1922 famine in the USSR. O'Grada (2007) presents the estimated death tolls of many major historical famines ill both relative and absolute terms.
(5) In 1985, the fungus p. infestans made its appearance in the Irish potato crop; potato output fell by half in that year. Potato growers mistakenly believed that the drop in output was temporary, but in fact the productivity of seed potato due to the fungus was permanent. In the second year, there was no significant decline in the planting of seed potato relative to the previous year, yet output fell by 80%. See Mokyr (1983), O'Grada (1989), and Bourke (1993) for additional economic analysis of the Irish famine.
(6) Lin's explanation for the abrupt collapse of Chinese agriculture provoked a heated debate over the nature of incentives within agricultural collectives. The articles that appeared in the 1993 symposium issue of the Journal of Comparative Economics were, in effect, criticisms and comments on Lin's paper. The debate focuses on two critical issues: (a) theoretically, whether the right to exit is necessary for high effort-supply among cooperative members, and (b) historically, whether there was voluntary participation during the collectivisation movement before the GLF movement (see additional evidence on this issue by Rung and Putterman (1997)). Although the symposium did not fully resolve the disputes, there was convergence on a key issue: that the elimination of exit right caused a decline in agricultural productivity during the collectivisation period of 1958-1978.
(7) Riskin (1987) gives a similar estimate of 41 million workers, or 21% of the agricultural labour force, who left farming between 1957 and 1958. Among these workers, approximately 17 million worked in the iron, steel, and other heavy industrial undertakings in the countryside, while close to 16 million migrated into cities, working in state industrial enterprises. See Ashton et al. (1984), Walker (1984), and Bernstein (1984) for additional information on sectoral labour allocation during the GLF, including the corrective policies of sending workers back to their rural homes for reducing urban food demand and increasing agricultural labour inputs.
(8) The national output of pig iron and steel was 11.3 million tons in 1957. It increased dramatically to 21.7, 35.8, and 45.8 million tons in the next 3 years, before it returned to the trend level of 14.7 million tons in 1962. Most of these increases in steel output were produced in China's rural areas.
(9) Anecdotal evidence used by Chang and Wen (1997) suggests that the total waste of food was enormous in some rural areas. For instance, 'in some rural areas the grain consumed by peasants in a three-month period amounted to what usually sufficed for six months' (Peng, 1987). 'In some places, three months' supply of grain was consumed in merely two weeks' (Yang, 1996). There is also anthropological evidence from Potter and Potter (1990): 'According to one peasant, everyone "irresponsibly" ate whether they were hungry or not, and in 20 days they had finished almost all rice they had, rice which should have lasted for six months'. While these descriptions could be true as isolated events, they cannot be used as scientific evidence to support the hypothesis that communal dining was a major cause of the famine.
(10) Chang and Wen (1997) provide a quantitative estimate made by a Chinese economist Xue Muqiao that the over-consumption of grain by peasants in 1958 amounted to 17.5 million tons, which was 8.78 % of the total domestic production in that year. However, the authors did not explain the method used to derive this estimate.
(11) Since information on the density of party membership is available only for 1956, the data are used for the analysis of the GLF period, with the implicit assumption that the ratio did not change much in the period of several years.
(12) As Lin and Yang (1998) and Rung and Lin (2003) point out, the range of variations in DPM was very narrow, with a low of 0.71 and a high of 3.41%, which raise the question of whether 1%-2% differentials in non-party membership would lead to significant differences in the degree of radicalism. Second, it is not clear whether party or non-party members were more zealous and loyal to the party centre. Third, it is likely that top provincial leaders played a much more influential role than low-level cadres in formulating local GLF policies.
(13) Note that the percentage of rural people in a province measures the percentage of people who did not have legally protected rights to food in that province. An alternative entitlement measure would be the state grain procurement and transfers from rural areas of a province, representing the deprivation of food entitlement of that province. However, the procurement and transfer information was not available to the authors at the time of their research.
(14) prior to agricultural collectivisation, two other large-scale famines also occurred in the Russian Empire and the USSR in recent history. The 1891-1892 famine resulted in 0.4-0.5 million excess deaths; the death toll in the 1918-1922 famine reached as many as 10-14 million. These two famines, which were primarily attributable to natural calamities, rural revolution, civil wars, and famine-related infectious diseases (Davies and Wheatcroft, 2004, pp. 402-406), are not covered in the subsequent discussion.
(15) See O'Grada (2007) for a survey of the major studies and their findings.
(16) See Davies and Wheatcroft (2004, pp. 432-433) for a summary of the estimates of substantial decline in rural food availability from several independent sources. They also point out a more severe decline in the consumption per head of meat and diary products during the famine years.
DENNIS TAO YANG (1,2)
(1) Department of Economics, Virginia Tech, Blacksburg, VA 24061, USA. E-mail: firstname.lastname@example.org
(2) Department of Economics, The Chinese University of Hong Kong, Shatin, NT, Hong Kong. E-mail: email@example.com
Table 1: China's grain output and agricultural inputs: 1952-77 Year Grain Grain Retained output procurement grain (million (million per capita tons) tons) (kg/person) (1) (2) (3) 1952 164 33 270 1953 167 47 257 1954 170 51 265 1955 184 48 278 1956 193 40 306 1957 195 46 295 1958 200 52 286 1959 170 64 223 1960 143 47 212 1961 148 37 229 1962 160 32 241 1963 170 37 245 1964 188 40 270 1965 195 39 271 1966 214 41 287 1967 218 41 287 1968 209 40 265 1969 211 38 266 1970 240 46 289 1971 250 44 298 1972 241 39 286 1973 265 48 303 1974 275 47 307 1975 285 53 315 1976 286 49 317 1977 283 48 313 Year Rural Area sown Draft labour to grain animals (million) (million (million heads) heads) (4) (5) (6) 1952 173 124 76 1953 177 127 81 1954 182 129 85 1955 186 130 88 1956 185 136 88 1957 193 134 84 1958 155 128 78 1959 163 116 79 1960 170 122 73 1961 197 121 69 1962 213 122 70 1963 220 121 75 1964 228 122 79 1965 234 120 84 1966 243 121 87 1967 252 119 90 1968 261 116 92 1969 271 118 92 1970 278 119 94 1971 284 121 95 1972 283 121 96 1973 289 121 97 1974 292 121 98 1975 295 121 97 1976 294 121 95 1977 293 120 94 Year Farm Chemical Sown area machinery fertiliser hit by (million (million calamity HPs) tons) (%) (7) (8) (9) 1952 0.3 0.08 2.9 1953 0.4 0.12 4.9 1954 0.5 0.16 8.5 1955 0.8 0.24 5.2 1956 1.1 0.33 8.2 1957 1.7 0.37 9.5 1958 2.4 0.55 5.2 1959 3.4 0.54 9.7 1960 5.0 0.66 15.3 1961 7.1 0.45 18.6 1962 10 0.63 11.9 1963 12 1.0 14.3 1964 13 1.3 8.8 1965 15 1.9 7.8 1966 17 2.3 6.7 1967 20 2.4 ... 1968 22 2.7 ... 1969 26 3.1 ... 1970 29 3.4 2.3 1971 38 3.8 5.1 1972 50 4.3 11.6 1973 65 4.8 5.1 1974 81 5.4 4.4 1975 102 6.0 6.7 1976 117 6.8 7.6 1977 140 7.6 10.2 Data source: Columns (1)-(2), (4)-(6), and (9) are from MOA (1984); (3)=(1)-(2)+(grain resales from rural to rural areas) divided by rural population, grain resale data are from Lardy (1987) and NBS (1983); and, (7)-(8) are from Wen (1993). Table 2: Death rates of the Chinese provinces: 1954-1966 (unit = 0.1%) Province 1956-1957 1958 1959 Average Anhui 11.7 12.3 16.7 Fujian 8.2 7.5 7.9 Gansu 11.1 21.1 17.4 Guangdong 9.8 9.2 11.1 Guangxi 12.5 11.7 17.5 Guizhou 8.2 13.7 16.2 Hebei 11.3 10.9 12.3 Heilongjiang 10.3 9.2 12.8 Henan 12.9 12.7 14.1 Hubei 10.2 9.6 14.5 Hunan 11.0 11.7 13.0 Inner Mongolia 9.2 7.9 11.0 Jiangsu 11.7 9.4 14.6 Jiangxi 12.0 11.3 13.0 Jilin 8.3 9.1 13.4 Liaoning 8.0 6.6 11.8 Ningxia 10.9 15.0 15.8 Qinghai 9.9 13.0 16.6 Shaanxi 12.2 11.7 12.8 Shandong 12.1 12.8 18.2 Shanxi 10.1 11.0 12.7 Sichuan 11.3 25.2 47.0 Yunnan 15.8 21.6 18.0 Zhejiang 9.4 9.2 10.8 Nation 11.1 12.0 14.6 Province 1960 1961 1962-1963 Average Anhui 68.6 8.1 8.1 Fujian 15.3 11.9 7.9 Gansu 41.3 11.5 9.4 Guangdong 15.2 10.8 8.5 Guangxi 29.5 19.5 10.2 Guizhou 45.4 17.7 9.9 Hebei 15.8 13.6 10.2 Heilongjiang 10.6 11.1 8.6 Henan 39.6 10.2 8.7 Hubei 21.2 9.1 9.3 Hunan 29.4 17.5 10.3 Inner Mongolia 9.4 8.8 8.8 Jiangsu 18.4 13.4 9.7 Jiangxi 16.1 11.5 10.4 Jilin 10.1 12.0 9.7 Liaoning 11.5 17.5 8.2 Ningxia 13.9 10.7 9.4 Qinghai 40.7 11.7 6.9 Shaanxi 14.2 12.2 11.4 Shandong 23.6 18.4 12.1 Shanxi 12.3 8.8 10.0 Sichuan 54.0 29.4 13.7 Yunnan 26.3 11.8 12.5 Zhejiang 11.9 9.8 8.3 Nation 25.4 14.2 10.0 Source: NBS (1990) Table 3: Mess hall participation rate, party membership density, and time of liberation Density of Mess hall party Time of Province participation membership liberation rate at end of in (month/year) 1959 (%) 1956 (%) Anhui 90.5 0.83 01/1949 Fujian 67.2 1.06 08/1949 Gansu 47.7 1.54 08/1949 Guangdong 77.6 0.93 11/1949 Guangxi 91.0 0.85 11/1949 Guizhou 92.6 0.86 11/1949 Hebei 74.4 3.14 11/1947 Heilongjiang 26.5 1.38 10/1948 Henan 97.8 1.08 06/1948 Hubei 68.2 0.77 05/1949 Hunan 97.6 0.80 08/1949 Inner Mongolia 16.7 1.78 09/1949 Jiangsu 56.0 1.37 04/1949 Jiangxi 61.0 1.39 05/1949 Jilin 29.4 1.62 10/1948 Liaoning 23.0 1.75 11/1948 Ningxia 52.9 N.A. 09/1949 Qinghai 29.9 1.04 09/1949 Shaanxi 60.8 1.15 05/1949 Shandong 35.5 2.14 09/1948 Shanxi 70.6 2.92 10/1948 Sichuan 96.7 0.71 11/1949 Yunnan 96.5 0.98 12/1949 Zhejiang 81.6 0.78 05/1949 Source: the first two columns are from Yang (1996, p. 57); the last column is from Kung and Lin (2003)