Politics and Entitlement: Causes of the Great Chinese Famine of 1959-1961
James Kai-sing Kung
【Abstract】 Corresponding author: James Kung, Division of Social Science, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong. E-mail: SOJK@UST.HK. James Kung would like to thank the HKUST for a research grant (DAG97/98.HSS17) and Zhou Feizhou, Fang Huirong and Ho Yin-Yuk for helpful research assistance. The authors would like to thank participants at the American Economic Association's Meetings, Jan. 3-5, 1999 for their valuable suggestions. We alone are responsible for any errors and views in this paper.
The rate at which the world economy has prospered since Malthus' time has surely proven wrong this classical thinker's dismal prognosis of the human livelihood. Yet, despite the unimaginable progress in food production technology and the general consensus that the world at large has long solved the problem of food shortages, famine continues to afflict mankind across time and space. Thus far, while analysts continue to debate among themselves in the midst of their search for more conclusive evidence, the underlying causes of famine remain little understood.
Of all the great famines that have occurred, the Chinese famine should merit special attention, for two reasons. The first is its scale. China suffered the worst famine in history during the three years 1959-1961, at a time when the Communist hyper-euphoria reached its nadir in 1958. While the Irish famine of 1845-1847 may well be the "best-known historical famine of all" and was truly a "great famine" when compared with most modern famines in terms of the number of people killed, the Chinese famine was decidedly a "macabre league of its own" (O' Grada, 1999). Indeed, recent demographic estimations have put the magnitude of excess deaths around 23-30 million, and another 33 millions of lost or postponed births (Ashton, Hill, Piazza and Zeitz, 1984; Peng, 1987). The second is its cause(s). Despite the severity of this famine, however, the Chinese case remains "one of the least understood events" with respect to what actually caused it (Chang and Wen, 1997).
An important reason why it is difficult to pin down the exact cause of the famine is because this tragic event occurred in a period commonly known as the Great Leap, a time of major political and economic upheavals. It is the simultaneous occurrence of a multitude of changes □both policy and institutional □that has resulted in a lack of consensus regarding the cause of the famine. For instance, the fact that grain procurement increased precipitously in the three years 1959-1961 has led some to argue that this was the main cause of the famine, which in turn could be attributed to informational problems (Bernstein, 1984). By the same token, the observation that overall death rate was higher in the rural areas has given rise to the hypothesis that some systemic forces must be in place that biased against the rural populace (Lin and Yang, 1999). Third, the uneven geographic concentration of death tolls in provinces that had seemingly adhered most closely to the Party line reasonably suggests that the famine was tied to some extent to political circumstances specific to these provinces (Yang, 1996). Fourth, and this is a view that has recently provoked considerable debate, sees the famine to be rooted primarily in the institution of food consumption □a feature unique to the people's communes. Under the institutional arrangement of forced, communal dining, farmers were encouraged to eat as much as they wanted; the result was simply a rapid depletion of food (Chang and Wen, 1997; Yang, 1996). Fifth, it is amazingly surprising that only until recently have some researchers begun to attribute the Chinese famine to the more conventional explanations of famine, such as a decline in food availability (Lin and Yang, 1999).
While these are all plausible explanations, they have not been tested in a comprehensive analytical framework. In light of the vast, complex changes that had occurred to both the larger institutional environment and the specific institutional arrangements during the Leap (see below), it is reasonable to expect the famine as being caused by more than just one factor. Against this background, the overriding objective and contribution of this paper is to provide a rigorous empirical estimation of all the existing hypotheses in order to ascertain their relative importance in causing this major human catastrophe.
Our analysis of the Chinese famine, based on a panel data set for 24 Chinese provinces for the 1954-1966 period, reveals that the famine was conceivably caused by the following three factors; namely, a decline in food availability, excessive grain procurement caused by an urban-biased policy, and politics. While preliminary, there is evidence to suggest that, were it not due to politics and the resulting radical policies implemented during especially the latter part of the Leap, the famine's severity could be alleviated. To the contrary, our analysis rejects the thesis that communal dining had negatively affected death rate. Empirical evidence fails to support the unsubstantiated assumption that food was available both for free and in unrestricted quantity in the public mess halls.
This paper is organized as follows. The next section briefly outlines those salient changes in policy and institutions during the Great Leap □changes that had possibly caused the famine. In Section 3, we outline the list of competing hypotheses in some detail and in Section 4 we discuss the proxies employed, the estimation procedure and the data source. The estimation results are presented and analyzed in Section 5. In Section 6, we examine in some detail as to why communal dining did not impact negatively on death rate. Section 7 concludes the study.
Starting from the summer of 1958, China was to experience in the next few years one of the most radical revolutionary changes in the history of mankind □the Great Leap Forward. In a nutshell, the Leap was intended by Chairman Mao to hasten the pace of radically transforming China's economy from its predominantly agrarian nature into a powerful industrial state. Seen from this perspective, a specific objective of the Leap was to increase by manifolds both agricultural and industrial output; the former by means of expanding the acreage covered by irrigation and the latter a stepping up in the capacity of steel and iron production. Both were to be achieved by means of mobilizing vast quantities of China's rural labor force on an unprecedented scale.
To be sure, major institutional changes had already occurred to facilitate the transfer of an agricultural surplus to fuel industrial growth prior to the Leap. Most notable of such a transformation had been the collectivization of land and other major farm resources in a number of stages. Headed by local officials, agricultural collectives carried out planting and other instructions according to a plan specified by the state. Under such a plan, the state was to supply the major farm inputs to these cooperatives, who in turn were obligated to sell a corresponding amount of output to it at prices unilaterally determined by the latter. Despite the putting in place of such a system, agriculture remained a bottleneck in China's industrialization process during the First Five-Year-Plan, circa. 1953-1957, however, as the farm sector had been unable to provide an amount of food and other major farm products to the state at rates rapid enough to speed up industrialization. Indeed, it was against the recognition of such a "planning failure" that Mao had to rely upon mobilizing vast quantities of local resources □ both labor and non-labor □in achieving the intended developmental objectives.
It was indeed reckoned that the most cost-effective way of raising crop yields, given China's availability of a vast pool of human resources, was to expand the irrigated acreage by means of mobilizing vast quantities of the rural labor force. Consisting of fewer than 200 farm families, the agricultural cooperative was however considered too small for effectively organizing local construction projects of a larger scale. This would be especially the case where projects involved the potential cooperative effort of several townships, let alone villages. A commune, which was typically established by merging together a few cooperatives and consisting of thousands of households, was therefore seen as an optimal organization under which such public projects of the sort in question could be efficaciously coordinated.
Similarly, China's lack of capital and technology posed an equally severe constraint for it to produce a vast amount of industrial materials such as iron and steel. Being a much larger organization, the communes were supposedly capable of mobilizing large quantities of labor to work in the local industries, just as it was able to do so with water conservancy projects. The communes were more appropriate an organization, in addition, because the smaller cooperatives typically lacked the kind of managerial skills and expertise required for industrial organization and production (Perkins, 1966: 84). At the height of this frenzy in late 1958, some 60 million peasants were reportedly engaged in smelting iron and mining and transporting ore (Dutt, 1967), a number whose magnitude allegedly amounted to 30 to 50 per cent of the rural labor force (Donnithorne, 1967). The role of the commune, in short, was to provide an organizational device for mobilizing vast quantities of labor in order to achieve the "Leap's" goals (Donnithorne, 1967; Riskin, 1987).
The reliance upon the communes to carrying out the necessary activities intended to raise both farm and industrial output was achieved at huge economic and social costs, including individual work incentives, however. For instance, the zealous mobilization of a substantial quantity of farm labor to work in steel furnaces and iron ores in 1958 had led to the neglect on the part of some communes to harvest the autumn crops. Equally detrimental was that much of the steel produced in the backyard furnaces was of such poor quality that they could hardly be productively employed (Donnithorne, 1967; Perkins, 1966; Riskin, 1987, among others).
More severe though was the adverse incentive effect of having raised the basic unit of accounting and work organization from the collective to the commune level. While farm work was similarly organized in teams, farmers, or members of the collectives, were able to revert to private, independent farming should they consider the final team output fell short of their expectations. This "exit right", in the language of Lin (1990), served as a powerful check on those who may otherwise shirk their effort while working within the team context, insofar as they valued the longer-term economic payoffs derived from team farming. Exit right was, in short, a substitute that effectively replaces the need for diligent farmers to monitor the quality of effort of their team workers. But as the communes came to replace the collectives and removed farmers' right to exit from collective farming, the powerful check that once provided by this institutional arrangement no longer functioned. By organizing work in a highly centralized manner and distributing income in an exceedingly egalitarian fashion using a narrowly differentiated wage system among tens of thousands of farm households, the connection between individual effort and income was basically severed. This must undoubtedly have affected agricultural performance in the famine years in a big negative way.
Unfortunately, the deleterious effect of the commune organization went beyond those of production, but encompassed also major aspects of village social life. In an attempt to release vast quantities of female labor to work in the fields so that their male counterparts could be reallocated for work on irrigation projects and steel manufacturing, a communal dining system was set up. Under this system, farmers' share of the food was no longer distributed to them for individual preparation and consumption at home. Instead, they were compelled to dine in communal mess halls where food would be prepared by the attendant central kitchen and made available free of charge (the so-called "free supply" of food based not on one's merit accrued to her labor, but according to consumption needs). While some are of the view that such a system was in reality not qualitatively different from the one which it replaced (e.g., Perkins, 1966: 88), others maintain that the famine was caused precisely by this institutional arrangement of free and unlimited food supply (Chang and Wen, 1997; Yang, 1996).
Finally, a description of the Leap is incomplete without mentioning the decentralized character that underpinned decision making behind agricultural and industrial production in the countryside. As mentioned earlier, the overriding aim of raising the basic unit of administration from the collectives to the much larger communes was to mobilize the vast quantities of labor to engage in non-farm work. From the center's perspective, it was just as necessary to put pressures upon the local officials who were in charge of their local development projects for achieving the targeted spectacular economic results through exercising their own energy and ingenuity. This selective decentralization of authority was, however, not carefully orchestrated, as central planners had already lost control over resource allocation through the effective manipulation of economic levers. Deserving special emphasis is the purposive decentralization of the State Statistical Bureau's power to collect agricultural production statistics down to the commune organization. To the extent that local officials were subject to intense pressure to fulfill the vastly exaggerated output targets set by their superiors, the former had all the incentives to falsify the farm output data. It was precisely under these "winds of exaggeration" that it became impossible for the central government to ascertain with any remote degree of certainty the amount of farm output China was actually producing during the period in question. In the words of Perkins (1966), "statistics could no longer be used to appraise performance" (p. 86). This may explain, as some claim, why the government procured too much grain from the peasants (e.g., Bernstein, 1984). After hovering at a relatively modest procurement ratio of 23 per cent in 1957, it soared to almost 26 per cent in 1958, and a whopping 39 per cent in 1959. While the excessive procurement of grain in 1959 may largely be the result of misinformation, the continuing high levels of grain procurement in the ensuing two years could not be attributed entirely to such a problem, however. This is because Mao had already learned of the problems pertaining to too sizeable a procurement and the commune organization more generally (Bo, 1993). While procurement was being scaled back to 34 per cent in1960, it was still too high nonetheless and food shortages had already developed to such an extent that the number of cumulative death tolls peaked in China in the same year.
A number of arguments and hypotheses have thus far been put forth to explain the cause of the Chinese famine, yet they are invariably being analyzed in isolation. Given the vastly complicated environment within which the Great Leap occurred, it is highly unlikely that the famine was caused by just one factor, in which case these isolated hypotheses would be unable to account for the possible simultaneous effect of one another in jointly determining the catastrophe in question. The purpose of this section is therefore to outline both the existing hypotheses as well as our own ones in a more comprehensive fashion, discusses the proxies employed in each of these specifications, and the data employed in the analysis. Before doing so, it should be pointed out that the dependent variable to be explained is provincial variations in death rate during the Great Leap period 1958-1961, shown in Table 1.
HYPOTHESIS 1: FOOD AVAILABILITY DECLINE
This hypothesis suggests that the Chinese famine was partially caused by a decline in food availability (FAD), which in turn was triggered by a production failure (due, for example, to natural disasters). More specifically, this hypothesis states that death rate would increase when food supply and accordingly the caloric intake falls below the normal subsistence level, as was the case during the famine period of 1959-1961 (Lin and Yang, 2000).
HYPOTHESIS 2: "URBAN BIAS" POLICY
This hypothesis suggests that the Chinese famine was a result of government's development policy, which had an inherent "urban bias". According to this reasoning, the regime's predisposition to rapidly industrializing China's predominantly agrarian economy required a substantial transfer of agricultural surplus to support the state-sponsored industrial sector. While this "squeezing" of the peasantry would not cause excess death in normal times, the government's guarantee for the urban population of a minimum subsistence consumption in times of food shortages, as was the case during 1959-1961 left inadequate food supply for the rural residents (Lin and Yang, 2000).
HYPOTHESIS 3: COMMUNAL DINING
A salient feature of the communes was the centralization of food preparation and consumption in the communal dining kitchens. Because food was provided free of charge and, more importantly in unrestricted quantity in these mess halls, farmers were encouraged to consume as much as possible, as the benefits of so doing accrued to the individuals, whereas the costs were borne by the group as a whole. Doing so would however quickly deplete the scarce food supply. Such a behavior is made analogous to the "commons' tragedy" (Yang, 1996) and "irrational consumption", respectively (Chang and Wen, 1997).
HYPOTHESIS 4: THE LEAP'S "RADICAL POLICIES"
The radical nature of the Leap was multi-dimensional; consisting of, most prominently, producing steel and carrying out water conservancy works in exceedingly labor-intensive manner. The reason why radical policies may result in higher death rates is because these activities consumed more physical energies of the people at a time when food became less available (the prevailing mentality at the time was "production first, living the second"). Thus, to the extent that provinces did not uniformly enforced these policies, death rates would likely differ. This particular hypothesis is intended to capture the effect of those aspects of the Leap's radicalism not directly related to communal dining and excessive procurement. Consistent with H3, the more radical the Leap's policy was implemented in a province the higher the resulting death rate.
In order to put our hypotheses in a more conceptual perspective, we group them according to the manner in which they effected upon the death rate. For example, the hypothesis that the Chinese famine was negatively affected by a decline in grain output (Hypothesis 1) belongs to the conventional explanation of famine -- one which attributes the occurrence of extraordinary deaths to a decline in food availability (FAD). Conversely, famine can also be triggered by what Sen (1981) refers to as "entitlement deprivation". In the Chinese context, this was articulated in the institutional arrangement of guaranteeing the urban residents a minimum amount of food grain consumption, while leaving the "residual" to the peasants (Lin and Yang, 2000, Hypothesis 2). Hypothesis 3 examines one specific aspect of the Leap's radical policy, namely, communal dining, on the period's death rate, whereas Hypothesis 4 is included to pick up any "residual" radical policy effect not already captured in Hypothesis 3. These "residuals" may include a wide array of radical, energy consuming activities such as farmers' engagement in steel production and in large-scale irrigation projects, "deep plowing", and so forth.
Ideally, grain consumption would be the most appropriate proxy for food availability. Data limitations unfortunately preclude us from using such a variable to directly test this hypothesis (H1). As a surrogate, we employ per capita grain output as a crude measure of food availability (Table 2). With regard to the "urban bias" thesis, the existing test has been to correlate the relationship between the proportion of rural population in a province with its death rates (Table 2). The former variable is employed to proxy for that proportion of the population in that province who, in the words of Lin and Yang (2000), "did not have the legally protected rights to food (and is thus) a proxy for the degree of urban bias in that province" (p. 8). While our test will certainly include this variable, it has several limitations. On the one hand it is not a direct measure of food entitlement, yet on the other it captures a wider spectrum of the urban bias ranging from better medical and healthcare to education and housing provisions in the urban areas. In view of these considerations, we will test the urban bias hypothesis more directly by using the data series on grain procurement-to-output ratios as well (H2, cf. Table 2). Thus, from the urban bias perspective, procurement is excessive only when overall grain output becomes inadequate that in order to protect the urban residents with the guaranteed rights to minimum food consumption a greater proportion of food was procured.
Expressed as a percentage of a province's population participating in communal dining, mess hall participation rate (hereafter MHPR) has been employed to proxy for the varying extent to which a province had involuntarily subject its rural populace to the forced institution of communal dining. Variations in this particular indicator among the provinces are then used to account for the effect of this specific institution on death rate variations (Chang and Wen, 1997; Yang, 1996). While we will use this variable in conjunction with those of others to account for the famine's severity during the period of the Leap (H3), it should be pointed out that figures on mess hall participation rate are available only for the yearend of 1959. For this reason, its effect, if any on death rate other than the years 1959 and 1960 should be read with some caution. As will be pointed out later, it is because of this reason that we also perform individual year-estimations.
Finally, we will employ two different proxies to test the effect of radical policies on death rate (H4). Recall Yang's (1996) idea that the catastrophe's severity was patterned according to how radical a provincial government had actually pursued the Leap's policies: the more radical its proclivities the higher the death rate. The question is how to measure variations in political predisposition. According to Yang, they could be specifically captured by: 1) the depth of the Party base and 2) the length of revolutionary history in a given province. Against this background, he proposes to use Party membership density (hereafter PMD) -- a measure of how many Party members there were per 100 persons, as a proxy for measuring both of these dimensions. Yang's intuitive reasoning is that, the shallower the Party base and the shorter the revolutionary history the larger the proportion of cadres who were non-Party members. In order to "signal" their loyalty to Mao (dubbed the "politics of loyalty compensation"), these non-Party member cadres pursued courses of action that were more radical (Yang, 1996: 59).
The use of PMD to proxy for provincial radicalism is however problematic, in our view. First of all, provincial variations in PMD were rather narrow (they range merely between 0.77 and 3.14, Table 3); so narrow that it is unlikely for provinces with one or two fewer Party members per one hundred persons to end up pursuing policies that were noticeably more radical. More problematic though is that the use of PMD as a proxy of radicalism neglects the critically important issue of who in the ranks of the political hierarchy was primarily responsible for policy implementation. Yang (1996) simply assumes that cadres who were still not members of the Communist Party would rush to become one; and, in order to achieve that they pursued radical courses of action, thereby resulting in higher death rates. Available evidence, on the other hand, however suggests that local policies were to a considerable extent shaped by the predisposition and preferences of the provincial leader -□usually the First Party Secretary, as was richly demonstrated in the case of Henan Province (Xu, 1998). Moreover, the idea that party membership density mattered ignores the dissimilar interests that existed between cadres of different levels; in particular, those at the village levels were basically predisposed to protect local interests and chances for them to "signal" their loyalty to Mao were therefore unlikely (Bernstein, 1984). With these qualifications in mind, we seek an alternative proxy to better represent the underlying dynamics behind the variations exhibited by different provinces in pursuing the center's policy during the Leap.
To the extent that provincial authorities varied in their implementation of the Leap's policies, these variations may be patterned upon a province's affiliation with the center, which in turn was determined largely by its revolutionary history. In particular, owing to the support with which local people of the old liberated provinces lent to the CCP during the civil war, it would be neither necessary nor desirable for Mao to send senior cadres with tough political stance to these provinces to have them complied with Beijing's directive. The same cannot however be said with regard to the "newly liberated provinces" □most of which happened to have been closely affiliated with the Nationalist Party during the civil war. Seen from this perspective, it is thus quite possible that Mao had less trust in the local people in the newly liberalized provinces. To make these provinces comply with the Party line, Mao therefore sent a crop of senior cadres with tough political stance (nanxia ganbu) to these provinces. In this connection, political leaders who were put in charge of the newly liberated provinces were inclined to implement the Leap's policies more rigorously than their counterparts in the old liberated provinces did. If such a conjecture indeed better characterized the actual political economy of provincial radicalism during the Leap, then the timing of "liberation" (or TOL) would be a more appropriate variable than PMD in capturing the differences in radicalism between these two types of provinces. On the basis of this reasoning we thus include TOL as a dummy variable in our estimation as an alternative hypothesis to party membership density, the data of which are presented also in Table 3 (see also Fig. 1). Owing to the more radical policies adopted in the newly liberated provinces, death rates would by implication be higher. Conversely, death rates would likely be lower in the old liberated areas.
(b) estimation procedure and data
Our overriding goal is to account for provincial variations in death rate during the period of the Leap, 1958-1961 using the aforementioned list of explanatory variables. Methodologically we estimate the effects of the group of independent variables on death rate variations for the famine period as a whole. However, since data on MHPR are available for only a given year, namely, 1959, it is therefore likely that this variable will affect death rate in 1959 and/or 1960 only. For this reason, we also estimate the equation with individual year data.
In order to establish a benchmark case for subsequent comparisons, we first re-examine Sen's (1981) entitlement thesis by testing the importance of grain availability (H1) and urban bias (H2), following Lin and Yang (2000). Instead of using only the ratio of rural population as a proxy for urban bias, as the two authors have done, we employ also grain procurement □a more direct measure of urban bias -- in our estimation. After then we will extend our model by adding to our estimation the "political" variables respectively of mess hall participation (MHPR), party membership density (PMD) and time of liberation (TOL). The first two variables are used for the purpose of replicating the analyses undertaken respectively by Chang and Wen (1997) and Yang (1996). But as we have argued, the actual time when a province was "liberated" (TOL) may be more appropriate than PMD in measuring a province's revolutionary history, and should thus be included in the estimation. These three variables are summarized in Table 3. When all these exercises are performed, we will test the effect of the same explanatory variables on individual year-variations in provincial death rates, for the reason already rehearsed. Our full model may now be written as follows:
Dit = b0 + b1Qit + b2Rit + b3Pit + b4Lit + + b5Cit + b6Mit + b7Yt + Sit (1)
Where i indexes a province, t indexes a year, and Dit is the death rate. Qit stands for per capita grain output or food availability, Rit indexing the percentage of population that is rural, and Pit an index of per capita grain procurement. For the political variables, Lit stands for the timing of liberation, Cit is an index for mess hall participation rate, and Mit for party membership density. Yt are year dummies for the Leap period or the disaster years, namely, 1959, 1960 and 1961, and eit is a stochastic disturbance term. The one way fixed-effect specification in (1) assumes that certain characteristics unique to individual years can be captured by differences in the constant terms, causing shifts in provincial death rates.
Provincial data on grain, procurement and population are obtained from the publication A Compilation of Historical Statistical Data of Provinces, Autonomous Regions, and Municipalities (Quanguo Gesheng Zizhiqu Zhixiashi Tongji Ziliao Huibian, 1949-1989), published by the State Statistical Bureau of China. The series of data on death rates, also at the provincial level, are based on A Compendium of Materials on Population and Census Statistics, 1949-1985 (Renkou Tongji Ziliao Hubian, 1949-1985), also published by the State Statistical Bureau. Data on mess hall participation rate are obtained from A Compendium of Important Articles on Agricultural collectivization, 1949-1957, published by the Communist Party School Press, whereas those on the timing of liberation are compiled from various sources outlined in Table 3. Finally, data on party membership density are taken from Yang (1996). The primary reason why we choose to perform our analysis at the provincial instead of a lower (e.g., county) level is due entirely to data availability, which is most complete at this level.
Equation (1) of Table 4 provides the results of our benchmark estimation, against which subsequent estimations can be compared and contrasted. Consistent with the findings of Lin and Yang (2000), both food availability (H1) and entitlement (H2) - the latter measured directly by procurement and indirectly by the ratio of rural population, are found to have negative effect on death rate during the famine period. The sign of the three variables also turns out as expected. The negative coefficient of the GRAIN variable means that death rate would be lower the greater the grain availability, whereas the positive sign of the two urban bias variables (RATRUPO and PROCUREMENT) imply that death rate would be higher the greater the systemic and policy biases against the countryside. The YEAR variable 1960 is also highly significant, suggesting that there are other unexplained factors specific to that year that are not captured by our benchmark variables. For instance, the mobilization of sizeable proportions of the rural labor force to engage in steel production and water conservancy projects may be some such factors that had negatively impacted on the exceptionally high death toll in that year.
TABLE 4 about here
In our full model, the three political variables are first tested in isolation with the three benchmark variables (Equations 2-4), then in pairs (Equations 5-6), before they are all included in one equation (7). It can be seen that in isolation both party membership density (PMD) and time of liberation (TOL) are highly significant at the 1% level, whereas mess hall participation rate (MHPR) is moderately significant at the 5% level. The signs of the coefficients are also consistent with our expectations. The positive coefficient of TOL implies that death rate tends to be lower in the Old Liberated Areas (OLA), and the positive coefficient of MHPR implies a lower death rate in provinces where communal dining was less rigorously enforced. Conversely, the negative coefficient of PMD suggests that provinces with more Party members per capita tend to have a lower death rate.
The estimation results become vastly more interesting when we took turns testing the three variables in a pair-wise fashion, including only two at a time as a means to ascertain their relative importance (Equations 5-7). Two significant findings are observed from these exercises. First, the variable TOL dominates both MHPR and PMD as the single most important political factor in accounting for the provincial variations in death rate in both the estimations in which it is included (Equations 5 and 7). Second, and in sharp contrast to our first finding, MHPR is not significant when either TOL or PMD is included (Equations 5 and 6), suggesting that it may have partially captured the effect of, respectively PMD and TOL when the latter are excluded in the estimation. Note that in all three estimations the two benchmark variables of GRAIN and PROCUREMENT have remained significant across all estimations and their signs unchanged, and that the YEAR variable (1960) also remains significant whenever TOL is not included.
In our final estimation (Equation 8), where the effect of all three political variables are accounted for, of the three of them only TOL is significant; both MHPR and PMD have dropped their significance. This finding squarely suggests that the high death tolls in some provinces were caused by aspects of the Leap policy that were not specifically related to both communal dining (Chang and Wen, 1997; Yang, 1996) and "loyalty compensation" (Yang, 1996). Since we have already discussed the appropriateness with regard to the choice between TOL and PMD as one of the political determinants of death rate variations, we confine our discussions here to explaining why MHPR is unimportant when a number of analysts contend that it should? To a considerable extent we think that is due to the unsubstantiated nature of their allegation. It is simply assumed without qualifications from the political slogan of "open up your stomach and eat as much as you can" that food was not only provided for free, but also in unrestricted quantity. Our own research, based on a careful analysis of local newspaper archive, however reveals that food was in fact rationed to the dining hall members in the fall of 1958 when communal dining was first adopted based on age, sex, and labor grade (Kung and Lin, 1999). Moreover, communal dining policy was not being rigorously enforced for the most part of 1959, and in many areas peasants were in fact given the freedom to prepare for their own food and to eat at home. And while the government's policy on communal dining became tightened in the spring sowing season of 1960, mandating the harvested crop to be distributed to the mess hall instead of the household, the amount of food grains available had become so limited that they could hardly be wastefully consumed. In fact, available evidence suggests that the imperative by then had become one of raising the milling efficiency of grain so that more rice and noodles could be grind using the same amount of raw food grains.
The relative insignificance of these two political variables can also be borne out by the values of the Adjusted-R2 in the different estimations of Table 4. For example, when only the three benchmark variables are included, the value of the Adjusted-R2 is 0.309 (Equation 1). When we add the variable TOL, the value increases noticeably to 0.398 (Equation 3). But when we add also the two other political variables, as we have done in Equation (8), the value of the Adjusted-R2 only rises negligibly to 0.400. In other words, of the three political variables TOL is the one with the greatest explanatory power in terms of death rate variations. In this context, it is interesting to note that whenever the variable TOL is not included in the estimation, the YEAR variable 1960 is always significant [Equations (2), (4) and (6)]. Where it is included, however, TOL becomes significant, whereas 1960 becomes insignificant, suggesting that TOL and the YEAR variable 1960 are capturing some common, underlying effect. While the YEAR variable remains significant, it is now 1961 instead of 1960. In addition, the negative sign of the coefficient implies a significantly lower death rate in that year □a time coinciding with the end of the famine.
Regarding the robustness of the overall fit of our estimations, the reasonably high values of the Adjusted-R2 in all of our estimations suggest the consistently strong explanatory power of the same independent variables employed in the analyses. Moreover, their ability to explain the dependent variable has been consistent throughout the entire analyses even where one or more of the political variables are deliberately removed. Also worth noting is the observation that the coefficients of those significant variables do not change randomly but remain in close range across various estimations. Altogether, these characteristics suggest that the reported estimations are highly robust. To avoid the potentially biased estimation regarding the effect of MHPR and to gain further insights into the causes of the famine, we now turn to examine the individual year effects of the same set of explanatory variables.
In order to gauge the effects of specifically our political variables on each specific year of the famine period, in light of the restrictions that data for MHPR are available for the year 1959, we performed another set of regressions, and the results are presented in Table 5. The following results are worth noting. First, Consistent with our previous findings, food availability (GRAIN) is consistently significant in all the four years in accounting for death rate variations. With the exception of 1961, the same can be said of urban bias, or, specifically, PROCUREMENT. The reason why this variable is not significant for 1961 is straightforwardly intuitive. After death toll reached a nadir in 1960 the Chinese government scaled back its procurement ratio from a hefty 34% in 1960 to less than 25% in 1961 -- a measure that allowed more food to be retained in the countryside.
But it is the second observation that is most enlightening. Consistent with the analytical results presented in Table 4, both MHPR and PMD are not significant □neither for the period as a whole nor during any individual year within that period. Conversely, except for the YEAR 1960, TOL is significant in all the individual year-estimations. It is only rendered insignificant in the 1960 estimation by some specific policy or incident whose effect eradicated any inherent differences between provinces. What policy or incident had the all-empowering effect of converging provinces into uniform action? Consider the following possibility.
Before Marshal Peng Dehuai "challenged" Mao at the Lushan Conference, the Communist Party's main concern was to contain the excessive leftist tendency □so-called "winds of exaggeration" □from further developing (fangzuo). In response to a personal letter by Marshal Peng addressing to him the problems experienced during the Great Leap, Mao charged Peng as having "leaned towards the right for about 30 kilometers" and openly declared that "it was rightist tendency that had now become the main danger" (Bo, 1993). Not only had this famous Lushan incident interrupted with the Party's proclivity to rectify the prevailing leftist tendency, as Bo (1993) has noted, it inadvertently intensified the Great Leap. Under the slogan of "anti-right with all the might" (fanyouqing, guganjing), provinces were under unprecedented, surmounting pressure to mobilize substantial resources □both human and otherwise □in undertaking steel production, irrigation works, other industrial pursuits and enormous hog farms. For example, after having deliberated to scale down steel production target in 1960 to a moderate 13 million metric tons on May 23 1959, the target was being revised in January 1960 to 18.4 million metric tons. Similarly, grain output for 1960 was targeted to reach 300 million metric tons, when actual, realized output in 1958 was only 200. To the extent the Lushan incident and its consequences eliminated any original behavioral proclivities between the old and newly liberated provinces in implementing the Leap's policies in the face of strong political pressure from Beijing, the variable TOL loses its significance.
An attempt has been made in this paper to ascertain China's famine associated with its Great Leap Forward movement. According to our estimations, the famine was caused both by a decline in food availability (FAD) during the period of the Leap as well as by a systemic bias articulated in the form of excessive procurement of food grains against the rural population. Second, our analysis lends no support to the recently popular claim that the famine had been negatively impacted by the communal dining system. It is probably that not only had public mess halls not operated strictly on the basis of the "commons", as our evidence has shown, they were not being rigorously enforced during a good part of 1959. And while their operations were reinvigorated in 1960, per capita grain availability had been declined to such an extent that there was unlikely to be enough food available for the Chinese peasants to "open up their stomach and to eat as much as they can".
Third, a more interesting aspect of our finding is that the Chinese famine had been negatively affected by politics, as is manifestly borne out by the political variable TOL. Prior to 1960, or the Lushan incident, differences in the rigor with which provinces carried out the Leap's policies can be attributed to their varying degrees of affiliation with the Communist Party. In particular, due possibly to their weaker affiliation with the center, the newly liberated provinces were assigned with cadres inclined to pursue policies that were more radical and therefore more detrimental in consequence. After the Lushan incident, however, even provinces with the right "credentials" were not immune from the ultra-leftist policies forced upon them.
It is in this context that we find Sen's cognizance of factors other than a decline in food availability in causing famine provides an inspiring perspective from which our own analysis can draw a parallel. His own theory, generalized from observations in the context of a market economy, points to the possibility of how acute food shortages could result in a sharp rise in the relative price of food, which in turn deprives a certain section of the population from this basic "entitlement". While our analysis of the Chinese famine may be regarded as supporting Sen's theory of entitlement in the broad sense, the latter offers few clues as to why and how famines may occur and develop in the context of a non-market economy such as China. It is to this important issue that the remainder of our conclusion turns.
A salient feature of China's post-Independence development strategy was the disproportionate importance to which it accorded to the heavy industrial sector □a strategy motivated by the political ambitions of catching up with the advanced nations within an incredibly short period of time (Lin et al., 1996). In this context, both decline in food availability and urban bias may to a considerable extent be regarded as the institutional outcomes of this development strategy. For instance, were it not due to this "leapfrogging" attempt, it would not be necessary to adopt the commune system and the resulting sharp fall in grain production. By the same analysis, if development priority had not been disproportionately accorded to the heavy industrial sector, then urban workers would not be better protected in terms of their food consumption (and other) right(s), in which case procurement would not have been excessive in times of food shortages. All the same, the entire Great Leap Forward movement and with it the establishment of the people's commune and the hosts of failed attempts to raise the economy's output was deeply embedded in this heavy-industrialization strategy of transforming China's agrarian economy within an unrealistically short period of time. In short, were it not due to the wholesale adoption of the aforementioned politically motivated development strategy, it may be safe to conclude that both grain availability and an (urban)-biased procurement policy would not have resulted. It is in this sense that both the variables food availability and urban bias are perhaps just as "political" as the variable TOL.
As the effective implementation of this development strategy required the mass mobilization of (predominantly) the rural people, Mao was forced to rely on his own personal charisma but also, perhaps more importantly an authoritarian political structure for implementing his grossly erred visions and policies. Our analysis of the Chinese famine shows that the cost of such policy errors, committed in the course of a series of miscalculations on the part of the leadership, was unimaginably high. It is due perhaps to our analysis of a famine as it occurred in a non-market economy characterized by an authoritarian polity that inspires us to suggest the need of integrating politics into analyzing the causes of famine under the kind of circumstances in question. Our analysis, as it transpires, suggests that the relationship between politics and food availability and entitlement is much less straightforward that previous analyses have made out (e.g., Bernstein, 1984; Yang, 1996), although these works have undoubtedly provided a useful benchmark from which subsequent works such as this one has benefited. Without slighting our own contributions to the analysis of the Chinese famine, this work merely opens us a vast, uncharted research agenda concerning the intriguing relationship between an authoritarian polity, a politically well-intended but unrealistic development strategy, and the effect of revolutionary history on bringing about a deadly famine. We hope to have opened the right door to an important chapter.
Ashton, B., K. Hill, A. Piazza and R. Zeitz. (1984), "Famine in China, 1958-61", Population and Development Review, 10 (December 1984): 613-45.
Bernstein, T. (1984), "Stalinism, Famine, and Chinese Peasants," Theory and Society, 13:339-377.
Bo, Y. B. (1993), Several Important Policy Decisions and Events in Retrospect (Ruogan Zhongda Juece yu Shijian de Huigu). Beijing: The Chinese Communist Party School Press.
Bradley, M. and M. Clark, (1972), "Supervision and Efficiency in Socialized Agriculture," Soviet Studies, 23: 465-473.
Chang, G. H. and G. J. Wen, (1997), " Communal Dining and the Chinese Famine of 1958-1961," Economic Development and Cultural Change, 46 (1): 1-34.
A Compendium of Important Articles on Agricultural Collectivization, 1949-1957 (Nongye Jiti hua Zhongyou Wenjian Huibian, 1949-1957). Beijing: The Chinese Communist Party School Press.
Dutt, G. (1967), Rural Communes of China. New York: Asia Publishing House.
Economist, The (1999), "Who's Afraid of Genetically Modified Foods?" (June 19th): 17-18.
Kung, J. K. (1993), "Transaction Costs and Peasants' Choice of Institutions: Did the Right to Exit Really Solve the Free Rider Problem in Chinese Collective Agriculture?" Journal of Comparative Economics, 17 (2): 485-503.
Kung, J. K. and J. Y. Lin, (1999), " Food Availability versus Entitlement: The Cause(s) of China's Modern Famine Re-examined." Paper presented at the American Economic Association Annual Meetings, Jan 3-5, New York.
Kung, J. K. and L. Putterman, (1997), "China's Collectivization Puzzle: A New Resolution," Journal of Development Studies, 33 (6): 747-753.
Li Xin and Chen Tiejian (1997), The Final Showdown: A History of the New China Democratic Revolution, 1947-1949 (Zuihou de Quezhan: Zhongguo Xin Minzhu Zhuyi Geming Shi Changbian, 1947-1949). Shanghai Chubanshe.
Lin, J. Y. (1990), "Collectivization and China's Agricultural Crisis in 1959-1961," Journal of Political Economy, 98 (6): 1228-1252.
Lin, J.Y., Fang Cai, and Li Zhou (1996), The China miracle: Development Strategy and Economic Reform. Hong Kong: Chinese University of Hong Kong Press.
Lin, J. Y. and D. Yang, (2000), "Food Availability, Entitlements and the Chinese Famine of 1959-61" (forthcoming) Economic Journal.
Lipton, M. (1977), Why Poor People Stay Poor: A Study of Urban Bias in World Development. London: Temple Smith.
O' Grada, C. (1999), Black '47 And Beyond: The Great Irish Famine in History, Economy, and Memory. New Jersey: Princeton University Press.
Peng, X. Z., (1987), "Demographic Consequences of the Great Leap Forward in China's Provinces", Population and Development Review, 13 (4): 639-70.
Riskin, C. (1998), "Seven Questions About the Chinese Famine of 1959-1961," China Economic Review, Volume 9, No. 2: 111-124.
Riskin, C. (1987) China's Political Economy: The Quest for Development Since 1949. New York: Oxford University Press.
Sen, Amartya K. (1981) Poverty and Famine. Oxford, U. K.: Clarendon Press.
State Statistical Bureau (1990), A Compilation of Historical Statistical Data of Provinces, Autonomous Regions, and Municipalities, 1949-1989 (Quanquo Gesheng Zizhiqu Zhixiashi Tongji Ziliao Huibian, 1949-1989). Beijing: Zhongguo Tongji Chubanshe.
State Statistical Bureau (1988), A Compendium of Materials on Population and Census Statistics, 1949-1985 (Renkou Tongji Ziliao Hubian, 1949-1985). Beijing: Zhongguo Caizheng Jingji Chubanshe.
Symposium on Theoretical and Historical Perspectives on Chinese Agricultural Cooperatives (1993), Journal of Comparative Economics, Vol. 17, No. 2 (June): 471-559.
Yang, D. L. (1996), Calamity and Reform in China: State, Rural Society, and Institutional Change since the Great Leap Famine. Stanford: Stanford University Press.
Zhang Xiongru (1993), A Compendium of important Events in Twentieth-Century China (Ershi Shiji Zhongguo Dashi Chuanshu). Beijing Chubanshe.