Famine and Overweight in China

Review of Agricultural Economics—Volume 28, Number 3—Pages 296–304 DOI:10.1111/j.1467-9353.2006.00290.x

Famine and Overweight in China∗

Zhehui Luo, Ren Mu, and Xiaobo Zhang

T

here is an increasing body of literature that examines the association between restricted fetal growth and diseases in adulthood as proposed by Barker. One common way to test the hypothesis in humans is to make use of a natural disaster, such as famine, that happened during gestation and examine disease prevalence in later life. Most of the famine-based epidemiological studies use the 1944–5 Dutch Hunger Winter when a sharp decline in food intake occurred due to a German army blockade. Drawing on retrospective cohort analyses, these studies in general find that famine has a negative impact on various health outcomes. For example, prenatal exposure to famine is believed to be associated with antisocial personality disorder in early adulthood (Neugebauer, Hoek, and Susser), major affective disorders (Brown et al.), and schizophrenia (Hulshoff et al.) in adulthood, and higher BMI and waist circumference in fifty-year-old women (Ravelli et al.). However, studies (Stanner et al.) based on a small sample (less than 600 people) of survivors of the Leningrad siege of 1941–4 lead to opposite findings from those on the Dutch famine (Ravelli, van de Meulen, and Michels).1 In a word, the findings are inconclusive. Compared to the Dutch Famine, the Great Famine in China from the late 1950s to the early 1960s lasted much longer and affected more people. The estimated excess deaths numbered from 20 to 30 million (Johnson). The regional distribution of the famine was highly uneven. As shown in table 1, the percentage change of the highest mortality rate during 1959–62 relative to the average mortality rate prior to the famine in 1956–8 ranged from 14.9% in Tianjin to 474.9% in Anhui province (Yang). In addition, cities suffered much Zhehui Luo is a research associate at Michigan State University. Ren Mu is a consultant at the World Bank. Xiaobo Zhang is a senior research fellow at the International Food Policy Research Institute (IFPRI). ∗ This paper was presented at the Principal Paper session, “The Diet Transition and Obesity in Developing and Developed Countries: Determinants and Policy Options,” Allied Social Sciences Association annual meeting, Boston, January 6–8, 2006.

The articles in these sessions are not subject to the journal’s standard refereeing process.

Proceedings

297

Table 1. Mortality rates in China, 1956–62 (per 1,000 population)

China Tianjin Shanxi Shaanxi Shanghai Inner Mongolia Zhejiang Heilongjiang Ningxia Beijing Xinjiang Jiangxi Hebei Yunnan Jilin Guangdong Jiangsu Shangdong Fujian Hubei Liaoning Guangxi Hunan Gansu Henan Sichuan Qinghai Guizhou Anhui

1956

1957

1958

1959

1960

1961

1962

11.4 8.8 11.6 9.9 6.8 7.9 9.5 10.1 10.6 7.7 13.9 12.5 11.3 15.2 7.5 11.2 13 12.1 8.4 10.9 6.6 12.5 11.5 10.8 14 10.4 9.4 13 14.3

10.8 9.4 12.7 10.3 6 10.5 9.3 10.5 11.1 8.2 13.9 11.5 11.3 16.3 9.1 8.4 10.3 12.1 7.9 9.6 9.4 12.4 10.4 11.3 11.8 12.1 10.4 12.4 9.1

12 8.7 11.7 11 5.9 7.9 9.2 9.2 15 8.1 13.9 11.3 10.3 21.6 9.1 9.1 9.4 12.8 7.5 9.6 8.8 11.7 11.7 21.1 12.7 25.2 13 15.3 12.4

14.6 9.9 12.8 12.7 6.9 11 10.8 12.8 15.8 9.6 18.8 13 12.3 18 13.4 11.7 14.6 18.2 7.9 14.5 11.8 17.5 13 17.4 14.1 47 16.6 20.3 16.7

25.4 10.3 14.2 12.3 6.8 9.5 11.9 10.5 13.9 9.2

14.2 9.9 12.2 8.7 7.7 8.8 9.8 11.1 10.7 10.8

16.1 15.8 26.3 10.1 15.1 18.4 23.6 15.3 21.2 11.5 29.5 29.4 41.3 39.6 54 40.7 52.3 68.6

11.5 13.6 11.9 12.1 10.7 13.4 18.5 11.9 9.2 17.5 19.5 17.5 11.5 10.2 29.4 11.7 23.3 8.1

10 7.4 11.3 9.4 7.3 9 8.6 8.6 8.5 8.7 9.7 11 9.1 10.9 10 9.3 10.4 12.4 8.3 8.8 8.5 10.3 10.4 8.2 8 14.6 5.4 11.6 8.2

Average 1956–8

Percentage Change

11.4 9 12 10.4 6.2 8.8 9.3 9.9 12.2 8 13.9 11.8 11 17.7 8.6 9.6 10.9 12.3 7.9 10 8.3 12.2 11.2 14.4 12.8 15.9 10.9 13.6 11.9

122.8 14.9 18.3 22.1 23.5 25.5 27.5 28.9 29.2 35 35.3 36.8 44.1 48.6 56.4 57.8 68.8 91.4 92.9 111.3 111.7 141.8 162.5 186.8 208.6 239.6 272.3 285.5 474.9

Note: From Yang (p. 38). The bold rows are the provinces included in the CHNS surveys. The last column “percentage change” is the change in mortality rate from the average level in 1956–8 to the highest value over the period of 1959–62.

less from famine than rural areas because of preferential supply of food to cities (Lin and Yang). Most studies on the Great Famine focus on uncovering its magnitude and causes. The origins of the famine are traced to various factors, including excessive food consumption in collective dining halls (Yang; Yang and Su; Chang and Wen), lower production incentives due to the denial of the right to withdrawal from the collectives (Lin), preferential supply of food to cities, favoritism of industry over agriculture during this period (Lin and Yang), and a possible mix of all the above factors (Li and Yang). It is generally agreed that the catastrophe was primarily induced by policies adopted during the Great Leap Period (Johnson).

298

Review of Agricultural Economics

While these studies contribute to understanding the causes of this human tragedy, rather limited knowledge exists on the long-term health consequences of the Great Famine. To our knowledge, only two papers study the impact of the Chinese Great Famine on health outcomes of survivors in later life. Gørgens, Meng, and Vaithianathan find significant stunting of growth in those exposed to the famine by using the height of the second generation as an instrument. Clair et al. collect and analyze the psychiatric case records in a mental hospital in Wuhu of Anhui Province, one of the most severely famine-stricken regions. Their study shows that prenatal exposure to famine greatly increases the risk of schizophrenia in later life. The regional variation of China’s Great Famine provided us an opportunity to examine the long-term health impact of the famine with a different identification strategy from the above two studies. We follow Yang and classify provinces into severe and less severe famine areas based on mortality (table 1) and use the cohorts born after the famine, or in less severe famine provinces or urban areas, as control groups. In this paper, we present the impact of exposure to famine in early life on overweight in adulthood several decades later.2 Overweight and obesity rose rapidly in both rural and urban areas in China.3 According to China’s first official nutrition and health survey, the adult rate of overweight was 22.8% in 2002, a 39% increase from 1992 (China Daily). The prevalence of overweight and obesity was much higher in urban areas compared with rural areas (20.9% versus 14.2%) (Du et al.). It is well known that overweight and obesity are important risk factors of many adult diseases. Therefore it is imperative to understand the causes of rapid increase in overweight and obesity. Biologically, increase in body weight is the result of energy intake in excess of energy expenditure, regulated in part by peptide hormones in the brain and gut. Most studies attribute the changing body weight patterns among the Chinese to shifts in diet and activity, a large increase in fat from animal sources, reduced daily physical activity, urbanization, and westernization of lifestyle (Popkin et al.; Sundquist and Johansson; Paeratakul et al.; Popkin; Stookey et al.; Du et al.). Yet to our knowledge, no studies have linked the observed patterns of overweight and obesity among adults to exposure to severe malnutrition prenatally and in infancy, although malnutrition was prevalent in the central planning era when most of the prime age population was born. As this generation ages with access to more adequate food, their likelihood of becoming overweight or obese might be higher than those not subject to malnutrition in early life (Barker). In other words, having been constrained during early life, exposures to abundant food supply (high fat, sugar diets, etc.) in later life may increase one’s susceptibility or risk of developing obesity and other chronic diseases.

Data The China Health and Nutrition Survey (CHNS) is a longitudinal survey that covers nine provinces that vary substantially in geography, economic development, public resources, and health indicators in 1989, 1991, 1993, 1997, and 2000.4 These data have detailed information on various measures of health outcomes, such as height, weight, blood pressure, activities of daily living

Proceedings

299

(ADL), self-reported health status, morbidity, physical function limitations, and disease history. A multistage, random cluster process was used to draw the sample. Counties in the nine provinces were stratified by income (low, middle, and high) and a weighted sampling scheme was used to randomly select four counties (one in low, two in middle, and one in high-income levels) from each province. In addition, the provincial capital and a lower income city were selected. Villages and townships within the counties and urban and suburban neighborhoods within the cities were selected randomly. There are about 190 primary sampling units and 3,800 households covering approximately 16,000 individuals of all ages. In 1989, health and nutrition data were only collected from preschoolers and adults aged twenty to forty-five. We therefore use data from 1991 to 2000, excluding pregnant women. This data set has been widely used for studying diet and health patterns for the Chinese population (see a review by Du et al. and citations therein). However, none of these studies have linked famine with health outcomes.

Methodology and Results In this study, we make full use of the uneven distribution of famine across regions and time. Comparing health outcomes of the famine cohort across provinces with different degrees of severity, however, may not be good enough to isolate the consequence of nutritional deficiency, because socioeconomic factors at the provincial level may affect the severity of famine as well as health outcomes of the famine survivors in their later life. Comparing the famine cohort with other nonfamine cohorts may render an estimation bias for the simple reason that the famine effect, if there is any, cannot be separated from the cohort effect. Even if the famine is a dominant determinant of later health, the selection bias associated with the surviving famine cohort may mask the famine effect. We adopt a double-difference approach by comparing the difference between cohorts across provinces under different degrees of famine severities.5 The double-difference method enables us to a large extent to control for some unobserved endogenous factors as well as selection bias of the famine survivors. Tables 2 and 3 illustrate the main idea of the double-difference identification strategy.6 The famine cohort consists of those who were born during 1959–62 when exposure to famine happened in utero and in infancy. A younger cohort of those born in 1963–6 serves as controls. Table 2 presents the probability of overweight for women in different categories. Panel A compares the probability of overweight individuals who were born during 1963–6 and therefore had no exposure to famine, with that of individuals who were born during 1959–62 and exposed to famine, in both types of provinces.7 For the famine cohort, the probability of being overweight is higher in the provinces that experienced more severe famine. For the young cohort, however, the probability is higher in provinces subject to less severe degree of famine. In both types of provinces, the famine cohort has higher overweight probability than the younger cohort, probably reflecting the age effect. However, the double difference between cohorts and provinces is significantly larger in the famine cohort in severe provinces. The difference in these differences can be interpreted as the causal effect of the famine, under the

300


Table 2. Proportion of women overweight by cohort and severity of famine Level of Famine Severity Severe (1)

Less Severe (2)

Difference (1)–(2)

Panel A: Comparison of interest (rural) Cohort born during 1959–62

0.208 0.183 (0.033) (0.015) Cohort born during 1963–6 0.097 0.158 (0.018) (0.013) Difference of the above two rows 0.111 0.025 (0.035)∗∗ (0.019) Panel B: Control comparison I (rural) Cohort born during 1963–6 0.097 0.158 (0.018) (0.013) Cohort born during 1967–70 0.056 0.124 (0.015) (0.013) Difference of the above two rows 0.040 0.033 (0.024)∗ (0.019)∗ Panel B: Control comparison II (urban) Cohort born during 1959–62 0.262 0.183 (0.037) (0.022) Cohort born during 1963–6 0.179 0.128 (0.032) (0.017) Difference of the above two rows 0.083 0.056 (0.049)∗ (0.027)∗∗

0.025 (0.035) −0.061 (0.025)∗∗ 0.086 (0.042)∗∗ −0.061 (0.025)∗∗ −0.068 (0.023)∗∗ 0.007 (0.034) 0.079 (0.041)∗ 0.051 (0.034) 0.028 (0.053)

Note: Standard errors are in parentheses. The symbols ∗∗ and ∗ represent significance levels of 5% and 10%, respectively. The severe famine provinces are Guizhou and Henan; and the less severe provinces are Guangxi, Hubei, Hunan, and Jiangsu.

assumption that in the absence of famine, the differences between the cohorts would not have been systematically different in severe and less severe provinces. This estimator shows that exposures to famine increase the probability of overweight by 0.086, which is significantly different from zero at the conventional significance level of 5%. This identification strategy may be invalid if the pattern of differences in cohorts varies systematically across provinces. However, we can explicitly test this identification assumption by using younger cohorts who were born after the famine and therefore were not subject to famine. As shown in Panel B, the difference in the probability of overweight between younger cohorts does not differ systematically across provinces. The result of the difference in differences for the young cohort is close to zero (0.007) and insignificant. In Panel C, we also use the same cohorts in the urban sample as a control experiment because in general urban residents were not subject to severe famine due to favorable food entitlements. However, because of the inflow of rural residents to cities in the late years, some of the current urban residents may not

Proceedings

301

Table 3. Proportion of men overweight by cohort and severity of famine Level of Famine Severity Severe (1)

Less Severe (2)

Panel A: Comparison of interest (rural) Cohort born during 1959–62 0.189 0.210 (0.040) (0.016) Cohort born during 1963–6 0.072 0.107 (0.016) (0.012) Difference of the above two rows 0.117 0.102 (0.036)∗∗ (0.02)∗∗ Panel B: Control comparison I (rural) Cohort born during 1963–6 0.072 0.107 (0.016) (0.012) Cohort born during 1967–70 0.073 0.118 (0.017) (0.013) Difference of the above two rows 0.000 −0.011 (0.023) (0.017) Panel B: Control comparison II (urban) Cohort born during 1959–62 0.247 0.193 (0.048) (0.023) Cohort born during 1963–6 0.208 0.213 (0.040) (0.021) Difference of the above two rows 0.039 −0.02 (0.062) (0.031)

Difference (2)–(1) −0.02 (0.045) −0.035 (0.021)∗∗ 0.015 (0.045) −0.035 (0.021)∗ −0.046 (0.023)∗∗ 0.011 (0.031) 0.054 (0.051) −0.005 (0.045) 0.059 (0.068)

Note: Standard errors are in parentheses. The symbols ∗∗ and ∗ represent significant levels of 5% and 10%, respectively. The severe famine provinces are Guizhou and Henan; and the less severe provinces are Guangxi, Hubei, Hunan, and Jiangsu.

have been born in cities. As rural people enter cities, having access to more adequate food and changing to a more sedentary lifestyle may increase their likelihood of gaining weight more than other urban residents. The table shows a higher prevalence of overweight in urban areas of the severe famine provinces, suggesting an important role of migration from rural to urban areas in these provinces and the rapid changes in diets and lifestyle that the new migrants experience. Therefore, the results for the urban control group may serve as a lower bound of estimates. As shown in the last row of Panel C, the result of double difference for the urban comparison is 0.028, not statistically significant from zero. The results in Panels B and C suggest that the double difference estimator is based on valid assumptions. Table 3 is similar to table 2 except that it is for men. In Panel A, despite being positive at 0.015, the double difference is statistically insignificant. In other words, the effect on overweight in men is positive but not as discernible as in women. For the two comparison groups in Panels B and C, the double difference does not yield significant results. All in all, the results in tables 2 and 3 show that

302


exposures to famine at the early stage of life are associated with higher likelihood of overweight, in particular for women. This result replicates the findings by Ravelli et al. based on the Dutch Famine and lending support to the fetal origin hypothesis, but only for women.

Discussions Traditionally, mortality rate is often used as a measure of severity or cost of famine. However, negative health impacts on famine survivors indicate that mortality alone underestimates the true cost of famine. Famine and malnutrition have been chronic problems plaguing many developing countries. One of the key targets of the Millennium Development Goals is halving hunger and malnutrition by 2015. As the developing world strives to achieve the goals in the next few decades, the negative effect of malnutrition in early life on health outcomes in adulthood may increase along with the nutritional transition. It is therefore imperative to identify and quantify the prenatal and postnatal determinants of adverse adult health outcomes. Such link is also crucial to understanding the “external” consequence of the nutritional neglect of women (Sen), because it implies a causal pattern from maternal undernourishment to adult disease. By using a heterogeneous sample in a large developing country that suffered the largest famine in human history, our study makes a serious effort toward this direction. The relation between exposure to severe famine at the prenatal and infancy stages with overweight in the adulthood is found to be gender-specific. The effect appears to be more pronounced in women than men. In interpreting these results, we have to bear in mind several limitations in our study. First, we cannot distinguish prenatal (different trimesters during pregnancy) and postnatal effect, in part, due to lack of data and the fact that the Chinese Great Famine lasted a long time. The negative impact of early exposure to famine on health in adulthood may depend on the timing of exposure, as shown by other researchers. Secondly, even though among those surveyed in the CHNS only a small number of household heads and their spouses confirmed that they lived in a different province than where they were born, we do not know the degree of migration within the same province from rural to urban areas during the famine. As we know, the degrees of famine severity in rural and urban areas were different, this may compromise our ability to test the hypothesis using the urban sample. A future study to examine the urban population by tracing back to their place of origins seems to be warranted. Third, the double difference approach does not control for individual characteristics that may be systematically different between cohorts and across provinces, which in turn may be correlated with health outcomes. In further studies, a regression method is needed to provide a robust check on the findings of the paper.

Acknowledgments These are the views of the authors, and should not be attributed to the World Bank or any affiliated organization. The authors are grateful for the helpful comments from Corinna Hawks, Per Pinstrup-Anderson, Marie Rue, Parke Wilde, and participants in the AAEA session, “The Diet

Proceedings

303

Transition and Obesity in Developing Countries and Developed Countries: Determinants and Policy Options”, held in Boston, January 6, 2006.

Endnotes 1 One possibility for the difference is that the Dutch general population was well nourished prior to and after the blockade. Studies from a more heterogeneous sample in developing countries will shed more light on the debate. The less severe form of natural disaster such as drought is also found to have negative effects on child growth (Hoddinott and Kinsey). 2 A more comprehensive analysis including other health outcomes can be found in Luo, Mu, and Zhang. 3 Overweight is defined by body mass index (BMI) greater than or equal to 25 kg/m2 . BMI is measured by weight (kg) divided by height (m2 ). Different cutoff points for overweight and obesity for Chinese population have been suggested. We adopt the above cutoff, however, to be consistent with the current standard of studying overweight and obesity by the World Health Organization. 4 See http://www.cpc.unc.edu/projects/china for details. 5 The severe famine provinces are Guizhou and Henan; and the less severe provinces are Guangxi, Hubei, Hunan, and Jiangsu. We take the residence province as the province of birth partly because we only have information on the place of birth for household head and their spouse. In the CHNS, among those who had ever been a household head (n = 6,300), we know some of their place of birth (n = 5,295); and only a few (n = 310) lived in a different province at the time of interview. For spouses, we know only a small proportion of place of birth (n = 1,420) and only 131 spouses lived in different province from birth. 6 The number of observations for the each gender/cohort/province cell ranges from 94 to 816 in rural area, and 66 to 398 in urban area. 7 As shown in table 1, the highest mortality rate within each province occurred during 1959–61. The vast majority of the 1959–62 cohorts was subject to prenatal malnutrition. We do not have information on monthly mortality rate to narrow down the impact of famine based on pregnancy stage. The paper looks at exposure to famine “during early life” and outcomes at adulthood. The early life includes prenatal and/or first few years of life.

References Barker, D.J. Mothers, Babies, and Disease in Later Life, 2nd ed. New York: Churchill Livingstone. 1998. Brown, A.S., J. van Os, C. Driessens, H.W. Hoek, and E.S. Susser. “Further Evidence of Relation Between Prenatal Famine and Major Affective Disorder.” Amer. J. Psychiatry 157, no. 2(2000):190–5. Chang, G.H., and G.J. Wen. “Food Availability versus Consumption Efficiency: Causes of the Chinese Famine.” China Econ. Rev. 9, no. 2(1998):157–65. China Daily. “Assessing Status of Nation’s Health.” October 12, 2004. Clair, D., M. Xu, P. Wang, Y. Yu, Y. Fang, F. Zhange, X. Zheng, N. Gu, G. Feng, P. Sham, and L. He. “Rates of Adult Schizophrenia Following Prenatal Exposure to the Chinese Famine of 1959–1961.” JAMA 294, no. 5(2005):557–62. Du, S., B. Lu, F. Zhai, and B.M. Popkin. “The Nutrition Transition in China: A New Stage of the Chinese Diet.” In The Nutrition Transition: Diet and Disease in the Developing World, B. Caballero and B.M. Popkin, eds, pp. 205–22. London: Academic Press, 2002. Gørgens, T., X. Meng, and R. Vaithianathan. “Selection and Stunning Effects of Famine: A Case Study of the Great Chinese Famine.” Working Paper, 2004. Hoddinott, J., and B. Kinsey. “Child Growth in the Time of Drought.” Oxf. Bull. Econ. Stat. 63, no. 4(2000):0305–9049. Hulshoff Pol, H.E., H.W. Hoek, E. Susser, A.S. Brown, A. Dingemans, H.G. Schnack, N.E. van Haren, L.M. Pereira Ramos, C.C. Gispen-de Wied, and R.S. Kahn.“Prenatal Exposure to Famine and Brain Morphology in Schizophrenia.” Amer. J. Psychiatry 157, no. 7(2000):1170–2. Johnson, D.G. “China’s Great Famine: Introductory Remarks.” China Econ. Rev. 9, no. 2(1998):103–9. Li, W., and D.T. Yang. “The Great Leap Forward: Anatomy of a Central Planning Disaster.” J. Pol. Econ. 113, no. 4(2005):840–77. Lin, J.Y. “Collectivization and China’s Agricultural Crisis in 1959–1961.” J. Pol. Econ. 98, no. 6(1990):1228–52. Lin, J.Y., and D.T. Yang. “Food Availability, Entitlements and the Chinese Famine of 1959–61.” Econ. J. 110, no. 460(2000):136–58. Luo, Z., R. Mu, and X. Zhang. “The Long Term Health Impact of Chinese Great Famine.” DSG Discussion Paper, International Food Policy Research Institute, forthcoming.

304


Neugebauer, R.,H. W. Hoek, and E. Susser. “Prenatal Exposure to Wartime Famine and Development of Antisocial Personality Disorder in Early Adulthood.” JAMA 281, no. 5(1999):455–62. Paeratakul, S., B.M. Popkin, K. Ge, L.S. Adair, and J. Stevens. “Changes in Diet and Physical Activity affect the Body Mass Index of Chinese Adults.” Int. J. Obes. 22(1998):424–31. Popkin, B.M. “Urbanization, Lifestyle Changes and the Nutrition Transition.” World Dev. 27, no. 11(1999):1905–16. Popkin, B.M., S. Paeratakul, F. Zhai, and K. Ge, “A Review of Dietary and Environmental Correlates of Obesity with Emphasis on Developing Countries.” Obes. Res. 3, (Suppl 2)(1995):S145–53. Ravelli, A.C.J., J.H. van de Meulen, and R.P.J. Michels. “Glucose Tolerance in Adults after Prenatal Exposure to Famine.” Lancet 351(1998):173–7. Ravelli, A.C.J., J.H. van de Meulen, C. Osmond, D.J. Barker, and O.P. Bleker “Obesity at the Age of 50 in Men and Women Exposed to Famine Prenatally.” Amer. J. Clin. Nutr. 70, No. 5(1999):811–6. Sen, A. “The Argumentative Indian.” New York: Farrar, Strauss and Giroux. (2005). Stanner, S.A., K. Bulmer, C. Andres, O.E. Lantseva, V. Borodina, V.V. Poteen, and J.S. Yudkin “Does Malnutrition in Utero Determine Diabetes and Coronary Heart Disease in Adulthood? Results from the Leningrad Study, A Cross-Sectional Study.” Br. Med. J. 315(1997):1342–8. Stookey, J.D., L. Adair, J. Stevens, and B.M. Popkin “Patterns of Long-Term Change in Body Composition Are Associated with Diet, Activity, Income and Urban Residence Among Older Adults in China.” J. Nutr. 131, Suppl:S2433–40. Sundquist, J., and S.E. Johansson. “The Influence of Socioeconomic Status, Ethnicity and Lifestyle on Body Mass Index in a Longitudinal Study.” Int. J. Epidemiol. 27(1998):57–63. Yang, D.L. Calamity and Reform in China: State, Rural Society, and Institutional Change since the Great Leap Famine. Stanford CA: Stanford Univ. Press. 1996. Yang, D.L., and S. Fubing Su. “The Politics of Famine and Reform in Rural China.” China Econ. Rev. 9, no. 2(1998):141–55.