Market Transition, Education, and Family Strategies in Rural China
275
MARKET TRANSITION, EDUCATIONAL DISPARITIES, AND FAMILY STRATEGIES IN RURAL CHINA: NEW EVIDENCE ON GENDER STRATIFICATION AND DEVELOPMENT* EMILY HANNUM Two theoretical perspectives have dominated debates about the impact of development on gender stratification: modernization theory, which argues that gender inequalities decline with economic growth, and the “women in development” perspective, which argues that development may initially widen gender gaps. Analyzing cross-sectional surveys and time-series data from China, this article indicates the relevance of both perspectives: while girls’ educational opportunities were clearly more responsive than boys’ to better household economic circumstances, the era of market transition in the late 1970s and early 1980s failed to accelerate and, in fact, may have temporarily slowed progress toward gender equity.
umerous studies in sociology, demography, and economics have investigated how N gender stratification changes over the course of development. Proponents of classical modernization theory have argued that gender inequalities are likely to decline with industrialization or economic growth. In education, this trend is thought to be driven by a micro-level process: better economic circumstances erode incentives for families to make different decisions about investments in children on the basis of their gender. However, alternative perspectives, grounded in Boserup’s (1989 [1970]) widely cited work and the “women in development” (WID) framework, argue that development may initially widen gender gaps because men and boys are better positioned to benefit from new economic opportunities and improved economic circumstances. Extensions of the WID perspective have gained currency in studies of market transition in postsocialist Asian nations and structural adjustment elsewhere in the developing world (Lantican, Gladwin, and Seale 1996). While market transitions and structural adjustment policies were designed to enable growth and did so in many cases, they brought new direct and opportunity costs to bear on families’ educational decisions. In settings where families operate with strong survival incentives for investing in sons, these costs are likely to disproportionately affect decisions about daughters’ schooling. This article examines educational gender stratification in rural China in the era of the transition to a market economy. I address four specific questions: in the early years of reforms, was girls’ schooling more sensitive than boys’ to improved economic circumstances? Did more gender-neutral family decisions about educating children emerge with better economic circumstances? Did the pace of progress toward educational gender
*Emily Hannum, Department of Sociology, University of Pennsylvania, 3718 Locust Walk, Philadelphia, PA 19104-6299; E-mail:
[email protected]. The author was supported while working on this article by a fellowship from the National Academy of Education, funded by the Spencer Foundation. An earlier version of this article was presented at the 2002 annual meeting of the American Sociological Association, August 16–19, Chicago. Comments from Yu Xie, Claudia Buchmann, Peggy Kong, and James Demopolos are gratefully acknowledged. I thank Yu Xie for providing access to tabular data from the censuses of China. Demography, Volume 42-Number 2, May 2005: 275–299
275
276
Demography, Volume 42-Number 2, May 2005
equality speed or slow with the transition to a rapid-growth economy? Finally, is there evidence that costs introduced in the period of reforms disproportionately affected decisions about girls’ schooling? Answers to these questions illuminate the modernization and WID perspectives as explanations for trends in educational gender stratification in China. PERSPECTIVES ON GENDER, EDUCATION, AND DEVELOPMENT This article considers expectations about gender inequality in the wake of China’s market reforms that are grounded in two general theoretical perspectives, set out in Forsythe, Korzeniewicz, and Durrant (2000): the modernization (or neoclassical) approach, and the WID approach.1 The Modernization Approach A prominent perspective on the relationship between development and gender inequality in education is derived from “modernization” or “industrialization” theory. Modernization theory posits that economic development reduces the significance of ascribed characteristics, such as gender, and increases the significance of individual achievement in the status-attainment process (Treiman 1970). At the national level, modernization theory hinges on the idea that the development of a modern economic system creates incentives for rationalization of the allocation of schooling across ascribed traits, such as gender, that are presumably unrelated to productive capacity (Forsythe et al. 2000). Although evidence for the ameliorative impact of development on educational disparities that are associated with other ascribed characteristics has been tenuous, research on gender disparities in education has been surprisingly consistent (Buchmann and Hannum 2001). Indeed, in the burgeoning cross-national literature on educational stratification, one of the few common themes has been the tendency of girls’ education to catch up with and, in some cases, to surpass boys’ education over time with economic development (Knodel and Jones 1996; Schultz 1995; Shavit and Blossfeld 1993). Higher-income countries typically exhibit a smaller female disadvantage in years of schooling than do lower-income countries, and progress toward gender equity over the past several decades has been the most substantial in countries characterized by rapid economic growth (Schultz 1995). At the level of individuals, the mechanism that drives the favorable impact of development on gender disparities is conceptualized within a household welfare framework, as the process of development modifies family incentives for differential investments in boys and girls (Papanek 1985). The mechanism of change is as follows: where scarcity dictates careful considerations about returns to the family for investments in schooling and labor markets and family systems traditionally differentiated the value of such returns in a manner that privileged boys, there is a strong incentive for a preference for sons in decisions about children’s schooling (King and Hill 1993; Stromquist 1989). With economic development, perceptions of gender differences in labor market opportunities and in the likelihood of intergenerational coresidence may lessen in salience, although the empirical basis for this statement and its implications for family decision making have not been well established. More directly, as families gain economic resources, they face less-stringent economic pressure to choose who among their children will be educated. The findings of individual-level studies in many developing countries have been consistent with the notion that girls’ schooling responds more strongly than boys’ to wealth (Filmer 1999; Knodel and Jones 1996; Schultz 1995). 1. I did not consider a third approach defined by Forsythe et al. (2000:577), namely, the critical feminist or “gender and development” approach, which “emphasizes the continuing or rising vulnerability of women over the course of development.” Speaking narrowly about gender and basic education in China, not about gender inequality in China more broadly, the results presented here and earlier research suggest no empirical base for this hypothesis (see, e.g., Hannum and Xie 1994; Lavely et al. 1990).
Market Transition, Education, and Family Strategies in Rural China
277
The WID Perspective Alternatives to the modernization framework have emerged for explaining the relationship between gender stratification and development (for a review, see Forsythe et al. 2000). Most notably, Boserup’s (1989 [1970]) widely cited thesis argues that early stages of development increase the gap between women and men. Only after a threshold has been reached do gender disparities begin to narrow, as expected under the modernization framework. Boserup focused on occupations, using the argument that early stages of development reinforce traditional gender hierarchies in the workplace (see Forsythe et al. 2000). With reference to African and Asian data, Boserup argued that the mechanization of agriculture initially widened gaps between women and men. Paralleling this theme, in urban China and Vietnam, recent scholarship has argued that the market transition has increased the potential for gender discrimination in hiring and firing.2 For rural areas in transitioning Asia, an analogous concern has been the “feminization of agriculture.” In rural China, for example, research has linked the contracting of farmland and resources to individual households to a concentration of women’s work in the household and in agriculture as men have moved into rural industries (Entwisle, Henderson, and Short 1995; Jacka 1997; Judd 1994). Evidence that is more systematic can be found in Forsythe et al.’s (2000) test of Boserup’s (1989 [1970]) thesis in a crossnational analysis. Using a broad-based measure of gender inequality that was derived from scores on the Human Development Index and the Gender-Related Development Index, Forsythe et al. presented evidence of a longitudinal, curvilinear relationship between economic growth and changes in gender inequality, with inequalities first increasing and then decreasing. A corresponding argument has been extended to educational trends. For example, Lantican et al. (1996) found that in three countries in Asia, gender inequalities in elementary schooling first increased and then decreased as structural transformation proceeded.3 At the micro-level, the traditional WID perspective suggests that such trends would emerge as families with new resources first used them to expand opportunities available to sons, rather than to bring daughters to the level of sons. In South Asia, for example, certain evidence suggests the possibility that wealthier or better-educated households do not necessarily exhibit greater gender equality, and sometimes exhibit less gender equality, in investments in children.4 Yet, cross-national research using data from the Demographic and Health Surveys has found that in a substantial number of countries, including
2. Summerfield (1994) argued that the main threat to urban women in China in the wake of economic reforms occurred through discrimination in hiring and policies to reduce the pressure of surplus labor. Managers who were faced with numerous applicants for each position and with a growing awareness of the need to minimize costs frequently chose male applicants over equally qualified female applicants, citing the costs of maternity and child care benefits as the reason (see also Stockman 1994). In contrast, Rama (2001:7) argued that in Vietnam, the lower cost of women’s labor would make women more attractive with market reforms. However, Rama also showed that public-sector downsizing disproportionately affected women. 3. This result did not apply to higher levels of education, where gender differences narrowed with growth. However, an argument could be made that primary levels are the most relevant for two reasons. First, parents have the most-dominant influence at the primary level of education. Second, in societies characterized by high rates of early school leaving, early stages of schooling are the primary sites of social selection. In China, for example, gender disparities have historically been greater at lower than at higher levels of transition (Lavely et al. 1990). In Nepal, gender differences in primary school were the greatest at the stage of entry and much smaller for grade-to-grade transitions, conditional on school entrance and earlier transitions (Stash and Hannum 2001). 4. For an example for education in Nepal, see Stash and Hannum (2001). Regarding investments in children’s health, Das Gupta (1987) found that in rural Punjab, India, women’s education was associated with reduced child mortality but stronger discrimination against higher-birth-order girls.
278
Demography, Volume 42-Number 2, May 2005
South Asian countries, living in higher-socioeconomic-status households benefits girls more than boys (Filmer 1999:19).5 However, extending the WID perspective, there are reasons why a pattern may emerge in which income growth could fail to have the expected consequences for educational gender inequality, even when girls did benefit more from improved household economic circumstances. In the 1980s, the cost of education began to rise in many developing countries as a consequence of the market transition or structural adjustment. For example, accompanying market-oriented reforms, many nations in Asia experienced a growth in private education, increased direct costs to individuals for public education, and rising opportunity costs for education that were associated with new wage-earning opportunities for children (Bray 1996). For rural families in these societies, agricultural reforms often brought not only new economic opportunities but also new uncertainties and heightened opportunity costs for educating children. The implications of reforms for educational costs suggest a slightly different scenario for trends in gender inequality than do those that are offered by either modernization theory or the standard WID perspective. Where wealth may be associated with more-equitable household decisions about schooling and market reforms were successful in stimulating growth, gender inequality could initially fail to decline or could even increase if girls were particularly susceptible to cost constraints.6 In societies where families depend on sons for old-age support or the labor market is stratified along gender lines, there is a clear potential for girls to be particularly vulnerable to new costs, even if they would eventually benefit from increased incomes. THE CHINESE CONTEXT China offers several attractive attributes as a case with which to examine links among development, shifts in market-oriented policy, and educational gender inequality. The first strength of the Chinese case is that educational and economic policies that were associated with the market transition shared direct parallels with those in other transitioning economies in Asia. With the transition to markets, education was reconceptualized as an instrument more of economic development than of socialist political ends. Reforms that were aimed squarely at improving the efficiency and quality of schools, reportedly to the point of shutting down some low-quality rural schools in the early reform years, and fiscal decentralization tightened the link between school resources and local economic conditions (Cheng 1996; Hannum 1999; Park, Li, and Wang 2002; Tsang 1994). While the central government provided financial transfers to poor rural areas, the financing of education became the responsibility of local governments and communities, and poor areas that did not have the resources to finance education increasingly covered their costs by charging fees to families (Davis 1989; Lewin and Wang 1994; Piazza and Liang 1998). Coinciding with the rising direct costs for schooling were new opportunity costs that were brought about by market-transition economic policies. For rural areas, agricultural decollectivization and the return to family farming provided new economic opportunities but also heightened uncertainties that were associated with the lack of a safety net. Although growth in rural incomes allowed families to devote more resources to the education
5. Specifically, Filmer (1999) identified one group of countries with a low female disadvantage among the rich but a reasonably large (greater than about 9 but less than about 15 percentage points) disadvantage among the poor: Mozambique, Guatemala, Uganda, and Cameroon. He found another group of countries in North Africa and South Asia where the female disadvantage was small among the rich but large among the poor: Egypt, Pakistan, India, Central African Republic, Nepal, and Morocco. 6. Consistent with this notion, structural-adjustment policies in Africa led to substantial declines in public investment in education; this change increased the costs to families and prompted significant declines in female, but not male, secondary school enrollments (Assié-Lumumba 2000; Buchmann 1996). However, Forsythe et al. (2000) failed to show significant effects of structural adjustment on trends in gender inequality.
Market Transition, Education, and Family Strategies in Rural China
279
of children, the implementation of the rural responsibility system returned the productive function to the household and thereby increased the economic value of child labor (Summerfield 1994). Thus, starting in the 1980s, educating a child required the forfeiture of revenues from his or her economic contribution to the household (World Bank 1992). Most rural school leavers in the 1980s were working: Knight and Li (1993) estimated that more than 83% of all rural 14- to 19-year-old school leavers in 1988 were engaged in income-generating activities. Second, these changes were superimposed on a rural family system that paralleled those of a number of other less-developed settings, especially in Asia. Patrilocal marriage traditions meant that long-term returns on investments in daughters were more likely to be realized by marital, rather than natal, families, while the reverse was true for sons. Thus, the education of a son was more likely to be perceived as a necessary investment for support in old age (Lin 1993). Analyses of survey and census data from the early reform period confirmed that there were gender gaps in enrollment in rural settings and showed that the gaps occurred in more-impoverished settings and more-impoverished households (Brown and Park 2002; Connelly and Zheng 2003; Hannum 1999; for a review of recent Chinese scholarship on enrollment patterns, see Zheng, Niu, and Xing 2002). Third, the reputed impact of market reforms parallels many WID arguments. To some degree, market transition “feminized” agriculture because men had greater access to rural industrial jobs (Summerfield 1994:722). Because wage work was better paid and was more likely than agricultural labor to reward educational credentials directly, perceptions that sons would be more likely to get wage work would reinforce incentives to make their education a priority.7 Summerfield (1994:721) also argued that the household responsibility system in rural China reinforced the value of sons, who were expected to contribute to their families’ increasingly uncertain future welfare (for a parallel set of observations about Vietnam, see Goodkind 1995). Poorer prospects for accessing the earnings of adult daughters provided a clear financial incentive, especially to poorer families, to avoid new direct and opportunity costs that were associated with educating daughters and to allow daughters to contribute to the household economy until they married and left the family households (Greenhalgh 1994; Lin 1993). The World Bank (1992:83–84) reported that as late as 1990, more than 80% of the 4.8 million school-age children who dropped out were girls, mostly from rural and remote mountainous areas and from minority groups. Fourth, China’s turnaround in economic performance was unusually clearly demarcated. The pace of economic growth in China changed sharply after the Cultural Revolution in the late 1970s, as reforms stimulated unprecedented economic growth and the reduction of poverty in the early 1980s (Piazza and Liang 1998). The well-defined point of change in China offers an unusual opportunity to investigate the changing trajectory of trends in gender inequality. Finally, the case of China also contains one unique attribute: the one-child policy. One-child families have no incentive to discriminate against girls in education, and scholars have argued that the one-child policy has reduced educational gender differences in urban China (Tsui and Rich 2002). In the conclusion to this article, I consider the implications of fertility trends.
7. A 2000 survey of the attitudes of the mothers of 2,000 9- to 12-year-old children in rural Northwest China (Hannum and Kong 2004) showed that almost half the mothers thought that education influences boys’ income more than girls’ (although the vast majority also believed that education influences the earnings of both boys and girls). Parents face incentives to act on their perceptions of boys’ higher earning potential, since most mothers reported that they expected to receive at least economic support from their children, primarily their sons, in the future.
280
Demography, Volume 42-Number 2, May 2005
HYPOTHESES This article investigates the applicability of three sets of expectations about patterns and trends in gender inequality in reform-era rural China corresponding to different theoretical perspectives. Modernization perspective. For the case of China, a standard modernization hypothesis can be stated in two ways: first, at the micro level, girls’ schooling is more responsive to better household economic circumstances than boys’ because wealthier families make more-egalitarian decisions about investing in the schooling of girls and boys. Second, at the macro level, the period of market reforms, characterized by rapid economic growth, sped progress toward gender equality. WID perspective. A traditional WID perspective suggests an alternative set of hypotheses: first, that girls’ schooling was not more responsive to better household economic circumstances than boys’ and, second, that the early period of market reforms may have slowed, and certainly would not have sped, progress toward gender equality. Extended WID perspective. The policies of market reforms suggest an additional set of hypotheses that could be viewed as an extension of the WID perspective. Even if girls benefited from better family economic circumstances, as implied by modernization theory, the early transition to a rapid-growth economy could fail to speed progress toward gender equality, as expected under the WID framework. This trend could emerge if girls were particularly vulnerable in the early reform era to new direct and opportunity costs of schooling. Thus, the hypotheses that are consistent with this perspective are (1) that girls’ schooling is more responsive to better household economic circumstances than boys’, (2) that the market transition failed to speed progress toward gender equality, and (3) that girls’ schooling was particularly sensitive to costs. DATA AND METHODOLOGICAL APPROACH Because no single data set would allow a comprehensive investigation of these hypotheses, I drew on available evidence from several data sources. In the micro-level analyses, I used data from the 1988 Chinese Household Income Project (hereafter referred to as CHIP) and the 1992 National Survey on the Situation of China’s Children (hereafter referred to as the Children’s Survey). In addition, I used tabular data from UNESCO (n.d.a), the Ministry of Education (1984), and the 1990 and 2000 censuses (All China Marketing Research Co. 2003; Census Office 1993) to illustrate temporal changes in gender disparities. The rural CHIP sample was drawn from a larger National Statistics Bureau sample of 67,186 households using a probability-proportionate-to-size method with the population stratified by income. The resulting 10,258 households represented all provinces except the Tibet Autonomous Region, the Xinjiang Uygur Automous Region, and Taiwan. The questionnaire included detailed indicators of household socioeconomic status and an enrollment-status variable for children (Eichen and Zhang 1993). To examine gender, income, and enrollment, I used an extract of children aged 7–16 with valid responses on all relevant variables, yielding a sample size of 9,827 cases. With this extract, I used logistic regression models to estimate the scope of gender disparities in schooling and the interaction of gender disparities with resource constraints in the household. To examine family strategies for investing in the education of children, I used a second extract from the CHIP data. Since the focus of this section is on examining family choices about investments in sons and daughters, households with at least one school-aged daughter and one school-aged son were selected as the unit of analysis. For this purpose, a school-aged child was defined as a child in the sample age range (7–16). With this extract, I used ordinary least-squares regression to estimate logged household educational spending and its variation with the number of school-aged sons and daughters (N = 2,113). I then
Market Transition, Education, and Family Strategies in Rural China
281
used bivariate probit analysis to estimate the effects of the household-income quintile on enrollment decisions for daughters and sons (N = 2,579). The second individual-level data source is the 1992 Children’s Survey, a cooperative effort of UNICEF and the Chinese government (State Statistical Bureau 1992, 1993). Fieldwork for this survey was conducted in June 1992. The rural component encompassed 381,039 households. The sample was drawn using a stratified multistage, probabilityproportionate-to-size approach with equal-interval systematic sampling at each stage.8 The sample included all provinces except the Tibet Autonomous Region and Taiwan. This analysis focused on an extract of rural children aged 7–14 (N = 231,110). I used data from the Children’s Survey for two purposes: (1) to replicate the CHIP analysis of the links among gender, household resources, and enrollment and (2) to examine gender differences in reported reasons for children’s school leaving. For the second purpose, I estimated multinomial logit models of reasons for leaving school among all children and among those who were not currently in school, testing for significant gender differences in the reasons related to direct and opportunity costs. Finally, to examine changes in gender disparity, I constructed gender-equity measures at the primary and secondary levels using data from UNESCO (n.d.a) and the Ministry of Education (1984), as well as tabular data from the 1990 and 2000 censuses (All China Marketing Research Co. Ltd. 2003; Census Office 1993) to show national trends in primary- and secondary-level percentage female in the student population, the ratio of female gross enrollment ratios to male gross enrollment ratios, and female-to-male ratios of primary school entrance rates and junior high school transition rates. ANALYSES The analyses focused on investigating the viability of the modernization and WID perspectives. In the following sections, I investigate the responsiveness of girls’ schooling to improved economic circumstances, the impact of economic circumstances on gender bias in household educational choices, the trend in gender disparities in the years of reform, and the relative vulnerability of girls’ schooling to new direct and opportunity costs. Responsiveness of Girls’ Schooling to Economic Circumstances Table 1 shows rural enrollment rates tabulated by gender and family characteristics, including income quintile, for both the 1988 and 1992 data. Here, I focus on differences in enrollment that are associated with gender and income. Both surveys attest to a lower enrollment rate for girls. In 1988, boys’ enrollment rates were 85.0%, while girls’ were 75.2%. The 1992 data set, with a slightly younger sample, shows boys’ enrollment rates at 92.7% and girls’ at 86.5%. More to the point, both data sets indicate that the enrollment gap between wealthier and poorer children was greater for girls. The 1988 data show a 17.2% gap between girls in the bottom fifth and top income quintiles compared with a 5.4% gap for boys. Similarly, the 1992 data show an 11.7% gap for girls but only a 5.6% gap for boys. In a bivariate context, these results suggest that economic circumstances were more strongly associated with decisions about educating daughters than about educating sons. For other
8. Each of 29 provinces was taken as a universe. (Beijing and Tianjin were considered together as a universe, as were Hainan and Guangdong.) Sample size was allocated across provinces with the intention of facilitating analyses of infant mortality and infant and child nutrition. The sample was stratified into urban and rural areas according to population proportions. Rural areas, counties, and towns that were not in the suburbs or outskirts of cities were divided into three strata: flat areas, hilly areas, and mountainous areas. The sample within rural areas was apportioned across strata according to population proportions. Updated 1990 census lists were used to create the sampling frames.
Demography, Volume 42-Number 2, May 2005
282
Table 1.
Enrollment Rates Tabulated by Family Characteristics and Gender, Rural China, 1988 and 1992a 1988 CHIP Survey
___________________________________
Male
Family Characteristics Overall By Household Income Quintile Top fifth
Female
1992 Children’s Survey
___________________________________
Male
Female
________________
________________
________________
N
% Enrolled
N
% Enrolled
N
% Enrolled
________________
N
% Enrolled
5,096
85.0
4,731
75.2
121,245
92.7
109,865
86.5
1,086
87.3
879
83.3
20,707
94.8
19,305
91.2
Fourth
1,036
84.9
926
77.5
27,128
94.7
25,275
90.4
Third
1,043
85.2
923
76.8
24,215
93.6
22,017
88.1
Second
973
85.2
993
73.7
17,128
92.0
14,954
84.4
Bottom fifth
958
81.9
1,010
66.1
32,056
89.2
28,303
79.5
26
80.8
21
66.7
369
95.1
332
94.0
13,782
95.8
12,473
93.6
By Father’s Education Tertiary Senior secondaryb b
Lower secondary
2,281
88.3
2,148
82.6
36,515
95.3
32,393
91.4
Primary
1,919
84.4
1,832
72.3
48,702
92.6
44,203
86.2
870
77.6
730
61.2
15,397
84.2
14,255
70.0
By Cadre Household No 4,671
84.5
4,371
74.8
––
––
––
––
89.9
360
80.6
––
––
––
––
Less than primary
Yes By Children in the Household 1
425
245
84.9
110
78.2
30,279
94.3
19,457
87.7
2
1,524
88.5
1,095
84.6
55,453
94.4
45,776
90.3
3
1,729
84.7
1,602
75.0
26,094
90.4
30,052
84.8
4
1,004
82.4
1,091
71.2
7,269
84.9
10,913
77.9
594
81.0
833
68.2
2,150
80.7
3,667
71.4
5+
Sources: Chinese Household Income Project and Children’s Survey. a The 1988 sample includes 7- to 16 year olds. Selected children are children of household heads and the number of children in a sample child’s household is calculated as number of people who report that they are children of the household head. The 1992 sample includes 7- to 14 year olds; the children in the household measure employs a variable that picks up all children under age 14 in the household. b
For 1988, lower and upper secondary levels are combined.
household characteristics—father’s educational attainment and number of children in the household—there is a similar pattern of a stronger association with girls’ than boys’ schooling. A measure of political status of the household, whether or not the household is a township or village cadre household or a household of a local government official, shows a different pattern. Residence in a township or village cadre household conferred a similar, slight benefit on girls and boys. I tested observations about Table 1 with a multivariate statistical analysis of the determinants of enrollment. This approach allows tests for statistical significance of the
Market Transition, Education, and Family Strategies in Rural China
283
income effect with potentially confounding factors controlled. Using both data sets, I estimated logit models predicting enrollment (1 = enrolled, 0 = dropped out).9 Table 2 presents the results of the analyses of the 1988 survey data (top panel) and of the 1992 survey data (bottom panel). In the base model (Model 1), the higher rate of enrollment in 1992 is reflected in the larger constant term. However, there are significant, negative effects of being female for both years, of strikingly similar magnitudes of –0.62 and –0.69. Exponentiating the coefficients yields odds of enrollment for girls that are about half those for boys (.54 in 1988 and .51 in 1992).10 Model 2, the main-effects model, introduces linear and quadratic controls for age and terms for number of children in the household; father’s education; and, for 1988, cadre status. With these factors controlled, the coefficients for gender are virtually identical in the two surveys and again correspond to the odds of enrollment for girls of about 50% of those of boys. Of central interest for the questions that motivated this research are the gender-income interactions introduced in Model 3. In both surveys, significant interactions between gender and logged income indicate that girls’ enrollment was more strongly associated with income than boys’. Model 4 reestimates this model with a set of indicators for province, to block out possible regional differences and to provide a more conservative test of the household income effects. Although the logged income measure is insignificant with this specification in 1988, in both years, the gender coefficient and the interaction with income remain significant and stable in magnitude, suggesting that regional differences in poverty or gender norms did not explain the observed gender effects and gender-income interactions. Economic Circumstances and Family Gender Bias The results from Table 2 bolster the notion that, cross sectionally, girls’ schooling responded more strongly to better economic circumstances than boys’. However, the results do not provide a direct evaluation of household strategies, as conceptualized in the household welfare framework. In this section, I investigate gender bias in family expenditures for schooling and in the allocation of education among children, with attention to how strategies of investment change with family income. Since this section focuses on examining within-family variations related to the gender composition of children, households with at least one school-aged daughter and one school-aged son (with school-aged defined as ages 7–16) were selected as the unit of analysis. I used the 1988 data set, which offers measures of educational expenditures and comprehensive measures of family income. Parents’ greater commitment to investing in the education of sons than of daughters can be directly illustrated by the fact that having school-aged sons in the household increases yearly household educational expenditures more than does having school-aged daughters. Table 3 presents the results of linear regressions of logged yearly household educational expenditures on the number of school-aged boys and girls in the household.11 9. The model estimated is specified as follows: logit(Ui) = xieG eG,
(1)
where Ui is the probability of enrollment, xi is a vector of covariates, and G is a vector of regression coefficients. 10. These statements refer to the odds of enrollment for girls compared to boys, not the probabilities of enrollment (× 100) that are shown in Table 1. The same odds ratios can be obtained from Table 1 using the formula [pf / (1 – pf )]/ [pm / (1 – pm )], where pf is the probability of enrollment for girls and pm is the probability of enrollment for boys. 11. Models were estimated using the following specification: yi = F + G1xi1 + G2xi2 + Ji,
(2)
where i indexes cases; yi is logged household educational expenditures; xi1 and xi2 represent the number of school-aged boys and school-aged girls in the household, respectively; and G1 and G2 are associated coefficients.
Demography, Volume 42-Number 2, May 2005
284
Table 2.
Estimated Coefficients for a Logit Model Predicting Enrollment, Rural China, 1988 and 1992
Predictors
(2) (3) (4) (1) Main Gender Province Base Effects Interaction Controls _________________ _________________ _________________ _________________ Parameter SE Parameter SE Parameter SE Parameter SE
1988 CHIP Survey Analytic variables Gender (ref. = male)
–0.62**
0.05
Log income
–0.67**
0.06
–2.71**
0.64
–2.90**
0.65
0.37**
0.06
0.17*
0.08
0.13
0.09
0.34**
0.10
0.36**
0.11
Gender × income interaction Demographic variables Age
2.78**
0.09
2.78**
0.09
2.81**
0.09
Age squared
–0.12**
0.00
–0.12**
0.00
–0.13**
0.00
Number of children
–0.14**
0.03
–0.14**
0.03
–0.10**
0.03
SES controls Indicators for province No Father’s education (ref. = less than primary) Primary
No
No
Yes
0.40**
0.08
0.41**
0.08
0.37**
0.08
Secondary
0.87**
0.08
0.88**
0.08
0.80**
0.09
Tertiary
0.32
0.44
0.31
0.45
0.27
0.42
0.24†
0.12
0.24†
0.12
0.30*
0.13
0.64
–13.70**
0.72
–13.89**
0.75
Cadre household (ref. = no) Constant 1992 Children’s Survey Analytic variables Gender (ref. = male) Log income
1.73**
0.04
–14.85**
–0.69**
0.01
–0.66**
0.02
–1.01**
0.10
–1.05**
0.11
0.24**
0.01
0.22**
0.01
0.20**
0.01
0.05**
0.01
0.04**
0.01
Gender × income interaction Demographic variables Age
3.12**
0.03
3.12**
0.03
3.15**
0.03
Age squared
–0.15**
0.00
0.15**
0.00
–0.15**
0.00
Number of children
–0.38**
0.01
–0.38**
0.01
–0.27**
0.01 (continued)
A model is estimated using the full sample. The sample is then divided into income quintiles, and the same model is estimated separately on each subsample. Turning first to the full-sample model, Table 3 shows that, on average, additional school-aged boys in the household lead to significantly greater increases in expenditures than do additional school-aged girls: on average, an additional boy raises expenditures by 30% (exp[.26] = 1.30), while an additional girl raises expenditures by about 18%. A test
Market Transition, Education, and Family Strategies in Rural China
285
(Table 2, continued)
(2) (3) (4) (1) Main Gender Province Base Effects Interaction Controls _________________ _________________ _________________ _________________ Parameter SE Parameter SE Parameter SE Parameter SE
Predictors
1992 Children’s Survey (cont.) SES controls Indicators for province No Father’s education (ref. = less than primary) Primary
No
Junior secondary
No
Yes
0.93**
0.02
0.93**
0.02
0.73**
0.02
1.50**
0.02
1.50**
0.02
1.22**
0.03
Senior secondary
1.82**
0.04
1.82**
0.04
1.55**
0.04
Tertiary
1.78**
0.20
1.79**
0.20
1.57**
0.21
–14.87**
0.18
–14.67**
0.19
–14.59**
0.20
Constant
2.54**
0.01
Notes: Age, age squared, net income, and numbers of sons and daughters are coded as continuous variables; all other variables are categorical. Robust standard errors allow for clustering within households. Sources: Chinese Household Income Project and Children’s Survey. †
p < .10; *p < .05; **p < .01
Table 3.
Regressions Predicting Household Logged Yearly Educational Expenditures, Rural China, 1988 By Income Quintile
_____________________________________________________________
Full Sample
Low Fifth
Second Fifth
Third Fifth
Fourth Fifth
Top Fifth
0.26** (0.05)
0.39** (0.12)
0.26* (0.11)
0.12 (0.09)
0.15† (0.09)
0.38** (0.11)
Girls
0.17** (0.04)
0.02 (0.10)
0.13 (0.09)
0.08 (0.09)
0.12 (0.08)
0.30** (0.08)
Constant
3.23** (0.09)
2.86** (0.23)
3.11** (0.21)
3.49** (0.18)
3.63 (0.19)
3.29** (0.21)
Predictors Number of Schoolaged Children Boys
Model N
2,113
387
421
429
444
432
Model F
F(2, 2110) = 21.06
F(2, 384) = 5.65
F(2, 418) = 3.32
F(2, 426) = 1.17
F(2, 441) = 2.06
F(2, 429) = 10.83
Model Significance
.0000
.0038
.0370
.3098
.1291
.0000
Model R 2
.0196
.0286
.0156
.0055
.0092
.0481
Test Boys = Girls
F(1, 2110) = 2.72
F(1, 384) = 6.81
F(1, 418) = 0.90
F(1, 426) = 0.11
F(1, 441) = 0.12
F(1, 429) = 0.36
Significance Level
.0994
.0094
.3432
.7395
.7342
.5474
Note: Standard errors are shown in parentheses. Source: Chinese Household Income Project. †
p < .10; *p < .05; **p < .01
Demography, Volume 42-Number 2, May 2005
286
of the hypothesis that coefficients for boys and girls are the same is rejected with marginal significance (F(1, 2110) = 2.72, p = .0994). I next turn to the results tabulated by income quintile. The difference in results across income quintiles is striking. The coefficients for boys and girls are similar in magnitude for the top three quintiles. The magnitudes differ among the bottom two quintiles, but the difference is statistically significant only in the bottom quintile (F(1,384) = 6.81, p =. 0094). For this group, the gap is dramatic: on average, an additional girl in the household raises expenditures by about 2%, while an additional boy raises expenditures by about 48%. These results are consistent with the notion that improvements in economic circumstances promoted egalitarian decisions about investing in the education of sons and daughters. The link between economic circumstances and egalitarianism is also clear in patterns of educational allocation among children. Table 4 presents a bivariate probit model that jointly estimates two outcomes: whether or not all school-aged girls in the household are enrolled in school (coded 1 if all school-aged girls are enrolled, 0 otherwise) and whether or not all school-aged boys in the household are enrolled in school (coded 1 if all schoolaged boys are enrolled, 0 otherwise).12 For the purpose of this article, the primary benefit of joint estimation is that it enabled me to calculate probabilities of different combinations of outcomes (for example, the probability that a household has chosen to enroll all its boys and girls, to enroll all its boys but not all its girls, and so on).13 The predictor variable in Table 4 is income quintile. The top panel in Table 4 presents coefficients; the bottom panel presents W, the correlation of the errors of the boys’ and girls’ equations, and goodness-of-fit statistics. A significant, positive W indicates that net of controlled effects, the likelihood of enrolling all school-aged boys is positively related to the likelihood of enrolling all school-aged girls. In other words, unmeasured family effects exert a similar influence on educational decisions for sons and for daughters. An examination of the coefficients themselves reveals that households in the lowest income quintiles were less likely to enroll all school-aged girls or all school-aged boys, but that the impact on girls was stronger. Negative coefficients for the lowest two income quintiles were significantly larger for girls than for boys.14 Using the models to calculate predicted probabilities under different economic circumstances provides a more intuitive illustration of the range in decision making about the 12. The bivariate probit model begins with a system of two equations (Hardin 1996): yib = XibG + Jib
(3.1)
yig = XigG + Jig,
(3.2)
and
where
¬ J ib ¼ © ¬0¼ ¬ I ½ ~ bivariate normal, ª ½, X 2 ® J ig ½¾ « ®0¾ ®WI
WI ¼¹ ½º . I ¾»
Here, y is the underlying propensity for families to enroll all school-aged boys (Eq. (3.1)) and all schoolaged girls (Eq. (3.2)), i indexes cases, X is a vector of predictors, Gb and Jib are the coefficients and error term for the equation predicting the enrollment of all school-aged boys, Gg and Jig are the coefficients and error term for the equation predicting the enrollment of all school-aged girls, and r is the correlation between the error terms of the two equations (Jib and Jig). Because joint estimation of the two equations allows for the use of information about the correlation between the errors Jib and Jig , this approach provides more efficient estimates of the coefficients than does separate estimation (Zellner and Lee 1965). 13. The formulas for calculating joint probabilities for bivariate probit models can be found in Hardin (1996). 14. The difference in the coefficients in the two models was statistically significant for the poorest two quintiles of households. Details are given in the notes to Table 4.
Market Transition, Education, and Family Strategies in Rural China
Table 4.
287
Bivariate Probit Regression Jointly Predicting the Enrollment of Households’ School-aged Boys and Girls, Rural China, 1988 All Boys Enrolled
______________________
Predictors
All Girls Enrolled
______________________
Parameter
SE
Parameter
SE
–0.09 0.01 –0.14 –0.26**
0.08 0.09 0.08 0.08
–0.26** –0.13 –0.35** –0.49**
0.08 0.08 0.08 0.08
Constant
0.70**
0.06
0.47**
0.06
W Significance Level of W Model H2 Significance Level of Model N
0.19 .0000 55.37 .0000 2,579
Income Quintile (ref. = top fifth) Fourth fifth Third fifth Second fifth Low fifth
Notes: Tests of the difference in coefficients in the two models show that the effects of the lowest two income quintiles differed for boys and girls (for the lowest income quintile, H2 = 4.22 [df = 1], p = .0398; for the second lowest income quintile, H2 = 4.01 [df = 1], p = .0452). Source: Chinese Household Income Project.
**p < .01
enrollment of sons and daughters across levels of income. Figure 1 depicts a simulation of the difference in family allocation strategies that is associated with changes in income. The three bars represent the simulated set of choices for families at the bottom, middle, and top quintiles of the income distribution. Figure 1 demonstrates that the likelihood that all school-aged children in the household were enrolled in school was positively related to household income. More to the point, at the high-income level, about 22% of the families enrolled all school-aged boys but not all school-aged girls, while the comparable figure was 26% at the middle-income level and about 31% at the low-income level. These findings confirm the importance of income as a determinant of the extent to which parents chose to educate sons at the expense of daughters. Changes in Gender Disparities Under Market Reforms In short, the results indicate that girls’ schooling was particularly susceptible to improvements in economic circumstances and that family decisions about educating children became more gender neutral with greater incomes. These results fall naturally into the household welfare framework and lead to the macro-level question of whether progress toward gender equality accelerated with the transition to a rapid-growth economy in the late 1970s and early 1980s. Evidence from educational data. Figure 2 shows the percentage of the primary and secondary school populations who were female, by year, from 1972 to 1997. Data for primary schooling are missing for 1976–1979. However, the available data suggest a rising trend in the percentage of girls among students at both the primary and secondary levels in the 1970s, in the latter years of the Cultural Revolution. This trend stalled in the late 1970s or early 1980s, in the years around the beginning of market reforms, and then resumed in the mid-1980s. Figure 3 presents the female gross enrollment ratios, expressed as a percentage of the male gross-enrollment ratios, for the primary and secondary levels by year. Gross
Demography, Volume 42-Number 2, May 2005
288
Figure 1.
Simulated Enrollment Choices, by Household Income Quintile 100 90
Percentage Distribution of Boys’ and Girls’ Enrollment
80
14
13
10
11
22
26
14
19
70 60
31
50 40 30
54
50
20
36
10 0
Top
Middle
Bottom
Income Quintile All boys, all girls Not all boys, not all girls
All boys, not all girls Not all boys, all girls
Source: See Table 4.
enrollment ratios are calculated as the number of students, regardless of age, at a given level of schooling in a given school year, divided by the number of children in the official age range for that level of schooling in the same year.15 Data are available for 1971 to 1997, with no years missing. Like Figure 2, the ratios at both primary and secondary levels in Figure 3 show steep increases from the mid-1970s (the latter years of the Cultural Revolution), with a downturn at the time of market reforms and a subsequent resumption of the increasing trend by the mid-1980s. Evidence from census data. These period measures tell a consistent story. An important question, however, is whether the same trends play out in the educational experiences of the population. Figure 4 addresses this question, showing the rates of girls’ entry 15. Because of the lack of correspondence of age groups in the numerator and the denominator of gross enrollment ratios, these measures overestimate actual enrollment rates if children outside official age ranges are enrolled, as happens in many countries, including China. Net enrollment ratios, calculated as the number of students in the official age range divided by the number of children in the official age range, is a preferred measure. However, in China, as in most less-developed countries, data for calculating gross enrollment ratios are much more readily available. Primary net enrollment ratios by sex are available for China from the United Nations Common Database, but only for 1990 and 2000; I was unable to locate secondary net enrollment ratios by sex (United Nations n.d.). According to UNESCO (n.d.b), the official age ranges for primary and secondary schooling in China since 1975 are 7–11 and 12–17, respectively. Official age ranges for 1971–1974 data are not specified.
Market Transition, Education, and Family Strategies in Rural China
Figure 2.
289
Percentage of the Student Populations Who Were Female, by Level of School and Year
50 45
Percentage Female
40 35
95.12
30 25 20 15 10
1990
1985
1980
1975
0
Primary
1995
Secondary
5
Year Sources: Ministry of Education (1984); UNESCO (n.d.a).
Figure 3.
Female Gross Enrollment Ratios as a Percentage of Male Gross Enrollment Ratios, by Level of School and Year
100 90 80
Percentage
70
95.12
60 50 40 30 20
Year
Source: UNESCO (n.d.a).
Primary
1995
1985
1980
1975
1971
0
1990
Secondary
10
Demography, Volume 42-Number 2, May 2005
290
Figure 4.
Female-to-Male Ratios of Primary Entrance and Junior High School Transition, by Birth Year, 1990 and 2000 Census Reports
100
80 70 60 50 40 30
1990
1985
1980
Junior high school, 1990 Junior high school, 2000
1975
1960
1955
0
1950
10
1970
Primary school, 1990 Primary school, 2000
20
1965
Female-to-Male Ratio (Percentage)
90
Birth Year
Sources: All China Marketing Research Co. (2003); Census Office (1993).
into primary school and of transition to junior high school as a percentage of the boys’ rates, calculated from the 1990 and 2000 census reports of educational attainment. For each sex and birth cohort, rates of primary entrance were calculated as the number of people who reported primary or higher education, divided by all people. Rates of junior high school transition were calculated as the number of people who reported junior high school or higher education, divided by the number of people who reported primary or higher education. The rate for girls was then divided by the rate for boys and expressed as a percentage. The years listed along the horizontal axis are approximate birth years, calculated by subtracting ages from the census years. One point that is important to note is that the educational attainment questions differed in 1990 and 2000. Most notably for this figure, in 1990, illiteracy/semiliteracy was a category of the educational attainment variable, but in 2000, literacy was asked as a separate question. In 2000, those who did not claim primary school or better educational attainment had to select “never went to school” or “attended literacy classes.” This inconsistency in wording means that the results for the same cohorts in 1990 and 2000 are not directly comparable.16 As Figure 4 shows, girls’ primary entrance rates as a percentage of boys’ climbed fairly steadily over time. The 1990 data indicate that much of the improvement in gender equality predated the market reforms, with female-to-male ratios reaching 95% by the 16. Those who attended one or two years of primary school might have been more likely to report that they were illiterate/semiliterate in 1990 but to claim primary attainment in 2000. It is likely that women would be disproportionately affected by the changed wording.
Market Transition, Education, and Family Strategies in Rural China
291
1971 birth cohort, who would have entered school in the late 1970s, and hovering at 95% to 96% thereafter. The 2000 data indicate even higher female-to-male ratios, which reached the mid-90% range by the early 1960s birth cohorts, who would have completed primary school well before the onset of market reforms. The junior high school results show rising trends since the early 1950s birth cohorts. For both the 1990 and the 2000 estimates, a temporary peak was reached at about the 1963–1964 birth cohorts, who would have been in junior high school around the era of the market transition, after which birth year the female-to-male ratio moved temporarily in a negative direction. In the 1990 data, the female-to-male ratio reached a low point for the 1971 birth cohort, who should have entered junior high school around 1984, and then began to rise again. In the 2000 data, the low point was reached earlier for the 1967 cohort, who should have been in junior high school in the early 1980s. The divergence in the patterns in the two censuses for about the 1969–1975 cohorts— 15- to 21 year olds in 1990—is puzzling and may be a function of the inconsistent question wording, differences in coverage of the migrant population in the two years, or different measurement errors in the education or age variables.17 However, the earlier recovery of the female-to-male ratios in the 2000 reports than in the 1990 reports suggests the possibility that girls who had progressed slowly through school or dropped out in the early- to mid-1980s may have been catching up in the 1990s as China’s development continued and as major poverty-alleviation initiatives, such as Project Hope and the Spring Bud Program, helped a large number of poor rural children, especially girls, to return to or stay in school in the 1990s (“Charity Returns” 2002; “‘Spring Bud’ Program” 2002). Overall, although the pattern is stronger in the 1990 data than in the 2000 data, the indication that female-to-male ratios of junior high school transition flattened in the early reform period is consistent with the corresponding trends shown in Figures 2 and 3. In summary, there is no convincing evidence at the level of either primary or junior high school that the transition to a market economy initially spurred progress toward gender equity. At the primary level, period measures show downturns in gender-equity measures in the market-transition years, and census data indicate that the vast majority of progress in expanding girls’ access to primary school had emerged for cohorts who were of an age to enter school well before the market reforms. At the transition to junior high school, where costs to families are higher and access is still not universal, both the period measures and the 1990 and 2000 census reports are consistent with stalled progress toward gender equity in girls’ schooling around the time of the market transition and resumed progress thereafter. Gender Differences in the Vulnerability to Costs The lack of consistent evidence that the market transition coincided with accelerated progress toward gender equity contradicts expectations of the modernization perspective about the response of gender disparities to growth. Yet, the individual-level results suggest that higher incomes did promote gender equity, in contradiction to the traditional WID perspective. What could explain the inconsistent micro- and macro-level results? Much of the discussion of rural gender issues in the wake of market reforms has focused on what some have argued to be a disadvantageous set of circumstances brought about by the responsibility system of agriculture and the decentralization of educational finance. Both changes served to raise the costs of schooling, even as incomes rose, and
17. Higher rates of primary entrance and junior high school transition in 2000 than in 1990 were prevalent across the birth cohorts among rural women. Junior high school transition rates were fairly similar until about the 1980 transition cohort, at which point the 2000 estimates diverged and were substantially higher than the 1990 estimates, especially for rural men and women.
292
Demography, Volume 42-Number 2, May 2005
both were superimposed on a family system in which investments in boys were a necessity and investments in girls were a luxury. A logical question, then, is whether the new costs, in the face of increased economic uncertainty about the future and the greater potential long-term returns on the schooling of sons, presented disproportionate barriers to the education of girls. The 1992 data permit a direct investigation of decisions about children’s enrollment. Unlike all other national surveys that were conducted reasonably early in the reform period, the 1992 survey directly asked mothers of out-of-school children the reasons why the children were out of school. Table 5 presents the results of two multinomial logit models predicting school leaving for various reasons. Gender and a constant term are included in the models. The top panel uses all children as the estimation sample, while the bottom panel uses an estimation sample of children who were not attending school. The top panel uses children who were currently enrolled as the omitted category, while the bottom panel uses the category “other reasons” as the base category. Estimated probabilities (expressed as percentages) of each response for girls and boys are presented below the coefficients. The top panel shows positive gender effects for all categories, reflecting girls’ greater propensity to drop out of school. However, among specifically identified reasons, the gender differences were the greatest among the categories “economic difficulties” and “to work on farm.” In both cases, these reasons were more than 2.5 times as likely to be cited for girls as for boys, measured in odds or probabilities. In the bottom panel, out-of-school girls were actually less likely than out-of-school boys to report all reasons for leaving school except these two reasons and the reason of inconvenient transportation. However, inconvenient transportation is substantively less interesting than the two cost-related categories because it was cited for a small proportion of those who had left school (1.79% of out-of-school boys and 2.04% of out-of-school girls).18 Conversely, the two cost-related categories were the most frequently cited, excluding the “other reasons” category. Financial difficulties were cited for an estimated 27% of out-of-school boys and about 37% of out-of-school girls. Work was cited by about 7% of out-of-school boys and about 11% of out-of-school girls. In short, according to the reports of rural mothers, girls appear more vulnerable to family financial difficulty or opportunity costs than boys do.19 Although cross-sectional results cannot directly illuminate the causes of the change in gender disparities in the market-transition years, the direct and opportunity costs incurred by families for educating children in 1992 were a direct consequence of the reform-era policies. A plausible inference is that the new costs disproportionately affected girls.20
18. One concern is that school shutdowns were part of the early modernization drive in China, associated with attempts to upgrade the quality of schools (see Tsui 1997), and thus access problems may have affected a broader group of children before 1992. In many less-developed settings, having to travel long distances to schools has been thought of as exerting a disproportionate negative impact on girls’ educational opportunities. However, in the case of China, evidence is not clear on this issue. In additional analyses using the 1992 children’s data matched to village data, village access to junior high school did not significantly interact with gender in predicting enrollment in models that incorporated village and household socioeconomic and demographic characteristics and income-gender interactions. Similar analyses using a distance-to-school variable indicated that extremely long distances to junior high schools, 5 kilometers or greater, significantly affected the enrollment of children of both genders, but did not disproportionately affect girls (see table 4 and note 42 in Hannum 2003). 19. This interpretation is consistent with recent findings from Indonesia suggesting that households treat girls’ education as a luxury good that is cut when hardship strikes (Cameron and Worswick 2001). 20. This conclusion is consistent with Treiman’s (2001) analysis of a retrospective stratification survey showing that girls who were teenagers during the early reform period were disproportionately likely to leave school for work in agricultural jobs.
4.50
Estimated %, female 2.35
4.50
–1.94** (0.06)
–0.55** (0.08)
0.30
0.31
–5.72** (0.05)
0.03 (0.08)
Handicapped
36.63
27.03
–0.15** (0.03)
0.41** (0.04)
4.61
1.83
–3.93** (0.02)
0.99** (0.03)
Economic Difficulty
p < .10; *p < .05; **p < .01
†
Source: Children’s Survey.
10.62
7.45
–1.44** (0.05)
0.46** (0.06)
1.34
0.51
–5.22** (0.04)
1.04** (0.05)
To Work on Farm
Note: Robust standard errors, shown in parentheses, allow for clustering within households.
7.67
–1.41** (0.04)
Estimated %, male
Intercept
–0.43** (0.06)
0.57
Estimated %, female
School Leavers Only (N = 22,064; ref. = other reasons) Gender (ref. = male)
0.52
–5.19** (0.04)
Estimated %, male
Intercept
0.15** (0.06)
Sick
14.20
18.17
–0.54** (0.03)
–0.15** (0.04)
1.79
1.23
–4.33** (0.03)
0.44** (0.03)
Poor Student
Multinomial Logit Models Predicting Reasons for Leaving School, Rural China, 1992
All Children (N = 231,112; ref. = in school) Gender (ref. = male)
Sample
Table 5.
0.46
0.81
–3.65** (0.13)
–0.46** (0.17)
0.06
0.06
–7.43** (0.12)
0.12 (0.17)
No Teacher
0.90
1.28
–3.20** (0.11)
–0.25† (0.14)
0.11
0.09
–6.98** (0.11)
0.33* (0.14)
2.04
1.79
–2.86** (0.09)
0.23* (0.11)
0.26
0.12
–6.64** (0.09)
0.81** (0.10)
Insufficient Inconvenient School Transportation
28.31
31.32
––
––
3.56
2.13
–3.78** (0.02)
0.58** (0.03)
Other
Market Transition, Education, and Family Strategies in Rural China 293
294
Demography, Volume 42-Number 2, May 2005
DISCUSSION AND CONCLUSIONS Summary China’s experience encompasses elements of both the modernization and WID perspectives. Consistent with the expectations of the modernization perspective, the results confirmed a narrowing of the gender gap with more-favorable economic circumstances, and the time-series data show a secular trend toward equality that is probably related to longterm social and economic development. Yet, consistent with expectations of the WID framework, the data on education indicate that the early reform years brought about a temporary setback in the rapid progress toward gender equity that occurred during the latter, low-growth years of the Cultural Revolution; census data suggest that this setback left a discernable impact at the secondary level. The results suggest that direct and opportunity costs that were associated with the market transition in China had a disproportionate impact on families’ decisions about girls’ schooling. Thus, in China, the new costs of schooling that were associated with market reforms likely exacerbated gender disparities, even under the favorable circumstance that the reforms stimulated rapid economic growth, and girls’ schooling demonstrably benefited from improved economic circumstances.21 The key to understanding the immediate impact of the market transition in China may lie in the intersection of family survival strategies and the new uncertainties and costs that were introduced by market reforms, irrespective of the longer-term implications of the market transition for growth. In other words, the policies adopted under a growth-promoting agenda must be considered, as well as the growth itself, as an input to family decisions. In the short term, families responded to new costs with a preference for sons that was dictated by traditional strategies for ensuring long-term family welfare. In the long term, however, socioeconomic development that was enabled by the market transition is likely to erode the very family structures and labor market conditions in China that made those strategies functional. Trend data suggest that this process had already begun in China after the early 1980s; census data from 1990 and 2000 also suggest a rapidly diminishing gender gap. The Role of Fertility Decline An important question to consider is whether China’s rapid drop in fertility played a role in the results presented here. China’s national fertility-control policies started with the wan-xi-shao (later-longer-fewer) policy, dating from 1970, and continued with the onechild policy starting in 1979 (Hannum and Liu forthcoming; Information Office 1995, 2000). Fertility fell dramatically under the wan-xi-shao policy, from a total fertility rate of about 5.78 births per woman in 1970 to 2.6 by 1978. Under the one-child policy, fertility rates fluctuated during the 1980s and did not drop below 2 until 1992 (see Appendix 1). Fertility remains higher in rural than in urban areas because of exemptions and local modifications that were increasingly allowed throughout the 1980s and 1990s (Hannum and Liu forthcoming). Declining fertility, and especially the one-child policy, would be expected to ease the competition for resources within households and thus to reduce the educational gender gap. However, the timing of the fertility decline in China means that fertility cannot explain the trends in gender inequality that emerged in the immediate wake of the market reforms. All the children who reached junior high school age prior to the market-reform period and most of the children who reached primary school age prior to this period were
21. It is significant that rural China experienced the social shocks of reforms under favorable conditions: the reforms successfully promoted unprecedented economic growth and poverty reduction in the early 1980s (Piazza and Liang 1998).
Market Transition, Education, and Family Strategies in Rural China
295
born before the fertility transition.22 Children born during the drop in fertility entered school during the late 1970s and 1980s, when the trend in gender inequality was mixed. However, the low levels of fertility that were achieved by the time of the late 1970s birth cohorts, who would be attending school in the late 1980s and 1990s, suggest that low fertility was conducive to progress toward educational gender equity after the early transition period. Low fertility rates since the early 1990s are likely to help sustain the low levels of preference for sons in education that were observed in the most recent data.23 Implications Forsythe et al. (2000: note 1) observed that “there are emerging reservations about the utility and advisability of generalizing at all about the relationship between women and economic development.” The results of this study suggest two specific reasons for this caveat as it applies to educational gender stratification. First, in the short term, differences across and within countries in modes of family organization and the gendered division of labor produce different incentives for educating girls and boys and have direct implications for families’ strategies for survival under duress.24 In rural China, in the early years of the market transition, the incentives played out in a way that disadvantaged girls, but in other settings or at other times, incentives may play out in a different fashion.25 Second, and related to the first, different factors come into play when one considers the short- and long-term implications of development. In the short run, the particular policies that are associated with growth-promoting reforms, specifically the policies’ implications for educational costs as filtered through traditional family survival strategies, probably matter more than does the potential for rising incomes as influences on the gender gap in educational decisions. In contrast, in the long run, rising incomes may produce the “rationalization” of incentives for investments that are expected under the modernization framework.
22. The argument here is stronger for junior high school students than for primary school students. However, cross-cohort analyses using data from a 1985 fertility survey in Hebei, Shaanxi, and Shanghai found little evidence to support the idea that fertility trends explained trends in gender inequality in the entrance to primary school because gender disparities were not significantly affected by sibship size (Hannum and Xie 1994). 23. Both the wan-xi-shao policy and the one-child policy were explicitly framed as means of addressing the impact of unhindered population growth on economic development and modernization. Thus, fertility control in China could be viewed as another example of a growth-promoting policy with immediate negative implications for girls, in the form of rising sex ratios at birth and a large number of “missing” girls (Croll 2001), but potentially more-favorable long-term consequences, since reduced fertility and one-child families exert a longerterm equalizing effect on educational investments in boys and girls. 24. The degree of duress under which families make educational decisions in China increasingly varies. Interprovincial income inequality increased markedly from the late 1980s at least until 2000, and the urbanrural gap in income and living standards remains large (Carter 1997; Khan and Riskin 1998; Zhang and Kanbur 2003). One reader suggested testing the notion that regional variation within China in women’s opportunities in the labor force may affect educational gender disparities. To test this idea, I used census tabulations from the 10% sample of the 1990 census (available in State Council Population Census Office 1991) and the Children’s Survey data set. I calculated correlations between the provincial percentage female in the nonfarming labor force and male enrollment rates (0.43), female enrollment rates (0.53), and female-male odds ratios of enrollment (0.51). These moderate, positive correlations are consistent with the idea that locales with more opportunities for women’s nonfarm labor also have greater gender equality (as suggested by the odds ratios), as well as greater investments in education overall (as suggested by the enrollment rates). Of course, it is unlikely that such correlations represent the exclusive impact of women’s participation in the nonfarm workforce, given the likely relationships between women’s participation in that workforce and other indicators of economic development. However, this finding is consistent with the argument that egalitarian labor market opportunities tend to accompany egalitarian educational opportunities. 25. For example, in Mongolia, the market transition brought about economic declines, along with increases in the costs of schooling. Here, a male educational disadvantage emerged: because families depended economically on herding done by boys, boys rather than girls left school to work (Bray 1996; Griffin 1995; Smith 1995; Smith and Lannert 1995). Secondary gross enrollment ratios for boys dropped from 86.8% in 1986 to 47.5% in 1996. For girls, the comparable figures were 94.2% and 65.1% (UNESCO n.d.a).
Demography, Volume 42-Number 2, May 2005
296
Appendix 1.
Estimates of Total Fertility Rates, by Source and Year
Total Fertility Rate (per Woman)
7 World Development Indicators (WDI)
6
U.S. Census Bureau International Database (IDB) 5 4 3 2
2000
1995
1990
1985
1980
1975
0
1970
1
Year
Sources: U.S. Census Bureau (n.d.); World Bank (n.d.).
REFERENCES All China Marketing Research Co. 2003. “4-1 Population Aged 6 and Over by Age, Sex and Education Attainment.” China 2000 Population Census Data Assembly [Electronic version]. Beijing: All China Marketing Research Co. Assié-Lumumba, N’dri Thérèse. 2000. “Educational and Economic Reforms, Gender Equity, and Access to Schooling in Africa.” International Journal of Comparative Sociology 41:89–120. Boserup, E. 1989 [1970]. Woman’s Role in Economic Development. London: Earthscan. Bray, M. 1996. Counting the Full Cost: Parental and Community Financing of Education in East Asia. Washington, DC: World Bank. Brown, P. and A. Park. 2002. “Education and Poverty in Rural China.” Economics of Education Review 21:523–41. Buchmann, C. 1996. “The Debt Crisis, Structural Adjustment and Women’s Education: Implications for Status and Social Development.” International Journal of Comparative Sociology 37:5–30. Buchmann, C. and E. Hannum. 2001. “Education and Stratification in Developing Countries: A Review of Theories and Empirical Research.” Annual Review of Sociology 27:77–102. Cameron, L. and C. Worswick. 2001. “Education Expenditure Responses to Crop Loss in Indonesia: A Gender Bias.” Economic Development and Cultural Change 49:351–63. Carter, C.A. 1997. “The Urban-Rural Income Gap in China: Implications for the Global Food Market.” American Journal of Agricultural Economics 79:1410–18. Census Office Under the State Council and the Department of Population Statistics, National Statistical Bureau. 1993. “T4-1. Population By Age and Sex, 1990,” “T5-5. Population Aged 6 and
Market Transition, Education, and Family Strategies in Rural China
297
Over by Educational Level, Educational Attainment Status and Age.” Tabulation on the National Population Census of China (4 Volumes) [Electronic version]. Beijing: China Statistical Publishing House. “Charity Returns Over 2.5 Million Children to School.” 2002. People’s Daily. Available on-line at http://english.people.com.cn/200212/21/eng20021221_108838.shtml Cheng, K.-M. 1996. The Quality of Primary Education: A Case Study of Zhejiang Province, China. Paris: International Institute for Educational Planning. Connelly, R. and Z. Zheng. 2003. “Determinants of School Enrollment and Completion of 10 to 18 Year Olds in China.” Economics of Education Review 22:379–88. Croll, E. 2001. Endangered Daughters: Discrimination and Development in Asia. New York: Routledge. Das Gupta, M. 1987. “Selective Discrimination Against Female Children in Rural Punjab, India.” Population and Development Review 13:77–100. Davis, D. 1989. “Chinese Social Welfare: Policies and Outcomes.” China Quarterly 119:577–97. Eichen, M. and M. Zhang. 1993. “Annex: The 1988 Household Sample Survey—Data Description and Availability.” Pp. 341–46 in The Distribution of Income in China, edited by K. Griffin and R. Zhao. New York: St. Martin’s Press. Entwisle, B., G.E. Henderson, and S.E. Short. 1995. “Gender and Family Businesses in Rural China.” American Sociological Review 60:36–57. Filmer, D. 1999. “The Structure of Social Disparities in Education: Gender and Wealth.” Gender and Development Working Paper Series, No. 5. World Bank, Washington, DC. Forsythe, N., R.P. Korzeniewicz, and V. Durrant. 2000. “Gender Inequalities and Economic Growth: A Longitudinal Evaluation.” Economic Development and Cultural Change 48:573–618. Goodkind, D. 1995. “Rising Gender Inequality in Vietnam Since Reunification.” Pacific Affairs 68:342–59. Greenhalgh, S. 1994. “Controlling Births and Bodies in Village China.” American Ethnologist 21(1):3–30. Griffin, K. 1995. “Poverty: Concepts and Measurement.” Pp. 27–44 in Poverty and the Transition to a Market Economy in Mongolia, edited by K. Griffin. New York: St. Martin’s Press. Hannum, E. 1999. “Political Change and the Urban-Rural Gap in Education in China, 1949–1990.” Comparative Education Review 43:193–211. ———. 2003. “Poverty and Basic Education in Rural China: Villages, Households, and Girls’ and Boys’ Enrollment.” Comparative Education Review 47:141–59. Hannum, E. and P. Kong. 2004. “Family Sources of Educational Gender Inequality in Rural China: A Critical Assessment.” Unpublished manuscript. Department of Sociology, University of Pennsylvania, Philadelphia. Hannum, E. and J. Liu. Forthcoming. “Adolescent Transitions to Adulthood in China.” In Studies on the Transitions to Adulthood in Developing Countries, edited by J. Behrman, C. Lloyd, N. Stromquist, and B. Cohen. Washington, DC: National Academies of Science Press. Hannum, E. and Y. Xie. 1994. “Trends in Educational Gender Inequality in China: 1949–1985.” Research in Social Stratification and Mobility 13:73–98. Hardin, J.W. 1996. “Bivariate Probit Models.” Stata Technical Bulletin 33(16):15–20. Information Office of the State Council of the People’s Republic of China. 1995. “Family Planning in China.” Government White Paper, August 1995, Beijing. English translation available online at http://www.china.org.cn/e-white ———. 2000. “China’s Population and Development in the 21st Century.” Government White Paper, December 2000, Beijing. English translation available on-line at http://www.china.org.cn/ e-white Jacka, T. 1997. Women’s Work in Rural China: Change and Continuity in an Era of Reform. Cambridge, England: Cambridge University Press. Judd, E.R. 1994. Gender and Power in Rural North China. Stanford, CA: Stanford University Press.
298
Demography, Volume 42-Number 2, May 2005
Khan, A.R. and C. Riskin. 1998. “Income Inequality in China: Composition, Distribution and Growth of Household Income, 1988 to 1995.” China Quarterly 154:221–53. King, E. and M.A. Hill. 1993. Women’s Education in Developing Countries. Baltimore, MD: John’s Hopkins University Press. Knight, J. and S. Li. 1993. “The Determinants of Educational Attainment in China.” Pp. 285–323 in The Distribution of Income in China, edited by K. Griffin and R. Zhao. New York: St. Martin’s Press. Knodel, J. and G.W. Jones. 1996. “Post-Cairo Population Policy: Does Promoting Girls’ Schooling Miss the Mark?” Population and Development Review 22:683–702. Lantican, C.P., C.H. Gladwin, and J.L. Seale, Jr. 1996. “Income and Gender Inequalities in Asia: Testing Alternative Theories of Development.” Economic Development and Cultural Change 44(January):235–63. Lavely, W., Z. Xiao, B. Li, and R. Freedman. 1990. “The Rise in Female Education in China: National and Regional Patterns.” China Quarterly 121:61–93. Lewin, K. and Y. Wang. 1994. Implementing Basic Education in China: Progress and Prospects in Rich, Poor and National Minority Areas. IIEP Research Report, No. 101. International Institute for Educational Planning, Paris. Lin, J. 1993. Education in Post-Mao China. Westport, CT: Praeger. Ministry of Education Department of Planning. 1984. Achievement of Education in China: Statistics 1949–1983 [in Chinese]. Beijing: People’s Education Press. Papanek, H. 1985. “Class and Gender in Education-Employment Linkages.” Comparative Education Review 29:317–46. Park, A., W. Li, and S. Wang. 2002. “School Equity in Rural China.” Paper presented at the Association of Asian Studies, April 4–8, Washington, DC. Piazza, A. and E.H. Liang. 1998. “Reducing Absolute Poverty in China: Current Status and Issues.” Journal of International Affairs 52:253–64. Rama, M. 2001. “The Gender Implications of Public Sector Downsizing: The Reform Program of Vietnam.” Gender and Development Working Paper Series, No. 19. World Bank, Washington, DC. Schultz, T.P. 1995. “Investments in the Schooling and Health of Women and Men: Quantities and Returns.” Pp. 15–50 in Investments in Women’s Human Capital, edited by T.P. Schultz. Chicago: University of Chicago Press. Shavit, Y. and H.-P. Blossfeld. 1993. Persistent Inequality: Changing Educational Attainment in Thirteen Countries. Boulder: Westview Press. Smith, S. 1995. “Human Capital Formation Under Conditions of Acute Resource Scarcity.” Pp. 63–76 in Poverty and the Transition to a Market Economy in Mongolia, edited K. Griffin. New York: St. Martin’s Pres. Smith, S. and J. Lannert. 1995. “Human Capital: the Health and Well-being of the Population.” Pp. 77–89 in Poverty and the Transition to a Market Economy in Mongolia, edited by K. Griffin. New York: St. Martin’s Press. “‘Spring Bud’ Program Helps 1.15 Million Girl Dropouts.” 2002. Xinhua News Agency June 1. Available on-line at http://www.china.org.cn/english/SO-e/33668.htm Stash, S. and E. Hannum. 2001. “Who Goes to School? Educational Stratification by Gender, Caste and Ethnicity in Nepal.” Comparative Education Review 45:354–78. State Council Population Census Office, and State Statistical Bureau Department of Population Statistics, eds. 1991. Ten Percent Sampling Tabulation on the 1990 Population Census of the People’s Republic of China (Computer Tabulation) [in Chinese]. Beijing: China Statistical Publishing House. State Statistical Bureau. 1992. Work Handbook for the 1992 National Sample Survey on the Situation of China’s Children [in Chinese]. Beijing: State Statistical Bureau. ———. 1993. 1992 National Sample Survey on the Situation of Children: National Final Report [in Chinese]. Beijing: State Statistical Bureau.
Market Transition, Education, and Family Strategies in Rural China
299
Stockman, N. 1994. “Gender Inequality and Social Structure in Urban China.” Sociology 28: 759–77. Stromquist, N.P. 1989. “Determinants of Educational Participation and Achievement of Women in the Third World: A Review of the Evidence and a Theoretical Critique.” Review of Educational Research 59:143–83. Summerfield, G. 1994. “Economic Reform and the Employment of Chinese Women.” Journal of Economic Issues 28:715–32. Treiman, D.J. 1970. “Industrialization and Social Stratification.” Pp. 207–34 in Social Stratification: Research and Theory for the 1970s, edited by E.O. Laumann. Indianapolis, IN: BobbsMerrill. ———. 2001. “Trends in Gender Inequality in China.” Paper presented at the Harvard Conference on Education Reform in China, July 14–15, Cambridge, MA. Tsang, M.C. 1994. “Costs of Education in China: Issues of Resource Mobilization, Equality, Equity, and Efficiency.” Education Economics 2:287–312. Tsui, K.-Y. 1997. “Economic Reform and Attainment in Basic Education in China.” China Quarterly 149:104–27. Tsui, M. and L. Rich 2002. “The Only Child and Educational Opportunity for Girls in Urban China.” Gender and Society 16(1):74–92. UNESCO. n.d.a. UNESCO Institute for Statistics World Education Indicators. Available on-line at http://unescostat.unesco.org/en/stats/stats0.htm ———. n.d.b. UNESCO Institute for Statistics Database, National Education Systems, Country: China. Available on-line at http://www.uis.unesco.org/pagesen/DBSysCri.asp United Nations. n.d. United Nations Common Database. “Education Enrolment Ratio, Net, Primary Level, by Sex (UNESCO)” [code9954]. Available on-line at http://unstats.un.org/unsd/ cdb/cdb_series_xrxx.asp?series_code=29954 U.S. Census Bureau. n.d. International Data Base, Table 028. Age-Specific Fertility Rates and Selected Derived Measures, China 1990–2000. Washington, DC: U.S. Bureau of the Census. Available on-line at http://www.census.gov/ipc/www/idbacc.html World Bank. 1992. China: Strategies for Reducing Poverty in the 1990s. Washington, DC: World Bank. ———. n.d. World Development Indicators (WDI) On-Line, Total Fertility Rate (Births per Woman), China, 1960–2000. Available on-line at http://devdata.worldbank.org/dataonline Zellner, A. and T.H. Lee. 1965. “Joint Estimation of Relationships Involving Discrete Random Variables.” Econometrica 33:382–94. Zhang, X. and R. Kanbur. 2003. “Spatial Inequality in Education and Healthcare in China.” Working Paper 2003-35. Department of Applied Economic and Management, Cornell University, Ithaca, NY. Zheng, Z., R. Niu, and L. Xing. 2002. “Determinants of Primary and Middle School Enrollment of 10–18 Year Olds in China” [in Chinese]. Population and Economics 131(2):28–37.
