EXTENDED FAMILY INFLUENCE ON INDIVIDUAL MIGRATION DECISION IN RURAL CHINA Hao DONG, Yu XIE Princeton University INTRODUCTION This study aims to understand whether and how extended family members influence individual migration decision in rural China. Taking advantage of the recently available longitudinal survey data of the China Family Panel Studies (CFPS) 2010, 2012 and 2014, we focus on the influence of both co-resident and non-co-resident parents and siblings, in addition to own and nuclear family characteristics, on individual migration. We not only examine the effects of their concurrent health, education, and geographical proximity, but also employ a withinfamily comparison approach to identify individual characteristics differentiating the migration decision and distance between siblings a strategic division of labor within the extended family. This perspective of extended family, distinct from many previous studies, is important to understand the massive internal migration in China. Existing research on the decision-making of China s rural migrants often adopts analytical frameworks of Western and international labor migration and emphasizes the effects of characteristics of individuals and their nuclear families. However, as opposed to an individualistic simple (nuclear and stem) family tradition in many Western populations, there has been a long tradition of collective extended (joint) family in China (Ruggles 2010). Chinese extended families have the controlling yet cooperative nature to differentiate behavioral outcomes between siblings. This has been better understood in studies, for example, on educational sibling differentials (Chu et al., 2007), but less so in migration studies. Indeed, empirical evidence in China increasingly suggests that extended family matters to individual migration decision-making. Several recent studies confirm the importance to take account of residential household context, usually measured by the dependency ratio or number of 1
children and elderly in the household (e.g. Lu and Qin 2014). But due to the limitation of most household surveys in China, few are able to study the influence of non-co-resident parents and other kin. One exception is the study by Giles and Mu (2007), which make use of two longitudinal surveys, one for information of total number of siblings regardless of co-residence, the other for detailed health and socioeconomic information of household members. Their findings highlight the roles played by both non-co-resident and co-resident kin. But their data remain limited to provide accurate information of parental health and to distinguish co-residence of specific siblings, leaving alone the geographical proximity and other socioeconomic status of specific non-co-resident kin. From a different perspective on living arrangements of the elderly, Ma and Wen (2016) also provide evidence that who stays with parents and who lives apart is a collective strategy of the extended family based on comparative advantage. While we do not list our detailed research hypotheses and relevant rationales in this abstract, we have several expectations of the extended family influence for empirical examination. Co-resident parents, especially those of bad health, may decrease the probability of migration. But, the influence of having migrant siblings is likely to be mixed: on one hand, it could encourage migration because of the chained migration effect given advantageous access to information and resources for migration; on the other hand, it could discourage migration due to internal division of labor among siblings to balance the needs of economic well-being and elderly support for the extended family. As a result, we need to further compare between siblings from the same family to identify important individual characteristics shaping such familial division of labor. Considering Chinese familial norms of the differential obligations between boys and girls, and between elder and younger siblings, we particularly expect gender and birth order to shape chances and distance of migration between siblings. In addition, according to the comparative advantages and costs among siblings, individual education and health may positively correlate with chances and distance of migration, but being married and parenting may have negative impact. DATA China Family Panel Studies (CFPS) is nationally representative stratified sampling longitudinal survey on Chinese families, the largest and most comprehensive by far. Starting 2
from 2010, it follows 42590 individuals (33600 adults) from 14960 households, representing 95 percent of the national population from 25 provinces. Three waves of the CFPS data available in 2010, 2012, and 2014. For detailed introduction and evaluation of the CFPS, see Xie and Hu (2014). The CFPS, in contrast with most household surveys in China, collects detailed information of not only those who live in the sampled household, but also their close family members (parents, siblings, and children) outside the household. 1 This allows us for the first time to put individual migration decision-making into the context of extended family. Moreover, the CFPS also specifies the geographical proximity of each parent, sibling and child. In addition to the standard binary outcome of migration or not, we are also able to study the distance of migration between siblings. It provides an opportunity to further understand the collective decision made by siblings. ANALYTICAL STRATEGY Our analysis has two steps. We first examine extended family influence on the chances of becoming a new migrant, as identified between the three CFPS waves. Considering the data structure of biennially repeated surveys, we apply event-history analysis via logistic regressions, with standard errors adjusted for the clustering at extended family level. We focus on those adults (aged 15-55) who are currently living in the rural area and followed in the next wave of survey. The outcome variable is a dummy variable indicating whether an individual becomes a new migrant outside of the county in the next 2 years. As for key explanatory variables, we have two categorical variables for parental health and survival: dead, unhealthy, and healthy for mother and father, respectively. We also include a set of categorical variables for the coresidence of parents as well as geographical proximity of siblings. Other control variables include age and its squared term, gender, marital status, education, self-reported health, household dependency ratio of children and of the elderly, household socioeconomic status, and community migration rates, and fixed effects of provinces for other unobserved heterogeneity. 1 There may still be rare cases that some non-co-resident siblings of some members in the initially sampled household are not included in the survey. And in practice, the follow-up rate between survey waves declines in particularly among those later migrants. This means the within-in family comparison for some individuals may not include a complete set of their non-co-resident siblings. Nevertheless, the CFPS is by far the best available source for our proposed kind of study. 3
We then study factors differentiating the migration distance between siblings with a within-family comparison approach. Namely, we introduce fixed effects of common father into our model estimation, which allows us to take account of unobserved time-invariant difference between families. In other words, we only compare siblings within each family. Unlike our first set of analysis that only studies new migrants and takes geographical proximity of other siblings as given, the current analysis further disentangles important factors shaping sibling differences in migration distance (including those stayers). We again employ the discrete-time event-history approach. But, considering the outcome variable is an ordinal variable indicating whether an individual lives in the same place as registered, or migrates within the same county, city, province, or country, we apply fixed-effect (or in other words, conditional) ordinal logit regressions. In addition to a similar set of right-hand variables included in the first step of analysis, we include birth order as a key explanatory variable, and study its interaction with individual gender. This is because we would like to additionally test if the traditional obligation for the eldest brother to take care of the parents persist in making him staying rather than migrating. Moreover, as every sibling is now the subject of analysis, we include their education, occupation, and self-reported health, and nuclear family characteristics into our within-family comparison to identify the effects of their comparative advantage and cost in making the collective decision on who moves and who stays. As the project just starts in this later this month (Sept. 2016), empirical results are not available to report in this abstract. However, we have carefully examined the capability of the CFPS for the proposed research, and believe preliminary results and further refinements of our analysis could be done well before the PAA 2017. REFERENCES Chu, C. C., Xie, Y., & Yu, R. R. (2007). Effects of sibship structure revisited: Evidence from intrafamily resource transfer in Taiwan. Sociology of Education, 80(2), 91-113. Giles, J., & Mu, R. (2007). Elderly parent health and the migration decisions of adult children: Evidence from rural China. Demography, 44(2), 265-288. 4
Lu, Y., & Qin, L. (2014). Healthy migrant and salmon bias hypotheses: A study of health and internal migration in China. Social Science & Medicine,102, 41-48. Ma, S., & Wen, F. (2016). Who Coresides With Parents? An Analysis Based on Sibling Comparative Advantage. Demography, 53(3), 623-647. Ruggles, S. (2010). Stem families and joint families in comparative historical perspective. Population and Development Review, 36(3), 563-577. Xie, Y., & Hu, J. (2014). An introduction to the China family panel studies (CFPS). Chinese Sociological Review, 47(1), 3-29. 5