اثرات همسالان در وزن بدن نوجوان: شواهدی از مناطق روستایی چین

Peer effect is a potential determinant of individual weight gain that has drawn considerable attention recently. The presence of peer effect implies that policies targeted at changing bodyweight can have enhanced effectiveness through a multiplier effect. This study aims to measure the peer effects on adolescent bodyweight in China. Using the small community nature of the rural sample of the wave 2000 of the China Health and Nutrition Survey, we define plausible peer groups and assess the effect of the average BMI of his/her peer group on the BMI of an adolescent. An instrumental variable (IV) approach is applied to control for potential endogeneity of the peer group's BMI. We find evidence supporting peer effect on BMI in general. The peer effect is around 0.3 with slight variation between two alternative peer definitions. Split sample analysis shows that the peer effect is significant for females (0.32–0.37), and insignificant for male adolescents. Furthermore, we find strong influence of same-gender peers (0.34–0.42) for female adolescents. Conditional quantile regressions show that the peer effect in weight gain is mainly present at or below the median in the conditional BMI distribution for girls, and at the higher end of the BMI distribution for boys. Multiple tests show strong identification, and strong instruments in our IV estimation. Placebo tests suggest that our results are reasonably robust to the correlated effect, due to unobserved community- and province-level factors.

The recent rise in rates of overweight and obesity in adolescents in many developed countries has drawn considerable attention from the public health community. Given that adolescent obesity can lead to persistent overweight or obesity in adulthood, and the known association between obesity and other long-term health problems such as cardiovascular disease, diabetes, etc., it is crucial to identify contributing factors for health policy considerations. Weight gain, however, is not only a ‘rich country’ phenomenon. In developing economies with improved nutrition, sanitation, and economic conditions, weight gain lifts the population from a state of underweight to normal weight, before the concern of overweight steps into the spotlight. The weight gain in the former case has positive implications for the population health and the long-term human capital accumulation of these countries. Understanding the causes of weight gain at the lower end of the body weight distribution is also of equal importance. It is commonly agreed that weight gain is associated with increased consumption of calories and lack of exercise. It is much less clear about the causes of these behaviors associated with weight gain. It has been suggested that parental influence, food prices, access to fast food, environment, opportunity for physical activities, and school nutrition policies have important roles. (Kaestner & Xu, 2006; Koplan, Liverman, & Kraak, 2005; Powell, Auld, Chaloupka, O'Malley, & Johnston, 2007) One determinant of individual weight gain that has drawn more attention recently is peer influence. Influence from peers may come in various forms. The bodyweight of peers may affect a person psychosocially by changing his/her norms about acceptable weight (For a theoretical model, see Burke and Heiland (2007)). The peer group may also influence a person through affecting weight-related behavior such as food consumption and exercise. The presence of the peer effect implies that a multiplier effect exists, and so policies targeting bodyweight have the potential for greater influence than intended (Christakis & Fowler, 2007; Glaeser et al., 2002). This study aims to assess the peer effects on adolescent bodyweight in China. Two definitions of peer group are considered – one based on the age range for the same level of school, the other based on the age within an interval of ±2 years from the age of the adolescent. Using the small community nature of a rural sample in the Wave 2000 of the China Health and Nutrition Study (CHNS), we form plausible empirical peer groups, and estimate the effect of the average BMI of the peer group on the BMI of an adolescent. To control for the potential endogeneity of the peer group's BMI, an instrumental variable approach is applied. We consider as instruments the average BMIs of peers' mothers and fathers, and the proportion of peers' parents born during the years of the Great Famine (1959–1961). Our econometric specification also controls for correlated effects using community-level covariates. Possible correlated effects due to unobserved community-level factors are addressed using several falsification tests. We find evidence supporting the presence of peer effect in general. When peers are defined based on the age range for the same level of school (Peer Definition 1), the peer effect on the BMI is significant for males (around 0.3), but not for female adolescents. When the peer is defined based on the interval of ±2 years from the age of the adolescent (Peer Definition 2), the peer influence appears significant for females (0.32–0.37). Furthermore, we find that the influence of same-gender peers is stronger (0.34–0.42) and significant for female adolescents. We use conditional quantile regression to show that the peer effect in weight gain is mainly present at or below the median in the conditional BMI distribution for girls, and at the higher end of the BMI distribution for boys. Test results show strong identification in our instrumental variable estimation. Placebo tests suggest that our results are robust to the correlated effect due to unobserved community- and province-level factors under Peer Definition 1, but less so under Peer Definition 2. The rest of the paper is arranged as follows. Section 2 provides an overview of the trends in underweight and overweight in China in recent years. Section 3 reviews the literature on peer effect in weight gain. Section 4 discusses the data and methodology. Section 5 presents the results. Section 6 presents conclusions on the work.