Working Paper: NBER ID: w23719
Authors: John Cawley; Euna Han; Jiyoon June Kim; Edward C. Norton
Abstract: Estimating peer effects is notoriously difficult because of the reflection problem and the endogeneity of peer group formation. This paper tests for peer effects in obesity in a novel way that addresses these challenges. It addresses the reflection problem by using the alter’s genetic risk score for obesity, which is a significant predictor of obesity, is determined prior to birth, and cannot be affected by the behavior of others. It addresses the endogeneity of peer group formation by examining peers who are not self-selected: full siblings. Using data from the National Longitudinal Survey of Adolescent Health, we find evidence of positive peer effects in weight and obesity; having a sibling with a high genetic predisposition raises one’s risk of obesity, even controlling for one’s own genetic predisposition to obesity. Implications of the findings include that peer effects may be an explanation for continued worldwide increases in weight, and that, because of social multipliers, the cost-effectiveness of obesity treatment and prevention programs may have been underestimated.
Keywords: peer effects; obesity; genetic data; sibling influence
JEL Codes: D1; I1; I12; I18; J1; Z18
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Alter's Genetic Risk Score (C29) | Ego's BMI (D91) |
| Alter's Genetic Risk Score (C29) | Ego's Probability of Being Obese (D91) |
| Alter's Genetic Risk Score (C29) | Ego's Weight Outcomes (D91) |
| Ego's Age Relative to Alter (D91) | Peer Effects in Obesity (C92) |
| Nonwhite Siblings (J15) | Greater Peer Effects (C92) |
| Sibling's Weight Influence (C92) | Ego's Weight (D91) |