Social interaction in risky behavior by college student in Japan

Presenter: Sachiko Shimizu, Osaka University

Abstract

The prevalence of teenaged smoking, drinking, and excessive diet have increased over the last 20 years in Japan. Although numerous literatures in sociology and psychology suggest that the influence of neighborhood on behavior of disadvantaged youths may be important in recent years, there are controversial evidences in econometric research. For health policy, the potential existence and magnitude of peer effect is of interest, since peer effects may dominate the effects of policy interventions. The purpose of the study is to assessthe peer effect on risk behavior, and to specify the individual health behavior under social interaction.

The primary contribution of this study is in its design, which allows for a more accurate assessment of peer group. Previous studies have set the peer effect on only class-level effect. We can identify the peer effect in three ways; individual's perception of class-level peers behavior, actual class-level peers behavior, and three best friend actual behaviors
The secondary contribution is to consider individual risk preference in a lottery-choice experiment inspired by Holt and Laury (2002). The experiment presents subjects with a set of lottery choices, which allows us to measure risk aversion. The behaviors measured in our survey represent some of the most significant and well-known behavioral influences on health status; smoking, drinking, and excessive diet, and risky sexual behavior.

Besides information on risky behaviors, risk preference and background characteristics for all individuals between 18-22 years old in Japanese college students, the data reveals which school, class and reference group each individual participates in. Compared to national averages, these risky behaviors are less common in our sample because of higher education and a high proportion of student of the department of health sciece.

Instrumental variables regression, fixed effects-estimation, and multi level multi variate analysis were utilized. In all specification, significant peer effects were found for drinking, smoking, and dieting. The impact of peer behavior is larger among females in dieting. Moreover, our results reveal a positive relationship between preference for risk and peer effect.