Do peers matter? Evidence from the sixth grade experiment

The existence and magnitude of peer effects is debated throughout the economics, sociology, and psychology literatures. Intuitively, the influence of peer behavior on individual behavior makes sense, but estimating this relationship has proven difficult. I use two approaches to estimate peer effects on the adolescent decision to smoke, drink, or use drugs. The first approach utilizes variation in the type of school a child attends in sixth grade as an instrument for peer behavior. This approach is based on the concept that the type of school a child attends, middle school or elementary school, determines the type of peer behavior he or she observes. The second approach consists of estimating a model of adolescent behavior that includes family fixed effects using samples of siblings and twins.; Between 1986 and 2002, the percentage of sixth grade students who attended middle school rose from 32% to 55%. Sixth grade students who are placed in a middle school, as opposed to an elementary school, are exposed to older peers and potentially different peer behavior. Using this variation in school type as an instrument for peer behavior I find statistically and economically significant peer effects on adolescent smoking, drinking, and drug use. I also find that attending a middle school significantly increase peer smoking, drinking, and drug use rates. Using samples of siblings and twins to control for family influences on behavior I find significant peer effect estimates that are significantly smaller than those obtained using instrumental variables analysis.; I further investigate the effects of the type of school a child attends in sixth grade by analyzing state level data on smoking and drinking rates. I find mixed results for the effect of school type on state smoking and drinking rates. When the effect is significant and signed as expected the magnitude is modest. Using difference-in-difference analysis to compare states that experience a large increase in the percentage of sixth graders who attend school with older students to states that did not experience this increase yields insignificant results for the effect of school type on smoking and drinking rates.; States located in the south have historically placed a significantly larger percentage of their sixth grade students in schools that contain older adolescents. Smoking is also generally believed to be a larger issue in the south than in other regions of the countries with states like Kentucky, West Virginia, and Tennessee having some of the highest smoking rates in the country. I attempt to determine if influences on the adolescent decision making process are different in the south than they are in other census regions. I find that peer effect estimates are smaller in the south than in other regions. I also find that placing sixth grade students in schools with older students increases state smoking and drinking rates in non-south regions, but has the opposite effect in the south.