| Behavioral scientists are aware of the strong and persistent correlation between peer and individual behavior. Evidence suggests selection and socialization effects explain the correlation, but the processes, or the details of the ways in which these effects operate in an empirical model, remain relatively unexamined.;This dissertation uses data from waves 3-5 of the National Youth Survey and waves 4-6 of the Rochester Youth Development Study, and finds that the strength of peer influence depends on the bond to peers. The effect of peers increases with the intensity of the bonds to peers. Like any causal mechanism, the effect of peers fades over time, and detecting details of a lagged process becomes difficult. The effect of previous peer associations does not always depend on the bond to peers, and even an additive effect is not always demonstrated. However, the current effect of peers is consistently demonstrated, and evidence for the peer-bond dependency hypothesis, which is consistent with a peer influencing process, is produced. Selection effects, or the notion that delinquent peer associations are a function of delinquency, are also substantiated. However, evidence for empirical complexities consistent with a delinquent peer selection process, such as whether the effect of delinquency on associating with delinquents varies with previous delinquent peer associations, is rather limited and ultimately additive in nature.;Since conclusions vary depending on the estimation strategy, especially when concerning statistical interactions and censored outcomes, the dissertation discusses the limitations of usual methods including Ordinary Least Squares (OLS) and Tobit regression. Censored least absolute deviations (CLAD) models provide an alternative, but results from logistic regression models are emphasized. |
