Author:
André Ernst | University of Cologne | Germany
Adolescents’ involvement in delinquent behavior resembles that of their peers, and the same is true for substance use including smoking, alcohol consumption and the use of illicit drugs (Akers, 1998; Warr, 2002). Whether this similarity is due to influence processes within the peer group, i.e. friends’ behavior becoming more similar over time, or due to selection, i.e. adolescents with similar behavior choosing each other as friends, is still an open and hotly debated question with theoretical arguments for both causal directions (see e.g. Akers, 1973 ;Gottfredson, 1990; Knecht, 2008). Recent advances in statistical modeling for longitudinal network data, namely the development of stochastic actor-oriented models (SAOM) for network dynamics (Snijders, 1996; Snijders, Van de Bund and Steglich, 2010), have renewed interest in this longstanding empirical question. These models allow to simultaneously estimate selection and influence processes in longitudinal network data, disentangling these two possible causal processes. In the last two decades, SAOMs have been employed in a sizable number of empirical studies on delinquency and substance use, mainly analyzing school samples. Results of these studies are heterogeneous, though, with some studies only finding influence( e.g. Weermann, 2011), some only finding selection effects (e.g. Knecht, 2008), and some finding evidence supporting both causal pathways (e.g. Osgood, Feinberg and Ragan, 2015). In addition, some researchers have started to move beyond the question of selection vs. influence, addressing follow-up questions such as possible moderators and mediators (Veenstra, Dijkstra,Steglich and Van Zalk, 2013).
The current study systemizes the existing literature employing SAOMs. We follow the protocol set forth by Moher, Liberati, Tetzlaff and Altman (2009) to ensure that all relevant studies are found. By now more than 5300 journal abstracts are identified by search terms at the databases SSCI and PsycINFO.
We employ meta-analysis to integrate the heterogeneous results found in different studies regarding influence and selection effects and present preliminary results of our study.