Close Friends Vs. the Norm Effect in a Classroom: Effects of Proximal and Distal Alcohol Social Context.

Background: Applied work usually uses the concept of norm to explain how social context affects the individual’s decisions. However, the social organization inside a classroom is composed by heterogenous relationships that are not explained by general norms. In a classroom, students´ sub groups can share different values regarding alcohol use. These values could be independent to the overall classroom norms of alcohol use. An analytical tool to account this heterogeneity is the concept proximal peer-context, which enables us to identify specific peers or relationships that affect individuals directly. Proximal context is more specific than social norms because capture the significant relationships in a social context influenced by multiple factors. Applying social network analysis, I operationalize the concept of proximal peer context and norm classroom context (distal) and study its effect on individual choices of alcohol use in last 30 days.

Research question: Are there differences in the association between students´ social context (proximal and distal) and the likelihood that a student will drink alcohol?

Ho: The proportion of peers connected to an ego who drink alcohol (proximal context) is a stronger predictor of ego use alcohol in comparison with the observed norm of alcohol use.

Methods: I identified the ego network of 2566 student nested in 73 classrooms (5 Colombian schools) and calculated the proportion of “close friends” with alcohol consumption in the last month. I calculated the aggregate norm by classrooms and structural networks characteristics like density and transitivity. Finally, I estimated a regression models taxonomy to identify the relationships of these predictors with the probability of alcohol use.

Results.

The proximal context norm was the stronger predictor of individual alcohol use OR=3.97; p<0.05; CI= [1.5-10.05] even controlling for the alcohol norm inside the classroom OR=3.22; p > 0.05; CI= [0.96-10.82]. The taxonomy analysis showed different favorable criteria for models that include more social context information: the last model that included proximal and distal variable, gender, age and structural networks characteristics obtained a χ² greater than early models (χ² =205.78), BIC’ was smaller (BIC'=-17.73) and the sensitivity and sensibility analysis perform better (72.2% of cases correctly classified).

Conclusion.

Proximal context measured through social network analysis method is a better predictor of alcohol use compared to the norm of alcohol use in the classroom.

There is an interplay between different levels of social context that can explain with more accuracy student’s decisions.

The interaction between specifics social structures and individual characteristics can be represented with sociometric information to have more complex understanding of how peers influence individual’s decisions.