Bringing Social Context Back In

Abstract This article demonstrates the utility of combining individual survey data with social media content in order to analyze how context predicts behavior. The lack valid and reliable measures contexts interaction which individuals are embedded has remained an Achilles heel method. reason is that collection direct observation requires qualitative analysis context, which, hitherto, been too costly collect on a large scale. Instead, researchers have resorted indirect such as aggregate group composition, respondent reports institutional accounts. However, recent advent data, contemporary scientists now unprecedented To utilize these for quantitative analysis, transform text prose into good measurement. We combine supervised machine learning ensure both semantic validity accuracy our measure Facebook groups. test substantive performance interaction, we use it predict participation refugee solidarity activism Denmark. Additional testing indicates cannot easily be replaced by derived from composition also show respondents can effectively sampled using Finally, discusses limitations points alternate settings where design applicable.