This study examines the predictors of bond friendship formation among women in a rural Hamer community in Ethiopia. Using data from a closed social network of 91 women, we investigate how structural, dyadic, and individual-level variables influence the likelihood of bond friendship ties. Bond friendship is categorized as no tie, asymmetric tie, or dyadic tie. We aim to test hypotheses about homophily (similarities in wealth, age, anthropometrics, and reproductive history), social structural factors (natal clan, husband’s clan, and geographic proximity), and behavioral influences (shared activities and husband’s bond friendships) on bond friendship formation.
The primary statistical framework is Exponential Random Graph Models (ERGMs), complemented by additional modeling approaches such as dyadic regression and stochastic block models to explore specific hypotheses.