Peer effects in agricultural extension: Evidence of endogenous social interaction in the performance of community knowledge (extension) workers in Uganda

The reflection problem, as described by Manski (1993), is a major issue in the identification of peer effects due to endogeneity, which limits the possibility of separating the effects of the observed attributes of peers from that of social interaction. However, recent methodological advances in social network analysis (such as Cohen-Cole (2006); Moffitt (2001); Bramoull´e, Djebbari, and Fortin (2009); De Giorgi, Pellizzari, and Redaelli (2010); Sacerdote (2001); Mas and Moretti (2009)) now show that it is possible to separate these effects. I utilize these rich theoretical methods and the empirical applications (e.g., Krish- nan and Patnam (2014); Bramoull´e, Djebbari, and Fortin (2009); De Giorgi, Pellizzari, and Redaelli (2010)) to empirically identify and estimate peer effects in the performance of rural extension workers, known as community knowledge workers (CKWs) in Uganda. I exploit a unique data set which comprise of administrative monthly records of the total monthly performance of CKWs (in terms of the agricultural information provided to farmers) as they deliver extension services to smallholder farmers in rural Uganda by means of smartphones which contain an agricultural information database. I separate the effects of peer outcomes on individual performance from those associated with group or social interaction. I use a pool of these total monthly performance records for a 13-month period including December 2010 to December 2011. The identification strategy utilizes the assumption of intransitive triads (Bramoull´e, Djebbari, and Fortin, 2009) or partially overlapping peers (De Giorgi, Pellizzari, and Redaelli, 2010) within a network of interacting agents. I construct a panel data set that constitute the monthly performance of individual CKWs across 13 districts for 13 months of community extension operation by these CKWs. The approach identifies peer effects among CKWs in terms of their total monthly performance. This study contributes to the literature by being first to empirically estimate peer effects among extension workers based on their total monthly performance in the context of a developing country. The study has significant policy implications for improving agricultural extension in developing countries and elsewhere