Processed food plant in West Bengal, employs both Hindus (majority group) and Muslims (minority), with a history of conflict.
Group composition randomly assigned (with some limitations), with high- and low-dependency tasks (quasi-randomly assigned)
Key findings:
Religious diversity negatively affects team output, but only in HD tasks
The difference in output attenuates over time (20% gap to 1%), driven by gains in mixed HD teams
Reduction in negative out-group attitudes in HD mixed groups
Religious Divisions & Production Technology
Religious Divisions & Production Technology
"It is insightful and non-obvious that the largest positive effects of treatment on attitudes occurred in teams that also suffered the largest negative output shocks."
Some thoughts:
It might have been better to collect more information about beliefs and attitudes before endline
Does the best possible to reconcile design limitations (e.g. no Muslim only teams) with robustness checks using observables, but wouldn't have to do this in a perfect world