Realized link to paper didn't work before. Here it is manaswinirao.com/files/202207...

We rule out the role of physical distance and show that social distances matter in the context of India. Finally, we show that much of the diffusion is observed within higher and intermediate status jati groups even though low status farmers are more likely to experience floods.

We also find that this is driven by higher adoption among farmers in the same jati as the initial, randomly treated, recipient farmers in treated villages.

This average effects masks substantial heterogeneity based on the underlying social structures (jati structures) of the study villages. We find significantly higher diffusion since the start in homogeneous villages relative to fractionalized villages.

We find modest overall diffusion of new, flood resistant seeds introduced in coastal floodplains of Odisha (India) among non-treated farmers in treated villages relative to their counterparts in control villages- only 2pp more farmers say that they grow the variety.