Improving Smallholder Agriculture via Video-Based Group Extension

Abstract

Providing technical advice to low income households at scale poses operational challenges, typically with respect to identifying and training a sufficiently large staff. Technology can help deliver messages more broadly, but at the risk of reducing their efficacy as customized messaging and human interaction are diminished. We tested a video added onto standard human-provided extension services to promote a climate-smart practice, System Rice Intensification (SRI) in Bihar, India. Using frequentist statistical methods, we find large but imprecisely estimated effects; the 95% confidence interval for treatment effect on output is between 10 and 500 kilograms and on profit is between 717 and 9650 Rps. However, our data are not normally distributed; specifically, key outcomes such as output have fat tails. We, thus, also employ a Bayesian hierarchical model and find smaller but more precise treatment effects with 95% of the effect on output falling between -8 and 70 kilograms, and between -193 and 1380 Rps for profit. We also test two messaging sub-treatments designed to address commonly cited constraints to SRI adoption: labor needs and self-efficacy of the farmer. A frequentist analysis shows no added gains, while the Bayesian shows an added benefit when both messages are delivered in tandem.

Keywords: water, field experiment, agriculture, system rice intensification, video-based training, group extension, information

JEL Classification: D13, D83, O12, O13, O33 , Q01, Q12, Q25