Developing country governments often launch new agricultural programmes without collecting pre-implementation survey data, making it difficult to evaluate the effects of such programmes. Leveraging the flexibility of granular pixel-level satellite panel data and a well-developed quasi-experimental policy evaluation design, we study a programme where pre-implementation data is unavailable. We estimate the effect of cash transfers on agricultural productivity in Telangana, India. Treatment and control regions are within 10 km on either side of the state border. They are identical in all respects except for the difference in exposure to policy treatment. Agricultural productivity increased in the major monsoon cropping season due to the cash transfer programme. The findings also reveal that cash transfers helped reduce productivity gaps between irrigated and rainfed agricultural areas. Our results are robust to two different sources of satellite data, three alternative indicators of productivity, two rounds of full-scale resampling, 100 rounds of small-scale resampling and three alternative border designs. Placebo regressions of two previous years also confirm our results. This approach to policy evaluation is applicable anywhere satellite data are available in the world.

Authors: Muddasir Ahmad Akhoon, Abhishek Shaw, Vidya Vemireddy