Project Drishya - Picture-based crop services: A randomized control trial evaluating the impacts of using smartphone camera data for personalized agro-advisory and insurance claims verification in India

How to cite

Kramer, Berber; Kannan, Samyuktha; Ceballos, Francisco. 2017. "Project Drishya - Picture-based crop services: A randomized control trial evaluating the impacts of using smartphone camera data for personalized agro-advisory and insurance claims verification in India." AEA RCT Registry. AEARCTR-0002587. https://doi.org/10.1257/rct.2587-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Berber Kramer* , Samyuktha Kannan , Francisco Ceballos

* Primary investigator

Registry ID
AEARCTR-0002587
DOI
https://doi.org/10.1257/rct.2587-1.0
First registered
Last updated
Status
In development
Trial dates
2017-11-29 – 2018-12-31
Countries
Private
JEL codes
Agricultural Finance (Q14), Agricultural Extension Services (Q16)
Registry record
http://www.socialscienceregistry.org/trials/2587

Abstract

Smallholder farmers in India are increasingly exposed to climate change and extreme weather events. Timely and customized agricultural advice for specific crops grown by a farmer can improve management practices, productivity and profitability. However available services today do not incorporate potentially useful information on what farmers observe themselves, for instance crop color, texture and how the crop is growing over time. These features can potentially make advisories and loss monitoring more effective by providing information on crop growth stage, pests and diseases, and visible damage to the crops grown. We test the effectiveness of an extension model that provides personalized agro-advisory based on localized information and visible crop characteristics derived from a stream of farmers’ own smartphone pictures for wheat crop in Haryana and Punjab regions of India. Pictures will be collected from farmers on a periodic basis through a freely available, easy-to-use mobile application. A first treatment arm will test the effects of this picture-based advisory model. A second treatment arm will test the added impact of providing a novel crop insurance that uses the stream of pictures to estimate plot-level crop loss in a cost-effective manner. We hypothesize that real-time crop data obtained from smartphone pictures can strengthen agricultural advisory and reduce basis risk in crop insurance. We will measure impact on take-up, productivity and profitability.

Intervention

The trial consists of a control arm and two treatment arms. Interested farmers in both control and treatment villages will be enrolled to receive generic advisory messages relating to weather, pest and diseases and input management via SMS and IVR on a periodic basis throughout the wheat growing season. Farmers will also have access to the national crop insurance scheme offered by the Government of India at a subsidized premium and other available private insurance products.

In the two treatment arms, field staff make public announcements about the program and invite all interested farmers to enroll for the advisory service in late November and early December 2017. To enroll, farmers download our app from the Google Play Store, register personal details and plots to be enrolled, and submit initial pictures for these plots. Twice per week, farmers take repeat pictures of enrolled sites and complete a short survey on crop management. The smartphone app includes built-in features designed to facilitate this process, enforcing that pictures are always taken of the same site and on time. Field staff support the farmers throughout the season. A team of agronomic experts from CABI will for every insured plot (1) review localized information and regularly captured crop pictures, (2) personalize default customized messages, and (3) send these personalized messages through the smartphone app. All farmers, including those interested but without smartphones, can enroll to receive advisory through SMS and Interactive Voice Response (IVR). They also have access to insurance products commonly available in the market or through the Government scheme.

In addition, only in the second treatment arm, we will additionally conduct insurance awareness campaigns and offer enrolled farmers the opportunity to subscribe to a picture-based insurance policy underwritten by HDFC Ergo in December 2017. 10 farmers who have been regularly clicking pictures and are interested in the insurance policy will be offered a full subsidy on the premium for a site of 1 acre. Farmers interested in insuring more than 1 acre and other farmers will be allowed to do so by paying the full premium amount. The insurance policy covers visible damage not due to mismanagement (eg. natural calamities such as lodging or hailstorm, pests and diseases, excess rainfall) and invisible damage due to excess temperature (calculated at the nearest weather station). The team of agronomic experts involved in providing the picture-based advisory services will estimate the damage based on pictures. Farmers with more than 20% yield loss receive payouts proportional to the estimated degree of damage.

Experimental design

250 villages across 9 districts in Haryana and Punjab are randomly assigned to receive either generic advisory (100 villages), treatment consisting of picture-based advisory or PBA (75 villages) or treatment consisting of picture-based insurance and picture-based advisory or PBA+PBI (75 villages). Of these, 50 villages from 6 districts in Haryana and Punjab follow a different sampling frame retained from an earlier study, with 25 villages randomly assigned to PBA and 25 randomly assigned to PBI. The remaining 200 villages are from 3 districts in Haryana, with 100 villages randomly assigned to control receiving generic advisory, 50 villages assigned to PBA treatment and 50 to PBA+ PBI treatment. Although randomization was conducted at a geographical cluster level, care was taken to ensure that the number of villages was balanced across arms.

Design details

Randomization method
Randomization done in office using Stata.
Randomization unit
In 200 villages in Haryana, the unit of randomization is at the cluster-level and stratified by district, where the clusters were defined as a group of up to 4 villages within a geodetic distance of 2 km of each other. Although randomization was at the cluster level, it was done in a manner to ensure that the number of villages was balanced across arms. 50 villages in Haryana and Punjab retained an earlier village-level randomization strategy that was stratified by weather station.
Clusters
200 villages from 162 clusters in Haryana (new sample); 50 villages (unclustered) in Haryana and Punjab (from a previous study)
Observations
6250 farmers
Arms
100 villages (72 clusters) control, 75 villages (35 clusters) picture-based advisory, 75 villages (40 clusters) picture-based insurance and picture-based advisory
Primary outcomes
Satisfaction with and trust in the advisory messages; insurance awareness and insurance coverage; adoption of recommended agronomic practices.
Secondary outcomes
Agricultural production, agricultural productivity (yields); costs of production; and profitability.

Cited by

No citation data available for this registration yet.

Keywords