It's Raining, It's Pouring? A web and SMS application in Flask and Twilio to call Farmers.

10:00 AM - 10:25 AM on July 16, 2016, Room CR5

Katya Vasilaky

Audience level:

intermediate

Watch:

https://www.youtube.com/watch?v=q6FBHTZFCE4

Description

This talk is about the use of python in building a web application with which we call farmers in less developed countries to capture rainfall occurrence. These data are used to complement lower resolution satellite data, and to potentially update climate models. The application is built in Flask and uses Twilio's API to deliver voicemails to farmers in less developed countries. Twilio's API records farmers' touch tone responses regarding the occurrence of rainfall.

Abstract

Index Insurance is a type of micro-insurance that insures large covariate risks at the village level, where informal risk sharing (friends and family) cannot. Based on a weather index, its payouts indemnify farmers if rainfall falls below a pre-specified rainfall threshold. Because indemnities are a function of an exogenous and publicly available rainfall index, the product removes much of the moral hazard and adverse selection associated with yield indemnified agricultural insurance.

Two challenges exist in scaling weather insurance products. First (1), there is a high cost of signing up farmers with a door-to-door process, and second (2), satellite data have limited resolution. (1) Field visits are often necessary to meet and reach farmers via local partners by car and by foot. Each visit to a farmer village is costly in time and human and physical resources. Furthermore, some villages may not be accessible because of roads and poor infrastructure, especially during floods. Second (2), satellite imagery has limited resolution, which means variations in rainfall within a pixel are not captured. Thus, farmers who should receive a payout may not.

This SMS/Voicemail application that we build here aims to remedy (1) and (2) by calling farmers directly to gather information directly from the farmer and stored in a relational database. This limits the number of visits needed for sign-ups, and helps cross check satellite data with farmers' real-time experiences. We are gradually building the application with feedback from beta testers including comprehension of the automated voicemail, and what questions would be feasible to ask over the phone.