Soil Information and Farmer's Decision-making

2023-06-12

2023-08-31

I will look at several outcomes designed to capture farmers' beliefs about salinity on their soil and their intentions about agricultural practices next season. I will assess farmers' self-reports about whether they are worried about salinity, their expectations about salinity changes both over the past 10 years and the next 10 years, their beliefs about the level of salinity next Boro season, five years ago, and five years in the future, and their expectations about the marginal effectiveness of saline-tolerant varieties next season. Additionally, I will assess impacts on farmers' willingness to pay for a saline tolerant seed variety, whether they expect to harvest Boro rice next year and on how much land, their other expected income sources next Boro season, their earnings and harvest expectations next season, and key aspects of their expectations about next year's agricultural production function (seed choice, irrigation method, and water source). Because I am also measuring farmers' social learning networks within the sample of farmers I survey in each village, I will also test for spillovers within village over the course of the three days in which enumerators interview farmers, instrumenting for the order in which farmers are interviewed based on the order that farmers appeared on the tablet for scheduling the interviews over the phone.

First, I create strata within which I randomize farmers. To do so, I begin by calculating whether farmers over- or under-estimated the amount of salt in their soil using the beliefs measured during the baseline survey. Then, I create strata using the intersection of the direction of belief error with village. Within these strata, I first assign farmers to one of two arms: an upazila-only arm, and an upazila-plus arm. In the upazila-only arm, farmers are offered information about the average soil salinity level in their upazila. In the upazila-plus arm, farmers are offered that same information plus the information on soil salinity from their own plot. In the initial randomization, I allocated 25 percent of farmers to the upazila-only arm and 75 percent to the upazila-plus arm. The distribution of treatment was chosen on the basis of power calculations to estimate to objects of interest. First, the impact of information about soil salinity in general on behavior, and second, the extent to which farmers update differentially about information on their own plot vs. about plots in their upazila at large. These power calculations which hinge critically on the extent to which farmers treat upazila information differently from their own, a parameter about which I had substantial ex ante uncertainty. Therefore, after 25 percent, 50 percent, and 75 percent of the data is collected, I will revisit this allocation based on updated power calculations, defaulting to not changing the ratios. This information is then sold to farmers through an ascending price list version of the Becker-DeGroot-Marschak elicitation mechanism. In the second stage of randomization in the experiment, farmers are randomly assigned the ultimate price they face in this elicitation. The procedure for assigning treatment is as follows: first, within strata, farmers are randomly assigned to either a price of 0 or 500 BDT. Then, the price faced by each farmer is randomly assigned to each number between 10 BDT and 490 BDT in increments of 10, with probability .0001 each. Due to an issue with updating the tablets on the first day of data collection, nine households were given the price of 30 BDT, which was the amount used in the training for enumerators. Note that the order in which villages were visited was randomized within enumerator. All randomization was done via computer.

All randomization was done via computer.

Individual

No

2,279

2,279

25 percent of farmers to the upazila-only arm and 75 percent to the upazila-plus arm; approximately 50 percent to a price of 0 and 50 percent to a price of 500 BDT.