Experimental Design
This study will include two pilot surveys, a baseline survey, and an end line survey approximately one year after the end of the baseline. We will be running this experiment in 219 villages in Ahmedabad and Gandhinagar, two districts in the Indian state of Gujarat, which have 35,329 current and former SEWA members.
In both pilots, we conduct surveys in three representative villages in Ahmedabad and two representative villages in Gandhinagar. Within each village, we select 50 representative households, yielding a total of 250 households in the first pilot and 250 households in the second pilot. Villages selected for the pilots are excluded from the main study.
During the first pilot, we implement a willingness-to-pay exercise in five villages in Ahmedabad and Gandhinagar to determine which two products yield the most consumer surplus. We vary the products along two dimensions: indemnity rates (i.e., amount of money paid out when a claim is made) and coverage (i.e., number of people in a household covered). Within each village, we randomly assign 25 households to the insurance product that varies along the indemnity rate dimension, but is fixed along the coverage dimension, and 25 to the insurance product that varies along the coverage dimension, but is fixed along the indemnity rate dimension. We then choose the two products with the greatest consumer surplus.
Approximately two weeks after the first pilot ends, the second pilot will begin. During the second pilot, we conduct another willingness-to-pay exercise focused on the two products selected in the first pilot. This pilot is conducted on survey respondents in a different set of five villages in Ahmedabad and Gandhinagar to determine optimal prices for the two products we've selected after the first pilot.
After the pilots, we randomly assign the remaining 209 villages to one of the two possible prices for each of the two products offered. Again, within each village, we select 50 representative households, yielding a total of 10,450 households total. SEWA representatives will offer the two hospi-cash policies selected in pilot 1 to households in each village at the prices assigned to that village. SEWA will also continue to offer it's standard suite of non-hospi-cash policies to SEWA households. This hospi-cash marketing push will last 6 to 9 months.
At the beginning of the marketing push we will conduct a baseline survey of sample households. After the marketing push, we will gather data from SEWA on which households purchased a hospi-cash policy and which one they purchased. We will also obtain from SEWA claims data for one year for for all sample households that enroll in a hospi-cash plan . Approximately one year after the end of the baseline, we will conduct an end line.
We will estimate the causal impact of price on uptake of hospi-cash policies using an ITT design. We will back out bounds on the demand function using Tebaldi et al. (2018). We will examine the impact of enrollment in wage insurance on the uptake of SEWA's health insurance product, utilization of hospital care, employment, and household financial planning using randomized prices as an instrument in a TOT design. This design will also enable us to estimate adverse selection into and moral hazard from wage insurance.