Bundling Drought-Tolerant Seeds and Index Insurance: Impacts on Agricultural Investment and Risk Management in India

Setting: The study was conducted in six blocks across three districts of Odisha, India—a region characterized by rainfed rice cultivation and high exposure to drought risk. Participating households were smallholder farmers who cultivated rice as their primary crop and owned at least some of the land they cultivated.

Randomization: Villages were randomly assigned to one of three experimental arms within each block:

Control group: 38 villages (760 households) - No intervention; farmers continued with existing cultivation practices
DT treatment arm: 36 villages (720 households) - Offered drought-tolerant rice seeds
DT-WII treatment arm: 37 villages (740 households) - Offered drought-tolerant rice seeds bundled with weather index insurance
Randomization was conducted at the village level to minimize spillovers between treatment arms.

Drought-Tolerant Seeds: Households in the DT and DT-WII treatment arms were offered the opportunity to purchase up to three units of Sahbhagi dhan, a drought-tolerant rice variety developed by the International Rice Research Institute (IRRI). Each unit contained a 2 kg bag of seed, sufficient for cultivating 0.1 acre using manual transplanting methods.

Weather Index Insurance: Households in the DT-WII treatment arm were offered a bundled product consisting of DT seeds and a complementary weather index insurance policy. Each insurance policy covered 0.1 acre of land. The insurance was designed to provide payouts triggered by rainfall deficits measured at local rain gauges, independent of individual farm-level losses. The actuarially fair premium was calculated based on historical rainfall data and adjusted in the second year to reflect changes in the minimum support price and yield growth.

Pricing: To minimize price as a barrier to uptake and to enable estimation of demand elasticities, public lotteries were conducted in each treatment village to randomly assign village-level discounts on the products. These discounts were fixed for all purchases within a given village for the season.

Marketing and Training: Field agents from a local NGO partner (BSSS) conducted three marketing sessions in each treatment village prior to each planting season. These sessions: (1) explained the risks of drought and the potential role of the offered products in mitigating these risks; (2) provided product-specific information; and (3) facilitated sales. Agents also conducted experiential learning exercises to illustrate concepts of probability, drought risk, and weather unpredictability. In DT-WII villages, additional exercises explained the terms of the index insurance contract.

Duration: The intervention spanned two agricultural seasons (kharif 2015 and kharif 2016). Farmers could purchase products in both years, allowing observation of initial uptake and renewal decisions

2015-06-15

2016-10-15

1. Agricultural input expenditures (seeds, fertilizer, irrigation, labor)
2. Rice yields (tons per acre)
3. Household consumption expenditures
4. Food security status

1. Product uptake and renewal rates
2. Yield variability
3. Crop losses
4. Coping strategies in response to drought

This study employed a cluster randomized controlled trial design with randomization at the village level. Villages were randomly assigned to one of three experimental arms:

- Control group (38 villages, 760 households)
- DT treatment arm (36 villages, 720 households)
- DT-WII treatment arm (37 villages, 740 households)

Village-level randomization was chosen to minimize spillovers between treatment arms, as agricultural information and practices often diffuse within villages through social networks.

Treatment Arms
------------------
Control Group: Households in control villages received no intervention and continued with their existing cultivation practices. They had access to the same agricultural extension services and market opportunities as before the study.

DT Treatment Arm: Households in DT villages were offered the opportunity to purchase drought-tolerant (DT) rice seeds (Sahbhagi dhan variety). Each unit consisted of a 2 kg bag sufficient for cultivating 0.1 acre using manual transplanting. Households could purchase up to three units per season.

DT-WII Treatment Arm: Households in DT-WII villages were offered a bundled product combining DT rice seeds with weather index insurance (WII). Each bundle included one unit of DT seed (2 kg) and one WII policy covering 0.1 acre. The insurance policy provided payouts triggered by cumulative rainfall deficits measured at the nearest India Meteorological Department (IMD) rain gauge, independent of individual farm-level losses. Households could purchase up to three bundles per season.

Price Randomization
------------------------
To estimate demand elasticities and minimize price as a barrier to uptake, village-level price discounts were randomly assigned through public lotteries conducted in each treatment village. All households within a given village faced the same discounted price for the season. Discounts were re-randomized in the second year.

Insurance Contract Design
-------------------------------
The WII policy was designed with the following features:

- Trigger: Cumulative rainfall during the kharif (monsoon) season measured at the nearest IMD rain gauge
- Coverage period: Kharif season (June–November)
- Payout structure: Binary payout triggered when cumulative rainfall fell below predetermined thresholds indicating moderate or severe drought
- Independence: Payouts were determined solely by rainfall measurements, not by individual farm-level yields or losses
- Premium: Actuarially fair premium calculated from historical rainfall data, adjusted in Year 2 to reflect changes in minimum support price and expected yield

Marketing and training protocol
------------------------------------
Field agents from a local NGO partner (BSSS) conducted three marketing sessions in each treatment village prior to each planting season:

- Session 1: Explained drought risks and the potential role of offered products in mitigating these risks
- Session 2: Provided product-specific information (seed characteristics, insurance terms)
- Session 3: Facilitated sales and addressed remaining questions

Marketing sessions included experiential learning exercises using dice and spinners to illustrate concepts of probability, drought risk, and weather unpredictability. In DT-WII villages, additional exercises explained the terms of the index insurance contract, including the conditions under which payouts would and would not be triggered.

Identification Strategy
-------------------------
Intention-to-Treat (ITT) Effects: Primary analysis estimates ITT effects comparing outcomes across randomly assigned treatment arms using ANCOVA specifications that control for baseline values of outcome variables and household characteristics.

Local Average Treatment Effects (LATE): Secondary analysis estimates treatment effects on the treated using instrumental variables, where random assignment to treatment arms instruments for actual product purchase. Because products were not available to control households (no "always-takers"), LATE estimates represent average treatment effects on the treated.

Drought Classification
Drought severity was classified based on cumulative kharif season rainfall measured at IMD rain gauges relative to historical averages:

- No drought: Rainfall above moderate drought threshold
- Moderate drought: Rainfall 20-40% below historical average
- Severe drought: Rainfall more than 40% below historical average

Heterogeneous Effects by Drought: To examine whether treatment effects varied by drought exposure, we interact treatment indicators with drought severity classifications based on rainfall recorded at block-level rain gauges.

Randomization done in office in Stata

Village

Yes

111 villages

2100

- Control: 701
- DT only: 701
- DT-WII: 701

15% increase in total incomes