[Paper Review] Risk, Agricultural Production, and Weather Index Insurance in Village India

Jeffrey D. Michler, Frédéri Viens+1 more|arXiv (Cornell University)|Mar 19, 2021

Agricultural risk and resilience1 citations

TL;DR

This study uses parcel-level panel data from rural India to quantify the role of seasonal weather variation in crop yield variance, finding it accounts for only 19–20% of total yield variation. Using multilevel modeling and Bayesian estimation, the authors derive actuarially fair pricing for weather index insurance and identify excessive loading factors (λ = 0.47–3.86) as a key reason for low farmer uptake, suggesting overpricing—not lack of understanding—undermines demand for index insurance.

ABSTRACT

We investigate the sources of variability in agricultural production and their relative importance in the context of weather index insurance for smallholder farmers in India. Using parcel-level panel data, multilevel modeling, and Bayesian methods we measure how large a role seasonal variation in weather plays in explaining yield variance. Seasonal variation in weather accounts for 19-20 percent of total variance in crop yields. Motivated by this result, we derive pricing and payout schedules for actuarially fair index insurance. These calculations shed light on the low uptake rates of index insurance and provide direction for designing more suitable index insurance.

Motivation & Objective

• To quantify the contribution of seasonal weather variability to crop yield variance among smallholder farmers in rural India.
• To assess the actuarial fairness of weather index insurance contracts using a short-term, localized data perspective.
• To investigate whether excessive insurance pricing—rather than behavioral or informational barriers—explains low uptake of index insurance in village India.
• To evaluate basis risk in production-oriented index insurance by estimating the misalignment between weather indices and actual yield losses.
• To provide evidence-based guidance for designing more actuarially fair and demand-aligned index insurance products in developing countries.

Proposed method

• Employs multilevel/hierarchical regression models to decompose yield variance across multiple levels: parcel, household, seasonal weather, village, and time effects.
• Uses Bayesian estimation techniques to handle skewed distributions of disturbance terms, particularly for idiosyncratic and extreme weather effects.
• Applies a five-season panel dataset (including a drought season) from 11,942 parcel-level observations across geographically dispersed locations in Andhra Pradesh.
• Derives actuarially fair insurance premiums using rainfall data over a five-year window, contrasting with traditional long-horizon actuarial methods.
• Calculates loading factors (λ) as the ratio of observed premium prices to actuarially fair prices to assess pricing markups.
• Uses the formula 1 + λ = paid price / actuarially fair price to quantify transaction costs embedded in commercial contracts.

Experimental results

Research questions

• RQ1 How much of the total variance in crop yields in rural India can be attributed to seasonal weather variation?
• RQ2 What is the actuarially fair price for rainfall-based index insurance when calculated over a short-term, localized five-year window?
• RQ3 To what extent do observed insurance premiums in village India exceed actuarially fair prices, and what are the implications for uptake?
• RQ4 How does basis risk—especially the misalignment between weather indices and actual yield losses—affect the utility of index insurance for farmers?
• RQ5 Why do farmers in village India exhibit low uptake of weather index insurance despite its potential to hedge covariate risk?

Key findings

• Seasonal weather variation accounts for only 19–20% of total variance in crop yields, indicating that most yield variation stems from non-weather sources such as parcel-specific and idiosyncratic factors.
• The actuarially fair premium for a high-payout rainfall index contract is estimated at 191 Indian Rupees, based on a five-year, localized dataset.
• Commercial premiums for the same contract are observed at 280 Rupees, implying a loading factor (λ) of 0.47, significantly higher than typical values (0.10–0.20) in other global markets.
• Loading factors inferred from two villages in the Cole et al. (2013) study reach up to 3.86, indicating substantial pricing markups that likely deter farmer uptake.
• The study finds that basis risk is substantial due to low correlation between weather indices and actual yield losses, especially when extreme but infrequent weather events occur.
• Farmers may rationally forgo index insurance at any positive price, as they are already effective at managing idiosyncratic risk and the covariate risk from seasonal weather is relatively small.