Designing policy and encouraging adaptation for small-scale agriculture when self-reported and measured climate shocks differ

• In Event: Agricultural Policy, Climate Adaptation and Welfare: International Perspectives

Abstract

Introduction/Background

Policy-makers concerned with helping farmers manage climate risk make decisions based on meteorological models and direct observation. However, individuals, however, base their choices on perceived risk, a function of their own climate experiences.  In sub-Saharan Africa, small-scale producers face increasing exposure to catastrophic losses due to climate shocks (Azzarri and Signorelli, 2020). Field analyses have observed that self-reported climate shocks differ from the meteorological record (Guiteras et al., 2015, Cullen and Anderson, 2017; Nguyen and Nguyen, 2020); similarly, surveys of farmers across multiple low-income countries have shown nonuniformity in climate change perceptions among population clusters (Li et al. 2013; Ogalleh et al., 2013; Below et al., 2012). Understanding the basis of the discrepancy between perceptions and measured trends is vital for crafting policy that supports optimal climate change adaptation decision-making within agricultural communities.

Purpose/Research Question

The goal of this paper is to understand drivers of discrepancies between self-reported climate shocks and observed meteorological events in Ethiopia and Malawi. Specifically, we assess discrepancies arising from unique household characteristics and community level factors that would inform policy design.

Methods

We combine the Living Standard Measurement Study - Integrated Surveys on Agriculture (LSMS-ISA), a longitudinal household survey, with spatially interpolated daily rainfall measurements from the Climate Hazards Infrared Precipitation with Stations (CHIRPS; Funk et al., 2015) in Ethiopia and Malawi from 2009 to 2019. We define an observed drought as growing-season rainfall or consecutive dry days (CDD) departure from the 30-year mean of 1 standard deviation or more and compare to survey-reported crop losses blamed on drought. We classify survey responses on self-reported drought as a match, a false positive (reported drought without measured record signal), or a false negative (unreported drought with measured record signal). We estimate linear probability models of mismatches on household, farm, and hazard characteristics, number of self-reported non-drought shocks, exposure to drought shocks in previous years, and enumeration-area fixed effects.

Results/Findings

Households that recently experienced a non-drought shock, and Ethiopian households headed by women, report drought without a measured signal of low rainfall more often. This pattern suggests that vulnerability corresponds to greater sensitivity to departures from rainfall normals. Households with more plots or more diversified livestock holdings are less likely to report measured drought events, indicating that assets may buffer hazard sensitivity. Drought shocks in previous seasons raise the probability of false positives, which may show vulnerability carries over between seasons.

Conclusion/Implications

The findings indicate a need for differentiated interventions. For vulnerable households with heightened sensitivity to rainfall variations, promoting the adoption of improved production technologies, such as stress-tolerant seeds, could use climate concerns to strengthen resilience. For households where asset holdings buffer shocks, targeted risk communication can promote loss-mitigating action before the next shock. Matching interventions to these distinct groups can improve the efficiency of agricultural development investment. 

Author

• Joaquin Mayorga, University of Washington
  Presenting Author