Beyond calories: rainfall shocks, dietary composition, and the limits of standard food security indicators in Malawi

Background Climatic shocks are known to undermine food security in rainfed African farming systems, but the evidence base relies overwhelmingly on indicators of energy adequacy and consumption frequency. Whether shocks also alter the composition of diets, independently of their quantity, remains largely untested. Methods We use four waves of the Malawi Integrated Household Panel Survey (2010, 2013, 2016, and 2019; 5,866 household–wave observations), in which food consumption is recorded item by item in physical quantities and disaggregated by source. Household and wave fixed effects identify associations from within-household variation over time, and our preferred inference uses Conley standard errors that allow for the spatial correlation of rainfall. Results A 20 percentage point rainfall deficit is not robustly associated with lower energy intake but is associated with a consistent shift in diet composition. The share of dietary energy from nutrient-dense foods falls by about 0.7 percentage points (4.1% of its mean), the staple share rises correspondingly, and dietary energy concentrates on fewer food groups; fruits and pulses bear the adjustment while cereals are unaffected. Own production declines by 1.9 percentage points and households consume 0.7 fewer distinct food items, while real food expenditure does not rise—a pattern of non-compensation rather than substitution toward the market. These four indicators of composition are complementary transformations of the same vector of food-group energy shares rather than independent tests; they move in a consistent direction, but their magnitude is modest and, under our preferred spatially-robust inference, the effect is imprecise and a joint test is not significant (χ 2 (4) = 7.75, p = 0.10). Conclusion The results are consistent with households maintaining energy intake while the composition of the diet shifts toward staples and away from nutrient-dense food groups—a compositional margin that energy-based indicators capture poorly. Because this margin is both small and imprecisely estimated under our preferred spatially-robust inference, the study carries a methodological caution as much as a substantive finding: analyses of climate shocks on diets that rely on energy-based measures, or on standard errors that ignore spatial dependence, are liable to report effects that are larger and more statistically secure than the data support.