Food Price Swings Drive Under-5 Deaths in Ethiopia Over Decades

Food price volatility in Ethiopia raises the risk of death for children under five, but the damage is not immediate. A new open-access study published in Agriculture & Food Security analyzed annual national data spanning 1970 to 2020 and found that unpredictable swings in food prices increase under-five mortality mainly over medium and long time horizons, while short-run effects are weak and statistically insignificant. The work also uncovered a striking asymmetry: declines in volatility produce larger improvements in child survival than the mortality increases caused by comparable rises in volatility.

The research team, led by Abule Mehare with Shemelis Kebede Hundie and Lamessa T. Abdisa, set out to fill a gap in a literature dominated by cross-sectional and household-level studies. Most previous Ethiopian research examined food insecurity, stunting, or wasting rather than mortality directly, and existing national time-series analyses of child mortality tended to be descriptive or forecasting-oriented, treating food price dynamics as irrelevant. No earlier study had tested whether the effects of price volatility evolve across temporal horizons or respond asymmetrically to rising versus falling volatility.

To measure volatility itself, the authors constructed an annual series from monthly food price index data published by FAOSTAT. They estimated a GJR-GARCH(1,1) model on monthly logarithmic price changes, a class of models designed to capture time-varying conditional variance and volatility clustering, and then aggregated the resulting conditional variances to the annual level. The generated series shows moderate volatility through the 1970s through the 1990s, followed by a marked and persistent increase beginning in the early 2000s that continued through the late-2000s global food price crisis and rose again toward 2020, indicating sustained periods of heightened uncertainty rather than isolated short-lived shocks.

The mortality model included per capita GDP, life expectancy at birth, the proportion of the population using basic sanitation, a government effectiveness index, and out-of-pocket health expenditure per capita, drawn from the World Development Indicators, the WHO Global Health Observatory, and the Worldwide Governance Indicators. Because conventional unit root tests can mislead when structural shifts are present, the team applied break-adjusted tests including the Clemente–Montañés–Reyes and Zivot–Andrews procedures, plus a nonlinear Kapetanios–Shin test. All series were found to be either I(0) or I(1), with no I(2) behavior, satisfying the requirements for the Fourier-based autoregressive distributed lag estimation that followed.

The FA-ARDL bounds test confirmed a stable long-run relationship among the variables. In the long-run estimates, food price volatility carried a statistically significant positive coefficient of roughly 0.40, meaning a 1 percent increase in volatility was associated with about a 0.4 percent increase in under-five mortality. Economic growth showed a significant negative association, with a 1 percent increase linked to a reduction of about 0.54 percent in the mortality rate. A one-unit increase in the government effectiveness index corresponded to a reduction of about 0.069 in under-five mortality, while a 1 percent rise in out-of-pocket per capita health spending was associated with an increase of roughly 0.27 percent. Life expectancy at birth showed no statistically significant long-run association.

The short-run picture differed sharply. Except for the lagged mortality terms, the first-difference coefficients were close to zero, and the lagged error-correction term was also nearly zero, which the authors interpret as indicating great difficulty returning to long-run equilibrium after a disturbance. CUSUM and CUSUMSQ stability diagnostics stayed within 5 percent critical bounds, supporting the stability of the estimated model.

To test for asymmetry, the researchers applied a Fourier-based nonlinear ARDL framework that decomposes volatility into positive and negative partial sums. The long-run coefficients for the two components differed significantly in sign and magnitude, with the negative partial sum showing the larger response, meaning that reductions in volatility are associated with a larger long-run decline in mortality than the increase produced by equivalent rises. In the short run, a contemporaneous increase in volatility was associated with a slight reduction in mortality of about 0.0003, whereas a decrease corresponded to an increase of about 0.0015, and negative shocks showed significant effects extending over two lags while positive shocks displayed only one significant lagged effect. The authors note these directional short-run patterns alongside the dominant long-run finding that falling volatility yields larger survival gains.

Frequency-domain Granger causality tests, using the Toda–Yamamoto approach with a maximal integration order of one and an optimal lag length of four, added a temporal dimension. Food price volatility Granger-caused under-five mortality across short, medium, and long-term frequencies in the spectral analysis, while the discussion emphasizes that the causal link concentrates at medium and long-term frequencies with no causal link in the short term. Economic growth Granger-caused mortality only in the long and medium terms, government effectiveness in the medium and long terms, and out-of-pocket expenditure across multiple frequencies. Standard Toda–Yamamoto tests also revealed two-way Granger causality between mortality and most other variables in the model.

The discussion translates the elasticity into illustrative terms: the authors calculate that a 32.1 percent rise in volatility observed during the COVID-19 pandemic would be associated with roughly a 12.8 percent rise in under-five mortality, equivalent to about 96,000 additional child deaths based on 2022 population estimates. They frame the mechanisms through the Agricultural Household Model, consumption smoothing theory, and the UNICEF Child Survival Framework, in which volatile prices raise income risk, prompt households to cut dietary diversity and delay healthcare, and erode government fiscal capacity for health and sanitation investment. The study is careful to note that these transmission channels are theoretically justified but were not empirically estimated within its reduced-form framework.

The authors acknowledge important limitations. Reliance on national-level annual data obscures subnational heterogeneity and immediate household coping responses, the reduced-form design cannot statistically decompose nutritional, healthcare, and intergenerational pathways, and unmeasured confounders such as localized conflict may affect the long series. They call for mediation analysis with household survey data and high-frequency, spatially disaggregated research. On the policy side, the findings point toward stabilizing food prices through storage infrastructure and market information systems, reforming the Productive Safety Net Programme’s triggers to respond to price volatility, expanding Community Based Health Insurance to reduce out-of-pocket burdens, and investing in government effectiveness and basic sanitation, since reductions in volatility appear to deliver disproportionately large returns for child survival in Ethiopia and comparable low-income settings.

Subject of Research: The dynamic and asymmetric effects of food price volatility on under-five mortality in Ethiopia from 1970 to 2020

Article Title: The effect of food price volatility on under-5 child mortality in Ethiopia

Article References: Mehare, A., Hundie, S. K., & Abdisa, L. T. (2026). The effect of food price volatility on under-5 child mortality in Ethiopia. Agriculture & Food Security, 15(1), Article 36. https://doi.org/10.1186/s40066-026-00621-z

Image Credits: AI Generated

DOI: 10.1186/s40066-026-00621-z

Keywords: food price volatility, under-five mortality, Ethiopia, ARDL, asymmetry, Granger causality, child survival, sanitation, government effectiveness, out-of-pocket health spending, food security, time-series analysis