Crop-specific impacts of the 2023/2024 drought on Zambia’s agricultural production

The 2023/2024 El Niño event was one of the most significant climatic shocks to Zambia's agricultural sector, resulting in a severe mid-season drought across much of the country. This study quantifies crop-specific impacts of the drought on Zambia's agricultural production using district-level crop production and 30-year rainfall data used to construct the Standardized Precipitation Index (SPI). Exploiting spatial variation in the severity of the 2023/2024 drought, we estimate the effect of a one standard deviation rainfall deficit on crop yields and national production. The objective is to quantify how this single extreme event affected crop production in Zambia. Results reveal substantial heterogeneity in crop resilience. Among cereals and tubers, sorghum (9% yield reduction) and millet (12%) are more drought tolerant than maize (30%), rice (26%), and sweet potato (25%). Among pulses and oilseeds, sunflower (12%), cowpea (14%), common bean (15%), and groundnut (16%) show lower yield losses than soybean (21%) and Bambara nut (20%). We find significant heterogeneity in the effect of drought on cereal crop yield across agroecological zones, but less consistent heterogeneity for pulses. The results further indicate that approximately 76% of the observed 64% decline in national crop production value can be directly attributed to El Niño-induced drought conditions. We also show that simply comparing production in a drought year with non-drought years overstates the drought impact for most crops, suggesting that some losses reflect other production and management factors. Our findings support policy interventions focused on strategic crop diversification. Practical implications Climate adaptation policies that support smallholder farmers increasingly rely on quantitative estimates of drought impacts to guide resource allocation, social safety nets, and risk management decisions. Such policies include weather-index based insurance, forecast-based anticipatory action, and agricultural advisory services. In all cases, these interventions depend on understanding the relationship between the measured weather variable and agricultural loss, usually crop yields. This study provides empirical evidence directly applicable to three climate service domains in Zambia and similar contexts across Southern Africa: designing crop-specific drought insurance products, strengthening agricultural extension and advisory services, and improving early warning systems and anticipatory adaptation strategies such as crop diversification or switching. The findings address fundamental challenges in translating climate information into actionable decisions by demonstrating substantial heterogeneity in crop vulnerability and quantifying the proportion of production losses directly attributable to drought conditions. On the insurance side, as an example, Zambia's Farmer Input Support Programme includes a parametric drought insurance component where farmers pay premiums and receive payouts when drought indices exceed pre-defined triggers. However, determining the appropriate relationship between meteorological drought severity and expected agricultural losses remains a fundamental challenge. Our study demonstrates that a one standard deviation rainfall deficit during the 2023/2024 season results in substantially different yield losses across crops: sorghum loses about 9 percent and millet about 12 percent, while maize loses about 30 percent, rice 26 percent, and sweet potatoes 25 percent. For insurance designers, this finding has immediate practical implications. First, using a single crop-neutral drought index for premium calculation and payout triggers will systematically under-compensate farmers cultivating vulnerable crops while potentially over-compensating those growing more resilient crops. Insurance products should either develop crop-specific indices with differentiated triggers and payout rates or use weighted composite indices that reflect actual crop portfolios at the district level. Second, our finding that 76 percent of the observed 64 percent production-value decline is directly attributable to drought conditions shows why payout models should be calibrated using econometric estimates rather than simple comparisons across years. Third, the spatial heterogeneity across agroecological zones indicates that insurance products must be locally calibrated. For advisory and extension services, we argue that climate services should shift from general climate awareness to actionable advisories that identify which crops are more vulnerable or more resilient under specific drought conditions. For early warning systems, our findings enable impact-based forecasting that specifies expected consequences rather than generic alerts: when real-time monitoring or seasonal forecasts indicate rainfall deficits approaching one standard deviation below average, systems can issue differentiated impact forecasts showing relatively high expected yield losses for maize and rice, but lower losses for millet, sorghum, sunflower, cowpeas, common beans, and groundnuts. While this study focuses on Zambia, the methodology and findings are transferable to other drought-vulnerable regions in Southern Africa and beyond, as the analytical approach of using spatial variation in drought severity combined with district-level panel data to estimate crop-specific impacts can be replicated in other countries experiencing similar challenges.