What Drives Persistent Cost-Price Squeezes in Food Systems?

Cost–price squeeze (CPS) is an important but under-measured form of producer-side stress in food systems. It occurs when rising production costs are not sufficiently matched by output-price adjustments, resulting in compressed producer margins. While existing research has documented agricultural profitability pressures in specific commodities or countries, less is known about the conditions under which CPS emerges across global staple crop systems. This paper examines the drivers of CPS using a harmonized country–year panel for maize, wheat, and rice systems over 2000–2023. Building on an outcome-based CPS indicator developed in a companion pattern paper, we estimate fixed-effects models, fertilizer shock-exposure specifications, placebo tests, and machine-learning prediction models. The results show that exchange-rate pressure is the most stable external predictor of CPS risk, suggesting that macro-external price exposure plays an important role in producer margin stress. More importantly, global fertilizer price shocks disproportionately increase CPS risk in countries with higher pre-period fertilizer intensity. This shock-exposure result remains robust after controlling for lagged macroeconomic, food-trade, agricultural-structure, and climate variables, while a lead-shock placebo is statistically insignificant. Nutrient-specific estimates indicate that nitrogen exposure is the most consistent fertilizer-related channel. Machine-learning models provide complementary predictive validation: Random Forest, XGBoost, and LASSO-logit models achieve meaningful out-of-sample performance, and fertilizer shock exposure, nitrogen shock exposure, food-trade variables, and global fertilizer shocks rank among the most important predictors. The findings suggest that CPS risk emerges from the interaction between external cost shocks and country-specific exposure structures, highlighting the importance of monitoring input dependence as an early-warning indicator of food-system vulnerability.