Why Panel Data Models Systematically Fail to Detect Climate Effects: Evidence from African Cereal Production

This paper addresses a persistent and consequential puzzle in climate econometrics: the recurring observation of null climate effects in panel fixed-effects models for African agriculture, despite overwhelming agronomic and micro-level evidence of climate vulnerability. This identification failure has led to confusion regarding climate risk assessment among policymakers. Recent studies by Gebreslassie (2024) and Affoh et al. Robustness checks confirm null findings are stable across alternative standard error specifications, ruling out error misspecification as an explanation. (2024) continue to document these statistically insignificant effects once year fixed effects are included. Using cereal yields from ten African countries (2000-2023), we deliberately replicate this "null results paradox" -our baseline specification with year fixed effects produces no significant climate coefficients (p > 0.28) for temperature and precipitation). Through variance decomposition, we provide the first empirical quantification of this identification failure: we demonstrate that the annual fixed effects alone absorb 15 percentage points of the explanatory power for within-country yield variation (R2 reduction from 0.380 to 0.230). Crucially, we demonstrate that the annual fixed effects act as a statistical absorber, masking economically significant relationships. When year dummies are removed, the strong and theoretically sound signal of fertilizer use (coefficient = 0.202, p