Vegetation productivity and agricultural employment in Central Asia

Abstract Environmental change may influence rural labor allocation, but aggregate evidence from Central Asia is limited and sensitive to how environmental exposure and employment are measured. This study examines the association between annual vegetation productivity and the agricultural employment share in Kazakhstan, Kyrgyzstan, Tajikistan and Uzbekistan during 2001–2020. We construct quality-screened country-year net primary productivity (NPP) measures from the MODIS MOD17A3HGF.061 NPP and quality-control layers and merge them with a harmonized economic panel. The primary two-way fixed-effects specification estimates a non-causal within-country association between one-year-lagged NPP and the employment share, conditional on lagged controls. The balanced panel has 80 country-year observations and the primary sample has 74. The primary coefficient is − 0.320 percentage points per 0.01 kg C m −2 yr −1 (HC3 standard error = 0.446; enumerated wild cluster restricted (WCR) bootstrap p = 0.250). A two-year prior-average exposure yields a larger negative point estimate but no stronger four-cluster evidence. The coefficient changes sign in first differences and under the strict QC25 measure, while leave-one-country-out estimates vary in magnitude. Moreover, 92.6% of NPP variation is between countries and only 7.4% is within countries. The evidence therefore does not establish a robust directional association. The study contributes transparent environmental-data construction, small-cluster inference diagnostics and a clear separation of within-country from descriptive between-country evidence. The findings support finer-scale integration of environmental and labor-market monitoring, not causal claims or intervention ranking.