Quantifying heterogeneity, heteroscedasticity and publication bias effects on technical efficiency estimates of rice farming: A meta-regression analysis

In recent decades, numerous studies have focused on technical efficiency in rice farming, finding considerable variation in mean technical efficiency (MTE) estimates. We conducted a meta-regression analysis (MRA), using a random-effects meta-regression model, to understand the variation in MTE estimates due to study heterogeneity, heteroscedasticity and publication bias. We used 443 observations extracted from 175 primary studies published in English in the last three decades. The results show that MTE estimates are affected by study heterogeneity. Variable returns to scale specification yielded higher MTE scores than constant returns to scale ones. Panel data, secondary data and value data had lower MTE estimates than cross-sectional data, primary data and physical (quantity) data, respectively. Compared to Southeast Asia, countries in East and South Asia had higher MTE estimates, whereas African countries had lower MTE estimates. We suggest that practitioners and policy-makers should consider carefully estimation specifications, data types and geographical regions of empirical studies when comparing and interpreting empirical results. The average genuine (predicted) MTE score was 0.76 (range 0.54–0.89), indicating the potential to improve technical efficiency in global rice farming and the need for further research to bridge managerial ability gaps among farmers.