A “true” random effects stochastic frontier analysis for technical efficiency and heterogeneity: Evidence from manufacturing firms in Ethiopia

This study examines the technical efficiency of the Ethiopian manufacturing sector using establishment-level census panel data over the period of 2000 to 2009. The “true” random effects stochastic frontier model (Greene, 2005a,b), which can disentangle time-varying technical inefficiency from time-invariant unobserved heterogeneity, and the conventional fixed and random effects models are used to estimate efficiency for the aggregated and individual industry groups. The results indicate that efficiency estimates are sensitive to model specifications of firm-specific unobserved heterogeneity. We find a significant gap in efficiency estimates between the “true” random effects model and the fixed and random effects models, which would imply considerable heterogeneity of manufacturing firms in Ethiopia. Our results suggest that firm-specific heterogeneity would be particularly significant in the food and beverages, non-metals, and furniture industries. We also show that the production of the Ethiopian manufacturing sector is largely responsive to changes in intermediate inputs compared to labor and capital inputs. The estimated technical efficiency considerably varies across firms within an industry suggesting a significant potential for improving efficiency in the sector. We discuss that the major problem for the variation in efficiency is the inability of firms to operate at their full production capacity, which was mainly caused by shortage of raw material supply. Generally, it is important to differentiate between inefficiency and unobserved heterogeneity in a stochastic frontier framework when firms operate under diverse social, industrial and environmental conditions.