"background": "Efficiency diagnostics in manufacturing systems within developing economies often rely on single-level analyses, which fail to account for the hierarchical structure of plant data. This limitation obscures the distinct influences of firm-level management and sector-wide technological factors on performance. ", "purpose and objectives": "This working paper develops and applies a multilevel regression framework to decompose efficiency variances in Senegalese manufacturing, aiming to quantify the proportion of performance variation attributable to firm versus within-firm effects. ", "methodology": "We employ a two-level hierarchical linear model. The level-1 model for plant i in firm j is y{ij = \0j + \1jXij + rij, with rij \ N (0, \²). The level-2 model is \0j = \00 + \01Wj + u0j, where u0j denotes random intercepts. Estimation uses restricted maximum likelihood with robust standard errors. ", "findings": "The analysis indicates that approximately 65% of the variance in technical efficiency scores is attributable to persistent differences between firms. The random intercept for firm-level capital intensity was statistically significant, with a 95% confidence interval indicating a positive association with plant-level output. ", "conclusion": "The multilevel approach provides a more nuanced diagnostic tool, confirming that firm-level heterogeneities are the dominant source of efficiency variation in the studied context, overshadowing intra-fplant operational differences. ", "recommendations": "Policymakers and plant managers should prioritise firm-wide capability building over uniform, sector-level interventions. Future engineering efficiency studies should adopt hierarchical models where nested data structures are present. ", "key words": "hierarchical linear model, technical efficiency, industrial engineering, variance decomposition, developing economy", "contribution statement": "This paper introduces a novel application of multilevel modelling for
Diagne et al. (Sun,) studied this question.