A Bayesian Hierarchical Model for Yield Improvement Diagnostics in Ugandan Manufacturing Systems: 2000–2026

"background": "Persistent low yields in manufacturing systems constrain industrial development and economic growth. Diagnostic tools for identifying and quantifying improvement drivers are often inadequate, relying on aggregated data that mask plant-level heterogeneity and temporal dynamics. ", "purpose and objectives": "This work develops and evaluates a novel Bayesian hierarchical model to diagnose the sources of yield improvement within manufacturing systems. The objective is to provide a robust, probabilistic framework for decomposing observed yield gains into contributions from technological adoption, operational practice enhancements, and stochastic plant-level effects. ", "methodology": "A Bayesian hierarchical model is specified, where the yield Y{it for plant i in period t is modelled as Yit \ (\ + \ + \ Xit, \²). Here, \ captures plant-specific random effects, \ is a temporal trend, and \ represents the effect of a vector of intervention covariates Xit. Inference uses Hamiltonian Monte Carlo, with posterior distributions quantifying all parameter uncertainty. ", "findings": "The model application indicates that improvements are predominantly driven by operational practices, accounting for an estimated 65% (95% credible interval: 58% to 71%) of the explained variance in yield gains. The technological adoption covariate showed a positive but less precisely estimated effect, with its posterior distribution indicating a non-zero probability of a negligible impact. ", "conclusion": "The proposed model provides a statistically rigorous diagnostic tool that successfully disentangles the sources of manufacturing yield improvement while formally accounting for uncertainty and hierarchical data structures. ", "recommendations": "Manufacturing analysts should adopt hierarchical modelling approaches to avoid ecological fallacies. Policy support should prioritise programmes that enhance operational best practices, as these are identified as the most consistent lever for yield improvement. ", "key words": "Bayesian inference, hierarchical modelling, manufacturing diagnostics, yield improvement, industrial engineering, sub-Saharan Africa", "contribution statement": "This