Methodological Evaluation and Panel-Data Estimation of Power-Distribution System Yield in Ethiopia, 2000–2026

The operational yield of power-distribution systems in developing economies is often constrained by ageing infrastructure and methodological inconsistencies in performance evaluation, leading to significant technical and commercial losses. This study aims to develop and apply a robust panel-data econometric framework to evaluate the methodological soundness of yield assessments and to estimate the determinants of system yield improvement for a national power utility. A balanced panel dataset of operational districts was constructed. The core specification is a fixed-effects model: $Y_{it} = \alpha_i + \beta_1 X_{1,it} + ... + \beta_k X_{k,it} + \epsilon_{it}$, where $Y_{it}$ is the distribution yield. Robust standard errors were clustered at the district level to account for serial correlation. The methodological evaluation revealed systematic over-reporting in historical yield data. Estimation indicates that targeted infrastructure renewal explains approximately 40% of the observed yield improvement, with a coefficient of 0.15 (95% CI: 0.11, 0.19). Operational practices and load density were also significant positive determinants. The proposed panel-data methodology provides a more reliable framework for yield estimation than prior aggregate approaches. The findings confirm that strategic capital investment in distribution assets is a primary driver for reducing losses. Utilities should adopt panel-data methodologies for internal performance auditing. Investment planning should prioritise the replacement of obsolete feeder lines and substation components, informed by the empirical elasticities estimated. distribution losses, fixed-effects model, infrastructure investment, panel data, power system yield, technical losses This paper provides the first application of a district-level panel model to decompose the drivers of power-distribution yield in the national context, introducing a method to correct for systematic reporting bias in utility data.