Hidden Energy Poverty, the Dwelling Envelope, and the Limits of Income-Based Targeting: Household Evidence from the Coal Phase-Out Region of Western Macedonia, Greece

Coal phase-out regions concentrate the distributional risks of the energy transition, yet the tools used to identify energy-poor households in these territories still rely mainly on income and welfare criteria. This study asks whether such criteria can identify the households that suffer. Drawing on a survey of 706 households across six municipalities of Western Macedonia, Greece, the core territory of the national lignite phase-out, we validate a four-item Thermal Stress Index (polychoric ω = 0.885; loadings = 0.71–0.88; no differential item functioning by gender, income, or survey wave) and test nine hypotheses and one descriptive benchmark on prevalence, mechanisms, typologies, and targeting. Winter thermal inadequacy reaches 22.9%, which sits above the 19.0% national EU-SILC figure, reported as descriptive context. Dwelling energy features dominate all socioeconomic predictors of severe thermal stress (pseudo-R-squared 0.482 against 0.024), and no direct tenure association remains once envelope quality is included, while the protective association of income operates primarily through dwelling quality and is not observed across the range of inefficient dwellings. Latent class analysis identifies a hidden energy poverty class, 19.5% of households with severe experiential deprivation, above-average income, and no payment problems. A machine learning targeting audit shows that administrative criteria alone identify severely stressed households and show no discriminative capacity for identifying severely stressed households (AUC = 0.517), whereas adding dwelling and financial-strain information raises discrimination to an area under the curve of 0.938. Finally, in this sample, income-based screening is weakly associated with the households reporting severe thermal deprivation.