Average marginal effects from logistic regressions on belonging to the top 10% and top 20% of total energy consumers. Coefficients represent the change in probability of belonging to the top 10% / 20% of energy users compared to the base category. Base categories are: bottom income quintile; age group 35-64; male household representative; household representative not employed; no one in household has higher education; rural location. The significance of coefficients is assessed with the z statistic, and the significance of the overall model with the Wald Chi-square statistic. Data: EU HBS 2015, 2010; Exiobase 3.7. The calculations exclude the top and bottom 1% energy consumption and income outliers and countries for which information in education and employment status is missing (Supplementary Table 5 ). n = 208,539 observations.
