Predicting energy poverty in Greece through statistical data analysis.

A comprehensive statistical analysis of energy poverty indicators is undertaken in the present paper, in an attempt to further understand the roots and results of the problem in Greece. Specifically, time-series data sets were analysed using various objective indicators, i.e. 10%, 2M, 2M EXP, M/2, M/2 EXP, as well as subjective indicators. Chi-square tests of Independence were performed and binary logistic regression models were developed to predict energy poverty (indicators of 10%, 2M and M/2), based on critical socio-economic factors. The logit model based on the 10% indicator presented the highest performance, reaching 32%. According to this model, the types of households mostly exposed to energy poverty were single families with dependent children and households located in Macedonia, increasing the relative probability of energy poverty by 7.0 and 6.5 times per unit, respectively. The outcomes derived can help policy-makers towards designing more targeted policies for tackling energy poverty in Greece. [ABSTRACT FROM AUTHOR]