On the power of indicators: how the choice of the fuel poverty measure affects the identification of the target population

We propose a critical analysis of fuel poverty indicators and demonstrate that choosing a given fuel poverty indicator and, in particular, its threshold level, is central to the identification of the fuel-poor population. First, we conducted an inter-indicator analysis to show how profiles of fuel-poor households vary depending on the indicator selected. More specifically, after identifying groups of affected house- holds using a set of objective and subjective measures, we designed a multidimensional approach based on a combination of complementary methods, namely, a multiple correspondence analysis and a hierarchical and partitioning clustering analysis to analyse their characteristics. Through this framework, we highlight the difficulty of identifying a “typical profile” for fuel-poor households because of the significant variability in their characteristics and we show that the composition of the population depends on the choice of the indicator. Second, we applied an intra-indicator analysis using two objective expenditure-based indicators with thresholds. In particular, we conducted a sensitivity analysis based on a logit model including variables describing household and dwelling characteristics. We show that the profiles of fuel-poor households as well as the drivers of fuel poverty vary considerably with the chosen threshold level. Given these findings, we stress the need to review how we currently rely on conventional fuel poverty indicators to identify affected groups and give some recommendations.