Abstract
This article has three goals. First, we wish to compare three multidimensional approaches to poverty and check to what extent they identify the same households as poor. Second, we aim at better understanding the determinants of poverty by estimating logit regressions with five categories of explanatory variables: size of the household, age of the head of the household, her gender, marital status and status at work. Third, we introduce a decomposition procedure proposed recently in the literature, the so-called Shapley decomposition, in order to determine the exact marginal impact of each of the categories of explanatory variables. Our empirical analysis is based on data made available by the European Community Household Panel (ECHP). We used its third wave and selected five countries: Belgium, France, Germany, Italy and Spain.
Acknowledgements
A preliminary version of this article was presented at the 2004 conference of the International Association for Research in Income and Wealth (IARIW) in Cork, Ireland. We thank Gordon Anderson and other participants for their useful comments. This article was revised while J. Silber was visiting FEDEA, Madrid, and he thanks this institution for its warm hospitality. C. D'Ambrosio thanks MIUR (Prin 2007) for financial support.
Notes
1 More details are given in Deutsch and Silber (Citation2005) who, in this article, also used the so-called efficiency analysis approach.
2 For a discussion on the choice of indicators see, among others, Pérez-Mayo (Citation2005).
3 Similar results were obtained when computing the correlation coefficients between any two approaches.