Vulnerability and multiple deprivation perspectives on economic exclusion in Europe: A latent class analysis

ABSTRACT

In this paper we address issues relating to vulnerability to economic exclusion and levels of economic exclusion in Europe. We do so by applying latent class models to data from the European Community Household Panel for thirteen countries. This approach allows us to distinguish between vulnerability to economic exclusion and exposure to multiple deprivation at a particular point in time. The results of our analysis confirm that in every country it is possible to distinguish between a vulnerable and a non-vulnerable class. Association between income poverty, life-style deprivation and subjective economic strain is accounted for by allocating individuals to the categories of this latent variable. The size of the vulnerable class varies across countries in line with expectations derived from welfare regime theory. Between class differentiation is weakest in social democratic regimes but otherwise the pattern of differentiation is remarkably similar. The key discriminatory factor is life-style deprivation, followed by income and economic strain. Social class and employment status are powerful predictors of latent class membership in all countries but the strength of these relationships varies across welfare regimes. Individual biography and life events are also related to vulnerability to economic exclusion. However, there is no evidence that they account for any significant part of the socio-economic structuring of vulnerability and no support is found for the hypothesis that social exclusion has come to transcend class boundaries and become a matter of individual biography. However, the extent of socio-economic structuring does vary substantially across welfare regimes. Levels of economic exclusion, in the sense of current exposure to multiple deprivation, also vary systematically by welfare regime and social class. Taking both vulnerability to economic exclusion and levels of exclusion into account suggests that care should be exercised in moving from evidence on the dynamic nature of poverty and economic exclusion to arguments relating to the superiority of selective over universal social policies.

Keywords:

• economic exclusion
• multidimensionality
• vulnerability
• multiple disadvantage
• latent class analysis

We would like to thank the anonymous referees, Richard Breen and Pasi Moisio and participants at an ESRI seminar for comments on an earlier version of this paper. The paper is based on analysis of the European Community Household Panel survey. The data are used with the permission of Eurostat. The authors are entirely responsible for the analysis and interpretations contained in the paper.

Notes

1. See Hallerod, Kangas and Ritakallio (1998), Muffels et al. (1992), Nolan and Whelan, (1996b).

2. For a recent discussion of applications of latent class models see McCutcheon and Mills (1998).

3. The first adult in a household is given a value of 1, any additional adult a value of 0.5 and every child aged under 14 a value of 0.3.

4. We have chosen the 70 per cent rather than the 60 per cent threshold because, since the economically excluded will by definition be a sub-set of the income poor, we sought to avoid defining economic exclusion in an overly restrictive fashion so that conclusions relating to the determinants of economic exclusion in comparison with conventional income measures are not simply an artefact of the relative sizes of the groups identified.

5. The high value of this coefficient reflects the fact that the dimensionality of deprivation is uniform across countries in the ECHP data set (Whelan et al. 2001).

6. We should make clear that we are not attempting to develop an alternative measure of poverty based on deprivation indicators. To do so would require that we address issues such as the appropriate threshold and the manner in which income and deprivation might be combined to construct an alternative indicator.

7. The reference person in the household responds to the household questionnaire.

8. It should be stressed though that our results are set out in this fashion purely for purposes of communication and only one income indicator comprising four categories enters into the analysis.

9. In all models employing class and employment status we allow for within latent class effects on income poverty and economic strain.

10. In λEM syntax where A is income poverty, B deprivation and C economic strain, S social class and X the latent variable, the model is A∣XS B∣X C∣XS. Full details of the model fits are presented in Appendix Table A2. The percentage of cases misclassified varies between 0.6 and 2.7 per cent for social class and 1.1 and 2.7 per cent for employment status.

11. Full details of model fits are provided in Appendix Table A3. For all variables except gender the basic latent class model was fitted. For the latter a within class effect on income was permitted. For gender the percentage misclassified ranges between 1.0 and 1.9 per cent, for divorce between 1.4 and 1.9 per cent and for lone parents between 1.1 and 2.2 per cent.

12. See Paugam et al. (200: 269), Gallie et al. (2003) Tsakloglou and Papadopoulous (2002), Whelan et al. (2002).

13. See also Tsakloglou and Papadopoulous (2002) and Whelan et al. (2002).