Abstract
Theoretical and empirical studies have recently adopted a multidimensional concept of poverty. There is considerable debate about the most appropriate degree of multidimensionality to retain in the analysis. In this work we add to the received literature in two ways. First, we derive indicators of multiple deprivation by applying a particular multivariate statistical technique, the non-linear principal component analysis (NLPCA), which overcomes traditional limits of many of the mostly used methodologies for poverty measurement. Second, on the basis of the aforementioned indicators, we provide an accurate identification of the poor in Italy by analyzing deprivation both as a distinct phenomenon in different life domains and as a single multidimensional concept. The main determinants of poverty in Italy are then investigated by estimating logit regressions and an ordered probit model. Our empirical analysis is based on data from the Italian component of European Statistics on Income and Living Conditions (EU-SILC-2004).
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Mack and Lansley (1985) define poverty as an “enforced lack of socially-perceived necessities”. Their approach aims at distinguishing whether the lack of a commodity results from resource constraints or individual preferences. Only goods people would like to possess, but they cannot afford should be properly considered as reflecting deprivation, in order to exclude lifestyle preferences from the concept of poverty.
Bourguignon and Chakravarty (2003), for instance, support the need to maintain a multidimensional perspective and argue that existing attempts of aggregating various attributes into a single poverty index are a mere redefinition of the concept of poverty, which then essentially remains a one dimensional concept. On the other hand, a single indicator can be more easily understood by the general public (see Brandolini 2008 and the references here quoted).
Factorial techniques derive the latent structure from the data on the basis of the Pearson correlation-matrix of the original variables. When categorical and binary variables are used, “the contingency table of the variables is used in lieu of the correlation matrix and the assumptions of the factor analysis model based on the Pearson correlation matrix of the variables would be violated” (Kamanou 2005).
NLPCA has been carried out by using the program CATPCA (Meulman et al. 2004a).
An example is the number of missing values in the variable “Financial burden of the total housing cost”, which describes the extent to which housing costs are a financial burden to the household. Those who do not have housing costs, for instance because they do not possess a house, report a missing value on this variable.
We consider that respondents experience health problems when they answer “bad” and “very bad” to questions relating to their general health status, if they suffer from chronic illness and if they have been limited in their usual activities and in job search because of health problems.
It should be recognized that, as the variables’ selection is data driven, different dimensions could be obtained for different countries or for the same country over time, by raising problems of results’ consistency and comparability. As far as the first point is considered, however, Nolan and Whelan (2010) show that employing a common set of dimensions across EU countries performs as well as allowing the dimensions to differ, according to various measures of goodness of fit. Further, we believe that exploiting correlations among data allows us to capture country-specific consumption patterns that should be included in the analysis, specifically when deprivation is intended as a relative concept.
We have preserved only the first principal component by each analysis as we were looking for the best indicator representing the latent dimension of deprivation underlying each group of variables. Recall that the first component explains the most of the variability among observed variables.
We have derived an even more comprehensive indicator, which aggregates over dimensions of deprivation and the equivalent disposable income through PCA. This synthetic measure expresses a wider concept of standard of living, being “the result of an accumulation of deprivation in both resources and the way of life” (Ringen 1988; p. 173). Results from logit regressions for this indicator are not significantly different from the outcomes in Table 3. They can be provided on request.
It is worth to note that the complexity of adopting endogenous weights for deriving the composite index is more than counterbalanced by the reduced subjectivity, implicit instead in aggregation methods based on expert panel’s judgments (which are plagued by both conceptual and methodological problems, see Dewilde 2004).
As noted above, deprived people are those belonging to the bottom 20 per cent of each score distribution. Outcomes from logit regressions, however, have been proved to be robust to the choice of alternative thresholds, corresponding to 15 and 25 per cent. Poor individuals are identified as those having an income below the poverty line (corresponding to the 60 per cent of the median equivalent income). Since the dependent variable is binary, a binomial logit model is used where the parameters are estimated through the method of the maximum likelihood (see Maddala 2002).
Differences in deprivation probabilities across individuals are evaluated by assuming, as reference category, a man, aged 45, living in the North of Italy, graduate and self-employed, having a permanent job in the public administration sector and married with one child.
According to our estimates, 52.7 per cent of total population is deprived at least in one dimension. Specifically, almost 30 per cent is deprived in only one dimension, while 13.4 per cent in two dimensions. These percentages fall to 7 per cent for the deprived in three dimensions, below 3 per cent in four and 0.5 per cent in five dimensions.
References
Atkinson, A. B., Cantillon, B., Marlier, E., & Nolan, B. (2002). Social indicators: The EU and social inclusion. Oxford: Oxford University Press.
Bourguignon, F., & Chakravarty, S. R. (2003). The measurement of multidimensional poverty. Journal of Economic Inequality, 1, 25–49.
Brandolini, A. (2008). On applying synthetic indices of multidimensional well-being: Health and income inequalities in selected EU Countries. Bank of Italy Temi di discussione (Working Paper), n.668.
Calandrino, M. (2003). Low income and deprivation in British families. London: Department for Work and Pensions Working Paper No 10.
Callan, T., Nolan, B., & Whelan, C. T. (1993). Resources, deprivation and the measurement of poverty. Journal of Social Policy, 22, 141–172.
Cannari, L. & D’Alessio, G. (1993). Non reporting and under reporting behaviour in the Bank of Italy’s survey of Household Income and Wealth. In Proceeding of ISI, 49th Session (pp. 395-412). Firenze: ISI.
Cheli, B., & Lemmi, A. (1995). A “totally” fuzzy and relative approach to the multidimensional analysis of poverty. Economic Notes, 24, 115–134.
Costantini, P., Linting, M., Porzio, G. C. (2008). Mining performance data through non-linear PCA with optimal scaling. Applied Stochastic Models in Business and Industry.
D’Alessio, G., & Faiella, I. (2002). Nonresponse behaviour in the Bank of Italy’s Survey of Household Income and Wealth. Banca d’Italia, Temi di discussione, n. 462.
D’Amuri, F., & Fiorio, C. V. (2006). Tax evasion in Italy: An analysis using a tax-benefit microsimulation model. The ICFAI Journal of Public Finance, 19–37.
Dewilde, C. (2004). The multidimensional measurement of poverty in Belgium and Britain: A categorical approach. Social Indicators Research, 68, 331–369.
Ferro Luzzi, G., Fluckiger, Y., & Weber, S. (2006). A cluster analysis of multidimensional poverty in Switzerland, Cahier de recherche. CRAG—Centre de Recherche Appliquée en Gestion, N°HES-SO/HEG-GE/C–06/1/2–CH.
Gifi, A. (1990). Nonlinear multivariate analysis. New York: Wiley.
Guio, A. C. (2009). What can be learned from deprivation indicators in Europe? Eurostat, Methodologies and Working Papers, presented at the Indicator Subgroup of the Social Protection Committee.
Halleröd, B. (1995). The truly poor: Direct and indirect consensual measurement of poverty in Sweden. European Journal of Social Policy, 5(2), 111–129.
Kamanou, G. (2005). Methods of factor analysis for ordinal categorical data and application to poverty data. Paper presented at the International Conference. “The many dimensions of poverty”, Brasilia, Brazil, 29–31 August.
Layte, R., Maître, B., Nolan, B., & Whelan, C. T. (2000). Targeting poverty: Lessons from monitoring Ireland’s national anti-poverty strategy. Journal of Social Policy, 29, 553–575.
Linting, M., Meulman, J. J., Groenen, P. J., & van der Koojj, A.. J. (2007). Nonlinear principal components analysis: Introduction and application. Psychological Methods, 12(3), 336–58.
Lucchini, M., Pisati, M., & Schizzerotto, A. (2007). Stati di deprivazione e di benessere nell’Italia contemporanea. Un’analisi multidimensionale. In A. Brandolini & C. Saraceno (Eds.), Povertà e benessere. Una geografia delle disuguaglianze in Italia (pp. 271–303). Il Mulino, Bologna.
Mack, J., & Lansley, S. (1985). Poor Britain. London: George Allen and Unwin.
Maddala, G. S. (2002). Introduction to econometrics. England: Wiley.
Martínez, R., & Ruiz-Huerta J. (2000). Income, multiple deprivation and poverty: An empirical analysis using Spanish data. 26th IARIW General Conference. Cracow.
McKay, S. & Collard, S. (2003). Developing deprivation questions for the Family Resources Survey. IAD Research Division Working Paper No. 13, Department for Work and Pensions, London.
Meulman, J. J., Heiser, W. J., & SPSS. (2004a). SPSS categories 13.0. Chicago: SPSS Inc.
Meulman, J. J., Van der Kooij, A. J., & Heiser, W. J. (2004b). Principal components analysis with nonlinear optimal scaling transformations for ordinal and nominal data. In D. Kaplan (Ed.), The Sage handbook of quantitative methodology for the social sciences (pp. 49–70). London: Sage.
Nolan, B., & Whelan, C. (2010). Using non-monetary deprivation indicators to analyze poverty and social exclusion: Lessons from Europe?. Journal of Policy Analysis and Management, 29, 305–325.
Nolan, B., & Whelan, C. T. (1996). Resources, deprivation and poverty. Oxford: Clarendon Press.
Pérez-Mayo, J. (2005). Identifying deprivation profiles in Spain: a new approach. Applied Economics, 37, 943–955.
Perry, B. (2002). The mismatch between income measures and direct outcome measures of poverty. Social Policy Journal of New Zealand, 19, 101–127.
Ringen, S. (1988). Direct and indirect measures of poverty. Journal of Social Policy, 17, 351–365.
Townsend, P. (1979). Poverty in the United Kingdom. Penguin: Harmondsworth.
Tsakloglou, P., & Papadopoulos, F. (2002). Aggregate level and determining factors of social exclusion in twelve European countries. Journal of European Social Policy, 12(3), 211–225.
Whelan, C. T. (2007). Understanding the implications of choice of deprivation index for measuring consistent poverty in Ireland. Working Paper No.181, Economic and Social Research Institute, Dublin.
Whelan, C. T., Maitre, B., & Nolan, B. (2006). Measuring consistent poverty in Ireland, with EU-SILC data. Working Paper No. 165, Economic and Social Research Institute, Dublin.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Coromaldi, M., Zoli, M. Deriving Multidimensional Poverty Indicators: Methodological Issues and an Empirical Analysis for Italy. Soc Indic Res 107, 37–54 (2012). https://doi.org/10.1007/s11205-011-9825-6
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11205-011-9825-6