A MICRO-ECONOMETRIC ANALYSIS OF POVERTY: EVIDENCE FROM TUNISIA

Résumé

This dissertation is comprised of three empirical essays dealing with some aspects of poverty analysis. The first one combines both unidimensional and multi-dimensional approaches to analyze poverty. First of all, through the Foser-Greer-Thorbeek index we present a first analysis of the Tunisian poverty profile. Secondly, we estimate a classic logit model using the same variables. Finally, we introduce a new approach based on fuzzy logic (Zadeh, 1975) able a comparison between logit model and fuzzy approach by building a membership function based on the logistic function. The execution of this models require the choice of well-being indicator, generally, the income or he consumption. In this PhD thesis, we favor the reel per capita consumer expenditures of households. The principal reasons of this choice are twofold: The first one, income is hardly observable due to a bad quality of data. The second reason is that consumer expenditures are stable over time comparing to income. Finally, we test the robustness of our work and he choice of the poverty line using sensitivity test and the concept of stochastic dominance. The second essay, focuses on the determinants of household welfare in Tunisia. Welfare is measured by real per capita household expenditure (income). We examine the inequality gap in consumption and education in rural and urban areas. Our empirical analysis relies on the Tunisian Living Standards Surveys of 2010. Our paper makes two original empirical contributions to the literature compared to previous. In order to analyze the gap between the two regions, we apply the RIF- regression advanced by Firpo et al. (2011) and the new counterfactual decompositions developed by Chernozhukov et al. (2013). We also use censored and uncensored quantiles regressions for disaggregate consumption expenditures. This study offers another analysis of the influence of some indicators such as health, food expenditures and especially education on regional disparities in developing countries. It also examines the relationship between these indicators and development disparities across the two areas. The third essay of this thesis aims to explain the determinants of poverty dynamics. Unfortunately in Tunisia, as in most developing countries, panel data are not available. We introduce a new approach able us to execute a dynamic analysis of poverty using potential outcomes approach and multiple imputation. In fact, the study based on two steps. First of all, we use a Bayesian algorithm which combines the two approach, cited previously, in order to penalize the data base. In the second step, we assess poverty dynamics through two different methods. On the one hand, we use static measurement of poverty through the Jalan-Ravallion and the Equally-Distributed Equivalent Poverty Gaps approaches. On the other hand, we execute an econometric procedure by estimating a recursive bivariate probit model. This model allows to treat the initial condition of poverty as endogenous which avoids a probable bias-selection.

Citer ce document

Accès au document

Ce document est hébergé sur son dépôt institutionnel d'origine.

Auteur(s)

Statistiques

Consultations : 2

Téléchargements : 0