Ph.D. thesis

Title: Poverty, Inequality and Redistribution: An Econometric Approach
Period: 09/2015 - 09/2019
Role: doctoral student
Director: Michel Lubrano (Aix-Marseille School of Economics)
Hosting institution: Aix-Marseille School of Economics, Aix-Marseille University
Financing institution: French Ministry of Higher Education and Research
Jury: Christian Robert (Université Paris-Dauphine), Frank Cowell (London School of Economics), Stephen P. Jenkins (London School of Economics), Habiba Djebbari (Aix-Marseille School of Economics), Emmanuel Flachaire (Aix-Marseille School of Economics), Michel Lubrano (Aix-Marseille School of Economics)

My thesis develops statistical models for representing distributional changes and quantifying the uncertainty of estimation using econometric methods and Bayesian inference. These models are used to examine events which may have had profound consequences on the income distribution raising concerns about poverty and inequality while fostering debates on implications for redistributive policies. For instance, the first chapter entitled Bayesian inference for TIP curves: an application to child poverty in Germany examines the evolution of child poverty in Germany over the period 2002-2011 using TIP curves. TIP curves are cumulative poverty gap curves used for representing the three different aspects of poverty: incidence, intensity and inequality. After having related these curves to a broad class of ethical judgements, we provide Bayesian inference for TIP curves based on a semiparametric representation of the income distribution using a mixture of log-normal densities. We treat specifically the questions of zero-inflated income data and of survey weights, which are two important issues in survey analysis, and we provide statistical tests for TIP dominance. Using GSOEP data, we find that child poverty increased between 2002 and 2006 while this was the opposite after the Hartz reforms (2003-2005) with a decrease of child poverty and a greater convergence between the East and the West parts of Germany.

In another chapter of my thesis entitled Bayesian inference for parametric growth incidence curves, we develop Bayesian estimation and inference for comparing distributional changes using Growth Incidence Curves. We demonstrate that the usual distribution-free estimator of these curves behaves erratically with usual sample sizes leading to problems in the tails. Then, we propose a series of parametric models in a Bayesian framework to obtain more reliable conclusions. A first solution consists in modelling the underlying income distribution using simple densities for which the quantile function has a closed analytical form. This solution is extended by considering a mixture model for the underlying income distribution. However, in this case, the quantile function is semi-explicit and has to be evaluated numerically using an algorithm. The last solution consists in adjusting directly a functional form for the Lorenz curve and deriving its first order derivative to find the corresponding quantile function. We compare these models by Monte-Carlo simulations and using real data, and we find that even the most simple parametric models tend to be significantly more efficient than their distribution-free counterparts when samples have less than 10 000 observations.

Building on this methodology, we empirically examine the role of social norms on labour force participation and wage inequality using a natural experiment in the chapter entitled Bayesian Inference for Distributional Changes: The Effect of Western TV on Wage Inequality and Female Participation in Former East Germany. Precisely, we exploit an exogenous variation of TV exposure during the Cold War in East-Germany to circumvent the endogeneity of social norms. For comparing distributional changes occurring in the treatment and the control groups after the reunification, we provide Bayesian estimation and inference for the growth incidence curve of Ravallion and Chen (2003) and the poverty growth curve of Son (2004) based on a Mincer-type model of wages with sample selection and propose a set of tests related to stochastic dominance criteria. Using the GSOEP, we find evidences that Western television has significantly increased wage inequality among males while it has significantly affected female labour participation and led the less productive females to drop out from the labour market, hiding thus a large increase in wage inequality among females. The idea is that Western TV broadcasts have conveyed Western norms fostering risky behaviours on the labour market, but also, shaping individual incentives (e.g. to invest in education, to choose a career, to migrate, to marry and make children) affecting, in fine, wage inequality and its persistence over time. This paper contributes to the literature on treatment effects by providing an heterogeneous treatment effect model while developing statistical inference for it. In addition, this paper contributes to the vast literature on labour by identifying social norms and culture as being an important determinant of wage inequality.

Finally, I investigate the determinants of the non-take-up of social assistance. Indeed, this phenomenon has been receiving increased attention among policy makers as it would apparently underpin the effectiveness of public intervention in alleviating poverty. In my last thesis chapter entitled How family transfers crowd-out social assistance in Germany, I empirically examine whether receipt of family transfers – generally unobserved by the welfare office - affects the decision to take-up social assistance in Germany between 2009 and 2011. I exploit the follow-up of households in the GSOEP to reconstruct family links and, after having completed survey data by micro-simulation, I estimate a model of welfare participation with endogenous private transfers and sample selection of the instruments. I find that 20% of the non-take-up rate is attributable to the monetary substitution of family transfers lowering welfare program costs. However, results indicate that social assistance would have been more effective in alleviating poverty and its intensity than family transfers. This paper contributes to the growing literature on the determinants of non-take-up by identifying a new factor that partly explains the important non-take-up rate in Germany, but it also contributes to the debate on targeting versus universalism of welfare benefits by showing that even if the screening process of the welfare office underestimates household means, those households are less likely to claim social assistance.