<p>This study is concerned with the contradiction in liberal-democratic society between the persistence of group inequalities on the one hand, and the firm commitment to individualism on the other. Individualism requires that group inequalities be absent, and that "particularistic policies" be avoided. Consequently, liberal governments tend to either ignore the issue of group inequality or abandon the liberal framework in order to deal with it. By examining the communal structure in pre-liberal societies, and the nature of modernday support for particularism, it has been found that the norm of universalism is to be preferred. The Indians in Canada have been used as an example illustrating the effects of a particularistic policy when applied in a liberal context. An analysis of the Indian case suggests that legal group differentiation does not reduce group inequality even within a liberal framework. The argument is advanced that a "liberal" solution to this problem is possible. Since liberal theory has been mute on this point, a solution is worked out by exploring what is consistent with classical liberal foundations as laid down by John Locke. In general, group differentiation may be practised (sic) in a qualified way provided it remains the exception, and is not established as a new norm.</p> / Thesis / Master of Arts (MA)

Essa dissertação tem como objetivo apresentar uma medida de desigualdade distinta das usualmente utilizadas na literatura de desigualdade. Tal medida, aqui denominada índice J, além de possuir a característica de ser simétrica também possibilita conduzir análises distintas daquelas feitas com os índices usualmente utilizados. Para avaliar a aplicabilidade do índice J foram utilizados dados provenientes da PNAD referente aos anos de 2007, 2011 e 2015 . Apenas dados referentes ao estado de São Paulo foram incluídos. Além de avaliar a evolução da desigualdade no período, decomposição intra e entre grupos foram conduzidas. Testes de hipótese para a desigualdade entre grupos, uma das possibilidades apresentada pelo índice J, foram conduzidos. Também foi avaliado como a presença de erros de medidas não condicionais à renda influenciaria o resultado. Os resultados apresentados pelo índice J apontam para uma diminuição da desigualdade de renda no período analisado, sendo que a variável educação se apresentou como a característica com maior capacidade de explicar a desigualdade total a partir da desigualdade entre grupos, dentre as variáveis analisadas. A simulação de Monte Carlo conduzida para o teste de hipótese para desigualdade entre grupos também apontou à variável educação como a mais provável de gerar desigualdade. A presença de erros de medida não condicionais à renda não influenciou no resultado final do índice J, porém a simulação de tais erros contribuiu para amenizar o problema de discretização dos dados provenientes da PNAD. O índice J apresentou-se como uma alternativa viável aos índices usualmente utilizados na literatura de desigualdade, possibilitando algumas análises distintas e que podem auxiliar no estudo da desigualdade de renda. / This dissertation aims to present a measure of inequality distinct from those usually used in the inequality literature. Such a measure, here called the J index, besides having the characteristic of being symmetrical also makes it possible to conduct analyzes different from those made with the indexes usually used. To evaluate the applicability of the J index, data from the PNAD for the years 2007, 2011 and 2015 were used in the analysis. Only data referring to the state of São Paulo were included. In addition to assessing the evolution of inequality in the period, within and between group decomposition were conducted. Hypothesis tests for the inequality between groups, one of the possibilities presented by the index J, were conducted. It was also evaluated how the presence of errors of measures not conditional to the income would influence the result. The results presented by the index J point to a decrease in income inequality in the analyzed period, with the education variable being the characteristic with the greatest capacity to explain the total inequality from the inequality between groups, among the variables analyzed. The Monte Carlo simulation conducted for the hypothesis test for inequality between groups also pointed to the education variable as the most likely to generate inequality. The presence of measurement errors did not influence the final result of the J index, but the simulation of such errors contributed to soften the problem of discretization of PNAD data. The index J was presented as a viable alternative to the indexes usually used in theliterature of inequality, allowing some different analyzes and that can help in thestudy of income inequality.

Essa dissertação tem como objetivo apresentar uma medida de desigualdade distinta das usualmente utilizadas na literatura de desigualdade. Tal medida, aqui denominada índice J, além de possuir a característica de ser simétrica também possibilita conduzir análises distintas daquelas feitas com os índices usualmente utilizados. Para avaliar a aplicabilidade do índice J foram utilizados dados provenientes da PNAD referente aos anos de 2007, 2011 e 2015 . Apenas dados referentes ao estado de São Paulo foram incluídos. Além de avaliar a evolução da desigualdade no período, decomposição intra e entre grupos foram conduzidas. Testes de hipótese para a desigualdade entre grupos, uma das possibilidades apresentada pelo índice J, foram conduzidos. Também foi avaliado como a presença de erros de medidas não condicionais à renda influenciaria o resultado. Os resultados apresentados pelo índice J apontam para uma diminuição da desigualdade de renda no período analisado, sendo que a variável educação se apresentou como a característica com maior capacidade de explicar a desigualdade total a partir da desigualdade entre grupos, dentre as variáveis analisadas. A simulação de Monte Carlo conduzida para o teste de hipótese para desigualdade entre grupos também apontou à variável educação como a mais provável de gerar desigualdade. A presença de erros de medida não condicionais à renda não influenciou no resultado final do índice J, porém a simulação de tais erros contribuiu para amenizar o problema de discretização dos dados provenientes da PNAD. O índice J apresentou-se como uma alternativa viável aos índices usualmente utilizados na literatura de desigualdade, possibilitando algumas análises distintas e que podem auxiliar no estudo da desigualdade de renda. / This dissertation aims to present a measure of inequality distinct from those usually used in the inequality literature. Such a measure, here called the J index, besides having the characteristic of being symmetrical also makes it possible to conduct analyzes different from those made with the indexes usually used. To evaluate the applicability of the J index, data from the PNAD for the years 2007, 2011 and 2015 were used in the analysis. Only data referring to the state of São Paulo were included. In addition to assessing the evolution of inequality in the period, within and between group decomposition were conducted. Hypothesis tests for the inequality between groups, one of the possibilities presented by the index J, were conducted. It was also evaluated how the presence of errors of measures not conditional to the income would influence the result. The results presented by the index J point to a decrease in income inequality in the analyzed period, with the education variable being the characteristic with the greatest capacity to explain the total inequality from the inequality between groups, among the variables analyzed. The Monte Carlo simulation conducted for the hypothesis test for inequality between groups also pointed to the education variable as the most likely to generate inequality. The presence of measurement errors did not influence the final result of the J index, but the simulation of such errors contributed to soften the problem of discretization of PNAD data. The index J was presented as a viable alternative to the indexes usually used in theliterature of inequality, allowing some different analyzes and that can help in thestudy of income inequality.

Cette thèse a pour objet de fournir une méthode d'évaluation simple de l'inégalité et de la mobilité des revenus. Nous utilisons une méthode de décomposition récemment introduite dans la littérature et aujourd'hui connue sous le nom de emph{décomposition faible}, pour analyser la répartition des écarts de revenus entre les individus appartenant à un même sous-groupe et ceux situés dans des sous-groupes distincts. Nous nous intéressons en particulier aux contextes dans lesquels il peut être pertinent de faire appel à un tel outil de décomposition. Nous approfondissons nos recherches en précisant les propriétés normatives des indicateurs compatibles avec le schéma de décomposition faible en sous-groupes. Nous énonçons notamment des principes de transferts intra- et intergroupes afin de fournir des moyens d'actions représentatifs des préférences d'un décideur politique. Les fondements axiomatiques de cette propriété de décomposition en sous-groupes sont également abordés. Des fonctions de pondération plus générale mais néanmoins conformes aux schémas de décomposition en sous-groupes usuels sont introduites. Cette généralisation nous permet de caractériser axiomatiquement des mesures d'inégalité à 2 paramètres que nous qualifions de (alpha,delta)-Gini. Ces 2 paramètres alpha et delta permettent de capter l'ensemble des points de vue politique des décideurs publics (d'extrême gauche à extrême droite). Nous montrons de plus que l'application de la décomposition faible en sous-groupes se généralise aisément à l'étude de la mobilité des revenus. Nous proposons ainsi une adaptation de la propriété de décomposition en sous-groupes dans un cadre bidimensionnel et caractérisons 2 classes de mesures d'inégalité de croissance ajustée et de mouvements de revenus cohérentes avec cette condition. Diverses études empiriques sont également menées afin d'illustrer les différentes notions développées dans cette thèse. / This Ph.D. Dissertation aims at providing accurate and simple tool to evaluate income inequality and mobility. Our work relies on a subgroup decomposition property recently introduced in the literature as the $``$emph{weak decomposition}$"$ to break down total disparities into within-group and between-group disparities. A particular interest is given to the context in which subgroup decomposition can be applied. We introduce within-group and between-group transfer principles that can be modulated according to a decision maker's preferences. The axiomatic basis of the subgroup decomposition property are also dealt with. Some general weighing functions are introduced to caracterize a two-parameters class of inequality measures denoted denoted (alpha,delta)-Gini. The parameters alpha and delta allow capturing the set of decision's maker point of view (from extrem leftist to extreme rightist). Furthermore, we demonstrate the the application of the weak decomposition can easily be extended to income mobility. We propose an adaptation of the subgroup decomposition property to a bimensional framework and we characterize 2 classes of inequality adjusted growth and income movements measures consistent with such a property. Various empirical studies are also carried out to illustrate the various developed concepts.

Diese Dissertation befasst sich mit Arbeitsmarkterfolg und Konsum sozioökonomischer Gruppen. Die ersten zwei Kapitel untersuchen konjunkturelle Auswirkungen auf Arbeitsmärkten für Hoch- und Niedrigqualifizierte. Zunächst wird ein qualifikationsspezifisches Konjunkturmodell mit Suchkosten entworfen. Es zeigt, dass imperfekte Substitution zwischen hoch- und niedrigqualifizierter Arbeit ein Grund für Veränderungen auf den Teilmärkten ist. Gemeinsam mit qualifikationsneutralen und -verzerrten Technologieschocks ist das Modell in der Lage, fallende Beveridge-Kurven zu generieren. Das zweite Kapitel erweitert diesen Ansatz um eine Verbindung zwischen qualifikationsabhängigen Arbeitsmärkten mit endogenen Investitionen in Humankapital. Idiosynkratische Schocks wirken auf den Anteil qualifizierter Arbeit und verändern die Arbeitsmarktdichte auf den Teilmärkten. Neutrale Schocks wirken zweistufig auf die Gesamtarbeitslosigkeit: Zuerst reduzieren sie geringqualifizierte Arbeitslosigkeit, und dann verringern sie rapide hochqualifizierte Arbeitslosigkeit. Eine hohe Substitutions-Elastizität zwischen den beiden Qualifikationen führt zu einer höheren Volatilität und einer höheren Korrelation zwischen Arbeitslosigkeit und freien Stellen. Das dritte Kapitel untersucht die Verbindung zwischen Gruppen-Konsumwachstum und dessen Volatilität, wenn die Agenten heterogen sind und eine Konsumexternalität vorliegt. Die Präferenzen der Haushalte hängen mit der Konsumwachstumsvolatilität insofern zusammen, als diese Vermögensentscheidungen treffen müssen: Die Volatilität verringert sich mit der Geduld und steigt mit dem Wunsch, das Konsumniveau der Vergleichsgruppe zu halten. Darüber hinaus sollten Konsumwachstum und dessen Volatilität positiv korrelieren. Diese letzte Hypothese wird mit Daten aus dem Sozio-oekonomischen Panel und der Einkommens- und Verbrauchsstichprobe überprüft, wobei sich ein U-förmiger Zusammenhang zwischen Konsumwachstum kurzlebiger Güter und dessen Volatilität ergibt. / This dissertation addresses the labor market performance and consumption dynamics of different socioeconomic groups. The first part examines the connection between cyclical variations in skilled and unskilled labor markets. Using a business cycle model with search frictions in skill-specific markets, I find that imperfect substitution between skilled and unskilled labor creates an important channel for variations in the skill-specific markets. Together with a skill-neutral or -biased technology shock, the model generates downward-sloping Beveridge curves in aggregate and skill-specific labor markets. I extend the study to allow for a dynamic link between the skill-specific labor markets. Human capital investment is determined endogenously and idiosyncratic shocks shift the skilled labor share and change tightness in both skilled and unskilled markets. Upon a neutral shock, the decrease of total unemployment is two-staged: Firstly with a reduction in unskilled unemployment, and then with a sharp decline of skilled unemployment when skill substitution dominates. A larger elasticity of substitution between the two types of labor leads to higher volatility of the model variables and higher correlation between unemployment and vacancies. The second part studies the link between group-specific consumption growth and its volatility in a framework of heterogeneous agents, under the assumption of a consumption externality. Household preferences are related to the consumption growth volatility through asset holding decisions: The volatility decreases with groups'' patience, and increases with the eagerness to keep up with the group average. Moreover, consumption growth is expected to be positively related to its volatility. This last hypothesis is tested using household data imputed from the German Socio-Economic Panel and the German Income and Expenditure Survey, where a U-shaped relationship is found between nondurable consumption growth and its volatility.