This dissertation provides evidence from controlled laboratory experiments that, in contrast to Frohlich and Oppenheimer's widely quoted results, supports Rawls' (1971) conjecture about the use of the maximum decision rule or "difference principle" in the original position. In addition, focal points defining alternative levels of poverty are examined to determine whether this additional information affects the selection of rules of distribution in an experimental original position. The basic experimental approach developed by Frohlich and Oppenheimer (1992) is used in this dissertation to examine whether existing theories of distributive justice are consistent with observed decisions on the selection of alternative rules used to distribute income. My experiments are focused on a design where subjects are required to perform tasks to earn income as opposed to the income being exogenously determined. In addition, the instructions are purged of any references to justice or fairness that might influence subject behavior. The baseline experimental design utilized in this study was found to replicate the results obtained by Frohlich and Oppenheimer, who found no support for a Rawlsian rule of distribution. Two experimental treatments were performed to investigate questions on the selection of floor income levels and the role of risk aversion. Both treatments were found to significantly affect the results. The first treatment investigated the impact of giving the subjects information on alternative measures of need or poverty. These focal points were found to significantly narrow the distribution of floor constraints selected by groups of subjects. The second treatment introduced a random chance of excluding subjects from the production tasks, thereby thickening the "veil of ignorance". This "exclusion" treatment manipulated a factor related to the subjects' risk preferences and was found to dramatically shift the distribution of rules selected by the subjects. The exclusion treatment found significant support for both Rawlsian and egalitarian rules of distribution. Finally, to examine the effectiveness of our experimental design in inducing impartial decisions, we performed multinomial logit regressions on both group and individual decisions. The results suggest that our design was effective, since individual characteristics were unable to explain the group decisions.

The principal purpose of this thesis is to extend empirical likelihood (EL) based procedures to some statistical models defined by unconditional moment inequalities. We develop EL procedures for two such models in the thesis. In the first type of model, the underlying probability distribution is the (infinite-dimensional) parameter of interest, and is defined by a continuum of moment inequalities indexed by a general class of estimating functions. We develop the EL estimation theory using a feasible-value-function approach, and demonstrate the uniform consistency of the estimator over the set of underlying distributions in the model. Furthermore, for large sample sizes, we prove that it has smaller mean integrated squared error than the estimator that ignores the information in the moment inequality conditions. We also develop computational algorithms for this estimator, and demonstrate its properties in Monte Carlo simulation experiments for the case of infinite-order stochastic dominance. The second type of moment inequality model concerns stochastic dominance (SD) orderings between two income distributions. We develop asymptotic and bootstrap empirical likelihood-ratio tests for the null hypothesis that a given unidirectional strong SD ordering between the income distributions holds. These distributions are discrete with finite support, and, therefore, the SD conditions are framed as sets of linear inequality constraints on the vector of SD curve ordinates. Testing for strong SD requires that we consider as the null model one that allows at most one pair of these ordinates to be equal at an interior point of their support. Finally, we study the performance of these tests in Monte Carlo simulations. / Le principal objectif de cette thèse est d'étendre les procédures basées sur la vraisemblance empirique (EL) à des modèles statistiques caractérisés par des inégalités concernant des moments non conditionnels. On développe dans la thèse des procédures EL pour deux types de modèles. Pour le premier type, le paramètre d'intérêt, de dimension infinie, est la distribution de probabilité sous-jacente, définie par un continuum d'inégalités en correspondance avec une classe générale de fonctions d'estimation. L'approche utilisée afin de développer la théorie d'estimation s'appuie sur une méthode faisable de calcul de la fonction objectif. Elle permet de démontrer la convergence uniforme de l'estimateur sur l'ensemble des distributions du modèle. On démontre en outre que, pour une taille d'échantillon suffisamment grande, son erreur quadratique moyenne est inférieure à celle d'un estimateur qui ne se sert pas des informations fournies par les inégalités. Des algorithmes numériques pour le calcul de l'estimateur sont développés, et employés dans des expériences de simulation afin d'étudier les propriétés de l'estimateur dans le contexte de la dominance stochastique à l'ordre infini. Le second type de modèle concerne la dominance stochastique (SD) entre deux distributions de revenus. On développe des tests asymptotiques et bootstrap, basés sur le rapport de vraisemblance empirique, pour l'hypothèse nulle selon laquelle il existe un ordre de dominance stochastique forte unidirectionnelle entre les deux distributions. Celles-ci sont discrètes et avec support fini, ce qui permet de formuler l'hypothèse nulle en termes de contraintes d'inégalité sur le vecteur d'ordonnées des courbes de dominance. La dominance forte exige que l'hypothèse nulle n'admette qu'une paire d'ordonnées égales dans l'intérieur du support. Les performances des tests sont enfin étudiées au moyen de simulations Monte Carlo.

<p> This dissertation studies firm behavior and adaptation of firms in developing countries using firm-level data. Chapter 2 analyses the impact of business group membership on a firm's economic performance when operating in a suboptimal legal environment. The specific null hypothesis is that Indian firms in business groups have advantages in contract intensive industries, especially in Indian states with poorly functioning legal institution. A 2002 national judicial reform is used as quasi-natural experiment as an identification strategy. Chapter 3 further explores Indian business groups by looking at how ownership structure affects TFP, including the impact of family or corporate ownership. These two chapters suggest that business groups can have important economic affects at the firm level. Chapter 4 uses a World Bank firm level survey in 57 developing countries to examine whether MNC presence affects various measures of innovative output. We find that industrial, labor, and input proximities have a positive and significant impact on the likelihood of innovation when interacted with FDI concentration.</p>

[...] Men are divided subjectively by factors such as race, language, religion, and culture - i.e. ethnicity; and objectively by factors of distance, climate, economic situation, etcetera. Despite these differences, some degree of pilitical organisation has proven to be universal. Since the 18th century, modern technology and commerce have made political societies of alrger and larger scale more practicable. Furthermore, in many cases the desire for economic progress has made larger political units necessary. And finally, the formation of one large and stable political society often begets others. [...]

Unit root tests typically suffer from low power, severe size distortions, and dependence on tuning parameters. All three are particularly pronounced in situations deemed most relevant to empirical researchers, namely small samples, correlated errors, and nonlinear transformations of data. An important example of the latter arises with regulated (bounded) integrated time series which exhibit asymptotic distributions substantially different from their unregulated counterparts. In this regard, this thesis bridges a gap in the literature by generalizing the contributions of [Cavaliere, 2005] and [Cavaliere and Xu, 2011] to derive limiting distributions for regulated fractionally integrated series. Several algorithms are suggested for deriving the limiting distributions for these processes computationally. This theory is then applied to develop the fractional variance ratio statistics proposed in [Nielsen, 2009] for regulated processes. Moreover, results of [Troki ́c, 2013] are extended by deriving the asymptotic distribution of regulated (fractionally) integrated series with a linear trend. The power performance of these statistics depends crucially on the length, direction, and nature of the regulating interval. When the interval is sufficiently wide the results in [Nielsen, 2009] are replicated. In all other cases the statistic has unacceptably low power. This thesis also uses wavelet theory to propose a tuning parameter free family of unit root tests in the frequency domain. This test exploits the wavelet power spectrum of the observed series and its fractional partial sum to construct a test of the unit root based on the ratio of the resulting scaling energies. This statistic has good power and exhibits drastically reduced size distortions. Remaining size distortions are virtually eliminated using the proposed fast double wavestrap method. The final chapter extends the work of [Davidson and MacKinnon, 2007]to develop a theory for the general fast iterated bootstrap and demonstrates its efficiency in eliminating severe size distortions in unit root tests. / Les tests de racine unitaire souffrent en général d'une puissance faible, de sérieuses distorsions de niveau, et d'une dépendance de paramètres de réglage. Tous les trois de ces défauts sont particulièrement prononcés dans les situations jugées les plus pertinentes pour les chercheurs empiriques, à savoir les petits échantillons, les aléas corrélées, et les transformations non linéaires des données. Un exemple important de ce dernier survient avec les séries chronologiques intégrées mais bornées (ou régulées), dont les distributions asymptotiques diffèrent considérablement de leurs homologues à cet égard, cette thèse comble une lacune dans la littérature en non régulées. A généralisant les contributions de [Cavaliere, 2005] et [Cavaliere and Xu, 2011] par le calcul des distributions limite des séries intégrées à un ordre fractionnaire mais régulées. Plusieurs algorithmes sont proposés à cette fin. La théorie est ensuite appliqée pour développer les statistiques ratio-de-variance fractionnaires proposées dans [Nielsen, 2009] pour les processus régulés. En outre, des résultats de [Troki ́c, 2013] sont étendus par la dérivation de la distribution asymptotique des séries fractionnaires avec une tendance linéaire. La puissance de ces statistiques dépend essentiellement de la longueur, de la direction et de la nature de l'intervalle de régulation. Lorsque l'intervalle est suffisamment large, les résultats de [Nielsen, 2009] sont reproduits. Dans tous les autres cas, la statistique a une puissance beaucoup trop faible. Cette thèse utilise également la théorie des ondelettes afin de proposer une famille de tests de racine unitaire, tous indépendants de paramètres de réglage, dans le domaine fréquence. Ces tests exploitent le spectre de puissance en ondelettes de la série observée, et sa somme partielle fractionnaire, pour construire un test de racine unitaire basé sur le rapport des énergies aux échelles différentes. Les tests ont une bonne puissance et bénéficient d'une réduction importante des distorsions de niveau. Celles qui restent sont pratiquement éliminées par l'utilisation d'une méthode proposée de wavestrap double rapide. Le dernier chapitre étend le travail de ct davidson-2007 en développant une théorie générale du bootstrap rapide itéré, et en démontrant sa capacité d'éliminer les distorsions sévères de niveau dans les tests de racine unitaire.

<p> Urban producer theory introduces a production function which incorporates congestion in production with inputs possessing a quality component that influences productivity. These features yield cost-minimizing behavior in which firms respond to higher space rent by increasing the quality of the inputs used in production. This behavior generates demand-side sorting of high quality inputs into high rent areas. The prediction of sorting based on input quality is tested on attorneys employed in the Cleveland CBSA. Evidence of the sorting into high rent areas of attorneys based upon the national ranking of the law school attended is found. A 1% increase in rent leads to a 1.26% to 2.89% increase in the number of the highest quality attorneys employed in high rent districts. Ability sorting poses a significant risk in biasing the measurement of agglomeration economies based on wages.</p>

<p> This dissertation examines the implications of sticky rents on the measurement of owner-occupied housing in the Consumer Price Index (CPI). I argue that marginal and not average rents are the most theoretically justified measurement of owners' equivalent rent (OER), and that the current measurement of rental inflation using average rents is methodologically incorrect. I then discuss the literature on sticky rents and tenure discounts and present a theoretical model showing the implications of sticky rents for aggregate measures of inflation. Then I use two new data sources to construct marginal rent measures to compare to average rent measures. The results show that marginal rents reflect market turning points sooner, and show a larger post-housing bubble decline in rents. In addition, marginal rents are shown to forecast overall inflation better than average rents. Finally, the implications of these results for policy are considered using the Taylor Rule for optimal monetary policy. The results present suggestive evidence that the impacts of switching to marginal rents may be large enough to significantly impact monetary policy and allow the Federal Reserve to be more responsive to both the boom and bust of housing bubbles.</p>

This study challenges the use of the gross increase in female participation rate--prewar to postwar--as the sole indicator of changes in women's employment. It is hypothesized that the utilization of aggregate participation to measure progress in women's employment disguises significant variation in women's participation by industry and by occupation, during and after the war. The purpose of this dissertation, thus, is to clarify the impact of WWII on female employment. The method developed was a sytematic analysis of the wartime employment of women in terms of its growth and its industrial and occupational structures, 1940 to 1950 inclusive. Chapter II initiates the study by giving a brief summary of the background against which the increase in female employment occurred. The labor market during the defense period and early war years is described in terms of the rising demand for labor because of the drafting of men into the Armed Forces and the mounting production requirements. Then the three phases of the response to this demand is discussed: reemployment, overtime, and the new regard for women as a source of labor. Chapters III and IV investigate the nature of women's employment by industry and by occupation, and produce a more complete picture of female employment over the decade 1940-1950. It was shown that although the war exercised an immediate impact, causing certain compositional changes and determining certain trends in distribution, these effects were not evident by the end of the decade. In summary, the decade 1940-1950 was composed of two female employment trends. From 1940 to 1944, there were major aberrations in the structure of female labor with respect to aggregate participation rates, numbers, and industrial distribution. Women entered particularly the war manufacturing industries while prewar feminized industries such as personal services actually decreased in size and rank. The compositional change in the occupational structure during this period was derived largely from new female entrants to the labor force who entered primarily the Operative and Clerical occupations. Stability in the occupational structure, nevertheless, was evidenced in minimal shifting between occupations. . . . (Author's abstract exceeds stipulated maximum length. Discontinued here with permission of school.) UMI

This dissertation consists of two empirical papers that investigate the return to education in Jamaica and the effect of federal assistance on state spending on Old Age Assistance. Paper one investigates the wage determination process in Jamaica in light of the heavy subsidization of all three levels of education from public resources. The results imply that highly educated workers prefer public sector employment. Whereas the private sector appears to value workers with A-levels (comparable to the AP designation in the United States) the public sector is prepared to pay University graduates the highest premium. The second paper analyzes the efficacy of federal grants on state spending for Old Age Assistance (OAA). Under this program, states had two variables at their disposal: the number of recipients and the benefit level. States are found to be very responsive to changes in the price of offering additional benefits and of expanding coverage as well as to the income they receive from the federal government. The statistical significance of the cross and own price elasticities reinforces the importance of separating the two prices in investigating this relationship.

We introduce a model of market shares in aggregate markets beginning with the McFadden multiattribute consumer choice theory and extending to allow for lagged individual choice probabilities and unequally spaced historical data. Our market share model admits the use of Bayesian methods (in addition to traditional maximum likelihood methods) to deal with the problem of chronically short historical time series (particularly prices) in many countries for applications such as energy shares. Two applications are presented. In the first (United States electric power fuel shares), fuel shares are modeled as a function of oil, gas, and coal prices and installed generation capacity. We estimate own and cross elasticity parameters and use Bayesian methods both to address short data series and alternative prior beliefs about the parameters. In the second application, we use G8 industrial sector fuel price data to estimate oil, gas, and coal own and cross price parameters in those economies. Both applications employ Bayesian methods to address short historical time series, particularly in the latter application of fuel price data outside the United States.