Date of Award

May 2014

Degree Type


Degree Name

Doctor of Philosophy (PhD)




Jeffrey Kubik

Subject Categories

Social and Behavioral Sciences


This dissertation consists of three chapters and mainly focuses on the robust estimation of different important factors contributing changes to the U.S. income inequality over the last two decades. The primary objective is to precisely estimate different labor market outcomes when the behaviors of the tails of the distribution bear much importance. These studies are very relevant in the current context of the U.S. labor market because over the last three decades the U.S. income distribution have come very skewed and therefore, we are more interested at the behaviors of the the upper and lower-tails of the U.S. income distribution compared to the middle.

The first chapter of the dissertation proposes a semi-parametric procedure to determine the contribution of human capital variables in explaining the changes of U.S. wage distribution function over the period of 1990-2000. The effects of these factors are estimated by using the Chamberlain's two stage Box-Cox quantile regression approach. One of the main contributions of this chapter is that it relaxes the linearity assumption of the conditional quantile function to estimate the counterfactual wage distribution function consistently. This chapter also shows that the proposed method provides better estimates of capturing the effects of human capital variables in the two tails of the U.S. wage distribution while the results of other parts of the distribution are comparable with the estimates of the previous study which used the linear quantile regression approach.

The second chapter proposes a semi-parametric estimation method known as `Box-Cox Unconditional Quantile Regression' to explain the increasing trend of the U.S. wage inequality over the last two decades for men and women separately. Box-Cox Unconditional Quantile Regression is a generalization of the Linear Unconditional Quantile Regression model proposed by Firpo, Fortin and Lemieux (2009). The main contribution of this chapter is to determine the role of unionization in explaining the rising wage gap between the upper and lower tails of the U.S. wage distribution function during the period 1990-2010. I also show that proposed Box-Cox unconditional Quantile Regression model precisely estimates the parameters compared to the Linear Unconditional Quantile Regression model at the two tails of the U.S. wage distribution function while the results of the rest of the part of the distribution are comparable with the estimates of the previous study by Firpo et al. (2009). To summarize, this proposed approach is most applicable in cases where the behavior of the tails of the distribution bears much importance. I find that declining unionization can explain around 20-25 percent of the total fall in the 50/10 percentile wage gap and does not have much impact on the rise in the 90/50 percentile wage gap for men over the period 1990-2010. For women, unionization has very little impact on the rising wage gaps at different parts of the wage distribution over the last two decades.

In the third chapter I propose an extension of Rosen's (1986) theory of equalizing differences model by incorporating the role of different types of cognitive and noncognitive skills in worker's job preference function to explain the U.S. labor market sorting mechanism. I show that ignoring the impact of worker's skills on occupational choice decision leads to bias and inconsistent results because of sample selection specification error. This chapter proposes a solution of the problem by using workers' education level as the proxy of the cognitive skills. This chapter also tests the implication of the proposed labor market sorting model by using the current population survey and Occupational Information Network data sets and finds that a positive and statistically significant relationship exists between noncognitive skills and workers sorting behavior. The empirical results suggest that the relative employment share of women is higher in the occupation in which the people's task is more important because of their job preferences, which depend on the level of different types of noncognitive skills such as interpersonal and social skills. This type of sorting behavior can explain a large portion of the male-female wage gap.


Open Access