Pre-market characteristics, gender wage disparities, and the performance of minorities in the United States labor market: Application and comparison of non-parametric methodologies on a highly-educated sample

Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)




Dan Black


Gender wage disparities, Minorities, Labor market, Wage gap, Discrimination

Subject Categories

Econometrics | Gender and Sexuality | Labor Economics | Race and Ethnicity


A large number of studies have investigated the source of gender and racial wage gaps, seeking to distinguish the effect of discrimination from other productivity-related effects, such as the effects of the low educational level and labor force participation rates of women and minorities. Most of the previous studies have largely neglected the importance of restrictive parametric assumptions. If those parametric assumptions did not hold, the estimation would be presumably biased. Against this backdrop, I applied and compared three non-parametric methodologies---exact cell matching, kernel on age cell matching, and local constant kernel regression to estimate the role of pre-market factors in shaping the gender and racial wage differences. These methodologies are carefully applied to the 1993 National Survey of College Graduates (NSCG) for four ethnic groups: whites, blacks, Hispanics, and Asians. My estimated results reveal that pre-market factors are very important in explaining both the gender wage gaps and the male racial/ethnic wage gaps. The pre-market characteristics are attributable for 46-54% of the gender wage gap, though the effects vary among ethnicities. When the sample is Surface provides description only. Full text is available to ProQuest subscribers. Ask your Librarian for assistance. to single women, unexplained gender wage gaps shrink significantly for white, black, and Asian women to 75-90%. In terms of the racial wage gap, black highly-educated women earn more than white. My further examination suggests that, rather than the favorable pre-market factors of black women, this difference is more likely the high proportion of inner-racial marriage of blacks (97%) and low income of their spouses. This may explain black women's high labor force participation and high income. Both Hispanic and Asian women earn similar amount as white women. This similarity could be attributable to their assimilation to white women in terms of their English language skills and inter-racial marriage with white men. The empirical results show that the parametric estimates can be very misleading. The results also indicate the advantages and limitations of each of the methodology. I find that there is a trade-off between bias and variance. Local constant kernel regression does not perform well in the wage decomposition unless I give it some restrictions.


Surface provides description only. Full text is available to ProQuest subscribers. Ask your Librarian for assistance.