College major choice, occupational structure and demographic patterning by gender, race and nativity

Abstract

This paper asks whether and to what extent occupational segregation by gender, race/ethnicity and nativity at national level influences the patterning of individual college major choice. National Education Longitudinal Studies (NELS 1988–1994) provides college major information and the Public Use Micro data 1990 census 5% sample (PUMS 5%) provides occupational structure information, focusing on the demographic representation in broad occupational fields by gender, race and nativity. I match the occupational structure information from the PUMS data to the NELS data by group membership, after cross-classifying gender, race and nativity. This paper finds that the demographic group representations in technical and life/health science occupations at societal level have significant positive influence on choosing corresponding college major fields for students from the same groups. This may indicate that the seemingly individual choice of college major has deep structural roots at the societal level.

Notes

1 Since college major choice and occupation choice are inter-related, the issue of endogeneity may be a concern. However, this article examines the influence of occupational structure, measured by the group composition in occupational fields at societal level—not individual level occupation choice. The group composition at the societal level in various occupations happens before and outside the process of the individual level college major choice. Therefore, the issue of endogeneity should not be a concern.

2 NELS data, though already over-sampled Asian and Hispanic students, still does not allow for the analysis of their sub-groups. The strength of NELS data, which I discussed at length in the article, outweighs the limitations.

3 The multinomial model requires testing of the IIA (Independence of Irrelevant Alternatives). That is, to test whether the categories are truly distinct from one another. This study used Hausman test for IIA in STATA and did not find any evidence to violate IIA assumption.

4 To solve the identification problem of Heckman's model, I use the instrumental variable approach, which requires an instrumental variable associated with the selection outcome, but not considerably associated with the substantive process, in this context, the model for college major choice. I use the expectation to attend college as the instrumental variable in this article, which is closely associated with college attendance, thus included in the selection model, but not considerably relevant to college major choice.