A multinominal approach to estimating the determinants of occupational segregation

Graduation date: 1994The purpose of this study is to analyze the sources\ud of occupational segregation between men and women.\ud Specifically, three hypotheses are tested. First, do\ud stiff specific training requirements for an occupation\ud increase the likelihood that it is male-dominated, as\ud would be expected from human capital theory? Second,\ud do differences in perceived natural ability between the\ud sexes heavily influence the possibility of an occupation\ud being male- or female-dominated? Finally, do\ud differences in preferences for job attributes between\ud men and women influence the chance of whether an occupation\ud will be male- or female-dominated?\ud The data used in this study cover 495 occupations\ud from 1979 in which on-the-job worker requirements are\ud provided describing workers' level and type of education,\ud natural ability, on-the-job environmental conditions,\ud and physical demands. Occupations are divided\ud into three categories: female-dominated, male-dominated,\ud or neutral. The empirical model uses a multinomial\ud approach to estimate the log of the odds that an\ud occupation is male-dominated relative to being neutral\ud and the log of the odds that an occupation is female-dominated\ud relative to being neutral.\ud The results of this study imply that sizable specific\ud training requirements for an occupation increase\ud the likelihood that it is male-dominated and higher\ud levels of general education requirements increase the\ud possibility that an occupation is female-dominated.\ud There is evidence to support the hypothesis that differences\ud between the sexes' perceived abilities contribute\ud to occupational segregation. Most on-the-job\ud environmental conditions and physical demands do not\ud affect the likelihood that an occupation is male- or\ud female-dominated in contrast to the traditional claim\ud that women prefer more desirable working conditions\ud than men