Occupational segregation by sex

This thesis is an attempt to describe and explain occupational segregation by sex as evidenced by 1961 Canadian census data. Previous writers discussed the ethical question of whether men and women should be occupationally segregated or whether they should receive equal pay for equal work. This literature is reviewed. Irrespective of ethical issues, if men and women are equally productive but unequally paid, why should a profit maximizing firm hire any but the cheapest labour? This problem is known as Cassel's paradox. Previous attempts to resolve Cassel's paradox included the use of simple supply and demand models, barriers to competition, theories of monopsony, human capital theory and adjustments for quality differences. These approaches are criticized and alternative concepts of discrimination are reviewed. For the purpose of the thesis, statistical discrimination is defined as a situation in which employers draw inferences about productivity from unalterable attributes of individuals even though the attributes are not correlated with productivity. A model of occupational segregation by sex is developed that permits analysis of statistical discrimination. Employers are assumed to hire labour under uncertainty as to its qualifications. Hiring is assumed to involve a cost. Each occupation is characterized by the traits required to perform in the occupation. The probability that a person is qualified for an occupation is assumed to depend on the traits required for the occupation and the person's sex. From these assumptions the derived demand for the male-female employment ratio by occupation is determined as a function of employer investment, male and female wages, and the required traits. Statistical discrimination is said to be indicated if a trait is significantly related to the male-female employment ratio and yet there is no significant difference in its distribution by sex. A correlation coefficient of 0.78 is found in a relation between the logarithm of the male-female employment ratio and thirteen independent variables including a proxy for employer investment, the wage ratio, the male-female education ratio and ten traits. The education ratio, included in the regression analysis to adjust for quality differences, has the greatest impact of any variable. Its negative coefficient is opposite in sign to what was expected. It is possible that the negative education coefficient indicates discrimination. Data was not available for the actual distribution by sex for five of the ten traits. Three of the remaining traits, numerical aptitude, spatial aptitude and form aptitude, indicate the presence of statistical discrimination. The need for further research on how stereotypes affect occupational segregation is suggested by this study.