Lewis and Shorten (Applied Economics, 1991, 23, 167-77) have proposed that male and female earnings, male and female labour force participation rates and occupational segregation are simultaneously determined in the labour market. They estimate their model for Australia using 1981 Census data and find substantial evidence to support their hypothesis. However, there have been no subsequent studies to empirically test their hypothesis for other countries. In this paper their model for the United States using 1990 Census data is replicated. All but one of the coefficients on the endogenous variables have the same signs as those reported by Lewis and Shorten, which tends to support their model. The specification of Lewis and Shorten's model is tested and it is inappropriate for the US data. The model is then reformulated and re-estimated. The reformulated model also shows substantial evidence of simultaneity between occupational segregation and other labour market outcomes.