HOW A COURT ACCEPTED AN IMPOSSIBLE EXPLANATION

Although it is often true that the association between two variables m ay be due to some unobserved third variable, the plausiblity of such a rguments needs careful examination. In 1987, a Canadian court accepted an explanation offered by the National Revenue Service of Canada that , in a test for promotion, lower pass rates among women than among men were explained by differences in rates of college attendance, a varia ble that was not directly observed by the court. We show that this exp lanation is not merely implausible. It is impossible. We conclude with a brief discussion of the role of a statistician in an argument invol ving an unobserved variable.