Measurements of wage discrimination, based on gender, age, race or rel
igion, are important for empirical economists. Traditional approaches
to the measurement of wage discrimination do not set up the question o
f discrimination as a hypothesis that can be statistically tested. Fur
thermore, they face the so called 'index number problem'. In this pape
r, an alternative approach is presented where wage discrimination can
be tested as a statistical hypothesis. Furthermore, the proposed metho
d eliminates the index number problem.