J. Racine et P. Rilstone, THE REVERSE REGRESSION PROBLEM - STATISTICAL PARADOX OR ARTIFACT OF MISSPECIFICATION, Canadian journal of economics, 28(3), 1995, pp. 502-531
The usual approach to wage discrimination asks whether certain individ
uals receive lower wages for the same level of productivity characteri
stics. The reverse approach asks whether these individuals are more pr
oductive given the same wages. When these hypotheses are tested, incom
patible conclusions seem to result. To circumvent specification proble
ms, non-parametric techniques were used to estimate Canadian male-fema
le wage/experience profiles. The findings indicate that, when the corr
ect functional form is specified and the effects of childrearing activ
ities are controlled for, there exists a wage/experience gap favouring
men regardless of the approach, suggesting that the paradox may be si
mply an artefact of misspecification.