Non-parametric decomposition and the choice of reference group

Abstract

Purpose – The Ñopo (2008) method of non-parametric decomposition, a matching-based alternative to Oaxaca (1973) and Blinder’s (1973) method of wage gap decomposition, is subject to the so-called “index number problem” common to the Oaxaca-Blinder and many related methods: its results are sensitive to the (arbitrary) choice of either male or female sex as the reference category in decomposition. The purpose of this paper is to address this issue by proposing an extension to the method that is invariant to the choice of reference category. Design/methodology/approach – The Ñopo method is modified such that the wage structure of the average worker instead of either male or female worker’s is used as the reference, enabling one to distinguish the “male advantage” and “female advantage” portions of the gender wage gap. As an illustration, a decomposition of the gender wage gap is performed with the modified method, using data from 15 OECD countries. Findings – The empirical results using the Ñopo decomposition indicate substantial differences in estimates of the unexplained gap depending on which sex is used as the reference category. Moreover, this disparity varies significantly with the choice of covariates used in the decomposition. This confirms there is significant cross-country variation in the asymmetry between male advantage and female disadvantage and that a decomposition method making this explicit would be relevant in real world settings. Originality/value – The extension of the Ñopo method proposed in this paper offers a way of decomposing the wage gaps in a way that is not sensitive to the choice of the reference category.