Simpson’s Paradox in the interpretation of “leaky pipeline” data

Abstract

Abstract The traditional ‘leaky pipeline’ plots are widely used to inform gender equality policy and practice. Herein, we demonstrate how a statistical phenomenon known as Simpson’s paradox can obscure trends in gender ‘leaky pipeline’ plots. Our approach has been to use Excel spreadsheets to generate hypothetical ‘leaky pipeline’ plots of gender inequality within an organisation. The principal factors, which make up these hypothetical plots, can be input into the model so that a range of potential situations can be modelled. How the individual principal factors are then reflected in ‘leaky pipeline’ plots is shown. We find that the effect of Simpson’s paradox on leaky pipeline plots can be simply and clearly illustrated with the use of hypothetical modelling and our study augments the findings in other statistical reports of Simpson’s paradox in clinical trial data and in gender inequality data. The findings in this paper, however, are presented in a way, which makes the paradox accessible to a wide range of people.