In a recent paper in the Review of Economic Studies, Siwan Anderson and Debraj Ray (Anderson and Ray, 2010) develop and apply a new ‘flow’ measure of ‘missing women’ to estimate the extent of gender bias in mortality in developing countries. Contrary to the existing literature, they find that the problem of gender bias in mortality is as severe among adults as it is among children in India, that gender bias in mortality is larger in Sub‐Saharan Africa than in China and India, and that there was substantial evidence of gender bias in mortality in the US around 1900. These latter results are driven largely by the finding of substantial gender bias among adults. We show first that the data for Sub‐Saharan Africa used in the paper are generated by simulations in ways that deliver their findings on Africa (and the US in 1900) by construction. Second, we show that the analysis is entirely dependent on a highly implausible reference standard that is inappropriately applied to settings where the overall disease and mortality environment differ greatly; the attempt to control for the disease environment by the authors is not able to address these issues. When a more appropriate reference standard is used, most of the new findings of Anderson and Ray disappear. Instead, the findings from the existing literature relying on stock measures of missing women are confirmed. The one finding that remains and deserves further attention is some evidence of gender bias in mortality among young adults in Africa (though of much lower magnitude than suggested by Anderson and Ray).
