The cover of randomness: validating implicit methods for the study of sensitive topics.

Abstract

We review the methods we developed to study female genital cutting in Sudan and sex-selective abortion in Armenia. These methods were untested at the time of our original research, and here we compare the distinct but overlapping approaches we used to validate our methods for each of the two countries. Additionally, we repeat a number of analyses, including those related to validation, with previously unpublished data from Sudan. All results replicate previous findings. Replicating previous results is encouraging, but we nonetheless argue that validation for Armenia is more convincing than for Sudan. Specifically, even if female genital cutting and the preferential abortion of females are equally sensitive as research topics, son bias is inherently easier to study than cutting because biological sex determination is a random process with no natural analogue in the case of cutting. This randomness provides a kind of cover for research participants who are son-biased but want to create the impression that they are not. This cover, in turn, allows the researcher to resolve any trade-off between methods that produce explicit granular data and methods that produce untraceable, highly aggregated data in favour of methods producing the explicit and granular.