What can people tell us about their own gender?

Making use of a series of national surveys in the US, both online and live caller, we assess the extent to which respondents can accurately self-report their gender identities, using a baseline of Bem Sex Role Inventory scores. We show that respondents are able to accurately report the extent to which they fit the gender identity most closely associated with their asserted sex (masculinity for men, femininity for women), but do a poor job of self-reporting the extent to which they have characteristics more commonly associated with a different sex. Moreover, we show that uni-dimensional gender identity self-report scales, while less theoretically appealing than bi-dimensional scales, capture almost exactly the same variance for most respondents. We also explore the reactiveness of these scales, as well as mode differences in the reporting of gender between live called and online modalities in the same survey. The results show that indirect measures, like the BSRI, are relatively stable within respondents, while the self-report scales show some degree of variance based on priming and framing effects, leading to caution in how they should be placed within surveys and understood. We go on to show that these scales are useful in predicting socio-political attitudes and reported behaviors on a variety of dimensions, outperforming self-reported sex in predicting attitudes, with the biggest differences coming between men who assert a “completely” masculine gender identity and everyone else. Finally, we explore the extent to which such scales can be useful in understanding the socio-political attitudes of individuals who hold non-binary sexual identities.