Research comparing heterosexuals with bisexuals and homosexuals in economics and the social sciences typically relies on two strategies to identify sexual orientation in existing survey data of general populations. Probing respondents to self-report their sexual orientation is generally considered the preferred option. Since self-reports are unavailable in most large multidisciplinary surveys, often researchers infer sexual orientation from the gender-constellation of a respondent’s partnership instead. Based on German Socio-Economic Panel (SOEP) data, this article reviews both strategies empirically in the context of a household panel survey. The analysis shows that self-reported and partnership-inferred sexual orientation are not mutual substitutes, instead leading to substantively different conclusions about differences between heterosexuals and LGBs (Lesbian, Gays, and Bisexuals). The article discusses problems of non-coverage in partnership-inferred sexual orientation and also investigates measurement error in self-reported sexual orientation, finding notable mode and interviewer effects.