Are the Effects of Poor Survey Measurement of Sexual Identity on Estimated Associations between Sexual Identity and Health-Related Measures Moderated by the State-Level Policy Environment?

Including a "something else" response option for survey questions about sexual identity significantly moderates estimated differences between sexual identity subgroups in terms of behaviors and attitudes reported in U.S. health surveys. We hypothesize that these moderation effects will be larger in states with negative policies that fail to protect sexual minorities from discrimination. We tested this hypothesis by linking public-use data from the National Survey of Family Growth (2015-2019), which randomly assigned respondents to receive either a four-category measure of sexual identity, including "something else" as a response option, or a three-category measure omitting "something else," to year-specific state-level policy data from the Movement Advancement Project, which tracks state policies related to sexual minorities. In multivariable models for measures of substance use, reproductive health, family formation, and other correlates of health, we tested three-way interactions involving sexual identity, measurement type, and state-level policy classification. As hypothesized, we find repeated evidence (10 of the 32 health-related measures analyzed) of changes in the moderating effects of question type on sexual identity differences, with more prominent effects in negative policy environments. Suboptimal measurement of sexual identity may therefore have the largest negative effects on estimated health disparities in states with fewer protective policies.