Reevaluating the Substantive Representation of Lesbian, Gay, and Bisexual Americans: A Multiverse Analysis

Abstract

Social scientists are facing a crisis of confidence in quantitative results. Multiverse analysis provides concerned scholars a tool for verifying the robustness of findings. This article introduces political scientists to multiverse analysis through an application. It identifies how differing approaches to data processing led to divergent conclusions about the representation of lesbian, gay, and bisexual (LGB) Americans in Congress in a 2015 Journal of Politics article. The analysis casts doubt on the original conclusion that the size of the LGB population in a district is significantly associated with the bill sponsorship activity of its representative. More broadly, it demonstrates how researchers can keep a running tally of data-processing decisions and parsimoniously present the consequences of those decisions for the findings. Multiverse analysis can help scholars publish replicable results in original work as well as replicate and extend previously published work.