Academic work on the "motherhood shift" has defined two categories of women: mother and non-mother. Quantitative work on motherhood's effect on sociopolitical variables, such as voting, instantiates this problematic reductive by coding women into these two categories. This operationalization misconstrues the reality of motherhood, failing to take it as an transition which some women select or are selected into. Nearly all previous research on the motherhood shift in voting conflates women who will and will not become mothers by using cross-sectional data and/or random effects models, ignoring the effect of selection. In this article, I find that reconsidering the motherhood shift as a selected phenomenon and modeling it through fixed effects regression reverses our understanding motherhood's effect on voting. This article clarifies our conceptualization of the motherhood shift in voting and calls into question all cross-sectional and/or random effects work on the motherhood shift, which suffers from selection bias.