Whether selection processes contribute to group-level disparities or merely reflect pre-existing inequalities is an important societal question. In the context of observational data, researchers, concerned about omitted-variable bias, assess selection-contributing inequality via a kitchen-sink approach, comparing selection outcomes of different-group individuals net of various characteristics. We introduce a conceptual framework that clearly defines the quantity of interest and argue that researchers should only control for the extent to which individuals meet selection criteria. Informed by this framework, we use directed acyclic graphs and structural equation modeling to show that traditional practices can inaccurately represent selection-contributing inequality because chosen controls frequently capture selection-irrelevant characteristics, which we define as the correlated proxy problem. Using Black-White disproportionality in special education as a case study, we show that typical practices of using test scores as covariates likely drastically underestimate the influence of selection-contributing inequality to Black over-representation in special education.