This paper expands the empirical implementation of hierarchical segregation models to allow for the use of panel methods. We use firm level data collected between 1977 and 1985 from a regional grocery store that faced a title VII class-action lawsuit over gender discrimination much the same as Wal-Mart and Costco. Special problems arise in implementing decompositions in a fixed effects and random effects setting, especially when analyzing wage-level differences. We develop a variation of wage decompositions that takes into consideration an unbalanced design and extends the literature by explicitly formalizing the inclusion of the unobserved heterogeneous effects.