DETECTING DISCRIMINATION - ANALYZING RACIAL DISPARITIES IN PUBLIC CONTRACTING

In City of Richmond v. J. A. Croson Co. (1989), the Supreme Court esta blished strict scrutiny as the standard applicable to affirmative acti on programs which set aside quotas of public contracts for minority-ow ned businesses, and Aderand v. Pena (1995) extended the strict scrutin y standard to federal programs. Although the requirements of these dec isions clearly require multivariate statistical analysis, most ''dispa rity studies'' have used a univariate comparison between the expected and the observed shares of contracts going to minority-owned firms. We examine four statistical methods-ordinary least square multiple regre ssion, logit and tobit models, and a multivariate procedure for compar ing expected and observed outcomes. Because no data are presently avai lable at the level of specificity required by Croson, we constructed s ynthetic data sets to represent typical variations among large U.S. ci ties. Applying the statistical methods to each data set allows evaluat ion of the extent to which each method is able to both remove spurious and detect valid estimates of racial disparity when relevant control variables are added. Findings: (a) All four models removed apparent di sparities which, although significant in univariate analysis, were Kno wn to be spurious. (b) Tobit and logit models, whose underlying assump tions better fit the nature: of public contracting data. provided more accurate and more sensitive estimates than OLS regression. (c) Compar ison of expected and observed outcome within categories of control var iables yielded results very similar to legit and tobit models and, bec ause of the nature of the comparison specified in Croson, produced sli ghtly mon sensitive probability estimates. (C) 1996 academic Press, In c.