Detecting Discrimination: Analyzing Racial Disparities in Public Contracting

Area of Science:

• Public Policy Analysis
• Statistical Methods in Law
• Econometrics

Background:

• Supreme Court rulings (City of Richmond v. J. A. Croson Co., 1989; Adarand v. Pena, 1995) mandate strict scrutiny for affirmative action programs.
• Existing disparity studies often rely on univariate comparisons, potentially overlooking complex factors.

Purpose of the Study:

• To evaluate multivariate statistical methods for analyzing racial disparity in public contracting.
• To compare the effectiveness of OLS regression, logit, tobit models, and a multivariate outcome comparison procedure.

Main Methods:

• Construction of synthetic datasets representing U.S. cities due to data limitations.
• Application of ordinary least squares (OLS) multiple regression, logit, and tobit models.
• Utilizing a multivariate procedure for comparing expected and observed outcomes with control variables.

Main Results:

• All four methods successfully removed spurious disparities identified through univariate analysis.
• Tobit and logit models provided more accurate and sensitive disparity estimates than OLS regression.
• Comparing outcomes within control variable categories yielded results similar to logit/tobit, with slightly higher sensitivity.

Conclusions:

• Multivariate analyses are essential for accurate assessment of racial disparity in affirmative action programs.
• Logit and tobit models are recommended for analyzing public contracting data due to their suitability.
• The study highlights the importance of appropriate statistical techniques for legal and policy analysis.