Predicate Studies and Regression Analysis: A Look at Dade County
Lunn, John, Steen, Todd P., Bonnema, Thomas, Journal of Forensic Economics
The Supreme Court made state and local racial set-aside programs subject to strict scrutiny when it struck down Richmond, Virginia's program in 1989 (City of Richmond v. J. A. Croson Co., 488 U.S. 469). The decision made state and local governments undertake studies to determine whether discrimination existed in the relevant markets before a race-conscious program could be used. This requirement has generated a spate of such studies, usually statistical in nature, to provide the factual "predicate" for the race-conscious programs.
The predicate studies have evolved over the years since the Croson decision. The early studies relied mostly on anecdotal evidence combined with disparity analysis. Disparity ratios compare the utilization of groups that are believed to face discrimination with the proportion of the relevant population held by the groups. The use of regression analysis has been added in some recent studies, although disparity analysis remains the most commonly-employed tool in predicate studies.
Of course, statistical analysis is only as good as the data used in the analysis and the model. Empirical models must be derived from sound theoretical models if one is to have any confidence that the empirical results can be used to make inferences about the existence of discrimination in a market. In this note, we examine some of the regression work used in the predicate study for Dade County in Florida. It is an example that illustrates some of the pitfalls of statistical analysis when the connections between theory and practice are missing. As such, the example can be valuable for any econometric work performed for litigation.
I. Statistical Analysis in Predicate Studies
In the majority opinion of Croson, Justice O'Connor stated that it was inappropriate to compare the utilization of minority-owned firms with the proportion of the greater population that is minority, and that trying to determine the number of minority-owned firms that would exist but for past discrimination was "sheer speculation." Instead, she stated that the existence of disparity between the number of qualified minority-owned firms that were willing and able to perform a particular service and the number of firms actually utilized in the market could lead to an inference of discrimination. The researcher then must determine which firms are wining and able to perform a service to be able to make the proper comparisons.
Much of the post-Croson litigation and predicate studies have involved the construction industry, including the litigation involving Dade County. So, we will focus on the construction industry in our discussion. The typical method employed by practitioners of predicate studies is to take licensing as an indication of qualifications, and sometimes as an indication of "willing and able." Another way of determining "willing and able" is to look at firms that have tried to obtain contracts from the relevant government agency. Then disparity ratios are calculated, in which the utilization of minority- or women-owned firms relative to availability are compared to the utilization of firms owned by white males relative to capacity. Disparity ratios less than unity indicate that minority firms are not used as much as majority firms (relative to capacity). If the difference from one is statistically significant, then the inference of discrimination is often drawn.
There are several problems with inferring discrimination on the basis of disparity analysis alone. First, licensing may not be the appropriate measure to determine which firms are qualified. Licensing requirements do not consider business practices and experience that may be important. For example, a city or state that wanted to build a bridge may want to look at contractors that have built bridges before, or who have worked on similar projects. Second, size may matter. A small firm may be qualified for some types of jobs and still not be qualified for large and technically complex projects. …