Guidelines for Conducting a Disparity and Availability Study for the Federal DBE Program (2010)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

65 Large and adverse statistical disparities between minority- owned or women-owned businesses and nonminority male- owned businesses have been documented in numerous re- search studies and reports since Croson.200 Business outcomes, however, can be influenced by multiple factors, and it is im- portant in disparity studies to examine the likelihood of whether discrimination is an important contributing factor to observed gross disparities. One traditional way that the linkage between statistical dis- parities and discrimination has been established is through the introduction of anecdotal or qualitative evidence. If the thrust of such anecdotal evidence is consistent with the dis- parities observed, the case for the linkage is strengthened. Another traditional way that the linkage between statisti- cal disparities and discrimination has been established is to consider the size of the observed disparities. That is, the larger the disparity, the less likely it becomes that nondiscrimi- natory factors can account for the entire difference. It is this straightforward observation that underpins the Equal Employment Opportunity Commission’s long-standing “four-fifths rule” for triggering employment discrimination investigations.201 Some critics of race-conscious contracting programs and some courts have criticized the validity of the use of the four- fifths rule in combination with anecdotal evidence on grounds that the availability measure in the disparity statistic does not factor in “capacity” or, stated another way, because availabil- ity statistics may include firms that are not “qualified, willing, and able” to perform the work. One critic has called this “the most common disparity study fallacy.”202 For several reasons, such criticisms are unwarranted and unscientific. First, it is helpful to consider an extreme example where discrimination has prevented the emergence of any minority- owned firms. Suppose that racial discrimination was ingrained in the state highway construction market. As a result, few mi- nority construction employees are given the opportunity to gain managerial experience in the business; minorities who do end up starting construction firms are denied the oppor- tunity to work as subcontractors for nonminority prime con- tractors; and nonminority prime contractors place pressure on unions not to work with minority firms and on bonding companies and banks to prevent minority-owned construc- tion firms from securing bonding and capital. In this exam- ple, discrimination has essentially prevented the emergence of a minority highway construction industry with “capacity.” Excluding firms based on their “capacity” in a discriminatory market would preclude a government agency from doing anything to rectify the continuing support of such a system with public dollars. There is no recognition that discrimina- tion has prevented the emergence of “qualified, willing, and able” minority firms. Without such firms, there can be no sta- tistical disparity, and without a statistical disparity there can be no remedy. The government is not so helpless in the face of the current effects of discrimination, however. Now, one might argue that this result is correct. The state should try to prevent the discriminatory impediments, and by doing so it will stimulate the formation of qualified, will- ing, and able firms. Of course, that ignores the logic of affir- mative action. If injunctive relief could always remedy dis- crimination, affirmative action would never be necessary. Affirmative action is used as a tool for minimizing discrimi- nation against minorities and women without having to reg- ulate decisions at every step along the line. Second, terms such as “capacity,” “qualifications,” and “abil- ity” are not well defined in any statistical sense. Should “capac- ity” be defined as revenues, employment size, or certain bond- ing limits? Does “qualified” or “able” mean possession of a business license, or certain amounts or types of training or work A P P E N D I X B Understanding “Capacity” 200 See Enchautegui, et al. 1996. 201 The four-fifths rule says that any disparity ratios less than or equal to 0.8 (on a scale of zero to one, zero being perfect disparity and one being perfect parity) indicate the presence of discrimination. See 29 C. F. R. § 1607.4(d). 202 La Noue 1994, p. 490.

experience? How is a government agency supposed to obtain information about such factors for subcontractors? Also, does the meaning of these terms differ from industry to industry or state to state? Third, in dynamic business environments, and especially in the construction sector, such “qualifications” can be ob- tained relatively easily. It is well known that small construc- tion companies can expand rapidly as needs arise by hiring workers and renting equipment. Many general contractors subcontract the majority of a project. Subcontracting is one important source of this elasticity, as has been noted by sev- eral academic studies. Bourdon and Levitt, for example, in their study of construction labor markets, observed that: “One of the unique aspects of the construction industry is the prevalence of subcontracting. Construction projects are under- taken by a multitude of firms assembled for brief periods of time on a site then disbanded. General contractors can under- take projects of considerable scale without large amounts of di- rect labor or fixed capital; subcontractors can start with one or two employees and bid only on particularly highly specialized contracts.”203 Eccles also noted the importance of subcontracting in con- struction.204 He found that subcontracting could be explained as a response to uncertainty and complexity. He also found that the larger the project the more subcontracting and the more extensive the market the more subcontracting. Dowall and Barone draw a similar conclusion regarding the use of subcontractors.205 Academic studies have also found that, absent discrimina- tion, entry into the construction industry is not difficult. Bourdon and Levitt attribute this to subcontracting opportu- nities.206 Eccles observes that entry is easy based on the large number of small firms and that capital requirements for fixed assets are small.207 Gould, who followed the careers of six con- struction contractors, also demonstrates ease of entry.208 He also notes that there is movement between small and large firms not only via subcontracting, but also by experienced staff at larger firms leaving to form smaller new firms. Dowall and Barone, based on a survey of construction firms, note that there is “considerable diversification into other types of construction activities.”209 The construction market is dynamic, facing boom and bust periods. In response, the “capacity” and “qualifications” of firms in this sector remain highly elastic. Firms grow quickly when demand increases and shrink quickly when demand de- creases. Therefore, focusing on the “capacity” of businesses in terms of employment, revenue, bonding capacity, number of trucks, and so forth is wrong as a matter of economics and can potentially obscure the existence of discrimination. To see this, consider using revenue as the measure of qualifica- tions. Revenues simply measure the value of contracts that firms are receiving. If minority-owned and women-owned businesses are subject to marketplace discrimination, their revenues will be smaller than nonminority male-owned busi- nesses because they will be less successful at obtaining work. Using revenues as a measure of DBE availability in contract- ing is like using pay as a measure of qualifications in an equal- pay case. Revenue, like pay, measures the extent to which a firm has succeeded in the marketplace—it does not measure the ability to succeed. Fourth, suppose for the sake of argument that DBE avail- ability should be based on detailed “qualifications” or “capac- ity” measures like bonding capacity, working capital, years of experience, and other items. Where would one obtain the data? The critics do not tell us and neither do the courts. In the Concrete Works trial, for example, a plaintiff’s expert com- plained that the availability measures proffered by the defen- dant’s expert, which controlled for detailed industry affiliation and geographic location, nevertheless did not contain the de- tailed information on firm “qualifications” that he believed were necessary for a proper analysis.210 However, plaintiff pre- sented no data whatsoever on any detailed qualifications of firms or their owners, the share of minority-owned firms that have achieved a particular bonding capacity, the average amount of capital equipment, amount of working capital, suc- cess on previous jobs, or any other possible metric of qualifi- cations. Such information does not exist and would be diffi- cult if not impossible to collect in any systematic fashion. Indeed, plaintiff’s expert admitted this in his deposition: “[t]here isn’t data in any database that tells you what all those qualifications could be.”211 To the best of our knowledge, no plaintiff’s expert has ever introduced statistical evidence demonstrating that accounting for “qualifications” or “capac- ity” explains away large and statistically significant disparities facing minority-owned or women-owned firms. Fifth, although it is true that some disparity studies have not controlled for factors such as “capacity” or “qualifica- tions” or “willingness,” it is not necessarily their obligation to do so. Although Croson provides little guidance on this mat- ter, using a disparity study to consider whether there is a prima facie case of disparate impact against certain groups follows a long-established pattern in employment discrimi- 66 203 Bourdon and Levitt, 1980. 204 Eccles, 1981. 205 Dowall and Barone, 1993. 206 Bourdon and Levitt, 1980. 207 Eccles, 1981. 208 Gould, 1980. 209 Dowall and Barone, 1993. 210 La Noue, 1998a, pp. 31–37. 211 La Noue, 1998b, p. 140.

nation litigation.212 For example, by demonstrating that gross statistical disparities facing a given group of minority busi- ness owners were both large and statistically significant, the burden of proof shifts to the plaintiff, who must then demon- strate that the gross disparities in evidence diminish substan- tially in size or statistical significance (or both) once other influential factors that are unlikely to be correlated with dis- crimination have been accounted for.213 As we have already argued, most of these other factors are strongly correlated with discrimination. Moreover, in those cases where plaintiff’s experts have had the opportunity and the incentive to counter defendant’s disparity statistics with their own statistics regard- ing “capacity” and “qualifications,” they have failed to do so because such data do not exist. Sixth, even in cases where “qualification”-type factors have been controlled for in statistical analyses, results consistent with business discrimination are still typically observed. For example, as we noted above, Denver demonstrated that large and statistically significant differences in commercial loan de- nial rates between minority and nonminority firms were evi- dent even when detailed balance sheet and creditworthiness measures were held constant.214 Similarly, economists using the decennial census microdata have demonstrated that statistically significant disparities in business formation and business owner earnings between minorities and nonminori- ties remain even after controlling for a host of factors avail- able in the data, including educational achievement, labor market experience, marital status, locational mobility, num- ber of workers in the family, number of children, immigrant status, disability status, veteran status, interest and dividend income, labor market attachment, industry, geographic loca- tion, and local labor market variables such as the unemploy- ment rate, population growth rate, government employment rate, and per capita income.215 Noted labor economist and former U.S. Secretary of Labor Ray Marshall, in partnership with former Federal Reserve Board Governor Andrew Brimmer, conducted one of the first post-Croson disparity studies for the City of Atlanta in 1990. Marshall summarizes well the arguments against using the outcomes of discrimination to measure “capacity”:216 The problem of establishing statistical proof of whether or not minority contractors are “qualified, willing and able” is particu- larly challenging. Croson provides limited guidance on this ques- tion. . . . Unfortunately, this lack of guidance has made it possible for courts and opponents of [race-conscious contracting] pro- grams to argue that the failure to produce perfect statistical evidence—i.e., timely and highly specific, and methodologies that control for everything except discrimination—invalidates these programs despite the fact that the most reliable statistics and the most appropriate methodologies confirm the persistence of dis- crimination. Our evidence for Atlanta suggests that even highly qualified black contractors are disadvantaged relative to similarly situated white contractors. . . . It also is hard to know how to de- fine the qualifications of businesses in dynamic markets where ex- pertise can be purchased in the open market and where “virtual” companies are increasingly common. Once contractors are able to obtain contracts, they usually are able to expand their capacity. In a dynamic business environment, it would be difficult to argue, as some critics have, that qualifications are determined mainly by size. . . . Moreover, as the Tenth Circuit Court of Appeals ob- served in Adarand VII, there is no credible evidence that minor- ity contractors who have been hired under [race-conscious con- tracting] programs have lacked adequate qualifications. Nevertheless, analyses of available data for business owners that enable personal characteristics and other factors to be controlled for [generate results that are] compatible with racial exclusion. There therefore is no credible evidence that the large disparities in the utilization of minority contractors can be explained by the lack of qualifications or the unwillingness to contract. Indeed, strong historical, anecdotal and survey evidence . . . demon- strates that minority contractors are more willing than white males to contract with governmental entities, even though they recognize that public contracting is less desirable than the main- stream private sector, where their opportunities are greatly re- stricted. The greater participation of minorities and women is compatible with the concept of “crowding,” mentioned earlier. This is all the more reason not to use participation in these sec- tors as a measure of discrimination and why broader market areas are more appropriate. To summarize, the statistical analysis of the availability of minority firms compared to nonminority firms to examine the existence and effects of discrimination in disparity stud- ies should not adjust for “capacity” because: • “Capacity” has been ill defined; • Small firms, particularly in the construction industry, are highly elastic with regard to ability to perform; • Many disparity studies have shown that even when “capacity”- and “qualifications”-type factors are held constant in statistical analyses, evidence of disparate im- pact against DBE and M/WBE firms tends to persist; and • Most important, identifiable indicators of capacity are themselves impacted by discrimination. 67 212 See Connolly, et al., 2001, chs. 2–3. 213 In the present context, factors that are uncorrelated with discrimination are referred to as exogenous variables. Factors that are correlated with discrimina- tion are referred to as endogenous variables. Only exogenous variables should be included as explanatory factors in a statistical model testing for disparities. 214 See Chapter Two, Review of Existing Studies. 215 E.g., Wainwright, 2000, pp. 85–135. 216 Marshall, 2002, pp. 81–82