21.07.2013 Views

The State of Minority- and Women- Owned ... - Cleveland.com

The State of Minority- and Women- Owned ... - Cleveland.com

The State of Minority- and Women- Owned ... - Cleveland.com

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Statistical Disparities in Capital Markets<br />

the firm applied. 274 In total, these three columns add 176 variables to the more parsimonious<br />

specification reported in Column (2). 275 Nevertheless, the estimated disadvantage experienced by<br />

African American-owned firms in obtaining credit remains large <strong>and</strong> statistically significant. <strong>The</strong><br />

estimate from each <strong>of</strong> the three additional columns indicates that African American-owned firms<br />

are 24 percentage points more likely than nonminority male-owned firms to have their loan<br />

application denied even after controlling for the multitude <strong>of</strong> factors we have taken into<br />

consideration.<br />

<strong>The</strong> results also indicate that Asians/Pacific Isl<strong>and</strong>ers had significantly higher denial rates than<br />

nonminority males—12 percentage points. <strong>The</strong>re is little evidence in the 1993 national data,<br />

however, that denial rates for firms owned by Native Americans or Hispanics were significantly<br />

different from the denial rates <strong>of</strong> firms owned by nonminorities; or that denial rates for firms<br />

owned by nonminority women were significantly different from those for firms owned by<br />

nonminority men. 276<br />

In Table 6.9, we see results for the ENC division similar to those reported in Table 6.8 for the<br />

nation as a whole. <strong>The</strong> table shows that the results <strong>of</strong> our loan denial model in the ENC are not<br />

substantially different from the nationwide results reported in Table 6.8. <strong>The</strong> indicator variable<br />

for the ENC division is negative <strong>and</strong> significant, indicating that denial rates, on average, are<br />

lower in the ENC than in the nation as a whole, but its significance decreases as more control<br />

variables are included in the model. None <strong>of</strong> interaction terms between race, ethnicity or gender<br />

<strong>and</strong> the ENC division indicator are statistically significant. 277<br />

274 Approximately four out <strong>of</strong> five (80.5%) <strong>of</strong> the firms who required a loan applied to a <strong>com</strong>mercial bank. Overall,<br />

seventeen different types <strong>of</strong> financial institutions were tabulated, although only the following accounted for more<br />

than 1% <strong>of</strong> the (weighted) total: Finance Companies (4.9%); Savings Banks (2.5%); Savings & Loans (2.3%);<br />

Leasing Companies (2.1%); <strong>and</strong> Credit Unions (2.0%).<br />

275 One piece <strong>of</strong> information to which we did not have access in the 1993 NSSBF or the 1998 SSBF because <strong>of</strong><br />

confidentiality concerns was each firm’s credit rating. A working paper by Cavalluzzo, Cavalluzzo, <strong>and</strong> Wolken<br />

(1999) was able to incorporate Dun & Bradstreet credit ratings for each firm because the authors’ connection to<br />

the Federal Reserve Board enabled them to access the confidential firm identifiers. <strong>The</strong>y added these credit<br />

rating variables in a model <strong>com</strong>parable to that reported here <strong>and</strong> found the results insensitive to the inclusion.<br />

<strong>The</strong> 2003 SSBF includes Dun & Bradstreet credit ratings for each firm. Below, we discuss the impact <strong>of</strong><br />

incorporating them into a model similar to that presented in Table 6.8 (see Tables 6.27 <strong>and</strong> 6.28).<br />

276 It would be a mistake to interpret a lack <strong>of</strong> statistical significance (as opposed to substantive significance) in any<br />

<strong>of</strong> the Tables in Chapter 6 as a lack <strong>of</strong> adverse disparity. While tests for statistical significance are very useful for<br />

assessing whether chance can explain disparities that we observe, they do have important limitations. First, the<br />

fact that a disparity is not statistically significant does not mean that it is due to chance. It merely means that we<br />

cannot rule out chance. Second, there are circumstances under which tests for statistical significance are not<br />

helpful for distinguishing disparities due to chance from disparities due to other reasons (e.g., discrimination). In<br />

the particular statistical application presented in this chapter, the chance that a test for statistical significance will<br />

incorrectly attribute to chance disparities that are due to discrimination be<strong>com</strong>es greater when relatively small<br />

sample sizes are present for an affected group.<br />

277 <strong>The</strong> number <strong>of</strong> Native Americans <strong>and</strong> Hispanics in the ENC sample was too small to yield statistical results.<br />

NERA Economic Consulting 198

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!