In message 27 Oct 1994 12:31:38 -0700,
pat mason <[EMAIL PROTECTED]> writes:
> Doug,
>
> I don't know if he's still on the list, but Gary Dymski would be a good
> source for this question. Also, try and contact Tim Bates at Wayne State.
>
> patrick "Murray and Hernnstein are racist pigs" mason
>
Greetings, Banking fans,
Since my old partner Patrick "m.a.h.a.r.p." Mason has suggested I get off my
bottom and reply to your query (a second time through!), let me try. Credit
scoring is simply one of many possible methods of assessing
creditworthiness. There is no reason that scoring -- assigning numerical
values to loan applicants -- is per se legal or illegal. What would make
any creditworthiness assessment technique illegal is this: if it involves
using any of the protected categories under the US's anti-discrimination
laws. By and large, this refers to the amendments to the Constitution that
have determined that due process protects people from being singled out on
the basis of (cet. par) race, gender, and age. That is, it is illegal to
give people a "score" using this category.
Do banks do it? Might they? Sure. The reason is, that race (eg) could
easily be highly correlated with a variety of real "economic" factors such
as, probability of job loss, credit history, job history, etc. In effect,
by using "race" as a flag (giving it a "score") the lender could economize
on its information-gathering costs (and not have to gather other information
on applicants). The same logic would suggest that racial composition of an
area would be a proxy for economic conditions in a particular locale. This
very practice was singled out by Guttentag and Wachter in a 1981 report for
the NYU Business School (published, I think, in pamphlet as "Redlining and
Bank Credit"). This Guttentag/Wachter article was the first to thus
indicate the possibility of what these authors termed "rational
discrimination".
The idea of "rational discrimination" has just resurfaced in the latest
JMCB, in an article by Calomiris et al on federal credit programs. Now,
this rationality rests on the use of a protected category as a short-hand
way of economizing on information-extraction costs. Strictly illegal. But
possible.
No bank would admit to using this, because of the illegality involved. So
no banking text would discuss it either. Does it happen? You can't ask a
bank. So there is a black-hole aspect to Doug Orr's question. The experts
can't ask this question of the lenders, without getting a "cold shoulder" in
response.
Rumors that banks use all kinds of things, including neural-network
algorithms into which are poured variables ranging from applicant race to
zip code to age of housing stock to ... But try to get Bank of America or
even your local bank to tell you their formula. They won't. You'll get
reams of material, but none that will get to the heart of your question.
BTW, in a working paper for UCR I go into some of this. Included within is
a discussion of several items of possible interest: (1) the ambiguity of
the very definition of "discrimination" in a multi-factor market setting;
(2) the importance of strategic interaction and spillover effects in
credit-market outcomes; (3) the importance of path dependence in creating
racial differences in credit flows. There are a lot of matters to think
through, and the model I set out in my working paper only barely breaks the
surface of what appears to be a huge set of inadequately-explored issues.
People who are interested in the working paper can ask for it by e-mailing
me directly at [EMAIL PROTECTED]
Gary Dymski
Gary A. Dymski [EMAIL PROTECTED]
Dept. Of Economics-Highlander Hall (909) 787 - 5037 ext 1570
University of California
Riverside, CA 92521 USA
