From: "E. Allen Smith" <EALLENSMITH@ocelot.Rutgers.EDU>
Date: Wed, 11 Dec 1996 16:28:43 -0800 (PST)
To: ichudov@algebra.com
Subject: Re: Redlining
Message-ID: <01ICW8E2LDSWAEL2GZ@mbcl.rutgers.edu>
MIME-Version: 1.0
Content-Type: text/plain


From:	IN%"ichudov@algebra.com" 11-DEC-1996 18:34:46.50

>Thanks for an interesting reply.

	Quite welcome. Due to the links to my own field, I've been thinking
about these issues for a bit.

>It is the cross-section test that evidences discrimination: suppose
>we have two large groups with the same credit-related parameters
>(credit history, etc). If one group gets better treatment from a
>bank, I see a discrimination.

	Is it the credit history that should be the same... or the outcomes?
In other words, if two groups of different races are all granted credit and
have equal credit histories, are the default rates greater for one group as
opposed to the other? If so, then there _is_ a fiduciary responsibility by
the bank's directors to not loan to the group that has greater default rates.
	Now, there's the problem in testing this that if a bank (or, more
likely, some loan officers at the bank) is practicing discrimination, those of
the discriminated-against races who do get credit evidently had _something_
that convinced the bank to still offer the people credit. This something may
or may not be controlled for by the credit factors equalization I mentioned
above. If it isn't controlled for, you'll get a bias in the outcome toward
the discriminated-against race being _better_ credit risks than they actually
are, on the average. This would remove any actual evidence of poor credit
riskworthiness on the part of that race (such as for cultural reasons, to the
(highly unfortunate) degree that race and culture are correlated).
	In other words, banks with prejudices (which I do fully admit are
out there) will tend to make any actual differences much harder - close to
impossible, in fact - to see.

>The crucial question is: do we believe that the characteristic
>that we are considering directly linked with future performance?

	As I point out above, studies on this topic are somewhat difficult,
and are prone to being biased in favor of discriminated-against races. (This
isn't even mentioning political biases...) You may or may not be able to
say _why_ a given characteristic is linked with future performance; it's
easy to come up with plausible explanations in most cases.

>This is a valid concern.

	Quite... I seem to recall a story involving a "Handicapper General" -
perhaps someone would recall the title and author? In it, those with higher
IQs were forced to be less intelligent through distractions, those with
greater beauty were forced to wear ugly masks, etcetera.

>> 	The third topic is that one commonly applied idea used by the
>> proponents of absolute equality is that found in Rawls' _Theory of
>> Justice_, under which the just outcome is said to be found by a group
>> of people who do not know what situation they will be in. (This is
>> a vast oversimplification of the book(s) in question, which upon
>> closer examination may realize the idea I am about to write down.)
>> The simplistic conclusion is that everyone will want everything to be
>> the same, since any individual might be in a bad or good situation. But
>> if you have a choice between 49 dollars and a 50/50 chance of 0 or 100
>> dollars, you should take the latter. In other words, a situation in

>Not necessarily.

	With the exception of needing <=49$ to live, under what conditions
would the former choice be better than the latter choice?
	-Allen