Journal of Real Estate Finance and Economics, 19:2, 147±159 (1999)
# 1999 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
Differences in the Cost of Mortgage Credit
Implications for Discrimination*
GORDON W. CRAWFORD AND ERIC ROSENBLATT
Fannie Mae 3900, Wisconsin Avenue, NW Washington, DC 20016-2892
This paper estimates the mortgage interest rate differences paid by Asian, Hispanic, and African±American
borrowers to a national home mortgage lender in the years 1988±1989. Controlling for differences in market rates,
rate lock protection, and borrower risk factors, conventional loan interest rates are almost perfectly race-neutral.
The single deviation from race-neutrality is that when interest rates fall during the borrower's rate-lock period,
only African±American borrowers are unable to capture a share of this decline. Government (FHA and VA) credit
models show small premia paid by African±American borrowers of about $1.80 per month on average. In
government lending, Hispanic borrowers alone are unable to capture rate declines occurring during the
borrower's rate-lock period.
Most research on racial discrimination in home ®nance accepts, as a simplifying
assumption, that interest rates in home mortgage markets do not differ across borrowers.
Examples from the redlining literature include Avery and Buynak (1981), Shlay (1988),
Bradbury et al. (1989), Gabriel and Rosenthal (1991), Hula (1991), Holmes and Horvitz
(1994), Berkovec et al. (1994), and Rossi and Phillips-Patrick (1995). Examples in the
debate over differential denial rates in underwriting include Canner and Smith (1991,
1992), Kohn et al. (1992), Munnel et al. (1992), Liebowitz (1993), Glennon and Stengel
(1994), Horne (1994), and Ferguson and Peters (1995). While not always explicit, the basis
for this view of a single interest rate is the credit rationing model of Stiglitz and Weiss
(1981) and Williamson (1986, 1987). However, in this paper, we will present evidence of
signi®cant price differences between individual borrowers, even for a single lender, and
even holding product and time constant. The failure to observe this price dispersion in the
past is probably due to a de®cit of good data, particularly of mortgage terms and timing.
For instance, Benston and Horsky (1992) know only the year of the loan, though annual
market changes are often over one hundred basis points.
This paper accesses the mortgage origination processing system of a major national
*The views expressed in this paper are not necessarily those of Fannie Mae. No Fannie Mae data sources were
used in this paper. This paper was previously circulated with the title, ``Borrower and Minority Differences in
Home Mortgage Interest Rates.''