Boston HMDA Data Analysis Re-Visited

33 Pages Posted: 26 Dec 2012

Date Written: December 25, 2012


This paper revisits the Boston HMDA data and concludes the original findings were accurate and examine using a proxy for credit guidelines without the racial impact on the credit guidelines variable. Examining the variables closely and capping variables using common statistical techniques instead of throwing data away shows the racial impact to persist. In addition the paper also examines the racial impact via a matching approach in which a sample of matched majority applicants who are most similiar to minorities are selected based on the underwriting variables. This matching is based on a logistic regression predicting being a minority based on other variables. Unbiased re-cursive partitioning is also used to understand what it means to be a minority in terms of variables which might be used in underwriting to cause disparate impact. Having a holistic understanding of the interactions of the weak financial variables and minority status provide a fuller picture to understand the phenomenon at play and why so many studies can get misleading results given the correlated and confounding relationships in the data. The nuanced approach outline allows the full implication of the racial variable to be seen, which comes out larger once the racial bias in the meets credit guideline variable is addressed or matched sample of similar applicants is examined.

Keywords: fair lending, boston HMDA, discrimination, mortgage lending, disparate impact

Suggested Citation

Sharma, Dhruv, Boston HMDA Data Analysis Re-Visited (December 25, 2012). Available at SSRN: or

Dhruv Sharma (Contact Author)

Independent ( email )

2023 N. Cleveland St.
Arlington, VA 22201
United States


Here is the Coronavirus
related research on SSRN

Paper statistics

Abstract Views
PlumX Metrics