NCRC’s HMDA 2018 methodology: How to calculate race and ethnicity

In 2010, the Dodd-Frank Wall Street Reform and Consumer Protection Act instructed the Consumer Finance Protection Bureau (CFPB) to fundamentally change the data collected under the Home Mortgage Disclosure Act (HMDA). Among the changes described in the bill was guidance on collecting more detailed data about the ethnicity and race of borrowers seeking mortgage loans.

To accomplish this, the CFPB issued a final rule in 2015, amending Regulation C of HMDA to implement these changes. Prior to this time, HMDA collected a single ethnicity and up to five racial groups for each applicant and co-applicant. The possible ethnicities included:

  • Code 1—Hispanic or Latino 
  • Code 2—Not Hispanic or Latino 
  • Code 3—Information not provided by applicant in mail, internet or telephone application 
  • Code 4—Not applicable 
  • Code 5—No co-applicant (for use in the co-applicant fields only)

 

For each of the five racial groups, the possible choices were:

  • Code 1—American Indian or Alaska Native 
  • Code 2—Asian 
  • Code 3—Black or African American 
  • Code 4—Native Hawaiian or Other Pacific Islander 
  • Code 5—White 
  • Code 6—Information not provided by applicant in mail, internet or telephone application 
  • Code 7—Not applicable 
  • Code 8—No co-applicant  (for use in the co-applicant fields only)

 

The result of the 2010 Dodd Frank Act, the 2015 CFPB rule and the 2018 Congressional update was the expansion of ethnicity fields from one possible choice to five for both the applicant and co-applicant. In addition, new sub-categories were added to both the ethnicity and racial fields in order to better define the borrowers and conform better to census data used in conjunction with HMDA.

The new ethnicity codes were outlined by the bureau in their instructions to the industry as follows:

  • Code 1—Hispanic or Latino 
    • Code 11—Mexican
    • Code 12—Puerto Rican
    • Code 13—Cuban
    • Code 14—Other Hispanic or Latino
  • Code 2—Not Hispanic or Latino 
  • Code 3—Information not provided by applicant in mail, internet, or telephone application 
  • Code 4—Not applicable 
  • Code 5—No co-applicant (for use in the co-applicant fields only)

 

New racial subcategories now exist as well;

  • Code 1—American Indian or Alaska Native 
  • Code 2—Asian 
    • Code 21—Asian Indian
    • Code 22—Chinese
    • Code 23—Filipino
    • Code 24—Japanese
    • Code 25—Korean
    • Code 26—Vietnamese
    • Code 27—Other Asian
  • Code 3—Black or African American 
  • Code 4—Native Hawaiian or Other Pacific Islander 
    • Code 41—Native Hawaiian
    • Code 42—Guamanian or Chamorro 
    • Code 43—Samoan
    • Code 44—Other Pacific Islander
  • Code 5—White 
  • Code 6—Information not provided by applicant in mail, internet, or telephone application 
  • Code 7—Not applicable 
  • Code 8—No co-applicant  (for use in the co-applicant fields only)

 

In the 2018 data, 15,119,651 loan applications records (LAR) were reported under HMDA.   Using the following method, we have combined all of the ethnicity categories into a single “Combined Applicant Ethnicity” field in Tableau, although the formula should easily translate to other formats.

if [Applicant Ethnicity 1] > 10 then [Applicant Ethnicity 1] ELSEIF [Applicant Ethnicity 2] > 10 then [Applicant Ethnicity 2] ELSEIF [Applicant Ethnicity 3] > 10 then [Applicant Ethnicity 3] ELSEIF [Applicant Ethnicity 4] > 10 then [Applicant Ethnicity 4] ELSEIF [Applicant Ethnicity 5] > 10 then [Applicant Ethnicity 5] else [Applicant Ethnicity 1] END

This method favors the reporting of Hispanic subgroups over Hispanic in order to account for observed multiple uses of different subgroups by many applicants. Should [Applicant Ethnicity 1] be the only filed to hold data, then it is what is reported. Likewise, if [Applicant Ethnicity 1] is a subgroup designation then it is reported. However, in other circumstances, the logic looks at [Applicant Ethnicity 2] and if it is a subgroup designation that is reported instead of the data in the first field. This intentionally prioritizes subgroup data. It is apparent that many borrowers used multiple ethnicities and races to identify themselves. In many cases, they identified first as Hispanic in [Applicant Ethnicity 1] and then as a specific subgroup(s) in the subsequent fields.  This logic favors the first subgroup it encounters and then if no subgroup is identified it defaults to the data in [Applicant Ethnicity 1]. In 2018, HMDA identified the following ethnic disaggregation for Applicants.

  • Null (no data)
    • 4,347 applications
  • Code 1—Hispanic or Latino
      • 800,525 applications
    • Code 11—Mexican
      • 375,781 applications
    • Code 12—Puerto Rican
      • 88,532 applications
    • Code 13—Cuban
      • 41,036 applications
    • Code 14—Other Hispanic or Latino
    • 175,893
  • Code 2—Not Hispanic or Latino 
    • 10,387,980 applications
  • Code 3—Information not provided by applicant in mail, internet, or telephone application 
    • 1,796,191 applications
  • Code 4—Not applicable 
    • 1,449,365 applications
  • Code 5—No co-applicant (for use in the co-applicant fields only)
    • 1 applications (this is obviously a reporting error)

 

In total, out of borrowers that identified as Hispanic (1,481,767) a total of 681,242 (45.97%) reported some sort of subgroup.  A smaller subset reported multiple subgroups.

Applicant race was also combined using a similar method.

if [Applicant Race 1] > 20 then [Applicant Race 1] ELSEIF [Applicant Race 2] > 20 then [Applicant Race 2] ELSEIF [Applicant Race 3] > 20 then [Applicant Race 3] ELSEIF [Applicant Race 4] > 20 then [Applicant Race 4] ELSEIF [Applicant Race 5] > 20 then [Applicant Race 5] else [Applicant Race 1] END

Here is the disaggregated racial data.

  • Null
    • 2,834 applications
  • Code -1 (negative 1)
    • 1 applications
  • Code 1—American Indian or Alaska Native 
    • 112,295 applications
  • Code 2—Asian 
      • 320,999 applications
    • Code 21—Asian Indian
      • 139,655 applications
    • Code 22—Chinese
      • 111,317 applications
    • Code 23—Filipino
      • 75,189 applications
    • Code 24—Japanese
      • 19,921 applications
    • Code 25—Korean
      • 37,955 applications
    • Code 26—Vietnamese
      • 47,595 applications
    • Code 27—Other Asian
      • 60,954 applications
  • Code 3—Black or African American 
    • 1,034,828 applications
  • Code 4—Native Hawaiian or Other Pacific Islander 
      • 24,849 applications
    • Code 41—Native Hawaiian
      • 5,957 applications
    • Code 42—Guamanian or Chamorro 
      • 2,320 applications
    • Code 43—Samoan
      • 2,026 applications
    • Code 44—Other Pacific Islander
      • 21,558 applications
  • Code 5—White 
    • 9,846,536 applications
  • Code 6—Information not provided by applicant in mail, internet, or telephone application 
    • 1,804,759 applications
  • Code 7—Not applicable 
    • 1,448,103 applications

 

Among the Asian and Native Hawaiian applicants, the desire to report disaggregated data is obviously strong, with 60.26% of the applications from these groups including a subgroup in one of the ethnicity or race fields.

This data is obviously a groundbreakingly complex addition to the HMDA dataset and provides ample opportunities for a spectrum of new research and fair lending enforcement efforts. The enthusiastic response of borrowers to the collection of this data shows that despite lending industry protests, consumers feel this data is critical and not a burden to their loan experience.

Scroll to Top