Improving HMDA Data Quality: Is HMDA Data Quality Good Enough to Justify Enforcement Actions?

Link to PDF Version

Summary:

  • Regulators use of HMDA data to evaluate mortgage lending activity can result in significant fines and penalties to mortgage lenders[1].
  • Mortgage lenders use of HMDA data released by the Federal Financial Institutions Examination Council (“FFIEC”) can prevent, detect, and correct potential issues using advanced analytics and peer benchmarking. 
  • Publicly available HMDA data, however, is not updated for late and/or revised filings thus calling into question regulatory actions based the original data and compromising the effectiveness of mortgage lenders’ responsible conduct protocols.[2]

Analysis:

Our last HMDA Insight[3] highlighted the delay in releasing HMDA data based on a comparison of the HMDA file modification dates to the public release date for those files.  A year-over-year comparison of the HMDA file modification dates for the period 2010 – 2014 shows that the HMDA files were not changed after their release to the public.  This indicates that the HMDA data was not subsequently updated to incorporate late and/or revised HMDA filings.

The likely justification for not issuing revised HMDA data files is that the number of applications involved is immaterial.  While there are no public comments – or supporting data – from regulators to substantiate this justification, our analysis shows that the scope of incomplete and/or inaccurate HMDA data is sufficient to merit the issuance of revised HMDA files.  Specifically:

  • Our analysis of HMDA data for the period 2010 -2014 (Table 1) identified six cases indicative of incomplete data. Our review of the result found instances of non-compliance with HMDA reporting requirements.  In one case the lender was unaware that their filing, consisting of more than 25,000 applications, had not been accepted by the FFIEC and was not included in the FFIEC’s public HMDA file. [4]
  • Mortgage TrueView’s LenderScore [5] algorithms found that aberrant B Scores were attributable to inaccurate data involving more than 25,000 applications per year over the five-year period.

A recent discussion with regulators indicated the most likely reason HMDA data files aren’t revised, at least in the case of incomplete data, is the lack of regulatory analytics (such as those summarized in Table 1) to identify HMDA non-filers.  This rather surprising admission raises the possibility that regulatory actions based on HMDA data lack credibility.

It’s unclear if the credibility of regulatory actions is further compromised by revised HMDA filings.  If the database used by regulators is not updated for revised HMDA filings, then regulatory actions likely lack credibility.  If regulator databases are updated for revised HMDA files, the results are likely credible; however, mortgage lenders would be justified in arguing that they didn’t have the information needed to responsibly, and proactively, address the issues that led to a regulatory action.

Table 1 | Incomplete HMDA Data Cases and Assessment (2010 – 2014)

Case

Results Comment

Assessment

  1. 2010 Respondents reporting more than 100 applications with no reported 2011 applications

7 Respondents

All but one of the 7 respondents reported applications in 2012, 2013, and 2014 Probable that results indicate missing LAR from the identified Respondents.
  1. 2011 Respondents reported more than 100 applications with no reported applications in 2012.

7 Respondents

All but one of the 7 respondents reported applications in 2013 and 2014.  In addition, all but one reported applications in 2010. Probable that results indicate missing LAR from the identified Respondents.
  1. 2012 Respondents reported more than 100 applications with no reported applications in 2013.

4 Respondents

All four respondents reported applications in 2010, 2011, and 2014. Probable that results indicate missing LAR from the identified Respondents.
  1. 2013 Respondents reported more than 100 applications with no reported applications in 2014.

186 Respondents

All but 10 of the 186 respondents reported applications in 2012.The Respondent reporting the most applications was an acquisition. Unable to fully differentiate between Respondents that stopped originating mortgages and those that may have failed to submit a LAR until release of 2015 HMDA data.
  1. 2014 Respondents that reported more than 100 applications in 2014 with no reported applications in 2013.

10 Respondents

All Respondents reported applications in at least one other year with four (4) of the 10 reporting in 2010, 2011, and 2012. Probable that results indicate missing LAR from the identified Respondents.
  1.  2014 Respondents reporting more than 100 applications with no reported applications in any of the prior periods.

133 Respondents

The top Respondent based on application count was involved in an acquisition.There are 16 other respondents that reported more than 1,000 applications in 2014. Unable to differentiate between new Respondents and those that may have failed to submit a LAR in one or more of the preceding periods.

 * * *

“It is a capital mistake to theorize before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts.” ― The Adventures of Sherlock Holmes, A Scandal in Bohemia

 * * * *

[1] Examples include http://files.consumerfinance.gov/f/documents/201606_cfpb_bancorpsouth-consent-order.pdf (Accessed on July 27, 2016), http://files.consumerfinance.gov/f/201511_cfpb_hudson-city-consent-order.pdf (Accessed on July 27, 2016), http://files.consumerfinance.gov/f/201310_cfpb_consent-order_washington-federal.pdf (Accessed on July 27, 2016) and http://files.consumerfinance.gov/f/201310_cfpb_consent-order_mortgage-master.pdf).  (Accessed on July 27, 2016).

[2] See “Responsible Business Conduct: Self-Policing, Self-Reporting, Remediation, and Cooperation” located at http://files.consumerfinance.gov/f/201306_cfpb_bulletin_responsible-conduct.pdf. (Accessed on October 2, 2015).

[3] See HMDA Insights: Earlier Release of HMDA Data, Volume 7, July 18, 2016.

[4] The mortgage lender provided us with a copy of the data and it was loaded into Mortgage TrueView. The results presented in Table 1, therefore, exclude this instance of incomplete data

[5] See LenderScores.com.

Leave a Reply

Your email address will not be published. Required fields are marked *