Fair Lending Pricing Analytics

1
Fair Lending Pricing Analytics
Presented at the ABA Pricing Modeling Symposium,
June 2, 2007 Katherine Samolyk, Senior
Economist, Division of Insurance and Research,
Federal Deposit Insurance Corporation
2
Disclaimer

• These remarks reflect my personal observations
  and opinions and should not be attributed as
  reflecting official policies and practices of the
  FDIC

3
Fair Lending Analytics

• My Division (the Division of Insurance and
  Research) provides analytical support for fair
  lending examinations conducted by the FDICs
  Division of Supervision and Consumer Protection
• Statistical analysis to compare credit outcomes
  for a target group (a racial-ethnic minority
  group or females) with credit outcomes for a
  control group (NH Whites or males)

4
DIR Fair Lending Examination Support

• Offsite screening DIR developed and runs
  statistical screens using HMDA pricing data to
  identify banks that appear to be at risk in
  terms of disparities in mortgage loan pricing to
  racial/ethnic minorities or women.
• Analytical support for on-site exams Conduct
  statistical analysis of loan data to investigate
  potential pricing disparities identified by FDIC
  screens or by DSC examiners
• Statistical analysis
• What measure of pricing will be examined
  (dependent variable)
• What sample of loans to look at
• What variables to include in the model
• What statistical tests to use

5
FDIC-Supervised Banks

• Many relatively small institutions
• Many in rural areas
• Examiner-identified non-mortgage pricing cases
• Emphasis on assisting with preliminary screening
  analysis Use readily available data to evaluate
  disparity. If a HMDA outlier, does the suspicious
  pattern continue in recently collected HMDA data?

6
Modeling Pricing Outcomes

• Analysis depends on pricing policies and
  realities
• Criterion Interviews conducted by DSC examiners
  ascertain pricing policies
• Across lending units
• Across products offered
• Across markets
• Available data

7
Modeling Pricing Outcomes

• Pricing policies
• Are there clear non-discriminatory criteria for
  pricing?
• To what extent are outcomes automated (factors
  specified on rate sheets,) versus judgmental
  (factors considered in judgmental fashionsuch as
  customer relationship)?
• To what extent are rate sheet deviations
  permitted?
• YSP compensation agreements and compensation?
• What are the banks markets and the competitive
  factors ?
• Is reality consistent with bank policies?
• Is there documentation of factors used to price
  in loan files?
• Are exceptions to rate sheets documented?

8
Modeling Pricing Outcomes

• Dependent variableit depends
• Note rate or APR
• Rate sheet deviation
• Yield spread premium
• Incidence of loans extended in high-price unit
• Explanatory variablesit depends
• Examiner criterion interviews
• Information in electronic files or compiled by
  DSC examiners from loan files
• Hard to include information if the bank doesnt
  have it on file

9
Evaluating Outcomes

• Statistical significanceobserved differences in
  outcomes for the target and the control group
  would be very unlikely to occur if the source of
  the differences were just random chance
• Economic significancewhat is the magnitude of
  the pricing differential ?