The Assessment Area That Stops at the Interstate

Assessment area delineation. An assessment area must consist of whole geographies, may not reflect illegal discrimination, and may not arbitrarily exclude LMI geographies. A boundary drawn along an interstate that neatly separates the minority and LMI tracts is the kind of decision that draws scrutiny on its face, regardless of intent. The CRA regulation and the interagency questions and answers treat this as a red flag for redlining.

Lending distribution. Six percent versus a 19 percent peer share in a county where the bank has branches is a disparity that cannot be explained by lack of physical presence. Examiners compare applications, originations, and denials against peers and against owner-occupied housing units by tract.

Denial disparity. A 31 percent denial rate against 18 percent bank-wide, combined with a 62 percent reliance on credit history, suggests either a genuine credit-quality difference or a reason-for-denial problem. Reg B requires specific reasons; credit history alone, without documented specifics, is weak in a file review and weaker still in a comparative analysis.

Pricing discretion. Twenty-two percent of one loan officer's FHA loans priced above the rate sheet, with a 2.5-point discretionary band and no exception log, is uncontrolled discretion. Discretion is where disparate impact and disparate treatment live.

Marketing invisibility. Bank-wide spend tracking means the bank cannot demonstrate it marketed in County C at all. In a redlining analysis, absence of evidence of outreach is treated as evidence of absence.

Redlining analysis. Compare application and origination distribution by census tract against peers, against owner-occupied units, and against the bank's own branch and marketing footprint. Pull the loan officer assignment map. Review the assessment area file for the rationale behind the interstate boundary. This tests whether credit is being made available on equal terms across the MSA.

Denial disparity analysis. Compute denial rates by race, ethnicity, and tract income, then control for credit characteristics -- credit score bands, DTI, LTV. Then sample County C denial files and score whether the adverse action notice states a specific reason supported by the file. This separates a legitimate credit story from a notice-quality problem from a treatment problem.

Pricing regression or matched-pair analysis. Model rate spread or APR against credit risk variables and a prohibited-basis indicator. Separately, review every one of the loan officer's above-sheet loans: was the exception documented, was it approved, was a rate sheet provided, and did borrowers of different races receive different treatment? Reconstruct the exception log backward.

HMDA data integrity sample. Re-verify a statistically valid sample of LAR fields against source documents (rate spread, action taken, denial reason, ethnicity, tract). A data error rate above a few percent undermines every other analysis and is its own finding.

CRA performance context. Full lending test analysis by tract income and borrower income, plus a service test review of branch locations, hours, and any alternative delivery channels in County C.

Take now, before the analysis finishes:

• Suspend discretionary overages above the published rate sheet pending a policy review, or at minimum require documented exception approval with a reason code and centralized logging effective immediately.
• Revise the adverse action workflow so credit history cannot be recorded as the sole reason without supporting detail.
• Preserve documents: pricing worksheets, exception approvals, loan officer correspondence, marketing plans, and the assessment area file. Issue a litigation hold.

Take after the analysis, because the fix depends on the finding:

• Redraw the assessment area to include the whole MSA, including the tracts east of the interstate. This is a decision that should be documented with the analysis behind it, not merely reversed.
• Expand marketing and outreach in County C, with tracked spend and community partnerships; CRA credit follows documented activity.
• Consider targeted lending programs or special financing for LMI borrowers, and review whether the branch presence in County C is adequate in staffing and hours.
• Remediate identified pricing or denial disparities through borrower reimbursement where the analysis supports it.

Ongoing program. Build quarterly fair lending monitoring with a written risk assessment, and put results in front of the board. The point of the program is to detect the next disparity before an examiner does -- the current fact pattern suggests no monitoring existed at all.