Real Estate (CRE & Residential) — Article 12 of 12

Compliance for RESPA, TILA, and State-Level Mortgage Regulations

Mortgage compliance now spans federal disclosure rules, 54 state and territorial licensing regimes, and a CFPB enforcement docket that has produced over $17B in consumer redress since 2011. Lenders that have automated TRID testing, fair-lending analytics, and state high-cost loan screening report 70-85% fewer post-close cures and dramatically lower repurchase risk.

11 min read
Real Estate (CRE & Residential)

A single residential mortgage in the United States now passes through roughly 3,000 distinct compliance checkpoints between application and post-close QC. The federal stack — RESPA, TILA, TRID, ECOA, HMDA, HOEPA, SAFE Act, FCRA, GLBA — sits on top of 54 state and territorial licensing regimes, several hundred state-specific high-cost loan thresholds, and a CFPB that has extracted over $17 billion in consumer redress since 2011. For a top-25 lender funding 200,000 loans a year, the cost of compliance has grown from roughly $1,400 per loan in 2008 to over $3,200 in 2024 according to MBA quarterly performance data. This article — the closing piece of the Property as a Platform guide — examines how automation, model governance, and integrated compliance engines are restructuring that cost base.

$3,200+Per-loan compliance cost for residential lenders in 2024, up from $1,400 in 2008 (MBA Quarterly Performance Report)

The RESPA Surface Area: Section 8, Affiliated Business, and TRID

RESPA's enforcement core remains Section 8, the prohibition on kickbacks and unearned fees in connection with a federally related mortgage loan. The CFPB's 2015 PHH Corporation action ($109M penalty, later reduced on appeal) and the 2023 Freedom Mortgage consent order signaled that marketing services agreements, lead-purchase arrangements, and co-marketing with real estate agents remain in the enforcement crosshairs. Compliance teams now run automated screens against vendor master files to detect prohibited fee splits, with platforms like ACES Quality Management and LoanLogics flagging Section 8 patterns — for example, a settlement service provider receiving payments that exceed fair market value for the services rendered, or a referral arrangement disguised as a desk-rental agreement.

Section 9 (seller-required title insurance), Section 10 (escrow account analysis), and the affiliated business disclosure regime each carry their own automation patterns. The escrow analysis rule requires annual reconciliation with a tolerance generally capped at two months of cushion. Servicing platforms like Black Knight MSP, ICE Mortgage Technology Servicing Digital, and Sagent's CARE compute the cushion in real time and generate the annual escrow statement automatically; the failure mode that still triggers CFPB matters is not the math but the timing of refund checks when surpluses exceed $50.

TRID — the 2015 integrated disclosure rule that fused RESPA's GFE/HUD-1 with TILA's early and final TIL — is where most lenders concentrate their automated testing. The Loan Estimate must be delivered within three business days of application; the Closing Disclosure must be received by the borrower at least three business days before consummation; and fee changes after the LE are constrained by three tolerance buckets. As covered in our earlier piece on loan origination from 45 days to 5 days, the timing rules and tolerance math have become the gating logic for any cycle-time compression program.

TRID Fee Tolerance Buckets
BucketExamplesToleranceCure Mechanism
Zero toleranceLender fees, transfer taxes, fees paid to affiliates, fees for services borrower cannot shop0%Refund full overage within 60 days of consummation
10% cumulativeRecording fees, services borrower can shop where lender provided list10% aggregateRefund excess over 10% threshold
No tolerancePrepaid interest, property insurance, escrow deposits, services borrower shopped off-listUnlimited variance permitted if good faithNo cure required if estimate made in good faith

Automated TRID engines from Wolters Kluwer (ComplianceOne), ICE Mortgage Technology (Encompass Compliance Service, formerly Mavent), and SitusAMC (ComplianceEase) run the full disclosure stack in milliseconds: TRID timing, APR/finance charge recalculation, Section 32 HOEPA triggers, state high-cost tests, and federal preemption logic. A typical implementation runs the test suite at four checkpoints — application, LE issuance, pre-CD, and pre-funding — and again post-close as part of QC sampling. Lenders that have integrated these engines into their LOS workflow report 70-85% reductions in post-close TRID cures and a measurable drop in repurchase demands from Fannie Mae and Freddie Mac, which both reference TRID defects in their Selling Guide Section D1-3.

TILA, Regulation Z, and the QM Rulebook

Regulation Z's substantive provisions extend far beyond disclosure. The Ability-to-Repay rule (12 CFR 1026.43) requires lenders to make a reasonable, good-faith determination of repayment ability using eight specified underwriting factors. The Qualified Mortgage safe harbor, restructured by the CFPB's 2021 General QM Final Rule, replaced the 43% DTI hard cap with a price-based test: a loan is QM if its APR does not exceed APOR by more than 2.25 percentage points for first-lien loans at or above the conforming loan limit (with higher thresholds for smaller loans). The Average Prime Offer Rate is published weekly by the CFPB; automated pricing engines from Optimal Blue, Polly, and LoanPASS pull APOR in real time and block lock confirmations that would breach the QM threshold without explicit override and exception documentation.

⚠️The APR miscalculation tail risk
An APR understated by more than 0.125% (or 0.25% for irregular loans) is a TILA disclosure violation that can trigger borrower rescission rights for up to three years post-consummation on refinance transactions. The 2008-2010 wave of rescission litigation cost the industry an estimated $4-6B in settlements and buybacks. Modern APR engines test against the actuarial method specified in Reg Z Appendix J at every fee change; firms that rely on manual recalculation in their TRID workflow consistently show 3-5x higher post-close defect rates in Fannie Mae LQI sampling.

The High-Cost Mortgage rules (HOEPA, Section 32) and Higher-Priced Mortgage Loan rules add overlapping tests with different thresholds, escrow requirements, and counseling mandates. A compliance engine evaluates each loan against Section 32 APR triggers (APOR + 6.5% for first liens, APOR + 8.5% for subordinate liens, with adjustments), points-and-fees triggers (5% for loans of $24,866 or more in 2025), and prepayment penalty triggers. Loans that test positive for HOEPA generally exit the standard production channel — most secondary market investors will not purchase them — and route to a specialized desk. The economic value of catching a HOEPA flag at application rather than at the closing table is roughly $8,000-$15,000 per loan in avoided cure, rework, or scratch-and-dent disposition.

The 54-Jurisdiction Patchwork

State regulation is where mortgage compliance becomes genuinely combinatorial. The SAFE Act requires individual loan originator licensing through the NMLS in every state in which the originator takes applications or offers terms; entity licensing requirements vary by state, with separate licenses often required for first mortgages, subordinate liens, servicing, broker activity, and high-cost or non-QM lending. New York DFS Part 419 imposes servicing rules stricter than RESPA on early intervention and loss mitigation. California's DFPI enforces the Residential Mortgage Lending Act and the CFL with distinct fee restrictions. Texas has its constitutional Section 50(a)(6) home equity rules — a separate compliance regime where a single defect (such as exceeding the 3% fees cap or failing the 12-day cooling-off requirement) renders the lien unenforceable.

Per-Loan State Compliance Test Count (Top 10 by Complexity)

State high-cost loan laws add a second layer on top of HOEPA: roughly 35 states maintain their own thresholds, often more restrictive than federal Section 32. Massachusetts MGL Chapter 183C, New Mexico's Home Loan Protection Act, and North Carolina's predatory lending statute each define their own APR and points-and-fees triggers, prohibitions on financing single-premium credit insurance, and counseling requirements. Compliance engines maintain these rule sets as versioned policy artifacts; ComplianceEase and Wolters Kluwer publish monthly rule updates that are consumed via API into the lender's LOS. A medium-sized lender funding in 40 states will run 3,500-4,500 distinct state-level tests across its production volume each week.

State licensing operations themselves have become a software problem. NMLS Call Reports (filed quarterly), state-specific renewal cycles, individual MLO continuing education tracking, and surety bond management for entity licenses generate a workflow that mid-market lenders increasingly outsource to specialists like Mortgage Industry Advisory Corporation or run through dedicated GRC platforms (Resolver, LogicGate, OneTrust). The cost of a single missed renewal — typically a cease-and-desist on new applications in that state and a $5,000-$50,000 administrative penalty — has driven adoption of automated calendar systems that integrate with NMLS data feeds.

Fair Lending, HMDA, and AI Model Governance

ECOA (Reg B) and the Fair Housing Act prohibit discrimination on prohibited bases; HMDA requires lenders to collect and report 110 data fields on each application, including pricing data, debt-to-income, credit score, automated underwriting system, and the borrower's race, ethnicity, age, and sex. The CFPB and DOJ have moved aggressively against redlining patterns inferred from HMDA data: the 2022 Trident Mortgage settlement ($24M), the 2023 ESSA Bank action, and the 2024 Fairway Independent Mortgage matter ($1.9M) each began as HMDA analytics flags before progressing to formal investigations.

The compliance engineering response runs in two directions. First, HMDA data quality: lenders run pre-submission edit checks (the FFIEC's HMDA Platform validations plus internal logic tests) on every application during origination, not as a year-end exercise. Second, fair lending analytics on outcomes: regression-based pricing disparity analysis, BISG-based proxy race assignment for indirect auto-style techniques applied to mortgage marketing, and HMDA peer benchmarking against MSA-level lender groups. Vendors in this space include CRA Wiz / Fair Lending Wiz (Wolters Kluwer), RATA Comply, and Ncontracts. A typical bank pricing review now examines APR, total fees, exception frequency, and overage/underage patterns across roughly 18-24 control variables.

🎯Model governance under the 2024 CFPB AI guidance
The CFPB's 2024 circular on adverse action notices clarified that lenders using AI or complex models must provide specific principal reasons — generic statements like 'model score insufficient' are not compliant. This has direct implications for the AVM and AUS stack described in our piece on <a href="/in-focus/property-as-a-platform/avm-2-satellite-imagery-transaction-data">AVM 2.0</a> and on <a href="/in-focus/next-gen-loan-origination-automated-underwriting">automated underwriting</a>. Lenders need explainability infrastructure (SHAP values, monotonic constraints, or rule-based overlays) that maps model outputs to ECOA-compliant reason codes. Firms running Zest AI, FairPlay AI, or Stratyfy have generally addressed this; legacy black-box scorecards remain an enforcement target.

The Compliance Stack Across the Loan Lifecycle

Compliance Checkpoints in a Modern Mortgage Workflow
1
Pre-application

MLO licensing verification against NMLS, advertising compliance review (Reg Z §1026.24, state-specific), telephone consumer protection (TCPA) consent capture

2
Application (Day 0-3)

TRID application trigger test (six elements), ECOA notice of receipt, intent-to-proceed capture, RESPA Special Information Booklet delivery for purchase loans, Loan Estimate issuance within 3 business days

3
Processing & underwriting

ATR/QM testing, HOEPA and state high-cost screens at each fee change, ECOA adverse action timeline tracking (30 days), HMDA data capture, AVM validation per OCC SR 10-16

4
Pre-closing (Day -7 to -3)

Closing Disclosure issuance with 3-business-day waiting period, TRID tolerance reconciliation, final APR validation, state-specific disclosures (e.g., NY Section 6-l, MA right to cure)

5
Closing & funding

Right of rescission (3 business days for refinances per TILA §125), wet/dry state recording, e-closing audit trail capture per ESIGN/UETA — see our piece on digital closing

6
Post-close & servicing

QC sampling per Fannie Mae Selling Guide D1-3 (10% minimum), escrow analysis annually, periodic statement delivery (Reg Z §1026.41), loss mitigation timelines under Reg X §1024.41

Two architectural choices distinguish compliance leaders. The first is whether compliance testing runs as a synchronous gate inside the LOS workflow or as an asynchronous post-event audit. The leaders run synchronous: every state change to a loan application (fee added, locked rate changed, property address updated, borrower added) triggers a full re-test of the applicable rule set before the change is committed. The second is whether the rule library is maintained internally or sourced. Mid-market lenders almost universally source from Wolters Kluwer, ICE/Encompass, or SitusAMC; the top-10 banks typically run a hybrid where federal rules come from a vendor library and proprietary overlays (investor-specific guideline differences) are maintained internally.

UDAAP and the Servicing Compliance Frontier

Unfair, Deceptive, or Abusive Acts or Practices authority under Dodd-Frank §1031 has become the CFPB's most flexible enforcement tool, with no rulemaking required to bring an action. Recent UDAAP matters in mortgage servicing — including the 2023 Carrington Mortgage Services consent order ($5.25M civil penalty) and the 2024 Solo Funds matter — focused on misleading communications, dual-tracking foreclosure activity during loss mitigation review, and improper fee assessments. Servicing platforms now run UDAAP screens on outbound communications using NLP — vendor systems like Verint and NICE Actimize flag promises of action, statements about credit reporting, and references to amounts owed that may be inaccurate at the moment of the communication.

Our defect rate on TRID dropped from 4.1% to 0.6% in eighteen months — but the harder win was getting state high-cost screens to run in the same engine. We used to maintain three different rules databases; one outage in any of them could shut down originations in entire regions. Consolidation onto a single API-served rule library was the unlock.
Chief Compliance Officer, top-15 independent mortgage banker

Reg X servicing rules (12 CFR 1024.30-41) — early intervention, continuity of contact, loss mitigation review timelines, error resolution and information request procedures — translate into workflow constraints that servicing platforms encode as state machines. The 36-day early intervention contact, the 45-day written notice, the 5-day loss mitigation acknowledgment, and the 30-day evaluation decision deadlines each become SLAs with automated escalation. Failure rates here drive both regulatory exposure and investor scorecard impact; Freddie Mac's STAR program and Fannie Mae's Servicer Total Achievement and Rewards both factor compliance timeliness into pricing.

Implementation Pattern: From Audit-Heavy to Engineering-Led

The traditional compliance operating model — quarterly internal audits, manual file review, and reactive issue logs — produces defect rates in the 3-7% range and average issue-to-remediation cycles of 60-90 days. The engineering-led model uses compliance-as-code: rules expressed in versioned policy artifacts, tested against synthetic loan files in CI/CD, deployed atomically across origination and servicing systems, and monitored in production with the same observability tooling used for other application code. Lenders that have completed this transition (most of the top-10 by volume have, plus a growing set of fintech-native shops like Better, Rocket, and Lower) report defect rates below 1% and remediation cycles measured in days.

12-Month Compliance Modernization Plan

Twelve articles into this guide, the through-line should be visible: every layer of the real estate finance stack — from digital closing and lease abstraction to securitization, servicing, and now compliance — converges on the same architectural pattern. Data captured once at the source, validated continuously against versioned rules, exposed via APIs to downstream consumers, and instrumented with the metrics that matter to regulators, investors, and operators. The lenders who get this right will run at 30-40% of the per-loan cost structure of those who don't, and they will do it with measurably lower regulatory exposure. The compliance function stops being a brake on the business and becomes part of the engineering substrate that makes mortgage manufacturing competitive.

💡Did You Know?
The CFPB's HMDA Modified LAR public dataset contains over 12 million records per year and is downloaded by plaintiffs' firms, academic researchers, and competitor analytics teams within hours of release. The 2018 expansion of HMDA fields (under Dodd-Frank §1094) more than doubled the data points reported per application — and made statistical disparity analysis materially more tractable for outside parties, which is why fair lending enforcement matters have accelerated since 2021.

Frequently Asked Questions

What is the difference between RESPA and TILA in mortgage compliance?

RESPA (Real Estate Settlement Procedures Act, Reg X) governs settlement service costs, kickbacks, escrow accounts, and servicing transfers. TILA (Truth in Lending Act, Reg Z) governs cost-of-credit disclosure, APR calculation, ability-to-repay, and rescission rights. Since 2015, the TRID rule has integrated their disclosure requirements through the Loan Estimate and Closing Disclosure, but the substantive obligations under each statute remain distinct.

How do state high-cost loan laws interact with federal HOEPA?

Roughly 35 states maintain their own high-cost or predatory lending statutes with thresholds that are generally more restrictive than federal HOEPA (TILA Section 32). A loan must pass both the federal HOEPA tests and the applicable state test; failing either typically excludes the loan from standard secondary market execution. Modern compliance engines maintain both rule sets and screen loans at every fee change throughout the origination process.

What does a TRID tolerance violation cost the lender?

Zero-tolerance fee overages must be cured to the borrower within 60 days of consummation; 10% bucket overages must be cured for amounts above the 10% aggregate threshold. Beyond direct cure costs, repeated TRID defects drive investor repurchase demands (Fannie Mae and Freddie Mac both reference TRID in their Selling Guides), scratch-and-dent execution losses of 200-500 basis points, and CFPB exam findings that can escalate to consent orders.

How are AI underwriting models governed under ECOA and CFPB guidance?

The CFPB's 2024 circular on adverse action notices requires lenders to provide specific principal reasons for credit denials, even when AI or complex models drive the decision. Generic explanations are non-compliant. Lenders must maintain explainability infrastructure — typically SHAP values, monotonic feature constraints, or rule-based overlays — that maps model outputs to specific ECOA-compliant reason codes, and they must document fair lending testing on the model itself.

What is the most efficient way to manage 50-state mortgage licensing compliance?

Most lenders source the federal and standard state rule library from one of three vendors — Wolters Kluwer ComplianceOne, ICE Mortgage Technology's Encompass Compliance Service, or SitusAMC's ComplianceEase — via API into their LOS. NMLS Call Report assembly, MLO license tracking, and renewal calendars are typically handled in a dedicated GRC platform with 60/30/15-day alerts. Mid-market lenders often outsource NMLS operations to specialists rather than build internally.