Analyzing Mortgage Backed Securities

Evaluates MBS structures with prepayment modeling (CPR/CDR), collateral analysis, and tranche-level credit risk assessment.

When To Use

• Analyzing agency (Ginnie Mae, Fannie Mae, Freddie Mac) or non-agency RMBS deals
• Modeling prepayment and default scenarios on residential mortgage collateral pools
• Evaluating tranche-level credit enhancement, subordination, and cash flow waterfall mechanics
• Comparing MBS tranches for relative value across spread, WAL, and convexity profiles
• Assessing seasoned pools for re-REMIC structuring or secondary market trading
• Reviewing offering documents, prospectus supplements, or trustee reports for deal-level risk

Inputs To Gather

• Deal documents: Prospectus supplement, pooling and servicing agreement (PSA), trustee reports
• Collateral tape: Loan-level data including FICO, LTV, DTI, loan age, geography, occupancy type, documentation level
• Pool statistics: Current balance, WAC, WAM, WALA, average loan size, delinquency buckets (30/60/90+), REO pipeline
• Structure details: Tranche map (senior/mezzanine/subordinate), credit enhancement levels, OC/XS triggers, step-down dates
• Prepayment and default assumptions: Base-case CPR, CDR, loss severity, recovery lag; stress scenarios if applicable
• Market context: Current mortgage rates, refi incentive, HPA trends, [VERIFY] agency guarantee status and program eligibility
• Rating agency criteria: Applicable S&P, Moody's, Fitch, or DBRS loss/stress frameworks if rating-dependent analysis

Workflow

1. Map the deal structure

   • Identify all tranches: senior (A classes), mezzanine (M classes), subordinate (B classes), IO strips, residual
   • Document the cash flow waterfall: sequential vs. pro-rata pay periods, trigger events (delinquency/loss triggers), clean-up call provisions
   • Record credit enhancement: subordination percentages, overcollateralization targets, excess spread capture mechanisms

2. Profile the collateral pool

   • Stratify loans by vintage, FICO band, LTV bucket, geography (state/MSA concentration), loan purpose (purchase/refi/cash-out), and property type
   • Identify adverse selection risks: high LTV concentrations, low-doc loans, investor properties, geographic clustering
   • Calculate weighted-average characteristics and compare against benchmark pools for the same vintage/program

3. Model prepayment scenarios

   • Set base-case CPR using historical analogs (e.g., seasoned conventional 30yr at current coupon spread to market)
   • Run sensitivity across CPR vectors: slow (e.g., 6 CPR), base (e.g., 15 CPR), fast (e.g., 35 CPR), and ramp scenarios (PSA multiples)
   • For each scenario, compute WAL, yield, spread to benchmark, and effective duration for each tranche
   • Assess negative convexity exposure on premium-priced tranches and extension risk on discount tranches

4. Model default and loss scenarios

   • Set base-case CDR and loss severity using collateral characteristics and historical performance curves
   • Run stress scenarios: 2x CDR, elevated severity (e.g., 40%–60% on non-agency), delayed recovery timelines
   • Determine tranche-level loss absorption: at what CDR/severity combination does each tranche experience principal write-down
   • Evaluate trigger mechanics — will delinquency or cumulative-loss triggers divert cash flow from subordinate tranches

5. Assess credit enhancement adequacy

   • Compare current subordination levels to original levels and to rating-agency loss benchmarks
   • Evaluate OC build/release mechanics and whether excess spread is sufficient to maintain targets under stress
   • For seasoned deals, assess whether step-down conditions have been met and whether senior tranches benefit from de-leveraging
   • [VERIFY] Check for any amendments, modifications, or servicer advances that may affect collateral performance

6. Synthesize relative value and risk conclusions

   • Rank tranches by spread per unit of WAL risk, credit risk, and convexity exposure
   • Flag tranches with asymmetric risk profiles (e.g., thin mezzanine with cliff risk, IO strips with high prepay sensitivity)
   • Compare to similar deals in the market for relative value context
   • Identify key monitoring triggers: delinquency thresholds, cumulative loss benchmarks, servicer performance metrics

Output

Produce a structured MBS analysis report containing:

• Deal overview: Issuer, shelf program, closing date, original/current balance, servicer(s), trustee
• Collateral summary: Pool composition table with stratifications, weighted-average metrics, delinquency and loss performance to date
• Structure summary: Tranche map with current balances, coupons, credit enhancement levels, and priority of payments description
• Prepayment analysis: Table of WAL, yield, and spread across CPR scenarios for each evaluated tranche
• Credit analysis: Loss absorption capacity by tranche, stress-test results, trigger proximity assessment
• Relative value assessment: Spread comparison to benchmark deals, convexity-adjusted return analysis
• Risk flags and monitoring points: Concentration risks, servicer concerns, trigger events approaching thresholds
• Appendix: Key assumptions, data sources, model methodology notes

Quality Checks

• Verify that all tranche balances sum to total deal balance and that waterfall logic is internally consistent
• Confirm CPR/CDR assumptions are sourced from stated methodology, not arbitrary — cite historical analogs or agency benchmarks
• Ensure loss severity assumptions match collateral type (e.g., non-agency subprime vs. agency conforming have materially different severities)
• Cross-check credit enhancement percentages against trustee reports, not just offering documents, for seasoned deals
• [VERIFY] Rating agency criteria versions used — S&P, Moody's, and Fitch periodically update RMBS loss frameworks
• [VERIFY] Regulatory considerations: risk retention rules (Reg RR), QM/ATR status of underlying loans, Volcker Rule implications for trading book holdings
• Flag any data gaps in the collateral tape (missing FICO, undisclosed LTV) and note their impact on model reliability
• Mark all forward-looking projections as estimates subject to rate, HPA, and macroeconomic assumptions