Portfolio valuation
Mark-to-model practices
Mark-to-model practices
Most ABF positions do not have observable market prices. You cannot pull up a Bloomberg screen and see where your warehouse residual or private ABS tranche traded today. That means you are marking to model, not marking to market.
This page covers when you must model, how to do it defensibly, and how to bridge from market anchors to your final marks.
When mark-to-market works
Pure mark-to-market is appropriate when:
- There is frequent trading in identical or substantially similar instruments
- Quotes are executable, not indicative
- The market is orderly (not distressed or frozen)
- You can observe the price without making adjustments
Examples where MTM works:
| Asset Type | Why MTM Works |
|---|---|
| Agency RMBS | Active trading, dealer quotes are executable, tight bid-ask spreads |
| Liquid corporate bonds | TRACE reporting, frequent transactions |
| Large-issue ABS senior tranches | Established shelves with regular secondary trading |
| Treasury securities | Continuous pricing |
For these positions, the price is observable. You use it directly.
When you must model
You must rely on mark-to-model when:
| Condition | Example |
|---|---|
| No secondary trading exists | Private placement that has never traded |
| Available quotes are indicative only | Dealer provides “color” but would not transact at that level |
| Quotes are stale | Last dealer quote was 90 days ago |
| Position has bespoke features | Waterfall differs from any comparable deal |
| Market is dislocated | Quotes exist but do not reflect orderly transactions |
Positions that are typically Level 3 / mark-to-model:
- Warehouse residuals and first-loss positions
- Private placement tranches
- Participations in loan pools
- Subordinate tranches of bespoke securitizations
- Any position valued primarily through DCF with internal assumptions
For these positions, you are deriving fair value from a model that incorporates your assumptions about future cash flows and the appropriate discount rate.
Hybrid approaches
Pure MTM and pure MTM (model) are the extremes. Most ABF valuations sit somewhere in between, using a hybrid approach that anchors to market data while adjusting for position-specific factors.
The hybrid framework
Step 1: Anchor to market
Find the closest observable benchmark:
- Public ABS spreads for similar collateral
- Recent transaction in a comparable structure
- Executable dealer bid on related paper
- Pricing service mark for a similar position
Step 2: Adjust with model
Use your DCF model to adjust for:
- Structural differences (credit enhancement, waterfall priority, triggers)
- Performance divergence (your pool vs. the comparable)
- Deal-specific factors (size, issuer, seasoning)
- Liquidity premium
Step 3: Document the bridge
Show explicitly how you moved from the market anchor to your final mark. This bridge documentation is critical for audit defense and LP transparency.
Worked example: hybrid valuation
Position: BBB-equivalent tranche of a private consumer loan ABS
Step 1 - Market anchor: Public consumer loan ABS BBB tranches are trading at SOFR + 280 bps based on TRACE data and dealer runs.
Step 2 - Adjustments:
| Factor | Adjustment | Rationale |
|---|---|---|
| Higher advance rate (85% vs. 80%) | +15 bps | More aggressive structure |
| Weaker originator track record | +20 bps | Execution and servicing risk |
| No secondary liquidity | +25 bps | Cannot exit quickly if needed |
| Better recent performance | -10 bps | Loss curves below expectations |
| Net adjustment | +50 bps |
Adjusted spread: SOFR + 330 bps
Step 3 - Documentation:
| Element | Detail |
|---|---|
| Market anchor | Consumer ABS BBB tranches at SOFR + 280 bps (source: TRACE, 5/15/2024) |
| Position details | $8.5M face, 3.2 year WAL, 12% credit enhancement |
| Adjustments applied | +50 bps (see table above) |
| Final discount spread | SOFR + 330 bps |
| Model inputs | 6% CDR, 75% severity, 12% CPR |
| Mark | $8.15M (95.9 price) |
Model documentation requirements
A mark-to-model valuation must be fully documented. The documentation serves three purposes:
- Audit defense: Auditors will test whether your methodology is reasonable and consistently applied
- LP transparency: Investors need to understand how you are arriving at marks
- Internal governance: Future valuators need to understand what was done and why
Required elements
| Element | What to Include |
|---|---|
| Methodology | DCF, market comparable with adjustment, other |
| Key assumptions | Prepayment rate, default rate, severity, discount rate |
| Source of assumptions | Where each input came from (historical data, market observation, judgment) |
| Market data used | Comparables, dealer quotes, pricing service inputs |
| Adjustments | Each adjustment applied and the rationale |
| Sensitivity analysis | How the mark changes with different assumptions |
| Classification | Level 2 or Level 3, with justification |
| Approval | Who reviewed and approved the mark |
What good documentation looks like
Good: “Discount rate of SOFR + 350 bps based on BBB consumer ABS trading levels of SOFR + 275 bps (source: TRACE, 5 trades week of 5/13), plus 75 bps adjustment for illiquidity, subordination, and issuer factors (see adjustment table).”
Bad: “Discount rate of SOFR + 350 bps based on professional judgment.”
The first version is traceable. An auditor can verify the TRACE data, evaluate whether the adjustments are reasonable, and understand how you arrived at the rate. The second version provides no support.
Managing model risk
Model-based valuations carry inherent risks. Assumptions may be wrong, comparables may be inapplicable, or structural features may not be captured correctly.
Common sources of model risk
| Risk | Description | Mitigation |
|---|---|---|
| Assumption bias | Manager selects assumptions that support desired marks | Independent valuation review, governance oversight |
| Stale inputs | Model uses outdated market data or performance data | Regular data refresh, currency thresholds |
| Structural errors | Waterfall modeled incorrectly | Periodic model validation, deal-by-deal review |
| Comparable mismatch | Using comps that are not actually comparable | Document why comps are appropriate, sensitivity to comp selection |
| Anchoring | Prior mark influences current mark inappropriately | Fresh analysis each period, document changes |
Model validation
For material positions, models should be validated periodically:
| Validation Element | Frequency |
|---|---|
| Check model math (waterfall, cash flows) | At acquisition and after any modifications |
| Review assumption reasonableness | Each valuation period |
| Back-test assumptions vs. actual performance | Quarterly |
| Independent model review | Annually or at acquisition for large positions |
When marks diverge from carrying value
As your portfolio ages, marks will diverge from cost basis. This is expected, but the divergence requires explanation.
Documenting mark changes
When a mark moves significantly from the prior period:
| Scenario | Documentation Required |
|---|---|
| Mark up due to spread tightening | Document market move with data |
| Mark up due to better performance | Document pool performance vs. prior assumptions |
| Mark down due to spread widening | Document market move with data |
| Mark down due to deteriorating performance | Document performance trends, trigger risks |
| Mark down due to structural concerns | Document specific issue (trigger breach, originator stress) |
Audit considerations
Auditors pay attention to:
- Large mark changes without corresponding market or performance changes
- Marks consistently higher than third-party valuations
- Assumptions that are aggressive relative to historical experience
- Inconsistent methodology application across similar positions
Anticipate these questions and document proactively.
Best practices summary
| Practice | Why It Matters |
|---|---|
| Anchor to market when possible | Provides external discipline |
| Document the bridge from anchor to mark | Makes adjustments transparent and auditable |
| Use consistent methodology | Prevents cherry-picking approaches |
| Refresh inputs each period | Prevents stale marks |
| Run sensitivity analysis | Shows range of reasonable values |
| Separate valuation from investment function | Reduces bias |
| Retain full documentation | Supports audit and LP requests |