Get Started
Get a Demo
Blog Archives

How Has the First “Social” RMBS Performed – And What’s So Social About It?

Now that six months have passed since Angel Oak issued AOMT 2021-2 – hailed as the first U.S. non-Agency RMBS to qualify as a social bond [1] – we can compare preliminary collateral performance to other deals. Angel Oak’s 2021-1, from the same shelf and vintage – but without the social bond distinction – provides an apt control group. To set the stage for this performance comparison, we’ll first reexamine the compositional differences – and significant overlap – between the two collateral pools. What we will show:

  • The pool compositions are highly overlapping, with marginally greater risk concentrations of self-employment and alternative documentation in the social securitization, and the same WA (weighted average) coupon
  • The social collateral has outperformed the benchmark credit-wise in the early going
  • The social deal has exhibited some lock-in, i.e., slower refinancing, providing some very preliminary evidence that the borrowers are indeed underserved, and that investors may be rewarded if the social collateral’s credit performance holds
  • However, the credit mix of the social collateral has drifted riskier – more so than the benchmark – meaning the strong early credit performance of the social deal could reverse, and ongoing surveillance is warranted

New Loans or New Label?

The Social AOMT 2021-2 Is Similar to AOMT 2021-1

Figure 1 shows AOMT 2021-1 vs. 2021-2 in the Collateral Comparison screen of Edge, RiskSpan’s data and analytics platform. Clearly, the two pools were similar at origination, with highly overlapping distributions of FICO, LTV, and DTI and many other similar metrics.

So What’s Different – And How Different Is It?

The distinguishing principle of a social bond under Angel Oak’s framework is that it provides affordable home mortgages to those who often can’t get them because they don’t qualify under the automated underwriting processes of traditional lenders because of the exceptional nature of their sources of income. [2]

Angel Oak says the specific characteristic hindering the borrowers in the AOMT 2021-2 deal is self-employment. [3] Self-employed borrowers make up 94.4% of the pool (with a median annual income of $227,803) [4], up marginally from 86.5% in the 2021-1 deal [5]. As Figure 1 shows, the proportion of low documentation by balance was up from 87.5% in 2021-1 to 97.5% in 2021-2.

Also, Figure 1 shows that 2021-2’s FICOs and LTVs are slightly worse on average with slightly more tail risk, and the cash-out proportion is slightly riskier.

Compensating marginally for 2021-2 are slightly lower ARM proportions (0 vs. 0.8% for 2021-1), lower WA. DTI, and a higher proportion of owner-occupied (90% vs. 85%), which many view as credit-positive.

In summary, RiskSpan calculates 1.83 average risk layers per loan for the social 2021-2, slightly higher than 1.78 for 2021-1.

Notably the WA coupons for the two pools are the same.

Figure 1: Edge’s Collateral Comparison Screen Showing AOMT 2021-1 (aka AOAK 2101) vs. 2021-2 (aka AOAK 2102) at OriginationSource: CoreLogic, RiskSpan

Would you like to see the tool we used to perform this analysis?


In summary, it seems that most – though perhaps not all – of the loans that qualified for AOMT 2021-2 would have qualified for AOMT 2021-1 and other non-QM deals.

Kroll’s new issue report seems to acknowledge that what is new about 2021-2 is mostly the formal emphasis on the social benefits of the loans made, and less a change in the kinds of loans made: “While many of [Angel Oak’s] lending programs overlap meaningfully with other non-QM lender’s offerings, the actions taken by AOCA generally indicate management’s attention to ESG related matters. Specifically, AOCA’s SBF puts focus on the impact that credit availability for underserved borrowers can have.” [4]

A skeptical interpretation of the overlap between 2021-1 and the social 2021-2 collateral would be that the social claim is largely hollow. Another way of looking at it is that a financial market participant is finally taking credit for good work it has been largely doing all along. Angel Oak itself seems to take this latter view, saying, “Since 2011, AOCA has been implementing ESG principles within its non-qualified mortgage (non-QM) origination and securitization program to provide access to residential credit for underserved borrowers.” [2]

Either way, logical hypotheses would be that collateral performance will be similar between 2021-2 and 2021-1, with -2 showing (a) slightly more credit trouble and (b) slightly less able to refinance. Regarding the second hypothesis, logically it should challenge the premise that the deal serves underserved borrowers if its borrowers can refinance just as readily as others.

Early Performance of the Social Bonds

Let’s see how AOMT’s social 2021-2 has performed as benchmarked to 2021-1 during the first six and seven months, respectively, of available data.

Better Delinquency Trend Than the Benchmark

As Figure 2 shows, delinquencies opened higher for the social 2021-2 but have mostly cured. By contrast, delinquencies have trended up for 2021-1. So far, Angel Oak’s social origination is outperforming its non-social contemporary from a credit standpoint.

Figure 2: AOMT 2021-2 Delinquencies Began Higher, Have Mostly Cured; AOMT 2021-1’s Delinquencies Have Trended Up 60 day-plus delinquency share over time, AOMT 2021-2 vs AOMT 2021-1 Source: CoreLogic, RiskSpan

Significantly Better Credit Performance by the Social DSCR Investor Loans

A small slice of the deals driving outsized delinquencies in 2021-1 are the DSCR-based investor loans (Figure 3). In the social 2021-2, delinquencies among this cohort are zero. We plot the spreads at origination (SATO) of this cohort alongside delinquencies to show that the DSCR loans in 2021-2 had lower credit spreads by about 20bps. Perhaps the investor loans pooled into 2021-2 were managed to higher standards for DSCR, rent rolls or other attributes (their LTVs and ages are similar).

Figure 3: Delinquencies – and SATOs – Are Lower Among DSCR-Based Investor Loans in AOMT 2021-2 60 day-plus delinquency share and WA SATOs over time, AOMT 2021-2 vs. AOMT 2021-1, includes Detailed Doc Type = DSCR Investor Cash Flow.Source: CoreLogic, RiskSpan

Ironically, The Full Doc Loans Are the Social Deal’s Blemish

The slice of full doc loans in the social 2021-2 have a much lower WA FICO than the low doc loans in the same deal or either the low or full doc loans in 2021-1 (see the green dotted line in Figure 4). Correspondingly, these full doc loans have the highest delinquent share among the four cohorts in Figure 4 (green solid line). If this pattern holds, it highlights the viability of using tradeoffs to manage down the overall credit risk represented by loans with risky attributes.

Figure 4: AOMT 2021-2’s Full Doc Loans Are the Most Delinquent Doc Cohort from Either Deal 60 day-plus delinquency share and WA FICOs over time, AOMT 2021-2 vs. AOMT 2021-1 and Full Doc vs. Low Doc Source: CoreLogic, RiskSpan

Slower Refinances Than the Benchmark

While credit performance has been better for the social deal than we might expect, voluntary prepays so far (Figure 5) support our hypothesis that the social deal should prepay slower. Note that we plot voluntary prepays over loan age, and that all loans from this recent non-QM vintage have similar (and highly positive) refinance incentive. If the social deal’s refinances remain slower, that accomplishes two significant things: 1) it supports the claim that the social borrowers are indeed underserved; 2) if combined with sustained credit performance, it provides support in terms of financial risk and return for the price premiums that social bonds tend to command.

Figure 5: AOMT 2021-2 Is Refinancing Slower CRR over loan age, AOMT 2021-2 vs. AOMT 2021-1, July 2021-January 2022 Source: CoreLogic, RiskSpan

The Relative Refinance Slowness Is From the Large Balance Loans

The overall slowness of the social collateral in Figure 5 is driven by large loans. Figure 6 shows that, among loans <$417K, the prepay patterns of 2021-1 and 2021-2 are similar, while among loans > $417K, the prepays of 2021-2 are consistently slower. This may suggest that large loans with complex sources of income are particularly hard to underwrite.

Figure 6: The Social Deal’s Low-Balance Loans Refi Similar to Benchmark, But Large Balances Have Been Slower CRR over loan age, AOMT 2021-2 vs. AOMT 2021-1, bucketed by loan size, July 2021-January 2022 Source: CoreLogic, RiskSpan


Updated Collateral Mix

The Social Deal’s Credit Mix Has Drifted Riskier, Warranting Ongoing Monitoring

While the early performance of the social collateral is positive, Figure 7 provides reason for concern and ongoing watchfulness. Since origination, the composition of the social 2021-2 has drifted riskier in all respects except slight improvements in WA DTI and WA LTV. Its LTV tails, WA FICO, and FICO tails; proportions of cash-out, low doc, non-owner-occupied; and average overall risk layers are all somewhat riskier.

The drift for 2021-1 has been more mixed. Like 2021-2, it is safer with respect to WA DTI and WA LTV. Unlike 2021-2, it is also safer with respect to LTV tails, FICO tails, and cash-out proportion. Like 2021-2, it is riskier with respect to WA FICO; proportions of low doc and non-owner-occupied; and average overall risk layers.

We will continue to monitor whether this composition drift drives differential performance going forward.

Figure 7: Edge’s Collateral Comparison Screen Showing AOMT 2021-1 (aka AOAK 2101) vs. 2021-2 (aka AOAK 2102) updated to the Current Factor DateSource: CoreLogic, RiskSpan

Using Edge, you can examine prepay or credit performance of loan subsets defined by any characteristics, and generate aging curves, time series, or S-curves.


Value Opportunities in Private-Label Investor Loan Deals

The supply of investor loan collateral in private securitizations has surged in 2021 and projects to remain high (more on this below). To gain an informational edge while selecting bonds among this new issuance, traders and investors have asked RiskSpan for data and tools to dissect the performance of investor loans. Below, we first show the performance of investor loans compared to owner-occupied loans, and then offer a glimpse into a few relative value opportunities using our data and analytics platform, Edge.

As background, the increase of investor loan collateral in PLS was spurred by a new FHFA policy, recently suspended, that capped GSE acquisitions of investor and second home loans at 7% of seller volume. This cap forced originators to explore private-label securitization which, while operationally more burdensome than GSE execution, has been more profitable because it bypasses the GSEs’ high loan-level pricing adjustments. Now that this difficult but rewarding PLS path has been more widely traveled, we expect it to become more efficient and to remain popular, even with the GSE channel reopening.

Subsector Performance Comparison: Investor Vs. Owner-Occupied Loans

Investor Loans Promise Longer Collection of Above-Market Rates

Compared to owner-occupants, investors have historically paid above-market mortgage rates for longer periods before refinancing. Figure 1 shows the prepayment rates of investors vs. owner-occupants as a function of refinance incentive (the borrower’s note rate minus the prevailing mortgage rate). As their flatter “s-curve” shows, the rise in investor prepayments as refinance incentive increases is much more subdued than for owner-occupants.

Crucially, this relationship is not fully explained by higher risk-based pricing premiums on investor loans. Figure 2 shows the same comparison as Figure 1 but only for loans with spreads at origination (SATO) between 50 and 75 bps. The categorical difference between owner-occupied and investor prepay speeds is partially reduced but clearly remains. We also tried controlling for property type, but the difference persists. The relative slowness of investors may result from investors spreading their attention across many elements of their P&L besides interest expense, from higher underwriting obstacles for a rental income-driven loan, and/or from lenders limiting allocation of credit to the investor type.

While we plot these graphs over a five-year lookback period to balance desires for recency and sample size, this relationship holds over shorter and longer performance periods as well.

Figure 1: The Investor Loans S-Curve is Significantly Flatter Than the Owner-Occupied Curve
Investor s-curve vs. owner-occupied s-curve. Includes prime credit, no prepayment penalty, original loan size $200K-$400K, ages 6-48 months for the past 5yr period performance.

Source: CoreLogic’s Private-Label RMBS Collateral Dataset, RiskSpan. Note: because the increase in private-label investor loan volume is coming from Agency cutbacks, the historical performance of investor loans within both Agency and private-label datasets are relevant to private-label investor loan future performance. In this analysis we show private-label data because it straightforwardly parses voluntary prepays vs. defaults, which of course is a critical distinction for PL RMBS investors. Nonetheless, where applicable, we have run the analyses in both datasets, each of which corroborates the performance patterns we show.

Figure 2: Even Controlling for SATO, The Investor vs. Owner-Occupied S-Curve Difference Persists

Same as Figure 1 but includes only loans with SATO between 50-75 bps Source: CoreLogic, RiskSpan

Investor Loans Pose Comparable Baseline Risk, Greater Downside Risk to Credit Investors

Credit performance of investor loans has been worse than owner-occupied loans during crises, which justifies a pricing premium. During benign periods, investor loans have defaulted at similar or lower rates than owner-occupied loans – presumably due to more conservative LTVs, FICOs and DTIs among the investor loan type – and have therefore been profitable for credit investors during these periods. See Figure 3.

Figure 3: Investor Loans Have Defaulted at Greater Rates During Crises and Similar Rates in Other Periods vs. Owner-Occupied Loans
Default rates over time, investor loans vs. owner-occupied. Includes prime credit, ages 12-360 months.

Source: CoreLogic, RiskSpan

Relative Value Opportunities Within Investor Loans

California Quicker to Refinance

California has the largest share of U.S. investor mortgages, as it does with all residential mortgages. California borrowers, both investors and owner-occupieds, have exhibited a steeper response to refinance incentives than have borrowers in other states. Figure 4 shows the comparison focusing on investors. While historical home price appreciation has enabled refinances in California, it has done the same in many states. Therefore, the speed differences point to a more active refinance market in California. All else equal, then, RMBS investors will prefer less California collateral.

Figure 4: California Prepays Significantly Faster In the Money
Investor s-curves bucketed by geography (California vs. Other). Includes prime credit, no prepayment penalty, original loan size $200k-$400k, ages 6-48 months for the past 3yr performance period.

Source: CoreLogic, RiskSpan

For AAA Investors, Limited-Doc Investor Loans May Offer a Two-Sided Benefit: They Buoy Premium Bonds, and a Small Sample Suggests They Lift Discount Bonds, Too

Limited-doc investor loans offer senior tranche holders the chance to earn above-market rates for longer than full-doc investor loans, a relative edge for premium bonds (Figure 5). This is intuitive; we would expect limited-doc borrowers to face greater obstacles to refinancing. This difference holds even controlling for spread at origination. Based on a smaller sample, limited-doc investor loans have also turned over more (see greater prepay rates in the negative refinance incentive bucket). This may result from a correlation between limited documentation and more rapid flipping into the rising HPI environment we have had nationally throughout the past seven years. If so, this would mean that limited-doc investor loans also help discount bonds, relative to full-doc investor loans, accelerate repayments at par.

Because limited-doc investor loans are rare in the RMBS 2.0 era, we widened the performance period to the past seven years to get some sample in each of the refinance incentive buckets. Nonetheless, with all the filters we have put on to isolate the effect of documentation type, there are only a few hundred limited-doc investor loans in the negative refinance incentive buckets.

Figure 5: Limited-Doc Investor Loans Have Prepaid Slower In-The-Money and Faster Out-of-the-Money
Investor s-curves bucketed by doc type. Includes prime credit, no prepayment penalty, original loan size $400K-$800K, ages 6-48 months, SATO 25-125bps for the past 7yr performance period.

Source: CoreLogic, RiskSpan

Size Affects Refi Behavior – But Not How You Think

An assumption carried over from Agency performance is that rate-driven prepays get likelier as loan size increases. This pattern holds across conforming loan sizes, but then reverses and refinance response gets flatter again as balances cross $800K. This is true for investor and owner-occupied loans in both Agency and private-label loan data, though of course the number of loans above $800K in the Agency data is small. Figure 6 shows this pattern for private-label investor loans. As shown, in-the-money prepayments are slowest among loans below $200K, as we would expect. But despite their much higher motivation to refinance, loans above $800K have similar S-curves to loans of just $200K-$400K.

The SATO is generally a few basis points higher for these largest loans, but this does not explain away the speed differences. Figure 7 shows the same comparison as Figure 6 except only for loans with SATO between 50-75 bps. Except for a slightly choppier graph because of the reduced sample size, the same rank-ordering is evident. Nor does controlling for property type or geography remove the speed differences. The largest loans, we conclude, have fewer credit alternatives and/or face more stringent underwriting hurdles than smaller loans, hampering their refi rates.

Rate refinances are fastest among the mid-sized loans between $400K-$600K and $600K-$800K. That these last two groups have similar S-curves – despite the greater dollar motivation to refinance for the $600K-$800Kgroup – suggests that the countervailing effect of lower ability to find refinancing outlets is already kicking in for the $600K-$800K size range.

All of this means that high-balance collateral should be more attractive to investors than some traditional prepayment models will appreciate.

Figure 6: The Largest Investor Loans Refinance Slower Than Medium-Sized
Investor s-curves bucketed by loan size. Includes prime credit, no prepayment penalty, ages 6-48 months for the past 5yr performance period.

Source: CoreLogic, RiskSpan

Figure 7: Controlling For SATO, Largest Investor Loans Still Refinance Slower Than Medium-Sized
Same as Figure 4 but includes only loans with SATO between 50-75 bps

Source: CoreLogic, RiskSpan

Preliminarily, Chimera Has Lowest Stressed Delinquencies of Top Investor Shelves

For junior-tranche, credit-exposed investors in the COVID era, 60-day-plus delinquencies have been significantly rarer on Chimera’s shelf than on other top investor shelves. The observable credit mixes of the three shelves appear similar. We ran this analysis with only full-doc loans and from only one state (California), and the rank-ordering of delinquency rates by shelves remains the same. Further to this point, note that the spread at origination of Chimera’s shelf is nearly as high as Flagstar’s. All of this suggests there is something not directly observable about Chimera’s shelf that has generated better credit performance during this stressed period. We caution that differences in servicer reporting of COVID forbearances can distort delinquency data, so we will continue to monitor this performance as our data updates each month.

Figure 8: Chimera Posts Lowest COVID Delinquencies, with Nearly Highest SATO of Top Investor Shelves
Investor DQ60+ rates over time, bucketed by shelf. Includes prime credit, ages 12-60 months.

Source: CoreLogic, RiskSpan

The Greater Default Risk of Low-Doc Investor Loans Lasts About 10 Years

Low-doc investors default more frequently than full-doc investors, but only during the first roughly 120 months of loan age. Around this age, the default rates converge. For loans seasoned beyond this age, full-doc loans begin to default slightly more frequently than low-doc loans, likely due to a survivorship bias. This suggests that credit investors are wise to require a price discount for new issuance with low-doc collateral. For deals with heavily seasoned collateral, junior-tranche investors may counterintuitively prefer low-doc collateral — certainly if they can earn an extra risk premium for it, as it would seem they are not actually bearing any extra credit risk.

Figure 9: Low-Doc Investor Loans Default More Frequently Than Full-Doc Until Loan Age = 120
Investor default rates over time, bucketed by doc type. Includes prime credit, RMBS 2.0 era, for the past 7yr performance period.

Source: CoreLogic, RiskSpan


  • Investor loans face higher barriers to refinance than owner-occupied, offering RMBS investors the opportunity to earn higher coupons for longer periods.
  • For junior tranche investors, the credit performance of investor loans has been similar to owner-occupied loans during benign economic periods and worse during stressed times.
  • California borrowers respond more quickly to refinance incentives than borrowers from other states; investors will prefer less California collateral.
  • Limited-doc investor loans offer AAA investors a double benefit: slower refinances in the money, extending premium bonds; and faster turnover out of the money, limiting extension risk.
  • Low loan balances are attractive for their slow refinance response – as are non-conforming (high) loan balances above $800K. Traditional prepay models may miss this latter dynamic.
  • For credit investors, Chimera’s delinquency rates have been significantly better during the pandemic than other investor shelves. We will continue to monitor this as different ways of reporting COVID forbearances may confound such comparisons.
  • For credit investors, limited-doc investor loans default at higher rates than full-doc loans for about the first ten years of loan age; after this point the two perform very similarly, with limited-doc loans defaulting at slightly lower rates among these seasoned loans, likely due to survivor biases.

Contact Us

Contact us if you are interested in seeing variations on this theme. Using Edge, we can examine any loan characteristic and generate an S-curve, aging curve, or time series.

Estimating Credit Losses in the COVID-19 Pandemic

In the years of calm economic expansion before CECL adoption, institutions carefully tuned the macroeconomic forecasting approaches and macro-conditioned credit models they must defend under the new standard. Now, seemingly an hour before public entities are to record (and support) their first macro-conditioned credit losses, a disease with no cure and no vaccine sweeps the globe and darkens whole sectors of the economy. Truth is stranger than fiction. 

Institutions Need New Scenarios and Model Adjustments, and Fast 

Institutions must overhaul their projection capabilities to withstand audit scrutiny and with only nominal relief in CECL deadlines. 

Faced with this unprecedented crisis, many institutions will need to find new sources of macroeconomic scenarios. Business-as-usual scenarios that seemed sensible a few months ago now appear wildly optimistic. 

Credit and prepay models built on data prior to February 2020 – the models that institutions have spent so much time and effort validating – must now be rethought entirely. 

Institutions may not have a great deal of time to make the necessary adjustments. While the Coronavirus Aid, Relief, and Economic Security (CARES) Act (in Section 4014, Optional Temporary Relief from Current Expected Credit Losses) allows banks and credit unions a brief delay in CECL adoption, RiskSpan’s public bank clients are adopting as scheduled. One reason is the short and uncertain length of the delay, which expires either on 12/31/2020 or when the national coronavirus emergency is declared over, whichever comes first. The national emergency could be declared over at any time, and indeed we hope the national emergency does end soon. Another reason is that, as Grant Thornton has noted, eligible entities that defer adoption will need to retrospectively restate their year-to-date results when they adopt ASU 2016-13. Ultimately, the “relief” is anti-relief. 

The revised CECL approaches that institutions race into production will need to withstand the inspection not only of the normal sets of eyes, but many other senior stakeholders. Scrutiny on credit accounting will be more intense than ever in light of COVID-19. 

Finally, to converge on a new macroeconomic scenario and model adjustments, institutions will be prompted by auditors and senior management to run their portfolio many times under many different combinations of approaches. As you can imagine, the volume of runs hitting RiskSpan’s Allowance Suite has spiked this month, with institutions running many different scenarios, and institutions with available-for-sale bond portfolios sending more impaired bonds than anticipated. The physics of pulling off so many runs in such a short time are impossible for teams and systems not set up for that scale. 

How RiskSpan is Helping Institutions Overcome These Challenges 

RiskSpan helps clients solve the new credit accounting rules for loans, held-to-maturity debt securities, and available-for-sale debt securities. As we all navigate these unique and evolving times, let us share how we incorporate the impact of COVID-19 into the allowances we generate. The toolbox includes new macroeconomic scenarios that reflect COVID-19, adjustments to credit and prepay models, an ability to selectively bypass models with user-defined scenarios, and even – sparingly – support for the dreaded “Q-factor” or management qualitative adjustment. 

COVID-19-Driven Macroeconomic scenarios 

RiskSpan partners with S&P Global Market Intelligence (“Market Intelligence”), employing their CECL model within our Allowance Suite. Each quarter, we apply a new macroeconomic forecast from the S&P Global Ratings team of economists (“S&P Global Economics”). We feed this forecast to all credit models in our platform to influence projections of default and severity and ultimately allowance. S&P Global Economics recent research has focused significantly on coronavirus, including the global and US economic impact of containment and mitigation measures and the recovery timeline. When the credit models take in this bearish outlook for the 3/31/2020 runs, they will return higher defaults and severities compared to prior quarters when the macroeconomic forecast was benign, which in turn will drive higher allowances. Auditors, examiners, and investors will rightly expect to see this.  


RiskSpan advises clients to apply model enhancements or adjustments as follows:

C&I loans, Corporate Bonds, Bank Loans, CLOs 

Corporate & Industrial (C&I) loans often carry internal risk ratings that are ‘through-the-cycle’ evaluations of the default risk and highly independent of cyclical changes in creditworthiness. Corporate bonds carry public credit ratings that are designed to represent forward-looking opinions about the creditworthiness of issuers and obligations, known to be relatively stable throughout a cycle. (Note: higher ratings have been consistently more stable than lower ratings).  

During upswings (downturns), an obligor’s point-in-time or short-term default rates will fall (rise) as the economic environment changes, and credit expectations may be better (worse) than implied by stable credit indicators and associated long-term default frequencies.  

To appropriately reflect the impact of COVID-19 on allowances, most of our clients are now applying industry-specific Point-in-Time (PIT) adjustments, based on Market Intelligence’s Probability of Default Model Market Signals (PDMS). These PIT signals, which use recent, industry-specific trading activity, are used as a guide to form limited adjustments to stable (or in some cases lagging) internal risk ratings of commercial loans and the current credit ratings of corporate bonds. (Adjustments are for the purpose of the CECL model only.)  

Because these adjusted risk ratings are key inputs to Market Intelligence’s CECL Model that we apply to these asset classes, Market Intelligence’s PDMS can influence allowances. Since economic conditions impact certain industry sectors (e.g., airlines, oil and gas, retail) in different ways, the industry-specific notches tend to vary by industry – some positive, some neutral, some negative. Consequently, in a diversified portfolio, we would not ordinarily expect a directional bias to the overall allowance, even though the allowance by industry will be refined. But this assumes a normal economic climate. During a major market downturn like we experienced in the runup to March 31, 2020, notches were negative across almost all industries, and we saw higher allowances as a result. Given the environment, this result is to be expected.  

Resi Loans and RMBS 

Even if we forecast the macroeconomy exactly right, the models of how borrowers perform given different macroeconomic patterns were built on prior decades of experience. Some of the macroeconomic twists and turns that this crisis will unleash will take different shapes than the last crisis.  

For example, a model built on the past two decades of data can only extrapolate what borrowers will do if unemployment goes to 20%; the historical data doesn’t contain such a stress. Even if the macroeconomic patterns do resemble prior crises, policy response will be different, and so will borrower behavior. And then after some recovery period, we expect borrower behavior to fall back into its classic grooves. For these reasons, we recommend model tunings that, all else equal, boost or dampen delinquencies, defaults and recoveries during a time-limited recovery window to account for the near-term impacts of COVID-19. We help clients quantify these model tunings by back-testing model projections against experience from recent weeks. 

In the past month, we have observed slower prepays from housing turnover because social distancing has blocked on-site walkthroughs and therefore home sales. Refinance applications, however, continue to roll in (as expected in this low rate environment), and the refi market is adopting property inspection waivers and remote notarization to work through the demand. As noted under Credit Model Tuning above, we help clients quantify and apply prepay model tunings that act in the short-term and can phase out across the long-term forecast horizon. 


Conventionally, ABS research departments form expected-case projections for underlying collateral by averaging the historical default, severity, and prepay behavior of the issuer. Because CECL calls for expected-case projections, RiskSpan’s bond research team has applied the same approach to generate ABS collateral projections for clients. 

ABS researchers identify stress scenarios by applying multiples or standard deviations to the historical averages. In the current climate, the expected case is a stress case. Therefore, RiskSpan has refined its methodology to apply our stress scenarios as our expected scenarios in times – like now – when the S&P Global Economics baseline macroeconomic forecasts show stress. 


Where clients have their own views of how loans or bonds will perform, we have always empowered them to bypass models and define their own projections of default, severity, and prepayment. 

Resi Loans and RMBS 

For resi loans and RMBS collateral, we have rolled out new “build-a-curve” functionality that allows clients to use our platform to create their own default and severity paths by stipulating drivers such as: 

  • Peak unemployment rate,  
  • How long the peak will last and where unemployment rate will settle, 
  • Share of those unemployed who will roll delinquent, 
  • Length of external forbearance timelines, and 
  • Share of loans that roll delinquent that will eventually default. 


At many institutions, we have seen “Q-factors” (“qualitative” management adjustments on top of modeled allowance results) go from all but forbidden before this crisis to all but required during it. This is because the macroeconomic scenarios now being fed into credit models is beyond the data upon which any vendor or proprietary models were built.  

For example, unemployment rate gradually rose to 10% during the great recession. Many scenarios institutions are now considering call for unemployment rate to spike suddenly to 20% or more. Models can only extrapolate (beyond their known sample) to project obligor performance under these scenarios—there is nothing else they can do. But we know that such extrapolations are unlikely to be exactly right. This creates a strong argument to allow, or even encourage, management adjustments to model results. We are advising many clients to do just that, drawing on available data from natural disasters. 


As important as these quantitative refinements are, performing multiple runs to better understand the range of possible allowance results is equally important to meeting auditor expectations. Whereas before some institutions would use month-end allowances from a month before quarter-end because of tight reporting deadlines, now such institutions are running again at quarter-end, under a very tight timeline, to meet auditor demands for up-to-the-minute analysis. Whereas previously many institutions would run one macroeconomic scenario, now – at the prompting of auditors or their own management – they are running multiple. Institutions that previously did not apply Market Intelligence’s PDMS to their commercial loans and corporate bonds are now running with and without it to evaluate the difference. The dimensionality quickly explodes from one run per quarter to two, ten, or twenty. RiskSpan is happy to offer its platform to clients to support such throughput. 


We will be exploring these topics in greater detail in a webinar on May 28th, 2020 at 1:00 p.m. EDT. You can join us by registering here. I can also be reached directly at

Choosing a CECL Methodology | Doable, Defensible, Choices Amid the Clutter

CECL advice is hitting financial practitioners from all sides. As an industry friend put it, “Now even my dentist has a CECL solution.” With many high-level commentaries on CECL methodologies in publication (including RiskSpan’s ), we introduce this specific framework to help practitioners eliminate ill-fitting methodologies until one remains per segment. We focus on the commercially available methods implemented in the CECL Module of our RS Edge Platform, enabling us to be precise about which methods cover which asset classes, require which data fields, and generate which outputs. Our decision framework covers each asset class under the CECL standard and considers data availability, budgetary constraints, value placed on precision, and audit and regulatory scrutiny. Performance Estimation vs. Allowance Calculations Before evaluating methods, it is clarifying to distinguish performance estimation methods from allowance calculation methods (or simply allowance calculations). Performance estimation methods forecast the credit performance of a financial asset over the remaining life of the instrument, and allowance calculations translate that performance forecast into a single allowance number. There are only two allowance calculations allowable under CECL: the discounted cash flow (DCF) calculation (ASC 326-20-30-4), and the non-DCF calculation (ASC 326-20-30-5). Under the DCF allowance calculation, allowance equals amortized cost minus the present value of expected cash flows. The expected cash flows (the extent to which they differ from contractual cash flows) must first be driven by some performance estimation method. Under the non-DCF allowance calculation, allowance cumulative expected credit losses of amortized cost (roughly equal to future principal losses). These future losses of amortized cost, too, must first be generated by a performance estimation method. Next, we show how to select performance estimation methods, then allowance calculations. Selecting Your Performance Estimation Method Figure 1 below lays out the performance estimation methods available in RiskSpan’s CECL Module. We group methods into “Practical Methods” and “Premier Methods.” In general, Practical Methods calculate average credit performance from a user-selected historical performance data set and extrapolate those historical averages – as adjusted by user-defined management adjustments for macroeconomic expectations and other factors – across the future life of the asset. When using a Practical Method, every instrument in the same user-defined segment will have the same allowance ratio. Premier Methods involve statistical models built on large performance datasets containing instrument-level credit attributes, instrument-level performance outcomes, and contemporaneous macroeconomic data. While vendor-built Premier Methods come pre-built on large industry datasets, they can be tuned to institution-specific performance if the user supplies performance data. Premier Methods take instrument-level attributes and forward-looking macroeconomic scenarios as inputs and generate instrument-level, macro-conditioned results based on statistically valid methods. Management adjustments are possible, but the model results already reflect the input macroeconomic scenario(s). Check marks in Figure 1 indicate the class(es) of financial asset that each performance estimation method covers. Single checkmarks (✔) indicate methods that require the user to provide historical performance data. Double checkmarks (✔✔) indicate methods that, at the user’s option, can be executed using historical performance data from industry sources and therefore do not require the customer to supply historical performance data. All methods require the customer to provide basic positional data as of the reporting date (outstanding balance amounts, the asset class of each instrument, etc.) Figure 1 – Performance Estimation Methods in RiskSpan’s CECL Module [1] Commercial real estate [2] Commercial and industrial loans To help customers choose their performance estimation methods, we walk them through the decision tree shown in Figure 3. These steps to select a performance estimation method should be followed for each portfolio segment, one at a time. As shown, the first step to shorten the menu of methods is to choose between Practical Methods and Premier Methods. Premier Methods available today in the RS Edge Platform include both methods built by RiskSpan (prefixed RS) and methods built by our partner, Global Market Intelligence (S&P). The choice between Premier Methods and Practical Methods is primarily a tradeoff between instrument-level precision and scientific incorporation of macroeconomic scenarios on the Premier side versus lower operational costs on the Practical side. Because Premier Models produce instrument-specific forecasts, they can be leveraged to accelerate and improve credit screening and pricing decisions in addition to solving CECL. The results of Premier Methods reflect macroeconomic outlook using consensus statistical techniques, whereas Practical Methods generate average, segment-level historical performance that management then adjusts via Q-Factors. Such adjustments may not withstand the intense audit and regulatory scrutiny that larger institutions face. Also, implicit in instrument-level precision and scientific macroeconomic conditioning is that Premier Methods are built on large-count, multi-cycle, granular performance datasets. While there are Practical Methods that reference third-party data like Call Reports, Call Report data represents a shorter economic period and lacks granularity by credit attributes. The Practical Methods have two advantages. First, they easier for non-technical stakeholders to understand. Secondly, license fees for Premier Methods are lower than for Practical Methods. Suppose that for a particular asset class, an institution wants a Premium Method. For most asset classes, RiskSpan’s CECL Module selectively features one Premier Method, as shown Figure 1. In cases where the asset class is not covered by a Premier Method in Edge, the next question becomes: does a suitable, affordable vendor model exist? We are familiar with many models in the marketplace, and can advise on the benefits, drawbacks, and pricing of each. Vendor models come with explanatory documentation that institutions can review pre-purchase to determine comfort. Where a viable vendor model exists, we assist institutions by integrating that model as a new Premier Method, accessible within their CECL workflow. Where no viable vendor model exists, institutions must evaluate their internal historical performance data. Does it contain  enough instruments, span enough time ,and include enough fields  to build a valid model? If so, we assist institutions in building custom models and integrating them within their CECL workflows. If not, it’s time a begin or continue a data collection process that will eventually support modeling, and in the meantime, apply a Practical Method. To choose among Practical Methods, we first distinguish between debt securities and other asset classes. Debt securities do not require internal historical data because more robust, relevant data is available from industry sources. We offer one Practical Method for each class of debt security, as shown in Figure 1. For asset classes other than debt securities, the next step is to evaluate internal data. Does it represent (segment-level summary data is fine for Practical Methods) and to drive meaningful results? If not, we suggest applying the Remaining Life Method, a method that has been showcased by regulators and that references Call Report data (which the Edge platform can filter by institution size and location). If adequate internal data exists, eliminate methods that are not asset class-appropriate (see Figure 1) or that require specific data fields the institution lacks. Figure 2 summarizes data requirements for each Practical Method, with a tally of required fields by field type. RiskSpan can provide institutions with detailed data templates for any method upon request. From among the remaining Practical Methods, we recommend institutions apply this hierarchy:

  • Vintage Loss Rate: This method makes the most of recent observations and datasets that are shorter in timespan, whereas the Snapshot Loss Rate requires frozen pools to age substantially before counting toward historical performance averages. The Vintage Loss Rate explicitly considers the age of outstanding loans and leases and requires relatively few data fields.
  • Snapshot Loss Rate: This method has the drawbacks described above, but for well-aged datasets produces stable results and is a very intuitive and familiar method to financial institution stakeholders.
  • Remaining Life: This method ignores the effect of loan seasoning on default rates and requires user assumptions about prepayment rates, but it has been put forward by regulators and is a necessary and defensible option for institutions who lack the data to use the methods above.

Figure 2 – Data Requirements for Practical Methods (Number of Data Fields Required) [3] Denotes fields required to perform method with customer’s historical performance data. If the customer’s data lacks the necessary fields, alternatively this method can be performed using Call Report data. Figure 3 – Methodology Selection Framework Selecting Your Allowance Calculation After selecting a performance estimation method for each portfolio segment, we must select our corresponding allowance calculations. Note that all performance estimation methods in RS Edge generate, among their outputs, undiscounted expected credit losses of amortized cost. Therefore, users can elect the non-DCF allowance calculation for any portfolio segment regardless of the performance estimation method. Figure 5 shows this. A DCF allowance calculation requires the elements shown in Figure 4. Among the Premier (performance estimation) Methods, RS Resi, RS RMBS, and RS Structured Finance require contractual features as inputs and generate among their outputs the other elements of a DCF allowance calculation. Therefore, users can elect the DCF allowance calculation in combination with any of these methods without providing additional inputs or assumptions. For these methods, the choice between the DCF and non-DCF allowance calculation often comes down to anticipated  impact on allowance level. The remaining Premier Methods to discuss are the S&P commercial and industrial loans (C&I) – which covers all corporate entities, financial and non-financial, and applies to both loans and bonds – and the S&P commercial real estate (CRE) method. These methods do not require all the instruments’ contractual features as inputs (an advantage in terms of reducing the input data requirements). They project periodic default and LGD rates, but not voluntary prepayments or liquidation lags. Therefore, users provide additional contractual features as inputs and voluntary prepayment rate and liquidation lag assumptions. The CECL Module’s cash flow engine then integrates the periodic default and LGD rates produced by the S&P C&I and CRE methods, together with user-supplied contractual features and prepayment and liquidation lag assumptions, to produce expected cash flows. The Module discounts these cash flows according to the CECL requirements and differences the present values from amortized cost to calculate allowance. In considering this DCF allowance calculation with the S&P performance estimation methods, users typically weigh the impact on allowance level against the task of supplying the additional data and assumptions. To use a DCF allowance calculation in concert with a Practical (performance estimation) Method requires the user to provide contractual features (up to 20 additional data fields), liquidation lags, as well as monthly voluntary prepayment, default, and LGD rates that reconcile to the cumulative expected credit loss rate from the performance estimation method. This makes the allowance a multi-step process. It is therefore usually simpler and less costly overall to use a Premier Method if the institution wants to enable a DCF allowance . The non-DCF allowance calculation is the natural complement to the Practical Methods. Figure 4 – Elements of a DCF Allowance Calculation I believe the S&P ECL approach is always (even with added prepayment info) a method closely related to, but not a discounted cash flow method, since the allowance for credit losses in S&P approach is calculated directly from the expected credit losses and not as amortized cost minus(-) present value of future cash flows. But this is good since it requires less inputs and easier to relate to macro-economic factors than is a pure DCF. This is consistent with Figure 5. Figure 5 – Allowance Calculations Compatible with Each Performance Estimation Method Once you have selected a performance estimation method and allowance calculation method for each segment, you can begin the next phase of comparing modeled results to expectations and historical performance and tuning model settings accordingly and management inputs accordingly. We are available to discuss CECL methodology further with you; don’t hesitate to get in touch!

Get a Demo

CECL: DCF vs. Non-DCF Allowance — Myth and Reality

FASB’s CECL standard allows institutions to calculate their allowance for credit losses as either “the difference between the amortized cost basis and the present value of the expected cash flows” (ASC 326-20-30-4) or “expected credit losses of the amortized cost basis” (ASC 326-20-30-5). The first approach is commonly called the discounted cash flow or “DCF approach” and the second approach the “non-DCF approach.” In the second approach, the allowance equals the undiscounted sum of the amortized cost basis projected not to be collected. For the purposes of this post, we will equate amortized cost with unpaid principal balance. A popular misconception – even among savvy professionals – is that a DCF-based allowance is always lower than a non-DCF allowance given the same performance forecast. In fact, a DCF allowance is sometimes higher and sometimes lower than a non-DCF allowance, depending upon the remaining life of the instrument, the modeled recovery rate, the effective interest rate (EIR), and the time from default until recovery (liquidation lag). Below we will compare DCF and non-DCF allowances while systematically varying these key differentiators. Our DCF allowances reflect cash inflows that follow the SIFMA standard formulas. We systematically vary time to maturity, recovery rate, liquidation lag and EIR to show their impact on DCF vs. non-DCF allowances (see Table 1 for definitions of these variables). We hold default rate and voluntary prepayment rate constant at reasonable levels across the forecast horizon. See Table 2 for all loan features and behavioral assumptions held constant throughout this exercise. For clarity, we reiterate that the DCF allowances we will compare to non-DCF allowances reflect amortized cost minus discounted cash inflows, per ASC 326-20-30-4. A third approach, which is unsound and therefore excluded, is the discounting of accounting losses. This approach will understate expected credit losses by using the interest rate to discount principal losses while ignoring lost interest itself. Table 1 – Key Drivers of DCF vs. Non-DCF Allowance Differences (Systematically Varied Below)

Variable Definitions and Notes
Months to Maturity Months from reporting date until last scheduled payment
Effective Interest Rate (EIR) The rate of return implicit in the financial asset. Per CECL, this is the rate used to discount expected cash flows when using the DCF approach and, by rule, is calculated using the asset’s contractual or prepay-adjusted cash flows. In this exercise, we set unpaid principal balance equal to amortized cost, so the EIR is the same assuming either contractual or prepay-adjusted cash flows and matches the instrument’s note rate.
Liquidation Lag (Months) Months between first missed payment and receipt of recovery proceeds
Recovery Rate Net cash inflow at liquidation, divided by the principal balance of the loan at the time it went into default. Note that 100% recovery will not include recovery of unpaid interest.

  Table 2 – Loan Features and Behavioral Assumptions Held Constant

Book Value on Reporting Date Par(Amortized Cost = Unpaid Principal Balance)
Performance Status on Reporting Date Current
Amortization Type Level pay, fully amortizing, zero balloon
Conditional Default Rate (Annualized) 0.50%
Conditional Voluntary Prepayment Rate (Annualized) 10.00%

  Figure 1 compares DCF versus non-DCF allowances. It is organized into nine tables, covering the landscape of loan characteristics that drive DCF vs. non-DCF allowance differences. The cells of the tables show DCF allowance minus Non-DCF allowance in basis points. Thus, positive values mean that the DCF allowance is greater.  

  • Tables A, B and C show loans with 100% recovery rates. For such loans, ultimate recovery proceeds match exposure at default. Under the non-DCF approach, as long as recovery proceeds eventually cover principal balance at the time of default, allowance will be zero. Accordingly, the non-DCF allo­wance is 0 in every cell of tables A, B and C. Longer liquidation lags, however, diminish present value and thus increase DCF allowances. The greater the discount rate (the EIR), the deeper the hit to present value. Thus, the DCF allowance increases as we move from the top-left to the bottom-right of tables A, B and C. Note that even when liquidation lag is 0, 100% recovery still excludes the final month’s interest, and a DCF allowance (which reflects total cash flows) will accordingly reflect a small hit. Tables A, B and C differ in one respect – the life of the loan. Longer lives translate to greater total defaulted dollars, greater amounts exposed to the liquidation lags, and greater DCF allowances.
  • Tables G, H and I show loans with 0% recovery rates. While 0% recovery rates may be rare, it is instructive to understand the zero-recovery case to sharpen our intuitions around the comparison between DCF and non-DCF allowances. With zero recovery proceeds, the loans produce only monthly (or periodic) payments until default. Liquidation lag, therefore, is irrelevant. As long as the EIR is positive and there are defaults in payment periods besides the first, the present value of a periodic cash flow stream (using EIR as the discount rate) will exceed cumulative principal collected. Book value minus the present value of the periodic cash flow stream, therefore, will be less than than the cumulative principal not collected, and thus DCF allowance will be lower. Appendix A explains why this is the case. As Tables G, H and I show, the advantage (if we may be permitted to characterize a lower allowance as an advantage) of the DCF approach on 0% recovery loans is greater with greater discount rates and greater loan terms.
  • Tables D, E and F show a more complex (and more realistic) scenario where the recovery rate is 75% (loss-given-default rate is 25%). Note that each cell in Table D falls in between the corresponding values from Table A and Table G; each cell in Table E falls in between the corresponding values from Table B and Table H; and each cell in Table F falls in between the corresponding values from Table C and Table I. In general, we can see that long liquidation lags will hurt present values, driving DCF allowances above non-DCF allowances. Short (zero) liquidation lags allow the DCF advantage from the periodic cash flow stream (described above in the comments about Tables G, H and I) to prevail, but the size of the effect is much smaller than with 0% recovery rates because allowances in general are much lower. With moderate liquidation lags (12 months), the two approaches are nearly equivalent. Here the difference is made by the loan term, where shorter loans limit the periodic cash flow stream that advantages the DCF allowances, and longer loans magnify the impact of the periodic cash flow stream to the advantage of the DCF approach.

Figure 1 – DCF Allowance Relative to Non-DCF Allowance (difference in basis points) Liquidation Lag Table Conclusion

  • Longer liquidation lags will increase DCF allowances relative to non-DCF allowances as long as recovery rate is greater than 0%.
  • Greater EIRs will magnify the difference (in either direction) between DCF and non-DCF allowances.
  • At extremely high recovery rates, DCF allowances will always exceed non-DCF allowances; at extremely low recovery rates, DCF allowances will always be lower than non-DCF allowances. At moderate recovery rates, other factors (loan term and liquidation lag) make the difference as to whether DCF or non-DCF allowance is higher.
  • Longer loan terms both a) increase allowance in general, by exposing balances to default over a longer time horizon; and b) magnify the significance of the periodic cash flow stream relative to the liquidation lag, which advantages DCF allowances.
    • Where recovery rates are extremely high (and so non-DCF allowances are held low or to zero) the increase to defaults from longer loan terms will drive DCF allowances further above non-DCF allowances.
    • Where recovery rates are moderate or low, the increase to loan term will lower DCF allowances relative to non-DCF allowances.[1]

Note that we have not specified the asset class of our hypothetical instrument in this exercise. Asset class by itself does not influence the comparison between DCF and non-DCF allowances. However, asset class (for example, a 30-year mortgage secured by a primary residence, versus a five-year term loan secured by business equipment) does influence the variables (loan term, recovery rate, liquidation lag, and effective interest rate) that drive DCF vs. non-DCF allowance differences. Knowledge of an institution’s asset mix would enable us to determine how DCF vs. non-DCF allowances will compare for that portfolio. Appendix A: The present value of a periodic cash flow stream, as discounted per CECL at the Effective Interest Rate (EIR), will always exceed cumulative principal collected when the following conditions are met: recovery rate is 0%, EIR is positive, and there are defaults in payment periods other than the first. To understand why this is the case, note that the difference between the present value of cash flows and cumulative principal collected has two components: cumulative interest collected, which accrues to the present value of cash flows but not cumulative principal collected, and the cumulative dollar impact of discounting future cash flows, which lowers present value but does not touch cumulative principal collected. The present value of cash flows will exceed cumulative principal collected when the interest impact exceeds the discounting impact. The interest impact is always greater in the early months of a loan forecast because interest makes up a large share of total payment and value lost to discounting is minimal. As the loan ages, the interest share diminishes and the discount impact grows. In the pristine case, where book value equals unpaid principal balance and defaults are zero, the discount effect will finally catch up to the interest effect with the final payment. The present value of the total cash flow stream will thus equal the cumulative principal collected and equal the beginning unpaid principal balance. If there are any defaults in periods later than the first, however, the discount effect can never fully catch up to the interest effect. Table 3 provides one such example. Table 3 – Cash Flow, Principal Losses, Present Value and Allowance under 0% Recovery Loan Features and Assumptions:

  • Reporting-date amortized cost and unpaid principal balance = $10,000
  • 5-year, annual-pay, fully amortizing loan
  • Fixed note rate (and effective interest rate) of 4%
  • 10% conditional voluntary prepayment rate, 0.50% conditional default rate, 0% recovery rate

DCF allowance DCF allowance = $10,000 − $9,872 = $128 Non-DCF allowance = Sum of Principal Losses = $134 We make the following important notes:

  • First-period defaults effectively make the loan a smaller-balance loan and will not cause a difference between the DCF allowance and non-DCF allowance; only defaults subsequent to the first period will drive a difference between the two approaches.
  • Interest-only loans will exacerbate the advantage of DCF allowances relative to non-DCF allowances.
  • For floating-rate instruments, a projected change in coupon rate (based on the known level of the underlying index as of the reporting date) does not change the fact that DCF allowance will be lower than non-DCF allowance if the conditions of 0% recovery rate, positive EIR, and presence of non-first-period defaults are met.

Finally, the discounting approach under CECL is different from that used in finance to assess the fundamental value of a loan. A loan’s fundamental value can be determined by discounting its expected cash flows at a market-observed rate of return (i.e., the rate that links recent market prices on similar-risk instruments to the expected cash flows on those instruments.) As we have noted in other blogs, CECL’s DCF method does not produce the fundamental value of a loan.

    [1] We see just one case in Figure 1 that appears to be an exception to this rule, as we compare the lower-right corner of Table D to the lower-right corner of Table E. What happens between these two cells is that the DCF allowance grows from 36.8 basis points in Table D to 58.9 basis points in Table E (a 60% increase in ratio terms), while the non-DCF allowance grows from 28.4 basis points in Table D to 50.1 basis points in Table E (a 77% increase in ratio terms). Because the allowances rise in general, the subtractive difference between them increases, but we see more rapid growth of the non-DCF allowance as we continue moving from the lower-right corner of Table E to the same corner of Table F.

Get a Demo

Augmenting Internal Loan Data to Comply with CECL and Boost Profit

The importance of sound internal data gathering practices cannot be understated. However, in light of the new CECL standard, many lending institutions have found themselves unable to meet the data requirements. This may have served as a wake-up call for organizations at all levels to look at their internal data warehousing systems and identify and remedy the gaps in their strategies. For some institutions, it may be difficult to consolidate data siloed within various stand-alone systems. Other institutions, even after consolidating all available data, may lack sufficient loan count, timespan, or data elements to meet the CECL standard with internal data alone. This post will discuss some of the strategies to make up for shortfalls while data gathering systems and procedures are built and implemented for the future.  

Identify Your Data

The first step is to identify the data that is available. As many tasks go, this is easier said than done. Often, organizations without formal data gathering practices and without a centralized data warehouse find themselves looking at multiple data storage systems across various departments and a multitude of ad-hoc processes implemented in time of need and not upgraded to a standardized solution. However, it is important to begin this process now, if it is not already underway. As part of the data identification phase, it is important to keep track of not only the available variables, but also the length of time for which the data exists, and whether any time periods have missing or unreliable information. In most circumstances, to meet the CECL standard, institutions should have loan performance data that will cover a full economic cycle by the time of CECL adoption. Such data enables an institution to form grounded expectations of how assets will perform over their full contractual lives, across a variety of potential economic climates. Some data points are required regardless of the CECL methodology, while others are necessary only for certain approaches. At this part of the data preparation process, it is more important to understand the big picture than it is to confirm only some of the required fields—it is wise to see what information is available, even if it may not appear relevant at this time. This will prove very useful for drafting the data warehousing procedures, and will allow for a more transparent understanding of requirements should the bank decide to use a different methodology in the future.  

Choose Your CECL Methodology

There are many intricacies involved in choosing a CECL Methodology. Each organization should determine both its capabilities and its needs. For example, the Vintage method has relatively simple calculations and limited data requirements, but provides little insight and control for management, and does not yield early model performance indicators. On the other hand, the Discounted Cash Flow method provides many insights and controls, and identifies model performance indicators preemptively, but requires more complex calculations and a very robust data history. It is acceptable to implement a relatively simple methodology at first and begin utilizing more advanced methodologies in the future. Banks with limited historical data, but with procedures in place to ramp up data gathering and data warehousing capabilities, would be well served to implement a method for which all data needs are met. They can then work toward the goal of implementing a more robust methodology once enough historical data is available. However, if insufficient data exists to effectively implement a satisfactory methodology, it may be necessary to augment existing historical data with proxy data as a bridge solution while your data collections mature.  

Augment Your Internal Data

Choose Proxy Data

Search for cost-effective datasets that give historical loan performance information about portfolios that are reasonably similar to your go-forward portfolio. Note that proxy portfolios do not need to perfectly resemble your portfolio, so long as either a) the data provider offers filtering capability that enables you to find the subset of proxy loans that matches your portfolio’s characteristics, or b) you employ segment- or loan-level modeling techniques that apply the observations from the proxy dataset in the proportions that are relevant to your portfolio. RiskSpan’s Edge platform contains a Data Library that offers historical loan performance datasets from a variety of industry sources covering multiple asset classes:

  • For commercial real estate (CRE) portfolios, we host loan-level data on all CRE loans guaranteed by the Small Business Administration (SBA) dating back to 1990. Data on loans underlying CMBS securitizations dating back to 1998, compiled by Trepp, is also available on the RiskSpan platform.
  • For commercial and industrial (C&I) portfolios, we also host loan-level on all C&I loans guaranteed by the SBA dating back to 1990.
  • For residential mortgage loan portfolios, we offer large agency datasets (excellent, low-cost options for portfolios that share many characteristics with GSE portfolios) and non-agency datasets (for portfolios with unique characteristics or risks).
  • By Q3 2018, we will also offer data for auto loan portfolios and reverse mortgage portfolios (Home Equity Conversion Mortgages).

Note that for audit purposes, limitations of proxy data and consequent assumptions for a given portfolio need to be clearly outlined, and all severe limitations addressed. In some cases, multiple proxy datasets may be required. At this stage, it is important to ensure that the proxy data contains all the data required by the chosen CECL methodology. If such proxy data is not available, a different CECL model may be best.  

Prepare Your Data

The next step is to prepare internal data for augmentation. This includes standard data-keeping practices, such as accurate and consistent data headers, unique keys such as loan numbers and reporting dates, and confirmation that no duplicates exist. Depending on the quality of internal data, additional analysis may also be required. For example, all data fields need to be displayed in a consistent format according to the data type, and invalid data points, such as FICO scores outside the acceptable range, need to be cleansed. If the data is assembled manually, it is prudent to automate the process to minimize the possibility of user error. If automation is not possible, it is important to implement data quality controls that verify that the dataset is generated according to the metadata rules. This stage provides the final opportunity to identify any data quality issues that may have been missed. For example, if, after cleansing the data for invalid FICO scores, it appears that the dataset has many invalid entries, further analysis may be required, especially if borrower credit score is one of the risk metrics used for CECL modeling. Once internal data preparation is complete, proxy metadata may need to be modified to be consistent with internal standards. This includes data labels and field formats, as well as data quality checks to ensure that consistent criteria are used across all datasets.  

Identify Your Augmentation Strategy

Once the internal data is ready and its limitations identified, analysts need to confirm that the proxy data addresses these gaps. Note that it is assumed at this stage that the proxy data chosen contains information for loans that are consistent with the internal portfolio, and that all proxy metadata items are consistent with internal metadata. For example, if internal data is robust, but has a short history, proxy data needs to cover the additional time periods for the life of the asset. In such cases, augmenting internal data is relatively simple: the datasets are joined, and tested to ensure that the join was successful. Testing should also cover the known limitations of the proxy data, such as missing non-required fields or other data quality issues deemed acceptable during the research and analysis phase. More often, however, there is a combination of data shortfalls that lead to proxy data needs, which can include either time-related gaps, data element gaps, or both. In such cases, the augmentation strategy is more complex. In the cases of optional data elements, a decision to exclude certain data columns is acceptable. However, when incorporating required elements that are inputs for the allowance calculation, the data must be used in a way that complies with regulatory requirements. If internal data has incomplete information for a given variable, statistical methods and machine learning tools are useful to incorporate the proxy data with the internal data, and approximate the missing variable fields. Statistical testing is then used to verify that the relationships between actual and approximated figures are consistent with expectation, which are then verified by management or expert analysis. External research on economic or agency data, where applicable, can further be used to justify the estimated data assumptions. While rigorous statistical analysis is integral for the most accurate metrics, the qualitative analysis that follows is imperative for CECL model documentation and review.  

Justify Your Proxy Data

Overlaps in time periods between internal loan performance datasets and proxy loan performance datasets are critical in establishing the applicability of the proxy dataset. A variety of similarity metrics can be calculated that compare the performance of the proxy loans with the internal loan during the period of overlap. Such similarity metrics can be put forward to justify the use of the proxy dataset. The proxy dataset can be useful for predictions even if the performance of the proxy loans is not identical to the performance of the institutions’ loans. As long as there is a reliable pattern linking the performance of the two datasets, and no reason to think that pattern will discontinue, a risk-adjusting calibration can be justified and applied to the proxy data, or to results of models built thereon.  

Why Augment Internal Data?

While the task of choosing the augmentation strategy may seem daunting, there are concrete benefits to supplementing internal data with a proxy, rather than using simply the proxy data on its own. Most importantly, for the purpose of calculating the allowance for a given portfolio, incorporating some of the actual values will in most cases produce the most accurate estimate. For example, your institution may underwrite loans conservatively relative to the rest of the industry—incorporating at least some of the actual data associated with the lending practices will make it easier to understand how the proxy data differs from characteristics unique to your business. More broadly, proxy data is useful beyond CECL reporting, and has other applications that can boost bank profits. For example, lending institutions can build better predictive models based on richer datasets to calibrate loan screening and loan pricing decisions. These datasets can also be built into existing models to provide better insight on risk metrics and other asset characteristics, and to allow for more fine-tuned management decisions.

Choosing a CECL Methodology

CECL presents institutions with a vast array of choices when it comes to CECL loss estimation methodologies. It can seem a daunting challenge to winnow down the list of possible methods. Institutions must consider considering competing concerns – including soundness and auditability, cost and feasibility, and the value of model reusability. Institutions must convince not only themselves but also external stakeholders that their methodology choices are reasonable, and often on a segment by segment basis, as methodology can vary by segment. It benefits banks, however, to narrow the field of CECL methodology choices soon so that they can finalize data preparation and begin parallel testing (generating CECL results alongside incurred-loss allowance estimates). Parallel testing generates advance signals of CECL impact and may itself play a role in the final choice of allowance methodology. In this post, we provide an overview of some of the most common loss estimation methodologies that banks and credit unions are considering for CECL, and outline the requirements, advantages and challenges of each.

Methods to Estimate Lifetime Losses

The CECL standard explicitly mentions five loss estimation methodologies, and these are the methodologies most commonly considered by practitioners. Different practitioners define them differently. Additionally, many sound approaches combine elements of each method. For this analysis, we will discuss them as separate methods, and use the definitions that most institutions have in mind when referring to them:

  1. Vintage,
  2. Loss Rate,
  3. PDxLGD,
  4. Roll Rate, and
  5. Discount Cash Flow (DCF).

While CECL allows the use of other methods—for example, for estimating losses on individual collateral-dependent loans—these five methodologies are the most applicable to the largest subset of assets and institutions.  For most loans, the allowance estimation process entails grouping loans into segments, and for each segment, choosing and applying one of the methodologies above. A common theme in FASB’s language regarding CECL methods is flexibility: rather than prescribing a formula, FASB expects that the banks consider historical patterns and the macroeconomic and credit policy drivers thereof, and then extrapolate based on those patterns, as well as each individual institution’s macroeconomic outlook. The discussion that follows demonstrates some of this flexibility within each methodology but focuses on the approach chosen by RiskSpan based on our view of CECL and our industry experience. We will first outline the basics of each methodology, followed by their data requirements, and end with the advantages and challenges of each approach.  

Vintage Method

Using the Vintage method, historical losses are tabulated by vintage and by loan age, as a percentage of origination balances by vintage year. In the example below, known historical values appear in the white cells, and forecasted values appear in shaded cells. We will refer to the entire shaded region as the “forecast triangle” and the cells within the forecast triangle as “forecast cells.”[/vc_column_text][/vc_column][/vc_row]

A simple way to populate the forecast cells is with the simple average of the known values from the same column. In other words, we calculate the average marginal loss rate for loans of each age and extrapolate that forward. The limitation of this approach is that it does not differentiate loss forecasts based on the bank’s macroeconomic outlook, which is a core requirement of CECL, so a bank using this method will need to incorporate its macroeconomic outlook via management adjustments and qualitative factors (Q-factors).

As an alternative methodology, RiskSpan has developed an approach to forecast the loss triangle using statistical regression, developing a regression model that estimates the historical loss rates in the vintage matrix as a function of loan age, a credit indicator, and a macroeconomic variable, and then applies that regression equation along with a forecast for the macroeconomic variable (and a mean-reversion process) to populate the forecast triangle. The forecast cells can still be adjusted by management as desired, and/or Q-factors can be used. We caution, however, that management should take care not to double-count the influence of macroeconomics on allowance estimates (i.e., once via models, and again via Q-factors)

Once the results of the regression are ready and adjustments are applied where needed, the final allowance can be derived as follows:

Loss Rate Method

Using the Loss Rate method, the average lifetime loss rate is calculated for historical static pools within a segment. This average lifetime loss rate of a is used as the basis to predict the lifetime loss rate of the current static pool—that is, the loans on the reporting-date balance sheet.

In this context, a static pool refers to a group of loans that were on the balance sheet as of a particular date, regardless of when they were originated. For example, within an institutions’ owner-occupied commercial real estate portfolio, the 12/31/06 static pool would refer to all such loans that were on the institution’s balance sheet as of December 31, 2006. We would measure the lifetime losses of such a static pool beginning on the static pool date (December 31, 2006, in this example) and express those losses as a percentage of the balance that existed on the static pool date. This premise is consistent with what CECL asks us to do, i.e., estimate all future credit losses on the loans on the reporting-date balance sheet.

A historical static pool fully aged if all loans that made up the pool are either paid in full or charged off, where payments in full include renewals that satisfy the original contract. We should be wary of including partially aged static pools in the development of average lifetime loss estimates, because the cumulative loss rates of partially aged pools constitute life-to-date loss rates rather than complete lifetime loss rates, and inherently understates the lifetime loss rate that is required by CECL.

To generate the most complete picture of historical losses, RiskSpan constructs multiple overlapping static pools within the historical dataset of a given segment and calculates the average of the lifetime loss rates of all fully aged static pools.  This provides an average lifetime loss rate over a business cycle as the soundest basis for a long-term forecast. This technique also allows, but does not require, the use of statistical techniques to estimate lifetime loss rate as a function of the credit mix of a static pool.

After the average lifetime loss rate has been determined, we can incorporate management’s view of how the forward-looking environment will differ from the lookback period over which the lifetime loss rates were calculated, via Q-Factors.

The final allowance can be derived as follows:

PDxLGD Method

Methods ranging from very simple to very sophisticated go by the name “PD×LGD.” At the most sophisticated end of the spectrum are models that calculate loan-by-loan, month-by-month, macro-conditioned probabilities of default and corresponding loss given default estimates. Such estimates can be used in a discounted cash flow context. These estimates can also be used outside of a cash flow context; we can summarize these monthly estimates into a cumulative default probability and corresponding exposure-at-default and loss-given-default estimates, which yield a single lifetime loss rate estimate. At the simpler end of the spectrum are calculations of the lifetime default rates and corresponding loss given default rates of static pools (not marginal monthly or annual default rates). This simpler calculation is the method that most institutions have in mind when referring to “PD×LGD methods,” so it is the definition we will use here.

Using this PDxLGD method, the loss rate is calculated based on the same static pool concept as that of the Loss Rate method. As with the Loss Rate method, we can use the default rates and loss given default rates of different static pools to quantify the relationship between those rates and the credit mix of the segment, and to use that relationship going forward based on the credit mix of today’s portfolio. However, under PDxLGD, the loss rate is a function of two components: the lifetime default rate (PD), and the loss given default (LGD).  The final allowance can be derived as follows:

Because the PDxLGD and Loss Rate methods derive the Expected Loss Rate for the segment using different but related approaches, one of the important quality controls is to verify that the final calculated rates are equal under both methodologies, and that the cause of any discrepancies is investigated.

Roll Rate Method

Using the Roll Rate method, ultimate losses are predicted based on historical roll rates and the historical loss given default estimate.  Roll rates are either (a) the frequency with which loans transition from one delinquency status to another, or (b) the frequency with which loans “migrate” or “transition” from one risk grade to another.  While the former is preferred due to its transparency and objectivity, for institutions with established risk grades, the latter is an appropriate metric.

Under this method, management can apply adjustments for macroeconomic and other factors at the individual roll rate level, as well as on-top adjustments as needed. Roll rate matrices can included prepayment as a possible transition, thereby incorporating prepayment probabilities. Roll rates can be used in a cash flow engine that incorporates contractual loan features and generates probabilistic (expected) cash flows, or outside of a cash flow engine to generate expected chargeoffs of amortized cost. Finally, it is possible to use statistical regression techniques to express roll rates as a function of macroeconomic variables, and thus, to condition future roll rates on macroeconomic expectations.

The final allowance can be derived as follows:

Discounted Cash Flow (DCF) Method

Discounting cash flows is a way of translating expected future cash flows into a present value. DCF is a loan-level method (even for loans grouped into segments), and thus requires loan-by-loan, month-by-month forecasts of prepayment, default, and loss-given-default forecasts to translate contractual cash flows into prepay-, default-, and loss-given-default-adjusted cash flows. Although such loan-level, monthly forecasts could be derived using any method, most institutions have statistical forecasting techniques in mind when thinking about a DCF approach. Thus, even though statistical forecasting techniques and cash flow discounting are not inextricably linked, we will treat them as a pair here.

The most complex, and the most robust, of the five methodologies, DCF (paired with statistical forecasting techniques) is generally used by larger institutions that have the capacity and the need for the greatest amount of insight and control. Critically, DCF capabilities give institutions the ability (when substituting the effective interest rate for a market-observed discount rate) to generate fair value estimates that can serve a host of accounting and strategic purposes.

To estimate future cash flows, RiskSpan uses statistical models, which comprise:

  • Prepayment sub-models
  • Probability-of-default or roll rate sub-models
  • Loss-given-default sub-models

Allowance is then determined based on the expected cash flows, which, similarly to the Roll Rate method, are generated based on the rates predicted by the statistical models, contractual loan terms, and the loan status at the reporting date.

Some argue that an advantage of the discounted cash flow approach is lower Day 1 losses. Whether DCF or non-DCF methods produce a lower Day 1 allowance, all else equal, depends upon the length of the assumed liquidation timeline, the discount rate, and the recovery rate. This is an underdiscussed topic that merits its own blog post. We will cover this fully in a future post.

The statistical models often used with DCF methods use historical data to express the likelihood of default or prepayment as a mathematical function of loan-level credit factors and macroeconomic variables.

For example, the probability of  transitioning from “Current” status to “Delinquent” at montht can be calculated as a function of that loan’s loan age at  multiplied by a sensitivity factor β1 on the loan age variable derived based on the data in the historical dataset, the loan’s FICO multiplied by a sensitivity factor β2, and the projected unemployment rate based on management’s macroeconomic assumptions at montht multiplied by a sensitivity factor β3.  Mathematically,

Because macroeconomic and loan-level credit factors are explicitly and transparently incorporated into the forecast, such statistical techniques reduce reliance on Q-Factors. This is one of the reasons why such methods are the most scientific.

Historical Data Requirements

The table below summarizes the historical data requirements for each methodology, including the dataset type, the minimum required data fields, and the timespan.

In conclusion, having the most robust data allows the most options; for institutions with moderately complex historical datasets, Loss Rate, PDxLGD, and Vintage are excellent options.  With limited historical data, the Vintage method can produce a sound allowance under CECL.

While the data requirements may be daunting, it is important to keep in mind that proxy data can be used in place of, or alongside, institutional historical data, and RiskSpan can help identify and fill your data needs.  Some of the proxy data options are summarized below:

Advantages and Challenges of CECL Methodologies

Each methodology has advantages, and each carries its own set of challenges.  While the Vintage method, for example, is forgiving to limited historical data, it also provides limited insight and control for further analysis.  On the other hand, the DCF method provides significant insight and control, as well as early model performance indicators, but requires a robust dataset and advanced statistical expertise.

We have summarized some of the advantages and challenges for each method below.

In addition to the considerations summarized in the table, it is important to consider audit and regulatory requirements. Generally, institutions facing higher audit and regulatory scrutiny will be steered toward more complex methods. Also, bankers who intend to leverage the loan forecasting model they use for CECL for strategic decision-making (for example, loan screening and pricing decisions), and who desire granular insight and dials around their allowance numbers, will gravitate toward methodologies that afford more precision. At the other end of the spectrum, the methods that provide less precision and insight generally come with lighter operational burden.

Choosing Your CECL Methodology

Choosing the method that’s right for you depends on many factors, from historical data availability to management objectives and associated operational costs.

In many cases, management can gain a better understanding of the institutional allowance requirements after analyzing the results determined by multiple complementary approaches.

RiskSpan is willing to talk further with individual institutions about their circumstances, as well as generate sample results using a set of various methodologies.

CECL–Why Implement Now?

FASB permits early adoption of CECL for fiscal years beginning after December 15, 2018, including interim periods within the fiscal year. The decision of whether to pursue formal early adoption is a complex one hinging on specific factors that vary among institutions. We are finding, however, that early implementation of a CECL solution offers many potential benefits regardless of whether an institution opts for early adoption.

The benefits of implementing a CECL solution comfortably in advance of adoption are analogous to those of a flight simulator. Both enable professionals to gain insights into how a new system functions and reacts to various influences in a riskless environment. Like a flight simulator, early CECL system implementation enables institutions to move forward with this important standard in a controlled and well-considered manner. It benefits institutions by allowing time for analysis and supporting optimal decision-making.

It accomplishes this in the following ways.


Enhanced Data Readiness

Data readiness is a crucial factor in CECL implementation. Analysts and risk managers need to know whether their data will be sufficient to support effective functioning of a credit loss forecasting model specifically designed for their portfolio. Early implementation facilitates answering this question in advance by enabling institutions to assess the following:

  1. Internal data completeness
  2. Internal data reliability
  3. Internal data accessibility
  4. Externally sourced forecasting elements

Data preparation often requires a considerable investment of time and effort that can span months.  Obtaining data and developing interfaces to legacy systems and other sources requires planning and sufficient time to develop and test queries. Early implementation provides a safe environment for exploring data nuances and figuring out what works best.


Appropriate Cohort Segmentation

Loans and revolving credits are segmented into risk-related cohorts for credit loss forecasting purposes.  Clients frequently ask whether existing segments should be revisited or revised in a CECL implementation. The answer, like most, depends on many factors. However, when new segmentation is called for, it takes time to analyze the effect of risk factors on losses and to devise the new segmentation strategy. That work would ideally begin well in advance of formal CECL adoption, along with testing methodologies for each segment.


Early Operational and Internal Control Insights

Capturing the real-time data necessary to book lifetime losses at origination will pose new and daunting challenges. Traditional ALLL processes historically enabled banks to largely ignore originations occurring near the end of an accounting period. CECL changes all this by requiring day-one losses to be booked upon origination, thus potentially taxing data collection processes that likely will need to be streamlined in order to gather the data necessary to produce loss forecasts in time for the accounting close.

In addition to overcoming the obvious operational obstacles associated with this degree of data processing, early implementation gives banks an advance opportunity to assess the internal controls issues that will inevitably arise. Datasets and models that previously were not subject to SOX and financial reporting control testing may need to be reviewed in a new light. Controls around the databases and models supporting the life-of-loan calculation will almost certainly need to be enhanced to meet financial reporting expectations. Figuring out how to address all this in the relative safety of an early-implementation simulator is preferable to learning it on the fly.

Capital and Strategic Planning Benefits

Because CECL is generally expected to increase loan loss reserves, the effect on bank capital should be understood as soon as possible. Banks will want to understand the potential impact of CECL on capital to inform discussions with the board, auditors, regulators, and shareholders and to factor into capital planning in advance of formal adoption.

It is particularly important for banks that are active in mergers and acquisitions to understand the implications of CECL on earnings. This will inform efforts currently underway or planned for the next couple of years. Early CECL implementation also provides the ability to begin making lending and origination decisions informed by CECL.

Compared with the relatively complex decisions surrounding whether to pursue early CECL adoption, deciding whether to implement a CECL solution early is straightforward. Regardless of whether a bank opts for early adoption, early implementation promises a range of benefits that will support optimal risk management decisions going forward, regardless of when CECL is ultimately adopted.

Get Started
Get A Demo