COVID-19’s impact on the CRT markets
Executive Summary
Between 2017 and 2020, the mortgage credit risk transfer (CRT) issuances of Fannie Mae and Freddie Mac (together, the government-sponsored enterprises, or GSEs) reflected approximately a 75/25 split between capital markets executions and reinsurance executions.1 Recent financial market disruptions, caused by the rapid spread of and response to COVID-19, have resulted in stark changes to the market price and appetite for CRT, particularly on capital markets executions.
Placing a larger share of GSE CRT in reinsurance executions, not only in periods of stress but also in benign periods, could allow the GSEs to further develop more stable sources of private capital. Sharing a greater percentage of risk with these more persistent sources of capital, with long-term views of risk, could lead to beneficial impacts to Enterprise Capital relief through credit cycles (e.g., more steady access to private capital at potentially lower average costs).
In support of this view, this paper discusses COVID-19’s impact on the CRT market and reflects, more generally, on the use of capital markets executions and reinsurance executions to meet the GSEs’ goals. Specifically, this paper:
- Provides statistics on recent secondary CRT market price movements and compares these reactions to credit risk fundamentals.
- Compares new issuance pricing for both capital markets executions and reinsurance executions over the past two years.
- Provides a summary of capital markets and reinsurance CRTs from the perspective of implications for Enterprise Capital relief.
Introduction
CRT is an important component of the U.S. housing finance system. CRT facilitates the sharing of credit risk from pools of mortgages to private market participants. This transfer reduces the risk, and the amount of required capital, of mortgage finance companies, which in turn allows them to increase the availability of credit to mortgage borrowers.
Under conservatorship of the Federal Housing Finance Agency (FHFA), the GSEs have come to market with various structures to transfer mortgage credit risk to both investors, via securities offerings (capital markets execution), and reinsurers, via insurance offerings (reinsurance execution).
Fannie Mae’s and Freddie Mac’s securities offerings, Connecticut Avenue Securities (CAS) and Structured Agency Credit Risk (STACR), respectively, issue nonguaranteed fixed-income securities within a multi-tranche structure that allows institutional investors to gain exposure to specific diversified pools of mortgages. The securities are nonguaranteed in-that principal losses on the mortgage collateral are passed through to investors in the securities.
Fannie Mae’s and Freddie Mac’s credit insurance transaction offerings, Credit Insurance Risk Transfer (CIRT) and Agency Credit Insurance Structure (ACIS), respectively, are excess-of-loss credit insurance transactions whereby the GSEs cede portions of the credit risk associated with specific diversified pools of mortgages. Under these structures, reinsurers collect a reinsurance premium. In return, reinsurers reimburse the GSEs for credit losses incurred on the mortgage pool, according to the insurance terms of the offering.
Between 2017 and 2020, there has been approximately a 75/25 split between capital markets executions and reinsurance executions; although Freddie Mac has utilized the reinsurance execution to a greater degree than Fannie Mae.
Secondary market: COVID-19 disruptions
The economic and financial market stress as a result of the COVID-19 pandemic has wreaked havoc on CAS/STACR secondary market trading volume and market valuations. Using Milliman’s CRT analytics platform, M-PIRe, we are able to produce a fundamental view of potential credit losses for CRT securities and compare the estimate to recent trading levels.
COVID-19 impact on the CRT secondary market
Financial markets have experienced stress and sharp declines in market valuations as a result of the COVID-19 pandemic’s effect on the United States.
One measure to quantify this impact on fixed-income securities is to view the change in the index value of Markit Credit Indices. These indices track different segments of the debt markets (e.g., corporate) across various regions (e.g., North America) and are referenced in various tradable products such as credit default swaps (CDS), and exchange-listed products.2 A credit default swap is a derivative contract that offers counterparty protection against a credit event (e.g., default or bankruptcy of an issuer). The Credit Default Swap Index (CDX) is a benchmark index that tracks a basket of U.S. single-issuer CDS. The CDX is standardized and exchange-traded, unlike single CDS, which provides a high level of liquidity and transparency.
CDX.NA.HY (CDX HY) is an index based on a basket of 100 liquid North American (NA) single-name entities with high yield credit ratings. Comparing this index over March 2020 serves as a way to measure COVID-19's impact on the market value of corporate debt.
Figure 1 compares CDX HY to Vista Data Services’ CRT 2017 M Index. Vista’s CRT 2017 M Index aggregates all standard CAS and STACR M2 securities issued in 2017. Vista’s CRT Index is constructed based on weighting market pricing and other elements of each security by the current amount outstanding of each security. This index serves to view the impact COVID-19 on CRT market trading. Figure 1 shows that, despite initially widening in tandem (with CDX HY widening slightly quicker), CRT 2017 M experiences a swift and pronounced further widening as compared to CDX HY on approximately March 19, 2020. This decoupling was influenced by several different dynamics within the CRT secondary market.3
Figure 1: CRT M Index (Discount Margin) vs. CDX HY Indices, 2020 To-Date
Partially explaining the divergence between the high-yield index and the CRT 2017 M Index, the bid-offer spreads for CRT bonds widened considerably in late March. As the CRT market is smaller and less liquid relative to corporate debt, many of the trades were completed without the dealers absorbing the volumes into their positions—more specifically, dealers would try to find a party to take the other side of the trade before executing as opposed to executing and then finding a party to take the other side of the trade. Underscoring the market dislocation, around this time B-1 tranches stopped trading or being quoted by dealers altogether. B-1 tranches are subordinate to M tranches in the deal structure. The rare quotes for B-1 tranches were generally between dollar prices of 20 and 35. This market dislocation demonstrates the substantial challenge that market stresses can place on the liquidity of CRT bonds in the capital markets—impacting the ability of investors to trade in and out of positions.
Since March 25, the CRT 2017 M Index has tightened considerably. Peaking at around 1,600 basis points, discount margins have compressed for CRT 2017 M tranches, hovering around 700 basis points during the first week of April. However, it still remains higher than the CDX HY—showing that the decoupling of the indices is persisting.4
This observation that the CRT market has a relatively more adverse reaction—in terms of price and liquidity—to a broader market stress than other markets may lead to important consequences going forward. Lack of a well-functioning secondary market can threaten the viability of primary CRT issuances.
Evaluating fundamental value
Using relatively well-traded 2017 low and high loan-to-value (LTV) transactions—STACR 2017-DNA1 and STACR 2017-HQA1 respectively—Figure 2 shows historical price movements across time using secondary market trades. Both the B and M tranches, in the DNA and HQA transactions, traded within a dollar price range of 104 to 116 from January 2018 through December 2019.
Figure 2: STACR 2017-DNA1 and 2017-HQA1 Dollar Prices, Indices 2018-2019
However, over March 2020, the values of these securities dropped precipitously, as shown in Figure 3. By March 25, 2020, both the B1 and M2 tranches of the DNA1 and HQA1 deals were trading in the mid-60s. This steep drop translates to dollar price levels that were consistent with assuming the securities would incur future principal losses, despite seasoning of the underlying collateral and associated credit enhancement associated with pool amortization to-date.
Figure 3: STACR 2017-DNA1 and 2017-HQA1 Dollar Prices, 2020 To-Date
Milliman’s M-PIRe CRT Analytics Platform was used to evaluate the probability of principal loss given performance to-date and alternative future economic scenarios. M-PIRe is Milliman’s turnkey platform for evaluating and analyzing mortgage credit risk transfer transactions, including single-family bond executions, reinsurance executions, multifamily transactions, and private mortgage insurance transactions.
The table in Figure 4 shows M-PIRe’s projected percentage of issued principal returned by tranche for STACR 2017-DNA1 and 2017-HQA1. This table includes deterministic estimates using Moody’s Economy.com forecasts, as well as stochastic results from M-PIRe’s Monte Carlo simulation model.5 The Moody’s scenarios were published as part of Moody’s March Mid-Cycle Update and explicitly incorporate estimated impacts of the COVID-19 pandemic.6
By 2020 Q2, the Moody’s Baseline, S3, and S4 scenarios forecast the U.S. unemployment rate to reach 8.7%, 13.0%, and 16.9%, respectively. These unemployment forecasts (along with scenario-specific home price and interest rate forecasts) are used as inputs in M-PIRe’s loan-level mortgage performance model. M-PIRe’s stochastic approach centers the median of the distribution at approximately the Moody’s Baseline scenario and simulates variation around the Baseline scenario.
Figure 4: Percentage of Principal Returned by Scenario, STACR 2017-DNA1 and 2017-HQA1
Percentage of Principal Returned by Scenario
STACR 2017-DNA1 and 2017-HQA1
Source: Milliman M-PIRe; as-of 4/27/2020
| STACR 2017-DNA 1 | STACR 2017-HQA 1 | |||
| Scenario | B1 | M2 | B1 | M2 |
|---|---|---|---|---|
| Moody's Baseline | 100.0% | 100.0% | 100.0% | 100.0% |
| M-P IRe Stochastics Median | 100.0% | 100.0% | 100.0% | 100.0% |
| M-P IRe Stochastics Mean | 97.2% | 99.7% | 97.4% | 99.7% |
| Moody's S3 | 100.0% | 100.0% | 100.0% | 100.0% |
| Moody's S4 | 100.0% | 100.0% | 100.0% | 100.0% |
| M-P IRe Stochastics 97.5th Pct'l | 44.1% | 100.0% | 41.2% | 100.0% |
| M-P IRe Stochastics 99.0th Pct'l | 0.0% | 90.4% | 0.0% | 87.2% |
| M-P IRe Stochastics 99.5th Pct'l | 0.0% | 78.1% | 0.0% | 76.2% |
| M-P IRe Stochastics 99.9th Pct'l | 0.0% | 32.2% | 0.0% | 37.3% |
Source: Milliman M-PIRe as of April 27, 2020.
Under Moody’s Baseline, S3, and S4 scenarios, neither B1 nor M2 are estimated to incur principal write-downs. Under M-PIRe’s Monte Carlo simulation model, which produces more severe stresses than Moody’s deterministic scenarios, B1 tranches are approximately exhausted at the 99th percentile.
Due to the GSEs’ forbearance policies released in reaction to COVID-19, there is significant uncertainty regarding the timing of principal repayment for CRT bonds. If scheduled principal is not received from a borrower (e.g., a borrower in forbearance), then the principal payments passed through to bondholders are similarly reduced. However, once repaid after the forbearance period, the principal payments are similarly passed through to the bondholders. If there is a permanent loan modification or principal reduction, then actual losses may be passed through to the investors. As there is significant uncertainty and many assumptions that impact the timing of cash flows, this analysis focuses on the probability of ultimate principal losses.7
In the M-PIRe models, the M2 tranches are not exhausted under any of the 10,000 scenarios. The M2 tranches are projected to incur principal write-downs in very few of the simulated percentiles. As of March 2020, the collateral pool for these deals has amortized to under 70% of the original pool balance. Losses-to-date for the collateral has been less than 1 basis point, and the attachment point for the M2 tranche is 100 basis points for each deal. Combining these facts, future credit losses would need to exceed 1.42% (100 / 0.7) on the remaining collateral. Using M-PIRe models, this equates to credit losses approximately twice the level as predicted under a 2007-2010 global financial crisis repeat scenario. There is significant uncertainty associated with future mortgage performance in the current environment;8 however, the above calculation demonstrates that unprecedented losses would be required to produce cash flows reflective of current market prices.
In contrast to the reinsurance markets, which are accustomed to taking illiquid long-term risk positions, it is common for liquidity risk to become a significant contributor to bond prices,9 particularly in times of market stress, and the CRT market appears to be exhibiting that behavior. M-PIRe’s CRT Analytics Platform can provide insight into how current CRT prices, including a high liquidity premium, are much lower than would be implied via a reasonable fundamental value analysis.
New issuance market: STACR/ACIS relative pricing
Figure 5 displays the relationship between CDX HY and Vista Data Services CRT 2017 M Index in more benign times (as compared to Figure 1 above).
In late 2018, financial market softness, driven by increased trade war tensions between the United States and China, caused spreads of both indices to widen temporarily. In November 2018, both corporate spreads and CRT spreads increased, consistent with general market movements during the time period.
Figure 5: CRT M Index (Discount Margin) vs. CDX HY, Indices 2018-2019
Figure 6 displays how the above market spreads translated into relative STACR/ACIS new issue pricing across the DNA series.10 The chart shows the CRT 2017 M Index in red and new pricing on capital markets executions in blue. The chart demonstrates that new issuance pricing reflects market prices, with a slight lag.11 Specifically, towards the end of 2018, the Vista CRT 2017 M Index increased from 150 basis points to over 235 basis points. This corresponded with new issuance pricing of 450 basis points in January 2019, up from 360 basis points in October 2018 (prior to the market turmoil). Throughout the remainder of 2019, new issuance pricing on the capital markets executions (STACR transactions) declined to less than 300 basis points, consistent with strong bond market performance and views on fundamentals. In contrast, during this time period, the average premium rate for the reinsurance executions (ACIS transactions) remained relatively stable at around 450 basis points, reflecting the reinsurance community’s long-term view of risk.
Figure 6: STACR/ACIS DNA Series Issuance Spreads vs. Reinsurer Premium (2018-DNA3 through 2020-DNA2)
There has not been a DNA (low LTV) series transaction brought to market since the stress related to COVID-19 in March 2020. However, if the dynamic present in late 2018 holds true, it is expected that STACR spreads will widen to a larger degree relative to ACIS. We can see this beginning to occur in Figure 7—which shows the relative STACR/ACIS new issue pricing across the HQA (high LTV) series. Figure 7 shows a 63% increase in STACR spreads, relative to a 30% increase in ACIS pricing.
Figure 7: STACR/ACIS HQA Series Issuance Spreads vs. Reinsurer Premiums (2018-HQA2 through 2020-HQA2)
STACR/ACIS 2020-HQA2 closed on March 18, 2020 (with pricing occurring in the weeks prior). Looking at that data point as an early indication of how the market stress influences relative pricing between the two executions, STACR spreads have widened by a larger degree than ACIS premiums. This behavior is consistent with the trends observed in late 2018. If this trend continues (as the full brunt of the secondary market disruptions begin to impact new issuance pricing) then GSE reinsurance executions could become competitive relative to the capital markets executions.
Following the 2007-2010 global financial crisis, the reinsurance industry began insuring a larger amount of mortgage credit risk—both via the GSE reinsurance CRT transactions as well as via private mortgage insurer (PMI) reinsurance treaties. During this time, many participants in the industry (both brokers and reinsurers) invested heavily in teams of mortgage analytics professionals to underwrite and monitor this growing line of business. Taking a data-driven approach, reinsurance markets, via in-house analytics and service providers, have developed sophisticated analytical models to evaluate mortgage credit risk throughout the housing cycle. As a result, the GSEs have seen the number of unique reinsurers participating in these transactions grow since program inception despite relative issuance sizes still being four to five times smaller than the analogous capital markets issuances. This investment and commitment to mortgage credit risk contributes to the lower volatility in pricing seen in reinsurance executions.
Lacking GSE CRT transactions closing since March 18, we can view another index from Vista Data Services to further examine the dynamics above. Although the CRT 2017 M Index tightened considerably from the widest spreads of late March, plotting the CRT 2020 M Index shows that recent vintages have not responded similarly. Figure 8 displays that, in late March, the 2020 M Index showed a decoupling from earlier vintages (not tightening after initial COVID-related market dislocation). This shows the relatively lower market appetite for the on-the-run securities.
Figure 8: CRT M Index (Discount Margin), 2017 Vintage vs. 2020 Vintage
If this trend persists, and the dislocation of the capital markets makes new issuances untenable, the reinsurance execution could become a more economical option for GSEs to transfer risk. Furthermore, in order to continue to obtain the same levels of Enterprise Capital relief on new acquisitions, the reinsurance execution may need to be expanded.
Enterprise Capital: Evaluating the benefit of CRT
In 2018, FHFA began seeking comments on a proposed regulation on risk-based capital requirements for Fannie Mae and Freddie Mac.12 In the proposed rule, the GSEs obtain capital relief “equal to the reduction in credit risk capital from transferring all or part of a credit risk exposure.” While FHFA is planning on re-proposing the entire regulation on Enterprise Capital requirements in 2020,13 in the interim we discuss CRT’s benefit to the GSEs.
While the current regulatory framework is currently in flux, the 2018 proposed Enterprise Capital rule included a framework where the GSEs' credit risk capital requirements were reduced to reflect the benefit of CRT transactions. The differences in capital relief, between capital markets executions and reinsurance executions, will largely be a function of GSE regulation and capital rulemaking—which is beyond the scope of this paper. However, the difference would generally be attributable to the collateralization of the CRT structure (capital markets executions are fully collateralized, while reinsurance executions are partially collateralized).
Regardless of the particular nature of the capital relief, both the capital markets executions and reinsurance executions provide the GSEs a source of private capital and serve the same general purpose. At a high level, the GSEs’ net cost of Enterprise Capital relief from engaging in CRT is equal to the cost/benefit of the transaction (i.e., cost of coverage minus the credit losses transferred) divided by the amount of capital relief provided by the transaction.
The cost of coverage across the two executions fluctuates across time, with financial market stresses having a disproportionately adverse impact on capital markets executions versus reinsurance executions. If the dynamics described in the section above persist, then the cost of capital relief under reinsurance executions may become more economical for the GSEs relative to capital markets executions. Furthermore, if the capital markets disruptions persist, capital markets CRT, which has become a central part of the GSEs business, may not be available. Thinking holistically regarding Enterprise Capital relief costs/benefits of CRT throughout the cycle may further strengthen the case for increased use of reinsurance executions.
Conclusion: Charting a path forward
The current COVID-19 financial markets stress highlights potential negative consequences of the GSEs relying on capital markets executions for the majority of their CRT transactions. Capital markets tend to be more volatile and represent a less stable source of risk capital during market stress. Reinsurance markets heavily invest in understanding the risks of exposures during the underwriting process. In addition, reinsurers do not often “move in and out” of CRT reinsurance positions during market volatility. Therefore, they tend to have a longer-term perspective of the exposure and price risks through market cycles.
Consideration of the relative instability of secondary CRT trading capital markets, the potential for price dislocations relative to the fundamental values of the transactions, the relatively small difference in cost of coverage in good times, and the potential cost of lapses in capital market issuances during the bad times are all crucial when shaping GSE CRT execution strategies.
1For information from Freddie Mac, on STACR/ACIS issuances, see https://crt.freddiemac.com/offerings/. For information from Fannie Mae, on CIRT issuances, see https://www.fanniemae.com/resources/file/credit-risk/pdf/cirt-deal-pricing-information.pdf. For information from the Urban Institute, on CRT issuances, see https://www.urban.org/sites/default/files/publication/101926/housing-finance-at-a-glance-a-monthly-chartbook-march-2020.pdf.
2Markit (August 2016). Markit CDX High Yield and Markit CDX Investment Grade Index Rules. Retrieved May 10, 2020, from https://www.markit.com/Company/Files/DownloadFiles?CMSID=1ec08990b6814532ad075e242b34160d.
3CRT securities have numerous complex analytical and market-driven pricing drivers that may contribute to the patterns observed in this paper. Changing prepayment expectations are a major driver of price fluctuations. This paper focuses on a subset of pricing drivers and highlights the recent extraordinary market dynamics.
4On April 9, the Federal Reserve announced expanded plans to buy corporate bonds both at an investment-grade level as well as high-yield. This action may contribute to the observed differences—as the Fed does not buy CRT. See https://www.federalreserve.gov/newsevents/pressreleases/monetary20200409a.htm.
5M-PIRe’s simulation model performs 10,000 trials of alternative future collateral performance, accounting for model risk, economic risk, and changes in borrower behavior. The model captures a broader spectrum of future uncertainty relative to other third-party models reviewed by Milliman.
6Note that Moody’s subsequent updates have deteriorated relative to these scenarios. Furthermore, on May 8, the U.S. Bureau of Labor Statistics reported that the unemployment rate rose to 14.7% in April (see https://www.bls.gov/news.release/pdf/empsit.pdf). This data point is not incorporated in these forecasts. Milliman continues to use recent economic information to estimate the fundamental value of the securities and may update this analysis in the future.
7Interest and timing of cash flows will also factor into a more complete evaluation of the fundamental value of the securities. This analysis aims to provide a simple benchmark.
8Our estimates make no provision for extraordinary future emergence of new classes of losses or types of losses not sufficiently represented in historical mortgage performance databases or that are not yet quantifiable, including the potential impact of the emerging situation regarding the COVID-19 pandemic. There is substantial uncertainty regarding the impact of COVID-19 on the level of mortgage default, prepayment, and severity rates. Exposures, claim frequency, and claim severity will likely be affected in ways we cannot currently estimate. It is important to recognize that actual losses may emerge significantly higher or lower than the estimates in this analysis. It is unknown how the COVID-19 pandemic may affect the timeliness of future default losses.
9Liquidity risk’s contribution to bond prices for other bond issuers (e.g., corporate and municipal), in times of market stress, has been previously studied and quantified; see Acharya et al. (2010) and Ang et al. (2014), respectively.
10STACR/ACIS transactions were used in this analysis to control for structural differences across the two executions. STACR/ACIS deals share a single reference pool and multi-tranche structure. In contrast, Fannie Mae’s CAS and CIRT transactions have disparate reference pools and dissimilar structures.
11New issuance data points are plotted at a date 10 business days prior to the STACR. This more closely aligns the deals with the dates where investors or reinsurers are evaluating and underwriting the transactions.
12The full text of the 2018 proposed regulation is available at https://www.gpo.gov/fdsys/pkg/FR-2018-07-17/pdf/2018-14255.pdf.
13FHFA (November 19, 2019). FHFA will re-propose enterprise capital rule in 2020. Press release. Retrieved May 10, 2020, from https://www.fhfa.gov/Media/PublicAffairs/Pages/FHFA-Will-Re-propose-Enterprise-Capital-Rule-in-2020.aspx.