As we enter the fourth month of the COVID-19 pandemic, the full impact on the U.S. economy is still unfolding and likely won't be fully realized for months to come. In the article, "Credit FAQ: Assessing The Credit Effects Of COVID-19 On U.S. RMBS," published March 20, 2020, S&P Global Ratings examined how the economic disruption could affect the dominant subsectors within U.S. residential mortgage-backed securities (RMBS). The nonqualified mortgage (non-QM) RMBS sector may be particularly sensitive to the economic freeze that has gripped the country because it is characterized by (1) lower relative FICO scores, (2) the use of alternative income documentation, (3) self-employed borrowers, and (4) concentration within three states (California, New York, and Florida). As such, we analyzed the underlying loans of 85 non-QM securitizations rated by S&P Global Ratings as of their issuance dates, which ranged from February 2017 to February 2020.

S&P Global Ratings acknowledges a high degree of uncertainty about the rate of spread and peak of the coronavirus outbreak. Some government authorities estimate the pandemic will peak about midyear, and we are using this assumption in assessing the economic and credit implications. We believe the measures adopted to contain COVID-19 have pushed the global economy into recession (see our macroeconomic and credit updates here: www.spglobal.com/ratings). As the situation evolves, we will update our assumptions and estimates accordingly.

COVID-19 And Non-QM's Geographical Footprints Substantially Overlap

California, Florida, and New York have been among the states hardest hit by the pandemic. Approximately 70% of the borrowers whose loans are in non-QM pools also reside in either California (50%), Florida (12%), or New York (8%) (see chart 1). This geographical concentration is unique to non-QM. Credit-risk transfer (CRT) transactions have loan pools that are more dispersed, and jumbo pools lie somewhere in the middle. Given the high concentrations in non-QM pools, we typically assign them higher loss expectations (through pool-level geographic concentration adjustment factors). On average, our non-QM pool geographic adjustment factor is roughly 1.06x and ranges from as low as 1.00x to just over 1.25x. A CRT pool, on the other hand, is so diversified that our geographic concentration adjustment factor is generally neutral at 1.0x.

Chart 1

Examining non-QM pool concentrations by metropolitan statistical area (MSA), we found that roughly two-thirds of non-QM loans are to borrowers in the top-15 MSAs (of which there are 399 in the U.S.), and 12 of these are located in either California, Florida, or New York.

No overall home price bubble in COVID-19 and non-QM local markets

Elevated default levels during the Great Recession were driven not only by adverse economic conditions and the corresponding impact on borrowers' ability to pay, but also by the extreme home price declines in many areas, which influenced borrower incentives and behavior. We considered the degree of over-/under-valuation in the top-three states, as well as the top-15 MSAs, to understand what impact the health of local real estate markets might have on borrower behavior. By our construction, a market is overvalued if the ratio of affordability (defined as the average home price to per capita income) to its long-term average is greater than unity. (For a review of our valuation method, see the appendix of "Will the Froth in U.S. Housing Bubble Over Again? We Think Not," Feb. 22, 2019.)

Relative to long-term norms, New York and California appear to be undervalued at -12% and -4%, respectively, as of the end of 2019. Florida, on the other hand, appears to be somewhat overvalued at +5. Among the top 15 MSAs, the average degree of overvaluation is more or less neutral at 0.7%. There is a reasonably wide dispersion among the MSAs, however, ranging from NY/NJ (-13%) to Houston (+13%) (see chart 2). Unlike during the years running up to the Great Recession, the U.S. market does not currently appear to be in a bubble, although there may be pockets of overvaluation. Moreover, market consensus regarding home price appreciation in 2020 is for negligible growth rather than declines.

Chart 2

Breaking Down Impact By Loan-Level Characteristics And States

Because of the substantial overlap between COVID-19-affected states and regions that represent a large portion of the non-QM sector, we broke out certain loan-level characteristics for California, Florida, and New York by non-QM shelf (see table 1).

Table 1

Top three states vs. national

In addition to geographic exposure, it is important to understand borrower profiles and loan-level characteristics in order to anticipate future performance. The loan attributes that we believe may influence which loans are more susceptible to the economic fallout from the pandemic and the related lockdowns vary among the three states. New York stands out in terms of COVID-19 exposure, likely due to its population density.

Self-employed borrowers may face business cash flow strains arising from mandatory lockdowns. We find that loans to borrowers in California make up 50% of the closing principal balance in the S&P Global Ratings-rated non-QM RMBS universe, with 28% underwritten to self-employed borrowers. California has the largest share of self-employed borrowers at 56%. The state's non-QM loans, however, exhibit FICO scores and loan-to-value (LTV) ratios that are stronger than the respective national averages.

Florida's non-QM loans stand out as having the lowest average FICO score and the highest average LTV ratio. New York and Florida both have greater exposure to debt service coverage ratio (DSCR) loans (i.e., investor property loans underwritten to rental cash flows) than California. New York has the highest FICO and lowest LTV ratio among these three states. At this early stage in the economic downturn, we have seen DSCR loans exhibit about the same credit performance as self-employed borrowers, except in New York where they have performed worse. The higher degree of COVID-19 exposure in New York may be having an impact on the residential rental market to a greater extent. DSCR borrowers, however, tend to have substantial equity in their properties and might therefore be disinclined to default strategically. In addition, they tend to also have cash flow diversification via multiple rental cash flow streams, which could also be a factor supporting DSCR loan performance.

The unique credit characteristics of non-QM shelf exposure

Comparing some of the issuers in the non-QM space, we notice that certain shelves have more exposure to specific loan types in specific areas. These shelves may also have unique credit characteristics. For example, when looking at table 1, Angel Oak stands out compared to other shelves shown in terms of overall exposure to the three featured states at only 49% concentration, with minimal exposure to New York. Compared to its peers, however, Angel Oak borrower characteristics indicate slightly higher LTV ratios. Contrast this with Starwood, which has relatively higher exposure to New York. It also has more DSCR loans, but appears to adjust for this risk given that borrowers tend to have greater equity and higher FICO scores. Arroyo, meanwhile, is heavily concentrated in California, but has relatively lower exposure to self-employed borrowers and DSCR loans, and has a better profile of FICO scores and LTV ratios.

Economic environment of the top three states

As of May 16, 2020, Florida has the highest insured unemployment rate (IUR) at 25.0%, which jumped up recently from a relatively low IUR of 5.2%. One factor that could have been driving the earlier low figures is that Florida has been slower to process unemployment claims applications. A combination of rigorous qualification requirements (residents must recertify unemployment status every two weeks to continue to receive benefits) and electronic filing issues may have contributed to delays and backlogs, which could have influenced the earlier low IUR and its subsequent spike.

California and New York have IUR levels of 20.6% and 19.9%, respectively. These two states provide a standard unemployment duration of 26 weeks as compared to 12 weeks for Florida. Under the CARES Act, all states will be allowed to provide 13 additional weeks of federally funded Pandemic Emergency Unemployment Compensation (PEUC) benefits to those residents who exhaust their regular state benefits. Even with an additional 13 weeks, however, Florida would lag the standard unemployment duration of most other states. This means that Florida loans could be more affected by COVID-19-related income loss/curtailment than those in other states. On a positive note, Florida has begun reopening its economy earlier than both California and New York.

While the economic fallout from COVID-19 has affected all three states, it is important to consider the longevity of their respective lockdowns and subsequent timelines for reopening. For example, many businesses are already reopening in Florida, yet California and New York remain largely shuttered. Looking forward, it is reasonable to assume that, given its population density, New York could lag the other two states in reopening. Because some states are better equipped to handle unemployment surges, we expect a degree of regional variability in the time it takes to resume operations. Moreover, fear of a second wave of COVID-19 may suppress consumer demand, as people avoid markets and entertainment. This, in turn, could hamper the rate of economic recoveries in individual states.

Bringing It All Together

While April remittance data was muted because it reflected March payment information (at which time the effects of COVID-19 were milder), the May remittance reports provide some indication of credit behavior associated with the pandemic. Pre-pandemic 30+ day delinquency prints in February were around 3%, versus roughly 20% as of the May remittance for New York, California, and Florida (see table 2).

Remittance reports also suggest that adverse credit behavior was more pronounced in loans that were made to self-employed borrowers or loans underwritten to a DSCR. Overall, considering S&P Global Ratings-rated transactions across all states, 30+ day delinquent loans to self-employed borrowers and DSCR borrowers were, as of the May remittance, hovering around 20%, compared with 13% for non-self-employed and non-DSCR loans.

Although New York has the highest average FICO score and lowest average LTV ratio, delinquency figures are relatively high for the state, adjusted for the credit quality of the loans, likely due to the concentration of the COVID-19 crisis in the New York City area. The relatively high percentage of DSCR loans in New York is also contributing to some of the higher prints. In that sense, New York seems to be the epicenter of not only COVID-19, but also of non-QM loans most affected by the crisis. That said, we should acknowledge that any potential losses in the future may be less severe for DSCR loans than for other loans due to the relatively lower LTV ratios of DSCR loans.

Florida, on the other hand--even though it has the lowest FICO and highest LTV ratio among these three states--seems to be performing in line with New York and California. The higher 'AAA' loan-level loss coverages also suggest that Florida loans are performing relatively better when using the loss coverages as a benchmark for comparison. When looking at the ratio of 30+ day delinquencies to 'AAA' loan-level loss coverages, Florida, California, and New York are roughly 0.71, 1.25, and 1.28, respectively. In addition, the average loan balance of Florida loans is around $300,000, compared with $500,000+ for California and New York, which could mean less pressure on Florida borrowers due to lower debt service payments and a relatively bigger pick-up from the various government stimuli. Florida, however, is more dependent on tourism and appears slightly overvalued.