Over the past five years, one of the identifying characteristics of the non-qualified mortgage (non-QM) subsector within non-agency residential mortgage-backed securities (RMBS) has been consistently high prepayment rates. Indeed, because of the strong housing market, non-QM prepayment rates have become a more relevant topic than defaults and delinquencies. Nevertheless, non-QM prepayment rates have dampened in the past year or so as the spread between the non-QM rate and the conforming rate tightened. In our view, this tightening was largely attributable to greater supply and competition in the origination market, growing liquidity in the capital markets, and broadly increased confidence that the requirements of the ability to repay (ATR) rule could be met.

Earlier this year, we had expected non-QM prepayment rates to increase as conforming rates ticked up slightly and originators focused more on non-QM loan channels. However, the recent spike in mortgage rates removed the interest rate refinance incentive for a large portion of the non-QM portfolio. This unprecedented spike in mortgage rates has caused conforming and jumbo prepayment rates to migrate closer to levels at which only home sales drive prepayments (the baseline mobility/turnover rate). While non-QM borrowers have a stronger interest rate incentive, it might not be sufficient to sustain the relatively high prepayment rates given the additional loan qualification hurdle that needs to be overcome in the non-QM sector.

Interest rates are broadly expected to increase over the next several years as the Federal Reserve implements its monetary policy tools to combat inflation (see "Economic Outlook U.S. Q2 2022: Spring Chills," March 29, 2022). The 30-year fixed rate for conforming mortgages has already risen over 200 bps since the fall, and this has a direct impact on the non-QM and other non-agency fixed mortgage rates. For this article, we analyzed most of S&P Global Ratings' outstanding rated non-QM transactions to understand the potential impact of the RMBS extension risk that looms.

From The Fast Lane To A Slow Ride

Earlier this year, our view was that non-QM prepayment rates should increase as the focus of origination strategies shifted away from agency refinancing and toward the production of non-QM loans. However, the recent and rapid bump in rates has rendered a large portion of non-QM borrowers out of the money from a refinancing perspective. In other words, the prevailing mortgage rate is higher (or at least not low enough to offset frictional costs) than the rate on typical non-QM loans. There's been a rapid uptick in the conforming 30-year fixed rate and a slow decline in prepayment rates for credit risk transfer (CRT) and jumbo RMBS since refinance burnout (the phenomenon in which repeated incentives to prepay have a diminishing impact on the same pool over time) started setting in (see Chart 1). It also shows how roughly a year ago, speeds of non-QM RMBS prepayments diverged from those of CRT and Prime 2.0. but have recently showed a trend similar to CRT and Prime 2.0 speeds, which continue their downward trajectory.

There is a period of several months between non-QM loan origination and securitization. This leads to a lag in reporting for the average non-QM rate (see Chart 1), unlike the case of current agency mortgage rate readings, which are readily available. Chart 1 also shows that the spread of the non-QM rate to the conforming rate has typically been 150 bps-200 bps. The inertia differential between non-QM and agency loan rates means that it could be several months before the spread reflects the current rate environment.

Chart 1

The non-QM spread above the conforming rate is attributable to various factors, including the aggregate credit quality of the non-QM portfolio as well as the non-QM designation itself (which can affect liquidity, among other things). While this spread has been reasonably consistent over time, credit metrics of loans in the non-QM securitized portfolio have weakened. This suggests that the spread component relating to the non-QM designation could be tightening and offsetting the widening due to credit deterioration (see "Factors Affecting Non-QM Mortgage Interest Rate Spreads," Feb. 20, 2020). Indeed, the non-QM loan market has gained wider acceptance among investors, as many new originators have entered the market. However, the recent rapid rise in rates might lead to consolidation among certain non-QM originators.

Despite the consistent spread between the non-QM loan rate and the conforming rate, approximately 30% of the non-QM loans in the securitized portfolio are now out of the money (assuming an average conforming rate of 5%). That is, the rate on the non-QM mortgages for this subset is lower than the current conforming rate, which has increased dramatically over the past six months. Furthermore, for the loans in the portfolio that have a rate incentive of 100 bps or less (roughly two thirds of loans in the portfolio that have interest rates of 6% or less), frictional costs could deter borrowers from prepaying, which means loans with interest rates between 5% and 6% might also be effectively out of the money. In the aggregate, the weighted-average coupon (WAC) of the total portfolio is just over 6%, and the interest rate incentive (to refinance into a conforming loan) is just over 1% (see Chart 2).

Chart 2

Which Types Of Loans Prepay The Fastest?

While non-QM prepayment rates might be slowing overall, the prepayment tendencies of particular types of loans within that RMBS subspace have remained the same. There continue to be faster prepayment rates for loans with full income documentation, while loans with alternative income documentation (such as bank/P&L statement loans) remain slower. The slowest are the property-focused DSCR loans, which are underwritten to the expected cash flows of the mortgaged investment property. Part of the reluctance to prepay DSCR loans derives from prepayment penalties, which are typically in place for three to five years.

Despite this variability, we believe all non-QM subsectors will experience declines in prepayment rates as the conforming mortgage rate renders roughly a third of non-QM mortgages out of the money, with another third effectively out of the money (again, assuming a hurdle of 100 bps). Moreover, it's possible that recently paid high non-QM closing costs could act as a psychological deterrent from refinancing in the near term, even though some of these loans might be in the money.

Non-QM prepayment rates have always varied by subsector (see "Non-Qualified Mortgage Prepayments: Is It Life In The Fast Lane, Or Will They Start To Take It Easy?" Sept. 12, 2018). Table 1 breaks out total loan balances and average mortgage interest rates, loan-to-value (LTV) ratios, and FICO scores of our non-QM portfolio by full documentation loans, alternative documentation loans, and DSCR (including no-ratio) loans. Because of the variation in prepayment rates, there is a survival bias that reinforces the speed ordering described above: full documentation is the fastest, alternative documentation loans are slower, and DSCR loans prepay the slowest.

Table 1

The statistics show that the set of factors influencing prepayment rates within the non-QM sector still holds. However, certain buckets--which we formally broke out as stand-alone cohorts, such as loans to borrowers with prior credit events or debt-to-income (DTI) ratios over 43%--could be losing relevance as the foreclosure crisis fades into the past and specific DTI ratio requirements have been removed in the revised QM rule. We could also see a convergence in the prepayment rates of DSCR loans and other loan types in the short to medium term, prepayment penalties notwithstanding. We base this on the following factors:

• DSCR loans have typically carried higher interest rates than non-QM products; however, greater popularity (and therefore competition causing rate compression) in DSCR loan origination could create more in the money refinance options for borrowers within the DSCR lending market.
• DSCR borrowers are more likely to cash-out equity positions and redeploy the funds to acquire additional properties for their portfolios. DSCR loans within the non-QM sector are roughly 49% cash-out, whereas non-QM overall is 28%.

Again, the resulting impact on prepayment rates will become more apparent in the coming months because of the typical lag between loan behavior and securitization accounting/reporting. Also, rates at which non-QM borrowers ultimately executed were locked in several months ago, and the current data do not yet reflect the prevailing rates.

The Crossroads Of WAC And Excess Interest

Even though prepayment rates have fallen broadly within the non-QM RMBS subspace since June 2021, the one-month CPRs on average have been in the 20+ range, and the instances of WAC deterioration within pools have been less pronounced than those of WAC appreciation. At first glance, this might appear counterintuitive because it means loans with lower mortgage interest rates are prepaying more rapidly than those with higher rates. However, the reason is likely that loans with higher rates often have credit/qualification impairments that impede the ability of these borrowers to refinance into an agency loan or a more competitive (lower mortgage interest rate) non-QM loan. Moreover, loans with lower mortgage interest rates could have better refinancing opportunities, even though their interest rate incentive is lower. This is evident in Table 1, where the weighted average mortgage interest rates of the surviving loans are higher than those of the paid-off loans for all loan types. Chart 3 shows the change in pool WAC (current WAC as of March 31, 2022, minus WAC at securitization closing) of non-QM transactions in our population since January 2019. The chart shows that the majority of transactions have had an increase in WAC (WAC appreciation as opposed to depreciation).

Chart 3

What's The Credit Impact?

The lack of WAC deterioration within portfolios supports the generation of excess interest. In turn, this limits the extent to which bonds are written down, as excess interest can be used to cover credit losses. While this has benefited non-QM securitizations over the years--especially in the wake of the COVID-19 pandemic, which resulted in forbearance/deferral related losses--the more recent issuances have experienced a marked reduction in excess interest, with several transactions having no excess interest at the time of securitization. This has been a direct result of non-QM loans originated in the recent ultra-low rate environment and entering securitizations for which spreads have widened in the current economic environment. As a result, differences between hard subordination credit enhancement and the credit losses projected from the asset pools experienced material widening over the past several months compared to prior periods.

Table 2 shows two transactions, issued almost a year apart, with comparable asset loss projections. The subordination credit enhancement necessary to support the tranche rating in the more recent transaction is higher than the corresponding loss coverage at all rating levels. However, the transaction issued earlier has less credit enhancement than the corresponding loss coverage at certain rating levels because the transaction is supported by the availability of excess interest.

Table 2

In For The Long Haul? Or Can Calls Mitigate Extension Risk?

Securitization spreads have widened because of various macroeconomic factors, including rising interest rates. Non-QM asset pricing at origination does not respond to market conditions with the same swiftness as the (larger and more liquid) conforming loan segment. Therefore, non-QM securitizations will likely experience several months of significant excess spread compression because the loans now being securitized might have been originated when the conforming rates were still close to 3%. The prepayment rates of these securitizations are in for a substantial slowdown. Prepayment rates of securitizations issued prior to the recent rate spike can also be expected to slow given the lack of rate incentive in those transactions. Over time, though, prepayment rates of future transactions should normalize as the mortgage interest rates of newly originated non-QM loans adjust upwards and reflect the historical spread between conforming loans and non-QM loans. However, much depends on the future path of market interest rates.

Recent securitizations have used several features to manage bond extension risk, including step-up coupons, reallocation of cashflows to the more senior tranches from the more subordinate classes via step-down coupons on lower tranches, and full sequential structures. Table 3 contemplates a range of prepayment rate scenarios for a sample transaction with a modified/sequential structure under one of our 'B' rating level stress scenarios. If non-QM prepayment rates were to fall to 10% (roughly double the historical mobility/turnover rate), then extension would be inevitable, absent a call. The optional call provisions generally remain at the earlier of three years from closing and when the pool factor is less than 30%. Whether the call would be exercised, thus mitigating extension risk, would depend on various factors such as the path of interest rates, bond pricing, whether the XS tranches are receiving money, and other sponsor-specific considerations.

Table 3