COVID-19 and the Mortgage Market

For many of us, the scale of the coronavirus crisis is reminiscent of 9/11 or the 2008 financial crisis—the events may reshape people’s behavior in a long-lasting way. As the event continues to unfold, we’ve found ourselves in uncharted territory. In this article, we will address some policy changes related to the U.S. mortgage market and our model changes as well as a few tuning parameters that can be used to adjust our U.S. mortgage prepayment model to reflect the potential impact of COVID-19.

Mortgage Market Update

The coronavirus pandemic has put tremendous pressure on the U.S. economy. To stop people from spreading the virus, social distancing was introduced in early March. Many states have also implemented shelter-in-place or stay-at-home orders. The lockdowns have led to the collapse of the U.S. economy. Over the last eight weeks, a total of 36.5 million Americans have filed for unemployment insurance. According to the Congressional Budget Office, the unemployment rate is expected to exceed 10% during the second quarter, as high as the peak of the last recession. GDP is expected to decline by more than 7% for the second quarter. These stunning figures may not even reflect the complete impact on the economy, as many people who lost jobs do not qualify for unemployment benefits. The long-term impact will take months to unfold and is still uncertain at this point. 

With a record increase in jobless claims, mortgage delinquencies and missing rental payments are likely to rise. The impact on the rental market may be more significant because hourly workers tend to be laid off first and are more likely to be renters than homeowners. The first hard data related to the rental market that we have seen comes from the rent payment tracker published by the National Multifamily Housing Council (NMHC) on April 8. The tracker found that only 69% of renters had paid their rent through April 5, a 12% decrease compared to the previous month and the same time last year.

In anticipation of financial hardship related to COVID-19, the U.S. Department for Housing and Urban Development (HUD) and Federal Housing Finance Agency (FHFA) have rolled out several policies to help mortgage borrowers. Foreclosure and evictions are suspended for at least 60 days. Hardship forbearance for both single-family and multifamily borrowers is recommended for servicers and forbearance plans provide borrowers with payment relief for up to 12 months. Late charges and penalties are suspended as well. Forbearance is also eligible for a second home or an investment property. According to the Mortgage Bankers Association (MBA) Forbearance and Call Volume Survey released on April 7, the total number of loans in forbearance has jumped from 0.25% to 2.66% in the one month between March 2 and April 1. Ginnie Mae loans have seen the biggest growth, jumping from 0.19% to 4.25%.

The surge in delayed payments will not have an immediate impact on investors because mortgage servicers must advance scheduled payments to investors even if end borrowers fail to make their payments. The MBA estimated the surge in delayed mortgage payments will cost the mortgage servicing industry as much as $100 billion over the next nine months. Stronger servicers will not need much assistance, while many smaller servicers will require some level of support from the government given the scale of this forbearance program at the duration required. The mortgage industry and housing associations have called on the federal government to establish a liquidity facility for servicers as Ginnie Mae modified an existing program to provide short-term liquidity support for some of its servicers.

With an increased number of loans in forbearance, the growth of delinquent loans could be slower than the initial estimate. However, we still expect the early delinquency rate to rise given the magnitude of reported jobless claims in recent weeks. It will take a few months for early delinquent loans to transition into serious delinquent loans and many of them will qualify for payment forbearance. Payment forbearance won’t trigger a buyout event and therefore, buyout-related prepayment speed won’t rise significantly in the near term. The delinquency trigger of the credit risk transfer (CRT) deals, usually based on the level of seriously delinquent loans, will not be hit right away either. The newer transactions, which have lower credit enhancements, will more likely fail the delinquency test earlier than the seasoned transactions. Some forbearance loans may lead to further loan modification or a higher default rate later. The losses will then be passed to CRT investors in reverse sequential order.

Housing market activity dropped significantly after social distancing measures took effect. As of March 27, the purchase index decreased 24% versus the same week one year ago. Based on data from ShowingTime, showing activity through mid-March was higher year-over-year. Since then, showings are down roughly 60% versus the same time last year. The turnover rates are expected to drop to reflect a weak spring home-buying season.

Treasuries rallied at the beginning of March and so did mortgage rates. In the following two weeks, mortgage rates rose as the primary and secondary spreads widened despite the treasury rate remaining low. The erratic mortgage rate movement in March was probably the result of liquidity concerns related to the MBS market since lenders usually securitize mortgage loans to hedge risk. In response, the Federal Reserve has rolled out unlimited quantitative easing to inject liquidity into the market. The mortgage rate may start to drift lower as liquidity concern is lifted. In addition, FHFA has granted flexibilities for appraisal and employment verifications as well as loan processing. However, lenders’ pipelines will remain clogged and processing time will be prolonged. The overall rate incentive for the borrowers will remain high, but the s-curve will flatten due to extended processing time.

LIBOR Market Model Update

In response to historically low rates in the fixed income space, option-adjusted analytics (e.g., OAS, Coupon Curve Duration, Coupon Curve Convexity, Spread Duration) calculated for securities in the U.S. securitized market that currently uses Monte Carlo model (MC Model)/LIBOR market model (LMM) will undergo the following changes.

1. Switch caplet volatilities used in calibration for MC Model from using lognormal volatilities to normal volatilities. In a low or negative interest rate environment, lognormal volatilities start to break down due to the assumption that the underlying rates are lognormally distributed (and therefore cannot be negative). On the other hand, rates in the normal model can be both infinitely positive and negative and therefore, we will use normal volatilities for caplets moving forward (swaptions are already using normal volatilities) for MC Model calibration.
2. Increase lognormal shift in Shifted Lognormal LMM from 200 bps to 400 bps. The current form of instantaneous volatility used in LMM is shifted lognormal with a 200 bps shift. The current choice of shift makes the rate floor at -200 bps. Give the current low interest rate environment, we have determined that this rate floor is too high, especially when a minus 200 bps shock is applied in a scenario analysis. Thus, we decided to increase the shift to 400 bps.
3. Add an alternative MC Model with a normal instantaneous volatility assumption. The normal form of instantaneous volatility used in LMM does not impose a rate floor at any level. Thus, it is free from any changes in interest rate levels. This model can be accessed via a change in Fixed Income Settings (@FS) and will be released later in May.

A good way to compare different LMM set-ups is to look at how good the calibration is compared with market data. The metric we use internally is the residual sum of squares (RSS) of calibrated instrument volatilities compared to market input volatilities. The figure below shows that during March normal LLM became the best setup because of the lowest RSS. On the other hand, lognormal LMM with a 400 bps shift improved marginally over lognormal LMM with a 200 bps shift.

FactSet U.S. Prepayment Model Knobs and Dials

Given the uniqueness of this event and the rapid changes of various policies, we are recommending that clients use model dials to adjust our U.S. prepayment model to reflect the potential impact. Ramping or fading multipliers allow users to create a ramp of prepayment speed and adjust down for the coming months and let it ramp back to normal over a few months. For example, as illustrated in the next figure, a 30 percent multiplier can be applied to the prepayment speed and then a slow ramp back to 100 percent in six months.

Another available feature is the factor multipliers, which adjust individual components of the prepayment model. The refi twist multiplier is a new tuning knob, which allows the user to steepen or flatten the s-curve. The turnover multiplier can be used to lower the overall turnover rate to reflect the current market condition.

Conclusion

There remains much uncertainty in the mortgage market related to the impact of the coronavirus pandemic. The CARES Act and various policy changes will provide temporary relief for mortgage borrowers but borrowers’ financial hardships may last longer if the job market recovers slowly, which may lead to a higher default rate later. The prepayment speed of the U.S. mortgage pool may slow down due to the weaker housing market as a result of social distancing and clogged lenders’ pipeline. The payment forbearance plan at the current scale will put many smaller servicers under severe financial constraints without the support from the federal government; that presents a potential risk to the mortgage market.

In response to the low rates, we have updated our LIBOR Market Model with normal caplet volatility and increased the shift to 400. An alternative Monte Carlo Model with normal instantaneous volatility assumption will be released soon. For the U.S. agency prepayment model, we recommend several tuning options to adjust prepayment speed slower and will continue to introduce features that give users greater ability to tune the model.

Andy Yang, Senior Financial Engineer, Fixed Income and Derivatives, also contributed to this article.