- An extremely volatile 2020, with historically high prepayment rates and price premium in mortgage-backed securities (MBS), posed unprecedented challenges to MBS investors.
- Both accurate historical prepayment error tracking and model duration consistent with forward market-price dynamics are typically essential for proper risk management.
- As what appears to be a new COVID-19 regime for MBS continued into 2021, prepayment uncertainty due to refinance burnout remained a key risk factor.
The Federal Reserve’s aggressive quantitative-easing (QE) program increased the Fed’s holdings in mortgage-backed securities (MBS) by about USD 700 billion during a volatile 2020. The historically high MBS price premium and low mortgage interest rates, compounded by a wave of mortgage-underwriting policy changes and technology advancements, led to a historic refinance frenzy and posed an unprecedented challenge for MBS risk management. A comprehensive review of MSCI’s agency-MBS model for 2020 may help investors understand the market-price dynamics and assess the adequacy of the model-based risk measures.1
MBS Yield’s Realized Volatility Spiked in the COVID-19 Crisis
As the effects of COVID-19 started to hit the U.S. last spring, the 10-year Treasury’s realized volatility rose threefold, while the MBS current coupon yield jumped sixfold. Decisive actions by the Fed, together with government fiscal-stimulus packages, eventually calmed the market. Noticeably, MBS yield volatility remained significantly higher than that for U.S. Treasurys.
Historically High Prepayment in 2020
2020 emerged as a historic year for mortgage-origination volume, as mortgage rates hit multiple record lows.2 The following exhibit shows MSCI’s prepayment model for agency MBS, which closely tracked the realized prepayment speeds for major home-loan cohorts.
Model Closely Tracked 2020’s Month-by-Month Prepayment Surge
For each Fannie Mae 30-year cohort (coupon/vintage), error tracking for 12 monthly data points for 2020. Source: Fannie Mae, Recursion, MSCI
Using rank-based error tracking,3 we found the MSCI model captured the differentiation across an array of collateral attributes — i.e., a steep ranking curve with an average absolute error of 0.25 single monthly mortality — during 2020. The MSCI model continued to perform in line with the actual prepayment throughout the year, as a wave of mortgage-underwriting policy changes and technology enhancements helped refinance activity reach historical highs.
Rank-Based Error Tracking Showed the Model Differentiated Prepayments
Source: Fannie Mae, MSCI, Recursion
How MBS Prepayment Regimes Shifted in the Past Two Decades
We can use the empirical S-curve — which shows realized prepayments vs. economic incentive to refinance — to place 2020 prepayments in historical context. 2000 and 2001 are generally considered benchmark years for prepayment efficiency. Following the 2008 global financial crisis, 2009 was marked by very tight mortgage-credit availability, with a depressed S-curve. As the housing market and economy recovered, however, refinance efficiency improved, as we also saw in 2016 and 2019. 2020’s S-curve for the most creditworthy borrowers most resembled that of 2003, a year that stands out for extremely loose underwriting standards. Mortgage-underwriting digitization had steadily advanced by 2020, but the COVID-19 pandemic drastically accelerated adoption of streamlined technology for loan underwriting by the agencies. Meanwhile, various temporary underwriting flexibilities effectively loosened underwriting standards.4
The 2020 Prepayment S-Curve in Historical Context
Incentive is defined as weighted-average gross coupon of all Fannie Mae MBS pools minus the prevailing mortgage rate. Source: Fannie Mae, Recursion, MSCI
Although the refinance wave continued into 2021, how soon could it crest and break? Time will tell. Technological advancements, which are a factor for sustained fast prepayment, tend to continue moving forward, not backward. On the other hand, the U.S. mortgage agencies have expressed concern about underwriting standards5 and that the temporary flexibilities are indeed temporary, which are both factors for slower prepayment. Meanwhile, macroeconomically, the 10-year Treasury sold off 50 basis points (bps) since last August, while MBS yield essentially remained flat. A large MBS sell-off, following Treasurys, could significantly curtail the current media effect, which could slow the prepayment speed drastically.6
Market-Price Dynamics Provide a Reality Check on Model Durations
The prepayment forecast is far from being deterministic, but market-price dynamics provide a range of information we can use to apply a reality check on model durations. MSCI model duration, in general, moved in tandem with empirical duration throughout 2020. The deviation between modeled and empirical durations was mostly driven by two factors:
- Refinance fears tended to drive empirical duration shorter, such as in early 2020.
- QE tended to drive empirical duration longer after March 2020, when the Fed instituted its latest round of stimulus.
After months of wobbly MBS price discovery, the empirical duration steadied recently, converging with the MSCI model’s fundamental-driven duration. This implies the MSCI model’s forward prepayment-burnout assumption is in line with current market expectations. By contrast, a model that assumed an aggressively persistent refinance wave may lead to negative durations under the historically high MBS price-premium environment.
MSCI Model and Empirical Duration Moved Largely in Tandem Through 2020
In summary, 2020’s historic prepayment rates and price premium, amid extreme volatility, posed unprecedented challenges for MBS investors. As the MBS regime ushered in by COVID-19 continued into 2021, MBS investors may want to monitor their models’ performance in multiple ways, to navigate the remainder of the pandemic.
1Yu, Y. 2020 “MSCI Agency MBS Model Performance Review 2020.” MSCI Model Insight.
2Eisen, B. “Mortgage Originations Are on Pace for Best Year Ever.” Wall Street Journal, Dec. 10, 2020.
3Zhang, D. 2020. “Rank-Based Error Tracking for Agency MBS Prepayment Models.” Journal of Fixed Income.
4“Lender Letter (LL-2020-03).” Fannie Mae, Dec. 10, 2020.
5“Request for Information on Appraisal-Related Policies, Practices, and Processes.” Federal Housing Finance Agency, Dec. 28, 2020.
6Abundant media news of historically or recently low mortgage rates may raise more awareness of refinance opportunities for borrowers, hence leading to a nonlinear effect on prepayment. See: Yu, Y. 2018. “MSCI Agency Fixed Rate Refinance Prepayment Model.” MSCI Model Insight.