article

Anticipating Regime Changes in the Agency MBS Market

Regimes change over time, including in the mortgage-backed security (MBS) space; an unexpected regime change can be an unwanted surprise and, at worst, play havoc with a portfolio or investment strategy, even ones that previously worked well.

Accurate anticipation of a regime change not only facilitates timely defensive re-balancing in advance of the change, but also provides an opportunity to establish new profitable positions before their value has been fully reflected in market prices. Risk and opportunity are opposite sides of the same coin.

Loan Balance Regimes

One current example of a regime is loan-balance (LB) stories, cohorts of MBS pools reflecting certain (generally) lower loan balances on the underlying mortgages. With some similarity to credit stories, the lowest loan balance (LB) cohort, LB085, often prepays more slowly than higher balance cohorts. These slower conditional prepayment rates (CPRs) can result from the associated smaller refinancing resources and (absolute) incentives that lower loan balance borrowers are thought to possess, as described in more detail below.

“Often” is not “always,” as the charts below indicate. In Figure 1, we began with all of Infima's LB085 and LB200 coupon-vintage story cohorts. Then, to compare apples with apples, we eliminated any coupon-vintage combination that had been in existence significantly longer in one loan balance story than the other. We then took the arithmetic average of the constant prepayment rates (CPRs) of the remaining coupon-vintage cohorts in each of LB085 and LB200 for every monthly factor date starting in 2019.

Figure 1: Average Monthly CPR by LB Story

Figure 1 shows how at the beginning of 2019 the two stories prepaid at nearly the same speed but diverged as interest rates fell, with LB200 increasing its speed significantly relative to LB085. The CPR differential (the “spread”) between them widened through the pandemic, then narrowed into 2022 as the Federal Reserve hiked rates to combat inflation, with LB085 prepaying as fast or faster than LB200 towards the end of the observation period. One conventional narrative, based on the assumed relative financial resources of the two stories, is that the lesser financial resources of lower loan balance borrowers prevented them from taking full advantage of dropping mortgage rates, while compelling them to refinance or liquidate when rates rose.

Prepayment Speed Distribution

Reviewing coupon-vintage cohort prepayment speeds within these two loan balance stories confirms that the speed differential was distributed widely among the constituent cohorts. Figure 2 displays, for three calendar quarters, from the beginning of 2019 and 2021 and the end of 2022, respectively, the distribution of realized prepayment speeds within the approximately 25 coupon-vintage cohorts in each of the two loan balance stories, a total of nearly 150 observations per quarter for both stories combined. (The small vertical lines resembling a rug on the x-axis show the constituent cohort observations.)

The charts in Figure 2 track the averages in Figure 1 nicely, showing how:

  • In the center chart at the height of the pandemic in 2021Q1:

    • the LB200 constituent coupon-vintage cohort speeds extended to 50%, perhaps as a result of attractive refinancing opportunities at lower mortgage rates and increased interest in updating homes in which more time was spent, while

    • in the same period, the LB085 constituent CPRs clustered around 20% and lower, as for these smaller borrowers refinancing was both harder to find and of lower absolute benefit, and

  • in the final chart during the Federal Reserve rate hikes:

    • LB200 reverted back to the lower CPR speed range of the first, pre-pandemic chart while

    • LB085 reverted less strongly, ending somewhat above the LB200 distribution.

Figure 2: Regime Change over Three Quarters from Three Years

Profiting from Regime Change

To profit from regime change requires accurately predicting both the change’s timing and amount. Fortunately, Infima's prepayment predictions did precisely that. In Figure 3 below, we superimpose Infima's one-month ahead CPR predictions on the realized CPRs shown in Figure 2. Although Infima's CPR predictions were (of course) not perfect, our predictions accurately captured the transition of LB085's prepayment speeds compared to LB200 over the three selected quarters:

Figure 3: Regime Change over Three Quarters with Infima Predictions

Armed with these predictions, an Infima client could adjust their portfolio proactively to weight these LB stories appropriately through the pandemic and the subsequent inflationary periods. 

A more detailed quarter-by-quarter graph panel confirming this analysis appears in Figure 4 below. 

Figure 4: Sixteen Consecutive Quarterly Loan-Balance Cohort Predicted and Realized CPRs

In summary, the spread between the CPRs of different loan balance stories changed dramatically from 2019 to 2022, creating both risk and opportunity. Infima’s projections predicted the change, helping clients avoid the former and capitalize on the latter.

 --

Interested in learning more? Find Infima's research in loan balance and other MBS stories appears in the white paper “The Fast Path to the Best MBS.”

Nicholas Gunther

Chief Solutions Officer

Nicholas Gunther is infima’s Chief of Product. He has a background in advanced mathematical methods and financial markets, services, regulation and securitization. Nick has worked on the development, structuring and execution of financial products at AIG, Goldman Sachs, and other Wall Street firms. He founded registered broker-dealer GH Group and became its CEO until selling it. Nick is also affiliated with UC Berkeley’s Consortium for Data Analytics in Risk.

Nick received a PhD in Mathematics (1982) and JD (1986), both from Harvard University.