Mortgage Derivatives Case Study: State Housing Finance Authorities

The mortgage derivatives market is one of the only fixed income markets that allows for a symmetrical risk profile. Unlike traditional bonds, the upside is not limited to return of principal plus a set amount of interest. Instead, an investor can actually receive more cash flow than expected for longer than expected.

It is also one of the few financial markets left where deep fundamental analysis still has the potential to consistently provide an information advantage for investors.

The reason for this is simple: Mortgage derivatives already carry a complexity premium, and prepayment models are widely available and accurate enough that many players in the market don’t feel the need to really dig into what’s going on in the securities’ underlying mortgage pools in order to make attractive returns. 

But off-the-shelf prepayment models have two main weaknesses:

1. The future may not resemble the past: The models’ forecasts are only as good as what they can learn from historical prepayment speeds, which means that they cannot factor in the impact of upcoming changes in state or federal government housing policy, the housing market, or mortgage origination businesses. 

2. Smaller market segments can slip through the cracks: These models are used by investors in the multitrillion-dollar agency mortgage-backed security (MBS) market, where the bulk of mortgage pools don’t vary much. This causes forecasts to be biased towards a broad national average, which may overlook what makes a given mortgage pool unique, such as factors affecting specific geographies, specific lenders, or the material impacts of smaller agencies’ differing rules on borrower behavior. 

These weaknesses don’t meaningfully impede the models’ utility to the average participant in the mortgage derivatives market. However, they do create outperformance opportunities for investors willing to do deeper fundamental analysis. 

And, given that prepayment speeds coming in 1% lower than forecast can increase yield by over 1% for the life of the security, that fundamental analysis can be very rewarding. 

This case study represents one recent opportunity exploited by Voya’s mortgage derivatives team.

When most people think about qualifying residential mortgages, they think of the three biggest guarantee programs: Fannie Mae, Freddie Mac, and the Federal Housing Administration (whose guaranteed loans flow through Ginnie Mae mortgage pools). However, there are actually dozens—if not hundreds—of other mortgage guarantee and assistance programs. 

On the federal level, the Veterans Administration and the Rural Housing Service (part of the U.S. Department of Agriculture) both have programs, among others. Beyond that, almost every state has its own housing finance authority (HFA) that offers some combination of low-rate, low-downpayment loans and downpayment assistance programs for first-time homebuyers. 

The rules for all these programs differ. You may get a subsidized mortgage rate and/or lower downpayment options; you may get the state to loan you money towards your downpayment and closing costs. 

Because there are so many programs, and they are so varied, it takes considerable extra effort and time to model them accurately. And because these programs are also tiny compared with Fannie, Freddie, and Ginnie, it doesn’t make sense for the average investor, or for prepayment models, to put the work into doing that. 

However, one generalization is possible: Participants in these assistance programs for low-income homebuyers tend to have very slow prepayment speeds. 

Their loan sizes tend to be smaller, and their credit profile tends to be less than perfect, so even in a falling-rate environment, they have less incentive and less opportunity to refinance (which would result in a prepayment). Also, refinancing costs like appraisal fees and closing costs, which can run $10,000-15,000, represent a significant deterrent to many lower-income first-time homebuyers. 

Additionally, these programs may come with structural disincentives to prepay, such as downpayment assistance loans that are forgiven if the borrower stays in the home for 3-5 years. 

Mortgage derivatives based on pools of borrowers in these smaller programs are often good investments that consistently come in with prepayments slower than the models would predict.

Except after Covid hit.

In the direct aftermath of the Covid lockdown, two unusual things happened simultaneously to these state housing finance authority mortgage-backed securities: Record delinquency levels combined with record home price gains. In response, prepayment speeds rocketed upwards (Exhibit 1). 

Delinquency was driven by many of these lower-income borrowers having more fragile economic situations. This has always been the case, but in the aftermath of Covid it was exacerbated by the economic impact of lockdowns and nationwide special forbearance programs. If these borrowers experienced Covid-related income disruptions, they frequently had minimal savings cushions, and that caused them to take advantage of government forbearance programs.

If a borrower misses four payments, the agency guarantee kicks in and the mortgage may be bought out of its pool, which acts as a prepayment. Even though the borrower had forbearance and wasn’t getting foreclosed on, their mortgage—and its stream of interest income—no longer flows to investors.

At the same time, home prices were skyrocketing. Many of these borrowers put very little money down on their homes, because they received a subsidy from the state housing finance authority. As a result, these borrowers experienced an outsized effect from home price appreciation. 

Borrowers who previously had 0-5% equity in their home saw that number leap to 25% equity at the same time as mortgage rates decreased and their income situation became more precarious. So people refinanced in record numbers to tap into that equity. 

At the time, that wasn’t great for anyone holding securities containing these loans, because the collateral was prepaying very quickly. The conditional prepayment rate (CPR) of MBS with 6% coupons from state HFAs was, on average, double that of non-HFA 6% MBS in 2020 and 2022.

The pandemic—and its effects on mortgages—was temporary. However, the standard prepayment model can’t say “these conditions were temporary, we’re going to ignore those high 2021-2022 prepayment speeds when forecasting future ones.” It also isn’t well suited to distinguishing situations where a bunch of different kinds of collateral start prepaying quickly at the same time, but for different reasons. 

The model just averages it all in. That set the stage for this collateral to be underappreciated once these temporary situations resolved.

Fast forward to 2024. House prices aren’t going up rapidly anymore. Anyone with a pre-2022 mortgage is holding onto it for dear life because, though mortgage rates have come down, they are still significantly above 2015-2021 levels. The Covid-era forbearance programs have ended. 

But all the structural factors that created durable prepayment protection with state housing finance authority borrowers—significant subsidies (which may need to be repaid in the event of a prepayment), lower income, dented credit, high loan-to-value on the property, and so forth— they still apply. 

If this sector was neglected before due to its marginal status, imagine the tumbleweeds after 2021-2022. Nobody wanted to spend time going through 50 different states with multiple programs, because they just got burned on prepayments and the collateral looked really dangerous. 

And because of the late-2022 surge in mortgage rates, new data on borrowers’ refinancing behavior was suddenly very hard to come by. HFA mortgages kept their bad reputation for years after the Covid episode was over, because there was no new refinancing wave to help models confirm and quantify what had changed. 

Between prepayment models structurally getting their prepayment speeds wrong and buyers avoiding state housing finance authority mortgage derivatives, these securities were really, really cheap by 2023. 

This was when Voya started buying them—but not all of them. 

There was a lot of legwork necessary to analyze the differences in programs, with some providing higher structural barriers to prepayment than others. We worked to understand how the nuances of each program, combined with borrower characteristics, could ultimately impact prepayment behavior.

Early on in this case study we noted that prepayment speeds coming in 1% slower than forecast can increase yield by over 1% for the life of the security. When refinancing incentives finally began to reappear, the difference in borrower behavior from the 2021 environment was stark. But because prepayment models were still reflective of Covid’s impact on these HFA mortgages, speeds were still forecast to be quite high. 

In reality, HFAs—the same subsector that prepaid more than twice as fast as the broad cohort in the last refinancing wave—prepaid roughly half as fast as the broad cohort instead (Exhibit 2). 

Because of the shifts in the mortgage landscape, this unloved sector had become one of the best sources of prepayment protection—and thus outperformance—in agency mortgages.

The U.S. agency residential mortgage-backed security market is massive, and it’s filled with constantly changing variables that affect prepayments in subsectors of collateral. Voya’s mortgage derivatives team conducts extensive fundamental research into these variables and collateral subsectors to formulate more-accurate-than-market prepayment views, in order to achieve attractive risk-adjusted returns. 

This state housing finance authority trade is representative of the team’s research-first approach, and it is one of many such trades the team has on at any given time. 

The advantage of fundamental research in the mortgage derivatives market is not one we consider to be a finite exploit that will someday be closed by better models. There is simply a limit to how much predictive power the raw historical data has, even in the hands of an expert modeler. 

Backward-looking data cannot predict the next change in mortgage policy or servicing technology, and models cannot continuously add variables and complexity without running into problems like overfitting. The mortgage lending industry is constantly in flux, and an approach reliant on fitting historical data will never be a perfect guide to the future. A deeper understanding of the drivers behind the data is invaluable. 

Voya’s mortgage derivatives team is a multibillion-dollar institutional investor in mortgage-related assets, with decades of specialization in prepayment analysis. We would welcome the opportunity to discuss how this strategy complements investors’ portfolios.

A note about risk: The principal risks are generally those attributable to bond investing. Holdings are subject to market, issuer, credit, prepayment, extension, and other risks, and their values may fluctuate. Market risk is the risk that securities may decline in value due to factors affecting the securities markets or particular industries. Issuer risk is the risk that the value of a security may decline for reasons specific to the issuer, such as changes in its financial condition. The strategy invests in mortgage-related securities, which can be paid off early if the borrowers on the underlying mortgages pay off their mortgages sooner than scheduled. If interest rates are falling, the strategy will be forced to reinvest this money at lower yields. Conversely, if interest rates are rising, the expected principal payments will slow, thereby locking in the coupon rate below market levels and extending the security’s life and duration while reducing its market value.