Video: Mortgage Market Evolution

As any mortgage market veteran will attest, the distribution and structure of the mortgage market is constantly in flux. When rates fall, at-the-money coupons become premiums, staffing at originators rises, the volume of refis increase, and the distribution and seasoning of coupons change.

And then the cycle turns. Rates rise. Premiums become discounts. Originators cut staff and prepay speeds plummet. But this too changes, and longtime participants will recognize echoes of 1994 or 1999-2000 in today’s washout.

The brief video animation below tracks the evolution of the mortgage market since 2006, with an eye on distribution and seasoning of borrowers.

MSR & Loan Trading Insights

RiskSpan’s Edge Platform is the leading comprehensive data and mortgage analytics platform tailored for residential whole loan and MSR trading, empowering investors with advanced cloud technology and AI. By streamlining loan and MSR data management, providing customizable historical performance insights, and powering robust valuation and risk analysis, Riskspan’s Edge Platform automates complex data tasks and identifies critical, loan-level insights.

Looking for an edge? Read our latest whole loan trading and MSR-related insights below.

What do 2023 Originations Means for MSRs?

Are you investing in MSRs or considering doing so in the near future? If so, understanding current origination trends and loan characteristics is a critical component of predicting future MSR performance and prepayment risk. Read our latest research post, which looks into key characteristics of 2023 originations.

Why Accurate Loan Pool and MSR Cost Forecasting Requires Loan-by-Loan Analytics

Loan cohorting has been a useful strategy to limit the computational power necessary to run simulations. But advances in cloud compute and increasing heterogeneity of loan and MSR portfolios means better methods are now available.

5 foundational steps for investors to move towards loan-level analyses

It’s critical to leverage your full spectrum of data and run analyses at the loan level rather than cohorting. But what does it take to make the switch to loan-level analytics? Our team has put together a short set of recommendations and considerations for how to tackle an otherwise daunting project.

It’s time to move to DaaS — Why it matters for loan and MSR investors

The ability to analyze loan-level granular data is fast becoming the difference between profitable trades and near misses… but operating at the loan level means wading through an ocean of data. Learn about how you can get the most out of your data.

Case Study: How one investor moved to loan level analysis while reducing their costs

Are you looking to optimize investment decisions while reducing costs? Discover how one loan and MSR investor transformed their analytics using RiskSpan, in our latest case study.

Whitepaper: Improving the precision of MSR pricing using loan-level analytics

Incorporating both credit and prepayment modeling into an MSR valuation regime requires a loan-by-loan approach to capture the necessary level of granularity, but performing such an analysis has been historically viewed as impractical. Read RiskSpan’s deep-dive whitepaper to explore how today’s cloud-based, loan-level technology can make this not only practical, but cost effective.

Interested In learning about RiskSpan’s Edge Platform?

SCHEDULE A DEMO TODAY

Agency Social Indices & Prepay Speeds

Do borrowers in “socially rich” pools respond to refinance incentives differently than other borrowers?

The decision by Fannie and Freddie to release social index disclosure data in November 2022 makes it possible for investors to direct their capital in support of first-time homebuyers, historically underserved borrowers, and people who purchase homes in traditionally underserved areas. Because socially conscious investors likely also have interest in understanding how these social pools are likely to perform, we were curious to examine and learn whether mortgage pools with higher social ratings behaved differently than pools with lower social ratings (and if a difference existed, how significant it was). To the extent that pools rich in social factors perform better (i.e., prepay more slowly) than pools generally, we expect investors to put an even higher premium on them. This in turn should result in lower rates for the borrowers whose loans contribute to pools with higher social scores.

The data is new and we are still learning things, but we are beginning to discern some differences in prepay speeds.

Definitions

First, a quick refresher on Fannie’s and Freddie’s social index terminology:

Social Criteria Share (SCS): The percentage of loans in a given pool that meet at least one of the “social” criteria. The criteria are low-income, minority, and first-time homebuyers; homes in low-income areas, minority tracts, high-needs rural areas; homes in designated disaster areas and manufactured housing. As of December 2022, 42.12 percent of loans in the average pool satisfy at least one of these criteria.
Social Density Score (SDS): A measure of how many criteria the average loan in a given pool satisfies. For simplicity, the index consolidates the criteria into three categories – those pertaining to income, those pertaining to the borrower, and those pertaining to the property. A pool’s SDS can be zero, 1, 2, or 3 depending on the number of categories within which the loan satisfies at least one criterion. The average SDS as of December 2022 is 0.62 (out of 3).

Do social index scores impact prepay speeds?

While it remains too early to answer this question with a great deal of certainty, historical performance data appears to show that pools with below-average social index scores prepay faster than more “social” bonds.

We first looked at a high-level, simplistic relationship between prepayments and Social Density Score. In Figure 1, below, pools with below-average Social Density Scores (blue line) prepay faster than both pools with above-average SDS (black line) and pools with the very highest SDS (green line) when they are incentivized by interest rates to do so. (Note that very little difference exists among the curves when borrowers are out of the money to refi.)

Fig. 1: Speeds by Prepay Incentive and Social Density Score

See how easy RiskSpan’s Edge Platform makes it for you to do these analyses yourself.

Request a Trial

We note a similar trend when it comes to Social Criteria Share (see Fig. 2, below).

Fig. 2: Speeds by Prepay Incentive and Social Criteria Share

Social Pool Performance Relative to Spec Pools

Investors pay up for mortgage pools with specified characteristics. We thought it worthwhile to compare how certain types of spec pools perform relative to socially rich pools with no other specified characteristics.

Figure 3, below, compares the performance of non-spec pools with above-average Social Criteria Share (orange line) vs. spec pools for low-FICO (blue line), high-LTV (black line) and max $250k (green line) loans.

Note that, notwithstanding a lack of any other specific characteristics that investors pay up for, the high-SCS pools exhibit a somewhat better convexity profile than the max-700 FICO and min-95 LTV pools and slightly worse convexity (in most refi incentive buckets) than max-250k pools.

Fig. 3: Speeds by Prepay Incentive and Social Criteria Share: Socially Rich (Non-Spec) Pools vs. Selected Spec Pools

We observe a similar effect when we compare non-spec pools with an above-average Social Density Score to the same spec pools (Fig. 4, below).

Fig. 4: Speeds by Prepay Incentive and Social Density Score: Socially Rich (Non-Spec) Pools vs. Selected Spec Pools

See how social index scores affect speeds relative to other spec pools.

5 foundational steps for investors to move towards loan-level analyses

Are you curious about how your organization can uplevel the accuracy of your MSR cost forecasting? The answer lies in leveraging the full spectrum of your data and running analyses at the loan level rather than cohorting. But what does it take to make the switch to loan-level analytics? Our team has put together a short set of recommendations and considerations for how to tackle an otherwise daunting project…

It begins with having the data. Most investors have access to loan-level data, but it’s not always clean. This is especially true of origination data. If you’re acquiring a pool – be it a seasoned pool or a pool right after origination – you don’t have the best origination data to drive your model. You also need a data store, like Snowflake, that can generate loan-loan level output to drive your analytics and models.

The second factor is having models that work at the loan level – models that have been calibrated using loan-level performance and that are capable of generating loan-level output. One of the constraints of several existing modeling frameworks developed by vendors is they were created to run at a rep line level and don’t necessarily work very well for loan-level projections.

The third requirement is a compute farm. It is virtually impossible to run loan-level analytics if you’re not on the cloud because you need to distribute the computational load. And your computational distribution requirements will change from portfolio to portfolio based on the type of analytics that you are running, based on the types of scenarios that you are running, and based on the models you are using. The cloud is needed not just for CPU power but also for storage. This is because once you go to the loan level, every loan’s data must be made available to every processor that’s performing the calculation. This is where having the kind of shared databases, which are native to a cloud infrastructure, becomes vital. You simply can’t replicate it using an on-premise setup of computers in your office or in your own data center. Adding to this, it’s imperative for mortgage investors to remember the significance of integration and fluidity. When dealing with loan-level analytics, your systems—the data, the models, the compute power—should be interlinked to ensure seamless data flow. This will minimize errors, improve efficiency, and enable faster decision-making.

Fourth—and an often-underestimated component—is having intuitive user interfaces and visualization tools. Analyzing loan-level data is complex, and being able to visualize this data in a comprehensible manner can make all the difference. Dashboards that present loan performance, risk metrics, and other key indicators in an easily digestible format are invaluable. These tools help in quickly identifying patterns, making predictions, and determining the next strategic steps.

Fifth and finally, constant monitoring and optimization are crucial. The mortgage market, like any other financial market, evolves continually. Borrower behaviors change, regulatory environments shift, and economic factors fluctuate. It’s essential to keep your models and analytics tools updated and in sync with these changes. Regular back-testing of your models using historical data will ensure that they remain accurate and predictive. Only by staying ahead of these variables can you ensure that your loan-level analysis remains robust and actionable in the ever-changing landscape of mortgage investment.

Temporary Buydowns are Back. What Does This Mean for Speeds?

Mortgage buydowns are having a deja-vu moment. Some folks may recall mortgages with teaser rates in the pre-crisis period. Temporary buydowns are similar in concept. Recent declines notwithstanding, mortgage rates are still higher than they have been in years. Housing remains pricey. Would-be home buyers are looking for any help they can get. While on the other hand, with an almost non-existent refi market, mortgage originators are trying to find innovative ways to keep the production machine going. Conditions are ripe for lender and/or builder concessions that will help close the deal.

Enter the humble “temporary” mortgage interest rate buydown. A HousingWire article last month addressed the growing trend. It’s hard to turn on the TV without being bombarded with ads for Rocket Mortgage’s “Inflation Buster” program. Rocket Mortgage doesn’t use the term temporary buydown in its TV spots, but that is what it is.

Buydowns, in general, refer to when a borrower pays “points” upfront to reduce the mortgage rate to a level where they can afford the monthly payment. The mortgage rate has been “bought down” from its original rate for the entire life of the mortgage by paying a lumpsum upfront. Temporary Buydowns, on the other hand, come in various shapes and sizes, but the most common ones are a “2 – 1” (a 2-percent interest rate reduction in the first year and a 1-percent reduction in year two) and a “1 – 0” (a 1-percent interest rate reduction in the first year only). In these situations, the seller, or the builder, or the lender or a combination thereof put-up money to cover the difference in interest rate payments between the original mortgage rate and the reduced mortgage rate. In the 2-1 example above, the mortgage rate is reduced by 2% for the first year and then steps up by 1% in the second year and then steps up by another 1% in the 3^rd year to reach the actual mortgage rate at origination. So, the interest portion of the monthly mortgage payments are “subsidized” for the first two years and then revert to the full monthly payment. Given the inflated rental market, these programs can make purchasing more advantageous than renting (for home seekers trying to decide between the two options). They can also make purchasing a home more affordable (temporarily, at least) for would-be buyers who can’t afford the monthly payment at the prevailing mortgage rate. It essentially buys them time to refinance into a lower rate should interest rates fall over the subsidized time frame or they may be expecting increased income (raises, business revenue) in the future which will allow them to afford the unsubsidized monthly payment.

Temporary buydowns present an interesting situation for prepayment and default modelers. Most borrowers with good credit behave similarly to refinance incentives, barring loan size and refi cost issues. While permanent buydowns tend to exhibit slower speeds when they come in the money by a small amount since the borrower needs to make a cost/benefit decision about recouping the upfront money they put down and the refi costs associated with the new loan. Their breakeven point is going to be lower by 25bps or 50bps from their existing mortgage rate. So, their response to mortgage rates dropping will be slower than borrowers with similar mortgage rates who didn’t pay points upfront. Borrowers with temporary buydowns will be very sensitive to any mortgage rate drops and will refinance at the first opportunity to lock in a lower rate before the “subsidy” expires. Hence, such mortgages are expected to prepay at higher speeds then other counterparts with similar rates. In essence, they behave like ARMs when they approach their reset dates.

When rates stay static or increase, temporary buydowns will behave like their counterparts except when they get close to the reset dates and will see faster speeds. Two factors would contribute to this phenomenon. The most obvious reason is that temporary buydown borrowers will want to refinance into the lowest rate available at the time of reset (perhaps an ARM). The other possibility is that some of these borrowers may not be able refi because of DTI issues and may default. Such borrowers may also be deemed “weaker credits” because of the subsidy that they received. This increase in defaults would elevate their speeds (increased CBRs) relative to their counterparts.

So, for the reasons mentioned above, temporary buydown mortgages are expected to be the faster one among the same mortgage rate group. In the table below we separate borrowers with the same mortgage rate into 3 groups: 1) those that got a normal mortgage at the prevailing rate and paid no points, 2) those that paid points upfront to get a permanent lower rate and 3) those who got temporary lower rates subsidized by the seller/builder/lender. Obviously, the buydowns occurred in higher rate environments but we are considering 3 borrower groups with the same mortgage rate regardless of how they got that rate. We are assuming that all 3 groups of borrowers currently have a 6% mortgage. We present the expected prepay behavior of all 3 groups in different mortgage rate environments:

*Turnover++ means faster due to defaults or at reset

Rate	Rate Shift	6% (no pts)	Buydown to 6%(borrower-paid)	Buydown to 6% (lender-paid)
7.00%	+100	Turnover	Turnover	Turnover++*
6.00%	Flat	Turnover	Turnover	Faster (at reset)
5.75%	-25	Refi	Turnover	Refi
5.00%	-100	Refi (Faster)	Refi (Fast)	Refi (Fastest)

Overall, temporary buydowns are likely to exhibit the most rate sensitivity. As their mortgage rates reset higher, they will behave like ARMs and refi into any other lower rate option (5/1 ARM) or possibly default. In the money, they will be the quickest to refi.

Bumpy Road Ahead for GNMA MBS?

In a recent webinar, RiskSpan’s Fowad Sheikh engaged in a robust discussion with two of his fellow industry experts, Mahesh Swaminathan of Hilltop Securities and Mike Ortiz of DoubleLine Group, to address the likely road ahead for Ginnie Mae securities performance.

The panel sought to address the following questions:

How will the forthcoming, more stringent originator/servicer financial eligibility requirements affect origination volumes, buyouts, and performance?
Who will fill the vacuum left by Wells Fargo’s exiting the market?
What role will falling prices play in delinquency and buyout rates?
What will be the impact of potential Fed MBS sales.

This post summarizes some the group’s key conclusions. A recording of the webinar in its entirety is available here.

GET STARTED

Wells Fargo’s Departure

To understand the the likely impact of Wells Fargo’s exit, it is first instructive to understand the declining market share of banks overall in the Ginnie Mae universe. As the following chart illustrates, banks as a whole account for just 11 percent of Ginnie Mae originations, down from 39 percent as recently as 2015.

Drilling down further, the chart below plots Wells Fargo’s Ginnie Mae share (the green line) relative to the rest of the market. As the chart shows, Wells Fargo accounts for just 3 percent of Ginnie Mae originations today, compared to 15 percent in 2015. This trend of Wells Fargo’s declining market share extends all the way back to 2010, when it accounted for some 30 percent of Ginnie originations.

As the second chart below indicates, Wells Fargo’s market share, even among banks has also been on a steady decline.

GeT A Free Trial or Demo

Three percent of the overall market is meaningful but not likely to be a game changer either in terms of origination trends or impact on spreads. Wells Fargo, however, continues to have an outsize influence in the spec pool market. The panel hypothesized that Wells’s departure from this market could open the door to other entities claiming that market share. This could potentially affect prepayment speeds – especially if Wells is replaced by non-bank servicers, which the panel felt was likely given the current non-bank dominance of the top 20 (see below) – since Wells prepays have traditionally been slightly better than the broader market.

The panel raised the question of whether the continuing bank retreat from Ginnie Mae originations would adversely affect loan quality. As basis for this concern, they cited the generally lower FICO scores and higher LTVs that characterize non-bank-originated Ginnie Mae mortgages (see below).

These data notwithstanding, the panel asserted that any changes to credit quality would be restricted to the margins. Non-bank servicers originate a higher percentage of lower-credit-quality loans (relative to banks) not because non-banks are actively seeking those borrowers out and eschewing higher-credit-quality borrowers. Rather, banks tend to restrict themselves to borrowers with higher credit profiles. Non-banks will be more than happy to lend to these borrowers as banks continue to exit the market.

Effect of New Eligibility Requirements

The new capital requirements, which take effect a year from now, are likely to be less punitive than they appear at first glance. With the exception of certain monoline entities – say, those with almost all of their assets concentrated in MSRs – the overwhelming majority of Ginnie Mae issuers (banks and non-banks alike) are going to be able meet them with little if any difficulty.

Ginnie Mae has stated that, even if the new requirements went into effect tomorrow, 95 percent of its non-bank issuers would qualify. Consequently, the one-year compliance period should open the door for a fairly smooth transition.

To the extent Ginnie Mae issuers are unable to meet the requirements, a consolidation of non-bank entities is likely in the offing. Given that these institutions will likely be significant MSR investors, the potential increase in MSR sales could impact MSR multiples and potentially disrupt the MSR market, at least marginally.

Potential Impacts of Negative HPA

Ginnie Mae borrowers tend to be more highly leveraged than conventional borrowers. FHA borrowers can start with LTVs as high as 97.5 percent. VA borrowers, once the VA guarantee fee is rolled in, often have LTVs in excess of 100 percent. Similar characteristics apply to USDA loans. Consequently, borrowers who originated in the past two years are more likely to default as they watch their properties go underwater. This is potentially good news for investors in discount coupons (i.e., investors who benefit from faster prepay speeds) because these delinquent loans will be bought out quite early in their expected lives.

More seasoned borrowers, in contrast, have experienced considerable positive HPA in recent years. The coming forecasted decline should not materially impact these borrowers’ performance. Similarly, if HPD in 2023 proves to be mild, then a sharp uptick in delinquencies is unlikely, regardless of loan vintage or LTV. Most homeowners make mortgage payments because they wish to continue living in their house and do not seriously consider strategic defaults. During the financial crisis, most borrowers continued making good on their mortgage obligations even as their LTVs went as high as the 150s.

Further, the HPD we are likely to encounter next year likely will not have the same devastating effect as the HPD wave that accompanied the financial crisis. Loans on the books today are markedly different from loans then. Ginnie Mae loans that went bad during the crisis disproportionately included seller-financed, down-payment-assistance loans and other programs lacking in robust checks and balances. Ginnie Mae has instituted more stringent guidelines in the years since to minimize the impact of bad actors in these sorts of programs.

This all assumes, however, that the job market remains robust. Should the looming recession lead to widespread unemployment, that would have a far more profound impact on delinquencies and buyouts than would HPD.

Fed Sales

The Fed’s holdings (as of 9/21, see chart below) are concentrated around 2 percent and 2.5 percent coupons. This raises the question of what the Fed’s strategy is likely to be for unwinding its Ginnie Mae position.

Word on the street is that Fed sales are highly unlikely to happen in 2022. Any sales in 2023, if they happen at all, are not likely before the second half of the year. The panel opined that the composition of these sales is likely to resemble the composition of the Fed’s existing book – i.e., mostly 2s, 2.5s, and some 3s. They have the capacity to take a more sophisticated approach than a simple pro-rata unwinding. Whether they choose to pursue that is an open question.

The Fed was a largely non-economic buyer of mortgage securities. There is every reason to believe that it will be a non-economic seller, as well, when the time comes. The Fed’s trading desk will likely reach out to the Street, ask for inquiry, and seek to pursue an approach that is least disruptive to the mortgage market.

Conclusion

On closer consideration, many of these macro conditions (Wells’s exit, HPD, enhanced eligibility requirements, and pending Fed sales) that would seem to portend an uncertain and bumpy road for Ginnie Mae investors, may turn out to be more benign than feared.

Conditions remain unsettled, however, and these and other factors certainly bear watching as Ginnie Mae market participants seek to plot a prudent course forward.

Optimizing Analytics Computational Processing

We met with RiskSpan’s Head of Engineering and Development, Praveen Vairavan, to understand how his team set about optimizing analytics computational processing for a portfolio of 4 million mortgage loans using a cloud-based compute farm.

This interview dives deeper into a case study we discussed in a recent interview with RiskSpan’s co-founder, Suhrud Dagli.

Here is what we learned from Praveen.

Speak to an Expert

Could you begin by summarizing for us the technical challenge this optimization was seeking to overcome?

PV: The main challenge related to an investor’s MSR portfolio, specifically the volume of loans we were trying to run. The client has close to 4 million loans spread across nine different servicers. This presented two related but separate sets of challenges.

The first set of challenges stemmed from needing to consume data from different servicers whose file formats not only differed from one another but also often lacked internal consistency. By that, I mean even the file formats from a single given servicer tended to change from time to time. This required us to continuously update our data mapping and (because the servicer reporting data is not always clean) modify our QC rules to keep up with evolving file formats.

The second challenge relates to the sheer volume of compute power necessary to run stochastic paths of Monte Carlo rate simulations on 4 million individual loans and then discount the resulting cash flows based on option adjusted yield across multiple scenarios.

And so you have 4 million loans times multiple paths times one basic cash flow, one basic option-adjusted case, one up case, and one down case, and you can see how quickly the workload adds up. And all this needed to happen on a daily basis.

To help minimize the computing workload, our client had been running all these daily analytics at a rep-line level—stratifying and condensing everything down to between 70,000 and 75,000 rep lines. This alleviated the computing burden but at the cost of decreased accuracy because they couldn’t look at the loans individually.

What technology enabled you to optimize the computational process of running 50 paths and 4 scenarios for 4 million individual loans?

PV: With the cloud, you have the advantage of spawning a bunch of servers on the fly (just long enough to run all the necessary analytics) and then shutting it down once the analytics are done.

This sounds simple enough. But in order to use that level of compute servers, we needed to figure out how to distribute the 4 million loans across all these different servers so they can run in parallel (and then we get the results back so we could aggregate them). We did this using what is known as a MapReduce approach.

Say we want to run a particular cohort of this dataset with 50,000 loans in it. If we were using a single server, it would run them one after the other – generate all the cash flows for loan 1, then for loan 2, and so on. As you would expect, that is very time-consuming. So, we decided to break down the loans into smaller chunks. We experimented with various chunk sizes. We started with 1,000 – we ran 50 chunks of 1,000 loans each in parallel across the AWS cloud and then aggregated all those results.

That was an improvement, but the 50 parallel jobs were still taking longer than we wanted. And so, we experimented further before ultimately determining that the “sweet spot” was something closer to 5,000 parallel jobs of 100 loans each.

Only in the cloud is it practical to run 5,000 servers in parallel. But this of course raises the question: Why not just go all the way and run 50,000 parallel jobs of one loan each? Well, as it happens, running an excessively large number of jobs carries overhead burdens of its own. And we found that the extra time needed to manage that many jobs more than offset the compute time savings. And so, using a fair bit of trial and error, we determined that 100-loan jobs maximized the runtime savings without creating an overly burdensome number of jobs running in parallel.

Get A Demo

You mentioned the challenge of having to manage a large number of parallel processes. What tools do you employ to work around these and other bottlenecks?

PV: The most significant bottleneck associated with this process is finding the “sweet spot” number of parallel processes I mentioned above. As I said, we could theoretically break it down into 4 million single-loan processes all running in parallel. But managing this amount of distributed computation, even in the cloud, invariably creates a degree of overhead which ultimately degrades performance.

And so how do we find that sweet spot – how do we optimize the number of servers on the distributed computation engine?

As I alluded to earlier, the process involved an element of trial and error. But we also developed some home-grown tools (and leveraged some tools available in AWS) to help us. These tools enable us to visualize computation server performance – how much of a load they can take, how much memory they use, etc. These helped eliminate some of the optimization guesswork.

Is this optimization primarily hardware based?

PV: AWS provides essentially two “flavors” of machines. One “flavor” enables you to take in a lot of memory. This enables you to keep a whole lot of loans in memory so it will be faster to run. The other flavor of hardware is more processor based (compute intensive). These machines provide a lot of CPU power so that you can run a lot of processes in parallel on a single machine and still get the required performance.

We have done a lot of R&D on this hardware. We experimented with many different instance types to determine which works best for us and optimizes our output: Lots of memory but smaller CPUs vs. CPU-intensive machines with less (but still a reasonably amount of) memory.

We ultimately landed on a machine with 96 cores and about 240 GB of memory. This was the balance that enabled us to run portfolios at speeds consistent with our SLAs. For us, this translated to a server farm of 50 machines running 70 processes each, which works out to 3,500 workers helping us to process the entire 4-million-loan portfolio (across 50 Monte Carlo simulation paths and 4 different scenarios) within the established SLA.

What software-based optimization made this possible?

PV: Even optimized in the cloud, hardware can get pricey – on the order of $4.50 per hour in this example. And so, we supplemented our hardware optimization with some software-based optimization as well.

We were able to optimize our software to a point where we could use a machine with just 30 cores (rather than 96) and 64 GB of RAM (rather than 240). Using 80 of these machines running 40 processes each gives us 2,400 workers (rather than 3,500). Software optimization enabled us to run the same number of loans in roughly the same amount of time (slightly faster, actually) but using fewer hardware resources. And our cost to use these machines was just one-third what we were paying for the more resource-intensive hardware.

All this, and our compute time actually declined by 10 percent.

The software optimization that made this possible has two parts:

The first part (as we discussed earlier) is using the MapReduce methodology to break down jobs into optimally sized chunks.

The second part involved optimizing how we read loan-level information into the analytical engine. Reading in loan-level data (especially for 4 million loans) is a huge bottleneck. We got around this by implementing a “pre-processing” procedure. For each individual servicer, we created a set of optimized loan files that can be read and rendered “analytics ready” very quickly. This enables the loan-level data to be quickly consumed and immediately used for analytics without having to read all the loan tapes and convert them into a format that analytics engine can understand. Because we have “pre-processed” all this loan information, it is immediately available in a format that the engine can easily digest and run analytics on.

This software-based optimization is what ultimately enabled us to optimize our hardware usage (and save time and cost in the process).

Rethink Analytics Computational Processing – Solving Yesterday’s Problems with Today’s Technology and Access

We sat down with RiskSpan’s co-founder and chief technology officer, Suhrud Dagli, to learn more about how one mortgage investor successfully overhauled its analytics computational processing. The investor migrated from a daily pricing and risk process that relied on tens of thousands of rep lines to one capable of evaluating each of the portfolio’s more than three-and-a-half million loans individually (and how they actually saved money in the process).

Here is what we learned.

Could you start by talking a little about this portfolio — what asset class and what kind of analytics the investor was running?

SD: Our client was managing a large investment portfolio of mortgage servicing rights (MSR) assets, residential loans and securities.

The investor runs a battery of sophisticated risk management analytics that rely on stochastic modeling. Option-adjusted spread, duration, convexity, and key rate durations are calculated based on more than 200 interest rate simulations.

GET A FREE DEMO OR FREE TRIAL

Why was the investor running their analytics computational processing using a rep line approach?

SD: They used rep lines for one main reason: They needed a way to manage computational loads on the server and improve calculation speeds. Secondarily, organizing the loans in this way simplified their reporting and accounting requirements to a degree (loans financed by the same facility were grouped into the same rep line).

This approach had some downsides. Pooling loans by finance facility was sometimes causing loans with different balances, LTVs, credit scores, etc., to get grouped into the same rep line. This resulted in prepayment and default assumptions getting applied to every loan in a rep line that differed from the assumptions that likely would have been applied if the loans were being evaluated individually.

The most obvious solution to this would seem to be one that disassembles the finance facility groups into their individual loans, runs all those analytics at the loan level, and then re-aggregates the results into the original rep lines. Is this sort of analytics computational processing possible without taking all day and blowing up the server?

SD: That is effectively what we are doing. The process is not a speedy as we’d like it to be (and we are working on that). But we have worked out a solution that does not overly tax computational resources.

The analytics computational processing we are implementing ignores the rep line concept entirely and just runs the loans. The scalability of our cloud-native infrastructure enables us to take the three-and-a-half million loans and bucket them equally for computation purposes. We run a hundred loans on each processor and get back loan-level cash flows and then generate the output separately, which brings the processing time down considerably.

SPEAK TO AN EXPERT

So we have a proof of concept that this approach to analytics computational processing works in practice for running pricing and risk on MSR portfolios. Is it applicable to any other asset classes?

SD: The underlying principles that make analytics computational processing possible at the loan level for MSR portfolios apply equally well to whole loan investors and MBS investors. In fact, the investor in this example has a large whole-loan portfolio alongside its MSR portfolio. And it is successfully applying these same tactics on that portfolio.

An investor in any mortgage asset benefits from the ability to look at and evaluate loan characteristics individually. The results may need to be rolled up and grouped for reporting purposes. But being able to run the cash flows at the loan level ultimately makes the aggregated results vastly more meaningful and reliable.

A loan-level framework also affords whole-loan and securities investors the ability to be sure they are capturing the most important loan characteristics and are staying on top of how the composition of the portfolio evolves with each day’s payoffs.

ESG factors are an important consideration for a growing number of investors. Only a loan-level approach makes it possible for these investors to conduct the kind of property- and borrower-level analyses to know whether they are working toward meeting their ESG goals. It also makes it easier to spot areas of geographic concentration risk, which simplifies climate risk management to some degree.

Say I am a mortgage investor who is interested in moving to loan-level pricing and risk analytics. How do I begin?

SD: Three things:

It begins with having the data. Most investors have access to loan-level data. But it’s not always clean. This is especially true of origination data. If you’re acquiring a pool – be it a seasoned pool or a pool right after origination – you don’t have the best origination data to drive your model. You also need a data store that can generate loan-loan level output to drive your analytics and models.
The second factor is having models that work at the loan level – models that have been calibrated using loan-level performance and that are capable of generating loan-level output. One of the constraints of several existing modeling frameworks developed by vendors is they were created to run at a rep line level and don’t necessarily work very well for loan-level projections.
The third thing you need is a compute farm. It is virtually impossible to run loan-level analytics if you’re not on the cloud because you need to distribute the computational load. And your computational distribution requirements will change from portfolio to portfolio based on the type of analytics that you are running, based on the types of scenarios that you are running, and based on the models you are using.

The cloud is needed not just for CPU power but also for storage. This is because once you go to the loan level, every loan’s data must be made available to every processor that’s performing the calculation. This is where having the kind of shared databases, which are native to a cloud infrastructure, becomes vital. You simply can’t replicate it using a on-premise setup of computers in your office or in your own data center.

So, 1) get your data squared away, 2) make sure you’re using models that are optimized for loan-level, and 3) max out your analytics computational processing power by migrating to cloud-native infrastructure. Thank you, Suhrud, for taking the time to speak with us.

Quantifying the Impact of Climate Risk on Housing Finance

When people speak of the risk climate poses to housing, they typically do so in qualitative and relative terms. A Florida home is at greater risk of hurricane damage than an Iowa home. Wildfires generally threaten homes in northern California more than they threaten homes in New Hampshire. And because of climate change, the risk these and other perils pose to any individual geographical area are largely viewed as higher than they were 25 years ago.

People feel comfortable speaking in these general terms. But qualitative estimates are of little practical use to mortgage investors seeking to fine-tune their pricing, prepayment, and default models. These analytical frameworks require not just reliable data but the means to translate them into actionable risk metrics.

Physical risks and transition risks

Broadly speaking, climate risk manifests itself as a combination of physical risks and transition risks. Physical risks include “acute” disaster events, such as hurricanes, tornadoes, wildfires, and floods. Chronic risks, such as sea level rise, extreme temperatures, and drought, are experienced over a longer period. Transition risks relate to costs resulting from regulations promulgated to combat climate change and from the need to invest in new technologies designed either to combat climate change directly or mitigate its effects.

Some of the ways in which these risks impact mortgage assets are self-evident. Acute events that damage or destroy homes have an obvious effect on the performance of the underlying mortgages. Other mechanisms are more latent but no less real. Increasing costs of homeownership, caused by required investment in climate-change-mitigating technologies, can be a source of financial stress for some borrowers and affect mortgage performance. Likewise, as flood and other hazard insurance premiums adjust to better reflect the reality of certain geographies’ increasing exposure to natural disaster risk, demand for real estate in these areas could decrease, increasing the pressure on existing homeowners who may not have much cushion in their LTVs to begin with.

Mortgage portfolio risk management

At the individual loan level, these risks translate to higher delinquency risks, probability of default, loss given default, spreads, and advance expenses. At the portfolio level, the impact is felt in asset valuation, concentration risk (what percentage of homes in the portfolio are located in high-risk areas), VaR, and catastrophic tail risk.

VaR can be computed using natural hazard risk models designed to forecast the probability of individual perils for a given geography and using that probability to compute the worst property loss (total physical loss and loss net of insurance proceeds) that can be expected during the portfolio’s expected life at the 99 percent (or 95 percent) confidence level. The following figure illustrates how this works for a portfolio covering multiple geographies with varying types and likelihoods of natural hazard risk.

These analyses can look at the exposure of an entire portfolio to all perils combined:

SPEAK TO AN EXPERT

Or they can look at the exposure of a single geographic area to one peril in particular:

Accounting for climate risk when bidding on whole loans

The risks quantified above pertain to properties that secure mortgages and therefore only indirectly to the mortgage assets themselves. Investors seeking to build whole-loan portfolios that are resilient to climate risk should consider climate risk in the context of other risk factors. Such a “property-level climate risk” approach takes into account factors such as:

Whether the property is insured against the peril in question
The estimate expected risk (and tail risk) of property damage by the peril in question
Loan-to-value ratio

The most prudent course of action includes a screening mechanism that includes pricing and concentration limits tied to LTV ratios. Investors may choose to invest in areas of high climate risk but only in loans with low LTV ratios. Bids should be adjusted to account for climate risk, but the amount of the adjustment can be a function of the LTV. Concentration limits should be adjusted accordingly:

Conclusion

When assessing the impact of climate risk on a mortgage portfolio, investors need to consider and seek to quantify not just how natural hazard events will affect home values but also how they will affect borrower behavior, specifically in terms of prepayments, delinquencies, and defaults.

We are already beginning to see climate factors working their way into the secondary mortgage markets via pricing adjustments and concentration screening. It is only a matter of time before these considerations move further up into the origination process and begin to manifest themselves in pricing and underwriting policy (as flood insurance requirements already have today).

Investors looking for a place to start can begin by incorporating a climate risk score into their existing credit box/pricing grid, as illustrated above. This will help provide at least a modicum of comfort to investors that they are being compensated for these hidden risks and (at least as important) will ensure that portfolios do not become overly concentrated in at-risk areas.

GET STARTED

“Reject Inference” Methods in Credit Modeling: What are the Challenges?

Reject inference is a popular concept that has been used in credit modeling for decades. Yet, we observe in our work validating credit models that the concept is still dynamically evolving. The appeal of reject inference, whose aim is to develop a credit scoring model utilizing all available data, including that of rejected applicants, is easy enough to grasp. But the technique also introduces a number of fairly vexing challenges.

The technique seeks to rectify a fundamental shortcoming in traditional credit modeling: Models predicting the probability that a loan applicant will repay the loan can be trained to historical loan application data with a binary variable representing whether a loan was repaid or charged off. This information, however, is only available for accepted applications. And many of these applications are not particularly recent. This limitation results in a training dataset that may not be representative of the broader loan application universe.

Credit modelers have devised several techniques for getting around this data representativeness problem and increasing the number of observations by inferring the repayment status of rejected loan applications. These techniques, while well intentioned, are often treated empirically and lack a deeper theoretical basis. They often result in “hidden” modeling assumptions, the reasonableness of which is not fully investigated. Additionally, no theoretical properties of the coefficient estimates, or predictions are guaranteed.

This article summarizes the main challenges of reject inference that we have encountered in our model validation practice.

Speak With A MODEL VALIDATION EXPERT

Selecting the Right Reject Inference Method

Many approaches exist for reject inference, none of which is clearly and universally superior to all the others. Empirical studies have been conducted to compare methods and pick a winner, but the conclusions of these studies are often contradictory. Some authors argue that reject inference cannot improve scorecard models[1]and flatly recommend against their use. Others posit that certain techniques can outperform others[2] based on empirical experiments. The results of these experiments, however, tend to be data dependent. Some of the most popular approaches include the following:

Ignoring rejected applications: The simplest approach is to develop a credit scoring model based only on accepted applications. The underlying assumption is that rejected applications can be ignored and that the “missingness” of this data from the training dataset can be classified as missing at random. Supporters of this method point to the simplicity of the implementation, clear assumptions, and good empirical results. Others argue that the rejected applications cannot be dismissed simply as random missing data and thus should not be ignored.
Hard cut-off method: In this method, a model is first trained using only accepted application data. This trained model is then used to predict the probabilities of charge-off for the rejected applications. A cut-off value is then chosen. Hypothetical loans from rejected applications with probabilities higher than this cut-off value are considered charged off. Hypothetical loans from the remaining applications are assumed to be repaid. The specified model is then re-trained using a dataset including both accepted and rejected applications.
Fuzzy augmentation: Similar to the hard cut-off method, fuzzy augmentation begins by training the model on accepted applications only. The resulting model with estimated coefficients is then used to predict charge-off probabilities for rejected applications. Data from rejected applications is then duplicated and a repaid or charged-off status is assigned to each. The specified model is then retrained on the augmented dataset—including accepted applications and the duplicated rejects. Each rejected application is weighted by either a) the predicted probability of charge-off if its assigned status is “charged-off,” or b) the predicted probability of it being repaid if its assigned status is “repaid.”
Parceling: The parceling method resembles the hard cut-off method. However, rather than classifying all rejects above a certain threshold as charged-off, this method classifies the repayment status in proportion to the expected “bad” rate (charge-off frequency) at that score. The predicted charge-off probabilities are partitioned into k intervals. Then, for each interval, an assumption is made about the bad rate, and loan applications in each interval are assigned a repayment status randomly according to the bad rate. Bad rates are assumed to be higher in the reject dataset than among the accepted loans. This method considers the missingness to be not at random (MNAR), which requires the modeler to supplement the additional information about the distribution of charge-offs among rejects.

Proportion of Accepted Applications to Rejects

An institution with a relatively high percentage of rejected applications will necessarily end up with an augmented training dataset whose quality is heavily dependent on the quality of the selected reject inference method and its implementation. One might argue it is best to limit the proportion of rejected applications to acceptances. The level at which such a cap is established should reflect the “confidence” in the method used. Estimating such a confidence level, however, is a highly subjective endeavor.

The Proportion of Bad Rates for Accepts and Rejects

It is reasonable to assume that the “bad rate,” i.e., proportion of charged-off loans to repaid loans, will be higher among rejected applications. Some modelers set a threshold based on their a priori belief that the bad rate among rejects is at least p-times the bad rate among acceptances. If the selected reject inference method produces a dataset with a bad rate that is perceived to be artificially low, actions are taken to increase the bad rate above some threshold. Identifying where to establish this threshold is notoriously difficult to justify.

Variable Selection

As outlined above, most approaches begin by estimating a preliminary model based on accepted applications only. This model is then used to infer how rejected loans would have performed. The preliminary model is then retrained on a dataset consisting both of actual data from accepted applications and of the inferred data from rejects. This means that the underlying variables themselves are selected based only on the actual loan performance data from accepted applications. The statistical significance of the selected variables might change, however, when moving to the complete dataset. Variable selection is sometimes redone using the complete data. This, however, can lead to overfitting.

Measuring Model Performance

From a model validator’s perspective, an ideal solution would involve creating a control group in which applications would not be scored and filtered and every application would be accepted. Then the discriminating power of a credit model could be assessed by comparing the charge-off rate of the control group with the charge-off rate of the loans accepted by the model. This approach of extending credit indiscriminately is impractical, however, as it would require the lender to engage in some degree of irresponsible lending.

Another approach is to create a test set. The dilemma here is whether to include only accepted applications. A test set that includes only accepted applications will not necessarily reflect the population for which the model will be used. Including rejected applications, however, obviously necessitates the use of reject inference. For all the reasons laid out above, this approach risks overstating the model’s performance due to the fact that a similar model (trained only on the accepted cases) was used for reject inference.

A third approach that avoids both of these problems involves using information criteria such as AIC and BIC. This, however, is useful only when comparing different models (for model or variable selection). The values of information criteria cannot be interpreted as an absolute measure of performance.

A final option is to consider utilizing several models in production (the main model and challenger models). Under this scenario, each application would be evaluated by a model selected at random. The models can then be compared retroactively by calculating their bad rates on accepted application after the financed loans mature. Provided that the accept rates are similar, the model with the lowest bad rate is the best.

Conclusion

Reject inference remains a progressing field in credit modeling. Its ability to improve model performance is still the subject of intense debate. Current results suggest that while reject inference can improve model performance, its application can also lead to overfitting, thus worsening the ability to generalize. The lack of a strong theoretical basis for reject inference methods means that applications of reject inference need to rely on empirical results. Thus, if reject inference is used, key model stakeholders need to possess a deep understanding of the modeled population, have strong domain knowledge, emphasize conducting experiments to justify the applied modeling techniques, and, above all, adopt and follow a solid ongoing monitoring plan.

Doing this will result in a modeling methodology that is most likely to produce reliable outputs for the institutions while also satisfying MRM and validator requirements.

[1] https://www.sciencedirect.com/science/article/abs/pii/S0378426603002036

[2] https://economix.fr/pdf/dt/2016/WP_EcoX_2016-10.pdf

‹ Prev 1 234 5 6 Next ›Last »

Video: Mortgage Market Evolution

MSR & Loan Trading Insights

What do 2023 Originations Means for MSRs?

Why Accurate Loan Pool and MSR Cost Forecasting Requires Loan-by-Loan Analytics

5 foundational steps for investors to move towards loan-level analyses

It’s time to move to DaaS — Why it matters for loan and MSR investors

Case Study: How one investor moved to loan level analysis while reducing their costs

Whitepaper: Improving the precision of MSR pricing using loan-level analytics

Interested In learning about RiskSpan’s Edge Platform?

Agency Social Indices & Prepay Speeds

5 foundational steps for investors to move towards loan-level analyses

Temporary Buydowns are Back. What Does This Mean for Speeds?

Bumpy Road Ahead for GNMA MBS?

Optimizing Analytics Computational Processing

Could you begin by summarizing for us the technical challenge this optimization was seeking to overcome?

What technology enabled you to optimize the computational process of running 50 paths and 4 scenarios for 4 million individual loans?

You mentioned the challenge of having to manage a large number of parallel processes. What tools do you employ to work around these and other bottlenecks?

Is this optimization primarily hardware based?

What software-based optimization made this possible?

Rethink Analytics Computational Processing – Solving Yesterday’s Problems with Today’s Technology and Access

Could you start by talking a little about this portfolio — what asset class and what kind of analytics the investor was running?

Why was the investor running their analytics computational processing using a rep line approach?

So we have a proof of concept that this approach to analytics computational processing works in practice for running pricing and risk on MSR portfolios. Is it applicable to any other asset classes?

Say I am a mortgage investor who is interested in moving to loan-level pricing and risk analytics. How do I begin?

Quantifying the Impact of Climate Risk on Housing Finance

“Reject Inference” Methods in Credit Modeling: What are the Challenges?

Selecting the Right Reject Inference Method

Proportion of Accepted Applications to Rejects

The Proportion of Bad Rates for Accepts and Rejects

Variable Selection

Measuring Model Performance

Conclusion

Company

Products

Security & Compliance