By Tina Freeman, CFA

Recent market turmoil is challenging the hedging capability of mortgage bankers across the industry. Layer in the issues associated with the COVID-19 pandemic, and you have a whole new level of uncertainty. Let’s start with the basics. To craft the best hedge, you must first understand your pipeline’s fallout behavior. Fallout is simply the percentage of loans that don’t close. If your fallout modeling is accurate, you know how much hedge coverage to put on, and your hedge will be successful. If you get it wrong, you can suffer serious losses.

The challenge, of course, is in understanding how both operations and market conditions impact your organization’s fallout behavior and building a predictive model that makes reliable estimations of expected closings under different market conditions. This requires the study of a robust set of your historical data and observing closing rates for different loan characteristics such as product, loan purpose, property type, and credit quality.

It is essential to also understand how fallout will change if interest rates change. A small increase in closing rates might happen if rates rise, as borrowers become more motivated to get deals done. But if rates fall, closing rates can decline sharply and there will be more fallout as borrowers seek lower rates elsewhere or lenders may renegotiate locks, offering borrowers lower rates to keep them from leaving. These renegotiations are in essence a “soft fallout” which is a form of loss mitigation. The lender still loses money since the lock was hedged but can make some origination margin on the loan to offset the hedge loss, which in most cases is better than losing a deal altogether.

Fallout studies should also attempt to understand how locks perform after they pass key processing milestones. For example, a lock that has been approved in underwriting not only has a higher probability of closing, but it will also be less likely to “walk away” or renegotiate if rates drop. An effective fallout model should not only adjust for processing milestones but should recognize that fallout rates are most volatile early in the processing cycle and become less volatile the closer a transaction is to the closing table.

Once the fallout study is complete, it’s time to build an effective fallout model. The fallout model should be granular enough to recognize the key qualitative factors related to the rate lock and assign a fallout probability at a loan level. It should then be able to project a shock that provides a revised estimate of closing probability on each loan for rising and falling rate environments, and this determination should incorporate a measure of market movement from the inception date of each individual rate lock. For more than 30 years, MIAC has refined and enhanced its industry-leading fallout models, through both calm and volatile markets.

Common Fallout Modeling Problems Exposed in the Current Market

When a big rally comes along, available historical data for creating the fallout model may not provide the sampling information needed to populate expected closing probability curves. Common modeling problems that can be exposed by such a market include:

1.  Failure to recognize market conditions affecting historical performance: 

If a lender builds a fallout model with baseline closing percentages tied directly to historical data, and that history reflected a strongly trending rate environment, the baselines are not valid. For example, if a lender concludes they have very stable fallout and little renegotiation activity based upon a period of rising rates, they may fail to anticipate that baseline closing rates in a stable environment are actually a bit lower, or may falsely assume that in a falling rate environment they will have little exposure to increased fallout or renegotiations. Likewise, if baseline rates are determined from data collected during a rally, and the impact of market movement was not captured, baseline closing rates may be set too low and the lender may under-hedge in a rising rate environment.

2.  Failure to anticipate “out of sample” market behavior:

Many lenders industry-wide have suffered pipeline losses in 2020 due to fallout models that did not anticipate the degree of both absolute and “soft” fallout that has occurred due to falling interest rates. Some lenders may never even have experienced a big rally! A quality model must make estimates for potential increases in fallout in market rally scenarios that are outside the bounds of the rate movements that occurred during the time period captured by the historical data set.

3.  Failure to measure the impact of rate movements at a loan level: 

If the fallout study does not measure the impact of interest rate movements on every historical rate lock, the lender has no basis from which to forecast the anticipated change in closing rates when rates move. Once the history is obtained on each rate lock, it is important to assess the impact of rate movements not only on the population but on subgroups based on key loan characteristics.  It is also critical to assess how the relationships change on these subgroups as loans move through the processing cycle. Getting this right is critically important to maintaining an effective hedge, especially in a rally when the lender’s exposure to both absolute and “soft” fallout is at its greatest.

Some industry models bucket loans by loan characteristics, without properly measuring fallout volatility on these buckets at the loan level, and without building this expected volatility into the shock analysis and hedge construction. The risk here is that closing rates are overestimated in a rally, and the lender loses money by being over-hedged. MIAC’s models are based on loan level analysis including robust shock analysis.

4.  Failure to recognize that the fallout function is not linear:

Some models will recommend coverage based upon an expected closing percentage but fail to recognize that this closing percentage will change if the market moves and will change in a non-linear fashion. At the high end of the closing estimate curve, when rates are stable or rising, the only fallout anticipated is associated with loans that are declined or transactions that fall through for non-market related reasons. As rates start to fall, the anticipated closing rate will gradually go down and fallout will increase. But the more the market falls, the more likely borrowers are to walk away or renegotiate, and this fallout increases at an increasing rate the more that rates fall.

Why is this so important? If you think about your objective in hedging your pipeline, it is to preserve the profit margin you planned for when you locked the loan.  The hedge you place should not be based on simply a baseline measure of “net exposure”, but one that best offsets the change in the value of your pipeline in a variety of rate scenarios. By using MIAC’s MarketShield® and dynamically shocking the fallout rates on each loan in a granular fashion, and doing so with a properly non-linear fallout function, you get the best estimate of the value of each loan at each shock point. This allows you to create a more effective overall hedge.

Models that do not do this are doomed to leave the lender over-hedged in a rally, resulting in excessive pair-off losses and an erosion of the expected profit margin.

The Pandemic Question: What impact will it have?

There are undoubtedly loans in every pipeline that are subject to issues associated with the COVID-19 pandemic. Most obviously, there are applications in pipelines submitted by borrowers who will lose their jobs or have their income severely impacted by business closures and furloughs.

There will also be impacts on operations caused by social distancing and “stay at home” orders that may result in delays or even cancellations of transactions. Borrowers that are quarantined may not be able to attend a closing. Borrowers may not want to allow an appraiser or inspector into their property. County courthouses will be working with reduced staffing and will be limiting services. Vendors that are already backlogged may have staffing issues related to remote working arrangements or sick leave.

What do we do with our fallout models when we have no data? We adapt our models as the data comes in. But until then, it is wise and prudent to proactively adjust assumptions, using our best judgment in collaboration with each lender.

On Pull-through:

The geographic distribution of the pipeline is important to consider, as the degree of “lockdown” varies by locality, as well as the distribution of borrower occupations and income level. We would recommend considering a reduction in pull-through expectations of 5-10% for most lenders.

On Cost to Completion:

On a related note, when it comes to the valuation of interest rate lock derivatives, one important factor is the expected Cost to Completion. Lenders should increase the expected Cost to Completion to recognize the fact that there will be both market and operating costs associated with the pandemic. A reasonable estimate of between .125% and .50% for potential market rolls/extension costs would be prudent, particularly for locks that are at the early stages of processing. Furthermore, it is reasonable to assume that processing and underwriting costs may be higher than normal.

On Pipeline Cleanup: 

It is critically important for lenders to attempt to identify and remove applications that no longer qualify from the pipeline as quickly as possible in order to reduce exposure to hedge losses. While keeping a clean pipeline is always a best practice, not since 2008 have we had a situation that could adversely impact so many applicants at once. It may also be prudent, if possible, to flag applications that are “at risk” of decline, so that they may be proactively hedged at a lower closing probability.

On Reserves: 

Once you have made a reasonable adjustment to your pull-through model, adjusted your rate lock derivative to be more conservative with respect to costs, and made sure your pipeline is clean, you may decide it makes sense to set up an additional reserve from first-quarter earnings. Lenders took an enormous volume of rate locks that are going to generate derivative earnings that are at a much higher risk than normal. Setting up a reserve now can help reduce the risk of having to write off these derivatives later.

MIAC Can Help

In short, don’t get analysis paralysis, but make sure you are considering the impact of current market conditions on your hedge. The pandemic adds an entirely new layer of uncertainty to an already complicated process. MIAC can bring to bear the industry’s leading models, analytics and experienced professionals to help you navigate the uncertainty.

About MIAC

Since 1989, MIAC Analytics™ has been the preferred destination for sophisticated mortgage industry and capital markets participants, offering transaction execution services, secondary market hedge advisory solutions, third-party mortgage asset valuations, as well as state-of-the-art valuation and risk models incorporating a full range of consumer behavioral risk factors.

Follow MIAC on LinkedIn and Twitter!

Author

Tina Freeman, CFA
Managing Director, Secondary Solutions Group
(212) 233-1250 ext. 236
Tina.Freeman@miacanalytics.com

Author Bio

Tina Freeman, CFA, Managing Director of the Secondary Solutions Group, has devoted most of her 30-year mortgage banking career to risk management and profitability strategies in Secondary Marketing. Prior to joining MIAC in 2004, she spent 13 years at Everbank in Jacksonville, FL as Secondary Marketing Manager.

Ms. Freeman received her BA in Accounting, Summa Cum Laude, and her MBA from the University of North Florida. She received her Certified Public Accountant designation in 1995, and was an Elijah Watt Sells Award Winner. In 2002, she earned the Chartered Financial Analyst (CFA) designation.

Additional Contacts

Brad Eskridge
Managing Director, Secondary Solutions Group
(212) 233-1250 ext. 346
Bradley.Eskridge@miacanalytics.com

Steve Harris
Managing Director, MIAC Capital Markets Group
(908) 400-2615
Steve.Harris@miacanalytics.com