Federal banking regulators have proposed changes to the capital treatment of mortgage servicing assets (MSA) One rationale for this change is to increase mortgage origination and servicing activity by banks, who have lost market share to nonbank entities. What are mortgage servicing assets and what are their risks? And are these proposed changes likely to have a major impact?
What is a mortgage servicing asset?
When a bank sells a mortgage but retains the servicing rights, it creates a mortgage servicing asset (MSA.) More about the “create” part later. Servicers collect a fee, normally a percentage of the mortgage’s principal balance. The MSA equals the present value of these fees, plus other sources of servicing income, minus servicing costs.
What makes MSAs risky and complex?
MSAs have prepayment risk (on steroids). The value of the servicing depends in large part on the life of the mortgage. That life isn’t fixed but instead depends on the borrower’s willingness and ability to refinance or otherwise prepay the mortgage.
Contrast the MSA to the mortgage itself. Suppose you expect the mortgage to have an average life of seven years, but it prepays tomorrow. If you are carrying the loan at par, it has no immediate effect (though it may create reinvestment risk). If the loan is at a discount (say 97), you record a gain. If the loan is at a premium, you record a loss. Most mortgages are carried within a few percentage points of par, so the gain or loss is usually quite modest.
Prepayment of an MSA is quite a different story. Say you have a $100,000 mortgage and assume a value of the MSA of $2,000 based on a seven-year average life. If the mortgage prepays tomorrow, the MSA is worth zero. Zip. Zilch. You don’t just lose the premium; you lose the entire asset.
The valuation of MSAs depends primarily on a model rather than price quotes. Purchases and sales of MSAs are infrequent and often not publicly disclosed. MSA models hinge on assumptions related to mortgage prepayments, servicing costs, and ancillary income like late fees. They also make estimates of income from float on principal and interest and on tax and insurance escrows. The estimates further assume the required return on the MSA, in the form of a discount rate (or option-adjusted spread (OAS) for more sophisticated models). Most banks get third party appraisals to back up their estimates, but these valuations primarily compare the appraiser’s model to the bank’s. Appraisers may also incorporate market intelligence. But the dearth of actual transactions makes it difficult to calibrate the model’s assumptions to market data.
MSAs are marked-to-model from day one. Until the mid 1990s, most MSAs resulted from a purchase (purchased mortgage servicing rights or PMSR). Other servicing activity remained off balance sheet. That changed in 1996 when the Financial Account Standards Board (FASB) adopted FAS 140, which required assigning a value to retained servicing when recording a gain on sale of mortgages.
Say you originate $100 million in mortgage loans and sell them for the same amount. That results in zero gain or loss, right? Wrong! If you retained the servicing rights, then the cost basis for the mortgages also reflects the assumed value of the servicing. Say you estimate the servicing is worth 2% of the principal balance. Then the cost basis for the mortgage would equal 100/102, or 98.4% its face value. And you’ve just recorded a gain on sale of $1.6 million. But the $2 million you assign to the servicing is just an assumption. That’s much different from paying $2 million for the asset in an arm’s length transaction.
Hedging MSAs is tricky. MSAs usually have a negative duration. Unlike most bonds, an MSA’s value usually goes up when rates rise but goes down when rates fall. This sensitivity is not symmetrical. Like many mortgage-related assets, MSAs typically have negative convexity, meaning their downside is usually considerably more than their upside. This pricing profile, as well as other characteristics can make them difficult to hedge.
Most larger banks hedge MSAs, but that hedging trades off precision and market liquidity. Mortgage TBAs have lots of liquidity but hedging with TBAs also doubles down on negative convexity. Swaptions usually have decent liquidity and can mirror the embedded options of MSAs. But they also depend on accurate prepayment estimates, both in the current market and as rates change. Prepayment models are more mature than, say, deposit models. But they can and have been inaccurate at times, with the biggest misses occurring during turning points in the market. Mortgage Interest IOs closely mimic the cash flows of MSAs, so systematic model errors will largely offset each other.[1] However, the market for these instruments isn’t very liquid, making them less useful.
A Real-Life Case
MSA losses are not merely theoretical. In 2018, Citi’s exited from the mortgage servicing business, which included the sale of most of its MSA portfolio. Citi recognized a pretax loss of $331 million ($382 million if including other operating costs) while de-recognizing $1.046 billion in MSAs. That translates to a loss rate of 31.64%. And this was for a bank that had marked its MSA to market quarterly and actively hedged.
MSAs and Model Risk Management
The banking agencies recently issued revised guidance on model risk management. The revised guidance takes a decidedly more high-level approach, focusing on risk management principles rather than specific applications. Regulators cut the guidance’s length from twenty-one single-spaced pages to just twelve double-spaced pages. Moreover, the OCC rescinded the Model Risk Management section of the Comptroller’s Handbook and even disappeared the rescinded section from its website. No explanation was provided.
Since both the initial and ongoing valuations of MSAs as well as their risk management depend on models, examiners have historically placed much of their focus on the quality of those models. Regulators now seem to feel that examiners shouldn’t look too closely at a bank’s models and prefer they adopt a bobblehead doll approach. (Keep nodding.)
MSAs and Capital
MSAs receive draconian treatment under capital regulations. This is hardly surprising considering of the complexity and risk of these assets. MSAs receive a risk weight of 250%. For context, regulators assign risk weights to delinquent residential mortgages and other past due loans at only 100% and 150%, respectively. Current regulations also limit MSAs to 25% of common equity tier 1 (CET1) capital. For a handful of advance approaches banks, the limit is only 10% of CET1.
Banking regulators have proposed changes to the treatment of MSAs. The proposal would retain the 250% risk weight but would remove limits on MSAs as a percentage of capital. Supporters of this proposal have expressed concern over the migration of mortgage servicing to the nonbank sector. They maintain that these changes can reverse this trend. But how likely is that?
The table (below) shows MSAs as a percentage of common equity tier 1 capital (CET1) at large banks with over $1 billion in MSAs. Each bank is well below the current CET1 limit. In fact, MSAs could more than double and each bank would remain below current thresholds. Removal of the threshold is unlikely to unleash mortgage servicing activity since banks already have room to grow MSAs under current rules. MSA exposure can change significantly over time, however. As recently as 2016, Wells Fargo carried $16.2 billion in MSAs.
Removing the caps on MSAs won’t make servicing activity suddenly attractive. Although capital requirements can be fungible, the 250% risk weight remains a significant disincentive. There are also substantial costs with building out a large servicing platform. And banks may be more than a little gun shy following the Robo-Signing Scandal of the early 2010s. Although regulators now appear hellbent on banishing the phrase “reputation risk” from the lexicon, robo-signing certainly gave the banking industry a black eye. In more practical terms, five large banks wound up having to pay out $26 billion in relief to distressed homeowners.
Concluding Thoughts
The proposed capital changes are unlikely to bring mortgage servicing back to the banking sector in a meaningful way, but could they do significant harm? Perhaps. Excess concentrations have contributed to countless financial collapses. Risk-based capital requirements, as well as the other risks described above are likely to deter most banks from jumping into the mortgage servicing business in a big way. But failed banks are usually outliers. Concentration limits can rein in these outliers. While a relaxation of current MSA limits appears reasonable, eliminating those limits entirely (for the first time since the 1980s) goes too far. It’s just asking for trouble.
[1] Model risk in this case is limited to differences between prepayments on the specific pool of IOs and prepayments on the bank’s specific MSAs.
Leave a Reply