The failure of Silicon Valley Bank could have a lasting effect on markets for mortgage-backed securities. And not the ones that investors are most worried about.
In fact, the safest corners of mortgage markets could end up with permanently lower valuations, Morgan Stanley analysts write in a note this week. Their argument is an interesting one, because it tells us lots about both agency MBS markets and the way banks manage their balance sheets.
After the run on SVB, banks will probably need to reduce their holdings of mortgage-backed securities that are guaranteed by the likes of Fannie Mae and Freddie Mac. That’s because those bonds have high duration, and experience sizeable losses when interest rates rise.
This bodes poorly for agency MBS markets — mortgages on single-family homes and multi-family apartment buildings, basically — where banks are massive players. (The line “I’m not long your house anymore” doesn’t have the same ring as “I’m short your house”, but the change will affect markets nonetheless.)
From Morgan Stanley, with our emphasis:
Banks owned over 1/3 of the single-family agency MBS market and over half of the multi-family agency CMBS market as of 4Q22, so any changes to their demand function would clearly impact both short-term and longer-term spread levels. While we don’t know exactly what the changes will be, we have relatively high confidence that the events of the past few weeks will result in equilibrium spread levels on mortgages to be biased wider in perpetuity, and we move to underweight MBS (from neutral MBS) . . .
For context, the analysts provide a more granular look at banks’ presence in agency mortgage-backed security markets in the helpful charts below. (TL;DR they are indeed big.)
“But hey Alphaville,” you may be asking, “investors expect the Federal Reserve to stop raising rates this year, right? What happens when it cuts rates? Wouldn’t banks want to own agency MBS then?”
Good question! Whenever the Fed does eventually cut rates, agency MBS will appreciate in value, and be more attractive to own. But banks’ investment portfolios reflect more than their predictions about market returns.
Bank regulations, and the make-up of banks’ liabilities, are equally if not more important. Morgan Stanley points out that normal bank deposits — which don’t earn interest — are usually assigned a seven-year duration. The length of that time period sounds odd given everything that’s happened since early March, to be sure. But the thinking was that those deposits are used mainly for transactions, payroll and other daily expenses, making it disruptive and costly to change banks.
Still, after the run on SVB, it seems . . . increasingly unwise to rely on that seven-year duration. From Morgan Stanley’s bank equity analyst Betsy Graseck:
Expect banks will need to prove out or reduce the duration assumptions for [non-interest-bearing deposits]. Banks will likely need to review and re-justify their duration assumptions for [those deposits], and potentially a wider range of deposit types.
At a minimum, we think banks should be allowed to assign a long duration to the inflows and outflows of a transaction banking account, and the incremental deposit balance required to pay for these services. Incremental balances over that could come under more scrutiny and we would not be surprised if banks have to incorporate a higher tail risk of deposit flight into their duration assumptions on these incremental deposit balances.
In other words, banks won’t be able to treat cash as an equally reliable source of funding if it’s just sitting there without being used (ie, if the cash isn’t part of the steady inflows and outflows of day-to-day business).
That means banks will need to reduce the duration of their assets as well, Graseck writes:
Reducing the duration assumptions of non-interest-bearing accounts will directly reduce how much asset duration banks can take. Especially post-SIVB, we expect regulators to scrutinise how banks are assessing and managing their asset-liability gap. Shorter-duration liabilities will directly translate to increasing shorter-duration assets like cash and shorter-term Treasuries.
And how do banks manage their duration levels?
Well, agency MBS is an especially popular choice! That’s because of the market’s low credit risk, and as a result, its favourable treatment by regulators.
So Morgan Stanley’s fixed-income strategists — Jay Bacow and Zuri Zhao — do some back-of-the-envelope math on what lower deposit duration, and therefore lower asset duration, could mean for agency mortgages:
. . . every year of duration by which the banking industry lowers the [non-interest-bearing deposit] duration [assumptions] would equate to a $450bn reduction in the industry’s mortgage holdings, which we calculate using $4.7tr * 1yr duration diff *60% MBS haircut/6yr MBS duration.
Clearly, there are many assumptions in this analysis, but we want to provide some context to the potential demand shift just on the asset/liability side — specifically, banks net added $936bn of mortgages during 2020 and 2021.
If this does occur, it’ll be the first time since at least the GFC that banks and the Fed simultaneously reduce their holdings of agency MBS (remember the Fed is shrinking its balance sheet as well):
On the bright side, banks probably won’t be actively selling their agency MBS. Most of their holdings are in their held-to-maturity portfolios, and selling securities from those portfolios would force a bank to mark the whole thing to market:
We also note that many banks won’t be able to or do not want to actively reduce holdings — for instance, as of 4Q22, GSIBs owned $1.55tr in mortgages of mortgages, but only $253bn were in AFS portfolios. Furthermore, selling formalises losses, which flow through to earnings per share for all banks, and through capital for the banks that currently have the AOCI exemption (below $700bn in assets). In effect, we think it is likely that banks won’t reinvest proceeds and will simply allow their mortgage portfolios to wind down through amortisation and prepays.
The question is, then, how quickly banks will reduce the size of their agency MBS portfolios by letting bonds mature without reinvesting.
From the strategists, with our emphasis:
If banks allow their portfolios to run-off completely, the significant supply/demand mismatch results in equilibrium spread levels on mortgages that are likely to be about 25-30bp wider than they previously averaged . . .
Should their run-off be tempered, perhaps on the order of $5bn/month, then we’d expect equilibrium spread levels to be closer to 5-10bp wider than previously averaged.
In 2023, we expect banks to reduce their MBS holdings by approximately $5-10bn per month, and on top that, we see an additional reduction of $100bn from bank selling out of receivership. In 2024, we expect the pace of reduction to increase as we get closer to the regulatory changes.
Putting it all together, we think that it’s likely that equilibrium levels for the current coupon are likely to be about 15bp wider than previously averaged.
The strategists expect money managers to step up, and buy in size:
. . . we’re looking at over a trillion dollars of mortgages that need to get bought by money managers, overseas, and REITs over the next two years, by our estimates. If money managers buy 2/3 of that, that’s $650bn they would need to add on top of their roughly $1,500bn holdings they had at the beginning of the year.
But hey, what’s $650bn between friends?
