Powell's Duration Trap, Banks and the US Treasury
Updated: Apr 10
April 10, 2023 | In the next Premium Service issue of The Institutional Risk Analyst to be published this Wednesday, we’ll provide a pre-earnings look at the top-five depositories – JPMorganChase (JPM) at $3.6 trillion in total assets, Bank of America (BAC) at $3 trillion, Citigroup (C) at $2.4 trillion on balance sheet, U.S. Bancorp (USB) at $670 billion and Wells Fargo & Co (WFC) at $1.8 trillion. But first we make a couple of general market comments for all of our readers about the state of the banking industry and the Fed itself after a year of tightening by the FOMC.
America’s largest bank, JPMorgan, actually shrank in size 2% in 2022, after growing average assets 10% in 2021 and 25% in 2020 during the “go big” period of QE. In that year, the Federal Reserve Board decided to double the size of its balance sheet without considering the now apparent down-side risks. As a result of the Fed’s ill-considered actions, several large banks have failed and hundreds of billions of dollars of private equity and debt have been destroyed. The total cost of QT to investors and also the US Treasury, however, is likely to be far larger as the tightening process continues.
Before the bank liquidity crisis in March of 2023, our expectation was that JPM would continue to shrink in Q1 2023, but now it's likely that the largest bank in the US actually grew in 2023. Even as the size of the money center banks increased during the period of massive bond purchases by the FOMC, the book value multiple of these stocks has declined, a grim testament to the negative impact of QE on banks and investors generally. Notice that subprime lender Citi has been punished by investors, while JPM and USB lead the group. In that sense, nothing has changed in five years.

Source: Yahoo, Bloomberg
A couple of general observations about the group are in order. First, the Fed’s quantitative easing (QE) policy seems to have retarded lending even as banks grew in size. As banks now shrink in the world of post-QE and higher short-term interest rates, it is unlikely that we shall see significant loan growth or spread expansion. Total loans and leases held by all US banks rose 7% in 2022, but loan portfolios actually fell over the past five years vs total assets.

Source: FDIC
In March 2023, the Fed expanded its balance sheet by several hundred billion dollars to accommodate the cash needs of scores of smaller depositories following the collapse of Silicon Valley Bank and Signature Bank, as shown in the chart below. This extension of credit to banks was far smaller than that provided in 2008-2009, however, and the system open market account (SOMA) is already starting to decline, at least in nominal terms.

Source: Board of Governors
Since the Fed’s balance sheet is theoretically running off around $90 billion per month, including $60 billion in Treasury debt and an uncertain amount of mortgage-backed securities, the Treasury will eventually need to issue new bills and notes to refinance these redemptions. Because prepayments on MBS generally are very low, the redemptions from the SOMA are running well-below the $35 billion cap. The Fed’s projections for portfolio runoff are as shown in the chart below. Notice that the MBS barely declines.

The green area shows MBS and CMBS held by the SOMA. The residential MBS have a weighted average coupon (WAC) around 3%, These same MBS have a weighted average maturity (WAM) in excess of 15 years or 5x the WAM at the time of issuance in 2020-2021. As a result, we think projections of the rate of runoff of MBS from the SOMA are still too optimistic. Actual compounded prepayment rates (CPRs) are running in low single digits vs the 6% CPR minimum assumed in most commercial prepayment models.
"There is absolutely no way that homeowners are going to give up those 3% mortgages that went into UMBS/GNMA 2% 30r bonds," notes reformed mortgage banker and entrepreneur Alan Boyce. "The Fed will have to actively sell the MBS at a big loss to make the QT goals. All the regional banks have the same issue, you will lose the money now by selling or over time via negative carry," says Boyce.
Boyce notes that there is almost 100bp spread between the WAC on the mortgage loan and the bond coupon. "Half is a forever annuity held by the GSEs and the rest is massively undervalued as a mortgage servicing right (MSR) held by a mortgage bank," he observes. "Ironically, Basle III treats the MSR as an intangible in the same category as tax loss carryforwards. MSRs trade at a lower CAP rate than commercial real estate! The Fed screwed the financial system, forcing them to stuff their balance sheets with low Risk Capital Weighted assets like MBS while eschewing the best cash flows."
Simply stated, US banks are caught in a vice between rising short-term interest rates and the Fed’s $16 plus trillion effective long duration position in Treasury debt and MBS. How can SOMA be approaching $20 trillion in effective, duration adjusted size when the Fed’s own data show a nominal value just shy of $9 trillion today? Because of the extension risk of the MBS, risk that now resides inside every mortgage portfolio in the US.
The mortgage bonds owned by the Fed, which had an effective average life of 2-3 years at the time of issuance during QE, are now closer to 20 years when measured against actual prepayment rates. CMBS, which are generally interest-only affairs, where principal is rarely repaid and refinancing is assumed, are also extremely sensitive to changes in interest rates.
The cool chart below from the Bloomberg terminal shows the index of modified duration of Ginnie Mae securities over the past five years. In one striking image, we can see the true idiocy of QE and now QT as formulated by the FOMC. The huge manipulation of the duration of mortgage exposures by the FOMC caused the failure of Silicon Valley Bank and has badly damaged dozens of other banks.

Source: Bloomberg
Notice that after the market break in December 2018, the FOMC under Powell started to flood the markets with reserves based on the untested idea that this would preclude further market mishaps. This judgment turned out to be badly wrong, however, and forced the FOMC to spawn various band aids such as Reverse Repurchase Agreements to keep the money markets from imploding.
The FOMC pushed down the effective duration of MBS from the LT average of 6 (think of duration as the average time required to recover principal) down to three by the end of 2019 via massive market purchases of Treasuries and MBS. With the onset of COVID, the FOMC doubled down, driving the effective duration of Ginnie Mae MBS further down to 1 from Q1 2020 through Q1 2021. In March of 2020, T-bill rates were effectively zero.
Between Q1 of 2020 and Q1 of 2021, the US mortgage market originated over $5 trillion in new, low coupon mortgages with an average duration of say 1.5 at the time of issue. That's 40% of the entire residential mortgage market in 12 months. Those 2% and 3% MBS are now trading on average durations closer to 6 with weighted average maturities closer to 15-20 years. The average index value for new production MBS, securities with 5.5-6% coupons, is now back to a duration of 6.
Let's consider the example of Silicon Valley Bank (SVB). At the end of 2019, SVB, has about 25% of assets invested in MBS with a duration of ~ 6 and a WAM of 5-6 years. By the end of 2020, SVB had taken its MBS position up to 33% of total assets, but half of the MBS portfolio had prepaid during that first big year of QE. Now the SVB MBS portfolio had a duration of ~ 3 and a WAM of perhaps three years.

Source: FFIEC
By the end of 2021, another 50% of the SVB mortgage portfolio had prepaid, but the bank's management had purchased even more, lower coupon MBS, to take the total position up to over 45% of total assets. This MBS position had sharply lower cash flows, duration now approaching 2 and a WAM of perhaps 2 years. The volatility of the SVB MBS portfolio had now doubled compared to the end of 2019 and the cost of hedging had likewise increased dramatically.
Q: Did the management of SVB realize that they had killed the bank a year before it actually failed? It seems not. Of note, the Federal Reserve Bank of New York states in a February 2022 staff paper:
“Because MBS pay fixed coupons to investors and typically have 30-year maturities, duration is high and prices are very sensitive to interest rates. A key distinguishing feature of MBS is that the duration of the security is not fixed but rather uncertain because borrowers can prepay their loans at any time.”
The problem now for banks and the Fed itself is that virtually no borrowers are prepaying COVID era mortgages. The MBS that were being priced off of the 5-year Treasury note in 2021 are now priced off of the 15-year portion of the Treasury yield curve and extending. And even as the rally in the 10-year helps the mark-to-market on Treasury bond portfolios, the prepayment behavior of low coupon MBS may not change very much. The 3% jumbo we have sitting in a Fannie Mae high balance pool vintage March 2020 is not going anywhere, thank you.
The table below illustrates the SOMA holdings as of April 3, 2023 and the crude adjustment by WGA LLC to illustrate the market impact of the Fed’s MBS and commercial mortgage holdings. Fed Chairman Jerome Powell told Congress in his recent testimony that the SOMA portfolio is running off, yet in duration adjusted terms, the SOMA is actually growing and is now 2x the notional amount.

Source: Board of Governors, WGA LLC