Right now, big banks in the US are approaching their SLR-limited balance sheet size. The diagram below shows a simplified picture of how a sudden and excessive inflow of the Fed liquidity (i.e. bank reserves) into these US big banks (over which they have no control in timing and magnitude) may impact the price of risk assets.
Now, let's take a closer look at each pathway individually.
Crowding out by bank reserves
First, why a sudden and excessive inflow of bank reserve can sometimes crowd out other assets on banks' balance sheets? This is largely an unintentionally consequence of the supplementary leverage ratio rule in the Basel III accord, which aims to prevent those too-big-to-fail banks (GSIBs) from repeating their mistake in 2008 GFC of over-leveraging and loading their books with toxic assets. SLR rule stipulates that GSIB banks' balance sheet cannot exceed 20x of their tier-1 capital. Otherwise, they will be limited in their capability in issuing dividends, share buybacks and discretionary bonuses. Together, 5 of the 8 US GSIB banks currently provide more than 50% market share of prime brokerage service to hedge funds and family offices (see table below). Now all of them are operating fairly close to their SLR-limit, as the 3rd column of the table below shows their SLR on Jan 1st 2021 with the recent SLR exemption removed.
2021 looks to be an exceptional challenging year for the US GSIBs to keep the size of their balance sheets in check: the US government is running about $3Tn deficit in 2021, which adds around $1.4Tn bank reserves (from TGA) to the banks, and an additional $400-$900Bn UST to US banks' balance sheet. Fed's ongoing QE4 will further add $1440Bn bank reserves to the system. European and swiss banks have already breached their SLR ratio (red numbers in the table above), thus most of the burden will fall on the shoulders of US banks.
Now, with a relatively fixed upper bound for a GSIB (since it usually wants to pay dividend, conduct share buyback and pay bonuses to management/employees), when one asset class (bank reserve) is injected (with no control in timing and magnitude from the bank) by the Treasury or Fannie/Freddie via the Fed, the bank will have to trim or at least freeze the provision to other asset classes, such as UST (in its commercial bank subsidiary) or total return swaps and other derivative (in its broker/dealer subsidiary), as illustrated by the JPM example below. Josh Younger (JPM) has a good discussion on this issue in this BBG Odd Lots podcast (at 54:00 mark)
JPM has a typical balance sheet for a big bank with a significant prime brokerage subsidiary. As can be seen in its 2020 Q4 10k filing below, its derivative/repo/off-balance-sheet (FX basis swap etc) book contributed to a significant portion of its overall "total leverage exposure" (i.e. balance sheet). With the $682Bn bank reserve (referred as FRB deposits in the table below) and UST exemption removed now, plus the flood of bank reserve from TGA-unwind and additional QE that has happened and will happen in 2021, JPM's total leverage exposure is approaching its SLR limit very soon.
One objection that people may have is that banks can simply raise more capital to accommodate the flood of bank reserve to avert the crowding-out effect. However, bank reserve is only paying 10 bps (per current Fed IOR), but JPM needs to pay 17bps FDIC surcharge for small clients. Even levered at 20x, bank reserve pay at most 2% on the new capital that they can raise, which greatly dilutes their earning-per-share. Understandably, banks are not motivated to raise new capital in this environment. Additionally, banks have been put on a pause for dividend and share-buybacks since March 2020, and there is a lot shareholder pressure on the banks to resume their dividend and share-buyback program in 2021, which would further reduce their tier-1 capital.
Why are UST and derivatives are particularly prone to the crowding-out effects? Given its safety-asset status, UST is the 2nd lowest-margin asset a bank usually holds on its book (bank reserves being the lowest-margin). Naturally, banks would prefer to hold more profitable assets and shed UST before other assets. Derivatives are also prone to the crowding-out effect, as its exposure is proportional to the notional value of its underlying. For example, equity TRS' PFE (in the diagram below) is around 5-10% of its notional value, even if the TRS book is netted by banks' equity holding to have zero-delta exposure. Although conducting market-making of these derivatives is generally lucrative to banks, their balance sheet cost is disproportional large. When overall balance sheet space is squeezed by a sudden influx of reserves, the balance-sheet provision for derivatives may be frozen or even trimmed for a few days, until banks find a way to transfer the reserves to other banks who have more balance sheet space.
Since a large part of the stock market rally in the past 12 months has been fueled by the low-cost leverage from hedge funds and family offices using derivatives (TRS, options and futures) provided by major prime brokers, once the balance-sheet provision (for more derivatives) is frozen, the buying power of these hedge funds is effectively removed from the market, thereby setting up a situation where buyers are AWOL and a sell-off can ensue with a proper catalyst. Once 2-3 major prime brokers get affected/frozen by their balance-sheet constraints, the market becomes susceptible to a sell-off once sellers appear.
Over the long run, banks are profit-seeking and derivative contracts generate lucrative fees for them. They are generally highly motivated to shed those unexpected reserves to other banks that still has balance-sheet space. These banks with balance-sheet space include smaller/regional US banks (with a looser SLR requirement), UK and Japanese banks (for which reserve is exempted from SLR calculation). They can also shed reserve via prime money-market funds that have access to the Fed Reverse Repo Program. But these maneuvers take at least 1-3 days to complete/settle, which means the market dislocation can last for 2-3 days, before the eventual rebound.
Transmission via US Treasury notes/bonds
The shedding of (and the aversion of acquiring new) UST is more long-lasting than the temporary crowding-out effects on the derivatives, because current yields of UST makes it the second lowest yielding assets that a bank can hold. This aversion to UST is likely to continue, and pushes rates higher in the coming months, as more and more reserve is added into the system. Consequently rate-sensitive names (tech stocks and zero-earning bubble stocks) can be affected in a dramatic way. This dynamics occurred in the late February, and can repeat itself in late May-June and late-August-September. Rising rates also tend to elevate USD vs other major currencies, generating additional headwind for stocks.
Transmission via USD FX market
Over the timeframes of weeks, US banks would try to transfer their bank reserves to their foreign counterparts (e.g. by persuading their clients to move some of their deposits/money market funds to foreign banks via imposing fees). Two episodes of such transfer are highlighted by the arrows in the chart below, when the bank reserves of US banks have dropped disproportionally to those of foreign banks. This process is slow, because European and swiss banks have already breached their SLR limit and foreign central banks are conducting their own QEs as well, which increases the bank reserves of those foreign banks in their local currency. Nevertheless, this process continues on, and is likely to create downward pressure on USD, which in turns provides a small but steady tailwind for equities (stocks and DXY have been stay anti-correlated more often than not in the past year).
This pathway further contributes to the rebound of stocks after the initial sell-off caused by an acute "indigestion" of oversized bank reserve injection from the treasury or Fannie/Freddie.
Implications for trading
- SICO-induced sell-off are short-lived (3-5 days at most), and the dip will be bought soon after. As a result, I have been maintaining a long bias on stocks, and hedge my delta exposure prior to a potential SICO episode via relatively-cheap index puts. Timing of entries and exits is quite important to the eventual PnL.
- Effects of large liquidity withdrawal (drawdowns in bank reserve balance) is completely opposite to large liquidity injection (i.e. bullish for risk assets). Large liquidity withdrawal can occur on a major tax day or a Fannie/Freddie payment day, both of which are bullish for equity.
- Small liquidity injection (<$10Bn/day) is usually digestible for banks, and is bullish for risk assets
- A standalone Medium-sized liquidity injection ($40-60Bn/day) requires additional external catalysts for a sell-off to materialize.
- Large liquidity injection ($100+Bn/day) or persistent multi-day medium-sized liquidity injection can often create 1-2 days of sell-off by themselves. An external catalyst and negative option dealer gamma can sometimes creates large single-day sell-offs (3-4% days).
- The sell-off and the subsequent bounce are more pronounced in sectors that are more sensitive to interest rates (especially when rates are trending in the same direction of liquidity flow)
- Liquidity-induced sell-offs and rallies are enhanced by negative gamma and dampened by positive gamma from the positioning of option dealers.
- Fed liquidity affects not only equity market, but also the rates market over a longer timeframe (3-7 days) and the FX market over 5-15 days. Time-averaged liquidity flow (e.g. 5D SMA of the change in bank reserve ) is more appropriate for those markets.