Balance at the overnight reverse repo program (RRP) at New York Fed has been rising very rapidly lately, especially in the past week (see chart below). The magnitude of its daily fluctuation is now large enough to affect the balance-sheet pressure at the big banks and consequently the price of risk assets. Thus, it is very useful to have a good understanding of the role that RRP plays in the financial plumbing system that convert fiscal/monetary policies into market price actions.
Because of the complexity of this subject matter, I am going to split the discussion into a three-article series:
- What RRP is and is not in the current regulatory regime? (the focus of this note)
- Short-term trading implication of the RRP balance fluctuation (bullish or bearish and timing)
- Similarities and differences between the current RRP surge and the 2019 Repo crisis. (longer-term implications)
The goal of this note is to establish a couple of extremely important (and often completely misunderstood) features of Fed RRP:
- Fed RRP is currently used almost exclusively by money market funds
- Banks can only shed unwanted reserves through independent money market funds.
The Purpose of RRP
The Fed started the current RRP back in late 2013 when prolonged QE3 finally caused an oversupply of bank reserve in the system. QE3 was tapered off gradually between Jan 2014 and Oct 2014. RRP is a temporary overnight asset swap, where Fed takes in bank reserves (or GSE "other account" balance), while posting collateral (UST) to RRP counterparties: Banks (commercial bank subsidiaries of bank holding companies), Primary dealers (dealer/broker subsidiaries of bank holding companies), GSEs (mostly FHLMC and FNMA), MMFs (government money market funds, independent or as a bank subsidiary). The counterparties get paid interest at RRP rate in the form of newly created bank reserve.
From Fed's perspective, RRP is somewhat akin to a temporary quantitative tightening: taking reserves away from banks and giving collaterals back to the private sector. And its strength/magnitude is determined by market participants (banks, GSEs and MMFs), giving the Fed a buffer to overrun QE without stressing the system, because the excess reserve would be absorbed by RRP in principle. This is important to the Fed, because it favors making very slow adjustment with months of forward guidance in order "not to spook the market".
From the perspective of RRP counter-parties, RRP is, first and foremost, a special collateralized savings account at NY Fed, through which they get paid at RRP rate if they could not find a better rate in the money market (i.e. 0-1Y bond market, loosely speaking). MMFs and GSEs use RRP for this purpose only.
Banks, however, use RRP differently. Banks are paid IOR on bank reserves. Since IOR has always been 10-15bps higher than RRP rates, banks have no incentive to use RRP for yields. Banks, especially their primary dealer arms, used to use RRP to acquire additional UST collaterals, which used to be profitable for eurodollar rehypothecation. However, this was completely changed in Jan 2018, when Basel III banking regulations were 100% implemented. New liquidity requirement (LCR) and capital requirements (SLR, TLAC, GSIB surcharges etc) made RRP unattractive because it is costly from both liquidity and balance-sheet perspective. As a result, banks' usage of RRP, either by their commercial banking arms (denoted as Banks) or by their dealer-broker arms (denoted as Primary Dealers), has dropped to nearly zero soon after January 2018, as shown in the chart below.
Only for a few days during the acute phase of the COVID crisis, primary dealers employed a few $Bn of RRP. Banks otherwise stayed away from RRP completely.
GSEs has also stayed away completely from RRP when the Fed cut RRP rate to 0 in March 2020, as shown by the chart below. Why? Both FHLMC and FNMA already have special NY Fed accounts (reported as "Other" on Fed Balance Sheet) that pays 0%. Why bother moving money from "Other" to RRP, which does not pay a penny more?
So that leaves the only regular users of RRP to be the MMFs. Recognizing this insight is very important, because the entire RRP balance can now be interpreted as an single-purpose entity. And that purpose is to shed bank reserves for the big banks, as Fed originally intended.
Confused? What do MMFs have to do with bank reserves? Can't banks directly use RRP to shed bank reserves? Let's go through a concrete example to illustrate this point.
MMFs – the Key to the Balance-sheet Magic of RRP
Let's consider a hypothetical situation when JPM needs to shed $10Bn of excess bank reserve, which came from one of its clients, AAL, receiving $10Bn stimulus money from the Treasury. I am using AAL as an example here, because non-operating business deposit is the most unwanted deposit type for big banks under the current regulatory framework on bank liquidity and capital requirement.
The Treasury's payment of $10Bn into AAL's business checking account at JPM would cause TGA to fall by $10Bn, and JPM's bank reserve to grown by $10Bn, at NYFed. JPM's balance sheet would grow by $10Bn as a result, $10Bn of new bank reserve on the asset side, and $10Bn of new deposit on the liability side. This new deposit is undesirable for JPM, considering the fact that $1 of reserve increase would displace roughly $20 of total-return swap (TRS) at the broker-dealer arm of JPM, under current SLR calculation. TRS generates a lot more fees than the 10bps from holding the reserve. Exceeding the SLR threshold will limit JPM's ability in share buybacks, dividend payouts and bonus for the executives. So JPM would be highly motivated to shed this extra $10Bn.
What if JPM put this $10Bn into Fed RRP? Well, because a RRP transaction is an asset swap, JPM will receive $10Bn worth of UST when shedding $10Bn of bank reserve. JPM's overall balance sheet size remains unchanged, and the $10Bn UST it receives has the same balance-sheet cost for JPM as the $10Bn bank reserve it sheds. And the return on the asset (0bps for RRP) is worse than bank reserve (IOR at 10bps).
What can JPM do in this situation? It could suggest AAL to move its deposit to an non-bank MMF that is eligible for Fed RRP, say Fidelity. The "suggestion" is delivered in the form of a hefty fee on all business checking account (e.g. fees equivalent to a negative interest rate of -17bps). AAL would be motivated to move its money to a MMF which currently pays an interest at around 1bps. For simplicity, let's assume Fidelity's cash deposit account is also with JPM (also subject to the high fee from JPM). the balance sheets now becomes the chart below: JPM still needs a way to shed the $10Bn reserve, and Fidelity now wants to find a new home for the $10Bn, because keeping it at JPM's checking account will cost Fidelity money.
Well, Fidelity has access to Fed's RRP, which pays 0bps, but far better than the hypothetical -17bps JPM is charging. Now, to be able to use the Fed RRP, fidelity must use its deposit account at a bank, because the Fed only accepts bank reserve as the payment, and only banks have bank reserve with the Fed. So now Fidelity uses its JPM checking account to pay for the RRP transaction, resulting in JPM shedding $10Bn in both bank reserve (on the asset side) and deposit (on the liability side). The biggest difference in this scenario from the previous Fed-JPM transaction is that JPM will pass along the $10Bn UST it receives to Fidelity to complete the transaction. At the end of the day, JPM sheds the bank reserve successfully, while Fidelity carries the extra $10Bn on its balance sheet. Since Fidelity is not a bank and thus not subject to the SLR limit, there is no balance sheet cost for it to carry the extra $10Bn asset. Note that an MMF independent of the bank is absolutely necessary for JPM to shed the unwanted reserve. Had a JPM MMF been used instead, the extra $10Bn asset is still on the book of JPM (the bank holding company), and no balance sheet reduction is realized.
Banks' need for independent MMFs in order to shed reserve can also explain the dominance of MMFs' share of RRP usage between 2014 and 2017, as shown in Chart 2 (near the beginning of the note). Most of the RRP usage at the time, especially near quarter ends, has been recognized as balance-sheet window-dressing by Eurobanks, i.e. shedding reserves to reduce banks' balance-sheet for regulatory purposes. That practice stopped in 2018, when bank reserves became both needed for LCR purposes and scarce from the Fed quantitative tightening program.
One may ask here why would Fidelity choose to use the Fed RRP, which is paying 0bps? Why can't Fidelity invest in T-bills instead for higher yield? Of course, Fidelity would buy T-bills instead, if there were T-bills yielding better than 0bps available in the market. And it was indeed the case before Q2 2021. However, since the beginning of April 2021, in each week, Treasury has been letting ~$39Bn worth of T-bills (specifically, cash management bills) to mature to lower the TGA balance. T-bill has been scarce, in comparison to the large deposit balances MMFs are carrying, as a result of QE (which removes UST from the market) and Treasury's effort of replacing T-bills with longer duration T-notes and T-bonds. Fed RRP is now the highest yielding vehicle for many MMFs, and that's why RRP balance has been edging up every week lately.
Now, since the highest-yielding investment vehicle MMFs can find on the margin is the RRP with 0 interest rate, and MMFs cannot "break the buck". MMFs would not accept unlimited amount of deposit, even if there were no cap on RRP balance for each counterparty. There is a size limit for MMFs before they start to lose money from paying their clients at 1bps while receiving interest income at 0bps. In other words, the realistic capacity of RRP to absorb excess bank reserve is finite in the current 0-RRP-rate condition. Once MMFs are stressed from either hitting the RRP cap ($80Bn per counterparty) or bleeding too much money, no more bank reserve can be shed via this route.
It seems that we are getting pretty close to that point for some funds. Fidelity's FDRXX has already got $70Bn worth of RRP on its book on April 30th, when total RRP balance was at $183Bn, as opposed to a total of $351Bn on May 20th.
To fix this, the Fed could raise RRP rate, even 3-5 bps would motivate and help MMFs to put more money into RRP, thereby reducing the stress on banks' balance sheets. But I suspect that maybe the current quandary is exactly what the Fed intended: cranking up inflation pressure via MBS QE, while keeping financial speculation in check by throttling the use of leverage by speculators via banks' balance sheet pressure.
That said, for us traders, let's enjoy this new trading regime in which the price action in equity market is closely driven by the Fed/Treasury/GSE cash flow and the reactive MMF flow (rising RRP is bullish for stocks because it lowers banks reserve footprint, vacating precious balance sheet space for total-return swaps and other derivatives that enable leveraged speculative positions). This regime should last at least into the next Fed meeting (June 15-16th).
I will discuss how to interpret RRP balance changes in the context of my liquidity projection, specific trading setups, and timing issue in the next installment.