On Monday, the Federal Reserve formally stopped shrinking its balance sheet. Analysts think it’s going to expand again — and probably very soon.
The reason is that US money markets have remained choppy even after Fed announced it would stop the process of “quantitative tightening”, having shedded c. $2.4tn of balance sheet assets since its nearly-$9tn peak in 2022. QT has the effect of draining money from the US financial system.
The clearest sign that money is now a little uncomfortably tight is in the repo market, where rates have been unsettlingly volatile lately. Given the global systemic importance of the $12tn US repo market, that is not something the Fed is likely to tolerate for long.
Marco Casiraghi and Krishna Guha at Evercore ISI reckon the Fed will therefore announce the start of “reserve management purchases” at its meeting next week:
Money markets continue to signal the need for the Fed to advance to the next phase of balance sheet normalization and start reserve management purchases (RMPs) . . . Reference repo rates outside of the target range risk weakening monetary policy transmission and point to the need for RMPs to begin soon.
We think the Fed will start buying $35bn of T-bills per month in Jan to grow the balance sheet by about $20bn a month taking into account roughly $15bn of maturing MBS. We also think there is a good likelihood the Fed will need to buy an extra $100bn to $150bn of T-bills on a one-time basis in Q1 to nudge back up the ratio of reserves to bank deposits.
Hang on, what are RMPs? Is this just QE redux?
Good question. Alphaville is sure that we’ll see lots of shrill headlines conflating the two over the coming weeks and months, but the quantitative easing programmes of yore and the coming reserve management purchases DO actually differ — in substance, intent and impact.
Firstly and most importantly, while QE involved buying long-term bonds, RMPs only involve buying short-term government debts known as bills. Bonds typically mature in 2-30 years, while bills are like tradeable short-term IOUs.
The original purpose of QE was to spray money at the financial system back in 2008 and quell the inferno that was raging at the time. It then morphed into a broader tool to stimulate economic growth at a time when interest rates were already floored. The Fed did so by buying Treasuries and US-guaranteed mortgage-backed securities with reserves that it simply created, hoping to drive down government bond yields.
Despite being relatively controversial — to this day many economists disagree what the effect was, and even exactly how QE actually works — it was dusted off again and supercharged when Covid-19 struck.
When inflation later made a comeback, the Fed had to quickly reverse course and started jacking up rates and shrinking its balance sheet. This involves the opposite operation, either selling or letting a certain amount of the bonds it has bought mature, and contracts the amount of reserves in the financial system.
However, the financial piping nowadays needs a certain amount of money sloshing around to make sure things operate smoothly. Since last autumn there have been mounting signs that maybe what was once considered an “abundant” level of reserves was now becoming merely “ample” — and could quickly become, well, uncomfortable. Or “scarce” in central banker argot.
Here’s a chart showing the Secured Overnight Interest Rate (a broad measure of overnight secured borrowing costs) and the Tri-Party General Collateral Rate (a slightly narrower measure of one segment of the repo market) versus the corridor formed by the Fed’s two main interest rate tools.
As you can see, both SOFR and TGCR have occasionally jumped out of the corridor for a year now, but only around the end of quarters when money is usually tight for a variety of fiddly reasons.
However, the jumps have become more violent and lasting of late, which has made a lot of finance people a tad skittish. As the cliché goes, you really don’t want the plumbing to malfunction.
Is buying debt really the only thing the Fed can do?
The US central bank has pointed to one of its more newish tools — the Standing Repo Facility — as a backstop that will contain the repo market’s volatility and prevent accidents from happening.
The SFR is certainly a powerful facility, and designed to contain the kind of systemic mayhem we saw in March 2020. But it hasn’t been able to prevent repo market spikes, likely as there seems to be a stigma attached to its use in peacetime.
JPMorgan’s analysts suggest that tweaks could make it more effective, but therefore also believes the Fed will eventually have to start buying at least some Treasury bills to reflate its balance sheet. Here’s what the bank’s rate analysts Phoebe White and Molly Herckis wrote in their 2026 outlook:
. . . . The SRF remains an important stabilizer. However, despite increased usage, the facility has not been effective in policing rates on the upside. So, what can be done?
The Fed could consider changing the SRF to be offered on a continuous basis as opposed to via auctions twice a day. Doing so would allow banks to have access to reserves “on demand.” Lowering the SRF rate by 5bp could also incentivize greater participation, push overnight rates closer to the middle of the Fed funds corridor, and further distinguish the SRF as an administered facility for money market control, rather than a backstop akin to the discount window. Centrally clearing SRF operations could also boost usage, as was noted in the October meeting minutes, though implementation would take time.
Ultimately, the Fed will likely need to add assets to its balance sheet. We believe reserve management purchases will begin in January 2026, with the Fed buying roughly $8bn/month of T-bills to keep up with the pace of growth in currency.
The goal of these RMPs would not be to bring down government borrowing costs or stimulate economic growth, as was the aim of QE. Rather, the point is to ensure sufficient liquidity in the bowels of the financial system to prevent any messy accidents from happening.
This is why RMPs involve buying Treasury bills rather than bonds — the impact on markets should be more neutral. As the NY Fed’s president John Williams stressed at a conference a month ago (with Alphaville’s emphasis in bold):
Such reserve management purchases will represent the natural next stage of the implementation of the FOMC’s ample reserves strategy and in no way represent a change in the underlying stance of monetary policy.
He’s not wrong but . . .
Williams was guarded about when this begins (and how big the RMP programme might be) but hinted that he personally thought it might be started soon. With quantitative tightening now formally over as of Monday, it makes sense that the Fed starts sketching out the balance sheet reflation when it meets next week.
Of course, there will be plenty of other things to thrash out, given the growing divide between policymakers on what to do with interest rates. It’s therefore conceivable that the Fed punts any balance sheet decisions and instead elects some simple tweaks to its toolbox to try to dampen the repo market volatility.
But Bank of America’s Mark Cabana also reckons the Fed will have to say something, and begin RMPs as early as January 1. The real questions is how much they will do, he argued in a recent note. After all, funding markets already seem to be pretty tight — so some “backfilling” of money might be appropriate:
On top of MBS prepayment reinvestment into bills, we estimate that the Fed will buy $45b/mo in bills for natural balance sheet growth and to backfill some of the reserve over drain. Fed bill buying size of <=$30b/m would likely signal no Fed backfill interest & be negative for spreads. Fed total bill buying of >=$40b/m would suggest desire to RMP backfill, better anchor repo in target range, & be spread positive.
Evercore agrees that the Fed will/should aim to quickly ensure enough liquidity with an “extra” $100-150bn of Treasury bill purchases in the first quarter, on top of a steady $20bn a month bill buying programme for the foreseeable — on top of ca $10-20bn a month of maturing MBS, which will be shovelled into bills as well. Goldman Sachs is silent on any backfill, but also predicts about $20bn of net RMPs a month.
Evercore, Goldman and BofA estimates are notably higher than JPMorgan’s, and it’s not entirely clear why they are so different when their analyses of the situation are broadly similar. But there seems to be broad consensus across Wall Street that the new era of QE, sorry, RMP will start in January.
However, given the stubbornness of the recent repo market volatility, Alphaville would be surprised if the Fed doesn’t unveil a dose of “temporary open market operations” before then — just to make sure Christmas isn’t ruined by an end of year funding crunch shitshow.
