How Kevin Warsh could put the toothpaste back into the tube at the US Fed

We were driven to this point through a series of significant builds in the balance sheet as the Fed battled the Great Financial Crisis (GFC) and pandemic events through aggressive buying of bonds. Getting here was in a sense tolerably positive, as it involved the unleashing of an extensive comfort blanket over the system. It was a means to ease pain. Going back is not quite guaranteed pain, but would be managed as a damage limitation exercise.

But let’s be constructive and at least examine the mechanics. First, on the theory that the unwinding of the balance sheet could impliedly mean a return to a pre-GFC-type balance sheet, let's go back in time and view how things were then. There are three dimensions to consider: the Fed’s balance sheet, how the funds rate was set, and crucially, system plumbing. Let’s talk balance sheet first.

At the end of 2005, the Fed’s balance sheet was about 5.5% of GDP. Roll on 20 years, in and out of the GFC and pandemic, and it’s now 21% of GDP (quadruple). The driver was bond buying. Total bonds held by the Fed were around 5.5% of GDP 20 years ago. That now equates to 20% of GDP (almost quadruple). Bank reserves were purely regulatory in nature and a puny 0.1% of GDP 20 years ago. Today’s bank reserves are closer to 10% of GDP (100-fold).

Let’s continue with how the funds rate is set. A key change element here is the 100-fold increase in bank reserves – that's the key delta the Fed has had to grapple with.

So, what are bank reserves exactly? They are balances owned by the commercial banks, but sitting at the Federal Reserve. And why are they sitting at the Fed? Pre-GFC they were purely regulatory in nature. There was a tiering of rates, but system-wide banks held up to the equivalent of 2% of their deposits at the Fed, as a type of regulatory security balance. And as banks received a zero percent rate on those (regulatory) reserves, there was no reason to post more than the bare minimum.

Now here is the important point – where did the reserves come from? Well, banks either had vault cash on hand, or they had to borrow reserves. They could not use customer deposits, for example, as that is not theirs to use in that way. So they either borrowed from other banks that had excess reserves, or they borrowed from the Fed. And at what rate? At the Fed funds rate. And that’s how the system worked back then. This was a low reserves rate-setting environment, where the Fed would set the funds rate, and then adjust the reserves balance on a daily basis so that it cleared at that rate.

Then we had a number of big bangs; huge quantitative easing shocks to the system. Before we knew it, bank reserves had gone through the roof, as a consequence of the very simple transaction whereby the Fed bought bonds from banks, the proceeds of which got posted as bank reserves. In order to incentivise this process, the Fed had to remunerate the holding of reserves (as why otherwise would banks hold such reserves). The rate that the Fed then paid on those reserves effectively became the dominant rate – the rate paid on excess reserves.

Now, instead of the funds rate being set through supply and demand for (scarce) reserves among banks, it’s now being driven by the rate the Fed decides to pay on those reserves. There is a world where the Fed could in fact have chosen to employ the rate on excess reserves as the key policy rate (replacing the funds rate). But that did not happen, as the Fed preferred to stick with the funds rate given its insertion in numerous contracts, and its legacy status. So instead, it set a funds rate target range of 25bp around the rate on excess reserves.

So how is the effective funds rate set today? It’s effectively guided into place by the Fed, so that it's within the aforementioned 25bp corridor. This is done with specific reference to the various players operational in the market for liquidity. Banks that post reserves will clearly not lend at rates below the rate paid on reserves, so that sets something of a ceiling. The floor is then framed by the rate the Fed pays non-banks that have a need to post excess liquidity – the Fed facilitates this through its reverse repo rate. That rate represents the floor of the 25bp range.

The other competition for avenues for market liquidity is repo, as encapsulated by the SOFR rate (basically an amalgamation of repo rates). Different players in the market with varying rights in the reserves bucket or the reverse repo bucket will see SOFR as an alternative rate that gets deployed from a relative value perspective. This is important for framing where the effective fund rate actually sits. SOFR theoretically should trade below the effective funds rate, as SOFR is a collateralised rate. So there is a whole relative value trade going on in this space.

And in the funds market, we also need to account for the Federal Home Loan Banks. They provide funds to typically smaller banks. The Federal Home Loan Banks cannot post in the Fed’s reserves bucket, but they can post at the Fed funds rate. In fact, they tend to be dominant players in the current funds rate market, especially as the commercial banks will prefer to post at the higher excess reserves rate. Bottom line, they are effective funds rate influencers.

This whole combination ultimately determines where the effective funds rate ends up at. It’s a complex web, but it works. And it acquiesces to the excess reserves environment that the various crises and consequential buy bond buying crusades have left us with. The Federal Reserve has basically run with this, on a theory that if it’s not broken, then no need to fix it.

Fed Chair-elect Kevin Warsh has a different opinion. He wants to 'fix it', it seems. Whether that means doing the whole hog and going back 20 years or something less dramatic is not clear. But let's push things here and ask, what could that potentially look like?

The technicalities would require the sale of all mortgage-backed bonds (US$2tr) and at least half of the Treasury bonds (circa US$2.5tr). So that’s some US$4.5tr in total, and would bring Fed bond holdings back down to around 5.5% of GDP (pre-GFC proportions). That in itself is big, and raises questions about a doable pace. It could hardly be done promptly, and far more likely doable over a few years. And there are clear implications here for a larger marketable holding of Treasuries on the marketplace.

Then there is the complication of transition from the current excess reserves environment, potentially back towards a scarce reserves environment. This is 'putting the toothpaste back into the tube' issue. Selling the bonds is relatively straightforward; far less straightforward is how to get back to the prior regime for reserves management and how the funds rate is set; along the way, and in the future.

One aspect that can help here is the revision to the leverage ratio requirements as of 1 April 2026. The fact that the larger banks, as a result of this, will have more capacity to buy Treasuries and/or engage in repo is potentially of huge benefit. The weak link in the current structure has been the capacity for tightness in repo to bully the effective funds rate higher, primarily on a relative value play. Larger banks in the game could and should tame the tendency for repo to over-tighten at times.

There is a link here with reserves too, as the Federal Reserve has reverse engineered a logic that repo tightness was associated with the fall in excess reserves to a level that acted to exacerbate repo tightness. While a clear link between the two is not necessarily obvious, there is clearly a link. We saw the same when the Fed went through its first quantitative tightening exercise through 2019, which culminated in quite a severe repo tantrum, even if brief. It was calmed through a Fed decision to buy bills to help add bank reserves. Fast-forward to 2025, and we saw the same, as the Fed resumed buying bills on repo tightness.