Are We There Yet?
Scaring myself on the journey to balance sheet reduction.
A kids book published in 2017 entitled ‘Are We There Yet?’ describes a boring car journey “as the little boy feels the hours ahead weigh heavy on the horizon, his imagination (or is it?) takes the car further and further into the past until there's a very real possibility that he and his mum and dad might be eaten by a dinosaur!”
This feeling is familiar to those of us trying to assess the real-world impact of central bank balance sheet reduction. Schooled by the real financial market monsters of 1992, 2000, 2007-12, 2019 and 2020, it is not hard for imaginary financial monsters to loom under QT. Yet the boring journey of balance sheet reduction largely continues uninterrupted - with the significant exception of SVB and friends. After a year or so imagining the imminent appearance of liquidity monsters it feels a little foolish to continue to scare ourselves. But we can’t help it, and anyway we are not alone. It seems even some central bankers scare themselves with imaginings of financial instability.
This isn’t surprising. Estimates of what is required to preserve financial stability became a challenge with the move to ‘ample reserves’. What used to be a relatively precise science for the Fed, ECB and BoE was obliterated by the gush of reserves. While central bank balance sheets remained stable, that didn’t matter. But with the advent of QT, an estimate of liquidity demands of the banking system becomes a pressing matter. Unfortunately, ample liquidity meant the old science became an artful guessing game - an act of imagination. And while there have been many confident pronouncements that the future will require more reserves than pre-GFC, there are very few calculated estimations of exactly how many reserves. The current state of assessment on system liquidity limits could be described in a similar manner to the famous quote by Justice Potter Stewart, 'I’ll know it when I see it’. That is an alarming way to run a financial system, even though it is probably the way a financial system is often run.
On 23rd August, the Federal Reserve of St Louis published a short assessment of ‘balance sheet normalization’ (QT). The focus of the piece is the potential financial stability effects from changes to Fed liabilities caused by QT. The article concludes with several estimates of the lower bound of system liquidity needed for financial stability. “Something closer to $2 trillion could be the optimal level of reserves in the system before liquidity constraints begin to force money market rates higher. The optimal level could be even higher, though. Financial markets keep evolving and desired liquidity may be something closer to 10% to 12% of nominal GDP ($2.7 trillion to $3.3 trillion).”
For reference, current system reserves stand at $3.3 trillion, in line with the upper estimate for an optimal level of reserves quoted in the St Louis paper. However, the paper does not show their calculations for the lower boundary of reserve demand.
One notable attempt to calculate the boundary level of reserves is a paper by David Lopez-Salido and Annette Vissing-Jorgensen (hereafter LS-VJ). They suggested a safe level of reserves may be $3.34 trillion, based on bank deposit holdings in early 2022. I note that this figure is closely aligned to the upper estimate of the recent St Louis Fed, cited above, and close to current actual levels of reserves in the system.
So, it is worth considering how these dynamics may play out. Especially as the ongoing QT programme will, within a few months, reduce the level of system reserves well below the upper limit required for financial stability outlined both by the St Louis Fed paper and by LS-VJ.
It would help to ask those most intimately affected by system liquidity what they think. Conveniently, the Fed does this for us. Before every FOMC the Fed conducts a survey of primary dealers, asking questions about monetary policy and system liquidity. Results of the latest survey were submitted by dealers on 11th September (Monday of this week). The results will be used to inform the FOMC discussion in the coming week. The public will have to wait another 3 weeks or so before we get to see the survey results.
However, results from the last primary dealers survey are available. And responses on expected liquidity are especially interesting. Below is a construction of Fed’s balance sheet that satisfies the median responses given in the primary dealers survey. The items in red represent the median responses from the survey.
There are a big assumptions here: median responses may hide a lot of variation, we’re only dealing with ‘main’ components of the balance sheet, how the TGA (Treasury General Account) evolves is an important unknown and cash in circulation is calculated as a residual here, were in reality cash is most certainly not a residual.
That said, there are some pretty striking takeaways to ponder. The median primary dealer response suggests system reserves will fall $2.7 trillion in about a year’s time. This is well below the upper estimate of necessary reserves quoted by both the recent St Louis Fed paper and by LS-VJ. Of secondary importance, but interesting, the survey expected SOMA Treasury holdings will cease to contract around Q3 2024. In addition, holdings in the ONRRP are expected to fall to $1.1 trillion by end-2024 - from a level of $1.835 trillion currently.
The LS-VJ paper observed that despite the abolition of official reserve requirements, the US commercial banking system effectively operated an informal reserve requirement linked to the amount of total deposits held in the banking system. Their paper used regression estimates to model deviations in Interest-on-Reserves and Fed Funds rate, as a test for short-term rate sensitivity to liquidity. They showed that changes in deposits and system liquidity (reserves) provided an accurate gauge of how money-market spreads behaved from early 2009 to early 2022, including reasonable estimates of spread behaviour during the ‘Repo Crisis’ of September 2019. They had built a ‘Geiger counter’ for financial instability. Or a ‘monster’ detector.
Given the much higher level of deposits created by government and banking largesse over COVID, the LS-VJ study suggests the room for manoeuvre on reserves is considerably more constrained during QT2 than during QT1 (2017-2019), though this is implicit, rather than explicit.
If we take the expected reserve holdings for the next year reported in the primary dealers survey and apply the coefficients from the LS-VJ study then, absent a large decline in total bank deposits, the result suggests a high probability of widening money-market spreads - and so the possible appearance of monsters.
Bank deposits declined 4.2% from August 2022 till August 2023. Assuming that annual rate of decline continues and the profile of reserve holdings match that reported in the primary dealers survey, coefficients from the LS-VJ paper suggest upward pressure on money-market rates would appear towards the end of 2023, and continue throughout next year. This is represented by a move above the zero line in the chart below.
Note, their model also suggested that money-market spreads would tend to widen from end-2022 till March-2023. In fact, spreads between repo rates such as GCF and official rates did show signs of widening from about November 2022, though not by much. The subsequent fall in the modelled spread in March 2023 was associated with emergency liquidity provision following SVB. It is interesting, though fruitless, to speculate whether the tighter liquidity had any bearing on the banking difficulties of early 2023. We do know that additional liquidity was an important contributor to stemming the instability.
To turn the previous question around, if reserves fall at the rate expected by the primary dealers survey, deposits would need to fall at an annual rate of ~10% to offset pressure on money market rates, if the coefficients of the LS-VJ paper hold. This is considerably faster than the already unusual rate of decline of deposits seen in the last year and would be the greatest annual rate of decline in deposits since 1931, which heralded the beginning of Great Depression. At present, such a fall in deposits seems very unlikely to happen.
Conversely, if deposits actually rise, then the study suggests upward pressure on money market rates would become more pronounced.
There are too many assumptions in these calculations to have any confidence in the results. They remain imaginings. But that is exactly the point. It is clear from the recent St Louis Fed paper that officials share a similar state of confusion.
We cannot be surprised that the central banks have so little visibility into these situations given the sheer opacity of the entire process. Arguably, it is why they are always reactive, with changes in their actions merely a question of how long a delay, rather than if there is a delay between a catalyst and a response.