How You Can Understand Financial Conditions
- Zachary Cameron
- Oct 16, 2024
- 10 min read
Updated: Nov 24, 2024
The 2023 banking crisis was one of the largest in US history, and like the Great Financial Crisis 'no one' saw it coming. This note breaks down how I was able to predict this crisis (reaching out to my hedge fund contacts a week prior to Silicon Valley Bank failing).
Note that the material in this article is relatively advanced, as it was written initially for hedge fund managers. We will work up to understanding these concepts in depth; for now I simply wish to present some ways in which monetary conditions can be inferred.
“Collateral flows lie at the heart of any proper understanding of financial stability. No other market is so critical to the functioning of the financial system and yet so poorly understood” ~ Manmohan Singh (2016)
Introduction
In February 2023, “liquidity expert” Michael Howell claimed that a new liquidity super cycle had started months prior; these words echoed the sentiment of financial plumbing expert and “repo oracle” Zoltan Pozsar, who stated throughout 2022 that the financial system was flush with cash and that it was likely that we would not have a financial crisis for years.
This note attempts to elucidate (as briefly as possible) the variables that led the author to call for a March banking crisis. The intention of the above anecdotes is not to criticize- forecasting is incredibly difficult and the aforementioned analysts (Zoltan in particular) do some great work- but rather to illustrate that even ‘financial plumbing’ experts missed what were relatively obvious signals of monetary distress.
The point of doing so is to convey to traders that there is an incredible amount of alpha to be gained from understanding the plumbing of the monetary system- there are not many people that understand these markets and even fewer that can connect conditions in these markets to broader financial variables.
This note- while omitting variables the author considers proprietary- will provide publicly available, simple indicators and narrative discussion in order to illustrate to readers how the crisis was forecastable.
Collateral and Liquidity
QE adds liquidity; QT drains it. There are not many (any?) that would disagree with these statements. However, these statements aren’t necessarily true. In fact, they can be the polar opposite of the truth: QE can drain liquidity and QT can add it.
The March banking crisis occurred in an environment in which there were trillions of reserves but a clear scarcity of collateral. Why did liquidity experts fail to warn about the crisis? Because too much attention was paid to the former and not enough to the latter.
This is not surprising- when QE first began in the US, many well-known economists were scared of (hyper)inflation; when the 2019 September repo rumble occurred, the assigned blame by even renowned plumbing experts like Darrell Duffie was that the level of reserves was too low (leading to the acronym “LCLoR” for “Lowest Comfortable Level of Reserves”).
There has always been a bit of fetishization of banks reserves (stemming in part from monetarism and a belief in “fractional reserve” banking and the money multiplier), but blaming liquidity events on a dearth of reserves is way too simplistic of analysis. The monetary system is more complicated than that. So what else should warrant focus?
Collateral.
Lending between financial institutions is almost always secured: this is as true in countries like China as it is in the US.
Bank reserves can only be traded within the banking system and stay relatively inert on balance sheets (roughly $3,000 billion in reserves and $100 billion in Fed Funds volume...).
Collateral, on the other hand, can be used (and re-used) throughout the monetary system. This has become increasingly important given the rise of non-bank financial institutions (shadow banking).
As such, it is collateral, not reserves, that ‘lubricates the pipes’ of the financial system.
Here are a couple of quotes to support this claim:
The St. Louis Fed (2017):
“Reserves may be inferior assets to Treasury bills... T-bills are
more widely held and are useful as collateral in financial transactions in ways that reserves are not.”
Manmohan Singh (2019), arguably the world’s foremost “monetary plumber”:
“QT provides balance sheet space and helps the dealer community accommodate more transactions.”
Jerome Schneider (2013), head of the money markets desk at PIMCO:
“Repos provide the oil and the grease to the financial market, and
dealers don’t have the collateral. As the Fed goes and purchases more and more collateral in the marketplace, there is less to supply to the repo market.”
*Note that Schneider’s quote was from 2013, when repo rates on the 10 year UST went negative due to a QE induced collateral shortage
Although a full exposition of the liquidity effects of QE/QT- and, relatedly, reserves and collateral- is outside the scope of this note, the important takeaway is that collateral markets are incredibly important and that there is not a simple correlation between QE and liquidity.
What follows are some indicators that illustrate the shape of collateral markets before SVB...
Dealers fund themselves at the triparty rate via MMF lending: this rate is never supposed to be below the RRP rate (Zoltan claimed in 2018 that it was impossible for TGCR to go below RRP) as money funds could simply lend to the Fed and capture the higher rate instead.
The divergence between these rates can serve as a rough barometer for the collateral / reserve balance in the financial system: Triparty funding rates high above RRP (10-15 bps) signal that dealers are welcoming cash and are willing to part with collateral; TGCR below RRP signals that dealers are trying to deflect cash, unwilling to part with collateral, and unwilling to expand their balance sheets.
Relatedly, the Fed’s RRP facility can serve as an indicator for collateral conditions... While Zoltan often discusses the RRP as being the parking spot for “excess cash” (again utilizing a bank reserve centric framework), a more insightful way to understand this facility is as a proxy for balance sheet and treasury bill scarcity.
Note in the following images that spikes in the RRP often occur at times when there are deflationary monetary conditions- e.g., upticks in USD and a flattening UST yield curve- thus the narrative of ‘too much dollar liquidity’ misleads.
A 2022 report by FRBNY concurs with this narrative, pointing out that a scarcity of balance sheet space motivates banks to push deposits off their books towards MMFs. Likewise, balance sheet scarcity causes banks and dealers to reduce repo activity (as repo increases balance sheet size and ‘encumbers’ collateral, thereby reducing HQLA). The net result is an uptick in MMF assets and RRP usage.
Note the usage of the RRP was near an all-time high leading up to SVB:
Importantly, an unwillingness to utilize balance sheets can also be caused by risk-aversion (and not simply a lack of space caused by a glut of reserves).
In ‘risk-on’ periods, collateral chains lengthen- meaning that junkier collateral is accepted and more parties are willing to partake in repo transactions- thereby easing funding conditions. Furthermore, risk-on sentiment causes dealer-banks to adjust internal risk-metrics and deposit outflow assumptions, effectively reducing the amount of HQLA needed, enabling these institutions to participate in secured lending / borrowing.
Conversely, risk-aversion reduces the willingness of repo counterparties to roll-over transactions and increases the standard of ‘acceptable’ repo collateral, thus increasing the demand for pristine collateral.
The following image illustrates the percentage of equity collateral used in primary dealer repo transactions. It evidences that dealers were rejecting junkier collateral prior to SVB, thus indicating risk-aversion and more difficult funding conditions. This means that repo borrowers were forced to increase their demand for pristine collateral in order to roll-over funding.
A shortage of treasury bills (both in terms of reduced issuance and increased demand due to increased collateral standards) prompts MMFs to turn to the RRP facility, given the resulting yield differential between 4-week bills and the RRP rate.
In fact, one can learn a lot about monetary conditions by looking at the 4-week bill rate vs. the RRP rate: A significant amount of financial counterparties have access to the RRP facility (and counterparty caps are high), thus it ‘should’ be unusual for the 4-week rate to be below the RRP rate.
But this has occurred often... so what can we glean?
First, note that 4-week bills are better collateral- although the RRP is collateralized by USTs, the collateral received must stay within the triparty system and thus doesn’t offer the same re-use benefits, as most repo is bilateral. Furthermore, bills offer better intraday liquidity as the RRP facility returns cash late in the day (3:30 PM). As such, drops in 1m yields relative to the RRP is a huge red flag that there is an increase in demand for collateral and liquidity (i.e., a stressed monetary system), as the following image illustrates.
Note the 4-week fell below the RRP during March 2020 and leading up to SVB.
There is a significant amount of bargaining that occurs in repo markets: there is a hierarchy of repo counterparties and rates that reflect this hierarchy. The spread in repo rates thus can tell us something about the monetary system via the shifting bargaining power of different counterparties (a bit similar to how effective federal funds fell during the GFC giving the increased desire for transactions to occur between highly rated counterparties).
However, the spread in repo rates also tells us something even more telling about the monetary system: specifically, the gap between triparty repo and GCF repo indicates the willingness of dealers to utilize their balance sheets to deal money.
Generally speaking, dealers fund themselves at the triparty rate and lend to other (typically smaller) dealers at the GCF rate. Thus, a wide spread between these rates (emphasizing again that these aren’t monolithic rates) indicates that dealers need extra profits to be incentivized to utilize their balance sheets.
As the following image illustrates, the dollar strengthens and the global economy suffers when there is a wide spread between triparty and GCF rates (the dollar not necessarily causing global economic weakness as much as both variables being dependent on the independent variable of dealer risk-aversion / balance sheet capacity).
The following image, utilizing the BGCR spread (consider this the same as a combined triparty and GCF rate, with lower percentiles being triparty and higher percentiles being GCF rates), illustrates the stress in the interdealer system in March-
These are enormous spreads. They occurred often in the lead-up to SVB.
A widening of these spreads can occur both from limited balance sheet space (usually due to too many reserves) or from risk-aversion (liquidity, not default risk being the concern).
This next image evidences that dealers were incredibly risk-averse leading up to SVB-
As this image illustrates, it is extremely rare to see dealers hoarding t-bills like they were before SVB.
Why might risk-aversion cause dealers to hoard bills?
Because bills are always liquid, always acceptable repo collateral. It is only at periods of extreme monetary stress that the distinction between bills and coupons (and on-the-run / off-the-run) becomes important for repo purposes.
Note the yield spread between 4-week on-the-runs and off-the-runs prior to SVB-
Treasury auctions in the months preceding SVB also serve to illustrate the dire collateral conditions. For example...
It is extremely rare for indirect bidders to be a multiple of primary dealers for 4-week bills: the last time a 2x multiple occurred prior to March 2023 was March 2020.
Indirects are a proxy for foreign demand for USTs- if pristine collateral is not circulating / counterparties are demanding higher quality collateral, this increases the bid from foreign entities (often via their central banks) at auctions.
This leads us to a larger point...
We have a global monetary system, meaning that collateral scarcity is a global phenomenon. Exemplifying this-
Repo rates in Japan hit an all-time low at the beginning of 2023. Negative repo rates indicate collateral scarcity, as cash lenders are paying interest to obtain the collateral value of pristine securities.
Note the correlation between negative repo rates in Japan and the usage of the Foreign RRP below.
Unpacking this correlation is a bit outside the scope of this note, but the important takeaway is that collateral scarcity / monetary stress is a global condition. Expanding one’s scope globally can thus aid in forecasting conditions on a country level.
Finally, it is important to point out that there were clear signs of dollar funding stress leading up to the banking crisis- emphasizing again that collateral shortage and dollar shortage often going hand in hand given the relationship between collateral and funding.
For example....
On February 24, 2023, the FHLB Office of Finance reported Q4 operating results. The report states, “Advances totaled $819.1 billion... an increase of 133% resulting primarily from an increase in short-term advances, driven by depository member demand for liquidity...”
SVB and First Republic were reported as the top borrowers.
It is extremely common (almost a necessary condition) to have a drastic increase in FHLB advances before a financial crisis- e.g., there was a huge increase in advances before the GFC.
Thus, from this metric alone, it was clear that funding markets were strained.
But let’s turn briefly to other signs of dollar funding stress...
As borrowings from the Fed and Swiss dollar auctions illustrate, there was evidence of dollar funding stress heading into the Spring of this year.
This coincided directly with collateral market stress indicators-The first image below shows dealers turning to the Fed for UST borrowings; the second image shows the drastic drop in Japanese 3m yields during these same periods (Japanese bills are pristine collateral).
Note that the steep drop in Japanese bill yields that occurred in October of 2022 coincided with the drastic increase in SNB dollar auctions (ironically claimed to be a result of ‘easy profit taking’ and ‘not a function of dollar liquidity problems’ by Credit Suisse economists) as well as a sharp increase in borrowings from the Fed.
Although the borrowings themselves were trivial in size, the important point is that there were numerous signs of collateral and dollar funding stress leading up to SVB. These strains were ignored by ‘experts’ in financial plumbing due in large part to an excessive focus on bank reserves at the expense of an appreciation of collateral and wholesale dollar funding markets.
Conclusion
So the banking crisis was predictable, great, but does studying the ‘plumbing’ matter for more than predicting once in every decade(s) financial crises?
Yes. And this is the most important takeaway: we have a credit based (not fiat) monetary system, and the necessity of financial institutions to roll over their short-term (USD denominated) liabilities is the foundation of this system.
Dealers are the critical entities, the beating hearts that determine the flow of settlement claims, and money markets are where one can observe the pulse of this system.
As such, studying the ‘plumbing’ isn’t just some wonky activity that’s important for predicting financial crises- and it shouldn’t just be the purview of money market traders. The plumbing is essential for forecasting economic growth and variables like FX / rates, thus making it vital as an arena of analysis for all macro traders.