Hello everyone and welcome back to the Chartbook! Before getting started, a quick disclaimer on some of the content to follow. This issue will contain discussion of the financial data of several listed US companies (tickers JPM, BAC, C, GS, MS). The discussion is intended as informational commentary on money markets and not analysis of the relevant public equities. As always, nothing here is advice and please do your own research if you are interested in taking a position in large-cap US bank stocks (nobody involved in the writing of this Chartbook has any such positions currently or in the foreseeable future). Okay, now let’s get going!
Repo Specials
For this week’s issue, let’s save the usual discussion of US Treasuries for the end and start with the repo market instead. The US Treasury repo market allows counterparties to temporarily exchange cash and US Treasuries with the agreement to reverse the exchange at a later date. Sometimes this is the very next day (overnight repos) or other times it can be a while later (term repos). Usually, the counterparty providing cash will be paid back a slightly larger amount on the closing day, which makes this analogous to a collateralized loan secured by US Treasuries. However, the repo market can also be seen as a way to borrow US Treasuries from the other counterparty’s perspective. Since borrowing US Treasuries is useful sometimes, some participants are active in the repo market for this reason (not necessarily to borrow cash).
This gives rise to the concept of repo specials, which are specific US Treasury securities where demand to borrow is high enough that the pricing for repos on them diverges significantly for the rest of the market. Borrowing cash against a special security is cheaper than borrowing against general collateral (generic unspecified US Treasuries) because the counterparty on the other side of the trade has incentive to borrow the specific security in high demand. Repo specials are a fairly common occurrence, but pronounced and frequent ones can suggest some things about the market. For example, one possible cause would be heavy short-selling of US Treasuries. For short-sellers, borrowing Treasuries in repo is a convenient and efficient way to fund a trade. The security borrowed needs to be specific, causing specials in the most shorted maturities (most would not be satisfied with shorting a generic Treasury that could have any maturity). To get an idea of the specials activity recently in the Treasury repo market, let’s have a look at a composite repo rate (SOFR) broken down by percentiles of trading volume.
The chart above shows the composite SOFR rate (red) and the 25th (blue) and 1st percentile (white) of trading volume. The composite rate is calculated as a median of data collected from various repo trading venues, some of which is partly filtered to remove specials. The details of the methodology are beyond the scope of this discussion but curious readers can find it here if interested (and a helpful reference guide here for more market mechanics). The key point we’ll focus on in this chart is the white line starting to dip notably into negative territory over the past two months or so. This means that at least 1% of Treasury repos in the SOFR data are trading far below the median rate, suggesting increased specials activity. Since this would imply demand to borrow Treasuries is high, it would be intuitive to think that some participants holding them would be stepping up their lending to harvest the borrow cost. The next two charts will provide a theory on who is doing so and what securities they are lending.
In this chart, we see the reported average balances for the three most active US banks in the Treasuries lending business: JPMorgan (JPM), Bank of America (BAC), and Citigroup (C). While BAC and C remained relatively steady in their lending in Q1 2021, JPM has ramped up aggressively, taking over the top spot from BAC as the biggest bank lender of Treasury securities. This suggests that JPM is actively taking advantage of the specials activity in the repo market by lending out their Treasury portfolio (also the largest among US banks).
While borrowing Treasuries from banks in the repo market is the most efficient option for many market participants, primary dealers can also borrow directly from the Federal Reserve, as the NY Fed allows them to borrow some Treasuries overnight from the Fed’s holdings at a fairly low rate. The primary dealers can then re-lend or otherwise redistribute these Treasuries into the market, so looking at Fed securities lending activity can give us an idea of the maturities that are most in-demand.
This chart shows year-to-date activity in the NY Fed’s overnight securities lending program, aggregated by year of maturity. Notable here is the fact that most of the heavy borrowing came in the 0 to 10 year maturities, with spikes at the 3 year, 5 year, and 10 year points. The high volume in 2030 maturities is likely due to demand for 10 year securities that were issued last year, before new auctions created enough supply of 10 year Treasuries maturing in 2031. Past the 10 year point, there is a gap followed by lower spikes at 15, 20, 25 and 30 years. This all suggests demand to short the 3, 5, and 10 year US Treasuries has been high this year, with short-sellers paying high rates to borrow these maturities. Incentivized by the repo specials, large players have likely either been lending their own portfolios or borrowing from the NY Fed to re-lend. JPM has seemingly played a particularly active role in this action, as evidenced by their spike in repo securities lending in Q1 2021. Now, let’s move on to some other trends in US bank activity!
Bank Loss Absorbing Capacity
Since Q1 2021 ended and the much-discussed SLR rule exemptions expired on March 31, there has been a notable rush to issue bonds among some large US banks. Most prominent in this rush were JPM and BAC, with JPM pricing the largest bank bond deal up to that point on April 15 ($13B) only to be surpassed by BAC the next day ($15B). While the SLR rule expiry was floated as a possible reason, there may be a slightly more obscure rule regulating loss absorption capacity actually driving these sales. To start unraveling this, let’s first look at what the relevant bank regulations here actually require.
In this diagram, we have a stylized depiction of a generic bank balance sheet. On the left are the bank assets, which consist of high quality liquid assets (HQLA) such as reserves at the Fed and Treasuries, as well as other assets such as loans and interbank assets. On the other side of the balance sheet there is some shareholder equity at the top, unsecured debt in the middle, and other liabilities at the bottom (such as bank depositor’s accounts). The two rules relevant to this discussion are on the right, with the factors color-coded based on the balance sheet bucket they come from.
The SLR rule at the top is meant to put some limits on the size a bank can grow to without raising additional capital through equity. The rule defines a minimum for the ratio of Tier 1 Capital (which is approximately equal to the shareholders equity) to the Total Leverage Exposure (which is approximately equal to the total assets). The controversy earlier this year about SLR exemptions centered around whether or not banks should be allowed to exclude some HQLA (reserves and Treasuries) from the Total Leverage Exposure calculation, as had been temporarily the case since last year.
On the other hand, the more obscure TLAC rule (key text here) is meant to define how resilient bank balance sheets need to be against losses. For this calculation, most long term unsecured debt is included as loss absorbing since it can in theory be converted into equity in a severe crisis. The TLAC ratio is then calculated as Tier 1 Capital plus eligible long term debt divided by the Total Leverage Exposure, and held to a higher minimum than the SLR to avoid being redundant due to the added term in the numerator. Since the TLAC ratio also uses Total Leverage Exposure as the denominator, the SLR rule exemptions also affected the TLAC rule, allowing banks to exclude reserves and Treasuries from both calculations. Note that this is just a basic overview and there are many other limits and ratios in these two rules, but for this discussion I think it will suffice.
Finally, it is important to note that since 2008 all the major US banks are structured as bank holding companies (HoldCo) with bank operating subsidiaries (OpCo), and the SLR minimums are different between these two separate entities. The controversy around exemptions earlier this year affected almost exclusively the bank HoldCos, since the exemption at the OpCo level did not apply automatically and Goldman Sachs (GS) was the only major bank that chose to utilize it. Now, let’s look at some real data!
Here we have the the HoldCo (left) and OpCo (right) SLRs of five major US banks as of the end of 2020, broken down by the required portion (blue) and buffer. In the chart on the left, the buffer is further split into the portion that would remain regardless of SLR exemptions (orange) and the part that would disappear were the exemptions to be revoked (grey). In the chart on the right the only bank that would be affected is GS, which has a high enough OpCo SLR that the split can be excluded for clarity.
We see here that the two banks issuing the most bonds (BAC & JPM) also have the lowest OpCo SLRs, which suggests one possible reason for the issuance is to replenish Tier 1 Capital at the OpCo through an infusion of the proceeds from the HoldCo. To work through the balance sheet arithmetic here, we can move on to the actual dollar amounts provided by JPM’s OpCo in their year-end 2020 FFIEC 031 Report (which can be found here).
In these two extracts from the report, we see the SLR for JPM’s OpCo (officially known as JPMorgan Chase Bank, National Association) and the dollar values of the numerator and the denominator. At the close of last year, this was $234B of Tier 1 Capital and $3.69T of Total Leverage Exposure, the ratio of which gave 6.35% SLR. Since the SLR minimum at the OpCo level is 6% and the exemptions did not apply to this, they had a 0.35% buffer here regardless of rule changes. Translating this buffer into dollar amounts, JPM could grow its OpCo Total Leverage Exposure to $234B / 6% = $3.90T without starting to incur penalties, or about $210B in balance sheet growth at the OpCo.
Now let’s suppose JPM’s OpCo had grown to where the buffer was almost completely eroded (SLR approaching 6% at around $3.90T in Total Leverage Exposure), and they had done the $13B bond issuance at the HoldCo to infuse additional capital into the OpCo. Assuming Tier 1 Capital had remained unchanged otherwise, the SLR post-infusion would be ($234B + $13B) / $3.90T = 6.33%. Thus, in one move, they would reverse a fairly large quarter of balance sheet growth in SLR terms. Since there is no regulatory change here, only organic growth, it seems unlikely that this can explain the size of the bond issuances from JPM and BAC. While OpCo capital infusions may be part of the use of capital here, gradual tightening of buffers does not seem like sufficient reason to issue a huge multi-tranche bond deal all at once. For a possible alternative theory, we must turn to the HoldCo TLAC rule. Let’s look at some data provided by JPM’s HoldCo in their latest 10-K (which can be found here).
In this extract we see that JPM’s HoldCo (officially known as JPMorgan Chase & Co) had a minimum required TLAC ratio of 9.5% at the end of 2020 and an actual value of 12.4%. Since the denominator used for both ratios is the same, we can use the fact that JPM’s HoldCo SLR was 6.9% (see the chart comparing the five banks) to derive an approximate relationship between the HoldCo SLR and TLAC ratio. In the equation below, assuming constant numerators for both ratios, JPM’s HoldCo TLAC would breach minimums at a HoldCo SLR value of x:
9.5% / 12.4% = x / 6.9% -> solve for x to get x = 5.3%
Since x is higher than the HoldCo SLR minimum of 5%, this means JPM would run into the TLAC minimum before they run into the SLR minimum at the HoldCo level, which is automatically affected by the rule change unlike the OpCo. This seems more plausible as an explanation for the large bond issuance, because long term unsecured debt is classified as loss absorbing and would add to the TLAC numerator to offset the growth in the denominator due to the exemptions expiring!
Finally, in this chart of debt distribution by maturity, we see the effects of this likely TLAC-driven issuance on JPM’s overall debt structure. While it may seem insignificant compared to the $100s of billions of JPM debt already outstanding, the new issuance since the end of Q1 is very large considering it has been less than a month. However, considering that it fixes two problems at once (OpCo SLR and HoldCo TLAC) as opposed to just one, the size seems much more justified.
I have not worked through the same arithmetic with BAC, but am fairly sure that it would also show the HoldCo TLAC as a pressing regulatory constraint. Now, to wrap up, let’s take a step back and look at the Treasury market!
US Treasuries
In this chart, as usual, we see the US Treasury yield curve at the close of this week (April 23, solid green) and last week (April 16, dashed brown), with the bars below showing the change over the course of the week in basis points. While yields did drop across the curve this week, the move was relatively small (3 basis points or less) and led by the longer maturities as opposed to the middle of the curve in previous weeks. In fact, the theme of the 5 & 7 year maturities leading moves seems to have faded away for now at least, as has downward momentum in yields. While the partial reversal of the upward move in yields from Q1 has sparked much debate and possible explanations, I believe simple profit-taking may be a significant factor here given the amount of likely short-selling activity we observed in the first section.
A natural question to ask here is which way yields go next. While there are many moving parts, one major factor could be the willingness of foreign buyers to step back into the market in size. To get a gauge on these buyer’s sentiment, let’s check out some data on the weekly purchases of foreign bonds by Japanese institutions (major US Treasury holders).
In the chart above, we see the weekly net change in Japanese institutional holdings of foreign long and medium term bonds as the white bars (in ¥ billions). The red line in the bottom panel shows the 10 year US Treasury yield after adjusting for the cost of hedging for FX risk and the blue line shows the yield of a 10 year Japanese Government Bond. As we can see, the 10 year US Treasury has an attractive relative yield currently, but Japanese institutions are not buying any more foreign bonds than they were last fall, when the yield differential was much lower. It is true Treasuries are no longer being sold off in Japan as they were in late-February and early-March, but buying has remained fairly unimpressive for now. This hesitance from overseas investors should be a sign of caution for US Treasury holders expecting yields to fall further I believe, as it seems doubtful such a move could be sustainable without enthusiastic participation from abroad. For now, Treasuries remain in a bit of a limbo, with the possible short-covering-related rebound running its course but the logical buyers at these yields mostly staying out of the action.
That will be all for this week! Thanks for reading if you made it all the way here and hope to see you back next week for another Chartbook :)
Cheers,
DC
This set of posts is amazing. I read your explanation on how the TGA drawdown is partially responsible for the increase in the ON RRP rates, however you also mention in this chartbook that "repo specials are a fairly common occurrence, but pronounced and frequent ones can suggest some things about the market. For example, one possible cause would be heavy short-selling of US Treasuries." There is prevailing sentiment that Citadel and other HF's are behind this trade is a post here https://www.reddit.com/r/GME/comments/mgucv2/the_everything_short/. Do you think there is any validity to what is posted there?