Treasury Issuance is what really matters
Month end Outstanding Treasury Debt projections (including new issuance) through the end of fiscal year 24
TLDR
QT is just a monthly expense for the Treasury, and has only that indirect effect on supply of publicly held treasury duration. Treasury issuance composition is what should be paid attention to.
Ive built a model to project outstanding (including ongoing issuance) treasuries through end of fiscal year 2024. (basically projecting the Treasury Monthly statement of the public debt). Fed “add ons” and duration characteristics are included historically back to December 2018 and through September 2024. My methodology and issuance assumptions (modeled after TBAC charge) are detailed herein. Model output and configuration inputs are attached.
Fed will fail to reach the UST QT cap of 60b in June 2024 and again in September 2024
Volume of outstanding treasuries with 3 months or less to maturity will rise ~50% over the last 6 months of 2023 going from ~3 to ~4.6T.
Greetings!
For the last year or so I have been providing monthly reports on QT including the precise days and mostly precise amounts by which the Feds SOMA holdings would reduce. I think my track record for accuracy has been pretty good. Unfortunately, though, my reports focus on the entirely wrong thing. At least for the 60b a month of UST QT, the rolloff intramonth when and by how much just doesn’t really matter. Why? Because UST QT is just an expense to the Treasury, granted it’s a 60b a month expense (the repayment of principle on the maturing securities), but its an expense like any other that Treasury must pay for using funds in the TGA. So whether it has to pay 15b mid month or 45b mid month, as long as the TGA has sufficient funds to pay off the maturing securities (which aside from debt ceiling drama windows is never a problem), shrug.
On the other hand, the TGA does not gain the funds to pay for QT and other govt expenditures by magic. Funds get added to it when the Treasury collects taxes (and other various govt revenues) and from the proceeds of newly issued debt. QT raises the amount Treasury must raise through some means by up to 720b a year, but that’s it. Thinking about it this way helps because it isolates what actually is important, how Treasury raises the funds in the TGA to keep it roughly level and more specifically, because tax collections are what they are, how Treasury issues debt to cover deficit spending + QT. That debt issuance, and its composition, is all that really matters.
But wait, isnt Treasury spending on QT fundamentally different versus any other Treasury spending (interest on debt, social security checks, govt workers)? Yes, when Treasury pays off QT maturities, the reserves it transfers from the TGA are destroyed, whereas with every other expenditure those reserves are transferred to some commercial bank and excluding any payments directly to the bank (e.g. paying off a maturing UST held by say Wells Fargo) creates corresponding deposits for someone. Generally speaking, other Treasury spending increases the publicly held money supply and spending on QT does not. That certainly is an important difference, but liquidity is never reduced (though it might not be increased as with QT spend) with Treasury spending, only to some degree or another, when the Treasury collects money. So that is where the focus needs to be.
<Side Note. QT also reduces overall reserves in the system, eventually that reduction will lead to a problem in the aggregate for the banking sector, but with 1.6T in the RRP we are a loooong way from that impact of QT hitting home>
When Treasury raises cash for the TGA by issuing debt, they trade a UST in exchange for someone’s reserves/deposits. Literally, this reduces the publicly held money supply/liquidity (albeit temporarily until they spend it (on something other than QT)), but how much it effectively reduces it I think varies with the term of that security. The longer the term, the less “money like” the exchanged UST for the deposit/reserve is. Why? Well regardless of the term the credit risk is zero you will get repaid at par when the time comes, but the interest rate risk is very much different, for longer term UST you might take a significant haircut to par if you sell it before maturity (just ask the folks who bought 30yr UST bonds in 2021….looking at you banks) assuming interest rates rise in the interim. Whereas with a short term, you are close to maturity where the value of the UST will be par. What’s really the difference between a $100 bank deposit and a $100 treasury bill that matures tomorrow. The answer I think is not much. Sure if you *needed* to make a $100 payment (in bank deposits) to someone today, you would have to go through the hassle of selling that $100 treasury bill, but you would get very close to par (remembering that since this is a bill you paid less than par to acquire it) and there is no shortage of folks willing to buy that bill at very close to par because for them (who don’t *need* the deposit/reserves today), come tomorrow it will in fact become $100 in bank deposits/reserves.
Hassle is an important consideration, however, for short time to maturity treasuries MMFs will effectively handle all that hassle for you providing what is basically to you a bank deposit, that is backed in part but theoretically entirely, by short time to maturity treasuries. Short term to maturity UST are very money like and highly liquid, so when Treasury issues them in exchange for deposits/reserves they really aren’t reducing the effective money supply much if any at all.
Long dated treasuries are a different matter though. While they can certainly be sold (or repoed at a haricut and a cost), it may or may not be for anything near par and it’s a long time before the sovereign actually repays them at par, so while long dated treasuries have some degree of money-likeness/liquidity its considerably less than the publicly held deposits/reserves they replace when the Treasury issues debt.
Broadly, and all else equal, issuance that skews longer term decreases the effective money supply more than issuance that skews short term, inherently tightening, without even considering the derivative effects of things tied to long term rates.
Ok John, interesting way of looking at it, but you aren’t really breaking new ground here. Andy Constan has been saying for quite some time that QT has been neutered because it hasn’t caused (indirectly of course) an increase in long dated treasury issuance. Are you saying something different? Nope. Then what exactly is the point of this post. What are you adding to the conversation? This…
Modeling Outstanding Treasuries (and issuance) through the end of FY 24
I have constructed a detailed model of month end outstanding treasuries, historical through July 31st and projected through the end of fiscal year 24 (9/30/2024). Wait John, isnt the historical part of your model just copied in from the U.S. Treasury monthly statement of the public debt (available here https://fiscaldata.treasury.gov/datasets/monthly-statement-public-debt/detail-of-marketable-treasury-securities-outstanding) , well yes it mostly is and each month end projection in my model through the end of the year and beyond is essentially a projection of what the Treasury’s monthly statement will be. But! I have added interesting additional information to each security in that statement each month. Historically and projected.
First, I have added the amount of that security held by the Federal Reserve. I did this to be able to delineate between publicly held debt and that which the Fed owns. To determine this historically for a given month end, I match (by cusip) the historical month-end outstanding treasuries against the closest historical Wednesday SOMA treasury holdings and log the fed held amount of that security on that month end. As the model switches from historical to projected (after 7/31/23), I project the Fed reinvestment “add on” for each bill and coupon auction (the rules are somewhat intricate but programmatically implementable and I have done that. Most of the August reinvestment amounts for bills and coupons are known by now and the models output for August matches exactly) and log the Fed take (if any) on the projected issuance in the future.
Second, I have added some duration related details to each security (notes and bonds) each month. Specifically, the Market Price, Macaulay Duration, Modified Duration and DV01 at that particular point in time calculated using prevailing rates (for the discount rate) at that time. To determine the rates for a particular historical point in time I apply the most relevant rate for a particular security (depending on its remaining time to maturity) using the historical Nominal Yield Curve published here ( https://www.federalreserve.gov/data/nominal-yield-curve.htm). As the model turns from historical to projection after 7/31, I use the following rates going forward (Note I also use these rates to estimate what the coupon will be on future treasury issuance that have them).
1 - 5.43, 2 - 5.03, 3 - 4.70, 4.89, 5 - 4.40, 4.36, 7 - 4.34, 4.30, 4.25, 10 - 4.21, 4.23, 4.26, 4.29, 4.32, 4.35, 4.38, 4.41, 4.44, 4.46, 20 - 4.48, 4.46, 4.44, 4.42, 4.40, 4.38, 4.36, 4.34, 4.32, 4.30, 30 - 4.28
Which roughly reflected the yield curve from Friday (8/25) morning. It has changed some since and will doubtlessly continue to change some (maybe a lot) between now and the end of the year/next September. I used a fixed curve for the entire projection period (instead of trying to model future implied curves) because I wanted to get a sense of the additional DV01 that was being added with upcoming coupon auctions without having to account for the change in daily DV01 for existing bonds due to the fluctuations in interest/discount rates. Note that I only calculate the duration and DV01 for notes and bonds.
How am I projecting the issuance? By using the roadmap the TBAC lays out in the latest quarterly refunding documents. I apply the TBAC recommended coupon issuance exactly (with the exception of 3 times this month where actual issuance exceeded the TBAC schedule; 38B in 10yr issued on 8/15 instead of the 37b recommended, and in 45b and 46b offered for the 2 and 5yr 8/31 issuance instead of the 44b and 45b recommended) through the end of the year. Note that I did not upward adjust Q4 issuance for the 10s, 2s and 5s above TBAC recommended to propagate the Treasury’s increase over recommended this month. I had to choose a path and I decided to stick exactly to TBAC guidance for Q4. If Treasury could kindly just stick to the damn plan that will work out, otherwise my coupon issuance will be a touch light through the end of the year and then perhaps beyond.
For the issuance schedule in the first 3 quarters of 2024, I use a variation of the Neutral Issuance Scenario from the TBAC Charge document released with the Quarterly Refunding Documents (found here https://home.treasury.gov/system/files/221/TBACCharge1Q32023.pdf) I follow the issuance increase pattern of +3 across the 2,3,5,10, and 30 and +2 across the 7 and 20 in Q1 and then slow it down to +2 across the 2,3,5,10, and 30 and +1 across the 7 and 20 in Q2 and no increases in Q3. The net effect brings the total coupon issuance per quarter by Q3/24 within the range of the 3 issuance scenarios detailed in the charge and I think better extrapolates what was projected in the TBAC recommended issuance for the remainder of this year (again, if Treasury could kindly stick to the damn plan). I have included the detailed coupon issuance schedule the model is using.
For bills I start with the current regular issuance volumes and increase following the recent incremental increase patterns for each bill type. This includes both keeping and increasing the 42day CMBs since the quarterly refunding documents hinted these were going to stick around through at least end of year. I increase the bill issuance amounts until it results in a net bill increase on quarter end that matches the TBAC Net bill issuance for the quarter. Specifically, I apply the following increases to bill issuance through the end of the year.
The bill increases in October may well play out a little different but unlikely in a way that materially alters the 12/31/2023 projection assuming Treasury follows the TBAC guidance for net new bills. There is no pesky debt ceiling in play so they should, adjusting of course for any deltas in realized deficit when they update the guidance in November.
For 2024, I back into the net bill number by anticipating the total borrowing needed using the Median Primary Dealer estimate for the FY 2024 Deficit of 1600b (found on page 18 in the Treasury’s presentation to the TBAC in the refunding documents found here https://home.treasury.gov/system/files/221/TreasuryPresentationToTBACQ32023.pdf) I differ with the dealers median on SOMA redemption since I think there is little to no chance QT ends by 9/24 so I apply a 681b SOMA redemption to reach a total borrowing need of 2281b. Whoa whoa John doesn’t 60b redemption x 12 months = 720b SOMA redemption? It does, but only if the Fed can hit the cap every month and it wont next June and September (more on this later). From there, I subtract the net borrowing that will happen in Q1 (FY so oct,nov,dec 23) 2024 (~852b), subtract the net new coupons between the end of Q1 FY 2024 and Q4 FY 2024 (~1730b), account for the rise in the TGA steady balance from 650b to 750b (which for purposes of this calculation I assume will also be the Q4 FY 2024 quarter end level). The end result comes to negative 199b in net bill issuance over the first 3 quarters of calendar year 24. I achieve this in the model by reducing the 42d CMBs from 75b a month down to 38b a month starting on 1/4/2024 and keep the remaining bill issuance steady. Im sorta guessing at how they would implement the net bill reduction but probably doesn’t make much difference to any macro insights that might come out of evaluating the models output.
How does the model work? As the model makes its projections each month it starts by copying in the outstanding debt from the prior month, then as it cycles through the days of the month it both adds bill and coupon issuance as well as removing matured securities (sometimes both adding and removing the same security within a month as when a 4 week bill issues on the 1st of the month). What it ends up with at month end is a picture of the outstanding treasury debt, very similar to the Treasury’s monthly outstanding debt report (plus of course the Feds portion of those securities and the market price, duration/dv01).
What does the model tell us? Two interesting things (to me at least):
1. While the Fed has been no stranger to not meeting the MBS QT cap, it will for the first time not meet the 60b UST cap in June 2024 and then again in September 2024. But wait John wont the Fed still have nearly 195b in bills outstanding come early June 2024, why wouldn’t they be able to meet a bill redemption target of only ~24b . Yeah I thought so too, so much so that I chased a phantom bug in my model for most of Saturday before finally coming to realize that its just a quirk of the way the rollover mechanics play out. As of 5/31/24 the Fed will have 195b bills remaining on their balance sheet. This checks out against the amount of bills they will have rolled off since the start of QT due to months where less than 60b in coupons mature. Critically though, only ~12.6b of the 195b will mature in the month of June (the rest have later maturities) and with only ~35.7bB in coupons maturing, the result is the UST rolloff cap will not be hit (missing by about 12b). Its even more pronounced in September when the fed only has ~30.5b coupons maturing and a measly 4b bills maturing. This illustrates why it is important to model system mechanics from the bottom up; quirks resulting from the rules play out in a way you might otherwise miss taking a top-down approach. Interesting John, but are the failures to meet UST rolloff caps in June and September 24 truly material? Meh, not so much, combined they represent only ~39B less funding pressure on Treasury for FY 24. But!!!!!!! Think of the opportunities to impress your friends at parties by being able to accurately answer in advance when the Fed will first not meet a monthly UST QT cap. You are welcome!
2. We are in the midst of a significant rise in the level of outstanding ultra short remaining time to maturity treasuries (3m to maturity or less). Significant to the tune of an ~52% increase in outstanding volume from 5/31/2023 to 12/31/2023 rising from 3.081T to 4.68T, ~1T more than the previous high in September 2020, and more than double the pre-covid norm. The level of outstanding debt in this time to maturity bucket stays mostly level through the end of fiscal year 24 ending at a value of 4.717T. To put this into some historical context Ive charted the volume of publicly held (excludes the Feds holdings) treasury debt broken out into 9 time to maturity buckets: 20-30yrs to maturity, 10-20yrs, 7-10yr, 5-7yr, 3-5yr, 2-3yr, 1-2yr, 3m-1yr, and finally 3m or less.
What are the practical implications of this? Being honest, Im not exactly sure, maybe its really nothing more than a reflection of the significant growth in treasury issuance combined with the Treasury’s temporary flood of bills vs. coupons over the last 6 months of 2023 (breaking through the normal 20% max bills guidance) which they will reel back as coupons largely fund the deficit (and QT) in 2024. That said, given that treasuries in this maturity bucket are pretty damn money like, I think its probably a good idea to consider that there may be more “liquidity” lurking out there then meets the eye with money supply measures that don’t or only partially take these securities into account. I am very much interested in the opinions of others on the topic, so if you have one, please share in the comments.
Looking forward to FY 25, the Coupon issuance would be roughly 4672b assuming my/the TBAC charge’s issuance scenarios play out, the public coupon maturities are 2502b so it could fund a deficit of 2172b. Median Primary dealers estimate for FY 25 is 1688 but throw say 650b in QT in there (more and more months will likely not hit cap if QT goes through end of FY 25) and you have a total borrowing need of 2338 so a bit of net bill issuance. Bottom line, deficits being what they are, until the Fed stops QT, the level of 3m to maturity debt is not going to meaningfully decrease, so whatever its practical implications are, they will likely be with us for a while. When the Fed does stop QT though (or if meaningful deficit reductions come out of D.C, had to introduce a little levity) then unless Treasury ratchets down coupon issuance, negative net bill issuance and a corresponding drop in the 3m to maturity debt will ensue.
I will write more in the future as I consider how to interpret the duration and DV01 data (logged in the model output from December 2018 through the projection period (September 2024). If any readers have suggestions I would love to hear them. For now, I provide the entire model output here.
Along with the specific future issuance I used.
Feel free to use as you see fit. Do bear in mind though that at least for now, all my monetary plumbing projection model development work (Ill eventually return to updating the TGA model) and writing are just a personal hobby so nobody is “checking” this work. I dot I’s and cross t’s and I am confident the projected calculations based on the provided inputs are correct but you may want to do some diligence on your own if you intend to use it in any way (if you do discover any errors please alert me and I will promptly correct). If you do find it valuable I would be grateful if you would pass it along either generally (reposts/retweets/restacks whatever the hell they are being called these days) or specifically to folks who might be interested. Not everyone is going to care to peruse year out detailed monthly projections of the outstanding treasury debt, but for those who do, they might be keenly interested. Finally, if you have questions or suggestions about how to further develop/interpret the data please reach out to me here or on twitter.
As always, Thanks for reading!
John
Awesome work John - thank you!
Thx so much John. I've been looking for a net bills issuance outstand data/projection.