Reverse Engineering QT

TLDR – FED is executing QT exactly as it said it would

Many folks are unimpressed with the effect QT has had on the FED’s BS thus far. Some are accusing the FED of lying about conducting it altogether. Others accusing it of being too slow. Universally though, these folks don’t seem to understand its mechanics. This post seeks to explain them and demonstrate that the FED is executing QT as it said it would. Note though it isn’t trying to defend the FEDs plan for QT, personally, Im in the go faster camp, so for those who would call me a shill for the FED, im not, just an observer/explainer of reality (with sources).

Lets start with what the FED actually said (and did not say) about its plan for QT by reviewing the may FOMC minutes https://www.federalreserve.gov/monetarypolicy/files/fomcminutes20220504.pdf (page 10). “FED intends to reduce securities holdings over time primarily by adjusting amounts reinvested of principal payments received from securities held in SOMA. Starting 6/1 principal payments will be reinvested to extent they exceed monthly caps…” 30b a month for treasuries ramping to 60 after three months and 17.5 for MBS, ramping to 35b after three months. NOTE: FED did not make any promises about the size of the BS reduction, just the reduced reinvestment of principal. This is a key point of misunderstanding with many folks. These reduced reinvestment purchases will certainly lead to BS reductions, eventually, but with a very significant lag for MBS. The correct measure for QT vs what the FED promised is reduced reinvestment purchases NOT BS reduction, though it will lead to that eventually.

So how can we prove the FED has reduced its reinvestment purchases? By looking here: https://www.newyorkfed.org/markets/soma-holdings (SOMAH)

Starting with UST. Coupon USTs typically mature the 15th and last day of the month. You can see the specific UST the FED owns at SOMAH (e.g. click on T-Notes and T-Bonds or TIPS tab), including their value and maturity date. You can also adjust the week-of date for the SOMAH to see how it looked (anc changed) from week to week. Note from 7/6 to 7/13, no coupon UST were schedule to mature so their balances stayed the same. On 7/15 though, a 13.6b note/bond cusip 9128287C8 matured and 9.6b TIPS cusip 912828TE0 matured. Accordingly, on 7/20 SOMAH for TIPS we see its total value down 9.6b and nothing in the total change in par value from prior week because FED did not reinvest the maturing 9.6b TIPS (meeting 9.6b of the 30 cap). For notes and bonds though, while 13.6b rolled off, the FED reinvested 12.1b of it in 3 separate UST (cusips 91282CEY3, 91282CEP2, 912810TG3) (so net meeting another 1.5b of the 30 cap). FED has ~28b in notes/bonds maturing on 7/31. It will let ~18.9b roll off and will reinvest the remaining 9.1b. This will reflect on SOMAH for 8/3 when it is released. One can do the same detailed analysis on June and see the net 30b reduction in UST BS.

Reduced UST reinvestments immediately translate to BS reduction for UST. MBS reduced reinvestments take a few months. Lets dig in.

The FEDs detailed MBS holdings can also be viewed at SOMAH (MBS tab), but these do not account for MBS the FED has agreed to purchase but the purchase has not settled yet. The FED publishes the schedule of MBS it intends to and actually buys here. https://www.newyorkfed.org/markets/ambs/ambs_schedule. The operational result spreadsheets show in detail what the FED has already purchased and when it will settle (and accordingly hit the SOMAH/BS). If one filters on the settlement dates that fall within a SOMAH reporting week and sum the purchases netting any sales (not many but a few) you can anticipate how much new MBS will hit the BS each week, though you have to do it spanning the operational results for the prior 2-3 months since MBS can take a few months to settle. For example, last week 7/27 the GNMA II 30 yr MBS on the SOMAH grew from 547.8b to 557.7b so ~10b additional GNMA II MBS. Where did that come from? Look to the operational results for the period 5/13-6/13, filter on 7/21 settlement date and sum the rows (since they are all purchases in this case). ~6.5b of GNMA II MBS purchased between 5/18 – 6/13 was scheduled to settle on 7/21. Do the exact same thing for the operational results for 6/14-7/14 and you find another 5.4b for a total of 11.9b. So why only 9.9b hitting the BS. Im not totally sure but it is probably that there were some sales as part of roll operations that settled quickly and will show on the 7/15 – 8/11 operational results when they are released. Or possibly some of the MBS purchases did not settle.

So ~10b of GNMA II MBS was added to the SOMAH on 7/27 but the total MBS dropped by ~9b? What gives? Note that on the SOMAH MBS falls into one of four types, FHLMC GOLD, GNMA I, GNMA II, and UMBS (both FNMA and FHLMC). This matters because payments on those securities come in once a month but on different days. Gold and GNMA I come on the 15th, GNMA II on the 20th, and UMBS on the 25th. (or next business day). MBS doesn’t mature like UST which returns all the principal on the maturity date. MBS is amortized so a little principal is returned each month for each mortgage gradually winding down the UPB over the 15//30 yr maturity, unless the borrower prepays the mortgage because they refinanced it or sold the house to move. Then the remaining principal is returned all at once. Monthly borrower payments and any prepayments are collected by the mtge servicers, forwarded on to agencies (e.g. FNMA) who then remit that to investors like the FED once a month on the dates above. When the FED receives that payment the SOMAH/BS is reduced accordingly. For example, the SOMAH for 7/20 Fed had 942.5b of 30yr FNMA MBS, and on 7/27 933.5b. if you add up across all UMBS holdings the total was ~19b reflecting the payment the FED received on 7/25. Offsetting the 19b payment with the 10b new GNMA MBS that settled you get a net of -9b for the week.

So to review, MBS BS increases when MBS purchased between a few days and a few months ago settles (typically in the middle of the month) and drops when payments/prepayments are received from investors on set dates (BS can also drop when MBS sales, usually as part of rolls, settle). So just looking at week-to-week changes without knowing why it changes will be confusing. Plus, since the bulk of the MBS purchases takes 1-2 months to settle. There is a significant lag. As a result, almost all the MBS adding to the SOMAH in June and July was actually purchased prior to the start of QT. But that is all passed now so MBS BS reductions should reflect 17.5b in August. But do note that there will be a similar lag for when QT ramps to a 35b cap, will just reflect a 17.5b drop in September and October as the MBS purchased in July-Aug settles.

Speaking of that 35b cap, will the FED ever hit that? Maybe, but it couldn’t hit it today and wont hit it unless prepayment speeds rise a bit. Today the FED is receiving about ~29.5b in principal payments/prepayment from all its securities each month. So when the upped cap comes into play the FED will just stop buying MBS altogether (they are purchasing about 12b a month today to reinvest the ~29.5b payments received minus the cap). They might eventually actively sell MBS but they have announced no concrete plans to do so of yet.

One other note on UST, when the cap hits 60b, if there are not enough coupon maturities to hit that 60b then the FED will let Bills mature without reinvestment. Bills mature all month long so UST BS will drop outside of the 15th and 31st.

To sum up:

1.      Fed has reduced reinvestments of maturing treasuries by 30b a month initially. This is proven by comparing the SOMAH UST holdings on  June 1 vs. July 6. And will further prove out comparing July 6 and Aug 3.

2.      Fed has reduced reinvestments of maturing MBS by 17.5b a month initially. This is proven by the 21.5b reduction in FED MBS purchase 5/13-6/13 (34.5b) vs 6/14 – 7/14 (13b). 4b of that is slower prepayments, 17.5b is due to QT cap

3.      Reduced UST reinvestments hit the BS quickly with precise mechanics explained above.

4.      Reduced MBS reinvestments hit the BS with ~2 month lag with precise mechanics explained above.

Hopefully this post will help those brave enough to venture though its details better understand what is actually happening week to week on the BS for UST and MBS and settle any remaining controversy on whether the FED is doing QT as it said it would. One can disagree with what the FED is doing (more or less QT) but its nonsense to argue with the reality that it is doing it.

Thanks for reading, and feel free to provide feedback or note any inaccuracies.

John