How to calculate MBS principal payments to the Fed each month to evidence the FED is conducting QT as announced

Hello again,

TLDR – Actual MBS principal payments received by the FED in April, May and June align almost exactly with the next mid-month to mid-month reinvestment schedule offering further evidence that the FED is conducting QT as advertised and also conducted the post QE, pre-QT period as advertised (reinvesting only the MBS principal payments received)

For MBS, the FED is conducting QT by only reinvesting principal payments on MBS each month to the extent they exceed the cap 17.5->35b. Prior to the start of QT but after QE had fully tapered the FED aimed to keep its MBS BS level by fully reinvesting principal payments but not buying more. Thus, calculating the principal payments themselves is pretty important. This post intends to demonstrate how to do that using the  SOMA data https://www.newyorkfed.org/markets/soma-holdings (SOMAH) and the operational results from the trading desk https://www.newyorkfed.org/markets/ambs/ambs_schedule. (OPRESULTS)

Prior to that though, a quick review of how and when MBS principal payments come in is in order. MBS on the SOMAH falls into one of four types, FHLMC GOLD, GNMA I, GNMA II, and UMBS (both FNMA and FHLMC). The principal payments for each type are paid once a month by the agencies responsible for collecting them from the mtge servicers. Gold and GNMA I come on the 15th, GNMA II on the 20th, and UMBS on the 25th. (or next business day, which usually wont be a problem since the SOMA updates as of Wednesday each week but might occasionally enter in like Christmas)

Calculating the principal payment for each type is usually straightforward, all you have to do is subtract the amount of a particular type of MBS for the week that spans its payment  from the amount of that same type of MBS from the week before. So for the FHLMC Gold (GOLD) payment in May 2022. (GOLD pays on the 15th)

GOLD MBS on SOMAH week of 5/11 = 153,334,515,515.54 (week prior to payment)

GOLD MBS on SOMAH week of 5/18 = 150,086,164,958.80 (week including payment)

So the payment for GOLD on the 15th is 3,248,350,556.74. Simple enough, and usually that is all you have to do. However, occasionally MBS of the same type that was previously purchased (or potentially sold in a roll) will settle/hit the BS in the same week that the payment for that type comes in. In May this happened for GNMA II. (payment on 20th)

GNMA II MBS on SOMAH week of 5/18 = 563,324,370,789.26

GNMA II MBS on SOMAH week of 5/25 = 558,920,195,155.66

So at first glance it appears the GNMA II payment was 4,404,175,633.60 Except that a number of GNMA II MBS purchases made between 3/28 – 4/18 settled on 5/19 totaling 6,001,000,000 offset by a few GNMA II MBS sales (rolls) from 5/16-5/17 that also settled on 5/19 totaling  1,160,000,000 for a net addition of 4,841,000,000 that hit the BS that week. Adding 4,841,000,000 to the  SOMAH BS difference, 4,404,175,633.60 = 9,245,175,633 which is the true GNMA II payment received on the 20th of May.

Where did we look to see when MBS was going to settle? From the operational results spreadsheets (contractual settlement date column) from the trading desk. Fortunately, the FED is no longer buying GNMA I or GOLD MBS, so no need to look for them, and usually the settlement dates for GNMA II and UMBS fall outside the week that spans the payment date, but it can and does sometimes overlap so you have to be careful

Completing the payments for may

GNMA I MBS on SOMAH week of 5/11 = 9,361,502,960.54

GNMA I MBS on SOMAH week of 5/18 = 9,194,425,923.26

GNMA I May payment = 167,077,037.28

And

UMBS MBS on SOMAH week of 5/18 = 2,002,307,376,581.81

UMBS MBS on SOMAH week of 5/25 = 1,980,369,249,227.95

UMBS May payment = 21,938,127,353.86

Accordingly, the total principal payments in May are

21,938,127,353.86 -- UMBS

+ 167,077,037.28 –- GNMA I

+ 9,245,175,633 – GNMA II

+ 3,248,350,556.74 – GOLD

= 34,598,730,581.48 – Total May payments

What is interesting about the total May payment is that it matches exactly (rounded) to the 34.6B the FED announced (https://www.newyorkfed.org/markets/ambs_operation_schedule#all-schedules) MBS purchases for the period of 5/13-6/13 (last period before MBS reinvestment reductions started on 6/15.

Prepayment data for the month prior would be known to the FED by early in the Month, so it certainly appears that for May, the Fed announced MBS reinvestment = the amount of principal collected, which is consistent with their policy between QE and QT.

This same analysis applied to April yields 40.1b in total payments, a hair shy of the 40.2b purchases announced for mid-april to mid may but close enough for my mind.

Also, the same analysis applied to June yields Total June payments of 30,449,015,065.94 Subtracting the reinvestment cap of 17.5b = 12,949,015,065.94, just shy of the 13b of purchases announced 6/14 – 7/14.

Based on the actual MBS principal payments remitted to the FED in April, May and June matching nearly exactly to the announced purchases conducted by the fed for the next monthly period. It certainly seems the FED is conducting QT as advertised and conducted the period between QE and QT as advertised as well.

Hopefully you found this article helpful. Thanks for reading!

John

Technical Appendix.

You can repeat this analysis just using all the operational results spreadsheets from the trading desk and the SOMAH values week to week (with a whole bunch of filtering and autosumming). You will make your life vastly easier though,  if you pull all the operational results spreadsheets into a database so you can query it. The same holds true for the SOMAH. I personally used the full detailed holdings spreadsheets for each week of the SOMAH (you can find it at the bottom of the MBS tab in the View Details link next to the Grand Total. It gives you the full list of MBS securities held by the FED (cusip and all) as of that particular week. To group those securities into the four buckets described in this post you can use the following like filter on the [security description] column. (in t-sql)

where [security description] like 'UMBS%'

where [security description] like 'FHLMCGLD%'

where [security description] like 'GNMA I %'

where [security description] like 'GNMA II%'

one final note for totality, starting 4/27 there are a few securities with a [security description] of FANNIE. Im not really sure what this is, but the values are so small (2-20million in aggregate) that they make no meaningful impact on the results so I ignored them for now.