Introduction
In addition to providing pool-level A factors on GNMA MBS pools on the 4th business day (BD4) of each month, StoryBook also offers a loan-level data set that is derived from the loan liquidation data released by Ginnie Mae on the same day. Loan-level liquidation data is released around 9pm on BD4 by Ginnie Mae and is typically available in our application within an hour. The upshot is that users have the ability to perform loan-level analysis starting the night of BD4 instead of waiting until the evening of BD6, which is when Ginnie Mae releases a more complete set of loan-level data.
Constructing the BD4 Data Set
On BD6 of every month, Ginnie Mae provides loan-level detail on balances, prepayments, and a number of other fields for its single-family MBS pools. On BD4, however, the only loan-level disclosure currently provided by Ginnie Mae is a liquidation file which provides a snapshot of all loans that paid off during the relevant reporting period along with a code that specifies the payoff reason (voluntary/involuntary). No other loan-level fields are updated. BeginningNovember 2022, MachineSP has made available through StoryBook a BD4 Ginnie Mae loan-level file (which we can think of as a loan-level 'A' factor tape approximation to the 'B' factors of the BD6 loan-level set) by stitching together the previous factor month's BD6 loan data with the current month's loan-liquidation information. Essentially, we selectively update certain columns in the previous month’s loan-level data set to reflect the passage of one month plus incorporate the information contained in the current loan liquidation file. Empirically, our research shows that this synthetically constructed BD4 loan-level data set provides an excellent approximation at the aggregate level to realized prepayments and buyout rates (it is typically within 0.1% CPR/CBR of aggregate BD4 pool-level prepayments/buyouts, see below for further details). At a more granular level, the approximation may be less precise to the extent that there is substantial variability in curtailment rates from cohort to cohort. The BD4 loan-level data set is of course replaced in StoryBook by the full set of loan-level disclosures on BD6.
Specifically, the following loan-level attributes are currently updated in the BD4 set loan-level data set:
Loan age
Remaining Maturity
The current balance for non-paid off loans (the previous month’s current balance less (a) the scheduled payment and (b) a projected curtailment amount which is taken to be the previous month's realized curtailment rate)
Incentive
Current LTV
Other attributes are assumed to be the same as the previous month including the delinquency status, the servicer of record etc. In case of a servicing transfer, even though the new servicer will be incorrectly recorded in the BD4 data set, analyzing servicer-level prepayments and buyout rates will still provide accurate answers because these payoffs are often assumed to be driven by the servicer of record as of the previous factor date. On the other hand, delinquency-based analyses may be off to the extent that the current factor month's delinquency rates are very different from the previous factor date.
BD4 Estimates vs BD6
To give a sense for how accurate BD4 estimates are, Table 1 compares the BD4 numbers versus the realized BD6 prepayments for a selection of Ginnie Mae coupons; the buyout rate (CBR) and the curtailment rate (CCR) are also displayed. As we can see, the agreement between BD4 and BD6 is excellent; the 0.1% CPR differences are attributable to rounding and/or very small differences between forecasted and realized curtailments.
Table 1. Prepayments on select Ginnie Mae Coupons: Feb 2024 Factors
Source: MachineSP/Ginnie Mae
Even at fairly granular levels of disaggregation, the agreement between BD4 and BD6 still holds up (Table 2)
Table 2. Prepayments on G2M 9 WALA 6.5s by Agency: : Planet and Lakeview (Feb 2024 Factors)
Source: Machine SP/Ginnie Mae.