New Data Shed Lights on Agency Mortgage Originators

The secondary mortgage market is generally focused on mortgage servicers. These institutions bear substantial risk, and their strategies regarding prepayments and foreclosures impact borrower welfare and investor returns. Yet they are not the whole picture. A buyer of a mortgage pool owns a collection of loans serviced by a variety of financial institutions. However, there are borrower characteristics impacting loan performance that are out of the control of the servicer, notably the strength of underwriting beyond what is captured in disclosed statistics such as credit scores. The originators matter.

The issue we face is that loan-level data disclosed by the Agencies does not, in any case, correspond precisely to the notion of “originator” for either the GSEs or FHA besides servicers. In the case of FHA, we see the entities that pool loans to sell in accordance with Ginnie Mae regulations, known as “issuers”. For the GSEs, this is sellers, the firms that deliver loans to Fannie Mae and Freddie Mac to securitize. The sellers and issuers may be the loan originators, but many of these loans may have been purchased from other firms (through what is known as the “wholesale channel”).

The purpose of this note is to see to what extent we can utilize supplemental data regarding originators to enhance the information we have at our disposal to assess the valuation of Agency pools. Our ability to perform this analysis relies on our expertise in normalizing massive amounts of mortgage company information. We start with FHA and move on to the GSEs, paying close attention to the institutional differences.

FHA
For some time, we have mentioned the FHA portfolio snapshot data[1]. This data provides the originators for the universe of all FHA loans. We can usefully compare the distributions of the originators of the loans in this dataset to the issuers of Ginnie Mae pools containing FHA loans.

The GSEs
As we explored alternative datasets, we noticed that The Security and Exchange Commission (SEC) has for some time been collecting originator information at the pool level for loans securitized by Fannie Mae and Freddie Mac. This discovery has profound implications for enhancing transparency in mortgage production activity.
The first issue that arises is concerned with data coverage and quality. The most straightforward approach to this is accomplished by comparing the balance and loan counts of the loans originated in each pool to the number of loans serviced across pools securitized by each agency:

​It turns out that the two measures are quite close, with 98.9% of pool-level coverage within 10% and 95.5% within 5% of the Agency data from 2013 to Q1 2024. Fannie Mae’s ratio experienced some bumpiness in 2017 but more recently, has held close to 100%. Freddie Mac experienced a brief period (3 months) of a sharp drop in the ratio in 2014. Interestingly, ever since Covid struck in early 2020, Freddie Mac has experienced a small but persistent deviation averaging 2.7%. Much remains to be explained here.
​
In order to benchmark the new data, the table below provides comparisons between originator and seller/issuer across product types

In the case of the GSEs, the loan counts are extremely close, with a difference of under 700 loans out of almost two million originated. In the case of FHA, there is a difference of 3,800 or 0.5%. This difference primarily reflects the number of FHA loans originated that are held on depositor balance sheets. This factor arises in the FHA category because the data source provides total originations while the SEC data provides information only for originations of loans that are securitized into pools.
​
Bank/nonbank
As a final benchmarking exercise, we look at the bank/nonbank composition for originations and sellers of conforming loans:

The shares have been very similar since mid-2020, one year after UMBS implementation, implying that there is minimal bias in this regard in the originator data. The Freddie nonbank seller data has ticked up recently vs. nonbank originations data, however.

The Homebuilders
An interesting use case of the new data can be found in the category of homebuilders. These are an interesting subset of the mortgage universe for a variety of reasons. In particular, the “mortgage winter” environment, where existing home sales are suppressed by current residents who have low-rate mortgages, doesn’t apply in the new home market. As a result, the share of this segment out of total purchase originations and issuance in agency pools has risen sharply over the past two years:

A key takeaway is that we can now observe that the share of this sector in the securitized market is larger than can be obtained directly from the Agency disclosures.

One straightforward but interesting exercise is to calculate the average loan amount for home builders.  It is not surprising that the builder loan balances are significantly higher than other loans.
​
What next?
Looking forward, the application of the SEC originations data to conforming pools adds a new dimension to the analysis of pool performance. Our Cohort Analyzer will soon be able to break down prepayments and delinquencies by both servicer and originator, offering investors a new approach to determining valuation. We’ll be back.
​
Note: For a more complete version of this report including detailed information on the largest institutions in the various categories discussed above, reach out to [email protected].

---

[1] We referred to this dataset very recently in an analysis of the use of down payment assistance programs for FHA loans: https://www.recursionco.com/blog/a-deep-dive-into-fha-dpas.