We decided to take a look at lenders that have the option to sell loans as well as portfolio to see what we can learn about their decisioning and selection process.  Happily, we have access to LendingPatterns™.

The robust peer selection tool allowed us to create a custom peer group of 2,461 lenders that either sold or portfolioed between 1-99% of their originations.  Lenders that portfolioed 100% of their originations were excluded from the peer group since they were not in a position to decide which loans to sell or retain on the books.  Independent mortgage companies were also excluded. These filters applied throughout my analysis. The dataset we used is the 2018 HMDA Snapshot, published by the CFPB on August 30, 2019.

My analysis looks at:

  1. Investor code (Portfolioed vs GSEs vs other investors) by action taken;
  2. Percent held in portfolio within DTI and CLTV ranges; and
  3. Borrower/loan characteristics for portfolioed versus secondary market loans


In conclusion, there are countless ways to cut the data in LendingPatterns™ and learn about the mortgage market.  The expanded 2018 HMDA fields give us an opportunity to drill down in ways previously unavailable to examine the secondary marketing and asset decision process.

Investor code by action taken

Within the sample, about 51% of the originated and purchased loans, combined, were held in portfolio (sample size 5,071,107).

Interestingly, when we analyzed originated (sample size 3,787,248) and purchased loans (sample size 1,283,859) separately, the data indicates a preference to sell purchased loans: only 25% of purchased loans were portfolioed vs. 59% of originated loans.  So, when lenders buy loans that other lenders underwrote, these loans are more likely to be slotted for sale to other investors than when institutions underwrite and close the loan.

Combining originated and purchased activity, the most popular specific secondary market outlet was Fannie Mae (16%) followed by Freddie Mac (11%) and Ginnie Mae (10%).

If we isolate purchased loans, the share that went to GSEs is higher than among originated loans. This is due in part to the fact that, as noted, purchased loans are more likely to remain in portfolio. Ginnie Mae and Fannie Mae each purchased about 25% of the total purchased loan volume.

Percent held in portfolio within DTI and CLTV ranges (originated loans only)

We looked across the various Debt-to-Income (DTI) and Combined LTV (CLTV) ranges (e.g., CLTV over 100%) for substantial differences in the frequency of loans that were sold.

This section filters on originated conventional, first lien, owner occupied, and 1-4 family loans, and excludes commercial loans, lines of credit, and reverse mortgages.

We observed that there are attributes in the data that are associated with particularly high percentages of portfolioed loans. For example, consider these small segments of the market:

  • Among loans with a CLTV of 100% or greater, 74% were held in portfolio (sample size: 53,807); and
  • Among loans with a DTI over 60%, 78% were held in portfolio (sample size: 8,113).

These shares of originated loans that remain in portfolio are a bit higher than the 59% across all products, and DTI and CLTV ranges.

Borrower/loan characteristics for portfolioed versus secondary market loans (originated only)

The table below shows the profile of the application characteristics for portfolio loans compared to loans sold in the secondary market using this peer group.  The sample here is originated loans only.

Borrower/loan Characteristic Loans kept in Portfolio
(sample size 2,252,150)
Sold to Investor Outlets
(sample size 1,535,098)
Conv. Loan Type % 98% 78%
Purpose: Home Purchase % 31% 70%
Adjustable Rate Mortgage % 51% 3%
Average 1st Lien Loan Amount (thousands $) $367.57 $245.79
Jumbo Loan Amount % 11% 1%
Upper Applicant Income % 59% 46%
Wholesale plus Correspondent %* 4% 10%
High spread/subprime loans % 6% 4%

This table shows that, relative to portfolioed loans, loans that go to investors are more likely to be:

  • government products,
  • home purchase loans,
  • fixed-rate mortgages,
  • smaller loan amounts,
  • wholesale and correspondent* loans, and
  • loans to low/moderate/middle income borrowers, as opposed to upper income borrowers.

Meanwhile there is only a small difference between the two sets of loans in terms of the percentage that are above the high-spread threshold.

* Note: This means that the application was either submitted to a third party other than the reporting institution, or was initially payable to a third party other than the reporting institution, or both.