mortgagetape™How Call Reports telegraphed the biggest MSR migration in a decade
·
Written by Mortgage Tape Team — a group of industry analysts leveraging our proprietary mortgage-domain language models to synthesize and decode housing data.
📌 Takeaways by role
- MSR investors & agency MBS desks: Between year-end 2019 and year-end 2024, the bank-held 1-4 family mortgage servicing book shrank from $2,749B to $2,288B — $461B gone on an aggregate RC-S basis across the ~80 banks that file Call Reports. Wells Fargo alone ran off $585B. Bank of America ran off $102B. The next tier of exits — Santander, Banco Popular — follows the same 40-60% drawdown shape. On a share basis, banks lost 11.2 points of the attributable agency-servicing book over five years. The leading-indicator claim: on a 30-day post-quarter-end filing cadence, bank Call Reports quantify a trade’s actual UPB settlement before most trade press runs its post-close recap.
- Bank credit analysts & FIG investors: The migration is a split story. On GSE pools the flow was mostly bank-to-bank — JPMorgan Chase +$338B, PNC +$113B, Truist +$96B, USB +$74B on the 3-agency basis. On Ginnie it was almost entirely bank-to-non-bank — Freedom +$208B, Lakeview +$167B, Rithm/NewRez +$100B, Carrington +$88B of Ginnie book absorbed. Only Ginnie bled bank share meaningfully (-16.7 pts vs -11 to -13 pts on the GSEs). The capital-rules mechanism bites hardest on Ginnie MSR, and the data shows exactly that pattern.
- Housing-policy analysts & market-structure researchers: A substantial piece of Wells’s $585B drop doesn’t live in agency pools at all. The 3-agency subset of Wells’s book shrank ~$226B; the remaining ~$359B is private-label, portfolio, HELOC, and FHLB-held — invisible to any pool-level file but visible on RC-S. For Bank of America it’s the same pattern, more extreme: of its $102B drop, ~$106B is residual (BofA grew ~$4B on the 3-agency basis). FFIEC Call Reports see portions of the mortgage servicing market that agency pool disclosures cannot. The FFIEC + Fannie + Freddie + Ginnie cross-reference is the authoritative public picture; neither side alone captures the story.
The filing before the news
Wells Fargo’s Call Report at June 30, 2023 shows the bank’s 1-4 family serviced-for-others book at $624B — a $58B drop from the quarter before. That filing lands at the FFIEC in late July 2023. It is the single largest single-quarter decline in Wells’s servicing book in the five-year window.
By the time it hits the public FFIEC repository, trade press has run speculation about Wells’s intentions for seven months. Three dollar-figure anchors have been reported: the Jan 10, 2023 strategic-exit announcement (no dollars), Q4 2022 rumors of a $100-$150B conventional package plus a $100B Ginnie package (unconfirmed), and the April 14, 2023 Q1 earnings call at which CFO Mike Santomassimo confirmed that Wells had marketed “approximately $50 billion of loans serviced for others” and expected the sale to close later in 2023.
The $58B drop in that July 2023 filing is the first full-quarter settlement number. It confirms the dollar figure Wells revealed on the earnings call and quantifies what actually left the balance sheet.
That shape — Call Reports as the trade’s post-close quantitative record — repeats across the five-year window. Trade press runs on announcement dates and rumors. Call Reports confirm what actually closed, every quarter, for every reporting bank, including the trades that are never individually announced. For every one of Wells’s declines in the table above, there is an identifiable Call Report filing that lands four weeks after quarter-end and prints the drawdown explicitly. For most of them, no trade-press article ever ran with the right dollar figure.
The aggregate scope
The $461B aggregate decline is a story about two banks: Wells shed $585B, BofA shed $102B, together $687B on an RSSD-stable basis. That more than fully explains the industry total — meaning the rest of the ~80 banks together added about $226B back on new originations held for servicing.
The year-by-year shape has two distinct declines separated by a flat middle:
| Year-end | Bank serviced for others ($B) | Filers |
|---|---|---|
| 2019-12-31 | 2,749 | 72 |
| 2020-12-31 | 2,360 | 85 |
| 2021-12-31 | 2,331 | 84 |
| 2022-12-31 | 2,414 | 83 |
| 2023-12-31 | 2,455 | 81 |
| 2024-12-31 | 2,288 | 82 |
2020’s -$389B is refi-wave runoff. The COVID rate collapse to the 2.6-3.0% range pulled refinance volume forward from any book held long enough to be in the money; a bank’s servicing book shrinks mechanically when half its loans prepay. This leg is not an MSR-trade story.
2021-2023 is near-flat. New originations held for servicing roughly offset continued runoff. Wells was already shrinking its book through this period — the preceding four quarters of 2022 showed single-digit $B/quarter declines — but no clean industry-level signal.
2024’s -$167B is the second wave, driven by capital rules. The Basel III endgame NPR had been released in July 2023 and remained the single largest regulatory event hanging over bank balance sheets through 2024. Combined with G-SIB surcharge recalibration and the aftermath of SVB — which lifted the regulatory premium on any asset that could be sold at a lift — this wave of 2024 trades looks less like opportunistic rate-driven selling and more like forced-rebalancing. The 2024 decline is also the one where BofA and the second-tier exits (Santander, Banco Popular) show up on the aggregate.
Who exited, who held
RSSD-stable, 2019-Q4 vs 2024-Q4:
| Bank | 2019 ($B) | 2024 ($B) | Δ ($B) | Δ (%) |
|---|---|---|---|---|
| Wells Fargo | 1,079 | 495 | -585 | -54% |
| Bank of America | 178 | 76 | -102 | -57% |
| US Bank | 226 | 217 | -9 | -4% |
| Santander | 15 | 7 | -8 | -51% |
| Banco Popular PR | 14 | 8 | -5 | -38% |
| Citibank | 52 | 49 | -3 | -7% |
| Truist | 219 | 218 | -1 | -0.3% |
Wells and BofA are the exits. Santander and Banco Popular are smaller cuts at similar percentage severity — a profile that reads as the same mechanism (money-center-scale operational exit) applied to a smaller base. The rest of the top-filer list is operationally stable.
Bank share collapsed on all three agencies — Ginnie worst
Non-bank growth across the three agencies was dramatic — a $2.9T combined gain for the top-15 non-bank servicers between 2019 and 2024 — but a lot of that was organic, from new originations held for servicing while the agencies’ total books grew. The sharper cut is share.
| 2019-12 bank share | 2024-12 bank share | Δ (pts) | |
|---|---|---|---|
| Fannie SFP | 42.0% | 29.4% | -12.6 |
| Freddie SLLD | 44.3% | 33.2% | -11.1 |
| Ginnie | 25.2% | 8.5% | -16.7 |
Combined across all three sources, banks’ share of the attributable 3-agency servicing book went from 34.9% to 23.7% — a 11.2-point drop over five years. If banks had held 34.9% through 2024, they would be servicing $2.32T of attributable agency book instead of the $1.58T they actually service — a $740B forfeited-growth floor on how much agency book either moved to non-banks or was captured by non-banks on new originations.
The Ginnie leg is the sharpest story. Bank absolute dollars on Ginnie fell from $351B to $188B — a $164B drop — while non-banks absorbed every dollar of growth plus took bank share. On Fannie and Freddie, banks held or grew in absolute dollars but lost share to non-banks who grew faster. The mechanism is the same (capital-driven rebalancing), but Ginnie is where the clean bank-to-non-bank migration shows up.
The 16.7-point Ginnie share drop vs ~11-point GSE share drops is consistent with the capital-rules mechanism. Ginnie MSR carries a 250% risk weight under current standardized Basel vs 100% for GSE MSR, and the Basel III endgame proposals go further on buffer requirements for Ginnie-weighted exposures. A capital-constrained bank facing a choice of which part of its MSR book to shed shed Ginnie first. Over five years, they shed almost all of it.
Where it went
Top gainers on the 3-agency basis, 2019-12 → 2024-12:
| Servicer | Type | Δ ($B) | FN Δ | FR Δ | GN Δ |
|---|---|---|---|---|---|
| Mr. Cooper (Nationstar) | Non-bank | +506 | +255 | +221 | +30 |
| Lakeview | Non-bank | +437 | +144 | +126 | +167 |
| JPMorgan Chase | Bank | +338 | +128 | +229 | -19 |
| Freedom Mortgage | Non-bank | +312 | +34 | +69 | +208 |
| PennyMac | Non-bank | +298 | +40 | +174 | +83 |
| Rocket Mortgage | Non-bank | +256 | +109 | +108 | +40 |
| Rithm / NewRez | Non-bank | +255 | +11 | +145 | +100 |
| Onslow Bay (Annaly) | Non-bank | +187 | +110 | +77 | 0 |
| UWM | Non-bank | +182 | +77 | +87 | +18 |
| PNC Bank | Bank | +113 | +61 | +55 | -3 |
The split story is sharpest if you read the FN/FR/GN columns side-by-side.
On the GSE pools, the flow was mostly bank-to-bank. JPMorgan Chase picked up $357B of combined Fannie + Freddie book across the five-year window, more than any single non-bank. PNC picked up $116B of combined GSE book. Truist picked up $91B. US Bank picked up $72B. Collectively, these four big-bank GSE buyers absorbed ~$640B — comparable to all of Wells + BofA + Santander + Banco Popular’s combined decline on the bank side. A chunk of the GSE MSR that left Wells’s book went to other large banks who had room on the capital stack and infrastructure in place.
On Ginnie, the flow was almost entirely bank-to-non-bank. JPMC’s Ginnie line is actually negative. PNC’s is negative. The four big-bank GSE buyers either held flat or shed Ginnie book over the window. The Ginnie absorbers are a tight cluster of non-banks whose business models are built around Ginnie: Freedom (+$208B Ginnie), Lakeview (+$167B Ginnie), Rithm/NewRez (+$100B Ginnie), Carrington (+$88B Ginnie), PennyMac (+$83B Ginnie), Planet Home (+$61B Ginnie), Rocket (+$40B Ginnie). Lakeview’s position at the top of Ginnie absorption is itself a story — most of it is Ginnie co-issue aggregation that pool-level files encode as “MULTIPLE” and only loan-level attribution resolves cleanly.
A capital-rules reading explains why: Ginnie MSR’s higher risk weight makes it unattractive for any bank running tight on standardized-approach capital, but the Ginnie non-banks don’t face those weights — they fund MSR off warehouse-and-securitization rails that treat it as an asset-rich-asset-light balance sheet item instead of a risk-weighted exposure.
The invisible residual
Wells’s $585B drop is not primarily agency MSR. Decomposed:
| Source | 2019 ($B) | 2024 ($B) | Δ ($B) |
|---|---|---|---|
| Fannie SFP | 350 | 179 | -171 |
| Freddie SLLD | 55 | 108 | +53 |
| Ginnie | 188 | 80 | -108 |
| 3-agency total | 593 | 367 | -226 |
| Call Reports RC-S total | 1,079 | 495 | -585 |
| Private-label / portfolio / FHLB residual | ~486 | ~127 | -359 |
Wells actually grew its Freddie book over the window — probably reflecting the retail-focused book Wells kept — while shedding $171B on Fannie and $108B on Ginnie. The 3-agency sum accounts for 39% of Wells’s total RC-S decline. The remaining $359B lives outside agency pools entirely — jumbo, HELOC, portfolio, private-label trusts, FHLB-held. For Wells specifically, that residual is largely the Wachovia-legacy private-label book being wound down.
BofA shows the same pattern more extreme:
| Source | 2019 ($B) | 2024 ($B) | Δ ($B) |
|---|---|---|---|
| Fannie SFP | 0 | 0 | 0 |
| Freddie SLLD | 5 | 12 | +7 |
| Ginnie | 4 | 1 | -3 |
| 3-agency total | 9 | 13 | +4 |
| Call Reports RC-S total | 178 | 76 | -102 |
| Private-label / portfolio residual | ~169 | ~63 | -106 |
BofA technically grew its tiny 3-agency book and shed $106B of residual — the Countrywide-legacy private-label and whole-loan book wind-down continuing through the window. For a reader anchored on agency MBS disclosures, BofA’s $102B drop is invisible. For a Call Reports reader, it is one of the two defining stories of the window.
The operational takeaway is a methodology point: neither side of the FFIEC-plus-agency cross-reference captures the full picture alone. Agency pool disclosures catch the GSE and Ginnie flow with clean counterparty attribution. Call Reports catch the private-label, portfolio, and jumbo wind-downs that pool files cannot see. The two together are the authoritative public picture; either alone is incomplete.
Case studies: Wells, Wells, USAA
Three named trades span the window.
Wells → public strategic announcement, Jan 10, 2023. Wells Fargo released a strategic update titled “Wells Fargo Announces Strategic Direction for Home Lending — A Smaller, Less Complex Business” (Business Wire, Jan 10, 2023). The release named the exit from the Correspondent business and said Wells planned “to reduce the size of its Servicing portfolio”. No dollar figure attached. Trade press ran rumors over the next quarter of a $100-$150B conventional package plus $100B of Ginnie, none of them confirmed by Wells. The Call Report at March 31, 2023 showed a $13B drawdown — material but not yet the kind of number the rumors anticipated. Wells’s balance sheet was moving slower than the rumor cycle.
Wells → Mr. Cooper, ~$50B package, revealed Apr 14, 2023. On the Q1 2023 earnings call, CFO Mike Santomassimo confirmed that Wells had “successfully marketed mortgage servicing rights for approximately $50 billion of loans serviced for others” and expected the sale to close later in 2023. Bidding had closed in late February / early March. The buyer was later confirmed as Mr. Cooper. The Call Report at June 30, 2023 showed the $58B drop — the first filing to quantify the trade’s actual settlement. Call Reports are published on an approximately 30-day lag; the filing landed in late July. Mr. Cooper’s balance sheet grew correspondingly over 2023.
USAA → Mr. Cooper, 2021. USAA Bank sold roughly its entire MSR portfolio to Mr. Cooper, effectively exiting mortgage servicing. The Call Reports data shows USAA’s Ginnie book dropping from $36.5B to zero over the 2020-2021 window. Trade press coverage of the deal size and counterparty landed in late 2021; the balance-sheet record is one of the cleanest wind-down shapes in the full window.
For each of these, the Call Reports filing confirmed the trade’s actual settlement UPB at a 30-day lag after quarter-end — before most trade-press coverage ran its post-close recap. Trade press ran announcements and rumors ahead of the Call Report in each case; it also ran corrections, counterparty confirmations, and dollar-figure refinements months after. The Call Report is the primary quantitative record of what actually closed.
Why now
Two mechanisms, both acting through the back half of the window, make the 2023-2024 wave sharper than the 2020 refi-driven leg.
Rate shock turned MSR into a sellers’ market. The Fed’s 2022-2023 hiking cycle lifted MSR values dramatically — a 30-year mortgage servicing right is a short-prepay-option asset, so when the forward curve pushes refinance incentive out of the money, the asset is worth more. Banks sitting on long-duration books that were underwater during the refi wave of 2020-2021 found themselves holding the same book at a saleable premium by mid-2023. The economics finally lined up for a planned exit.
Capital rules tightened on the same timing. The Basel III endgame NPR was released July 27, 2023 and proposed meaningful increases in capital charges on MSR exposure, particularly for Ginnie-weighted MSR. G-SIB surcharge recalibration proposals in parallel threatened to raise the capital cost of a large mortgage franchise further. SA-CCR had already come into effect in 2022 and reshaped counterparty-credit charges for warehouse and gestation activity. The regulatory signal running through 2023-2024 was unambiguous: a dollar of MSR on a bank balance sheet was going to cost more capital next year than this year.
Combined, these mechanisms made MSR divestiture both forced and well-priced. The data pattern — Wells pulling the trigger first in January 2023, the industry aggregate taking its biggest step down in 2024 as the Basel III NPR comment period closed and capital planning for 2025 locked in — is what you would expect under this reading.
Closing — what Call Reports do that nothing else does
For a desk pricing MSR flow, forecasting non-bank balance-sheet capacity, or evaluating bank mortgage-franchise valuation, the quarterly Call Report is the primary quantitative record. It arrives on a known cadence (30 days after quarter-end), covers every reporting bank, and prints UPB serviced for others at a line-item granularity (RCFDB805) that no other source matches. It captures trades the trade press misses. It captures private-label and whole-loan declines that agency pool files cannot see. It quantifies in dollars what rumor cycles characterize in adjectives.
What the FFIEC data cannot do alone is tell you where the flow went. The counterparty resolution is in Fannie SFP’s servicer_name, Freddie SLLD’s servicer_name, and Ginnie servicer-of-record. Reconciled against RC-S, those sources resolve both sides of the migration — who shed, who absorbed, and (crucially) the residual that neither bank earnings calls nor agency pool files disclose.
For MSR investors, bank credit analysts, and anyone else trying to understand the structural reshaping of the mortgage servicing stack over the past five years, running the two sides of this cross-reference every quarter is the baseline. The last five years were a $461B migration out of the banking system — observable in near-real-time to anyone looking at the quarterly filing before the news cycle caught up.
Try it yourself. Ask Mortgage Tape:
- “Wells Fargo RC-S line RCFDB805 quarterly 2019 through 2024”
- “Top 3-agency servicer gainers 2019 to 2024 with Ginnie breakout”
- “Which banks hold the largest private-label residual outside agency pools”
The FFIEC + Fannie + Freddie + Ginnie cross-reference used here is reproducible on our analytical instance.