Fannie/Freddie MBS: pricing an implicit guarantee

On the page Fannie Mae devotes to its own securities, one sentence has stood for decades: the certificates and payments of principal and interest are not guaranteed by the US Government and do not constitute a debt or obligation of the United States. The text is plain. It does not invite interpretation.

On 19 May 2025, Moody’s nonetheless cut the senior debt ratings of Fannie Mae and Freddie Mac from Aaa to Aa1. The stated reason: the downgrade of the US government on 16 May. Three days apart, an identical move, no separate analysis of the two companies’ credit risk.

The whole question of privatising the two GSEs sits between those two facts. The law says there is no guarantee. Prices, ratings, prudential rules and federal budget accounting say there is one. Taking Fannie and Freddie out of conservatorship means asking the market to choose.

Five lines to keep

At 30 June 2026 the two companies carried $7,850.7 billion of assets and $194.3 billion of equity, or 2.5% of assets.
The Treasury holds senior preferred stock on them with a liquidation preference of $380.8 billion, more than their entire equity.
The January 2021 letter agreement bars any exit until CET1 capital reaches 3% of assets. It is currently negative by $47 billion.
No amendment to the Treasury-GSE agreements has been made since 2 January 2025, despite eighteen months of statements about an IPO.
The CBO puts the budgetary cost of new guarantees over 2026-2035 at $81.8 billion, which means the federal budget already counts the guarantee as real.

The law says no

The legal architecture dates from the crisis. On 30 July 2008, the HERA statute created the Federal Housing Finance Agency. On 6 September, the agency placed Fannie Mae and Freddie Mac into conservatorship with their boards’ consent. The next day, the Treasury signed a senior preferred stock purchase agreement with each, the SPSPAs.

Those agreements do not create a guarantee on the securities the two companies issue. They create a Treasury commitment to fill their negative net worth, in tranches, in exchange for senior preferred stock and warrants over 79.9% of the common equity at a nominal price. The distinction matters: an investor holding a Fannie Mae MBS has no recourse against the Treasury. It has recourse against Fannie Mae, which holds a solvency line with the Treasury.

In practice the commitment was drawn. The initial $100 billion per company was raised to $200 billion on 6 May 2009, then replaced by formulas on 24 December 2009. The third amendment of 17 August 2012 replaced the fixed dividend with a sweep of net worth. The letters of 21 December 2017 allowed a $3 billion reserve each; those of 30 September 2019 raised it to $25 billion for Fannie and $20 billion for Freddie.

Across the whole period, cumulative draws came to about $191 billion and dividends returned to the Treasury to about $301 billion, according to Laurie Goodman’s work for the Urban Institute published in April 2025. The balance does not extinguish the claim: dividends paid never reduced the liquidation preference, which has kept growing instead.

Five places where the guarantee is written

No federal text guarantees agency MBS. Five arrangements nonetheless treat them as if it did, and each leaves a measurable trace.

Sources: Fannie Mae (MBS page), Moody’s Ratings (19 May 2025), AGNC Agency MBS At-a-Glance 1Q26, 12 CFR 217.32(c), Federal Reserve H.4.1 of 13 August 2026, CBO (July 2025). l0g synthesis.

The first item is a rating. Moody’s did not run a standalone credit analysis of the two companies before downgrading them: it replicated the sovereign move. Fitch had already cut them to AA+ on 2 August 2023, the day after downgrading the United States. The method is explicit, and it amounts to treating these companies as an extension of the public balance sheet.

The second item is a yield spread. At 31 March 2026, the current coupon agency MBS spread over government bonds stood at 124 basis points according to AGNC’s quarterly dashboard, on a market of $9.5 trillion outstanding at end-December 2025. That spread mainly compensates prepayment risk and negative convexity, not issuer default risk. Corporate debt rated AA without public support would trade on other terms.

The third item is a prudential rule. Paragraph 217.32(c) of title 12 of the Code of Federal Regulations requires banks to apply a 20% risk weight to a GSE exposure other than equity or preferred stock. The same rule assigns 100% to preferred stock issued by a GSE. Within a single section, the regulator therefore separates the GSE’s debt and guarantee, treated as quasi-sovereign, from its capital, treated as ordinary private risk. It is the most explicit dividing line in existence between what the state covers and what it does not.

The fourth item is a stock. At 12 August 2026, the Federal Reserve held $1,930.9 billion of MBS out of $6,471.5 billion of securities held outright, nearly 30% of its portfolio. The subject was covered here from the central bank’s income statement angle, in the piece on how the Fed makes and loses money. A central bank does not hold $1,900 billion of an asset whose credit risk it judges to be private.

The fifth item is budgetary, and the heaviest. The CBO treats Fannie and Freddie as federal entities, because the FHFA controls them and the Treasury owns most of their securities. In its July 2025 note it puts the cost to the federal government of the new guarantees the two companies will issue between 2026 and 2035 at $81.8 billion, on a projected volume of about $15 trillion. That cost comes from underpricing it estimates at 6 to 10 basis points a year: the GSEs charge less for their guarantee than a private actor would at the same risk. The US federal budget therefore already incorporates the guarantee, and its price. The method is set out in the guide on reading CBO projections.

The claim that blocks the door

Public debate talks about an IPO. The accounts talk first about a ranking of claims.

At 30 June 2026, Fannie Mae reported $4,332.3 billion of assets, $116.5 billion of equity and a thirty-fourth consecutive profitable quarter, with $4.0 billion of net income for the quarter and $7.7 billion for the half. Freddie Mac reported $3,518.3 billion of assets, $77.8 billion of equity and $3.8 billion of quarterly net income, up 61% year on year. Serious delinquencies remain low: 0.58% on single-family at Fannie, 0.60% at Freddie.

These companies make money, quarter after quarter, and have for years. Yet they hold no regulatory capital.

On Fannie Mae’s balance sheet, the Treasury’s senior preferred stock is carried at $120.8 billion, with a liquidation preference of $234.2 billion, against $227.0 billion six months earlier. At Freddie Mac the carrying value is $72.6 billion and the liquidation preference $146.6 billion at 30 June, rising to $150.4 billion on 30 September 2026. The mechanism is automatic: since the January 2021 letter, every dollar of retained earnings raises the Treasury’s claim by the same amount.

As a result Fannie Mae’s available CET1 capital is a $33 billion deficit and Freddie Mac’s a $14 billion deficit, $47 billion negative in total. Two companies that have accumulated $194.3 billion of book equity remain below regulatory thresholds, because almost all of that equity is already promised to their controlling shareholder.

Sources: second quarter 2026 releases and financial statements of Fannie Mae and Freddie Mac, FNMA and FMCC closing prices at 17 August 2026 (stockanalysis.com). The 3% threshold applies the January 2021 letter agreement condition to reported combined assets: l0g calculation, the exact regulatory base may differ.

The market prices that subordination without ambiguity. At 17 August 2026 Fannie Mae stock traded at $6.11 for a $35.2 billion market capitalisation, Freddie Mac at $5.57 for $18.0 billion, $53.1 billion between them. Against a $380.8 billion senior claim, common shareholders are not valuing a dividend stream: they are buying an option on a political decision.

The 3% threshold, and the step to clear

The exit condition is written in black and white. The January 2021 letter agreement provides that no exit will occur until all material litigation relating to the conservatorship is resolved or settled, and until the company holds common equity tier 1 capital of at least 3% of its assets.

Applied to combined assets reported at 30 June 2026, the 3% bar represents $235.5 billion. Available CET1 capital is negative by $47 billion. The gap to close therefore reaches $282.5 billion. The two companies generated about $15.1 billion of net income in the first half of 2026 alone, an annual run rate near $30 billion. At that pace, and at a constant balance sheet, the gap closes in about a decade. Since assets keep growing, the 3% bar rises alongside, which stretches the horizon further.

What matters is where those earnings go. The Treasury’s preferred stock sits on the balance sheet at its original value, $120.8 billion at Fannie and $72.6 billion at Freddie, not at its liquidation preference. Regulatory capital does rebuild, dollar by dollar. But since the January 2021 letter, every dollar retained raises the Treasury’s liquidation preference by the same amount. The companies are accumulating capital whose counterpart is already owed to their controlling shareholder: on the day of release, the equity counter and the public claim counter will have advanced by the same figure.

Hence the pivot in every serious scenario. The CBO models two in its July 2025 note, and both start with the same operation: the Treasury gives up its claim in its current form. In the first, it converts its preferred stock into common shares, exercises its 79.9% warrants, and the companies raise $162 billion of new capital from private investors. In the second, the FHFA places the companies into receivership, creates new entities, transfers assets and liabilities to them, and sells shares in the new corporations. In both cases the CBO puts the combined market value after recapitalisation and release at $368 billion, with a $162 billion capital shortfall to be filled by outside investors.

The question is therefore not whether the state will sell its stake. It is how much it agrees to write off before selling. A decision worth several hundred billion dollars, taken administratively, without a vote in Congress.

Eighteen months of announcements, zero amendments

The contrast between the volume of statements and the stillness of the documents is striking.

On governance, the move was fast. On 19 March 2025, FHFA Director Bill Pulte replaced fourteen directors across the two boards and named himself chairman of each. The same person then held the conservatorship power and chaired the bodies he is meant to supervise. On 2 June 2026 he was named acting Director of National Intelligence while keeping his FHFA role and both chairmanships.

On announcements, the sequence is long. An IPO floated by the president in May 2025, with continued implicit guarantees claimed. A combined valuation around $500 billion and a 5 to 15% stake sale discussed in autumn 2025. On 6 June 2025, fourteen senators led by Elizabeth Warren asked for a pause and a full briefing of Congress. On 25 June 2026, a package of three bills tabled by Representative Scott Fitzgerald finally proposed a legislative framework for release, without an explicit guarantee. It went to committee.

On binding documents, nothing. The last amendment to the Treasury-GSE agreements dates from 2 January 2025, under the previous administration, and it runs against release. It restores the Treasury’s right to consent to any exit from conservatorship and to any receivership. It requires the FHFA to publicly solicit input on options for ending the conservatorships and their effects on housing markets, then to brief the Financial Stability Oversight Council before any release. It leaves the expiry of the Treasury’s warrants unchanged at 7 September 2028.

To date, no such public solicitation has been launched. The January 2021 conditions on capital and litigation remain in force. The door is locked by texts nobody has amended.

One event did change the nature of the conservatorship. On 8 January 2026, the president directed the two companies to buy $200 billion of MBS in order to bring mortgage rates down. The 30-year rate fell to 5.99% the next day. Retained portfolios followed: $150 billion at Fannie and $139 billion at Freddie in February 2026, against $93 billion a year earlier at Fannie, with $11 billion of net purchases in February alone after $15 billion in January, according to monthly data reported by National Mortgage News. The retained portfolio cap set at $225 billion per company since end-2022 becomes a live constraint.

The operation says something more important than its size. Two companies supposedly preparing for private life were used as a substitute monetary policy instrument, with the money they were meant to accumulate to reach their capital threshold. The move pushes their balance sheet closer to the Fed’s at the exact moment official rhetoric wants to pull them away from it.

The likely price of an exit

A less certain guarantee costs something. The only question is how much, and to whom.

The main channel runs through the guarantee fee charged on every securitised loan. It averaged 65.2 basis points in 2024 on single-family, against 65.5 in 2023, according to the FHFA’s annual report published on 22 December 2025. The Urban Institute estimates that an exit from conservatorship would take that fee to about 76 basis points under utility-style pricing, and about 92 basis points under market pricing, excluding the cost of any Treasury credit line.

The gap, 11 to 27 basis points, looks modest. Against the 30-year rate of 6.67% recorded by Freddie Mac’s weekly survey at 13 August 2026, it represents about 1.6 to 4% of the interest burden. Against $15 trillion of new guarantees over ten years, it changes in nature.

All these estimates assume an orderly transition, with federal backing maintained in some form. They do not measure the scenario in which the market stops believing in the guarantee. In that case, the five arrangements described above turn at once: the rating decouples from the sovereign, the spread widens, the 20% bank weighting loses its justification, the Fed loses the legal basis for crisis purchases, and the CBO has to reclassify $7,800 billion of commitments. Nobody can size that shift, because it has never happened. The general method for reading such a move is set out in the guide on credit spreads.

Three paths, three bills

Sources: CBO, “Seven Things to Know About CBO’s Budgetary Treatment of Potential Changes to Fannie Mae and Freddie Mac”, July 2025; NAR, presidential directive of 8 January 2026. l0g synthesis.

The comparison brings out a point the IPO debate obscures. All three routes move the bill, none removes it, and none turns the implicit guarantee into a voted commitment. The Treasury itself wrote as much in its January 2021 blueprint on next steps for reform: it supports legislation authorising an explicit, full faith and credit federal guarantee on MBS, and considers legislative reform preferable to administrative action. Five and a half years later, that legislation does not exist.

Checkpoints

Four indicators allow this file to be tracked without relying on statements.

The first is documentary. Any real exit starts with an amendment to the Treasury-GSE agreements, published by the FHFA and the Treasury. As long as the SPSPA page stops at 2 January 2025, nothing substantive has moved.

The second is procedural. The FHFA must publicly solicit input and brief the FSOC before any release. Publication of that solicitation would be the first operational signal.

The third is accounting. The Treasury’s liquidation preference is disclosed quarterly in both companies’ financial statements. The CBO projects it at nearly $400 billion by end-2026, against $341 billion in December 2024. A stabilisation or a fall would signal a renegotiation under way.

The fourth is market-based. The current coupon MBS spread, at 124 basis points at end-March 2026, continuously measures the value investors attach to the guarantee. A durable widening without an equivalent move in private credit would mark a repricing of public support.

One deadline deserves noting. The Treasury’s warrants expire on 7 September 2028. In January 2025 the Treasury said it anticipated a future agreement to push that date back and avoid a disorderly exit. No such agreement has been published. If the deadline approached without extension, the administration would have to choose between exercising, letting a right over 79.9% of the equity of two companies holding $7,850 billion of assets lapse, or amending the agreements in haste.

The guarantee nobody signed has survived eighteen years of conservatorship, two sovereign downgrades and three administrations. It will probably survive this one. Its strength lies precisely in what makes it unavowable: written nowhere, it has never had to be defended before a vote.

On the same ground, the l0g investigation into the Federal Home Loan Banks describes the third floor of the same architecture, the one whose public subsidy quietly changed beneficiary. US mortgage credit was also covered through the home insurance premium and through the commercial real estate refinancing wall.

Sources

Limits

Balance sheet and capital data come from quarterly disclosures as at 30 June 2026 and incorporate no later event. Share prices cited are those of 17 August 2026, on a thin over-the-counter market where both stocks trade off exchange.

The 3% exit threshold is applied here to the consolidated total assets reported by the two companies. The January 2021 letter agreement refers to 3% of assets without adopting the base definition of the ERCF prudential framework, which applies its own adjustments. The $282.5 billion gap shown is therefore an order of magnitude, not an enforceable regulatory requirement.

The estimates of the effect on guarantee fees come from an April 2025 Urban Institute paper and assume an orderly transition with federal backing maintained. They do not model a loss of confidence in the guarantee, a scenario for which no reliable public estimate exists.

The $81.8 billion budgetary cost is computed by the CBO on a fair-value basis, a method distinct from the Federal Credit Reform Act used elsewhere in the federal budget. It rests on guarantee volume projections made in July 2025 and on the assumption of law in force at that date.

Finally, any assessment of the implicit guarantee is indirect by nature. It is inferred from observable behaviour, ratings, spreads, prudential rules, holdings and accounting conventions, not from a state commitment that could be cited and enforced.