The Second Policy Rate

Abstract

Research summary

A central bank sets the overnight price of money. It does not set the compensation that private balance sheets require to hold the sovereign's debt for ten or thirty years, and when that compensation moves independently of the expected policy path it tightens or loosens the economy without a committee vote. This paper calls that compensation, in its macroeconomic role, the Second Policy Rate, and asks when it has been doing the work of the first.

Using the Federal Reserve Bank of New York's Adrian, Crump and Moench term-premium estimates through 25 September 2026, we find two distinct episodes. Between September 2024 and December 2025 the Federal Reserve lowered its policy rate by 175 basis points while the 10-year Treasury yield rose 64 basis points; the model attributes a 39 basis point decline to the expected policy path and a 103 basis point rise to the term premium. In 2026 the two components have alternated. In the first half the expected path rose 47 basis points while the term premium fell 24. From the end of June to early September the pattern returned to that of 2024 to 2025: the 10-year yield rose 31 basis points, of which the term premium accounted for 25. The Federal Reserve then raised its target range in September for the first time since July 2023, lifting the expected path. The evidence strengthens the proposition of Chapra Research No. 02, The Fiscal-Supply Regime, and shows that the second price of money now sits on top of a rising first one.

We set out a measurable framework, place the present in a regime matrix, examine the fiscal channel through the gap between the effective and marginal cost of federal debt, and propose tests for credit, equities, gold and Bitcoin, each specified with the result that would refute it.

Principal findings

1. The clearest instance of long-end restriction independent of policy occurred during the 2024 to 2025 easing cycle, when a 103 basis point rise in the term premium more than offset a 39 basis point decline in expected short rates, leaving the 10-year yield higher despite 175 basis points of cuts.
2. Across 2026 as a whole the expected path accounts for most of the 89 basis point rise in the 10-year yield, but the year has two distinct phases. Between 30 June and 3 September, before the Committee acted, the term premium explained 25 of a 31 basis point rise: the long end tightened on its own again. The September increase then added to the expected path.
3. Against the Committee's own projections, the 10-year yield carries a wedge of about 1.7 percentage points. Roughly half is a gap between the policy path the market expects and the path the Committee projects; the remainder is term premium.
4. The fiscal channel operates with a lag. The average rate on marketable federal debt is 3.48 per cent, against a 10-year constant-maturity yield of 4.68 per cent on average in August and a 10-year zero-coupon yield above 5 per cent in late September. The effective rate is still below nominal growth, but representative longer-duration marginal borrowing costs are above a plausible trend rate of nominal growth. Whether the aggregate marginal funding cost crosses that threshold depends on issuance composition; the gap narrows as the debt stock reprices.
5. The decisive transition remains the Spread Handoff, when private credit spreads stop absorbing sovereign tightening and begin adding to it. High-yield spreads widened by about 60 basis points between the end of August and 1 October, partly after our cut-off, which we treat as an early warning rather than evidence of a new regime.

Chapra & Co. Research, London. This paper is research, not investment advice. Scenario probabilities are house priors and are revised as evidence changes. Figures marked Illustrative are constructed for exposition and are not observed data.

Citation convention

The paper uses author-date citations in the text and numbered notes for official and contemporaneous sources. Term-premium results are model estimates from the Federal Reserve Bank of New York's ACM dataset and are stated as one estimate among several. Chapra & Co. calculations and illustrative exhibits are labelled as such. Evidence dated after the 25 September 2026 cut-off is marked post-cut-off.

01.

Two prices of money

For most of the period after 2008, investors read monetary conditions through a single instrument. The central bank moved first, long yields followed, and asset prices responded to both. The arrangement rested on the assumption that the long yield is mainly the expected path of short rates, plus a small and stable charge for holding duration.

That assumption is reliable only while the charge is stable. A long yield can be written, conceptually, as the average expected short rate over its life plus a term premium, the additional return investors require to bear the uncertainty of holding the bond rather than rolling short-dated paper. The first component belongs to monetary policy, inflation and growth. The second reflects inflation risk, real-rate uncertainty, the supply of duration, the balance-sheet capacity of those asked to hold it, and the extent to which government bonds still hedge other risks in a portfolio.

When the second component moves on its own, households and firms feel it directly. Mortgages, investment-grade borrowing, infrastructure finance and equity discount rates are priced off the long end of the curve, not the overnight rate. The sovereign itself refinances into the long end as its debt matures. A central bank can therefore hold or lower its rate while effective financial restriction continues to rise. We refer to the long-end price, in this macroeconomic role, as the Second Policy Rate. The term is an analogy rather than an equivalence: the long end is not set by anyone, it carries no mandate, and its effect on activity depends on where long-duration pricing enters actual financing decisions.

Exhibit 1 shows why the question has become practical. From the first cut in September 2024 the policy rate fell steadily, yet the 10-year and 30-year yields drifted higher. In 2026 the effective rate was unchanged until the September increase while long yields rose further. The two periods look similar on a chart of yields, but the decomposition in section 4 shows that they are economically different.

02.

Related literature and contribution

The observation that long rates can tighten conditions on the central bank's behalf is not new. In October 2023 Federal Reserve officials, among them Lorie Logan of the Dallas Fed, argued that a higher term premium could reduce the need for further increases in the policy rate.¹ The supply channel through which public borrowing raises term premia has a well-developed theory in preferred-habitat models (Vayanos and Vila, 2021) and empirical support from Greenwood and Vayanos (2014), Laubach (2009) and Krishnamurthy and Vissing-Jorgensen (2012), while Hanson and Stein (2015) show that monetary policy itself moves long real rates substantially through term premia. Most recently, Bi, Phillot and Zubairy (2026) identify Treasury supply shocks from Treasury futures price changes in narrow windows around auction announcements and find that debt-expansion shocks raise yields across the curve by increasing term premia, tighten financial conditions and reduce private investment and production, with stronger effects when debt is accumulating quickly; maturity-extension shocks behave differently.² The distinction between default risk and risk appetite in credit spreads draws on the excess bond premium of Gilchrist and Zakrajšek (2012).

The paper contributes in four ways. First, it dates the episodes in which the long end has acted independently of policy since 2024, using a published term-structure decomposition rather than inference from yield levels. Second, it compares the model's expected path with the Committee's own projections, which separates the term premium from a component we call the Committee and market gap. Third, it proposes that the composition of real yields, rather than their level, governs the response of gold, and states the test. Fourth, it offers a sequential hypothesis for Bitcoin with dating rules fixed in advance. Each claim is paired with the evidence that would retire it, and section 12 tracks the propositions of Chapra Research No. 02 against the new evidence in the manner of Research No. 03.

03.

Framework

3.1 Decomposition

Let y_t⁽ⁿ⁾ denote the yield on an n-year zero-coupon Treasury. Following Adrian, Crump and Moench (2013), we write

where the first term is the risk-neutral yield, the average expected short rate over the bond's life, and TP is the term premium. Both components are latent. We use the Federal Reserve Bank of New York's published ACM Treasury Term Premia data, which provide fitted yields, risk-neutral yields and term-premium estimates from 1961 to the present, and treat them as one estimate among several. Kim and Wright (2005) offer an alternative with survey information, and it was the Kim and Wright series that Research No. 02 used to show the term premium moving from negative values in late 2021 to decisively positive ones in 2025 and 2026. The ACM estimate tells the same broad story with different timing, turning briefly positive in October 2023 and durably positive only from late 2024; the conclusions below should be checked against both before they are relied upon.

3.2 The Policy and Duration Wedge

The model's risk-neutral yield reflects market expectations. The Committee publishes its own. We therefore define the wedge relative to the median path in the Summary of Economic Projections, ρ_t^SEP:

The wedge is the term premium plus G, the gap between the path the market expects and the path the Committee projects. Under the model's own expectations the wedge and the term premium coincide, so the definition adds information only through G. A positive G is consistent with investors expecting more restriction than the Committee projects, whether because they doubt its inflation forecast or its reaction function. It is not a clean measure of credibility: it also contains any bias in the model's risk-neutral yield, the difference between the model's short rate and the policy-rate midpoint, and the interpolation used to build the Committee path. These components can have either sign, so G is an imprecise measure of credibility rather than a bound on it.

On 25 September the Committee path averaged 3.47 per cent over ten years, the model's market-implied path 4.36 per cent, and the fitted 10-year yield 5.14 per cent; components are rounded independently. The wedge of 1.67 percentage points divides almost equally between a Committee and market gap of 0.90 and a term premium of 0.77. Subject to the caveats above, about half of the long-end restriction now in place appears to reflect doubt that the Committee will return to its longer-run rate of 3.2 per cent as quickly as it projects; the other half is the price of duration itself. The gap is insensitive to reasonable changes in the interpolation, moving by about one basis point if the longer-run rate is reached in mid-2030 rather than at the end of that year.

3.3 A regime matrix

Crossing the direction of the expected policy path with the state of the term premium gives four states. We classify duration restriction as high when the term premium either rises by more than 50 basis points over the period or stands above its post-2000 median of 0.58 percentage points for most of it. On this definition a period is placed by its decomposition rather than by the level of yields. The 2024 to 2025 episode qualifies through the change, a rise of 103 basis points from minus 0.28; 2026 qualifies through the level, which stayed between 0.46 and 0.89 and above the median for most of the year. A classification by changes alone would place 2026 in conventional tightening; we include levels because it is the level of compensation that enters financing costs.

04.

Evidence: two episodes, one mechanism

Exhibit 4 places the current term premium in its sixty-five year context. It was persistently positive and large through the 1980s and 1990s, compressed after 2008 under central-bank asset purchases and low inflation uncertainty, and reached a low of minus 1.65 percentage points on 9 March 2020 (minus 1.29 on a monthly average basis, in July 2020). Its rise since then, of more than two percentage points, returns the long end to a regime in which investors require positive compensation to hold duration.

The level matters less for this paper than the decomposition of changes. Exhibit 5 compares six episodes. The 1994 sell-off and the 2017 to 2018 tightening were driven mainly by expected policy. The 2013 taper episode and the 2023 repricing were driven almost entirely by the term premium; in the latter the expected path fell slightly while the 10-year rose 116 basis points.

4.1 September 2024 to December 2025: easing without transmission

The Federal Reserve began cutting on 18 September 2024 and had reduced its target range by 175 basis points by December 2025. Over the same period the model's expected path fell 39 basis points, roughly what one would expect from a cutting cycle that markets had partly anticipated. Yet the 10-year yield rose 64 basis points, because the term premium rose 103. Long-end financing costs therefore rose while the policy rate fell, and the term premium offset the whole of the decline in expected short rates. Whether the net effect on activity was restrictive cannot be read from yields alone, since the front and long ends enter different financing decisions; but the transmission of easing to long-term borrowing costs failed. This is the configuration Chapra Research No. 02 described as the fiscal-supply regime, and it is the period in which the Second Policy Rate was most clearly operating.

4.2 2026: the first rate returns, the second reasserts itself

The current year is more complicated. Headline consumer prices rose 3.4 per cent over the year to August, while core prices rose only 2.4 per cent, the slowest pace since March 2021: the inflation shock is concentrated in energy, with crude oil above 100 dollars a barrel as the conflict involving Iran escalated. At its meeting of 15 and 16 September the Committee nonetheless raised its range to 3.75 to 4.00 per cent, with sixteen of eighteen participants projecting at least one further increase this year. Over the year as a whole the expected path accounts for 87 of the 89 basis point rise in the 10-year since the end of 2025, but the aggregate conceals two phases.

In the first half of the year the expected path rose 47 basis points while the term premium fell 24. Between 30 June and 3 September the configuration of 2024 to 2025 returned: the expected path barely moved, rising 6 basis points, while the term premium rose 25 and carried the 10-year yield 31 basis points higher, before the Committee had acted. The long end was tightening on its own account again, at the same time as the Treasury was expanding its buybacks of long-dated bonds, a point we return to in section 5. Only after early September, as the increase became certain, did the expected path take over. The Bank for International Settlements estimates that the US term premium rose by about 10 basis points following the end-June escalation of the conflict involving Iran, around 40 per cent of the subsequent move in yields.³ The ACM decomposition attributes a larger share to the term premium after the escalation, about 80 per cent between 30 June and 3 September. Over the BIS's full review period from 1 June the ACM decomposition attributes 39 per cent, but that coincidence in numbers compares different windows and should not be read as agreement. The gap illustrates how sensitive such attributions are to the model and to the choice of window. The ACM model measures zero-coupon yields, which rose less over the window than the constant-maturity yield cited by the BIS.

The thesis that the long end has been doing the work of policy is therefore well supported for 2024 to 2025 and for the summer of 2026; it does not explain the first half of 2026 or the September repricing, which belong to the expected path. The regime has not ended: the term premium remains near its highest level since 2014, it led the summer sell-off, and it is now combined with a rising expected path. We place the present in the double-restriction quadrant of Exhibit 3, not in the duration-supply squeeze. The squeeze becomes the relevant state again only when the Committee next eases.

05.

The fiscal channel

The Congressional Budget Office projects a federal deficit of 5.8 per cent of GDP in fiscal 2026, against a fifty-year average of 3.8 per cent, with debt held by the public rising from 101 per cent of GDP to 120 per cent by 2036. Over the same decade the primary deficit narrows from 2.6 to 2.1 per cent of GDP while net interest rises from 3.3 per cent in 2026 to 4.6 per cent in 2036.⁴ The growth in borrowing is increasingly a matter of carrying an existing stock at a higher price rather than of new discretionary spending.

The channel works through the average effective rate on the debt rather than any single marginal yield. As Exhibit 8 shows, the average rate on marketable debt was 3.48 per cent in August 2026, against a 10-year yield of 4.68 per cent that month. Because marketable debt has an average maturity measured in years, higher yields raise the average cost of the stock gradually as bills roll and coupons mature. The debt ratio is governed by the gap between the effective rate and nominal growth (Blanchard, 2019). This year's nominal growth, of the order of 6 per cent if the Committee's fourth-quarter projections for real GDP (2.3) and headline PCE inflation (3.7) are combined, comfortably exceeds the effective rate. Trend nominal growth is lower: the Committee's longer-run real growth projection of 2.0 per cent and its 2 per cent inflation target imply about 4.0 per cent, below representative longer-duration marginal borrowing costs. Whether the government's aggregate marginal funding cost crosses that threshold depends on issuance composition across bills, notes and bonds. As the stock reprices toward prevailing funding costs, the favourable gap narrows. The feedback is genuine but slow; it reduces fiscal space over years rather than producing an abrupt spiral.

The behaviour of the debt manager is itself evidence. Gross federal debt passed 40 trillion dollars in August 2026. On 19 August the Treasury announced larger liquidity-support buybacks at the long end, and from 9 September at least doubled the size of each operation, from a maximum of 2 billion dollars to at least 4 billion, through early November.⁵ In the operation of 25 September holders offered 10.5 billion dollars of 20 to 30-year bonds and the Treasury accepted 4.1 billion. Offers routinely exceed acceptances in reverse auctions, since the Treasury takes only the cheapest bonds, so the ratio is weak evidence on its own. The more telling fact is the decision itself: the issuer chose to at least double its support for liquidity in its longest bonds, which is consistent with the price-sensitive marginal holder on which the fiscal-supply proposition rests. Buybacks of this size do not materially reduce net duration supply; they are better read as a symptom of the regime than as a remedy for it.

Quantity also needs care, because a deficit and a supply of duration are different things. The amount of interest-rate risk private balance sheets must absorb depends on how the Treasury finances the deficit, through bills or coupons, and on official holdings. The Federal Reserve ended the run-off of its securities portfolio in December 2025 and has since purchased Treasury bills to manage reserves, which affects the short end far more than the long. A properly specified supply variable is net coupon issuance in ten-year duration equivalents, net of official purchases, and the maturity choice is itself a policy lever, which is the second half of the Bi, Phillot and Zubairy title.

06.

Competing explanations

An attribution of the long end to fiscal supply has to be weighed against the alternatives, five of which we consider.

Energy and geopolitics. The 2026 move coincided with an oil shock and a Gulf conflict. Inflation compensation remained stable, according to the BIS, which argues against a pure inflation-expectations story, but real-rate uncertainty and the risk of a renewed supply shock are plausible drivers of both the expected path and the rebound in the term premium from July. The energy channel and the fiscal channel are not exclusive: a supply shock raises the uncertainty premium on duration most when the marginal holder is already price-sensitive.

The reaction function. Kevin Warsh became Chair in May 2026. A new leadership raises uncertainty about how the Committee will trade inflation against activity, and the Committee and market gap in Exhibit 2 is consistent with investors pricing a less predictable reaction function. Fiscal supply cannot be separated from this channel using yields alone.

A higher equilibrium rate. If artificial intelligence raises expected productivity, the neutral rate may be higher than the Committee's longer-run policy-rate estimate of 3.2 per cent, equivalent to about 1.2 per cent in real terms. That would raise long real yields without implying any fiscal stress. The test is whether productivity, profits and investment remain strong while real yields stay high.

The hedging value of bonds. Since 2021 the correlation between bond and equity returns has been positive for long stretches. When government bonds fall alongside equities they no longer hedge portfolio risk, and investors rationally demand a higher premium to hold them (Campbell, Sunderam and Viceira, 2017). This mechanism raises term premia independently of supply and should be included in any regression that claims a supply effect.

Global repricing. Exhibit 9 shows that 10-year yields rose by similar amounts across the major markets during the BIS review period. The common movement favours global factors, including higher sovereign issuance across economies and weaker structural demand for long duration. The Bank of England's experience is informative: 10-year gilt yields have risen about 350 basis points since quantitative tightening began in February 2022, and the Bank attributes only 20 to 30 basis points of that rise to quantitative tightening.⁶

For investors holding dollar assets from Japan and Europe the comparison should be made on a currency-hedged basis. As domestic yields normalise, the yield at which a foreign investor is indifferent between home bonds and hedged Treasuries rises. Because hedging costs reflect the short-rate differential, the Committee's increases in 2026 make hedged Treasuries less attractive still, which reinforces the global argument.

07.

The Spread Handoff

A corporate borrower pays the Treasury yield plus a credit spread. If the Treasury yield rises 75 basis points while spreads narrow 25, the all-in cost rises 50 and part of the shock is absorbed. If spreads instead widen 125, the same sovereign move becomes a 200 basis point increase in funding costs (Exhibit 10). We call the transition from the first state to the second the Spread Handoff. Before it, high Treasury yields are mainly a valuation problem for long-duration assets. After it, they become a refinancing problem for the private sector. In the language of Research No. 03, the handoff is the point at which the cash-demand clock of leveraged borrowers and credit vehicles begins to run faster than their adjustment clock.

The concept requires three qualifications. First, some of the spread compression seen at high Treasury yields is mechanical: insurers and annuity writers buy all-in yield, so spreads narrow as the risk-free component rises. The handoff therefore begins when demand for all-in yield is exhausted, not at a fixed Treasury level. Second, 2022 is a counter-example that must be handled: Treasury yields and spreads rose together for much of that year without a recession or a default cycle. A useful definition requires persistence, and should use the excess bond premium to separate a change in risk appetite from a change in expected defaults. Third, the mortgage spread over Treasuries is a parallel handoff for households and deserves its own measure.

08.

Cross-asset implications

8.1 Equities: what kind of yield

The shorthand that higher yields are bad for equities does not survive historical testing. The effect depends on why yields rose. Exhibit 11 sets out the channels.

The hypothesis is refuted if, controlling for earnings revisions, rises in long yields attributed to the term premium are associated with no greater compression of equity multiples than rises attributed to the expected path. The BIS describes risk appetite as broadly resilient through the summer while leadership changed beneath the index: valuations of the large technology firms fell by about 15 per cent on a price-to-earnings basis and those of semiconductor manufacturers by about 40 per cent. Because the BIS window coincides with the term-premium-led summer, this is consistent with a discount-rate shock that had not yet become a growth shock. Since early September the expected path has led, so the risk for equities shifts toward the earnings channel through tighter policy, on top of the term-premium channel that dominated 2024 to 2025 and the summer.

8.2 Gold: the composition of real yields

The conventional relationship holds that gold falls when real yields rise, because the opportunity cost of a non-yielding asset increases (Erb and Harvey, 2013). Central banks bought 289 tonnes of gold in the second quarter of 2026, up from a revised 56.5 tonnes in the first. Yet the average London price fell 8 per cent from the first quarter to the second, and exchange-traded funds saw outflows of 45 tonnes, which the World Gold Council attributes to higher inflation and interest-rate expectations and a stronger dollar.⁸ Official buying did not prevent the price from responding to the expected-path move, and the conventional relationship remains visible.

Our hypothesis is that gold responds differently to the two components of real yields. A rise in the expected real policy rate should remain a headwind. A rise in the real term premium, to the extent that it reflects concern about the issuer rather than confidence in growth, should be a smaller headwind and may become supportive. A single quarter of average prices cannot test either half; the quarter does not contradict the first, and the second is untested. The empirical specification appears in the appendix. Because term premia are model estimates, ordinary regressions will suffer from measurement error, and we propose instrumenting the term premium with the announcement-surprise series of Bi, Phillot and Zubairy (2026), which isolates supply-driven variation.

8.3 Bitcoin: a sequential hypothesis

Bitcoin behaves at different times as leveraged exposure to global liquidity, as technology beta, and as a scarce non-sovereign asset. These identities can produce opposite reactions to the same fiscal development, which is why full-sample correlations with yields are unstable. We propose that its response to a fiscal-supply episode is sequential, and we fix the dating rules in advance so that the hypothesis can fail.

The sample is short and the holder base changed materially with the approval of spot exchange-traded funds in January 2024, so results should be read as exploratory. The proposition is narrow: Bitcoin may not hedge a fiscal-supply shock when it first tightens conditions, but may respond to the policy reaction that the tightening eventually provokes. The hypothesis is consistent with the house framework of Research No. 02, which treats digital assets as a liquidity, adoption and risk-appetite exposure inside the regime rather than a hedge against fiscal risk, and with Research No. 01, in which the July fall in long yields after the Treasury's buyback announcement produced an immediate response in Bitcoin without establishing a durable change in financing conditions. Gold and Bitcoin should therefore not be treated as a single hard-asset trade. Gold responds earlier to reserve and credibility concerns; Bitcoin, if the hypothesis holds, responds later and to policy.

09.

How long-end sell-offs end

A common forecasting rule holds that the 10-year yield peaks when the central bank finishes tightening. In none of the six cycles shown in Exhibit 13 did the 10-year peak coincide exactly with the final policy increase. In 1994, 2000 and 2018 the long end peaked 2.5 to 4.1 months before the final increase, as markets became convinced that policy would succeed or began to price the damage it would do. In 1981 and 2023 it peaked later, by 4.7 and 2.8 months. The 2007 case is economically a tie: the yield on 26 June 2006, three days before the final increase, was within one basis point of the June 2007 peak, so its sign depends on the window.

The mechanism that ends a sell-off is more informative than its timing, and we classify episodes accordingly. Credibility tops (1981, 1994) occur when markets become convinced that inflation will be contained. Growth-destruction tops (2000, 2007, 2018) occur when tighter conditions visibly damage activity and risk assets. Term-premium resets (2013, 2023) end when the new yield becomes attractive enough to bring in buyers. Intervention tops (March 2020, the gilt market in 2022) end because an official balance sheet supplies capacity the market cannot. The present episode most resembles a combination of 1994 and 2023: the expected path has risen, the term premium is elevated, and neither credit nor activity has yet weakened enough to create a strong demand for safe duration.

9.1 The crossover

The useful object is an observable crossover: the point at which the growth damage created by high yields begins to dominate the fiscal and duration-supply forces keeping them high. We make that idea testable with a decision rule fixed in advance.

We standardise each indicator, average within each basket, and define a peak signal as the first month in which the three-month change in the growth-destruction score exceeds the three-month change in the fiscal-pressure score by more than one standard deviation of their historical difference. The rule is fixed before testing. It is supported if, across the episodes in Exhibit 13, the signal precedes or coincides with the peak and rarely fires outside peak windows. It is rejected if it performs no better than the final-increase rule.

10.

Scenarios to the end of 2027

The following are conditional regimes, not point forecasts. Each is defined by the mechanism that would end the sell-off and the indicators that would confirm it, so that probabilities can be revised mechanically as data arrive.

11.

A monitoring framework

The framework is useful only if it can be revised against data. We will publish the following measures with each update, without combining them into a composite score until the weighting has been tested.

Each update will classify the evidence as strengthening, leaving unchanged or weakening the thesis, and record the reason in the revision ledger.

12.

Revision, refutation and house view

12.1 Place in the Chapra research programme

This paper is the fourth in a sequence. Chapra Research No. 01, Bitcoin 2026: Countertrend Rally, Cyclical Context and the Path to a Durable Bottom (21 August 2026) observed that long-duration yields fell abruptly after the Treasury's buyback announcement of 29 July and that high-beta assets responded at once, while cautioning that an isolated improvement in market liquidity does not establish a persistent fall in real financing costs. Chapra Research No. 02, The Fiscal-Supply Regime (2 September 2026), defined the regime as the interaction of recurrent supply disturbances, persistent sovereign borrowing, an elevated price for duration and reduced monetary-policy room, alongside private balance sheets intact enough to prevent a household-led deleveraging; its second conditional path was monetary easing that does not restore the old duration regime. Chapra Research No. 03, When Liquidity Becomes Structure (18 September 2026), located fragility in the race between the cash-demand clock and the adjustment clock, and concluded that supply determines what must be held while structure determines how violently holdings may have to change.⁹

The present paper supplies the price that connects them. No. 02 asked who absorbs growing sovereign duration and at what discount rate; the decomposition here measures that discount rate and dates the episodes in which it moved independently of policy. The Spread Handoff is the point at which the Second Policy Rate begins to run the cash-demand clocks of leveraged credit holders faster than their adjustment clocks, the mechanism No. 03 documented in March 2020, the gilt market in 2022 and Credit Suisse.

Our working view since mid-September has been that the Committee is tightening into an energy supply shock while core consumer prices are rising at their slowest pace since 2021, and that the risk markets have not priced is a forced reversal of policy in 2027. Such a reversal would provide the conditions for the decisive test of No. 02. Because No. 02's cut-off was 1 September 2026, the 2024 to 2025 episode lies within its sample: this paper quantifies it with a different model but cannot count it as an out-of-sample confirmation. The first genuine out-of-sample test is the next easing cycle. If the Committee is obliged to ease in 2027 and the long end does not follow, No. 02's second conditional path will have been confirmed out of sample; if the long end compresses normally, the structural interpretation should be scaled back.

12.2 Research No. 02 revisited

Following the convention of Research No. 03, each proposition is tracked against the evidence in this paper and classified as an established mechanism, case-consistent evidence, live or unresolved evidence, or counterevidence. Evidence dated after No. 02's cut-off of 1 September 2026 is marked as such.

12.3 What would retire the thesis

We would abandon the Second Policy Rate framework if any two of the following occur. During the next easing cycle, long yields converge toward the expected path while sovereign issuance remains heavy; term-premium estimates from several model families fall and stay low; private financing costs respond almost entirely to the policy rate rather than the long curve; credit, housing and investment prove insensitive to elevated long real yields; the decomposition of real yields adds no explanatory power for gold; or the staged hypothesis adds nothing for Bitcoin beyond conventional liquidity measures.

12.4 House view as of 5 October 2026

1. The long end is materially restrictive. Real 10-year yields averaged 2.40 per cent in August, a level matched since 2008 only in October 2023, and the 30-year real yield averaged 2.97 per cent, the highest since that series began in 2010.
2. Research No. 02 is strengthened. The 2024 to 2025 cycle showed the term premium offsetting policy easing at the long end, and the summer of 2026 showed it tightening conditions before the Committee acted. The first half of 2026 and the September repricing were led by the expected path.
3. About half of the current wedge against the Committee's projections is a gap between market and Committee paths. Part of that gap is likely to reflect doubt about the Committee's projections, so credibility as well as supply is being priced, although the measure is imprecise.
4. Fiscal supply is part of the explanation but not all of it. Energy, a new Chair's reaction function, a possibly higher neutral rate, the loss of bonds' hedging value and global issuance all contribute.
5. Representative longer-duration marginal borrowing costs exceed plausible trend nominal growth, while the aggregate marginal funding cost depends on issuance composition. Fiscal space narrows as the debt stock reprices.
6. There is no broad credit event. The widening in high-yield spreads into 1 October, partly post-cut-off, is an early warning, and credit is the indicator to which we give most attention, pending a tested weighting.
7. We do not call the top. The balance of risks leaves room for a term-premium overshoot, which is our second most likely scenario. A Committee forced to reverse course in 2027 while the long end holds would confirm Research No. 02.
8. Gold and Bitcoin are different trades. Gold has so far responded conventionally to the expected-path move; Bitcoin is more likely to suffer in the tightening phase and to respond to the eventual policy reaction.

13.

Conclusion

The question is no longer whether the Federal Reserve controls interest rates; it never controlled the whole curve. The more useful question is whether the part it does not control has become sufficiently independent to frustrate the transmission of the part it does. Between September 2024 and December 2025 it did: the Committee eased by 175 basis points while the term premium rose enough to leave the 10-year yield higher. The summer of 2026 repeated the mechanism in miniature before the Committee itself returned to tightening. Monetary conditions are therefore being set simultaneously by a committee and by the marginal balance sheet willing to warehouse sovereign duration.

That does not make the Treasury yield a policy instrument. It makes it a constraint. The next easing cycle will show how structural that constraint has become. If long rates follow the Committee lower, this episode belongs principally to the history of monetary tightening. If they do not, the fiscal-supply regime will have crossed a more consequential threshold, and the sovereign financing requirement will have acquired a monetary character of its own.

14.

Methodological appendix

A1. Data and decomposition

Zero-coupon yields are from Gürkaynak, Sack and Wright (2007), updated by the Federal Reserve Board. Term premia and risk-neutral yields are from the Federal Reserve Bank of New York's published Treasury Term Premia dataset (ACM), with observations through 25 September 2026. Monthly policy and constant-maturity yields are from the H.15 release. Because term premia are latent and model-dependent, every result in sections 4 and 8 should be checked against Kim and Wright (2005) and survey-based expectations before publication of the empirical companion.

A2. Committee and market gap

The Committee path is constructed by linear interpolation between the midpoint of the prevailing target range and the year-end SEP medians, reaching the longer-run median at the end of 2030 and holding it thereafter. The ten-year average is compared with the ACM risk-neutral 10-year yield on the same date. Because the SEP is a median of individual projections rather than a forecast of the Committee's decisions, the gap should be read as indicative.

A3. Gold specification

Monthly gold returns in dollars are regressed on changes in the expected real policy component, changes in the real term premium, the trade-weighted dollar, a fiscal-risk proxy and official purchases. The primary hypothesis is that the coefficient on the expected component is negative and that the coefficient on the term premium is less negative and may turn positive in fiscal-stress regimes. The term premium is instrumented with announcement-surprise supply shocks to address measurement error.

A4. Spread Handoff

Episodes are classified by the joint sign of three-month changes in the 10-year yield and in investment-grade and high-yield spreads, requiring persistence of at least two consecutive quarters for a handoff. Outcomes tested are investment growth, default rates, employment, equity returns and recession probability over the following four quarters, with the excess bond premium substituted for raw spreads as a robustness check.

A5. Exhibits marked Illustrative

Exhibit 8's two nominal growth bars and Exhibit 10's spread arithmetic are constructed for exposition. All other figures are drawn from the sources stated beneath each exhibit.

15.

Notes

1. Lorie K. Logan, Financial conditions and the monetary policy outlook, speech, Federal Reserve Bank of Dallas, 9 October 2023.
2. Huixin Bi, Maxime Phillot and Sarah Zubairy, Treasury Supply Shocks: Propagation Through Debt Expansion and Maturity Adjustment, Federal Reserve Bank of Kansas City Research Working Paper 26-04, 2026.
3. Bank for International Settlements, Yields climb, yet risk appetite holds firm, BIS Quarterly Review, September 2026. Review period 1 June to 3 September 2026.
4. Congressional Budget Office, The Budget and Economic Outlook: 2026 to 2036, February 2026.
5. U.S. Department of the Treasury, Treasury Announces Increased Sizes of Nominal Long-End Liquidity Support Buybacks Beginning September 9, press release, August 2026; operation results of 25 September 2026; Treasury FiscalData, Debt to the Penny, gross federal debt.
6. Bank of England, Dave Ramsden, speech on quantitative tightening, September 2026, citing the July 2026 Monetary Policy Report estimate that cumulative quantitative tightening added about 20 to 30 basis points to the 10-year gilt yield.
7. Freddie Mac, Primary Mortgage Market Survey, 1 October 2026; Federal Reserve Bank of St. Louis, FRED, ICE BofA US Corporate Index OAS and ICE BofA US High Yield Index OAS; underlying source ICE Data Indices, LLC.
8. World Gold Council, Gold Demand Trends: Q2 2026.
9. Chapra Research No. 01, Bitcoin 2026: Countertrend Rally, Cyclical Context and the Path to a Durable Bottom, 21 August 2026; Chapra Research No. 02, The Fiscal-Supply Regime, 2 September 2026 (cut-off 1 September); Chapra Research No. 03, When Liquidity Becomes Structure, 18 September 2026. All at chapraandco.com/research.

16.

References

1. Adrian, T., Crump, R. K., & Moench, E. (2013). Pricing the term structure with linear regressions. Journal of Financial Economics, 110(1), 110-138.
2. Bi, H., Phillot, M., & Zubairy, S. (2026). Treasury supply shocks: propagation through debt expansion and maturity adjustment. Federal Reserve Bank of Kansas City Research Working Paper 26-04.
3. Blanchard, O. (2019). Public debt and low interest rates. American Economic Review, 109(4), 1197-1229.
4. Campbell, J. Y., Sunderam, A., & Viceira, L. M. (2017). Inflation bets or deflation hedges? The changing risks of nominal bonds. Critical Finance Review, 6(2), 263-301.
5. Erb, C. B., & Harvey, C. R. (2013). The golden dilemma. Financial Analysts Journal, 69(4), 10-42.
6. Federal Reserve Bank of New York. (2026). Treasury Term Premia, Adrian, Crump and Moench model, daily and monthly data. Applied Macroeconomics and Econometrics Center.
7. Gilchrist, S., & Zakrajšek, E. (2012). Credit spreads and business cycle fluctuations. American Economic Review, 102(4), 1692-1720.
8. Greenwood, R., & Vayanos, D. (2014). Bond supply and excess bond returns. Review of Financial Studies, 27(3), 663-713.
9. Gürkaynak, R. S., Sack, B., & Wright, J. H. (2007). The U.S. Treasury yield curve: 1961 to the present. Journal of Monetary Economics, 54(8), 2291-2304.
10. Hanson, S. G., & Stein, J. C. (2015). Monetary policy and long-term real rates. Journal of Financial Economics, 115(3), 429-448.
11. Kim, D. H., & Wright, J. H. (2005). An arbitrage-free three-factor term structure model and the recent behavior of long-term yields and distant-horizon forward rates. Finance and Economics Discussion Series 2005-33, Federal Reserve Board.
12. Krishnamurthy, A., & Vissing-Jorgensen, A. (2012). The aggregate demand for Treasury debt. Journal of Political Economy, 120(2), 233-267.
13. Laubach, T. (2009). New evidence on the interest rate effects of budget deficits and debt. Journal of the European Economic Association, 7(4), 858-885.
14. Vayanos, D., & Vila, J.-L. (2021). A preferred-habitat model of the term structure of interest rates. Econometrica, 89(1), 77-112.