Unconventional monetary policy spillovers and the (in)convenience of Treasuries

Layer 1 — Overview

Research Question

The paper asks why unconventional monetary policy (UMP) spillovers from the European Central Bank (ECB) to the U.S. Treasury yield curve vary so substantially over time, and whether the time-varying “convenience” of Treasuries — their non-pecuniary premium as the world’s preeminent safe asset — can explain that variation. The core claim is that a declining convenience yield on Treasuries makes them more substitutable with other safe sovereign bonds, thereby amplifying the portfolio-balance channel through which foreign large-scale asset purchases (LSAPs) depress U.S. term premia.

Data and Methodology

The authors use high-frequency identification of ECB monetary policy surprises following Altavilla et al. (2019), defined as the first principal component of intraday changes in 1-, 3-, 6-, 12-, and 24-month euro OIS rates plus 5- and 10-year German and French bond yields, measured in the 10-20 minute window bracketing each ECB decision press conference. Surprises are normalized so that one unit raises the 24-month euro OIS by 10 basis points. The sample runs from March 2001 to December 2023, covering approximately 265-268 ECB announcement dates. U.S. zero-coupon Treasury yields come from Gürkaynak et al. (2007); the yield is decomposed into an expected short-rate path and a term premium using the shadow-rate term structure model (SRTSM) of Wu and Xia (2016). The convenience yield on Treasuries is proxied by the spread between the 10-year Treasury yield and the maturity-matched overnight index swap (OIS) rate, so that a positive (and rising) spread indicates declining convenience. Structural breaks in the convenience yield are identified via the Bai-Perron test.

The empirical strategy has three main components: (i) 700-business-day rolling regressions of Treasury yields and their decomposition on ECB surprises to document time variation; (ii) interaction regressions (following equation 5/9) that condition the ECB shock effect on lagged convenience-yield proxies, net Treasury supply, intermediary balance-sheet constraints (proxied by G10 covered-interest-parity deviations), and inflation-anchoring indicators; and (iii) a policy decomposition following Swanson (2021) that decomposes ECB surprises into “target,” “forward guidance,” and “LSAP” components. These empirical findings are rationalized in a two-country preferred-habitat model, extending Gourinchas, Ray, and Vayanos (in press) (GRV) by allowing the demand-slope parameter governing investor price elasticity to vary with the convenience yield. Functional derivatives and Malliavin calculus are used to characterize dynamic impulse responses to elasticity shifts.

Main Findings with Quantitative Magnitudes

1. Rising spillovers post-GFC, concentrated at long maturities. Rolling regressions show that ECB-to-U.S. spillovers were statistically indistinguishable from zero during the conventional-policy era but grew significantly after 2010, well before the ECB’s Expanded Asset Purchase Programme (EAPP) launched in 2015 and before “whatever it takes” (summer 2012). Spillovers began to dissipate not when ECB purchases ended (March 2022) but when the Fed announced tapering in November 2021 — consistent with the convenience channel rather than mere co-movement in LSAP volumes. A Bai-Perron test detects five structural breaks in the relationship between ECB surprises and 10-year Treasury yields.

2. Term-premium dominance, amplified by inconvenient Treasuries. At average convenience-yield levels, a one-standard-deviation ECB loosening shock (lowering the 24-month euro OIS by 10 basis points) reduces the 10-year Treasury yield by approximately 4.4 basis points (column 5, Table 2). When the Treasury convenience yield is one standard deviation below its historical average (i.e., Treasuries are less convenient), the spillover increases by 1.64 basis points, making the total effect approximately 6.1 basis points — a shift from the bottom 20th to below the 12th percentile of the unconditional distribution of daily Treasury yield changes. This amplification operates entirely through the term premium; the expected path of short rates shows no statistically significant sensitivity to the convenience yield interacted with ECB shocks.

3. Net Treasury supply amplification. Conditional on the net publicly available U.S. debt stock (Treasury debt less Fed holdings, as a percent of GDP), a one-standard-deviation ECB shock at average supply reduces the 10-year yield by approximately 3.9 basis points; when net supply is one standard deviation above its historical average (approximately 7.6 percentage points of GDP), the same shock generates a 5.35 basis-point decline — a 50-percent amplification (Table 5, column 5).

4. Intermediary constraints amplification. Conditioning on the first principal component of G10 CIP deviations against the dollar (a proxy for intermediary balance-sheet tightness), a CIP deviation one standard deviation above average amplifies the ECB spillover from approximately 3.9 basis points to 6.2 basis points (Table 7).

5. Inflation anchoring. Periods when inflation expectations lie outside the interquartile range of the historical distribution are associated with larger spillovers to 10-year Treasury yields, an effect that is statistically significant both above the 75th and below the 25th percentile of expectations, with point estimates of the interaction coefficient reaching approximately 5.0-5.3 basis points (Table 6).

6. Policy asynchronicity. Spillovers are especially pronounced when the Federal Reserve is tightening while the ECB is easing. The rolling regressions show term-premium spillovers become dominant (relative to expected-path spillovers) post-2014, coinciding with U.S. normalization. The calibrated model shows that, during policy asynchronicity combined with lower convenience, the home short-rate tightening is partially offset by capital inflows induced by foreign QE, with the attenuation especially pronounced at intermediate and long maturities and persistent across multiple periods.

7. Alternative channels ruled out. Horse-race regressions against the VIX, MOVE index, Economic Policy Uncertainty (EPU) index, Monetary Policy Uncertainty (MPU) index, and 30-day EUR/USD spot variance show none of these candidates displaces the convenience channel. Short-rate-risk decompositions (Bundick et al. 2017) and equity-orthogonal risk premium shocks (Leombroni et al. 2021) cannot explain the post-Taper Tantrum timing pattern of rising term-premium spillovers.

Scope Conditions

• All empirical results apply to ECB-to-U.S. spillovers; the paper explicitly leaves Bank of England-to-U.K. Gilt spillovers for future work.
• The portfolio-balance amplification through convenience is specific to unconventional monetary policy (LSAP shocks); target and forward-guidance components drive spillovers through different channels (expected short-rate path) and do not exhibit the same convenience-contingent amplification.
• The mechanism operates through preferred-habitat investors demanding sovereign-grade credit; the Bund convenience yield does not amplify U.S. spillovers, consistent with Bunds being an imperfect representation of the full portfolio requiring substitution under ECB capital-key-based purchases.

Layer 2 — Q&A

Q1: How do the authors measure ECB monetary policy surprises, and why do they prefer this measure?

A1: Surprises are the first principal component of intraday changes in 1-, 3-, 6-, 12-, and 24-month euro OIS rates plus 5- and 10-year German and French bond yields, measured from 10-20 minutes pre-announcement to 10-20 minutes post-press conference. This cross-section of yields is preferred because it summarizes shocks to the overall stance of policy both at and away from the effective lower bound, including effects on different parts of the yield curve. The composite measure therefore subsumes both conventional rate actions and unconventional (LSAP, forward guidance) dimensions. Surprises are normalized so one unit raises the 24-month euro OIS by 10 basis points.

Q2: What is the key empirical fact about the timing of spillover emergence and dissipation?

A2: Rolling regressions show ECB spillovers to U.S. Treasury yields became statistically significant when the rolling window began integrating observations starting in approximately 2010 — substantially before the ECB’s EAPP (2015) and even before “whatever it takes” (summer 2012). Moreover, spillovers began to dissipate not when the ECB’s Pandemic Emergency Purchase Programme ended (March 2022) but when the Fed announced tapering in November 2021. This timing pattern is inconsistent with a simple “both central banks doing QE simultaneously” explanation and instead points to the importance of Federal Reserve balance sheet behavior for the convenience of Treasuries.

Q3: How do the authors decompose the Treasury yield, and what does the decomposition reveal about the channel of transmission?

A3: Following standard term-structure decomposition, the n-year yield equals the expected path of short-term rates over the maturity plus a maturity-specific term premium. Rolling regressions on this decomposition show that term-premium spillovers dominate expected-path spillovers, especially post-2014 when the Federal Reserve is out of sync with other advanced economies. Early ECB UMP spillovers showed a more even mix of expected-path and term-premium effects; later spillovers loaded much more heavily on the term premium. This is consistent with the portfolio balance channel — LSAPs remove duration risk and compress term premia, and this effect transmits internationally.

Q4: How is the convenience yield proxied, and why does the paper use this proxy in particular?

A4: The authors use the spread between the sovereign bond yield and the maturity-matched overnight index swap rate (Y − OIS), expressed so that a larger spread (sovereign yield higher than OIS) reflects less convenience. Prior to the GFC, Treasury yields ran below swap rates (negative spread, high convenience); post-GFC, the spread reversed and turned positive, reflecting deterioration in Treasury specialness. This proxy is preferred because it captures the relative convenience as priced by the marginal investors the model focuses on — those with sovereign credit quality preferences and arbitrageurs — rather than broader measures such as the Treasury-to-corporate spread.

Q5: What is the quantitative impact of convenience yield variation on the size of ECB spillovers to U.S. yields?

A5: In the most conservative specification (Table 2, column 5), an ECB loosening shock that lowers 24-month euro OIS by 10 basis points reduces the 10-year Treasury yield by 4.4 basis points when the convenience yield is at its historical average. When the convenience yield falls one standard deviation below average (Treasuries are less convenient), the spillover increases by 1.64 basis points to approximately 6.1 basis points. A one-standard-deviation change in 10-year Treasury yields in the sample is 5.86 basis points; the 4.4 bp response falls in the bottom 20th percentile of unconditional daily yield changes, while the 6.1 bp response falls below the 12th percentile.

Q6: Does the amplification of spillovers from ECB shocks by Treasury inconvenience operate through the term premium or the expected short-rate path?

A6: The amplification operates entirely through the term premium. In Table 2, columns 7 and 8, the interaction coefficient between the ECB shock and the convenience yield proxy is positive and statistically significant for the 10-year term premium but is not statistically different from zero for the expected path of short rates. The authors interpret this as confirming the portfolio balance channel: displaced Bund investors substitute into Treasuries, raising Treasury prices and compressing term premia, with no mechanical connection to market participants’ updating of expected future Federal Reserve policy rates.

Q7: How does net Treasury supply interact with the size of ECB spillovers?

A7: Net U.S. Treasury supply (debt outstanding as a percent of GDP, less Fed holdings) is strongly positively correlated with the swap spread, confirming the link between supply and convenience. Interaction regressions (Table 5) show that a one-standard-deviation ECB shock at average net supply reduces 10-year yields by 3.9 basis points. When net supply is one standard deviation above average (approximately 7.6 percentage points of GDP), the same shock generates a 5.35 basis-point decline — roughly a 50 percent amplification. The point estimates suggest this operates primarily through term premia, though those interaction coefficients are statistically insignificant in the term premium specification.

Q8: How do intermediary balance-sheet constraints relate to Treasury convenience and ECB spillover amplification?

A8: The authors follow Du, Hébert, and Huber (2023) in using deviations from covered interest parity (CIP) among G10 currencies against the dollar as a proxy for the shadow cost of intermediary balance-sheet constraints. When CIP deviations are at historical average, the ECB spillover to 10-year Treasury yields is approximately 3.9 basis points; when CIP deviations are one standard deviation above average, the spillover rises to approximately 6.2 basis points. The authors also use the plausibly exogenous variation from quarter-end “window dressing” (per Correa, Du, and Liao 2020): LSAP-type ECB surprises landing near quarter-end generate larger spillovers to the term premium, and the further into the quarter an announcement occurs, the larger the LSAP shock’s effect on the term premium — consistent with balance-sheet constraints amplifying the portfolio balance channel.

Q9: What is the theoretical model, and what is the key innovation relative to the baseline GRV framework?

A9: The paper extends the two-country preferred-habitat model of Gourinchas, Ray, and Vayanos (in press), in which segmented investors demand bonds of specific maturities and currencies while capital-constrained global arbitrageurs partially bridge the segmentation. The key innovation is allowing the demand-slope parameter α_j(τ) — which in GRV is fixed and governs how inelastic investors are with respect to price — to vary over time as a function of the convenience yield. When Treasuries are special (high convenience), α_H(τ) is large, demand is inelastic, and foreign shocks have limited pass-through. When convenience falls, α_H(τ) shrinks, demand becomes more elastic, investors reallocate more aggressively in response to yield differentials, and U.S. term premia respond more strongly to ECB purchases. Functional derivatives and Malliavin calculus are used to characterize both instantaneous and dynamic amplification effects.

Q10: What does the calibrated model predict about the maturity structure of spillover amplification?

A10: In the calibration exercise (Figure 4), the elasticity perturbation is modeled as a smooth function (transformed Cauchy distribution) centered at the 10-year maturity, and the ECB QE shock is a purchase concentrated at the 5-year maturity amounting to 10 percent of euro-area GDP. The marginal change in the home yield impulse response (the quantity ∂²_{α_H,b} log P^τ_{Hs}) is positive across nearly all maturities and horizons, but is most pronounced around the 5-year maturity and during the first few periods after the shock — where the ECB purchase profile and the demand perturbation are most closely aligned in tenor. Amplification effects are persistent across horizons due to the dynamic multiplier in Theorem 3.1.

Q11: How does the model rationalize the 2019 yield curve inversion?

A11: In August 2019, the 10-year Treasury yield fell below short-term rates despite a robust domestic labor market, while the Fed was raising rates and the ECB remained accommodative. The model’s asynchronicity exercise (Section 3.3) shows that combining a home short-rate increase with ongoing foreign QE and a contemporaneous decline in Treasury convenience produces attenuated or even reversed yield curve responses. More elastic investors facing a flatter demand curve shift into longer-term Treasuries — whose relative yields remain attractive globally — resulting in a yield-curve inversion driven not by recession expectations but by asymmetric monetary policy and a time-varying convenience premium.

Q12: Do alternative explanations — risk sentiment, policy uncertainty, exchange rate volatility — explain the time variation in ECB spillovers?

A12: No. Horse-race regressions in Table 9 condition the ECB shock on lagged VIX, MOVE index, Economic Policy Uncertainty (Baker et al. 2016), Monetary Policy Uncertainty (Husted et al. 2020), and 30-day EUR/USD spot variance. None of these measures displaces the baseline convenience-yield interaction, which remains statistically significant across all specifications. Elevated EPU is associated with smaller spillovers (consistent with uncertainty impairing substitution), but this does not reduce the magnitude or significance of the convenience-yield interaction. Exchange-rate variance does not alter spillover size. A rolling regression decomposing the term premium into a short-rate-uncertainty component (Bundick et al. 2017) and a residual shows the empirical pattern is more consistent with the residual — not the short-rate-volatility channel. An equity-orthogonal risk premium shock (Leombroni et al. 2021) explains some term premium effects in the early GFC period (2008-2012) but cannot rationalize the post-Taper Tantrum pattern of growing term-premium spillovers.

Q13: How does the Swanson (2021) decomposition confirm the portfolio balance channel?

A13: Following Swanson (2021), the authors decompose ECB surprises into a “target surprise” (change in 3-month OIS futures), a “forward guidance surprise” (residual from projecting 24-month futures onto the target surprise), and an “LSAP surprise” (residual from projecting French and German 10-year bond yields onto target and forward guidance). In the full sample (Table 3), LSAP shocks drive spillovers to U.S. yields exclusively at higher maturities and exclusively through the term premium; they have no statistically significant impact on the expected path of short rates. Conditioning LSAP shocks on the convenience yield (Table 4, panel c) shows that it is specifically LSAP-type announcements combined with Treasury inconvenience that generate larger medium- and long-term term-premium spillovers, confirming the portfolio balance mechanism.

Q14: What are the implications for fiscal and monetary policy?

A14: The paper argues that the persistently low long-term rates and yield curve inversions observed between the GFC and the COVID-19 pandemic were driven partly by ECB LSAPs amplified by U.S. quantitative tightening, which increased net Treasury supply, reduced Fed absorption, constrained dealer balance sheets, and lowered Treasury convenience. Simultaneously, U.S. monetary tightening raised short-term rates while ongoing ECB easing depressed long rates, reshaping the yield curve in a manner consistent with the model. More broadly, the effectiveness of conventional domestic monetary policy tightening is attenuated when the convenience yield is compressed and foreign QE is ongoing — not because the short rate fails to move, but because more elastic investors reallocate around it. This suggests policy asynchronicity, combined with declining convenience, creates a constraint on monetary independence that may require more forceful or coordinated policy action.

Key Concepts

Convenience yield (Treasury convenience premium) The non-pecuniary value that investors derive from holding U.S. Treasury securities over and above cash flows and credit risk — arising from their deep and liquid markets, broad regulatory compatibility, high-quality collateral function, and reserve-currency status. Operationalized in this paper as the spread between the n-year Treasury yield and the maturity-matched overnight index swap (OIS) rate; a positive and rising spread indicates declining convenience, not increasing yield risk.

Portfolio balance channel (of unconventional monetary policy transmission) The mechanism by which large-scale asset purchases by one central bank displace investors from their target allocations, inducing them to substitute into other assets — including foreign sovereign bonds — thereby compressing yields and term premia in those markets. Distinguished from the signaling/expected-path channel in that it operates through changes in duration risk (term premia) rather than revisions to expected future short rates, and is unique to UMP because it targets long-duration assets.

Preferred habitat investors Investors with persistent, institutionally determined demand for bonds of specific maturities and issuers (e.g., insurance companies, pension funds), arising from regulatory constraints, risk management practices, or balance sheet matching. Their demand is modeled as relatively price-inelastic when assets command a convenience premium, and more elastic when that premium erodes.

Demand-slope parameter α_j(τ) In the extended GRV preferred-habitat model, the parameter governing the price elasticity of preferred-habitat investor demand for country-j bonds of maturity τ. Large values imply inelastic demand (strong habitat preferences), small values imply elastic demand and greater cross-border substitutability. The paper’s key innovation is treating this parameter as time-varying — specifically, as a function of the observed Treasury convenience yield rather than a fixed structural constant.

Policy asynchronicity The condition in which the Federal Reserve is tightening monetary policy (raising rates or conducting quantitative tightening) while other advanced-economy central banks (specifically the ECB) are simultaneously easing through LSAPs. The paper argues that asynchronicity interacts with a declining convenience yield to amplify ECB spillovers to U.S. term premia and attenuate the effectiveness of Federal Reserve tightening at the long end of the yield curve.

Swap spread (as inconvenience proxy) The spread of the sovereign bond yield over the maturity-matched OIS rate (Y − OIS). Expressed so that a larger positive value indicates greater Treasury inconvenience. Prior to the GFC, 10-year Treasury yields ran below swap rates (negative spread); post-GFC, this relationship reversed, with the spread turning persistently positive and exhibiting structural breaks consistent with Bai-Perron tests.

Exorbitant privilege The benefit the United States accrues from the global dominance of its sovereign debt and currency, which structurally insulates U.S. financial markets from foreign monetary policy shocks through inelastic global demand for Treasuries. The paper argues this insulation is not structural but endogenous and state-dependent: erosion of exorbitant privilege — operationalized as a declining convenience yield — substantially increases U.S. vulnerability to foreign monetary shocks.

Gâteaux/Malliavin functional derivative (as used in the model) Mathematical tools used to characterize how the impulse response function of the yield curve to policy shocks changes when the demand-slope parameter α_k(τ) is perturbed. The mixed Gâteaux differential ∂²_{α_k,b} log P^(τ)_{js} captures both the instantaneous amplification (direct pass-through increase) and the intertemporal propagation (dynamic multiplier) of a foreign policy shock under lower convenience, enabling a tractable decomposition of state-contingent spillover magnitudes across maturities and horizons.