US Treasuries have long been the benchmark dollar safe asset, and the convenience yield they command is the price the world pays for dollar liquidity. This column shows that the convenience yield on ten-year Treasuries relative to corporate bonds has collapsed from around 60 to 10 basis points since 2019, while the premium on repo has held near 90. Repo, not the Treasury, is now the highest-convenience dollar safe asset. The same shift helps explain why the dollar has been losing ground as the global reserve asset even as domestic convenience yields hold up.
The convenience yield on dollar safe assets (‘dollar safe asset premium’) is the price investors pay for the liquidity, collateral, and safety services those assets provide (Krishnamurthy and Vissing-Jorgensen 2012). It matters for the capital structure of financial institutions, for the transmission of monetary policy, and for the dollar’s position in the international monetary system. Thus, when commentators ask whether dollar convenience is disappearing, much depends on the answer.
The question, it turns out, has two different answers depending on whose portfolio you have in mind. For a domestic investor choosing among dollar assets, convenience yields are roughly where they were in 2019, and swap pricing indicates the market expects them to stay there. For an international investor choosing between dollar safe assets and the safe assets of other currencies, the dollar safe asset’s premium has eroded substantially. So, is the dollar convenience disappearing? The answer is no for domestic liquidity purposes; yes for global liquidity purposes.
Inside dollar markets, the premium holds — but the ranking flips
To measure the convenience yield within dollar markets, we need to benchmark relative to a dollar asset with a plausibly lower convenience premium than Treasuries or repo. We use the ten-year BBB corporate bond index hedged with the matched-maturity investment-grade CDX index, which strips out default risk and leaves a yield on a dollar bond that is safe but less liquid than Treasuries or repo. This follows Mota (2023).
Figure 1 plots the spread between the CDX-hedged 10–year corporate bond yield and both 10-year Treasuries and the 10-year SOFR swap rate. Before delving into the data, it’s useful to understand the connection between overnight repo rates and the SOFR swap rate. A SOFR swap is an exchange of a fixed cash flow, the quoted SOFR swap rate, against a floating (variable) cash flow that depends on the path of overnight repo rates. Note that an investment in an over-collateralised overnight repo is safe and liquid and will command a convenience premium. Then, the time path of overnight repo rates will pick up this premium and, via expectations hypothesis logic, the fixed SOFR swap rate will reflect the expected convenience premium over the term of the swap. Now, the fixed SOFR swap rate will also reflect a bond risk premium – a term premium – following standard bond pricing logic. Importantly, this term premium will also be reflected in the CDX-hedged 10–year corporate bond yield. Thus, the spread extracts the expected convenience premium on repo relative to corporate bonds over the next 10 years.
The orange line — the premium on repo, measured through the corporate-CDX-hedged to SOFR-swap spread — has fluctuated around 90 basis points and sits close to where it was in 2019 and in 2023. The green line is the premium on the ten-year Treasury, and it has fallen steadily since 2023, from roughly 60 basis points to around 10 in October 2026. The two series have pulled apart, and the gap has widened since 2022.
Figure 1 The convenience yield on repo has held up; the Treasury premium has collapsed
Note: Spread of the CDX-hedged ten-year BBB corporate yield over the ten-year SOFR swap (orange) and over the ten-year Treasury (green), in basis points.
Source: Chumbo and Krishnamurthy (2026).
The widening gap between long-term Treasury yields and SOFR swap rates indicates that, relative to a repo claim, long-term Treasuries have lost convenience. This could be because they are less demanded directly by liquidity/safety investors, or indirectly, as they have become costlier to hold on an intermediary balance sheet, or both. We can put a number on the balance-sheet piece. A strategy that buys the ten-year Treasury, finances it in repo, and pays fixed on the ten-year SOFR swap earns positive carry throughout our sample, averaging about 44 basis points. So, a large component of the spread reflects the cost of using a dealer balance sheet, with the rest reflecting a reduction in the direct convenience of holding long-term Treasuries. The short end tells a similar story, albeit with smaller deviations. The average spread between 90-day AA financial commercial paper and 90-day Treasury bills has halved from around 20 basis points to 10 basis points since 2023. The spread between the 90-day Treasury bill and the 90-day SOFR swap, which historically had been negative (premium on the bill; Nagel 2015) has turned positive at times this year — a premium for repo over the bill. These spreads are smaller than those in Figure 1 because the gap in liquidity and safety is much narrower among short-term assets (commercial paper versus Treasury bills or repo) than among long-term assets (corporate bonds versus Treasury bonds or SOFR swaps). Overall, on both the long and the short end, the Treasury is no longer the richest dollar safe asset.
Across currencies, the dollar safe asset premium has gone
Now take the perspective of a foreign reserve manager deciding whether to hold liquidity in dollars or in another currency. A standard comparison in the literature is the covered-interest-parity (CIP) wedge (Du et al. 2019): roll euros at eSTR, swap the proceeds into dollars at one year, and compare with holding a one-year dollar claim directly. But it is important to note, as analysed by Jiang et al. (2021), that this CIP deviation is not the same as the premium on the dollar exchange rate. The latter is the lower return on holding a safe dollar-denominated asset versus a safe foreign-currency-denominated asset; that is, an uncovered interest parity (UIP) deviation. Jiang et al. (2021) provide a methodology to extract this unobserved dollar premium.
Figure 2 shows the 1-year CIP wedge. Measured against the 1-year Treasury bill, the basis ran between 20 and 60 basis points until mid-2023, then declined, turned negative in mid-2024, and remained negative for large periods of 2025. It became more negative on the April 2025 tariff announcement — a flight away from Treasuries rather than the customary flight towards them (see also Acharya and Laarits 2025). For context, the same spread constructed against G10 government bonds averages 22 basis points from 1988 to 2025 and is almost uniformly positive, typically rising in recessions and financial disruption events. A zero spread has happened before, but it is uncommon (Jiang et al. 2021).
Figure 2 The dollar safe asset premium against euro safe assets has shrunk to near zero
Note: One-year EUR/USD CIP basis measured against the one-year Treasury bill (green) and the one-year SOFR swap (red), in basis points. Dashed vertical lines mark the November 2024 election and the April 2025 tariff announcement.
Source: Chumbo and Krishnamurthy (2026).
The red line is the same spread measured against dollar repo. It has followed the same downward path but has stayed positive throughout, recently at a few basis points. Moreover, in contrast to the bill spread, it widened in April 2025 (vertical yellow bar). Finally, in Chumbo and Krishnamurthy (2026) we compute that the erosion extends to other currencies — the dollar safe asset premium has also fallen against yen and Danish krone safe assets, though it remains positive against both.
Why? The pool of safe dollar assets has shrunk
How can convenience yields hold up inside dollar markets while falling against foreign currencies? Our hypothesis is that the set of dollar assets investors treat as genuinely safe and liquid has narrowed to Treasury bills and repo. Within dollar markets, those surviving instruments stay rich. But because the pool itself is smaller and less liquid, the dollar becomes a less compelling place to park global reserves, and foreign investors reallocate towards the safe assets of other currencies. Coppola et al. (2025) argue that reserve currency status has historically been housed wherever the capacity to create safe and liquid assets was greatest, so a contraction in that capacity should show up exactly as the international erosion we document.
To provide some sense of this shrinkage, consider the following computation. If in 2019 all Treasuries counted as safe assets, the supply of public dollar safe assets was 70% of GDP. If by 2025 only Treasury bills and repo qualify, that supply falls to 20% of GDP — this despite more than $10 trillion of additional Treasury issuance over the same period. The comparison surely overstates the decline, since nothing turns off at a threshold, but the direction is the point: large deficits funded by long-term issuance appear to have reduced rather than expanded the effective supply of safe dollar assets.
Two pieces of quantity evidence support this reading. First, the share of outstanding long-term Treasuries used as repo collateral has fallen since the global financial crisis: the stock of Treasuries has grown while repo volumes have been more stable, so a shrinking fraction of the debt is being used as collateral. Second, and more directly, the composition of repo collateral has shifted.
Figure 3 Money market funds have shifted repo collateral away from Treasuries
Note: Reverse repo positions of the money market fund sector, split by collateral type, in trillions of dollars.
Source: SEC Form N-MFP, via Chumbo and Krishnamurthy (2026).
Figure 3 plots money market fund repo holdings by collateral class. Treasuries have been the dominant collateral since 2014, but since 2023 Treasury collateral has fallen while Agency and GSE collateral has risen. Decomposing Treasury collateral by maturity sharpens the picture further: since 2023 the primary collateral has been Treasuries with one to five years to maturity, with longer maturities used less. Collateral demand has moved away from long-term Treasuries and towards short-term Treasuries and repo backed by short-dated Treasury and Agency MBS collateral — which is precisely where the convenience yield has held up.
A second candidate explanation also deserves mention. Through the 2010s, the zero lower bound and credible forward guidance suppressed long-term Treasury volatility. When rates and inflation surged in 2022, that volatility returned, which may itself have reduced the safety of long-term Treasuries independent of any fiscal channel. Our data are consistent with both of these hypotheses.
Three implications
First, dollar convenience has not disappeared; it has migrated. Any analysis that measures dollar safe-asset demand off the Treasury alone will now understate it, and one that measures it off repo will miss the international erosion. The two margins need to be tracked separately.
Second, if repo is the marginal dollar safe asset, then the Federal Reserve’s operating framework is already in the right market — the standing repo facility and the overnight reverse repo facility place the central bank natively inside the instrument that now carries the highest convenience yield (Krishnamurthy 2026). That looks less like an accident of post-2008 plumbing and more like the operating framework following the money.
Third, the supply channel reflects competing forces. Issuing more long-term Treasury debt did not expand the stock of safe assets; on the evidence here it may have contracted the effective stock by degrading the safety and liquidity of the marginal long bond. For a sovereign whose reserve currency status rests on its capacity to manufacture safe assets, that is an unwelcome finding.
Source : VOXeu
