How the Treasury Clearing Rule for Repo Might Affect SOFR

Views expressed are those of the authors and do not necessarily represent official positions or policy of the Office of Financial Research or the U.S. Department of the Treasury.

The Securities and Exchange Commission (SEC) rule requiring the central clearing of certain U.S. Treasury-secured repurchase agreements (repos) is scheduled to go into effect on June 30, 2027.¹ While there are exceptions, this rule will cover up to 85% of non-centrally cleared repos.² As these repos move to clearing, a portion of their associated rates will be used to calculate the Secured Overnight Financing Rate (SOFR), which serves as a benchmark for borrowing cash overnight in the United States.

The inclusion of repo-rate information from newly cleared repos could affect the level and behavior of SOFR and affect the price of many loans and swaps. One way to assess the potential impact is to estimate a hypothetical SOFR that includes representative non-centrally cleared bilateral repos (NCCBR). In this blog, the authors describe their estimation methodology and results. They use a 2022 pilot collection of NCCBR data from the Office of Financial Research (OFR) and apply the SEC rule that governs which NCCBR will need to be cleared. The authors also use tri-party repo data from the Federal Reserve Bank of New York (FRBNY) and GCF and DVP repo data from the OFR, as well as the FRBNY’s methodology for calculating SOFR.³

The OFR’s pilot collected NCCBR data from nine sell-side participants on three days during June 2022, totaling about $900 billion in outstanding repos per day. The estimation shows that if the SEC rule had been in effect at the time of the pilot, 42% of the sampled volume would have been cleared under the rule, and about 9% of the volume would have been included in the reference rate. Including the volume in the SOFR calculation would have no effect on SOFR for those three days in June 2022. However, the new volume affects the SOFR distribution’s tails, indicating that including NCCBR could make SOFR slightly more volatile.

This analysis does not account for the equilibrium effects of the SEC clearing rule. Dealers with constrained balance sheets may behave differently after the rule takes effect.

Methodology

Estimating a hypothetical SOFR requires identifying the subset of NCCBR that the SEC rule would cover and qualify to be included in SOFR. Figure 1 shows how filtering the pilot data based on the SEC rule and the SOFR eligibility criteria results in a fraction of the NCCBR volume being included in SOFR. The starting point is the total volume from the OFR pilot collection (Figure 1, row 1).

The SEC rule covers only repos using Treasury securities as collateral (Figure 1, row 2). It also excludes repos with certain exempt counterparties, such as affiliates, central banks, and local governments (rows 3 and 4).⁴ Those filters reduce the NCCBR pilot volume that can be included in SOFR to less than $400 billion daily.

Furthermore, the Fixed Income Clearing Corporation (FICC), the only repo clearinghouse in the United States, does not clear repos with optionality, meaning repos for which the maturity date is flexible. FICC also does not clear repos with a floating rate, meaning a rate that can change based on some benchmark or index. Removing these repos reduces the eligible NCCBR volume to around $336 billion daily (row 5).

SOFR is calculated by finding the volume-weighted median rate across eligible tri-party, GCF, and DVP repos.⁵ Of the remaining NCCBR volume, only repos that are overnight and same-day settled are eligible for inclusion in SOFR. When the NCCBR volume is filtered by this additional criterion, about $85 billion remains to be cleared and included in the SOFR calculation (row 6).

In the DVP segment, cash lenders offer lower rates to obtain repos known as specials—repos that require a specific security as collateral. Specials are excluded from SOFR by removing DVP repo trading at rates below the DVP 25th volume-weighted percentile rate.⁶ NCCBR, which are also collateralized by specific securities, are added to the hypothetical SOFR under the assumption that they are cleared in the DVP segment. This means that the NCCBR are subject to the specials trim. This filter is not included in Figure 1 to preserve the confidentiality of the pilot data.

At the time of the OFR pilot, primary dealers’ total exposures to the NCCBR market, at about $2.1 trillion, were about twice as large as the OFR’s NCCBR pilot collection.⁷ To make the sample of repos used to estimate the hypothetical SOFR more representative of the market, a second, identical repo with the same terms is added to the data before filtering for each repo in the collection.

Results

Figure 2 shows the FRBNY’s SOFR compared with the hypothetical SOFR including NCCBR. Rates are shown for the median and the four percentiles that the FRBNY publishes. There is a single basis-point change to the 25th and 75th volume-weighted percentile rates across most pilot collection days. There are larger divergences in the tail rates: 55 basis points (bps) in the 1st volume-weighted percentile rate and 8 bps in the 99th volume-weighted percentile rate.

In conclusion, on the three pilot data collection dates in June 2022, including eligible NCCBR would not have impacted SOFR. While the eligible NCCBR rates would have moved the extreme percentiles published by the FRBNY, SOFR is a median and would have been undisturbed by differences in the tails of the rate distribution.

The differences in the extremal tails show that including NCCBR could affect the distribution of SOFR, which could raise its volatility slightly. However, the inclusion exercise does not account for any of the equilibrium effects of the SEC clearing rule. Dealers with constrained balance sheets may demand more extreme rates to do repos that are balance-sheet intensive. As a counter, central clearing can relieve the balance-sheet intensity of repo through netting, so the NCCBR tail rates might become less extreme once cleared. Balance sheets are especially constrained on quarter-end dates such as June 30, 2022, when the hypothetical SOFR was 55 bps higher at the 1st volume-weighted percentile rate. If netting were available, including NCCBR rates in SOFR might not result in such large changes in the tails of the SOFR distribution.

---