Constructing high-frequency monetary policy surprises from SOFR futures

Finance and Economics Discussion Series Federal Reserve Board, Washington, D.C. ISSN 1936-2854 (Print) ISSN 2767-3898 (Online) Constructing high-frequency monetary policy surprises from SOFR futures Miguel Acosta, Connor M. Brennan, and Margaret M. Jacobson 2024-034 Please cite this paper as: Acosta, Miguel, Connor M. Brennan, and Margaret M. Jacobson (2024). “Constructing high-frequencymonetarypolicysurprisesfromSOFRfutures,”FinanceandEconomicsDiscussion Series 2024-034. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2024.034. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Constructing high-frequency monetary policy surprises from SOFR futures ∗ † ‡ Miguel Acosta, Connor M. Brennan and Margaret M. Jacobson May 2024 Eurodollar futures were the bedrock for constructing high-frequency series of monetary policy surprises, so their discontinuation poses a challenge for the continued empirical study of monetary policy. We propose an approach for updating the series of Gu¨rkaynak et al. (2005) and Nakamura and Steinsson (2018) with SOFR futures in place of Eurodollar futures that is conceptually and materially consistent. We recommend using SOFR futures from January 2022 onward based on regulatory developments and trading volumes. The updated series suggest that surprises over the recent tightening cycle are larger in magnitude than those seen over the decade prior and restrictive on average. ∗Federal Reserve Board. Miguel.Acosta@frb.gov †Federal Reserve Board. Connor.M.Brennan@frb.gov ‡Federal Reserve Board. Margaret.M.Jacobson@frb.gov. The views are those of the authors and not of the Federal Reserve Board, System or governors. The authors thank Andrea Ajello, David Bowman, Don Kim and Ben Johannsen and audiences at the Federal Reserve Board for helpful comments and feedback. 1

Understanding the effects of monetary policy on economic and financial outcomes is a longstandingobjectiveofmonetaryeconomics. Identifyingthecausaleffectsofmonetarypolicy hinges on isolating changes in policy that are exogenous—that is, attributed to monetary policy actions or communications beyond the endogenous response to economic outcomes. Gu¨rkaynak et al. (2005) and Nakamura and Steinsson (2018) (henceforth, GSS and NS, respectively) designed series of monetary policy surprises that capture exogenous variation in the current and, crucially, the future path of policy—a central feature of modern monetary policy. These series rely on changes in market prices of Eurodollar futures contracts in short time windows around FOMC announcements. However, with the phaseout of LIBOR, Eurodollar futures no longer exist, making it impossible to directly update these series of monetary policy surprises.1 The contribution of this paper is to document how these workhorse series of monetary policysurprisescancontinuetobeupdatedusingSecuredOvernightFinancingRate(SOFR) futures as an alternative for Eurodollar futures. We describe institutional features of Eurodollar and SOFR futures and show that between mid-2018 and mid-2023—the time period where both were traded—intraday changes around FOMC announcements are nearly perfectly correlated. In practice, we recommend using SOFR futures starting in January 2022 due to the regulatory phase out of LIBOR at the end of 2021 and adequate trading volumes needed to reliably calculate changes in tight time windows around FOMC announcements. Using our methodology, we update the GSS and NS series of monetary policy surprises to the present. Over the recent monetary tightening cycle, we find that surprises are larger in magnitude than those seen over the prior decade. Since March 2022, surprises are restrictive, on net, with a couple of meaningful pivots. While other papers have paired Eurodollar and SOFR futures to study the effects of monetary policy—e.g., Kroner (2024) and Boehm and Kroner (2023)—we are the first to document how to construct and update canonical series of monetary policy surprises and discuss the associated practical and institutional issues. 1 Eurodollar Futures and monetary policy surprises Nearly all high-frequency series of monetary policy surprises rely on changes in expected interest rates at different horizons around central bank policy announcements. The changes in these expected paths of interest rates can be measured via the variation in prices of interest rate futures in a narrow window around each announcement. More specifically, the GSS and NS series span roughly the first year of the interest-rate term structure via a linear combination of federal funds futures and Eurodollar futures. The market for federal funds futures is liquid enough to measure expectations of rates roughly three months in the future. Eurodollar futures provide an instrument for measuring expectations beyond three months. Figure 1 contains a schematic overview of quarterly Eurodollar and SOFR futures contracts that provide exposure to the interest rate term structure at an arbitrary time t. A Eurodollar futures contract expiring in quarter q was an agreement to exchange 100 − r , τ where τ represents that contract’s expiration day and r the three-month LIBOR prevailing τ at day τ. Because the price of Eurodollar futures was based on a three-month term rate, 1“LIBOR”isanabbreviationfortheU.S.dollarBritishBanker’sAssociationLondonInterbankOvernight Rate. 2

Start of quarter q q +1 q +2 q +3 q +4 q +5 t 1st SOFR 2nd SOFR 3rd SOFR 4th SOFR 5th SOFR 1st ED 2nd ED 3rd ED 4th ED Futures used in series of monetary policy surprises Figure 1: Interest rate exposure of Eurodollar (ED) and SOFR futures. they incorporated market participants’ expectations of the future path of interest rates in the three months after settlement. For example, in Figure 1, a market participant trading on date t and seeking interest rate exposure in quarter q + 1, would have traded the first-outstanding Eurodollar future, with settlement towards the end of q. Eurodollar futures were listed quarterly on the Chicago Mercantile Exchange (CME) from 1981 to April 2023 and were one of the most actively traded futures in the world until early 2022. Despite their widespread use, trading volume declined over the past decade, as shown in Figure 2. This can be attributed, in part, to the declining use of LIBOR leading up to its termination in June 2023. Although LIBOR was used in hundreds of trillions of dollars of financial instruments at its height of popularity, scandals in 2012 decelerated its usage as a reference rate. 2 Using SOFR Futures In 2014, the Federal Reserve convened private sector participants to recommend a new reference rate in place of LIBOR. SOFR was endorsed in June 2017 and first released in April 2018.2 A month later, SOFR futures began trading on the CME. In 2021, the CME announced that by April 14, 2023 all Eurodollar futures would be automatically converted to SOFR futures at a fixed spread.3 As shown in Figure 2, declining trading volume of Eurodollar futures coincides with increasing trading volumes of SOFR futures, especially after the announcement in 2021. SOFR futures are structured differently than Eurodollar futures. The settlement price of the quarter-q SOFR future at day τ is 100-R , where R is compounded daily SOFR r q q τ over the three months preceding the settlement date:   100×360 (cid:89) D b q (cid:26) (cid:18) δ (cid:19) (cid:16)r (cid:17) (cid:27) τ−d τ−d R q = Dq  1+ 360 100 −1, (1) d=1 where Dq is the number of calendar days in q, Dq is the number of US government securities b 2Seetheoverview(https://www.newyorkfed.org/markets/reference-rates/sofr),andrelatedworkbyHeitfield and Park (2019). 3The 26.161 basis point spread is based on the historical differences between LIBOR and SOFR due to the former being unsecured and the latter secured (https://www.isda.org/a/BcJgE/Progress-on-Global- Transition-to-RFRs-in-Derivatives-Markets.pdf). 3

Monthly totals, millions Daily totals on FOMC announcement days, thousands 60M 150K May 2022 May 2022 100K 40M 50K 20M 0 2018 2020 2022 2024 2018 2020 2022 2024 2nd ED 3rd ED 4th ED Quarterly Eurodollar Futures (ED) 3rd SOFR 4th SOFR 5th SOFR Quarterly Secured Overnight Financing Rate Futures (SOFR) Note: dashed black line is the total volume of Eurodollar futures shown Figure 2: Trading volumes of futures used in the GSS and NS series market business days in q, d indexes business days, r SOFR, and δ the number of calendar τ τ days to which r applies. τ Care must be taken to properly align SOFR and Eurodollar futures contracts. While Eurodollar futures were based on expected interest rates over three months after the settlement date, SOFR futures are based on interest rates over the three months before. As Figure 1 shows, both the first-outstanding Eurodollar future and the second-outstanding SOFR future are called the q + 1 contract. Because the CME named both Eurodollar and SOFR futures based on the quarter of their interest rate exposure, they can be matched based on their contract names. Alternatively, one can match the nth-outstanding SOFR contract with the (n−1)st-outstanding Eurodollar contract. 3 Series Construction We now turn to updating the GSS and NS series of monetary policy surprises beyond the 2023 termination of Eurodollar futures. We use CME Datamine data to construct both series using the original 30-minute changes in the prices of federal funds futures and either Eurodollar or SOFR futures around FOMC announcements. The federal funds futures contracts used for a given announcement have typically been those expiring at the end of the month of that announcement and the subsequent scheduled announcement, while the Eurodollar futures have been the second-, third-, and fourth-outstanding contracts, as shown in Figure 1. Brennan et al. (2024) provide details on federal funds futures and the selection of exact trades for both types of futures. We ensure conceptual similarity between our updated series and the originals by replacing Eurodollar contracts with SOFR contracts that measure interest rate expectations for the same horizons. Based on the discussion in Section 2, we use the third-, fourth-, and fifth-outstanding SOFR contracts in place of the second-, third-, and fourth-outstanding Eurodollar contracts. As long as researchers use SOFR futures with horizons longer than the first-outstanding contract—as is the case in both the GSS and NS series—then differences in prices around announcements can be used without additional adjustments. Using the firstoutstanding SOFR contract would require researchers to account for the pre-determination 4

.2 3rd outstanding SOFR future 2nd outstanding Eurodollar future .1 0 -.1 Correla‹tion = 0.962 Correlafition = 0.997 -.2 2018 2019 2020 2021 2022 2023 2024 .2 4th outstanding SOFR future 3rd outstanding Eurodollar future .1 0 -.1 Correla‹tion = 0.941 Correlafition = 0.996 -.2 2018 2019 2020 2021 2022 2023 2024 .2 5th outstanding SOFR future 4th outstanding Eurodollar future .1 0 -.1 Correla‹tion = 0.946 Correlafition = 0.998 -.2 2018 2019 2020 2021 2022 2023 2024 Figure 3: 30-minute changes around FOMC announcements, percentage points of some daily SOFR r used to calculate the settlement rate R in equation (1).4 Alternative τ q series of monetary policy surprises relying on the first-outstanding Eurodollar contract—e.g., those of Bauer and Swanson (2023)—can be updated using the second-outstanding SOFR contracts without any concerns about pre-determination. Given that Eurodollar and SOFR futures are designed to provide nearly the same interest rate exposure, we simply append the SOFR-contract series to the Eurodollar-contract series afteracertaindate. Choosingwhichdateisnotobvious, sinceintradaychangesinEurodollar futures and SOFR futures were not identical in the years immediately following SOFR’s launch, as shown in the 30-minute changes around FOMC announcements plotted in Figure 3. We recommend switching to SOFR futures in January 2022. Ideally, one would switch in July 2021 when the Market Risk Advisory Committee encouraged market participants to transact in SOFR derivatives to avoid the risks of declining liquidity for LIBOR instruments 4(https://www.cmegroup.com/education/articles-and-reports/three-month-sofr-futures-rates-andfuture-sofr-levels.html). 5

like Eurodollar futures. After all, regulations aimed for US banks to cease entering new LIBOR-based contracts by late 2021.5 Despite these regulatory developments, Figure 2 shows that although trading activity in SOFR futures was rising, it still remained quite low until 2022. January 2022 is therefore a date that assures sufficient trading volume in SOFR futures.6 Figure 3 shows that, over the sample between our chosen switch date and the end of Eurodollar futures trading, intraday changes in SOFR and Eurodollar futures around FOMC announcements are nearly perfectly correlated. With replacements for Eurodollar futures in hand, we present the updated NS and GSS series of monetary policy surprises in Figure 4. Through 2021, the updated and original series are nearly identical, despite the longer sample used for estimating the loadings of these factors on the underlying assets. FOMC announcements since the end of Eurodollar futures—which includes the most-recent tightening cycle—have induced restrictive surprises to the path of interest rates on net. For example, the sum of the NS series since the cycle started in March 2022 is about 20 basis points. Two meaningful pivots are large negative surprises: the first in November 2022 when markets interpreted FOMC communications as signaling a forthcoming slowdown in the pace of rate increases and the second in December 2023 when markets viewed the tightening cycle as drawing to a close.7 4 Conclusion Estimatingtheeffectsofmonetarypolicyoneconomicandfinancialoutcomesisacornerstone of research in monetary economics. The contribution of this paper is to document how workhorse series of monetary policy surprises can be updated after the discontinuation of underlying futures data. We find that intraday changes in SOFR futures can be substituted for Eurodollar futures with little material effect on the series of monetary policy surprises. We recommend substituting SOFR futures for Eurodollar futures starting in January 2022 due to regulatory developments in 2021 and ample trading activity. 5(https://www.newyorkfed.org/arrc/sofr-transition). 6We also tried alternative switch dates, and a state-space model approach. Ultimately, all series were highly correlated, so we opted for the approach based on institutional knowledge. 7See media coverage: (https://www.bloomberg.com/news/live-blog/2023-12-13/fomc-rate-decision-andfed-chair-news-conference) and (https://www.reuters.com/markets/us/fed-set-another-big-rate-hike-maytamp-down-future-tightening-2022-11-02). 6

Target series of Gürkaynak et al. (2005), percentage points .2 Eurodollar and SOFR futures starting in January 2022 Eurodollar futures only, ending in April 2023 .1 0 -.1 Correla‹tion = 0.998 -.2 1995 2000 2005 2010 2015 2020 2025 Path series of Gürkaynak et al. (2005), percentage points .2 Eurodollar and SOFR futures starting in January 2022 Eurodollar futures only, ending in April 2023 .1 0 -.1 Correla‹tion = 0.998 -.2 1995 2000 2005 2010 2015 2020 2025 Series of Nakamura and Steinsson (2018), percentage points .2 Eurodollar and SOFR futures starting in January 2022 Eurodollar futures only, ending in April 2023 .1 0 -.1 Correla‹tion = 0.998 -.2 1995 2000 2005 2010 2015 2020 2025 Figure 4: Series of monetary policy surprises 7

References Bauer, Michael D. and Eric T. Swanson, “An Alternative Explanation for the ‘Fed Information Effect’,” American Economic Review, March 2023, 113 (3), 664–700. Boehm, ChristophE. andT. NiklasKroner,“MonetaryPolicywithoutMovingInterest Rates: The Fed Non-Yield Shock,” 2023. Working Paper. Brennan, Connor M., Margaret M. Jacobson, Christian Matthes, and Todd B. Walker, “Monetary Policy Shocks: Data or Methods?,” 2024. Working Paper. Gu¨rkaynak, Refet S., Brian P. Sack, and Eric T. Swanson,“DoActionsSpeakLouder than Words? The Response of Asset Prices to Monetary Policy Actions and Statements,” International Journal of Central Banking, 2005, 1, 55–93. Heitfield, Erik and Yang-Ho Park, “Inferring Term Rates from SOFR Futures Prices,” 2019. Working Paper. Kroner, T. Niklas, “InflationandAttention: EvidencefromtheMarketReactiontoMacro Announcements,” 2024. Working Paper. Nakamura, Emi and Jo´n Steinsson, “High-Frequency Identification of Monetary Non- Neutrality: The Information Effect,” Quarterly Journal of Economics, 2018, 133, 1283– 1330. 8