Let me address a few of these points quickly -
comeinvest wrote: ↑Fri Apr 14, 2023 8:44 am
SCraw wrote: ↑Thu Apr 13, 2023 5:44 am
This might help understand - here are the actual forward curves
Edit: Stripping SFRH5 (Mar25) to SFRH8 below. I've roughly matched it with the tenors for actual 3m forwards for SOFR fixed/USD Swaps/Treasuries. There's a few things going on in this decomposition - there's a difference between Treasuries and the USSW (USD Swaps) actually used in FRAs, then a difference between termed SOFR and those USD swaps (should be 29bps as of now), and then also a gap between the SFR futures contracts I've used and the tenor I've matched it with.
Thank you very much. Can you please explain what are the data sources for the columns that are not calculated? The SOFR fixed 3m forward, USSW fixed 3m forward, and Treasury 3m forward? And how exactly did you strip? Can you point me to a web site explaining the process? Thanks!
Do you use SOFR futures yourself?
I’m using Bloomberg data for each and stripping it from that. I don’t know if there are public data sources for you to bootstrap this – honestly anyone who needs to do it will have a Bloomberg terminal. It should be straightforward if you have access to the curves.
I do hold SFR futures myself and trade them as part of my job. I have about ~$1mm NAV in a strategy similar to this – at the beginning of March I switched from holding 23 contracts of /TN (and one /ZF contract) to holding 91 SFR contracts (dv01 neutral); it worked quite well given the move in front end rates (about ~105k added return given the 45bps steepening in 2s10s).
comeinvest wrote: ↑Fri Apr 14, 2023 8:44 am
Ultimately my goal really is to verify my method theoretically and empirically - that the SOFR futures' returns will mimic treasury term premiums i.e. that treasuries, treasury futures, and SOFR futures are interchangeable implementations.
Let's add LIBOR futures and fed funds futures the mix. Theoretically they all should have the same returns on average?
comeinvest wrote: ↑Fri Apr 14, 2023 2:58 pm
I still have to wrap my head around this, and I'm getting increasingly confused. Where do you see a 0.4% spread?
SOFR rates are usually higher than 3m T-bill spot rates, not the other way like in SCraw's table.
I'm also not sure why there are CDS on risk-free assets.
As you know, SOFR is a measure of the effective overnight rate – to be precise, it’s a measure of the rate paid for lending out Treasuries as collateral. Where 3m T-bills trade is (theoretically) a function of the expected effective rate over the next three months, and whether that is higher or lower than overnight SOFR depends on market expectations for Fed hikes (and some idiosyncrasies which I won’t get into) – the market is currently pricing an 85% chance we get to 500-525bps in May, then a 65% chance we cut at least once in September (the first month where a cut is expected). As such, we’d expect the 3m T-bill rate to exceed the overnight rate.
But what we care about is *3m* SOFR and 3m T-bills, and to understand the difference between these two things we need to understand the intermediate steps in terming out overnight SOFR to a 3m fixed rate. Let’s start with SOFR’s predecessor, LIBOR. LIBOR was designed to capture interbank risk, which we could calculate against the risk-free rate - we called the spread between 3m Libor and 3m T-bills the “TED spread”, which is quite noisy but was typically in the 15bps to 50bps range, though it blew past 300bps in 2008.
3m T-Bills - 3m Libor
When we term out SOFR (i.e., construct a fixed rate curve from the single overnight data point which we can call risk-free), we must account for the peculiarities of FRA (forward-rate agreement) markets. When we go long a SFR future, we’re agreeing to lend money for 3 months starting at that date for three months. If you get to that date and rates are tighter, you’ve made money – e.g. you buy Dec24 SFRZ4 at 95.5 (100-95.5=4.5% rate) and in December 2024 3m SOFR (3m Libor for ED futures) is only 3.0%, you make money because you can borrow at 3.0% and lend at 4.5%. They’re cash settled so you just pay the difference, and these instruments are fine to hold to maturity.
Term SFR is essentially a synthetic LIBOR which incorporates the SOFR compounded in arrears to get a term rate, plus an appropriate spread to account for the small credit risk and liquidity premium that can’t exist with an overnight rate. Until now, that spread was based on the historical difference between Libor and the relevant risk-free rate, but with ICE having stopped polling banks for Libor (and the ED to SFR transition complete), ISDA have specified IBOR fallback rates for each tenor. Overnight and 3m are the ones we care about, which are 0.00644% and 0.26161% adjustments, respectively. The above should you give you a basic idea of the distinction between rates for T-bills, SOFR/Libor (o/n, 3m), and Fed Funds futs.
SFR futures are really the only option for front-end rates hedging; T-bill futures trading is incredibly thin, and open interest on SFR is c 9.3m contracts. If you’re hedging a fixed income/equities portfolio, Eurodollar futures exhibit much higher correlation with yields on spot securities that T-Bills, which are a bit more peculiar and are simply an inefficient way to hedge that rate exposure. I still care foremost about the shape of the curve (expecting 2s10s and 10s30s steepening), and I’m comfortable with splitting my rates exposure between SFR and the 10y Ultra. I've become more wary of using term-premia/carry/rolldown and so forth for anticipating expected Treasury returns (aside from some high-level concerns I have wrt positioning premia)
For going long SFR futures in practice there’s a few important things to note. The basic thing is convexity; ED futures are linear instruments with a fixed dv01 of $25 per bps, but for Treasuries we discount the cashflows and are therefore s.t. convexity as yield changes (dv01 increases if yield decreases; given market value increases your rate sensitivity goes up). This will be important for rebalancing; your rate sensitivity won’t adjust naturally.
Here's a quick example for a dv01-indifferent implementation across the rate instruments I'd use
• We want to target the same dv01 on the rates side for whatever we buy; let’s assume a case of a $1m NAV with 45% (450k) 30y treasuries, that’s roughly 1913.52 dv01 (dollar value of a 1bps move); for a 1bps move in yield, you make $1913.52. Depending on convexity, this will be smaller when yields are higher and larger when yields are lower; 80bps in tightening will earn you more than 80bps in widening will lose you. SOFR futures always have a $25 dv01; they are not subject to convexity as yield changes.
• Whatever contract you choose for the Treasury leg will experience (depending on convexity) an identical impact with respect to parallel movements across the yield curve (assuming futures basis holds). Say at 2.5x leverage, an 80bps tightening in the 30y while holding 45%*2.5x = 112.5% of UB will be the same as holding (dv01UB / dv01ZF)*45%*2.5x =~337.5% of ZF
• Expectations on curve movements/term premia will determine your chosen contract. Honestly, going forward I might end up keeping a mix of the 10y and 30y ultras (/TN and /UB), though I do expect steepening in the 10s30s.
• Below is an example, where we want to match the dv01 for 45% of /UB across each alternative instrument. For instance, 3.16 contracts of /UB (239.19 dv01 per contract) is equivalent to ~30.28 SFR contracts (25 dv01 per contract).
(this post is a pretty rushed high-level explanation of quite a lot, let me know if any further questions)