Term SOFR rate formula

The following website gives the specifications of the CME Term SOFR reference rates: CME Term SOFR.

Point 1 in the link above specifies that the tenors that are currently supported are 1m, 3m, 6m, and 12m. Point 2 specifies that CME SOFR futures of various maturities are used to imply the SOFR Term rates (the maturities differ, but the underlying accrual period is always either 1m or 3m). Point 7 specifies that the Alternative Reference Rate Committee (ARRC) supports the introduction of SOFR Term rates.

My question is this: why do you think that the CME uses SOFR futures, rather than SOFR OIS swaps, to imply the SOFR Term rate?

January 2022 data show that EuroDollar futures (i.e. USD Libor futures) were still 3-times more liquid in terms of volume than the SOFR equivalent. On the other hand, SOFR OIS swaps have by now far exceeded LIBOR swaps in terms of liquidity. Why not just use the OIS SOFR swaps?

Why is the market obsessed with trying to come up with some "fancy" SOFR term rate (such as the CME SOFR term rate, which mixes various SOFR futures and uses an opaque formula based on VWAP): I totally understand that the market wants a forward looking credit sensitive rate (for example to index lending to): but why not just take points on the SOFR OIS yield curve as the forward-looking SOFR term rate?

My only possible explanation is that the SOFR OIS curve is based purely on a compounded rate, whilst the market might fancy some type of an arithmetic average rate: when you undertake overnight financing, you do it on a non-compounded basis (you borrow a fixed amount, pay prevailing interest rate, and then might chose to borrow the same fixed amount again and pay the new prevailing overnight rate: therefore arithmetic average might be better for hedging, rather than a compounded average).

The SOFR futures link above specifies that the CME 1m SOFR futures use an arithmetic average daily SOFR during the delivery month to compute the settlement rate, whilst the CME 3m SOFR futures use the compounded daily SOFR during the delivery quarter to compute the settlement rate.

To my knowledge, the SOFR OIS swaps use the compounded average for the SOFR floating leg (so presumably the same formula as the 3m CME futures).

So the 1m SOFR futures might be better for a forward-looking SOFR Term rate based on an arithmetic average.

Is it the arithmetic vs. the compounded rate argument, or is there a different reason for not using the SOFR OIS swap curve to imply a SOFR Term rate?

• I think it is because there isnt much liquidity in SOFR ois swaps with very short maturities like 1m, 3m etc. in this sector , the futures may be more active than the swaps.
– dm63
Mar 28 at 17:23
• True about outright short maturities OIS not being heavily traded: but I was thinking about 1-year swaps that were traded (say) 11m, 9m, or 6m ago: these would be liquid enough and could be used to extract the 1m, 3m or 6m implied compounded forward SOFR "term" rates. Mar 28 at 18:25

The issue is data ownership and transparency in my view.

Whilst OIS-swaps do give a more accurate view of the daily RFR rates that will compound to yield the Term SOFR reference rates, the OIS swaps are not reliably visible and they and are out of CME's control.

CME SOFR futures are directly within CMEs control and their pricing and transactions are regulated by the CME

Regulatory reporting of SEF trades requires OTC OIS swaps be reported but timings are not entirely robust and much more difficult to coordinate across all register SEF entities. CLOB are also difficult or impossible for CME to gain access to determining the OIS-mid swap rates at any given time.

The SOFR futures contain enough information to determine the SOFR tenor indices except for convexity adjustments, which are obviously assumed to be zero, since the CME also has no transparent way of determining this information, except to offer options on SOFR futures from which it could determine a transparent volatility parameter (within its own sphere of influence) and devise a public formula that will include convexity adjustments.

• Thank you for the nice answer. I was thinking that the volume of cleared swaps at both, LCH and CME, would be sufficient to use these for a term SOFR. I understand that large volumes are still traded OTC, but I think these are mostly dealer to client. I think the majority of Dealer-to-Dealer swaps are now cleared. Mar 29 at 8:45
• yes but you dont have to transact swaps at mid market to put them thrpugh CME or LCH, so the whole process is subject to arbitrary abuse, and timings of swaps are not recorded correctly if that matters for consistency, also CME has no visibility of LCH swaps and vice versa.
– Attack68
Mar 30 at 11:03
• Ok, final question then: how about market-making consensus on (say) Bloomberg? Just like we used to be able to get the yield curve based off the Libor IRS, we can get a yield curve (even zero coupon) based off the SOFR OIS. I guess the one thing that matters is volume, which you can't see on these BBG yield-curves, and also like you said: data ownership. Mar 30 at 13:54
• The curves in Bloomberg are typically constructed with the data source pnemonic BGN (or related) that stands for Bloomberg composite price. It is usually composed of a multitude of dealers who stream prices to bloomberg's multitude of electronic platforms. Problems from a CME perspective: relies on an external data provider, and is ultimately dependent upon bloombergs composite algorithm, and then on the prospective dealers streaming prices (which are not necessarily actual trades but rather bids and offers, and dealers may be subject to change over time)
– Attack68
Mar 30 at 17:17