Using Compounded Rates – A comparison of different methodologies in high volatility

My previous blog looked at the ways to use SOFR to calculate a term rate. Because SOFR is an overnight rate, users are obliged to transform this into a term rate such as 1-month to align with the settlement frequencies on financial products. This is typically done by averaging or compounding SOFR over the relevant period.

This blog looks at the impact of different methodologies during the periods of high SOFR volatility in 2008, 2019 and 2020. These dates were selected because they have large moves in SOFR over a short time frame resulting in significantly different outcomes for the compounded term rate. While this may be a relatively rare event, should we enter a period of increasing US overnight rates (Fed Funds) then this will be reflected in SOFR. Even minor differences in the methodology can lead to significant differences in outcomes.

Of course, you can set the SOFR rate upfront in the same way as LIBOR. This is Term SOFR and is available for licensing from CME. I will look at developments in Term SOFR in following blogs.

The term SOFR (small t) here is set in arrears once the final SOFR input is known.

A recap on the methodologies

I used this table in the previous blog to describe the basic methodologies used in creating a term SOFR.

In the previous blog I showed Lookback and Observation Period shift are quite similar in outcome for each hi.

In this blog I look at the Payment Delay and Lookback options only to show their relative performance in different conditions and for different lookback days.

2008 – the GFC effect

SOFR was not published until 2018 but the NY Fed has kindly provided a set of proxy SOFR rates derived from market inputs from much earlier dates including 2008. While this is not actually SOFR, it appears to be a good representation of the repo rates (SOFR proxy) from 1 August 2008 to 31 December 2008.

The following chart shows the rates over that period followed by 2 tables with the results for 17 September and 19 September 2008. I selected these 2 dates to demonstrate how a few days difference in the number of days for the lookback can give very different outcomes for the 1 and 3-month term rates.

SOFR was quite flat around 2.00% until 11 September 2008 (LHS red circle) when it spiked on 12 September and fell to 0.25% on 17 September. It then bounced to 1.82% on 19 September (RHS red circle) before continuing a volatile fall to near zero by 31 December 2008.

The grey shading for the 5-day lookback demonstrates how volatility can impact the outcomes for the methodologies.

Both the 1 and 3-month term SOFR rates differ because the input rates to the calculation are very volatile, and the difference of a few days can include or exclude outliers in the data set.

2019 – the SOFR spike in September

SOFR was setting at around 2% for most of September 2019 but spiked to 5.25% on 17 September on technical liquidity issues. While this was unusual it still had an impact on the outcomes for term SOFR rates which did or did not include the spike. The chart and table are below.

The payment delay does not include the spike but the lookbacks both have this feature in the data sets. This results in a 17-19 basis point increases in the 1-month term SOFR and 8-9 basis point increases in the 3 -month term SOFR.

Again, we see the impact of the choice of methodology on the outcomes for the term SOFR.

2020 – the COVID-19 impact in March

This period was quite volatile and presented 3 rapid falls in SOFR from 1.6% before 2 March 2020 to 0.00% by 18 March 2020. The red circle outlines the largest of these falls in SOFR between 13 and 16 March where SOFR moved from 1.20% to 0.26% over the weekend. The impacts are captured in the chart and table below.

Yet again, the methodology does matter. The 1-month difference between the payment lag (o days) and the 5-day lookback is almost 25 basis points! The 2 and 5-day lookbacks both include SOFR rates at 1.6% while the payment delay does not use that data. The outcomes are as expected with the lookbacks showing higher term SOFR rates than the payment delay.

Does this matter?

Last time we saw that the choice of methodology was not important when SOFR is not volatile: this has been the case in the recent past.

But the methodology does matter in periods when SOFR is volatile (2008), subject to a liquidity event (2019) or has a sudden change based on Fed activity (2020).

In these cases, we can see that the use of a 0, 2 or 5-day lookback can have very significant impacts on the term SOFR rate applied to a contract or trade.

Read the fine print carefully and align your hedges

Although the actual methodology does not significantly impact the term SOFR rate when markets are calm and the Fed is not moving the target band, this is not always the case. We have seen in this blog that a simple change of the lookback days can lead to important and meaningful differences in the final term SOFR.

So, as we stated in the previous blog, read the contractual terms very carefully, it does matter! And make sure you align the hedges as well.

Do you really want the risk of a mismatch in the timing of the SOFR data used for the calculation of the term SOFR? As we have seen, this can matter a great deal and can be eliminated by correct alignment of the hedge to the risk.

The next blog will look more at CME Term SOFR and how it performs relative to the compounded term SOFR.