muRisQ Advisory offers quantitative finance advisory services. Its services focus on trainings and workshops, benchmark transitions, interest rate model development and validation, product design and risk management strategies.
Looking at last week SOFR numbers, the only description that comes to mind is "unchanged to slightly lower". No big push since "SOFR First" date four weeks ago.
LCH figures slightly lower, ISDA figures slightly lower, ISDA outright SOFR as portion of LIBOR slightly lower.
We still have roughly four months to go to the LIBOR transition. USD-LIBOR rates will still be published for 18 months after that, but the goal is to stop adding more LIBOR exposure from the beginning of 2022 onward.
A fallback mechanism for legacy LIBOR has been proposed by ISDA, based on a historical spread approach between the in-advance credit risky LIBOR and the in-arrears secured SOFR. The results of that historical spread computation has been announced on 5 March 2021. FOR USD-LIBOR-3M, the result was a spread of 26.161 basis points.
In USD, other mechanisms to replace LIBOR have been proposed. They include, among others (see this blog for more alternatives) Bloomberg BSBY, CME SOFR Term rates, and AMERIBOR. In this blog we propose recent historical data related to those benchmarks.
The data themselves are IBA USD-LIBOR-3M (2019-01-02 to 2021-08-20), Bloomberg BSBY 3M (2019-01-02 to 2021-08-20), SOFR Term rate 3M (2019-01-03 to 2021-08-19), SOFR compounded using ISDA mechanism (computed from Fed raw data) (2019-01-02 to 2021-05-17), and AMERIBOR 90D (2021-05-17 to 2021-08-20). The SOFR compounded time series are shorter by 3 months due to the in-arrears feature. The AMERIBOR 90D is published only since May this year. The time series are displayed in Figure 1.
The drop of rates after COVID-19 pandemic spreads from Wuhan to contaminate the world is clearly visible in the graph at the beginning of 2020. The way the different rates reacted was different.
In Figure 2, we display the spread between LIBOR and the other rates. The BSBY spread has also jumped around that time, but remained contained. The Term SOFR had one big jump that can be interpreted as a market credit concern. At the opposite, the compounded in-arrears ISDA style had two large jumps, probably one due to the delayed nature of in-arrears and one related to credit.
For the period in the graph, the statistics for the spreads are an average of 5 bps and a standard deviation of 4.6 bps for BSBY. For SOFR Term, the statistics are an average of 23 bps and a standard deviation of 21 bps. For SOFR in-arrears with ISDA methodology, the statistics are an average of 32 bps and a standard deviation of 30 bps. For AMERIBOR 90D, the statistics are an average of -1.7 bps and a standard deviation of 1 bps; the period of reference is very different and those statistics cannot be compared to the previous ones that are reported on a substantially more unstable period.
The fallback method is based on a fixed spread of 26.161 bps (and an implicit standard deviation of 0).
Added 2021-08-28: The spread versus term SOFR is presented in Figure 3. LIBOR, BSBY and Ameribor spreads varied widely but are all positive, except one date for LIBOR (2 March 2020). Over the last year, the spreads have been well below the ISDA spread fixing.
SOFR First started on 26 July. We now have three weeks of data. The results are still somehow mixed.
On the LCH front, the volume is again significantly below the 30 July week level (SOFR First initial week). The volume is roughly the same as at the end of May.
On the ISDA figures side (from US regulatory figures), the absolute volume is up for the 6th week in a row; the volume has tripled in those weeks. But the figures reported by ISDA (USD 166 bn) are well below the figures reported by LCH (USD 520 bn).
A temptative interpretation could be that the US local market is moving slowly to SOFR (SOFR volume is now around 11% of LIBOR volume). Even if that move represents a real chance in behavior, it impacts only a small portion of the USD market. The figures published indicate that ISDA actually captures less than a third of the SOFR market in its SwapInfo analysis.
Note: More volume analysis over the last years available on Marc's multi-curve framework blog.
SOFR First started on 26 July. We now have two weeks of data. The result is somehow mixed.
On the LCH front, the volume has decrease significantly last week with respect to the previous one (SOFR First initial week). The volume is roughly the same as in end of May.
On the ISDA figure side (from US regulatory figures), the absolute volume is slightly up but the proportion of SOFR with respect to LIBOR is down; it is below 8%.
Still far away from SOFR First!
Note: Blog initially published at: https://multi-curve-framework.blogspot.com/2021/08/sofr-first-two-weeks-on.html. More volume analysis over the last years available on Marc's multi-curve framework blog.
A new working paper related to convexity adjustment in the overnight ransition has been made available on SSRN. The paper title is
The paper is available on SSRN at
This note analyses derivative pricing in the context of a collateral rate switch during the life of a financial product or the existence of two overnight rates. In particular we analyse the impact of forward change of collateral, the impact on OISs when the collateral rate is different from the OIS underlying, and the impact of bilateral swaptions collateral rate different from the one implied by the cleared market. In each case, we evidence new convexity adjustment impacts previously not accounted for. The order of magnitude of those impacts is also proposed for realistic values of model parameters; the individual relative impacts are not huge, but when applied on trillions of notional, as it is the case, the monetary impacts can be substantial.
Marc was quoted in the press in an article related to the LCH steering away from ISDA fallback: Industry broadly supports LIBOR big bang switch (subscription required).
The quotes are
While the proposal crafted by LCH may look like an ideal solution for many market participants, it could come at a price, which some deem too high. “I encouraged my clients to consider this approach even before it was proposed by LCH, but we should bear in mind that clearinghouses look at what is in their books – not at the market,” said Marc Henrard, managing partner at muRisQ Advisory.. “They want to push the bifurcation problem out to end-users, who in turn will have to deal with the issue on their bilateral trades and manage the bilateral-to-cleared basis risk."
Bringing forward this proposal earlier and presenting a common mechanism that can also apply to bilateral trades would have made for a smoother process, Henrard added.and
In addition, there is a risk that LCH's solution could fragment the swaps market, as liquidity will be split between the trades referencing RFRs and those using ISDA’s fallback rates. “Market participants should make sure they have the right tools in place to compute any small residual valuation difference,” said Henrard. “LCH’s proposal may not work out to everyone’s advantage, and some may want to adjust their positions ahead of the big bang.”
Marc Henrard will present a two half-day workshop
organized by the CQF Institute on 18 May 2021 and 19 May 2021 (on-line 13:00 - 17:00 BST).
The agenda of the workshop can be found on the organizer website:
The agenda of the workshop will be similar to the one proposed on our training page.
The workshop is also availabe as a in-house course for financial institutions. If you want to be ahead in the transition game, contact us for LIBOR fallback and discontinuation: trainings, workshops, advisory, tools, developments, solutions.
Marc Henrard will present a seminar at
which will take place on-line from Monday 22 February to Friday 26 February 2021. The agenda of the conference can be found on the organizer web site:
Marc's talk will take place on Wednesday 24 November at 13:00 GMT and will be titled Benchmarks in transition: Hidden convexities.
Agenda:
With the recent consultations on different dates for USD-LIBOR and other currencies LIBOR cessation, the discrepancy in spread computation mechanism has been discussed.
To our opinion, there will be no date discrepancy as we expect the LIBOR cessation announcement date to be the same for all currencies, even if the cessation date is not the same.
There are other more important inconsistencies. In the cross-currency market, there is the obvious LIBOR v (IBOR but not LIBOR). EUR-EURIBOR will not cease and the major cross, the one between USD and EUR, will be the prime example of this inconsistency.
But the other, less talk about, inconsistency that we want to point out in this post is the tenor inconsistency within a single currency. The spread for all tenors will be computed on the same day using a look-back period of 5 years. The period for all currencies is nominally the same 5 years, but those 5 years represent different economical realities for each tenor. The differences are coming from the fundamental flaw in the fallback spread methodology which is comparing a forward looking rate to a backward looking composition. To compare those two rates, you have by definition to use different dates. For LIBOR there is one fixing date while for the overnight, there is a tenor-full of dates.
Take the example of the USD-LIBOR-12M. Suppose that the announcement date is on 2021-01-05. What is the meaning of 5 years? This is the most recent 5 years of LIBOR fixing for which we have the corresponding 5 year of SOFR composition over the LIBOR tenor. The period starts on 2015-01-05 and finishes on 2020-01-05. The end is in 2020 and not 2021, even if the announcement is in 2021. This is not a typo but the impact of the in-arrears. To compute the spread, we need the in-arrears compounded on a 12 months period, i.e. the period from 2020-01-05 to 2021-01-05. This is a little bit more complex than that, because of the -2 business day shift in the fallback, but for this post we ignore it (all our data analysis and graphs are produced with the exact rules implemented in a production grade library).
But if we look at the same question for USD-LIBOR-1M, the LIBOR period is 2015-12-05 to 2020-12-05, a 11 month difference with USD-LIBOR-12M, i.e. almost 20% of the period is different. This is very significant when the economical situation is not stable. It turns out that the last 12 months have been very unstable from an economical perspective.
Let's look at it through the lens of the spread data. In Figure 1, we have represented the spread computed on a daily basis of the LIBOR/SOFR compounded in-arrears for tenors 1M, 3M, 6M, and 12M using the ISDA methodology.
Figure 1. Historical spread between USD-LIBOR and SOFR-compounded for different tenors.
Contrary to the intuition of the spread as being a credit spread, we see that the curves cross in several places. The 12M spread can be in some periods below the 6M and even the 3M spread. The April credit spread increase is clearly visible in the 1M, 3M and 6M spreads. Is it visible in the 12M curve? The answer is definitively no. The USD-LIBOR-12M data for the March/April period is not used, only the LIBOR data to 2020-01-05 is used. What is the spike seen at the end of the curve? This is the market misestimation of the monetary policy. SOFR, which closely follow monetary policy, over the last year crisis is strongly impacting the spread, but not for credit risk reason. The spread there is the "wrong" one if the goal is to represent the credit spread, it only represents the market inability to predict the future rates. Is what we want to represent in the spread? Usually a floating rate instrument is there to avoid the future's prediction. In this case, the spread represent the opposite of what we would like. A second spike will appear in the next 3 months and represent the credit spread effect.
This discrepancy can also be viewed in Figure 2.
Figure 2. Distribution of spread for USD-LIBOR-12M.
We have represented the distribution of the spread with the most recent data in a lighter colour. The recent spreads are high while today all the rates are low, the forward spreads are low due to the injection of liquidity by central banks and all short term historical spreads are low. The same economical reality produces completely opposite effects on different figures that are supposed to provide a fallback to the same LIBOR reality.
Recent related post: Historical spread adjustments: GBP figures, Historical spread adjustments: USD figures and CME steers away from ISDA fallback.
Fallback transformer series: LIBOR Fallback transformers!
If you want to be ahead in the transition game, contact us for LIBOR fallback and discontinuation: trainings, workshops, advisory, tools, developments, solutions.
We compute estimates of the historical spread adjustment for GBP-LIBOR on a regular basis. The data below is as of 2021-01-04.
| Spread type | GBP-LIBOR-1M | GBP-LIBOR-3M | GBP-LIBOR-6M | |
| Current Median | 3.4773 | 12.2818 | 27.8621 | |
| Minimum Median | 3.4305 | 11.4454 | 25.2390 | |
| Maximum Median | 3.5505 | 12.4071 | 28.9121 | |
| Mean | 3.6657 | 13.8936 | 27.095 | |
| Mean-Median | 0.1884 | 1.6118 | -0.7671 |
Contact us for more details on the hypothesis and methodology used.
We can provide production grade code that run those estimates on a daily basis.
Figure: Distribution of GBP-LIBOR-1M / GBP-SONIA compounded in arrears over 1M.