Synthetic LIBOR, genuine manipulation
I have expressed my opinion about synthetic LIBOR on many occasions in the last couple of years, in conferences, seminars, informal discussions, etc., but I have not directly posted anything substantial on this blog. This post uses as an example the USD LIBOR to be created next June but applies in similar manners to GBP and JPY LIBOR created at the beginning of this year. I already mentioned synthetic LIBOR in my first blog after the official announcement of LIBOR discontinuation: “Alea iacta est: LIBOR non est”.
For a transition that was announced by the same FCA in its ill-titled “the future of LIBOR” speech more than 5 years ago and that was described as at a date that ``is far enough away”, if the need for a synthetic LIBOR sounds like a failure it is because it is.
What is synthetic LIBOR? It means continuing to publish a rate, still call LIBOR, in the way and with the mechanism used for LIBOR, but not with an actual meaningfully economical number but with a number generated in a formulaic way using a formula decided by regulators without financial justification.
From a legal perspective, one can understand what they are trying to achieve. The last 5 years have not been enough to ensure that all the existing contracts referring to LIBOR do it in a way that can endure the discontinuation of the publication of the number. By imposing the synthetic LIBOR, regulators ensure that the “tough legacy” contracts are not frustrated.
But those who are naturally frustrated by this manoeuvre are the good faith LIBOR users.
Remember that in July 2017 Andrew Bailey said "It would therefore no longer be necessary for us to sustain the benchmark through our influence or legal powers." Any discussion for synthetic LIBOR or non-representativeness is thus a failure by FCA in general and A. Bailey in particular to truly understand what LIBOR represents to the market. The least we can say is that they have been economical with truth. I have not seen an official FCA’s document saying “we were wrong in 2017, we did not understand the LIBOR market, we have to change our approach to the transition” or a similar comment by Mr. Bailey.
The formula used in the synthetic LIBOR is based on a floating rate generated from some overnight-indexed market and a fixed spread. This spread is supposed to represent the credit and liquidity risk difference between the LIBOR mechanism and this new overnight-indexed mechanism.
The history of this spread is worth mentioning.
The spread was decided in an ISDA consultation by a simple majority of strongly conflicted respondents – I was one of them. The spread was designed to be used in new trades for which a new LIBOR definition would be created with ample time for users to review the new definition implication.
The ISDA decided spreads are fair in a very restricted sense. For a new trade, if you know the rules, you can price them - if you don’t know how to price them, you can always get advice from highly qualified independent advisors – like myself ;) – and incorporate them in your risk management strategies. The spreads are fair because they are known before the trades are entered into.
In no way they can be considered as fair for existing trades. ISDA explicitly indicated that there will be “losers and winners” from such a fallback. To understand how representative those spread are, just consider the following facts:
• the spreads were computed with past 5Y of data to extract one number to be applied to the end of time (in practice for the next 30 or 50 years)
• the spreads are computed as median; changing the single word “median” by “mean” has an impact of more than 6 basis points on USD-LIBOR-3M. This is on the notional of several hundred trillion dollars. See for example here.
• the spreads are between forward looking LIBOR and backward looking compounded SOFR on different periods and have ranged from 0 to more than 100 bps in the historical data used.
• when discussing “fair”, it is often understood as fair in terms of equal value. Valuation in derivatives is done by replication argument and is based on hedging. Where is the hedging when you compare forward looking LIBOR to backward looking composition?
• the valuation of derivatives is based on risk neutral probability, not historical probabilities; it involves a mean, not a median
• the LIBOR rates and overnight-index rates used in the spread computation were not for the same period.
Note that it appears that in the consultation used to justify the spread, the respondents appear incoherent with their understanding of median. The results of the consultation were that “consistency is critical or very important for 70% of the respondents”, nevertheless, the responses themselves are not consistent:
• Use median 55/90 - 61% Yes
• Exclude the outliers: 44/90 - 49% No
Median is, by definition, the exclusion of all outliers, actually it uses only 1 or 2 data points. Do the answers indicate that respondents don't understand what median means (pun intended)?
Notwithstanding those issues related to the spread, the regulators have decided to impose it on the world.
It was imposed on existing professional trade by black mailing the banks: sign protocol or face difficult questions. I have discuss this in several blogs, for example here.
It was later imposed on the general public through the synthetic LIBOR by manipulating lawmakers in order to give regulators exorbitant power that does not exist in the EU version of BMR. The Article 21, 3 of EU Regulation REGULATION (EU) 2016/1011 says “the competent authority shall have the power to compel the administrator to continue publishing the benchmark“. Nowhere there is the power for the authority to change the definition/formula of the benchmark.
If one of Brexit's goals was to remove EU imposed regulatory power by unelected bureaucrats, it failed dramatically in the BMR case. The FCA power is not only to force administrators to continue publishing an existing benchmark but also the superpower to impose a new formula/fixing mechanism for that benchmark. The FCA can, at their own whim, decide that LIBOR is 20%, or that RPI is fixed at 100 for the next 10 years, or decide that the official temperature in London is now 25 degrees everyday! The last one is perhaps a slight exaggeration, but only because nobody has (to my knowledge) yet registered a benchmark representing London temperature to be used in financial products that have been deemed to be “critical”.
The FCA considers that a consultation run by a private company with less that one hundred respondents, most of them conflicted, is a solid enough basis to decide on a new law (creating synthetic LIBOR power) to be used in a different context for all end users, including retail customers.
By such a definition of fair, LIBOR has always been a fair number, especially when it was manipulated by panel banks collusion. If the bank consensus made the fallback adjustment spread fair, the bank collusion around LIBOR manipulation, which is a form of consensus, made LIBOR "fair" also at that time.
When banks want to change a minute detail of a model used for a couple of insignificant trades, they need a model validation. Was there any in depth model validation for this major change potentially impacting billions or trillions of notional? Was there a precise review of which trades are impacted by the change? To the best of my knowledge the answer is no! Regulators appear to request — with reason — a lot from banks but very little from themselves.
Note: I have not been able to find the full text of the UK BMR, even on the FCA benchmark pages. I contacted the FCA “Benchmark supervision” and their answer was “We do not have the full text of the UK Benchmark Regulation document on our website, however as far as we are aware, all such information is in the public domain including legislation.gov.uk.” I have not found the actual text allowing for the “synthetic LIBOR” on the mentioned site. Probably not searching with the correct keywords. If you know where to find it, don’t hesitate to let me know.
Note added 2022-12-22: After many attempts, I finally found the place where the new power to change the benchmark nature (Article 23D) is located: Financial Services Bill 2020. The new Article 23D says "The FCA may by written notice impose requirements on the benchmark administrator relating to any of the following (a) the way in which the benchmark is determined, including the input data (b) rules of the benchmark [...]". The FCA has indeed to power to decide that the temperature in London is 25 degree (Celsius or Fahrenheit for that matter) every day for financial benchmarks related matters.
Note added 2022-12-26: The “Synthetic LIBORs” proposed by the FCA are based on CME term rate. Those term rates are based on futures. The volume weighted average of prices throughout the day is used. The conversion between futures prices and forward rates does not use any convexity adjustment. From the price/rates a “best fit” method is used, i.e. not an exact fit. A specific interpolation method with flat rate between FOMC meeting is used even if actual historical data for SOFR is not flat between meetings.
How do you hedge volume weighted average? How do you hedge interpolation? How do you hedge actual data being different from the simplistic model used by the benchmark administrator? Actual OIS would provide an exact benchmark but are not currently used.
I would prefer a benchmark based on a precise point in time data and based directly on a market product. LIBOR is based on 11:00am London time term deposit with an exact term and a clear market convention.
Comments
Post a Comment