A quant perspective on IBOR

fallback consultation results

Market Infrastructure Analysis

muRisQ Advisory January 2019

Electronic copy available at: https://ssrn.com/abstract=3308766

 Abstract

With the increased expectation of some IBORs discontinuation and the increasing
regulatory requirements related to benchmarks, a more robust fallback provision for
benchmark-linked derivatives is becoming paramount for the interest rate market.
Several options for such a fallback have been proposed and ISDA held a consultation
on some of them. The results of the ISDA consultation has been to privilege the
“compounding setting in arrears” option. This note, which can be view as the version
2.0 of a previsous note, note present the different options briefly and analyses the
privileged option in details. It also presents an alternative option supported by
different working groups. The note’s focus is on the quantitative finance impacts for
derivatives. To our opinion, the option selected by the consultation fails the basic
achievability criterion in many cases. Even when achievable, the option can lead to
significant valuation and risk management complexities.

Copyright c 2018 by Marc Henrard. No quotation of this document is allowed

without properly identifying the source and the author.

Electronic copy available at: https://ssrn.com/abstract=3308766

Contents
1 Introduction 1

2 Relevant benchmarks and notations 4

3 Fallback provisions – adjusted RFR 6

3.1 Option 1: Spot Overnight Rate . . . . . . . . . . . . . . . . . . . . . 7
3.2 Option 2: Convexity-adjusted Overnight Rate . . . . . . . . . . . . . 9
3.3 Option 3: Compounded Setting in Arrears Rate . . . . . . . . . . . . 9
3.4 Option 4: Compounded Setting in Advance Rate . . . . . . . . . . . 11
3.5 Option X1: OIS Benchmark rate . . . . . . . . . . . . . . . . . . . . 12
3.6 Option X2: Futures based rate . . . . . . . . . . . . . . . . . . . . . . 14

4 Fallback provisions – spread adjustment 15

4.1 Option 1: Forward Approach . . . . . . . . . . . . . . . . . . . . . . 15
4.2 Option 2: Historical Mean/Median Approach . . . . . . . . . . . . . 16
4.3 Option 3: Spot-Spread Approach . . . . . . . . . . . . . . . . . . . . 17
4.4 One day spread capture . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.5 Period spread capture . . . . . . . . . . . . . . . . . . . . . . . . . . 17

5 Fallback provisions – analysis 18

5.1 Spread levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.2 Historical spread . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
5.3 Valuation: forward-looking v backward-looking . . . . . . . . . . . . . 21
5.4 Note on dates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.5 Contamination and value transfer . . . . . . . . . . . . . . . . . . . . 25
5.6 Non vanilla products implied by the fallback . . . . . . . . . . . . . . 27
5.7 Risk profile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.8 Clearing and margin . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.9 Litigation and disruption . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.10 Leg compression at CCPs? . . . . . . . . . . . . . . . . . . . . . . . . 36
5.11 Fallback of fallback and multi-stage fallback . . . . . . . . . . . . . . 37

6 Personal preference 37

A Gaussian HJM one-factor model 38

Electronic copy available at: https://ssrn.com/abstract=3308766

Author(s): Marc Henrard

Electronic copy available at: https://ssrn.com/abstract=3308766

1 Introduction
Since July 2017 speech by Andrew Bailey, the CEO of the U.K. Financial Conduct
Authority (FCA), on The future of LIBOR 1 , there is an increased expectation that
some IBOR benchmarks will be discontinued in a not too distant future. On a dif-
ferent front, several public/private working groups have been working on alternative
reference rates with the goal of replacing some of the current IBOR benchmarks with
benchmarks embedding les bank credit risk, the so-called Risk-Free Rates (RFR).
A phase-out of existing benchmarks is a major undertaking. As billions of no-
tional for derivatives with very long maturities are linked to those benchmarks, a
robust fallback procedure is certainly required. The fallback provisions in current
derivative’s master agreements do not seem robust enough to be sustainable in case
of discontinuation. In most derivative’s cases, the current fallback would be a daily
pooling of banks in the relevant market. If the banks have stopped contributing to
the organised IBOR pooling, it is dubious that they would contribute voluntarily in
an informal pooling.
Adding to the questions related to the benchmarks is the recent EU Benchmark
Regulations2 which requires enhanced rules for benchmarks providers and users.
There is no guarantee that the currently liquid benchmarks will fulfil the European
rules at the end of the transition period on 1 January 2020. It is expected that the
EONIA benchmark will not fulfil the rules but that EURIBOR will.
There have been several proposals on how to amend the current master agree-
ments to deal with the potential discontinuations. Some descriptions of the potential
proposals for fallback in case of discontinuation of IBOR rates have been proposed
by an ISDA working group3 . ISDA has published a consultation on benchmark fall-
backs to market participants on 12 July 20184 . The consultation is regarding GBP
LIBOR, CHF LIBOR, JPY LIBOR, TIBOR, Euroyen TIBOR and BBSW. The
results of consultation have been published on 27 November 2018.5 Even if USD
was not included in the main part of the consultation, we will use USD LIBOR in
some examples of this note. Other consultations covering other benchmarks and
currencies will be launched later.
The goal of this note is to remind briefly the different alternatives in the con-
sultation and to discuss the selected options in details. The note is a quantitative
finance view on the issue and follow-up on a previous note Henrard (2018b). We
try to provide as much details on the pricing and risk management impacts of the
different alternatives. We also propose an alternative option that has been excluded
0
First version: 25 April 2018; this version: Version 2.0 – 1 January 2019
1
Available at https://www.fca.org.uk/news/speeches/the-future-of-libor.
2
https://www.esma.europa.eu/policy-rules/benchmarks
3
Slide of presentation available at https://www.isda.org/a/vKiDE/
development-of-fallbacks-for-libor-and-other-key-ibors.pdf.
4
The consultation is available at https://www.isda.org/2018/07/12/
interbank-offered-rate-ibor-fallbacks-for-2006-isda-definitions.
5
https://www.isda.org/2018/11/27/isda-publishes-preliminary-results-of-benchmark-consultation/

Electronic copy available at: https://ssrn.com/abstract=3308766

from the consultation but appears to have a broad support in the market.
This note does not represent in any way an official description of the alterna-
tives or their actual legal wording. It represents a personal opinion of the author
to start the discussion around those issues and their quantitative finance details.
Also this note does not discuss the legal aspects of changing the fallback proce-
dures and updating the master agreement definitions. The objective triggers of the
discontinuation are not discussed in this note.
Currently there are a several alternative reference rates that have been proposed
for the main currencies: SOFR for USD, reformed SONIA for GBP, SARON for
CHF, TONAR for JPY and very recently ESTER for EUR. Note that the SOFR
rate definition itself has already an explicit and extensive fallback language in its
ISDA definition6 .
The potential discontinuation of IBORs will affect many products: derivatives,
syndicated loans, retail lending, mortgages, floating rate notes, etc. Each of those
products has a different legal framework. In this note we focus on derivatives under
ISDA master agreements. The definition of IBOR, the triggers for the fallbacks, and
the fallback procedures themselves vary significantly, even within the same product
set. For example in some cases, the fallback procedure is to move from a floating
rate to a fixed rate; in case of no publication of the benchmark, the rate applied is in
some cases the previously used rate in the same contract, in other cases it is a fixed
rate explicit in the contract or another benchmark altogether. Most of the current
fallback language has been written with the case of a temporary non-availability of
the benchmark rate in mind and not a permanent discontinuation.
An aspect that will need to be analysed carefully in this period of turbulence
about benchmark existence is the fallback of the fallback. If some USD LIBOR rates
do fallback to SOFR and later the SOFR benchmark is itself discontinued in favour
of OBFR, do the legacy USD LIBOR trades now fall under OBFR or are they in
some type of limbo state?
In a presentation from 20177 , ISDA indicated some criteria against which the
fallback rates will be measured. Among those, it says that the fallback should
mitigate

1. litigation risks

2. value transfers

3. manipulations

4. threats to market liquidity, market disruption and financial stability

In the consultation document, ISDA identifies the following criteria:

6
See exact wording at https://www.isda.org/a/kKHEE/Supplement-57-USD-SOFR-COMPOUND.
pdf.
7
Available at https://www.newyorkfed.org/medialibrary/microsites/arrc/files/2017/
OMaliaDarraspresentation.pdf.

Electronic copy available at: https://ssrn.com/abstract=3308766

 5. simplicity and ease of calculating

6. data requirements

7. similarity with the structure of overnight index swaps that reference the RFRs.

I will add my own criteria:

8. consistency

9. risk management continuity

10. valuation simplicity

For me the criterion on consistency is very important and even a must have. The
consistency should be preferably between asset classes but certainly inside the deriva-
tive asset class with a common master agreement. For derivatives the consistency
should be between products for a given benchmark family – e.g. USD LIBOR – and
between market participants. The procedure proposed should be the same for IRS,
FRA, and Libor exotics. The IBORs fixing at one given date used in the different
transactions should be unique. The fallback procedure will create an implicit risk
transfer and probably a value transfer but it should be consistent between instru-
ments. The fallback procedure should not create a basis risk that was not present
before the fallback. I’m completely opposed to any option that break the consistency.
To parody a recent famous sentence: LIBOR means LIBOR!
The fallback problem is not an overnight or a generic interest rate problem, it is
a specific term rate replacement problem. It is difficult to imagine that a coherent
answer that would not also be a term rate could exist.
The introduction of a fallback procedure may have a regulatory impact, on mar-
gin requirement in particular. All new bilateral trades are now under a mandatory
Variation Margin (VM) regime and some institution are under a mandatory Initial
Margin (IM) regime. Many vanilla interest rate derivatives are under a mandatory
clearing regime. The margin framework applies to all new trades, and an amendment
to an existing trade is to be considered as a new trade. Changing the benchmark def-
inition gives, without doubt, a new economic value to the trade; the change should
trigger the mandatory bilateral VM/IM and clearing. There is some expectation
that regulators may propose an exception or some relief for those amendments but
this has not been officially confirmed.
Any comments and suggestions for improvement to this note are welcome. It
is expected that the note will be enhanced over time with alternative proposals or
further details. This note has already benefited from the insightful comments of
Damiano Brigo and Marco Bianchetti.

Electronic copy available at: https://ssrn.com/abstract=3308766

2 Relevant benchmarks and notations
An IBOR rate itself is characterised by three dates. The derivative payments refer-
encing an IBOR rate are themselves characterised by (packages of) four dates. This
means a total of seven (actually six as explained below) dates have to be taken into
account in the different descriptions. The terminology and notations of this note
are the one used in Henrard (2014).
Let’s j designate an IBOR benchmark, like “CHF-LIBOR-3M”. The three dates
that characterise a j-IBOR are its fixing date, the effective date of the underlying
deposit and the maturity date of the same deposit. For all IBOR benchmarks under
consideration, the benchmark number is published on the fixing date and we do not
distinguish between the fixing date and the publication date. For derivatives, the
effective and maturity dates of the underlying deposit are theoretical dates as no
actual deposit takes place. But those dates are the one used by the panel banks to
estimate the answer to the LIBOR question “At what rate could you borrow funds,
were you to do so by asking for and then accepting interbank offers in a reasonable
market size just prior to 11 am? We will denote those dates by t0 for the fixing
date, u for the effective date and v for the maturity date. The accrual factor between
dates u and v in the day-count convention of the benchmark is denoted δ j (t0 ) or
δ j (t0 , u, v) if we need to clarify all dates. The relation between those dates are fixed
by the conventions associated to the benchmark. For example for CHF-LIBOR-3M,
for a given fixing date t0 , the effective date is two London good business days after
the fixing date, except if that date is not a Zurich good business date in which
case it is moved to the following good business day for both London and Zurich.
The maturity date is three-month after the effective date using a modified following
rule with the Zurich calendar. The dates are represented Figure 1. Those rules are
very explicit and depend only of the IBOR convention and are not related to the
derivatives in which the IBOR fixing may be used. The fixing rate for date t0 and
index j is denoted I j (t0 ).

t0 u v

Underlying deposit

Figure 1: Representation of the dates associated to a IBOR fixing.

The published fixing are then used in derivatives. Derivatives payments are them-
selves characterised by (multiples of) four dates: the fixing date, the start accrual
date, the end accrual date and the payment date tp . In some derivatives, several
fixing are grouped together to form a unique payment, this is why we have used the
term “multiple of” in the above sentence. But for the purpose of the description in
this section, we can focus on one fixing at a time, even if for pricing purposes, the

Electronic copy available at: https://ssrn.com/abstract=3308766

full payment details and the possible aggregation of several fixings have to be taken
into account. The fixing date is the one making the link between one derivative
payment and one (or several) published numbers. The start accrual date and end
accrual date are used to compute a multiplicative factor, the accrual factor, used in
the derivative’s payment. Even if computing this number correctly is important in
practice, this is only a static number and does not (should not?) influence the fall-
back. The payment date is the most important date. It indicates when the amount
computed on the fixing date has to be paid. Certainly the amount to be paid has
to be known before the payment date and this creates several constraints on what
is possible in the fallback procedure. The dates are represented Figure 2.

AS AE tp

Accrual period

Figure 2: Representation of the dates associated to a IBOR derivative.

Let’s O denote an overnight benchmark, like SONIA. The four dates that charac-
terise an overnight benchmark are the fixing date, the publication date, the effective
date of the underlying deposit and the maturity date of the same deposit. We will
denote those dates by t0 for the fixing date, t1 for the publication date, w for the
effective date and x for the maturity date. The accrual factor between date w and
x in the day-count convention of the benchmark is denoted δ O (t0 ). The relation
between those dates are fixed by the conventions associated to the benchmark. For
example for SONIA, for a given fixing date t0 , the publication date is the next Lon-
don good business date (at 9:00 am), the effective date is the fixing date and the
maturity date is the next London good business day. The fixing for date t0 and
overnight index O is denoted I O (t0 ).
Taking those date into account precisely is important, as this is what will de-
termine if some approaches are achievable or not and which one fulfil the criteria
mentioned in the introduction.
Some approaches relate in some way to compounded overnight, which is the way
the floating leg of an OIS is computed in most cases. The floating cash flow of an
OIS is described below.
If we denote by [si−1 , si ] the different overnight periods and δi the accrual factor
associated to each of these periods, the payment of the OIS coupon on the period
[s0 , sn ], for a swap of notional N , is
n
!
Y
1 + δi I 0 (si−1 ) − 1 .


N (1)
i=1

In theory the amount should be paid in sn , but in practice, to insure a smooth

settlement, the payment is lagged and takes place in most currencies two days after

Electronic copy available at: https://ssrn.com/abstract=3308766

the last fixing publication. In means that in EUR/EONIA, the payment take place
one business day after sn and in USD/EFFR, the payment take place two business
days after tn . In GBP/SONIA, the payment take place on the last publication date
itself without lag. In the above formula, we suppose that the day count convention
of the swap payments is the same as the day count convention of the index. This is
the case in practice for all standard OIS. When we compare an OIS with a IBOR
period, we will take s0 = u and sn = v. The dates are represented Figure 3.

s0 s1 s2 si−1 sn−1

u v tp

Accrual/compounding period

Figure 3: Representation of the dates associated to an OIS.

3 Fallback provisions – adjusted RFR

The fallback procedure would work as follows. When the IBOR benchmark discon-
tinuation is announced, some market values are measured. The values measured are
related to the discontinued benchmark and the replacement benchmark used in the
fallback procedure. In the wording of the consultation, the number computed from
the new benchmark is called the adjusted RFR and the spread above that rate to
match the original benchmark is called the spread adjustment. For all the cases for
which an actual procedure is already under discussion, the fallback benchmark is an
overnight-linked benchmark. For the adjusted RFR, the rate is function of one or
several overnight-linked benchmark fixing. The exact timing and mechanism of the
adjusted RFR measurements are described in the rest of this section. The details
for the spread adjustment are described in the next section. For all approaches, we
denote the adjusted RFR used for benchmark j at fixing date t0 by FRj (t0 ) – the
notation FR is the one used in the ISDA consultation document.
The values are computed for dates just before the announcement of the discon-
tinuation, but the fallback procedure will apply only from the discontinuation date
itself. In some extreme circumstances, those two dates could coincide, but it is ex-
pected that there will be several months between the two. From the discontinuation
date on, the fallback for each IBOR fixing date is the adjusted RFR measured in
the fixing date plus the fixed spread adjustment measured on the announcement
date. The spread can be IBOR fixing date dependent. This means that the spread
between the discontinued IBOR rate and the new overnight-linked reference is now
fixed and not changing anymore.

Electronic copy available at: https://ssrn.com/abstract=3308766

 In all cases, a spread between the estimated IBOR rate and some overnight-linked
rate is computed around the announcement date. That spread for a given IBOR
fixing date is recorded on the day of the IBOR rate discontinuation announcement.
When the payment date associated to the fixing date actually comes, the IBOR fixing
rate is replaced by the adjusted RFR rate plus the previously recorded spread. As
some IRS have maturities up to 50 years, a 50-year period may elapse between the
date the spread is computed and the date it is effectively used.
The change of reference benchmark through the fallback has at best no impact
on the valuation of vanilla swaps on the discontinuation date, but has immediately
a risk management impacts. The risk of IBOR is not the same than the risk of an
overnight-linked rate and all users have to review their risk management process.
Moving to overnight-linked benchmarks as the dominant benchmarks may simplify
the valuation of vanilla instruments by simplifying the vanilla contracts; in the best
case, both the collateral/discounting and the reference benchmark will be the same.
But it will only simplify the bank’s balance sheets on the vanilla derivative side,
there is no guarantee that it will simplify the balance sheet globally. The funding
term structure of a bank will become the IBOR of tomorrow but without its liquidity.
The generalised VM/IM and its associated costs have pushed the funding issue into
the valuation of derivatives (FVA, MVA). There is no need to see the IBOR or the
funding cost to appear directly in the instrument term sheet to need it in derivatives
valuation. The users relying on the risk of IBOR-based products to match in part
the risk of other products, like bank borrowing (directly or through xVA), will need
to readjust their hedging portfolio beyond the one-off static spread computation but
in a dynamic market way.
Since the announcement of the results of the consultation,8 it is expected that
the adjusted RFR will follow the compounded setting in arrears mechanism (Option
3). Nevertheless we remind briefly what the other options were and add some other
market proposed options that were not in the consultation.

3.1 Option 1: Spot Overnight Rate

In this approach the rate replacing the IBOR rate fixing at date t0 is simply the
overnight rate fixing on the same date

FRj (t0 ) = I O (t0 ). (2)

The advantages are: it is easy, all the data is available readily and the adjusted
RFR is an interest rate.
The main disadvantage is also clear: the interest rate is not related to the tenor
of the original benchmark j. This will means no differentiation between tenors, the
economics of the original period are not taken into account, the idiosyncrasies of
8
The related documents are available at https://www.isda.org/2018/12/20/
isda-publishes-final-results-of-benchmark-fallback-consultation/

Electronic copy available at: https://ssrn.com/abstract=3308766

the overnight period are imposed on a term rate, e.g. end-of-month jumps, central
bank policies changes and monthly seasonality. Moreover the simple rate over a
term deposit is not equal to the same rate for overnight deposits compounded over
the period. The dates are represented Figure 4.
t0 = w

x u v = tp

Depo = 1d Accrual

Figure 4: Representation of the dates associated to the Spot Overnight Rate option.

This would certainly create large risk management disruptions in the system, by
removing the implicit averaging mechanism present in term deposits. The valuation
of derivatives would require very precise curve calibration procedures taking into
account daily idiosyncrasies of the market. Those procedures have been developed9
and are used by many financial institutions. Imposing them on market participants
that did not have the requirement to deal with them in the past would certainly
create risk management disruptions. The risk disruptions would in turn lead to
market disruptions and instability in the derivative market. Moreover the legacy
trades after the fallback provision would not be similar to any OIS that are based
on compounded rates. This creates a risk management discontinuity at the dis-
continuation date but also the quasi impossibility to risk manage the legacy trades
short of cancelling them. The idiosyncrasies of the overnight rate would need to be
embedded in the valuation and risk management systems where today some market
participants may rely on the averaging features of term rates. The features like
end-of-month and central bank policy meetings dates would need to be introduced
in the curve description and calibration tools. This would create very important
operational challenges for many existing systems and the less sophisticated users.
On top of this, the valuation would require a timing/convexity adjustment as
the rate is not paid on its natural period, an important valuation complexity. This
adjustment were discussed in more details in Section 5.4 of the version 1.1 of this
note (Henrard (2018b)) but have been removed from this version 2.0 as they are less
relevant now with the results of the consultation.
It is not surprising that in the consultation, very few respondents (2 out of 142)
selected this option.

9
See for example (Henrard, 2014, Chapter 5)

Electronic copy available at: https://ssrn.com/abstract=3308766

3.2 Option 2: Convexity-adjusted Overnight Rate
Like for option 1, only one rate is used in this option, but that rate is adapted to
try to take into account some of the natural composition of overnight rates over a
longer period. I would have preferred a name like Composition-adjusted Overnight
Rate.
For that option, the formula is

δj (t0 )/δO (t0 )
FRj (t0 ) = I O (t0 ) 1 + δ O (t0 )I O (t0 ) . (3)

I have not been able to create a clear explanation for this formula. If one wants
to take the effect of composition, one could use the continuous compounding ap-
proximation,
1 
δj (t0 )/δO (t0 ) 
FRj (t0 ) = j 1 + δ O (t0 )I O (t0 ) −1 (4)
δ (t0 )
or better the composition using the correct market dates:
n
!
1 Y
FRj (t0 ) = j 1 + δi0 I O (t0 ) − 1 .


(5)
δ (t0 ) i=1

The advantage of this option is to try to correct one of the disadvantage of the
previous one, which is the composition of constant simple rate does not produce an
equal simple rate. But that disadvantage is to my opinion the least of them. The
correction comes at the cost of non-linearity in the formula and more exotic valuation
framework. The disadvantages of the previous option in term of risk management
and valuation are amplified. If any composition is used, one could use the one with
the actual market date of Formula (5).
It is not surprising that in the consultation, very few respondents (2 out of 142)
selected this option.

3.3 Option 3: Compounded Setting in Arrears Rate

For this option, the consultation text indicates “The fallback could be to the rele-
vant RFR observed over the relevant IBOR tenor and compounded daily during that
period.” This is translated in the consultation into the formula
n
!
1 Y
FRj (t0 ) = j 1 + δi0 I O (si−1 ) − 1 .


(6)
δ (t0 ) i=1

The consultation document does not precise which dates related to the IBOR
tenor have to be used as s0 and sn . To keep a consistency, one should obviously use
the start and end accrual date of the synthetic underlying IBOR deposit (u and v
in our notations), not the one of the derivatives. But then the overnight fixing may

Electronic copy available at: https://ssrn.com/abstract=3308766

not be known on the payment date, making it impossible to do the actual payment.
This issue of ambiguity on dates between the IBOR dates and the derivatives dates
is detailed in Section 5.4. To make the answer to the consultation meaningful, the
ambiguity on dates should have been removed before the answers were received. If
the dates are the one of the IBOR deposit, as it should be, the consultation should
have indicated how the computation would be achieved in practice without requiring
the knowledge of the future fixing occurring after the payment date. I have asked
the question to the consultation email and have received an acknowledgment answer
but no clear answer to the question. The details of the problem have been clearly
indicated in my answer to the consultation. The document presenting the results of
the consultation, referenced earlier, is simply a count of the number of answer and
quotes from some of them, but not an analysis of the quality of the answer.
The advantage of this option is that, when it is achievable and the derivative
dates align with the deposit dates, the rates are for the correct period and there is
a direct economic relation between the original IBOR and the fallback.
The huge disadvantage is that it is either unachievable or break the coherency
requirement. This approach has a very strong requirement in term of dates related to
the coupons. It requires that the dates related to the composition are all on or before
the payment date. Unfortunately this requirement is seldom met in practice, even
for vanilla swaps, due to non-good business dates, and completely impossible for
in-arrear swaps and FRAs. Even in the best case, this would require participants
to realise, in some currencies, the payment on the day when the amount itself is
known. In the IBOR coupon, there is no lag between the end accrual date and the
payment date, this would translate into coupons where there is no lag between the
date the amount is known and the payment. In GBP – reformed SONIA –, USD –
EFFR or SOFR – and EUR – ESTER –, the benchmark value is announced only
on the end date of the overnight period in the morning. It would create, even in
the best case, an important constraint on back-office for messaging and payments.
Further comments on those dates are provided in Section 5.4.
Would a payment delay, like the one existing in most OIS markets, be applied
to the fallback Libor payments? Such a lag has been proposed in answers to the
consultation and could be introduced in the fallback provisions. If yes, would it
be applied only to the Libor payments or also the other payments of the same
derivatives. Not applying the same payment lag to all payments may introduce
extra credit risk. Introducing payment lags may jeopardise the hedging purpose of
the derivative. The change of payment dates could introduce breaks in cash-flow
hedging. Maybe a lock-out period, like for Fed Funds swaps, would be introduced.
This lock-out period has also been proposed in some answers to the consultation.
Those lag adjustment and lock-out period would only solve the marginal side of the
issue, not its main problem in the case when the payment is required just after fixing
like for FRA and in-arrear fixing IBOR payments. Also the introduction of lag and
lock-out would create artificial convexity adjustments which did not exists in the
original IBOR derivatives.

10

Electronic copy available at: https://ssrn.com/abstract=3308766

 It is surprising that in the consultation, so many respondents (127 out of 142)
selected this option given the non-achievability of the option for many instruments
and the lack of details on what it means exactly in the consultation document. Ac-
tually this significant majority has to be alleviated by the fact that in the comments
many respondents indicate that it is not achievable for some trades or indicated
that some modification, like lock-out period, are required. We will come back to
potential modification latter.
The proposed fallback is not a simple LIBOR fallback, but a trade term sheet
fallback. The instrument term sheets have to be revisited at the same time as the
fallback.

3.4 Option 4: Compounded Setting in Advance Rate

In this option, “the observation period would be equal in length to the IBOR tenor,
it would end immediately prior to the start of the relevant IBOR tenor ”. In this
case also the formula proposed by the consultation is ambiguous. It indicate that
the composition period has to start in “T − f ” with f the IBOR tenor. The IBORs
are defined from their effective date to their maturity date (using a given conven-
tion, usually modified following); that rule does not defined the computation of the
effective date from the maturity date, i.e. it defines the “+” between one dates
and a period to one other date but not the “−”. Several effective dates can lead to
the same maturity date. For example a deposit starting on the 18 July 2018 will
matures on 20 August 2018 as 18 August 2018 is a Saturday. But deposits starting
on 19 July 2018 or 20 July 2018 will also mature on 20 August 2018. It is also
possible no effective date leads to the required date.
The advantages of this option is that it is achievable – the fixing available on the
fixing date – and the overnight rates are averaged, removing some of the idiosyn-
crasies of the overnight market with respect to the term market.
In term of disadvantages, the rate is fixed on the wrong period. The averaging
advantage is done on the wrong economic reality. To some extend it is not even
important to do the compounding on the same tenor as the IBOR tenor. It could
be argued that for a one-year tenor, it would be better to average on a one or three-
month period before the fixing instead of a one year. The last month is more likely
to be representative of the next year that the previous full year in term of interest
rates. If such a method is used, my suggestion would be to always use a one-month
period, this would smooth the end-of-month and tax payment idiosyncrasies but not
create a too large payment delay.
One other important disadvantage is that it introduces a timing adjustment for
currently vanilla swaps as the rates are paid for the wrong period. All the vanilla
swaps would become some new sort of exotics with convexity/timing adjustment
and path dependency. It certainly breaks the valuation simplicity criterion. It is
doubtful that some market participants dealing only with linear products would
have the systems and expertise to value and risk manage those trades.

11

Electronic copy available at: https://ssrn.com/abstract=3308766

3.5 Option X1: OIS Benchmark rate
This option and the following were not proposed in the ISDA consultation. To my
opinion they should be part of the general discussion on fallbacks. This opinion is
shared by many market participants, as can is evidenced by the different working
group analysing this mechanism in EUR10 , GBP11 , USD12 and JPY13 .
In t0 the rate of the OIS associated to the IBOR deposit start and maturity
dates is measured to create an OIS benchmark. Those OIS benchmarks may not
exists currently or may only be seldom used. Certainly at this point there is very
little scrutiny on those benchmark and they cannot be consider as significant bench-
marks. With the increase importance of overnight benchmark and the proposal
from different private/public working groups to replace IBOR term benchmarks by
overnight-linked benchmarks, it seems appropriate to work on the creation of those
benchmarks. The underlying OIS market seems to be more liquid and transparent
than the current unsecured term deposits, at least different electronic trading venues
(i.e. Multilateral Trading Facilities – MTF) provide real time tradable quotes.
In the collateral framework, the forward or spot OIS rate in t for the period [u, v]
can be computed from the overnight collateral pseudo-discount factors as:

P c (t, u)
 
c 1
F (t; u, v) = −1 (7)
δ P c (t, v)

This make the computation of those numbers very easy and the valuation and risk
management of related derivatives straightforward. Section 5.3 also shows that in
term of valuation before the fixing date, it is equivalent to Option 3, when the latter
is achievable. When the overnight rate is based on repo-like transaction, as is the
case for USD-SOFR and CHF-SARON, the rate can also be obtained from longer
term repos as explained in Henrard (2018c).
In most currencies the OIS payments are lagged by two business days. The lag is
required to have a couple of days to prepare a smooth payment. Such a lag does not
exist for IBOR coupons. Would the standard OIS lag be taken into account? The
benchmark could be based on the direct market quote, even if that market quote
is for a lagged payment. One could also try to adjust the OIS quote to take the
IBOR absence of lag into account. The analysis of the impact of the lag and its
convexity adjustment was done in Henrard (2004). The conclusion regarding the
two-day payment lag was that the impact was negligible.
10
ESTER-based term structure methodology as a fallback in EURIBOR-linked contracts: https:
//www.ecb.europa.eu/paym/cons/html/wg_ester_term_structure_methodology.en.html
11
Consultation on term SONIA reference rates: https://www.bankofengland.co.uk/-/media/
boe/files/markets/benchmarks/consultation-on-term-sonia-reference-rates.pdf
12
FAQ 6 on “term SOFR” date 20 September 2018: https://www.newyorkfed.org/
medialibrary/Microsites/arrc/files/2018/ARRC-Sept-20-2018-FAQ.pdf
13
Term rate and TONAR liquidity: https://www.risk.net/derivatives/6229596/
qa-japan-rfr-group-head-on-term-rates-and-tonar-liquidity

12

Electronic copy available at: https://ssrn.com/abstract=3308766

 The very important advantages of this option is that it matches the current
IBOR process with a rate known at the same time as the IBOR rate and it is easy
to compute. It creates a consistency, no valuation complexity and very little risk
management complexity and discontinuity. Moreover when used with the correct
spread, it reduces the value transfer for vanilla products. The underlying instrument
is the already exisiting OIS market.
The main disadvantage of this option is to require the creation of a new bench-
mark, one based on OIS. The data may not be readily available at this stage. But
as overnight benchmarks are becoming more important than IBORs, it would make
sense to have a such a benchmark. The IBOR swap markets have such swap bench-
marks, called ICE Swap Rate14 , which are used daily for cash settle swaptions and
CMS. The liquidity of the OIS market should increase and the option should become
feasible soon. Some of the documents referred to above contain the analysis of the
OIS markets in the different currencies. From those documents, it seems that a
benchmark based on committed quotes would be feasible.
In a recent document the Financial Stability Board (2018) expressed its believe
“that transition of most derivatives to the more robust overnight RFRs is important
to ensuring financial stability.” It also acknowledged that “An overnight RFR may
not, however, be the optimal rate in all the cases where term IBORs are currently
used.” Those opinions appear compatible with this option. The market would tran-
sition to compounded overnight rates setting in arrears for new products and in par-
allel, for the case of the discontinuation of an existing term rate setting in advance,
to a RFR-linked term rate. From a valuation perspective and risk management
perspective, as demonstrated in Section 5.3, the two mechanisms are similar, but
from an achievability perspective, the two are diametrically opposed. The FSB’s
position described in the document referenced above does not appear as negative as
the explanation in the ISDA’s FAQ15 related to the consultation on why this option
was not in the consultation. The FSB’s document goes as far as stating “The FSB
supports the exploration of the potential to create new RFR-derived term rates in
these jurisdictions.” Moreover in a recent panel discussion at a ISDA meeting, a
FCA representative, Edwin Schooling-Latter, positively commented on the achiev-
ability of term versions of RFRs. His comments are presented in the recent Risk
article Wood (2018).
The option proposed above has not as a goal to move the bulk of the trading
of the most liquid interest derivatives to RFR-linked term rates, but to ensure a
smooth discontinuation process from a valuation and risk management perspective.
The FSB’s document also state that the “RFR-derived term rates [...] should be
used only where necessary. It is my believe that they are necessary for the fallback
provisions. The document adds that “the FSB supports a focus on the overnight
RFRs as a primary IBOR fall back rate” but does not provide any detail on how
such a fallback procedure would work and does not exclude an indirect focus through
14
https://www.theice.com/iba/ice-swap-rate
15
https://www.isda.org/a/RNjEE/Fallback-Consultation-FAQ.pdf

13

Electronic copy available at: https://ssrn.com/abstract=3308766

derivatives.
One version of the spread adjustment was the forward approach, presented in
Section 4.1. To compute the adjustment, a full overnight and a full IBOR forward
curve need to be calibrated. The calibration of the curves required to have the full
term structure of IBOR swaps and OISs as inputs. Those inputs are as important
as the benchmark itself, even if it is not clear to me if they qualify as a benchmark.
If the forward approach to spread adjustment is deemed acceptable, using only one
of such swap rate – where the market is the most liquid – in the fixing should also
have been deemed acceptable for the consultation purpose.
As mentioned, such a mechanism already exists for swap benchmark with ICE
Swap Rate. A similar mechanism for OIS should be possible. Is there more liquidity
on GBP IRS 9Y or on EUR OIS 3M? I don’t know the exact figures, but I would be
very surprised if the liquidity on the GBP IRS 9Y was an order of magnitude larger
that the EUR OIS 3M. The OIS benchmark could also be based on futures as pro-
posed by IBA (ICE (InterContinental Exchange) Benchmark Administration)). The
new term benchmark16 is computed based on ICE traded futures and a (simplistic)
curve model.

3.6 Option X2: Futures based rate

In the absence of OIS benchmark to represent the term rate, other traded instru-
ments with public prices could be used. In some currencies (USD and GBP), some
overnight-linked futures are semi-liquid. There is no futures with expiry exactly on
each fixing date, so the exact IBOR period can not be replicated exactly without a
model. But one could imagine a fallback where the best combination of futures (in
a sense to be specified) would be used. One possibility would be to take the rate
from the futures with the most days overlap with the IBOR replaced. To have a
simple adjusted RFR, the futures prices would not be adjusted for convexity. For a
one-month IBOR, the one-month futures with most overlap would be used. There
would be between 15 days and a full month overlap. On average, 75% of the period
is in common. For three-month IBOR, the three-month futures with most overlap
would be used. There would be between 45 days and full three-month overlap. On
average, 75% of the period is in common. For six-month IBOR, the two three-month
futures with most overlap would be used. There would be between 135 days and
the full six-month overlap. On average, 87.5% of the period is in common.
The obvious disadvantage would be that the rate inferred from the futures would
not be the rate for the exact IBOR period. This is already the case for the Options 1
to 4 proposed by ISDA. The second disadvantage is that the rate would jump when
the fixing pass from one futures to the other. This problem is already present with
Options 1 and 2 as the overnight rate usually follow central bank policies and jump
on change of policies.
16
https://ir.theice.com/press/press-releases/all-categories/2018/
10-10-2018-140016614

14

Electronic copy available at: https://ssrn.com/abstract=3308766

 The advantages are that the rates are publicly traded, they figures are published
on a daily basis and have a long history, the rates are in large part a realistic
economic replacement of the interest rate part of IBOR (with the bank credit taken
into account by the spread adjustment).
This is not my preferred option, but in the absence of OIS benchmark (forward
looking), this seems to be a good compromised on using readily available data and
a sensible economic replacement.

4 Fallback provisions – spread adjustment

Like for the adjusted RFR, the spread adjustment preferred option from the consul-
tation has been relatively clear. The majority has selected the Historical mean/median
approach. We leave the other options in this note, specially the forward approach
for comparison purposes.

4.1 Option 1: Forward Approach

In this option for the spread adjustment computation, the spread adjustment [is]
calculated based on observed market prices for the forward spread between the relevant
IBOR and the adjusted RFR in the relevant tenor at the time the fallback is triggered.
The forward rates are measured on the discontinuation announcement date for
each fixing date up to the most distant fixing date of all financial instruments re-
lated to the discontinued IBOR benchmark. For each of those fixing dates, the
spread between the estimated forward IBOR and estimated forward adjusted RFR is
recorded. This guarantees the absence of value transfer for standard IBOR coupons
at the moment of the measurement if the spread estimates are correct.
From a valuation perspective, on the date of the discontinuation, the value of
vanilla swaps will be unchanged. A rate is replaced by another rate and the necessary
spread is added to insure that the coupons have the same value.
After the discontinuation date, the IBOR rates are not published anymore, so it
is not possible to measure a spread anymore. But if we think of the IBOR fixing
as still existing, but becoming unaccessible to the market, we could say that in an
abstract way the spreads are still evolving but we are not able to observe them
anymore and we keep the last visible value from there on. The change of reference
benchmark through the fallback has no impact on the valuation of vanilla swaps on
the discontinuation date, but has immediately a risk management impact and do
not lead to the same cash flows in the future. The risks and cash flows of a IBOR
coupon are not the same than the risks and cash flows of an OIS and all users have
to review their risk management process.
The absence of value transfer is only the case for standard IBOR coupons on
the discontinuation date. Many other products, even plain vanilla, may not enjoy
the same absence of value transfer.

15

Electronic copy available at: https://ssrn.com/abstract=3308766

 As some of the adjusted RFR are not traded in the market, the notion of rate
calculated on observed market prices may be very theoretical. The only case where
one could argue that there is a market is for Option X1 mentioned in Section 3.5.
The disadvantages of this method is that the spread computation for each day
is not an easy task and depends heavily on non-market-observable hypothesis like
interpolation. Some examples of this are provided in Section 5.1. Also the data used
for the spread calibration need to be available, this may not be the case for very
long dated instruments.
Today’s markets, even for the most liquid instruments, trade in silos depending
on the market infrastructure in which they settle (various CCPs, bilateral with IM,
bilateral without IM). The forward rates implied by the curves in those different
market segments can be different. Which of those silos would be used for the spread
measurement?

4.2 Option 2: Historical Mean/Median Approach

In this option for the spread adjustment computation, the spread adjustment [is]
based on the mean or median spot spread between the IBOR and the adjusted RFR
calculated over a significant, static lookback period (e.g., 5 years, 10 years) prior to
the relevant announcement or publication triggering the fallback provisions.
Obviously this would mean that the long term historical data is available. That
may not be the case for the USD SOFR (published only since April 2018), the new
EUR ESTER benchmark (not published yet), or SONIA that was reformed in 2018.
The clear disadvantage of this option is that it potentially introduce very large
market value changes. The market value change will be analyses in more details
in Section 5.5. In particular we will review the impact that the fallback procedure
announcement had already on the market.
The consultation paper also proposes a transitional period as: This spread ad-
justment could then be used from the end of a one-year transitional period after the
fallback takes effect. During the transitional period, the spread to be used would be
calculated using linear interpolation between the spot IBOR/adjusted RFR spread at
the time the fallback takes effect (i.e., the spot IBOR-adjusted RFR spread on the
last date that the relevant IBOR is published) and the spread that would apply after
the end of the transitional period.
The consultation document claims that the transitional period would mitigate
against a “cliff effect” at the time the fallback takes effect if the spot IBOR/adjusted
RFR spread at that time differs from the historical mean/median. To my opinion
this is not really the case. The mitigation will have an impact only on the very near
payments (within one-year) but none on the other payments which have a more
important impact on valuation. The cliff effect on valuation would be clearly there
immediately for any medium to long term instrument. Only the very short term
cash flows would have the cliff effect reduced. This seems to be a extra complexity
that is adding very little value in the grand scheme of thing related to value transfer,

16

Electronic copy available at: https://ssrn.com/abstract=3308766

except maybe to reduce potential manipulation and insider information impacts.

4.3 Option 3: Spot-Spread Approach

In this option for the spread adjustment computation, The spread adjustment [is]
based on the spot spread between the IBOR and the adjusted RFR on the day pre-
ceding the relevant announcement or publication triggering the fallback provisions.
This option can be implemented with the variant One day spread capture de-
scribed in Section 4.4 or with the variant Period spread capture described in Sec-
tion 4.5.
One can see a continuum of options between Option 2 and Option 3, with the
reference period varying from one day to 10-year.
The advantages are the simplicity and the direct availability of the required
data. The clear disadvantage of this option is that it potentially introduces very
large market value changes like for the previous option.

4.4 One day spread capture

In this version of the fallback, all the measurements related to IBOR and overnight
linked instruments are performed on a unique day. The exact date is the date before
the announcement of the discontinuation.
On that date all the required figures are recorded. Depending of the Spot (Sec-
tion 4.3), historical (Section 4.2) or Forward version (Section 4.1), this means that
one number or a full family of numbers indexed by the future fixing dates is captured.
That date will take a very significant importance as the full stock of existing
IBOR linked products will be modified by what happen on that single date.
Moreover, that single day where the measurement will be done, will certainly not
be the day where the dying IBOR market will be the most liquid. Expecting that
IBOR will stop to be published soon, what liquidity would be available in the forward
IBOR market? Would there be a threats to market liquidity, market disruption and
financial stability on that very specific date? What economical signification should
be attributed to a measure of the expectation of a number that will cease to exist a
moment after the expectation is estimated? Specially in the forward version of the
IBOR-OIS spread, the one day spread captured just before the discontinuation is
not certain to provide very reliable numbers.

4.5 Period spread capture

In this version of the fallback, all the measurements related to IBOR and overnight
linked instruments are performed over several days during a given period and aver-
aged out. Typically the measurement of the spot/forward spreads would be done
daily for a period of one week, two weeks or one month preceding the IBOR discon-
tinuation announcement.

17

Electronic copy available at: https://ssrn.com/abstract=3308766

 The main reason behind such a proposal is to smooth out the numbers over
multiple days. As the figures will be used over a very long period, it seems reasonable
not to rely on a unique measure.
One of the impact of such an average is that the number is partly known before
the end of the period. Suppose that the average is done over the last 20 business days
before the discontinuation. The day before the announcement, a portion of 19/20 of
the spread is already known. If someone is aware of the forthcoming announcement,
the trades related to IBOR on that last date will be more impacted by the already
known figures than by the economic reality of the IBOR-OIS spread estimation of
that day. A swap user will have no incentive to express his view of IBOR-OIS spread
through trading, as even if his view is correct, his position will be replaced the next
day by a position with a fixed spread, the value of which is already decided in great
part and not related to the economic reality of the day or of the future. This impact
on existing long term trades is even larger in the historical mean/median approach.

5 Fallback provisions – analysis

5.1 Spread levels
In this section we give some examples of spread levels computed by several methods.
The goal of the section is not to propose a figure to be used in the fallback provision,
but to show the extend to which the figures can differ between seemingly similar
method using historical data sets.
Some forward spread levels for USD, EUR and GBP are presented in Figure 5.
The numbers in USD are the spread between LIBOR-3M and OIS, the numbers in
EUR are the spread between EURIBOR-6M and OIS and the numbers in GBP are
the spread between LIBOR-6M and OIS. We have selected the IBOR tenor used in
the most liquid swaps.
One technical issue that impacts the spread computation is the choice of inter-
polation scheme. The graph of Figure 5 has been done with natural cubic spline on
zero-rates.
In Figure 6, we have computed the forward spread between USD-LIBOR-3M and
USD-FED-FUNDS at the end of December 2017. The spreads are computed using
different interpolation mechanisms: natural cubic spline interpolation and linear
interpolation on zero-rates and monotonic cubic spline on log-discount-factors. It
is well documented that the interpolation mechanism as an impact of the level of
forward rates – see for example Ametrano and Bianchetti (2009), Andersen and
Piterbarg (2010) and (Henrard, 2014, Section 4.2). Here we describe the impact on
the spreads, even if the same interpolation mechanism is used for both curves. This
type of graph on the spread can found for example in Ametrano and Bianchetti
(2013). For the date presented, the difference in spread is up to 3 basis points.
In Figure 7, we display only the difference in forward rates – with linear interpo-
lator and natural cubic spline on zero rates – but this time at different dates. The

18

Electronic copy available at: https://ssrn.com/abstract=3308766

 Forward spreads
50
USD
45 EUR
GBP
40

35
Spread (in bp)

30

25

20

15

10

0
20

25

30

35

40

45
n-

n-

n-

n-

n-

n-
Ja

Ja

Ja

Ja

Ja

Fixing date Ja

Figure 5: Recent IBOR-OIS spreads in the main currencies. Data as of end Decem-
ber 2017.

differences change through time and in the small sample selected, the difference is
up to 10 basis points. Note that the jumps, if any, occur at the same period and
correspond to the curve nodes.
In the above examples, we have used only the simplest approach to spread com-
putation. The curves use the same nodes and the same interpolation methods. In
practice, curves are often calibrated on instruments leading to different nodes. For
example one could use STIR futures for a three-month forward curve and OIS for
the overnight curve. The difference in nodes between the IMM related dates and the
tenor related dates would add noise to the spread if not taken into account properly.
One way to deal with those issues is to use spread curves for calibration as described
in (Henrard, 2014, Section 5.9). Some may want to introduce end-of-month spikes
or central bank meeting dates in the curve calibration procedure, adding to the
diversity in the spread estimates.
One suggestion regarding data to use has been to use instead of OIS benchmark
rates data from overnight linked futures and interpolation between the quoted dates.
As described above, interpolation is technically simple but very difficult to agree on.
It is almost a matter of taste and an art more than a science. The market would
have to agree on an interpolation mechanism not only to capture the spread on

19

Electronic copy available at: https://ssrn.com/abstract=3308766

 USD Forward spreads
60
Natural Cubic Spline zero-rate
Linear Zero-rate
50 Monotone Cubic Spline on Log DF
Difference NCS ZR-Lin ZR
Difference Monot CS DF-NCS ZR
40
Spread (in bps)

30

20

10

-10
20

25

30

35

40

45
n-

n-

n-

n-

n-

n-
Ja

Ja

Ja

Ja

Ja

Date Ja

Figure 6: Impact of interpolation on spread computation. USD data as of end

December 2017.

the discontinuation date but also on a different interpolation mechanism to capture

the overnight-linked reference rate. Moreover the futures would need a convexity
adjustments. For the currently traded overnight linked futures – futures with daily
margin without collateral interest up to the end accrual date – even on the start date
of the accrual period there is a convexity adjustment. The convexity adjustment for
those futures is discussed in Henrard (2018a). Moreover for some currencies, like
GBP, the futures are based on a 3-month periods while the main term benchmark
is 6-month.

5.2 Historical spread

As mentioned earlier, historical data is not available in a significant lookback period
for the two main currencies USD and EUR as one benchmark (SOFR) has only been
published for 8 months and the other will start to be published only in 10 months
time. The following analysis is done for GBP-LIBOR and SONIA. The first results
can be seen in Figure 8. Historically in the last 10 years the spread GBP-LIBOR-
3M and SONIA compounding setting in arrears has moved from -3 basis points to
almost +60 basis points. The negative figure is not an error. The spread is measure

20

Electronic copy available at: https://ssrn.com/abstract=3308766

 Dec 2016
0.1 Jun 2017
Dec 2017

0.05
Differences (in %)

-0.05

-0.1
0

10

15

20

Period to fixing (in years) 25

Figure 7: Impact of interpolation on spread computation. USD data as of end

December 2016, June 2017 and December 2017.

between a forward looking rate and a backward looking rate on the same period.
The spread represent a credit impact (forward looking) and a rate impact (change
between expected rate and actual rates).
The range of spread is very large and the spread itself moves rapidly. In the last
6 months, it has moved from 35 down to 10 basis points. Note that todays spread
is well below the historical average. We will also compare the historical average to
the level of the basis market in Section 5.5.

5.3 Valuation: forward-looking v backward-looking

In this section, we show that OIS benchmark forward-looking and compounded
setting in-arrears rates have the same valuation before the fixing date – at least
in the case where the dates are aligned. In particular the choice does not have
an impact on the value transfer issue. Obviously the settlement mechanisms are
different and the final cash flows may be different, but viewed from a date before
the fixing date, the present values are the same. We also insist that, as describe in
Section 5.4, this is true only when the dates are aligned and the backward-looking
mechanism is not achievable in all cases.

21

Electronic copy available at: https://ssrn.com/abstract=3308766

 Overnight compounding in arrears fallback - past rates
1.4 LIBOR-3M
SONIA-CMP-3M
Spread
1.2 Average

0.8
Rate (in %)

0.6

0.4

0.2

-0.2
12

13

14

15

16

17

18
n-

n-

n-

n-

n-

n-

n-
Ja

Ja

Ja

Ja

Ja

Ja

Ja
Date

Figure 8: Historical spreads and historical average for GBP-LIBOR-3M v GBP-

SONIA compounded over 3-month.

If the benchmark is fixed in advanced, based on a OIS benchmark, the floating

leg amount paid is the spot OIS rate as measured at the fixing date t0 for the period
[u, v] and it is paid in v. We denote by N the notional and δ the payment accrual
factor. The value today is, using the measure associated to the collateral account,

EX (Nvc )−1 N δF c (t0 ; u, v) = N EX (Ntc0 )−1 δNtc0 EX (Nvc )−1 F c (t0 ; u, v) Ft0
     
  c  
c −1 P (t0 , u) c
X
= N E (Nt0 ) − 1 P (t0 , v)
P c (t0 , v)
= N EX (Ntc0 )−1 (P c (t0 , u) − P c (t0 , v))
 

= N (P c (0, u) − P c (0, v))

The first equality used the tower property of conditional expectation. The second
equality uses the fact that F c (t0 ; u, v) is Ft0 -measurable, Formula (7) for the forward
and the definition of P c (t0 , v). The last equality uses the martingale property of
P c (., x).
The cash flow in the backward-looking option is also paid in v but based on the
composition of the daily rates ri compounded over the period. Today’s value of the

22

Electronic copy available at: https://ssrn.com/abstract=3308766

payment is given by
" n
!#
P c (0, u)
Y  
E X
(Nvc )−1 N (1 + δi ri ) − 1 c
= N P (0, v) −1
i=1
P c (0, v)
= N (P c (0, u) − P c (0, v))

The first equality is obtained from (Henrard, 2014, Theorem 2.4). The reasoning
leading to that result is a little bit long and involves multiple nested conditional
expectation and for that reason is not reproduced here but we refer to the above
reference for the details.
We have proved that the two approaches lead to the same valuation before the
fixing. The choice between the two is not a value transfer question but an achiev-
ability and availability question.

5.4 Note on dates

From a dates perspective, the replacement of a IBOR-linked derivative payment over
a period by an OIS or compounded setting in arrears (plus a spread) is ambiguous.
Many dates are involved in LIBOR and the derivatives payments as described at
the start of Section 2. Some dates are related to the synthetic deposit underlying
the IBOR definition and some dates are related to the derivative schedule.
The schedules of vanilla swaps are based on periodic payment – e.g. quarterly –
on the floating leg. Due to the presence of non-good business days, all the period
are not exactly of the same length. In the quarterly case, one period may have a
three-month plus two-day length and the next a three-month minus two-day length.
The IBOR fixing periods are themselves for a fixed tenor, e.g. three-month, and
always use the modified following rule in case the end day is on a non-good business
day. The IBOR fixing consequently always refer to a deposit of the stated period or
longer. Comparing those two schedules together, it means that the swap schedules
do not always correspond to the underlying IBOR deposits schedules, even for plain
vanilla swaps.
An example of such dates in proposed below in a very simple case. It is a two-
month swap with two one-month periods and a GBP-LIBOR-1M index. We select
GBP for this example to avoid the extra complication of the effective day lag. The
swap starts on 18-Jul-2018 and its unadjusted payment dates are 18-Aug and 18-Sep.
As 18-Aug is a Saturday, it is adjusted to Monday 20-Aug; the final date 18-Sep is
a Tuesday and is not adjusted. The first IBOR synthetic deposit starts on 18-Aug
and ends on 20-Aug (also adjusted). The second IBOR synthetic deposit starts on
20-Aug and ends on 20-Sep. The swap’s second period is paid on 18-Sep even if the
synthetic underlying deposit only matures on 20-Sep.

23

Electronic copy available at: https://ssrn.com/abstract=3308766

 O O

j j
δIX (18-Jul) δIX (20-Aug)

18-Jul 18-Aug 20-Aug 18-Sep 20-Sep

1 1
IBOR Deposit 1 IBOR Deposit 2

When the IBOR benchmark is replaced by an OIS benchmark, which OIS dates
should be selected? The dates corresponding to the swap dates or to the IBOR
deposit dates? Selecting the IBOR dates is the only approach in line with the
consistency criteria. All the IBOR fixings on the same date are replaced by the
same overnight-linked rate. But such a consistency requirement would exclude the
compounded setting in arrear version of the fixing described in Section 3.3. When the
IBOR fixing and the swap dates of a vanilla swap do not coincide, the compounded
setting in arrear overnight rate will not be known yet on the date it has to be paid.
There will still be a period on which the number has to accrue.
The date problem will be even more obvious when FRA with FRA discount-
ing settlement mechanism or IBOR with fixing in-arrears need to be replaced by
overnight. The FRA are paying in advance of the accrual period and the fixing
in-arrear IBOR are fixing in-arrear of the accrual period, which means that both are
paid immediately after the fixing – with a spot lag of typically two days in EUR and
USD. None of the information about the overnight composition would be known, by
a margin of several months, when it has to be paid. A backward-looking overnight
fixing is not possible for those instruments. The term rate fallback problem is a term
rate problem, not an overnight problem, even if the fallback is overnight based. I
don’t see any solution to the problem without a term rate – a rate known fully
on the fixing date – in the answer. To some extend, even the option 1 with spot
overnight rate has to be assessed carefully. The overnight rate are known only at
the end of the date – EUR EONIA – or the next day – EUR-ESTER, USD-SOFR
and GBP-SONIA reformed – while the IBOR fixing are known at 11:00 am. Is there
some provisions in some contracts that require to know the fixing by 11:00 am? For
example the cap/floor are exercised – even if it is automatic exercise – at 11:00 am,
can the exercise mechanism be delayed to the end of the day or even the next day?
People advocating for the backward-looking approach do not provide a detailed
description of how such a solution would work. The proposal looks attractive at
a cursory glance but actually impossible when the actual details are worked out.
For example in the Risk.Net article Becker (2018), a LIBOR expert advocates the
method, but no details of what that would mean exactly in practice is provided.
The ISDA consultation does not provide the required details either. Those details

24

Electronic copy available at: https://ssrn.com/abstract=3308766

have requested in the answer to the consultation but have not been provided yet.
To achieve consistency between IBOR replacements with the same fixing date,
the fixing cannot be different depending of how the IBOR is used – standard period,
adjusted by business days, in-arrears, FRA, etc. This appears to exclude backward-
looking/compounded setting in arrear approach to fallback.

5.5 Contamination and value transfer

Is a IBOR contract still an IBOR contract if a fallback provision is included? Or
has it been contaminated by the existence of the fallback provision itself? When
the fallback provision is included, the contract could be viewed as a weighted IBOR
and OIS+spread contract with the weight of the OIS+spread increasing with the
probability of IBOR discontinuation. What sense should we give to the figures
computed? Should we try to extract the true IBOR out of the weighted average?
Once a new fallback provision is introduced and the new contracts refer to it, we
could see a divergence in price between legacy contract and new contracts. Which
type of contracts should be used to estimate the IBOR-OIS spreads?
In some of my blogs17 , I have indicated that I was surprised by ISDA consultation
result to use “historical mean/median approach” for the spread adjustment. This
fallback method will create some value transfer between the different sides of the
existing trades. To my opinion, that value transfer does not take place at the LIBOR
discontinuation announcement date but at the fallback methodology announcement,
which was pubic on 27 November 2018. Note that this potential value transfer on the
announcement date was not mentioned in the ISDA document on the consultation
which only indicates that it could not be “present-value neutral on the calibration
date”.
The full details of the exact spread computation methodology have not been
announced yet, but we can already check some ballpark figures. The following
paragraphs look at the GBP figures. The reason is that in USD there is no historical
data for SOFR (not the multi-year history required by the methodology) and no
liquid SOFR-OIS trading and in EUR, the new benchmark is not even published
yet.
The most liquid GBP spread is the LIBOR-3M/SONIA spread. The analysis
start from the hypothesis that the LIBOR will be discontinued on 1-Jan-2022. The
analysis could be adapted for other discontinuation dates. The historical spread for
LIBOR-3M v SONIA-compounded-in-arrears-over-3M (using historical data for the
benchmarks, all correct market conventions, holidays, dates, etc.) is computed. The
proposed lookback periods in the consultations were 5 and 10 years. We take 10
years to start. The 10-year period is from the discontinuation announcement. For
simplicity, we suppose that it will be the 1-Jul-2021, which leave a 6-month period
between announcement and actual discontinuation. We don’t know yet what will
17
The first one on that subject was: https://multi-curve-framework.blogspot.com/2018/
11/isda-published-preliminary-results-of.html

25

Electronic copy available at: https://ssrn.com/abstract=3308766

happen between now and Jul-2021, but we have already a good view of the mean as
we have almost 7 years of the 10 years. The mean we obtain for that 7-year period
is 17.40bps. We have to compare that figure to the current spread in long term basis
swaps. We have selected the 30Y tenor as an example. Just before the announce-
ment, the spread was at 22.5 bps. The graph of several basis spreads for 30Y tenors
are proposed in Figure 9 together with the methodology announcement date (verti-
cal red line). The spread before the announcement date was significantly above the
historical average. In the couple of days after the announcement, the market spread
moved lower. Since, the movement in the same direction has continued.

25

20
Spread (in bps)

15

10

5 LIBOR3M-SONIA
LIBOR6M-LIBOR3M
LIBOR6M-LIBOR1M
LIBOR1M-SONIA
0
8

8
-1

-1
ov

ec
N

Date

Figure 9: Historical value of long term basis spread.

The same analysis was done for other spreads. For the LIBOR-6M v LIBOR-3M,
the average historical spread (using the same method as above) was 15.40 bps. The
market 30Y basis spread was at 6.0 before the announcement is and increase in two
days by 1.8 bps. For the LIBOR-6M v LIBOR-1M, the average historical spread
was 25.00 bps. The market 30Y basis spread was 12.2 bps before the announcement
and was up by 1.7 bps in the two days after.
We don’t know yet the exact technical details that will be proposed for the spread
computation: exact length of the look-back period (5 or 10 years), look-back period
based on IBOR fixing date or maturity date (due to in-arrears), mean or median,
holidays (the overnight benchmarks may not have the same good business days than

26

Electronic copy available at: https://ssrn.com/abstract=3308766

the IBORs), rounding, cut-off days, etc. So there is still some range for the known
part of the spread and there is certainly also some uncertainty about the future
values of the fixings (between now and discontinuation announcement).We have ran
several scenarios for historical data. For the LIBOR-3M v SONIA, we obtained a
range of spreads from 9 to 18 bps; for the LIBOR-6M v LIBOR-3M, from 8 to 16
bps and for LIBOR-6M v LIBOR-1M, from 14 to 25 bps. All the scenarios are in
the same direction with respect to the current basis spread market for long tenors.
In all cases, just after the announcement, the market has moved in the direction
indicated by the above analysis. In all cases the market has moved to the be very
close to one of the bound of the ranges indicated above. In two cases out of three,
the basis spread is near its maximum over the last 3 years. In Figure 10 we have
reproduce the previous figure for three spreads, but this time adding the range of
potential mean/median.

25

20
Spread (in bps)

15

10

LIBOR3M-SONIA average range

LIBOR1M-SONIA average range
5 LIBOR6M-LIBOR1M average range
LIBOR3M-SONIA historical
LIBOR1M-SONIA historical
LIBOR6M-LIBOR1M historical
0
8

8
-1

-1
ov

ec
N

Date

Figure 10: Historical value of long term basis spread and range of historical average
of the spot spread.

5.6 Non vanilla products implied by the fallback

Most of the originally proposed options involved some timing or convexity adjust-
ment in the valuation mechanism. The exception is the Option 3: Compounded
Setting in Arrears Rate – in the cases where it is achievable – and the Option X1:
Forward looking OIS rate described in Section 3.5.

27

Electronic copy available at: https://ssrn.com/abstract=3308766

 Those adjustment comes from the fact that the rates are not paid on their natural
period or with natural formulas. Those issues would transform all IBOR linked
products, even the most vanilla ones, into exotic derivatives. From a valuation
perspective, this is better avoided as it creates valuation complexity for instruments
that were previously plain vanilla and priced by “forwarding and discounting”.
The convexity adjustment is similar to an exotic contingent claim with the impact
depending significantly on the maturity. There is an upfront valuation impact from
the introduction of the timing issue. If this adjustment is not hedged with other
optional products, like cap/floor or other products with payment timing features,
the adjustment will be bled through time. Even if the adjustment is compensated at
the discontinuation date through the spread, if it is not hedge continuously through
time, the adjustment compensation can be lost (or increased) through time. This
type of feature would add volatility to the derivative portfolio P/L.
In the version 1.1 of this perspective, we analysed in more details the convexity
adjustments of the spot overnight rate and the compounding setting in advance.
As the consultation result has selected the compounding setting in arrears, in this
version, we will look only at adjustments related to that option.
One of the vanilla instruments that would require an adjustment with this option
are the caps/floors. In a vanilla IBOR caplet with strike K̄, the pay-off is known in
t0 and paid in v with a simple amount (for a cap and ignoring notional)

δ j (t0 )(I j (t0 ) − K̄)+ .

What is the amount in the case of the option compounding setting in arrears?
The amount is still paid in v and is given, for a spread S, by

n
! !+
1 Y
1 + δi0 I O (si−1 ) − 1 + S − K̄


j
δ (t0 ) i=1

What is the very important difference between an IBOR caplet and a new RFR
based caplet after fallback, beyond the change from one benchmark to another?
The striking difference (pun intended) is not the strike (which is spread adjusted)
but the exercise date. The actual rate with compounding in arrears is know only at
the end of the period (in arrears). The dynamic of the rates need to be measured
not to the original exercise date t0 but to the maturity date v (or more exactly
to the last overnight fixing before the maturity date). The rate is averaged (in a
special way, called compounding) over the accrual period. The dynamic of the rates
between the start accrual date and the end accrual date is still important and this
importance changes each day. The change of dynamic of the final rate will depend
of the number of day left in the period. The vanilla IBOR cap/floor are becoming
path-dependent Asian options using compounding as averaging method on rates.
Such products with cap/floor on composition have been analysed in the relatively
old paper Henrard (2007) in a context not related to the IBOR fallback. When

28

Electronic copy available at: https://ssrn.com/abstract=3308766

applied to our current problem, the paper analyse the pricing of the payments of
the type !
n−1
Y
1 + δi F O (si , si , si+1 ) , 1 + δK − 1


max
i=0

with different models. To simplify the writing, the spread introduced in K and we
use δ = δ j (t0 ).
The results of the above paper specialised to our overnight composition cap in
the one-factor HJM model can be expressed as

Theorem 1 Let 0 = s−1 ≤ s0 < s1 < s2 < · · · < sn . In the a HJM one-factor
model, the price of an instrument paying in sn the maximum of a fixed amount
1 + δK and of a principal gross-up by the discrete compounding of interest rates
Qn−1
over the periods [si , si+1 ] fixed in si (i.e. i=0 P (si , si )/P (si , si+1 )) is given in 0 by

F0 = P (0, s0 )N (κ + σ) + (1 + δK)P (0, sn )N (−κ)

where
n−1 X
X n−1 Z min(si ,sj )
2
σ = (ν(τ, si+1 ) − ν(τ, si ))(ν(τ, sj+1 ) − ν(τ, sj ))dτ.
i=0 j=0 0

and    
1 P (0, t1 ) 1 2
κ= ln − σ .
σ (1 + δK)P (0, tn ) 2

The result takes into account the daily fixing through the accrual period, hence
the volatility formula σ that include a different term for each business day. The
formula can be related to the convexity adjustment for overnight compounded fu-
tures where the convexity adjustment has also one term for each day in the accrual
period. Those formulas can for example be found in Henrard (2018a).
The above formula is a version in a simple term-structure model. To be more
realistic we should include a multi-factor feature in the term structure and a volatility
smile.
This issue of Asian option created by the fallback procedure would not be present
if the OIS Benchmark option was selected.

5.7 Risk profile

Most of the arguments proposed in this section where originally presented in a series
of blogs titled LIBOR Fallback Transformers! 18
18
Available at https://murisq.blogspot.com/2018/10/libor-fallback-transformers.
html

29

Electronic copy available at: https://ssrn.com/abstract=3308766

 The first and obvious risk profile impact of the fallback is that the risk is changed
from a IBOR risk to a overnight risk. In the following examples we use a non-ATM
USD IRS 6Mx7Y to illustrate the impacts.
In Figure 11 we represent the impact of the fallback to OIS benchmark on the
bucketed PV01 (or delta) of the above swap. We suppose that the impact is immedi-
ate with all IBOR coupons transformed immediately into overnight-linked coupons.
As proved in Section 5.3, the fallback profile would be the same for OIS benchmark
or overnight setting in arrears options when looked at from before the first fixing.

Figure 11: Delta (bucketed PV01) for a IBOR swap and for the same swap after
fallback to OIS benchmark.

Note that the total PV01 is very similar and that the nodes tenors are also
similar. The difference is the transfer from the LIBOR-3M curve to the OIS curve.
It is likely that the discontinuation will not happen without some warning. Today
we can already look at the risk profile of a swap (or of a full book) under the
hypothesis that the discontinuation take place at the end of 2021. In that case,
the IBOR coupons up to 2021 stay IBOR coupons and the ones after that date are
becoming overnight-linked coupon. The PV01 risk profile for such an hypothesis is
described in Figure 12. We see the shift from IBOR to OIS between 3 and 4 years
from now. The sensitivity look like the one of a standard IBOR swap up to 2021
and a forward OIS from that date.
Next we look at the evolution through time of a swap exposure. We start with
the simplest portfolio, composed of a single swap on a single period. The date of
the analysis is 30-Aug-2018 as previously. The swap has a start date in 12 months
and a 3-month tenor on USD-LIBOR-3M. The notional is 100m.
We look at the risk through the glasses of PV01. We compute the market quotes
bucketed PV01 with respect to each tenors and then sum them by curve (OIS and
LIBOR3M). This gives us two numbers for each date. Those numbers have to be

30

Electronic copy available at: https://ssrn.com/abstract=3308766

Figure 12: Delta (bucketed PV01) for a IBOR swap viewed as of August 2018 under
the hypothesis of a discontinuation at the end of 2021.

taken with a pinch of salt as they are obtained by adding sensitivities to different
market realities (market quotes from different instruments with different conven-
tions). They are enough for the qualitative analysis we perform in the following
paragraph, but may not be perfect for all purposes.
First we look at the trade risk in absence of LIBOR discontinuation. The risk is
composed of the risk to the LIBOR fixing for roughly 2,500 USD/bp (100m/10,000/4)
and a very small discounting amount from the fact that the swap is not ATM. The
Y axis of the graph represent the PV01 in K USD/bp for the LIBOR and the OIS
curves. The X axis is the date on which the risk is computed. To avoid complicating
the picture, we have used the market rates as of end of August and computed the
implied forward curves for each day in the following year. The risks are computed
with those forward curves. If we had used the actual market curves for each day,
there would be on top of the changes described here some small ups and downs due
to market fluctuations. The graph of the profile is provided in Figure 13.
We now introduce the Announcement Date and the Discontinuation Date. We
suppose that the announcement is 30-Dec-2018 and the actual discontinuation is
28-Feb-2018. Those dates do not affect our previous risk graph but we reported the
dates for visualisation purposes.
Now we introduce a fallback option, starting with the OIS Benchmark option.
The reason to start with that option, even if this is not in the ISDA consultation,
is that this is the one the closest to the actual LIBOR in term of risk profile.
The big change happens on the announcement date. The only fixing in our
swap is after the discontinuation date, it is then replaced by a fixing to the OIS
benchmark (plus a spread). In term of risk, the OIS benchmark is on the Discount-
ing/Overnight/OIS curve. On that date, the risk jumps from the LIBOR curve

31

Electronic copy available at: https://ssrn.com/abstract=3308766

 2.5

Discontinuation Date
Announcement Date
PV01 (KUSD/bp)

1.5
Legacy LIBOR3M
Legacy DSC-ON
1

0.5

0
19

9
8

9
l-1
-1

r-1

-1
n-
ct

ct
Ju
Ap
Ja
O

O
Date

Figure 13: Time profile of the total PV01 with respect to the LIBOR-3M and OIS
curves for a single coupon.

(dashed light blue) to the OIS curve (dashed dark blue). Then nothing spectacular
happens to the risk up to the fixing date. On that date the risk decreases dramat-
ically when the rate is known, leaving only a residual small OIS risk (coming from
the difference between the fixing and the fixed rate of the trade) which disappears
completely at maturity three months later.
Note also that it is possible that the OIS fixing and LIBOR fixing dates will be
slightly different because of non-good business days. For example USD-LIBOR is
fixing according to the London calendar but SOFR according to the US Government
Securities calendar (and obviously this is not yet defined for the OIS Benchmark
finance fiction we use here). The graph of the profile is provided in Figure 14.
Once the profile of one fallback option is understood we can add the other three.
The LIBOR profile will be the same for all options. It goes from something before
the announcement date, and that something is the same for all options, to nothing.
We do not repeat that part to avoid overloading the graph.
The other options included are Spot Overnight, Compounding Setting in Advance
and Compounding Setting in Arrears. For the Spot Overnight, the fixing is also on
one date, so the profile is very similar around the fixing date to the OIS Benchmark.
The total PV01 risk between the announcement date and the fixing date is quite
similar to the previous one. As discussed in a previous episode, this is not true when
looking at the tenors/buckets level. For the Compounding Setting in Advance,
the risk start to decrease three months before the actual fixing date. The fixing
is obtained by compounding the rates over the three-month period preceding the
fixing. So each day that is passing a small piece of the rate is know and there is
no risk anymore on it. Each day the risk is decreasing slightly. Finally for the
Compounding Setting in Arrears, the risk is roughly constant up to the start of the

32

Electronic copy available at: https://ssrn.com/abstract=3308766

 2.5

Discontinuation Date
Announcement Date
PV01 (KUSD/bp)

1.5
Legacy LIBOR3M
Legacy DSC-ON
Fallback LIBOR3M
OIS Benchmark DSC-ON
1

0.5

0
19

9
8

9
l-1
-1

r-1

-1
n-
ct

ct
Ju
Ap
Ja
O

O
Date

Figure 14: Time profile of the total PV01 with respect to the LIBOR-3M and OIS
curve for a single coupon (legacy coupon and fallback coupon with OIS benchmark
option).

theoretical deposit underlying the fixing and slightly decrease up to the maturity
date of the same fixing. This is a translated version of the previous description. The
graph of the profiles are provided in Figure 15. Note that between the fixing date and
the end of the accrual date the risk profile is significantly different between the OIS
benchmark (and IBOR) and the compounding setting in arrears. The equivalence
between OIS benchmark and compounding in arrears proved in Section 5.3 is true
only to the fixing date.

We now change the underlying instrument to a two-year swap starting in three

months. The announcement date for this example is 30-Dec-2018 and the discontinu-
ation date is 30-Sep-2019. The swap has eight 3-month periods. The announcement
date is in the first period and the discontinuation date is in the fourth period.

The profile in the absence of discontinuation is a standard profile with a small

discounting risk and a LIBOR risk that steps down at each fixing date (yellow). The
graph of the profiles is provided in Figure 16.

When we introduce the fallback, on the announcement date, the LIBOR risk of
all the fixing after the discontinuation date (four of them) are transferred to the OIS
curve. The light blue dashed line drops on the announcement date by the equivalent
of 4 fixings risk and the OIS risk jumps in the opposite direction. The 3 fixings that
are between the announcement date and the discontinuation date are not affected
by the fallback, this is why they is still three quarterly drops on the LIBOR light
blue line.

33

Electronic copy available at: https://ssrn.com/abstract=3308766

 2.5

Discontinuation Date
Announcement Date
PV01 (KUSD/bp)

Legacy LIBOR3M
1.5 Legacy DSC-ON
Fallback LIBOR3M
OIS Benchmark DSC-ON
Spot DSC-ON
1 In Advance DSC-ON
In Arrears DSC-ON

0.5

0
19

9
8

9
l-1
-1

r-1

-1
n-
ct

ct
Ju
Ap
Ja
O

O
Date

Figure 15: Time profile of the total PV01 with respect to the LIBOR-3M and OIS
curve for a single coupon (legacy coupon and fallback coupon with all options).

5.8 Clearing and margin

The payment related to swaps (and FRA) cleared at a CCP are done according to
the rule book of each CCP. The CCP’s rule book incorporate definition similar to
the one used in the ISDA master agreement. But in the case where a benchmark is
discontinued, the CCP can determine the rate in its sole discretion.19 The situation
for the cleared swap in case of discontinuation and in absence of fallback introduction
is easier from a legal point of view but probably not better for the end users.
Some clearing houses have indicated their support for the reform of the current
benchmarks and the work on fallback reform. Even if this is the case, to my knowl-
edge, no CCP has explicitly indicated that they will implement the new fallback
procedure as decided by ISDA. A description of LCH’s position can be found on its
Circular No 3999.20 At the end of December 2018, CME has send a mail titled CME
Group Supports ISDAs LIBOR Fallback Provisions, in which it says CME reserves
the right to make necessary adjustments based on consultations with our clients 21 .
Also, if/when the CCPs adopt the new fallback provision, the rule will apply to
all trades: the new trades transacted after the rule change but also the old trades
transacted before. The rule will apply retroactively to all existing cleared trades.
So we can already guarantee a fork in the swap definitions between cleared and
uncleared. If the new ISDA rules are not adopted in a CCP rule book, there is an
19
For example the LCH rule 1.8.12 states [...] provided that where the rate for a Reset Date (i) is
unavailable (including where such rate ceases, or will cease, to be provided by its administrators),
[...] the Clearing House will determine an alternative rate at its sole discretion. .
20
Available at https://www.lch.com/membership/ltd-membership/ltd-member-updates/
lchs-position-respect-isdas-recommended-benchmark.
21
I have not found a similar text on CME website.

34

Electronic copy available at: https://ssrn.com/abstract=3308766

 20

18

16

Discontinuation Date
14 Announcement Date
PV01 (KUSD/bp)

Legacy LIBOR3M
12 Legacy DSC-ON
Fallback LIBOR3M
10 OIS Benchmark DSC-ON
Spot DSC-ON
8 In Advance DSC-ON
In Arrears DSC-ON
6

0
19

20
9

0
8

0
l-1

l-2
-1

r-1

-1

r-2

-2
n-

n-
ct

ct

ct
Ju

Ju
Ap

Ap
Ja

Ja
O

O
Date

Figure 16: Time profile of the total PV01 with respect to the LIBOR-3M and OIS
curve for a multi-periods swap (legacy coupon and fallback coupon with all options).

obvious fork for the new trades; if the new ISDA rules are adopted by a CCP, they
apply to legacy trades also and there is a fork for the legacy trades. Note that this
fork somehow already exists in the current situation as in case of discontinuation the
CCP determine an alternative rate at its sole discretion but up to now this situation
was very unlikely to occur.
It seems odd that there are three different streams looking at the fallback (ISDA,
CCPs and working groups) and they don’t have a common position on all the details.
As discussed in Section 3.5, the working groups in the main currencies are working on
term RFR rates, but those have already been excluded from the ISDA consultation.
On the other side, the CCPs appear in favour of a new fallback provision, but reserve
the right not to apply it. In the ISDA consultation, one CCP answered that the
composition setting in advance was ”incompatible with the contractual requirements
of Forward Rate Agreements (FRAs)” (as explained in Section 5.4 of this note) and
they have a ”potential mitigating arrangements to cater for these”. This could lead
to different term sheets for cleared and uncleared FRAs.

5.9 Litigation and disruption

A new paradigm – the possibility of the IBOR kings disappearing is certainly some-
thing new – is, without doubt, introducing the possibility of litigation and disruption
for existing agreements. The best way to settle those issues is to take each agree-
ment, put all the parties involved around a table and come to a agreement on how to
navigate this new paradigm. That requires a lot of preparation and time. Hopefully
this note is a small piece of the preparation. Time between now and 2020 or 2021
will not be too much to solve all the individual issues.

35

Electronic copy available at: https://ssrn.com/abstract=3308766

 The most efficient way to deal with changes of fallback and the fallbacks them-
selves is to avoid them. This would mean closing all the existing position with the
current weak fallback provision and open new positions with a fallback provision
the market participants are comfortable with or with more robust benchmarks. In
the extreme, this may mean a different fallback provision or different benchmarks
for each trade. The participants would close IBOR-linked derivatives well in ad-
vance of the discontinuation and open new positions linked to benchmarks with a
longer life expectancy. This note does not enter into discussion on how to put in
place the infrastructure to deal with those questions in the most efficient way. Some
mechanism, involving auctions, are described in Duffie (2018).
But it appears that to facilitate the transition some flexibility by regulators
would be welcome. This would include rewriting part of the mandatory clearing
and bilateral margin regulation to allow the transition and the removal of existing
risks to be done in simple ways. It may also be useful to postpone some of the EU
Benchmark Regulation deadlines.

5.10 Leg compression at CCPs?

Currently the compression between instruments at CCPs is done on a swap basis.

Two swaps need to have the same schedule to be compressed. With the potential
discontinuation of existing benchmark, the increased importance of new benchmarks
and fallback procedures, we can expect that some market participants will anticipate
those issues and enter into new trades to move from one benchmark to another.
Standard products to achieve this would be basis swaps. For example the first SOFR
trades cleared at LCH were basis swaps SOFR-EFFR. But if a clearing member has a
fixed for IBOR swap and trade a IBOR for SOFFR swap, even if the IBOR cash flows
cancel exactly, they will never be compressed on a swap basis. One way to achieve
an efficient compression would be to generalised the trading convention used in EUR
and trade only fixed to floating swaps and trade the basis swaps as two swaps. This
would be a way to achieve the simplification going forward but does not solve the
problem of the legacy books. The existing basis swaps have been traded as one swap,
and the couple of years to the expected discontinuation of some IBOR benchmarks
is too short to have the bulk of the trades to reach maturity before that. One way to
improve dramatically the ease of compression would be to perform the compression
on a leg basis instead of on a full swap basis. Up to some years ago, the compression
could take place only if both parties to the original trade were compressing at the
same time. Since, the CCPs have introduced the “un-legging” of the two sides of the
trade and coupon blending to simplify the compression procedure. Maybe a similar
procedure could be introduce on a leg basis. Each leg of a trade could be compressed
independently of the other. This could facilitate the compression of disappearing
benchmarks and basis swaps.

36

Electronic copy available at: https://ssrn.com/abstract=3308766

5.11 Fallback of fallback and multi-stage fallback
Most of the fallback discussion is about direct fallback when one benchmark is
discontinued and a previously selected one replaces it.
As the discontinuation of benchmark is an important question, one has also to
discuss the potential discontinuation of the benchmark used in the fallback. Would in
this case the fallback procedure cascades down. One benchmark is discontinued and
fallback in theory to another benchmark which is itself discontinued later. Would the
trade referencing the first benchmark be also included into the fallback procedure
of the second benchmark? What if the target fallback benchmark is discontinued
before the discontinuation of the original benchmark?
One could also imagine a multi-stage or conditional fallback. My personal pref-
erence for the adjusted RFR is for Option X1 (Section 3.5) but there is no good
OIS benchmark currently available. The fallback procedure could reference a OIS
benchmark to be created. If it is created before the IBOR discontinuation, it is used,
if not another option is used. Financial markets are used to waterfall, one could be
created for the fallback procedure.

6 Personal preference
Based on the above discussion, there is, to my personal opinion, no perfect way to
introduce a fallback procedure without value transfer, without loss of consistency
and without risk of manipulation.
Nevertheless, if an IBOR-overnight spread approach is used to introduce a new
fallback provision, my proposal on how to achieve it would have been the following.
The adjusted RFR should be based on forward-looking derivative benchmarks
and the adjustment spread should be based on forward spreads. This seems the only
way to generally mitigate the value transfer. Any figure used to replace IBOR should
keep the same structure in order to create a fallback only for the rate and not for
the trades term-sheets (date on which the final rate is known). The overnight-linked
figure should be term rate, i.e. based on forward-looking derivatives benchmark.
The backward-looking proposals appear to fail either the very important consistency
criterion or the existential question of achievability and in any case do not respect
the existing trades term-sheets. The compounded backward-looking amounts are
standard for overnight-linked markets, but cannot be used as a replacement of term
IBORs that are fixed in advance. The problem is not an overnight or a generic
interest rate problem, it is a specific term rate replacement problem. The use of
those forward-looking term rates would have only a minor impact on common risk
management of standard OIS and legacy trades resulting from fallback and extend
the transition from term-rate to overnight rate where appropriate to a longer period.
In the absence of already available forward-looking derivative benchmark, my
preference would go to a multi-stage fallback where the option of the forward-looking
derivative benchmark is the first step of the waterfall and only in its absence a second

37

Electronic copy available at: https://ssrn.com/abstract=3308766

best option is selected. The second best solution could be a future based solution
or a composition setting in advance with a one month tenor, in all cases a term rate
available at the original IBOR rate fixing time.
The results of the ISDA consultation, which designated the historical mean/median
approach as the adjustment spread, seems to have already created a value transfer
between market participants. That value transfer started on the day before the
consultation result announcement.
Previous document by ISDA insisted on the absence of value transfer. The
proposed spread approach leads to the existence of value transfer. To attenuate the
impact of transfer and the asymmetry of information between market participants,
the methodology for the announcement of the consultation results should have been
detailed in advance. For example the date of the announcement could have been
provided at least one week in advance, the detailed results made available to all at the
same time, the publication done on a Saturday to leave all participants the possibility
to analyse them before the opening of the market (announcement outside trading
hours would not have been possible as the different currencies involved implied that
in all cases at least one of the impacted market would have been opened).
To decrease the risk of manipulation and litigation, all the procedures related
to the fallback should be described with all the required details. At the very least
all the technical details related to the selected procedure – e.g. all the relevant
dates, formulas, and technical questions included in this note – should be part of
the fallback procedure. For the moment, those details have not been provided by
ISDA or its working groups and are still under discussion.
In a recent Risk column related to transition to new rates, Bakkar and Brigo
(2018) indicated that This should involve the quant community in an active role; I
hope they will be heard.

A Gaussian HJM one-factor model

Let Z u
ν(t, u) = σ(t, s)ds.
t
and σ satisfies a separability condition
σ(t, u) = g(t)h(u). (8)
In a separable Gaussian HJM one-factor, the dynamic of the pseudo-discount
factors is given by
P c (t, u) P c (s, u)
 
1 2
= exp −α(s, t, u)Xs,t,u − α (s, t, u) .
Ntc Nsc 2
with Xs,t,u a standard normally distributed random variable independent of Fs and
Z t
α(s, t, u)Xs,t = ν(τ, u)dWτ
s

38

Electronic copy available at: https://ssrn.com/abstract=3308766

References
Ametrano, F. and Bianchetti, M. (2009). Modeling Interest Rate, chapter Boot-
strapping the illiquidity: Multiple yield curves construction for market coherent
forward rates estimation. Risk Books, Incisive Media. 18
Ametrano, F. and Bianchetti, M. (2013). Everything You Always Wanted to know
About Multiple Interest Rate Curve Bootstrapping But Were Afraid to Ask.
SSRN Working paper 2219548, Banca IMI/Banca Intesa Sanpaolo. 18
Andersen, L. and Piterbarg, V. (2010). Interest Rate Modeling – Volume I: Foun-
dations and Vanilla Models. Atlantic Financial Press. 18
Bakkar, I. and Brigo, D. (2018). Model risk in the transition to risk-free rates.
Risk.net, published online 4 June 2018. 38
Becker, L. (2018). Libor expert: don’t rely on forward RFR rates for transition.
Risk.net, published online 29 May 2018. 24
Duffie, D. (2018). Compression auctions with an application to libor-sofr swap
conversion. Technical report, Graduate School of Business, Stanford University.
36
Financial Stability Board (2018). Interest rate benchmark reform - overnight risk-
free rates and term rates. Technical report, Financial Stability Board. 13
Henrard, M. (2004). Overnight Indexed Swaps and Floored Compounded Instrument
in HJM one-factor model. Ewp-fin 0402008, Economics Working Paper Archive.
12
Henrard, M. (2007). Skewed Libor Market Model and Gaussian HJM explicit ap-
proaches to rolled deposit options. The Journal of Risk, 9(4). Preprint available
at SSRN: http://ssrn.com/abstract=956849. 28
Henrard, M. (2014). Interest Rate Modelling in the Multi-curve Framework: Foun-
dations, Evolution and Implementation. Applied Quantitative Finance. Palgrave
Macmillan. ISBN: 978-1-137-37465-3. 4, 8, 18, 19, 23
Henrard, M. (2018a). Overnight futures: Convexity adjustment. Working Paper
Series 3134346, SSRN, Available at https://ssrn.com/abstract=3134346. 20, 29
Henrard, M. (2018b). A quant perspective on IBOR fallback proposals - ver-
sion 1.1. Market Infrastructure Analysis, muRisQ Advisory, Available at
https://ssrn.com/abstract=3226183. 1, 8
Henrard, M. (2018c). A short note on long-term repos. Market Infrastructure Anal-
ysis, muRisQ Advisory. Available at SSRN: https://ssrn.com/abstract=3258690.
12

39

Electronic copy available at: https://ssrn.com/abstract=3308766

Wood, D. (2018). Term versions of RFRs will work – FCA official. Risk. 13

40

Electronic copy available at: https://ssrn.com/abstract=3308766

placeholder

Electronic copy available at: https://ssrn.com/abstract=3308766

 muRisQ Advisory

Web: murisq.com
Email: [email protected]

We offer quantitative finance advisory services. Our blend of

cutting edge academic research and practitioner experience in
this specialised field provides to our clients a rapid access to the
foundations on which to build their added value.
muRisQ Advisory’s services include trainings and workshops, model
validation, product design and risk management strategies.

Electronic copy available at: https://ssrn.com/abstract=3308766