SOFR So Far:

Modeling the LIBOR Replacement

Fabio Mercurio

Bloomberg L.P., New York

Columbia Financial Engineering seminar

October 15, 2018

1 / 27
Introduction

On June 22, 2017, the Alternative Reference Rates Committee (ARRC)

identified a treasuries repo financing rate, which they called SOFR, as
the best replacement for LIBOR.

On July 27, 2017, Andrew Bailey, the head of the Financial Conduct
Authority, said that LIBOR is not sustainable because of a lack of
transactions providing data, and that it will be phased out in 2021.

On November 29, 2017, Mark Carney, the Bank of England (BOE)

Governor, said that the BOE has stepped up efforts to replace LIBOR
with SONIA by the end of 2021.

On July 12, 2018, Andrew Bailey announced that banks should look to
move off of LIBOR sooner than 2021. He also said that interest rate
derivatives do not need term rates, and that synthetic solutions created
to replace LIBOR were not viable.
2 / 27
Introduction
On each business day, starting April 2, 2018, the New York Fed has
been publishing the SOFR on the New York Fed website.
The SOFR is calculated as a volume-weighted median of
transaction-level tri-party repo data collected from the Bank of New
York Mellon as well as GCF Repo transaction data and data on bilateral
Treasury repo transactions cleared through FICC’s DVP service.

3 / 27
Introduction
On May 7, 2018, CME launched 1-month and 3-month SOFR futures
contracts.
The contract listings of the 1-month futures comprise the nearest 7
calendar months.

4 / 27
Introduction

The contract listings of the 3-month futures comprise the 20 March

quarterly months, which started with June 2018.

5 / 27
Introduction

On July 16, 2018, LCH cleared first SOFR-based derivatives.

The first trade was a SOFR-Fed-funds basis swap arranged by TP ICAP.
The trade was followed by a SOFR swap and another SOFR-Fed-funds
basis swap.
On July 26, 2018, Fannie Mae issued the market’s first-ever SOFR
FRNs.
In September, CME will begin clearing SOFR fixed/float and basis
swaps.
Caveat: LCH decided to use Fed-fund for PAI, CME will instead use
SOFR.
Different collateral rates imply valuations based on different discount
curves.
6 / 27
SOFR-linked product issuance
ISSUER DATE TYPE

Fannie Mae 07/26/2018 FRN

World Bank 08/14/2018 FRN

Credit Suisse 08/20/2018 CD

Barclays 08/24/2018 CP

MetLife 08/30/2018 FRN

MTA 09/20/2018 FRN

7 / 27
A minimal multi-curve model

Our purpose is to build a SOFR curve using quoted SOFR derivative

data.

To this end, we introduce a simple multi-curve model, which extends

that of Moreni and Pallavicini (2010), to simultaneously price SOFR
futures and Eurodollar futures.

We assume that:
OIS rates evolve according to the Hull-White one-factor (1990) model
The SOFR-OIS basis is deterministic
Forward LIBORs follow a shifted-lognormal LMM

We then assume OIS discounting, and that numeraires are defined by

the OIS curve.

SOFR discounting can be modeled as well.

8 / 27
A minimal multi-curve model
The instantaneous OIS short rate r(t) is assumed to follow, under the
risk-neutral measure Q, the Hull-White one-factor (1990) model
r(t) = x(t) + α(t)
where α is a deterministic function, and
dx(t) = −ax(t) dt + σ(t) dZ(t), x(0) = 0,
where a is a positive constant and σ is deterministic.
We model the instantaneous SOFR short rate s(t) by assuming that
s(t) = x(t) + β(t)
where β is deterministic and used to calibrate the time-0 SOFR curve.
This implies that
s(t) − r(t) = β(t) − α(t) =: γ(t)
9 / 27
A minimal multi-curve model
Let us denote the price at time t of the OIS zero-coupon bond with
maturity T by P(t, T), that is:
h RT i
P(t, T) = E e− t r(u) du |Ft

The SOFR zero-coupon bond at time t with maturity T is then defined

by:
h RT i
Ps (t, T) = E e− t s(u) du |Ft
h RT RT i
= E e− t r(u) du e− t γ(u) du |Ft
h RT i
T − t γ(u) du
= P(t, T) E e |Ft

Since γ is assumed to be deterministic, then

RT
Ps (t, T) = e− t γ(u) du
P(t, T)
10 / 27
A minimal multi-curve model
Assuming continuous compounding, we define the simply-compounded
SOFR forward rate Fjs (t) for the interval [Tj−1 , Tj ) by
 R Tj 
s Tj Tj−1 s(u) du
1 + τj Fj (t) = E e |Ft

which can be expressed as

 RT R Tj 
1 − t j r(u) du Tj−1 s(u) du
1+ τj Fjs (t) = E e e |Ft
P(t, Tj )
 RT R Tj 
1 − t j−1 r(u) du Tj−1 γ(u) du
= E e e |Ft
P(t, Tj )
P(t, Tj−1 ) Tj−1 TTj−1
 R j 
γ(u) du
= E e |Ft
P(t, Tj )
Since γ is deterministic, then
Ps (t, Tj−1 )
1 + τj Fjs (t) =
Ps (t, Tj )
11 / 27
A minimal multi-curve model

Forward LIBORs Lj , j = 1, . . . , n, are defined by:

Lj (t) = ETj L(Tj−1 , Tj )|Ft

 

where L(t, T) is the time-t spot LIBOR with maturity T.

Lj is assumed to evolve under the Tj -forward measure according to:

dLj (t) = σj (t)[Lj (t) + αj ] dWj (t)

where σj is deterministic and αj is constant, for all j.

We assume a one-factor model for simplicity, that is we set

dWi (t) dWj (t) = dt for all i, j.

Forwards Lj are shifted-lognormal also under Q. So, Eurodollar futures

can be priced in closed form, see Henrard (2014) and Mercurio (2018).
12 / 27
The model calibration
The LIBOR shift parameters αj can be set to be αj = 1/τj .
Alternatively, they can be calibrated to the corresponding caplet skews.
The LIBOR volatilities σj can be calibrated to caplets ATM volatilities
σjATM . Assuming constant σj , and lognormal σjATM , we have:
σjATM Lj (0)
σj ≈
Lj (0) + αj

The LIBOR-OIS correlation can be calibrated to Eurodollar futures, or

in a way to maximize smoothness of the corresponding LIBOR curve.
The OIS-SOFR volatility can be defined so that a given LIBOR-OIS
basis has minimal volatility, see Mercurio (2018).
Alternatively, the OIS-SOFR volatility can be fine tuned to maximize
smoothness of the corresponding SOFR curve.
13 / 27
The pricing of CME 1m-SOFR futures
We consider a 1m-SOFR futures contract with maturity T, and whose
delivery month is represented by [T − δ, T), where δ ≈ 1/12.
We approximate the arithmetic average of daily SOFR during the
delivery month by:
1 T
Z
s(u) du
δ T−δ
Assuming T − δ ≥ 0, the 1m SOFR futures rate f s (0; T − δ, T) is
calculated as follows:
Z T 
s 1
f (0; T − δ, T) = E s(u) du
δ T−δ
1 Ps (0, T − δ)
= ln + C1m (0; T − δ, T)
δ Ps (0, T)
where C1m (0; T − δ, T) is the 1m-SOFR futures convexity adjustment,
which is an analytic function of a and σ(t).
14 / 27
The pricing of CME 1m-SOFR futures

Equivalently, we can write:

f s (0; T − δ, T) = Rs (0; T − δ, T) + C1m (0; T − δ, T)

where Rs (0; T − δ, T) denotes the continuously-compounded SOFR

forward rates for the interval [T − δ, T).
In the case of a constant σ(t) ≡ σ, the convexity adjustment is
explicitly given by

σ2
 
1m 2 −aT aδ 1 −2aT 2aδ
C (0; T − δ, T) = δ + e (1 − e ) − e (1 − e )
2δa2 a 2a
σ2
= [3T 2 − 3Tδ + δ 2 ] + O(a)
6
Since δ ≈ 1/12, the maximum T ≈ 7/12 and σ typically below 1%,
then C1m (0; T − δ, T) is likely to be a fraction of a basis point.
15 / 27
The pricing of CME 3m-SOFR futures
We consider a 3m-SOFR futures contract with maturity Tj , and whose
reference quarter is represented by the interval [Tj−1 , Tj ).
We approximate the compounded daily SOFR interest rate during the
reference quarter by:
 R Tj 
1 s(u) du
e Tj−1 −1
τj

The 3m-SOFR futures rate fjs (0) is calculated as follows:

 R Tj 
s Tj−1 s(u) du
1 + τj fj (0) = E e

Assuming Tj−1 ≥ 0, we have:

Ps (0, Tj−1 ) Uj
1 + τj fjs (0) = e
Ps (0, Tj )
where Uj is an analytic function of a and σ(t).
16 / 27
The pricing of CME 3m-SOFR futures
The SOFR forward rate Fjs (0) can then be obtained from the quoted
futures rate fjs (0) as follows:
h1 i 1
Fjs (0) = + fjs (0) e−Uj −
τj τj
so, the 3m SOFR futures convexity adjustment is given by
h1 i
Cj3m (0) := fjs (0) − Fjs (0) = + fjs (0) 1 − e−Uj


τj

In the case of a constant σ(t) ≡ σ, Uj is explicitly given by:

σ 2  −a(Tj +Tj−1 )
− e−2aTj + e−a(Tj −Tj−1 ) + . . .

Uj = 3
e
2a
σ2  3 2 3

= 2Tj − 3Tj Tj−1 + Tj−1 + O(a)
6
1 1 σ2 2
Since τj >> fjs (0) and Uj is small, then Cj3m (0) ≈ τj Uj ≈ 2 Tj .
17 / 27
Stripping discount factors from futures
SOFR discount factors Ps (0, T) can be stripped from 1m and 3m
futures rates using the previous formulas.
A SOFR curve can then be extrapolated by assuming, for instance, a
deterministic basis between SOFR and OIS swap rates.

18 / 27
The valuation of a SOFR fixed-floating swap
Consider a swap where the floating leg pays at times Tj ,
j = a + 1, . . . , b, and where the fixed leg pays the fixed rate K on dates
0 , . . . , T 0 , with T 0 = T and T 0 = T .
Tc+1 d c a d b

The floating-leg payment at time Tj is approximately given by

R Tj
Tj−1 s(u) du
e −1

The value of this payment at time t ≤ Tj−1 is

 R Tj 
Tj Tj−1 s(u) du
P(t, Tj ) E e − 1|Ft = τj P(t, Tj )Fjs (t)

Then, the SOFR swap value to the fixed-rate payer, at time t ≤ Ta , is

b
X d
X
τj P(t, Tj )Fjs (t) −K τj0 P(t, Tj0 )
j=a+1 j=c+1

where τj0 0 , T 0 ).
denotes the year fraction for the fixed-leg interval [Tj−1 j 19 / 27
The valuation of a SOFR fixed-floating swap

The corresponding forward swap rate is then given by:

Pb s
j=a+1 τj P(t, Tj )Fj (t)
S(t) = Pd 0 0
j=c+1 τj P(t, Tj )

Rt
When Ta < t ≤ Ta+1 , Ta s(u) du is known, so formulas must be
modified accordingly.

Equivalent formulas can be derived under SOFR discounting.

CME is switching to SOFR discounting.

LCH stays with Fed funds.

Alternative SOFR fixed-floating swaps could be offered to please the

buy side.
20 / 27
The new valuation of a LIBOR fixed-floating swap
Consider a standard LIBOR-based swap where the floating leg pays at
times Tj , j = a + 1, . . . , b, and where the fixed leg pays the fixed rate K
on dates Tc+10 , . . . , T 0 . We set T 0 = T and T 0 = T .
d c a d b

The swap value to the fixed-rate payer at time t < Ta+1 is given by
b
X d
X
τj P(t, Tj )Lj (t) − K τj0 P(t, Tj0 )
j=a+1 j=c+1

where we set Lj (t) = L(Ta , Ta+1 ) if Ta ≤ t < Ta+1 .

This valuation relies on LIBOR being published at least until the last
LIBOR fixing date Tb−1 , so that forwards Lj (t) can be defined
accordingly.
However, soon enough this may no longer be the case, because LIBOR
is very likely to be discontinued before the end of 2021.
21 / 27
The new valuation of a LIBOR fixed-floating swap
If LIBOR is to be discontinued, then swaps like the above are standard
up to some payment time Tk (included), and from Tk (excluded) on they
become swaps written on a new interest rate index.
Assuming Tk > Ta , the valuation of the above swap must then be
modified as follows:
k
X b
X d
X
τj P(t, Tj )Lj (t) + τj P(t, Tj )L̂j (t) − K τj0 P(t, Tj0 )
j=a+1 j=k+1 j=c+1

where L̂j (t) denotes the forward at time t of the new LIBOR fallback
L̂(Tj−1 , Tj ), that is:
L̂j (t) = ETj L̂(Tj−1 , Tj )|Ft
 

The methodology for the new LIBOR fallback L̂(Tj−1 , Tj ) has not been
decided yet, but ISDA started a consultation.
22 / 27
The new valuation of a LIBOR fixed-floating swap

The LIBOR fallback will likely be defined as the sum of a SOFR-based

term rate R(Tj−1 , Tj ) and a LIBOR-SOFR basis spread S(T ∗ )
calculated at the time T ∗ < Tk when an official announcement of
LIBOR discontinuation will be given.
Therefore, we can write:

L̂j (t) = Rj (t) + ETj S(T ∗ )|Ft

 

where Rj (t) is the time-t forward of R(Tj−1 , Tj ).

Forwards L̂j (t) can be calculated using a multi-curve model where
SOFR and LIBOR (and possibly OIS) rates are jointly modeled:
This will allow us to calculate Rj (t), should the choice of term rate
R(Tj−1 , Tj ) generate a convexity adjustment for Rj (t).
It will also allow us to calculate expected values of S(T ∗ ), should it be
modeled as stochastic.
23 / 27
The valuation of a LIBOR-SOFR basis swap

A LIBOR-SOFR basis swap is a swap with two floating legs, one being
the floating leg of a LIBOR fixed-floating swap, the other being the
floating leg of a SOFR-based swap with the same maturity and payment
frequency.

Let us denote by Ta the start date of the swap, and by Tj ,

j = a + 1, . . . , b its payment dates.

Assuming the same day count convention for the two legs, the value at
time t of the basis swap to the LIBOR payer is:
b
X k
X b
X
τj P(t, Tj )Fjs (t) − τj P(t, Tj )Lj (t) − τj P(t, Tj )L̂j (t)
j=a+1 j=a+1 j=k+1

where, for simplicity, we also assume t ≤ Ta < Tk .

24 / 27
Conclusions
We have introduced a simple multi-curve model to price SOFR futures,
as well as SOFR swaps, with the purpose of building a SOFR curve.
We have also valued LIBOR based swaps under the new LIBOR
fallback, and basis swaps.
There are still many outstanding questions:
How will a risk-free term rate be calculated?
How will LIBOR fallbacks be defined?
Will there be LIBOR fallback bases?
Will SOFR-based derivatives be liquid enough?
Will there be a new LIBOR proxy?
Will there be a “zombie” LIBOR?
When will the market start to trade SOFR-based non-linear derivatives?
How to transition from a LIBOR-based contract to a SOFR-based one?
What about currencies other than USD, GBP, CHF, JPY and EUR?
... 25 / 27
Appendix A: the minimal basis volatility
We define the multiplicative LIBOR-OIS basis Bj as:
Lj (t) − Fj (t)
Bj (t) :=
1 + τj Fj (t)

The Qj -dynamics of Bj is:

 
1 Lj (t) + αj
dBj (t) = · · · dt + Bj (t) + σj (t) dWj (t)
τj Lj (t) + τ1j

− (B(t, Tj ) − B(t, Tj−1 ))σ(t) dZj (t)

Assuming a constant σ, minimizing the basis volatility of at time 0

yields:
ρσj (0) Lj (0) + αj
σ=
B(0, Tj ) − B(0, Tj−1 ) Lj (0) + τ1
j

26 / 27
Appendix B: the pricing of Eurodollar futures

The Eurodollar futures rate at time t for the same interval [Tj−1 , Tj ) is
defined by
fj (t) = E[L(Tj−1 , Tj )|Ft ]
and is associated with the Eurodollar-futures contract, with unit
notional, that pays out 1 − L(Tj−1 , Tj ) at time Tj−1 .

The futures convexity adjustment is defined by:

Cj (t) = fj (t) − Lj (t)

In our simple multi-curve model, Eurodollar-futures convexity

adjustments can be calculated exactly and in closed form:
  Z Tj−1  
Cj (0) = [Lj (0) + αj ] exp ρ σj (t)σ(t)B(t, Tj ) dt − 1
0

27 / 27