Professional Documents
Culture Documents
Multiple Curve Disocunt PDF
Multiple Curve Disocunt PDF
Abstract
After the credit and liquidity crisis started in summer 2007 the market has rec-
ognized that multiple yield curves are required for estimation of both discount and
FRA rates with different tenors (e.g. Overnight, Libor 3 months, etc.), consistently
with the large basis spreads and the wide diffusion of bilateral collateral agreements
and central counterparties for derivatives transactions observed on the market.
This paper recovers and extends our previous work [1] to the modern single-
currency multiple-curve bootstrapping of both discounting and FRA yield curves,
consistently with the funding of market instruments. The theoretical pricing frame-
work is introduced and modern pricing formulas for plain vanilla interest rate deriva-
tives, such as Deposits, Forward Rate Agreements (FRA), Futures, Swaps, OIS, and
Basis Swaps, are derived from scratch.
The concrete EUR market case is worked out, and many details are discussed
regarding the selection of market instruments, synthetic market quotes,, smooth
interpolation, effect of OIS discounting, possible negative rates, turn of year effect,
local vs non local delta sensitivities, performance and yield curve sanity checks.
The implementation of the proposed algorithms is available open source within
the QuantLib framework.
∗
Financial Engineering, Banca IMI, Piazzetta G. Dell’Amore 3, 20121, Milan, Italy,
ferdinando.ametrano(AT)bancaimi.com.
†
Market Risk Management, Banca Intesa Sanpaolo, Piazza G. Ferrari 10, 20121, Milan, Italy,
marco.bianchetti(AT)intesasanpaolo.com, corresponding author.
Keywords: crisis, crunch, liquidity risk, funding risk, credit risk, counterparty risk,
collateral, CSA, fixed income, interest rates, negative rates, Libor, Euribor, Eonia, yield
curve, FRA curve, discount curve, single-curve, multiple-curve, bootstrapping, no-arbitrage,
pricing, hedging, interest rate derivatives, Deposit, FRA, Futures, Swap, IRS, OIS, Basis
Swap, turn of year, spline, greeks, sensitivity, QuantLib.
Disclaimer: the views expressed here are those of the authors and do not represent the
opinions of their employers. They are not responsible for any use that may be made of
these contents.
3
1 Introduction
Pricing complex interest rate derivatives requires modeling the future dynamics of the
yield curve term structure. Hence, most of the literature assumes the existence of the
current yield curve as given, and its construction is often neglected, or even obscured,
as it is considered more a question of art than a science. Clearly, any yield curve term
structure modeling approach will fail to produce good/reasonable prices if the current
term structure is not correct.
Financial institutions, software houses and practitioners have developed their own
proprietary methodologies in order to extract the current yield curve term structure from
quoted prices of a finite number of market instruments (see e.g. [2, 3, 4, 5, 6, 7]). Broadly
speaking, there are two classes of yield curve construction algorithms: best-fit and exact-
fit.
Best-fit algorithms assume a functional form for the term structure and calibrate its
parameters using a selection of calibration instruments quoted on the market, such that
to minimize the repricing error. For instance, the European Central Bank publishes yield
curves on the basis of the Soderlind and Svensson model [8], which is an extension of
the Nelson-Siegel model (see e.g. refs. [9], [10] and [11]). Such approach is popular due
to the smoothness of the curve, calibration easiness, intuitive financial interpretation of
functional form parameters (level, slope, curvature) and correspondence with principal
component analysis. On the other side, the fit quality is typically not good enough for
trading purposes in liquid interest rate markets, where 0.25 basis points can make the
difference.
In practice “exact-fit” algorithms are often preferred: they fix the yield curve on a
time grid of N points (pillars) in order to exactly reprice N pre-selected market instru-
ments. The implementation of such algorithms is often incremental, extending the yield
curve step-by-step with the increasing maturity of the ordered instruments, in a so called
“bootstrap”approach. Intermediate yield curve values are obtained by interpolation on
the bootstrapping grid. Different interpolation algorithms are available, but little atten-
tion has been devoted in the literature to the fact that interpolation is often used already
during bootstrapping, not just after that, and that the interaction between bootstrapping
and interpolation can be subtle if not nasty (see e.g. [12], [13], [14]).
Whilst naive algorithms may fail to deal with market subtleties such as date conven-
tions, the intra-day fixing of the first floating payment of a Swap, the turn-of-year effect,
the Futures convexity adjustment, etc., even very sophisticated algorithms used in a naive
way may fail to estimate correct FRA rates in difficult market conditions, as those ob-
served since the summer of 2007 during the so-called subprime credit crunch crisis, or the
very low (even negative) rates regimes in late 2012. Namely using just one single yield
curve is not enough to account for discount and FRA rates of different tenor, such as 1,
3, 6, 12 months, because of the large Basis Swap spreads presently quoted on the market.
A further critical point triggered by the crisis is the coherent construction and usage
of the discounting curve. Since, by no-arbitrage, the present value of any future cash
flow must be unique, the same discounting curve must be consistently used during the
construction of all the FRA curves. The construction of the discounting curve must
be consistent with the funding of the bootstrapping instruments quoted on the market.
Since Futures are traded on regulated markets, and OTC instruments (Forward Rate
4
Agreements, Swaps, and Basis Swaps) are traded between counterparties under bilateral
collateral agreements, or with central counterparties, with daily margination at the over
night collateral rate, the natural bootstrapping instruments selected for the discounting
curve construction are the Overnight Indexed Swaps (OIS).
Finally, we stress that bootstrapping interest rate yield curves is nothing else than
a recursive application of pricing formulas associated with the bootstrapping financial
instruments within some pricing model. Thus a solid theoretical framework for pricing
interest rate derivatives under collateral is required.
The plan of the paper is as follows. In section 2 we start by reviewing the mar-
ket practice for interest rate yield curves construction and pricing and hedging interest
rate derivatives, both in the traditional (old style) single-curve version, and in the mod-
ern multiple-curve version triggered by the credit crunch crisis. In section 3 we fix the
notation and we briefly derive the theoretical framework for pricing derivatives including
funding and collateral. In section 4, that constitutes the central contribution of this work,
we describe in great detail the modern multiple-curve approach, in the single currency
case. In particular in its numerous subsections we discuss the general features of the
bootstrapping procedure, we review in detail the (EUR) market instruments available for
yield curves construction, and we deal with some issues crucial for bootstrapping, such
as interpolation, negative rates, turn of year effect, multiple delta hedging, performance
and yield curve checks. Finally, in section 5 we show an example of numerical results
for the Euribor1M, 3M, 6M and 12M FRA curves bootstrapping using the open source
implementation released within the QuantLib framework. The conclusions are collected
in section 6. A few final appendices reports mathematical details used in the main text.
1. Interbank credit/liquidity issues do not matter for pricing, Libors are a good proxy
for risk free rates, Basis Swap spreads are negligible (and neglected).
3. Select one finite set of the most convenient (i.e. liquid) vanilla interest rate instru-
ments traded in real time on the market, with increasing maturities. For instance,
5
a very common choice in the EUR market is a combination of short-term EUR
Deposit, medium-term Futures on Euribor3M and medium-long-term Swaps on Eu-
ribor6M.
4. Build one yield curve using the selected instruments plus a set of bootstrapping
rules (e.g. pillars, priorities, interpolation, etc.).
5. Compute on the same curve FRA rates, cash flows1 , discount factors, and work out
the prices by summing up the discounted cash flows.
6. Compute the delta sensitivity and hedge the resulting delta risk using the suggested
amounts (hedge ratios) of the same set of vanillas.
For instance, a 5.5Y maturity EUR floating Swap leg on Euribor1M (not directly quoted
on the market) is commonly priced using discount factors and FRA rates calculated on
the same Depo-Futures-Swap curve cited above. The corresponding delta sensitivity is
calculated by shocking one by one the curve pillars and the resulting delta risk is hedged
using the suggested amounts (hedge ratios) of 5Y and 6Y Euribor6M Swaps2 .
We stress that this is a single-currency-single-curve approach, in that a unique yield
curve is built and used to price and hedge any interest rate derivative on a given currency.
Thinking in terms of more fundamental variables, e.g. the short rate, this is equivalent
to assume that there exist a unique fundamental underlying short rate process able to
model and explain the whole term structure of interest rates of any tenor.
It is also a relative pricing approach, because both the price and the hedge of a
derivative are calculated relatively to a set of vanillas quoted on the market. We notice
also that the procedure is not strictly guaranteed to be arbitrage-free, because discount
factors and FRA rates obtained through interpolation are, in general, not necessarily
consistent with the no-arbitrage condition; in practice bid-ask spreads and transaction
costs virtually hide any arbitrage possibility.
Finally, we stress that the first key point in the procedure above is much more a
matter of art than of science, because there is not an unique financially sound choice of
bootstrapping instruments and, in principle, none is better than the others.
The pricing & hedging methodology described above can be extended, in principle,
to more complicated cases, in particular when a model of the underlying interest rate
evolution is used to calculate the future dynamics of the yield curve and the expected
cash flows. The volatility and (eventually) correlation dependence carried by the model
implies, in principle, the bootstrapping of a variance/covariance matrix (two or even three
dimensional) and hedging the corresponding sensitivities (vega and rho) using volatility
and correlation dependent vanilla market instruments. In practice just a small subset
of such quotations is available, and thus only some portions of the variance/covariance
matrix can be extracted from the market. In this paper we will focus only on the basic
matter of yield curves and leave out the volatility/correlation dimensions.
1
within the present context of interest rate derivatives we focus in particular on cash flows dependent
on a single FRA rate.
2
we refer here to the case of local yield curve bootstrapping methods, for which there are no sensitivity
delocalization effect (see sec. 4.5.
6
2.2 The Modern Multiple-Curve Approach
Unfortunately, the pre-crisis approach outlined above is no longer consistent, at least in
this simple formulation, with the present market approach.
First, it does not take into account the market information carried by the Basis Swap
spreads, now much larger than in the past and no longer negligible. Second, it does not
take into account that the interest rate market is segmented into sub-areas correspond-
ing to instruments with different underlying rate tenors, characterized, in principle, by
different dynamics (e.g. short rate processes). Thus, pricing and hedging an interest
rate derivative on a single yield curve mixing different underlying rate tenors can lead to
“dirty” results, incorporating the different dynamics, and eventually the inconsistencies,
of different market areas, making prices and hedge ratios less stable and more difficult
to interpret. On the other side, the more the vanillas and the derivative share the same
homogeneous underlying rate, the better should be the relative pricing and the hedging.
Third, by no-arbitrage, discounting must be unique: two identical future cash flows of
whatever origin must display the same present value; hence we need an unique discount-
ing curve. Fourth, there is no mention to the collateral agreements and funding rates
associated with the bootstrapping instruments.
The market practice has thus evolved to take into account the new market information
cited above, that translate into the additional requirement of homogeneity and funding.
The homogeneity requirement means that interest rate derivatives with a given underlying
rate tenor must be priced and hedged using vanilla interest rate market instruments
with the same underlying. The funding requirement means that the discount rate of
any cash flow generated by the derivative must be consistent, by no-arbitrage, with the
funding rate associated with that cash flow. In case of OTC derivatives traded between
counterparties subject to collateral agreements (CSA, Credit Support Annex to the ISDA
Master Agreement), the funding rate is the collateral rate, normally the over night rate
in case of daily margination (see sec. 3 for more details).
The post-crisis standard market practice for the construction of multiple interest rate
yield curves can be summarized in the following procedure.
1. Interbank credit/liquidity issues do matter for pricing, Libors are risky rates, Basis
Swap spreads are no longer negligible.
2. The collateral does matter for pricing, OIS discounting is adopted.
3. Decide the appropriate funding rates of the derivatives to be priced, then select the
corresponding market instruments and build one single discounting curve using the
classical single-curve bootstrapping technique.
4. Select multiple separated sets of vanilla interest rate instruments traded in real time
on the market with increasing maturities, each set homogeneous in the underlying
rate (typically with 1M, 3M, 6M, 12M tenors).
5. Build multiple separated FRA curves using the selected instruments plus their boot-
strapping rules and the unique discounting curve.
6. Compute the relevant FRA rates and the corresponding cash flows from the FRA
curve with the appropriate tenor.
7
7. Compute the relevant discount factors from the discounting curve with the appro-
priate funding characteristics.
9. Compute the delta sensitivity and hedge the resulting delta risk using the suggested
amounts (hedge ratios) of the corresponding set of vanillas.
For instance, the 5.5Y floating Swap leg cited in the previous section, in case of
standard CSA with daily margination and Eonia rate, must be priced using Euribor1M
FRA rates computed on an “pure” 1M FRA curve, bootstrapped on Euribor1M vanillas
only, plus discount factors computed on the discounting curve, bootstrapped on Eur
OIS. The corresponding delta sensitivity should be calculated by shocking one by one
the pillars of both yield curves, and the resulting delta risk hedged using the suggested
amounts (hedge ratios) of 5Y and 6Y Euribor1M Swaps plus the suggested amounts of
5Y and 6Y OIS.
The improved approach described above is more consistent with the present market
situation, but - there is no free lunch - it does demand much more additional efforts. First,
the discounting curve clearly plays a special and fundamental role, and must be built
with particular care (see e.g. the discussion in ref. [17]). Second, building multiple curves
requires multiple quotations: much more interest rate bootstrapping instruments must
be considered (Deposits, Futures, Swaps, Basis Swaps, FRA, etc.), which are available on
the market with different degrees of liquidity and can display transitory inconsistencies.
Third, non trivial interpolation algorithms are crucial to produce smooth FRA curves
(see e.g. refs. [13], [16]). Fourth, multiple bootstrapping instruments implies multiple
sensitivities, so hedging becomes more complicated. Last but not least, pricing libraries,
platforms, reports, etc. must be extended, configured, tested and released to manage
multiple and separated yield curves for forwarding and discounting, not a trivial task for
quants, developers and IT people.
roughly corresponding to the schedules of a fixed and floating leg of an Interest Rate
Swap (IRS). The schedules will collect all the relevant contract dates (period start/end
8
dates, fixing dates, accrual dates, cash flow dates, etc.) and year fractions known at the
beginning of the contract, written in or derived from the termsheet. We will denote the
fixed and floating year fractions as
τfix,j := τ (Sj−1 , Sj , dcfix )
τfloat,i := τ (Tx,i−1 , Tx,i , dcfix ), (2)
respectively, where dc is the day count convention, that will be dropped whenever un-
necessary. In particular, the floating leg year fractions τFloat,i will be consistent with the
corresponding floating Libor rate tenor x, e.g. if the rate is Libor6M, with tenor x = 6M
then the floating leg frequency is 2y −1 and the times Tx,i are six-month spaced.
The payoff of any derivative Π at each cash flow date T will be denoted by Π(T ; p),
where p is a generic set of parameters specifying the contract details, such as nominal
amount, schedules, etc. The price of the same derivative at any previous date t < T
will be denoted consistently by Π(t; p). Finally, we will consider only single currency
derivatives, dropping any currency index.
where rα (t) is the (short) funding interest rate, related to the cash amount Bα (t), fixed
at time t, spanning the time interval [t, t + dt], and exchanged at time t + dt.
In particular, we identify two sources of funding associated with interest rate deriva-
tives3 .
• The generic funding (or treasury) account, denoted with Bf , associated with the
standard (unsecured) money and bond market funding at rate rf , typically Libor
plus spread, operated by a trading desk through a treasury desk.
3
we omit here the secured repo funding.
9
• The collateral account, denoted with Bc , associated with a collateral agreement
at collateral rate rc , typically over night, operated by a trading desk through a
collateral desk.
Real CSA are regulated mostly under the Credit Support Annex (CSA) of the ISDA
Standard Master Agreement. They are characterized by various clauses and parameters,
such as margination frequency, margination rate, threshold, minimum transfer amount,
eligible collateral (cash, bonds, etc.), collateral currency, collateral asymmetry (e.g. one-
way collateral), etc. We refer to ISDA documentation [19] for the precise definition of the
terminology used here.
For pricing purposes it is useful to introduce an abstract ’perfect’ collateral agreement,
characterized as follows.
Definition 3.1 (Perfect CSA). We define “perfect CSA” an ideal collateral agreement
with the following characteristics:
• zero initial margin or initial deposit
• fully symmetric
• cash collateral
• zero threshold
• continuous margination
• instantaneous settlement
We stress that, in case of default of one of the counterparties, we assume that the
perfect collateral exactly matches the realized loss, and there are neither close-out amount
nor legal risk relative to the closing of the deal or availability of the collateral.
The best market proxy available to a perfect CSA is the new ISDA Standard Credit
Support Annex (SCSA) as described e.g. in [20], [21]. The main features of the SCSA
are daily margination frequency, flat over night margination rate, zero threshold, zero
minimum transfer amount, cash collateral in the same currency of the trade, collateral
asymmetry.
Collateral Agreements are very diffused on the OTC market, according with the annual
margin survey conducted by ISDA [22]. In particular, all the major financial institutions
operating in the interest rate money and derivative markets are covered by mutual col-
lateral agreements. As a consequence, we may safely assume that market quotations of
derivatives, such as FRA, Swaps, Basis Swaps, OIS, etc. reflect collateralized transactions.
10
3.3 Pricing Under Collateral
Pricing derivatives under collateral implies that we must take into account, other than the
cash flows generated by the derivative, also the cash flows generated by the margination
mechanism provided by the CSA. As a consequence, we must extend the basic no-arbitrage
framework described in standard textbooks (see e.g. [18, 14, 23, 24]) introducing the
collateral account Bc defined in the previous section 3.2, and the corresponding collateral
cash flows generated by the margination mechanism. This extension is discussed in a
number of papers, see e.g. refs. [25, 26, 27, 28, 29, 30], also in connection with counterparty
risk, see e.g. refs. [31, 32, 33, 34, 35]. We report here the basic result, assuming that under
perfect CSA the default of both counterparties is irrelevant, and there are no Credit/Debit
Valuation Adjustments (CVA/DVA) to the value of the trade.
Proposition 3.2 (Pricing under perfect collateral). Let Π be a derivative with maturity
T written on a single asset X following the Geometric Brownian Motion process
where t ∈ R+ , X ∈ R, µP : R+ × R → R, σ : R+ × R → R+ , W P ∈ R is a 1-dimensional
independent standard Brownian motion in the probability space (Ω, F, P ), and P is the
real, or objective, probability measure. Assuming perfect collateral as in def. 3.1, the
derivatives’ price Π(t, X) at time t < T obeys the PDE
where Qf is the probability measure associated with the funding account Bf such that
11
underlying asset S(t), and the amount Π(t, S), but the funding rates are different. This
term generates the discount and the drift terms in eqs. 10 and 12.
We conclude that the collateral rate is the correct discount rate in case of perfect
collateral.
holds, where QTf is the probability measure associated with the collateral zero coupon bond
Pc (t, T ).
Proof. The result is obtained by applying the usual change of numeraire technique (see
e.g. [18]), from Bc (t) under Qf to Pc (t, T ) under QTf .
Eq. 13 above is the general pricing formula for collateralized derivatives.
12
T0 Ti −1 K Ti Fixed leg
τL (Ti −1,Ti )
Floating leg
L (T ,T )
T0 t TiF−1 Ti −1 x i −1 i Ti
Figure 1: representation of a standard (or textbook) FRA contract. Cash flow date is at
Ti , the end of the interest rate period.
We conclude that the FRA rate is equal to the expectation Fx,i (t) defined in eq. 17 above,
that will be called FRA rate as well.
Proposition 3.4 (Properties of FRA rate). The FRA rate Fx,i (t) defined in eq. 17 above
has the following properties.
1. At fixing date Ti−1 , Fx,i (t) collapses onto the spot Libor rate Lx (Ti−1 , Ti ),
T
Qf i
Fx,i (Ti−1 ) = ETi−1 [Lx (Ti−1 , Ti )] = Lx (Ti−1 , Ti ). (19)
13
2. By definition, Fx,i (t) is a martingale under QTf i measure
T T
Qf i Qf i
Fx,i (t) := Et [Lx (Ti−1 , Ti )] = Et [Fx,i (Ti−1 )] . (20)
3. In the single-curve limit Fx,i (t) recovers the classical single-curve value Fi (t),
T
Q i Ti
Fx,i (t) := Et f [Lx (Ti−1 , Ti )] −→ EQ
t [L(Ti−1 , Ti )]
Ti
= EQ
t [F (Ti−1 ; Ti−1 , Ti )] = F (t; Ti−1 , Ti ) := Fi (t). (21)
We observe that FRA contracts are quoted on the market in terms of the FRA equi-
librium rates. Thus FRA rates can be directly read “on the screen” (see sec. 4.3.2). FRA
rates are also included into Futures, IRS and IRBS rates (see sec. 4.3.3, 4.3.4 and 4.3.6).
Hence, a FRA rate term structure can be stripped from market quotations. In conclusion,
the FRA rate is the basic building block of the modern theoretical interest rate frame-
work. We do not even need to talk about the classical “forward rates” anymore. The
pricing expressions for interest rate derivatives may be derived within this framework, as
discussed in sec. 4.2.
Given the Libor rate, we can define the corresponding risky zero coupon bond Px (t; T )
1
Px (Ti−1 ; Ti ) := , (22)
1 + Lx (Ti−1 , Ti )τL (Ti−1 , Ti )
as a (Libor) discount factor, where τL (Ti−1 , Ti ) denotes the year fraction associated with
the time interval [Ti−1 , Ti ] with some day count convention dcL . We stress that the zero
coupon bond Px (t; Ti ) is risky because it is associated with a risky Libor rate with tenor
x. We can think to this bond as issued by an average Libor bank, representing the banks
belonging to the Libor panel (see e.g. the discussion in [36]). We stress also that Px (t; T )
differs, in general, from the collateral zero coupon bond Pc (t; T ), associated with the
collateral account as in eqs. 14.
We may invert the definition of risky zero coupon bond in eq. 22 above,
1 1
Lx (Ti−1 , Ti ) = −1 , (23)
τL (Ti−1 , Ti ) Px (Ti−1 , Ti )
and generalize it to (risky) forward Libor rates using the classical expression
1 Px (t; Ti−1 )
Fx,i (t) := Fx (t; Ti−1 , Ti ) := −1 . (24)
τL (Ti−1 , Ti ) Px (t; Ti )
Clearly eq. 24 reduces to eq. 22 in the limit of spot rates, t → Ti−1 . We observe that
eq. 24 can be rewritten as a no-arbitrage relation between discount factors over the three
time intervals [t, Ti−1 ], [Ti−1 , Ti ], and [t, Ti ],
Px (t; Ti ) = Px (t; Ti−1 )Px (t; Ti−1 , Ti ),
1
Px (t; Ti−1 , Ti ) = , (25)
1 + Fx (t; Ti−1 , Ti )τL (Ti−1 , Ti )
as represented in fig. 2, but we stress that the relations 24 and 3.4 above are just a
recursive definition of risky zero coupon bonds Px (t; T ), not a consequence of no-arbitrage
between consecutive spot/forward rates in the classical sense.
14
t Ti −1 Ti
or FRA rates
CxF (t) = {T −→ Fx (t; T, T + x), T ≥ t} , (27)
where t is the reference date (e.g. today, or spot date). The index x will take values
corresponding to the underlying rate tenors, e.g. x = {ON, 1M, 3M, 6M, 12M }.
We also define continuously compounded zero coupon rates zx (t, T ) and simply com-
pounded instantaneous FRA rates fx (t, T ) such that
Z T
Px (t, T ) = exp [−zx (t, T )τz (t, T )] = exp − fx (t, u) du , (28)
t
where
τz (T1 , T2 ) := τ (T1 , T2 ; dcz ) (30)
and dcz is the day count convention specific for the zero rate. Eq. (28) or (29) allows to
define a zero rate curve Cxz and an instantaneous FRA rate curve Cxf , such that
where
1
zx (t, T ) = − log Px (t, T ), (33)
τz (t, T )
∂
fx (t, T ) = − log Px (t, T )
∂T
∂
= zx (t, T ) + τz (t, T ) zx (t, T ), (34)
∂T
15
respectively. Swap par rates could also be used to represent yield curves. We do not
use them here as they are not frequently used and would not provide additional benefit
anyway.
From the relationships above it is immediate to observe that:
• zx (t, T ) is the average of fx (t, u) over the time interval [t, T ],
Z T
1
zx (t, T ) = fx (t, u) du; (35)
τz (t, T ) t
16
4.2 Modern Bootstrapping of Multiple Yield Curves
An yield curve is a complex object bearing many properties resulting from multiple dif-
ferent choices. We collect here the complete set of features that concur to shape an yield
curve. Most of these features will be discussed in great detail in the following sections.
We refer in particular to the EUR market case.
Zero coupon rate conventions: since the discount curve is normally observed to be
exponentially decreasing, as expected when the interest rate compounding is made
so frequent to be practically continuous, the zero rates compounding rule is chosen
to be continuous, as in eq. (28). The associated year fraction dcz in eq. (30) must be
monotonically increasing with increasing time intervals (non increasing convention
would lead to spurious null FRA rates), and additive, such that
The day count convention satisfying the above conditions that is normally adopted
by the market is the dcz = actual/365(f ixed) [19], such that:
T2 − T1
τz (T1 , T2 ) := τ [T1 , T2 ; actual/365(f ixed)] = . (37)
365
Typology: parameter specifying the type of the yield curve, as defined in section 3.5,
i.e.
Reference date: parameter t0 specifying the reference date of the yield curve, such that
Px (t0 , t0 ) = 1. It can be, for instance, t0 = today date, or t0 = spot date (which in
the EUR market is two business days after today according to the chosen calendar)
or, in principle, any business day after today (see sec. 4.3).
Time grid: the predetermined vector of dates, also named pillars, or knots, for which
the yield curve bootstrapping procedure returns a value. It is defined by the set of
maturities associated with the selected bootstrapping instruments.
4
as discussed in sec. 3.4, we will dismiss the classical expression forward rate/curve and we will always
talk about FRA rates/curves.
17
Bootstrapping instruments: the set of market instruments selected as input for the
bootstrapping procedure (see sec. 4.3).
Currency: parameter specifying the reference currency of the yield curve, corresponding
to the currency of the bootstrapping instruments.
Calendar: parameter specifying the calendar used to determine holidays and business
days. In the EUR market the standard TARGET5 calendar is used.
Side: parameter specifying the bid, mid or ask price chosen for the market instruments,
if quoted.
Best fit vs exact fit: as discussed in the introduction, best fit and exact fit algorithms
can be used to bootstrap an yield curve. We will adopt an exact fit algorithm
because it ensures exact repricing of the input bootstrapping instruments.
Hence, the first step for multiple yield curve construction is a very careful selection of the
corresponding multiple sets of bootstrapping instruments. Different kinds of instruments
can be selected for bootstrapping an yield curve term structure, and whilst they roughly
cover different maturities, they may overlap in some sections. Therefore it may be impos-
sible to include all the available instruments, and the subset of the mostly non-overlapping
contracts is selected, with preference given to the more liquid ones (with tighter bid/ask
spreads), used for delta sensitivity representation and hedging purposes. The mispricing
5
Trans-european Automated Real-time Gross settlement Express Transfer.
18
level of the market instruments excluded from the bootstrapping must thus be monitored
as safety check (see sec. 4.11).
The set of bootstrapping instruments also defines the time grid used for bootstrapping.
Market practitioners usually consider bootstrapping time grids from today up to 30Y-60Y.
The first point in the time grid is the reference date t0 of the yield curve Cx (t0 ). While it
makes perfectly sense to consider the first point {t0 , Px (t0 , t0 ) = 1} for the discount curve
CxP , the corresponding choices for {t0 , zx (t0 , t0 )}, {t0 , Fx (t0 , t0 + x)} and {t0 , fx (t0 , t0 )}, for
the zero curve Cxz , FRA curve CxF , and the instantaneous FRA curve Cxf , respectively, are
less significant and to some extent arbitrary, being just limits for shrinking T → t0 , and as
such must be handled with care. The reference date for all the EUR market bootstrapping
instruments (except Over Night and Tomorrow Next Deposit contracts, see section 4.3.1
below) is t0 = spot date. Once the yield curve at t0 is available, the corresponding yield
curve at t0 = today date can be obtained using the discount between these two dates
implied by ON and TN Deposits.
In the following sections we examine these market instruments in detail. In order to fix
the data set once for all, we always refer consistently to the EUR market quotes observed
on the Reuters platform as of 11 Dec. 2012, close time (around 16.30 CET6 . Obviously
the discussion holds for other EUR market data sets and can be remapped to other major
currencies with small changes.
where the expectation on the r.h.s. has obviously no effect because the inner value is
already fixed at time t. In fig. 3 we show a representation of the Deposit contract.
We stress that the Deposit is not a collateralized contract. Hence, in eq. 40 we have not
used the pricing expression for collateralized financial instruments, given in proposition
3.3, but we have used a discount factor Px (t; Ti ) based on a rate tenor x consistent with
the Deposit rate tenor, as in eq. 23.
6
Central European Time, equal to Greenwich Mean Time (GMT) plus 1 hour
19
Lx (T0,Ti )
T0F T0 t Ti
Figure 3: representation of a Deposit contract. The timeline reports the schedule, with
fixing, start, maturity and evaluation dates, T0F , T0 , Ti , t, respectively.
Deposit contracts are quoted and exchanged on the interbank OTC market for various
currencies. The EUR market quotes at time t0 = today a standard strip of Deposits
based on Euribor rates, with fixing date T0F = t0 , start date T0 = spot date = t0 +
2 business days, and maturity dates T1 , ..., Tn from 1 day up to 1 year. Thus the schedule
F
of the Deposit strip is T = T0 , T0 , T1 , ..., Tn = {t0 , T0 , T1 , ..., Tn } and the Deposit rate
is, in our notation, RxDepo (T0F ; Ti ) = Lx (T0 , Ti ), as shown in fig. 3.
In fig. 4 we show an example of the Euro Deposit strip quoted as of t0 = 11 Dec. 2012.
We see that the quotation is expressed directly in terms of the Deposit rate RDepo (t0 ; Ti ),
instead of the contract value in eq. 40. In particular the first two Deposits in the strip
are “irregular”: the first Deposit, denoted with ON (Over-Night Deposit), starts today
(T0 = t0 ) and matures tomorrow (T1 = t0 + 1 business day); the second Deposit, denoted
with TN (Tomorrow-Next Deposit), starts tomorrow (T0 = t0 + 1 business day) and
matures 1 day after (T2 = t0 + 2 business days). The next Deposit, denoted with SN
(Spot-Next Deposit) is regular. The following Deposits are denoted according with their
maturity, e.g. SW (Spot-Week Deposit), 1M (1-Month Deposit). Thus the Deposit strip
covers with neither holes nor overlaps the two business days interval between today and
spot dates. The maturity date of Deposits shorter than one month obeys date rolling
convention following. For longer Deposits the convention is modified following, end of
month. The TARGET calendar is adopted. Notice that the ON, TN and SN Deposits
insist on over night rates, with one day tenor, while the other Deposits in the strip admit
underlying rates with increasing tenors, depending on their maturities. For example the
3M Deposit admits Euribor3M, the 6M Deposit admits Euribor6M, etc.
Market Deposits can be selected as bootstrapping instruments for the construction
of the short term structure section of the yield curves Cx . Hence each Deposit may be
selected, in principle, for the construction of one single different curve with that tenor.
The first pillar at time Ti of the FRA curve at spot date T0 , CxF (T0 ) (actually a spot
rate) is obtained directly using the i-th market Deposit rate RxDepo (t0 ; Ti ) = Lx (T0 , Ti )
associated with the i-th Deposit with maturity Ti and underlying rate tenor x = Ti − T0
months. Instead, the discount curve CxP (T0 ) pillar at Ti is obtained using the following
relation
1
Px (T0 , Ti ) = Depo
, T0 < Ti , (41)
1 + Rx (t0 ; Ti ) τL (T0 , Ti )
where the year fraction τL is computed using the actual/360 day count convention given
in eq. 38.
20
Quote Settlement Business Day End of Month
Instrument Underlying Start Date Maturity Date
(ask, %) rule Convention Convention
Depo ON 0.040 Euribor1D Tue 11 Dec 2012 Wed 12 Dec 2012 Today Following False
Depo TN 0.040 Euribor1D Wed 12 Dec 2012 Thu 13 Dec 2012 Tomorrow Following False
Depo SN 0.040 Euribor1D Thu 13 Dec 2012 Fri 14 Dec 2012 Spot Following False
Depo 1W 0.070 Euribor1W Thu 13 Dec 2012 Thu 20 Dec 2012 Spot Following False
Depo 2W 0.080 Euribor2W Thu 13 Dec 2012 Thu 27 Dec 2012 Spot Following False
Depo 3W 0.110 Euribor3W Thu 13 Dec 2012 Thu 03 Jan 2013 Spot Following False
Depo 1M 0.110 Euribor1M Thu 13 Dec 2012 Mon 14 Jan 2013 Spot Mod. Following True
Depo 2M 0.140 Euribor2M Thu 13 Dec 2012 Wed 13 Feb 2013 Spot Mod. Following True
Depo 3M 0.180 Euribor3M Thu 13 Dec 2012 Wed 13 Mar 2013 Spot Mod. Following True
Depo 4M 0.220 Euribor4M Thu 13 Dec 2012 Mon 15 Apr 2013 Spot Mod. Following True
Depo 5M 0.270 Euribor5M Thu 13 Dec 2012 Mon 13 May 2013 Spot Mod. Following True
Depo 6M 0.320 Euribor6M Thu 13 Dec 2012 Thu 13 Jun 2013 Spot Mod. Following True
Depo 7M 0.350 Euribor7M Thu 13 Dec 2012 Mon 15 Jul 2013 Spot Mod. Following True
Depo 8M 0.390 Euribor8M Thu 13 Dec 2012 Tue 13 Aug 2013 Spot Mod. Following True
Depo 9M 0.420 Euribor9M Thu 13 Dec 2012 Fri 13 Sep 2013 Spot Mod. Following True
Depo 10M 0.460 Euribor10M Thu 13 Dec 2012 Mon 14 Oct 2013 Spot Mod. Following True
Depo 11M 0.500 Euribor11M Thu 13 Dec 2012 Wed 13 Nov 2013 Spot Mod. Following True
Depo 12M 0.540 Euribor12M Thu 13 Dec 2012 Fri 13 Dec 2013 Spot Mod. Following True
Figure 4: EUR Deposit strip. Source: Reuters page KLIEM, as of 11 Dec. 2012.
T0 Ti −1 K Ti Fixed leg
τL (Ti −1,Ti )
Floating leg
T0 t TiF−1 Ti −1Lx (Ti −1,Ti ) Ti
Figure 5: representation of a market FRA contract. Cash flow are exchanged at time
Ti−1 , the beginning of the interest rate period (compare with standard FRA in fig. 1).
discussed the standard (or textbook) FRA, with payoff, price and equilibrium rate given
by eq. 15, 16, and 17, respectively. The actual FRA traded on the market, has a different
payoff, such that, at payment date Ti−1 (not Ti ), we have
[Lx (Ti−1 , Ti ) − K] τF (Ti−1 , Ti )
FRAMkt (Ti−1 ; T , Lx,i , K, ω) = N ω
1 + Lx (Ti−1 , Ti )τF (Ti−1 , Ti )
FRAStd (Ti ; T , Lx , K, ω)
= . (42)
1 + Lx (Ti−1 , Ti )τL (Ti−1 , Ti )
In fig. 5 we show a representation of the market FRA contract.
Since FRA are traded OTC between collateralized counterparties, we may apply the
pricing under collateral approach discussed in sec. 3.2. The price and the FRA equilibrium
rate, derived in appendix C.1, are given by eq. 125 and 126,
1 + Kτx (Ti−1 , Ti ) Cc,x FRA (t;T )
FRAMkt (t; T , K, ω) = N ωPc (t; Ti−1 ) 1 − e i−1
, (43)
1 + Fx,i (t)τx (Ti−1 , Ti )
FRA 1 n FRA
o
Rx,Mkt (t; T ) = [1 + Fx,i (t)τx (Ti−1 , Ti )] eCc,x (t;Ti−1 ) − 1 , (44)
τx (Ti−1 , Ti )
FRA
where Cc,x (t; Ti−1 ) is a convexity adjustment, depending on the particular model adopted
for the dynamics of the two rates Fd,i (t) and Fx,i (t). In ref [42] it is proved that, for typical
post credit crunch market situations, the actual size of the convexity adjustment results
21
Quote Quote Quote
Instrument Underlying Start Date Maturity Date
(bid, %) (ask, %) (mid, %)
FRA Tod3M 0.156 0.206 0.181 Euribor3M Thu 13 Dec 2012 Wed 13 Mar 2013
FRA Tom3M 0.154 0.204 0.179 Euribor3M Fri 14 Dec 2012 Thu 14 Mar 2013
FRA 1x4 0.140 0.190 0.165 Euribor3M Mon 14 Jan 2013 Mon 15 Apr 2013
FRA 2x5 0.116 0.166 0.141 Euribor3M Wed 13 Feb 2013 Mon 13 May 2013
FRA 3x6 0.104 0.154 0.129 Euribor3M Wed 13 Mar 2013 Thu 13 Jun 2013
FRA 4x7 0.101 0.151 0.126 Euribor3M Mon 15 Apr 2013 Mon 15 Jul 2013
FRA 5x8 0.099 0.149 0.124 Euribor3M Mon 13 May 2013 Tue 13 Aug 2013
FRA 6x9 0.096 0.146 0.121 Euribor3M Thu 13 Jun 2013 Fri 13 Sep 2013
FRA Tod6M 0.291 0.341 0.316 Euribor6M Thu 13 Dec 2012 Thu 13 Jun 2013
FRA Tom6M 0.287 0.337 0.312 Euribor6M Fri 14 Dec 2012 Fri 14 Jun 2013
FRA 1x7 0.268 0.318 0.293 Euribor6M Mon 14 Jan 2013 Mon 15 Jul 2013
FRA 2x8 0.247 0.297 0.272 Euribor6M Wed 13 Feb 2013 Tue 13 Aug 2013
FRA 3x9 0.235 0.285 0.260 Euribor6M Wed 13 Mar 2013 Fri 13 Sep 2013
FRA 4x10 0.231 0.281 0.256 Euribor6M Mon 15 Apr 2013 Tue 15 Oct 2013
FRA 5x11 0.227 0.277 0.252 Euribor6M Mon 13 May 2013 Wed 13 Nov 2013
FRA 6x12 0.223 0.273 0.248 Euribor6M Thu 13 Jun 2013 Fri 13 Dec 2013
FRA 7x13 0.229 0.279 0.254 Euribor6M Mon 15 Jul 2013 Wed 15 Jan 2014
FRA 8x14 0.236 0.286 0.261 Euribor6M Tue 13 Aug 2013 Thu 13 Feb 2014
FRA 9x15 0.242 0.292 0.267 Euribor6M Fri 13 Sep 2013 Thu 13 Mar 2014
FRA 10x16 0.254 0.304 0.279 Euribor6M Mon 14 Oct 2013 Mon 14 Apr 2014
FRA 11x17 0.266 0.316 0.291 Euribor6M Wed 13 Nov 2013 Tue 13 May 2014
FRA 12x18 0.278 0.328 0.303 Euribor6M Fri 13 Dec 2013 Fri 13 Jun 2014
FRA 13x19 0.293 0.343 0.318 Euribor6M Mon 13 Jan 2014 Mon 14 Jul 2014
FRA 14x20 0.310 0.360 0.335 Euribor6M Thu 13 Feb 2014 Wed 13 Aug 2014
FRA 15x21 0.327 0.377 0.352 Euribor6M Thu 13 Mar 2014 Mon 15 Sep 2014
FRA 16x22 0.346 0.396 0.371 Euribor6M Mon 14 Apr 2014 Tue 14 Oct 2014
FRA 17x23 0.364 0.414 0.389 Euribor6M Tue 13 May 2014 Thu 13 Nov 2014
FRA 18x24 0.384 0.434 0.409 Euribor6M Fri 13 Jun 2014 Mon 15 Dec 2014
FRA 12x24 0.482 0.532 0.507 Euribor12M Fri 13 Dec 2013 Mon 15 Dec 2014
FRA IMMF3 0.264 0.314 99.7110 Euribor6M Wed 16 Jan 2013 Tue 16 Jul 2013
FRA IMMG3 0.244 0.294 99.7310 Euribor6M Wed 20 Feb 2013 Tue 20 Aug 2013
FRA IMMH3 0.234 0.284 99.7410 Euribor6M Wed 20 Mar 2013 Fri 20 Sep 2013
FRA IMMJ3 0.230 0.280 99.7450 Euribor6M Wed 17 Apr 2013 Thu 17 Oct 2013
Figure 6: EUR FRA strips on Euribor3M, Euribor6M, and Euribor12M. Source: Reuters
page ICAPSHORT2, as of 11 Dec. 2012.
to be below 1 bp, even for long maturities. Hence, in any practical situation, we can
discard the convexity adjustment and use the classical pricing expressions
Notice that in the single-curve limit we recover the classical single-curve expressions
FRA contracts are quoted and exchanged on the interbank OTC market for various
currencies. For example the EUR market quotes at t0 = today three standard strips of
Euribor FRA, starting at spot date T0 = t0 + 2 business days with different forward start
and end dates, Ti−1 and Ti , and tenors x = 3M, 6M, 12M . The corresponding FRA is
denoted shortly with Ti−1 ×Ti FRA. The year fraction τL is computed using the actual/360
day count convention given in eq. 38. The underlying Euribor FRA rate fixes at time
22
F
Ti−1 , two business days before the forward start date Ti−1 , and shares the same tenor of
the FRA. The FRA is conventionally denominated payer or receiver with respect to the
fixed leg (we enter in a payer FRA if we receive the fixed rate interest). For example, the
3x6 FRA starts at T = 3 months and matures at T = 6 months, the underlying being
Euribor3M. In particular, the FRA end dates are set with the same convention used for
the end dates of Deposits, such that FRA dates do concatenate exactly: the 3x6 FRA
above matures at T = 6M where the following 6x9 FRA starts. For the single market
F
FRA with maturity Ti we will use the schedule notation Ti = t0 , Ti−1 , Ti−1 , Ti , with
F FRA
t0 < Ti−1 < Ti−1 < Ti , i = 1, ..., n, and Rx,Mkt (t0 ; Ti ) for the equilibrium FRA rate.
In fig. 6 we show an example of the four FRA strips on Euribor3M, Euribor6M,
and Euribor12M rates, quoted as of 11 Dec. 2012. As for Deposits, the FRA quotation
FRA
is expressed directly in terms of the equilibrium FRA rate Rx,Mkt (t0 , Ti ), such that the
quoted FRA has zero value, FRAMkt (t0 ; Ti ) = 0, i = 1, ..., n. We notice also that the 3M
strip includes also four IMM FRA, struck on the fixed IMM dates as Futures (see section
4.3.3 below). Furthermore, there are 2+2 “irregular” FRA at the beginning of the two
3M and 6M strips, respectively, bridging the two business days interval between today
and spot dates.
Market FRA quotations in fig. 6 provide direct empirical evidence that a single-curve
cannot be used to estimate FRA rates with different tenors. We can observe that, for
instance, the level of the market 1x4 FRA3M (spanning from 14th Jan to 15th Apr. 2013,
FRA
with τL (1x4) = 0.25278) was R3M,Mkt (t0 , 1x4) = 0.165%, the level of market 4x7 FRA3M
FRA
(spanning from 15th Apr. to 15th Jul., τL (4x7) = 0.25278) was R3M,Mkt (t0 , 4x7) = 1.580%.
If one would compound these two rates to obtain the level of the implied 1x7 FRA6M
(spanning from 14th Jan. to 15th Jul., τL (1x7) = 0.50556) would obtain, using eq. 3.4,
FRA
R6M,Implied (t0 , 1x7)
FRA
FRA
1 + R3M,Mkt (t0 , 1x4)τL (1x4) × 1 + R3M,Mkt (t0 , 4x7)τL (4x7) − 1
=
τL (1x7)
= 0.146%, (49)
mkt
while the market quote for the 1x7 FRA6M was F1x7 = 0.293%, 14.7 basis point larger.
As discussed in section 3.4, this is not surprising, because eq. 3.4 is not a no-arbitrage
relation fulfilled by market rates, but just a recursive definition of risky zero coupon bonds.
The 14.7 basis points are the price assigned by the market to the different liquidity/default
risk implicit in the two investment strategies, using a 1x7 FRA6M or using two 1x4 and
4x7 FRA3M. See [36] for further details.
Market FRA with tenor x can be selected as bootstrapping instruments for the con-
struction of the short term structure section of the yield curve Cx . The FRA curve CxF (T0 )
FRA
pillar at Ti is obtained directly using the i-th market FRA rate Rx,Mkt (t0 , Ti ). Instead, the
P
discount curve Cx (T0 ) pillar at Ti is obtained recursively, given the previous point Ti−1 ,
from eq. 46 and 24 as
Px (T0 , Ti−1 )
Px (T0 , Ti ) = FRA
, (50)
1+ Rx,Mkt (t0 , Ti )τL (Ti−1 , Ti )
where τL is given by eq. 38. Notice that FRA collapse to Depos in the limit Ti−1 → t0
FRA
lim Rx,Mkt (t; Ti ) = RxDepo (t0 , Ti ) , (51)
Ti−1 →T0
23
and eq. (50) reduces to eq. (41).
4.3.3 Futures
Interest rate Futures are the exchange-traded and margined contracts equivalent to the
over-the-counter (market) FRA. The Futures’ payoff at payment date Ti−1 (as for the
market FRA) is given, from the point of view of the counterparty paying the floating
rate, by7
Futures(Ti−1 ; T ) = N [1 − Lx (Ti−1 , Ti )] , (52)
where N is the nominal amount. This payoff is a classical example of “mixing apples and
oranges” because, clearly, the Libor rate in eq. 52 is not an adimensional number that can
be directly summed to 1, and an year fraction τL (Ti−1 , Ti ) is lost. Thus we must consider
this definition merely as a rule to compute the margin.
The Futures’ price and rate at time t < Ti−1 are given in app. C.2 as
Hence the pricing of Futures requires the computation of the convexity adjustment in
the Futures’ rate. The expression of the convexity adjustment depends on the particular
model adopted. Convexity adjustment arises because of the daily marking to market and
margination mechanism of Futures. An investor long a Futures contract will have a loss
when the Futures price increases (and the Futures rate decreases) but she will finance
such loss at lower rate. Viceversa, when the Futures price decreases, the profit will be
reinvested at higher rate. This means that the volatility of the FRA rates and their
correlation to the spot rates have to be accounted for, and a convexity adjustment is
needed to convert the rate RxFut (t; T ) implied in the Futures price to its corresponding
FRA rate Fx,i (t) (see e.g. ref. [43]). From a mathematical point of view, the trivial unit
discount factor implied by daily margination introduces a pricing measure mismatch that
generates a volatility-correlation dependent convexity adjustment (see e.g. ch. 12 in ref.
[18]).
Different approaches to Futures’ convexity adjustment can be found in the financial
literature. Hull-White model approach [44], Libor Market Model [43], [18], two-factors
short rate gaussian model [18], stochastic volatility model [45], one-factor HJM model
[46]. The most recent and advanced derivation is given by [42] under the multiple-curve
Libor Market Model, as discussed in app. C.2, eq. C.2. The most common practitioners’
recipe is that of [44], based on a simple short rate 1 factor Hull & White model [47].
This approach has been used in fig. 7 to calculate the adjustments, using the Hull-White
parameters values given in table 1.
While FRA are traded OTC and have the advantage of being more customizable,
Futures are highly standardized contracts traded on exchanges. In the EUR market the
most common contracts (so called IMM 8 Futures)
insist on Euribor3M. The contract’s
F e
schedule of the i-th Futures, Ti = Ti , Ti , Ti , includes, fixing, expiry and maturity dates
7
See for example the NYSE Euronext definition for Euribor Futures at https://globalderivatives.
nyx.com/en/contract/content/29045/contract-specification.
8
International Money Market of the Chicago Mercantile Exchange.
24
such that τ TiF , Tie = settlement lag (two business days), and τL (Tie , Ti ) = 3M . This
schedule is in common with the underlying FRA3M. Standard expiry dates are March,
June, September and December, called IMM dates. The IMM Futures fix and stop trading
the third Monday of the expiry month, and expiry the following Wednesday. Notice that
e
such date grid is not regular: in general, Ti 6= Ti+1 , so Futures’ dates do not concatenate
exactly. The traded amount is measured in lots, with unitary size of 1.000.000e. The
market trades also so called serial Futures, expiring in the upcoming months not covered
by the quarterly Futures. Since these contracts are traded on exchanges, any profit and
loss generated by a Futures position and market movements is regulated through daily
marking to market. The margination amount is computed, for quarterly Futures, as
N0
∆(t, t − 1; T )) = [Futures(t; T ) − Futures(t − 1; T )] , (55)
n
where N 0 is the number of lots, and n is the Futures’ tenor (in months, e.g. n = 4 for
quarterly futures).
Such standard characteristics reduce the credit risk and the transaction costs of Market
Futures, thus enhancing a very high liquidity. The first front contract is one of the most
liquid interest rate instruments, with longer expiry contracts having very good liquidity
up to the 8th-12th contract. Also the first serial contract is quite liquid, especially when
it expires before the front contract.
In fig. 7 we report the quoted Futures strip on Euribor3M rate up to 3 years maturity.
As we can see, Futures are quoted in terms of prices instead of rates, the relation being
HW parameter Value
Mean reversion 3%
Volatility 0.3526%
Instead, the discount curve CxP (t0 ) pillar at Ti is obtained recursively, given the previous
point Ti−1 , from eq. 46 and 24 as
Px (t0 , Ti−1 )
Px (t0 , Ti ) = . (58)
1+ − CxFut (t0 ; Tie )] τL (Tie , Ti )
[RxFut (t0 ; Ti )
25
Quote Quote Quote Convexity Underlying Start Underlying End
Instrument Underlying
(bid, %) (ask, %) (mid, %) adjustment Date Date
FUT 3MZ2 99.8200 99.8250 99.8225 0.0000% Euribor3M Wed 19 Dec 2012 Tue 19 Mar 2013
FUT 3MF3 99.8450 99.8500 99.8475 0.0000% Euribor3M Wed 16 Jan 2013 Tue 16 Apr 2013
FUT 3MG3 99.8450 99.8650 99.8550 0.0001% Euribor3M Wed 20 Feb 2013 Mon 20 May 2013
FUT 3MH3 99.8700 99.8750 99.8725 0.0001% Euribor3M Wed 20 Mar 2013 Thu 20 Jun 2013
FUT 3MM3 99.8750 99.8800 99.8775 0.0003% Euribor3M Wed 19 Jun 2013 Thu 19 Sep 2013
FUT 3MU3 99.8700 99.8750 99.8725 0.0006% Euribor3M Wed 18 Sep 2013 Wed 18 Dec 2013
FUT 3MZ3 99.8400 99.8450 99.8425 0.0009% Euribor3M Wed 18 Dec 2013 Tue 18 Mar 2014
FUT 3MH4 99.8000 99.8050 99.8025 0.0013% Euribor3M Wed 19 Mar 2014 Thu 19 Jun 2014
FUT 3MM4 99.7400 99.7450 99.7425 0.0018% Euribor3M Wed 18 Jun 2014 Thu 18 Sep 2014
FUT 3MU4 99.6850 99.6900 99.6875 0.0024% Euribor3M Wed 17 Sep 2014 Wed 17 Dec 2014
FUT 3MZ4 99.6150 99.6200 99.6175 0.0030% Euribor3M Wed 17 Dec 2014 Tue 17 Mar 2015
FUT 3MH5 99.5550 99.5600 99.5575 0.0036% Euribor3M Wed 18 Mar 2015 Thu 18 Jun 2015
FUT 3MM5 99.4750 99.4800 99.4775 0.0044% Euribor3M Wed 17 Jun 2015 Thu 17 Sep 2015
FUT 3MU5 99.3800 99.3850 99.3825 0.0051% Euribor3M Wed 16 Sep 2015 Wed 16 Dec 2015
FUT 3MZ5 99.2750 99.2800 99.2775 0.0060% Euribor3M Wed 16 Dec 2015 Wed 16 Mar 2016
FUT 3MH6 99.1650 99.1700 99.1675 0.0069% Euribor3M Wed 16 Mar 2016 Thu 16 Jun 2016
FUT 3MM6 99.0350 99.0400 99.0375 0.0079% Euribor3M Wed 15 Jun 2016 Thu 15 Sep 2016
FUT 3MU6 98.9050 98.9150 98.9100 0.0090% Euribor3M Wed 21 Sep 2016 Wed 21 Dec 2016
FUT 3MZ6 98.7700 98.7850 98.7775 0.0100% Euribor3M Wed 21 Dec 2016 Tue 21 Mar 2017
FUT 3MH7 98.6500 98.6650 98.6575 0.0111% Euribor3M Wed 15 Mar 2017 Thu 15 Jun 2017
FUT 3MM7 98.5200 98.5450 98.5325 0.0124% Euribor3M Wed 21 Jun 2017 Thu 21 Sep 2017
FUT 3MU7 98.4000 98.4250 98.4125 0.0136% Euribor3M Wed 20 Sep 2017 Wed 20 Dec 2017
Figure 7: EUR Futures on Euribor 3M. Both main cycle Futures (denoted by “H”, “M”,
“U”and “Z”, standing for March, June, September and December expiries, respectively)
and two serial Futures (denoted as F3, G3 for Jan. and Feb 2013 expiries) are displayed.
Source: Reuters page 0#FEI, as of 11 Dec. 2012. In column 5 are reported the corre-
sponding convexity adjustments, calculated as discussed in the text.
The expression above can be used to bootstrap the yield curve Cx at point Ti once point
Ti−1 is known.
We stress that Futures contracts have fixed, not rolling, expiration dates, gradually
shrinking to zero as today’s date approaches the expiration date. As such, Futures used
as bootstrapping instruments in the yield curve generate rolling pillars that periodically
jumps and overlap the fixed DepoDeposit and FRA pillars9 . Hence some priority rule must
be used in order to decide which instruments must be excluded from the bootstrapping
procedure.
26
S0 S j −1 τK ( S j −1, S j ) Sj Sm Fixed leg
K
Lx (Ti −2 ,Ti −1 ) Lx (Ti −1,Ti )
Floating leg
L τ T T
( i −2 , i −1 ) τ L (Ti − 1 , Ti )
t T0F T0 TiF−2 Ti −2 TiF−1Ti −1 TiF Ti Tn
Figure 8: representation of an IRS contract. In this case, the frequency of the floating leg,
indexed to Libor Lx with tenor x (bottom), is twice the frequency of the fixed leg (top),
indexed to the fixed rate K. The coupon frequency of the floating leg is also consistent
with the Libor tenor (e.g. Libor3M, quarterly coupon).
We stress that the classical telescopic property of IRS rates, such that
n
X
Pc (t; Ti )Fx,i (t)τL (Ti−1 , Ti ) ' Pc (t; T0 ) − Pc (t; Tn ), (64)
i=1
does not hold, even as an approximation, because in the modern multiple-curve world the
discount rate and the FRA rate belongs to two different yield curves. As a consequence,
each leg of an IRS starting at T0 = S0 = today at par (with null total value) does not
need to be worth par (when a fictitious exchange of notionals is introduced at maturity),
that is
n
X
Pc (T0 ; Ti )Fx,i (t)τL (Ti−1 , Ti ) + Pc (T0 , Tn )
i=1
m
X
= Pc (S0 ; Sj )τK (Sj−1 , Sj ) + Pc (S0 , Sm ) 6= 1. (65)
j=1
27
Quote Quote Quote
Instrument Underlying Start Date Maturity
(bid, %) (ask, %) (mid, %)
AB6E1Y 0.266 0.306 0.286 Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2013
AB6E15M 0.225 0.275 0.250 Euribor6M Thu 13 Dec 2012 Thu 13 Mar 2014
AB6E18M 0.268 0.318 0.293 Euribor6M Thu 13 Dec 2012 Fri 13 Jun 2014
AB6E21M 0.257 0.307 0.282 Euribor6M Thu 13 Dec 2012 Mon 15 Sep 2014
AB6E2Y 0.304 0.344 0.324 Euribor6M Thu 13 Dec 2012 Mon 15 Dec 2014
AB6E3Y 0.404 0.444 0.424 Euribor6M Thu 13 Dec 2012 Mon 14 Dec 2015
AB6E4Y 0.556 0.596 0.576 Euribor6M Thu 13 Dec 2012 Tue 13 Dec 2016
AB6E5Y 0.742 0.782 0.762 Euribor6M Thu 13 Dec 2012 Wed 13 Dec 2017
AB6E6Y 0.934 0.974 0.954 Euribor6M Thu 13 Dec 2012 Thu 13 Dec 2018
AB6E7Y 1.115 1.155 1.135 Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2019
AB6E8Y 1.283 1.323 1.303 Euribor6M Thu 13 Dec 2012 Mon 14 Dec 2020
AB6E9Y 1.432 1.472 1.452 Euribor6M Thu 13 Dec 2012 Mon 13 Dec 2021
AB6E10Y 1.564 1.604 1.584 Euribor6M Thu 13 Dec 2012 Tue 13 Dec 2022
AB6E11Y 1.683 1.723 1.703 Euribor6M Thu 13 Dec 2012 Wed 13 Dec 2023
AB6E12Y 1.789 1.829 1.809 Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2024
AB6E13Y 1.881 1.921 1.901 Euribor6M Thu 13 Dec 2012 Mon 15 Dec 2025
AB6E14Y 1.956 1.996 1.976 Euribor6M Thu 13 Dec 2012 Mon 14 Dec 2026
AB6E15Y 2.017 2.057 2.037 Euribor6M Thu 13 Dec 2012 Mon 13 Dec 2027
AB6E16Y 2.066 2.106 2.086 Euribor6M Thu 13 Dec 2012 Wed 13 Dec 2028
AB6E17Y 2.103 2.143 2.123 Euribor6M Thu 13 Dec 2012 Thu 13 Dec 2029
AB6E18Y 2.130 2.170 2.150 Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2030
AB6E19Y 2.151 2.191 2.171 Euribor6M Thu 13 Dec 2012 Mon 15 Dec 2031
AB6E20Y 2.167 2.207 2.187 Euribor6M Thu 13 Dec 2012 Mon 13 Dec 2032
AB6E21Y 2.180 2.220 2.200 Euribor6M Thu 13 Dec 2012 Tue 13 Dec 2033
AB6E22Y 2.191 2.231 2.211 Euribor6M Thu 13 Dec 2012 Wed 13 Dec 2034
AB6E23Y 2.200 2.240 2.220 Euribor6M Thu 13 Dec 2012 Thu 13 Dec 2035
AB6E24Y 2.208 2.248 2.228 Euribor6M Thu 13 Dec 2012 Mon 15 Dec 2036
AB6E25Y 2.214 2.254 2.234 Euribor6M Thu 13 Dec 2012 Mon 14 Dec 2037
AB6E26Y 2.219 2.259 2.239 Euribor6M Thu 13 Dec 2012 Mon 13 Dec 2038
AB6E27Y 2.223 2.263 2.243 Euribor6M Thu 13 Dec 2012 Tue 13 Dec 2039
AB6E28Y 2.227 2.267 2.247 Euribor6M Thu 13 Dec 2012 Thu 13 Dec 2040
AB6E29Y 2.231 2.271 2.251 Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2041
AB6E30Y 2.236 2.276 2.256 Euribor6M Thu 13 Dec 2012 Mon 15 Dec 2042
AB6E35Y 2.275 2.315 2.295 Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2047
AB6E40Y 2.328 2.368 2.348 Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2052
AB6E50Y 2.401 2.441 2.421 Euribor6M Thu 13 Dec 2012 Wed 13 Dec 2062
AB6E60Y 2.443 2.483 2.463 Euribor6M Thu 13 Dec 2012 Tue 13 Dec 2072
Figure 9: EUR IRS on Euribor6M. The codes “AB6En”in col. 1 label IRS receiving
yearly a fixed rate and paying semi-annually a floating rate on Euribor6M with maturity
in n months/years. Source: Reuters page ICAPEURO, as of 11 Dec. 2012.
However, this is not a problem, since the only requirement for quoted spot-starting IRS
is that their total initial value must be equal to zero10 .
The EUR market quotes at t0 = today standard strips of plain vanilla IRS, with start
date T0 = spot date = t0 + 2 business and maturities ranging from 1Y to 60Y , annual
fixed leg versus floating leg indexed to x-months Euribor rate exchanged with x-months
frequency. The day count convention for the quoted (fair) IRS rate RxIRS (t; T , S) is 30/360
(bond basis) [19]. Also IMM IRS are quoted. In figures 9, 10 and 11 we report the quoted
IRS strips on 6M, 3M and 1M Euribor rates, respectively.
Market IRS on x-tenor Euribor can be selected as bootstrapping instruments for the
construction of the medium-long term structure section of the yield curve Cx (T0 ). Setting
the market IRS schedule as Ti = {T0 , · · · , Ti }, Sj = {S0 , · · · , Sj }, with t = T0 = S0 , Ti =
10
both these notions are hard-coded into (classical) financial operators and thus very hard to die.
28
Quote Quote Quote
Instrument Underlying Start Date Maturity
(bid, %) (ask, %) (mid, %)
AB3E1Y 0.116 0.166 0.141 Euribor3M Thu 13 Dec 2012 Fri 13 Dec 2013
AB3E15M 0.119 0.169 0.144 Euribor3M Thu 13 Dec 2012 Thu 13 Mar 2014
AB3E18M 0.128 0.178 0.153 Euribor3M Thu 13 Dec 2012 Fri 13 Jun 2014
AB3E21M 0.143 0.193 0.168 Euribor3M Thu 13 Dec 2012 Mon 15 Sep 2014
AB3E2Y 0.161 0.211 0.186 Euribor3M Thu 13 Dec 2012 Mon 15 Dec 2014
AB3E3Y 0.260 0.310 0.285 Euribor3M Thu 13 Dec 2012 Mon 14 Dec 2015
AB3E4Y 0.412 0.462 0.437 Euribor3M Thu 13 Dec 2012 Tue 13 Dec 2016
AB3E5Y 0.598 0.648 0.623 Euribor3M Thu 13 Dec 2012 Wed 13 Dec 2017
AB3E6Y 0.792 0.842 0.817 Euribor3M Thu 13 Dec 2012 Thu 13 Dec 2018
AB3E7Y 0.975 1.025 1.000 Euribor3M Thu 13 Dec 2012 Fri 13 Dec 2019
AB3E8Y 1.146 1.196 1.171 Euribor3M Thu 13 Dec 2012 Mon 14 Dec 2020
AB3E9Y 1.299 1.349 1.324 Euribor3M Thu 13 Dec 2012 Mon 13 Dec 2021
AB3E10Y 1.434 1.484 1.459 Euribor3M Thu 13 Dec 2012 Tue 13 Dec 2022
AB3E11Y 1.557 1.607 1.582 Euribor3M Thu 13 Dec 2012 Wed 13 Dec 2023
AB3E12Y 1.667 1.717 1.692 Euribor3M Thu 13 Dec 2012 Fri 13 Dec 2024
AB3E15Y 1.908 1.958 1.933 Euribor3M Thu 13 Dec 2012 Mon 13 Dec 2027
AB3E20Y 2.074 2.124 2.099 Euribor3M Thu 13 Dec 2012 Mon 13 Dec 2032
AB3E25Y 2.131 2.181 2.156 Euribor3M Thu 13 Dec 2012 Mon 14 Dec 2037
AB3E30Y 2.161 2.211 2.186 Euribor3M Thu 13 Dec 2012 Mon 15 Dec 2042
AB3E40Y 2.263 2.313 2.288 Euribor3M Thu 13 Dec 2012 Fri 13 Dec 2052
AB3E50Y 2.342 2.392 2.367 Euribor3M Thu 13 Dec 2012 Wed 13 Dec 2062
AB3EZ2 0.113 0.163 0.138 Euribor3M Wed 19 Dec 2012 Thu 19 Dec 2013
AB3EH3 0.109 0.159 0.134 Euribor3M Wed 20 Mar 2013 Thu 20 Mar 2014
AB3EM3 0.126 0.176 0.151 Euribor3M Wed 19 Jun 2013 Thu 19 Jun 2014
AB3EU3 0.158 0.208 0.183 Euribor3M Wed 18 Sep 2013 Thu 18 Sep 2014
AB3EZ3 0.158 0.208 0.183 Euribor3M Wed 18 Dec 2013 Fri 18 Dec 2015
AB3EH4 0.183 0.233 0.208 Euribor3M Wed 19 Mar 2014 Mon 21 Mar 2016
AB3EZ4 0.258 0.308 0.283 Euribor3M Wed 17 Dec 2014 Mon 18 Dec 2017
Figure 10: EUR IRS on Euribor3M, receiving yearly a fixed rate and paying quarterly a
floating rate on Euribor3M. At the bottom IMM starting IRS are reported, denoted with
their corresponding IMM code (see sec. 4.3.3). Source: Reuters page ICAPSHORT2, as
of 11 Dec. 2012.
Figure 11: EUR IRS on Euribor1M, receiving yearly a fixed rate and paying monthly a
floating rate on Euribor1M. Source: Reuters page ICAPSHORT2, as of 11 Dec. 2012.
29
Sj , the FRA curve CxF (T0 ) and the discount curve CxP (T0 ) at pillar Ti = Sj are obtained,
shorting the notation in eqs. 62 and 63 above, as
Pi−1
RxIRS (T0 ; Ti )Ac (T0 ; Ti ) − α=1 Pc (t; Tα )Fx,α (t)τL (Tα−1 , Tα )
Fx,i (T0 ) = , (66)
Pc (T0 ; Ti )τL (Ti−1 , Ti )
Pc (T0 ; Ti )Px (T0 ; Ti−1 )
Px (T0 ; Ti ) = ,
RxIRS (T0 ; Ti )Ac (T0 ; Ti ) − i−1
P
α=1 P c (t; Tα )F x,α (t)τ L (Tα−1 , Tα ) + P c (T0 ; Ti )
(67)
j
X
Ac (T0 ; Sj ) = Pc (T0 ; Sβ )τK (Sβ−1 , Sβ ) = Ac (T0 ; Sj−1 ) + Pc (T0 ; Sj )τK (Sj−1 , Sj ). (68)
β=1
The expressions above can be used, in principle, to bootstrap the yield curve Cx at pillar
Ti recursively, once the curve at previous pillars is known. We stress that both formulas
66, 67 require inputs from the discounting curve Cc (T0 ).
In practice, since in the market IRS the fixed leg frequency is annual and the floating
leg frequency is given by the underlying Euribor rate tenor x, we have that [S1 , ..., Sj ] ⊆
[T1 , ..., Ti ] for any given fixed leg date Sj . Hence some points [Ti−2 , Ti−1 , · · ·] inside the last
fixed leg interval [Sj−1 , Sj ] in eq. 67 may be unknown, and one must resort to interpolation
and, in general, to numerical solution of eqs. 66, 67. For example the bootstrap of
IRS
Euribor6M curve C6M (T0 ) from 9Y to 10Y pillars using the quotation R6M (T0 ; 10Y ) =
3.488% in fig. 9 is given by
IRS IRS
R6M (10Y )Ac (10Y ) − R6M (9Y )Ac (9Y ) − Pc (9.5Y )τL (9Y, 9.5Y )
F6M,10Y (T0 ) = , (69)
Pc (10Y )τL (9.5Y, 10Y )
Pc (10Y )P6M (9.5Y )
P6M (10Y ) = IRS IRS
,
R6M (10Y )Ac (10Y ) − R6M (9Y )Ac (9Y ) − Pc (9.5Y )τL (9Y, 9.5Y ) + Pc (10Y )
(70)
where T = [T0 , 0.5Y, ..., 9.5Y, 10Y ], S = [S0 , 1Y, ..., 9Y, 10Y ] are the floating and fixed
leg schedules. Since P6M (T0 , 9.5Y ) in eq. 70 above is unknown, it must be interpolated
between P6M (T0 , 9Y ) (known) and P6M (T0 , 10Y ) (unknown). We thus see, as anticipated
in the introduction, that interpolation is already used during the bootstrapping procedure,
not only after that.
where N is the contract’s nominal, τK , τon are the year fractions with the fixed and floating
rate conventions, and Ron (Ti ; Ti ) is the coupon rate compounded from over night rates
30
S0 S j −1 τK ( S j −1, S j ) Sj Sm
Fixed leg
K
Ti ,0
Ti k − Ti k
, 1 , Ti n
, i
Floating leg
t T0 Ti −1 Ron (Ti k − ,Ti k )
, 1 , Ti Tn
τon (Ti k − ,Ti k )
, 1 ,
Figure 12: representation of an OIS contract. In this case, the frequency of the two legs is
the same. The frequency of the floating leg is not consistent with the frequency of the over
night rate. The floating leg (bottom) is indexed to the over night rate Ron (Ti,k−1 , Ti,k )
with daily fixings and complex schedule, as described in eq. 72.
where Ti denotes the sub-schedule for the coupon rate Ron (Ti ; Ti ), and Ron (Ti,k−1 , Ti,k )
are the single over night rate spanning the over night time intervals (Ti,k−1 , Ti,k ]. Notice
that the i−th coupon rate Ron (Ti ; Ti ) is completely fixed only at the fixing date Ti,ni −1 of
the last rate Ron (Ti,ni −1 , Ti,ni ). In fig. 12 we show a representation of the OIS contract.
The price and the equilibrium rate of the OIS are given in app. C.4, eqs. 149 and 152,
respectively, as
c
OIS
OIS(t; T , S, Ron , K, ω) = N ω Ron (t; T , S) − K Ac (t; S), (73)
Pn
OIS Pc (t; Ti )Ron (t; Ti )τon (Ti−1 , Ti ) Pc (t; T0 ) − Pc (t; Tn )
Ron (t; T , S) = i=1 = . (74)
Ac (t; S) Ac (t; S)
The last term on the r.h.s of equation 74 above assumes that, under perfect collateral,
the over night FRA rates Ron (Ti,k−1 , Ti,k ) may be replicated using collateral zero coupon
bonds Pc (t; T ). Hence, the OIS floating leg value is drastically simplified to the classical
single-curve expression Pc (t; T0 ) − Pc (t; Tn ) above, to be compared to the multiple-curve
expression for the IRS in eq. 62.
The EUR market quotes at t0 = today standard strips of plain vanilla OIS, with start
date T0 = spot date = t0 + 2 business and maturities ranging from 1W to 60Y , annual
fixed leg versus annual floating leg indexed to Eonia rate. The day count convention is
the same for IRS. Also Eonia FRA on IMM and ECB dates are quoted. In figure 13 we
report the quoted OIS strips on Eonia as of 11 Dec. 2012. We observe very low and even
negative quotations for short term OIS, that will be discussed in sec. 4.6.
Market OIS can be selected as bootstrapping instruments for the construction of the
yield curve Cc (T0 ) used for discounting collateralised cash flows. In fact, as discussed in
31
Quote Quote Quote
Instrument Underlying Start Date Maturity
(bid, %) (ask, %) (mid, %)
EON1W 0.020 0.120 0.070 Eonia Thu 13 Dec 2012 Thu 20 Dec 2012
EON2W 0.019 0.119 0.069 Eonia Thu 13 Dec 2012 Thu 27 Dec 2012
EON3W 0.028 0.128 0.078 Eonia Thu 13 Dec 2012 Thu 03 Jan 2013
EON1M 0.049 0.099 0.074 Eonia Thu 13 Dec 2012 Mon 14 Jan 2013
EON2M 0.036 0.086 0.061 Eonia Thu 13 Dec 2012 Wed 13 Feb 2013
EON3M 0.022 0.072 0.047 Eonia Thu 13 Dec 2012 Wed 13 Mar 2013
EON4M 0.008 0.058 0.033 Eonia Thu 13 Dec 2012 Mon 15 Apr 2013
EON5M -0.001 0.049 0.024 Eonia Thu 13 Dec 2012 Mon 13 May 2013
EON6M -0.007 0.043 0.018 Eonia Thu 13 Dec 2012 Thu 13 Jun 2013
EON7M -0.012 0.038 0.013 Eonia Thu 13 Dec 2012 Mon 15 Jul 2013
EON8M -0.016 0.034 0.009 Eonia Thu 13 Dec 2012 Tue 13 Aug 2013
EON9M -0.020 0.030 0.005 Eonia Thu 13 Dec 2012 Fri 13 Sep 2013
EON10M -0.022 0.028 0.003 Eonia Thu 13 Dec 2012 Mon 14 Oct 2013
EON11M -0.024 0.026 0.001 Eonia Thu 13 Dec 2012 Wed 13 Nov 2013
EON1Y -0.035 0.035 0.000 Eonia Thu 13 Dec 2012 Fri 13 Dec 2013
EON15M -0.023 0.027 0.002 Eonia Thu 13 Dec 2012 Thu 13 Mar 2014
EON18M -0.017 0.033 0.008 Eonia Thu 13 Dec 2012 Fri 13 Jun 2014
EON21M -0.004 0.046 0.021 Eonia Thu 13 Dec 2012 Mon 15 Sep 2014
EON2Y 0.001 0.071 0.036 Eonia Thu 13 Dec 2012 Mon 15 Dec 2014
EON3Y 0.092 0.162 0.127 Eonia Thu 13 Dec 2012 Mon 14 Dec 2015
EON4Y 0.239 0.309 0.274 Eonia Thu 13 Dec 2012 Tue 13 Dec 2016
EON5Y 0.421 0.491 0.456 Eonia Thu 13 Dec 2012 Wed 13 Dec 2017
EON6Y 0.612 0.682 0.647 Eonia Thu 13 Dec 2012 Thu 13 Dec 2018
EON7Y 0.792 0.862 0.827 Eonia Thu 13 Dec 2012 Fri 13 Dec 2019
EON8Y 0.961 1.031 0.996 Eonia Thu 13 Dec 2012 Mon 14 Dec 2020
EON9Y 1.112 1.182 1.147 Eonia Thu 13 Dec 2012 Mon 13 Dec 2021
EON10Y 1.245 1.315 1.280 Eonia Thu 13 Dec 2012 Tue 13 Dec 2022
EON11Y 1.369 1.439 1.404 Eonia Thu 13 Dec 2012 Wed 13 Dec 2023
EON12Y 1.481 1.551 1.516 Eonia Thu 13 Dec 2012 Fri 13 Dec 2024
EON15Y 1.729 1.799 1.764 Eonia Thu 13 Dec 2012 Mon 13 Dec 2027
EON20Y 1.904 1.974 1.939 Eonia Thu 13 Dec 2012 Mon 13 Dec 2032
EON25Y 1.968 2.038 2.003 Eonia Thu 13 Dec 2012 Mon 14 Dec 2037
EON30Y 2.003 2.073 2.038 Eonia Thu 13 Dec 2012 Mon 15 Dec 2042
EONECBFEB13 0.021 0.071 0.046 Eonia Wed 16 Jan 2013 Wed 13 Feb 2013
EONECBMAR13 -0.009 0.041 0.016 Eonia Wed 13 Feb 2013 Wed 13 Mar 2013
EONECBAPR13 -0.032 0.018 -0.007 Eonia Wed 13 Mar 2013 Wed 10 Apr 2013
EONECBMAY13 -0.038 0.012 -0.013 Eonia Wed 10 Apr 2013 Wed 08 May 2013
EONECBJUN13 -0.039 0.011 -0.014 Eonia Wed 08 May 2013 Wed 12 Jun 2013
EONECBJUL13 -0.041 0.009 -0.016 Eonia Wed 12 Jun 2013 Wed 10 Jul 2013
Figure 13: EUR OIS, receiving yearly a fixed rate and paying yearly a floating rate on
Eonia. Bottom section: forward starting OIS at known ECB dates. Negative OIS rates
are enlightened in red color. Source: Reuters pages ICAPSHORT1 and ICAPEURO2, as
of 11 Dec. 2012.
32
secs. 3.2 and 3.3, the over night rate underlying the OIS is also the collateral rate for collat-
eralised market instruments. Setting the market OIS schedule as Ti = {T0 , · · · , Ti } = Si ,
the FRA curve CcF (T0 ) and the discount curve CcP (T0 ) at pillar Ti are obtained from eq.
74 above as
1
Fc (T0 ; Ti−1 , Ti ) =
τ (T ; T )
(on i−1 i )
OIS
Pc (T0 ; Ti−1 ) 1 + Ron (T0 ; Ti )τK (Ti−1 ; Ti )
× OIS (T ; T OIS
−1 ,
[Ron 0 i−1 ) − Ron (T0 ; Ti )] Ac (T0 ; Ti−1 ) + Pc (T0 ; Ti−1 )
(75)
OIS OIS
Ron (T0 ; Ti−1 ) − Ron (T0 ; Ti ) Ac (T0 ; Ti ) + Pc (T0 ; Ti−1 )
Pc (T0 ; Ti ) = OIS
. (76)
1 + Ron (T0 ; Ti )τK (Ti−1 ; Ti )
The expressions above may be simplified using the simpler single-curve expression for the
OIS rate on the r.h.s. of eq. 74,
( )
OIS
1 Pc (T0 ; Ti−1 ) 1 + Ron (T0 ; Ti )τK (Ti−1 ; Ti )
Fc (T0 ; Ti−1 , Ti ) = OIS (T ; T )A (T ; T )
− 1 , (77)
τon (Ti−1 ; Ti ) 1 − Ron 0 i c 0 i
OIS
1 − Ron (T0 ; Ti )Ac (T0 ; Ti )
Pc (T0 ; Ti ) = OIS
. (78)
1 + Ron (T0 ; Ti )τK (Ti−1 ; Ti )
IRBSletx,float (Tx,i ; Tx,i−1 , Tx,i , Lx ) = N Lx (Tx,i−1 , Tx,i )τL (Tx,i−1 , Tx,i ), i = 1, ..., nx ,
IRBSlety,float (Ty,j ; Ty,j−1 , Ty,j , Ly ) = N Ly (Ty,j−1 , Ty,j )τL (Ty,j−1 , Ty,j ), j = 1, ..., ny ,
IRSletfix (Sk ; Sk−1 , Sk , K) = N KτK (Sk−1 , Sk ), k = 1, ..., m, (80)
where N is the contract’s nominal, Lx (Tx,i−1 , Tx,i ) and Ly (Ty,j−1 , Ty,j ) are the Libors
with tenors x and y spanning the coupon periods (Tx,i−1 , Tx,i ] and (Ty,j−1 , Ty,j ],
respectively, and K is the common fixed rate of the two IRS.
33
Ty,0 Ty, j −1 τL (Ty j − ,Ty j )
, 1 , Ty, j Ty,m
Fixed leg y
Ly (Ty j − ,Ty j )
, 1 ,
Lx (Tx i − ,Tx i − )
, 2 , 1
Lx (Tx i − ,Tx i ) T
, 1 ,
i ,n i
Floating leg x
τL (Tx i − ,Tx i − )
Tx ,i −L1 x i − x i Tx ,i
, 2 , τ (T 1 ,T ) , 1 ,
t Tx ,0 Tx ,i −2 Tx ,n
Figure 14: representation of an IRBS contract, in the form of a single IRS with two
floating legs indexed to different Libors Lx , Ly with tenors x, y, respectively. In this case,
the frequency of the x leg is twice the frequency of the y leg, and they are consistent
with their corresponding Libor tenors. The IRBS as two IRS (not represented here) is a
portfolio of two IRS as in fig. 8, indexed to Libors Lx and Ly .
The price and the equilibrium basis spread of the IRBS as two IRS are given in app.
C.5, eqs. 153, 154, as
where IRSx,float (t; Tx , Lx ), IRSy,float (t; Ty , Ly ) are the values of the two IRS floating
legs as in eq. 139.
IRBSletx (Tx,i ; Tx,i−1 , Tx,i , Lx ) = N Lx (Tx,i−1 , Tx,i )τL (Tx,i−1 , Tx,i ), i = 1, ..., nx ,
IRBSlety (Ty,j ; Ty,j−1 , Ty,j , Ly , ∆y,ny )
= N Ly (Ty,j−1 , Ty,j ) + ∆y,ny τL (Ty,j−1 , Ty,j ), j = 1, ..., ny , (83)
where ∆y,ny in the second leg is a (constant) basis spread on Ly (Ty,j−1 , Ty,j ) for
maturity Tx,nx = Tx,ny . The price and the equilibrium basis spread of the IRBS as
single IRS are given in app. C.5, eqs. 155, 156,
34
In fig. 14 we show a representation of the IRBS contract. A particular case is the OIS-
Libor IRBS, where one floating leg is an OIS leg, indexed to compounded over night rate
as discussed in sec. 4.3.5.
In both eqs. 81, 85 above, the IRBS spread is positive when the fixings of the first
floating rate are higher, on average, than the fixings of the second floating rate on the
r.h.s.. The two definitions are financially equivalent, but the two IRBS spreads are slightly
different because of the different annuities involved (see eq. 157). The definition of IRBS
as two IRS is sometimes preferred because, in this case, the basis rate shares the same
rate conventions of the swap rates, while in the definition of IRBS as single IRS the basis
rate shares the same conventions of the FRA rate.
The EUR market quotes standard plain vanilla IRBS as two IRS11 with annual fixed
leg vs floating legs indexed to Euribors with different tenors, e.g. 3M vs 6M, 1M vs
6M, 6M vs 12M, etc. Also the 3M vs Eonia IRBS is quoted. In fig. 15 we report three
quoted IRBS strips. The quotation convention is to provide the difference (in basis points)
between the fixed rate of the higher frequency IRS and the fixed rate of the lower frequency
IRS. Normally such difference is positive, reflecting the tenor-dependent counterparty risk
included in the underlying rate (e.g. ON instead of 3M, 3M instead of 6M, etc.).
IRBS are a fundamental element for long term multiple-curve bootstrapping, because,
starting from the quoted IRS on Libor 6M (e.g. fig. 9), they allow to imply levels for
non-quoted IRS on different underlying Libor tenors (e.g. Euribor 1M, 3M, and 12M), to
be selected as bootstrapping instruments for the corresponding yield curves construction.
If, for example, ∆(t; Tx , T6M , S) is the quoted basis spread for an IRBS (as two IRS)
receiving Libor xM and paying Libor 6M for maturity Ti , we simply have
with the obvious caveat that ∆(t; Tx , T6M , S) = −∆(t; T6M , Tx , S). In fig. 16 we show all
the possible basis combinations obtained from figs. 15 and 13. Notice that IRBS in fig. 15
are quoted up to 30 years, while swaps on Euribor6M in fig. 9 are quoted up to 60 years.
Thus the bootstrapping of yield curves different from C6M over 30 years maturity requires
extrapolation of basis swap quotations. When the long term shape of the basis curves is
smooth and monotonic as in fig. 15, such extrapolation is not particularly critical.
35
Quote Quote Quote Underlying Underlying
Instrument Start Date Maturity Date
(bid, %) (ask, %) (mid, %) 1st leg 2nd leg
1E6E1Y 22.20 Euribor1M Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2013
1E6E2Y 22.60 Euribor1M Euribor6M Thu 13 Dec 2012 Mon 15 Dec 2014
1E6E3Y 23.80 Euribor1M Euribor6M Thu 13 Dec 2012 Mon 14 Dec 2015
1E6E4Y 24.60 Euribor1M Euribor6M Thu 13 Dec 2012 Tue 13 Dec 2016
1E6E5Y 25.00 Euribor1M Euribor6M Thu 13 Dec 2012 Wed 13 Dec 2017
1E6E6Y 25.00 Euribor1M Euribor6M Thu 13 Dec 2012 Thu 13 Dec 2018
1E6E7Y 24.80 Euribor1M Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2019
1E6E8Y 24.50 Euribor1M Euribor6M Thu 13 Dec 2012 Mon 14 Dec 2020
1E6E9Y 24.10 Euribor1M Euribor6M Thu 13 Dec 2012 Mon 13 Dec 2021
1E6E10Y 23.70 Euribor1M Euribor6M Thu 13 Dec 2012 Tue 13 Dec 2022
1E6E11Y 23.30 Euribor1M Euribor6M Thu 13 Dec 2012 Wed 13 Dec 2023
1E6E12Y 22.80 Euribor1M Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2024
1E6E15Y 21.10 Euribor1M Euribor6M Thu 13 Dec 2012 Mon 13 Dec 2027
1E6E20Y 18.90 Euribor1M Euribor6M Thu 13 Dec 2012 Mon 13 Dec 2032
1E6E25Y 17.50 Euribor1M Euribor6M Thu 13 Dec 2012 Mon 14 Dec 2037
1E6E30Y 16.30 Euribor1M Euribor6M Thu 13 Dec 2012 Mon 15 Dec 2042
3E6E1Y 14.00 15.00 14.50 Euribor3M Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2013
3E6E2Y 13.30 14.30 13.80 Euribor3M Euribor6M Thu 13 Dec 2012 Mon 15 Dec 2014
3E6E3Y 13.45 14.45 13.95 Euribor3M Euribor6M Thu 13 Dec 2012 Mon 14 Dec 2015
3E6E4Y 13.40 14.40 13.90 Euribor3M Euribor6M Thu 13 Dec 2012 Tue 13 Dec 2016
3E6E5Y 13.45 14.45 13.95 Euribor3M Euribor6M Thu 13 Dec 2012 Wed 13 Dec 2017
3E6E6Y 13.25 14.25 13.75 Euribor3M Euribor6M Thu 13 Dec 2012 Thu 13 Dec 2018
3E6E7Y 13.00 14.00 13.50 Euribor3M Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2019
3E6E8Y 12.70 13.70 13.20 Euribor3M Euribor6M Thu 13 Dec 2012 Mon 14 Dec 2020
3E6E9Y 12.35 13.35 12.85 Euribor3M Euribor6M Thu 13 Dec 2012 Mon 13 Dec 2021
3E6E10Y 12.00 13.00 12.50 Euribor3M Euribor6M Thu 13 Dec 2012 Tue 13 Dec 2022
3E6E11Y 11.65 12.65 12.15 Euribor3M Euribor6M Thu 13 Dec 2012 Wed 13 Dec 2023
3E6E12Y 11.20 12.20 11.70 Euribor3M Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2024
3E6E15Y 9.95 10.95 10.45 Euribor3M Euribor6M Thu 13 Dec 2012 Mon 13 Dec 2027
3E6E20Y 8.35 9.35 8.85 Euribor3M Euribor6M Thu 13 Dec 2012 Mon 13 Dec 2032
3E6E25Y 7.30 8.30 7.80 Euribor3M Euribor6M Thu 13 Dec 2012 Mon 14 Dec 2037
3E6E30Y 6.50 7.50 7.00 Euribor3M Euribor6M Thu 13 Dec 2012 Mon 15 Dec 2042
3E6E40Y 5.50 6.50 6.00 Euribor3M Euribor6M Thu 13 Dec 2012 Fri 13 Dec 2052
3E6E50Y 4.90 5.90 5.40 Euribor3M Euribor6M Thu 13 Dec 2012 Wed 13 Dec 2062
6E12E1Y 26.20 Euribor6M Euribor12M Thu 13 Dec 2012 Fri 13 Dec 2013
6E12E2Y 20.70 Euribor6M Euribor12M Thu 13 Dec 2012 Mon 15 Dec 2014
6E12E3Y 17.90 Euribor6M Euribor12M Thu 13 Dec 2012 Mon 14 Dec 2015
6E12E4Y 16.40 Euribor6M Euribor12M Thu 13 Dec 2012 Tue 13 Dec 2016
6E12E5Y 15.10 Euribor6M Euribor12M Thu 13 Dec 2012 Wed 13 Dec 2017
6E12E6Y 13.90 Euribor6M Euribor12M Thu 13 Dec 2012 Thu 13 Dec 2018
6E12E7Y 13.00 Euribor6M Euribor12M Thu 13 Dec 2012 Fri 13 Dec 2019
6E12E8Y 12.30 Euribor6M Euribor12M Thu 13 Dec 2012 Mon 14 Dec 2020
6E12E9Y 11.80 Euribor6M Euribor12M Thu 13 Dec 2012 Mon 13 Dec 2021
6E12E10Y 11.30 Euribor6M Euribor12M Thu 13 Dec 2012 Tue 13 Dec 2022
6E12E11Y 10.90 Euribor6M Euribor12M Thu 13 Dec 2012 Wed 13 Dec 2023
6E12E12Y 10.60 Euribor6M Euribor12M Thu 13 Dec 2012 Fri 13 Dec 2024
6E12E15Y 9.30 Euribor6M Euribor12M Thu 13 Dec 2012 Mon 13 Dec 2027
6E12E20Y 8.00 Euribor6M Euribor12M Thu 13 Dec 2012 Mon 13 Dec 2032
6E12E25Y 7.20 Euribor6M Euribor12M Thu 13 Dec 2012 Mon 14 Dec 2037
6E12E30Y 6.60 Euribor6M Euribor12M Thu 13 Dec 2012 Mon 15 Dec 2042
Figure 15: EUR IRBS. The codes “xEyEnY”in col. 1 label basis swaps receiving Euribor
xM and paying Euribor yM plus basis spread with n years maturity. Source: Reuters
page ICAPEUROBASIS, as of 11 Dec. 2012.
36
60 EUR Basis Swaps
ON vs 12M
ON vs 6M
ON vs 3M
50 ON vs 1M
1M vs 12M
3M vs 12M
1M vs 6M
40 6M vs 12M
Basis spread (bps)
3M vs 6M
1M vs 3M
30
20
10
0
Dec-12
Dec-15
Dec-18
Dec-21
Dec-24
Dec-27
Dec-30
Dec-33
Dec-36
Dec-39
Dec-42
Term
Figure 16: EUR Basis spreads from fig. 15. The spreads not explicitly quoted there
have been deduced using eq. (86) and the corresponding IRBS, IRS and OIS quotations.
Continuous lines: IRBS Eonia vs Euribor. dashed lines: IRBS Euribor vs Euribor.
37
secs. 5.2, 5.3, and 5.5, respectively. Clearly, extrapolation of market quotes is, in general,
much more subjective, and must be handled with care.
38
Quote OIS FRA-OIS
Instrument Tenor Start date Maturity Date Alpha Beta
(mid, %) (%) basis (%)
Depo 1M 1M Thu 13 Dec 2012 Mon 14 Jan 2013 0.1100 0.0740 0.0360
FRA 1x2 1M Mon 14 Jan 2013 Wed 13 Feb 2013 0.1017 0.0460 0.0557 0.0258 0.0003
AB1E2M 1M Thu 13 Dec 2012 Wed 13 Feb 2013 0.1060 0.0605 0.0455
FRA Tom3M 3M Fri 14 Dec 2012 Thu 14 Mar 2013 0.1790 0.0461 0.1329
Fut 3MZ2 3M Wed 19 Dec 2012 Tue 19 Mar 2013 0.1775 0.0415 0.1359 0.1148 0.0002
FRA Tom6M 6M Fri 14 Dec 2012 Fri 14 Jun 2013 0.3120 0.0170 0.2950
FRA 1x7 6M Mon 14 Jan 2013 Mon 15 Jul 2013 0.2930 0.0027 0.2903 0.3166 -0.0001
Depo 12M 12M Thu 13 Dec 2012 Fri 13 Dec 2013 0.5400 0.0042 0.5358
FRA 12x24 12M Fri 13 Dec 2013 Mon 15 Dec 2014 0.5070 0.0677 0.4393 0.5839 -0.0001
OIS 1M 3M 6M 12M
Instrument Tenor Start date Maturity Date
(%) (%) (%) (%) (%)
Synth. Depo 1D 1D Thu 13 Dec 2012 Fri 14 Dec 2012 0.0400 0.0261 0.1150 0.3165 0.5838
Synth. Depo 1W 1W Thu 13 Dec 2012 Thu 20 Dec 2012 0.0700 0.0280 0.1161 0.3158 0.5830
Synth. Depo 2W 2W Thu 13 Dec 2012 Thu 27 Dec 2012 0.0690 0.0303 0.1175 0.3150 0.5821
Synth. Depo 3W 3W Thu 13 Dec 2012 Thu 03 Jan 2013 0.0780 0.0325 0.1189 0.3142 0.5811
Synth. Depo 1M 1M Thu 13 Dec 2012 Mon 14 Jan 2013 0.0740 0.0360 0.1211 0.3129 0.5797
Synth. Depo 2M 2M Thu 13 Dec 2012 Wed 13 Feb 2013 0.0605 0.1270 0.3094 0.5757
Synth. Depo 3M 3M Thu 13 Dec 2012 Wed 13 Mar 2013 0.0466 0.1325 0.3061 0.5720
Synth. Depo 4M 4M Thu 13 Dec 2012 Mon 15 Apr 2013 0.0320 0.3022 0.5677
Synth. Depo 5M 5M Thu 13 Dec 2012 Mon 13 May 2013 0.0236 0.2989 0.5640
Synth. Depo 6M 6M Thu 13 Dec 2012 Thu 13 Jun 2013 0.0172 0.2953 0.5599
Synth. Depo 7M 7M Thu 13 Dec 2012 Mon 15 Jul 2013 0.0133 0.5557
Synth. Depo 8M 8M Thu 13 Dec 2012 Tue 13 Aug 2013 0.0107 0.5519
Synth. Depo 9M 9M Thu 13 Dec 2012 Fri 13 Sep 2013 0.0085 0.5478
Synth. Depo 10M 10M Thu 13 Dec 2012 Mon 14 Oct 2013 0.0067 0.5437
Synth. Depo 11M 11M Thu 13 Dec 2012 Wed 13 Nov 2013 0.0054 0.5398
Synth. Depo 12M 12M Thu 13 Dec 2012 Fri 13 Dec 2013 0.0042 0.5358
Figure 17: Construction of synthetic EUR Deposits on Euribor 1M, 3M, 6M, 12M. Top:
input instruments and quotes as of 11 Dec. 2012. Coefficients α and β are those of
eq. 90. Bottom: output synthetic quotes. Black: market instruments. Black: market
instruments. Green italics: synthetic instruments.
39
Quote
Instrument Value Date Maturity Date
(mid, %)
Synt Depo 1MD Euribor12M Thu 13 Dec 2012 Fri 13 Dec 2013 0.5400
FRA 12x24 Fri 13 Dec 2013 Mon 15 Dec 2014 0.5070
IRBS 3Y ON vs 12M Thu 13 Dec 2012 Mon 14 Dec 2015 0.6030
Figure 18: Construction of synthetic EUR FRA 12M. Top: input instruments and quotes
as of 11 Dec. 2012. Bottom: output interpolated FRA 12M. Black: market instruments.
Green italics: synthetic instruments.
40
4.5 Interpolation
The interpolation scheme adopted for bootstrapping a given parametrization is of paramount
importance to obtain reasonable yield curves (see e.g. refs. [12], [13], [16], [14]). For in-
stance, linear interpolation of discount factors is an obvious but extremely poor choice.
Linear interpolation of zero rates or log-discounts are popular choices, leading to stable
and fast bootstrapping procedures, but unfortunately they produce horrible FRA curves,
with a sagsaw or piecewise-constant shape.
We show in fig. 19 some examples of such poor interpolation schemes. Discount zero
rate curves (top panel) display similar and smooth behaviors, giving the false impression of
a good, interpolation independent, bootstrapping. Simple visual inspection of FRA curves
(bottom panel) may reveal non-smooth behaviors, with nasty oscillations up to tens of
basis points. Such discontinuities in the FRA curves correspond to angle points in the
zero curves, generated by linear interpolation that forces them to suddenly “turn around”
a market point. Notice that only the most liquid IRS from fig. 9, with maturities 3-10,
12, 15, 20, 25, 30, 35, 40, 50, 60 years have been included in the bootstrapping of curve
C6M . Thus the most irregular sections of the yield curve are those with less dense market
data: 13-14, 16-19, 21-24, 26-29, 31-34, 36-39, 41-49, 51-59 years. Often the remaining less
liquid quotations for maturity are included in the linear interpolations schemes to reduce
the amplitude of the FRA curve oscillations. If the market does not quote an annual
time grid, such as for IRS 3M, OIS, and IRBS shown in figs. 10, 13, 15, respectively, the
market quotes are often directly interpolated using some smooth algorithm, as discussed
in sec. 4.4.
In fig. 19 the monotonic cubic spline interpolation on log-discounts is shown too,
clearly ensuring a smooth and financially sound behavior of the FRA curve. This choice
is not straightforward, because spline interpolations must be handled with care. Simple
splines (see e.g. [49]) suffer of well-documented problems such as spurious inflection
points, excessive convexity, and lack of locality after input price perturbations (distributed
sensitivities). Recently, Andersen [16] has addressed these issues through the use of shape-
preserving splines from the class of generalized tension splines, while Hagan and West [12]-
[13] have developed a new scheme based on positive preserving forward interpolation. We
found the classic Hyman monotonic cubic filter [48] applied to spline interpolation of log-
discounts to be the easiest and best approach: its monotonicity ensures non-negative FRA
curves and actually removes most of the unpleasant waviness. Notice that the Hyman
filter can be applied to any cubic interpolants: this helps to address the non-locality of
spline using alternative more local cubic interpolations. In fig. 20 we show an example of
particularly nasty curve taken from Hagan and West [12] (p. 98 and fig. 2, bottom right
panel). The FRA curve obtained through Hyman monotonic cubic spline [48] applied on
log discounts (bottom panel) is always non negative (there is a unique minimum at 20Y),
contrary to the findings in ref. [12].
There are three penalties for using non-local interpolation schemes inside the boot-
strapping procedure.
1. During the bootstrapping, the shape of the part of the yield curve already boot-
strapped is altered by the addition of further pillars. This is usually remedied by
cycling in an iterative fashion: after a first bootstrap, which might even use a local
interpolation scheme and build up the pillar grid one point at time, the resulting
41
3.5% EUR yield curve 6M 1.0
3.0%
0.8
2.5%
Discount factor
Zero rate (%)
0.6
2.0%
1.5%
0.4
1.0%
0.2
0.5% Linear on zero rates
Linear on log discounts
Monotonic cubic spline on log discounts
0.0% Serie4 0.0
Dec 12
Dec 16
Dec 20
Dec 24
Dec 28
Dec 32
Dec 36
Dec 40
Dec 44
Dec 48
Dec 52
Dec 56
Dec 60
Term
3.0%
2.5%
FRA rate 6M (%)
2.0%
1.5%
1.0%
Dec 16
Dec 20
Dec 24
Dec 28
Dec 32
Dec 36
Dec 40
Dec 44
Dec 48
Dec 52
Dec 56
Dec 60
Term
Figure 19: Examples of bad (but very popular!) interpolation schemes. Top panel:
z P
discount and zero rate curves, C6M , C6M , obtained using different interpolation schemes,
display very similar and smooth behaviors, giving the false impression of a good, interpola-
F
tion independent, bootstrapping. Bottom panel: FRA curves C6M may reveal non-smooth
behaviors, with nasty oscillations up to tens of basis points. The most affected sections
of the yield curve are those with less dense market data. The monotonic cubic spline
interpolation on log-discounts (continuous black line) appears to be a better approach.
42
Zero Log Discrete
Capitalizat Discount Discount
Term rate ion factor factor FRA
factor forward
0.0 0.00% 1.000000 1.000000 0.000000
0.1 8.10% 1.008133 0.991933 0.008100 8.1000% 8.1329%
1.0 7.00% 1.072508 0.932394 0.070000 6.8778% 7.0951%
4.0 4.40% 1.192438 0.838618 0.176000 3.5333% 3.7274%
9.0 7.00% 1.877611 0.532592 0.630000 9.0800% 11.4920%
20.0 4.00% 2.225541 0.449329 0.800000 1.5455% 1.6846%
30.0 3.00% 2.459603 0.406570 0.900000 1.0000% 1.0517%
12%
10%
Forward rates (%)
8%
6%
4%
2%
0%
0 5 10 15 20 25 30
-2%
Term (Y)
Figure 20: Example of nasty curve taken from ref. [12] (p. 98 and fig. 2, bottom right
panel). Top panel: the example curve. Bottom panel: the FRA curve obtained through
Hyman monotonic cubic spline [48] applied on log-discounts. The FRA rate at 20Y is
null but no negative rates appear, contrary to the findings in ref. [12].
43
complete grid is altered one pillar at time using again the same bootstrapping al-
gorithm, until convergence is reached within a given precision. The first cycle can
be even replaced by a good grid guess, the most natural one being just the grid
previous state in a dynamically changing environment.
2. The non-local interpolation implies a non-local delta sensitivity, such that the shock
of, e.g., the 10Y market quote produces a delta sensitivity peaked at the 10Y pillar
but distributed onto the surrounding pillars (see sec. 4.9).
We stress that the focus on smooth discrete FRA rate is the key point of state-of-the-
art bootstrapping. For even the best interpolation schemes to be effective the FRA rate
curve must be smooth, i.e. any jump must be removed, and added back only at the end
of the smooth curve construction. The most relevant jump in FRA rates is the so-called
turn of year effect, discussed in sec 4.8.
44
0.20% EUR yield curve ON (0-2Y)
1.0002
0.15%
1.0000
FRA rate ON (%)
Discount factor
0.10% 0.9998
0.9996
0.05%
0.9994
Jun 14
Dec 12
Mar 13
Sep 13
Dec 13
Mar 14
Sep 14
Dec 14
Term
f
Figure 21: Short term FRA curve CON on Eonia as of 11 Dec. 2012 (from fig. 26,
top panel). Left scale: ON-tenor FRA rates FON (t0 ; t, t + 1d, act/360 ), t daily sampled.
The two outlier dots are the turn of the year jumps discussed in sec. 4.8. Right scale:
discount factors P1D (t0 ; t, act/360 ). Negative FRA rates up to −2 bps and increasing
discount factors are visible in the 3M-12M window. See comments in the text.
45
3 Multiple-curve vs single-curve bootstrapping
Euribor FRA curves
2
Difference (bps)
FRA Euribor1M
FRA Euribor3M
-1
FRA Euribor6M
FRA Euribor12M
-2
Dec-12
Dec-14
Dec-16
Dec-18
Dec-20
Dec-22
Dec-24
Dec-26
Dec-28
Dec-30
Dec-32
Dec-34
Dec-36
Dec-38
Dec-40
Term
2
Difference (bps)
0 Swap 3Y Euribor6M
Swap 5Y Euribor6M
Swap 10Y Euribor6M
-1
Dec-12
Dec-14
Dec-16
Dec-18
Dec-20
Dec-22
Dec-24
Dec-26
Dec-28
Dec-30
Dec-32
Dec-34
Dec-36
Dec-38
Dec-40
Term
Figure 22: Effect of exogenous bootstrapping. Top panel: difference (in basis points)
between exogenous and endogenous bootstrapping for FRA Euribor curves. Each FRA
rate has a tenor consistent with the underlying Euribor tenor. The spikes in the 1M and
3M curves are due to turn of the years jumps (see sec. 4.8) included in the exogenous
bootstrapping and excluded in the endogenous bootstrapping. Bottom panel: effect on
Swap curves.
46
order of the (non-negligible) differences observed in fig. 22. Nevertheless, such approach
is still used in old-style analytics libraries and softwares, sometimes unconsciously.
47
5.5 Euribor1M Time Series
5.0
4.5
Rate (%)
4.0
3.5
3.0
2.5
May 08
Nov 07
Mar 08
Apr 08
Aug 08
Nov 08
Oct 07
Dec 07
Jan 08
Feb 08
Jun 08
Jul 08
Sep 08
Oct 08
Dec 08
Jan 09
Date
Figure 23: Turn of year effect on Euribor1M. The historical time series in the Oct. 2007 -
Feb 2009 window is displayed. Three jumps can be identified: the 2007 turn of year (29th
Nov. 2007, 64 bps, left rectangle); the 2008 turn of year (27th Nov. 2008, 22 bps, right
oval); a smaller “end of semester effect”(29th May 2008, 9 bps, middle rhombus). Source:
Reuters.
48
the yield curve bootstrapping. An yield curve term structure up to N years including
the turn of year effect should contain, in principle, N discontinuities; in practice only the
first two jumps expected at the end of the current and of the next year can be taken
into account. The jump effect can be modeled simply through a multiplicative coefficient
applied to discount factors, or, equivalently, an additive coefficient applied to zero rates,
corresponding to all the dates of the yield curve affected by a possible jump date. In this
way we are allowed to estimate the coefficient using instruments with a given underlying
rate tenor (e.g. those on Euribor3M used for C3M ), and to apply it to any other curve Cx
taking into account the proper weights. Notice that, as stressed in the previous section,
starting from a smooth and continuous yield curve is crucial for correctly take into account
the discontinuity at the turn of year.
The jump coefficient can be estimated from market quotations using different ap-
proaches, discussed below.
• Jump in the Futures 3M strip: the (no-jump) end of year crossing FRA rate is
obtained through interpolation of non-crossing FRA rates; the rate jump coefficient
is given by the difference between the latter and the quoted value. This approach
always allows the estimation of the second turn of year. The first turn of year can
be obtained only up to the third Wednesday of September, when the corresponding
Futures expiries. In the period October-December there are no non-crossing Futures
to interpolate and the first turn of year should be extrapolated from the second,
making this method not very robust.
• Jump in the FRA 6M strip: this is equivalent to the approach above but it allows
the estimation of the first turn of year up to June (included).
• Jump in the IRS 1M strip: this is equivalent to the approaches above and it allows
the estimation of the first turn of year up to November (included).
• Jump in the FRA strip quoted by brokers each Monday: this approach is valid all
year long, but it allows only a discontinuous weekly update, and is more brokers’
dependent.
The empirical approaches above, when available at the same time, normally give estimates
in excellent agreement with each other. A numerical example is given in fig. 24, where
f f f
we collect the short term bootstrapping of FRA curves CON , C1M , and C3M recovered from
f
figs. 26, 28, and 30 in section 5. The CON yield curve displays both the 2013 (10.2 bps)
f
and the 2014 (8.5 bps) turn of year jumps. The C1M yield curve displays the 2014 turn
of year jump between 1st Dec. 2013 (+1.8 bps) and 2nd Jan. 2014 (−1.6 bps) with size
f
roughly equal to 1/20 of the ON jumps. The C3M yield curve displays the 2014 turn of
year jump between 1st Oct. 2013 (+0.6 bps) and 2nd Jan. 2014 (−0.5 bps), with size
roughly equal to 1/3 of the 1M jump. The jumps are also observable in the zero rate
z z z
curves CON , C1M , and C3M (not shown in fig. 24). We stress that a single turn of the year
induces one discontinuity in the zero rate and discount curves, and two discontinuities in
the FRA rate curve (the FRA rate depending on a ratio of two discounts).
The yield curve discontinuities induced by the turn of year effect may appear, to a non
market-driven reader, a fuzzy effect broking the desired yield curve smoothness. On the
contrary, we stress that they are neither a strangeness of the market quotations nor an
49
0.25% EUR yield curves ON, 1M, 3M (0-2Y)
0.20%
0.15%
FRA rate (%)
0.10%
0.05%
Jun 14
Dec 12
Mar 13
Sep 13
Dec 13
Mar 14
Sep 14
Dec 14
Term
F F F
Figure 24: Yield curves CON (red crosses, bottom), C1M (blue dots, middle) and C3M
(green crosses, top), as of 11 Dec. 2012 (details from figs. 26 and 28, and 30, top panels).
The the turn of year jumps are indicated by arrows: vertical red for ON, horizontal blue
for 1M, and oblique green for 3M.
accident of the bootstrapping, but correspond to true and detectable financial effects that
should be included in any yield curve used to mark to market interest rate derivatives.
50
The delta sensitivity of Π is given by
n
X
Π
∆ (t; R) = ∆Π
j (t; R), total delta sensitivity,
j=1
mj
X
∆Π
j (t; R) = ∆Π
j,k (t; R), partial delta sensitivity for curve Cj ,
k=1
∂Π
∆Π
j,k (t; R) := , partial delta sensitivity for pillar Tj,k . (91)
∂Rj,k
Actually the value of Π depends indirectly on the bootstrapping market rates R and
pillars T through discount factors and FRA rates appearing in the corresponding pricing
formulas14 . Since FRA rates can be written in terms of their associated discount factors
in the sense discussed in sec. 3.4, and since discount factors can be written in terms of
their corresponding zero coupon rates (see sec. 3.5), we may think, in general, that Π
depends directly on a vector of discount factors P = {P1 , . . . , Pn }, with Pj,k := P (t; Rj ),
or zero rates z = {z1 , . . . , zn }, with zj,k := z(t; Rj ). Notice that we take into account
possible non-local effects in bootstrapping, such that a single discount factor or zero rate
may depend on more than one single component in Rj . Since zero rates and market
rates are similar quantities, it’s common practice to consider zero rates as intermediate
variables. In this case the delta sensitivity in eqs. 91 may be written as
mj
X ∂Π
∆Π
j,k (t; R) = Jj,k,a , (92)
α=1
∂zj,α
∂zj,α
Jj,k,a := , (93)
∂Rj,k
or, in compact matrix notation
∆Π
j (t; R) = Jj · ∇j Π,
∂z
j,1
∂zj,mj ∂
∂Rj,1
. . . ∂Rj,1 ∂zj,1
. . .. ..
Jj := .
. . . . , ∇j := . ,
(94)
∂zj,1 ∂zj,mj ∂
∂Rj,m
. . . ∂Rj,m ∂zj,mj
j j
where Jj and ∇j are the jacobian matrix and the gradient operator for yield curve Cj ,
respectively. Notice that the sum in eq. 92 above may include a few terms surrounding
∂zj,k
∂Rj,k
(with index α ' k), depending on the degree of non-locality of the bootstrapping
methodology adopted (see sec. 4.5). In case of local bootstrapping the Jacobian is a
diagonal matrix and we have
∂zj,k
Jj,k,α = δα,k ,
∂Rj,k
∂Π ∂zj,k
∆Π
j,k (t; R) = , (95)
∂zj,k ∂Rj,k
14
this is true also in case of financial instrument more complex than the plain vanillas discussed in secs.
4.3.2-4.3.6.
51
where δα,k is the Kronecker’s symbol15 .
Finally, we take into account the multiple-curve (exogenous) bootstrapping (see sec.
4.2), such that the forwarding yield curves Cx depend on the discounting yield curve CON .
Assuming that CON = C1 in the notation above, we have the Jacobian
∂zj,α ∂zj,α ∂zj,α
Jj,k,α = δj,1 + (1 − δj,1 ) + . (96)
∂Rj,k ∂Rj,k ∂R1,k
∂z
In case of single-curve (endogenous) bootstrapping we have ∂Rj,α 1,k
= 0 and eq. 96 reduces
to eq. 93.
We observe that the computation of the delta sensitivity in eqs. 92, 94, is factorised
into the computation of the Jacobian J and of the gradient ∇Π. The computation of the
gradient depends on the possible complexity of the portfolio Π. Instead, the computation
of the Jacobian depends
on yield
curve bootstrapping, following the logical chain: shock
Rj,k → bootstrap C1 , . . . , Cmj → compute Jj,k,α . Hence, once the yield curves structure
has been fixed (pillars, bootstrapping instruments, etc.), the Jacobian J is relatively stable
with respect to first order movements of the market rates R, and can be recomputed less
frequently than the gradient ∇Π.
is null. Typically the set of hedging instruments for yield curve Cj is a subset of the
most
H liquid Hbootstrapping with marketPquotes denoted by RH =
instruments of Cj , P
n n
R1 , . . . , Rn ⊆ R = {R1 , . . . , Rn }, with j=1 hj ≤ m . The hedging in-
Hj
j=1
struments define an hedging time grid T such that T = R1 , . . . , TnH ⊆ T =
H H
{T1 , . . . , Tn }. The selection of the hedging instruments is clearly subjective, more an art
of the interest rate trader than a science.
In practice, the delta sensitivity of Π is re-distributed on the hedging time grid T H ,
n
X
Π Π H
∆ (t; R) = ∆ (t; R ) = ∆Π H
j (t; R ), total delta sensitivity,
j=1
hj
X
∆Π
j (t; R) = ∆Π H
j,h (t; R ), partial delta sensitivity for curve Cj ,
h=1
∂Π
∆Π H
j,h (t; R ) := H
, partial delta sensitivity for pillar Tj,h . (98)
∂Rj,h
15
such that δα,k = 1 if α = k and 0 otherwise.
52
Then, the hedge amount (or hedge ratios) are computed as
∆Π H
j,h (t; R )
Hj,h (t, RH ) = − Π
,
δj,h (t; RH )
H
Π
∂πj,h
δj,h (t; RH ) := H
, (99)
∂Rj,h
such that the total portfolio satisfies the zero delta condition,
∆Tot (t; RH ) = ∆Π (t; RH ) + ∆H (t; RH )
hj
n X hj
n X H
X X ∂πj,h
= ∆Π H
j,h (t; R ) + H
Hj,h (t, R ) H
j=1 h=1 j=1 h=1
∂Rj,h
hj
n X
X Π
∆j,h (t; RH ) + Hj,h (t, RH )δj,h
Π
(t; RH ) = 0,
= (100)
j=1 h=1
where in the second line we have computed the delta sensitivity ∆H (t; RH ) of the hedge
portfolio ΠH assuming static hedge ratios, such that
∂Hj,h (t, RH )
H
' 0. (101)
∂Rj,h
4.10 Performance
A good computational performance of yield curve bootstrapping algorithms is a key fea-
ture of real-time applications in liquid markets. This is particularly important when delta
sensitivities are required. The bootstrapping algorithms discussed in the previous sections
may display several bottlenecks, as discussed below.
• Analytical vs numerical interpolation (see sec. 4.5: while simple interpolation
schemes may use analytical formulas (e.g. linear interpolation), more general schemes
require a solver or zero-finding numerical algorithm. Hence, efficient algorithms (e.g.
Newton-Raphson [49]) and careful choice of their parameters (e.g. initial guess and
convergence criteria) are required for fast interpolation.
• Non-local interpolation: non-local schemes, such as splines, require an iterative
algorithm to converge to the final solution. Again careful control of numerical
parameters is required for fast convergence.
• Exogenous bootstrapping (see sec. 4.7: this bootstrapping scheme implies a cross-
dependence of the forwarding yield curves Cx on the discounting yield curve Con .
Hence, a market change in a single OIS quotation triggers the recalculation of all
the dependent yield curves. Since this dependence may be often very small, the
cross recalculation may be disabled.
53
• Delta sensitivity (see sec. 91): this is clearly the most expensive computation, since
each single pillar shock triggers the recalculation of the entire yield curve (or of the
whole set of yield curves, in case of OIS shock and exogenous bootstrapping), and
the revaluation of the corresponding portfolio of derivatives. This huge task may be
made more efficient by pre-computing and storing the Jacobian in eq. 96.
We conclude that a good design and careful test of the basic yield curve bootstrapping
component will affect the overall performance of any pricing system.
2. Visual inspection: the bare shape of the yield curve is a good sentinel of possible
bootstrapping problems. All the yield curve typologies (discount, zero/FRA/instantaneous
FRA rates) must be monitored all together, since they convey different and com-
plementary information. Typical examples are reported in figs. 19 and 35.
4. Repricing of market instruments: finally, instruments quoted on the market but not
included in the bootstrapping should be repriced within their bid-ask window. A
typical example are quotes of long term Futures and forward starting IRS.
We conclude that any good yield curve bootstrapping system should include, for each
curve generated, a real-time snapshot of yield curve shapes and a real-time pricer of
market instruments.
54
The numerical results have been obtained using the QuantLib framework16 . The
basic classes and methods (iterative bootstrapping, interpolations, market conventions,
etc.) are implemented in the object oriented C++ QuantLib library [54]. The QuantLib
objects and analytics are exposed to a variety of end-user platforms (including Excel
and Calc) through the QuantLibAddin [55] and QuantLibXL [56] libraries. Market data
are retrieved from the chosen provider and real time is ensured by the ObjectHandler
in-memory repository [57]. The full framework described above is available open source17 .
Clearly, the yield curves reported here are the final result of a complex chain of choices,
and many alternatives are possible. We do not claim that our choices are the best solution
to any problem, they are just our preferences and experience.
55
Eonia ON Yield Curve
Bootstrapping Instrument Rate Start Date End Date
EUR_YC_OND 0.0400% Tue 11 Dec 2012 Wed 12 Dec 2012
EUR_YC_TND 0.0400% Wed 12 Dec 2012 Thu 13 Dec 2012
EUR_YC_SND 0.0400% Thu 13 Dec 2012 Fri 14 Dec 2012
EUR_YC_EONSW 0.0700% Thu 13 Dec 2012 Thu 20 Dec 2012
EUR_YC_EON2W 0.0690% Thu 13 Dec 2012 Thu 27 Dec 2012
EUR_YC_EON3W 0.0780% Thu 13 Dec 2012 Thu 03 Jan 2013
EUR_YC_EON1M 0.0740% Thu 13 Dec 2012 Mon 14 Jan 2013
EUR_YC_EONECBJAN13 0.0460% Wed 16 Jan 2013 Wed 13 Feb 2013
EUR_YC_EONECBFEB13 0.0160% Wed 13 Feb 2013 Wed 13 Mar 2013
EUR_YC_EONECBMAR13 -0.0070% Wed 13 Mar 2013 Wed 10 Apr 2013
EUR_YC_EONECBAPR13 -0.0130% Wed 10 Apr 2013 Wed 08 May 2013
EUR_YC_EONECBMAY13 -0.0140% Wed 08 May 2013 Wed 12 Jun 2013
EUR_YC_EON15M 0.0020% Thu 13 Dec 2012 Thu 13 Mar 2014
EUR_YC_EON18M 0.0080% Thu 13 Dec 2012 Fri 13 Jun 2014
EUR_YC_EON21M 0.0210% Thu 13 Dec 2012 Mon 15 Sep 2014
EUR_YC_EON2Y 0.0360% Thu 13 Dec 2012 Mon 15 Dec 2014
EUR_YC_EON3Y 0.1270% Thu 13 Dec 2012 Mon 14 Dec 2015
EUR_YC_EON4Y 0.2740% Thu 13 Dec 2012 Tue 13 Dec 2016
EUR_YC_EON5Y 0.4560% Thu 13 Dec 2012 Wed 13 Dec 2017
EUR_YC_EON6Y 0.6470% Thu 13 Dec 2012 Thu 13 Dec 2018
EUR_YC_EON7Y 0.8270% Thu 13 Dec 2012 Fri 13 Dec 2019
EUR_YC_EON8Y 0.9960% Thu 13 Dec 2012 Mon 14 Dec 2020
EUR_YC_EON9Y 1.1470% Thu 13 Dec 2012 Mon 13 Dec 2021
EUR_YC_EON10Y 1.2800% Thu 13 Dec 2012 Tue 13 Dec 2022
EUR_YC_EON11Y 1.4040% Thu 13 Dec 2012 Wed 13 Dec 2023
EUR_YC_EON12Y 1.5160% Thu 13 Dec 2012 Fri 13 Dec 2024
EUR_YC_EON15Y 1.7640% Thu 13 Dec 2012 Mon 13 Dec 2027
EUR_YC_EON20Y 1.9390% Thu 13 Dec 2012 Mon 13 Dec 2032
EUR_YC_EON25Y 2.0030% Thu 13 Dec 2012 Mon 14 Dec 2037
EUR_YC_EON30Y 2.0380% Thu 13 Dec 2012 Mon 15 Dec 2042
EU R
Figure 25: Bootstrapping instruments selected for yield curve CON construction as of 11
Dec. 2012.
56
0.20% EUR yield curve ON (0-2Y)
0.15%
FRA rate ON (%)
0.10%
0.05%
0.00%
Jun 14
Dec 12
Mar 13
Sep 13
Dec 13
Mar 14
Sep 14
Dec 14
Term
3.0%
2.5%
FRA rate ON (%)
2.0%
1.5%
1.0%
Dec 16
Dec 20
Dec 24
Dec 28
Dec 32
Dec 36
Dec 40
Dec 44
Dec 48
Dec 52
Dec 56
Dec 60
Dec 64
Dec 68
Dec 72
Term
EU R,f
Figure 26: FRA curve CON on Eonia as of 11 Dec. 2012, plotted with ON-tenor
FRA rates FON (t0 ; t, act/360 ), t daily sampled. Top panel: short term structure up to 2
years. Bottom panel: whole term structure up to 60 years. Blue dots (continuous line):
monotonic cubic spline interpolation on log discounts. Red crosses (discontinuous line):
piecewise constant interpolation on log discounts. See comments in the text.
57
5.2 Euribor1M Yield Curve
EUR
Our Euribor1M curve C1M is bootstrapped using the Euribor1M market instruments
selected in fig. 27 and the corresponding pricing formulas discussed in secs. 4.3.1-4.3.6.
The results are shown in fig. 28.
The first section of the term structure is covered using four synthetic Deposits, from
SND to 3WD, obtained as discussed in sec. 4.4. These instruments are selected to ensure
a smooth short curve in the very first section of the term structure, below the first market
pillar (1M in case of Euribor 1M underlying). We stress that such particular choice is
rather subjective and possibly unconventional, more an art of the interest rate trader
than a science. The 1M Deposit, insisting on Euribor 1M, is taken from fig. 4. The
term structure up to 1Y is built with short term IRS 1M from fig. 11. The medium/long
term structure up to 30Y is built with IRS 1M obtained from IRS 6M in fig. 9 and IRBS
1M vs 6M in fig. 15. The last four pillars between 35Y and 60Y are obtained by flat
extrapolation of the 30Y pillar quote of IRBS 1M vs 6M in fig. 15. Notice that the yield
curve reference date is spot date (t0 = 13 Dec. 2012) and not today’s date. This is not a
problem, since this curve is used for computing FRA rates, not for discounting.
The resulting yield curve term structure displays the mixed upward/downward sloping
behavior frequently observed in recent years. The jumps observed in Dec. 13 correspond
to the turn of year for 1M tenor FRA rates discussed in sec. 4.8.
58
Euribor 1M Yield Curve
Bootstrapping Instrument Rate Start Date End Date
EUR_YC1M_SND 0.0661% Thu 13 Dec 2012 Fri 14 Dec 2012
EUR_YC1M_SWD 0.0980% Thu 13 Dec 2012 Thu 20 Dec 2012
EUR_YC1M_2WD 0.0993% Thu 13 Dec 2012 Thu 27 Dec 2012
EUR_YC1M_3WD 0.1105% Thu 13 Dec 2012 Thu 03 Jan 2013
EUR_YC1M_1MD 0.1100% Thu 13 Dec 2012 Mon 14 Jan 2013
EUR_YC1M_2X1S 0.1060% Thu 13 Dec 2012 Wed 13 Feb 2013
EUR_YC1M_3X1S 0.0960% Thu 13 Dec 2012 Wed 13 Mar 2013
EUR_YC1M_4X1S 0.0850% Thu 13 Dec 2012 Mon 15 Apr 2013
EUR_YC1M_5X1S 0.0790% Thu 13 Dec 2012 Mon 13 May 2013
EUR_YC1M_6X1S 0.0750% Thu 13 Dec 2012 Thu 13 Jun 2013
EUR_YC1M_7X1S 0.0710% Thu 13 Dec 2012 Mon 15 Jul 2013
EUR_YC1M_8X1S 0.0690% Thu 13 Dec 2012 Tue 13 Aug 2013
EUR_YC1M_9X1S 0.0660% Thu 13 Dec 2012 Fri 13 Sep 2013
EUR_YC1M_10X1S 0.0650% Thu 13 Dec 2012 Mon 14 Oct 2013
EUR_YC1M_11X1S 0.0640% Thu 13 Dec 2012 Wed 13 Nov 2013
EUR_YC1M_12X1S 0.0630% Thu 13 Dec 2012 Fri 13 Dec 2013
EUR_YC1M_AB1EBASIS2Y 0.0980% Thu 13 Dec 2012 Mon 15 Dec 2014
EUR_YC1M_AB1EBASIS3Y 0.1860% Thu 13 Dec 2012 Mon 14 Dec 2015
EUR_YC1M_AB1EBASIS4Y 0.3300% Thu 13 Dec 2012 Tue 13 Dec 2016
EUR_YC1M_AB1EBASIS5Y 0.5120% Thu 13 Dec 2012 Wed 13 Dec 2017
EUR_YC1M_AB1EBASIS6Y 0.7040% Thu 13 Dec 2012 Thu 13 Dec 2018
EUR_YC1M_AB1EBASIS7Y 0.8870% Thu 13 Dec 2012 Fri 13 Dec 2019
EUR_YC1M_AB1EBASIS8Y 1.0580% Thu 13 Dec 2012 Mon 14 Dec 2020
EUR_YC1M_AB1EBASIS9Y 1.2110% Thu 13 Dec 2012 Mon 13 Dec 2021
EUR_YC1M_AB1EBASIS10Y 1.3470% Thu 13 Dec 2012 Tue 13 Dec 2022
EUR_YC1M_AB1EBASIS11Y 1.4700% Thu 13 Dec 2012 Wed 13 Dec 2023
EUR_YC1M_AB1EBASIS12Y 1.5810% Thu 13 Dec 2012 Fri 13 Dec 2024
EUR_YC1M_AB1EBASIS15Y 1.8260% Thu 13 Dec 2012 Mon 13 Dec 2027
EUR_YC1M_AB1EBASIS20Y 1.9980% Thu 13 Dec 2012 Mon 13 Dec 2032
EUR_YC1M_AB1EBASIS25Y 2.0590% Thu 13 Dec 2012 Mon 14 Dec 2037
EUR_YC1M_AB1EBASIS30Y 2.0930% Thu 13 Dec 2012 Mon 15 Dec 2042
EUR_YC1M_AB1EBASIS35Y 2.1320% Thu 13 Dec 2012 Fri 13 Dec 2047
EUR_YC1M_AB1EBASIS40Y 2.1850% Thu 13 Dec 2012 Fri 13 Dec 2052
EUR_YC1M_AB1EBASIS50Y 2.2580% Thu 13 Dec 2012 Wed 13 Dec 2062
EUR_YC1M_AB1EBASIS60Y 2.3000% Thu 13 Dec 2012 Tue 13 Dec 2072
EU R
Figure 27: Bootstrapping instruments selected for yield curve C1M construction as of 11
Dec. 2012. Black: market instruments. Green italics: synthetic instruments.
59
0.75% EUR yield curve 1M (0-2 Y)
FRA 1M rate
FRA 1M rate (%)
0.50%
0.25%
0.00%
Jun 2013
Jun 2014
Dec 2012
Feb 2013
Apr 2013
Aug 2013
Oct 2013
Dec 2013
Feb 2014
Apr 2014
Aug 2014
Oct 2014
Dec 2014
Term
2.50%
FRA 1M rate (%)
2.00%
1.50%
1.00%
0.50%
FRA 1M rate
0.00%
Dec 2012
Dec 2017
Dec 2022
Dec 2027
Dec 2032
Dec 2037
Dec 2042
Dec 2047
Dec 2052
Dec 2057
Dec 2062
Dec 2067
Dec 2072
Term
f
Figure 28: FRA curve C1M on Euribor1M as of 11 Dec. 2012, plotted with 1M-tenor FRA
rates F (t0 ; t, t + 1M, act/360 ), t daily sampled. Top panel: short term structure up to
2 years. Bottom panel: whole term structure up to 60 years. The jump observed in the
curve correspond to the turn of years for 1M tenor FRA rates spot starting at 2nd Dec.
2013.
60
Euribor 3M Yield Curve
Bootstrapping Instrument Rate Start Date End Date
EUR_YC3M_2WD 0.1865% Thu 13 Dec 2012 Thu 27 Dec 2012
EUR_YC3M_3WD 0.1969% Thu 13 Dec 2012 Thu 03 Jan 2013
EUR_YC3M_1MD 0.1951% Thu 13 Dec 2012 Mon 14 Jan 2013
EUR_YC3M_2MD 0.1874% Thu 13 Dec 2012 Wed 13 Feb 2013
EUR_YC3M_TOM3F1 0.1790% Fri 14 Dec 2012 Thu 14 Mar 2013
EUR_YC3M_FUT3MZ2 0.1775% Wed 19 Dec 2012 Tue 19 Mar 2013
EUR_YC3M_FUT3MH3 0.1274% Wed 20 Mar 2013 Thu 20 Jun 2013
EUR_YC3M_FUT3MM3 0.1222% Wed 19 Jun 2013 Thu 19 Sep 2013
EUR_YC3M_FUT3MU3 0.1269% Wed 18 Sep 2013 Wed 18 Dec 2013
EUR_YC3M_FUT3MZ3 0.1565% Wed 18 Dec 2013 Tue 18 Mar 2014
EUR_YC3M_FUT3MH4 0.1961% Wed 19 Mar 2014 Thu 19 Jun 2014
EUR_YC3M_FUT3MM4 0.2556% Wed 18 Jun 2014 Thu 18 Sep 2014
EUR_YC3M_FUT3MU4 0.3101% Wed 17 Sep 2014 Wed 17 Dec 2014
EUR_YC3M_AB3E3Y 0.2850% Thu 13 Dec 2012 Mon 14 Dec 2015
EUR_YC3M_AB3E4Y 0.4370% Thu 13 Dec 2012 Tue 13 Dec 2016
EUR_YC3M_AB3E5Y 0.6230% Thu 13 Dec 2012 Wed 13 Dec 2017
EUR_YC3M_AB3E6Y 0.8170% Thu 13 Dec 2012 Thu 13 Dec 2018
EUR_YC3M_AB3E7Y 1.0000% Thu 13 Dec 2012 Fri 13 Dec 2019
EUR_YC3M_AB3E8Y 1.1710% Thu 13 Dec 2012 Mon 14 Dec 2020
EUR_YC3M_AB3E9Y 1.3240% Thu 13 Dec 2012 Mon 13 Dec 2021
EUR_YC3M_AB3E10Y 1.4590% Thu 13 Dec 2012 Tue 13 Dec 2022
EUR_YC3M_AB3E12Y 1.6920% Thu 13 Dec 2012 Fri 13 Dec 2024
EUR_YC3M_AB3E15Y 1.9330% Thu 13 Dec 2012 Mon 13 Dec 2027
EUR_YC3M_AB3E20Y 2.0990% Thu 13 Dec 2012 Mon 13 Dec 2032
EUR_YC3M_AB3E25Y 2.1560% Thu 13 Dec 2012 Mon 14 Dec 2037
EUR_YC3M_AB3E30Y 2.1860% Thu 13 Dec 2012 Mon 15 Dec 2042
EUR_YC3M_AB3EBASIS35Y 2.2308% Thu 13 Dec 2012 Fri 13 Dec 2047
EUR_YC3M_AB3EBASIS40Y 2.2880% Thu 13 Dec 2012 Fri 13 Dec 2052
EUR_YC3M_AB3EBASIS50Y 2.3670% Thu 13 Dec 2012 Wed 13 Dec 2062
EUR_YC3M_AB3EBASIS60Y 2.4126% Thu 13 Dec 2012 Tue 13 Dec 2072
EU R
Figure 29: Bootstrapping instruments selected for yield curve C1M construction as of 11
Dec. 2012. Black: market instruments. Green italics: synthetic instruments.
alternative choice with respect to Futures would be the selection of the market FRA 3M
and IMM FRA 3M in fig. 6. In particular, the FRA 3M would allow for a constant time
grid. Futures are often preferred due to their high market liquidity. Again, this choice is
rather subjective. The other bootstrapping instruments are medium/long term IRS 3M
from 3Y to 30Y taken from fig. 10. The very long-end pillars between 35Y and 50Y are
obtained from IRS 6M quotes in fig. 9 and IRBS 3M vs 6M quotes in fig. 15. The last
pillar at 60Y is obtained through linear extrapolation of previous IRBS quotes. The yield
curve reference date is spot date (t0 = 13 Dec. 2012).
We observe in fig. 30 a behavior similar to Euribor1M yield curve. The jumps observed
in the window Oct. 13 - Dec. 13 correspond to the turn of year for 3M tenor FRA rates
discussed in sec. 4.8. Finally, we stress that FRA 3M quotes shown in fig. 30 are not
included in the bootstrapping, but perfectly repriced. This is the fourth kind of yield
curve check discussed in sec. 4.11.
61
0.75% EUR yield curve 3M (0-2 Y)
FRA 3M rate
FRA 3M quotes
Futures 3M quotes
0.50%
FRA 3M rate (%)
0.25%
0.00%
Jun 2013
Jun 2014
Dec 2012
Feb 2013
Apr 2013
Aug 2013
Oct 2013
Dec 2013
Feb 2014
Apr 2014
Aug 2014
Oct 2014
Dec 2014
Term
2.50%
FRA 3M rate (%)
2.00%
1.50%
1.00%
0.50%
FRA 3M rate
0.00%
Dec 2012
Dec 2017
Dec 2022
Dec 2027
Dec 2032
Dec 2037
Dec 2042
Dec 2047
Dec 2052
Dec 2057
Dec 2062
Dec 2067
Dec 2072
Term
f
Figure 30: FRA curve C3M on Euribor3M as of 11 Dec. 2012, plotted with 3M-tenor FRA
rates F (t0 ; t, t + 3M, act/360 ), t daily sampled. Top panel: short term structure up to 2
years. Bottom panel: whole term structure up to 60 years. Quotes for Futures 3M and
FRA 3M (not included in bootstrapping) are also reported.
62
5.4 Euribor6M Yield Curve
EUR
Our Euribor6M curve C6M is bootstrapped using the Euribor6M market instruments
selected in fig. 31 and the corresponding pricing formulas discussed in secs. 4.3.1-4.3.6.
The results are shown in fig. 32.
The first section of the term structure is covered using nine synthetic Deposits from
EUR
SND to 5MD, obtained as discussed in sec. 4.4. In this case, similar to C!M bootstrap-
ping, we include also the shortest SND instrument. Next, we include the strip of nineteen
market FRA 6M, from FRA Tom6M to FRA 18 × 24, up to 2Y maturity, taken from
fig. 6. We complete the term structure with medium/long term IRS 6M from 3Y to 60Y
taken from fig. 9. In this case neither IRBS nor extrapolations are used, since the market
quotations cover up to 60Y maturity. The yield curve reference date is spot date (t0 =
13 Dec. 2012).
We observe in fig. 32 a behavior similar to Euribor1M and 3M yield curves. No turn
of the year jumps are included, because their market quotation is negligible. The IMM
FRA 6M quotes shown in fig. 32 are not included in the bootstrapping, but perfectly
repriced.
63
Euribor 6M Yield Curve
Bootstrapping Instrument Rate Start Date End Date
EUR_YC6M_SND 0.3565% Thu 13 Dec 2012 Fri 14 Dec 2012
EUR_YC6M_SWD 0.3858% Thu 13 Dec 2012 Thu 20 Dec 2012
EUR_YC6M_2WD 0.3840% Thu 13 Dec 2012 Thu 27 Dec 2012
EUR_YC6M_3WD 0.3922% Thu 13 Dec 2012 Thu 03 Jan 2013
EUR_YC6M_1MD 0.3869% Thu 13 Dec 2012 Mon 14 Jan 2013
EUR_YC6M_2MD 0.3698% Thu 13 Dec 2012 Wed 13 Feb 2013
EUR_YC6M_3MD 0.3527% Thu 13 Dec 2012 Wed 13 Mar 2013
EUR_YC6M_4MD 0.3342% Thu 13 Dec 2012 Mon 15 Apr 2013
EUR_YC6M_5MD 0.3225% Thu 13 Dec 2012 Mon 13 May 2013
EUR_YC6M_TOM6F1 0.3120% Fri 14 Dec 2012 Fri 14 Jun 2013
EUR_YC6M_1x7F 0.2930% Mon 14 Jan 2013 Mon 15 Jul 2013
EUR_YC6M_2x8F 0.2720% Wed 13 Feb 2013 Tue 13 Aug 2013
EUR_YC6M_3x9F 0.2600% Wed 13 Mar 2013 Fri 13 Sep 2013
EUR_YC6M_4x10F 0.2560% Mon 15 Apr 2013 Tue 15 Oct 2013
EUR_YC6M_5x11F 0.2520% Mon 13 May 2013 Wed 13 Nov 2013
EUR_YC6M_6x12F 0.2480% Thu 13 Jun 2013 Fri 13 Dec 2013
EUR_YC6M_7x13F 0.2540% Mon 15 Jul 2013 Wed 15 Jan 2014
EUR_YC6M_8x14F 0.2610% Tue 13 Aug 2013 Thu 13 Feb 2014
EUR_YC6M_9x15F 0.2670% Fri 13 Sep 2013 Thu 13 Mar 2014
EUR_YC6M_10x16F 0.2790% Mon 14 Oct 2013 Mon 14 Apr 2014
EUR_YC6M_11x17F 0.2910% Wed 13 Nov 2013 Tue 13 May 2014
EUR_YC6M_12x18F 0.3030% Fri 13 Dec 2013 Fri 13 Jun 2014
EUR_YC6M_13x19F 0.3180% Mon 13 Jan 2014 Mon 14 Jul 2014
EUR_YC6M_14x20F 0.3350% Thu 13 Feb 2014 Wed 13 Aug 2014
EUR_YC6M_15x21F 0.3520% Thu 13 Mar 2014 Mon 15 Sep 2014
EUR_YC6M_16x22F 0.3710% Mon 14 Apr 2014 Tue 14 Oct 2014
EUR_YC6M_17x23F 0.3890% Tue 13 May 2014 Thu 13 Nov 2014
EUR_YC6M_18x24F 0.4090% Fri 13 Jun 2014 Mon 15 Dec 2014
EUR_YC6M_AB6E3Y 0.4240% Thu 13 Dec 2012 Mon 14 Dec 2015
EUR_YC6M_AB6E4Y 0.5760% Thu 13 Dec 2012 Tue 13 Dec 2016
EUR_YC6M_AB6E5Y 0.7620% Thu 13 Dec 2012 Wed 13 Dec 2017
EUR_YC6M_AB6E6Y 0.9540% Thu 13 Dec 2012 Thu 13 Dec 2018
EUR_YC6M_AB6E7Y 1.1350% Thu 13 Dec 2012 Fri 13 Dec 2019
EUR_YC6M_AB6E8Y 1.3030% Thu 13 Dec 2012 Mon 14 Dec 2020
EUR_YC6M_AB6E9Y 1.4520% Thu 13 Dec 2012 Mon 13 Dec 2021
EUR_YC6M_AB6E10Y 1.5840% Thu 13 Dec 2012 Tue 13 Dec 2022
EUR_YC6M_AB6E12Y 1.8090% Thu 13 Dec 2012 Fri 13 Dec 2024
EUR_YC6M_AB6E15Y 2.0370% Thu 13 Dec 2012 Mon 13 Dec 2027
EUR_YC6M_AB6E20Y 2.1870% Thu 13 Dec 2012 Mon 13 Dec 2032
EUR_YC6M_AB6E25Y 2.2340% Thu 13 Dec 2012 Mon 14 Dec 2037
EUR_YC6M_AB6E30Y 2.2560% Thu 13 Dec 2012 Mon 15 Dec 2042
EUR_YC6M_AB6E35Y 2.2950% Thu 13 Dec 2012 Fri 13 Dec 2047
EUR_YC6M_AB6E40Y 2.3480% Thu 13 Dec 2012 Fri 13 Dec 2052
EUR_YC6M_AB6E50Y 2.4210% Thu 13 Dec 2012 Wed 13 Dec 2062
EUR_YC6M_AB6E60Y 2.4630% Thu 13 Dec 2012 Tue 13 Dec 2072
EU R
Figure 31: Bootstrapping instruments selected for yield curve C1M construction as of 11
Dec. 2012.
64
0.75% EUR yield curve 6M (0-2 Y)
0.50%
FRA 6M rate (%)
0.25%
FRA 6M rate
FRA 6M quotes
IMM FRA 6M quotes
0.00%
Jun 2013
Jun 2014
Dec 2012
Feb 2013
Apr 2013
Aug 2013
Oct 2013
Dec 2013
Feb 2014
Apr 2014
Aug 2014
Oct 2014
Dec 2014
Term
2.50%
FRA 6M rate (%)
2.00%
1.50%
1.00%
0.50%
FRA 6M rate
0.00%
Dec 2012
Dec 2017
Dec 2022
Dec 2027
Dec 2032
Dec 2037
Dec 2042
Dec 2047
Dec 2052
Dec 2057
Dec 2062
Dec 2067
Dec 2072
Term
f
Figure 32: FRA curve C6M on Euribor6M as of 11 Dec. 2012, plotted with 6M-tenor FRA
rates F (t0 ; t, t + 6M, act/360 ), t daily sampled. Top panel: short term structure up to 2
years. Bottom panel: whole term structure up to 60 years. Quoted FRA 6M and IMM
FRA 6M (not included in bootstrapping) are also reported.
65
Euribor 12M Yield Curve
Bootstrapping Instrument Rate Start Date End Date
EUR_YC12M_1MD 0.6537% Thu 13 Dec 2012 Mon 14 Jan 2013
EUR_YC12M_3MD 0.6187% Thu 13 Dec 2012 Wed 13 Mar 2013
EUR_YC12M_6MD 0.5772% Thu 13 Dec 2012 Thu 13 Jun 2013
EUR_YC12M_9MD 0.5563% Thu 13 Dec 2012 Fri 13 Sep 2013
EUR_YC12M_12MD 0.5400% Thu 13 Dec 2012 Fri 13 Dec 2013
EUR_YC12M_3x15F 0.4974% Wed 13 Mar 2013 Thu 13 Mar 2014
EUR_YC12M_6x18F 0.4783% Thu 13 Jun 2013 Fri 13 Jun 2014
EUR_YC12M_9x21F 0.4822% Fri 13 Sep 2013 Mon 15 Sep 2014
EUR_YC12M_12x24F 0.5070% Fri 13 Dec 2013 Mon 15 Dec 2014
EUR_YC12M_15x27F 0.5481% Thu 13 Mar 2014 Fri 13 Mar 2015
EUR_YC12M_18x30F 0.6025% Fri 13 Jun 2014 Mon 15 Jun 2015
EUR_YC12M_AB12EBASIS3Y 0.6030% Thu 13 Dec 2012 Mon 14 Dec 2015
EUR_YC12M_AB12EBASIS4Y 0.7400% Thu 13 Dec 2012 Tue 13 Dec 2016
EUR_YC12M_AB12EBASIS5Y 0.9130% Thu 13 Dec 2012 Wed 13 Dec 2017
EUR_YC12M_AB12EBASIS6Y 1.0930% Thu 13 Dec 2012 Thu 13 Dec 2018
EUR_YC12M_AB12EBASIS7Y 1.2650% Thu 13 Dec 2012 Fri 13 Dec 2019
EUR_YC12M_AB12EBASIS8Y 1.4260% Thu 13 Dec 2012 Mon 14 Dec 2020
EUR_YC12M_AB12EBASIS9Y 1.5700% Thu 13 Dec 2012 Mon 13 Dec 2021
EUR_YC12M_AB12EBASIS10Y 1.6970% Thu 13 Dec 2012 Tue 13 Dec 2022
EUR_YC12M_AB12EBASIS12Y 1.9150% Thu 13 Dec 2012 Fri 13 Dec 2024
EUR_YC12M_AB12EBASIS15Y 2.1300% Thu 13 Dec 2012 Mon 13 Dec 2027
EUR_YC12M_AB12EBASIS20Y 2.2670% Thu 13 Dec 2012 Mon 13 Dec 2032
EUR_YC12M_AB12EBASIS25Y 2.3060% Thu 13 Dec 2012 Mon 14 Dec 2037
EUR_YC12M_AB12EBASIS30Y 2.3220% Thu 13 Dec 2012 Mon 15 Dec 2042
EUR_YC12M_AB12EBASIS35Y 2.3610% Thu 13 Dec 2012 Fri 13 Dec 2047
EUR_YC12M_AB12EBASIS40Y 2.4140% Thu 13 Dec 2012 Fri 13 Dec 2052
EUR_YC12M_AB12EBASIS50Y 2.4870% Thu 13 Dec 2012 Wed 13 Dec 2062
EUR_YC12M_AB12EBASIS60Y 2.5290% Thu 13 Dec 2012 Tue 13 Dec 2072
EU R
Figure 33: Bootstrapping instruments selected for yield curve C1M construction as of 11
Dec. 2012.
66
0.75% EUR yield curve 12M (0-2 Y)
FRA 12M rate (%)
0.50%
0.25%
0.00%
Jun 2013
Jun 2014
Dec 2012
Feb 2013
Apr 2013
Aug 2013
Oct 2013
Dec 2013
Feb 2014
Apr 2014
Aug 2014
Oct 2014
Dec 2014
Term
2.50%
FRA 12M rate (%)
2.00%
1.50%
1.00%
0.50%
FRA 12M rate
0.00%
Dec 2012
Dec 2017
Dec 2022
Dec 2027
Dec 2032
Dec 2037
Dec 2042
Dec 2047
Dec 2052
Dec 2057
Dec 2062
Dec 2067
Dec 2072
Term
f
Figure 34: FRA curve C12M on Euribor12M as of 11 Dec. 2012, plotted with 12M-tenor
FRA rates F (t0 ; t, t + 12M, act/360 ), t daily sampled. Top panel: short term structure
up to 2 years. Bottom panel: whole term structure up to 30 years. FRA 12M are also
reported.
67
60 EUR FRA basis
12M vs ON
6M vs ON
50
3M vs ON
FRA basis spread (bps)
1M vs ON
40
30
20
10
0
Dec-12
Dec-15
Dec-18
Dec-21
Dec-24
Dec-27
Dec-30
Dec-33
Dec-36
Dec-39
Dec-42
Term
1M vs ON
12M vs 1M
40 12M vs 3M
6M vs 1M
30 12M vs 6M
6M vs 3M
3M vs 1M
20
10
0
Dec-12
Dec-15
Dec-18
Dec-21
Dec-24
Dec-27
Dec-30
Dec-33
Dec-36
Dec-39
Dec-42
Term
Figure 35: Forward IRBS curves as of 11 Dec. 2012. Top panel: difference between OIS
vs Euribor FRA rates, with FRA tenor equal to Euribor tenor. Bottom panel: difference
between FRA rates both with over night tenor.
68
year jumps included in some yield curves, and to the over night FRA rate tenor. On the
other side, this representation is less significative from a financial point of view, because
we compute a FRA rate with tenor x using an Euribor curve CyF with tenor y 6= x.
The top panel shows the same difference between two FRA rates, but in this case
the length of both FRA is equal to the tenor x of the Euribor rate. This is equivalent
to a series of forward starting zero coupon IRBS with length x. This representation is
financially sound and much more smooth, with a residual irregularity due the different
local shapes of the pair of yield curves involved.
69
Using eqs. 8 and 4, we obtain the corresponding price, dividend and gain dynamics
P
µ (t)S(t)dt + σ(t)S(t)dW P (t)
X (t) = rf (t)Bf (t)dt , dD(t) = 0, dG(t) = dX (t). (103)
rc (t)Bc (t)dt
∂Π
dΠ(t, X) = D̂µ (t)Π(t, X)dt + σ(t)X(t) (t, X)dW P (t),
∂X
∂ ∂ 1 ∂2
D̂µ (t, X) : = + µP (t)X(t) + σ 2 (t)X 2 (t) . (104)
∂t ∂X 2 ∂X 2
We replicate the derivatives’ price Π(t, X), ∀t ≤ T with the trading strategy
θ1 (t)
θ(t) = θ2 (t) , (105)
θ3 (t)
such that the corresponding strategy’s price, dividend and gain dynamics are given, re-
spectively, by
+ θ1 (t)σ(t)X(t)dW P (t),
dD(t, X , θ) = 0,
dG(t, X , θ) = dX (t, X , θ). (106)
We apply now the relevant conditions: self-financing, replication, perfect collateral, and
risk free. The self-financing condition, D(t, X , θ) = 0, is automatically satisfied by the
strategy by construction, thanks to the absence of dividends. The replication conditions
are
where GΠ (t, X) is the cumulative gain process associated with derivative Π. Using the
first replication condition in eq. 107 we obtain
where we have used the perfect collateral condition, Π(t, X) = Bc (t), given in Def. 3.1.
Setting, without loss of generality, θ3 (t) = 1, we obtain the equation
70
Using the second replication condition in eq. 108, the derivative’s price dynamics in eq.
104, the strategy’s gain dynamics in eq. 106, and eliminating θ2 (t) using eq. 109 above,
we obtain
∂Π
dGΠ (t, X) = dΠ(t, X) = D̂µP Π(t, X)dt + σ(t)X(t) (t, X)dW P (t)
∂X
= dG(t, θ, X) = dX(t, X, θ)
= θ1 (t)µP (t)X(t)dt + θ1 (t)σ(t)X(t)dW P (t) + dΓ(t, X), (111)
dΓ(t, X)
:= −rf (t)θ1 (t)X(t) + rc (t)Π(t, X). (112)
dt
Rearranging the terms we obtain
∂ 2Π
∂Π P ∂Π 1 2 2
+ µ (t)X(t) − θ1 (t) + σ (t)X (t) dt
∂t ∂X 2 ∂X 2
∂Π
+ σ(t)X(t) − θ1 (t) dW P (t) = dΓ(t, X). (113)
∂X
∂Π
Applying the market risk neutral condition θ1 (t) = ∂X
we obtain the PDE
D̂µ (t, X)f (t, X) = A(t, X)f (t, X) + B(t, X), (115)
∂ ∂ 1 ∂2
D̂µ (t, X) := + µ(t, X) + σ 2 (t, X) , (116)
∂t ∂X 2 ∂X 2
f (T, X) = ψ(X),
71
also known as backward Kolmogorov equation, where t ∈ [0, T ], X ∈ R, ψ : R → R,
f ∈ C 1,2 {[0, T ] × R → R}, A, B : [0, T ]×R → R, µ : [0, T ]×R → R, σ : [0, T ]×R → R+ ,
we have, under appropriate technical conditions on functions f, ψ, A, B, µ, σ,
Z T
Q
f (t, X) = Et D(t, T, X)ψ(X) − D(u, T, X)B(u, X)du , (117)
t
Z T
D(t, T, X) := exp − A(u, X)du , (118)
t
under the probability measure Q such that the asset X follows the Geometric Brownian
Motion process
Remark B.2 (Probability measure). We stress that the dynamics in eq. 119 differs from
those in eq. 8 for the drift, µ instead of µP , and the associated probability measure, Q
(risk neutral) instead of P (real). In particular, the risk neutral drift µ in eq. 119 is
dictated by the coefficient µ in the differential operator D̂µ in eq. 116, while the real drift
µP disappears in the final eqs. 117 and 119. Since the stochastic terms in eq. 119 and
8 are, instead, the same, by Girsanov theorem (see e.g. [23], [24]) the Feynman-Kac
theorem implies a change of probability measure.
C.1 FRA
The payoff of the market FRA is given in eq. 42. Using the pricing under collateral
approach discussed in sec. 3.2, proposition 3.3, we have that the price of the market FRA
T
at time t < Ti−1 is given, under the payment forward measure Qf i−1 , by
T
Qf i−1
FRAMkt (t; T , K, ω) = Pc (t; Ti−1 )Et [FRAMkt (Ti−1 ; T , K, ω)]
T
Qf i−1[Lx (Ti−1 , Ti ) − K] τL (Ti−1 , Ti )
= N ωPc (t; Ti−1 )Et
1 + Lx (Ti−1 , Ti )τL (Ti−1 , Ti )
T
Qf i−1 1
= N ωPc (t; Ti−1 ) 1 − [1 + KτL (Ti−1 , Ti )] Et . (120)
1 + Lx (Ti−1 , Ti )τL (Ti−1 , Ti )
72
T
Switching from Qf i−1 to QTf i , the expectation above becomes
T
Q i−1 1
Et f
1 + Lx (Ti−1 , Ti )τL (Ti−1 , Ti )
Pc (t, Ti−1 ) QTf i 1 1
= E ,
Pc (t, Ti ) t Pc (Ti−1 , Ti ) 1 + Lx (Ti−1 , Ti )τL (Ti−1 , Ti )
T
1 Qf i 1 + Lc (Ti−1 , Ti )τc (Ti−1 , Ti )
= E . (121)
1 + Fc,i (t)τc (Ti−1 , Ti ) t 1 + Lx (Ti−1 , Ti )τL (Ti−1 , Ti )
Thus we have the pricing expression
We observe that both the price and the equilibrium rate of the market FRA depend on
the expectation of the ratio between the two rates Lc (Ti−1 , Ti ) and Lx (Ti−1 , Ti ) under the
forward measure QTf i . This quantity depends on the particular model chosen for the joint
distribution of these two rates. In general, for any model, we have
T
Qf i 1 + Ld (Ti−1 , Ti )τd (Ti−1 , Ti ) 1 + Fd,i (t)τd (Ti−1 , Ti ) Cc,x
FRA (t;T
i−1 )
Et = e , (124)
1 + Lx (Ti−1 , Ti )τx (Ti−1 , Ti ) 1 + Fx,i (t)τx (Ti−1 , Ti )
such that
1 + Kτx (Ti−1 , Ti ) Cc,x FRA (t;T
i−1 )
FRAMkt (t; T , K, ω) = N ωPc (t; Ti−1 ) 1 − e , (125)
1 + Fx,i (t)τx (Ti−1 , Ti )
FRA 1 n FRA
o
Rx,Mkt (t; T ) = [1 + Fx,i (t)τx (Ti−1 , Ti )] eCc,x (t;Ti−1 ) − 1 , (126)
τx (Ti−1 , Ti )
where CxFRA (t; Ti−1 ) is a convexity adjustment, whose detailed expression depends on the
chosen model for the dynamics of Fd,i (t) and Fx,i (t). A possible modeling choice is that
of [42], in which the two FRA rates are modeled as shifted lognormal martingales under
the collateral forward measure QTc i ,
T
dFc,i (t) Qf i
1 = σc,i dW c (t), (127)
Fc,i (t) + τc (Ti−1 ,Ti )
T
dFx,i (t) Qf i
1 = σx,i dWx (t), (128)
Fx,i (t) + τL (Ti−1 ,Ti )
T T
Q i Q i
dWc f (t)dWx f (t) = ρc,x,i dt. (129)
73
In this case the convexity adjustment assumes the form
FRA 2
Cc,x (t; Ti−1 ) = σx,i − σx,i σc,i ρc,x,i τ (t, Ti−1 ). (130)
For typical post credit crunch market situations, the actual size of the convexity adjust-
ment results to be below 1 bp, even for very long maturities (see [42]).
C.2 Futures
The payoff of Futures is given in eq. 52. Since Futures are collateralized financial instru-
ments, we may use proposition 3.2 to obtain the price of the Futures at time t < Ti−1
as
Q
Futures(t; T ) = Et f [Dc (t; t)Futures(Ti−1 ; T )] (131)
n o
Qf
= N 1 − Et [Lx (Ti−1 , Ti )] = N 1 − RxFut (t; T ) ,
(132)
where RxFut (t; T ) is the Futures’ rate and Qf is the risk neutral funding measure. Notice
that the Futures daily margination mechanism implies that the payoff is regulated every-
day, thus generating the unitary stochastic discount factor Dc (t; t) = 1 appearing in the
first line above. Hence the pricing of Futures requires the computation of the Futures’
rate
Q Q
RxFut (t; T ) := Et f [Lx (Ti−1 , Ti )] , = Et f [Fx,i (Ti−1 )] . (133)
Since the FRA rate Fx,i (t) is not a martingale under the risk neutral measure Qf , such
computation requires the adoption of a model for the dynamics of Fx,i (t), similarly to the
case of the market FRA discussed in app. C.1. In general, we obtain that the Futures’
rate is given by the corresponding FRA rate corrected with a convexity adjustment
T
Q Qf i
RxFut (t; T ) := Et f [Lx (Ti−1 , Ti )] = Et [Lx (Ti−1 , Ti )] + CxFut (t; Ti−1 )
= Fx,i (t) + CxFut (t; Ti−1 ). (134)
The expression of the convexity adjustment will depend on the particular model
adopted, and will contain, in general, the model’s volatilities and correlations. We re-
port here the result of the most recent derivation given by [42] under the multiple-curve
74
Libor Market Model, where the convexity adjustment takes the form
Z Ti−1
Fut
Cx (t; Ti−1 ) ' Fx,i (t) exp µx,i (u) du − 1 ,
t
T
dFc,i (t) Q i
= µc,i (t)dt + σc,i dWc f (t),
Fc,i (t)
T
dFx,i (t) Q i
= µx,i (t)dt + σx,i dWx f (t),
Fx,i (t)
i
X σc,j ρc,c,j,i τc,j Fc,j (t)
µc,i (t) := σc,i ,
j=1
1 + τ c,j F c,j (t)
i
X σx,j ρx,c,j,i τx,j Fx,j (t)
µx,i (t) := σx,i ,
j=1
1 + τx,j Fx,j (t)
Z Ti−1 i
X σc,j ρx,c,j,i τc,j Fc,j (t)
µx,i (u) du ' σx,i τ (t, Tj−1 ),
t j=1
1 + τ c,j F c,j (t)
T
Q j 1 Pc (t; Tj−1 )
Fc,j (t) := Et f [Lc (Tj−1 , Tj )] = −1 , (135)
τc,j Pc (t; Tj )
where σx,i , σc,j , ρx,c,i,j ρc,c,i,j are the instantaneous (deterministic) volatilities and corre-
lations of Fx,i (t), Fc,j (t), respectively.
C.3 IRS
The payoff of IRSlets are given in eq. 60. Using the pricing under collateral approach
discussed in sec. 3.2, proposition 3.3, we obtain that the price of the IRSlets at time
t < Ti−1 is given, under the payment forward measure QTf i , by
Ti
Q
IRSletfloat (t; Ti−1 , Ti , Lx ) = Pc (t; Ti )Et f [IRSletfloat (Ti−1 ; Ti−1 , Ti , Lx )]
= N Pc (t; Ti )Fx,i (t)τL (Ti−1 , Ti ), (136)
S
Qf j
IRSletfix (t; Sj−1 , Sj , K) = Pc (t; Sj )Et [IRSletfix (Sj−1 ; Sj−1 , Sj , K)]
= N Pc (t; Sj )KτK (Sj−1 , Sj ). (137)
75
where ω = +/ − 1 for a payer/receiver IRS (referred to the fixed rate K), and the legs
values, the equilibrium IRS rate and the annuity have been defined as
n
X
IRSfloat (t; T , Lx ) : = IRSletfloat (t; Ti−1 , Ti , Lx )
i=1
n
X
=N Pc (t; Ti )Fx,i (t)τL (Ti−1 , Ti ), (139)
i=1
m
X
IRSfix (t; S, K) : = IRSletfix (t; Sj−1 , Sj , K)
j=1
respectively.
C.4 OIS
The payoff of the OIS is given in eqs. 71, 72. Using the pricing under collateral approach
discussed in sec. 3.2, we obtain that the prices of OISlets at time t < Ti−1 are given by
Ti
Q
OISletfloat (t; Ti−1 , Ti , Ron ) = Pc (t; Ti )Et f [OISletfloat (Ti−1 ; Ti−1 , Ti , Ron )]
= N Pc (t; Ti )Ron (t; Ti )τon (Ti−1 , Ti ), (143)
OISletfix (t; Sj−1 , Sj , K) = IRSletfix (t; Sj−1 , Sj , K), (144)
where we have used the tower rule for nested conditioned expectations, and we have
defined the over night FRA rate, insisting onto the over night time interval [Ti,k−1 , Ti,k ],
as
T
Q i
Ron (t; Ti,k−1 , Ti,k ) : = Et f [Ron (Ti,k−1 , Ti,k )] . (146)
76
Under the perfect collateral assumption the over night FRA rates Ron (Ti,k−1 , Ti,k ) may be
considered martingales under the funding measure QTf i , such that the classical single-curve
expression,
1 Pc (t; Ti,k−1 )
Ron (t; Ti,k−1 , Ti,k ) = −1 , (147)
τon (Ti,k−1 , Ti,k ) Pc (t; Ti,k )
holds. In other words, eq. 147 assumes that the over night FRA rates Ron (Ti,k−1 , Ti,k )
may be replicated using collateral zero coupon bonds Pc (t; T ). In this case the over night
FRA rates assume a simpler expression
(n )
i
1 Y Pc (t; Ti,k−1 )
Ron (t; Ti ) = −1
τon (Ti−1 , Ti ) k=1 Pc (t; Ti,k )
1 Pc (t; Ti,0 ) Pc (t; Ti,1 ) Pc (t; Ti,ni −1 )
= ··· −1
τR (Ti−1 , Ti ) Pc (t; Ti,1 ) Pc (t; Ti,2 ) Pc (t; Ti,ni )
1 Pc (t; Ti−1 )
= −1
τon (Ti−1 , Ti ) Pc (t; Ti )
= Fc (t; Ti−1 , Ti ), (148)
where we have used the telescopic rule for the zero coupon bonds inside the product. The
price of the complete OIS is thus given by
c
OIS(t; T , S, Ron , K, ω) = ω [OISfloat (t; T , Ron ) − OISfix (t; S, K)]
OIS
= N ω Ron (t; T , S) − K Ac (t; S), (149)
where the OIS legs values and the equilibrium OIS rate are
n
X
OISfloat (t; T , Ron ) : = OISletfloat (t; Ti−1 , Ti , Ron )
i=1
n
X
=N Pc (t; Ti )Ron (t; Ti )τon (Ti−1 , Ti )
i=1
n
X
=N [Pc (t; Ti−1 ) − Pc (t; Ti )]
i=1
= N [Pc (t; T0 ) − Pc (t; Tn )] , (150)
Xn
OISfix (t; S, K) : = OISletfix (t; Sj−1 , Sj , K)
i=1
= IRSfix (t; S, K)
= N KAc (t; S), (151)
Pn
OIS Pc (t; Ti )Ron (t; Ti )τon (Ti−1 , Ti )
Ron (t; T , S) = i=1
Ac (t; S)
Pc (t; T0 ) − Pc (t; Tn )
:= . (152)
Ac (t; S)
77
C.5 IRBS
IRBS as Two IRS
The schedule and payoff of IRBS as two IRS are given in eqs. 79, 80, respectively. The
IRBS price is given, using eq. 138, by
and the corresponding equilibrium IRBS spread is defined as the difference between the
two IRS rates
We observe that the two definitions of IRBS are equivalent, but the two basis spreads in
eqs. 154 and 156 are not equal, because of the two different annuities involved
Ac (t; Ty )
∆(t; Tx , Ty , S) = ∆(t; Tx , Ty ). (157)
Ac (t; S)
References
[1] Ferdinando M. Ametrano and Marco Bianchetti. Smooth Yield Curves Bootstrapping
For Forward Libor Rate Estimation and Pricing Interest Rate Derivatives. In Fabio
Mercurio, editor, Modelling Interest Rates: Latest Advances for Derivatives Pricing,
pages 3–42. Risk Books, May 2009.
[2] Peter Jurcaga. SwapClear Zero-Coupon Yield Curve Construction. Technical Infor-
mation Package 4.10, LCH.CLEARNET, June 2010.
78
[3] H. Lipman and Fabio Mercurio. The New Swap Math. Bloomberg Markets, March
2010.
[5] Richard White. The analythic framework for implying yield curves from market data.
Quantitative research, OpenGamma, March 2012.
[6] Mark Gibbs and Russel Goyder. The past, present and future of curves. Technical
paper, Fincad, October 2012.
[8] Paul Soderlind and Lars E.O. Svensson. New techniques to extract market expec-
tations from financial instruments. Journal of Monetary Economics, 40:383–429,
1997.
[9] Charles R. Nelson and Andrew F. Siegel. Parsimonious modeling of yield curves.
Journal of Business, 60:473–489, 1987.
[10] Jens H. E.Christensen, Francis X. Diebold, and Glenn D. Rudebusch. The affine
arbitrage-free class of nelson-siegel term structure models. Working Paper 2007–20,
FRB of San Francisco, 2007.
[11] Laura Coroneo, Ken Nyholm, and Rositsa Vidova-Koleva. How arbitrage free is the
nelson-siegel model ? Working paper series 874, European Central Bank, 2008.
[12] Patrick. S. Hagan and Graeme West. Interpolation methods for curve construction.
Applied Mathematical Finance, 13(2):89–129, June 2006.
[13] Patrick. S. Hagan and Graeme West. Methods for constructing a yield curve. Wilmott
Magazine, pages 70–81, 2008.
[14] Leif B.G. Andersen and Vladimir V. Piterbarg. Interest Rate Modeling. Atlantic
Financial Press, 1st edition, 2010.
[15] Uri Ron. A practical guide to swap curve construction. Working Paper 2000-17,
Bank of Canada, August 2000.
[16] Leif B.G. Andersen. Discount curve construction with tension splines. Review of
Derivatives Research, 10(3):227–267, December 2007.
[17] Peter Madigan. Libor under attack. Risk Magazine, June 2008.
[18] Damiano Brigo and Fabio Mercurio. Interest-Rate Models - Theory and Practice.
Springer, 2nd edition, 2006.
[20] ISDA. The Standard CSA: a Proposal for Standard Bilateral CollateraliZation of
Over-the-Counter Derivatives, Release 2, May 2011.
79
[21] Nick Sawyer. Standard CSA: Industry’s solution to novation bottleneck gets nearer.
Risk Magazine, September 2011.
[23] Darrel Duffie. Dynamic Asset Pricing Theory. Princeton University Press, 3 edition,
2001.
[24] Tomas Bjork. Arbitrage Theory in Continuous Time. Oxford Finance. Oxford Uni-
versity Press, 3 edition, 2009.
[26] Vladimir V. Piterbarg. Cooking With Collateral. Risk Magazine, August 2012.
[27] Masaaki Fujii and Akihiko Takahashi. Choice of Collateral Currency. Risk Magazine,
pages 120–125, January 2011.
[28] Antonio Castagna. Pricing of derivatives contracts under collateral agreements: Liq-
uidity and funding value adjustments. SSRN working paper, 12 2011.
[30] Meng Han, Yeqi He, and Hu Zhang. A Note on Discounting and Funding Value
Adjustments for Derivatives. SSRN working paper, February 2013.
[31] Damiano Brigo, Andrea Pallavicini, Cristin Buescu, and Quing Liu. Illustrating a
problem in the self-financing condition in two 2010-2011 papers on funding, collateral
and discounting. SSRN working paper, July 2012.
[32] Stephane Crpey. Bilateral counterparty risk under funding constraints part i: Pric-
ing. April 2012.
[33] Stephane Crpey. Bilateral counterparty risk under funding constraints part ii: Cva.
April 2012.
[34] Christoph Burgard and Mats Kjaer. In the balance. Risk, pages 72–75, November
2011.
[35] Christoph Burgard and Mats Kjaer. A generalised cva with funding and collateral.
SSRN working paper, Aug 2012.
[36] Massimo Morini. Solving the Puzzle in the Interest Rate Market. SSRN working
paper, 2009.
[37] Marco Bianchetti. Two Curves, One Price. Risk Magazine, pages 74–80, August
2010.
[38] Fabio Mercurio. Post Credit Crunch Interest Rates: Formulas and Market Models.
SSRN working paper, 2009.
80
[39] Interest Rate Instruments and Market Conventions Guide. Quantitative research,
OpenGamma, April 2012.
[40] John C. Hull. Options, Futures and Other Derivatives. Prentice Hall, 8th edition,
January 2011.
[41] Riccardo Rebonato. Interest-Rate Option Models. John Wiley & Sons, 2nd edition,
1998.
[42] Fabio Mercurio. Libor Market Models with Stochastic Basis. SSRN working paper,
March 2010.
[43] Peter Jackel and Atsushi Kawai. The future is convex. Wilmott Magazine, pages
2–13, February 2005.
[44] George Kirikos and David Novak. Convexity conundrums. Risk Magazine, 10(3):60–
61, March 1997.
[45] Vladimir V. Piterbarg and Marco A. Renedo. Eurodollar futures convexity adjust-
ments in stochastic volatility models. Journal of Computational Finance, 9(3), 2006.
[46] Marc P. A. Henrard. Eurodollar futures and options: Convexity adjustment in HJM
one-factor model. SSRN working paper, March 2009.
[47] John Hull and Alan White. Pricing interest rate derivative securities. The Review of
Financial Studies, 3:573–592, 1990.
[49] William H. Press, Saul A. Teukolsky, William T. Vetterling, and Brian P. Flannery.
Numerical Recipes: the Art of Scientific Computing. Cambridge University Press,
3rd edition, 2007.
[50] Ralph Atkins and Claire Jones. Investors eye possible negative ecb rates. Financial
times, December 2012.
[51] Laurie Carver. Negative rates: dealers struggle to price 0 percent floors. Risk Mag-
azine, November 2012.
[52] Chris Whittall. LCH.Clearnet re-values $218 trillion swap portfolio using OIS. Risk
Magazine, June 2010.
[53] M. Cameron. CME and IDCG revalue swaps using OIS discounting. Risk Magazine,
October 2011.
[54] QuantLib, the free/open-source object oriented c++ financial library. Release 0.9.9-
2009, 2009.
81
[56] QuantLibXL, the free/open-source library for exporting QuantLib to microsoft excel.
[58] Andrea Pallavicini and Marco Tarenghi. Interest-Rate Modeling with Multiple Yield
Curves. SSRN eLibrary, 2010.
[59] Masaaki Fujii, Yasufumi Shimada, and Akihiko Takahashi. On the Term Structure
of Interest Rates with Basis Spreads, Collateral and Multiple Currencies. SSRN
eLibrary, 2010.
[60] Masaaki Fujii, Yasufumi Shimada, and Akihiko Takahashi. Collateral Posting and
Choice of Collateral Currency - Implications for Derivative Pricing and Risk Man-
agement. SSRN eLibrary, 2010.
[61] Masaaki Fujii, Yasufumi Shimada, and Akihiko Takahashi. Modeling of Interest Rate
Term Structures Under Collateralization and its Implications. SSRN eLibrary, 2010.
[62] Masaaki Fujii and Akihiko Takahashi. Clean Valuation Framework for the USD Silo -
An Implication for the Forthcoming Standard Credit Support Annex (SCSA). SSRN
eLibrary, 2011.
[63] Antonio Castagna. Pricing of collateralized derivatives contracts when more than
one currency are involved: Liquidity and funding value adjustments. SSRN working
paper, 06 2012.
[65] J.M. Harrison and D.M. Kreps. Martingales and arbitrage in multiperiod securities
markets. Journal of Economic Theory, 20:381–408, 1979.
[66] J.M. Harrison and S.R. Pliska. Martingales and stochastic integrals in the theory of
continuous trading. Stochastic Processes and their Applications, 11:215–260, 1981.
[67] J.M. Harrison and S.R. Pliska. A stochastic calculus model of continuous trading:
Complete markets. Stochastic Processes and their Applications, 15:313–316, 1983.
[68] Ioannis Karatzas and Steven E. Shreve. Brownian Motion and Stochastic Calculus.
Springer, 1997.
82