A Simple Way To Find Prices, Greeks, and Static Hedges For Barrier Options
Uploaded by
seanwu95A Simple Way To Find Prices, Greeks, and Static Hedges For Barrier Options
Uploaded by
seanwu95A Simple Way to Find Prices, Greeks, and
Static Hedges for Barrier Options
Rolf Poulsen
Abstract
We give simple proofs of some pricing and hedging results for barrier options.
Following the probabilistic approach in Andreasen (2001), we prove some pricing and hedging results
for single barrier zero-rebate options. The results are all well-established, but the proofs are selfcontained and simple; no reference to first-hitting-time densities or the joint distribution of Brownian
motion & its maximum is needed, be this explicit (as for instance in Bj
ork (1998)[Chapter 13]) or
implicit (as in Carr & Chou (1997)[Appendix]). All that is needed is a suitable change of measure.
We look at the Black-Scholes model for a dividend-paying stock, i.e. an asset whose price dynamics
are
dS(t) = (r q)S(t)dt + S(t)dW Q (t),
under the equivalent martingale measure Q (with the bank-account as numeraire). Put p = 1
2(r q)/ 2 . Consider a simple claim with a payoff at time T specified by a (suitably nice) payoff
function g (a g-claim for short). Its arbitrage-free time-t price
r(T t)
f (S(t), t),
g (t) = er(T t) EQ
t (g(S(T ))) = e
where of course f (S(t), t) = EQ
t (g(S(T ))), and the Markov property of S ensures that this is nondeceptive notation.
Let B be a constant, define a new function fb
b t) = (x/B)p f B 2 /x, t ,
f(x,
and look at a simple claim with payoff-function gb(x) = fb(x, T ).
Define the process Z by
Z(t) =
This
S(t)
B
p
version: April 9, 2003
of Statistics and Operations Research, Institute for Mathematical Sciences, Universitetsparken 5,
Department
University of Copenhagen, DK-2100, Denmark. Email: [email protected]
Using the Ito formula we get
dZ(t) = pZ(t)dW Q (t),
so Z(t)/Z(0) is a positive, mean-1 Q-martingale. Here the exact form of p is needed, the result does
not hold if isnt constant. (That is, unless it happens that r = q, as is the case in Andreasen
(2001).) Hence
dQZ
Z(T )
=
on FT
dQ
Z(0)
defines a probability measure QZ Q.
Now recall the Abstract Bayes Formula for conditonal means, see ksendal (1995)[Lemma 8.24]
for instance. To recap the theorem: Let and be equivalent probability meausures on a space
(, F), and let f be their Radon/Nikodym-derivative, ie.
d
= f.
d
Let X be a random variable and H a -algebra such that H F. Then
E (X|H)E (f |H) = E (f X|H).
(1)
Now apply the formula with the QZ in the role of , Q as , Z(T )/Z(0) as f , F = FT and H = Ft .
Because Z(t)/Z(0) is a martingale this gives us that the price of the gb-claim can be written as
p 2
B
S(T )
g
b
r(T t) Q
g
(t) = e
Et
B
S(T )
2
p
B
S(t)
QZ
r(T t)
Et
g
= e
.
B
S(T )
Girsanovs theorem (ksendal (1995)[Theorem 8.26]) tells us that
Z
dW Q (t) = dW Q pdt
(2)
defines a QZ -Brownian motion. Put Y (t) = B 2 /S(t). Then the Ito formula and (2) gives us that
(again, the particular form of p is needed)
Z
dY (t) = (r q)Y (t)dt + Y (t) dW Q (t) ,
which means the law of Y under QZ is the same as the law of S under Q. Therefore
Z
(g(Y (T ))) = f (Y (t), t) = f (B 2 /S(t), t),
EQ
t
and the gb-claims price is
gb(t) = er(T t) fb(S(t), t) = er(T t) (S(t)/B)p f (B 2 /S(t), t).
And why is this useful?
Consider a 0-rebate knock-out (at a barrier B) version of a claim with payoff function g (the barrier
2
option in the following). Without loss of generality we may assume that g(x) = 0 for x B (in the
down-and-out case) or g(x) = 0 for x B (in the up-and-out case). Consider a simple claim with
payoff function h = g gb (the adjusted payoff in the language of Carr & Chou (1997)). Note that
h(x) = g(x) if x B (x B for up-and-out). The time-t price of the h-claim is
h (t) = er(T t) f (S(t), t) (S(t)/B)p f (B 2 /S(t), t) .
(3)
We see that if S(t) = B, then the h-claim has a price of 0. Suppose now that we buy the h-claim,
sell it again if the stock price hits the barrier, and if this doesnt happen, simply hold it until expiry.
If the stock price stays above (below) the barrier, we receive g(S(T )) at expiry, otherwise we get 0.
In other words, exactly the same as the barrier option, so we can read off its price directly from (3).
In this way we can establish all formulas from Bj
ork (1998)[Chapter 13.2-3] (possibly trough in-out
parities). We also easily get information about the various partial derivatives of the price wrt. input
variables (often called greeks). For instance we immediately see that the Vega (i.e. price/)
of a knock-out barrier call is lower than that of the plain vanilla call; it could even be negative.
Further, since any simple claim can be statically hedged by a portfolio of plain vanilla puts & calls,
we can also devise static hedges for the barrier option. Note, however, that for a general p, the
h-function is not piecewise linear, so the static hedge portfolio involves a continuum of options. In
fact, h could be discontinuous at the barrier level (this happens for up-and-out calls), in which case
matching the pay-off becomes problematic in practice (unless some sort of digital options exist).
If we do assume that r = q (this could be because were really looking at forward prices, or because
were in an exchange rate setting with equal foreign and domestic interest rates) and consider a call
option, g(x) = (x K)+ , then we find that
gb(x) = (x/B)(B 2 /x K)+ = (K/B)(B 2 /K x)+ ,
which is the payoff of K/B puts with strike B 2 /K. So the knock-out call can be hedged by buying
the plain vanilla call and shorting K/B strike-B 2 /K puts.
References
Andreasen, J. (2001), Behind the Mirror, Risk Magazine 14(November).
Bj
ork, T. (1998), Arbitrage Theory in Continuous Time, Oxford.
Carr, P. & Chou, A. (1997), Breaking Barriers, Risk Magazine 10(September).
ksendal, B. (1995), Stochastic Differential Equations, 4. edn, Springer-Verlag.
