Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
then
d (S a )
=
Sa
a (a 1) 2
a ( ) +
dt + adZ (t)
2
Forward:
a(r)+
F0,T (S a ) = S (0) e
P
F0,T
(S ) = e
rT
a(r)+
S (0) e
a(a1) 2
2
a(a1) 2
2
T
T
r
r
=
C = Sa,
a
a
2
S (t)
t as w +(1a) r+ a
2
W (t) = W (0)
e
S (0)
For
Proportional
dX1
X1
dX2
X2
AT/ADD/Lara/Halfass
dX
r
= ( ) dt + dZ =
Sign of the denominator is important
X
c1 X1 (0) 1 + c2 X2 (0) 2 = 0
= (1 1 ) dt + 1 dZ
then risk free
or
= (2 2 ) dt + 2 dZ
( ) dt + dZ = ( ) dt + dt + dZ =
Cr = CS S (r ) +
Black-Scholes Equation
dt + (dt + dZ) = (r ) dt + dZe
S 2 2 CSS
+
2
Ct+h Ct = 4 +
Delta-gamma-theta approximation
2
+ h
2
ito's lemma:
dC = CS dS +
1 2 2 2
S v
2
CSS (dS)
+ Ct dt
2
No Prot Threshold:
Martingale
If
dS
= ( ) dt + dZ
S
1 2 2 2
S v n
2
Sv
dt's coecient is 0
t,T
dF
Ft,T = ( r) dt + dZ
then
dFPt,T = dt + dZ
F
t,T
Arithmetic Brownian
Geometric
X (t) = X (0) e
Call
Put
Comment
2
dX
= dt + dZ Cov (X (u) , X (t)) = E [u] E [t] e t 1
X
2
2
d ln X (t) =
dt + dZ ln X (t) = ln X (0) +
t + Z
2
2
E [X (t)] = X (0) et
2
2 t+Z
= CS
> 0, N (d1 ) eT
< 0, N (d1 ) eT
S shaped
= CSS
> 0,
Vega
> 0,
same
Symmetric Hump
Stock:
same
Assymetric Hump
4 = 1,
C
C
S
S
<0
usually, dierent
Upside-Down Hump
S
C
S
S
Rho
(r)
> 0,
< 0,
up
Psi
< 0,
> 0,
up
Multiply by
0.01
==
()
down
down
Multiply by
S
S
C = S pay attention to signs of
C
C
X
Stotal total
portfolio: simple sum based on shares Ftotal =
ni Fi vs Elasticity for portfolio:
Ctotal
1, Sx
2, x
Estimate Annual :
S2
3, Annual = x
+ x n
Elasticity: Instantaneous
Greek for
r
r
Theta
0.01
PtEuropean
+ CTAmerican CtEuropean
= CallonPut
1
1
then
U=
S8
S4
8
S4
S2
12
S2
S0
14
whose strike is
S014 ln U = 8 ln
D K 1 e(T t1 )r
S8
S4
S2
+ 12 ln
+ 14 ln
+ 14 ln S0
S4
S2
S0
Var (ln U )
)=
Correlation
Var (aX
Var (ln X
+ bY ) = a
(
x xi ) (
y yi )
==
Var (X)
+b
Cov (X, Y
x y
Var (Y
) + 2abCov (X, Y )
2
ln Y ) = Var (ln X) + Var (ln Y ) 2Cov (ln X, ln Y ) 2 = X
+ Y2 2X Y
ev
d
p
ev
Forward
Lognormal/Jarrow-Rudd
Future
2
r 2 h+v
e
2
r 2 hv
ev
e(r)h+v
e(r)hv
e
e(r)h d
p = ud
u d
u Cd h
= C
and =
SuSd e
SuSd
d
1d
p = ud
Cu Cd
4 = F uF d
ev
and
1d
1
p = ud
= 1+u
rh
B=C=e
(p Cu + (1 p ) Cd )
Call
Put
American=European
American>European
CEU 0, T, Ke
U = Called
(
Qu = ptrue Uu
Qd = (1 ptrue ) Ud
Option
r(T t)
CEU (0, t, K)
(Qu + Qd ) (1 + reective ) = 1 p =
Qu
S0 = Qu Su + Qd Sd
Qu + Qd
if C = C (S | S > K)
if C = C (S | S < K)
then
> K) = N (d2 )
and
and
4et = N (d1 ) +
1
e v 2 2
q
1
2
ed1 v 2 2
d2
1
4et = N (d1 )
q
4 can be
1
rt
and 4e
=K
e 2 S 2 v 2 2
1
rt
12
X
ui
or
z = N 1 (u)
where
Start option:
uniform, then
more than
K = S0 e(r)t
e+z
i=1
Anthetic:
uses
Stratied Sampling:
Break down into intervals, then scale into them in the given order
Non-Control:
Control:
Control Variate:
Boyle Modication:
Y
X
Y = y + (E (x) x
) Var (Y ) = Var (
y ) + Var (
x) 2Cov (
x, y)
Y = y + a (E (x) x
) Var (Y ) = Var (
y) + a
Var (
x)
2aCov (
x, y)
1
n1
where Cov (
x, y)
where
a=a
=r
in TI
and Cov (
x, y)
=r
in TI
Variance/Covariance used, if not given by the calculator, are just the naive variance, so they must be multiplied with
a=
Var (
x)
Var (Y
Take
Quadratic variation:
Cov(X,Y )
Cov (
x, y)
Var(X)
n
Cov (X, Y
Var (Control)
) = Var (
y ) 1 2XY
2
n
X
iT
(i 1) T
lim
X
X
n
n
n
i=1
(dX (t))
or
T
2
0 (dt) = 0
2
T
dt2 = 0
0
2
T
dZ 2 = T
0
and
A Brownian motion has the innite crossing property: the path corsses its starting point innitely often with prob
Ornstein-Uhlenbeck
1
n1
>1
Y (t) = X (t)
dY = Y () dt + dZ
Y = Y0 et + 0 e(st) dZ
1+K
A caplet is equivalent to
1
1+K
dr = ardt + rdZ
P =
1
2
Rendleman
If
(
a(T t)
dr
=
a
(b
r)
dt
+
dZ
Vasicek
B = 1e a
For Vasicek
Br
P = Ae
For
and
1
r1
dr = a (b r) dt + rdZ CIR
2 = r2 with same T t
(
(
=
In general
2
dP
= a (b r) B + 21 (r) B 2 + PPt For Vasicek
= dt qdZ where
dr = a (r) dt + (r) dZ (t) then
P
q = PPr (r)
In general
q = B (r) and B = Pr
For Vasicek
P
(r) =
r
,
q
and CIR
and CIR
Risk Neutral:
b +
1 2
a 2 a
Yield-to-maturity on innity
2ab
2ab
a+a
2a2
BPS/FOAM/
2ab
a+
(a)2
2ab
a+
(a)2 +2 2
for CIR
(SOA)2
for Vasicek
2
e x / 2
e x / 2
e x / 2
f Z ( x) = N ( x) =
=
=
,
2
2 3.14159 2.50663
< x < .
Let Y be a lognormal random variable. Assume that ln(Y) has mean m and standard
deviation v. Then, the density function of Y is
fY ( x ) =
1 ln( x) m 2
exp
,
2
v
xv 2
x > 0.
ln( x) m
FY ( x) = N
,
v
x > 0.
Also,
1
k
E[Y ] = exp km + k 2v 2 ,
2
which is the same as the moment-generating function of the random variable ln(Y)
evaluated at the value k.