1.

Put-call parity

C(K, T ) − P(K, T ) = e−rT (F0,T − K) = F0P − Ke−rT

P = S e−δT , so C(K, T ) − P(K, T ) = S e−δT − Ke−rT .
Continuous: F0,T 0 0

Discrete: F0,T = S 0 − dt e = S 0 − PV(dividends), so C(K, T ) − P(K, T ) = S 0 − dt e−rt − Ke−rT .

P P −rt P

C(S ,K)−P(S ,K) Ke−rT Ke−rT

To create one unit of synthetic stock: S 0 = e−δT
+ e−δT
, so buy 1
e−δT
calls, sell 1
e−δT
puts, buy e−δT
bonds.

2. Comparing options
∂C ∂P
For a Call: higher strike =⇒ lower price. For a Put: higher strike =⇒ higher price. −1 ≤ ∂K ≤ 0 and 0 ≤ ∂K ≤ 1.
American options are worth more than European.

S ≤ C Amer (S , K, T ) ≤ C Eur (S , K, T ) ≤ max(0, F0,T

P
− Ke−rT )
K ≤ PAmer (S , K, T ) ≤ PEur (S , K, T ) ≤ max(0, Ke−rT − F0,T
P
)

The only rational time to exercise an American call option early is just before a dividend. If no dividends, then never exercise early.
Early Call exercise may be rational if PVt,T (Div) ≥ K(1 − e−r(T −t) ). Early Put exercise may be rational if Call + PVt,T (Div) ≥ K(1 − e−r(T −t) ).
An American option with exercise time T is ≥ one with exercise time t < T .
A European call option on a non-dividend stock with exercise time T is ≥ one with exercise time t < T .
A European option on a non-dividend stock with strike price Ker(T −t) is ≥ one with strike price K.

3. Binomial trees — stock, one period

Option payoff at upper, lower nodes: Cu , Cd . Must have d < F = S e(r−δ)h < u to prevent arbitrage.
Replicating portfolio: solving S u∆ + er B = Cu and S d∆ + er B = Cd gives

Cu − Cd −δh uCd − dCu −rh e(r−δ)h − d

# shares to buy: ∆= e , # bonds to buy: B= e , risk-neutral probability: p∗ =
S (u − d) u−d u−d

These formulas can also be used to pull back the replicating portfolio (∆, B) at each node.
√
For a tree built using forward rates: u/d = e(r−δ)h±σ h
. Option price: C = S ∆ + B = e−rh (p∗ Cu + (1 − p∗ )Cd )
To pull back to previous node, discount with e−rh , not with the e−(r−δ)h that appears in the expression for p∗ !

4. Binomial trees — general

Pn n

For European trees, use binomial theorem shortcut: k=0 k pk (1 − p)n−k .
For American options, compare pulled-back value to exercise value (S node − K) ∧ 0 or (K − S node ) ∧ 0.
For currency options, use δ = rforeign .
1−d Cu − Cd
Futures: p∗ = , ∆= , B = C = e−rh (P∗ Cu + (1 − P∗ )Cd
u−d Fu − Fd
For computing the forward price, F = S e(r−δ)T where T is the length of the contract (nothing to do with the period h of the tree).
Stock/index Currency Futures Commodity Bond
δ dividend yield rforeign 0 lease rate coupon yield
risk-free rate r rdomestic r r r

5. Risk-neutral pricing

5.1. Pricing with true probabilities. α is the expected return on a stock, and γ is the corresponding discounting rate for the option α > r.
γ is also called the compound annual return for the option. For puts, usually γ < 0.
p∗ is not really the probability that the stock increase in value. It is the probability that makes α = r.
e(α−δ)h − d S ∆ αh B
p= eγh = e + erh Ceγh = S ∆eαh + Berh = pCu + (1 − p)Cd
u−d S∆ + B S∆ + B

1
5.2. Risk-neutral pricing. Ui is PV(1) for stock in state i. Qi is conditional value of PV(1), so Qu = pUu and Qd = (1−p)Ud and Qu + Qd = .
1+r
p 1− p
Current value of a stock: C0 = Qu Cu + Qd Cd = pUu Cu + (1 − p)Ud Cd If risk-neutral, Qu = , Qd =
1+r 1+r
pCu + (1 − p)Cd pCu + (1 − p)Cd
Effective annual rate of return: 1 + α = =
pUu Cu + (1 − p)Ud Cd Qu C u + Qd C d
Qu pUu Qd
Risk-neutral probabilities: p∗ = = = Qu (1 + r) 1 − p∗ =
Qu + Qd pUu + (1 − p)Ud Qu + Qd
1
 2

6. Binomial trees: miscellany

For early exercise to be optimal: div − int ≥ put, or S (1 − e−δt ) − K(1 − e−rt ) ≥ Put(S , K)
√
Standard tree centered on (r − δ)h, so u/d = e(r−δ)h±σ h
. Based on forward prices.
√ √
Cox-Ross-Rubenstein tree centered on 1, so u/d = e ±δ h
. Allows arbitrage iff eσ h < e(r−δ)h .
σ2
√ σ2
√
Lognormal tree centered on (r − δ − σ2 )h, so u/d = e(r−δ−
2
2 )h±σ h
. Allows arbitrage iff e(r−δ− 2 )h+σ h
< e(r−δ)h .
Also called Jarrow-Rudd. Makes p very close to 0.5.
Estimating volatility: given stock prices S ti for i = 0, 1, . . . , n, compute τi := ln(S ti /S ti−1 ), for i = 1, . . . , n. Then
n n √
1X 1X 2 n
x̄ = τi and µ2 = τ and s2 = (µ2 − x̄2 ) −→ unbiased annualized variance = s/ h
n i=1 n i=1 i n−1

7. Modeling stock prices with the lognormal distribution

Weaknesses/assumptions: (i) constant

 volatility, (ii) stock
 returns for different periods are independent, (iii) stock prices don’t jump.
σ2 σ2
For S t lognormal, ln(S t /S 0 ) ∼ N (α − δ − 2 )t, σ t
2 , so the annual continuously compounded mean return is α − δ − 2 .

σ2
ln(S 0 /K) + (α − δ − 2 )t
Pr(S t > K) = N(d̂2 ) and Pr(S t < K) = N(−d̂2 ) , where d̂2 has α in place of r: d̂2 = √
σ t
σ2 σ2
Median stock price = S 0 e(α−δ− 2 )t
= E(S t )e− 2 )t
< E(S t ) = S 0 e(α−δ)t = mean stock price.
σ2 √ σ2 √
)t+σ t·N −1 (p/2) )t+σ t·N −1 (1−p/2)
To have S L ≤ S t ≤ S U with probability p: S L = S 0 e(α−δ− 2 and S U = S 0 e(α−δ− 2

N(d̂1 ) N(−d̂1 )
Conditional expected stock price: E[S t |S t > K] = S e(α−δ)t and E[S t |S t < K] = S e(α−δ)t
N(d̂2 ) N(−d̂2 )

Expected payoff on a Call: E[max{0, S t − K}] = S e(α−δ)t N(d̂1 ) − KN(d̂2 ) Put: E[max{0, K − S t }] = KN(−d̂2 ) − S e(α−δ)t N(−d̂1 )
R∞ d̂1 ) d̂1 )
C = e−rt K (S t − K)g∗ (S t ) dS t = e−rt E∗ [S − K|S > K]P∗ [S > K] = S e(α−r−δ)t N(
N(d̂ )
N(d2 ) − Ke−rt N(−
N(−d̂ )
N(−d2 ) = B-S, for α = r.
2 2

8. Fitting stock prices to a lognormal distribution

Estimate α (with dividends removed) with

s
σ2
!
√ n 1X 2
α̂ = µ̂ + µ̂ = p x̂ σ̂ = p xi − x̄2
2 n−1 n

where n is number of observations and p is number of periods per year and xi = ln(S i /S i−1 )

9. The Black-Scholes formula

Assumptions:
•Continuously compounded returns on the stock (δ) are normally distributed and independent over time.
•Continuously compounded returns on the strike asset (r) are known and constant.
•Volatility is known and constant.
•Dividends are known and constant.
•There are no transaction costs or taxes.
•It is possible to short-sell any amount of stock and to borrow any amount of money at the risk-free rate.

ln F P (S )/F P (K) + σ2 T
  2
√
General Black-Scholes C(S , K, σ, T, r, δ) = F P (S )N(d1 ) − F P (K)N(d2 ) , where d1 = √ and d2 = d1 − σ T .
σ T
      √
ln KS + r − δ + σ2 T and d2 = d1 − σ T and
2
For stocks, F P (S ) = S e−δT and F P (K) = Ke−rT , so d1 = √ 1
σ T

Call(S , K) = S e −δT
N(d1 ) − Ke −rT
N(d2 ), Put(S , K) = Ke−rT N(−d2 ) − S e−δT N(−d1 ).

For currency, let rf = rforeign and let x be the amount of foreign currency to buy. Then δ = rf and

Call(x, K) = xe−rf T N(d1 ) − Ke−rT N(d2 ), Put(x, K) = Ke−rT N(−d2 ) − se−rf T N(−d1 ).
    √
ln KS + σ2 T . When K = S , this becomes d1 = σ T /2. The Black formula is:
2
For futures, δ = r, so d1 = √ 1
σ T

Call(x, K) = Fe−rT N(d1 ) − Ke−rT N(d2 ), Put(x, K) = Ke−rT N(−d2 ) − Fe−rT N(−d1 ).
 3

10. The Black-Scholes formula: Greeks

∆ ∆ put = ∆call − e−δT S-shaped curve
Γ Γ put = Γcall symmetric hump, peak to left of stock price, further left with higher T
vega vega put = vegacall asymmetric hump, peak like Γ
θ θ put = θcall + 365
1
(rKe−rT − δS e−δT ) Sulcus for short lives, gradual decrease for long lives.
ρ ρ put = ρcall − 100 Ke−rT
T
increasing curve (pos for calls, neg for puts)
Ψ Ψ put = + 100
T
S e−δT Ψcall decreasing curve (neg for calls, pos for puts)
To convert between Put and Call greeks, differentiate both sides of C − P = S e−δt − Ke−rt with respect to the appropriate variable.
θ is almost always negative for calls; negative for puts unless far in the money.
A call option can be replicated by buying ∆ shares of stock and borrowing Ke−rT N(d2 ). ∆call = e−δT N(d1 ) and ∆put = −e−δT N(−d1 ) = ∆call − e−δT
S∆
= ε∆/C
%option change
Elasticity of the option with premium C is Ω = ε/S = %stock change = percentage risk. Dollar risk is ≈ ε∆.
C
For calls, Ω ≥ 1 because ∆ ≥ 0 and S ∆ = S e−δT N(d1 ) > C(S ). For puts, Ω ≤ 0 because ∆ ≤ 0.
Volatility of the option is σoption = σstock |Ω| .
γ−r α−r
Sharpe ratio is = . γ − r = Ω(α − r) is true instantaneously, and follows from eγh = erh + Ω(eαh − erh ).
σoption σstock
X
For a portfolio consisting of ai of options Ci : A greek for the porfolio is computed by ∆portfolio = ai ∆Ci .

S ∆portfolio S ai ∆Ci
P
Elasticity for the porfolio is computed by Ωportfolio = = P .
Cportfolio ai Ci

11. The Black-Scholes formula: applications, implied volatility

Calendar spread Sell C(S , K, t) and buy C(S , K, T ), with t < T (or buy C(S , K, t) and sell C(S , K, T )).
If Ct << C0 , then loss on C(S , K, T ) outweighs profit on sale of C(S , K, t).
If Ct >> C0 , then obligation from sale of C(S , K, t) outweighs profit when exercising C(S , K, T ).
To find holding period profit/calendar spread profit for [0, t], where 0 < t < T i , use BS to compute
n
X m
X
(portfolio value at time t) − (portfolio value at time 0)ert = Ci (S , K, T i ) + (cash settlements) − C j (S , K, t)ert . (1)
i=1 j=1

Implied volatility (i) Allows pricing other options on the same stock, without market prices. (ii) Is a quick way to describe option prices. (iii)
Volatility skew measures accuracy of Black-Scholes model. Volatility skew: implied volatility tends to be lower for high strike prices.

12. Delta hedging

If a market-maker sells an option, he buys ∆ of the stock to hedge so there will be no profit or loss if the stock price changes.

12.1. Overnight profit. Ignoring dividends, profit is change in option value & ∆ (change in stock price) & interest on borrowed money:
 
market-maker profit = − (C(S t ) − C(S 0 )) + ∆(S t − S 0 ) − ert/365 − 1 (∆S 0 − C(S 0 )), where t is in days (usually t = 1). This is a special case of (1):

√ √
market-maker profit = (∆S t − C(S t )) − ert/365 (∆S 0 − C(S 0 )) which is positive iff S − σ t < St < S + σ t .

12.2. Delta-gamma-theta approximation. C(S t + ε) = C(S ) + ∆ε + 12 Γε2 + θt + Taylor remainder error.

√ √
The whole point of ∆-hedging is to separate ε (or σ h) from ∆. For ε = σ h, get
ε σ 2
" 2 # " 2 #
market-maker profit = − Γ + rh(S ∆ − C(S )) + θh = − S Γ + r(S ∆ − C(S )) + θ h (2)
2 2

12.3. Rehedging. Buy ∆t − ∆0 of the stock, where ∆t = N(d1 ) is computed at time t, and ∆0 = N(d1 ) computed at time 0. If negative, sell.
Let Rh,i be period-i return to a delta-hedged market maker who has written a call. Then

σ2 2 2  2  2
Rh,i = S (xi − 1)Γh, and Var(Rh,i ) = 1
2 σ2 S 2 Γh and annual variance = 1
2 σ2 S 2 Γ h
2

12.4. Hedging multiple greeks. Sell option Csell , then ∆-Γ hedge by buying x1 shares of stock and buying x2 of another option Cbuy :

stock Cbuy Csell

stock Csell
Delta-gamma hedging ∆ : x1 + ∆buy x2 = ∆ sell compare to Delta-hedging
∆: x1 = ∆ sell
Γ: Γbuy x2 = Γ sell
 4

13. Asian and barrier options

q
13.1. Asian options. Arithmetic average A(S ) = S t and geometric average G(S ) =
1 PT T QT
T t=1 t=1 S t . Ignore initial price (exclude S 0 )

arithmetic geometric
average price Call = max(0, A(S ) − K), Put = max(0, K − A(S )) Call = max(0, G(S ) − K), Put = max(0, K − G(S ))
average strike Call = max(0, S T − A(S )), Put = max(0, A(S ) − S T ) Call = max(0, S T − G(S )), Put = max(0, G(S ) − S T )
Asian is cheaper than European, since less volatile. Similarly, average over more items is cheaper.
Daily average price < monthly average price. Monthly average strike < daily average strike.
G(S ) ≤ A(S ) =⇒ Geometric average price call < arithmetic average price call. Reverse inequality for puts, also reverse for average strikes.

13.2. Barrier options. Rebate options: pay a fixed amount when the barrier is hit.
Parity: knock-in option + knock-out option = ordinary option .

13.3. Compound options. For binomial tree models, work out the binomial tree for the underlying first. Then for the compound option, work out
a second tree with initial vertices given by the prices of the underlying.

13.3.1. Compound option parity.

CallOnCall(C(S , K, T ), x, t) − PutOnCall(C(S , K, T ), x, t) = C(S , K, T ) − xe−rt

CallOnPut(P(S , K, T ), x, t) − PutOnPut(P(S , K, T ), x, t) = P(S , K, T ) − xe−rt

13.3.2. American options on stocks with one discrete dividend. If a dividend D is paid at time t, then value of an American call at time t is greater
of exercise value and option for remaining period: max{S t + D − K, C(S t , K, T − t)}. If S t is cum-dividend and S t0 = S t − D is ex-dividend, then

time t : S t0 + D − K + max{0, P(S t0 , K, T − t) + K(1 − e−r(T −t) ) − D} =⇒ time 0 : S 0 − Ke−rt + CallOnPut[S , K, D − K(1 − e−r(T −t) )] .

Early exercise is not rational if x = D − K(1 − e−r(T −t) ) < P(S t0 , D − K(1 − e−r(T −t) )) . For P, use F P (S ) = S 0 − De−rt and F P (K) = Ke−rT .

14. Compound, gap, and exchange options

Option S |S > K S |S < K c|S > K c|S < K

14.1. All-or-nothing options.
Value S e−δT N(d1 ) S e−δT N(−d1 ) Ke−rT N(d2 ) Ke−rT N(−d2 )

14.2. Gap options. K splits into Kstrike and Ktrigger . Occurrence of payoff is determined by Ktrigger , so use it to determine the probabilities N(di ).
Amount of payoff is determined by Kstrike , so use it to compute option price:
Call = S e−δt N(d1 ) − Kstrike e−rt N(d2 ), Put = Kstrike e−rt N(−d2 ) − S e−δt N(−d1 ).
    √
ln(S /Ktrigger ) + r − δ + σ2 T and d2 = d1 − σ T . Decompose gap in terms of AoNs:
2
where d1 = 1
√
σ T

C(S , Kstrike , Ktrigger ) = C(S , Ktrigger ) + (Ktrigger − Kstrike )|S > Ktrigger P(S , Kstrike , Ktrigger ) = P(S , Ktrigger ) + (Ktrigger − Kstrike )|S < Ktrigger

14.3. Exchange options. Let S be the asset you receive, with dividend rate δ1 , and K be the asset you may exchange for it, with dividend rate δ2 .
The volatility of S − K is σ2 = σ2S + σ2K − 2ρσS σK

14.4. Chooser options. V = C(S , K, T ) + e−δ(T −t) P(S , Ke−(r−δ)(T −t) , t)

14.5. Forward-start options. Let diT −t be di computed with time T − t instead of t. Then
discount to t=0
S t e−δ(T −t) N(d1T −t ) − S t e−r(T −t) N(d2T −t ) −−−−−−−−−−−−−−−−→ S 0 e−δT N(d1T −t ) − S 0 e−r(T −t)−δT N(d2T −t )
Z di (S ) −(x)2 /2
e−(di ) /2 ∂di e−(di ) /2 e−(di ) /2
2 2 2
∂N(di ) ∂
! !
e 1
14.6. For hedging: differentiate using = √ dx = √ = √ √ = √ .
∂S ∂S −∞ 2π 2π ∂S 2π S σ T S σ 2πT
15. Monte Carlo valuation
−1 (u)
To simulate a lognormal random variable, let u ∼ U(0, 1) be uniform. Then N −1 (u) ∼ N(0, 1) and eN is lognormal.
n
e−rT X
If V(S T , T ) is option payoff at T , the Monte Carlo time-0 price is V(S 0 , 0) = V(S Ti , T )
n i=1

n
e−rT X √
( )
σ2
For a European call, this would be C = max 0, S 0 e(r−δ− 2 )T +σ T Zi − K , where Zi ∼ N(0, 1) for i = 1, . . . , n.
n i=1
 P 1/2 sn
x̄ is sample mean and σC = sn = 1 n
n i=1 (xi − x̄)2 is sample stdev for one draw. Then σn = √ is stdev of the Monte Carlo estimate.
n

To attain a given target standard error of σn , need n = (sn /σn )2 trials.

 5

16. Brownian motion

At is stock price at time t, α is continuous rate of return, δ is continuous dividend rate, σ is volatility, N(x) is the normal (cumulative) distribution
function. So total drift is α − δ =“capital gains return”=“contin. compounded expected incr.”.
If α or σ is given in terms of a time unit, use this to denominate time (e.g., σ = 2 per quarter =⇒ 1 year would be t = 4).
(t)
 (t) 
For stock to exceed a continuously compounded annual return (yield) of y means SS (0) ≥ eyt , or ln SS (0) ≥ yt.

16.1. Arithmetic BM:. X(t) = αs + σZ(t). The increment X(t + h) − X(t) has mean µ = (α − δ)h and var= σ2 h, so
At+h − At − µ
!
P[X(t + h) < At+h |X(t) = At ] = N √ , dX(t) = α dt + σdZ(t)
σ h

16.2. Geometric BM:. Y(t) = eX(t) . The increment Y(t + h) − Y(t) has mean µ = (α − δ − σ2 )h and var= σ2 h, so
2

Y(t + h) ln At+h − ln At − µ
" # !
At+h
P[Y(t + h) < At+h |Y(t) = At ] = P < |Y(t) = At = N √ , dY(t) = αY(t) dt + σX(t)dZ(t)
Y(t) At σ h
σ2
To go from geometric to arithmetic, subtract 2 and replace X(t) with ln X(t) .
Covariance. For standard BM: hZ s , Zt i := Cov(Z s , Zt ) = min{s, t}.
σ2
For Xt = X0 eµt+σZt , (eσ s ,Zt i
2 hZ
For Xt = X0 + αt + σZt , hX s , Xt i = σ2 hZ s , Zt i. hZ s , Zt i = X02 e(µ+ 2 )(s+t)
− 1).

17. Differentials

Watch for variance (σ2 ) given instead of volatility (σ). Var(ln S (t)|S (0)) = Var(ln F0,T (S )) = Var(ln F0,T
P (S )) = σ2 t.

  σ2
σ2
ln X(t)|X(0) ∼ N X(0) + (ξ − 2 )t, σ t
2
X(t) = X(0)e(ξ− 2 )t+σZ(t)

σ2 σ2
ln X(t) − ln X(0) = (ξ − 2 )t + σZ(t) d (ln X(t)) = (ξ − 2 ) dt + σdZ

18. Itô’s lemma

dC = CS dS + 12 CS S (dS )2 + Ct dt. If dS is arithmetic BM, (dS )2 = σ2 dt. Don’t forget σ2 !

(dt)2= 0, dt × dZt = 0, (dZt = dt, dZt × dZt0 = ρdt, where ρ is correlation coefficient.
)2
Ornstein-Uhlenbeck process: dXt = λ(α − Xt ) dt + σ dZt , where λ is speed of reversion to mean α.

19. The Black-Scholes equation

σ2 σ2 2
Set (2) equal to 0 to obtain: 2 CS S S
2
+ rCS S + Ct = rC(S ) or 2 S Γ + rS ∆ + θ = rC(S )
Use it to price a claim C or to determine the parameters of a derivative security.

20. Sharpe ratio

  α(t, X) − r
Express process: dX
= α(t, X) − δ(t, X) dt + σ(t, X)dZ. Then φ = .
X σ(t, X)
dS 1 (t) 2 (t)
Risk-free portfolios: buy x1 of S 1 (t)= α1 dt + σ1 dt and buy x2 of dS S 2 (t) = α2 dt + σ2 dt, to solve either of:
 
α1 S 1 (0)x1 + α2 S 2 (0)x2 = S 1 (0)x1 + S 2 (0)x2 r or σ1 S 1 (0)x1 + σ2 S 2 (0)x2 = 0.

21. Risk-neutral pricing and proportional portfolios

α−r
Risk-neutral process for stocks (Girsanov’s theorem): let φ = σ be the Sharpe ratio, so dZ = dZ + φ dt is arithmetic BM. Then
dS (t) dS (t)
geometric BM = α dt + σ dZ 7→ risk-neutral BM = r dt + σ dZ
S (t) S (t)
!η 2
 
dW(t)
  W(t) S (t) ηδS −δW +(1−η)(r+η σ2 t
Blended portfolio: W(t) = ηα + (1 − η)r dt + ησ dZ. Then = e
W(0) S (0)

22. Monomial securities Sa

σ2 σ2
The process is S ta = S (0)a ea(α−δ− 2 )t+σaZt and its expected value is E S (T )a = S (0)a ea(α−δ+(a−1) 2 )T
 

σ2 σ2
Forward price is F0,T S (T )a = S (0)a ea(r−δ+(a−1) 2 )T . Prepaid forward price is F0,T P
S (T )a = e−rT S (0)a ea(r−δ+(a−1) 2 )T

d(S a ) σ2
!
The Itô process C = S a is given by a
= a(α − δ + (a − 1) )T dt + aσdZ.
S 2
Sharpe ratios γ−r α−r
aσ = σ show that C = S earns γ = a(α − r) + r. Then ln F 0,T (S ) = ln E [S ] − γ relates the above formulas.
a P a a

Use α for S ta and E [S (T )a ]; use r for F0,T (S (T )a ) and F0,T

P (S (T )a ). For options on S a , use σ̂ = aσ in BS .
 6

23. Stochastic integration

∂
R t 
For differentiating an integral: ∂t 0
f (s)dZ s = f (t) dZt . Don’t forget the dZt !
R
To solve dXt + A(t)Xt dt = B(t) dt, use integrating factor ρ = e A(t) dt
.
n 
X 2
Quadratic variation: for [0, T ], let ti = i
nT, so equally spaced increments. Then QV(X, 0, T ) = lim X(ti ) − X(ti−1 ) .
n→∞
Rt i=1
Ornstein-Uhlenbeck: dXt = λ(α − Xt )dt + σ dZt ↔ Xt = X0 e−λt + α(1 − e−λt ) + σ 0
eλ(s−t) dZt .

24. Binomial trees for interest rates

gggg ruu ddddddd e

rates prices −ruu
gUgggg
U
hh ru
UUUU e−ru
e−(ru +ruu ) + e−(ru +rud )
 
hhhh UUU r YYYYYY
hhhh ddddddd
2
r∅ VVVVV =⇒ YY e−rud
iiii rdu
ud e−r∅
VVVV e−rn ZZZZZZZ fffff e−rdu
PQ
V r Wiiiii f ff
|Paths|
d WWW
WWWWW e−rd
e−(rd +rdu ) + e−(rd +rdd )
 
rdd
2 ZZZZZZZ
e−rdd

24.1. Generic (nonBDT) interest rate trees. Given a continuously compounded interest rate tree with entries r (or prices P(t, t + 1) = e−r ),
X Y
P(0, T ) = P(γ) e−rn , where rn are the rates appearing in the path γ (3)
γ∈Paths n∈γ

To obtain the continuously compounded yield to maturity, solve P(0, T ) = e−rT for r, where T =number of periods (columns).
To compute the premium of a call at time t, strike K: compute (P(0, T ) − K) ∧ 0 for column t. Then walk back by average & discount.
Prices at the end of the tree are just given by e−r ; for intermediate prices, use (3).
For American options: immediate exercise at a node with rate r is worth e−r , so exercise a call early if e−r − K > (pulled-back value).

1 1  1
24.2. Black-Derman-Toy trees. Use P j (t − 1, T ) = P j (t, T ) + P j+1 (t, T ) to walk back, and P = to convert.
1 + R j (t − 1, t) 2 1+R

√
g R2t e4σ t
gggggg
BDT tree Construct with: !
√ P(0, 2) 1 1 1
XXXXXX = +
ff Rt e2σ t
fffff
√
X P(0, 1) 2 1 + R1 1 + R1 e2σ t
XfXXXXX
R0 f
√
XXXXX ffffffff R2t e
2σ t
Ratio in column t: √
Rt YYYYYY
rate j
YYYY = e2σ t
rate j+1
R2t
 
R j (t,T )
To get volatility for yields, use σ = 1
2 ln R j+1 (t,T ) from column ratio. For t , T − 1, compute rates from P j (t, T )
To compute P(0, T ), start with prices P j (T − 1, T ) = 1
1+R j (T −1,T ) in column T . Then walk back by average & discount.
standard
To compute standard F0,T , start with discounted rates R j (T − 1, T ) = R
1+R in column T . Then walk back by average & discount.
standard
standard P(0, T ) P(0, T ) − P(0, T + 1) F0,T
RA = −1= =
P(0, T + 1) P(0, T + 1) P(0, T + 1)
Eurodollar
To compute Eurodollar-style F0,T , start with rates R j (T − 1, T ) = 1
P j (T −1,T ) − 1 in column T . Then walk back by average & discount.

 
Equivalently, start with discounted prices 1 1
1+R j (T −2,T −1) 2 R j (T − 1, T ) + R j+1 (T − 1, T ) in column T − 1 and walk back by average & discount.
Eurodollar
Eurodollar F0,T
RA =
P(0, T )

The forward rate RA for an FRA (=“interest rate for computing payoff”) is the rate that makes the expected value of the FRA equal to 0.
!T −t
1
If R(t, T ) is the rate of a loan starting at t and ending at T , then FRA pays R(t, T ) − RA at time T , or (R(t, T ) − RA ) at time t.
1 + R(t, T )

25. The Black formula

P(0, T )
25.1. Bond options. Black formula for bond options: Ft = F0,t [P(t, T )] = is the price at t to purchase a bond maturing at T .
P(0, t)
Call : C(Ft , K) = P(0, t)[Ft · N(d1 ) − K · N(d2 )] ln(F/K) + σ2 √
2 t
d1 = √ , d2 = d1 − σ t (4)
Put : P(Ft , K) = P(0, t)[K · N(−d2 ) − Ft · N(−d1 )] σ t
 7

25.2. Caps via the Black formula. Each caplet is (1 + K) puts with strike 1
1+K . A Put with strike 1
1+K and exercise time t has value P(Ft , 1
1+K ).
h i
Cap price = (1 + K) P0 + P(F1 , 1+K
1
) + · · · + P(FT , 1+K
1
) Caplet price = (1 + K)P(Ft , 1
1+K ) from (4)

where P0 = ( bond price

1
at 0 − 1) ∧ 0 = R0 ∧ 0 is the initial payoff, and P(F t ,
1
1+K ) is computed using F0,t and σt .
P(0, t)
The strike K is constant throughout, but Ft = and t changes for each caplet . Remember to discount by P(0, t)!
P(0, t − 1)

26. Equilibrium interest rate models

26.1. The impossible model. Use continuously compounded interest P(0, T ) = e−rT .
t1 P(0, t1 )
Hedge ratio for duration-hedge: −N = − . (< 0 =⇒ sell), N = Nt2 =number of t2 -bonds
t2 P(0, t2 )

26.2. Equilibrium models. P = P(r, T ) is the price of a zero-coupon bond when the short-term rate is r, and dr = a(r) dt + σ(r) dZ.
1 σ2 1
dP = αP dt + qP dZ, α = α(r, t, T ) = (aPr + 2 Prr + Pt ) q = q(r, t, T ) = σPr
P P
The Sharpe ratio is

α(r, t, T ) − r σ2 a σPrr Pt − rP
!
1
φ(r, t) = = aPr + Prr + Pt − rP = + +
q(r, t, T ) σPr 2 σ 2Pr σPr

Risk-neutral process for bonds: short-term interest rates dr = a dt + σ dZ 7→ risk-neutral interest rates dr = (a − σφ) dt + σ dZ
Black-Scholes equation: σ2 S 2 Γ + rS ∆ + θ = rC(S ) σ2
2
7→ 2 P rr + (a − σφ)P r + Pt = rP

26.2.1. Rendelman-Bartter: dr = ar dt + σr dZ . Advantages: r ≥ 0, vol ∼ r. Disadvantages: no mean reversion, r unbounded.

26.2.2. Vasicek: dr = a(b − r) dt + σ dZ . Advantages: mean reversion. Disadvantages: vol is constant, can have r < 0.

2 σ2 σ σ2
P(r, t, T ) = Ae−rB , where A(t, T ) = er[B−(T −t)]−B 4a , B = B(t, T ) = a1 (1 − e−a(T −t) ) , r =b+ φ− 2 . (5)
a 2a

Also, B(t, T ) = āT −t a and r is the formula for the yield to maturity on an infinitely-lived bond. The price model (5) satisfies

σ2
Prr + (a(b − r) − σφ)Pr + Pt = rP, for Pr = ∆ = −BP and Prr = Γ = B2 P.
2
2 σ2 B3
For the case a = 0: dr = σdz, r̄ is undefined, B = T − t, and A = exp(σφ B2 + 2 3 ).

√
26.2.3. Cox-Ingersoll-Ross: dr = a(b − r) dt + σ r dZ . Advantages: mean reversion, vol ∼ r, r ≥ 0.

P(r, t, T ) = Ae−rB ,
#2ab/σ2
2γe(a+ϕ+γ)(T −t)/2 2(eγ(T −t) − 1)
"
where A = A(t, T ) = , B = B(t, T ) = ,
(a + ϕ + γ)(eγ(T −t) − 1) + 2γ (a + ϕ + γ)(eγ(T −t) − 1) + 2γ

2ab
and γ = (a + ϕ)2 + 2σ2 and ϕ = φ(r, t)σ(r, t)/r and r =
p
is yield to maturity on an infinitely-lived bond.
a+ϕ+γ

26.2.4. Facts for Vasicek and CIR models.

Higher volatility =⇒ lower yield. “Instantaneous rate of change” = drift term of dr.
For CIR: higher risk premium ϕ =⇒ lower yield (same for Vasicek when a = 0).
A = A(t, T ) = A(T − t) and B = B(t, T ) = B(T − t) depend only on T − t.
σPr
∆ = Pr = −BP and Γ = Prr = B2 P , so q = P =⇒ q(r, t, T ) = −σ(r)B .
σ2 Pt α−r
Vasicek: α(r, t, T ) = −a(b − r)B + 2 B2 + P and the Sharpe ratio φ = −σB does not vary with r or t.
√  √  √ 
, so φ(r1 , t) , φ(r2 , t) and φ(r, t)σ(r) = ϕ σ σ r = ϕr, and √φr does not vary with r or t.
r r
CIR: Sharpe ratio φ = ϕ σ

26.3. Delta-hedging.
X t1 P(r, 0, t1 )
duration-hedge : (bond value) (bond duration) = 0 =⇒ need to sell N=− of bond 2
t2 P(r, 0, t2 )
X Pr (r, 0, t1 ) ∂P
delta-hedge : (bond value) (bond delta) = 0 =⇒ need to sell N=− of bond 2, Pr =
Pr (r, 0, t2 ) ∂r
8
 D v Theta for a call
q r Psi for a call
Y
Delta for a call Vega for a call or a put Rho for a call
1 25 0 80 0
0.1 year 0.1 year
0.8 20 0.5 year 0.5 year −20
60
1.2 year −5 1 year
0.6 15 −40
30 year 40 2 years
0.1 year 0.1 year 0.1 year
0.4 0.5 year 10 −60 0.5 year
−10 0.5 year
1 year 1 year 20 1 year
0.2 5 −80
2 year 2 years 2 years
0 0 −15 0 −100
20 40 60 80 20 40 60 80 20 40 60 80 20 40 60 80 20 40 Stock price 60 80
Stock price Stock price

Call premium as a function of price S Call premium as a function of volatility s Call premium as a function of t Call premium as a function of r Call premium as a function of d
40 20 20 25 35
0.1 year 0.1 year 30 0.1 year
0.1 year
30 0.5 year 15 20 0.5 year 0.5 year
0.5 year 15 25
1 year 1 year 1 year
1 year Strike K=40
20 2 years 10 15 2 years 20 2 years
2 years Strike K=50
10 Strike K=55 15
10 5 Strike K=65 10
10
Strike K=80
0 5 0 5 5
20 40 Stock price 60 80 0 0.1 0.2 0.3 0.4 0.5 0 1 2 3 0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2
volatility Risk−free rate r
Time to expiry Dividend yield rate d

D G q r Y
Delta for a put Theta for a put Rho for a put Psi for a put
0 5 0 80
Gamma for a call or a put 0.1 year
−0.2 0.08 −20
0 60 0.5 year
0.1 year 1 year
−0.4 0.06 0.5 year −40
−5 40 2 years
0.1 year 1.2 year 0.1 year 0.1 year
−0.6 0.5 year 0.04 −60 0.5 year
10 year 0.5 year
1 year −10 1 year 1 year 20
−0.8 0.02 −80
2 years 2 years 2 years
−1 0 −15 −100 0
20 40 60 80 20 40 60 80 20 40 60 80 20 40 60 80 20 40 60 80
Stock price Stock price Stock price Stock price Stock price

Put premium
Put premium as a function of price S Delta for a call or a put Put premium as a function of t Put premium as a function of r Put premium as a function of d
30 15 10 0.25
0.1 year 0.1 year 0.1 year
25 8 0.2
0.5 year 0.5 year 0.5 year
20 1 year 10 1 year 1 year
Strike K=40 6 0.15
15 2 years 2 years 2 years
0.1 year Strike K=50
0.5 year 4 0.1
10 5 Strike K=55
1 year Strike K=58
5 2 0.05
2 years Strike K=65
0 0 0 0
20 40 60 80 20 40 60 80 0 1 2 3 0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2
Stock price Stock price Risk−free rate r
Dividend yield rate d
 Normal Distribution Table
0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09

0.0 0.50000 0.50399 0.50798 0.51197 0.51595 0.51994 0.52392 0.52790 0.53188 0.53586
0.1 0.53983 0.54380 0.54776 0.55172 0.55567 0.55962 0.56356 0.56749 0.57142 0.57535
0.2 0.57926 0.58317 0.58706 0.59095 0.59483 0.59871 0.60257 0.60642 0.61026 0.61409
0.3 0.61791 0.62172 0.62552 0.62930 0.63307 0.63683 0.64058 0.64431 0.64803 0.65173
0.4 0.65542 0.65910 0.66276 0.66640 0.67003 0.67364 0.67724 0.68082 0.68439 0.68793

0.5 0.69146 0.69497 0.69847 0.70194 0.70540 0.70884 0.71226 0.71566 0.71904 0.72240
0.6 0.72575 0.72907 0.73237 0.73565 0.73891 0.74215 0.74537 0.74857 0.75175 0.75490
0.7 0.75804 0.76115 0.76424 0.76730 0.77035 0.77337 0.77637 0.77935 0.78230 0.78524
0.8 0.78814 0.79103 0.79389 0.79673 0.79955 0.80234 0.80511 0.80785 0.81057 0.81327
0.9 0.81594 0.81859 0.82121 0.82381 0.82639 0.82894 0.83147 0.83398 0.83646 0.83891

1.0 0.84134 0.84375 0.84614 0.84849 0.85083 0.85314 0.85543 0.85769 0.85993 0.86214
1.1 0.86433 0.86650 0.86864 0.87076 0.87286 0.87493 0.87698 0.87900 0.88100 0.88298
1.2 0.88493 0.88686 0.88877 0.89065 0.89251 0.89435 0.89617 0.89796 0.89973 0.90147
1.3 0.90320 0.90490 0.90658 0.90824 0.90988 0.91149 0.91309 0.91466 0.91621 0.91774
1.4 0.91924 0.92073 0.92220 0.92364 0.92507 0.92647 0.92785 0.92922 0.93056 0.93189

1.5 0.93319 0.93448 0.93574 0.93699 0.93822 0.93943 0.94062 0.94179 0.94295 0.94408
1.6 0.94520 0.94630 0.94738 0.94845 0.94950 0.95053 0.95154 0.95254 0.95352 0.95449
1.7 0.95543 0.95637 0.95728 0.95818 0.95907 0.95994 0.96080 0.96164 0.96246 0.96327
1.8 0.96407 0.96485 0.96562 0.96638 0.96712 0.96784 0.96856 0.96926 0.96995 0.97062
1.9 0.97128 0.97193 0.97257 0.97320 0.97381 0.97441 0.97500 0.97558 0.97615 0.97670

2.0 0.97725 0.97778 0.97831 0.97882 0.97932 0.97982 0.98030 0.98077 0.98124 0.98169
2.1 0.98214 0.98257 0.98300 0.98341 0.98382 0.98422 0.98461 0.98500 0.98537 0.98574
2.2 0.98610 0.98645 0.98679 0.98713 0.98745 0.98778 0.98809 0.98840 0.98870 0.98899
2.3 0.98928 0.98956 0.98983 0.99010 0.99036 0.99061 0.99086 0.99111 0.99134 0.99158
2.4 0.99180 0.99202 0.99224 0.99245 0.99266 0.99286 0.99305 0.99324 0.99343 0.99361

2.5 0.99379 0.99396 0.99413 0.99430 0.99446 0.99461 0.99477 0.99492 0.99506 0.99520
2.6 0.99534 0.99547 0.99560 0.99573 0.99585 0.99598 0.99609 0.99621 0.99632 0.99643
2.7 0.99653 0.99664 0.99674 0.99683 0.99693 0.99702 0.99711 0.99720 0.99728 0.99736
2.8 0.99744 0.99752 0.99760 0.99767 0.99774 0.99781 0.99788 0.99795 0.99801 0.99807
2.9 0.99813 0.99819 0.99825 0.99831 0.99836 0.99841 0.99846 0.99851 0.99856 0.99861

3.0 0.99865 0.99869 0.99874 0.99878 0.99882 0.99886 0.99889 0.99893 0.99896 0.99900
3.1 0.99903 0.99906 0.99910 0.99913 0.99916 0.99918 0.99921 0.99924 0.99926 0.99929
3.2 0.99931 0.99934 0.99936 0.99938 0.99940 0.99942 0.99944 0.99946 0.99948 0.99950
3.3 0.99952 0.99953 0.99955 0.99957 0.99958 0.99960 0.99961 0.99962 0.99964 0.99965
3.4 0.99966 0.99968 0.99969 0.99970 0.99971 0.99972 0.99973 0.99974 0.99975 0.99976

3.5 0.99977 0.99978 0.99978 0.99979 0.99980 0.99981 0.99981 0.99982 0.99983 0.99983
3.6 0.99984 0.99985 0.99985 0.99986 0.99986 0.99987 0.99987 0.99988 0.99988 0.99989
3.7 0.99989 0.99990 0.99990 0.99990 0.99991 0.99991 0.99992 0.99992 0.99992 0.99992
3.8 0.99993 0.99993 0.99993 0.99994 0.99994 0.99994 0.99994 0.99995 0.99995 0.99995
3.9 0.99995 0.99995 0.99996 0.99996 0.99996 0.99996 0.99996 0.99996 0.99997 0.99997
Inverse lookup: P(Z<z) 0.800 0.850 0.900 0.950 0.975 0.990 0.995 0.999
z 0.84162 1.03643 1.28155 1.64485 1.95996 2.32635 2.57583 3.09023