Article

The Pricing of Vanilla Options with Cash Dividends as

a Classic Vanilla Basket Option Problem
Jherek Healy

* Correspondence: jherekhealy@protonmail.com

Abstract: In the standard Black-Scholes-Merton framework, dividends are represented as a

continuous dividend yield and the pricing of Vanilla options on a stock is achieved through the
well-known Black-Scholes formula. In reality however, stocks pay a discrete fixed cash dividend at
each dividend ex-date. This leads to the so-called piecewise lognormal model, where the asset jumps
from a fixed known amount at each dividend date. There is however no exact closed-form formula
for the pricing of Vanilla options under this model. Approximations must be used. While there exists
many approximations taylored to this specific problem in the litterature, this paper explores the use
of existing well-known basket option formulas for the pricing of European options on a single asset
with cash dividends in the piecewise lognormal model.

Keywords: European options; discrete dividends; quantitative finance; pricing.

1. Introduction

2. The piecewise-lognormal model

We assume that the stock price S follows a lognormal process between dividend dates, and jumps
at each dividend ex-date ti from the cash dividend amount αi . The corresponding stochastic differential
equation reads
(
dS = r R (t)S(t) dt + σ (t)S(t) dW, ti ≤ t < ti+1 ,
(1)
S(ti ) = S(ti− ) − αi ,

where (ti )i denotes the full ordered set of dividend dates, r R is the stock lending rate (also known as
repo rate), and W is a Brownian motion.
The price of a European call option of strike K, maturity T on S reads

VC ( T ) = B(0, Tp )EQ [max(S( T ) − K, 0)] , (2)

where B(0, Tp ) is the discount factor to the payment date (typically two business days after the maturity
date), calculated using the risk free rate, or the collateral rate. Let ṼC = B(VC
0,T )
be the undiscounted
p
option price. Integrating Equation 1 leads then to
" !#
RT σ2 ( s ) RT n RT σ2 ( s ) RT
r R (s)− ds+ σ (s) dW (s)
ṼC ( T ) = EQ max S(0)e 0 r R (s)− 2 ds+ 0 σ (s) dW (s)
− ∑ αi e ti 2 ti
− K, 0 ,
i =1
(3)

where n is the number of dividends whose ex-date falls between the valuation date and the maturity
date.
 2 of 6

3. Recasting the problem in terms of a basket option

Let
Z t
W̃ (s) = W (t) − σ (s)ds , (4)
0
Rt Rt
σ (s) dW (s)− 21 σ2 (s) ds
M(t) = e 0 0 . (5)

We have
σ2 ( s )
  R ti R ti 
n ds− σ (s) dW (s)
αi e 0 2 0 K
ṼC ( T ) = C (0, T )EQ  M ( T ) max S(0) − ∑ − , 0  ,
i =1
C (0, ti ) M( T )C (0, T )

RT
with C (0, T ) = e 0 rR (s) ds . The Girsanov theorem [1, Theorem 5.2.3] tells us that W̃ is a Brownian
motion under the probability measure Q̃ defined by M. The Radon-Nikodym theorem [1, Lemma
5.2.2] leads to
  R ti σ2 ( s ) Rt RT 2 RT 
n − ds+ 0 i σ (s) dW̃ (s) − 0 σ 2(s) ds+ 0 dW̃ (s)
αe 0 2 Ke
ṼC ( T ) = C (0, T )EQ̃ max S(0) − ∑ i − , 0  .
i =1
C ( 0, t i ) C (0, T )
(6)
The expectation in Equation 6 corresponds to the undiscounted price under the standard
Black-Scholes model of a European put option of strike S(0) and maturity T on a basket composed of
the lognormal assets (Si ) with drift µi and volatility σi defined by

Si ( 0 ) = α i , (7)
1 ti
Z
µi = − r R (s) ds , (8)
T 0
Z t
1 i
σi2 = σ2 (s) ds , (9)
T 0

for 1 ≤ i ≤ n.
If tn = T, we let Sn = αn + K instead of Sn = αn and the basket in composed of n assets. Otherwise,
RT RT
we define Sn+1 = K, µn+1 = − T1 0 r R (s) ds, σn2+1 = T1 0 σ2 (s) ds and the basket is composed of n + 1
assets.
Thus, we have shown that any classic basket option approximation may be used to price European
options on a stock paying discrete cash dividends, under the piecewise lognormal model.

4. Remarks

4.1. Dividend policy

In absence of a dividend policy, the piecewise lognormal model allows for negative stock prices,
if αi > S(t− j ). Haug et al. [2] explore two natural choices : the liquidator policy where S ( ti ) =
max(0, S(ti− ) − αi ) and the survivor policy where S(ti ) = max(S(ti− ), S(ti− ) − αi ).
The liquidator dividend policy does not impact the call option price, because we have

σ2 ( s )
  R ti R ti  
n ds− σ (s) dW (s)
αi e 0 2 0 K
max max S(0) − ∑ , 0 − , 0 =
i =1
C (0, ti ) M( T )C (0, T )
 R ti σ2 ( s ) Rt 
n ds− 0 i σ (s) dW (s)
αe 0 2 K
max S(0) − ∑ i − , 0
i =1
C ( 0, t i ) M ( T ) C (0, T )
 3 of 6

for K ≥ 0.
It will however impact the put option price VP , and put options must then be priced using the
put-call parity formula VC − VP = B(0, Tp )( F (0, T ) − K ) with forward equal to the undiscounted call
option price of strike zero F (0, T ) = ṼC ( T, K = 0).

4.2. Affine dividends

In addition to cash dividends, we may include proportional dividends where S jumps as S(t+
j )=
S(t− −
j ) − β j S ( t j ). The formulae will not change, except for the definition of C (0, T ) which then must
include the proportional dividends as follows
RT
C (0, T ) = e 0 r R (s) ds
∏ (1 − β j ) . (10)
j:0<t j ≤ T

5. Numerical tests
We assess the accuracy of the basket approach against the modern and accurate cash dividend
approximation of Sahel and Gocsei [3], the second-order approximation of Zhang [4] and the
third-order formula of Le Floc’h [5] respectively named "GS", "Zhang-2" and "LL-3". As basket
option approximation, we evaluate the one of Ju [6] named "BB-Ju" the more standard Curran based
approximation of Deelstra et al. [7] named "BB", and the related lower bound approximation named
"BB-LB" in the figures and tables.
Deelstra et al. [7] present several variations. While those do not significantly change the results,
using their notation, we choose δi = δ3 = er(T − j) and f s (Λ) = f 3 (Λ) = FGF as it led to the best results
in [7, Figures 1 and 2].

5.1. Single dividend case

We consider a European call option of maturity one year on a stock with spot price S = 100, and a
single dividend α1 = 7, varying the ex-date, for three different strikes. To stay in line with the setup
of Haug et al. [2], we let the interest rate be r = 6% and the volatility σ = 30%. We look at the error
in terms of Black-Scholes implied volatility. This is a good scale to compare out-of-the-money with
at-the-money or in-the-money options as well as to compare options of short maturity with options of
long maturity. This is also a very natural measure in the context of the volatility surface construction.
The Ju approximation is very accurate when the dividend ex-date is close to maturity, but
exhibits a large error when the ex-date is close the valuation date (Figure 1). The error of the Deelstra
approximation is much more contained, and is largest when the dividend falls at t = 0.4. Overall,
the error of the basket approach with the Deelstra approximation is found to be the most accurate,
more accurate than the third-order method of Le Floc’h. The formula from Sahel and Gocsei is around
two orders of magnitudes less accurate on this single dividend example. The lower bound basket
approximation is competitive with the formula of Sahel and Gocsei in terms of accuracy.

5.2. Many dividends case

We reproduce the example of Sahel and Gocsei [3], of a call option of maturity 10 years, on a
stock of price S = 100, paying dividends of amount αi = 2 on a semi-annual schedule, with the first
dividend starting one day from the valuation date at t1 = 1/365. The interest rate is taken to be r = 3%
and the volatility σ = 25%.
Figure 2 shows that the basket approach, with a maximum absolute error of 0.0002 volatility point,
is the most accurate. With a maximum error of 0.0078 volatility point, GS is slightly more accurate
than Zhang-2, the basket approach with the Ju formula leads to the largest error and the lower bound
approximation leads to a maximum error of 0.0252 volatility points at the lowest strike.
 4 of 6

50 100 150

1e-02
Absolute Error in Volatility (%)

1e-03

1e-04

1e-05

1e-06
0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00
Dividend ex-date in years

BB BB-LB LL-3
Model
BB-Ju GS Zhang-2

Figure 1. Error of the various approximation on the case of the single dividend α1 = 7, varying the
dividend ex-date for strike K = 50, K = 100 and K = 150.

In Table 1, we reproduce the example of a 7-year option, with varying yearly dividend amounts,
first presented in [8] and in [9]. Our reference, named "FDM" is the price obtained by the TR-BDF2 finite
difference method using 10,000 space steps and 3,650 time-steps and is accurate to the fifth decimal.
We also give the values of the Richardson extrapolated binomial tree from Vellekoop and Nieuwenhuis
[8], named "VNRE", and of the third order approximation from Etore and Gobet [9], named "EG-3".
On this example, the maximum relative error in price (MRE) is largest for GS (2 · 10−3 ), and smallest

Table 1. Price of a call option of maturity 7 years, on an asset with spot price S = 100 paying yearly
dividends of respective amounts 6, 6.5, 7, 7.5 8, 8, 8, starting at t1 = 0.9, with r = 6% and σ = 25%.

Strike FDM VNRE EG-3 LL-3 GS BB-Ju BB BB-LB

70 27.21395 27.21 27.20498 27.21768 27.26457 27.18748 27.21394 27.20683
100 19.48229 19.48 19.47843 19.48478 19.51519 19.47156 19.48245 19.47511
130 14.13026 14.13 14.12933 14.13139 14.15233 14.13643 14.13045 14.12143
MRE N/A N/A 3E-4 1E-4 2E-3 8E-4 1E-5 6E-4

for BB (around 1 · 10−5 , at the limit of our reference price accuracy obtained by the TR-BDF2 finite
difference method).

5.3. Extreme settings

We now explore the extreme settings from [4] where the dividend amount is very large, and the
volatility is high. Figure 3 shows the absolute error in the price of a call option of maturity 1 year, on an
asset of spot price S = 100 paying two dividends of size αi = 25 at t1 = 0.3 and t2 = 0.7. The interest
rate is r = 5% and the volatility is σ = 80%. Again, the basket approach BB is the most accurate one,
and is significantly more accurate than any other approximations. The lower bound approximation
BB-LB is slightly more accurate than GS

5.4. Performance
The basket approximation from Deelstra et al. [7] requires a one-dimensional numerical integration
and as such is slower than the alternatives. We use an adaptive Simpson quadrature for the integration,
 5 of 6

1e-01
Absolute Error in Volatility (%)

1e-02 Model
BB

1e-03 BB-Ju
BB-LB
GS
1e-04 LL-3
Zhang-2

1e-05

50 100 150 200

Strike

Figure 2. Error of the various approximation in the case of semi-annual dividends αi = 2, for an option
of maturity 10 years, varying the strike price.

1e-01
Absolute Error in Price

1e-02 Method
BB
BB-LB
1e-03 GS
LL-3
Zhang-2
1e-04

1e-05

50 100 150 200

Strike

Figure 3. Error of the various approximation in the case of two large dividends αi = 25 and high
volatility σ = 80%, for an option of maturity 1 year, varying the strike price.

with an error tolerance set to 10−7 , such that the result of the quadrature may be considered exact for
all practical purposes. For a large number of dividends, this basket approximation will however be
faster than third-order approximations as the computational complexity is in O(n2 ) where n is the
number of dividends vs. O(n3 ) for third-order methods. While the Deelstra-based approximation is

Table 2. Time to price 1000 vanilla options with 100 dividends yo the option maturity.

Approximation BB-LB GS Zhang-2 BB-Ju LL-3 BB-L BB

Time (s) 0.10 0.12 0.15 2.08 4.74 5.82 10.20

around 70 times slower than other second-order methods, it is unlikely to be a real bottleneck in a
financial system. The use of a Gauss-Legendre quadrature ("BB-L" in Table 2) with 33 points results in
a speed-up by a factor of two, while keeping the same accuracy overall. Instead of the one-dimensional
integration, it is also possible to use the corresponding lower-bound approximation (named "BB-LB"
in Table 2), where no numerical integration is required. The method is then the fastest, but its accuracy
becomes similar to the least accurate second-order method, GS.
 6 of 6

6. Conclusion
We have shown that any basket option approximation for the standard Black-Scholes model may
be directly used to price a European option on single aasset under the piecewise-lognormal model.
For a single cash dividend, the Curran based basket approximation leads to prices as accurate
as advanced second-order or third-order cash dividend specific approximations. In the context of
many dividends, the basket approach is found to be the most accurate one while the computational
cost evolves as the square of the number of dividends, in similar fashion as the second-order cash
dividend approximations. The need for a numerical integration makes it slower than other existing
cash-dividend approximations in general. This is however unlikely to be a real issue in a financial
system. The lower bound basket approximation, while less accurate, does not require a numerical
integration and was found to be of similar accuracy as the approximation of Sahel and Gocsei, while
being faster the the latter.
The technique presented in this paper may be easily extended to to price vanilla basket options
under the piecewise lognormal model through the classic Black-Scholes basket option approximations,
something that specific discrete dividend approximations typically do not allow.

1. Shreve, S. Stochastic Calculus For Finance II, Continuous-Time Models; Springer Finance, 2004.
2. Haug, E.G.; Haug, J.; Lewis, A. Back to basics: a new approach to the discrete dividend problem. Wilmott
magazine 2003, 9, 37–47.
3. Sahel, F.; Gocsei, A. Matching Sensitivities to Discrete Dividends: A New Approach for Pricing Vanillas.
Wilmott 2011, 2011, 80–85.
4. Zhang, K. Fast Valuation for European Options With Cash Dividends. Wilmott magazine 2011, 5, 46–51.
5. Le Floc’h, F. More Stochastic Expansions for the Pricing of Vanilla Options with Cash Dividends. Available
at SSRN 2698283 2015.
6. Ju, N. Pricing Asian and Basket Options Via Taylor Expansion 2002.
7. Deelstra, G.; Diallo, I.; Vanmaele, M. Moment matching approximation of Asian basket option prices.
Journal of computational and applied mathematics 2010, 234, 1006–1016.
8. Vellekoop, M.H.; Nieuwenhuis, J.W. Efficient pricing of derivatives on assets with discrete dividends.
Applied Mathematical Finance 2006, 13, 265–284.
9. Etore, P.; Gobet, E. Stochastic expansion for the pricing of call options with discrete dividends. Applied
Mathematical Finance 2012, 19, 233–264.