Sie sind auf Seite 1von 2

Equity options

l Cutting edge

A discrete question
How should discrete dividend options be modelled in an equity option pricing framework? As Volf Frishling warns, unthinking use of certain models to solve this problem can lead to significant mispricing in some situations
he formula for valuing European-style options with fixed discrete dividends is well known and can be found in the literature (see, eg, Hull, 2000, and Haug, 1998). A number of commercial systems also use it. The logic is simple as the dividends are a risk-free component of the stock price dynamics they can be taken away from the current stock price upfront and the usual Black-Scholes (1973) arguments can then be applied. But is this true? The model separates the capital price process from the dividends and models it as a continuous geometric Brownian motion. Denoting the price process as St, the capital price process as Ct, the dividend payable at time ti as Dti and expiry as T, the equations for the price and capital price processes will be:
dC t = rC tdt + C tdw t St = Ct +
T > ti > t

It is instructive to look at the terminal stock prices at time T for each of the three models. For simplicity, we will assume that only one dividend D is payable at time t < T. Model 1:
S T = C T = S 0 De r t e S0 e

(r 0.5 ) T +w
2

(r 0.5 ) T + w
2 2

De (

r T t ) 0.5 T + w T

Model 2:
S T = A T De (
r T t)

= S 0e

(r 0.5 ) T +w
2

r T t De ( )

Model 3:
ST = St D e

r t t D t ie ( i )

ST = CT

In this model, the stock price process does not follow the geometric Brownian motion, so the standard Black-Scholes formula cannot be applied. The payout of a call option, however, can be written as (ST K)+ = (CT K)+ and the standard Black-Scholes formula can be applied to the process Ct with the spot C0 = S0 Dtier(t i t). Musiela & Rutkowski (1997), following Heath & Jarrow (1988), consider another model, where the accumulation price process At is assumed to follow a geometric Brownian motion: r tt dA = rA dt + A dw S =A D e ( i) S =A
t t t t t t 0 < ti < t

(r 0.5 2 ) t + w t D e(r 0.5 2 )(T t )+ w T t = S 0e S0 e

(r 0.5 ) (T t )+ w
2

Tt

(r 0.5 ) T + w
2

De

(r 0.5 )(T t )+ w
2

T t

ti

The payout of a call option can then be written as:

(S T K )+ = (A T D t er (t t ) K )
i i

= AT K +

( (

D t e (
i

r t ti )

))

As with the above, the stock price process does not follow the geometric Brownian motion, but AT does and the standard Black-Scholes formula with the adjusted strike K = K + Dt ier(t t i) can be used. Yet another approach models the price process as a discontinuous process with deterministic jumps at deterministic times and continuous geometric Brownian motion between the jumps:
dS t = rS tdt + S tdw t,
+

t i < t < t i +1

S t+i = S ti D t i

Here, S ti and S ti are stock prices immediately before and after the dividend respectively. For processes of this type, no convenient closed-form formula exists (except in the case of one dividend) and a numerical method (eg, finite-difference, lattices) must be used. A detailed description of this approach can be found, for example, in Wilmott (1998).

w where Brownian motion ~ is independent of w. While the expected values (forward price), E(ST) = S0erT Der(T t), are the same for all three models (as they should be to avoid arbitrage), their distributions are different. Different terminal distributions will generate different option prices, which are the subject of our analysis. We give a hypothetical example that assumes that option prices are given via their volatilities. We disregard the issue of the volatility smile. Table A compares European-style option prices for all three models with S0 = 1, r = 5%, = 30%, T = 1 and various strikes and dividends (paid at six months) per $1 million face value. Note the call and put price differences are the same due to the call-put parity relationship. What we see is that the three methods generate very different prices, ie, they are not consistent. This should come as no surprise as the underlying models are very different. There is a simple argument that may help to explain the differences between the three models. Consider a call with a spot price of $2, a strike of $2, a dividend of $1 and a zero interest rate, where the dividend is paid just before expiry. Let us assume that the stock price doubles. The first model brings the dividend to the start of the process, ie, the price of the stock drops to $1 immediately. Then, on this Brownian motion path, the stock ends up being worth $2, thus rendering the option worthless. For the second model, the price will double to $4 first. As the dividend is paid out, the price drops to $3 and the option will be worth $1. Using the same logic, it is easy to see that any path that generates a payout for the first approach will gen-

A. Vanilla options price analysis


European K = 1, D = 0.1 K = 1, D = 0.05 K = 0.5, D = 0.1 K = 0.5, D =0.05 K = 1.5, D = 0.1 K = 1.5, D = 0.05 Model 1 87.8 113.4 428.0 476.3 9.3 14.1 Call ($ 000) Model 2 99.3 119.2 429.7 476.8 13.3 16.6 Model 3 93.8 116.4 428.6 476.5 11.5 15.5 Model 1 136.6 113.4 1.1 0.7 533.7 489.7 Put ($ 000) Model 2 148.0 119.2 2.8 1.2 537.7 492.2 Model 3 142.5 116.4 1.8 0.9 535.8 491.1 Diffs (in $ 000 from model 3) Model 1 Model 2 6.0 (5.5) 3.0 (2.9) 0.6 (1.0) 0.2 (0.3) 2.2 (1.8) 1.3 (1.1)

WWW.RISK.NET JANUARY 2002 RISK

Cutting edge

Equity options

B. Down-out-call price analysis


Expiry 1m 2m 3m 4m 5m 6m 1 day 6m +1 day 7m 8m 9m 10m 11m 12m Call prices $36,950 $52,966 $66,124 $77,482 $86,743 $96,197 $56,094 $62,419 $69,166 $75,466 $81,769 $87,873 $93,625 Implied volatility model 3 30.0% 30.0% 30.0% 30.0% 30.0% 30.0% 30.0% 30.0% 30.0% 30.0% 30.0% 30.0% 30.0% Implied volatility model 1 30.0% 30.0% 30.0% 30.0% 30.0% 30.0% 33.1% 32.7% 32.4% 32.1% 31.9% 31.8% 31.6% Barrier prices model 3 $36,715 $50,726 $60,269 $67,241 $72,220 $76,776 $47,698 $51,422 $54,631 $57,175 $59,428 $61,404 $63,115 Model 1 closed form $36,715 $50,733 $60,280 $67,283 $72,307 $76,916 $2,039 $2,122 $2,204 $2,275 $2,340 $2,400 $2,453 Model 1 cont yield $36,712 $50,733 $60,280 $67,282 $72,306 $76,914 $43,836 $46,556 $49,323 $51,792 $54,159 $56,361 $58,362 Model 1 numeric $36,712 $50,726 $60,269 $67,240 $72,219 $76,774 $52,135 $55,087 $57,752 $59,940 $61,920 $63,685 $65,234

erate a larger payout for the second, thus making the option price higher. The values generated by the third model lie between the other two. This discrepancy arises from the fact that the price process is a geometric process and is not invariant with respect to linear adjustments. In practice, prices of vanilla options of different expiries are observed in the market and a term structure of volatilities is generated from those using one or another model. We show that this way of approaching the problem of inconsistency between the models described above does not solve it. Consider table B. The deal parameters are as above and a dividend of 0.1 is paid in six months. The call prices are given (in fact generated using model 3 and flat volatility term structure at 30%) and the barrier is set at 0.9. Now, say, the trader using model 1 to price vanilla options and calibrate his/her volatility term structure wishes to price an exotic option, such as a down-and-out barrier. He/she can reason in three plausible ways: As the dividends are taken care of, I can price barrier options using the closed-form formula with no dividends. The prices obtained are very low for expiries past the dividend date. The reason is obvious: the adjusted spot price of 0.9025 is very close to the barrier. If the dividend was bigger, then the spot could have breached the barrier immediately, thus rendering the option worthless. In reality, however, as the dividend is paid later the option has a reasonable chance to survive. Noticing this, the trader decides to change the approach. Ill start with the current spot, calculate forwards and the implied continuous dividend yield, and then apply the closed-form formula. First, the trader has already compromised the model by switching to a different one. Second, the implied dividend yield is very large immediately after the dividend date, making the price lower than it should be. Third, the closedform formula may not be very accurate for non-constant term structures. Because the term structure is not flat and I have a discrete dividend, Ill use a numeric method lattice or finite-difference. As before, the method is compromised and, due to higher volatilities, the generated prices will be different (higher in this case) than they should be. A question arises: why use model 1 at all if exotics are to be priced by a numeric method anyway? Model 3 takes the timing of the dividend into the account. It is unclear how to adjust either model 1 or 2 for this. Moreover, the implied volatility term structure using model 1 exhibits discontinuity across the dividend date. If the data is calibrated the other way around, ie, the prices are generated using model 1 and a smooth volatilities term structure, then the spot volatilities implied by using model 3 will show discontinuity. In fact, the situation will be even worse as the discontinuity will be downwards, thus generating negative squares of forward-forward volatility an impossible local volatility curve. Another example of potential mispricing caused by model inconsistencies occurs with American-style options. Consider American-style versions of the options analysed above. If closed-form approximations are used naively, American-style call options will have the same values as their EuRISK JANUARY 2002 WWW.RISK.NET

ropean-style counterparts. This could be expected because calls are not optimal to exercise under models 1 or 2 and, again, short of reverting to numeric methods, it is unclear how to adjust the pricing to take account of discrete dividends. Hull (2000) and Haug (1998) provide a closed-form formula for pricing American-style options with one known dividend. The formula, however, appears incorrect as the price of the American-style call expiring immediately after the dividend is $88,173 less than the $96,197 of the European-style option expiring just before. In any case, the formula is not applicable as the volatility is not constant. The answer to the question of which model reflects reality better and provides consistent pricing across a range of options is almost self-evident. It seems that the third model is more in agreement with the actual evolution of the price process and should be used, particularly if pricing of exotics is required. It should be pointed out that the three models compared in this note are well known, although the second one, which models the accumulation process, is less familiar than the others. However, no source known to us analyses the differences between these models or recommends one in preference to another. Moreover, it appears that some commercial systems use or at least provide these models quite arbitrarily and inconsistently, thus leading to potentially dangerous mispricing and mishedging of the portfolios. We have no doubt that many practitioners and quantitative analysts grappled with this problem at some time in their careers and may have arrived at conclusions similar to those outlined. If this article makes the quantitative analyst community more aware of these pitfalls, and helps them to price their positions correctly, then it will have fulfilled its purpose. Volf Frishling is head of quantitative analysis of financial markets at the Commonwealth Bank of Australia and adjunct professor at the University of Technology, Sydney. He would like to thank an anonymous referee for helpful comments
Comments on this article may be posted on the technical discussion forum on the Risk website at http://www.risk.net

REFERENCES
Black F and M Scholes, 1973 The pricing of options and corporate liabilities Journal of Political Economy 81, pages 637659 Haug E, 1998 The complete guide to option pricing formulas McGraw-Hill Heath D and R Jarrow, 1988 Ex-dividend stock price behaviour and arbitrage opportunities Journal of Business 61, pages 95108 Hull J, 2000 Options, futures and other derivatives Fourth edition, Prentice-Hall International Musiela M and M Rutkowski, 1997 Martingale methods in financial modelling Springer Wilmott P, 1998 Derivatives John Wiley & Sons

Das könnte Ihnen auch gefallen