Scribd Logo
Upload

Welcome to Scribd!

  • 168 views

    Equity Smile A Monte-Carlo Approach

    Uploaded by

    Sol Fernández
    This document discusses using a weighted Monte Carlo approach to price complex derivatives that incorporate volatility smiles observed in markets. It presents the methodology, which assigns non-uniform weights to Monte Carlo paths to match implied volatility surfaces. Numerical examples show the approach provides stable calibration and pricing of path-dependent options like Asian options, with faster computation times than other methods.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd

    Equity Smile A Monte-Carlo Approach

    Uploaded by

    Sol Fernández
    168 views2 pages
    This document discusses using a weighted Monte Carlo approach to price complex derivatives that incorporate volatility smiles observed in markets. It presents the methodology, which assigns non-uniform weights to Monte Carlo paths to match implied volatility surfaces. Numerical examples show the approach provides stable calibration and pricing of path-dependent options like Asian options, with faster computation times than other methods.

    Original Title

    Equity Smile a Monte-Carlo Approach[1]

    Copyright

    © Attribution Non-Commercial (BY-NC)

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document discusses using a weighted Monte Carlo approach to price complex derivatives that incorporate volatility smiles observed in markets. It presents the methodology, which assigns non-uniform weights to Monte Carlo paths to match implied volatility surfaces. Numerical examples show the approach provides stable calibration and pricing of path-dependent options like Asian options, with faster computation times than other methods.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    168 views2 pages

    Equity Smile A Monte-Carlo Approach

    Uploaded by

    Sol Fernández
    This document discusses using a weighted Monte Carlo approach to price complex derivatives that incorporate volatility smiles observed in markets. It presents the methodology, which assigns non-uniform weights to Monte Carlo paths to match implied volatility surfaces. Numerical examples show the approach provides stable calibration and pricing of path-dependent options like Asian options, with faster computation times than other methods.

    Copyright:

    Attribution Non-Commercial (BY-NC)
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 2

    SPONSORS STATEMENT

    Equity smile, a Monte Carlo approach


    Jean-Nol Dordain and Christophe Patry, members of Sophis financial research team, present the use of alternative Monte Carlo algorithms for the pricing of complex derivatives with volatility smile

    arket prices observed for listed options on indexes or shares show that the Black-Scholes time-deterministic log-normal volatility assumption is not verified. As the vega hedging is based on listed options, to ignore this fact when valuing structured products can lead to dramatic losses. To take into account the market smile, several methods re-defining the continuous-time share dynamics have been proposed; from the complete market Dupires modelling [Dup94] to stochastic volatility models [Hest93] or mixed jump processes [And00]. In practice, these methods, although theoretically attractive, involve complex and extremely time-consuming numerical schemes. Monte Carlo remains one of the most flexible algorithms when pricing path-dependent complex derivatives. Avellaneda et al [Ave01] proposed a KullbackLeibler probability change on a Monte Carlo sampling eg, assign a nonuniform weight to each trajectory of the sampling to fit a volatility matrix. As opposed to other numerical schemes, this weighted Monte Carlo method offers a simple and flexible way to incorporate the smile in the pricing. But, this method relies heavily on the choice of the input of the model (the choice of the initial asset dynamics, the choice of a strike/maturity volatility grid). First, we will discuss the precise relevance of the chosen method. Second, we will detail the initial guess used to perform the algorithm and, finally, give numerical examples on Asian options.

    For a given grid in maturities/strikes (ti, Kj)i,j, EP*((Sti Kj)+) = EP((Sti Kj)+) (matching the implied volatility grid) We minimise over all probabilities P* equivalent to P, the following
    ln dP * regularity criterion dP . This so-called Kullback-Leibler regularity criterion ensures that P* is the closest probability to P, which fits the implied volatility and the term structure of forward prices.
    dP *

    Weighted Monte Carlo setup


    The weighted Monte Carlo is simply a discrete version of this continuous time setting. However, this algorithm relies heavily on the choice of the input parameters. A careful study gave rise to the following choices: The dynamics of St under P is the standard time-deterministic volatility log-normal process. The prior guess for St was chosen because the Monte Carlo discretization of a log-normal process is extremely efficient, which yields to a minor error on the discretization of the sampling probability space underlying the stochastic process (St). Furthermore, as P* is close to P, the stochastic process ((St),P*) should be close to a locally log-normal process. The set of forward volatilities (1, 2, ,n) used in the simulation of the asset dynamics under P is obtained by the following minimisation where (ti)i is the whole set of dates of the Monte Carlo sampling:
    1 ,..., n i , j

    Why choose a weighted Monte Carlo algorithm?


    The classical Monte Carlo approach used to calibrate the volatility smile is heavy and very time-consuming. Simulation is usually performed via an Euler method, which yields to an error in 0(t) where t is the discretization step. To get accurate results, t needs to be small. Hence, for a one-year option, these algorithms demand at least 300 discretization time steps! We focused our attention on the second category of Monte Carlo, which assigns different probabilities to each path of the simulation. Instead of a direct calibration of the parameters of the stock dynamics, we change the probability distribution of the asset dynamics. The price of any contingent claim is then obtained as the discounted expectation of the payout under this new probability. More precisely, the weighted Monte Carlo approach is the discretization of the following continuous time setting. Denote by P the initial probability, P* the smile probability and St the asset process. P* is defined by: EP*(St) = EP(St) for all t (matching the term structure of forward prices)

    min

    BS ( (1 t1 + 2
    2 2

    (t2 t1 ) + ...

    + i (ti ti 1 ), K j ) BS ( impl 2 (ti , K j ), K j ))


    where BS(2T,K) is the Black-Scholes price of a call option with maturity T, strike K and a volatility , impl(ti,Kj) is the implied volatility with maturity ti and strike Kj. Performing such a minimisation of the volatility of the prior guess ensures the validity and correctness of the KullbackLeibler criterion: we perturb P to obtain P* but this perturbation is fully meaningful if P is not too far from P*. Hence, this minimisation step is essential to ensure the numerical performance of the algorithm. Given a set of option prices and the underlying dynamics under P, we have to find the different probabilities to apply to each path that are

    Implied volatility smile (%)

    compliant with the observed smiles at all maturities. As the problem is under-determined, we impose the following minimisation criterion over all the probability vector (p*)i (sum of the weighted least square residi uals and the relative entropy):

    Figure 1. Fitted volatility smile


    0.33 0.31 0.29 0.27 0.25 0.23 0 50 100 Strike (cash) 150 200 Calibrated implied volatility Black-Scholes volatility

    E wi ( p* (hi ) BS(impl
    i

    (ti , K j ), K j )) + pi* ln( pi* )


    i

    where (wi)i is a vector of positive weights. This problem is reformulated as a dual formulation using Lagrange multipliers and solved by a pseudo-Newton (the BFGS) algorithm.

    Numerical results
    We present below three different numerical test series: Test series 1: First, we test the stability of the smile calibration. We calibrate at two extreme dates and value the implied volatility at an intermediate date and then compare the results obtained with the direct bilinear interpolation of the volatility surface. Test series 2: We then test the stability of the calibration in terms of pricing. We calibrate the smile on different time mesh and value at each time the same Asian option. Test series 3: We compare the results obtained between a BlackScholes-like model and Smile Monte Carlo for a series of Asian options and we also provide for each pricing the calibration CPU time.

    on

    Asian options are all with maturity 1 year and strike = 100, averaging the spot is made on the whole option lifetime: 2 fixings per month, series 1 4 fixings per month, series 2 6 fixings per month, series 3 8 fixings per month, series 4 10 fixings per month, series 5
    Series 1 2 3 4 5 Smile Monte Carlo 7,71498 6,81295 6,7862 6,50858 6,49045 Black-Scholes 7,66056 6,73843 6,73116 6,54835 6,53331 CPU time 1,4 s 2,1 s 2,8 s 3,2 s 4,1 s

    Test series 1
    We calibrate two given identical smiles at T = 9 months and T = 15 months and compute the implied volatility smile obtained with the calibrated Monte Carlo at T = 1 year (see figure 1).

    Test series 2
    We now consider an Asian option (average on the final spot) (T = 1 year, strike = 100) with five fixings per month. We take a flat at-the-money volatility set to 25% and a smile at 1 year for a spot price equal to 100 defined by
    Cash strike 80 100 120 Volatility (%) 28 25 24

    Conclusion
    The weighted Monte Carlo algorithm is certainly one of the most fruitful approaches to equity smile issues for path-dependent options. Once properly defined in terms of input data, it gives sensible figures in a very small CPU time compared with any other numerical scheme. There is much research yet to do to fully exploit the ideas underlying this algorithm especially to speed up Greek computations as an alternative to Malliavin Calculus. References: [And00] Andersen L, Andreasen J. Jump-diffusion processes: volatility smile fitting and numerical methods for pricing, Review of Derivatives Research, Volume 4, pages 231262, 2000. [Ave01] Avellaneda M, Buff R, Friedman C, Grandchamp N, Kruk L, Newman J. Weighted Monte Carlo: A New Technique for Calibrating Asset-Pricing Models, International Journal of Theoretical and Applied Finance, Volume 4, No. 1, pages 91119, 2001. [Dup94] Dupire B. Pricing with a smile, Risk, Volume 7, No. 1, January 1994. [Hest93] Heston S. A closed-form solution for options with stochastic volatility with application to bond and currency options, Review of Financial Studies, Volume 6, pages 327343, 1993.

    We then give the prices (and their comparison with Black-Scholes prices) obtained for 1 calibration date 1 year, series 1 2 calibration dates 6 months, 1 year, series 2 3 calibration dates 6 months, 9 months, 1 year, series 3 4 calibration dates 3 months, 6 months, 9 months, 1 year, series 4.
    Series 1 2 3 4 Smile Monte Carlo 6,64669 6,59975 6,61435 6,61759 Black-Scholes 6,57574 6,57574 6,57574 6,57574

    CONTACTS
    Jean-Noel Dordain, Christophe Patry Quantitative Analysts, Sophis Tel: +33 1 4455 3773 e-mail: [email protected] Website: www.sophis.net

    Test series 3
    We take the same stock as in test series 2 and consider several Asian options. We then give the price delivered by Smile Monte Carlo and BlackScholes. Furthermore, for each of them, the CPU time is also provided.

    You might also like