Supported by Workshop on Stochastic Analysis and Computational Finance, November 2005 Imperial College (London) G.N. Milstein and M.V. Tretyakov Numerical analysis of Monte Carlo evaluation of Greeks by finite differences J. Comp. Fin. 8, No 3 (2005), 1-33
MC evaluation of Greeks by finite differences Plan Model Model Other approaches Other approaches Finite difference approach Finite difference approach Numerical integration error Numerical integration error Monte Carlo error Monte Carlo error Other Greeks Other Greeks Numerical examples Numerical examples Conclusions Conclusions
Model
Model
Model
Other approaches Broadie, Glasserman (1996); Milstein, Schoenmakers (2002)
Other approaches Fournie, Lasry, Lebuchoux, Lions, Touzi (1999, 2001); Benhamou (2000)
Finite difference approach Standard finite difference formulas Weak-sense numerical integration of SDEs Monte Carlo technique
Finite difference approach Newton (1997); Wagner (1998); Milstein, Schoenmakers (2002); M&T (2004)
Weak Euler scheme
Estimator for the option price
Estimator for deltas
Estimators for deltas
Assumptions
Numerical integration error Proof. It is based on the Talay-Tubaro error expansion (Talay, Tubaro (1990); M&T (2004))
Numerical integration error: proof
Monte Carlo error: price
Monte Carlo error: deltas If all the realizations are independent
Monte Carlo error: deltas Boyle (1997); Glasserman (2003), Glasserman, Yao (1992), Glynn (1989); L’Ecuyer, Perron (1994)
Monte Carlo error: deltas
Main theorem
Higher-order integrators
Non-smooth payoff functions Bally, Talay (1996)
Non-smooth payoff functions
Other Greeks
Other Greeks: theta
Numerical tests: European call
Numerical tests: variance reduction Newton (1997); Milstein, Schoenmakers (2002); M&T (2004)
Numerical tests: variance reduction
Numerical tests: binary option
Numerical tests: Heston stochastic volatility model
Supported by Approximate deltas by finite differences taking into account that the price is evaluated by weak-sense numerical integration of SDEs together with the MC technique Exploit the method of dependent realizations in the MC simulations Rigorous error analysis Conclusions