No Arbitrage Criteria for Exponential Lévy Models A.V. Selivanov Moscow State University Financial Mathematics 3 ”columns” (Z. Bodie, R.C. Merton ”Finance”): Pricing Resource Allocation Utility maximization Equilibrium Arbitrage Pricing Theory Risk Management Concept of No Arbitrage Practical: Mathematical: air A L0 {0}, A – set of incomes No Free Lunch (NFL) (J.M. Harrison, D.M. Kreps 1979) No Arbitrage condition in discrete time (, F ,P), ( S n ) nN0 – any sequence, Fn ( S k , k n) M {Q ~ P : S is an ( Fn , Q) - martingale } Fundamental Theorem of Asset Pricing: NFL M • J.M. Harrison, S.R. Pliska 1981 – finite • R.C. Dalang, A. Morton, W. Willinger 1990 – general case No Arbitrage conditions in continuous time No Free Lunch with Vanishing Risk (NFLVR) F. Delbaen, W. Schachermayer 1994 No Generalized Arbitrage (NGA) A.S. Cherny 2004 Definition of sigma-martingales The definition is given by T. Goll and J. Kallsen A semimartingale M is a sigma-martingale if there exist predictable sets ( Dn )nsuch that 1 • Dn Dn 1 , • Dn [0, T ] or [0, ) n 1 I Dn dSt is a uniformly integrable 0 martingale for any n Sigma-martingales and local martingales sigma-martingales local martingales positive sigma-martingales Classes of Martingale Measures (, F ,P), S – any process, Ft ( Su , u t ), t 0 M {Q ~ P : S is an ( Ft , Q) - - martingale } M {Q ~ P : S is an ( Ft , Q) - martingale } UM {Q ~ P : S is an ( Ft , Q) - uniformly integrable martingale} Fundamental Theorem of Asset Pricing no arbitrage condition finite time horizon infinite time horizon M M M UM F. Delbaen, W. Schachermayer NFLVR S – semimartingale A.S. Cherny NGA S – any positive process absence of arbitrage existence of certain martingale measure completeness of the model uniqueness of the measure Models under consideration exponential Lévy model: St e Lt (1) time-changed exponential Lévy model: St e ( L )t e L t (2) L – nonzero Lévy process – independent increasing non-constant process Black-Scholes and Merton models Black-Scholes model Bt t St e B – Brownian motion, , 0 Merton model t t St e – Poisson process, , 0 Theorem for models with finite time horizon statement M NFLVR MNGA Let M . Then | M | 1 | M | 1 condition for model (1) condition for model (2) process S is not monotone Black-Scholes or Black-Scholes or Merton model; Merton model is deterministic and continuous Theorem for model (1) with infinite time horizon statement condition for model (1) M NFLVR either the process S is a P-martingale, or E S1 1 and the jumps of S are not bounded from above UM GA always Let M . Then | M | 1 Black-Scholes or Merton model Theorem for model (2) with infinite time horizon Suppose that P-a.s. Then always UM GA. . An example: NFLVR and GA St e Bt t / 2 , t0 NFLVR is satisfied; NGA is not satisfied Strategy: H t I(t ) X 0.5 inf{t 0 : St 0.5} Conclusions We have obtained: the criteria for the NFLVR and the NGA conditions for models with finite time horizon; for these models NFLVR NGA the criteria for the NFLVR and the NGA conditions for models without time change and with infinite time horizon; for these models the NGA is never satisfied, while the NFLVR is satisfied in certain cases