This paper shows the links between the different methods currently used for pricing contingent claims. These methods are all grounded on the no arbitrage assumption. The first approach changes the historical probability into a martingale measure. The second starts from a state variable framework and yields prices as solutions of a partial differential equation. We show how these two methods can be derived from a general state variable-APT model and how the latter leads to the expression of the appropriate martingale measure through the Radon-Nicodym derivative. Hence, all major pricing models are presented within a unified framework.
BAJEUX-BESNAINOU, I. and PORTAIT, R. (1992). Méthodes probabilistes d'évaluation et modèles à variables d'état : une synthèse. ESSEC Business School.