We model information flows in continuous time that are generated by a number of information sources that are switched on and off at random times. In a novel approach, we explicitly relate the discovery of relevant new information sources to jumps in the conditional expectation process of a partially observed signal. We derive the resulting endogenous jump-diffusion, and show that it is a solution to the stochastic filtering problem associated with the Markov-modulated multivariate information process. We give a Feynman-Ka\'c representation for the endogenous jump-diffusion, and produce an explicit expression for the size of the jumps. The jump-size distribution is a function of a weighted sum of those information processes that are activated at the jump time. The proposed approach can be applied broadly in signal processing, and an example in Mathematical Finance is provided. We consider a vanilla option and find that its Merton-type price can be expressed by a weighted sum of vanilla prices based on the possible combinations of active information processes at option maturity. The constructed information-flow models lend themselves to the quantification of informational advantage in a competition where agents have asymmetric information flows.
↧