To investigate the role of trading commissions in broker-investor relationships, I study a dynamic cheap talk communication game in which a receiver's payoff depends on a stochastic observable state (portfolio position) and a fixed state, which is privately known to the Sender. The position evolves as a Brownian motion, whose evolution can be controlled by the Receiver at a fixed cost, in an (s,S) policy - the adjustment cost being paid to the Sender. I show that the position has an `informational' value to the Receiver in addition to its inherent value, because it determines the sender's incentives to truthfully report the state. I study Markov Perfect Equilibria where the sender has an option to send uninformative messages, and show that this restricts the set of equilibrium `truth-telling' positions, even though the sender is not able to commit to delaying communication. I extend the model to show analogous results for proportional commissions and stochastic states.

We assume that financial traders remember certain days - for example, those when the trader was actively trading - and that this memory does not fade over time. Unlike standard Bayesians, whose beliefs converge over time in the absence of private information, disagreement can persist in our setup due to idiosyncratic recall. Asset prices are determined by a market maker attempting to clear markets, resembling a classical tatonnement process. The unique equilibrium, where trader actions are optimal given memory-constrained beliefs, is characterized by a Levy process with heavy-tailed price increments. Prices are volatile despite the absence of any change in fundamentals.

To investigate the role of trading commissions in broker-investor relationships, I study a dynamic cheap talk communication game in which a receiver's payoff depends on a stochastic observable state (portfolio position) and a fixed state, which is privately known to the Sender. The position evolves as a Brownian motion, whose evolution can be controlled by the Receiver at a fixed cost, in an (s,S) policy - the adjustment cost being paid to the Sender. I show that the position has an `informational' value to the Receiver in addition to its inherent value, because it determines the sender's incentives to truthfully report the state. I study Markov Perfect Equilibria where the sender has an option to send uninformative messages, and show that this restricts the set of equilibrium `truth-telling' positions, even though the sender is not able to commit to delaying communication. I extend the model to show analogous results for proportional commissions and stochastic states.