An agent-based modeling approach to study price impact

Price impact models are important for devising trade execution strategies. However, a proper characterization of price impacts is still lacking. This study models the price impact using an agent-based modeling approach. The purpose of this paper is to investigate whether agent intelligence is a necessary condition when seeking to construct realistic price impact with an artificial market simulation. We build a zero-intelligence based artificial limit order market model. Our model distinguishes limit orders according to their order aggressiveness and takes into account some observed facts including log-normal distributed order sizes and power-law distributed limit order placements. The model is calibrated using trades and orders data from the London Stock Exchange. The results indicate that agent intelligence is needed when simulating an artificial market where replicating price impact is a concern.