Stochastic resonance in the recovery of signal from agent price expectations

Contributions that noise can make to the objective of detecting signal in agent expectations for price in financial markets are examined. Although contrary to most assumptions on exogenous noise in financial markets as increasing both risk and uncertainty in the detection of signal, a basis for the contribution that noise can have to agent objectives in signal detection through stochastic resonance (SR) is well-documented across disciplines. After reviewing foundations for the micro-processing of expectations, a multi-component model of networked agents that includes a component of bounded rational processing and a component that has been cited as generating “herding” behavior in financial markets is offered. The signal-to-noise ratios in the proposed models provide a basis to investigate SR in an application to financial markets. Results with both deterministic and stochastic forms of the proposed model support SR as a process in which randomness can contribute to the recovery of signal in agent expectation. Additionally, predictive models that indicate the sensitivity of the occurrence of SR to the parameters of the models of agent expectations were estimated and cross-validated. The discriminative ability of the models is reported through Area Under the Receiver Operating Curve (AUROC) methodology. These results extend the cross-discipline demonstrations of SR to models of price in financial markets. •Stochastic resonance in the detection of signal in price expectations.•Multi-component dynamic model for agent price expectations.•Deterministic and stochastic models of price expectation.•Small world networks of interacting agents.•Signal-to-noise ratio in expectations for market price.•AUROC predictive model of stochastic resonance in the recovery of signal.