A behavioral stock market model (Q2482683)

This paper proposes a behavioural stock market model in the form of a nonlinear stochastic feedback system. The model consists of a part which models the market dynamics, which the prices of the assets with the demand quantity and the bid price for the participating agents, and a part which models the transaction requests, which connects the observed stock prices process with the demand quantities and bids of the agents. These two parts are composed in a feedback loop, thus closing the system. As the behaviour of the agents is based on anticipated future price movements of the market, an important role in the model is played by the price anticipation operator, and the dynamics of the updating of this predictor. In the mathematical form of the model, the equilibrium beliefs in the market are expressed as the fixed point of an operator equation, and the dynamics leading to this equilibrium are provided by an iterative scheme. The paper investigates a class of such belief updating schemes, based on linear price predictors; such predictors include e.g. \(AR(k)\) type approximation etc. These prediction schemes translate to general problem of approximation of the best predictor from a given parameterized model, which in real life translates to minimization of appropriately chosen functionals, which include the observations of market data. These minimization problems are treated with the use of stochastic gradient algorithms. The main results of the paper are convergence results of these algorithms for the problem at hand, under general linear growth and Lipschitz conditions on the price formation functions and the demand functions of the agents. The proposed model is general enough to incorporate realistic market function rules, heterogeneity of agents of different behavioural types e.g. loss averse agents, risk seeking traders, rational agents etc, and is a very interesting attempt in modelling market dynamics tested as driven by the actions of heterogeneous agents and their expectations. A simulation experiment is performed using the opening rules of the Budapest Stock Exchange.