Predictable markets? A news-driven model of the stock market

Abstract

We attempt to explain stock market dynamics in terms of the interaction among three variables: market price, investor opinion and information flow. We propose a framework for such interaction and apply it to build a model of stock market dynamics which we study both empirically and theoretically. We demonstrate that this model replicates observed market behavior on all relevant timescales (from days to years) reasonably well. Using the model, we obtain and discuss a number of results that pose implications for current market theory and offer potential practical applications.We conclude Section 2 by comparing the characteristic behaviors of theoretically-modeled, empirically-modeled and observed market prices and by discussing the possibility of market forecasts. Finally, we provide an overall summary of conclusions in Section 3.The present study has been carried out with a view toward potentially predicting stock market returns. This view is supported by the empirical research over the last 30 years, which suggests that returns are predictable, especially over long horizons (see Fama and French (1988, 1989), Campbell and Shiller (1988a, Cochrane (1999Cochrane ( , 2008, Baker and Wurgler (2000), Campbell and Thompson (2008)). This empirical research has primarily focused on identifying potentially predictive variables, such as the dividend yield, earnings-price ratio, credit spread and others, and verifying or refuting their correlation with subsequent returns, typically applying regression methods.Our objective is to capture the basic mechanisms underlying this predictability. 1 We show that the theoretical model (Section 2), which treats information, opinion and price as endogenous variables, can reproduce observed market features reasonably well, including the price path and the return distribution, under the realistic choice of parameter values consistent with the values obtained using the empirical data (Section 1). Most importantly, this model permits market regimes where deterministic dynamics dominate random behaviors, implying that returns are, in principle, predictable. Because this model is fundamentally nonlinear, the causal relation among the variables 1 The existing literature examines the long-term predictability (e.g. monthly and annual returns) and links it to economic fluctuations and changes in risk perception of investors. Our findings support the view that long-term predictability exists and, furthermore, indicate that returns may already be predictable on intervals of several days. We reach this conclusion using a framework which attributes price changes to the dynamics of investor opinion. Interestingly, we find that, over long horizons, there is a connection between the evolution of investor opinion and economic fluctuations ( Fig. 9, Section 1.4.1). * *

An endogenous mechanism of business cycles

“…It follows that the feedback-interlinked common feature of these models is that being inherently nonlinear they look for explanations of market behaviors as nonlinear phenomena. We also take note of Franke (2014) who proposed a generic model for investor opinion dynamics, which exhibits certain behaviors similar to those found in Gusev et al (2015).…”

“…Page | 9 information, expectation and price may be the relevant macro variables for describing market dynamics, and thus with the correct choice of hypotheses about how interactions occur at the micro level a dynamic model for the evolution of these variables can be developed that may be able to explain the observed market behaviors. Gusev et al (2015) developed, and Kroujiline et al (2016) extended, such a dynamic model.…”

“…Finally, we demonstrate that the model contains the efficient market as a particular case (Section 1.3). Gusev et al (2015) introduced an agent-based model in which investors' decisions are influenced by opinions exchanged with other investors and by opinions accessed through mass media. The latter are assumed to come from financial analysts, market commentators, newspaper journalists, finance bloggers and other participants who interpret relevant information, opine on how the market might react and make their views available to investors through media outlets (for convenience we refer to them collectively as analysts).…”

“…Appendix A in Gusev et al (2015) provides a formal treatment of this problem and detailed derivation of equations governing the evolution of macro variables. This section offers a simplified derivation to illustrate the modeling approach and enable the reader to develop intuition about the micro-macro connection in this model.…”

“…To this end, we consider a news-driven stock market model developed in Gusev et al (2015). This model, based on micro-level interactions among agents, yields at a macro level a closed-form nonlinear dynamical system that explains price formation as stemming from endogenous dynamics between information flow, investor expectation and the market price itself.…”