Abstract This paper investigates the nature of financial market fluctuations by empirically testing three competing models of instability. We contrast a linear state-space model and a nonlinear Markov-switching model – both rooted in heterogeneous behavioral heuristics and capable of generating endogenous dynamics – with a benchmark linear random walk model that assumes exogenous shocks. Using monthly S&P 500 data from 1990 to 2019, we find strong evidence supporting endogenous sources of instability. In particular, models incorporating behavioral nonlinearities significantly outperform both the linear behavioral model and the random walk in short-, medium-, and long-term forecasting. Our findings underscore the importance of accounting for heterogeneous expectations and regime-switching behavior in explaining asset price dynamics.
Gatti et al. (Wed,) studied this question.