Forecasting Prices in Regime-Switching Markets

Abstract

Linear autoregressive (LAR) models poorly predict asset prices in nonlinear, regime-switching markets. In this article, the authors use SETAR, a threshold model that accounts for nonlinearities, to test for the existence of regime-switching in global equity markets. A comparison of SETAR’s predictive power against that of LAR models suggests that SETAR yields more accurate long forecasts, in both emerging and developed stock markets. The authors discuss extensions of threshold models into portfolio management, corporate valuation, and the long-term forecasting of financial indicators.