54

In this paper, we implement a bivariate Markov-Regime-Switching (MRS) model to filter time-varying size risk premia relying on five international equity markets. Our statistical framework identifies two endogenous size states exhibiting distinct characteristics. The upward (downward) size state is characterized by strongly positive (low or negative) size spreads, and is positively (negatively) correlated with the lagged changes of the Composite Leading Indicator (CLI). These results hold for all the considered markets, time intervals, and are robust to alternative factor sets used to adjust for risk. They directly challenge the statistical fluke hypothesis stating that the size effect may be a matter of pure coincidence. JEL classification: C58, G11, G12