In-sample and out-of-sample properties of international stock return dynamics conditional on equilibrium pricing factors

We conduct a comprehensive analysis of the in-sample and out-of-sample properties of stock return dynamics in 14 developed and 12 emerging markets. We start by formulating a theoretically founded asset-pricing model that decomposes log stock returns into equilibrium pricing factors (accounting and discount factors) and short-run (vector) autoregressive dynamics. Based on this model, we design both in-sample and out-of-sample panel modeling techniques to investigate international stock market returns at short and long horizons. Our findings show that (i) there is evidence of in-sample signaling from the equilibrium relations but this feature does not appear to translate into out-of-sample forecasting, (ii) a rolling window forecasting scheme can better approximate the distributional features of returns in comparison with a recursive method, (iii) forecasting with single-lagged equilibrium relationships does not play a uniformly significant role in anticipating returns, (iv) forecasting with a full model containing all lagged equilibrium relations can outperform both a random walk model and a VAR(1) model and (v) linear combinations of alternative forecasts reduce ex-ante uncertainty.