International asset prices and portfolio choices under Bayesian learning

We calculate equilibrium asset prices and portfolio choices from a two-country OLG international asset pricing model under the assumption that investors are on a Bayesian learning path. Investors from both countries receive identical information flows, but domestic investors start off with less precise priors concerning foreign fundamentals. Learning is shown to produce first-order effects on the properties of asset prices, in the form of increased equity returns, volatility clustering, and time-varying correlations across national stock markets. Moreover, on a learning path, estimation risk generates portfolio biases similar to those observed empirically, i.e. a strong preference towards domestic securities and excessive turnover in foreign securities. These findings are robust to changes in prior beliefs, the calibration of initial information asymmetries, and the parameterization of the model. We use real GDP data for the US and Europe to calibrate the model and show that in the event of a financial liberalization during the 1970s, high excess returns, time-varying volatility, substantial home bias, and excess turnover should have been observed.