Non-Linear Predictability of Stock Market Returns : Comparative Evidence from Japan and the US

Using smooth transition regression model analysis, we examine the non-linear predictability of Japanese and US stock market returns by a set of macroeconomic variables between 1981 and 2012. The theoretical basis for investigating non-linear behavior in stock returns can be based on the interaction between noise traders and arbitrageurs or behavioral finance theories of non-linear risk aversion. As heterogeneity in investors' beliefs gives reason to suspect a smooth transition between extremes, rather than abrupt, a smooth transition regression model is estimated. Our findings support differences in non-linearity of stock returns in Japan and the US that might be linked to different shareownership of the Japanese stock market compared to the US. In addition, differences in the legal system might have some influence over our findings as well. The US results also suggest greater heterogeneity in the relationship between stock returns and macro variables in the US data relative to the Japanese data. The reasons behind the differences in our results, both between countries and between regimes are probably due to the different economic conditions faced by Japan and the US over our sample, to the possible existence of bubbles in the data and to investor behavior consistent with ‘behavioral finance' theories of investor behaviour