Detecting predictable non-linear dynamics in Dow Jones Islamic market and Dow Jones industrial average indices using nonparametric regressions

Abstract

This study performs the challenging task of examining the forecastability behavior of the stock market returns for the Dow Jones Islamic market (DJIM) and the Dow Jones Industrial Average (DJIA) indices, using non-parametric regressions. These indices represent different markets in terms of their institutional and balance sheet characteristics. The empirical results posit that stock market indices are generally difficult to predict accurately. However, our results reveal some point forecasting capacity for a 15-week horizon at the 95 per cent confidence level for the DJIA index, and for nine- week horizon at the 99 per cent confidence for the DJIM index, using the non-parametric regressions. On the other hand, the ratio of the correctly predicted signs (the success ratio) shows a percentage above 60 per cent for both indices which is evidence of predictability for those indices. This predictability is however statistically significant only four-weeks ahead for the DJIM case, and twelve weeks ahead for the DJIA as their respective success ratios differ significantly from the 50 percent, the expected percentage for an unpredictable time series. In sum, it seems that the forecastability of DJIM is slightly better than that of DJIA. This result on the forecastability of DJIM adds to its other findings in the literature that cast doubts on its suitability in hedging and asset allocation in portfolios that contain conventional stocks.