Time-varying stock market return predictability: Do we have what it takes?

We use a dividend-yield model from Campbell and Shiller (1988) to forecast the future stock market return on the U.S and Norwegian data from 1984-2018. We use the method from Cochrane (2008), by regressing a Vector Autoregression (VAR)-system and check for forecasting power in the long-run. We find that return gives stronger evidence against unforecastable null-hypothesis for return in the U.S data than the Norwegian data. Norwegian market gives stronger evidence for the dividend growth. R2 increases in the long-run for dividend growth in the Norwegian data, while R2 decreases for return. The opposite appears for the U.S data. We conclude that stock market predictability using the dividend yield model from Campbell and Shiller (1988) and Cochrane (2008) method gives different results for Norwegian data compared to the U.S data