摘要：In this article we examine markets dominated by retail investors where herding behavior can be prevalent. We consider a multi factor model to forecast stock returns that we suppose affected by size, book to market and herding behavior. Applying the model to Saudi stock market on daily data from 7th January 2007 to 1st March 2016, we construct three type of weighted portfolio: large, mid and small capitalization. The result of a logistic regression shows that our model can estimate stock returns with a higher precision of more than 70%. Using our model we estimate the out-of-sample Value at Risk using historic simulation. Finally we conduct a back testing which confirm the precision of the forecasted VaR.
