Stock Price Forecasting Using A Dependence Structure

It is important to incorporate diverse dependence structures between stocks when managing a stock portfolio. Copulas are a useful statistical tool to capture dependence structure, dealing with both the linear and non-linear association that may occur in the tails of data. Financial time series datasets often exhibit volatility clustering that affects price forecasting accuracy. This work proposes the initial use of the principal component analysis followed by a copula and GARCH model that filters the effect of the volatility clustering in the series. For illustration, we consider ten banks from which Bank of America and PNC Financial Services Group are chosen, and then we project their future price movements through simulations. Since they are selected in terms of the principal component analysis, the procedures could help the proposed model to become a more widely used tool in forecasting financial stock performance.