Bayesian State Space Modelling Of Factor Investing: A Quantitative Equity Strategy Based On Kalman Filter

Abstract

This document focuses on the implementation and development of a momentum investment strategy. Instead of considering the past movements of stock prices as signals for investments, the positions are based on forecast estimations of asset prices. Predictions are based on a dynamic linear representation of the Fama French three factors model. Kalman recursion allows to filter out noise free estimates of the regression coefficients and to predict asset returns. The simulated investment strategies are able to overperform the market portfolio with statistical significance in a control environment, but performances start to suffer once market frictions are introduced.