Date of Award
Doctor of Business Administration (DBA)
Jack Welch College of Business
Dr. W. Keener Hughen
Dr. Lior Menzly
Dr. Michael Gorman
This thesis examines the efficacy of alternative modeling techniques to predict stock market returns modeled with time-varying coefficients with the goal of developing and implementing a trading strategy that yields excess returns. First, we determine the modeling technique with the smallest forecast error using historical predictors: the differenced dividend-price ratio, lagged S&P 500 returns, and the change in implied volatility. The candidate modeling techniques include both constant and recursive ordinary least squares (OLS) regression methods and diverges from previous return forecast literature with the comparison of a state-space model (SSM) cast as a VAR(1) process to each OLS technique. The state-space model is found to be the superior modeling technique with the smallest RMSE 3.76% and greatest out-of-sample of 2.62% using delta VIX as the forecasting variable. Second, we demonstrate economic significance, using 1) monthly stock return forecasts in a market timing strategy, and 2) daily price forecasts in a simulated live pairs trading strategy taking into account implementation shortfall. In both trading strategies, the state-space model Kalman filter significantly outperforms the alternative OLS modeling techniques with an annualized total return of 21.64% in the market timing strategy and an annualized total return of 13.21% unlevered in the pairs trading strategy.
C6; C15; C32; C88; G11; G17; Y40
Dempsey, H. E. (2021). Kalman filter vs alternative modeling techniques and applied investment strategies. Jack Welch College of Business & Technology dissertation, Sacred Heart University, Fairfield CT. Retrieved from https://digitalcommons.sacredheart.edu/wcob_theses/24/
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.