344 – Methods in Financial Econometrics
Bootstrap Prediction Intervals for Fractionally Integrated Generalized Autoregressive Conditionally Heteroscedastic (FIGARCH) Models
Rukman Ekanayake
Missouri University of Science and Technology
The Generalized Autoregressive Conditional Heteroscedastic (GARCH) formulations are inadequate to model the persistent volatility found in certain financial assets. The integrated version of the GARCH formulation, namely the IGARCH model, was developed to handle such situations. Fractionally Integrated Generalized Autoregressive Conditionally Heteroscedastic (FIGARCH) models, however, provide a more flexible alternative to modeling long-term dependence of volatility, providing a leptokurtic unconditional distribution for returns having long memory behavior. We propose a method based on the residual bootstrap to obtain prediction intervals for the returns of FIGARCH processes. A Monte-Carlo simulation study, conducted using a variety of distributions for the error terms, show that the proposed intervals have good coverage probabilities in most cases.