APPLICATIONS OF VARYING PARAMETER REGRESSION TO INVESTMENT THEORY
Abstract
Over the past two decades, the notion of time-varying models has been an important focus in the economic literature. In particular, the focus has been on the derivation of efficient estimation procedures for models exhibiting time-varying parameters. Many different types of variation have been introduced but most fall within one of three basic categories: random, sequential or systematic parameter variation. This thesis is comprised of three distinct essays. The later two essays are characterized by both theoretical and empirical sections whereas the first is entirely theoretical in nature. Moreover, each essay addresses the notion of time-varying parameters from an investment-related prospective. In other words, portfolio and/or security return generating processes are analyzed within dynamic settings. The first essay concerns estimation of a multi-index return generating process characterized by sequentially varying coefficients or "index sensitivities". Estimation procedures are derived that impose less structure on the model than that currently expoused in the literature. Consequently, a more practical and efficient estimation methodology is likely to result. The second essay is focused on the dynamic nature of security risk parameters during a time interval or "event period" surrounding a stock split. Economic rationale and estimation methodology for temporally concave (event period) risk parameters are presented. Finally, hypothesis testing for event period concavity of the security risk parameters is conducted over a sample of split stocks. Lastly, varying parameter regression techniques are used in the final essay to improve upon existing methods of portfolio performance evaluation. In particular, portfolio performance is dichotomized and analyzed separately for its micro-and macroforecasting components. A sample of mutual funds is used in the empirical examination.
Degree
Ph.D.
Subject Area
Finance
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.