An empirical study of a simultaneous-equation approach to models of covariance.
Item
-
Title
-
An empirical study of a simultaneous-equation approach to models of covariance.
-
Identifier
-
AAI8820852
-
identifier
-
8820852
-
Creator
-
Chiou, Jengren.
-
Contributor
-
Adviser: Harry M. Markowitz
-
Date
-
1988
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Economics, Finance
-
Abstract
-
In portfolio analysis, since the number of covariance grows rapidly as the number of securities under consideration increases, some models of covariance have been proposed to reduce the required amount of covariance input. In this study, another type of model of covariance is proposed using a simultaneous-equation approach to improve the prediction of future covariances or correlation coefficients between security returns.;Simultaneous-equation approach is the most fundamental way to incorporate industry interaction in a model of covariance. Three forms of models of covariance with industry interaction were experimented with in this study: a stock index-based Multi-Index Econometric Model, an economic factor-oriented Multi-Factor Input-Output Model and an index/factor-mixed Market Model with Input-Output Analysis. These three covariance models with industry interaction were compared to other models of covariance with respect to their ability to estimate the dependence structure of security returns.;The evaluation of the dependence structure of security returns predicted by alternative models of covariance is based on MSE criterion and its decomposition. The Theil Inequality Coefficient measures whether a model of covariance can predict better than simple historical extrapolation.;It is found that as long as the choice of underlying factors in a security return-generating equation is reasonably good, more complicated models of covariance typically performed better than simpler models in tracking the historical dependence structure of security returns. However, without the Bayesian coefficient adjustment, simpler models of covariance outperformed more complicated models in predicting the future correlation structure of security returns. But the Bayesian approach enables the Multi-Index Econometric Models to surpass all the other models of covariance in predicting the future correlation structure of security returns in the more stable, second forecasting period.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
