Econometric Modeling of Processes with Varying Structural Parameters

点击翻译按钮获取中文摘要

This project continues a successful line of econometric research on a type of model especially suited for time series analysis, in particular for examining financial sector price movements. Traditional statistical methods for time series analysis assume that the variances of forecast errors remain constant from one observation to the next. Professor Engle and his colleagues have developed a class of models know as Autoregressive Conditional Heteroskedasticity (ARCH) models which relax this restrictive assumption. In the ARCH model the variances of a variable are themselves stochastic and conditioned on the information contained in the entire variable set. These models have been used very successfully in analyzing asset price movements, volatility in the stock market, in forecasting inflation, and in measuring risk. In this project Professor Engle extends his past work to include multivariate time series models. He allows the variances to change over time in a manner that does not decrease the generality of the model, but results in a parsimonious specification. Also included in this work is the concept of co-integration, which is a statistical characterization of several series of data which might reasonably contain common trends or growth tendencies. This is an important extension in that such models have a straightforward economic interpretation. Professor Engle also empirically validates his theoretical work by applying the ARCH specification to data on asset prices in foreign exchange markets, interest rates, options contracts, and equity prices.