Macroeconomic Modelling and Forecasting Using Non-Stationary Data

Important aspects of macroeconomic modelling and
forecasting in the presence of non-stationarity are
examined in this book. Three forms of
non-stationarity are assessed: explosive,
structural-break, and unit root non-stationarity.
First, testing for unit-root non-stationarity in the
presence of explosive non-stationarity is considered.
Numerical difficulties are circumvented using
approximations before the finite-sample properties of
the unit-root test are assessed. Secondly the use of
model averaging given non-stationarity is
investigated. While model averaging can provide
competitive forecasts and parameter estimates,
selection is required, and often a single selected
model will perform best. Because averaging does not
avoid the need to select, methods of selection are
discussed. Third, regression models in the presence
of unit-root non-stationarity are estimated. Previous
empirical studies of monetary and fiscal policies
have made little reference to non-stationarity. A
cointegrated
vector-autoregressive model is used to combat this
and evidence for policy interactions is found.

Autorentext

J James Reade is a Postdoctoral Research Fellow at the Universityof Oxford. He completed his PhD at the University of Oxford inJune 2007 under the supervision of Prof. David F. Hendry. Hismain research pursuit is the application of econometric methodsto interesting fields of economic theory.

Klappentext

Important aspects of macroeconomic modelling andforecasting in the presence of non-stationarity areexamined in this book. Three forms ofnon-stationarity are assessed: explosive,structural-break, and unit root non-stationarity.First, testing for unit-root non-stationarity in thepresence of explosive non-stationarity is considered.Numerical difficulties are circumvented usingapproximations before the finite-sample properties ofthe unit-root test are assessed. Secondly the use ofmodel averaging given non-stationarity isinvestigated. While model averaging can providecompetitive forecasts and parameter estimates,selection is required, and often a single selectedmodel will perform best. Because averaging does notavoid the need to select, methods of selection arediscussed. Third, regression models in the presenceof unit-root non-stationarity are estimated. Previousempirical studies of monetary and fiscal policieshave made little reference to non-stationarity. Acointegratedvector-autoregressive model is used to combat thisand evidence for policy interactions is found.