Please use this identifier to cite or link to this item: https://hdl.handle.net/10419/90715 
Year of Publication: 
2012
Series/Report no.: 
Memorandum No. 27/2012
Publisher: 
University of Oslo, Department of Economics, Oslo
Abstract: 
An autoregressive fixed effects panel data equation in error-ridden endogenous and exogenous variables, with finite memory of disturbances, latent regressors and measurement errors is considered. Finite sample properties of GMM estimators are explored by Monte Carlo (MC) simulations. Two kinds of estimators are compared with respect to bias, instrument (IV) validity and model fit: equation in differences/IVs levels, equation in levels/IVs in differences. We discuss the impact on estimators' bias and other properties of their distributions of changes in the signal-noise variance ratio, the length of the signal and noise memory, the strength of autocorrelation, the size of the IV set, and the panel length. Finally, some practical guidelines are provided.
Subjects: 
Panel data
Measurement error
ARMA model
GMM
Signal-noise ratio
Error memory
IV validity
Monte Carlo simulation
Finite sample bias
JEL: 
C21
C23
C31
C33
Document Type: 
Working Paper

Files in This Item:
File
Size
291.34 kB





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.