A0272
Title: Heteroskedasticity and autocorrelation robust inference for a system of regression equations
Authors: Denise Osborn - University of Manchester (United Kingdom)
Ralf Becker - University of Manchester (United Kingdom)
Robert Anderson - Newcastle Univeristy (United Kingdom) [presenting]
Abstract: Standard single equation heteroskedasticity and autocorrelation (HAC) robust inference methods are extended to allow consistent inference for a system of vector moving-average correlated equations also accommodating contemporaneous correlations. This is of particular relevance to the examination of inflation forecast errors, as forecasts for different groups are contemporaneously correlated, while any proposed forecasting model utilising a time-series of multi-period forward-looking expectations data will suffer from overlapping errors inducing a moving-average error structure. The proposed methodology is a generalisation of previous work. Monte Carlo simulations confirm that the method performs well in large samples. Applications testing the rationality of male versus female inflation forecasts, and those of defined educated groups, are also included.
