Frequency domain and time series properties of asymmetric error correction terms

Abstract

The now familiar error correction model has recently been extended to allow for the modelling of asymmetric behaviour, resulting in the development of the asymmetric error correction model. Previous workers have undertaken studies upon which the ‘classic’, implicitly symmetric, error correction model was developed. It was found that the Escribo-Pfann method approach detected significant asymmetries while the Granger-Lee did not. In this paper the properties of the alternative methods are analysed in detail to examine why one method met with success, but not the other. Via the use of estimated sample spectral densities and correlograms, a significant property of the Granger-Lee method is uncovered. This property, particularly relevant in application to consumers' expenditure, has a wider relevance, helping explain why the approach has met with limited success.