Abstract
We revisit equilibrium-correction modelling of aggregate real consumers’ expenditure in the UK, using Autometrics applied to the data in Davidson, Hendry, Srba and Yeo (DHSY; 1978). The many selection decisions involved in developing viable empirical models are discussed in a setting where there are more candidate explanatory variables than observations, here due to Impulse-Indicator Saturation (IIS) for detecting breaks, outliers and data contamination. Additional tests of the selected model include whether it encompasses the original specification, evidence of nonlinearity and if the conditioning variables are super exogenous. We consider how IIS affects economic interpretations of models, and conversely, and the implications of robust forecasting devices.
Acknowledgements
This research was supported in part by grants from the Open Society Foundation and the Oxford-Martin School. I am grateful to Jennifer L. Castle, Jurgen A. Doornik, Sören Johansen, Grayham E. Mizon and Bent Nielsen for invaluable contributions to the approach adopted, and to Jennie and Grayham for helpful comments on an earlier draft.