Abstract
When the possibility of an unknown structural break is allowed and it is taken into account we find a significant long-run relationship between Swedish money demand and its determinants that is not found when no break is considered. The estimated elasticities show that money demand is more responsive to its determinants in the period after the break than before. Possible underlying reasons for the occurrence of this break and its implications are explained.
Acknowledgements
The authors thank Allan Gregory and Bruce Hansen for providing the GAUSS programme corresponding to their 1996 paper through Hansen's home page. The authors were able to use it for their purposes with little modification. The authors are grateful to Hansen for additional source information that he provided.
Notes
Gregory and Hansen (Citation1996) also dealt with US data using the same technique as was used in the current paper for finding breakpoints (which they developed in that study), but their empirical application seemed to be more for demonstrative purposes rather than for the purpose of making an empirical contribution, given their presentation. They found money demand cointegrated with its determinants regardless of whether a break is taken into account.
The Gregory and Hansen technique could also be used with the augmented Dickey-Fuller test, as Gregory and Hansen demonstrated. For the purposes of this paper, we have chosen not to apply this test since it could be sensitive for the presence of autocorrelation.
The estimation for the long-run variance utilizes an automatic bandwidth estimator and a prewhitened quadratic spectral kernel with a first-order autoregression for the prewhitening. For details, Gregory and Hansen (Citation1996) refer us to Andrews (Citation1991) and Andrews and Monahan (Citation1992).
The LM test was used for testing autocorrelation because this test has better size properties compared to other tests according to Hatemi-J (Citation2004).
The above historical information in this paragraph comes from SAF (Citation1998) and Graflund (Citation2001).