Abstract
A method of information-criterion-based cointegration detection using dynamic factor models is proposed. The results of the data-based and non data-based Monte Carlo simulations suggest that this method is as effective as conventional hypothesis-testing methods. In the proposed method, an observed multivariate time series is described in terms of common stochastic trends plus stationary autoregressive cycles. Then the best model is selected from among alternative models obtained by changing the number of common stochastic trends, on the basis of information criteria. Consequently, the cointegration rank is determined on the basis of the selected model. Two advantages of the proposed method are also discussed.
Acknowledgment
The author is grateful to the associate editor for helpful comments. Needless to say, any remaining errors are mine.
Notes
Note: The figure in the parentheses stands for standard error.
Note: *shows the frequency count of the correct selection.
Note: *denotes a rejection of the null hypothesis at the 5% level. R i denotes the ith interest rate.
Note: * denotes a selection of the model using the minimum AIC or BIC procedure.
Note: Di denotes DGP i and Mi denotes Model i.
Note: Di denotes DGP i and Mi denotes Model i.