Killing four unit root birds in the US economy with three panel unit root test stones

Abstract

This study tests for the presence of unit roots in four US macroeconomic time series using panel unit root tests. The Im, Pesaran and Shin (Journal of Econometrics, 115, pp. 53–74, 2003) test, the Multivariate Augmented Dickey-Fuller test (Taylor and Sarno, Journal of International Economics, 46, pp. 281–312, 1998) and the Johansen (Journal of Economic Dynamics and Control, 12, pp. 231–54, 1988) likelihood ratio test are applied to unemployment, the real exchange rate, the nominal interest rate and inflation. The three tests all have ways of controlling the obvious cross-sectional dependence in the panel. Using monthly data from 1960 to 2002 there is evidence that all time series are generated by stationary processes.

Acknowledgements

Financial support from Jan Wallander's and Tom Hedelius’ foundation and Sparbankernas forskningsstiftelse is gratefully acknowledged.

Notes

¹ See for instance O’Connell (1998) and Maddala and Wu (1999).

² The results from both these tests are robust to the choice of lag length; regardless of whether the lag length was set to 11, 12, 13, 14 or 15, the result was still found to be the same.

³ The same lag length was also chosen using a likelihood ratio test and a final prediction error criterion.

⁴ See for instance Cheung and Lai (1993).