Cointegration analysis and the choice of lag length

This article investigates the effects of the choice of lag length on the estimation of long run cointegration relationships using the Johansen estimation procedure. This issue is of particular interest to applied researchers using time series data, see for example Awokuse (2005 Awokuse, T. 2005. Exports, economic growth and causality in Korea. Applied Economics Letters, 12: 693–6. [Taylor & Francis Online], [Web of Science ®] , [Google Scholar]), Bacchiocchi et al . (2005 Bacchiocchi, E, Florio, M and Grasseni, M. 2005. The missing shock: the macroeconomic impact of british privatizations. Applied Economics, 37: 1585–96. [Taylor & Francis Online], [Web of Science ®] , [Google Scholar]), Gallegati (2005), Gomes and Paz (2005 Gomes, FAR and Paz, LS. 2005. Can real exchange rate devaluation improve the trade balance? the 1990–1998 Brazilian case. Applied Economics Letters, 12: 525–8. [Taylor & Francis Online], [Web of Science ®] , [Google Scholar]), Hasan (2005 Hasan, M. 2005. The information content of M0 in the United Kingdom. Applied Economics Letters, 12: 711–7. [Taylor & Francis Online], [Web of Science ®] , [Google Scholar]) and Pieroni and Ricciarelli (2005 Pieroni, L and Ricciarelli, M. 2005. Testing rational expectations in primary commodity markets. Applied Economics, 37: 1705–18. [Taylor & Francis Online], [Web of Science ®] , [Google Scholar]), among many others. An empirical example is used to demonstrate some of the issues that applied researchers face when they wish to use cointegration analysis. First, the number of lags to include in the model must be determined. However, different lag length selection criteria often lead to a different conclusion regarding the optimal lag order that should be used. Second, as demonstrated in this article, the choice of lag length can drastically affect the results of the cointegration analysis.