More Powerful Threshold Cointegration Tests

• Activity Number: 32
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Dong-Yop Oh*+ and Hyejin Lee and Junsoo Lee
• Companies: University of Texas Pan American and University of Alabama and University of Alabama
• Keywords: Cointegration ; Efficient estimation ; Asymmetric adjustment ;

Abstract

This paper proposes a modified threshold cointegration tests of Enders and Siklos (2001; ES) in order to improve testing power. We conjecture that an appropriate way to increase the power of Engle and Granger (1997; EG) test will probably lead to higher power for ES test as well; since ES tests can be considered as an extended EG approach to the analysis of cointegration, allowing for the possibility of asymmetry. In particular, we consider a simple modification procedure to resolve the issue related to Common Factor Restriction, which can lead to significant loss of power when the signal-to-noise ratio increases. The asymptotic results yield different range of critical values which depends on the long run covariance estimator between the residuals of tests. Interestingly, it shows that the minimum level of critical value is the same as those of the usual threshold cointegration test. This implies that we are more likely to reject the null to all corresponding cases. Throughout the simulation, we confirm that the power gain of extended threshold cointegration tests is substantial as the assumption that the same long run and short run coefficient become invalid.