The article examines two common tests used to check for long-term relationships between variables. It finds that one test is more reliable than the other in certain situations, such as when variables have long-memory properties or when their error covariance matrix is singular. The study suggests using both tests to avoid potential pitfalls in cointegration analysis.