This paper demonstrates that all rank test statistics are functions of implicit null space estimators. The paper proposes a novel theory of null space estimation that allows for standard asymptotics, polynomial regressions, and cointegration asymptotics. The paper proves that the behaviour of rank test statistics is completely governed by the implicit null space estimators through a plug-in principle. This allows for a general theory of rank testing that simplifies the asymptotics of rank test statistics, clarifies the relationships between the various rank test statistics, makes full use of the numerical analysis literature, and motivates numerous new rank test statistics. A brief Monte Carlo study illustrates the results.
