1: \begin{abstract}

2: This paper presents necessary and sufficient conditions for tests to have

3: trivial power. By inverting these impractical tests, we demonstrate that the

4: bounded confidence regions have error probability equal to one. This

5: theoretical framework establishes a connection among many existing

6: impossibility results in econometrics, those using the total variation

7: metric and those using the L\'{e}vy-Prokhorov metric (convergence in

8: distribution). In particular, the theory establishes conditions under which

9: the two types of impossibility exist in econometric models. We then apply

10: our theory to Regression Discontinuity Design models and exogeneity tests

11: based on bunching.

12: \end{abstract}

13: