We propose a sequential test for predictive ability for recursively assessing whether some economic variables have explanatory content for another variable. In the forecasting literature it is common to assess predictive ability by using “one-shot” tests at each estimation period. We show that this practice leads to size distortions, selects overfitted models and provides spurious evidence of in-sample predictive ability, and may lower the forecast accuracy of the model selected by the test. The usefulness of the proposed test is shown in well-known empirical applications to the real-time predictive content of money for output and the selection between linear and nonlinear models.