We propose a new no-change benchmark to evaluate forecasts of series that are temporally aggregated. The new benchmark is the last high-frequency observation and reflects the null hypothesis that the underlying series, rather than the aggregated series, is unpredictable. Under the random walk...