Central bank economists have to understand and forecast macroeconomic time series. A serious problem that they face is that those series are often trended or a.ected by persistent innovations to the process. To try to get round this problem, or at least to understand its possible e.ects, it is common to test whether series are stationary. These tests are often called unit-root tests.1 In this handbook we discuss such testing. A model-builder should use appropriate econometric techniques. In order to choose between alternative estimators, the model-builder needs to think carefully about the relevant theory and the available data. But economic theory is rarely unambiguous in its implications for the data generating process. Subjecting the data to pre-estimation testing can help to gauge the relevance of different theories and possible data problems.
