Visual, Unit Root and Stationarity Tests and Their Power and Accuracy

The study at hand concentrates on existing stationarity tests as well as some of their variants and generalizations. It also focuses on the results of applying time series methods to univariate non-stationary data in order to stress the importance that stationarity plays in generating accurate and reliable estimates and forecast models to describe the data. Part A of the study focuses on various stationarity and unit root tests such as visual tests, correlograms, the Dickey-Fuller test and its augmented version, the Phillips-Perron test, as well as the KPSS test and some generalizations of it. The tests are then applied, in Part B, to a few demographic time series data sets in order to assess the stationarity assumption and the eventual discrepancy between the various test results. Finally Part C of the study focuses on the power and accuracy of these tests when the autoregressive coefficients as well as the number of observations in the sample take on different values. 1 A Review of Existing Stationarity Tests