Granger-Kausalitätsprüfung : eine anwendungsorientierte Darstellung

For the testing of Granger-causality stationarity of data is obligatory. But economic time series data often contain trend and/or seasonal factors. Several procedures are known to model or eliminate these influences. Unit root tests are often used to construct trend- or differencestationary processes in a VAR-model. Considering the error correction approach (VARECM) it is possible to distinguish between short- and long-run Granger-causality. Toda/Yamamoto reveal a way how traditional tests for Granger-causality can be used in a VAR-model with the original variables.