Granger causality tests in the presence of structural changes

Granger causality tests are widely used in applied economics as a way of establishing if a variable has been a leading indicator of another over the past. However, like most statistical tests, Granger causality tests require that the relationship between the variables remains stable over the sample period being tested. This paper illustrates that, if significant structural change occurs, these tests can provide misleading results. The paper then goes on to describe a statistical method that identifies structural breaks in a given data sample. Having identified them, Granger Causality tests are adjusted to make them 'robust' to those breaks. The paper also presents an application of the method to Canadian GNP and M1.