Magdalena Osinska, Ph.D., Department of Econometrics, Nicolaus Copernicus University of Torun, Gagarina 13 A, Torun, 87-100, Poland
The problem concerns out-of-sample Granger causality testing developed by Chao, Corradi and Swanson (2001) as well as by Clark and McCracken (2001). The question is whether relationships indicated as causal within the sample can still be considered in the same manner out-of-sample. This is related to the problem of stability of causal relations in economics, or more precisely and widely, their forecasting ability. We act within the nested models framework. The problem is considered in two ways: the first one intuitive, based on rolling or recursive application of the Granger causality test and the second one, based on the formal statistical background. The mentioned methodology is applied to macroeconomic and financial phenomena, observed in the last years in Poland.