Three essays on causality

Abstract

Nowadays, Granger causality tests are standard tools to investigate causal relationships between financial and economic time series. Econometric advances in the field have shown that the causal relationship between two variables is not invariant to the integration and cointegration properties of the processes nor the relevant information that is available and included in the analysis. Hence, various notions of Granger non-causality are developed in the context of linear bivariate or multivariate stationary or nonstationary discrete time processes. Several of these causality concepts are reviewed in this thesis. Their extended concept is contrasted to the standard Granger causality concept. A wide range of causality tests have been used to investigate the independence between the second moments of the time series. There is currently much interest in testing causality-in-variance by policy makers, portfolio managers, and academic researchers. […]