Granger causality testing for Argentina MERVAL index and the major world stock markets

Abstract

In this paper are analyzed the causal links among a selected group of global stock market indices, with special focus on the role of Argentina MERVAL index. With this objective in mind, two types of non-conventional Granger causality test are performed in order to avoid the theoretical limitations of the traditional test which requires stationary time series. The first test is based in a surplus-lag VAR model and allows testing for Granger causality in thecontext of non-stationary processes. The second test rests on the estimation of a VARX model and is robust to non-stationarity; long memory; and non-modeled structural breaks. This second test also admits conditioning on endogenous modeled control variables. The estimations are performed using daily data for a long time period, being both testing procedures implemented in the programming language R. Finally the results from both tests are compared and interpreted in order to capture their economic meaning.