Linear and Nonlinear Granger Causality: Evidence from the Ft-Se 100 Stock Index Futures and Cash Markets
Posted: 20 Dec 1998
Date Written: September 1994
Several studies have examined the nature of the lead-lag relationship between returns in the stock index futures and the cash market. In a perfect market, both assets, which reflect the same underlying value, should react simultaneously to new information. In practice it is observed that the futures market largely leads the cash and this has been ascribed to the presence of market frictions and differing institutional structures in the two markets. This lead-lag relationship has been tested using the traditional linear concept of Granger causality. However recent evidence suggests the existence of significant non-linear behaviour in the return generating process in both markets. This study uses five minute matched data on the FT-SE 100 Index futures and cash index returns to examine the linear causal as well as the presence of any nonlinear causal relationship between these two markets. The paper uses a new test for nonlinear Granger causality proposed by Baek and Brock (1992) and modified by Heimstra and Jones (1994). The results of the linear Granger causality tests are similar to earlier work. However the results of the nonlinear Granger causality are in sharp contrast and indicate evidence of bi-directional nonlinear causation between the two markets both for residuals of the linear Granger regression tests as well as residuals of linear tests on ARMA filtered returns. However after using an EGARCH filter on both the return series, the presence of a bi-directional nonlinear causal effect is somewhat weaker though still significant. The results suggest the possible misspecfication of the cost-of-carry model in the presence of market frictions and provide an empirical basis for incorporating possible nonlinear effects in a model linking the futures and cash markets.
JEL Classification: G10, G13, C10
Suggested Citation: Suggested Citation