“nonparametric Nonlinear Causality Testing with Stepwise Multivariate Filtering”

The present study investigates the linear and nonlinear causal linkages in exchange, equity and derivatives markets. Specifically, in case of exchange markets, among six currencies denoted relative to United States dollar (USD), namely EUR, GBP, JPY, CHF, AUD and CAD. The prime motivation for choosing these exchange rates comes from the fact that they are the most liquid and widely traded, covering about 90% of total FX trading worldwide. The data spans two periods (PI: 3/20/1991 – 3/20/1997, PII: 3/20/2003 – 3/20/2007) before and after the structural break of the Asian financial crisis, which set a platform for departure for causality testing. In case of derivatives markets, between daily spot and futures prices for maturities of 1, 2, 3 and 4 months of West Texas Intermediate (WTI) crude oil. The data cover two periods October 1991-October 1999 and November 1999-October 2007, with the latter being significantly more turbulent. Finally in case of stock markets, among NIKKEI 255, FTSE100 and NYSE. The data spans four periods (PI: 3/20/1991 – 3/20/2007, PII: 3/20/1991 – 3/20/1997, PIII: 3/20/2003 – 3/20/2007, PIV: 3/20/2000 – 3/22/2004), namely the total period, a preand post-bubble period and an in-bubble period. The structural break point is again the Asian financial crisis. We apply a new nonparametric test for Granger non-causality as well as the conventional linear Granger test on the return time series after controlling for cointegration. In addition to the traditional pairwise analysis, we test for causality while correcting for the effects of the other variables in a multi-variate formulation. To check if any of the observed causality is strictly nonlinear in nature, we also examine the nonlinear causal relationships of VAR or VECM filtered residuals. Finally, we investigate the hypothesis of nonlinear noncausality after controlling for conditional heteroskedasticity in the data using a GARCHBEKK model. Our approach allows the entire variance-covariance structure of the time series interrelationship to be incorporated in order to explicitly capture the volatility spillover mechanism. In case of exchange markets, whilst the nonparametric test statistics are smaller in some cases, significant nonlinear causal linkages persisted even after GARCH filtering during both the preand post-Asian crisis period. This indicates that currency returns may exhibit asymmetries and statistically significant higher-order moments. Similar 1 E-mail addresses: [email protected] , [email protected] (corresponding author); Tel.: +39-055-4685-698. This study was conducted at the Center for Nonlinear Dynamics in Economics and Finance (CeNDEF), Department of Quantitative Economics, University of Amsterdam.