BRIC and GIPS – who drives who? Evidence from newly developed asymmetric causality tests

ABSTRACT

We investigate the asymmetric causal interaction between the stock markets of the GIPS (Greece, Ireland, Portugal and Spain) and those of the BRIC (Brazil, Russia, India and China) based on a newly developed asymmetric causality test by Hatemi-J (2012) [Hatemi-J, A. 2012. “Asymmetric Causality Tests with an Application.” Empirical Economics 43: 447–456. doi:10.1007/s00181-011-0484-x]. We confirm a significant stock market interaction between the two blocs in which the BRIC drives the GIPS but not vice versa. Thus, the BRIC seems to be more influential on the GIPS than the GIPS on the BRIC. However, this interaction occurs only during downmarket conditions but not during upmarket times. The BRIC pulls down the GIPS during bad times but does not pull them up during good times. These results have significant implications for international policymakers and provide further evidence on the existence of asymmetric causal interactions between financial markets.

KEYWORDS:

• BRIC
• GIPS
• asymmetric causality test
• stock markets
• bootstrapping

JEL CLASSIFICATION:

• G15
• C32

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Data from the World Trade Organisation reveal that the BRIC had registered a surplus every year from 1996 to 2010 in their trade with the GIPS.

² It should be mentioned that we tested the underlying data set for multivariate normality by using the Doornik and Hansen (2008) test as well as for multivariate ARCH effects by using the Hacker and Hatemi-J (2005) test. The results showed the data set is non-normal and ARCH effects prevail. Thus, using bootstrap simulations with leverage adjustments is necessary in order to create reliable critical values when the asymmetric causality tests are implemented.

³ For asymmetric generalized impulse response functions see Hatemi-J (2014). For dealing with deterministic trend parts the interested reader is referred to Hatemi-J and El-Khatib (forthcoming).

⁴ In conducting tests for causality between negative cumulative components, the vector $P_t^{-}=\left(P_{1t}^{-},P_{2t}^{-}\right)$ is used.

⁵ One additional unrestricted lag is included in the VAR model in order to account for one unit root, as suggested by Toda and Yamamoto (1995).

⁶ The HJC is recommended in Hatemi-J (2003, 2008). Via Monte Carlo simulations, the author showed that this information criterion is successful in picking the correct lag order even if ARCH effects prevail. The conducted simulations also showed that this information criterion has good forecasting properties.

⁷ It should be mentioned that initial values are assumed to be available. For more information on the reason behind this assumption, see Lutkepohl (2005).

⁸ For additional details on the correction of potential heteroscedasticity via leverage adjustment see Davison and Hinkley (1999) for univariate analysis and Hacker and Hatemi-J (2006) for multivariate analysis.

⁹ See Hatemi-J (2011) for the statistical software component.

¹⁰ The results of these diagnostic tests are not presented here in order to save space. These results are available on request.

Additional information

Funding

This work was supported by the UPAR-UAE University [grant number UPAR-2013(2)].