Intraday bidirectional volatility spillover across international stock markets: does the global financial crisis matter?

Abstract

This article studies volatility spillover between the US and the three largest European stock markets (Frankfurt, London and Paris) around the time of the recent Subprime crisis. In order to investigate the impact of the latter, we break our sample down into two sub-periods: a pre-crisis period and a post-crisis period, using a structural break test that has the advantage of endogenously testing for further breaks in the data. Unlike previous studies that have frequently investigated this issue using low frequency data, our article makes use of intraday data. Accordingly, using Threshold generalized autoregressive conditional heteroscedasticity (GARCH) model estimations, we find weak evidence of volatility transmission between the two regions before the Subprime crisis. However, during the post-crisis period, we record returns and volatility spillover from US to European markets and vice versa at different times of the trading day, indicating that the two regions became more dependent during the recent Subprime crisis, a finding that supports the contagion hypothesis between the US and European stock markets.

Keywords:

• volatility spillover
• threshold GARCH model
• intraday analysis
• Subprime crisis

JEL Classification:

• G10
• G15

Acknowledgement

This article was presented at the Third International Symposium in Computational Economics in Finance (ISCEF, Paris, 10–12 April 2014, www.iscef.com). We would like to thank the ISCEF participants for their comments and the two anonymous referees for their constructive comments and suggestions on an earlier version of the article. All further errors or omissions are ours.

Notes

¹ For the DAX30, we consider only Xetra trading hours, that is, the period between 8 am and 4.30 pm (GMT). Nevertheless, after the end of Xetra trading, the DAX is traded based on floor trading until 9 pm GMT (Late-DAX). Moreover, the X-DAX is traded between 7 am and 8 am GMT and between 4.45 pm and 9 pm (GMT) on the basis of the DAX futures.

² Our test detected different breaks, but some breaks are not statistically significant. In order to divide our sample into two sub-samples (before and after the subprime crisis), we retained only one break that this the most significant. Results for structural break tests are not reported but available upon request.

³ We would like to thank the anonymous referee for this suggestion that considerably improved our analysis.

⁴ In all our models, we only present the specification for the CAC40 as an example to save space. The same model is estimated for the DAX30 and the FTSE100 and the relevant specification can be provided upon request.