ARCH effects, trading volume and the information flow interpretation: empirical evidence from the Chinese stock markets

Abstract

This study revisits the relation between ARCH effects and trading volume. We extend the specification of the VA-GARCH (1, 1) model by using various volume variants and constructing contrast equity groups. We verify that the information flow assumed to be contained in the four trading volume variants has a starkly different explanatory power compared with the ARCH effects. Successive improvement of the model's empirical fit and the reduction of the fat-tailedness in the model residuals in the sequence of volume adjustment imply an increase in the strength of explaining the static aspects of volatility dynamics by the further adjusted volume variants.

Keywords:

• ARCH effects
• information flow interpretation
• volume variants
• contrast equity group

JEL Classifications:

• G1
• C5

Acknowledgement

This paper is supported by the National Social Science Foundation of China (ID:10zd&035).

Notes

1. Literature on price change limits present conflicting points. The advocates (for example, Blume, MacKinlay, and Terker 1989; Greenwald and Stein 1991) claim that price limits decrease stock price volatility, counter overreaction, and do not interfere with trading activity. Conversely, price limit critics (for example, Fama 1989; Kyle, Kurserk, and Locke; Kuhn et al. 1991; Lehmann 1989) claim that price limits cause higher volatility levels on subsequent days (volatility spill-over hypothesis), prevent prices from efficiently reaching their equilibrium level (delayed price discovery hypothesis), and interfere with trading due to limitations imposed by price limits (trading interference hypothesis).

2. It is expanded to a more general specification of GARCH (1, 1) model with stochastic covariates in Fleming, Kirby, and Ostdiek (2008).

3. Qianlong is one of the most widely used online security trading system in China and it holds market share more than 50% among both institutional and individual investors. See: http://www.qianlong.com.cn/.

4. All the inferences are implemented at 5% significance level.

5. The details of this preliminary study on the dataset are not presented in this paper owing to limited space but can be provided upon request.