Abstract
The volatility of financial markets has long been studied by researchers from a variety of academic background, with ultimate objective being uncovering its underlying cause. Particularly, the volatility of daily stock returns draws much attention of the researchers in empirical finance. Amongst the various theories proposed and numerous empirical techniques developed, the mixture of distributions hypothesis (MDH) provides a formal base that daily return volatility is explainable by trading volume, and the GARCH family models facilitate empirically investigation on the well acknowledged facts of volatility persistence and volatility clustering (ARCH effects). Although the MDH hypothesis experienced extensive empirical and theoretical examination during recent decades, overall evidence is still mixed. None of the empirical findings can be treated as overwhelming conclusion. This study revisit the relations between return volatility, ARCH effects and trading volume within the MDH framework in the context of the Chinese stock market, by employing a three-stage modelling strategy, i.e. applying the volume-augmented GARCH (1,1) model, the two component GARCH model and a GARCH equivalent stochastic autoregressive volatility model. Model specifications are extended by: (1) using four volume variants with each containing its own information contents in the perspective of market microstructure and behavioural finance; (2) constructing contrast equity groups in terms of turnover velocity and market capitalization; (3) embedding a price limit dummy. Empirical investigations are implemented on market index (portfolio) level as well as individual equity level. It is justified that the central prediction of the MDH framework is basically at odds with the data of the Chinese stock market. The VA-GARCH model can not adequately capture ARCH dynamics, i.e. trading volume can not explain ARCH effects. Trading volume is strongly correlated with contemporaneous return volatility, but the correlation is driven largely by transitory volatility shocks that have little to do with the highly persistent component of volatility captured by standard volatility models. Distinctive differences between the four volume variants, between the contrast equity groups, between the two study levels, between the two types of shares, A-share and B-share regarding the bivariate dynamics of return volatility and trading volume are verified. Overall, the complicated relation between return volatility, ARCH effects and trading volume is uncovered to a reasonable extent.