Much research has empirically examined the positive relationship between volatility and trading volume. Nonetheless, a consensus on the relationship has not been reached. In search of a feasible cause for this mixing outcome, this notes that the exist ...
Much research has empirically examined the positive relationship between volatility and trading volume. Nonetheless, a consensus on the relationship has not been reached. In search of a feasible cause for this mixing outcome, this notes that the existing literature ignores the characteristic of surprising information that influences the relationship negatively. Therefore, this report has applied the concept of surprising information with the sign effect to modify the MDH theory. The report has also, from a practical perspective, proposed a suitable method for picking up on unobservable surprising information on the basis of the quantile regression of trading volume on realized volatility.
To detect surprising information in empirical work, we have used four-minute frequency data on the KRW/ USD spot exchange rates, covering the period between April 9, 2004, and October 31, 2006. Trading-volume data in the Korean foreign exchange market are systematically reliable. Further, due to the sign effect of return shock, generally observed in real markets, it is plausible to classify surprising information into two types: positive surprising information and negative surprising information. Therefore, the information-type-switching GARCH(1,1)-V model has been estimated to analyze the relationship between volatility, trading volume, and the GARCH effect and to verify whether the modified MDH with surprising information is reliable in empirical work. The estimation results worth mentioning are as follows: First, absolute return residuals widely used to estimate daily volatility are noisy estimates as compared to realized volatility. Second, from estimating the quantile regression model of realized volatility on trading volume at six representative quantiles , we discover that, as we pass from the median to upper quantiles, the relationship coefficient becomes less significant while its estimate tends to increase consistently. This finding is quite consistent with the modified version of the MDH in which surprising information that influences the relationship negatively, is distinguished from general information. Further, as expected, the negative effect of negative surprising information on the relationship is greater than that of positive surprising information. Third, trading volume alone cannot sufficiently absorb the GARCH effects. This may arise because although the trading-volume series is generally regarded as a proxy for the rate of general information arrival, for surprising information, it becomes a poor proxy because it cannot account for the arrival rate of surprising information that has conflicting effects on trading volume and volatility. As such, our empirical results are sufficiently supportive of the modified version of the MDH with surprising information classified into two types: positive and negative surprising information..
With respect to future research, our study suggests that it would be interesting to replace the variance equation in GARCH(p,q) models with a sophisticated nonlinear specification that sufficiently accounts for the role of surprising information in a volatility model. We might also consider a stochastic volatility representation for returns as a more reliable model accounting for surprising information arrivals.