As said by the old Wall Street Adage, “It takes volume to move the price.” For any market securities, when new information begins to disseminate into the market, trading volume will be formed and lead to price change. Therefore, understanding the relation between volatility and volume tends to help investors in making better investment decisions.

In the study of finance, volume-volatility relation has been a popular subject of recent considerate research. Karpoff (1987) provides four reasons that explain the significance of price-change and volume relation. First, empirical relations between prices and volume help to explain the dissimilarities between various market structure hypotheses, henceforth providing an insight to the financial market structure. Second, for studies which use price and volume data to draw inferences from, such relation is vital. Third, the price-volume relation is critical to the debate over the empirical distribution of speculative prices. Finally, the impact of price-volume relations on futures markets research is considerable.

In the past, numbers of studies attempt to explore the contemporaneous relationship between price changes and volume. (Karpoff, 1967) Dynamic price-volume relationship has been frequently tested and examined.

In Gündüz, et al. (2005) paper, dynamic price-volume relationship is tested for stock markets in the Czech Republic, Hungary, Poland, Russia, and Turkey. Through the Granger causality test, the results conclude a continuing and steady relationship between stock prices and volume, except for the case of Czech Republic. In both Hungary and Poland cases, it is found that there is bidirectional causality between stock prices and volume. However, for Russia and Turkey cases, results show that there is only unidirectional causality from stock prices to volume.

Chiang et al. (2010) uses intraday NASDAQ data to go through the linear and nonlinear Granger causality tests in order to examine the causal relationship between volatility and volume. The results from the linear Granger causality tests reveal that volume does not Granger-cause volatility but there is an ambiguous causality relation for the reverse direction. On the contrary, strong bidirectional casual relation between return volatility and trading volume is found through the use nonlinear Granger causality tests. Our results provide strong support to the sequential information theory, and our forecasting results show that it is useful to use lagged values of trading volume to predict return volatility with nonlinear models.

Lee and Rui (2002) study the casual relation between volume and volatility for both domestic and cross-country markets by using the daily data of the three largest stock market: New York, Tokyo, and London. They also find that trading volume does not Granger-cause stock market return on each of three stock market. Meanwhile, there exists a positive feedback relationship between trading volume and return volatility in all three markets. That is volume helps predict return volatility and vice versa. US financial trading volume contains extensive predictive power for UK and Japanese financial market variables. Sub-sample analyses show evidence of stronger spillover effects after the 1987 market crash and an increased importance of trading volume as an information variable after the introduction of options in the US and Japan.

Llorente et al. (2002) analyzes the daily volume-return dynamics of individual stocks traded on the NYSE and AMEX. The empirical results support the predictions of the model on the nature of the dynamic volume-returnr elation. Stocks that are associated with a high degree of informed trading exhibit more return continuation on high-volume days, and stocks that are associated with a low degree of informed trading show more return reversals on high-volume days. Our results are robust to various econometric specifications, potential data problems, alternative definitions of volume, and changes in the lengths of the measurement intervals and the estimation period. The empirical findings support the general notion that volume does tell us something about future price movements. The analysis also suggests that the actual dynamic relation between volume and returns depends on the underlying forces driving trading. Explicitly modeling these driving forces allows us to use volume effectively in making an inference about returns.

Chen et al. (2001) empirically examine the dynamic (causal) relation between stock market returns, trading volume, and volatility. positive correlation between trading volume and the absolute value of the stock price change. Granger causality tests demonstrate that for some countries, returns cause volume and volume causes returns. Our results indicate that trading volume contributes some information to the returns process. The results also show persistence in volatility even after we incorporate contemporaneous and lagged volume effects. The results are robust across the nine national markets. U.S. Japan, the U.K, France, Canada, Italy, Switzerland the Netherlands, and Hong Kong. Our results suggest the EGARCH models reflect an appropriate representation of the returns in stock index data. Our evidence indicates that trading volume contributes some information to the returns processes of stock indexes. However, in contrast to Lamoureux and Lastrapes (1990), we find that the persistence in volatility remains even after incorporating contemporaneous and lagged volume effects (both of which are proxies for information flow). Our findings suggest that more can be learned about the stock market through studying the joint dynamics of stock prices and trading volume than by focusing only on the univariate dynamics of stock prices. We find that our results are robust across all nine major stock markets, implying that there are similar returns, trading volume, and volatility patterns across these markets.

According to McMillan (2007), significant amount of evidence shown that there is a negative relationship between trading volume and returns. He quoted Wang and Chin (2004), low volume tends to typified by momentum behavior while high volume exhibits reverting behavior in return. Daily national stock index and volume data were obtained for the UK, US, France and Japan. when volume is low returns do indeed exhibit positive serial correlation or momentum behavior, whilst when volume is high returns appear to exhibit random behaviour or weak reversion.

Wang and Chin

strong evidence of predictable patterns of cross-sectional returns significant momentum profits present in low-volume stocks but not in high-volume stocks. The characteristics of China’s stock market, for example, the short-sales prohibition and the dominance of unsophisticated individual investors, imply that China’s stock market is more prone to behavioral concerns, greatly influencing the return–volume dynamics. Our findings suggest that the informational content of past prices and past trading volume for future market movements can be market specific. Therefore, additional out-of sample evidence is certainly beneficial for academics to better understand volume–return dynamics in asset markets. Our results on the interaction between return predictability and trading volume tend to be consistent with the behavioral finance theories

What models are more accurate in predicting volatility?

Brooks (1998) has outlined several incentives for the line of inquiry. First, volatility is used as a rudimentary measure of the risk of a financial instrument by the computing the standard deviation or variance its returns. Second, volatility of underlying stock price is one of the important elements in the option pricing model, Black-Scholes formula. Lastly, volatility trading may not be implausible if combinations of options are used.

There is a growing body of evidence claims that sophisticated time-series models contribute to more precise forecasts compared to a primitive one. In the past, measures of volatility was simpler and mostly univariate in nature.

McMillan (2007),He also reported that among its models, there are two smooth transition non linear models better than the linear models. linear random walk and AR models, the alternative univariate (LSTAR) non-linear model and the TAR and MTAR non-linear models, in three of the four cases. LSTR-forecast volume model appears to provide positive trading returns across all series and could possibly act as a generic trading rule,

Examine, test, study, analyze, access, investigate, consider, question, explore, experiment, put to the test, validate, verify, determine, report