Time Series Analysis and Volume

Generic
1

Hi there:



 I was wondering, since Volume is a measure of the liquidity of a stock, and  Stocks are usually modeled using ARIMA time series models, How can we incorporate the volume information to have a more accurate  forecast of a stock?? is that possible??





Thanks 

2

Hi Ghery, 



First, let's question ourselves –how would volume affect the stock prices?



As you have mentioned, at the most fundamental level, volume can represent the liquidity of a stock. 



But, does liquidity impact the stock price?



In a way, YES!. 



In general, lower liquidity usually results in an increase in volatility.

Similarly, an increase in liquidity makes the price moment stable and decreases the volatility.



However, this is not always the case and only a substantial increase/decrease in the liquidity can affect the volatility. 



So, we can take the following approach to incorporate the effect of volumes on volatility and improve the prediction of stock returns using the ARIMA model.



Step-1:



Check if the errors of the ARIMA model are homoskedastic or not.

If they are heteroskedastic, then we can conclude that the variance in error terms ( representing volatility of stock returns) is possible and this variance should be modelled.



You can check this by checking if the error terms of ARIMA follow the GARCH structure.



Step-2:



Identify the order of the GARCH structure and fit the GARCH model of that order on the error terms of the ARIMA model.



Step-3:



Predict the error term using the GARCH model and add it to your predicted value from the ARIMA model.



Instead, you can also fit the ARIMA+GARCH model in one shot after confirming that the ARIMA error terms have a GARCH structure.



In this way, you can capture the volatility caused by the volumes and improve the results of stock returns predicted by ARIMA model.



(It should be noted that the volumes can't explain all variance in the error terms, however, using the above approach, you can capture the variance caused by the volumes while predicting stock returns with ARIMA)



Please check this paper in which the team added volume as an additional explanatory variable in the GARCH model to examine if volume can capture GARCH effects. (Equation 8 for GARCH model with volume variable)



References:



a. The bivariate GARCH approach to investigating the relation between stock returns, trading volume, and return volatility



b. GARCH AND VOLUME EFFECTS IN THE AUSTRALIAN STOCK MARKETS



Hope this helps!!



Thank you

3

Thanks to you… I'll review those links…