Choosing Time-Series Models for Trading Volume
Summary
The document considers whether there is a standard econometric model for trading volume and argues that model choice should follow the properties of the particular market and security. It suggests examining trends, mean reversion, persistence, and seasonality before choosing a specification. AR or GARCH models may suit data with mean-reverting or persistent behavior, while seasonal indicators can represent calendar effects. One response also suggests fitting separate models by security rather than imposing a universal form.
Other proposed approaches include ARIMA, exponentially weighted averages, and ARMA after a logarithmic transformation; long-memory models are raised if simpler models do not fit. Intraday data may need adjustment for time-of-day patterns, and daily data can have expiry-related or event-related effects. These are suggestions from responses, not a comparative study: the document supplies no systematic evaluation, forecasting results, or evidence that one model works best across markets.
Key ideas
- There is no universal volume model because volume behavior differs across securities and markets.
- Inspect trend, persistence, mean reversion, and seasonal patterns before selecting a model.
- AR, GARCH, ARIMA, ARMA, and exponentially weighted approaches are suggested as candidates.
- Intraday volume may require time-of-day seasonal adjustment, while daily data can have other calendar effects.
- The recommendations are empirical suggestions rather than results from a model comparison.
Tags
Full text
# What is a commonly accepted econometric model for volume? # What is a commonly accepted econometric model for volume? What is the gold standard econometric model for volume? Base model for price changes is the autoregressive (AR) model and GARCH(1,1) for volatility. Is there any survey about econometric models used in volume modeling ? ## Answer by Ram Ahluwalia (score 9) https://quant.stackexchange.com/a/3522 GARCH will work if volume has memory with some decay. AR will work if volume has mean reversion properties. Both of these are empirical questions and depend on the market. You should also consider if there are seasonal (day-of-week, monthly, quarterly effects) in which case you would want to add dummy variables. MA models will work well if volume behaves like a random-walk (not the case). There is no "gold standard" since markets have different volume characteristics (for example, emerging markets have rising volumes; developed markets more recently are seeing less volume traded year-over-year with the rise of crossing networks and dark pools). I would start by observing the volume of interest to see what properties hold (i.e. trend, mean reversion, persistence, seasonality) for starters. You may want to consider using the auto.arima function in the package forecast to fit a volume model to each security rather than looking for a global functional form for all securities. ## Answer by aajajim (score 3) https://quant.stackexchange.com/a/9512 I deal recently with some analysis of the Volume time series, daily volume in € for European stocks. I found out that an ARIMA model works well. But, some EWMA could also provide good forecast if it's well parameterized. You can also face some seasonality effect due to macroeconomic events, some you may need to clean you data and treat these days in a different way. ## Answer by htrahdis (score 2) https://quant.stackexchange.com/a/9518 Try the following : - perform the logarithmic transformation of the volume data. - check if the transformed data fits the normal distribution nicely. - if you are working with intraday volume, then adjust for the seasonality for time of the day effect, if using daily data, in some cases some special seasonalities like expiry day, etc might be applied but it may not be compulsory. - fit an ARMA model. - if you are still not satisfied, try using a long memory process.
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.