The Impact of Volume on Daily Follow-Through
In the last post we looked at the recent impact of volume on SPY returns. In the last 60 days, this variable was more predictive than many other short-term variables or indicators. As it turns out, the short-term volume effect is persistent across 3000 bars. Average daily returns are six times higher than buy and hold (.01%) when SPY volume is higher than 5 days ago (5d ROC V>0), in contrast returns are much lower when SPY volume is lower than 5 days ago (5d ROC V<0). This effect is not some feature unique to using ROC as an indicator, similar effects are seen using the simple 5-day moving average of SPY volume.
But this is not the most interesting aspect of the effect. In fact the most interesting effect is that short-term trends in volume have a significant impact on daily follow-through. Returns to a standard mean-reversion strategy (buy on down days, sell on up days) were virtually subsumed by the impact of this short-term volume effect. Interestingly, the linkage is not connected in a statistically significant manner to factors such as historical volatility. As such, this is a unique and additional source of information. In the tables below, you can see that the volume effect was stronger in moderating the returns for buying after down days. When volume was rising over the past week and today was a down day, returns were much stronger than average. When volume was falling over the past week and today was a down day, returns were actually negative despite the fact buying after down days has delivered a solid edge over time. The logical explanation for this effect is that rising volume is a proxy for either buying interest or a form of capitulation. Falling volume is a sign of a lack of buying interest, and certainly does not indicate panic-selling. Thus buying after down days in these situations is riskier than average. But what about the impact of volume on up days? In fact there was little or no relationship between volume and buying/selling after up days over the last 3000 bars. However, in the last 3 months as we showed in the last post the volume effect on up days is in the predicted direction–falling volume is negative and rising volume is positive.
|5d ROC V>0||V>5smaV||5d ROC V<0||V<5smaV|
|Average Daily Return||0.06%||0.05%||-0.04%||-0.02%|
|Avg Daily Return vs Vol||5.57%||4.97%||-3.82%||-2.37%|
|down day||up day|
|Average Daily Return||0.06%||-0.04%|
|Avg Daily Return vs Vol||6.12%||-3.96%|
|5d ROC V>0||5d ROC V>0||5d ROC V<0||5d ROC V<0|
|down day||up day||down day||up day|
|Average Daily Return||0.15%||-0.04%||-0.03%||-0.04%|
|Avg Daily Return vs Vol||17.95%||-5.87%||-5.26%||-5.36%|