Main reason is the practical one: if you can predict returns, you can earn money. Predicting volume might have some benefits, but much smaller than predicting returns.

@Peter: Let's not forget Return by itself has not predictability, it is lagging indicator.

It's just a measure.

Volume is rather a leading indicator. Think about it, if there is low breadth then it must be only few investors investing in huge volumes because either they are market makers or they are informed traders.

So, Volume has high info value to predict a value sensitive disclosure.

It would be interesting to rank-order predictability based on volume and price under varrying conditions of market efficiency. Or rather volume and price together provide an efficiency-adjusted ptedictor of returns. Since  in a strongly efficient market, predictability is close to impossible. As the conventional notion of efficient markets is said to be tautological.

https://uneasymoney.com/2015/08/30/excess-volatility-strikes-again/

@Shian: Yes, Rank ordering Price & Volume is insightful view. Since Volume and Price can have many Combinations in terms of Trend & Liquidity, it's best to to do rank ordering at Combination level (segment). 

Dear Ankur,

Especially when volume may shed some light on momentum, which in turn provides some indication as to the predictability of returns.

http://lsvasset.com/pdf/research-papers/Price-Momentum-Trad-Vol-2000.pdf

@Shian: I have read and heard a lot that if the momentum is not supported by an abnormal volumes then the Prices will revert to Mean. Based on my understanding I would simply refute this myth. 

One possible reason could be, there was liquidity crunch / low liquidity.

The reason why i said this is, without normal/ abnormal volume how can there be momentum? Hence, Looking at Momentum alone is half job done, unless it's interpreted with Liquidity Profile of those trading sessions.

Both together has higher info value for short-term abnormal return.

Dear Ankur,

Do agree with the view that  we are often just scratching the tip of the iceberg with models and predictions. Often more so towards the end of their shelf life. Sounds tautological, but due to reflexivity in the financial markets (this idea is partly foreshadowed by EMH), every "new" factor we isolate as pertinent for the times will eventually be endogenized as part of the specified model. More so if the idea catches on and the real economy enters a phase that makes such conceptualization reasonable. So that the tasks of modeling the financial markets is one of pushing the envelope of the known and acceptable. To that end, contrarians/technicians will always be rewarded for taking a bet on what seems to be mere noise. Not forgetting as you have highlighted, the need for sufficient liquidity.Either supplied by an individual or precipitated by a bandwagon effect.So what we recognize as volume (or an abnormal volume in the form of an event) and momentum is actually a signal (our ability to make sense of what others take for noise). Timing of an analysis is also important as many have observed that one cannot recognize a bubble when one is in it. It is often when the bubble has burst that we recognize it as such. Neither is prediction easy because the makings of bubbles are often veiled by inexplicable events (the unknown/unacceptable), at the inception of what some will eventually recognize as a momentum.