Is there any theory which is able to unify and/or falsify existing explanations on the causes of asset price/return momentum? The prevailing theory is that behavioral and cognitive biases lead to momentum, but I strongly suspect that momentum has multiple causes and this conflation between variables explains why market theoreticians and practitioners have such a difficult time explaining it. For example, I find it highly plausible that the causes of observed momentum in individual securities differ greatly from that which is observed at the industry/sector and asset class levels.
I am aware of the definition of momentum, the statistical arguments for its existence, as well as the numerous ways in which momentum can be measured. I am also aware of some of that it has been observed for a number of asset classes including equities, sectors of equities, commodities, and others. Furthermore, I am aware of some of the theoretical underpinnings for momentum. However, I am not aware of any unifying theory which is able either subsume and/or falsify existing arguments.
Some of these theoretical arguments are summarized below:
- Informational asymmetry (e.g., "price leads fundamentals").
- Limits to market efficiency (e.g., slower diffusion of information than expected under EMH; institutional constraints; liquidity constraints; etc...)
- Various behavioral arguments (e.g., cognitive dissonance; sentiment; greed/envy; bandwagon effect; "under(over)-reaction to good (bad) news")
- Style/factor/asset-class rotation (e.g., reaction to the confluence of economic/business/market cycles)
- Capital gains deferral (i.e., tax loss selling)
While existing narratives all sound plausible, they are neither broadly encompassing nor falsifiable. I.e., none are mutually exclusive and/or inclusive. While I am willing to accept that momentum is simply the confluent result of several distinct factors and interactions between factors (in which case there would be no broader explanation), I also believe that there should be a way to better articulate observed phenomena.
I ask because I am trying to figure out a practical solution to a factor-based returns attribution model. Specifically, I am trying to determine whether to group momentum under an existing classification, make it its own classification (i.e., it subsumes multiple classifications), or include different aspects of momentum under multiple classifications (e.g., security price momentum under asymmetry; sector momentum under sentiment/rotation; etc..).
As a bonus question, how does any of this articulate with observed mean reversion?