Crowdedness for divergent and convergent strategies - How can it be measured?

Crowding is increasingly important topic with discussion of factor strategy returns. Not all factor strategies will behave the same from increases in crowdedness. Greater crowdedness will affect divergent and convergent strategies differently in the short and long-run. We have discussed the concept of convergent and divergent strategies (see: The "3 x 5 index card" on "Divergent" and "Convergent" hedge fund strategies), but the paper, "The Impact of Crowding in Alternative Risk Premia Investing", by Nick Baltas takes a different perspective and explains how different strategies will be affected by increased interest, crowdedness. 

A momentum strategy does not have a fundamental anchor. It is mean-fleeing and will gain from divergences through continued trend following. Hence, new flows can be a self-reinforcing mechanism which can lead to destabilizing behavior and a potential crash. In this case, more new capital will actually increase returns in the short run but will lead to distorts in performance the longer a trend lasts. 

On the other hand, a convergent strategy like value investing has a natural anchor, the fundamental valuation. In this case, there is a self-correcting mechanism, the long should increase and short decrease to close the value gap. As more capital flows into a value strategy the value gap should be closed. 

In both cases the turnover from rebalancing tells us something about performance. In the case of momentum, a positive feedback strategy, lower turnover will have higher performance from reinforcing behavior. This will not exist for value strategies.

Now we could measure crowdedness through money flows, but this information is difficult to obtain. However, we can learn something about crowdedness by the pairwise correlation of assets in the long and short portfolios constructed to make momentum and value factor strategies. In the case of momentum, if there is more pairwise correlation, a sign that investors are holding the same portfolio and thus crowded, there should be lower returns over time. A low pairwise portfolio should have higher returns. The research finds this result across different asset classes. It also finds the opposite result for value portfolios; high pairwise correlation in a portfolio suggests similar behavior that will lead to self-correction.

Everyone thinking the same for a momentum portfolio will start well but ultimately end with a bad result. On the other hand, acting the same will close the value gap for names in a value portfolio which will lead to more turnover and new opportunities to again exploit value.