In this model by Capital Fund Management (www.cfm.fr), high levels of copying behavior leads to bubbles and then crashes. This is regardless of positive or negative news stories.
Although only a mathematical model, this is interesting to me. The key here is the definition of low vs high level of copying behavior. If an inflection points exists for copying behavior, then what is that inflection point and how does it reflect itself in the market. For example, can large changes in volume or open interest reflect mimicking behavior? Or do you need to look at bid volume on the upside or short lending on the downside?
Again, just a mathematical model, but another sign that clustering behavior is an important aspect of modeling.