Perhaps THE biggest question in Finance is what makes expected returns vary from one security to the other. The idea that getting higher expected returns cannot be generated without bearing more risk has driven the revolution and spurred a million papers, either trying to show that it works or it doesn't.
Avanidhar Subrahmanyam has just released a paper on SSRN in which he reviews many variables and methods that people have used to predictor returns (like P/E, size, liquidity, etc.)
He writes "our learning about the cross-section is hampered when so many predictive variables accumulate without any understanding of the correlation structure between the variables, and our collective inability or unwillingness to adequately control for a comprehensive set of variables."
With so many people, testing so many variables, with so many different methods, it is often difficult to really know whether a given variable can truly predictive future returns or it is just reflecting correlation with something else.
It would be nice to see a paper trying to run a "horse-race"of lots of variables at the same time (I mean a lot, not just 4-5).
PS - The article cited that shows that garbage production in the US is a better proxy for consumption than standard measures is really cool.
No comments:
Post a Comment