My question is does the data bear it out (presumably the data exists) or is it just an assumption based on the idea that insiders _should_ have a pretty good idea of the value of the company.
http://books.google.com/books?id=iYuh-gfnwb0C&pg=PA361&#...
Basically, it seems that extreme insider buying is a very good sign for the market as a whole, but it's fairly rare. (31.7% returns in one three month period w/ insider buying signal, vs. avg. 4.2% returns in 75 3-month periods w/ insider selling signal)
In terms of individual stocks, it seems that the predictive power of insider buying as a metric is best when mixed with taking into account the company's dividend yield, and they claim that this gives you a 6.2% edge.
For individual stocks the difference is too small for me to feel comfortable.
Of course companies act for the well being of the company itself, while directors only for theirs, so yeah, insider buying by a company or a director is totally different.