Econophysics is the ultimate in an empirical approach. It starts with
data, and chooses the least number of assumptions they can about
individuals.  Theory comes only after the data have spoken.  In a recent
paper in the Santa Fe III book edited by Brock and Durlauf, Doyne Farmer
(et al) developed a zero intelligence model of agent behavior that was
able to match financial data reasonably well.  

It is an interesting approach that I think will take off in finance, but
because of its enormous data requirements, it will have a harder time in
other macro areas.  


David Colander