Theres a neat definiton of the Lucas critique: “One of the main purposes of simultaneous equations estimation is to forecast the effect of changes in the exogenous variables on the endogenous variables. However, if the exogenous variables are changed and agents see the changes coming (and especially if they believe that the changes are permanent), they would modify their behaviour accordingly. Thus the coefficients in the simultaneous equations models cannot always be assumed to be independent of (changes in) the exogenous variables.”
(Theres a less academic version of that in this posting about an interview with a Wall Street quant.)
This is the basic issue with all modelling: a model is a historical description, and cant contain the completely unexpected.
The Melbourne page is maintained by Dr Robert Dixon.