Many micro econometric analyses take a single equation approach. However, economists typically consider economic sectors are interrelated. The general (or partial) equilibrium of (a sector of) the economy is achieved through the automatic operation (say, free competition) of the parts that compose the system. We argue that without considering if an equation is not identified in the system, the estimation of such an equation could lead to meaningless curve fitting. Moreover, economic modeling is a simplification of reality, not a mimic of reality. Panel data, being multi-dimensions, provide the possibility to control the omitted variable effects through decomposing the stochastic errors into components, thus introducing restrictions on the error variance and covariance which provide the possibility to relax the identification conditions derived by the Cowles Commission. We also consider inference methods for an endogenous single equation panel interactive model including the limited information maximum likelihood and the methods of moments estimators such as the two stage least squares, the profile GMM and the transformed GMM. A small-scale Monte Carlos is conducted to demonstrate the feasibility of implementing these methods in a finite sample.
Key words: simultaneous equations, identification, panel interactive effects, maximum likelihood, generalized method of moments.