A Monte Carlo Study of the Estimation of an Overidentified Model with Temporally Dependent Residuals

When building an econometric model, the equation residuals may be both simultaneously and temporally correlated. In the following Monte Carlo Study I try to visualize the effect of that problem on the estimators generated by six different methods of estimation. I postulate a two-equation model without a lagged endogenous variable. In this system I change various parameters: the sample period size, the level of simultaneous correlation, the level of autocorrelation, and the residual variances. This study is the second part of my MA thesis, which is written in French, presented at the University of Montreal.