1 Evidence of Variance Covariance Matrix Sensitivity to Correlated

The design of Monte-Carlo experiment to compare alternative estimators of simultaneous equation model requires an absence of correlation between the pairs of normal deviates generated. This is a very difficult requirement except several pairs of normal deviates are generated and then pairs with very low correlation coefficients are drawn. This study designed a 2-equation simultaneous system with a specified covariance matrix and examined the correlation between the pairs of normal deviates generated and the implied covariance matrix. The result showed that the random disturbances generated from normal deviates with feeble correlation coefficient (negative or positive) reproduced the specified variance-covariance matrix while others failed. The result suggested the need to screen the normal deviates generated for the purpose of Monte-Carlo studies especially in the context of simultaneous equation systems