Regression Model: Endogeneity and Collinearity

In a linear regression model, regressors are assumed to be exogenous implying that the conditional mean of random error is zero. This assumption is essential for having unbiased estimator. When this assumption is violated, the problem of endogeneity will appear. Another important assumption of a multiple linear regression model is the full rank assumption implying that regressors are not correlated. When regressors are highly correlated, the problem of multicollinearity appears. Multicollinearity is one of the several problems in regression analysis. The term multicollinearity was first introduced by Ragnar Frisch (1934).