In: Math
Conditions for instruments to be valid include all of the following except: (A) each one of the instrumental variables must be normally distributed. (B) at least one of the instruments must enter the population regression of ?? on the ??s and the ??s. (C) perfect multicollinearity between the predicted endogenous variables and the exogenous variables must be ruled out. (D) each instrument must be uncorrelated with the error term
Instrumental variable regression works on the assumption that -
At least one of the instruments must enter the regression equation (This is assumed since if no instrumental variable enters the equation, the equation wold reduce to standard one)
Perfect multicollinearity between the predicted endogenous variables and the exogenous variables must be ruled out (Otherwise the coefficients of variables might give a false estimate about the average change in outcome per unit change in variable)
Each instrument must be uncorrelated with the error term (Instrumental regression splits the regression equation in two parts, one part which is correlated with the error term and other which is not, when we isolate the part with no correlation, coefficients can be estimated)
But, there is no such assumption which states instrumental variables to be normally distributed, they can or can not be normally distributed