RANDOMIZATION AS AN INSTRUMENTAL VARIABLE

This paper discusses how randomized social experiments operate as an i nstrumental variable. For two types of randomization schemes, the fund amental experimental estimation equations are derived from the princip le that experiments equate bias in control and experimental samples. U sing conventional econometric representations, I derive the orthogonal ity conditions for the fundamental estimation equations. Randomization is a multiple instrumental variable in the sense that one randomizati on defines the parameter of interest expressed as a function of multip le endogenous variables in the conventional usage of that term. It ort hogonalizes the treatment variable simultaneously with respect to the other regressors in the model and the disturbance term for the conditi onal population. However, conventional ''structural'' parameters are n ot in general identified by the two types of randomization schemes wid ely used in practice.