The problem of identification is defined in terms of the possibility of characterizing parameters of interest from observable data. This problem occurs in many fields, such as automatic control, biomedical engineering, psychology, systems science, the design of experiments, and econometrics. This article focuses on identification in econometric models, which typically involve random variables. Identification in general parametric statistical models is defined, and its meaning in a number of specific econometric models is considered: regression (collinearity), simultaneous equations, dynamic models, and nonlinear models. Identification in nonparametric models, weak identification, and the statistical implications of identification failure are also discussed.