This paper presents a novel self-report approach to identify a general causal model with an unobserved covariate, which can be unobserved heterogeneity or an unobserved choice variable. It shows that a carefully designed noninvasive survey procedure can provide enough information to identify the...