Statistical inference teaches us how to learn from data, whereas identification analysis explains what we can learn from it. This paper proposes a simple unifying theory of identification, encouraging practitioners to spend more time thinking about what they can estimate from the data and assumptions before trying to estimate it.
↧