Identification and Causal Order

This paper looks at one important part of econometric methodology: identification. Put generally, relations that hold among observable and unobservable entities are identifiable if given some a priori knowledge about the relations, previously unknown characteristics of those relations can be deduced from observations. Typically in the econometrics literature, the identification problem is presented as the problem of inferring unknown coefficient values in systems of equations from observations. Yet a natural question comes to mind when these equations are structural (taken to denote a causal structure): what does requiring identifiability for a set of structural equations require of the causal structure represented? What features of causal structures ensure that they can be denoted by identifiable equations?