In structural econometrics elements of models are endowed with economic meaning and knowledge is sought of the magnitudes and distributions of these elements and of the relationships amongst them.
There may be many possible economic interpretations of a body of data. Identification analysis seeks to determine what interpretations can be sustained under more or less restrictive assumptions concerning the economic process that generates data, taking into account the way that the process is observed.
Identification analysis builds the bridge between economic theory and econometric practice. It is attention to identification issues that gives econometrics its distinctive flavour relative to many other areas of applied statistics. The focus of these lectures is the principles and practice of identification analysis in structural econometrics.
Identification issues were a central theme when the foundations of modern econometrics were laid in the 1940’s and 1950’s. The topic remains at the core of econometrics but it fell somewhat out of the spotlight until a recent renaissance. Now identification issues are at the fore and new identifying models are a major stimulus to development of new econometric methods and inferential procedures. This is one of the most active areas of econometric research.
Study of the identifying power of econometric models is worthwhile because it gives insight into many issues including:
- the nature of the information about economic processes contained in
- the role of “assumptions” in econometric inference,
- the extent to which the restrictions of models are falsifiable,
- econometric architecture – the design of procedures such as surveys and experiments that deliver information about economic magnitudes and relationships,
- the impact of measurement – in particular what is measured and how it is measured – in determining the information content of economic data and the limits to econometric inference.
These issues will be touched on in the six lectures in this series which start with the parametric Cowles models of the 1940’s and trace the development of the subject through to some semi- and non-parametric extensions that are the subject of current research.