New Developments in Identification Analysis
|Speaker:||Andrew Chesher, UCL|
|Date:||Wednesday 2 May 2012|
Recent developments in identification analysis allow the characterisation of the identifying power of a rich class of econometric models in which observed endogenous and observed and latent exogenous variables can be scalar or multi-dimensional and discrete or continuous along different dimensions.
The models place restrictions on (i) structural relationships that deliver values of endogenous outcomes given values of observed and unobserved exogenous variables and (ii) the co-variation of observed and unobserved exogenous variables. The structural relationships admitted by these models may be incomplete in the sense that they can deliver non-singleton sets of values of outcomes. Such cases arise in models of strategic interaction with multiple equilibria and also in classical instrumental variable models and their recent extensions. I will give an outline of these developments, aiming to put across the main ideas and avoiding some of the technical detail, and I will show how they deliver results in some old and some new problems. One application is covered in detail in the paper "An Instrumental Variable Model of Multiple Discrete Choice", Andrew Chesher, Adam Rosen and Konrad Smolinski, CeMMAP WP39/11