Andrew Chesher (University College London)

“Structural Econometrics with Discrete Outcomes”

abstract

Discrete outcomes are ubiquitous in econometrics. Much information about human behaviour in response to economic stimuli and motivations is provided by data recording variously: choices amongst discrete, ordered or unordered alternatives, attitudes, preferences and strengths of response recorded on ordinal scales, counts of occurrences of events, and so forth.

I consider classical structural econometric analysis which uses models in which discrete outcomes are delivered by a structural function with: observed arguments, which may be endogenous; and unobserved arguments which are restricted to be distributed independently of some observed exogenous variables.

I review three main approaches to such structural analysis of discrete outcomes from the 1970’s to today. I explain why, without very strong restrictions, structural models of discrete outcomes are partially, not point, identifying and I provide a general result characterising the information content of discrete responses when viewed through the lens of a structural econometric model.