統計学輪講（第７回）

日時      2010年05月11日(火)    15時50分～16時40分
場所      経済学部新棟３階第３教室
講演者    菅原 慎矢 (経済D2)
演題      Refinint the incoherency problem via compatibility: With
application to entry game and binary peer effect models

概要

This paper presents a new perspective on an identification problem of a
binary outcome model with mutual dependencies among
observations. Previous econometric studies have characterized a
statistical problem for this model called incoherency by
Gourieroux et al (1981), which causes non-identification of model
parameters.  We refine the incoherency problem using a concept of
compatibility, which has been developed in spatial statistics as a
necessary condition for identification. Unlike the previous studies, our
analysis shows an explicit distributional assumption which is necessary
for identification. Based on this finding, we provide new insights for
two empirical fields, entry game models with multiple Nash equilibria
and peer effect models with binary outcomes, in which the incoherency
problem has been considered as a source of the identification
problem. First, it is revealed that the identification problem of entry
game models has a different nature from the incoherency, and this
confusion is caused by misunderstanding on the information structure of
the underlying models. Second, we find that binary peer effect models
must be separated to two distinct economeric problems according to the
information structures.