sdurlauf
commented
5 years ago Great question. Answer will depend on data set. Note that the assumption is that the x's that agents observe are also observed by analyst. There are other individual characteristics (the linear model errors) that where agent observes her own, but no one else's while analyst does not observe any of them. There is an intermediate case where analyst sees subject of x's. There are tools in the Handbook paper to extend the results in that paper and others. But your point is well taken.
Thanks for presenting! In section III (From a Theoretical to an Econometric Model) of your paper “Linear Social Interactions Models”, you propose an identification assumption K.1 that requires the analyst to “observe both the outcomes and the characteristics of each member of the population”. I was thinking that this might be a quite strict assumption for empirical work. How would you justify whether the analysts in the real world meet this assumption? Thank you.