Social Interactions - Is There Really an Identification Problem?

It is an everyday experience that the behavior of individuals belonging to the same social group tends to be correlated. In his seminal work, Manski differentiates two basic types of feedback between group and individual and he maintains that it is not possible to discriminate between the two by mere observation. What is more: Only under very favorable conditions can social effects be distinguished from other reasons for correlations within social groups, such as selectivity. Manski's forceful critique challenges not only the numerous empirical efforts to understand the nature of social interactions. In the light of his arguments many theoretical disputes in the social sciences suddenly appear to be rather futile. Thus, a further analysis of his position seems well justified. The result is quite encouraging. Manski himself renders the solution to his identification problem impossible by imposing a very special assumption. In his econometric model, social effects do not flow from the outcomes realized within the group, but from their respective conditional mathematical expectations. By substituting this critical assumption by a more realistic formulation, a fully identified model is obtained. For this modified model, FIML estimators of all parameters are explicitly derived. The new estimator allows to differentiate clearly between endogenous social effects, exogenous social effects and correlated effects.