The importance of being clustered

(I A Few-Cluster-Robust Test for Weak Instruments) - ABSTRACT - In the usual IV regression models, the quality of estimation and inference dra- matically depends on the relevance of the set of instruments. To test such condition, most empiricists rely on the first-stage F-statistic as suggested by Stock and Yogo (2005a). However, this method is not valid for clustered data, especially when there are few clusters. Recently, weak instrument-robust methods for inference have been proposed, but none of them deals with the latter case. In this paper, we propose a simple method to perform inference which is robust both to the presence of weak instruments and few clusters in the case of a single endogenous regressor. (II Sharing or gambling over losses?) - ABSTRACT - This paper investigates experimentally whether individuals prefer to share an exogenous loss in a deterministic way or to gamble over it. In particular, in some scenarios subjects face an equal allocation of the loss, in others a social lottery with an equal chance to suffer it entirely. The loss is implemented after the endowment is earned. We find that the loss domain does not affect subjects' behavior directly, but it decreases their probability to play in a competitive way.