| We propose a new methodology to perform parameter inference for partially identified models specified in terms of a finite number of equality as well as inequality unconditional moment restrictions. In particular, we propose a methodology to build confidence sets for the parameters of interest in this context. The proposed confidence sets are shown to be asymptotically valid and conditions under which they are uniformly consistent in levels are established for a certain class of probability measures. We assess their finite sample performance using Monte Carlo experiments. Finally, an empirical application illustrates the use of the methodology to estimate the effects of competition on the decision of adopting an MRI by U.S. hospitals. |
