Research

This study investigates the identification of marginal treatment responses within multi-valued treatment models. Extending the hyper-rectangle model introduced by Lee and Salanié (2018), this paper relaxes restrictive assumptions, including the requirement of known treatment selection thresholds and the dependence of treatments on all unobserved heterogeneity. By incorporating an additional ranked treatment assumption, this study demonstrates that the marginal treatment responses can be identified under a broader set of conditions, either point or set identification. The framework further enables the derivation of various treatment effects from the marginal treatment responses. Additionally, this paper introduces a hypothesis testing method to evaluate the effectiveness of policies on treatment effects, enhancing its applicability to empirical policy analysis.

In this paper, I utilize the Bayesian inference framework developed by Mele (2017) to investigate the determinants of pairwise stable network formation. Specifically, I examine how social relationship networks are constructed and influenced in rural Indian villages. One of the key findings is that in areas with limited financial accessibility, individuals with access to micro-finance often connect with those who do not have access. However, as financial accessibility increases, this trend weakens. In addition, I conduct a counterfactual experiment to demonstrate that introducing a financial facilitator may not necessarily increase the indirect financial coverage rate due to the complex dynamics of the network formation. This highlights the importance of understanding the entire network structure when making policy decisions. I also extend the inference framework to incorporate aggregate relational data, which can be applied to settings where the researcher cannot observe the entire network but can observe only aggregated features.

This paper develops a two-stage method for inference on partially identified parameters in moment inequality models with separable nuisance parameters. In the first stage, the nuisance parameters are estimated separately, and in the second stage, the identified set for the parameters of interest is constructed using a refined chi-squared test with variance correction that accounts for the first-stage estimation error. We establish the asymptotic validity of the proposed method under mild conditions and characterize its finite-sample properties. The method is broadly applicable to models where direct elimination of nuisance parameters is difficult or introduces conservativeness. Its practical performance is illustrated through an application: structural estimation of entry and exit costs in the U.S. vehicle market based on Wollmann (2018).