Boston University
Essays on causal inference, structural estimation, and their applications
Abstract
dc:description.abstractThis dissertation comprises three chapters that explore two interconnected areas: the development of innovative econometric tools to reduce computational complexities and the analysis of strategic behaviors for actionable policy insights. The first two chapters introduce new statistical approaches that link advanced econometric methods with empirical research, while the third chapter connects economic theory to practical applications by leveraging big data techniques. When conducting causal inference or designing policy, researchers are often concerned with the existence and extent of interference between units. However, complex correlations across units pose significant challenges for inference. Chapter 1 introduces the pairwise imputation-based randomization test (PIRT), a novel framework for testing interference in experimental settings. PIRT employs a design-based approach, combining unconditional randomization testing with pairwise comparisons to facilitate straightforward implementation and ensure finite-sample validity under minimal assumptions about network structure. To illustrate the method’s broad applicability, I apply it to a large-scale experiment by Blattman et al. (2021) in Bogotá, Colombia, which evaluates the impact of hotspot policing on crime using street segments as units of analysis. The results indicate that increasing police patrolling time in hotspots has a significant displacement effect on violent crime but not on property crime. Chapter 2, coauthored with Jean-Jacques Forneron at BU, studies the Generalized Method of Moments and the Simulated Method of Moments for estimating structural economic models. These methods are often reported to pose optimization challenges, largely because the corresponding objective functions are non-convex. For smooth problems, Chapter 2 shows that convexity is not required: under a global rank condition involving the Jacobian of the sample moments, certain algorithms are globally convergent. These include a gradient-descent and a Gauss-Newton algorithm with appropriate choice of tuning parameters. The results are robust to 1) non-convexity, 2) one-to-one non-linear reparameterizations, and 3) moderate misspecification. In contrast, Newton-Raphson and quasi-Newton methods can fail to converge because of non-convexity. The condition precludes non-global optima. Numerical and empirical examples illustrate the condition, non-convexity, and convergence properties of different optimizers. Chapter 3, coauthored with Daniele Paserman at BU and Angela Crema at the University of Rochester, develops a model of discrimination that helps interpret observed outcome differences across groups, conditional on passing a screening test, as taste-based (employer), statistical, or customer discrimination. The framework is applied to examine non-white underrepresentation in the US motion picture industry. Leveraging a novel dataset that provides racial identifiers for the casts of 7,000 motion pictures, we show that, once a movie is produced, non-white movies tend to have higher average revenues and a smaller variance. These findings are consistent with the model if non-white movies face higher production standards.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy
- Level thesis:degree_level
- doctoral
- Discipline thesis:degree_discipline
- Economics
- Grantor
- Boston University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Zhong, Liang
- Advisor dc:contributor.advisor
-
- Kaido, Hiroaki
Subjects
dc:subject × 1Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/2144/53398