This dissertation addresses three distinct topics in development economics. The first chapter assesses the role of entry and exit in the measurement of misallocation in India. In the last decade, misallocation of productive inputs across firms has been proposed as a primary driver of differences in aggregate productivity over time and across countries. By focusing on the distribution of resources across existing market participants, the literature has ignored interaction between forces that distort the distribution of inputs and distort entry and exit decisions, simultaneously, delivering possibly misleading conclusions about the dynamics of misallocation. In this chapter, I develop a model of misallocation with endogenous entry to illustrate this phenomenon. I then use an Indian manufacturing firm panel to show that entry and exit decisions are correlated with drivers of misallocation, and that the spike in misallocation during the global recession would have been more muted (about 23% less) had there been no endogenous market participation. The results of this exercise point to the importance of considering firm participation decisions when comparing measured misallocation across time and space. The second chapter develops a tool to detect treatment effect variation in randomized trials. I develop a novel method to detect variation in treatment effects in a randomized controlled trial. The variation need not be dependent on observed characteristics. I conduct simulations and power calculations to show that the method can be applied to detect variation with relatively modest data requirements. I apply the method to two flagship studies: Banerjee et al. 2015 and Crepon et al. 2015. I find evidence of treatment effect variation indicating that both programs studied help some subjects more than others. The third chapter, based on work co-authored with Samuel Bazzi, Robert Blair, Christopher Blattman, Oeindrila Dube, and Matthew Gudgeon, develops predictions of violent events in Indonesia and Colombia. Policymakers can take actions to prevent local conflict before it begins, if such violence can be accurately predicted. We examine the two countries with the richest available sub-national data: Colombia and Indonesia. We assemble two decades of fine-grained violence data by type, alongside hundreds of annual risk factors. We predict violence one year ahead with a range of machine learning techniques. Models reliably identify persistent, high-violence hot spots. Violence is not simply autoregressive, as detailed histories of disaggregated violence perform best. Rich socio-economic data also substitute well for these histories. Even with such unusually rich data, however, the models poorly predict new outbreaks or escalations of violence. "Best case'' scenarios with panel data fall short of workable early-warning systems.

Creator

• Peck, Richard

DOI

• https://doi.org/10.21985/n2-h2sc-7m02

Subject

• Economics

Language

• en

Alternate Identifier

• etdadmin_upload_744020
• http://dissertations.umi.com/northwestern:15134

Date created

• 2020-01-01

Resource type

• Dissertation

Rights statement

• In Copyright