Working Papers
Land ownership is an important determinant of intra-household bargaining power in low-income countries, yet women are often legally barred from inheriting property. Granting equal access to land tenuring has the potential to improve women’s ability to make decisions within the household, particularly regarding their children. This paper examines the effect of women’s land inheritance rights on fertility and child mortality in India. I develop a household bargaining model in which granting mothers inheritance rights may affect child mortality and fertility through a land channel and a human capital channel. I empirically estimate the effect of each channel using quasi-random variation from a natural experiment in which four Indian states enacted equal rights for women to inherit joint family property between 1986 and 1994. I construct difference-in-differences estimators using variation in eligibility across marriage cohorts and religions. Using retrospective life history and fertility history data, hazard model estimates show that the reforms reduced child mortality through the land channel and reduce fertility through the human capital channel. Children with eligible mothers have a 57% lower hazard of dying before age five. Eligible women are more likely to delay their first birth and have a 32% lower hazard of having more than two children. The results imply a decrease in overall under-five mortality in reform states from about 63 deaths per 1000 live births to 59 deaths. This corresponds to 344,169 children who were saved between the reform passage years and 2005, the survey collection year.
“The Three-Gap Model of Health Worker Performance” with Kenneth Leonard, Luke Bawo, and Rianna Mohammed-Roberts, 2019. World Bank Policy Research Working Paper; no. WPS 8782.
The Three-Gap Model examines the determinants of low-quality health care by examining the patterns and determinants of three gaps. Using four measures of performance—target performance, actual performance, capacity to perform, and knowledge to perform—this paper defines three gaps for each health worker: the gap between target performance and what they have the knowledge to do (the know gap), the gap between their knowledge and their capacity to perform (the know-can gap), and the gap between their capacity and what they actually do (the can-do gap). The paper demonstrates how the patterns of these gaps across health workers in a sample can be used to diagnose failures in the system as well as evaluate the outcomes of policy experiments. Using data on pediatric care from hospitals in Liberia, the paper illustrates how the model can be used to investigate the potential for improvements in the quality of care from several possible policy interventions. The analysis of the relationships between these gaps across health workers in a health system help to paint a better picture of the determinants of performance and can assist policy makers in choosing relevant policies to improve health worker performance.
Work in Progress
“Effects of Click-to-Cancel on Consumer Welfare”
This study evaluates the effect of self-cancellation on cancellation rates and consumer welfare using data from a large subscription company. I exploit time and state variation from two self-cancellation policies implemented by the company: (1) a “multi-step” cancellation method that required customers to click through multiple screens and take a survey before cancellation, and (2) a “one-click” cancellation method that gave customers the ability to cancel by clicking a single button. Before either self-cancellation method was introduced, customers could only cancel via phone, chat, or email. Event study results indicate that multi-step caused a small, temporary increase in cancellation, while one-click resulted in a much larger, highly persistent increase. I apply these treatment effects in a policy counterfactual analysis to estimate how much consumers could have saved if they had access to click-to-cancel. This study indicates that purposefully making cancellation difficult drives a large amount of revenue for subscription companies, and that requiring businesses to make it as easy to cancel a subscription as it is to sign up may have large effects on consumer welfare in specific settings.
“Law Enforcement, Facial Recognition Technology, and Police Violence”
This study evaluates the effect of law enforcement’s use of facial recognition technology (FRT) on police use of lethal force, including lethal and non-lethal outcomes. We combine police department-level data on law enforcement’s use of a major FRT with county-level data on police use of lethal force to create a novel panel dataset linking intensity of FRT use to police violence. We construct a cohort of “control” non-FRT police departments by matching on baseline measures of the outcomes of interest as well as other covariates that may correlate with outcomes. This study aims to provide empirical evidence for ongoing debates about the impact of FRT. Critics argue that relying on FRT may lead to more frequent police encounters, and that more encounters, especially under uncertainty, often correlate with more force. However, if FRT use is largely tightly constrained, audited, and limited to non‑coercive investigative roles as some departments argue, police may be limited in how they can act on FRT results, which may lead to no effect on police violence.
“User Perception of Low-Quality Facial Recognition Technology Search Results”
Facial recognition technology (FRT) is increasingly used by law enforcement agencies to identify potential suspects in post-event investigations. A growing number of wrongful arrests due to inaccurate search results raises the question of how best to regulate FRT companies in order to mitigate user misuse of the technology. FRT companies often argue that providing users with full disclosures about input image and subsequent search result quality reduces users’ reliance on low-quality results. This study is the first to empirically investigate whether this is true using search-level data from a major FRT company used by law enforcement. I use a sharp regression discontinuity design to evaluate the impact of seeing a warning banner triggered by low-quality input images on users’ self-reported satisfaction with search results. Preliminary results show that there is no change in user satisfaction, indicating that users are not more cautious or skeptical of low-quality search results even when given a warning.
Fabliha Ibnat 