Research
Job Market Paper
Gender Equity in the US Civil Service: Evidence from the Classification Act of 1923
(with Socorro Pardo Martinez)
Abstract: Pay standardization schemes are commonly implemented in organizations to address pay gaps based on gender, race, or ethnicity. These schemes limit managerial discretion to determine wages, thereby limiting gender inequality within job titles. However, this may shift inequality to other margins such as position quality or promotions. This study examines the effects of the Classification Act of 1923, which standardized pay grades and position categories and required ``equal pay for equal work'' in the US Civil Service, on women's earnings relative to men's. Using a difference-in-differences approach, we exploit the differential application of the policy to civil servants in Washington, D.C., compared to those in federal field offices outside of DC. For the Classification Act to improve relative pay, we expect compensation to rise faster for women in DC versus women in field offices, compared to men in DC versus field offices. We find that the law led to a small, although insignificant, improvement in women's relative compensation within job titles. However, without controlling for job title, the law lowered relative compensation, suggesting changes in women's positions. We indeed find large negative effects on the position margin. In response to the law, departments downgraded women to lower-quality positions. These negative consequences predominantly affected newly hired women. Our findings underscore the importance of unintended margins of adjustment for policies related to gender equity.
Publications
(with Travis Lybbert, Ashish Shenoy, Rupika Singh, and Daniel Stein). Journal of Development Economics 166 (2024): 103199.
Abstract: Ubiquitous mobile phone ownership makes phone surveying an attractive method of low-cost data collection. We explore differences between in-person and phone survey measures of agricultural production collected for an impact evaluation in India. Phone responses have greater mean and variance, a difference that persists even within a subset of respondents that answered the same question over both modes. Treatment effect estimation remains stable across survey mode, but estimates are less precise when using phone data. These patterns are informative for cost and sample size considerations in study design and for aggregating evidence across study sites or time periods.
Working Papers
What are the Barriers to Pulse Cultivation in India? Evidence from a Randomized Controlled Trial
(with Rupika Singh, Daniel Stein, and Kate Sturla)
Abstract: India is the largest producer and consumer of pulses, but in recent years has needed to import pulses to meet domestic demand. In order to keep prices stable and control the balance of trade, increasing domestic pulse production has become a national policy priority. However, farmers face multiple constraints to pulse cultivation. We design an experiment to measure the relative importance of these constraints to design an optimal short-term policy for pulse promotion. We conduct an RCT testing three different implementation models designed to ease different constraints: the “high intensity” provides seed distribution, extension services, and demonstration plots, the “medium intensity” provides seed distribution and extension services, and the “low intensity” only provides seed distribution via a voucher system. Overall, we find that all three models are effective in promoting pulse cultivation, with no statistically significant differences between the models. Treatment increased farmers’ propensity to grow pulses over two seasons by 12 and 15 percentage points, respectively. These results suggest that, at least for the initial phase of pulse adoption, access to quality seeds is the key constraint, as opposed to information. Therefore, seed distribution is likely a cost effective way to quickly increase the adoption and production of pulse crops.