The US Social Security Administration administers retirement, disability, and low-income safety net programs supported by their network of field offices providing in-person customer service. I use employee microdata and a long-differences approach to analyze how staffing decreases during the Reagan era affected program enrollment. I find that a 10% decrease in field office employees in a county led to decreases in enrollment for Old-Age, Survivors and Disability Insurance (OASDI) benefits of 0.06%, driven primarily by disability insurance, and 0.32% for Supplemental Security Insurance (SSI) benefits. These reductions imply 79,027 fewer enrollments nationwide than would have occurred absent the staffing cuts.

Paper: PDF Press: UCI , The American Prospect, Chicago Sun Times

Poppy cultivation for opium production accounted for 14% of Afghanistan's GDP. However, in 2022, the Taliban imposed a nationwide ban on the cultivation and sale of all opium products, representing a major exogenous shock to the country's economy and a sudden shift in factors determining conflict. We exploit this natural experiment to examine how resource depletion affects power consolidation and conflict dynamics in a fragile state. We first formalize and empirically test a traditional two-agent conflict model, with the Taliban and poppy farmers, and show that it cannot explain the conflict patterns observed after the ban. We then introduce a novel three-agent conflict model, with the Taliban, poppy farmers, and resistance fighters, which more accurately represents the nature of conflict in this context. Using a triple difference-in-difference design, we find that this framework better explains the observed post-ban conflict and power dynamics. The model is widely applicable to settings where agents do not switch between production and fighting but instead form distinct groups.

States frequently implement generalized or targeted tax credit programs to spur job creation. These programs display a wide degree of heterogeneity in their provisions, however little is known about the influence of balance of power in the legislature on the provisions included in these laws. Using a database of all U.S. state-level hiring credit programs from 1969 through 2014, I find that a provision for full-refundability is 31% less likely to be included in new programs when legislatures have a veto-proof Republican majority. Recapture provisions, which require the recipient to pay back excess tax credit if they do not create the promised number of jobs, are 50% more likely to be included under Democrat veto-proof majorities and 20% more likely under split majorities. This may represent genuine policy preferences amongst political parties with and without the presence of negotiation, or may reflect a tendency of legislatures to base new policies off of existing policies of similarly-aligned legislatures.

As part of major pension reform, the U.S. government switched from a classic defined benefit plan (Civil Service Retirement System) to a new, at-the-time undefined replacement pension scheme. Eligibility for the plans was based on a hiring date before or after January 1, 1984, announced several months beforehand. Using federal employee microdata, I find that compared to other years that see decreased hiring in December followed by a spike in hiring in January, the hiring of civil servants was more uniform between December and January around the cutoff. This points to manipulation of hiring dates for some new employees in response to the change, through agreement between the employee and the hiring agency in a two-party collusion. Because the replacement system was unknown, manipulation was driven by beliefs about the generosity of the new system or avoidance of uncertainty, rather than reflecting the relative value of the plans.