UC Berkeley

This dissertation studies various ways in which federally funded social safety net programs impact low-income households. Over 38 million people in the US live in poverty, and the federal government dedicates nearly one-sixth of its annual budget towards programs designed to reduce the negative consequences of poverty. In this dissertation, I use various empirical techniques and policy settings to explore the causal impact of increasing access to federal assistance benefits on outcomes for those in need.

In Chapter 1, I study how reducing the costs of participation in the Supplemental Nutrition Assistance Program (SNAP, previously called Food Stamps) influences program participation and downstream measures of infant health. Incomplete take-up is common in many transfer programs: more than 7 million individuals who were eligible for SNAP benefits in 2017 did not participate in the program. An understanding of the causes and downstream consequences of incomplete take-up is necessary in order to understand whether low take-up of public programs is limiting their effectiveness in reducing economic disparities. In this chapter, I use a unique setting to shed light on the role that participation costs play in determining food assistance take-up, and quantify the effects of increasing take-up on infant health at birth. I do this by studying a reform that made it easier to receive and use food assistance benefits in the US: the adoption of the Electronic Benefit Transfer (EBT) debit card for SNAP benefit disbursement.

Prior to the adoption of EBT, SNAP benefits were distributed via food stamp coupons. Thus, the switch to EBT reduced both the time that it took to receive benefits each month and the visibility (or stigma) associated with using benefits at the grocery store. I use the staggered county level rollout of the EBT card in California between 2002 and 2004 to estimate event study regressions of its effect. I find that EBT adoption led to a large and persistent increase in caseloads and applications for the program, as well as higher retailer participation in high poverty neighborhoods. I document that this rise in food stamp benefit take-up led to a meaningful increase in average birth weight for births most likely impacted by the policy, with effects concentrated in the bottom half of the birth weight distribution. These estimates provide new evidence that reducing the barriers to participation in food assistance programs can lead to potentially large gains in health for disadvantaged children.

The adverse consequences of childhood poverty are extensive and long-lasting, suggesting an important role for government policies that have direct impacts on early-life outcomes. There is a large literature on the incentive effects of cash tax and transfer programs in the US, yet less is known about how these programs impact the children of recipients. In Chapter 2, I estimate the causal effect of unconditional means-tested cash assistance on outcomes for low-income children by studying the first federally funded cash welfare program in the US: Aid to Dependent Children (ADC), a means-tested program implemented as part of the Social Security Act of 1935.

Using newly digitized data on the full population of the US in 1940, I leverage discontinuities in ADC benefit generosity across state borders to estimate the effects of cash assistance within contiguous county pairs. I find that a one standard deviation increase in the maximum ADC benefit was associated with a 2.5 percentage point increase in the probability of school enrollment for disadvantaged children, reductions in overcrowded living arrangements, and evidence of substitutions away from work towards schooling for older children. Consistent with the existing literature on the labor supply disincentives of unconditional transfers, I also find that higher benefit generosity is associated with reductions in maternal labor supply.

In Chapter 3, I study the effect of geographic access to authorized retailers in the Supplemental Nutrition Assistance Program (SNAP). When determining the optimal design of means-tested transfer programs, policymakers must consider the features of policy design that may unintentionally serve as barriers to limit the effectiveness of the program. Some barriers, such as those discussed in Chapter 1, may prevent eligible people from participating in the program, while others, like those I explore in Chapter 3, may impact how households use their benefits. Given the substantial geographic variation in access to grocery stores authorized to accept SNAP benefits, it is important to understand whether policies to maximize the number of retailers authorized to accept SNAP benefits are needed.

Using a difference-in-differences design, I estimate the effect of a food retailer gaining SNAP authorization in a zip code on household grocery spending behavior. I find that when a small grocer or convenience store gains authorization to accept SNAP benefits, eligible households in the area shift some of their food expenditures away from supermarkets and supercenters and towards these smaller SNAP retailers. I find mixed evidence of the impact of SNAP retailer access on the healthfulness of food purchases. Lastly, I examine whether increased access to a SNAP authorized retailer has any effect on intra-month cyclicality in food spending. I find that access to a new SNAP supermarket or super center does reduce the severity of this "SNAP cycle" in food expenditures, with less evidence of any impact for smaller stores. These results suggests that there could be gains to incentivizing increased retailer participation in SNAP.