The Supplemental Nutritional Assistance Program (SNAP) in the US provides low-income families with assistance to purchase groceries at selected grocery stores. With a budget of $65B in 2018, it has historically been a large component of the social safety net in the US. However, Finkelstein and Notowidigdo (2019) note that program take up is relatively low among the elderly: only 42% of eligible elderly individuals receive SNAP, compared with 83% of eligible individuals overall. The authors note that this low take-up may be due to a lack of information about the program’s benefits and eligibility, or to application transaction costs such as gathering documents, filling out forms, and visiting application centers.
Finkelstein and Notowidigdo (2019) conduct an RCT with 30,000 likely eligible individuals in Pennsylvania to evaluate the impact of two types of treatments to increase SNAP take up by the elderly. They divide the sample into a control group and two treatment groups. The first treatment group (information only) receives a letter and follow up postcard reminder from the secretary of Pennsylvania’s Department of Human Services (DHS) explaining likely eligibility and encouraging SNAP application. The second treatment group receives the same information treatment plus a phone number they can call to get phone-based assistance to apply. Authors find that the information only treatment increases take up, and that information plus assistance increases it even more, however with a higher cost per beneficiary.
The authors provide a framework to translate their estimates into their implied MVPFs for both experiments. This page outlines the MVPF for the information treatment.
MVPF = 0.9
To compute costs, the authors divide the population into high-income (H) and low-income (L) elderly. If accepted in the program, high-income individuals receive lower benefits ($16/month), while low-income individuals receive $178/month. Given that the program requires re-application after 36 months, authors assume conservatively that application acceptance gives 36 months of benefits, and that both types of individuals have the same 75% probability of being accepted if they apply. This leads to expected total transfers of $4,806 for L and $432 for H if they apply. Fiscal externalities come not from lower work supply, given that the population is elderly and mostly out of the labor force, but from application processing costs of $267 incurred by the government. The experimental results in the paper show that the information only intervention increases the probability of application by 4% for individuals in the L group and by 3% for individuals in H. The total expected cost to the government of providing the information intervention is then ($4,806 + $267)*0.04 + ($432 + $267)*0.03 = $234
The authors also divide the population into H and L types to compute the WTP of each group separately. For the information intervention to have any value, each type must be initially misinformed about the benefits of applying. To calibrate the level of misperception, authors assume a time cost of application and use a utility function approximation to compute the expected utility of applying. They then use that to obtain an estimate of the level of misperception that rationalizes the application rate observed in the population. A large fraction of the eligible population does not apply, and their model parametrization implies large misperceptions. H types perceive the probability of acceptance into the program as 17% of the true probability, and L types as 2% of their true probability. Once accepted in the program, enrollees value transfers on a dollar-for-dollar basis. Each type’s WTP for the information intervention is the degree of misperception multiplied by the expected value of applying, which gives WTP = (0.98*$4,806*0.04) + (0.83*$432*0.03) = $208.
Combining the cost and WTP estimates, the MVPF for the information intervention is 0.89.
Finkelstein, A., & Notowidigdo, M. J. (2019). Take-up and Targeting: Experimental Evidence from SNAP. The Quarterly Journal of Economics, 134(3), 1505-1556.
https://academic.oup.com/qje/article-abstract/134/3/1505/5484907