Economics of Nutrition

Food Assistance Programs

One way for governments to address nutrition concerns is directly, through food assistance and nutrition programs. This section provides some background about food assistance programs, explains how such programs affect household budgets, and describes research that seeks to measure program effects.

Background On Food Assistance Programs

Food assistance programs may provide food through several mechanisms, including the following: (1) broadly targeted food benefits that low-income consumers may use to purchase food through normal retail channels, (2) more narrowly targeted food vouchers for purchase of specific foods and beverages with particular nutritional qualities, and (3) direct provision of free meals. In addition to food assistance programs, more general income support programs may have benefits for nutrition or food security. This section principally uses US food assistance programs as examples of each type of program, because food assistance plays a bigger role in the social safety net in the United States than in other developed countries.

  1. Broadly targeted food benefits. In the United States, the largest food assistance program is the Supplemental Nutrition Assistance Program (SNAP), formerly known as food stamps. SNAP provides targeted benefits for food and nonalcoholic beverages from authorized grocery retailers through Electronic Benefit Transfer (EBT) cards similar to debit cards. It served 40.3 million people per month on average during fiscal year 2010 at a total cost of $68.2 billion. Program eligibility historically has depended largely on having income less than 130% of the federal poverty standard, so the program is counter-cyclical, and caseloads have recently risen to record levels during the recent recession. The primary purpose of the program is to prevent hunger and promote food security.
  2. Narrowly targeted food vouchers. The Special Supplemental Food Program for Women, Infants, and Children (WIC) provided nutrition counseling, services, and a package of particular high-nutrient foods and infant formula to approximately 9.2 million people per month, at a cost of $6.8 billion in fiscal year 2010. Only pregnant and postpartum women, infants, and children through the age of 4 years are eligible. Eligibility also requires household income less than 185% of the federal poverty standard or participation in one of several other safety net programs, plus evidence of nutrition risk broadly defined.
  3. Direct provision of meals. The National School Lunch Program served 31.6 million lunches, and the smaller and newer School Breakfast served 11.6 million breakfasts, on average, each school day in fiscal year 2010, at a cost of $13.3 billion. A free meal requires income less than 130% of the federal poverty standard, though all school provided meals in schools participating in the federal school meals programs are subsidized to some extent. The Child and Adult Care Food Program served meals in centers and home day care settings, costing $2.6 billion. These programs have primary nutrition goals, but antihunger effects are acknowledged as important secondary purposes.
  4. Cash assistance, cash-based social insurance, and tax credits. Finally, governments provide cash welfare, which can be used for food as well as other products. In 2009, 1.8 million families with children obtained Temporary Assistance for Needy Families cash benefits, with total cash benefit payments amounting to $9.3 Billion. Supplemental Security Income (SSI) provided $41 billion in cash assistance for 6.4 million low-income individuals who were disabled, blind, or elderly. Unemployment insurance is typically viewed as a social insurance program but is also another important part of the safety net, with payments of approximately $131 billion in 2009. Tax credits such as the Earned Income Tax Credit may also play a role.

How Food Assistance Affects Household Budgets

The effect of food assistance on food choices depends on the role of program benefits in the household budget. To understand this role in a rigorous way, it helps to compare the effects of: (1) providing a targeted food assistance benefit or (2) providing a hypothetical equivalent cash subsidy. For example, consider a monthly voucher that provides a family with $50 for use in purchasing qualifying food products. The hypothetical comparison program provides $50 in cash.

  • If a family with the hypothetical cash benefit would have spent less than $50 per month on qualifying foods, the family is called extramarginal or constrained by the form of the benefits, because the voucher program causes more food spending than would otherwise have occurred.
  • In contrast, if a family with the hypothetical cash benefit would have spent more than $50, the family is called inframarginal or unconstrained by the form of the benefits, because it is able to purchase its desired amount of qualifying foods under either the targeted voucher or the hypothetical cash program.

Economic theory predicts that a marginal increase in targeted food assistance benefit will strongly affect food spending for extramarginal participants, and the increase will only weakly affect food spending for inframarginal participants. Empirical research commonly finds bigger program effects than expected for inframarginal participants in targeted food assistance benefits programs (Meyerhoefer and Yang, 2011).

Measuring The Effects Of Food Assistance Programs

A challenge is obtaining causal estimates of the effects of food assistance programs. Due to selection bias in who takes up these programs, one cannot merely compare the outcomes of recipients and others. Given that participation is tied to low income and asset holdings or poor nutritional status or both, it is clear that comparisons of outcomes for program recipients to those of the general population are likely to be biased estimates of the effects of these programs. Even among those eligible for the programs, recipients may be positively or negatively selected compared with eligible nonparticipants due to the fact that participation is a choice variable. If recipients are healthier, more motivated, or more knowledgeable about the programs than nonparticipants, comparisons may suggest the program has a more positive effect than it actually does. Alternatively, if participants are more disadvantaged than eligible nonparticipants, comparisons of these two groups could lead to underestimates of the effects of the program. For example, skeptics often attribute many of the positive effects of the WIC program to positive selection among women participating in the program, although Bitler and Currie (2005) find little evidence than WIC participants who also participate in Medicaid are positively selected.

There are several approaches that researchers have taken to avoid selection bias (Gundersen et al., 2011). Just four approaches are mentioned here. One approach is to compare outcomes among individuals in geographic areas with different program rules. This approach is comparatively less useful for SNAP policies that are national in scope and comparatively more useful for SNAP policies that have substantial state-to-state variation (Ratcliffe et al., 2011).

A second approach, when program rules do not vary across areas is to look at the effects of the introduction of programs; comparing otherwise similar individuals in places before and after programs are introduced. The introduction of the food stamp program led to increases in food consumption (Hoynes and Schanzenbach, 2009), whereas the introduction of the WIC program led to an increase in average birth weight (Hoynes et al., 2011). A limitation with this approach is that program rules may change over time, potentially raising questions about the ongoing validity of historical estimates for evaluating programs today.

A third approach to studying the effects of programs is to use random assignment. If program administrators are able to randomly assign an offer of program participation to otherwise identical eligible individuals, then comparisons of those assigned to be eligible for the program with those denied the option to participate can yield unbiased estimates of the effects of the program. In the case of food stamps, there have been several demonstration projects funded by USDA, which have yielded evidence about the effects of cashing out food stamp benefits on food spending.

A fourth approach to studying the effects of such programs is to use variation in program rules that use exogenous thresholds in income, age, or other characteristics to assign program eligibility. These regression discontinuity approaches compare otherwise similar individuals who because of small differences in a characteristic such as age face different program rules, while controlling for age. These ‘regression discontinuity’ approaches typically bring both considerable internal validity and limited external validity given the local nature of the estimates they yield.

Much progress has been made on the effects of food assistance programs, yet there are outstanding questions. For example, how does a program like WIC obtain positive results with relatively low benefit levels? What channels do effects of these programs work through? What is the role of information in the effects of these programs?

Determinants of Mental Health
Peer Effects in Health Behaviors