This chapter explores the public policy responses to childhood hunger. This topic includes the impact of the food safety net and the extent to which childhood hunger and food insecurity persist due to gaps in program coverage, the inability of potential participants to access programs, and the insufficiency of program benefits or services. The moderator was Judith Bartfeld, Department of Consumer Science, University of Wisconsin–Madison. The main speaker was David Ribar, Economics Department, University of North Carolina at Greensboro, and the Institute for the Study of Labor (IZA), Bonn, Germany. Ribar was followed by two formal discussants: Lara Shore-Sheppard, Department of Economics, Williams College, and Jim Weill, president of the Food Research and Action Center.
STATEMENT OF DAVID RIBAR1
Ribar set the context with a quote from Nord and Parker (2010:1179): “With one important exception, the major determinants of food insecurity are fairly well understood. The exception is the effects of food and nutrition assistance programs.” Ribar then defined a conceptual model, followed by a descriptive typology of existing public and private food assistance programs. He discussed the evidence available on the effective-
1Ribar (2013) wrote a commissioned background paper on the topic for the workshop.
ness of these programs, and pointed out program and methodological gaps. He ended his presentation with his own suggestions.
Ribar observed that the use of a conceptual model helps in the understanding of how children get fed and why some go hungry, provides insight into how various programs work, and helps identify potential challenges for program effectiveness. Ribar pointed to the Gundersen discussion in Chapter 2 on the development of conceptual models, to the Institute of Medicine and National Research Council (IOM/NRC) report (2013), and specifically to Barrett’s (2002) model of how household food security is determined. The Barrett model adapted Becker’s (1965) household production model, and is very similar to Grossman’s health production model (1972). Ribar used the Barrett model in his discussion.
The model assumes a household that faces a life-cycle utility function with two objectives in each period. One is to advance its physical wellbeing, and the other is to consume things based on a preference function that incorporates tastes and culture. Finally, the model assumes that the household will discount the future and that the future will be uncertain.
Ribar explained that in the model, physical well-being in a given period is based on the previous status of physical well-being augmented with inputs of nutrition, activities (such as rest and exercise), nonfood consumption, financial constraints, and other things. It is subject to shocks from illness or injury. Nutritional inputs are based on inputs of food and the household member’s time. Their effectiveness is conditioned on the member’s health as well as his/her skill and knowledge.
Ribar then described potential outcomes. The household in this model chooses work, activities, and the consumption of food and nonfood items so as to maximize its objectives subject to its constraints. Through its decisions, the household might achieve one of three levels of food security: sufficient for survival, sufficient for nonimpairment, or sufficient for health. U.S. policies typically focus on achieving good health, while survival and nonimpairment are important focuses in developing countries.
He said that within this standard framework for household decision making, Barrett (2002) pointed out six types of static structural threats to food security: (1) low labor productivity (a limited ability to work or to earn, which results in fewer resources available to the household); (2) adverse terms of trade for a given level of work or abilities (the household member is not able to command a very high wage and/or may face high food prices); (3) lack of access to markets where household members could engage in paid labor or purchase goods; (4) asset poverty (low lev-
els of savings and other assets); (5) borrowing constraints; and (6) weak availability of public and private safety nets.
In a dynamic framework, other factors put households at risk, such as operating close to one of the constraint levels. In this situation, a bad shock may push the household into risk. Some households have a susceptibility to adverse shocks (either social or economic) that put them at higher risk. Finally, households with inadequate insurance will also be at greater risk of food insecurity.
The model identifies numerous coping strategies that households use to avoid hunger. The strategies include the use of transfers and loans, foraging, disposal of nonproductive assets, reduced consumption and energy expenditure, disposal of productive assets, expropriation of other’s assets, and migration. These coping strategies complicate the measurement of hunger, because even if a household faces a bad shock, it usually does not experience an immediate hunger outcome.
Ribar said that Barrett’s general model does not give special consideration to children and their circumstances. Children have limited capacities to work, are dependent on other family members, and have little or no ability to influence decision making. A child’s capacity and dependency will vary with age. Within this framework, children are very vulnerable, he said. Instead, the standard model (Becker, 1983) assumes the existence of caring and capable parents. The standard economic assumption is that parents are both rational and altruistic, leading to Ricardian results: Parents will be protective of children and will mitigate the relationship between shocks and/or programs and the outcomes for their children. In particular, if the government does not step in to help children, parents generally will fill the gap. Conversely, if the government adds support, parents may withdraw some of their own support in response. This leads to the household coping strategy in which children typically are the last to go hungry. Ribar said there is strong evidence that this is the typical behavior in households (Edin et al., 2013; McIntyre et al., 2003), but it is not necessarily the only behavior in households.
There are examples of other types of parental actions, as discussed in Chapter 5. Some parents could have limited food preparation capabilities or habits (McLaughlin et al., 2003) or parenting problems brought forward from their own childhood circumstances (Chilton and Rabinowich, 2012). Gundersen also described financial management problems in households (see Chapter 3; also Gundersen and Garasky, 2012). Finally, some children are difficult to parent, causing circular problems in which food problems cause bad outcomes for children, which in turn make those children harder to parent (Kleinman et al., 1998; Pérez-Escamilla and Pinheiro de Toledo Vianna, 2012).
The evidence regarding these threats generally supports the general model. Additional literature includes Nord and Parker (2010), which discussed low income and unemployment, low skills or disability, single parenthood, large household, minority or noncitizen status, and poor local economic conditions and institutions. Kimbro et al. (2012) also described the effect of disadvantaged neighborhoods.
Public and Private Food Assistance Programs
There are three general strategies for providing food assistance in the United States. One is to supplement household resources, which effectively lifts the budget constraint and gives households more opportunity to produce better outcomes for their children. A second strategy is to provide a household or its individual members with specific types of foods, thereby directly supplying the nutritional inputs. The advantages of this type of program are that it is easier to target, the program is harder for people to undermine, and benefits can go directly to children. The third strategy is to help households be more productive with the resources they have. This is the objective of educational programs that help households do more with a given level of resources to lead to better outcomes. All three strategies are being used in the major U.S. food assistance programs.
Ribar described the major U.S. food assistance programs. The largest is the Supplemental Nutrition Assistance Program (SNAP), formerly known as the Food Stamp Program. Next are the National School Lunch Program (NSLP) and the School Breakfast Program (SBP), followed by the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) and the Child and Adult Care Food Program, which provides assistance through child and adult care facilities. Ribar noted eight additional U.S. Department of Agriculture (USDA)-funded programs that also deliver food assistance for children.2 Several of them fill gaps in the major programs. For example, the Food Distribution Program on Indian Reservations and the Nutrition Assistance Block Grants are alternatives to SNAP that operate in Indian reservations and U.S. territories. The WIC Farmers’ Market Nutrition Program addresses a gap in the WIC Program.
The federal government provides the resources for all these programs, but state and local governments and school food authorities, and in some cases community organizations, run them. This is an opportunity for success, but also a source of weakness with these programs, according to Ribar. The states contribute substantial amounts of administrative resources and, in some cases, fund modest supplementary programs. For instance, North Carolina funds a universal free school breakfast program
2Ribar (2013) provides more detail about these programs.
for all kindergarten students. Washington, DC, has just moved to universal free breakfast in its public schools, and Washington State, seeing a hole in assistance for immigrant families, funds a special supplemental program for immigrants. Numerous states have commodity support programs, such as the New Jersey State Food Purchase Program.
Chapter 6 discussed private food assistance programs that provide general assistance, such as food banks and food kitchens. They also help leverage available resources. They may provide special child-oriented programs, such as backpack programs, school pantry programs, and “Kids Cafés” (Tapper-Gardzina and Cotuga, 2003). Though privately run, these programs usually heavily depend on the federal government for resources (Mabli et al., 2010).
All of this leads to a very complex and uneven food assistance landscape, according to Ribar. Depending on where children live and where they attend school, there may be ample potential resources and lots of flexibility, but substantial potential for overlap and inefficiency. The landscape itself is uneven, he noted, often depending on whether the state and local governments apply for grants to run various programs. In some sense, he said, children are at the mercy of state and local governments.
Discussions often highlight the best examples of what state and local governments are doing with the flexibility that they are given and overlook some of the negative impacts of that flexibility, Ribar stated. For example, as mentioned above, North Carolina funds a universal free breakfast program in its public schools; however, the state’s charter schools are not required to offer any school meal programs. Thus, he said, these charter schools are effectively discouraging attendance by poor children and are becoming racially segregated. Programs that offer voucher assistance to private schools but do not require those schools to provide school meal assistance are likely to have similar negative results. Food assistance (or lack of it) can become a tool for discrimination, according to Ribar.
Evidence of Effectiveness
Ribar next considered whether existing programs prevent food insecurity and hunger. He emphasized the term “existing,” and stated that he does not believe that they do. He said evidence presented at this workshop demonstrates that even with the $100 billion food safety net, there are many examples of children living in households with very low food security. He referred to Coleman-Jensen et al. (2012), who reported that in 2011 11.5 percent of children lived in households with low food security among children, 1.5 percent lived in households with very low food security among children, 23 percent of SNAP households had very low
food security, 17 percent of households receiving NSLP meals had very low food security, and 14 percent of WIC households had very low food security. Even households receiving benefits have high levels of reports of food problems. He said that this is prima facie evidence that the existing network has holes.
A separate question, he said, is whether the existing programs reduce food insecurity and hunger, and here, he posited, the answer is probably. It is hard to imagine how food given to children and households does not help in some way, he commented, but the evidence is surprisingly weak. He cited a simple descriptive comparison of households that are receiving benefits versus households that are not receiving benefits and pointed out the perverse results that the households on assistance tend to report higher levels of food hardships than the nonparticipating households (Coleman-Jensen et al., 2012).
Numerous empirical studies have investigated these relationships more carefully, including studies with multivariate controls and quasi-experimental designs. These same problematic relationships were reported from many of these studies as well (for example, Barrett, 2002; Colman et al., 2012; Currie, 2003; Fox et al., 2004; Institute of Medicine and National Research Council, 2013). There is evidence in particular studies that food assistance programs increase expenditures on food, but these expenditure increases are less than a dollar for a dollar. Ribar said there is also evidence of positive consumption and specific nutrition effects, especially within the WIC Program, although Besharov and Germanis (2000) strenuously disputed many of the results from the WIC studies.
When specifically addressing food insecurity and hunger, the evidence becomes even more equivocal, Ribar said. For instance, Colman et al. (2012) described as “mixed” the evidence coming from a small number of WIC studies. The 2013 IOM/NRC study took a more positive view of SNAP. However, Ribar commented that while the report pointed to a handful of studies that gave positive results, the authors tended to overlook other studies with less positive results. Ribar stated that it can be very difficult to get a study published showing negative results associated with SNAP, and he believes that the IOM/NRC study overlooked this publication bias in its analysis.
According to Ribar, a standard set of questions to ask when assessing gaps in assistance programs will help to illuminate the issues: Are the benefits the correct size to do the job; do the programs cover the right people; and do the right people enter the programs? The issue of the sufficiency of benefits is particularly salient for SNAP. Other programs are not
intended to feed an entire household. WIC is a supplemental program, as are the school lunch and breakfast programs.
The maximum SNAP benefits are set each year based on the cost of the USDA Thrifty Food Plan, a basket of foods that can be obtained at low cost while still providing a nutritious diet. Ribar expressed concerns with this approach. A lag exists between the time that the cost of food is measured and when the benefit number is set, and food prices can rise during this time. Even more problematic to him, the process relies heavily on assumptions about households’ capabilities and time availability to convert raw ingredients into meals in the way that the Thrifty Food Plan intends. Often these capabilities and time availability do not exist in at-risk households. Third, the plan leaves little margin for such things as spoiled food, utility disruptions, appliance breakdowns, or pest infestation.
In addition, Ribar pointed to coverage gaps. Certain households are not eligible for SNAP, such as noncitizen immigrant households, who are generally prohibited from receiving SNAP benefits. Other rules exist that restrict access to benefits, such as those that apply to felons and individuals with drug convictions. There are limitations on the use of SNAP electronic benefit transfers and WIC vouchers to authorized retail establishments. This limitation could restrict the purchases of otherwise eligible food items.
School and childcare meals are limited to enrolled children, they are generally only provided when the children are in school, and they are not offered at all schools. Ribar termed universal coverage of the SBP in all public elementary schools as elusive, and he also reiterated the gap in coverage in some charter schools.
Ribar cited Currie (2004), who reviewed research on program take-up and offered three principal reasons for incomplete take-up. One is that households may not know that they are eligible or may not have the information to apply. There is evidence that when households receive more information, they do take up the program. Second, the administrative burden to apply for these programs may deter participation. Ribar and Edelhoch (2008) found that many people who end up leaving SNAP because of administrative hurdles are lower in income distribution and face substantial challenges and unstable circumstances. The third reason found for incomplete take-up is stigma (Moffitt, 1983). Taking away the stigma, as with the Universal Free Breakfast Program, increases participation by free-eligible children. Haldeman and Ribar (2011) found that free-eligible participation went up 7–13 percent. In an eligibility-based system, everybody knows who is going into the cafeteria and why. When the program is open to everyone, or better yet served to everyone in a school classroom, it no longer has the stigma attached, Ribar said. A fourth rea-
son is program complexity. With a long list of programs, it is difficult for households to understand and distinguish among them.
Ribar stated household behavior can affect program outcomes. On the one hand, there are the Ricardian results he described earlier, in which protective parents will mediate the effects of withdrawals of program support. On the other hand, certain types of support like SNAP and WIC require a great deal of capability on the part of parents. If parents lack that capability, they may not be in a position to convert the assistance that they are getting into nutritional outcomes for their children. Other, nontargeted household members may share targeted resources such as WIC benefits.
Ribar stated that assistance programs are complex, and measuring their effectiveness is even more complex. Households may receive benefits from multiple programs, making it difficult to judge the effectiveness of an individual program. He stated that the interaction of benefits from multiple programs is typically not modeled in research work. Additionally, food hardships rarely appear as the only problem in a household (Joyce et al., 2012). Instead, food problems often co-occur with hospitalizations and poor health of family members, housing insecurity, and energy insecurity. Households may be receiving assistance from other programs, which in turn positively impact food insecurity. As discussed earlier, Ribar noted that some people advocate looking at food systems more broadly to include problems in communities, reduced economic opportunities, neighborhood food resources, and emergency assistance (Ganapathy, Duffy, and Getz, 2005). Little of this complexity is captured in quantitative research, Ribar observed.
Methodological Gaps and Challenges
Ribar said the biggest methodological challenge to examining childhood hunger is the low statistical power in existing data for quantitative analysis. With very small numbers of observations, high-quality multivariate analysis with appropriate controls and disaggregation is difficult. The estimates are not precise, and it may be difficult to correctly do the analysis, because most statistical theory depends on asymptotic assumptions. When looking at only a few dozen positive responses, some of these assumptions are questionable, Ribar opined.
There are many issues associated with the measurement of food hardships. The 2013 IOM/NRC study pointed to several limitations in measurement, and Ribar stated that USDA is considering that report’s recommendations.
Ribar identified additional issues in the measurement of childhood food hardships. An important issue is the social undesirability of parents admitting that they let their children go hungry. Survey methodological
research has shown that questions about socially undesirable actions often are underreported.
The Household Food Security Survey Module (HFSSM) uses extensive screening questions that assume standard household coping strategies. They serve a good purpose, Ribar said, in that they reduce the respondent burden and screen out certain types of reporting errors, but they may also lead to underreporting of child hunger. In order to be asked questions regarding children, the respondent must have screened into that module by affirming other hardships. If the coping strategies for the household are different from the standards assumed, child hunger may be missed.
Ribar said the measures themselves are typically used ineffectively. Researchers rarely utilize all of the information contained in the HFSSM, but instead use a binary indicator for a particular threshold of hardship. Ribar suggested a richer set of results might be beneficial (see DePolt et al., 2009; Wilde and Nord, 2005).
Ribar identified alternative ways to measure program effectiveness, such as pantry inventory checklists (Bryant and Stevens, 2006). He said one advantage to the checklist approach is that it is harder for people to know the purpose of the questions. Diary methods and inventory methods have less scope for social desirability. Another approach used in numerous local studies is an eight-item measure focused more on hunger (developed by the Food and Research Action Center; Wehler et al., 1992). Ribar said there are also problems with measuring “participation” in food assistance programs, and lack of statistical power is a big issue. Participants are a modest proportion of the total population. There are small groups of nonparticipating eligibles and near-eligibles about which little information is available.
Ribar observed that the entire measurement process is made more difficult by the many combinations of different programs (federal, state, local, and private). Most researchers use methods that examine one program at a time, particularly, he said, in economic studies. Ribar termed household self-selection as another methodological challenge. Food program participation is not randomly assigned, and take-up requires a household to take several active and time-consuming steps. Some households with certain characteristics or living in certain conditions are more likely to take these steps and to successfully enroll than others. The challenge in quantitative analyses, he said, is that the very conditions that may determine whether a household enrolls in a program are also likely to affect children’s food insecurity and hunger.
In conclusion, Ribar provided a number of suggestions for research. First, he suggested testing new and alternative measures of food inse-
curity and program participation, using split-sample modules on the Current Population Survey (CPS) HFSSM for cost-effective testing. For example, most respondents would be asked the standard items from the HFSSM while a minority would be asked a version that has been modified with alternative questions. Ribar also proposed the development of measures for hunger-specific items, other food outcomes, and examination of other periodicities. In addition he pointed to the need for enhanced use of information from existing measures, expanding beyond the use of simple binary measures, and consideration of behavioral item response theory models. He also suggested making more and better use of food assistance program administrative data.
Ribar suggested a stronger focus on intermediaries in the program delivery process, and said research to understand the roles of intermediaries and their effects on the outcomes would be important. He noted federal assistance relies on government intermediaries, school authorities, local organizations, and parents to supply nutrition, but researchers do not typically look at these as actors in models.
He further suggested a closer focus on households and how they make decisions. He said households are often viewed as a black box: Inputs, including food assistance program benefits, enter in one side and nutrition somehow appears on the other side. A better understanding about how nutrition for specific household members is produced and the accompanying challenges to this production are important. He said Chilton’s presentation in Chapter 5 provided enlightening information about how households might operate. Ribar suggested that research focusing on participation in multiple programs will also be important. Multiple program use is widespread (Newman et al., 2011), and SNAP eligibility leads to categorical or adjunctive eligibility in other programs. Food assistance program recipients also commonly use community resources (Mabli et al., 2010). Again, the inteconnection between programs appears in some studies, but, in Ribar’s view, not often enough. He said researchers should investigate the range and combinations of food assistance program and include these combinations in empirical work.
Food assistance programs are helping households that exist in the context of multiple problems, including health issues, housing insecurity, and energy insecurity. The household may be receiving assistance from programs other than food assistance programs, and is certainly making decisions within the context of the multiple problems. Ribar said understanding this concept and including it in research efforts is key. He commented that qualitative methods applied to especially vulnerable populations may make quicker advances than the current quantitative methods.
Finally, Ribar said, research focused on programs that directly help children is important. Most children are doing okay, but children are
especially vulnerable in these models, and some are falling through the cracks. He characterized as useful research a focus on programs that build additional capacities for children in the situations where their parents or some other institutions may not be as capable. Specifically, he suggested a focus on new programs that (1) can feed children when schools and childcare centers cannot, and (2) empower children to produce nutrition by cooking and preparing food for themselves.
STATEMENT OF LARA SHORE-SHEPPARD
Shore-Sheppard described her research on food security as documented in Schmidt et al. (2012), in which they examined the effects on food insecurity of five major safety net programs. The first three were “cash” programs: Temporary Assistance to Needy Families (TANF), Supplemental Security Income (SSI), and the Earned Income Tax Credit (EITC). Also examined were a health program (Medicaid/Children’s Health Insurance Program [CHIP]) and SNAP.
She said that the goal of this research was to pull these programs together in a multiprogram examination. She said she agreed with Ribar that this approach is atypical of current research. Thus, not much is known about the effect of nonfood safety net programs on food insecurity. However, she noted, nonfood programs expand resources available to the household, and they may also change the household’s allocation toward food. In addition, enrollment in nonfood programs can affect eligibility for, or enrollment in, food programs.
Shore-Sheppard reported that the interactions are complex. If a household becomes eligible for one program, it might automatically become eligible for another program. Often, she described a “crowding-out effect” in which a household that receives benefits from one program may see its benefits reduced in a different program. For household decision making, the overall effect of these program interactions is somewhat ambiguous. The income effect gives households more resources, making it possible to buy more or higher quality food. However, a substitution effect is also at play in which a household that gets resources not targeted towards food might purchase other goods instead.
She reported that Schmidt et al. (2012) used data from the CPS HFSSM from 2001–2009. The subsample used in the study included households with at least one child younger than 18 years and a reference person between 18 and 64 years of age, and it had a focus on single-parent and low-income households. Immigrant families were excluded because of the added complexity of determining their program eligibility.
She reported that survey respondents reported how much their household spent on food, their use of food assistance programs, and whether
they were able to afford enough food. She noted that the income measure collected in the CPS is crude for this type of analysis, because it includes post-benefit income rather than the pre-benefit income. To offset this, they matched respondent data to other rotations of the CPS to capture earnings data collected when the household rotated out of the sample.
She explained the overall model regresses the outcomes of interest, such as food security, on measures of benefits for which the household would be eligible by demographic cell, state, and year.
Eligibility and benefit levels for the five major programs were imputed based on a set of “calculators” developed by Shore and her colleagues, using program rules and trying to account as carefully as possible for the interactions between the programs. She noted that benefits are likely endogenous. She and her colleagues developed simulated average benefits that are arguable exogenous variables as the difference between the household’s imputed benefit and the average imputed benefit computed over all households in the cell (Currie and Gruber, 1996). Detail can be found in Schmidt et al. (2012).
Shore-Sheppard reported that the take-up of safety net programs is low, and it depends on unobservable variation. She and her colleagues examined the relationship between participation from the March CPS Annual Social and Economic Supplement (2002–2010) and the participation as predicted by their model. For each program, their predicted participation positively predicted actual participation in that program.
She summarized the findings: A more generous cash and food safety net reduces low food security in families with children; there is no evidence that the distribution between cash and food affects food security; and there is no evidence for an effect of health insurance provision.
Shore-Sheppard went on to describe data limitations and gaps. First, she said that immigrants face a diverse and complex set of rules that are difficult to model. She suggested adding a variable to the CPS food security supplement to collect length of time an immigrant has been in the United States. This would help support analysis because many state rules are based on length of time in this country. Second, she and her colleagues combined multiple datasets in order to have a measure of food insecurity, family economic circumstances, and program participation in one place. She suggested a research project to create a single database for researchers to use or to develop an approach that would make linking easier and more direct.
Shore-Sheppard went on to address other knowledge gaps, saying that it would be useful for researchers to consider how the public safety net combines with the private safety net. For example, economists have thought a lot about “crowd-out,” referring to Gruber and Hungerman
(2005) who looked specifically at religious charitable giving versus state programs.
In trying to model household eligibility, Shore-Sheppard and her colleagues noted that the current process places a premium on parents who can manage complexity, and there is little knowledge of why some low-income households have low food security and others do not. She wondered about seasonality of food security status, an aspect that will not be picked up in current measures where surveys are conducted once a year. She noted that researchers know little about how resources are translated into nutrition and health. She suggested that better measures of how the food outcomes that are measured, like food security, play into hunger and nutrition at the individual level would be useful and may indicate a role for parental and child education. She went on to suggest that a single dataset with both nutrition measures and resource measures, along with multiple measures during the year would be invaluable.
In closing, Shore-Sheppard provided two big-picture suggestions. One was to more support for experiments within the programs to get around the problem that in many safety net programs, there is not much exogenous variation in eligibility that can be used to determine effects of the programs. The Massachusetts Healthy Incentives Pilot3 is one example of an experiment. Experiments could take the form of providing information, like a large-scale version of Daponte, Sanders, and Taylor (1999). Her other suggestion was to use the opportunity presented by the implementation of the Patient Protection and Affordable Care Act (P.L. 111-148). Extensive information will be gathered concerning eligibility from Medicaid and subsidies that involves real-time linkages between employer databases and government databases. This new database may provide instantaneous information about a household’s eligibility. If this information could be used for a household to enroll in SNAP, that could eliminate a lot of the issues related to application and re-enrollment.
STATEMENT OF JAMES WEILL
Weill presented four categories of what he considers potentially fruitful research, as well as some independent comments about topics discussed in the workshop. He first pointed to the importance of state and local policy choices as they affect food security, not just as helping experimental design, but also in pointing a path toward what works in these programs and what does not. Praising the research of Nord and others on this topic, he said not enough has been learned from either the good or
3See http://www.mass.gov/eohhs/consumer/basic-needs/food/snap/hip/ [August 5, 2013].
the bad state-level policies. He said there is probably more state variation than the research community picks up, and more interaction between the advocacy and research communities could help identify where the variation exists.
He reminded the audience that Alaimo challenged researchers to address the question of growing inequality and household economic struggles (see Chapter 6). Weill said the importance of that challenge cannot be overstated. He said that there is a risk of asking food assistance programs to do too much if the underlying economic situation is not first addressed.
Weill noted that he overheard a colleague state that the best use of $10 million to address the causes and consequences of child hunger would be to learn to communicate with politicians. Weill said while he would not address that statement, polls show that public support of SNAP and similar programs remains strong. He said while this workshop has considered how research plays into political programs, it is important for researchers to understand that there is a very strong backdrop of public support.
Returning to the four categories of potentially fruitful research, Weill stated that the first is to explore more deeply food insecurity’s adverse consequences and the role of public policy responses in averting such consequences. Weill pointed out that the workshop discussions have dealt little with consequences, an area that he said would benefit from further discussion. Second, he suggested research that is particularly timely and important from the viewpoint of struggling low-income families themselves. Third, he suggested taking advantage of new opportunities created by recent policy and economic changes. Fourth, he suggested research to focus on immigrant families.
Weill said his first suggested research category is looking more at the consequences of food insecurity to child development, health, school readiness, mental health, school achievement, adult workplace productivity, and other effects, which is the central and explicit part of the Section 141 mandate. Congress’ interest in this area is not accidental, he said, characterizing U.S. politics and policy as typically utilitarian. The idea that religious, moral, or ethical reasons indicate a collective desire to avoid hunger among children or adults has some resonance, he said, but seldom moves policy or politics. He stated that what has more impact is research and findings on costs and benefits to address the outcomes of the increased prevalence or severity of food insecurity, how health and ability burdens associated with food insecurity affect private and public systems, and the impact of private and public costs. He noted there has been considerable research in the past on these issues, but that a new generation of research on the consequences of food insecurity will be important.
Weill’s second category of suggested research relates to the importance of more research on the cumulative long-term human and social
costs of allowing people to suffer food insecurity for extended periods of time or for several times during several years. He referred to Children’s HealthWatch research (for example, Joyce et al., 2012), and noted longitudinal studies and qualitative studies are key. Some of the work that has been done by Bartfeld and her colleagues and by Children’s HealthWatch4 starts to show a path, he said.
He noted that more analysis of marginal food insecurity’s impact on health and well-being will be important, referring to research that demonstrates the detrimental effect of marginal food security. The more the cumulative impact of food insecurity and consequences to children can be demonstrated, the more likely it is to have an impact on moving public policy. More research on the food insecurity and nutrition impacts of the new low-wage, part-time work, contingent-worker economy will also be important, according to Weill.
Weill said that some of workers’ lost income and benefits have been replaced by the combination of the EITC and the Refundable Child Tax Credit, as well as by Medicaid, SNAP, CHIP, school meals, and childcare supports. That substitution has been inadequate, and he suggested more research of the type Shore-Sheppard discussed. Weill said analysis of the impact of the change in the nature of low-income work itself, the increasingly contingent nature of jobs, the increase in part-time work, erratic employment, and nonstandard hours will be important. He cited research by Coleman-Jensen (2011) on this topic that suggests greater food insecurity when wages come from nonstandard work arrangements.
He stated that the workshop did not sufficiently address the adequacy of benefits. He reported that the 2013 IOM/NRC study recently concluded that SNAP benefits are too low and identified flaws in how they are calculated, and USDA followed up by asking for more research. Weill argued that both research and action are critical.
Weill noted that with continued wage stagnation for the bottom third of the population, research is needed to figure out how to make SNAP a more adequate support that will carry families, including low-wage working families, through a month and improve outcomes. This inadequacy issue underscores the importance of being cautious about overstating shortcomings and understating the positive impacts of the existing programs. Those programs are the strongest strands in the safety net for children, and they are crucial in preventing hunger and increasing food security. They certainly could be structured and managed to do much more, he observed. There are shortcomings in preventing hunger as shown in some research with mixed results, but some of that could well be the selection bias and much of it is probably due to inadequate benefits.
He reported that another area of importance to struggling families is the interaction of food insecurity, low wages, inadequate community and family resources, stress, and the harm caused both to parents and often through parents to children. He referred to Ribar’s comment about the importance of research looking at outcomes inside a family’s “black box.” Weill added that it is important to look at how different family members bear the consequences of household food insecurity and how it changes over time.
Weill’s third suggested category of research related to recent policy and economic developments, such as the Patient Protection and Affordable Care Act and the boost to SNAP benefits enacted by Congress in the American Recovery and Reinvestment Act of 2009 (P.L. 111-5). Whether or not it continues, from a research point of view, the boost is hugely important as a source of potential information. Weill said that Nord and Prell (2011) reported that the boost increased food expenditures and reduced food insecurity and a Children’s HealthWatch policy action brief on October 20115 stated that the boost protected young children’s health, noting that much more can be done to ascertain the impact of better benefits. He also suggested research to consider whether improvements in benefits in SNAP and SNAP-like programs can be viewed as key mechanisms to boost food security. He called for a close examination of the impact of 2008 and 2009 congressional actions that significantly increased the EITC value and particularly the value of the Child Tax Credit; growth in participation in the SBP; and changes in school meal standards as other areas of potential research.
Weill noted the recession and its impact on family incomes and food expenditures and relation to food security and hunger is hugely important. As an example, he pointed to recent data that the median African American family and the median Hispanic family spend less than the Thrifty Food Plan amount on food. This is true not just of SNAP families or food-insecure families, he stressed, but all families in these racial and ethnic groups. Weill commented that the recession has been a tragedy, but noted that it will be important for researchers to aggressively seek out the lessons to be learned from it. Lastly, Weill talked about the importance of researching the extent of hunger and the relation of food security to the public policy environment for families with immigrant members and for Hispanics as a community with significant numbers immigrant members. He said that hunger spiked in the immigrant community after the 1996 Welfare Law terminated SSI, Food Stamps, Medicaid, and TANF for almost all documented immigrants. Undocumented immigrants were always ineligible.
5See http://www.childrenshealthwatch.org/upload/resource/snapincrease_brief_oct11.pdf [August 13, 2013].
He said that food insecurity in Hispanic households remains high, and noted immigration reform may have an impact. Weill said his organization and others are advocating for access for immigrants to key health and nutrition programs. What this means for the workshop discussion, he said, is the potential for important research looking at the food insecurity impact on immigrant families and on the programs for which they are and are not eligible. Research on the question of how to ensure that needy immigrant families can access benefits, even if only some members are eligible, will be important in the future.
Craig Gundersen (University of Illinois at Urbana-Champaign) responded to Ribar’s presentation, saying that it is almost a stylized fact that SNAP leads to reductions in food insecurity once self-selection is properly addressed. He pointed to DePolt et al. (2009) concerning research on childhood hunger programs. This research showed that SNAP leads to reduction in food insecurity, not even controlling for selection. He also referred to Kreider et al. (2012a), which showed that while imposing relatively minor and innocuous assumptions, SNAP leads to up to a 14.2 percentage point decline in food insecurity. His point, he concluded, is that the literature has many examples showing that SNAP leads to reductions in food insecurity.
Ribar, a coauthor of DePolt et al. (2009), responded that he stands behind the work and noted that he and his colleagues were able to use some different research methods, including (1) longitudinal measures that used fixed effects controls, (2) multiple program data that could be examined in another study, and (3) the opportunity to examine response relationships. This study was one of the few that came out in the anticipated direction, and he pointed out that putting it in the context of many other studies that did not get the anticipated result is also important.
Mark Nord (Economic Research Service) supported Gundersen’s comments, saying many recent, well-constructed studies have showed fairly conclusive evidence of the effectiveness of SNAP. He also clarified a point made by Ribar: The questions about children’s food security in the CPS HFSSM are now asked in any household with children, regardless of how the household respondent answered questions about adult food security. Households with incomes under about 200 percent of the poverty line are asked three food security questions about children. Analysis shows that there is essentially no bias on children’s food insecurity measures caused by the questionnaire screening.
James Ziliak (University of Kentucky) commented that, as a member of the panel that prepared the 2013 IOM/NRC study, looking at the
adequacy of SNAP benefits, he supported Gundersen’s comments and noted the panel members could be objective because they were not direct contributors to the relevant research. Ziliak acknowledged other studies find the opposite effect, but it was the panel’s perspective that most well-crafted studies on this issue showed that these programs reduce the likelihood of food insecurity. Ziliak reported that the panel found that when the researcher appropriately dealt with the issues of self-selection and mismeasurement, there was substantial evidence that these programs reduce the risk of food insecurity.
Ribar responded to Ziliak’s remarks, saying that the panel reviewed numerous studies that had mixed results, pointing to the methodological challenges with this area of research. He reiterated that it is difficult to argue that giving people more food does not somehow lead to better food results. Edward Frongillo (University of South Carolina) asked Ribar whether the result that “children go hungry less” was a logical consequence of the standard rational economic model, or whether there was direct evidence. Ribar responded that it is a standard economic assumption that parents are altruistic, which leads to the within-home crowd-out effect in the theoretical model. However, he said that there are studies that have looked at coping strategies and the typical behavior is that the children do go hungry less. Frongillo noted that he would like to see citations to that research.6
Jay Hirschman (Food and Nutrition Service) noted that researchers who focus on SNAP and general food security issues may not have seen the evaluation of the Summer Electronic Benefit Transfer for Children documented in Collins et al. (2012). He noted that this analysis is the product of an $85 million appropriation from Congress to conduct demonstrations with rigorous evaluations. He said that it is a clear demonstration that very low food security for children is subject to improvement through a known form of changing benefits. In this case, the benefit was $60 per summer month for each school-aged child in the household. Weill responded that Collins et al. (2012) is a great study for SNAP benefits but a mixed study for summer food purposes. Ultimately, it is common knowledge that children are better off if they are in programs over the summer. Funneling money into the family in lieu of building programs where they can be fed meals and also get mentoring, tutoring, and activity is an artificial choice. Both are important.
6One citation discussed during the workshop was Edin et al. (2013). In her statement, Zapolsky stated that this result showed that SNAP helps parents protect children. Another reference provided by Ribar during his talk was McIntyre et al. (2003).