What is the methodology of the research, the hypothesis, the type of researchqualitative, quantitative, statistical, case study, etc, and the way the research was conductedsurveys, interviews, content analysis, ethnography, etc?

There is an active literature on the impact of SNAP on the short-term outcomes for beneficiaries, such as consumption, food insecurity, and labor supply. There is also a good deal of work on the longer-term impact of SNAP on children's health and economic outcomes and on the determinants of the SNAP caseload. This brief summary is a subset of the broader review of the literature over the past few decades, published in Hoynes and Schanzenbach (2016).

As illustrated in Figure 1, participation in and expenditures on SNAP have varied significantly over time. At a macro level, this has generally aligned with fluctuations in the unemployment rate. While the macroeconomy is the largest contributor to changes in SNAP caseloads, SNAP and welfare policies have also played a role. For example, welfare reform and reductions in the length of SNAP certification periods led to reductions in SNAP caseloads in the 1990s (Currie and Grogger 2001; Ziliak, Gundersen, and Figlio 2003; Figlio, Gundersen, and Ziliak 2000). Changes in immigrant access to the safety net during the welfare reform period also led to reductions in SNAP participation (Borjas 2004; Haider etal. 2004; Kaestner and Kaushal 2005; Bitler and Hoynes 2013). Ganong and Liebman (2018) find that state policy changes explain a share of the increase in participation in the early 2000s. There has been particular interest in the sharp increase in the caseload during the Great Recession. Ganong and Liebman (2018) find that unemployment explains most of the increase, estimating that a 1 percentage point increase in the state unemployment rate increases enrollment by 15 percent. Ziliak (2015) confirms a large role for unemployment and investigates a wider range of policies, finding a larger impact of policy, perhaps accounting for 30 percent of the caseload increase. Bitler and Hoynes (2016) find that the countercyclical effect of SNAP as measured by the effect of the unemployment rate on the SNAP caseload was larger during the Great Recession compared to the early 1980s recession (although the difference was not statistically significant). Impacts of SNAP on food insecurity An important goal of SNAP is to address food insecuritythat is, having inadequate or uncertain access to enough food for an active, healthy lifestyle. It has Figure 3 Relationship between Expansive SNAP Policies and Participation, 2016 212 THE ANNALS OF THE AMERICAN ACADEMY been a challenge to identify the causal impact of SNAP on food insecurity, and the literature is characterized by a wide range of estimates. Several studies employ instrumental variable approaches, using as instruments state SNAP policies such as certification length and treatment of immigrants (Yen etal. 2008; Mykerezi and Mills 2010; Shaefer and Gutierrez 2013; Ratcliffe, McKernan, and Zhang 2011). The results vary across studies and typically find that SNAP participation leads to decreases in food insecurity, but many are not statistically significant. Borjas (2004) and Schmidt, Shore-Sheppard, and Watson (2016) investigate the impacts of public assistance, including SNAP and other programs, and find that program participation or higher benefit amounts lead to statistically significant reductions in food insecurity. Other, less credible studies using approaches such as household fixed effects (DePolt, Moffitt, and Ribar 2009; Wilde and Nord 2005) or propensity score matching (Gibson-Davis and Foster 2006) have found a positive association between SNAP and food insecurity. Overall, the literature finds sharply mixed results, with some studies finding a positive association between SNAP and food insecurity, some finding a negative association, and some finding insignificant results. Gregory, Rabbitt, and Ribar (2015) recently reviewed this literature and conducted a replication exercise, implementing propensity score matching, longitudinal, and instrumental variables approaches in one sample. They find a range of estimates, illustrating well the difficulty in finding a way to isolate the causal impact of SNAP on outcomes, such as food security. Impacts of SNAP on labor supply By providing unearned income in the form of benefits, SNAP is predicted to have work disincentive effects. The (sparse) empirical literature using strong identification strategies tends to find that these work disincentives are modest. East (2018b) examines SNAP's impact on labor supply, leveraging variation across states and over time in immigrants' eligibility for SNAP in the years after a 1996 rule that removed them from the program. She finds that SNAP causes single women immigrants to reduce their employment rates by 2 to 5 percentage pointsa relatively modest impact among a group in which 65 to 70 percent are employed. While married male immigrants do not reduce their probability of employment in response to SNAP access, they do measurably reduce the number of hours that they workon average, they work about 1.5 fewer hours per week when eligible for SNAP and are 4 to 4.5 percentage points less likely to work full time (on a base rate of 72 percent). There are no detectable impacts on the labor supply of unmarried male immigrants. Hoynes and Schanzenbach (2012) use county variation in the rollout of the program to identify its impact on labor supply. They find no significant impacts on the overall sample, but among single-parent households with a female heada group with high SNAP participation ratesthey find that access to SNAP (then called the Food Stamp Program) in one's county of residence reduces employment by 183 hours annually. They find no significant impacts of the program on earnings or family income, though the estimates are imprecise. Exploring Options to Improve Snap 213 Impacts of SNAP on consumption Since SNAP increases the total resources that a family has to purchase food, economic theory predicts that it should increase food spending and, in most cases, nonfood spending as well. The empirical literature confirms that SNAP increases food spending and appears to increase the variety and quality of recipients' diets (Anderson and Butcher 2016). The literature is unsettled about the magnitudes of the increases in food and nonfood spending. Some studies find that recipients treat the additional resources in the same manner as they would an equivalent cash transfer (Hoynes and Schanzenbach 2009; Schanzenbach 2007), while others find that participants are especially more likely to purchase food with SNAP (Beatty and Tuttle 2014; Hastings and Shapiro 2018). SNAP serves as social insurance, propping up consumption when families face declines in or low levels of income. Studies show that SNAP reduces volatility in consumption and income (Blundell and Pistaferri 2003; Gundersen and Ziliak 2003). Shaefer and Gutierrez (2013) also find that SNAP receipt reduces a family's likelihood of experiencing types of economic distress, such as falling behind on bills or forgoing medical care due to cost. SNAP's impacts on long-term health and economic outcomes The additional resources provided by SNAP likely would be expected to also improve a range of health and education outcomes. In general, studies find improvements in children's health and education in both the short and long run due to SNAP. The literature is more mixed on the short-run impact on adult obesity. There is a small set of studies that examine the effect of SNAP on birth outcomes. Currie and Moretti (2008) use cross-county introduction of the program in California and find that SNAP is associated with a reduction in birth weight, a surprising result driven by increased births among teens and by birth weight declines in Los Angeles County. Almond, Hoynes, and Schanzenbach (2011) examine the effects of the program rollout across all counties in the United States, finding that SNAP improves infant outcomes, especially at the bottom of the birth weight distribution. East (2018a) utilizes changes in immigrants' eligibility across states and over time and finds that the mother's access to SNAP during pregnancy improves health at birth. In addition, she finds that increases in SNAP access during early life improve parent-reported health at ages 6 to 16 (with suggestive evidence of reductions in school days missed, doctor visits, and hospitalizations). Hoynes, Schanzenbach, and Almond (2016) use the cross-county rollout to estimate the relationship between childhood access to the program and adult health and human capital outcomes. They find that early childhood access to SNAP leads to a large and statistically significant reduction in the incidence of "metabolic syndrome" (obesity, high blood pressure, heart disease, diabetes) as well as an increase in the share reporting to be in good health. They also find for women, but not men, that access to food stamps in early childhood leads to an increase in economic self-sufficiency. Bailey etal. (2019) use data from the Social 214 THE ANNALS OF THE AMERICAN ACADEMY Security Administration and Census Bureau and the same identification strategy, finding positive impacts of early life access to SNAP on a number of outcomes including human capital, self-sufficiency, neighborhood quality, longevity, and a reduction in the likelihood of being incarcerated. The predictions of SNAP's impact on body weight are ambiguous. If SNAP improves the quality of family diets, we might expect SNAP to reduce body mass index (BMI) or obesity outcomes. On the other hand, if SNAP causes participants to purchase more calorie-dense foods, then it may worsen these outcomes. The literature on this topic is unsettled. Schmeiser (2012) uses an instrumental variables approach, using state SNAP policies as instruments, and finds that SNAP reduces BMI among children for most gender-age groups. Kreider etal. (2012) address selection into SNAP and measurement error using a bounding approach, finding rather wide bounds that generally cannot rule out positive or negative effects of SNAP on BMI for children. Turning to adults, Fan (2010) uses a combination of fixed effects and propensity score matching and finds no significant effect of SNAP on obesity, being overweight, or BMI. Meyerhoefer and Pylypchuk (2008) combine individual fixed effects and an instrumental variables approach and find SNAP increases obesity among women but has no significant impact on men. Kaushal (2007) finds insignificant effects of SNAP on obesity among immigrants in the postwelfare reform era. Potential Reforms within the Current Structure Overall, extant evidence tells us that SNAP is an effective and efficient program that is well designed to work with the market, target benefits, and provide a phase-out that minimizes cliffs and other nonlinearities that may induce participants to respond to the program in suboptimal ways. Nonetheless, there are a number of potential reforms to SNAP that are often discussed in policy circles and are worth exploring further. Below, I outline a series of potential policy reforms organized around problems to be addressed and analyze them in terms of theory and research. Of course, policy in practice includes not only questions of evidence, but also of politics. I will leave discussions of the latter to those more skilled in that arena. Reforms aimed at reducing food insecurity While SNAP increases families' resources available for food purchases, nonetheless 31 percent of those reporting SNAP participation also experienced food insecurity in 2017. There are several types of policies that can address high levels of food insecurity, including providing more resources (through increased benefit levels or higher participation rates) or helping families to better smooth their consumption between paychecks or benefits payments. One type of policy to attempt to reduce food insecurity is to increase benefit levels broadly among SNAP participants so that they have more money to spend Exploring Options to Improve Snap 215 on food. One way to do this is to increase the maximum SNAP benefit. As Ziliak (2016) explains, the USDA's Thrifty Food Plan (TFP) has become increasingly out of line with actual consumption patterns and assumes an unreasonable amount of time spent preparing meals from scratch. To address these shortcomings, he concludes that a 20 percent increase in SNAP's maximum benefit is needed in the short term and that, in the longer run, an evidence-based SNAP benefit would likely justify a similar maximum benefit level.3 A $30 monthly benefit increase is predicted to decrease food insecurity by approximately 1 percentage point and improve the quality of diets consumed by SNAP participants (Anderson and Butcher 2016). Broad changes to SNAP payments are expensive; for example, Ziliak's proposal would increase SNAP spending by 24 percent but would also have additional spillover benefits to the macroeconomy as well as to dietary quality. Others have proposed replacing the USDA's TFP as the basis of SNAP benefits with its low-cost food plan, which costs about 30 percent more for a family of four (Food Research and Action Center 2012). An alternative approach would be to increase benefits in a more targeted manner, aimed at families experiencing temporary increases in food costs. For example, families of the 22 million children who receive free or reduced-price school lunches (and the 12.5 million who receive free or reduced-price school breakfasts) on an average day typically lose access to these programs when school is out of session in the summer months. To offset these losses, currently the USDA offers the Summer Food Service Program, which provides free meals to children in low-income areas when school is out. The program is funded by the USDA and run by nonprofit community organizations (schools, camps, faith-based groups, hospitals, etc.) that can ensure that children receive meals in a supervised environment. Access to summer meals varies widely across locations, and on average only 2.7 million children participated in the program each day in July 2018 (USDA 2019b). One option to increase coverage would be to temporarily increase SNAP benefits to families with children during summer months. The USDA ran a series of high-quality randomized experiments to test the impact and feasibility of providing additional benefits to families with children during summer months through a Summer EBT program and found that benefits significantly reduce levels of children's food insecurity and very low food security (VLFS) (Collins etal. 2016). When families were provided an additional $60 in electronic benefits similar to SNAP payments, children's food insecurity declined by 20 percent, and VLFS declined by one-third. The following summer, they tested additional policies, including a $30 per month benefit and a Women, Infants, and Children (WIC)-based benefit, which is prescriptive on the types of food purchased. Comparing families who received $60 per month to those who received $30 per month demonstrated that the higher benefit level reduced food insecurity by an additional 10 percent but had no marginal impact on the rate of VLFS (suggesting that $30 per month was sufficient to ameliorate VLFS). The WIC model had better improvements in dietary quality for children, and similar impacts on food insecurity despite lower take-up rates. Which model (the SNAPbased or WIC-based) to expand would depend on the relative administrative costs at scale, as well as other policy preferences. In either case, increased 216 THE ANNALS OF THE AMERICAN ACADEMY summer food support benefits to families with children reaches a higher share of children than is currently served by the summer meals program. Of course, such a program would not convey the same community benefits (i.e., benefits from camps or other enrichment activities paired with the community summer meals), so a hybrid model of community-based and SNAP-based (or WIC-based) resources could be designed in a straightforward manner. Another way to structure a targeted benefit increase would be to temporarily increase benefits to families with teenagers. Children's dietary requirements and food intake both increase during their teen years (U.S. Department of Health and Human Services 2015; Anderson and Butcher 2016), and families with teenagers experience higher rates of food insecurity than those with only younger children (Anderson etal. 2016). Note that these impacts are similar whether the teen is a male or female. Increasing SNAP benefits to families with teenagers would offset the increased costs of feeding older children and would help to reduce food insecurity among this group. Studies have documented the so-called monthly SNAP cycle in which at the end of the benefits month dietary quantity or quality declines, or food insecurity increases, as participants run out of resources and await their next benefit payment (Todd 2015; Whiteman, Chrisinger, and Hillier 2018).4 This has detrimental impacts on other outcomes as well, including children's test scores (Gassman-Pines and Bellows 2018) and emergency department visits among diabetics (Basu, Berkowitz, and Seligman 2017). Some advocate for breaking up SNAP payments from once to twice per month to encourage participants to smooth their consumption across the month. Recent work by Zaki and Todd (2019) find that SNAP participants obtain better value per dollar spent closer to the payment date, for example, by searching for bargains more effectively and purchasing larger item sizes that come with volume discounts. It is an empirical question whether more frequent payments, and how frequent they are, encourages more smoothing overall and whether there are particular groups made better or worse off. For example, Zaki (2016) finds sharp declines in consumption at the end of a two-week pay period. An experimental pilot program may be worth pursuing to study the trade-offs. Reforms aimed at improving dietary quality Another line of proposed reforms attempts to use the program to help improve the dietary quality of participants. In considering this issue, it is useful to think about the economic model of how SNAP works in conjunction with prices and individuals' tastes and preferences. In the absence of SNAP, consumers choose how much to spend on food and which mix of items they want to purchase subject to the prices that they face, as well as their own tastes and preferences.5 For participants to change what they purchase fundamentally requires changing one of these: the prices that they face, their total food budget, or their underlying tastes and preferences. How does SNAP impact consumer choice, then? As illustrated in Figure 2, by providing additional resources for food, SNAP is predicted to increase the Exploring Options to Improve Snap 217 purchase of food (and, to the extent that SNAP benefits are fungible, to increase the purchase of all normal goods). Theory predicts not only the quantity but also the quality of food purchased to increase, because when a household is consuming at subsistence level, it typically prioritizes calories over variety; but as income increases, consumption bundles shift toward more variety and less calorie-dense foods. One way to increase families' total food budgets, which in turn improves the variety and quality of their diets, is to increase benefits levels. Anderson and Butcher (2016) simulate the expected impact of a $30 increase in SNAP benefits and predict increases in consumption of milk, grains, and vegetables, and a decrease in fast food consumption. Applicable lessons also can be drawn from the summer EBT program described above, which found that additional benefits improved dietary quality for children, increasing consumption of fruits and vegetables, dairy, and whole grains, and reducing consumption of added sugars.6 A different approach would be to decrease the relative prices of healthy foods through subsidies or other mechanisms, which would be predicted to cause families to shift their consumption toward the foods with now lower prices. Targeted price subsidies for healthy foods have been shown to be effective in the USDA's randomized controlled trial of the Healthy Incentives Pilot in Massachusettsa program that gave SNAP recipients an immediate 30-cent rebate for every dollar they spent on a narrowly defined group of fruits and vegetables (Olsho etal. 2016). In response to this price rebate, consumption of the targeted healthy foods increased by 25 percent. In recent years, many local areas and a few states have taken a similar approach by awarding bonus dollars for benefits used at farmers' markets, allowing recipients to stretch their food budget further when they buy fresh produce. This approach would raise many difficult policy decisions, including which foods should be eligible for subsidies and what level of subsidy is best; without question, more research is needed into these issues. Nonetheless, the concept of subsidizing healthy foods has many merits, is a market-based solution that respects consumers' decision-making, and is feasible through SNAP. Another option that garners support from some policy analysts is to disallow the purchase of soft drinks or sweetened beverages with SNAP benefits. Proponents hypothesize that by banning the purchase of sweetened beverages with SNAP, participants would curtail or even eliminate their purchase and consumption of these goods. Recall that SNAP benefits are modestapproximately $4.50 per person per dayand nearly all families supplement their SNAP purchases with groceries purchased from their cash income. At the same time, spending on sweetened beverages is also relatively low, averaging $12 per month (McGranahan and Schanzenbach 2011). Economic theory thus predicts that banning purchase of sweetened beverages would not change consumption of these goods, because the policy does not alter prices or consumers' tastes and preferences, and typically does not alter the budget constraint. A typical family that spends a modest amount on sweetened beverages, and importantly supplements their SNAP spending with cash resources that exceed the amount spent on sweetened beverages, would be expected to continue to purchase the identical basket of goods regardless of the banbut they would have to make certain to 218 THE ANNALS OF THE AMERICAN ACADEMY pay for the soft drinks out of their own cash resources instead of their SNAP benefits. In other words, a sweetened beverage ban would be expected to increase the administrative costs of the program to both the USDA and retailers as they would be required to categorize sweetened beverages and monitor their purchase, and it could increase the stigma faced by recipients when they use the benefits, but it would not be expected to induce behavioral changes or dietary improvements. In general, it is difficult to substantially improve dietary quality because dietary intake is based on tastes and preferences that are hard to alter. Policies that change prices or budget constraints show some promise to improve dietary quality through market mechanisms. Other more drastic reforms, such as moving to a more restrictive basket of goods, undermine many of the strengths of SNAP and are discussed further below. Reforms aimed at encouraging work Economic theory demonstrates that providing unearned income such as SNAP benefits is expected to reduce an individual's work effort, though in practice the effects of SNAP on work effort tend to be modest (East 2018b; Hoynes and Schanzenbach 2012). The SNAP benefit formula already attempts to reduce the disincentive to work by providing a 20 percent earned income deduction, and one way to further promote work is to increase this deduction (Schanzenbach 2013). Another way that SNAP reduces the disincentive to work is through waiving the gross income test under BBCE policies. Waiving the gross income test allows households to participate in SNAP that have higher levels of income but also high levels of allowable deductions such as childcare expenses and work deductions. Schanzenbach (2017) finds that SNAP families with gross incomes above standard gross income test level make up a small share of recipients approximately 4 percent of households receiving 1.3 percent of total SNAP paymentsbut that 97 percent of them have earnings. Preserving such families' access to SNAP through waiving the gross income test reduces the benefit tax rate that they would otherwise face, promotes work, and provides needed support to these families. Some analysts advocate for changing the current work requirement rules to further encourage work among SNAP participants.7 The changes most frequently discussed include (1) expanding the groups that are subject to work requirements; (2) alternatively, eliminating work requirements for those currently exposed to them; and (3) changing the eligibility criteria under which states may request temporary waivers from the work requirements. Under current law, able-bodied adults without dependents (known as ABAWDs, who are those between ages 18 and 49 who have no dependents and are not receiving disability benefits) may only receive SNAP for three months in a three-year period if they do not meet work requirements after this time limit. To be eligible after the time limit, an ABAWD must work at least 80 hours per month or participate in a state-approved workfare program. States may request temporary waivers from these time limits when unemployment is high or when Exploring Options to Improve Snap 219 there are insufficient jobs available in an area. Congress temporarily expanded the circumstances under which an area could qualify for a waiver during the Great Recession and suspended the time limit nationwide for part of 2009 and 2010, though states had the option to retain the time limit if they offered work opportunities to those subject to the rule. Economists generally think about the effectiveness of work requirements and incentivesthat is, carrots or sticks approaches to encouraging workin the context of the local labor market. If individuals can increase their employment through exerting more effort, such as by searching for a job more diligently, or being willing to accept a lower-paying job, or working more hours, then incentives and/or requirements can potentially be quite effective. For example, the mid-1990s increase in the Earned Income Tax Credit (EITC) substantially increased earnings among the targeted group of unmarried mothers. On the other hand, the EITC likely had less of an incentive value during the Great Recession because unemployment was more of a function of factors outside of the individual's control such as macroeconomic conditions (Bitler and Hoynes 2010; Bitler, Hoynes, and Kuka 2017). The effectiveness of work requirements in SNAP will similarly vary by factors that influence whether participants can obtain a job, including the local labor market conditions and the individual's work readiness. Bauer, Schanzenbach, and Shambaugh (2018) investigate labor market patterns that inform the likely impact of current SNAP work requirements and potential expansions of work requirements that were under debate in the 2018 Farm Bill. They demonstrate that variability in employment and hours among SNAP participants is high; and as a result, a large share of those who would be sanctioned under SNAP work requirements themselves have substantial work histories and may be falling below the required number of hours due to fluctuations on the low-wage labor market that are out of their direct control (see also Butcher and Schanzenbach 2018). Some of the sanctions against workers could be ameliorated by expanding the time period covered (e.g., instead of a monthly requirement, set a requirement for quarterly or annual hours) or otherwise smoothing across time periods, but such reforms would add complexity to an already complicated policy area.8 Reforms to SNAP's employment and training programs are another approach to promoting work among SNAP participants. The USDA has commissioned several high-quality randomizedcontrolled trials of reforms to these programs, and results are expected in fall 2019. Reforms aimed at targeting benefits or controlling costs Economic growth is often described as the "right way" to reduce program costs. As shown in Figure 1, SNAP costs in large part vary with macroeconomic conditions, going down during economic expansions when more individuals are working and in turn fewer families require the assistance of the safety net. There are several additional policy options that could be adopted to reduce SNAP spending, including reducing eligibility or participation or further improving program integrity. As shown in the previous section and the discussion of Table 3 above, many states have used BBCE to increase eligibility among households with gross 220 THE ANNALS OF THE AMERICAN ACADEMY incomes above 130 percent of the poverty threshold. Nationally in 2017, as shown in Table 1, SNAP families with gross incomes above 130 percent of the poverty threshold make up approximately 5 percent of households and receive 1.2 percent of total SNAP payments, and they are disproportionately families that have earnings; have child, elderly, or disabled members; or both. The CBO estimates that limiting eligibility to those with gross incomes at or below 200 percent of the poverty line (for households with an elderly or disabled member) or 130 percent of the poverty line (for households without an elderly or disabled member) would remove eligibility for 400,000 households and reduce spending by $5 billion over 10 years (Congressional Budget Office 2018).9 Preserving such families' access to SNAP through waiving the gross income test reduces the benefit tax rate that they would otherwise face, promotes work, and provides needed support to these families. Other states have used BBCE to extend SNAP access to those with assets above the federal limit of $2,250 in countable assets (excluding the value of vehicles).10 The asset test reduces the population eligible for SNAP by 14 percent (Ratcliffe etal. 2016). Low-income households with assets above the test limit tend to have assets substantially above the limits, with median liquid asset levels of $36,000, and are more likely to include elderly members. Eliminating waivers to the asset test in SNAP would reduce participation, although the increased administrative cost burden associated with collecting asset data would offset potential savings. As described above, participation rates are currently at 85 percent of the eligible population. Policies to reduce participation rates include adding administrative hurdles to application or renewal, or introducing policies that increase stigma on participants. The literature suggests that such policies will discourage the neediest from participating (Mills etal. 2014; Ribar and Edelhoch 2008). Another way to reduce costs without limiting eligibility or take-up is to reduce fraud and error, although SNAP already performs well on this dimension. One approach is to improve monitoring for dual enrollment in SNAP across multiple states, which may, for example, occur if parents living in different states and sharing custody each claim the same children on their SNAP application, or if a participating family moves across state lines and enrolls in SNAP in the new state but fails to discontinue enrollment from the prior state. The National Accuracy Clearinghouse (NAC) currently monitors for dual SNAP enrollment across multiple states, and a recent demonstration project in southeastern states showed that the NAC could reduce the approximately two thousand cases per month of dual enrollment identified. If implemented nationwide, the NAC is predicted to save $114 million per year in erroneous payments. Reforms aimed at improving SNAP's role in stimulating the economy A stated goal of SNAP is to strengthen the agricultural economy. Since SNAP benefits are quickly spent by recipients, they provide a rapid fiscal stimulus to the economy. During normal economic times, moderately higher benefit levels that are still quickly spent would do more to stimulate the economy, as well as address Exploring Options to Improve Snap 221 food insecurity and related issues as described above. SNAP plays a particularly important stimulus role during economic downturns, with caseloads and spending expanding rapidly along with need. According to the Congressional Budget Office, SNAP is one of three programs that constitute the majority of the automatic stabilization aspect of federal spending, along with Unemployment Insurance and Medicaid (Russek and Kowalewski 2015). SNAP reaches a different population than is served by Unemployment Insurance, including those who do not meet minimum thresholds of hours or wages prior to job loss (Anderson, Butcher, and Schanzenbach 2015). Reforms that harm the program's ability to expand rapidly in times of economic downturns, such as block grants or broad work requirements, would blunt the program's effectiveness as a countercyclical stimulus. SNAP's stabilization impact could be even greater if the benefits schedule is increased during recessions. This was demonstrated in the aftermath of the Great Recession, when for five years Congress temporarily increased maximum benefits and also awarded states additional administrative funds to serve an increased caseload (Keith-Jennings and Rosenbaum 2015). For example, maximum benefits to a family of four were increased by 13.6 percent, from $588 to $668. Blinder and Zandi (2015) estimate that every dollar of increased SNAP benefits spurred $1.74 in economic activity in the first quarter of 2009, and $1.22 in the first quarter of 2015the highest multiplier of any of the policies adopted during the Great Recession. Potential Major Reforms As described in the previous sections, SNAP is a well-designed program that props up family food spending and stabilizes the economy when negative shocks occur. The program works with the market, providing resources that allow participants to obtain food from our highly efficient retail food system through normal channels. Proposals for major reforms to SNAP almost all make the program worsein terms of being less effective or efficientand not better. Some commonly proposed major reforms are considered below. Block granting the program There is some discussion of block granting the program, which would provide a set amount of funding to states and give them more flexibility on how to administer the program. Some lessons can be drawn from the 1996 block grant of Temporary Assistance to Needy Families (TANF), the cash welfare program. After the block grant, there were dramatic decreases in participation (which has been explained by a combination of macroeconomic strength and policy reforms). States shifted their TANF spending away from core supports (basic assistance, work activities and supports, and childcare) and toward a wider range of programs, some of which serve families that are not low income. The program's 222 THE ANNALS OF THE AMERICAN ACADEMY ability to respond to changes in needdue to economic downturns or to shifts in low-income populations across stateswas blunted (Bitler and Hoynes 2016). Block granting SNAP would harm its ability to serve its core purposes. As discussed above, a crucial role of SNAP is to serve as an automatic fiscal stabilizer in times of economic downturn. Its current structure allows the program to expand quickly, providing benefits to families that become eligible for the program due to job loss or other economic shocks. If the program were block granted and the funds were not promptly increased in proportion to economic downturns, some families made eligible during downturns may not be able to receive benefits. This would fundamentally undermine its stabilizing impact on the macroeconomy. Even though Congress could allocate additional resources in times of economic need, this would necessarily come with delays, reducing the ability to quickly stimulate the economy. More complicated block grant structures are possible, such as tying the block grant to population shifts and to inflation, or automatically increasing the grant during economic downturns. Such a structure would mitigate some of the drawbacks of a TANF-style block grant. The Trump administration has proposed that SNAP be converted to a matching grant, where states must pay for a portion of benefits. This would also be expected to dampen its countercyclicality, depending on how the match is set, since states that are worse hit by the recession would struggle to meet their share of benefits at the same time that need is highest in those states. Many states have balanced budget requirements, which would exacerbate their inability to provide adequate matching funds during economic downturns. Switching to a WIC- or commodities-based model In their desire to improve the nutritional intake of participants, some advocate for replacing the current SNAP with a program that is more closely aligned with WIC, restricting benefits use for a specific bundle of goods. There are many ways to design this. For example, one could retain the basic SNAP system, in which participants receive benefits in the form of a dollar-value electronic voucher payment that can be redeemed at regular retail outlets at the regular prices charged. In this case, the list of permissible goods to purchase would be (dramatically) narrowed. Another option would be to adopt a model like the current WIC program, in which participants are given vouchers for quantities of particular goods (e.g., four gallons of milk per month). In this case, participants still use regular grocery stores, but no longer face prices. A third option would be for the government to directly provide goods to consumers.11 Each of these approaches would require difficult policy choices about which items should be included that would be sufficient to serve the wide range of people who participate in SNAP. It is worth noting that WIC participation rates drop dramatically after age one, when the infant formula benefit discontinues, and range from 33 to 15 percent (Schanzenbach and Thorn 2019). A key reason for SNAP's success is that it relies on the private sector to provide efficient access to food, through grocery stores and other retail outlets. Each of these proposals would dramatically reduce the efficiency of the program by Exploring Options to Improve Snap 223 raising administrative costs. To the extent that they move the procurement of goods away from a market-based system where consumers face prices, there will be new administrative inefficiencies and moral hazards as well. Such proposals would also diminish or eliminate the program's ability to act as a local economic stimulus. Students of economics will recognize that these reforms would also be expected to decrease participants' utility. Replacing SNAP with a cash transfer Others question the need for restricting SNAP benefits to purchasing food, arguing that these resources should be available to purchase other necessary goods such as housing and medical care. Some argue that it would be preferable to pay out benefits in cash, which can be spent on any good that the consumer wants to purchase. As described above, economists generally argue that if in-kind benefits like SNAP are paid at relatively modest levels, and most participants would like to spend at least as much on food as their SNAP benefits are worth, then the potential consumption distortion caused by providing benefits in-kind is modest.12 Since the vast majority of participants spend more on food than their SNAP benefits are worth, replacing the program with an equivalent cash transfer would have only a small impact, with a small share of households getting more fungibility. As suggested by Table 1, many of these households are extremely disadvantaged, with little or no cash income. Barriers to work and mental health problems are also high among this population (Anderson etal. 2016). While it is clear that the very poor are not adequately served by the current structure of the safety net, in my opinion this likely would not be improved by replacing SNAP with cash. Having resources earmarked for food among this population likely has benefits that outweigh the efficiency loss in constraining their choices. Conclusion As currently structured, SNAP is an effective and efficient program. Its strength lies in its structure, which is based on a classic means-tested income transfer program as outlined by Milton Friedman (1962) in his negative income tax proposal and is now part of the economics canon. The program makes use of the highly efficient private market for distributing and obtaining food. It provides additional resources for food, while respecting consumer sovereignty in making decisions subject to their incomes, prices, and their own tastes and preferences. Its structure and efficiency allow the program to quickly respond to increased need during economic downturns, stimulating and stabilizing the economy. The inefficiencies caused by its being an in-kind transfer that can only be used to purchase food are modest and, in fact, likely serve to protect the most vulnerable. This basic structure has enabled it to serve a wide range of participants, from the elderly to infants, workers and those not employed, across the entire nation. It has been able to adapt to broad macroeconomic trends, including our aging 224 THE ANNALS OF THE AMERICAN ACADEMY demography, increasing labor force participation among women, and stagnating wages for workers with lower education levels. To be sure, there are ways to modestly improve the program as described above. For example, policies to reduce food insecurity among families with children, especially during the summer months, are worth pursuing. There are also market-based approaches to improving dietary quality, for example by subsidizing healthy foods, that show promise. Straightforward changes to the benefits formula during economic downturns would also strengthen the program's ability to stabilize the economy during recessions. However, any potential reforms need to be carefully weighed so that they do not inadvertently reduce the many strengths of the program. The major reforms considered here generally move the program away from a marketbased program that can quickly respond to changes in need and are not advisable. SNAP has been effective for the past half century, and with its current structure it should also be well-suited to meet the challenges of the next half century