Supermarket Purchases Over the SNAP Benefit Month: A Comparison Between SNAP Participants and Non-participants

Abstract

Introduction:

The Supplemental Nutrition Assistance Program (SNAP) provides financial assistance for food and beverage purchases to approximately one in seven Americans, with benefits distributed once monthly. The majority of SNAP benefits are spent early in the month, leading to decreased caloric intake later in the month. The effects of this early benefit depletion on the types of foods and beverages purchased over the course of the month is unclear.

Methods:

Using individually tracked sales data from 950 participants enrolled in two supermarket-based RCTs in Maine (October 2015–April 2016 and October 2016–June 2017), purchases of selected food categories by SNAP participants (n=248) versus non-participants (n=702) in the first 2 weeks compared with the last 2 weeks of the SNAP benefit month were examined. Analyses were completed in 2019.

Results:

For SNAP participants, adjusted mean food spending decreased 37% from the first 2 weeks to the last 2 weeks of the SNAP benefit month (p<0.0001) compared with a 3% decrease (p=0.02) for non-participants. The decline in spending by SNAP participants occurred in all examined categories: vegetables (–25%), fruits (–27%), sugar-sweetened beverages (–30%), red meat (–37%), convenience foods (–40%), and poultry (–48%). Difference-in-difference estimators comparing SNAP participants with non-participants were statistically significant (p<0.05) for all examined categories.

Conclusions:

In the second half of the SNAP benefit month, individuals reduced purchases of all examined categories. More research is needed to understand the impact of these fluctuations in spending patterns on the dietary quality of SNAP participants.

INTRODUCTION

The Supplemental Nutrition Assistance Program (SNAP) aims to reduce poverty and food insecurity by supplementing the food budgets of low-income households. In 2018, SNAP had an annual budget of more than $65 billion and served one in seven Americans each month.¹ Benefits are issued to participants once monthly, but most households spend the majority of benefits in the first 2 weeks after issuance^2–4 and may not have sufficient funds for groceries at the end of the month. The resulting food deprivation has important health effects, especially in the most vulnerable, high-risk populations, such as children and adults managing diet-related diseases.^5–7 For example, prior studies have documented declines in energy intake and diet quality,^2,8–10,13 increased hospital admissions for hypoglycemia,^5,6 and lower standardized test scores and more disciplinary actions for students over the benefit cycle.^11,12

However, there is a lack of evidence documenting fluctuations in purchasing patterns of specific food and beverage groups using individually linked sales data. Specifically, it is important to determine if once-monthly distribution may differentially affect foods and beverages purchased throughout the month. The use of individually linked sales data, which provide an objective measure of all supermarket purchases including produce, bulk items, and deli and other prepared foods over an extended time period, provides the opportunity to build on previous studies examining fluctuations in purchasing patterns related to issuance and can provide evidence for the design of future supermarket-based interventions and policy proposals. Other data sources, such as FoodAPS or Nielsen Homescan, either have fewer included product categories, rely on customer self-report, or only have limited time periods of purchases. Given evidence that SNAP participants will spend more on food using SNAP dollars than non-SNAP income,¹⁴ an important next step is to determine whether this relationship holds true for purchases of specific food and beverage categories of public health importance.

To address this gap in the literature, the goal of this paper is to examine fluctuations in purchasing patterns of selected food and beverage categories over the course of the SNAP benefit month using individually tracked sales data from 950 shoppers in Maine, half of whom were selected randomly for a double dollar fruit and vegetable discount. The hypothesis was there would be greater spending during the first 2 weeks versus the last 2 weeks following benefit issuance among SNAP households compared with non-SNAP households, and that the composition of the average SNAP transaction would change over the course of the month as benefits were depleted, perhaps with more convenience, packaged, or temperature-stable foods early in the month and perishable items throughout the month or while funds last. An additional aim was to explore whether any fluctuations in patterns in shopping were impacted by the fruit and vegetable discount program.

METHODS

Study Population

This study is a secondary analysis of 950 adult shoppers who participated in two separate randomized trials to test a same-day double dollar fruit and vegetable incentive of up to $10 per shopping trip.¹⁵ Each study took place in one of two large supermarkets from the same chain, which were located in low-income Maine communities.

Methods for recruitment, randomization, data access from point-of-sales systems, and food categorization have been described previously.¹⁵ Briefly, shoppers were recruited in the entryway of the two study stores. Participants were given a loyalty card at the time of enrollment that provided a 5% discount on all purchases at the study store during the study period and allowed for tracking of their purchases. Each study consisted of a several-month baseline period (when the 5% discount began and purchases were tracked), followed by randomization to a double dollar fruit and vegetable incentive intervention group or a control group.

To be eligible for the studies, adult shoppers had to have a child aged ≤18 years living in the household. Data from these studies provided up to 8 months of linkable sales data collected from the participants. Participants were excluded from the present analysis if they never used their loyalty card (n=52), if a duplicate ID number was assigned (n=2), or if they participated in both trials (n=2) (Figure 1).

Figure 1.

CONSORT flow diagram.

Measures

Item-level scanner data were obtained from each of the two study stores. Scanner data included price and quantity of all items purchased in each transaction, including packaged foods and beverages (identified via universal product codes) and fresh meats and produce (identified via product lookup codes), as well as payment method (e.g., SNAP or cash) and the date of purchase. Purchases were linked to participants via loyalty card ID and the date of purchase was used to aggregate purchases to the daily and weekly level.

This analysis focused on six categories given prior research showing that, in transactions where SNAP benefits were used, there was higher relative spending on sugar-sweetened beverages (SSBs), red meat, and convenience foods (prepared or packaged foods, such as boxed macaroni and cheese) and lower relative spending on fruits, vegetables, and poultry compared with transactions made without SNAP benefits.¹⁶ From the daily point-of-sale data, spending was aggregated in each category by day and week for each participant and assigned to the specific day and week of their monthly SNAP benefit cycle.

An enrollment survey gathered information on SNAP participation status, date of monthly SNAP benefit receipt, race/ethnicity and sex of the primary respondent, and size of household. SNAP participants receive benefits on the same day each month; the time between issuance periods is referred to herein as the “SNAP benefit month” and each week since issuance as the 1st, 2nd, 3rd, and 4th week of the benefit month, respectively. Day 1 of the SNAP benefit month was based on participant self-report of the day of benefit receipt. For non-participants and those SNAP participants who did not self-report their day of benefit receipt, Day 1 of the monthly cycle was defined as the 10th of each month (in Maine, SNAP benefits are issued between the 10th and the 14th of the month¹⁷; 94% of SNAP participants self-reported their day of benefit receipt).

Statistical Analysis

Difference-in-differences models were used to compare mean spending per transaction on all food and on selected categories from the first 2 weeks after SNAP benefit issuance with the last 2 weeks of the SNAP benefit month by SNAP participants versus non-participants. The primary independent variables were indicators for SNAP participation status, time of the benefit month (first or last 2 weeks of the SNAP month), and a SNAP X time interaction term (the difference-in-differences, which represents the differential effect of the time since SNAP benefit issuance on SNAP participants versus non-participants). Models were adjusted for demographic covariates (race/ethnicity, sex, age, household size), season, store, and intervention arm, and included random intercepts for shopper to account for repeated observations; covariates were selected based on previous publications on purchasing patterns and theoretical association with purchasing behaviors. For the primary model, SNAP participation status was defined based on self-report at study enrollment. Sensitivity analyses were also conducted: (1) using an indicator for monthly SNAP participation status (i.e., based on whether or not point-of-sale data indicated a shopper used SNAP benefits during a given month rather than self-report at baseline to account for people who may have gone on or off SNAP during the study period); (2) comparing the first week of the month to the remainder of the month; (3) restricted to sales during the pre-intervention period only; and (4) separate models stratified by intervention arm and restricted to the intervention period. All analyses were completed in 2019 using SAS, version 9.4.

The Harvard T.H. Chan School of Public Health’s IRB for the Protection of Human Subjects determined that this study was not human subjects research under 45 CFR subsection 46.102(f).

RESULTS

Overall, the majority of participants were non-Hispanic white (94%) and female (84%), had an average household size of 3.9 people, and had 27.3 transactions (3.4 per month on average) during the study period (Table 1). SNAP recipients spent an average of $49.96 per shopping trip on SNAP-eligible items and purchased an average of 14 items per trip, whereas non-participants spent an average of $71.01 per shopping trip and purchased an average of 19 items per trip.

Table 1.

Participant Characteristics of Shoppers in Maine Enrolled in Two Separate Supermarket-based Fruit and Vegetable Incentive Programs^a

Participant characteristics	Overall	SNAP participants	Non- participants
	N=950	N=248	N=702
Trial randomization
Intervention (incentive) group	471 (50)	111 (45)	360 (51)
Control group	479 (50)	137 (55)	342 (49)
Sex
Female	783 (84)	214 (89)	569 (82)
Male	153 (16)	26 (11)	127 (18)
Race/ethnicity
Non-Hispanic white	881 (94)	229 (94)	652 (94)
Non-white (all other racial/ethnic groups)	53 (6)	14 (6)	39 (6)
Age, years
18–29	176 (19)	82 (34)	94 (14)
30–39	393 (42)	90 (38)	303 (44)
40–49	283 (30)	47 (20)	236 (34)
50–59	65 (7)	16 (7)	49 (7)
≥60	17 (2)	5 (2)	12 (2)
BMI (kg/m2), mean (SD)	28.1 (6.6)	29.8 (7.0)	27.5 (6.3)
Household size, mean (SD)	3.9 (1.3)	4.0 (1.7)	3.9 (1.1)
Percent federal poverty linec
≤100%	201 (23.7)	150 (67.6)	51 (8.2)
≤130%	266 (31.4)	176 (79.3)	90 (14.4)
≤150%	307 (36.3)	193 (86.9)	114 (18.2)
≤180%	366 (43.2)	203 (91.4)	163 (26.1)
≤200%	419 (49.5)	212 (95.5)	207 (33.1)
Shopping patterns, mean (SD)
Total transactions per shopper	27.3 (25.6)	22.0 (25.2)	29.2 (25.6)
Spending per trip on SNAP eligible itemsb	66.58 (70.08)	49.96 (65.76)	71.01 (70.53)
Spending per trip on sugar-sweetened	2.44 (5.29)	3.02 (6.10)	2.29 (5.04)
beverages
Spending per trip on fruits	7.43 (14.80)	3.19 (8.78)	8.57 (15.85)
Spending per trip on vegetables	6.38 (10.50)	3.66 (8.91)	7.10 (10.77)
Spending per trip on red meat	6.00 (19.86)	6.32 (22.88)	5.92 (18.97)
Spending per trip on poultry	6.98 (24.92)	4.80 (20.56)	7.56 (25.93)
Spending per trip on cold convenience	4.10 (7.04)	4.38 (7.98)	4.03 (6.76)
foods
Items (food only) per transaction	18.2 (16.1)	14.4 (15.3)	19.3 (16.2)
Monthly SNAP benefit (self-reported, $)	n/a	292.5 (168.6)	n/a
SNAP use (used benefits during study period)	286	227	59

^aValues are number (%) unless noted otherwise. Values may not sum to 100% due to missing data.

^bExcluded transactions ≥$1,000.

^cIf annual income was reported, percent poverty was calculated by dividing the median of the annual household income category by the annual federal poverty guideline for the household size in 2016; if only weekly income was reported, the median of the income category was multiplied by 4.35 to obtain monthly income, which was divided by the monthly federal poverty guideline for the household size in 2016.

There was an observable spike in transactions made by SNAP participants at the beginning of each month and differences in the use of SNAP benefits to make purchases over the month (Figure 2). The proportion of SNAP participants who paid for their transactions with SNAP benefits (versus with other means such as cash or credit card) markedly declined in Weeks 3 and 4 of the benefit month. Half of SNAP participants did not use benefits in the last week of the month, and 31% only did so once or twice over the entire study period. Demographics of SNAP participants who continued to shop with benefits in Week 4 did not appreciably differ from the overall sample (data not shown).

Figure 2.

Proportion of SNAP participants paying with SNAP benefits by time since distribution of monthly benefits.

Adjusted total mean food and beverage spending with and without SNAP dollars decreased 37% from the first 2 weeks to the last 2 weeks of the SNAP benefit month (from $58.06 to $36.37, p<0.0001) for SNAP participants, compared with a 3% decrease (from $63.46 to $61.56, p=0.02) for non-participants (Table 2). For the 248 SNAP participants, adjusted mean total spending significantly decreased across all examined categories from the first 2 weeks to the last 2 weeks of the SNAP benefit month (p<0.01). By contrast, adjusted mean spending for the non-SNAP participants only decreased significantly for two categories (–9% for meat and –15% for poultry, p<0.05). The decline in spending by SNAP participants occurred in all examined categories: vegetables (–25%), fruits (–27%), SSBs (–30%), red meat (–37%), convenience foods (–40%), and poultry (–48%). This decline was greater for SNAP participants than for non-participants (i.e., the difference-in-differences) for all examined categories (p<0.05).

Table 2.

Adjusted Mean Spending (Before Discounts) on Selected Food Groups, by Shoppers at Two Maine Supermarkets During the First 2 Weeks Versus the Last 2 Weeks of the SNAP Cycle^a

	SNAPb				Non-SNAP				Difference-in-difference
Mean spending per transaction ($)	First 2 weeks	Last 2 weeks	Diff(% change)	p-value	First 2 weeks	Last 2 weeks	Diff(% change)	p-value	First 2 weeks	Last 2 weeks	Diff	p-value
Foods and beverages overall	58.06	36.37	21.69 (−37)	<0.0001	63.46	61.56	1.90 (−3)	0.02	−5.40	−25.19	19.79	<0.0001
Sugar-sweetened beverages	3.66	2.57	1.09 (−30)	<0.0001	2.21	2.10	0.11 (−5)	0.10	1.45	0.47	0.98	<0.0001
Fruits	3.91	2.85	1.06 (−27)	0.005	8.31	8.09	0.22 (−3)	0.24	−4.40	−5.24	0.84	0.04
Vegetables	4.68	3.49	1.19 (−25)	<0.0001	7.33	7.27	0.06 (−1)	0.64	−2.65	−3.78	1.13	0.0001
Red meat	9.11	5.77	3.35 (−37)	<0.0001	6.49	5.87	0.61 (−9)	0.03	2.63	−0.10	2.73	<0.0001
Poultry	5.72	2.95	2.78 (−48)	<0.0001	6.91	5.88	1.03 (−15)	0.004	−1.19	−2.93	1.75	0.03
Cold convenience foods	5.23	3.13	2.11 (−40)	<0.0001	3.61	3.75	−0.14(4)	0.12	1.62	−0.63	2.25	<0.0001

Note: Boldface indicates statistical significance (p<0.05).

^aModels adjusted for race/ethnicity (white vs non-white), sex, age, household size, season, store, and intervention arm.

^bShoppers who use SNAP benefits defined as individuals who self-reported receipt of benefits at enrollment.

Treatment arm (intervention versus control) was included as a covariate and was non-significant in all models. In sensitivity analyses, results did not change appreciably when using an indicator for monthly SNAP participation status (i.e., based on whether a shopper used SNAP benefits during a given month, rather than based on self-report at baseline; Appendix Table 1), when comparing the first week of the month to the remainder of the month (Appendix Table 2), when restricted to sales made only during the pre-intervention period (results not shown), or when stratified by intervention group (Appendix Table 3).

DISCUSSION

For SNAP participants, declines in overall food and beverage spending were observed as well as for all examined categories over the course of the SNAP benefit month. Relative declines varied by category, with larger relative decreases for red meat, poultry, and convenience foods than for SSBs, fruits, and vegetables. The monthly decline in spending was greater for SNAP participants than for non-participants within all examined food categories, indicating there are fluctuations in spending specifically associated with SNAP issuance policy.

There are several possible reasons for the observed spending fluctuations by SNAP participants over the course of the month. One reason is the once-monthly frequency of benefit issuance. The larger declines observed for red meat, poultry, and convenience foods would be expected given that these items are often shelf-stable or can be frozen, facilitating shoppers’ ability to “stock up” on items when benefits are newly available at the beginning of the SNAP benefit month. The lesser declines in purchases of fruits and vegetables indicate that shoppers continue to purchase produce at the end of the month when resources are scarce. Previous work has shown that shoppers prefer fresh produce over frozen or canned fruits and vegetables, despite the difference in price and perishability.¹⁵ The incentives were available to shoppers in the intervention groups at each visit to the supermarket throughout the benefit month. It is possible that timing the incentive to coincide with times of decreased benefit availability (i.e., the end of the month) could further influence the observed declines in spending at the end of the monthly cycle. More research is needed to understand the effects of financial incentives for fruits and vegetables on fluctuations in spending over the course of the benefit month.

A second possible reason for the observed fluctuations is retailer response to the timing of benefit issuance. Larger spikes in purchases of convenience foods and red meat were observed at the beginning of the benefit month relative to fruits and vegetables, which could be related to fluctuations in retail prices and marketing of certain food and beverage categories in response to SNAP benefit issuance.^3,19 All 50 states are federally mandated to distribute SNAP benefits to individual participants once per month, but the scheduling for this is determined at the state level. Several states distribute benefits to all beneficiaries on a single day (i.e., single-day issuance) and others distribute benefits over a longer timeframe (from 3 to 28 days each month, depending on state; i.e., extended issuance).¹⁷ Prior research demonstrates that food spending peaks sharply immediately following issuance^2–4; thus, grocers in those states with single-day or compressed (i.e., ≤5 days) issuance schedules can more easily target SNAP customers on these days by changing prices or promoting certain high-demand items such as highly processed foods and SSBs.^18,19

Finally, the spending fluctuations could be due to inadequacy of the total benefit. Previous work has shown that a documented decline in caloric consumption at the end of the benefit month went away during the period of increased SNAP benefits under the American Recovery and Reinvestment Act.²⁰ An increase in overall benefits could have a similar smoothing effect on monthly spending patterns. Future studies that collect comprehensive purchasing data, including item-level payment information, would help to fully evaluate this question.

There are several policy alternatives that could address the fluctuations in spending that are associated with monthly SNAP issuance. For example, benefits could be issued to participants twice per month (“split issuance”), or participants could be given the choice to receive their benefits once or twice based on individual preferences and needs. In addition, all states could be required to distribute benefits over longer timeframes (as many, but not all, currently do) to alleviate the incentive for stores to conduct predatory marketing practices such as promoting SSBs during the dates of benefit issuance.¹⁹ If issuance remains the same, incentives for healthful purchases could be incorporated to align with times when benefits are likely depleted. Alternative issuance policies need to be tested to determine whether they would have the desired effect of promoting healthier purchases. It is also critical to consider the possibility of unintended consequences of such changes. For example, twice-monthly issuance may be preferred by some participants as a budgeting tool, but it may force other participants to shop more frequently than is feasible or preferable.²¹

Finally, it is important to note that both SNAP participants and non-participants had very low purchases of fruits and vegetables throughout the month, much lower than recommended by the Dietary Guidelines for Americans. All shoppers would benefit from additional strategies that lead to more purchase and consumption of fruits and vegetables and other healthy foods to improve health.

This is the first known study to consider the impact of the SNAP issuance cycle on individually linked purchases of specific food and beverage categories of both SNAP participants and non-participants. This study builds on previous research^2,3,5,8–14 with greater detail on total dollars (SNAP and non-SNAP) spent on specific food and beverage categories, a control group consisting of SNAP non-participants, and substantial prospective follow-up over multiple seasons. Use of individually linked sales data and use of difference-in-differences models comparing SNAP participants with non-participants over time allows for stronger inferences regarding the effect of the current once-monthly SNAP issuance policy on shopper behaviors.

Limitations

This study also has limitations. Because SNAP eligibility and other demographics are self-reported, there could be inaccuracies due to under-reporting or to real changes in SNAP status over the study period. When validated information based on sales data was used and included a floating indicator based on use of monthly SNAP benefits, results did not appreciably differ. It is possible that secular trends or current events may have influenced the results, though this impact was likely minimized owing to the inclusion of many months of sales. Only purchasing data were analyzed and consumption patterns by month could not be determined; conclusions address only purchasing behaviors. If food choices shifted within categories (e.g., if customers purchased less-expensive items) spending does not necessarily reflect the quantity purchased. Information regarding what proportion of a transaction, or which specific items, were paid for with SNAP benefits versus other tender types was unavailable. Another limitation is that sales data are drawn from two stores from a single supermarket chain in Maine, which limits generalizability. Shoppers also may shop at multiple stores, and only data from a single store were analyzed for each study. Previous work has shown that 92.4% of traditional food sales take place at grocery stores and supermarkets as opposed to convenience stores or specialized markets.²² Also, participants were recruited who reported doing the majority of their grocery shopping at the store, and all participants were incentivized to take part in these trials through a 5% discount for every shopping trip. Participant feedback showed this incentive was a very successful motivator for shoppers to preferentially use the study store. The study data were not collected for the purpose of this analysis.

CONCLUSIONS

In this study of the monthly shopping habits of SNAP participants and non-participants, fluctuations in spending patterns were observed whereby SNAP participants substantially reduced the purchase of all examined categories over the course of the benefit month, but more so for red meat, poultry, and convenience foods than for fruits, vegetables, and SSBs. More research is needed to understand the implications of once-monthly distribution of SNAP benefits on dietary quality of low-income individuals on government food assistance. Demonstration projects of alternative benefit structures are necessary to determine whether they might alleviate this seemingly detrimental pattern of purchasing that is observed under the current program structure.