Does distributing SNAP benefits later in the month smooth expenditures?

Abstract

In this paper, we explore whether the alignment of the date a household receives Supplemental Nutrition Assistance Program (SNAP) benefits with the start of the calendar month affects the smoothness of household monthly expenditures. Across states, what day in the calendar month households receive SNAP benefits varies substantially. Further, other income sources (including transfer payments) and many regular household expenditures (e.g., rent) tend to arrive near the beginning of the calendar month. If alignment of SNAP benefit receipt with the start of the calendar month reduces expenditure smoothing, there may be important health and behavioral impacts for the SNAP population. Our main result is that increasing the median SNAP distribution date in a state by 10 days reduces the standard deviation of weekly spending by 0.23 dollars or approximately 6.5% for SNAP-eligible households. Results are robust to alternative measures of SNAP date distribution and expenditure smoothing.

Introduction

The majority of businesses pay workers either biweekly or semimonthly, frequently aligning with the beginning of a month (Burgess, 2014)¹. Further, most government payments (e.g., Temporary Assistance for Needy Families (TANF)) occur on or near the beginning of each calendar month². Many regular monthly payment (e.g., rent, utilities) due dates also cluster near the beginning of the month. As a result, there is a monthly “income availability” cycle that is particularly salient for low-income households.

While the Supplemental Nutrition Assistance Program (SNAP) is a national program (formerly known as the Food Stamps program), it is run by the states and they are left to determine the distribution schedule for its participants. While each SNAP household only receives benefits on a single day in a month, states may choose to spread the distribution of benefits to their population of recipients throughout the month. This creates a range of different SNAP benefit cycles within and across states (which is also changing over time). Some households receive benefits at the beginning of the month, while in others they may arrive in the middle or, in recent years, even near the end of the month. While SNAP benefits are restricted in terms of what they can be used to purchase, economic theory predicts there should be fungibility between SNAP benefits and cash, although empirical works suggests that the degree of fungibility differs substantially across households (Smith et al., 2016).

According to the Permanent Income Hypothesis, the timing of anticipated income should not matter as forward-looking households will engage in optimal saving behavior and choose to smooth their consumption spending over different intervals of time (e.g., month, year, etc.). This suggests that the timing of anticipated income or transfers should not affect consumption. Yet, there is ample evidence that the timing of transfers (Stephens, 2003) and the timing of income (Stephens, 2006) alter household purchasing patterns, particularly among income-constrained households. So, while theory suggests households should “smooth” their consumption, many households do not, especially those of limited means.

This is particularly salient for “high-frequency” expenditure cycles (e.g., monthly or weekly purchases of non-durable goods, such as food) because a lack of smoothing expenditure could lead to increased risk of food insecurity as a household moves later into its benefit cycle. Indeed, research that addresses high-frequency consumption patterns has shown that caloric consumption declines notably over the calendar month for income constrained households (Wilde and Ranney, 2000, Todd, 2015) and that households seek out other food sources (e.g., food banks) later in the benefit cycle (Byrne and Just, 2021). Shapiro (2005), for example, finds a 10–15% decline in overall consumption in SNAP households during the monthly SNAP benefits cycle, while some households exhibit even larger declines even after the adoption of electronic benefit cards (Todd, 2015). Further, households are more likely to be food-insecure near the end of the benefit cycle (Gregory and Smith, 2019) and households are more likely to have a day with no food consumption later in the benefit month (Hamrick and Andrews, 2016). Further, the degree of cyclicality in the consumption differs substantially across households (Dorfman et al., 2019). In these households with the most pronounced cycles, the effects could be much larger.

Taken together, these results suggest that the alignment of the timing of SNAP benefits and the calendar month might have real effects on the ability of the household to effectively smooth consumption and impact health or behaviors. For example, the arrival of TANF at the beginning of a calendar month and SNAP in the middle of the month might induce a greater degree of income smoothing than is experienced by households who receive both TANF and SNAP at the beginning of the month, and hence, must manage spending more carefully. Indeed, recent research has found suggestive evidence that when in the calendar month households receive SNAP benefits affects healthcare utilization (Arteaga et al., 2018, Cotti et al., 2020), drug overdoses (Allen and Atwood, 2019), and domestic violence (Carr and Packham, 2021). Further, the timing of benefits relative to when bills are received affects the likelihood they are paid (Barrage et al., 2019). While all these papers suggest that the alignment of benefits with the beginning of the calendar month have important implication, there is scant evidence of an effect on actual consumption and/or expenditures. Beatty et al. (2019) look specifically at the timing of other income sources on the SNAP cycle and find some suggestive evidence that it is more pronounced among those paid on a weekly or monthly basis, but no evidence that the arrival of other income alters the SNAP cycle.

In this project, rather than try to understand the effect of the benefit cycle on expenditure smoothing, we seek out to identify whether differences in state distribution schedules alter household expenditure smoothing for households that receive benefits later in the month rather than earlier. To do this we use detailed scanner data on all shopping purchases recorded in the Nielsen Consumer Panel to create average daily spending for each week in each month, i.e., average spending per day in the first 7 days, days 8–14, etc. Our primary measures of smoothing used in our regression analysis are the difference in the week-to-month average daily expenditure ratio between the first week and last week in a month and the standard deviation of household weekly spending. Unfortunately, we do not know when a household receives their SNAP benefit, so instead, we use the median day of benefit receipt for that state in that year as our main treatment variable, but also the percentage of distribution days that occur after the first week of the calendar month. Our main result is that increasing the distribution of SNAP benefits throughout the calendar mouth (as represented by increasing the median SNAP distribution date in a state or the percent of distribution days after the first week of the month) reduces the standard deviation of weekly spending and the difference in the week-to-month spending ratio. For example, we find that a 10-day increase in the median SNAP day is estimated to reduce the standard deviation of weekly spending by approximately 7%.

This work is most similar to the previously mentioned paper by Beatty et al. (2019), which researchers the degree to which having other sources of income alters the SNAP cycle for food purchases using detailed data on household income, SNAP receipt, and a combination of scanner data and food diaries from the FoodAPS data. This dataset has the advantage of precise information about receipt date and household characteristics, but it also has some limitations. For example, while FoodAPS oversamples SNAP recipients, there are still <1400 SNAP households in the data, which limits its power. Further, households only record food purchases for the interview week, and Kirlin and Denbaly (2017) indicate that there are some technical problems in the implementation of the scanning devices, which affected the accuracy of that data. In our paper, we use Nielsen data, which has a much larger scale in both the number of households and length of time as well as detailed data on purchases. However, it does not include information about when in a month a household receives benefits beyond the state the household resides within. Overall, we aim to provide further evidence to this line of research.

Given the observed consequences of poor smoothing among low-income households, this investigation has important policy implications. In particular, altering the alignment of benefit transfers could help reduce food insecurity and improve health outcomes, without increasing actual public expenditures.