10

Food Expenditure Analysis

In phase I, the committee was tasked with planning and implementing an analysis of food expenditures for the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) population using national data. This analysis is designed to provide estimates of the total food expenditures and expenditures on food groups¹ for WIC households to assess the relative contribution of the WIC food packages to their food expenditures. Analysis reported in the phase I report focuses on the level and contributions of at-home and away-from-home food expenditures and the WIC food package to total food expenditures for three kinds of households: (1) households that receive WIC benefits, (2) households that are eligible for WIC but do not participate in the program, and (3) higher-income households that would meet the eligibility criteria (e.g., having a pregnant woman or child in the household under the age of 5 years old) except for income. The analysis relied on recently released data from a national survey of households that was conducted by the U.S. Department of Agriculture’s (USDA’s) Economic Research Service (ERS) (USDA/ERS, 2015a). The analysis also provided insights into household demographics (e.g., presence of pregnant women, infants, and young children), food insecurity, and share of food acquired with WIC benefits (i.e., for households receiving WIC benefits). This chapter summarizes the methods and results of the committee’s phase I food expenditure analysis.

__________________

¹ Expenditures on food groups will be presented in phase II because the data were not available soon enough to complete these analyses. See Chapter 3 for additional detail.

DATA AND METHODS

Description of the Survey Dataset

The National Household Food Acquisition and Purchase Survey (FoodAPS) served as the primary dataset for this analysis, accessed under a third-party agreement with USDA-ERS and through a confidential Web system.² Data were collected from a nationally representative, stratified sample of 4,826 households between April 2012 and January 2013 (USDA/ERS, 2015a,b). The survey design had four target groups defined in terms of Supplemental Nutrition Assistance Program (SNAP) participation and total reported household income. The three target income groups were income less than 100 percent of the federal poverty-to-income ratio (PIR); income greater than or equal to 100 percent and less than 185 percent of the PIR; and income greater or equal to 185 percent of the PIR. Sampled households were selected through a multi-stage sample design. Prior to release, several quality controls were exercised, including verification of reported SNAP participation by linking responses on SNAP participation and purchases to SNAP administrative records. Information on SNAP participation status (as revised in the match to administrative data) was used in constructing the final sampling weights. Each household was given a final sampling weight to be used in making the sample nationally representative of all non-institutionalized households in the contiguous United States, and the analyses and standard errors account for the complex sampling design.

With respect to food purchases and acquisitions, the survey is a unique source of information on foods eaten both at home (FAH) and away from home (FAFH), as well as extensive information on the sample households. Purchase and acquisition data for each household were collected over a 7-day period. The data provide information on quantities; prices; expenditures for all foods eaten both at home and away from home and purchased or acquired from all sources (including large and small grocery stores, mass merchandisers, convenience stores, gas stations, and food marts); and source of payment for foods consumed at home (i.e., WIC voucher or other method of purchase). During a shopping event where more than one type of tender was used to purchase foods, although it is impossible to determine exactly which foods were obtained with the WIC voucher, it is possible to determine how much the WIC voucher contributed to the total purchase cost and what food items were purchased during the shopping event.

In addition to food purchase and acquisition data, the survey collected

__________________

² The data survey and access procedures are available through USDA-ERS at http://www.ers.usda.gov/data-products/foodaps-national-household-food-acquisition-and-purchase-survey.aspx (accessed August 15, 2015).

information on income, food security status (a scale based on responses to 10 questions used to assess household food security; USDA/ERS, 2015b), and reported WIC participation as well as some demographic information (e.g., presence of a pregnant woman, infant, or young child) that were used here to differentiate among the three types of households. Of note, households were not asked about breastfeeding or postpartum women. Forty-eight percent of the WIC households also participated in SNAP (USDA/ERS, 2015c).

Application of the FoodAPS Dataset for This Report

The household served as the unit of analysis for this study. For example, if a WIC-participating mother and her infant lived in a household with the mother’s (non-WIC-participating) parents, the entire household (mother, infant, and mother’s parents) was coded as “WIC,” not just the mother and her infant. As mentioned previously, three types of households were analyzed for this report: (1) WIC-participating households (i.e., households reporting either having a member participate in the WIC program or a purchase event with use of WIC voucher), (2) eligible non-WIC households (i.e., households with a pregnant woman or a child less than 5 years old and with income ≤ 185 percent PIR), and (3) non-eligible non-WIC households (i.e., households having a pregnant woman or a child younger than 5 years old and with income greater than 185 percent PIR). Although the survey did not cover breastfeeding or postpartum women, it is likely that these women were captured as part of WIC-participating households with infants.

All household members were asked to track and report their food purchases or acquisitions during the survey week, including all foods eaten at home and away from home. Each purchase or acquisition was considered a separate “event.” The sum of events across food purchased and acquired for at-home and away-from-home use constitutes the total food expenditures for that household (for the survey week). Some households (4 percent of WIC participating households, 2 percent of eligible, non-WIC-participating households, and 3 percent of higher-income households) reported no food purchases or acquisitions (FAH or FAFH) during the survey week. Nonetheless, they were included in the analysis to generate a representative “average” amount of food expenditures for all households, because, over the course of 1 month, households vary widely in the frequency and size of their food purchases and acquisitions. Among all households examined in the FoodAPS, 7 to 9 percent reported no FAH expenditures in the interview week (including 8.7 percent of WIC households).

With respect to the 10 FoodAPS questions related to food insecurity, the sum of affirmative responses (“yes,” “often,” “sometimes,” and “occurring 3 or more days” were all coded as affirmative) on questions related

to food insecure conditions were used to assign USDA 30-day Adult Food Security Scale values. A raw score of 0 (none of the 10 questions eliciting an affirmative response) was assigned a value of 1 (high food security), a raw score of 1–2 (1–2 affirmative responses) was assigned a value of 2 (marginal food security), a raw score of 3–5 was assigned a value of 3 (low food security), and a raw score of 6–10 was assigned a value of 4 (very low food security). For this analysis, households with assigned values of 3 and 4 were identified as food insecure.

Data were weighted in estimates of mean values and standard errors using the household weights; all standard errors account for oversampling and the complex survey design of FoodAPS.³ Mean food expenditures and food insecurity scores were compared using t-tests, and distributions of demographic and food insecurity data were compared using the Pearson chi-square statistic between WIC households and both types of non-WIC households (Rao and Scott, 1984).

RESULTS

Demographic Characteristics of the Survey Households

Compared to both types of non-WIC households, WIC households were more likely to have infants and pregnant woman but less likely to have young children (children ages 1 to less than 5 years old) (see Table 10-1). The latter finding is consistent with program data that find a decline in participation of children after the child’s first year (USDA/FNS, 2013). The WIC households did not differ from non-WIC but eligible households with respect to either participation in SNAP or household size (see Table 10-1).

Thirty-four percent of WIC households were identified as food insecure (see Table 10-1). Although WIC households were more likely to be food insecure than demographically similar but higher income households, WIC households and non-WIC but eligible households reported similar levels of food insecurity (see Table 10-1). These estimates of food insecurity are higher than estimates from other national surveys of food insecurity during the previous 30 days among U.S. households (e.g., estimates from the 2012 Current Population Survey [CPS] and the 2012 National Health Interview Survey [NHIS]) (see also USDA/ERS, 2015c). This may be caused by the food security questions being administered differently in the FoodAPS than

__________________

³ Sampling weights were constructed based on the FoodAPS survey stratification of households with the survey target groups determined by SNAP receipt and poverty status, and used to produce estimates that are nationally representative of U.S. households. To apply sampling weights, the committee used the svyset command in STATA (a data analysis and statistical software package).

TABLE 10-1 Households Examined in the Food Expenditure Analysis: Household (HH) Composition and Food Security Scores for WIC and Other Households, FoodAPS

NOTES: FoodAPS = Food Acquisition and Purchase Survey; HH = household; SE = standard error; SNAP = Supplemental Nutrition Assistance Program. Subgroup definitions are as follows:

^a All HH reporting participation in WIC regardless of income level.

^b Low-income HH (≤ 185 percent of the PIR) that did not report participation in WIC.

^c Higher-income HH (> 185 percent of the PIR) that did not report participation in WIC.

^d Significantly different from the WIC households (p < 0.01). Levels of significance (tested between WIC HH and eligible non-WIC HH or higher-income HH) by t-test (for mean raw food insecurity scores) or by Pearson chi-squared (for household characteristics and the food security categories), and the Type I error rate was adjusted to account for multiplicity.

^e Numbers represent the sum of raw scores. Lower numbers represent higher food security.

SOURCE: USDA Economic Research Service, National Household Food Acquisition and Purchase Survey (FoodAPS), data collected April 2012–January 2013 (USDA/ERS, 2015a). Population weights were applied.

in other surveys, particularly the CPS and the NHIS. In the FoodAPS, the period of reference was the 30-day period directly preceding the interview day. In the CPS, in contrast, households were first asked whether they experienced food-insecure conditions in the prior 12 months and only afterward, based on an initial affirmation of any food-insecure condition, were they asked about the past 30 days. The NHIS used a 30-day reference period but the survey was administered throughout the year, unlike the FoodAPS, which was administered between April and January. Also unlike the CPS and the NHIS, the FoodAPS was administered in the context of each household providing a detailed record of information on food purchases and acquisitions that is not collected in the other surveys (USDA/ERS, 2015b). (See also the section on food insecurity in Chapter 9.)

Food Expenditures

WIC households spent, on average, $184.80 per week on total food expenditures, mostly for FAH (see Table 10-2). Although the total food expenditures for WIC households were higher than those of eligible households that did not receive WIC benefits (significant at p < 0.1), differences in FAH and FAFH expenditures were not statistically significant. The higher-income households with similar demographic compositions (i.e., having a pregnant woman or child younger than 5 years old) spent more on total food, FAH, and FAFH, compared with the WIC households.

Nearly one-third of WIC households redeemed their WIC benefits to acquire food during the reporting week.⁴ At the time of the FoodAPS, most WIC households had benefits provided in the form of paper vouchers. Some WIC products are provided in relatively large sizes or in forms (gallons of milk, dozen eggs, or 36 oz of ready-to-eat cereals) that could last for more than 1 week. Therefore, it is not expected that all WIC households would redeem some vouchers every week. Across all WIC households, the value of WIC benefits used was $10.80 per week, on average, representing almost 9 percent of FAH expenditures. Among the nearly one-third (32.3 percent) of WIC households using WIC benefits for purchases during the interview week, the average value of acquisitions made using WIC vouchers was $33.30 and represented 24 percent of FAH expenditures.

__________________

⁴ Data-collection weeks for FoodAPS were distributed across each month based on when the household was determined eligible for the survey and when the initial interview could be scheduled. For food assistance programs like WIC and SNAP, data collection may have occurred up to 3 weeks after benefits were distributed.

TABLE 10-2 Weekly Food Expenditures for WIC and Other Households (HH) in FoodAPS

NOTES: FoodAPS = Food Acquisition and Purchase Survey; HH = households; PIR = poverty-to-income ratio; SE = standard error; SNAP = Supplemental Nutrition Assistance Program. Subgroup definitions are as follows:

^a All HH reporting participation in WIC regardless of income level.

^b Low-income HH (≤ 185 percent of the PIR) that did not report participation in WIC.

^c Higher-income HH (> 185 percent of the PIR) that did not report participation in WIC.

^d Significantly different from the WIC households (p < 0.01).

^e Significantly different from the WIC households (p < 0.1) Levels of significance (tested between WIC HH and eligible non-WIC HH or higher-income HH) by t-test, and the Type 1 error rate was adjusted to account for multiplicity.

SOURCE: USDA Economic Research Service, National Household Food Acquisition and Purchase Survey (FoodAPS), data collected April 2012–January 2013 (USDA/ERS, 2015a). Population weights were applied.

SUMMARY AND LIMITATIONS

FoodAPS data provide a unique source of information on food expenditure patterns of U.S. households, including households that participate in the WIC program. The strengths and unique features of the data include being a nationally representative survey, sampling to represent SNAP participants and other households in three income groups, and having a sample with SNAP participation and expenditures verified through administrative records. The results of the committee’s analysis show that, in any week of the month, an important share of WIC households redeem their WIC vouchers and, for these households, the value of the WIC foods is relatively high—almost one-fourth of the value of the foods acquired for home use.

There are a couple of noteworthy limitations to the data. The number of households surveyed is relatively limited (4,826 households). Although nationally representative, the FoodAPS relies primarily on self-reported participation in the WIC program to establish program participation. There were some households that redeemed WIC vouchers but did not report that they were currently participating in WIC. This difference may be due in part to lags in enrollment and benefit issuance. Also, as would be expected and as previously mentioned, some households had no food expenditures or no food expenditures for food at home during the survey week. The committee assumed that these households were similar to other households and that their lack of expenditures for that week reflected weekly variation in food purchase and acquisition patterns. That is, households with no food purchases captured in the survey week have purchased food at a different time in the month that WIC foods were prescribed.

REFERENCES

Rao, J. N. K., and A. J. Scott. 1984. On Chi-squared tests for multiway contingency tables with cell proportions estimated from survey data. Annals of Statistics 12(1):46-60.

USDA/ERS (U.S. Department of Agriculture/Economic Research Service). 2015a. FoodAPS National Household Food Acquisition and Purchase Survey. Washington, DC: USDA/ERS. http://www.ers.usda.gov/data-products/foodaps-national-household-food-acquisitionand-purchase-survey.aspx (accessed August 26, 2015).

USDA/ERS. 2015b. National Household Food Acquisition and Purchase Survey. User’s guide to survey design, data collection, and datasets. http://www.ers.usda.gov/datafiles/FoodAPS_National_Household_Food_Acquisition_and_Purchase_Survey/Codebooks/UserGuideFoodAPS.pdf (accessed October 22, 2015).

USDA/ERS. 2015c. Where do Americans usually shop for food and how do they travel to get there? Initial findings from the National Household Food Acquisition and Purchase Survey. Washington, DC: USDA/ERS. http://www.ers.usda.gov/media/1807325/eib138.pdf (accessed March 25, 2015).

USDA/FNS (U.S. Department of Agriculture/Food and Nutrition Service). 2013. WIC participant and program characteristics 2012 final report. Alexandria, VA: USDA/FNS. http://www.fns.usda.gov/sites/default/files/WICPC2012.pdf (accessed December 20, 2014).