Appendix R

Development of the Food Package Nutrient and Cost Profiles

To develop the nutrient and cost profiles of the food packages that are evaluated in this report, the committee created a series of linked spreadsheets representing the current and revised sets of food packages. These spreadsheets allowed the committee to determine the average per-participant cost difference between the current and revised packages to meet the charge of cost-neutrality. The basis for these profiles was a set of detailed assumptions based on the U.S. Department of Agriculture’s Food and Nutrition Service (USDA-FNS) price and redemption dataset, the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) Food Package Report series, and information provided by individual states. The assumptions applied to develop composites for WIC food categories included ratios of substitution options based on available data or a conservative assumption. Assumptions also differed for some WIC food categories depending on the food package. All details are included in the tables that follow. Details related to costs at the program level are provided in Chapter 10, “The Regulatory Impact Analysis.” Spreadsheets generated for this report were reviewed by the committee members, as well as internally cross-checked by staff. The regulatory impact analyses used the revised package data, with the base year of 2015 as the starting point, and therefore re-created the revised package cost profiles and compared them to the profiles that were created for the cost-neutral assessment.

TABLES

TABLE R-1 Cost Data and Assumptions Used in Analyses of Current and Revised Food Packages for Women and Children

TABLE R-2 Cost Data and Assumptions Used in Analyses of Current and Revised Food Packages for Infants

TABLE R-3 Assumptions Used in Nutrient Analyses of Current and Revised Food Packages for Infants

TABLE R-4 Assumptions Used in Nutrient Analyses of Current and Revised Food Packages for Women and Children

TABLE R-5 Redemption Rates Applied to Generate the Set of Cost-Neutral Food Packages

TABLE R-1 Cost Data and Assumptions Used in Analyses of Current and Revised Food Packages for Women and Children

NOTES: c-eq = cup equivalent; CPI = Consumer Price Index; CVV = cash value voucher; ERS = USDA’s Economic Research Service; FNS = USDA’s Food and Nutrition Service; FP = food package; IRI = Information Resources, Inc., Consumer Network data 2014; PC = WIC Participant and Program Characteristics Report. Food costs were found to be comparable to those reported in the fiscal year (FY) 2010 WIC Food Cost Report (USDA/FNS, 2013).

* FNS redemption data: The average prices of WIC foods redeemed in six states (Personal communication, K. Castellanos-Brown, USDA-FNS, April 7, 2016, and June 30, 2016) were adjusted for inflation to an average price for FY2015. The six-state prices provided were an average over the fiscal year August 2013 to July 2014. It was therefore assumed that the six-state average price was equivalent to the price for January 2014, the approximate midpoint of the August 2013 to July 2014 12-month period. This assumption is approximately equal to assuming the average of the nominal prices for the August to July 12-month period is equal to the price from redemption data. The average January 2014 price was then inflated to the average price for fiscal year 2015 using U.S. Bureau of Labor Statistics (BLS) CPI adjustments (see http://data.bls.gov/cgi-bin/dsrv?cu [accessed December 22, 2016]). This procedure was carried out using the FNS average prices for juice, milk, cheese, breakfast cereal, eggs, whole wheat bread and alternatives, canned fish, legumes, peanut butter, jarred infant fruits and vegetables, jarred infant food meat, and infant cereal. CPI inflation factors were matched to the closest BLS food category for each item.

TABLE R-2 Cost Data and Assumptions Used in Analyses of Current and Revised Food Packages for Infants

NOTES: CPI = Consumer Price Index; FP = food package; PC = WIC Participant and Program Characteristics Report. Values were multiplied by a redemption factor, based on FNS redemption data for current packages and projected values for revised packages (see Table R-5). Food costs based on redemption were found to be comparable to those reported in the FY 2010 WIC Food Cost Report (USDA/FNS, 2013). Note that the maximum amount of formula was assumed for the purpose of these analyses, although USDA/FNS supports provision of the least amount of formula possible or that is needed.

* FNS redemption data: The six-state average prices (Personal communication, K. Castellanos-Brown, USDA-FNS, April 7, 2016, and June 30, 2016) were adjusted for inflation to an average price for FY2015. The six-state prices provided were an average over the fiscal year August 2013 to July 2014. It was therefore assumed that the six-state average price was equivalent to the price for January 2014, the approximate midpoint of the August 2013 to July 2014 12-month period. This assumption is approximately equal to assuming the average of the nominal prices for the August to July 12-month period is equal to the price from redemption data. The average January 2014 price was then inflated to the average price for fiscal year 2015 using U.S. Bureau of Labor Statistics (BLS) CPI adjustments (see http://data.bls.gov/cgi-bin/dsrv?cu [accessed March 10, 2017]). CPI inflation factors were matched to the closest BLS food category for each item.

TABLE R-3 Assumptions Used in Nutrient Analyses of Current and Revised Food Packages for Infants

NOTES: FNB = full nutrition benefit; NDB = National Nutrient Database for Standard Reference, Release 28 (USDA/ARS, 2016); PC = WIC Participant and Program Characteristics Report.

* State data were incorporated into weighting only when the committee judged the data to be reasonable and complete.

TABLE R-4 Assumptions Used in Nutrient Analyses of Current and Revised Food Packages for Women and Children

NOTES: CVV = cash value voucher; FP = food package; NDB = National Nutrient Database for Standard Reference, Release 28 (USDA/ARS, 2016); PC = WIC Participant and Program Characteristics Report.

^a State data were incorporated into weighting only when the committee judged the data to be reasonable and complete.

^b Weighting for milk and milk alternatives for each food package do not total to 1 due to the 1:3 substitution of cheese for milk.

DESCRIPTION OF METHOD FOR CALCULATING REDEMPTION RATES

Two sets of redemption rates were applied in the cost analysis: rates for the current set of food packages and rates for the revised set of food packages. The primary data source for the current package redemption rates was data provided to the committee by USDA-FNS (herein referenced as the FNS redemption data) (personal communication, K. Castellanos-Brown, USDA-FNS, April 7, 2016 and June 30, 2016). The FNS redemption data included 12 months (August 2013 through July 2014) of price and redemption data from a convenience sample of six WIC state agencies, representing five of the seven regions of the country. The identity of the agencies was not known to the committee.¹ Redemption rates were available for the following WIC foods: juice, breakfast cereal, whole grains, CVV, eggs, legumes/peanut butter, and fish. For foods for which redemption data were not available in the FNS redemption dataset, redemption rates were calculated as the average (unweighted) from several states that provided the committee with information (see Chapter 2, Table 2-14). Together, these sources resulted in a consensus list of redemption rates that were applied to develop the current food package nutrient and cost profiles.

Redemption rates for milk were a special case. The FNS redemption dataset did not include a separate redemption rate for whole milk. Considering that redemption might be different for milk in food package IV-A (because of the different fat level of the product, and the different age group of children), the average redemption rate from three states of 75 percent was applied to the current food package IV-A. For low-fat milk, the average redemption rate from the FNS redemption data was 65 percent. This represented average redemption of low-fat milk for women and children, combined. The difference in redemption rates for whole milk and low-fat milk, in combination with the committee’s data on dairy intake of children and women, suggested that redemption rates may be higher for children ages 2 to less than 5 years, compared to women. The group average rate of 65 percent redemption for low-fat milk was decomposed into values of 71 percent for children ages 2 to less than 5 and 56 percent for women, rates that were consistent with the observed 65 percent redemption when

__________________

¹ To keep maintain anonymity of the state agencies, USDA-FNS inputted average monthly participation by participant category for each state into a spreadsheet containing the redemption equations created by the committee, and returned the overall unweighted average redemption across the state agencies per food package item. Through this process, USDA-FNS identified one of the six states as a clear outlier, and removed it from the averages (personal communication, K. Castellanos-Brown, USDA-FNS, June 22, 2016). As such, redemption estimates represent five of the six state agencies included in the FNS redemption dataset. Sensitivity analysis (see Chapter 8) explores the effect of shifting specific redemption rates up or down.

weighted by the population proportion of WIC-participating children and women.

Accounting for the Distribution of Redemption Rates

The redemption rates for WIC foods in the revised set of food packages are based on a distribution of redemption practices among WIC participants that ranged from no redemption of a food to partial redemption to full redemption of the package allotment. In 2012, Altarum conducted a study of redemption rates in three states (Kentucky, Michigan, and Nevada) (USDA/ERS, 2014) that provided basic information on the redemption distribution. All three states had implemented EBT issuance systems. The study classified redemption practices in three groups: full redemption, partial redemption and nonredemption. As stated in the Altarum report:

If the amount remaining was greater than the approved minimum amount but less than the amount issued, this was considered partial redemption. If none of the amount issued was redeemed, this was nonredemption. The redemption amounts were calculated for more than 14 individual food categories, for each of the food categories, and over all of the categories. Overall, 12.6 percent of the families receiving WIC benefits redeemed all of their benefits.

The information on redemption rates by product category from Altarum’s report can also be used to develop data-based assumptions for redemption behaviors, specifically those used to guide committee choices of redemption rates to apply for the revised packages. The committee used a combination of the Altarum and FNS data to develop a range of revised redemption rates, and the final rates used were close to the implied new FNS-based rates with some adjustments.

WIC-participating families that did not redeem a food category in the current food package (the nonredemption group) were assumed not to redeem under the revised package if there was no change to the food category or if the only change was to change the amount offered. This seems like a plausible assumption given that there was relatively little change in the type of foods offered for each food group. If WIC-participating families fully redeemed foods in the category (the full redemption group), the families were assumed to continue to fully redeem the benefits in the food category with the revised packages as long as the amount offered was less than in the

current package. This was also the case if the amount of the food provided increased, as in the case of the CVV, the full redemption rate was assumed to stay approximately unchanged. The CVV is the main category with a large increase in the amount of food provided. If there were partial redemption in the food category, the committee compared the revised amount offered in the package with the average partial redemption amount of the food before the change. (The average partial redemption amount redeemed can be derived from the redemption rates, the full amount offered, and the total average amount redeemed.) If the revised package amount was less than the average partial redemption amount, the partial redeemers were treated as were the full redeemers and assumed to fully redeem the revised package. If the revised amount was more than the average partial redemption amount, then a new implied average redemption rate consistent with the other numbers was calculated. The degree to which partial redemption affects the projected redemption rates can be derived from the available information.

Application of the Altarum Redemption Distributions (None, Partial, Full): Assumptions

To apply the Altarum redemption distribution information to estimate the revised redemption rates, several assumptions were applied, as listed below:

• The redemption rate patterns follow the distributions—full, partial, and none—observed in the Altarum study.
• The redemption patterns are constant and independent of the overall percentage redemption observed. That is, the Altarum distribution of rates of no, partial, and full redemption was applied to all redemption rates assumed for each food item, including the redemption rates based on the FNS redemption data provided to the committee as well as the observed state average rates included in the current food package consensus rates.²

__________________

² Note that the ideal data for this task would be nationally-representative redemption distributions across food items by package type. These data were not available. Although the Altarum redemption data and distribution of full, partial, and no-redemption are the best available, there are several limitations that should be noted. First, the Altarum report indicates the share of full, partial, and nonredeemers, but these percentages are not adequate to back-calculate anything other than the average amount redeemed by the partial redeemers without further more complex assumptions. Second, although there may be differences in the redemption shares (full, partial, and nonredeemers) by group across locations (e.g., states), the committee assumes that the shares are constant across locations. Finally, the committee assumes that the partial redemption average amount calculated from the Altarum study is consistent with the overall partial redemption mean and can be applied to overall redemption rates based on the FNS average redemption rates and state average rates.

• Implied adjustments occur because of changes in quantities only; adjustment for changes attributable to new products or available substitutions are made separately from the adjustments to the quantities.

The committee was interested in obtaining estimates for the implied new redemption rate (Redemption%New) for each food item. The relationship observed from Altarum is:

Where:

In this equation, all of the numbers are known from the data except PartialAmtAlt. Solving for PartialAmtAlt (the implied average partial redemption amount in Altarum):

Rearranging terms:

Similarly, the committee knows the overall average redemption rates in the FNS (or state average) data and can combine it with the no, full, and partial rates from the Altarum data to calculate the implied average partial redemption amounts assumed for the FNS data as well (PartialAmtAssumed). As the same redemption shares are assumed to hold for different overall redemption levels across all locations, the value

Redemption%Assumed can be substituted in the below calculation for the overall redemption rate observed in the Altarum study (Redemption%Alt) to solve for PartialAmtAssumed if this leads to nonnegative amounts. In this case, replacing Redemption%Alt, the reported overall Altarum redemption rate, with Redemption%Assumed, the FNS reported (consensus) redemption rate, yields:

As long as the PartialAmtAssumed, our calculated assumed partial redemption amount on average is positive, this calculation is not inconsistent with observed patterns and yet incorporates the distribution of redemptions rates across full, no, and partial rates.

Now when the implied partial amount assumed is nonnegative, we use this calculation of the assumed Partial Amount (PartialAmtAssumed) to adjust our overall redemption rate for the revised package. Then we consider a further adjustment to our revised rates to account for whether the distribution of full and partial redeemers (based on the mean) is above the newly projected amount in the new package. If the revised amount is below the current amount and the current average redeemed partial amount, we assume the new redemption rate is 1 minus the share that has no redemption (that is, all participants except the nonredemption group will redeem the revised full amount). However, if the revised full amount is more than the current amount redeemed by partial redeemers, then the redemption amount is down-weighted to incorporate their lower levels.

Following this procedure generates implied new redemption rates based on the changes between the current and revised food packages (see Table R-5). These implied rates do not account for behavioral changes, such as those resulting from the offering of new substitutions. Slight adjustments were made (see the “Revised” column, Table R-5) to account for these changes.

TABLE R-5 Redemption Rates Applied to Generate the Set of Cost-Neutral Food Packages

NOTES: BF = breastfeeding; CVV = cash value voucher; FP = food package; veg/fr = vegetables and fruits; NA = not applicable; P = pregnant; PP = postpartum; RTE = ready-to-eat.

^a These values were applied to estimate the costs of the current set of food packages. Values represent the unweighted mean values from five states provided by FNS, except for whole milk, cheese, and infant foods and formula. For these foods, the unweighted average of available state data was used.

^b These values were generated using the method described in this appendix, and account for the redemption distributions reported in USDA/ERS, 2014. These values account for changes in amounts, but they do not account for changes in composition or substitution options.

^c These values include upward adjustments to account for changes in composition or substitution options. Values represent the anticipated mean redemption over the course of the revised program.

^d USDA-FNS provided the committee with 12 months (August 2013 through July 2014) of price and redemption data from a convenience sample of six WIC state agencies, representing five of the seven regions of the country (herein referenced as the FNS redemption data). The identity of the agencies was not known to the committee. The states were diverse in terms of size and did not include Indian Tribal Organizations or territories. Redemption values from the five states may be slightly underestimated: The PC2014 report indicates that 17.7 percent of minimally breastfeeding women are included in reported participants, but do not receive a food package. FNS indicated that these women are included in administrative data for participation. Because these women are not issued and, therefore, do not redeem WIC foods, the redemption rates may be slightly deflated.

^e Bimodal distribution refers to cases where the tails of the distribution (none and full redemption) had a relatively large share of the redemptions and the “partial redemptions” were a small share (USDA/ERS, 2014).

^f The group average rate of 65 percent redemption for low-fat milk (women and children, combined) was decomposed into values of 71 percent for children ages 2 to less than 5 and 56 percent for women, rates that were consistent with the observed 65 percent redemption when weighted by the population proportion of WIC-participating children and women.

SOURCES: Personal communication, USDA/FNS, June 30, 2016; USDA/ERS, 2014; redemption data provided to the committee by individual states, available in the public access file for the study (Email: paro@nas.edu).

REFERENCES

This page intentionally left blank.