Statement of Work.docx

UNITED STATES DEPARTMENT OF AGRICULTURE

ECONOMIC RESEARCH SERVICE

STATEMENT OF WORK FOR

EXTENDING THE FOOD-AT-HOME MONTHLY AREA PRICES (F-MAP) DATASET

I. OBJECTIVE:

This study will extend the previously estimated Food-at-Home Monthly Area Prices (F-MAP) data set for 2023, 2024, and 2025. This project responds to the intense interest in access to healthy and affordable food in the United States. Specifically, it will (1) estimate weights and statistical replicate weights for retail scanner data for 2023 and 2024, (2) map UPCs in retail scanner data to 90 ERS Food Purchase Groups (EFPG), and (3) estimate panel price indexes for the 90 EFPGs in 10 metropolitan areas, 4 Census regions and the United States. Additionally, the results of this and previous research about the food price environment (4) will be disseminated to ERS employees through training, and (5) help ERS staff to estimate the 2025 F-MAP and address inconsistencies in past deliverables.

Proprietary retail scanner data are the backbone of many USDA high-profile and high-impact research products and are used widely across the agency and academic institutions to answer research questions relevant to the USDA mission. For example, research using retail scanner data includes (1) real-time economic surveillance of food prices that are of interest to the administration; (2) estimating the affordability of healthy foods; (3) evaluation of the Food Safety Modernization Act, (4) estimation of the relationship between food choices and food prices. According to a recent ERS report, the number of stores and the sales coverage in the retail scanner data obtained by ERS are underreported in some geographic areas. To overcome this challenge, ERS developed retail scanner weights for 2012-2022 that allow projection of the retail stores included in the data purchased by USDA to the population of stores in the United States. This allows price measures like the F-MAP and other analyses using the retail scanner data to be more representative of the national food retailing environment.

Retail scanner data are also used to construct price indices that vary geographically to understand the role of the local food price environment and food availability on Americans’ healthy food choice decisions. For example, researchers have studied how local food access and pricing affect household behavior, how regional variation in prices affects the value of the WIC fruit and vegetable voucher, and how geographic variation in food prices affects healthy food choices for participants of USDA food assistance programs. The 2012-2022 F-MAP incorporates these newly developed retail weights and are more representative of the U.S. food market and of local food environments. These improved data products will allow USDA to calculate more accurate food basket costs, conduct a broader range of studies using scanner data, and improve measures of the food price environment using the most recently available data.

II. APPROACH:

ERS intends to award a contract to an organization (hereafter the “contractor”) with extensive survey and data management experience who will research, develop, and evaluate methods; generate computer programs; calculate weights; and estimate price environment variables. The work comprises the following tasks:

· Task 1. The contractor will estimate the 2023 F-MAP data set. This includes several parts: (a) developing weights and replicate weights for statistical inference for 2023 retail scanner data that will allow sales volume and quantities reported by participating retailers to be projected to the total stores within the United States; (b) developing a crosswalk between the UPCs in the 2023 retail scanner data and the EFPGs; and (c) estimating monthly weighted and unweighted total quantities, total sales, mean unit values (and standard errors) and rolling-window GEKS price indices over select areas.

Part a. The Circana OmniMarket Core Outlet retail scanner data that ERS purchases does not contain weights that can be used to inflate sales volumes and quantities reported by participating stores to the population of stores in the United States. The unavailability of weights limits the representativeness of the retail scanner data in analyses, including calculating store-level food basket unit values and price indices, since only a subset of food stores agrees to release their data. In the past, these retailer scanner weights were constructed using NielsenIQ TDLinx data, a census of U.S. food store sales. ERS will provide the contractor with the “Linker” file that spatially assigns stores in the Circana retail scanner data to stores in the NielsenIQ TDLinx for constructing the weights and the contractor will evaluate whether the match rate is consistent with that of previous years. If the contractor finds the match rate in the 2023 Linker file is significantly less than previous years, the contractor will improve the match rate in the Linker file using a method that can be easily applied to other years of the data (e.g., fuzzy matching or a machine-learning algorithm). ERS will provide the contractor with R code used to construct the weights in previous years. The deliverables for this part of task 1 include (1) a dataset (R and CSV) of weights for all stores in 2023 retail scanner data, extending the series already developed for 2012-2022; (2) R programming code used to calculate the weights; (3) an improved dataset (CSV and R) that links stores in 2023 Circana retail scanner data to stores in the 2023 TDLinx with annotated programming code that is used to improve the match rate; (4) user documentation describing methods, results, and limitations.

Part b. The taxonomy of the UPCs in the Circana OmniMarket Core Outlet retail scanner data does not align well with the Dietary Guidelines for Americans. ERS developed the EFPGs to align with the Dietary Guidelines for Americans, allowing it to be more useful for nutrition and food policy analysis. ERS will provide the contractor SAS code from previous years, which will need to be converted to R, to construct the crosswalk. The deliverables for this part of task 1 include (1) a dataset (CSV and R) that maps UPCs with positive sales in the retail scanner data to EFPGs, extending the series already developed for 2012-2022; (2) annotated R programming code used to assign the UPCs to the EFPGs; (3) user documentation describing the methods, results and any major changes from the previous years.

Part c. The F-MAP is a suite of weighted and unweighted price, quantity and expenditure measures based on the store scanner data that is available to the public. The updated F-MAP will be estimated with the 2023 Circana OmniMarket Core Outlet store scanner data that ERS purchases; be on a monthly basis; use the 90 EFPGs to categorize foods; include weighted and unweighted total quantities, total sales, mean unit values (and standard errors) and rolling-window GEKS price indices; be calculated for 10 metropolitan areas, 4 Census regions and nationally as in the current F-MAP; and cover 2023. Methods of eliminating price outliers and converting count data and liquid measures to weights consistent with the prior versions of the F-MAP will be applied. ERS will provide the contractor SAS code from previous years, which will need to be converted to R, to construct F-MAP. The deliverables for this part of task 1 include (1) a dataset (R and CSV) of mean unit values and their distribution, total quantities, total sales and price indices by EFPG, time period, and geography, (2) Excel spreadsheets suitable for web posting, and (3) user documentation describing the methods, results, and limitations. The contractor will work with ERS to prepare data files and documentation suitable for posting materials on the web.

· Task 2. The contractor will provide training and technical guidance to ERS staff on calculating weights (i.e., conducting imputation and control totals, and developing weights) and using weights in analysis, performing mapping, and calculating price indexes. The contractor will transmit their practical understanding of completing Task 1 through virtual trainings. In addition, the contractor will provide statistical knowledge and expertise regarding questions on previous and future deliverables to improve the overall data product. The deliverables for this task are (1) virtual sessions available to current and future ERS staff to disseminate and discuss documentation and code; and (2) written responses and/or ad hoc virtual meetings to address questions related to current or past deliverables.

· Optional Task 3. The contractor will develop the 2024 F-MAP by conducting task 1 with the 2024 Circana retail scanner data and NielsenIQ TDLinx. The deliverables for this task include code, user documentation and datasets outlined in task 1.

· Optional Task 4. The contractor will assist ERS staff in conducting task 1 for producing the 2025 F-MAP. Ultimately, ERS staff will produce the F-MAP in-house. ERS staff may need the contractor’s statistical expertise and knowledge to troubleshoot issues in the data or code for the first year of production. The deliverable for this task is written responses or ad hoc virtual meetings to assist ERS in producing the 2025 F-MAP.

III. PERIOD OF PERFORMANCE: Two years from date of award.

IV. DELIVERABLES:

Task 1, part a: Retail scanner data weights, code and documentation/user’s guide

Optional Task 3. 2024 F-MAP

In addition, the contractor will provide quarterly progress reports through the duration of the contract. Teleconferences will be scheduled as needed.

5 | P a ge