ATTACHMENT 3 FY 2023 QC Tech Doc.pdf

ATTACHMENT 3

April 2025

Technical Documentation for the Fiscal Year 2023 Supplemental Nutrition Assistance Program Quality Control Database and the QC Minimodel

FINAL REPORT

Technical Documentation for the Fiscal Year 2023 Supplemental

Nutrition Assistance Program Quality Control Database and the QC Minimodel

Final Report

April 2025

Submitted to: Submitted by:

U.S. Department of Agriculture Food and Nutrition Service 1320 Braddock Place Alexandria, VA 22314

INTENTIALLY BLANK

Disclaimer

The findings and conclusions in this report are those of the authors and should not be construed to represent any official USDA or U.S. Government determination or policy.

ii

Acknowledgments

INTENTIONALLY BLANK

Table of Contents iii

Contents

Acknowledgments .......................................................................................................................... ii

I. Introduction

A. Key program changes since FY 2022

B. Key changes to the FY 2023 SNAP QC database

1. Differences in the QC sample

2. Changes to the raw data file

3. Federal policy changes affecting the contents of the file

II. Overview of the SNAP QC Database

A. The QC System

B. The raw data file

C. Creation of the SNAP QC database

1. Preliminary processing

2. Data editing

3. Variable construction

4. Weighting

D. Final SNAP QC database

III. FY 2023 SNAP QC File Development Process

A. Developing the SNAP QC file

Step 1. Obtain data

Step 2. Read in and prepare file

Step 3. Conduct quality assurance (QA) review of the data

Step 4. Set SNAP parameters

Step 5. Define metropolitan areas

Step 6. Recode and standardize variables

Step 7. Stratify data

Step 8. Update stratified data

Step 9. Calculate weights

Step 10. Add weights

B. Obtaining file consistency

1. Standard editing procedures iv

2. State variations to editing procedures

C. Derivation of sampling weights

IV. Development of the 2023 QC Minimodel

A. Create MATH-style version of SNAP QC database

1. Introduction

2. User parameters

3. Programmer’s guide

4. Technical description of procedures

B. QC-specific portion of the QC Minimodel

1. Introduction

2. User parameters

3. Programmer’s guide

4. Technical description of procedures

V. Codebook for the FY 2023 SNAP QC Database

A. Overview of variables on the QC file

1. Origin: Reported versus constructed

2. Missing values

3. Using the SNAP QC database

B. Codebook

References

APPENDIX A Assessment of the Quality of the Selected Variables in the FY 2023 SNAP QC Database................................................................................................... A.1

APPENDIX B Automated Edits to SNAP Units ........................................................................ B.1

APPENDIX C New Variables and Variables That Changed in the FY 2023 SNAP QC Database............................................................................................................ C.1

APPENDIX D Derivation of Weights by State and Month ........................................................ D.1

APPENDIX E State and Region Codes ................................................................................... E.1

APPENDIX F FY 2023 SNAP Parameters................................................................................F.1

APPENDIX G Quality Control Review Schedule ..................................................................... G.1 v

Tables

II.1. Number and percentage of cases sampled, dropped from the edited file, and included in the edited file, FY 2023

II.2. FY 2023 weighting targets versus SNAP QC data file weighted totals

II.3. Comparison of program data to edited SNAP QC database, FY 2023

II.4. Averages in program data compared to edited SNAP QC database, FY 2023

III.1. SSI-CAP programs with standard benefits

III.2. States with special rules for identifying, recoding, and calculating benefits for SSI-CAP participants

III.3. SSI-CAP programs with standard shelter expenses

III.4. States with standard medical deduction demonstrations

V.1. Constructed variables that are frequently used in the Characteristics of SNAP Households report series

V.2. Codes for missing data in the restricted use SNAP QC database

V.3. Quick-reference codebook

A.1. Overview of variable recommendations ........................................................................ A.2

D. 1a. Calculated weighted unit counts by State (October 2022 to April 2023) ....................... D.3

D.1b. Calculated weighted unit counts by State (May 2023 to September 2023) and FY average ......................................................................................................................... D.4

D.2a. Calculated weighted individual counts by State (October 2022 to April 2023).............. D.5

D.2b. Calculated weighted individual counts by State (May 2023 to September 2023) and FY average............................................................................................................. D.6

D.3a. Calculated weighted benefit amounts by State (October 2022 to April 2023)............... D.7

D.3b. Calculated weighted benefit amounts by State (May 2023 to September 2023) and FY average............................................................................................................. D.8

D.4a. Adjustments to weighted unit counts by State (October 2022 to April 2023) ................ D.9

D.4b. Adjustments to weighted unit counts by State (May 2023 to September 2023).......... D.10

D.5a. Adjustments to weighted individual counts by State (October 2022 to April 2023) ..... D.11

D.5b. Adjustments to weighted individual counts by State (May 2023 to September 2023) ........................................................................................................................... D.12

D.6a. Adjustments to weighted benefit amounts by State (October 2022 to April 2023) ...... D.13 vi

D. 6b. Adjustments to weighted benefit amounts by State (May 2023 to September

2023) ........................................................................................................................... D.14

D.7. Stratification and weight calculation by State, October 2022 ...................................... D.15

D.8. Stratification and weight calculation by State, November 2022 .................................. D.17

D.9. Stratification and weight calculation by State, December 2022 .................................. D.19

D.10. Stratification and weight calculation by State, January 2023 ...................................... D.21

D.11. Stratification and weight calculation by State, February 2023..................................... D.23

D.12. Stratification and weight calculation by State, March 2023 ......................................... D.25

D.13. Stratification and weight calculation by State, April 2023............................................ D.27

D.14. Stratification and weight calculation by State, May 2023 ............................................ D.29

D.15. Stratification and weight calculation by State, June 2023 ........................................... D.31

D.16. Stratification and weight calculation by State, July 2023 ............................................ D.33

D.17. Stratification and weight calculation by State, August 2023........................................ D.35

D.18. Stratification and weight calculation by State, September 2023 ................................. D.37

E.1. State FIPS codes (STATE) ........................................................................................... E.2

E.2. SNAP region codes (REGIONCD) ................................................................................ E.3

E.3. Census region codes (REGION) ................................................................................... E.4

F.1. SNAP gross income screen, FY 2023............................................................................F.3

F.2. SNAP net income screen, FY 2023 ...............................................................................F.3

F.3. Deduction amounts, FY 2023.........................................................................................F.4

F.4. Standard medical deduction demonstration, FY 2023 ...................................................F.5

F.5. Maximum monthly SNAP benefit, FY 2023 ....................................................................F.6

F.6. Minimum monthly SNAP benefit, FY 2023 .....................................................................F.6

F.7. Standard utility allowances, FY 2023 .............................................................................F.7

F.8. Minnesota Family Investment Program (MFIP) benefits, FY 2023...............................F.10

F.9. Arizona SSI-CAP (AZSNAP) benefit criteria, FY 2023.................................................F.10

F.10. Kentucky SSI-CAP (KYSAFE) benefit criteria, FY 2023 ..............................................F.10

F.11. Louisiana SSI-CAP (LaCAP) benefit criteria, FY 2023.................................................F.11

F.12. Maryland SSI-CAP (MSNAP) benefit criteria, FY 2023................................................F.11

F.13. Michigan SSI-CAP (MiCAP) benefit criteria, FY 2023..................................................F.11 vii

F.14. Mississippi SSI-CAP (MSCAP) benefits by income and shelter expense patterns, FY 2023 ........................................................................................................................ F.12

F.15. New Jersey SSI-CAP (NJ SNAS) benefit criteria, FY 2023 .........................................F.12

F.16. New York SSI-CAP (NYSNIP) benefit criteria, FY 2023 ..............................................F.13

F.17. North Carolina SSI-CAP (NCSNAP) benefit criteria, FY 2023 .....................................F.14

F.18. Pennsylvania SSI-CAP (PACAP) benefit criteria, FY 2023..........................................F.14

F.19. South Carolina SSI-CAP (SCCAP) benefits by income and shelter expense patterns, FY 2023.........................................................................................................F.14

F.20. South Dakota SSI-CAP (SD IN) benefit criteria, FY 2023 ............................................F.15

F.21. Texas SSI-CAP (SNAP-CAP) benefit criteria, FY 2023 ...............................................F.15

F.22. Virginia SSI-CAP (VaCAP) benefit criteria, FY 2023....................................................F.15

F.23. Florida (SUNCAP), Massachusetts (BAY STATE CAP), and Washington SSI- CAP (WASHCAP) shelter allowances, FY 2023 ..........................................................F.15

Figures

III.1. FY 2022 SNAP QC file development process

I. Introduction The Supplemental Nutrition Assistance Program (SNAP) is the largest of the domestic nutrition assistance programs administered by the Food and Nutrition Service (FNS) of the U.S. Department of Agriculture (USDA). SNAP provides millions of Americans with the means to purchase food for a nutritious diet. During fiscal year (FY) 2023, SNAP served an average of 42.2 million people monthly and paid out $107 billion in benefits, including emergency allotments to supplement SNAP benefits during the COVID-19 public health emergency.1

The characteristics of SNAP participants and households and the size of the SNAP caseload change over time in response to changes in program rules as well as economic and demographic trends. To quantify these changes or estimate the effect of adjustments to program rules on the current SNAP caseload, FNS relies on data from the SNAP Quality Control (QC) Database. This database is an edited version of the raw data file of monthly case reviews that are conducted by State SNAP agencies to assess the accuracy of eligibility determinations and benefit calculations for their SNAP caseloads.2

This document describes how the raw data are cleaned and edited to create the SNAP QC database. It also describes how the QC Minimodel—one of FNS’s SNAP microsimulation models—uses the SNAP QC database to simulate the effect of various policy changes to SNAP on current SNAP participants.

This chapter provides a road map to the report and summarizes key program and database changes since

FY 2022.

Chapter II provides an overview of the SNAP QC System, the resulting raw data file, and the creation of the SNAP QC database. The overview is designed to give analysts and new users of the data enough information to be able to analyze and interpret the results of SNAP QC data tabulations and policy change simulations from the QC Minimodel.

Chapter III describes the process for developing files for the SNAP QC database. We discuss the file development programs used to transform the raw data into the SNAP QC database, the algorithms used to edit the data for consistency, and the development of sampling weights.

Chapter IV provides a technical description of the procedures used to transform the FY 2023 SNAP QC database into the format required by the QC Minimodel and to document the QC-specific portions of the QC Minimodel.3

Chapter V contains the codebook for the FY 2023 SNAP QC database and explains how to use the database. For each variable, the codebook lists the variable name, the variable origin (whether it came from the raw data file or was constructed), and a description (including all valid values of the variable).

1 The estimates of 42.2 million participants and $107 billion in benefits were based on FNS administrative records, available at https://www.fns.usda.gov/pd/supplemental-nutrition-assistance-program-snap. They differ from the other estimates in this documentation, which come from the edited SNAP QC Database, because the database is adjusted to exclude ineligible households issued benefits in error and households that received disaster assistance (including COVID-19 emergency allotments).

2 This report refers to the original data file as the raw data file and the edited version as the SNAP QC database.

3 The portions of the QC Minimodel code that apply to all of FNS’s SNAP microsimulation models are documented in the 2011 MATH SIPP+ Microsimulation Model: Programmer’s Guide, Technical Description, and Codebook (Schechter et al. 2014).

https://www.fns.usda.gov/pd/supplemental-nutrition-assistance-program-snap

Chapter I Introduction

Appendix A provides an assessment of the quality of selected variables in the FY 2023 SNAP QC database. Users should read this appendix before using the SNAP QC database. The appendix recommends against the use of some variables and cautions against or provides a disclaimer for the use of others because of apparent miscoding, high prevalence of missing or unknown values, or small sample sizes. Appendix B describes automated edits used to improve the quality of the edited SNAP QC database. Appendix C provides information on new and changed variables in the FY 2023 SNAP QC database. Appendix D shows how the monthly sampling weights were derived. Appendix E lists the State and region identification codes used in the file. Appendix F contains the parameter values used to determine SNAP eligibility in FY 2023, including gross and net income eligibility thresholds, deduction amounts, and maximum benefit amounts. Appendix G presents the QC review schedule—the coding form on which the raw data are originally recorded by the State QC System reviewers.

A. Key program changes since FY 2022 Since the start of the COVID-19 public health emergency in March 2020, several pieces of legislation have contained changes, most temporary, to Federal SNAP rules. The provisions affecting SNAP eligibility rules and participants in FY 2023 are summarized below and discussed in more detail in the Characteristics of Supplemental Nutrition Assistance Program Households: Fiscal Year 2023 report (Monkovic and Ward forthcoming). Comprehensive detail is also available on FNS’s website.4 Three changes most directly affected the SNAP QC database:

1. SNAP emergency allotments. The Families First Coronavirus Response Act (FFCRA) of 2020 authorized emergency supplemental appropriations in response to the COVID-19 public health emergency. Beginning in April 2020, SNAP households receiving less than the maximum SNAP benefit for their household size were eligible for emergency allotments that brought their benefits up to the maximum. Beginning in April 2021, all SNAP households, regardless of benefit level, were eligible for an emergency allotment of at least $95, or the difference between the calculated benefit amount and the maximum if this difference was greater than $95. By the start of FY 2023, 17 States—Alaska, Arizona, Arkansas, Florida, Georgia, Idaho, Indiana, Iowa, Kentucky, Mississippi, Missouri, Montana, Nebraska, North Dakota, South Dakota, Tennessee, and Wyoming—had returned to regular benefit amounts without emergency allotments. South Carolina returned to normal benefit amounts in January 2023. Emergency allotment benefits ended for all remaining States and territories in February 2023, although some issued February 2023 benefits in March 2023. See Section B for information on how SNAP emergency allotments are handled in the database.

2. Time limits on SNAP benefits for adults ages 18–49 without disabilities in childless households. Adults ages 18–49 without disabilities who do not live with a household member younger than age 18 (called able-bodied adults without dependents, or “ABAWDs”) are normally subject to time limits on their participation. The FFCRA temporarily and partially suspended time limits, beginning on April 1, 2020. This suspension continued through the end of June 2023.

4 Food and Nutrition Service. “FNS Documents & Resources.” https://www.fns.usda.gov/resources.

https://www.fns.usda.gov/resources

3. Increase in the upper age limit for adults without disabilities in childless households subject to time limits. The Fiscal Responsibility Act of 2023 (FRA) gradually increased the age at which adults who do not meet certain exceptions, such as having a disability or dependents, could be subject to time-limited SNAP benefits. The upper age limit increased from age 49 to age 50, effective September 1, 2023; age 52, effective October 1, 2023; and age 54, effective October 1, 2024. See Section B for information on how this change is handled in the database.

Some key State policy changes between FY 2022 and FY 2023 included the following:

• Effective October 2022, Connecticut increased the gross income limit of its broad-based categorical eligibility (BBCE) policy from 185 percent to 200 percent of poverty.

• Effective October 2022, Minnesota increased the gross income limit of its BBCE policy from 165 percent to 200 percent of poverty.

• Effective October 2022, Pennsylvania increased the gross income limit of its BBCE policy for households without a member who is elderly or has a disability from 160 percent to 200 percent of poverty, similar to its BBCE gross income limit for households with a member who is elderly or has a disability.

• Effective January 2023, New Hampshire increased the gross income limit of its BBCE policy from 185 percent to 200 percent of poverty and removed the requirement that the household include a child under age 22 and a relative of the child.

• Effective July 2023, Michigan removed the asset limit from its BBCE policy.

• Effective April 2023, Louisiana implemented a standard medical deduction demonstration program

(SMD).

Although not noted in the previous documentation in this series, effective July 2022, Louisiana increased the gross income limit of its BBCE policy from 130 percent to 200 percent of poverty. No change occurred in the gross income limit of Louisiana’s BBCE policy between FY 2022 and FY 2023.

See Chapter III for more information about State SMDs and Appendix B for more information about State BBCE policies.

B. Key changes to the FY 2023 SNAP QC database

The contents of the FY 2023 SNAP QC database differ in several ways from earlier databases. The changes are the result of three factors: (1) differences in the QC sample; (2) changes to the raw data file, and (3) Federal and State policy changes.

1. Differences in the QC sample

As with the FY 2022 database, the FY 2023 database contains data from all 12 sample months of the fiscal year (October 2022 through September 2023). Most States and territories contributed data for all 12 months in each file. However, Delaware does not have any eligible sample data for July 2023 through September 2023. In the FY 2022 database, the exceptions are Alaska, Delaware, the District of Columbia, Rhode Island, and the Virgin Islands, all of which lack sample data for at least one month.

2. Changes to the raw data file

Person-level variables in the raw data file usually include data for up to 16 household members. However, the FY 2023 data file included one 17-person household. We added variables for a 17th household member to accommodate this household.

The QC review codes for the person-level sex variable (SEXi) will change in FY 2024 to add a new code for individuals who prefer not to answer. However, three individuals in FY 2023 were assigned the new code. Chapter V describes the new code in the codebook.

3. Federal policy changes affecting the contents of the file

Effective September 2023, the FRA increased the upper age limit of adults subject to time-limited benefits to 50. This change led to changes to the NDICSCAi variable for that month. The NDICSCAi variable defines this population.

II. Overview of the SNAP QC Database The SNAP QC database is an edited version of the raw data file generated by SNAP’s QC System. The FY 2023 SNAP QC database contains detailed demographic, economic, and SNAP eligibility information for a nationally representative sample of 43,776 SNAP units.5 The SNAP QC data, produced annually, are well suited for tabulating characteristics of SNAP units and simulating the impact on SNAP units of various policy changes to the program. Accordingly, the SNAP QC database is the source for FNS’s annual report, Characteristics of Supplemental Nutrition Assistance Program Households, and FNS’s QC Minimodel, a microsimulation model that estimates the effect of proposed changes to SNAP on currently participating units. In this chapter, we provide an overview of the raw data file and the processing and edits that convert the data file to the SNAP QC database.

A. The QC System

The raw data file is generated from the monthly reviews of SNAP cases conducted by State SNAP agencies as part of the QC System (SNAP-QCS). The primary objective of QC reviews is to assess the accuracy of eligibility determinations and benefit calculations in sampled cases.6 Participating units, or active cases, are reviewed to determine whether they are indeed eligible to participate and are receiving the correct benefit amount. Units that had their participation denied, terminated, or suspended, called negative cases, are reviewed to determine whether the denial, termination, or suspension was correct. This information is factored into States’ overall case and procedural error rate (CAPER).7 The SNAP QC database is normally based on the sample of active cases drawn each month for the 50 States, the District of Columbia, Guam, and the Virgin Islands. The FY 2023 database had data for all States and territories except for Delaware, which did not collect data on any SNAP eligible household for three months—July 2023 through September 2023.

State QC reviewers check data for the sampled cases. They gather financial and demographic information from the sampled unit’s case file, visit the household to reinterview participants, and then determine whether the SNAP unit received the correct SNAP benefit amount. Information from the review is either uploaded or entered directly into the SNAP-QCS by State agencies. FNS regional offices conduct a Federal re-review of a subsample of each original State sample. The Federal re-review data are also entered into the SNAP-QCS and are used in conjunction with the State review data to calculate the official payment error rate for each State. States can be sanctioned on the basis of their official payment error rates.

5 In this technical documentation, “SNAP unit” or simply “unit” refers to individuals who together are certified for and receive SNAP benefits. A household may contain multiple SNAP units and/or individuals who do not receive SNAP benefits. However, because QC sampling is done at the unit level, each record contains data on only one SNAP unit.

6 QC reviewers follow guidelines from the SNAP Quality Control Review Handbook 310 and the Quality Control Review Schedule 380-1.

7 The CAPER accounts for both the accuracy of the State agency’s determination and their compliance with federal procedural requirements around the determination. See https://www.fns.usda.gov/snap/qc/caper for more information.

https://www.fns.usda.gov/snap/qc/caper

Chapter II Overview of the SNAP QC Database

Most of the data in the raw data file are the financial and demographic data collected during the review.

The issued benefit amount and eligibility status determined by the caseworker are also on the file, along with the error amount and eligibility status determined by the reviewer.8 The reviewer-determined entries are defined as follows:

• If the SNAP unit was eligible and the authorized benefit amount determined by the reviewer equaled the issued benefit, then the error amount is zero and the case finding is entered as “amount correct.”

• If the SNAP unit was eligible and the authorized benefit amount varied from the issued benefit, then the difference between the two amounts is recorded as the error amount and the case finding is either an “overissuance” or “underissuance.” In FY 2023, error amounts of $54 or less were not included in the calculation of State error rates.9

• If the reviewer determines that the SNAP unit was ineligible, then the issued benefit amount is recorded as the error amount and the case finding is “ineligible.”

State QC reviewers also check the negative cases to decide whether proper procedures were used to deny or terminate a case. Because these cases are not participating in SNAP, they are not included in the SNAP QC database or the QC Minimodel.

B. The raw data file

Although most participating SNAP units in the active case file are subject to sampling, certain types of units that are not appropriate for review are excluded. Specifically, the active case universe excludes the following types of cases:

• Dropped as a result of oversampling

• Listed in error as active cases, including but not limited to the following:

− Negative cases incorrectly included in the active case file

− Cases that did not participate in SNAP for the sample month, including suspended cases and those that were eligible for zero benefits before any recoupments were made

− Cases receiving restored benefits that were not otherwise participating

− Cases receiving retroactive benefits for the sample month

• Receiving benefits solely through a Disaster SNAP program authorized by FNS

• Pending a hearing for an adverse action

• Under investigation for SNAP fraud (including those with pending fraud hearings)

• Where all members have died or moved outside the State

• Where no member could be interviewed because of the following:

− All members had been hospitalized, incarcerated, or placed in a mental institution and were expected to remain there for 95 days after the end of the sample month

− Members could not be located

8 The SNAP benefit does not include the emergency allotments authorized as part of the FFCRA.

9 This error amount, called the tolerance threshold, is adjusted each year to account for inflation. The tolerance threshold increased from $48 in FY 2022 to $54 in FY 2023.

The sampling unit within the active case universe is the SNAP unit, as defined in an FNS-approved State manual. State sampling plans must conform to accepted principles of probability sampling. A State may use either a simple random sampling plan or a more complex sampling design that better meets its needs.

FNS must approve all sampling designs, including simple random sampling.

In a typical year, including FY 2023, the standard minimum annual State sample sizes range from 300 reviews to 2,400 reviews, depending primarily on the size of the monthly participating caseload. States must use the following guidelines when determining their standard annual QC sample sizes:

• If the average monthly caseload is under 10,000, the standard minimum sample size is 300 cases per year.

• If the average monthly caseload is 60,000 or greater, the standard minimum sample size is 2,400 cases per year.

• If the average monthly caseload is between 10,000 and 60,000, the standard minimum sample size is derived by the following formula:

Standard minimum = 300 + 0.042 (N − 10,000), where N is the average monthly caseload.

A State may choose an optional minimum sample size if it agrees not to dispute later payment error rate findings and the associated sanctions on the basis of the precision of the estimates. Optional minimum sample sizes are determined as follows:

• If the average monthly caseload is under 12,942, the optional minimum sample size is 300.

• If the average monthly caseload is 60,000 or greater, the optional minimum sample size is 1,020.

• If the average monthly caseload is between 12,942 and 60,000, the optional minimum sample size is derived by the following formula:

Optional minimum = 300 + 0.0153 (N – 12,941), where N is the average monthly caseload.

In FY 2023, all States chose to use the optional minimum sample size. FNS applies adjustments to State payment error rates when the State’s QC review completion rate falls below a threshold of 98 percent.

C. Creation of the SNAP QC database

We create the SNAP QC database from the raw data file by following four steps: (1) preliminary processing, (2) data editing, (3) variable construction, and (4) weighting.

1. Preliminary processing

After converting the raw data file into a SAS file, we generate and inspect a series of quality assurance counts and frequency distributions for the values of each variable on the file. We assign missing value codes to data that are illogical or out of range, missing from the file, or coded as unknown in the source file.10 We remove records from that file for the following reasons:

10 See the codebook in Chapter V for the valid values for each variable.

• Coded as not subject to review (REVDISP = 2), incomplete (REVDISP = 3), or deselected due to oversampling (REVDISP = 4)

• Coded with review findings of ineligible (STATUS = 4)

• Missing all data except error and status information, identified as those coded with 0 case members (CERTHHSZ = 0), or had unresolved inconsistencies, as detailed in later sections

• Found by the reviewer to be eligible but not qualifying for a positive benefit or as having a benefit overissuance equal to or exceeding the recorded benefit (STATUS = 2 and RAWBEN <= AMTERR)

Table II.1 shows the number and percentage of cases that were dropped from the FY 2023 edited SNAP QC database.

Table II.1. Number and percentage of cases sampled, dropped from the edited file, and included in the edited file, FY 2023

Category

FY 2023

SNAP QC

sample

Percentage of cases sampled

Percentage of cases subject to review Number of cases sampled 55,115 100.0 n.a.

Cases not subject to review 2,378 4.3 n.a.

Cases deselected to correct for oversampling 0 0.0 n.a.

Cases subject to review 52,737 95.7 100.0

Incomplete cases 6,812 12.4 12.9 Cases completed 45,925 83.3 87.1

Not eligible for SNAP 1,509 2.7 2.9 Not eligible for a positive benefit 484 0.9 0.9 Eligible for a positive benefit 43,932 79.7 83.3

Dropped due to unresolved inconsistencies 156 0.3 0.3 SNAP units in the final SNAP QC database 43,776 79.4 83.0

Source: FY 2023 SNAP QC sample.

n.a. = not applicable.

2. Data editing

Consistent measures of SNAP unit size, income, and benefit level are critical to any analysis of SNAP units. However, data for these measures are not always consistent in the raw data file. For instance, the sum of the income of each person in the unit may not equal the reported unit-level gross income. Such inconsistencies may be rooted in the initial case record information or the data entry process. During data editing, we resolve the inconsistencies described below. We drop the small number of SNAP units with unresolved inconsistencies from the edited file.

The overall strategy of the editing process is to ensure that certain relationships hold for all cases. The two most basic relationships are the following:

• Net income must equal gross income minus the total deductions for which the unit is eligible, and it must not be negative.

• The SNAP benefit level must equal the maximum benefit for that unit size minus 30 percent of net income (or be set to the minimum benefit if appropriate), and it must not be negative.

In addition, several important relationships must hold for some final and intermediate variables. For example:

• Gross unit income must equal the sum of all countable person-level income amounts.

• The earned income deduction must equal the specified percentage (rounded down) of countable earned income.

• The excess shelter expense deduction must equal shelter costs above 50 percent of gross income minus all other deductions up to a cap. Units with elderly members or with non-elderly individuals with disabilities are not subject to the cap. Units with a homeless household shelter deduction will not have an excess shelter expense deduction.11

• Total deductions must equal the sum of the following:

− Standard deduction

− Earned income deduction

− Dependent care deduction

− Medical expense deduction

− Child support payment deduction12

− Excess shelter expense deduction or homeless household shelter deduction

Households participating in the Minnesota Family Investment Program (MFIP) or a Supplemental Security Income Combined Application Project (SSI-CAP) are subject to different eligibility and benefit determination rules, and their data are edited accordingly.

In Chapter III, we describe the complex process by which we determine whether a case is internally consistent and, if not, perform the needed edits.

3. Variable construction

We construct several variables from the reported data once the file is edited. Some of the constructed variables (for example, unit-level gross income, net income, and unit size) are edited versions of raw variables, while others (such as non-elderly individuals with disabilities) are created to more easily identify units and individuals with certain characteristics. The major classes of constructed variables are unit-level countable income variables, SNAP eligibility and benefit determination variables, and characteristics flags:

• Unit-level countable income variables. The total SNAP unit income variable for each type of income (for example, Temporary Assistance for Needy Families [TANF] or Social Security) is constructed by summing the person-level income of that type over all individuals in the SNAP unit.

The total SNAP unit gross income, earned income, and unearned income variables are constructed by summing all the appropriate unit income variables.

11 The 2018 Farm Bill made mandatory the existing State option to provide a standard shelter deduction to homeless households that had qualifying shelter expenses and that were not claiming the excess shelter expense deduction.

The 2018 Farm Bill also indexed the homeless shelter deduction to inflation. In FY 2023, the value of the mandated homeless shelter deduction was $166.81.

12 In some cases, child support payments are excluded from gross income and are not taken as a deduction.

• SNAP eligibility and benefit determination variables. Variables used to determine eligibility and benefits—such as SNAP unit deductions, SNAP unit net countable income, and SNAP unit benefits—are constructed on the basis of SNAP unit countable income and unit demographic characteristics.

• Characteristics flags. Characteristics flags identify SNAP units with certain features, such as the presence of an elderly individual or a non-elderly individual with a disability. In addition, we merge data from Census Bureau files to identify whether a SNAP unit resides in a metropolitan, micropolitan, or rural area.13

4. Weighting

We weight the observations in the raw SNAP QC data file such that the weighted totals match as closely as possible three adjusted SNAP Program Operations totals: (1) the monthly number of SNAP units by State and sampling stratum, (2) the monthly number of SNAP participants by State, and (3) the monthly total benefits issued by State. SNAP Program Operations totals are generated from FNS’s National Data Bank (NDB) and reflect actual levels of participation and benefit issuance. Through FY 2022, we adjusted the data as needed to remove units receiving benefits issued through the SNAP disaster assistance program, as well as disaster benefits for ongoing SNAP recipients, COVID-19 emergency allotments, and replacement benefits because these households are not included in the SNAP QC database. Beginning with the FY 2023 file, we used NDB data that already excluded these units and benefits. We used Form 388 (State Issuance and Participation Estimates) data for SNAP units and individuals14 and Form 46 (Issuance Reconciliation Report) data for SNAP benefit issuance.15 We used Form 388 data for SNAP benefits in States when Form 46 benefits data were missing (Connecticut and Massachusetts in September 2023 and Rhode Island in August and September 2023) or in States when FNS indicated Form 388 data were more reliable (Minnesota for all months).

For FY 2023, we further revised the disaster-adjusted values for units, individuals, or benefits when we suspected errors in the program data due to larger than average month-to-month changes in the average per person benefit. Specific adjustments were as follows:

• Units and individuals. In consultation with FNS, we made adjustments to Program Operations data for units or individuals for six States (Arkansas, New Jersey, North Carolina, Oklahoma, Tennessee, and Texas) and Guam in one or more months. In the data for Guam, North Carolina, and Oklahoma, units appeared to be overestimated in the months we adjusted. In Guam, individuals also appeared to be overestimated. In the data for Arkansas, New Jersey, Tennessee, and Texas, both units and individuals appeared to be underestimated in the months we adjusted. FNS provided revised numbers of units and individuals to use for Guam, New Jersey, and Tennessee based on correspondences with

13 A micropolitan statistical area has at least one urban cluster of at least 10,000 people but fewer than 50,000 people and includes adjacent territory that has a high degree of social and economic integration with the core, as measured by commuting ties.

14 Specifically, we used Form 388, part 3a, which is limited to regular ongoing SNAP participating individuals, and part 4a, which is limited to regular ongoing SNAP participating households. Part 3a and 4a exclude D-SNAP-only participants, and those receiving only disaster supplements, replacement benefits, or other issuances that are not considered regular ongoing.

15 Specifically, we used Form 46, part 6a, which is limited to regular ongoing benefits. It excludes D-SNAP, disaster supplements, replacements, State/Federal investigator benefits, or other issuances that are not considered regular ongoing.

or notes from the States or territory. For the remaining States, we adjusted the counts of units and individuals by using the average values for the adjacent months for the State.

• Benefits. We made adjustments to Program Operations data for benefits in 12 States, Guam, and the District of Columbia in one or more months. The 12 States were Alabama, Arkansas, Georgia, Hawaii, Massachusetts, Michigan, Minnesota, Nevada, Ohio, Oklahoma, Pennsylvania, and Tennessee. In Arkansas, Georgia, Guam, Michigan, Nevada, and Oklahoma, benefits appeared to be overestimated in the months we adjusted, compared with benefit amounts in adjacent months during the fiscal year. In Minnesota and Pennsylvania, the benefits appeared to be underestimated. In Alabama, the District of Columbia, Hawaii, Massachusetts, Ohio, and Tennessee, benefits in one month appeared to be overestimated and those in another month appeared to be underestimated. In three States—Arkansas, Michigan, and Ohio—FNS provided revised benefit amounts. In the remaining States or territories, we adjusted total benefits in one of three ways, depending on the State’s data and prioritizing the simplest approach in the order listed below: (1) by using the average values for the adjacent months for the State, (2) by using the average values of consecutive months if more than one month required adjustments, or (3) by using the average fiscal year value for the State.

The criteria used to determine whether an adjustment was needed for a particular month and State was based on the mean absolute deviation of the average per person benefit and information from FNS. After finalizing adjustments to the State program data, we adjusted the data to remove units that were ineligible for benefits, because these households are not included in the SNAP QC database. The rates of SNAP units and individuals receiving benefits in error, as well as total benefits received in error, are estimated from the raw QC data file. This process for the FY 2023 database was consistent with that for prior file years.

As a result of these adjustments, the totals used to weight the FY 2023 SNAP QC database do not match FNS administrative records. In addition, the QC System sample-based estimates differ slightly from the target numbers for the QC database. The weighting program was unable to match the disaster- and error-adjusted program targets for individuals and benefits in Alaska in March 2023 and so reverted to using the same weight for all households in that State and month. As shown in Table II.2, this approach resulted in negligible differences in the national weighted totals for individuals and benefits, and differences of approximately 1 and 3 percent, respectively, in the fiscal year weighted totals for individuals and benefits in Alaska. Although the draft edited FY 2023 SNAP QC data file contains samples of fewer than 10 households in Alaska in October and December 2022 and Guam in September 2023, the weighting program was able to match the weighting targets in these months. Thus, we kept all months of data in Alaska and Guam on the database. However, we caution against using monthly tabulations in the months in Alaska or Guam that are sub-optimally weighted or too small to produce informative data on subgroup distributions.

Table II.3 compares the aggregate program participation data for FY 2023 to the QC System sample-based estimates. Table II.4 compares average unit size, benefit per person, and household size in the Program Operations data to the QC sample estimates. Appendix Tables D.1 through D.3 present the weighted unit, individual, and benefit totals by State and month. Appendix Tables D.4 through D.6 show the corresponding adjustments to the Program Operations data that yielded the target numbers for those weighted totals. In Chapter III, Section C, we describe the derivation of the sampling weights in detail.

Table II.2. FY 2023 weighting targets versus SNAP QC data file weighted totals

Number of individuals Amount of benefits

Error-adjusted target

SNAP QC

weighted total

% difference from target

Error-adjusted target

SNAP QC

weighted total

% difference from target

National 40,064,877 40,065,125 0.00 7,102,945,012 7,102,643,960 0.00

Alaska 29,086 29,334 0.85 8,603,454 8,302,402 -3.50

Source: FY 2023 SNAP QC database.

Table II.3. Comparison of program data to edited SNAP QC database, FY 2023

Average monthly values

Category

Number of households

Number of participants

Value of benefits (dollars)

Program data 22,303,632 42,166,077 7,747,836,756 Adjustments to program data for the following:

Disaster assistance a 37,670 95,399 23,675,590 Smoothing the datab (5,478) 865,778 (15,026,075) Excluded State-monthsc (91,312) (1,073,541) (24,721,689) Ineligible SNAP units 987,472 2,213,561 660,963,919

Target numbers for edited SNAP QC database 21,375,279 40,064,879 7,102,945,012 Edited SNAP QC database 21,375,279 40,065,127 7,102,643,960

Source: FY 2023 Program Operations data and SNAP QC database.

a Program data values are based on data received from FNS on January 23, 2025, and include regular ongoing SNAP and D-SNAP. These numbers differ from those on FNS’s website, which also include disaster supplements, investigator issuances, and replacements. As discussed above, we used NDB data that already excluded disaster assistance households, participants, and benefits and replacement benefits to weight the FY 2023 database.

b Disaster assistance represents D-SNAP households, participants, and benefits (including D-SNAP benefits to ongoing households). It may also include return issuances for D-SNAP participants.

c We made smoothing adjustments when we suspected errors in the program data due to larger than average month-to-month changes in the average per person benefit or, when requested by FNS, based on their correspondences with or notes from a State or territory d As discussed in Chapters I and II, July through September data for Delaware are not included in the FY 2023 SNAP QC database. This row shows the aggregate effect on the monthly average program totals when the months not included in the SNAP QC database are removed from the calculation.

Table II.4. Averages in program data compared to edited SNAP QC database, FY 2023

Average monthly value

Category

Average SNAP unit size

Average benefit per person (dollars)

Average benefit per household (dollars)

Program data 1.89 184.30 348.22 Target numbers for edited SNAP QC database

1.87 177.29 332.30

Edited SNAP QC database 1.87 177.28 332.28 Sources: FY 2023 Program Operations data and SNAP QC database.

D. Final SNAP QC database

We create two versions of the final SNAP QC database: (1) a restricted-use version that includes all variables and (2) a public-use version that, for privacy reasons, excludes the QC review number (REVNUM) and four geographic variables: COUNTYCD, LOCALCOD, AK_AREA, and URBRUR.

We provide a more detailed explanation of the variables on the file in Chapter V.

After we develop the SNAP QC database, we create SAS, Stata, and SPSS versions that may be used to tabulate characteristics of SNAP units, as well as a binary file that serves as the underlying database for FNS’s QC Minimodel.

III. FY 2023 SNAP QC File Development Process

A. Developing the SNAP QC file

In this chapter and in Figure III.1, we describe the programs and data used to develop the FY 2023 SNAP QC file.16

Step 1. Obtain data

We received the data from FNS in an ASCII (or text) format.

INPUT FILE FY2023

Record length 2,250 55,115 records

(ASCII file)

Step 2. Read in and prepare file

We converted to SAS format the specified fields from the raw FNS file and created the unique record identifier (HHLDNO).

PROGRAM NAME 10_SASIFY.SAS

INPUT FILE FY2023 (ASCII; 55,115 records) OUTPUT FILE QCFY2023_1.SAS7BDAT (55,115 records; 721 variables)

Step 3. Conduct quality assurance (QA) review of the data

We ran preliminary frequencies on the SAS file and examined them for data corruption, consistency across States and months, and the extent of missing and out-of-range data. In addition, we calculated means and compared them with means for the previous year.

PROGRAM NAMES 01_FREQS.SAS

02_FREQSA.SAS

03_FREQS_ELG.SAS

04_COMPARE.SAS

05_OBS_STATE_MONTH.SAS

INPUT FILE QCFY2023_1.SAS7BDAT (55,115 records; 721 variables)

16 Copies of the file development programs are available from FNS upon request.

Chapter III FY 2023 SNAP QC File Development Process

Figure III.1. FY 2023 SNAP QC file development process

Step 4. Set SNAP parameters

We obtained relevant SNAP policy parameters, including maximum and minimum benefit amounts, income screens, Standard Utility Allowance (SUA) amounts, and values for the MFIP and SSI-CAPs by State.17 We entered them into a SAS format library and used the formats for the program in Step 6.

OUTPUT PROGRAM 31_FORMAT.SAS

Step 5. Define metropolitan areas

We added geographic information to the file. Using the local agency code in the raw data file, we assigned a county Federal Information Processing Standards (FIPS) code to each SNAP unit. We flagged unknown local agency codes for correction or addition to a concordance of local agency codes by county and State. We then merged each unit to the 2020 and July 2023 Census Bureau files of metropolitan and micropolitan areas by using State and county codes. We added 87 new counties present in the July 2023 file. For 56 counties that had a changed metro-micro status, we decided to retain the status from the 2020 file due to uncertainty about precisely when the change occurred. We coded units as metropolitan or micropolitan, depending on their match to one of the Census Bureau files. We coded those not found in either file as rural, except for those with State-wide local codes, which we coded as missing metropolitan status. We assigned Alaska units with missing or unknown local agency codes a metropolitan status based on the unit’s region (Alaska Urban, Alaska Rural I, or Alaska Rural II). We did not include cases not subject to review or incomplete cases in the output file.

PROGRAM NAME 20_URBAN.SAS

INPUT FILES QCFY2023_1.SAS7BDAT (55,115 records; 721 variables) METRO2_20_23.TXT (ASCII; 1,284 records; 4 variables) (Census

2020 and 2023 Metropolitan File) MICRO2_20_23.TXT (ASCII; 719 records; 4 variables) (Census

2020 and 2023 Micropolitan File) FIPS_LAC.TXT (ASCII; 5,276 records; 6 variables)

(Concordance of local area codes) OUTPUT FILE URBAN23.SAS7BDAT (45,925 records; 5 variables)

Step 6. Recode and standardize variables

We edited the file to resolve inconsistencies between variables within a unit and created several unit-level variables pertaining to SNAP affiliation, income deductions, the shelter limit, benefit amounts, assets, poverty status, and types of income.