Nutrition Assistance Program Report Series
The Office of Analysis, Nutrition and Evaluation
Special Nutrition Programs Report No. WIC-01-WICVM
WIC Vendor ManagementStudy, 1998
Final Report
icpvA United States Food and July 2001
UjL/rA Department of Nutrition
Agriculture Service
&
WU'W**'21*
Non-Discrimination Policy
The U.S. Department of Agriculture (USDA) prohibits discrimination in all its programs and activities
on the basis of race, color, national origin, gender, religion, age, disability, political beliefs, sexual
orientation, and marital or family status. (Not all prohibited bases apply to all programs.) Persons
with disabilities who require alternative means for communication of program information (Braille,
large print, audiotape, etc.) should contact USDA's TARGET Center at (202)720-2600 (voice and
TDD).
To file a complaint of discrimination, write USDA, Director, Office of Civil Rights, Room 326-W,
Whitten Building, 14th and Independence Avenue, SW, Washington, DC 20250-9410 or call (202)
720-5964 (voice and TDD). USDA is an equal opportunity provider and employer.
USDA United States
Department of
Agriculture
Food and
Nutrition
Service
July 2001
Special Nutrition Programs
Report No. WIC-01-WICVM
WIC Vendor Management Study,
1998
Authors:
HSR:
Loren Bell
Mary Harkins
Vivian Shayne
Susan Schreiber
Chris Miller
RTI:
Donald Smith
Paul Moore
Submitted by:
Health Systems Research, Inc.
1200 18* Street, NW
Suite 700
Washington, DC 20036
Project Director: Loren Bell
Submitted to:
Office of Analysis, Nutrition and Evaluation
USDA, Food and Nutrition Service
3101 Park Center Drive
Alexandria, VA 22302-1500
Project Officers:
Patricia McKinney
Boyd Kowal
This study was conducted under Contract number 53-3198-07-012 with the Food and Nutrition Service.
This report is available on the Food and Nutrition Service web site: http://www.fns.usda.gov/oane.
Suggested Citation:
U.S. Department of Agriculture, Food and Nutrition Service, Office of Analysis, Nutrition and Evaluation,
WIC Vendor Management Study, 1998 Final Report, by Loren Bell, et al. Project Officers, Patricia
McKinne/, Boyd Kowal. Alexandria, VA: 2001.
ACKNOWLEDGMENTS
This report was prepared by Loren Bell, Vivian Shayne, Mary Harkins, Susan Williams, and
Christopher Miller of Health Systems Research, Inc. under contract to the Food and Nutrition Service
(Contract No. FCS-53-3198-7-012). Other staff contributing to this project included Donald Smith,
Paul Moore, William Yeager, and Gina Kilpatrick ofthe Research Triangle institute. Thanks also go
to all of the field data collectors who conducted over 4500 compliance purchases across the country.
Their tireless efforts helped to keep the project on schedule and provided the study with quality data
for analysis.
We also wish to acknowledge the support and assistance we received from Marie Mitchell, Boyd
Kowal and Patricia McKinney who served as our Project Officers for this study. Their insight and
guidance over the course of the study proved invaluable.
Most importantly, we would like to thank the State WIC Directors and Vendor Managers who
contributed their time and resources in providing the study with grocer demographic data, food
instruments, and reconciliation reports. Without their contribution, this study would not have been
possible.
Table of Contents
List ofFigures iv
List of Tables vii
Executive Summary. xiv
Chapter I. Introduction and Overview 1
A. Background 2
B. Overview ofWIC Retail Vendors 3
C. Potential Uses of the Data from the Vendor Management Study 4
D. Overview of the Study Methodology 5
E Organization of this Report 6
Chapter II. Study Methodology 7
A. Defining the Survey Population 7
B. Constructing the Sampling Frame 8
C. Selecting the Sample 10
D. Developing the Data Collection Instrument 13
E Assembling the Data Collection Staff 14
F. Training the Field Staff 14
G. Equipping the Field Staff 15
H. Quality Control 17
I. Survey Weight 18
J. Overview of Statistical Analysis Methods 22
w
Chapter ID. Demographics ofthe Study Population 27
A. Metropolitan and Non-metropolitan Areas 28
B. Vendor Classifications 28
C. Vendor-Specific Versus Open Food Instrument Systems 29
D. Vendor-to-Participant Ratio 29
E Vendor Size 30
F. Vendor Use of Scanning Equipment 32
G. Cashier Familiarity with WIC Transactions 34
Chapter IV Findings Related to Vendor Administrative Errors 37
A Requiring the Participant to Sign the WIC FI Prior to Entering the
Purchase Price 38
B. Insufficient Stock 41
C. Vendors Offering Rain Checks and Requiring Buyers to Pay
Additional Cash 45
D. Provision of Receipts to Buyers 46
Chapter V Findings Related to Vendor Overcharge and Undercharge 48
A Vendor Overcharges 49
B. Models Describing Factors Contributing to Overcharge 57
C. Vendor Proclivity to Overcharge 64
D. Findings Related to Undercharge 65
E Approximating a National Estimate of Overcharge 68
F. Approximating a National Estimate of Undercharge 71
Chapter VI. Findings Related to Vendor Allowance of Substitutions 73
A Overall Results Related to Substitution Buys 74
B. Relationships between Type of Substitution and Vendor and
State Demographics and Administrative Errors 75
Chapter VII. Conclusion and Discussion 81
A Challenges Facing WIC State Agencies 81
B. Opportunities for WIC State Agencies to Use Study Findings to
Improve Vendor Management Systems 83
C. Conclusion 85
References 86
/
Appendix A:
Appendix B:
Appendix C:
Appendix D:
Appendix E:
Appendix F:
Appendix G:
Appendix H:
Tables Related to Demographics of the Study Population
Tables Related to Administrative Errors
Tables Related to Vendor Overcharge and Undercharge
Tables Related to Vendor Allowance of Substitutions
Tables Related to Administrative Errors for the Safe Buys
Tables Related to Overcharges for the Safe Buy
Sample Design, Selection, and Weighting
Compliance Buy Forms
IU w
List of Figures
Demographics of the Study Population
III-1. Distribution ofWIC Vendors Included in the Study by Locale
III-2. Distribution ofWIC Vendors Included in the Study by Type of Food Instrument
System
III-3. Distribution ofWIC Vendors Included in the Study by Vendor-to-Participant Ratio
Categories
III-4. Number ofWIC Vendors Included in the Study by Number of Cash Registers
III-5. Distribution ofWIC Vendors Included in the Study by Store Size
III-6. Distribution ofWIC Vendors Included in the Study by Use of Scanning Equipment
III-7. Distribution of WIC Vendors Included in the Study with No Scanning Equipment by
Vendor Size
III-8. Distribution of WIC Vendors Included in the Study by Cashier's Indication of
Unfamiliarity with WIC Transaction Procedures
III-9. Distribution of WIC Vendors Included in the Study with a Cashier Indicating a Lack
of Familiarity with WIC Transactions by Means of Indication
III-10. Percentage ofWIC Vendors Included in the Study with Cashiers Indicating
Unfamiliarity with Conducting a WIC Transaction by Vendor Size
Administrative Errors
IV-1. Percentage of WIC Vendors Committing Administrative Errors by Type of Error
Across All Buys
IV-2. Percentage ofWIC Vendors by Frequency of Occurrences of Deviation from
Countersignature Procedures Across All Buys
IV
l/il
IV-3.
IV-4.
rv-5.
IV-6.
rv-7.
IV-8.
rv-9.
Percentage ofWIC Vendors Deviating from Countersignature Procedures by Food
Instrument System and Locale across All Buys
Percentage of WIC Vendors with Insufficient Stock by Type of Food Instrument
System across All Buys
Percentage ofWIC Vendors with Insufficient Stock by Locale across All Buys
Percentage ofWIC Vendors with Insufficient Stock by Type of Food Package
across All Buys
Percentage ofWIC Vendors by Frequency of Insufficient Stock across All Buys
Percentage of WIC Vendors by Frequency of Occurrences of Cashier Not Providing
a Receipt across Three Buys
Distribution of Administrative Errors Committed by WIC Vendors Who Did Not
Overcharge, Undercharge, or Substitute Unauthorized Items Across All Buys
Vendor Overcharge and Undercharge
V-l.
V-2.
V-3.
V-4.
Percentage of WIC Vendors Overcharging Across All Buys
Percentage ofWIC Vendors by Frequency of Occurrences of Overcharging Across
All Buys
Average Amount of Overcharge by Type ofBuy
Distribution ofWIC Vendors Overcharging as a Function of Use of Scanning
Equipment Across All Buys
V-5. Distribution of WIC Vendors Overcharging as a Function of Vendor Si2e Across All
Buys
V-6. Distribution ofWIC Vendors Overcharging as a Function of Countersignature Timing
Across All Buys
V-7.
V-8.
Average Amount of Undercharge by Type of Buy
distribution ofWIC Vendors Undercharging as a Function of Vendor Size Across All
Buys
l////
Vendor Acceptance of Substitutions
VI-1. Outcome ofWIC Vendors Accepting Buyer-Initiated Substitutions by Type of
Substitution
VI-2. Outcome of WIC Vendors Accepting Buyer-Initiated Minor Substitutions by
Vendor's Use of Scanning Equipment
VI-3 Outcome ofWIC Vendors Accepting Buyer-Initiated Major Substitutions by
Vendor's Use of Scanning Equipment.
VI-4. Outcome of WIC Vendors Accepting Buyer-Initiated Major Substitutions by Vendor
Size
VI-5. Outcome of WIC Vendors Accepting Buyer-Initiated Major Substitutions by Cashier
Experience
VI
A
List of Tables
Chapter II
Study Methodology
II-1. Vendor Eligibility Categories for Ali Vendors Included in the Sample
II-2. Study Response Rates for All Vendors Included in the Study
II-3. Weighting Class Categories for All Vendors Included in the Study
II-4. Adjusted Survey Weight Categories
Chapter III
Demographics ofthe Study Population
III-1. Source of Demographic Variables Included in the Study
Appendix A
Demographics ofthe Study Population
A-1. National Estimate of the Proportion of WIC Vendors by Locale
A-2. National Estimates of the Proportion ofWIC Vendors by Type of Food Instrument
System
A-3. National Estimate ofWIC Vendors by Store Type
A-4. Distribution ofWIC Vendors by Average Vendor-to-Participant Ratio Category
A-5. National Estimate ofWIC Vendors by Size
Vll
X
A-6. Number and Percentage of WIC Vendors by Use of Scanning Equipment Across All
Buys
A-7. Distribution ofWIC Vendors By Cashier's Indication of Unfamiharity with WIC
Transaction Procedures Across All Buys
A-8. Distribution ofWIC Vendors By Cashier's Type of Indication of Unfamiliarity with
Proper WIC Transaction Procedures Across All Buys
Appendix B
Administrative Errors
B-1. Number and Percentage ofWiC Vendors Committing Administrative Errors by Type
of Errors Across All Buys
B-2. Number and Percentage ofWIC Vendors by Frequency of Occurrences of
Administrative Errors
B-3. Number and Percentage ofWIC Vendois Committing Administrative Errors for Each
Locale and Type of Error Across All Buys
B-4. Number and Percentage ofWIC Vendors Cornmitting Administrative Errors for Each
Type ofFood Instrument System and Type of Error Across All Buys
B-5. Number and Percentage of WIC Vendors Committing Administrative Errors for Each
Type of Food Package Across All Buys
B-6. Number and Percentage ofWIC Vendors Committing Administrative Errors for Each
Type ofBuy
B-7. Number and Percentage ofWIC Vendors Who Committed Administrative Errors,
but Did Not Substitute, Overcharge or Undercharge, by Type of Error A JTOSS All
Buys
B-8. t-Statistics Describing WIC Vendors with Insufficient Stock by Vendor
Characteristics Across All Buys
B-9. t-Statistics Describing WIC Vendors Who Violate Countersignature Procedures by
Vendor Characteristics Across All Buys
B-10. t-Statistics Describing WIC Vendors Who Provide Rainchecks for WIC Foods by
Vendor Characteristics Across All Buys
XI
B-11. t-Statistics Describing WIC Vendors with Administrative Errors by Type ofBuy
Appendix C
Vendor Overcharges and Undercharges
C-\. National Estimate of Undercharge and Overcharge Rates of Occurrence across All
Buys
C-2. National Estimate of Undercharge and Overcharge Rates of Occurrence for the Safe
Buy
C-3. National Estimate of Undercharge and Overcharge Rates of Occurrence for the
Partial Buy
C-4. National Estimate of Undercharge and Overcharge Rates of Occurrence for the
Minor Substitution Buy
C-5. National Estimate of Undercharge and Overcharge Rates of Occurrence for the
Major Substitution Buy
C-6. Number and Percentage ofWIC Vendors by Frequency of Occurrence of
Undercharging or Overcharging
C-7. Nationil Estimate of Undercharge and Overcharge Rates of Occurrence for Each
Type ofBuy
C-8. Average Amount of Undercharge and Overcharge for Each Type ofBuy
C-9. Number and Percentage ofWIC Vendors that Undercharged and Overcharged for
Each Type of Food Package Across All Buys
C-10. Number and Percentage ofWIC Vendors that Undercharged and Overcharged for
Each Use of Scanning Equipment
C-11. Number and Percentage of WIC Vendors that Undercharged and Overcharged for
Each Vendor Size
C-12. National Estimates of Undercharge and Overcharge Occurrences for
Countersignature Timing
C-13. National Estimates of Undercharge and Overcharge Occurrences for Receipt
Provision
*'/
C-14. National Estimates of Undercharge and Overcharge Occurrences for Each Locale
C-15. National Estimates of Undercharge and Overcharge Occurrences for Each Type of
Food Instrument System
C-16. Over All Buys: Single Variable Models of Overcharge
C-17. Safe Buy: Single Variable Models of Overcharge
C-18. Partial Buy: Single Variable Models of Overcharge
C-19. Minor Substitution Buy: Single Variable Models of Overcharge
C-20. Major Substitution Buy: Single Variable Models of Overcharge
C-21. Logit Models for Overcharge
C-22. Logistic Odds Ratios to Overcharge for Repeat Offenders
C-23. t-Statistics from Contrast Analyses Describing Overcharge Across All Buys as a
Function of Type of Food Package anc" Type ofBuy
C-24. t-Statistics from Contrast Analyses Describing Overcharge During Safe Buys as a
Function of Type of Food Package
C-25. t-Statistics from Contrast Analyses Describing Overcharge During Partial Buys as a
Function of Type of Food Package
C-26. t-Statistics from Contrast Analyses Describing Overcharge During Minor Substitution
Buys as a Function of Type of Food Package
C-27. t-Statistics from Contrast Analyses Describing Overcharge During Major Substitution
Buys as a Function of Type of Food Package
C-28. t-Statistics from Contrast Analyses Describing Overcharge Across All Buys as a
Function of Potential Administrative Error and Vendor Size
C-29. t-Statistics from Contrast Analyses Describing Overcharge During Safe Buy as a
Function of Potential Administrative Error and Vendor Size
C-30. t-Statistics from Contrast Analyses Describing Overcharge During Partial Buys as a
Function of Potential Administrative Error and Vendor Size
Y///
C-31. t-Statistics from Contrast Analysis Describing Overcharge During Minor Substitution
Buys as a Function of Potential Administrative Error and Vendor Size
C-32. t-Statistics from Contrast Analysis Describing Overcharge During Major Substitution
Buys as a Function of Potential Administrative Error and Vendor Size
C-33. t-Statistics from Contrast Analyses Describing Undercharge Across All Buys as a
Function of Type of Food Package and Type ofBuy
C-34. t-Statistics from Contrast Analyses Describing Undercharge During the Safe Buy as a
Function of Type of Food Package
C-35. t-Statistics from Contrast Analyses Describing Undercharge During Partial Buys as a
Function of Type of Food Package
C-36. t-Statistics from Contrast Analyses Describing Undercharge During Minor
Substitution Buys as a Function of Type of Food Package
C-3 7. t-Statistics from Contrast Analyses Describing Undercharge During Major
Substitution Buys as a Function of Type of Food Package
C-38. t-Statistics from Contrast Analyses Describing Undercharge Across All Buys as a
Function of Potential Administrative Error and Vendor Size
C-39. t-Statistics from Contrast Analyses Describing Undercharge During Safe Buy as a
Function of Potential Administrative Error and Vendor Size
C-40. t-Statistics from Contrast Analyses Describing Undercharge During Partial Buys as a
Function of Potential Administrative Error and Vendor Size
C-41. t-Statistics from Contrast Analyses Describing Undercharge During Minor
Substitution Buys as a Function of Potential Administrative Error and Vendor Size
C-42. t-Statistics from Contrast Analyses Describing Undercharge During Major
Substitution Buys as a Function of Potential Administrative Error and Vendor Size
C-43. t-Statistics from Contrast Analyses Describing Undercharge Amount Differences
Across All Buys
C-44. t-Statistics from Contrast Analyses Describing Overcharge Amount Differences
Across All Buys
XlV
Appendix D
Vendor Acceptance ofSubstitutions
D-1. National Rate ofWIC Vendors Accepting Buyer-Initiated Substitutions
D-2. Number and Percentage ofWIC Vendors Accepting Buyer-Initiated Minor
Substitutions for Use of Scanning Equipment
D-3. Number and Percentage ofWIC Vendors Accepting Buyer-Initiated Major
Substitutions for Use of Scanning Equipment
D-4. Number and Percentage ofWIC Vendors Accepting Buyer-Initiated Major
Substitutions for WIC Vendor Size
D-5. Number and Percentage ofWIC Vendors Accepting Buyer-Initiated Major
Substitutions for Cashier's Indication ofUiifamiliarity with WIC Transactions
D-6. t-Statistics from Contrast Analyses Describing Minor Substitution Buys by WIC
Vendor Demographics
D-7. t-Statistics from Contrast Analyses Describing Major Substitution Buys by WIC
Vendor Demographics
Appendix E
Administrative Errorsfor the Safe Buy
E-1. Number and Percentage of WIC Vendors Committing Administrative Errors by Type
of Error During the Safe Buy
E-2. Frequency of Administrative Errors for Locale during the Safe Buy
E-3. Frequency of Administrative Errors for Type of Food Listrument System and Type of
Error during the Safe Buy
E-4. Frequency of Administrative Errors for Type of Food Package during the Safe Buy
E-5. National Rate of WIC Vendor Administrative Errors among Vendors Who Did Not
Overcharge, Undercharge, or Substitute During the Safe Buy
JrV
Appendix F
Vendor Overchargesfor the Safe Buy
F-l.
F-2.
F-3.
F-4.
F-5.
F-6.
F-7.
Number and Percentage of WIC Vendors Undercharging or Overcharging by Type
of Food Package During the Safe Buy
Number and Percentage ofWIC Vendors Undercharging or Overcharging by Use of
Scanning Equipment During the Safe Buy
Number and Percentage ofWIC Vendors Undercharging or Overcharging by Size of
Vendor During the Safe Buy
Number and Percentage ofWIC Vendors Undercharging or Overcharging by Timing
ofCountersignature During the Safe Buy
Number and Percentage of WIC Vendors Undercharging or Overcharging by
Provision of Receipt During the Safe Buy
Number and Percentage ofWIC Vendors Undercharging or Overcharging by Locale
During the Safe Buy
Number and Percentage of WIC Vendors Undercharging or Overcharging by Type
of Food Instrument System During the Safe Buy
JQD
X*7
Executive Summary
A. Overview
The purpose of the study was to learn the extent to which retail grocers, defined as "vendors" in the
WIC Program, authorized to provide food to WIC participants, were violating program rules and
procedures, and to determine which programmatic and/or demographic variables could be associated
with vendor violations. The study examines three critical research questions in the area ofWIC
vendor management
• To what extent do WIC vendors commit vendor violations and administrative errors when
conducting a WIC transaction at the point of sale?
• To what extent do WIC vendors overcharge or undercharge the WIC Program?
• To what extent do WIC vendors allow participants to substitute unauthorized items for their
WIC-authorized food items?
These questions were answered through a national data collection effort involving data collectors
posing as WIC participants and conducting compliance buys at a nationally representative sample of
1,565 WIC retail vendors. Data collected and analyzed for this study can be useful to Federal and
State officials in evaluating the extent to which vendors comply with program rules. Key areas in
which these data may be usefiil are described below:
• Quantifying the Level of Vendor Errors;
• Identifying Administrative Practices on Which Vendor Training Should be Focused; and
• Identifying Vendor Demographics Associated with WIC Program Compliance.
B. Methodology
The population of interest for the study was defined as all vendors operating in States with retail food
delivery systems. Excluded from the study were States with direct food delivery systems
(Mississippi), home food delivery systems (all of Vermont and part of Ohio), State-run WIC vendors
(parts of Illinois), military commissaries, and pharmacies which only provided WIC participants with
exempt infant formula and/or WIC-eligible medical foods. Vendors operating in Alaska, Hawaii,
Puerto Rico, and the U.S. temtories, as well as vendors authorized by Indian Tribal Organization
State agencies were also excluded from the study population.
XIV
X</ll
The study sample was designed to meet the precision constraints of estimating national proportions
within 3 percentage points and estimating subgroup proportions within 5 percentage points, with 95
percent confidence. A total sample of at least 1,500 vendors was needed to meet the study's
precision requirements. Vendors were oversampled to ensure the study had a sufficient number of
vendors.
To successfully perform the required compliance buys, it was essential that the data collectors
embody the physical characteristics of women who receive WIC benefits. This meant, for example,
that all data collectors had to be females of childbearing age. In addition, if data collectors were to
perform their assignments without creating suspicion among vendors, it was also necessary for the
data collectors to belong to one of the racial or ethnic groups of customers who regularly shop at
those vendors.
Each data collector was responsible for completion of three compliance buys at each assigned
vendor. Data collectors were assigned an average of 18 vendors, although some had considerably
more and a few had less. The assigned buys at each vendor were performed as follows:
Buy#l: Safe Buy Buyer purchased all food items listed on the food
instrument in the quantities and types listed.
Buy #2: Partial Buy Buyer attempted to purchase some, but not all, of the
food items listed on the food instrument
Buy #3A: Minor Substitution Buyer attempted to substitute an unauthorized food item
within an approved food category.
Buy #3B: Major Substitution Buyer attempted to substitute an unauthorized item clearly
outside an approved food category.
Three buys were attempted at each vendor. The third buy was either a "Buy 3A" or a "Buy 3B," as
preprinted on the compliance buy form. To avoid arousing suspicion among vendor staff, data
collectors were instructed to allow five or more days between buys at each sampled vendor. The
primary tasks associated with a compliance buy entailed selecting the correct foods for the buy type
being undertaken, obtaining the shelf price of each item, presenting the food instrument (FT) at the
checkout counter, and observing any administrative violations ofWIC procedures.
Data were collected and reviewed for accuracy. Once a complete database was developed, weights
were assigned to each vendor, and data were prepared for analysis using SAS and SUDDAN
software. Statistical analysis was preformed on the database using a combination of descriptive
analysis and mulitvariate analysis Results were men organized into four categories: descriptions of the
study population, administrative errors, overcharge/undercharge, and substitutions.
XV
/////
C. Description of the Study Population
Vendor demographics were divided into two categories: descriptions of the physical location of the
vendor, and descriptions of the vendors' ability to conduct a WIC transaction. With regard to
location, 70 percent of the study vendors were located in metropolitan areas as compared to non-metropolitan
areas. Almost 80 percent of the vendors were located in States with open FI systems
and slightly over 20 percent were located in States with vendor-specific FI systems.
With regard to descriptive information about the vendors' ability to conduct a WIC transaction, two
areas were examined. First, vendors were grouped by physical size using the number of cash
registers as a proxy. Thirty-one percent of the vendors were classified as small vendors, 35 percent
were classified as medium-sized and 33 percent were classified as large. Use of scanning equipment
was also examined, with 69.1 percent of study vendors using scanners, 27.4 percent lacking scanning
equipment, and 3.6 percent having scanning equipment, but choosing not to scan.
D. Findings Related to Administrative Errors
This study examined some universal factors related to conducting a WIC transaction to determine the
extent to which vendors were not following proper WIC transaction procedures and the extent to
which such administrative errors were associated with overcharging, undercharging, and allowing
substitutions. For the purpose of this study, not following proper WIC transaction procedures was
considered an administrative error. The following categories of administrative errors were examined:
• Requiring the data collector to sign the WIC FI prior to the cashier entering the purchase
price;
• Having insufficient stock thereby preventing the data collector from obtaining her authorized
foods;
• Offering rain checks for foods not available; and
• Asking the data collector to pay cash in addition to the FI for WIC food items.
In addition, while not considered an administrative error in all States, the study examined the
percentage of all WIC vendors that provided a receipt to the data collector.
The most common error noted in the study was the failure of the cashier to have the data collector
sign the WIC FI after the cashier entered the purchase price. A total of 35.4 percent of all WIC
vendors failed to follow the proper countersignature procedures. To a lesser extent, vendors being
out of particular WIC foods was a problem. A total of 5.5 percent of the vendors were not able to fill
the food prescription because they did not cany at least one of the WIC food items on the data
collector's FI. Other variables examined were far less significant. Less than 0.5 percent of the
vendors issued rain checks or asked the data collector to pay cash in addition to the FI.
XVI
X/K
E. Findings Related to Overcharges and Undercharges
As was true in past vendor studies, vendors in this study both overcharged and undercharged the
buyers for items purchased Vendor overcharges and undercharges were examined in total and as a
function of several variables. Significant findings include:
•
•
Across all three buy types, an average of 8.7 percent of all vendors overcharged. When
vendors were examined for frequency of overcharge, 81.9 percent never overcharged, 12.4
percent overcharged only once, 4.2 percent overcharged twice, and 1.5 percent overcharged
three times.
Vendors were most likely to overcharge on a partial buy. In addition, vendors who
overcharged on the partial buy overcharged a larger dollar amount than on other types of
buys. The average amount of overcharge was $0.19 for safe buys, and $0.47 for partial
buys.
When logistic regression models were run for overcharge as a function ofvariables, results
indicated that vendors who failed to provide a receipt were ten times more likely to
overcharge than those providing a receipt. Other variables that seem associated with
overcharge include vendor size, with small vendors being three times as likely to overcharge
than middle-sized or large vendors; and countersignature timing with those vendors that failed
to have the data collector sign the FI prior to entering the purchase price, being four to six
times more likely to overcharge than those who had the data collector sign the FI after the
purchase price was written in.
As noted above, vendors also undercharged An average of almost seven percent of all vendors
undercharged over the three buys. Ofthe vendors where three buys were completed, 83.7 percent
never undercharged, 13.4 percent undercharged only once, 2.3 percent undercharged twice, and less
than 1 percent undercharged all three times.
Approximations of national estimates of total vendor overcharge and undercharge were also
developed The estimates are approximations because data were analyzed over all three buy types,
but it is unknown how often WIC participants make partial purchases or attempt to substitute foods.
When the amount of overcharge is calculated based on all three buys, it is estimated that 1.6 percent
of the total 1998 WIC redemptions nationally are attributed to overcharge. When only the safe buy is
used to calculate the estimate, the percent drops to 0.9 percent of the national WIC redemptions
being attributable to overcharge.
An approximation of national estimates for vendor undercharges was also developed When
examined across all three buys, 0.6 percent of the 1998 WIC redemptions nationally were attributable
to vendors undercharging. When only the safe buy was used to calculate the undercharges, the rate
dropped to 0.4 percent of 1998 national WIC redemptions.
xvn
**
F. Fir. Jings Related to Substitutions
The vendors' willingness to accept substitution of unauthorized foods for die WIC prescription was
also examined. Data collectors were asked to conduct a substitution buy on the third and final buy in
the series. Half the vendors were selected for a minor substitution buy, that is a substitution of
unauthorized foods within a WIC food category (e.g., unauthorized cereals and juices); while the
other half of the vendors were selected for a major substitution buy, which is an attempt to purchase
an item outside of the WIC food category (e.g., soda instead ofjuice). Findings were as follows:
• A large number (34.7 percent) ofvendors allowed minor substitutions. It is interesting to note
that most vendors who allowed minor substitutions also scanned die items. Because scanning
equipment can be programmed to screen out unauthorized purchases, this problem may be
best addressed through stricter requirements for vendors who have scanners to do such
screening.
• Just under four percent of the vendors allowed major substitutions. Vendor size, cashier
familiarity with WIC transactions, and use of scanning equipment were all associated with
major substitutions.
XVUl
Xx/
CHAPTER I
Introduction and Overview
In the fall of 1997 the Food and Nutrition Service (FNS) of the United States Department of
Agriculture commissioned a study to examine the practices of grocers providing supplemental foods to
participants in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC).
The purpose of the study was to learn the extent to which retail grocers (called "vendors" in the WIC
Program) were violating program rules and regulations, and to determine ifprogram management or
demographic variables could be associated with vendor violations. The study was a follow-up to the
1991 Vendor Issues Study, published by FNS, and examines three critical research questions in the
area of WIC vendor management:
• To what extent do WIC vendors commit procedural and administrative errors when
conducting a WIC transaction at the point of sale?
To what extent do WIC vendors overcharge or for that matter undercharge the WIC
Program?
To what extent do WIC vendors allow participants to substitute unauthorized items for their
WIC purchases?
This study examined these questions through a national data collection effort involving data collectors
posing as WIC participants conducting compliance buys at a nationally representative sample of
1,565 WIC retail vendors. This chapter provides an overview of the WIC Program, issues related to
vendor compliance, and potential uses for the data collected. Subsequent chapters will detail the
methodology, statistical analysis, and findings of the 1998 study.
A. Background
The WIC Program, was established in 1972 through an amendment to the Federal Child Nutrition
Act Its purpose is to provide low-income pregnant, breastfeeding, and postpartum women, infants,
and children .p to age five with supplemental foods, nutrition education, and health care referrals to
counteract the adverse effects of poverty on their nutrition and health status.
WIC Program regulations require that each State agency develop a food delivery system to provide
authorized supplemental foods to WIC participants. Three food delivery systems are used by State
agencies: retail purchase, home delivery, and direct distribution. In the home food delivery system,
supplemental foods are delivered directly to the participant's home. In the direct food delivery
system, supplemental foods are purchased by the State agency or a entity acting on its behalf, and
distributed to participants at a warehouse or other facility. The focus ofthis study is the retail food
delivery system, in which participants obtain supplemental foods at authorized vendors, e.g., grocery
stores, pharmacies, and WIC-only stores (stores serving WIC participants only). The vast majority of
WIC State agencies use the retail food delivery system to provide supplemental foods to WIC
participants. Exceptions include Mississippi which distributes food directly to participants from State-operated
warehouses, Vermont which uses a home food delivery service to provide WIC participants
with foods, and Ohio which provides home food delivery to participants in some counties. Alaska
uses direct distribution of foods when participants live in areas without access to retail vendors.
In the retail fjod delivery system food instruments (FIs) are issued to participants in the form of a
check or a voucher.' These FIs must be used within 30 days of issuance. Local WIC clinics issue
the FIs to participants. FIs may combine a variety of food categories; for example, a participant may
be issued certain FIs that authorize the purchase of milk, eggs, cereal, and juice, and other checks that
allow the purchase of peanut butter and cheese. Within each food category, participants are given a
1 Some States use a check system in which food instruments are processed like a check through the private banking system;
other State agencies use a voucher system in which food instruments are processed by the State, which acts as its own bank. In
Wyoming their check system is being replaced with an electronic benefits transfer (EBT) system that requires participants to
use a "smart card" (a card containing a computer chip) to purchase WIC foods.
choice of food products. For example, participants can select different types of cheese, such as
Colby, Cheddar, or Swiss, or various brands of WIC-approved cereals. Participants must redeem
their FIs at vendors that are authorized by the State or local WIC Program. Some States operate a
"vendor-specific" retail food delivery system, in which participants are required to select a single
vendor and to transact their FIs at that vendor. Other States operate an "open" system, in which
participants may use their FIs at any authorized vendor. Among geographic State agencies with retail
food delivery systems, 11 are vendor-specific and 35 operate open systems.
The following section provides an overview of the characteristics ofWIC vendors operating in States
with retail food delivery systems and the requirements under which they operate.
B. Overview of WIC Retail Vendors
State agencies use vendor selection criteria to determine which vendor applicants to select for
program authorization. Selection criteria may include competitive prices, shelf stocking requirements
for a minimum variety and quantity of supplemental foods, and no history ofcompliance problems. If
a vendor applicant meets or exceeds the State agency's selection criteria, it will likely be offered a
vendor agreement By signing the vendor agreement, the vendor agrees to comply with State rules
and regulations.
States face the ongoing challenge ofensuring effective management, oversight, and review of their
vendor population and ensuring adequate participant access to program benefits. State agencies
monitor their vendors for compliance with their vendor agreements. To accomplish this, States have
developed sophisticated vendor management systems designed to ensure that vendors comply with
the Program's rules and regulations and to prevent or detect vendor fraud and abuse.
In some cases, WIC vendors and participants mayjoindy be involved in violating the Program's rules,
such as exchanging FIs for unauthorized food items, non-food items or cash. In other cases, the
vendor is solely responsible for violating program rules, such as overcharging die WIC Program and
not following proper transaction or redemption procedures.
Some of the approaches used by States to manage vendors include strict vendor selection criteria,
comprehensive vendor training, routine monitoring, high-risk vendor identification systems, and
compliance investigations including compliance buys and inventory audits.
C. Potential Uses of the Data from the Vendor Management Study
Data collected and analyzed during the Vendor Management Study can be useful to Federal and State
officials in evaluating the extent to which vendors are complying with WIC transaction and redemption
procedures. Key areas in which these data may be useful are described below:
1. Quantifying the Level of Vendor Administrative Errors
States are required to ensure that proper procedures are followed when a participant transacts a FI to
obtain authorized WIC foods. Because a WIC transaction is a somewhat complicated process, a
number of errors can occur. Reliable estimates of the frequency of errors may assist FNS and States
to determine priorities for targeting limited vendor management resources. Data from this study will
provide estimates of the types of adininistrative errors being committed, as well as the extent to which
these errors can be associated with vendor overcharges and vendor acceptance of participant-initiated
substitutions.
2. Identifying Administrative Practices on Which Vendor Training Should be
Focused
The data from this study will enable FNS and State officials to identify the administrative errors that
occur most frequendy. In addition, data collectors were asked to report whether there was any
indication as to the cashiers' lack of familiarity with conducting a WIC transaction, such as asking a
co-woiker or manager for assistance, or making a statement that they were not sure how to conduct
the transaction. These data were collected to determine the extent to which cashier familiarity with the
WIC transaction affected administrative errors or other program violations. Officials may review this
information and use it to determine how best to strengthen or revise vendor training objectives and
programs.
3. Identifying Vendor Demographics Associated with WIC Program Compliance
States have implemented a number of strategies to maintain integrity in the WIC Program By having
solid data on which vendor demographic profiles are most associated with vendor errors, FNS may
encourage States to review their own vendor selection criteria and replace any ineffective criteria with
ones that are more likely predictors of fraud and abuse. This may also assist States with prioritizing
their compliance efforts. Reliable information may assist State directors in better managing the WIC
Program and allow them to develop better methods to detect and prevent vendor fraud and abuse.
D. Overview of the Study Methodology
Data for this study was collected at WIC-authorized retail veudors selected from a nationally
representative, probability sample ofWIC retail vendors. Data collectors were females of
childbearing age and belonged to one of die racial or ethnic groups of customers that regularly shop at
the vendors. An intensive training session was held for data collectors and comprehensive training
materials were distributed. A data collection form, called a compliance buy form (CBF), was used to
record data. A copy of this form can be found in Appendix ft.
Data collectors conducted compliance buys that consisted of obtaining WIC authorized and
unauthorized foods using a WIC food instrument at WIC retail vendors. Three buys were typically
conducted at each vendor over a period of one month. The first buy was a safe buy during which the
data collector obtained only WIC-authorized items and attempted to obtain all food items listed on the
food instrument During the second buy, the partial buy, the data collector purchased only a portion
of the foods listed on the food instrument. The third buy was a substitution buy, during which the data
collector attempted to obtain an unauthorized food product or item with the Fl. During all three buys
the data collectors attempted to capture the actual shelf price of the food being obtained. After
conducting a buy, the data collector completed the CBF and donated the items obtained with the FI
to a charitable organization, as designated by each WIC State agency.
The CBFs were reviewed by staff and entered into a dataset. The amount for which each FI was
redeemed was entered into the dataset and linked to a buy according to FI number. The dataset was
thoroughly reviewed, cleaned, and analyzed.
E. Organization of this Report
Chapter II describes the study methodology in greater detail and describes the statistical analysis
methods employed in the data analysis. Chapter III outlines the demographic characteristics of the
vendors selected in the sample. Chapters IV, V, and VI, discuss the study findings. The final chapter
summarizes the report and describes issues that may require further research.
This report is designed to highlight the findings of the WIC Vendor Management Study. Thus,
graphics and tables are designed to best describe a finding, and often contain only information relevant
to the particular finding being presented. Detailed data tables, including distributions, frequencies,
standard errors and t tests, supporting the findings of this study are contained in the appendices.
CHAPTER n
Study Methodology
This chapter describes the methods and procedures employed to identify a nationally representative
sample of WIC-approved food vendors; to prepare for and execute the data collection activities; to
process the collected data and develop an electronic data file; and to establish statistical weights for
each sampled vendor. In addition, the methods by which data were analyzed are described. The
detailed sampling plan is provided in Appendix G.
The 1998 WIC Vendor Management Study employed a nationally representative probability sample
ofWIC vendors. The sampling frame was constructed from complete lists of vendors provided by
the WIC State agencies. A cluster sample of 1,800 vendors in 100 primary sampling units (PSUs)
was selected. The response goal was to obtain complete study data from three compliance buys with
at least 1,500 vendors. After sample loss for vendors that were under State investigation, out of
business, or no longer authorized, 1,625 eligible vendors from the original cluster sample were
selected for the study. Complete study data for three compliance buys was obtained from 1,565 of
these eligible vendors.
A. Defining the Survey Population
The population of interest for trie study was defined as all vendors operating in States with retail food
delivery systems. Excluded from the study were States with direct delivery systems (Mississippi),
home delivery systems (all of Vermont and part of Ohio), State-run WIC vendors (parts of Illinois),
military commissaries, and pharmacies which only provided WIC participants with exempt infant
formula or WIC-eligible medical foods. Vendors in Alaska, Hawaii, Puerto Rico, and the U.S.
territories, as well as vendors authorized by Indian Tribal Organization State agencies were also
excluded from the study population. This decision was made because ofthe high cost of collecting
data from these areas and alternative food delivery systems, which only provide food benefits to a
small fraction ofprogram participants.
The study sample was designed to meet the precision constraints of estimating national proportions
within 3 percentage points and estimating subgroup proportions within S percentage points, with 95
percent confidence. A total sample of 1,500 vendors was expected to meet the study's precision
requirements at the most reasonable data collection cost Clustering the sample of vendors within 100
primary sampling units, which were counties or groups of counties, limited the number of data
collectors who had to be recruited and trained to conduct the survey, and reduced their travel costs.
B. Constructing the Sampling Frame
1. Obtaining Lists ofWIC Vendors
In January 1998, current lists of authorized retail vendors were requested from 46 States and the
District of Columbia. In addition to vendor name and address, information about WIC monetary
redemption amounts was also obtained for use in stratification. States were asked to identify any
home food delivery contractors, State-run vendors, military commissaries, ar d pharmacies that only
provide exempt infant formula or WIC-eligible medical foods. The vendor lists were received from the
States during the period from February through April 1998. Virtually all lists obtained were in
machine-readable formats.
The vendor lists were standardized to adjust for fonr&tting differences across States. Edit checks
included comparing the number of vendor; per State and the reported average monthly redemption
dollars for each State with similar past irformation for reasonableness. Questions and problems noted
in editing the frame information were raised with the States. Clarifications obtained were used to
update the frame file.
It was necessary to determine the county in which each vendor was located to complete the sampling
frame. Since most of the States did not identify the counties on the vendor lists provided, county
location was imputed based upon the zip codes in the vendor mailing addresses. A small number of
vendors with addresses outside the State were attached to nearby in-State counties. Vendors
identified as home food delivery contractors, State-run vendors, military commissaries, and
pharmacies providing only exempt infant formula or WIC-eligible medical foods were not included in
the vendor frame. Reported redemption dollars covering more than one month were converted to
one-month equivalent amounts. The final vendor list contained a total of41,007 W1C vendors.
2. Constructing Primary Sampling Units
Primary sampling units (PSUs) were defined as either individual counties or as groups of
geographically contiguous counties. Since comparisons were planned for different State vendor
management practices, PSUs also had to be defined so that each one included vendors only from a
single State. The number ofWIC retail vendors was determined for each county and used to assure
that each PSU in the sampling frame contained at least the target number of 70 vendors. The District
of Columbia and each county within the 46 study States were included in one, and only one, WIC
PSU. Counties with fewer than 70 WIC retail vendors were combined with geographically adjacent
counties to form PSUs that met or exceeded this minimum requirement.
A computer program using geographic information system (GIS) data was used to form PSUs. The
program allowed the user to group adjacent counties into PSUs within a State until each PSU
contained at least the minimum number of vendors. The program displayed the number of WIC
vendors in each county on a State-level county outline map. In order to form practical PSUs for field
visits, major highway routes were also shown on the computer screen. A highway atlas was used to
identify major mountain ranges, lakes, and other map features.
There were only a few cases where all of the PSU construction objectives could not be met The list
from the District of Columbia contained only 21 WIC vendors; it was combined with two adjacent
Maryland counties to form a PSU with 89 total vendors.
This PSU was included in the stratum for vendor-specific States wiii high participant-to-vendor
ratios. Delaware had only 67 vendors; in this case, the entire State wa.~ defined as a single PSU. In
total, only seven of the 366 PSUs in the sampling frame contained fewer than 70 vendors each. Thus,
the final WIC PSU sampling frame contained 366 PSUs that were contiguous geographic areas;
which in most cases contained at least 70 WIC retail vendors; which did not cross FNS region
boundaries; and that, with one exception, did not cross State boundaries. Each WIC retail vendor
was associated with only one PSU in the WIC sampling frame.
C. Selecting the Sample
A nationally representative sample of 1,800 WIC retail vendors was initially selected to complete the
study. The study team first selected 100 sample PSUs from the total of 366 available PSUs. Within
each of the 100 PSUs identified, a total of 18 sample vendors were selected for inclusion in the study,
for a total of 1,800 vendors. Because there were likely to be vendors selected in the sample that
either were no longer in business or were no longer authorized by the State to accept WIC FIs, a
backup sample of two vendors per PSU was also identified.
1. Stratification Variables
PSUs in the sampling frame were stratified to reduce sampling variability and to ensure adequate
sample sizes for key analysis comparisons. The PSUs in the sampling frame were stratified based on
the following three variables:
• Vendor-specific vs. open-food instrument systems;
• States with high, medium, and low vendor-to-participant ratios based on FY 1995 Vendor
Activity and Monitoring Profile (VAMT) data; and
• Metropolitan location - within a metropolitan statistical area (MSA) or not within .a MSA
(based on the largest population county within the PSU).
10
FNS was interested in comparing groups of States by their vendor management practices, which
included contrasting States with large and small numbers of WIC vendors (based on the proxy of
vendor-to-participant ratios). State-level vendor and participant counts from the fiscal year 1996
VAMP report were used to divide the population ofWIC vendors into three approximately equal
sized strata, based on the average number of participants per vendor for each State.
To determine whether fewer violations typically occur in States with vendor-specific FI systems, there
was interest in comparing States using vendor-specific and open-food delivery FI systems. Crossing
these two State-level stratification variables defined six primary strata. Appendix G displays the
States that were assigned to each of the six primary strata, the number of vendors in the sampling
frame, and the average State-level vendor-to-participant ratio. Also displayed in Appendix G is the
distribution of the vendcs in the sampling frame by the same six strata.
In addition, it was also important to control the sample of PSUs by whether or not they were located
in a metropolitan area PSUs in the sampling frame were classified as metropolitan if the largest
population county of the PSU was part of a metropolitan statistical area (MSA). PSUs that were
entirely composed ofnon-MSA counties were classified as non-metropolitan.
2. Selecting the Sample PSUs
Vendor-specific States included only about 20 percent of the vendors in the sampling frame. Equal
overall selection probabilities would have led to selecting about 20 PSUs in these States and obtaining
complete study data for only about 300 vendors from vendor-specific States. To meet the precision
constraint for this analysis domain, sample PSUs in the vendor-specific States were sampled at twice
the rate used for the States with open-food-instrument systems. This oversampling was implemented
by adjusting the PSU size measures (number ofWIC retail vendors) prior to selecting the sample
PSUs.
The sample of 100 PSUs was selected using probability non-replacement sampling with probabilities
proportional to size of the PSU. The size of the PSU was proportional to the number ofWIC
vendors in the PSUs, except for the 2:1 over-sampling in those States using vendor-specific WIC
food instruments.
The PSUs within each stratum were sorted by their metropolitan status prior to selecting the sample
PSUs, effecting an implicit stratification by metropolitan status. A probability minimum replacement
selection procedure developed by Chromy (1979) was used to select 100 sample PSUs. The
method allows multiple hits for those units whose expected sample size exceeds unity and restricts the
realized number ofhits for each unit
3. Selecting the Sample Vendors
Following the selection of 100 sample PSUs, a probability sample of 1,800 vendors and a 200-
vendor reserve sample was selected. First, a total sample of 20 vendors was selected from the
vendor list within each of the 100 PSU sample selections. New York City, Los Angeles County, and
San Diego County were multiple-hit PSUs, meaning more than one PSU was selected within their
boundaries. Prior to the selection, vendors within each PSU were sorted by their monthly WIC
redemption dollar amounts.
The 20 vendors were selected within each PSU using ^ - kriatic sampling with equal probabilities and
without replacement, effecting an implicit size stratificatio, i of the vendors. Then 18 ofthe 20 selected
vendors within each PSU were randomly selected for the study sample, yielding a main study sample
of 1,800 vendors and a 200 vendor reserve sample. The names and addresses of the 1,800 sample
vendors, except those identified by States as out-of-business, no longer authorized, or under State
investigation, were sent to the field for compliance buys.
12
D. Developing the Data Collection Instrument
A draft data collection instrument, called a compliance buy form (CBF), was developed which
contained individual data elements to be collected St the vendor. The CBF was pretested in the
Raleigh-Durham area of North Carolina on April 21-22,1998. Sixteen WIC vendors were selected
for one compliance buy each Selections were made to ensure a cross-section of vendor types,
including inner-city, suburban, rural, large, small, chain, and independent. Four data collectors were
assigned to complete four different kinds of buys each These included:
• A safe buy, where the buyer obtained all of the food items listed on the FI;
• A partial buy, where the buyer omitted some of the food items listed on the FT;
• A minor substitution, where the buyer attempted to obtain an unauthorized food item
within a WIC food category (e.g., unauthorized cereals or juices); and
• A major substitution, where the buyer attempted to obtain an unauthorized item not within
the WIC food categories (e.g., pasta instead of cereal).
After the data collectors finished their assigned compliance buys, donated the items purchased, and
completed the CBF, they participated in a pretest debriefing on April 22. Several recommendations
for improving the wording of certain questions and for facilitating accurate form completion were
obtained at the debriefing. These recommendations were incorporated into a revised CBF that was
submitted for final approval. Following several iterations to enhance the content and appearance of
the CBF, it was finalized in July 1998 (see Appendix H).
13
E. Assembling the Data Collection Staff
Several unique challenges were presented in assembling a national staffof "compliance buyers," the
title assigned to data collectors for the 1998 WIC Vendor Management Study. To successfully
perform the required compliance buys, it was essential that the buyers reflect the physical
characteristics ofwomen who receive WIC benefits. This meant, for example, that all buyers had to
be females of childbearing age. In addition, ifdata collectors were to perform their assignments
without creating suspicion among vendors, it was also necessary for them to belong to one of the
racial or ethnic groups of customers who regularly shop at those vendors. A total of 103 data
collectors and six field supervisors were recruited during July and August 1998.
F. Training the Field Staff
During June, July, and August 1998, draft training manuals and other materials necessary to ensure die
application of standardized data collection procedures were developed. Among the documents
prepared were the following:
• Compliance Buyer Manual;
• Field Supervisor Manual;
• Compliance Buyer Pre-Training Study Package;
• Field Supervisor Training Agenda; and
• Compliance Buyer Training Agenda
All field supervisors (6) and data collectors (103) were required to attend and complete a three-day
training program in Raleigh, NC. To reduce the trainer/trainee ratio to an effective level, two training
sessions were conducted during successive weekends in late August 1998. Halfthe field staff
attended the first session; the other half attended the second.
14
One noteworthy component required each trainee to complete a "practice buy" at an authorized WIC
vendor in the Raleigh area. On the final training day, trainees were required to locate and travel to the
specified vendor, correctly conduct the compliance buy, properly complete the CBF, and report back
to their field supervisor, who in turn reported to a data collection manager. Following successful
completion of the practice buy and all other training requirements, trainees were certified as
"compliance buyers" and declared ready to begin their field assignments.
G. Equipping the Field Staff
Each WIC State agency included in the sample was contacted to determine the appropriate food
package to be included on a series of three food instruments for each sampled vendor and also to
inform the State of data collector names to be imprinted on the FI. WIC State agencies issued the FI
in the quantities required. Food instrument serial numbers were entered into a database. Food
instruments were designated for use at a specific vendor and for the exact compliance buy for which it
was to be used For example, "Compliance Buyer #88335 will use FI # 987654321 at vendor
#1234 for comp'i>nce buy #3A (minor substitution)." This information, along with the food
instruments, was sent to the subcontractor on a flow basis, as food instruments were received from
States.
Six thousand CBFs were printed, three each for the 1,800 sampled vendors and 200 reserve
vendors. To simplify the data collector's role to the maximum degree possible, each form was pre-printed
with the following identifying information:
• Vendor name, address, and zip code;
• Four digit vendor number (first two digits identified the PSU number, last two digits identified
the vendor number within the PSU);
• Food Package (woman, infant, or child);
• Type of Buy (safe, partial, minor substitution, or major substitution); and
• FI serial number.
15
In addition to the above information, the foods listed on the FIs assigned for each compliance buy
were manually pre-entered, along with their quantities and sizes (e.g., "Similac With Iron: 15,13-
ouncecans")- The correct FI to be used was also attached to each CBF. Each CBF was prepared
through a process designed to eliminate decision-making by the field staff, which substantially
enhanced data accuracy, and facilitated standardized buying procedures at all sampled vendors.
Data collectors were equipped with other WIC materials that enabled them to complete their
purchases without arousing suspicion by vendor staff. States issued valid WIC identification cards for
buyers, provided official copies of approved food lists, and identified charitable organizations to which
data collectors could donate the items purchased.
A monetary advance was given to each compliance buyer for a cash purchase of $5 or less of non-
WIC items. This procedure was implemented to replicate the normal buying patterns of WIC
participants. Items purchased with cash were also donated to charitable organizations.
Each compliance buyer was responsible for the completion of three compliance buys at each assigned
vendor. Data collectors, on average, were assigned 16-18 vendors, although some had considerably
more and a few had less. Three buys were attempted at each vendor. The third buy was either a
"Buy 3A" or a "Buy 3B," as preprinted on the CBF. The three assigned buys at each vendor were
performed as follows:
Buy #1: Safe Buy Buyer purchased all foods listed on the food instrument in
the quantities and types listed.
Buy #2: Partial Buy Buyer attempted to purchase some, but not all of the
items listed on the food instrument.
Buy#3A: Minor Substitution Buyer attempted to substitute an unauthorized food item
within an approved food category.
Buy #3B: Major Substitution Buyer attempted to substitute an unauthorized item clearly
outside an approved food category.
16
WIC FIs could be used only during a specified 30- or 31 -day period. Data collectors were required
to complete all three buys within the transaction period printed on the food instruments. To a^oid
arousing suspicion among vendor staff, buyers were instructed to allow five or more days between
buys at each sampled vendor.
The primary tasks associated with a compliance buy entailed selecting the correct food items for the
type of buy being undertaken, obtaining the shelf price of each item, presenting the FI at the checkout
counter, and observing any violations ofWIC program procedures. In addition, buyers purchased
less than $5 ofnon-WIC items with cash. Immediately after the compliance buy, and away from
vendor premises, buyers completed the CBF on which they recorded all pertinent details associated
with the WIC transactioa All items purchased, WIC and non-WIC, were donated to charitable
organizations, with one exception: in several States, buyers delivered their purchased infant formula to
local WIC clinics.
Data collectors sent completed CBFs to their field supervisors twice a week. Field supervisors
reviewed and approved the CBFs and shipped them via overnight freight for processing. In addition
to sending CBFs, buyers reported progress to their field supervisors on a weekly basis, and
supervisors, in turn, reported weekly to the data collection manager.
Data were collected during the Fall of 1998. The 30- or 31-day purchasing period varied, depending
upon when States issued their food instruments. More than 90 percent of all compliance buys were
completed during September and October, and all data collection was completed by mid-December.
H. Quality Control
Early in the data collection period, quality control teams made visits, some announced and some
unannounced, to several data collectors to verify that the standardized buying procedures were being
implemented and to debrief the buyers to determine whether any adjustments were in order.
17
Prior to conversion of handwritten information contained on the CBFs to an electronic data hie, each
CBF received three levels of review. Field supervisors reviewed and approved each of their data
collectors' CBFs before sending them for processing. The data collection managers then reviewed
and approved each CBF received from their field supervisors. Finally, data receipt staff edited each
CBF for consistency and legibility before sending it on for keying. Any CBF failing approval at any of
the three levels was returned to the data collector to be corrected.
Data entry was performed with 100 percent verification; that is, each CBF was independentiy keyed
by two keyers. If both entered identical data, the system accepted the CBF as complete. Ifany
differences arose between keyers, the system required successful resolution of the problem before the
CBF was accepted into the data file.
FI redemption data were received electronically from some States, while others sent the processed
FIs, which then needed to be read by a contract bank. Some States produced several files as FIs
were processed periodically. Eventually, all files were merged to create a combined redemption file,
which contained the serial number and amount of each FI, as well as date of redemption and the
State-assigned WIC vendor number, if provided by the State.
After the CBF data were keyed, a computer program checked for errors and inconsistencies and
calculated numerous variables from the data on the CBF (e.g., the product of the quantity and shelf
price for each food item purchased, sum of the cost of all items in the purchase table). All CBFs with
inconsistencies were pulled and manually reviewed. Following resolution of data inconsistencies, the
file was corrected.
I. Survey Weights
The initial sampling weights for the 1,800 selected vendors were calculated based on the expected
PSU sample sizes and the conditional vendor selection probabilities. If complete study data were
obtained for all of the sampled vendors, these unadjusted weights would be appropriate for analyzing
18
the survey results. This was not the case, however, as some vendors were found to be ineligible for
the survey and it was not possible to complete all of the proposed data collection activities for Khers
(see Table n-1).
Table III.
Vendor Eligibility Categories for All Vendors Included in the Sample
Eligibility Categories Vendors Percent
1. Out ofbusiness at first buy attempt 20 1.1%
2. Not authorized to accept WIC at first buy attempt 27 1.5%
3. Dropped - under State Investigation or Sanction 127 7.1%
4. Other non-eligible 1 0.1%
5. Eligible for inclusion in the study 1,625 90.2%
6. Total Sample Vendors 1,800 100.0%
The response rate for the 1,625 eligible vendors was determined for each of three buys (see Table II-
2).
Table D-2.
Study Response Rates for All Vendors Included in the Study
Study Response Vendors Percent
1. Completed buy 1 (safe buy) 1,600 98.5%
2. Completed buy 2 (partial buy) 1,594 98.1%
3. Completed buy 3A or 3B (substitution) 1,580 97.2%
4. Completed all 3 buys 1,565 96.3%
19
Ineligible vendors were identified at fee time of the first buy attempt, and their adjusted sampling
weights were set to zero. The eligible in-sample vendors were partitioned into eight weighting classes
(see Table D-3) so that those within each weighting class were as similar as possible. The weighting
classes were defined using the State-level stratification variables of metropolitan classification, type of
FI system, and ratio of WIC Vendors to WIC participants.
Table H-3.
Weighting Class Categories for All Vendors Included in the Study
Class Metro Food Instrument
System
Vendor/Participant Ratio
1 Metro Open Low
2 Non-metro Open Low
3 Metro Open Medium
4 Non-metro Open Medium
5 All Open High
6 M Vendor-specific Low
7 AU Vendor-specific Medium
8 AH Vendor-specific High
The metropolitan classification variable was not used to subdivide classes 5 - 8 into separate
weighting classes because the number of non-metropohtan vendors responding would have been too
small, which could possibly lead to unstable adjustments for non-response.
20
The weights for the eligible in-sample vendors were adjusted by multiplying the initial weights for each
vendor in weighting class-k (where k = 1,2,..., 8) by the ratio R(k) where:
R(k) = [sum of initial weights for eligible vendors in weighting class
k]/[sum of initial weights for all completed eligible vendors in weighting
class k].
This weighting class procedure adjusts the sum of the survey weights, to compensate for those eligible
vendors for which complete survey data were not obtained (i.e., those in which the compliance buys
were not completed).
To the extent that the responses of respondents and non-respondents within the same weighting class
tend to be similar, the adjustment procedure reduces missing data biases.
Several weights were computed to facilitate the planned analysis. The weighting class methodology was
applied separately to compute adjusted survey weights (see Table D-4). A detailed description of the
weighting procedures used in this study may be found in Appendix G.
Table EM.
Adjusted Survey Weight Categories
Weight Used for Analysis of:
WTBUY1 Data from buy 1 (safe)
WTBUY2 Data from buy 2 (partial)
WTBUY3A Data from buy 3A (minor substitution)
WTBUY3B Data from buy 3B (major substitution)
WTBUYS Data from all 3 buys
21
J. Overview of Statistical Analysis Methods
The primary purpose of the WIC Vendor Management Study was to describe program violations
committed by WIC vendors. In order to examine each of the areas described in Chapter I, two
approaches to a quantitative description were used:
• An examination of how WIC vendors conducted the transaction in response to the
compliance buys. The response of the vendors as it relates to properly conducting a WIC
transaction were examined, particularly in regard to the vendors' disposition to overcharge or
undercharge, commit administrative errors, and allow buyers to make minor or major
substitutions. These statistics take the form of frequencies and distributions, showing the
vendors' actions over a wide variety of demographic variables.
• The relationship, if any, between a WIC vendor's improper conduct of a WIC
transaction and variables associated with State vendor management systems.
Multivariate analysis was conducted to examine whether store demographics (e.g., store size,
locale) or State vendor system demographics (e.g., vendor-to-participant ratio, open versus
vendor-specific) had any statistically significant relationship to vendor practices.
A total of 36,754 vendors are represented in the analysis. These vendors represent the survey
population: all WIC retail vendors in the 48 contiguous States and the District of Columbia excluding
States with direct distribution food delivery systems (Mississippi) and home food delivery systems
(Vermont and parts of Ohio).
In addition, after the sample was drawn, it became apparent that one additional State needed to be
excluded. The North Dakota State Agency utilizes a system by which all of the milk issued to a
participant is placed on a single FI. Ifparticipants choose not to purchase all ofthe milk at one time,
they are given a special "raincheck" by the store which allows them to return at a later date to pick up
thefr remaining milk. Because the methodology used for this study nxjuired that different types of
buys must be conducted at each visit, and because the approach used by North Dakota was
determined to be unique, vendors in a single PSU located in North Dakota were excluded from the
study.
22
As previously described, weights were calculated for vendors where each type of buy occurred and
for vendors where all three buys were completed. When results are described as computed across
all buys this indicates that calculated weights are describing vendors where all three buys were
completed. Since this type of analysis essentially multiplies the number of vendors by three (since
there were three buys made at each vendor), results for each vendor were averaged (divided by
three) in order to generate an estimate that for each vendor reflects the actual number of vendors who
participated in this study. For example, if a vendor had insufficient stock during one of three buys that
result (i.e., insufficient stock) was divided by three.
Totals generated in this context are labeled as "average totals" to refer to this process. In the
instances for which totals were generated for each buy type, these totals are simply labeled as totals.
1. Descriptive Analysis
Descriptive analysis used for this study entailed use ofcommon summary statistics, mostly estimated
frequencies, standard errors associated with weighted estimates, percentages and the standard error
of percentages. The focus ofattention in this analytic context is restricted to a description of the
proportion ofvendors mat can be categorized as problematic (i.e., the percentage of vendors
committing errors or violations ofWIC program procedures). The statistics reflecting compliant
actions on the part of the vendor are often omitted but can easily be determined by subtracting the
non-compliant responses from the total number of responses.
In this study the descriptive analysis addressed the frequency of occurrence and percentage of
vendors who overcharged, undercharged, and committed various administrative errors and other
recognized violations ofWIC program procedures. The violations were further examined as a
function of demographics, other types of errors, and other types ofcommon practices. For example,
after describing the distribution of overcharges, the frequency of an overcharge as a function of vendor
size or location is described. Similarly, the distribution of an overcharge as a function of administrative
errors (e.g., insufficient stock and violation ofthe FI countersignature procedures) was addressed.
23
Finally, differences in fee frequency of a violation as a function ofcommon vendor demographics, such
as use of scanning equipment and provision of receipts, were examined.
Differences in W..C vendor responses to the aforementioned factors were also subjected to statistical
testing, most often t statistics derived from contrast analysis. A significant t statistic indicates that the
difference in proportions between various variable categories were probably not due to chance or
random fluctuations. Levels of significance for this study are set at 0.0S and 0.01. Throughout this
report * and ** denote 0.05 and 0.01 levels of significance, respectively.
Essentially, these t statistics, like the more commonly used chi square statistic, effectively describe a
difference ofproportions test For variables that have more than two categories, contrast analysis
was especially advantageous (compared to the chi square analysis) since it permitted comparison of
specific variable values or categories. For example, in the comparison of overcharge as a function of
food package (FI type), a contrast analysis permitted the specific comparison of infant food packages
to the woman food packages, and/or each of those to die child food packages, while the chi square
simply indicates that there is a significant difference in overcharge as a function of food package.
Although it is always possible to simply describe the distribution of overcharge as a function of specific
food package categories, a contrast analysis enables ready identification of which differences are
statistically significant For a simple two-by-two comparison, results obtained by contrast analysis
and chi square were essentially equivalent
2. Multivariate Analysis
Multivariate statistical techniques facilitate identification of relevant associations between variables.
For example, when examining a variable such as vendor overcharge, it is of interest to analyze
whether or not there is a relationship between die size of a vendor and their proclivity to overcharge.
Through the use of multivariate statistical techniques, it is possible to estimate both the relationship
between one variable and another, such as the raw number of vendors of different sizes who
24
overcharge, as well as the extent to which a variable such as vendor size contributes to overcharging
(statistical significance).
It us also possible to develop multivanate models. For example, State agency vendor managers may
be interested in predicting the likelihood that a vendor will overcharge on a WIC transaction. The
State might predict whether a vendor will overcharge with one variable (e.g., vendor location), but a
more accurst* prediction may be made if more than one variable is analyzed (e.g., vendor location,
vendor history ofovercharging WIC, and vendor's use of scanners). Therefore, multivariate models
examine the extent to which a number of variables combine to predict an outcome. In the above
example, it may be found that the presence of three variables (e.g. rural vendors of small size who do
not use scanners) may be highly predictive of whether a vendor will overcharge.
In this study, multivariate statistical techniques were used to derive estimates of the relative
contribution of assorted independent variables (e.g., vendor demographics such as locale, size, and
vendor-to-participant ratio) to some dependent variable (e.g., overcharge, undercharge,
administrative error, or other program violation). The model building approach used entails
comparison of single variable models developed in accordance either with past performance in similar
studies, or hypothesized performance in this context Selected variables were then concurrently
entered into a new equation. Variables were eliminated from consideration in multivariate models
when it was determined that they lacked predictive power, had no conceptual justification for
retention, or appeared methodologically problematic for some other reason (e.g., questionable
reliability and/or validity).
Several criteria were used to evaluate model performance. First, consideration was given to
parsimony. The simplest models (with the fewest number of variables), explaining substantial amounts
of variation in the dependent variable of interest, were preferred Consideration was also given to the
dynamic interplay between component variables. Identification and discussion of various interactive
processes that affected the interplay between variables also constituted a focus ofattention.
25
Multivanate results are presented after the discussion of a descriptive analysis of overcharge by
vendors. Discussion of single variable models is introduced as a preamble to more complicated
model building. These a> aiyses can facilitate discussion of measurement problems that may be inherent
to a given variable. In other words, some variables may have proven to be ambiguous or unreliable as
valid measures. As previously discussed, such variables could be eliminated from further
consideration or continue to receive qualified consideration in subsequent model development In
addition, ancillary analysis, designed to elucidate the significance of variable interactions, is introduced
as necessary.
Results of the data analysis are detailed in the subsequent chapters. To facilitate displaying of the
findings, each chapter contains text descriptions and graphic representations. Detailed data tables
supporting the findings are included in the appendices.
26
CHAPTER ffl
Demographics of the Study Population
This chapter describes the demographic variables of the vendors selected for inclusion in the study.
Data to support each of the variables described were collected through a variety of means. For
example, data used to determine the location of the vendor (i.e., metropolitan/non-metropolitan) were
collected when developing the sampling plan through use of zip codes. Data regarding whether the
vendor operated in a State using a vendor-specific or open system were obtained from the States.
Finally, much of the data describing the vendor were collected in the field by the data collector at the
time of the compliance buy. Table ID-1 below displays the major variable categories and the data
sources for each.
Table mi.
Source of Demographic Variables Included in the Study
Variable Source
Open vs. vendor-specific FI system State Plans
Vendor location (Urban/Rural)
Vendor redemption volume (Vendor size)
Vendor Lists and Activity Reports From State
Number of cash registers
Vendor use of scanning equipment
Cashier familiarity with WIC transactions
Vendor stock levels
Provision of receipt
Field Data Collection
27
A. Metropolitan and Non-metropolitan Areas
The sampling plan was designed to ensure vendor representation from both metropolitan and non-metropolitan
geographic areas. Figure IQ-1 displays the distribution of the study vendors with regard
to their locale. As can be seen from the figure, 70 percent ofthe WIC vendors surveyed were
located in metropolitan locations (see Table A-1 in Appendix A).
Figure 111-1.
Distribution of WIC Vcadon Included la the Stady by Locale
MrtmnnliMn Area Nfm-irwtrrinnliMn Arra
B. Vendor Classifications
Ofthe vendors selected for the study, 97.8 percent were classified by States as "grocery stores"
while 2.2 percent were classified as "pharmacies" (see Table A-3 in Appendix A). Pharmacies were
included in the study as most States permit pharmacies to provide infant formula to participants. For
purposes of analysis and weighting, pharmacies were treated the same as any other WIC vendor
visited by a data collector.
28
C. Vendor-Specific Versus Open Food Instrument Systems
One of the key variables examined in the study was the relationship between the type of food
instrument system selected by die State and die extent to which it might contribute to vendor
violations. As indicated in Chapter II, vendors located in vendor-specific States were oversampled to
ensure adequate representation. Figure DI-2 shows the distribution between vendor-specific and
open system States selected in die study sample (see Table A-2 in Appendix A).
wj—mi
Distribution of WIC Vendors Included in the Stady by Type of Food Instrument System
' from * VrnrinrrSpfrifk.
D. Yendor-to-Participant Ratio
One of die important factors that some States consider when authorizing vendors is the overall ratio of
vendors to participants. FNS has traditionally believed that having fewer vendors to manage is a key
factor in improving State vendor management systems. To examine whether or not the vendor-to-participant
ratio has an impact on vendor practices, data was collected from the 1996 VAMP report
to establish vendor-to-paiticipant ratios for all of the study States. Vendor-to-participant ratios were
29
then divided into four categories, each with approximately 25 percent of the study population (see
Table A-4 in Appendix A). Figure IE-3 displays the distribution of vendors by the four categories of
vendor-to-participant ratio.
me
Figarc III-3.
Distribatioi of W1C Vnrion lacladed ia the Stady by Veador-to-Partkipaat Ratta Categories
I -=112 1:112-157 1:151-192 1:>I92
Vcaaw-te-Tankhwal Rao*
E. Vendor Size
The number of cash registers in the store was used as a proxy for the physical size of the WIC
vendor. A frequency polygon was developed to display the range of vendors with different numbers
of cash registers. This polygon is displayed in Figure HI-4, showing the number of cash registers
ranged from 0 to 32 with a mean of 6.0 registers.
30
FigartIO-4.
.Namber of WIC Vaadors Included in the Study by Number of Cash Registers
700 s
600|
500
e 1■ 1 U 400
5eI
' y/
sKK 1 300
a
Z \/
\
200'r
y
1001" whrtw_ 0*^
I 2 3 4 5 6 7 g 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
Namber of Cask Registers
30 31 32
The number of cash registers was categorized into three levels to create small (0-2 registers), medium
(3-7 registers) and large-sized vendors (more than 8 registers). These size levels were determined
early in the analysis and based on information contained in WIC State Plans, which described
categorization ofWIC vendors by size for vendor monitoring purposes. Figure IE-5 displays the
distribution of vendors by size categories, each of which accounted for about one-third of the vendors
surveyed (see Table A-5 in Appendix A).
31
Heart OI-S.
Ditfribwioa af WIC Vcadon Iachdcd la tac Stwiy by Start Sitt
JJ.4%
~CT
, JU%
ma
Small-Sized Vendor Medium-Sized Vendor Large-Sized Vendor
F. Vendor Use of Scanning Equipment
The vendors' use of scanning equipment when conducting a WIC transaction was also examined to
determine how the use of scanning equipment i« related to program compliance. Ofthe total vendors
surveyed, 27.4 percent of vendors did not have the equipment to scan the WIC transactions.
Another 3.6 percent of WIC vendors had scanning equipment but chose not to use it during the
compliance buys, fhe remaining 69.1 percent of the WIC vendors used scanning equipment during
all three compliance buys (see Table A-6 in Appendix A). Figure ni-6 displays the distribution of all
WIC vendors in the study based upon their use of scanning equipment As shown in Figure HI-?,
among vendors with no scanning equipment, 86.4 percent were small sized vendors, and 02 percent
were large vendors. The number of vendors with scanning systems programmed to flag WIC-approved
foods was not examined in this study.
32
Fbjnre III-6.
Distribation of WIC Vendors lacluded in the Stady by Use of Scanning Equipment
3.4%
tn.iv.
27.4%
1 No Scanning Equipment B Scanned Purchased WIC Items a Did Not Scan Purchased WIC Items |
Figure DI-7.
Distribution ofWIC Vendors Included in the Study with No Scanning Equipment by Vendor Sin
Snuil Medium
Vender 9s*
L*r
33
G. Cashier Familiarity with WIC Transactions
The cashier's familiarity with WIC transactions was also examined While it was not possible to
determine actual cashier experience, the study identified situations that might lead one to conclude that
the cashier may not be familiar with WIC transactions. These situations follow:
• The cashier indicated to the data collector that he/she was a new employee or wore a special
badge indicating he/she was in training;
• The cashier indicated that he/she had never completed a WIC transaction;
• The cashier required assistance from another co-worker to complete the WIC transaction;
and
• The cashier indicated in some other way (e.g., asked buyer what to do or made comments
that indicated lack of familiarity) that he/she was not familiar with the WIC transaction.
Using the above criteria for determining cashier inexperience or unfamiliarity, Figure EQ-8 shows that
8.0 percent of the vendors had cashiers providing some indication they were unfamiliar with the
conduct of a WIC transaction (see Table A-7 in Appendix A). Ofinterest also was how the data
collector determined that the cashier was not familiar with a WIC transaction. Figure HI-9 displays
the percentage of vendors with a cashier indicating a lack of familiarity with a WIC transaction, by the
type of indicator identified by the data collector (see Table A-8 in Appendix A).
34
Finrem-8.
Distribution of Vendors Included in the Study by Cashier's Indication of Inexperience
with W1C Transaction Procedures
«%
92%
'indication of Inexperience with W1C Transaction 1 NO Indication of Inexperience with WIC Transacoot
Figure Hl-9.
Distribution of WIC Vendors Included in the Study with a Cashier Indicating a Lack of Familiarity
with WIC Transactions by Means of Indication
>
•s
80%
70%
60%
50%
40%
30%
20%
10%
Indicated was a New
Employee
Indicated Had Never
Completed a WIC
Transaction
Received Assistance from a
Cc-worker
(MM
Mwioilnilicatim
35
Also examined was the relationship of vendor size to cashier inexperience with WIC transactions. As
shown in Figure HI-10, the relationship between size of vendor and indication of inexperience with
conducting a WIC transaction was examined. For each of die levels of vendor size less than 10
percent of die vendors had cashiers indicating inexperience with WIC transactions.
25.0V
20.0%'
Fifarc III-I0.
Percentage «f WIC Veadan Included in the Study with Cashiers Iadkatiag UBfaauliarity
Ceadactin; ■ WIC Transaction by Vendor Size
This chapter has presented die demographic profiles of die WIC vendors included in the study.
Additional information can be found in Appendix A In the next three chapters, die relationships
between these demographic variables and the findings related to vendor administrative errors, vendor
overcharges, and vendors allowing substitutions of ineligible foods are examined.
36
CHAPTER IV
Findings Related to Vendor Administrative Errors
In an attempt to curb vendor fraud and abuse, all States establish administrative procedures that the
vendor must follow in order to complete a WIC transaction. While some ofthese administrative
procedures vary slightly from State to State, a limited number of program rules and administrative
procedures are common to all States included in the study. In this chapter, an identified set of
program rules were examined to determine the extent to which vendors were following proper WIC
transaction procedures.
For the purpose of this study, the improper conduct ofa WIC transaction was considered an
administrative error. The following categories of administrative errors were examined:
• Having insufficient stock to allow the buyer to purchase the food items listed on die FI;
• Requiring the participant to sign the FI prior to entering the purchase price;
• Offering rain checks for foods not available; and
• Asking the buyer to pay cash in addition to the FI for obtaining the WIC food items listed on
theFI
In addition, while not considered an administrative error in all States, the study examined the provision
of a receipt to the buyer.
Actministrative errors were examined over three buys at each vendor. Figure IV-1 displays the
overall results for each of the four administrative errors examined Overall, the percentage ofvendors
committing administrative errors over the three buys was relatively low with the exception of following
37
proper FI countersignature procedures (Table B-2 in Appendix B), which occurred among 3S.4
percent of the vendors.
«rwicv<
HpnlV4.
Admtaistratiw Errors by Typ* of Error Acre* All B«y»
1
40%
UK
»%'
av
20%'
1»%
10%
—■
Insaflicicm Slock Failed to Counusiign Before
Price »» Entered
Offered Ramcbcdi Aikcd lo Pay Caia in Addition
10 FI
AdaMMraaviIrnn
As no significant differences in the level of administrative errors were found as a function ofdie type of
buy conducted, analysis for variables by individual buy type was not included. The extent to which
vendors committed each of the above categories of administrative errors are discussed in the
remaining sections of this chapter.
A. Requiring the Participant to Sign the Food Instrument Prior to the
Cashier Entering the Purchase Price
During the compliance buys, the buyers were asked to observe at what point during the WIC
transaction the cashier asked the buyer to sign the FI. Failure to follow proper transaction procedures
was noted if the cashier had the buyer sign the FI prior to ringing up the WIC purchase; had the buyer
38
sign flic FI after the WIC purchase was rung up, but prior to entering the purchase price; or if the
buyer was not asked to sign the FI. This is important because the failure of the WIC participant to
sign die FI after the purchase price was entered was found to be a significant predictor of vendor
overcharge in prior studies.
When estimating the percentage of vendors who failed to follow proper FI countersignature
procedures, two different aspects of this problem were examined. These two methods are described
below:
■ First, the rate of vendor errors was examined The term >ate" was defined as the
percentage of vendors who were likely to violate the FI countersignature procedures on
any single round of buys, over a three-buy period. Because some vendors violate the
procedures only on one round ofbuys, while other vendors violate twice or all three
times, it is necessary to average the number of violations that occurred each of the three
rounds. This lYate" of violation predicts the percentage of vendors who are likely to
violate during any single round ofbuys. The rate of vendor violations ofproper FI
countersignature procedures was 35.4 percent This means that if a single round of
compliance buys were to be conducted on 1000 WIC vendors, 354 would be likely to
violate proper FI countersignature procedures.
■ Second, thefrequency of vendor errors was examined Frequency was defined as
the likelihood that a vendor would violate the FI countersignature procedures one or more
times over a three-buy period. To measure diefrequency of vendor violations, vendors
were grouped into the four categories displayed in Figure IV-2. A total of47.6 percent
of the vendors never violated the FI countersignature procedures, while 24.5 percent
violated once, 18.7 percent violated on two of the three buys, and 9.2 percent violated
the procedures during each ofthe three buys (Table B-2 in Appendix B).
The difference between the rate and frequency is explained by viewing the rate as the likelihood of a
vendor violating the FI countersignature procedures on a single round ofbuys, and the frequency is
the likelihood of a vendor violating the FI countersignature procedures one or more times o\er three
compliance buys. If one wishes to predict the likelihood of a vendor violating the FI countersignature
procedures on any given set of buys, the best method is to use die rate.
39
rifart IV 2.
f «rctni|i «f WIC vc«d»ri by Prcqatacy *t Occarrcacn »l Davialiaa trim Caaaltriigaatart
Pracadarw Acran All Bay
Figure IV-3 displays the percentage of vendors not following proper FI countcrsignature procedures
by locale and by type of food instrument system. For open-FI systems, 78 percent of die vendors
who failed to follow proper FI countersignature procedures were located in a metropolitan locale
while 22 percent were located in a non-metropolitan locale. For vendor-specific FI systems, 922
percent ofthe vendors who failed to follow proper FI countersignature procedures were located in a
metropolitan locale while only 7.9 percent were located in a non-metropolitan locale.
40
100%'
90%"
Fignre IV-J.
Percentage of WIC Vextors Deviating from Coaatersignatare Procedure!
by Food Instrument System aad Locale Across All Bays
—.—_ ■
Metropolitan Noo-metropolitan Metropolitio Noonnetropolii
B. Insufficient Stock of WIC-Authorized Foods
Most States require that WIC vendors maintain a sufficient stock of WIC-authorized foods in order
to ensure that participants can purchase all of their prescribed foods without having to repeatedly
return to the vendor or visit multiple vendors. Insufficient stock violations include not having sufficient
stock of a particular food item (e.g., milk, cereal) to till the participant's FI or not having the correct
brand or proper size as prescribed on the FI. As noted previously in Figure IV-1, 5.5 percent of all
WIC vendors did not have sufficient stock to allow the buyer to purchase all of the WIC-authorized
foods prescribed on the FI.
41
FigarcIV-4.
Percentage of WIC VtaaVtn with laiufficieat Stock by Type of Food lutramcat
System
Vifci SpiiHlr M—
i^*t~*
Because participants in vendor-specific FI systems must use their FIs only at a single specified
vendor, it was of interest to examine the percentage ofvendors with insufficient stock by the type of
FI system in which they operated. As can be seen by Figure IV-4, vendors operating in open FI
system States (1.6 percent) were more likely to have insufficient stock during at least one of the buys
than those in vendor-specific FI system States (0.8 percent) (Table B-4 in Appendix B).
The study also compared the percentage ofWIC vendors with insufficient stock by vendor locale.
Figure IV-5 shows that vendors located in metropolitan areas (3.6 percent) were more likely to have
insufficient stock during at least one ofthe buys than vendors located in non-metropolitan areas (1.9
percent) (Table B-3 in Appendix B).
42
F^wcIV-S.
•erccatage of WIC Vndon With IastSkkot Stock by Loc«k Acrou All Bays
20*
IWt
Metropolitan Nofl-mcuopoliun
Analysis was also conducted to determine the type of food package for which the vendor was most
likely to be out of stock. Figure IV-6 displays the percentage of vendors who were out of stock for
each of the three WIC food packages. WIC vendors were more likely to be out of stock for items in
the infant food package (3.7 percent) as compared to items in the woman food package (1.0 percent)
or the child food package (0.8 percent) (see Table B-S in Appendix B).
43
FfajarcIV-t.
Pcrccatagt of WIC Vcaoton With lmfllcfcat Stock by Typ. of Food Packagt Acrow AH B«yi
!
20H"
15%
I OS'
0%
Typo of Food rocfcaft
This study also examined how well WIC vendors maintained sufficient stock over all three of the
compliance buys. Figure IV-7 displays the percentage of these vendors who either always had
sufficient stock during all three buys, or had insufficient stock during one or more of the three buys.
As shewn, 87.8 percent of these vendors always had sufficient stock while 1.0 percent violated the
requirement during each of the three buys (Table B-2 in Appendix B).
44
Rtwe IV-7.
Percentage of WIC Venders by Frequency of Insufficient Slock Acroe* All Buy*
One Two
Naattf ■TOMMMMI at IM«V<M stack
Three
C. Vendors Offering Rain Checks and Requiring Buyers to Pay Additional
Cash
As previously noted in Figure IV-1, WIC vendors offering rain checks to buyers or asking buyers to
pay additional cash had occurred infrequently. As this resulted in a small unit of comparison, limited
analysis was conducted. However, there was an interest in examining the types ofWIC food items
for which vendors were most likely to offer rain checks. The percentage of all vendors offering rain
checks by the type of food package was also examined. WIC vendors were more likely to offer a
rain check for items in die infant food package (0.4 percent) as compared to items in the woman food
package (0.1 percent) or the child food package (0.01 percent).
45
D. Provision of Receipts to Buyers
The percentage ofWIC vendors providing a receipt at the time of transaction was also examined for
WIC vendors where all three buys were completed. This variable was examined to determine ifthe
lack of providing a receipt might be a contributing factor to overcharges and substitutions. Figure IV-
8 displays the percentage ofvendors who did not provide a receipt for each ofthe three buys. As
shown, 48.9 percent of these vendors always provided a receipt, while 33.0 percent ofthe vendors
never provided a receipt
l|pnlV4L
ftntmatf of WIC Vendors by Freqaeacy of Occarrrnces of Cashier
Not rrovidiag a Receipt Across Three Bays
It was interesting to note that there were a number of vendors who committed administrative errors,
but did not commit one of the major violations examined in this study. Figure IV-9 displays the
percentage of all WIC vendors who committed an administrative error but did not overcharge,
undercharge, or allow a substitution.
46
Figare IV-9.
Distrib«tio« of Administrative Erron Committtd by WIC Vendors
Who DM Not Overcharge, Undercharge, or Sabstitate WIC Food items Across AD Buys
35%
30%"
25%'
20%"
15%
10%"
Iwufncieal Stock Failed to Coummitn Before Price w«i lUincheck
AimMnmhttmr
This chapter has presented the findings related to WIC vendor administrative errors. Additional
information can be found in the tables located in Appendix B. The next two chapters examine findings
related to overcharging/undercharging and substitution of unauthorized items.
47
CHAPTER V
Findings Related to Vendor Overcharge and Undercharge
This chapter discusses the study findings related to whether the vendor correctly charged die WIC
Program for the actual cost ofWIC foods provided to participants. Overcharge or undercharge was
calculated as the difference between the dollar amount redeemed by the vendor as reported to the
WIC State agency and the actual retail price ofthe foods provided to participants. The retail price
was determined through a combination of data collecting efforts at the time of the compliance buy.
First, ifthe buy a .eceived a receipt, the receipt amount was entered as the retail price. If a receipt
was not received, the shelfprice was used as the retail price. The data collectors collected
information on the shelfprice ofthe foods they purchased, either by recording a posted shelfprice or
by recording die amount rung up on the cash register. If no information was available through either of
these two means, the data collector returned later and purchased the same food items with cash. The
price from the cash purchase was used to determine the actual retail price. Where scanning was used,
the scanned price was used as the retail price, even if there was a difference between the shelfprice
and the amount scanned.
The results obtained with regard to the percentage of vendors who overcharged and undercharged
are based upon an average across all three types ofbuys. This allows for an examination of the
number of vendors who overcharge or undercharge for each of the individual types ofbuys conducted
(safe, partial, minor substitution, major substitution) as well as the overall rate of vendor overcharge
regardless of the type ofbuy.
Since it is not known how often a WIC participant will attempt and actually make a partial buy or
attempt to substitute unauthorized items, only the safe buy was used to derive a national estimate of
the rate of vendor overcharge. It is important to note that the safe buy is not necessarily a control
condition that approximates the true state of WIC vendor transactions. Results obtained across all
three buys closely parallel results obtained from the safe buy alone. When results diverge, those
48
derived from the safe buy typically appear to represent a lower limit while results obtained from the
combination of all three buys appear to represent an upper limit
A. Vendor Overcharges
As was true with examining proper FI countersignature procedures, vendor overcharges can be
explained both in terms of a rate and a frequency. The rate ofvendor overcharging is expressed in
terms ofthe likely percentage of vendors who would overcharge on a single round ofbuys. The
frequency of vendor overcharge is the likelihood that a vendor will overcharge one or more times over
a series offeree buys. When examining the probability ofa vendor overcharging on any given buy,
the rate is the proper percentage to use.
When the rate ofovercharging was examined, 8.7 percent of all WIC vendors overcharged at least
once during the three buys (see Table C-l in Appendix C; Figure V-l). The frequency of
overcharge is best examined by measuring the percentage of vendors who overcharged only once,
and those mat repeatedly overcharged To examine mis issue, analysis was conducted to determine
the number ofvendors mat had never overcharged, the number that overcharged only once, and the
number mat overcharged more than once. This analysis revealed that 81.9 percent ofthe vendors
never overcharged, while 18.1 percent of the vendors overcharged one or more times, including 12.4
percent that overcharged once, 4.2 percent that overcharged twice, and 1.5 percent that overcharged
for all three buys (see Table C-6 in Appendix C). Figure V-2 displays the frequency of repeat
vendor overcharges.
49
npnv-i.
PercenU„e of WIC Vendors Overcbarfiag Across AB Bays
S.7%
*1J%
" WIC Vrruir,™ OvCTthnrfin, WIC Ycmtan Nffl OYtrdiirfint.
FijireV-2.
Percentage of WIC Vendors by Frequency ofOccurrence* of Vendor Overcnar£ia
AcroaAUBny.
90%1
*~T~.—rnr
Two
Of Veaaor Overcharge
Thre.
50
An analysis was conducted examining vendor overcharge for each type of buy. For die safe buy, only
7.0 percent of all WIC vendors appeared to have overcharged.2 In contrast, for the partial buy
condition, 9.S percent ofWIC vendors appeared to have overcharged. For the minor substitution
buys, 9.7 percent ofWIC vendors overcharged, and for the major substitution buys, 10.4 percent of
WIC vendors overcharged (see Tables C-2-5, Appendix Q. In both types of substitution buys, the
vendors overcharged whether they allowed the substitution or not
When differences in overcharge as a function ofthe type ofbuy were evaluated across WIC vendors
where all buys were completed, differences in overcharge were observed in comparing the four
different types of buys. Vendors appeared more inclined to overcharge during partial and substitution
buys than for safe buys. The differences were statistically significant between partial and sate buys
and between major substitution and safe buys. The difference between minor substitution and safe
buys was not statistically significant (see Table C-23 in Appendix C).
The actual net dollar amount of vendor overcharges was also examined by taking the total dollar
amount of vendor overcharges and subtracting vendor undercharges. The overall difference between
total dollar amount of FIs redeemed and the actual retail price of the foods purchased was
$22,156.75, or an average of $0.21 per vendor. More specifically, 6.8 percent of vendors
undercharged compliance buyers for a total of $14,674.59; an average of $0.14; while 8.7 percent of
vendors (9,287) overcharged compliance buyers $36,831.35, an average of $0.35 per buy (see
Table C-8 in Appendix C).
This entailed calculation of vendor overcharge rates using weights that were developed to approximate the national estimate of WIC vendors
using the sample of WIC vendors Out participated in the safe buy only, or partial buy only, or rmnor/major substitutions buy only
51
IJJ
When undercharge information is removed and data re-analyzed, true overcharge differences were
higher for partial buys compared to all other buys. In addition, minor substitution and major
substitution overcharge amounts exceeded safe buy overcharge amounts. Figure V-3 displays die
average overcharge amount for each ofdie four buys.
SOJO
$0.43
$0.40
S0J5
|
I $030
f $025
■ $020
$0.15
$0 10
$0.05
$000
Frjure V-3.
Average Anoaat of Vendor Overcharge by Type of Bay
Safe Major
Trp«.f««y
The impact ofdemographic and administrative variables on whether or not a vendor overcharged was also
examined. Those variables that seemed to have a statistically signifcant relationship to vendor overcharge
are discussed on the following pages.
52
1. Overcharge as a Function of Type ofWIC Food Package
When differences in vendor overcharge as a function of type ofWIC food package were examined,
vendors appeared more inclined to overcharge for food packages for women and children man for
infants. More specifically, infant food packages accounted for 27 percent ofovercharges compared
to 37 percent for women and 36 percent for children.
When this estimation process was replicated for the safe, partial, and minor or major substitution
buys, no statistically significant differences in overcharge as a function of food package type could be
identified except for minor substitutions, where the difference between woman and infant food
packages was statistically significant across all buys (see Tables C-23-C-27 in Appendix Q.
2. Overcharge as a Function of the Use of Scanning Equipment
When vendor overcharge as a function of the use of scanning equipment was examined, fewer
overcharges occurred if a vendor had scanning equipment, even ifthe cashier chose not to scan the
WIC food items. Figure V-4 displays the comparison of vendors' use of scanning equipment across
all four buy types (see Table C-10 in Appendix C). Statistically significant differences were found in
the safe buy for vendors who scanned WIC items compared to vendors who chose not to scan WIC
items or who did not have scanning equipment. The differences were also statistically significant in the
partial buy for vendors who lacked scanning equipment compared to vendors who scanned die WIC
items. Examining the minor substitution buy, vendors who lacked scanning equipment appeared more
likely to overcharge than vendors who scanned. Also in tile minor substitution buy, vendors who
chose not to use their scanning equipment were more likely to overcharge than vendors who scanned.
The major substitution buy displayed statistically significant differences between vendors who lacked
scanning equipment compared to vendors who scanned WIC items (see Tale C-28-C-32 in
Appendix Q.
53
20%
IJ*
* 10K I
Heart V-4.
Di»irib»rto» «TWIC Vcadors OvcrchargUs at a Faactioa of Use of Scaaaiat Eqaipnwat Acrou Al
3. Overcharge as a Function of Vendor Size
More overcharges occurred for small-sized vendors compared to medium-sized or large vendors (see
Table C-l 1 and C-28 in Appendix C; Figure V-S) The same pattern of results was also observed
when the estimates were derived on the basis of the type ofbuy (see Table C-28-C-32 in Appendix
C).
54
Figu-eV-5.
Distribatiea of WIC Vendor* Overdurpnt if a F««ctio« ofVendor Size
Acrota AH Biyi
20%
IJ%"
I
f
10%
Srmll-Sized Vendor Medium-Sized Vendor large-Sizad Vendor
4. Overcharge as a Function of Proper FI Counter-signature Procedures
One of the key elements examined was the relationship between following proper FI countersignaiure
procedures and the rate of vendor overcharge. Vendors who improperly asked die buyer to sign the
FI prior to entering the purchase price were more likely to overcharge than those who followed the
proper procedures. Again, the same pattern of results is observed when estimates were derived for
each buy type and separately analyzed (see Tables C-28-C-32 in Appendix Q. Figure V-6 displays
the distribution ofWIC vendors overcharging by timing of countersignature (see Table C-12 in
Appendix Q.
55
FigarcV-t.
DistiibntioB on WIC Vendors Overcharging as a Fuactioa of CoaatersigaatBie Timing Across All Bays
20%
15%
5. Overcharge as a Function of Providing the Buye; with a Receipt
Failure of the vendor to provide a receipt for the WIC transaction was identified as a predictor of
overcharge in the 1991 WIC Vendor Issues Study. To test this hypothesis, all vendors who
overcharged (8.7 percent of the total number of vendors) were grouped into two categories, those
that provided a receipt, and those that did not provide a receipt The study found that of all vendors,
7.5 percent both overcharged and failed to provide a receipt, while only 1.3 percent of all vendors
overcharged and provided a receipt (see Table C-13 in Appendix C). Almost identical results were
obtained for the safe buy alone as compared to the results across all three buys. The same pattern of
results also characterized partial and substitution buys (see Table C-28-C-32 in Appendix C).
56
6. Overcharge as a Function of Other Demographic and Administrative
Variables
The type of FI system and locale were examined as they relate tc vendor overcharges. When
examining the type ofFI system in which the vendors operated, no differences in overcharge were
observed either across all three buys or with just the safe buy. When locate was examined, a trend
for vendors from metropolitan areas to overcharge more frequently than non-metropolitan vendors
raited to reach significance.
B. Models Describing Factors Contributing to Overcharge
Additional analysis was conducted in an effort to develop a model describing factors contributing to
the frequency ofovercharge. As an initial step in this direction, a series oflogistic regression equations
were generated. Subsequently, the data were reorganized to distinguish between vendors where
errors occurred occasionally from vendors where errors occurred more frequently. Finally, a series of
mululog analyses were generated in order to better account for repetition in overcharge across the
three compliance buys.
The muftilog results appeared to yield few additional insights, other than to confirm results previously
obtained. Results ofthemultibg analysis are discussed below, with statistically significant findings
detailed in Tables C-16-C-22 in Appendix C. Models were developed both on the basis of the
variables detailed earlier in this chapter and in accordance with models previously developed in the
1991 WIC Vendor Issues Study. However, some of the variables used in previous studies (e.g., the
extent to which a vendor might have been "busy") were not collected in this study. At the same time,
new variables were also introduced in this study, such as indication of cashier inexperience. The
following discussion addresses the dynamics of variable interaction and, where appropriate, the
methodological limitations ofresults obtained.
57
1. Logistic Regression Models
Logistic regression is used to generate model parameter estimates, their standard errors, tests of the
null hypothesis in which individual regression coefficients associated with each variable in the model
are equal to zero, as well as tests ofoverall model and individual parameters' significance. This
technique can also be used to generate odds ratios describing the ratio of odds for a one-unit change
in the independent variable. The odds ratio for a single regression coefficient Is the quantity In
addition, an R-square statistic based on Cox and Snell (1989) describes the proportion of die log-likelihood
that is explained by the model. This statistic will be frequently detailed as a summary
statistic to describe models that were tested.
Logistic regression is particularly useful in evaluating dichotomous outcome variables in its use of
maximum likelihood techniques. The logistic model can be described mathematically as:
log(p/l-p)=X?
p=prob(Y=l\X)
where Y is the response outcome (1= presence of overcharge; (^absence), X is the covariant design
matrix, and Is the vector of the parameters to be estimated. Essentially, overcharge was defined as
the dependent binary variable where "1" signified overcharge, and "0" signified the absence of
overcharge. Generalized estimating equations (GEE) were used to develop models (Zeger and Liang,
1986; Liang and Zeger, 1986).
Consideration was given to examining the individual variable contributions to overcharge. In
Appendix C, Tables C-l 6-C-22 list variables which appear to account for some of the variation in
overcharge. More specifically, these tables detail relevant variables, R2, odds ratios, overall model
statistics and p-values, and relevant statistics on parameter values.
Failure to provide a receipt, use of scanning equipment, vendor size and not following proper FI
countersignature procedures all appear to contribute to overcharge. Over all buys, WIC vendors
58
who did not provide a receipt were about 10.5 times more likely to overcharge than vendors who
provided a receipt In terms of the use of scanning equipment, vendors who had no scanning
equipment were about 6.0 times more likely to overcharge than vendors who used scanning
equipment, while vendors who simply chose not to scan (but had equipment) were 4.7 times more
likely to overcharge compared to vendors who used scanning equipment With regard to the
contribution of vendor size, small-sized vendors were 4.6 times more likely to overcharge than
medium-sized vendors, and 6.5 times more likely to overcharge than large vendors. Vendors who did
not require the buyer to countersign after the purchase price was entered were 4.9 tiroes more likely
to overcharge than vendors who had the buyer sign die FI after the purchase price was entered (see
Table C-16 in Appendix Q.
In general, the relationships described above appeared somewhat attenuated for the safe buys
compared to other buy types, and somewhat stronger for the partial buy. However, receipt provision
appeared particularly important in the minor substitution (see Tables C-17-C-20 in Appendix C).
Other single variable models analyzed included vendor-to-participant ratio, indication of cashier
inexperience, food package type, and cashiers providing the buyer with incorrect information related
to authorized foods. These models either accounted for only minuscule portions of the variance or
failed to meet all cntena for statistical significance.
Multiple variable model analyses were conducted for each variable across all buy types. The two
strongest three-variable models predicting variance were the model including error in countersignature
timing, failure to provide a receipt, and failure to use scanning equipment (R2=0.212); and the model
containing error in countersignature timing, failure to provide a receipt, and small vendor size
(R2^)209). Both models satisfied criteria for a statistically significant model across all buy types.3
When a four-variable model describing overcharge as a function of countersignature errors, failure to
provide a receipt, small vendor size and failure to use scanning equipment when conducting a W1C
transaction was simplified from a 10 parameter model (where scanning was defined as no scanning
In this context. Beta! far medium-sized vendors did not appear K> differ from O when smaller vendors constituted the lefoeuce level
(p-0.06)
59
equipment, scanning, or chosr not to scan) to a 9 parameter model (where scanning was denned as
no scanning versus scanning), it appeared to meet all criteria for a successful model (R2=0.213)
across all buy types (see Table C-21 in Appendix Q.
Vendors who overcharged on all three buys are the most likely to be intentionally overcharging the
W1C Program. For each of the three aforementioned models, odds ratios were calculated to examine
the contribution ofeach individual model variable when applied to vendors who overcharged on all of
the three buys (see Table C-22 in Appendix C). Each ofthese models is discussed below.
• Three-Variable Model: Vendors who did not scan, failed to provide a receipt, and
violated FI countersignature procedures. When the contribution of individual variables
contained in the above three-variable model (R2=0.212) were examined, vendors who did not
scan were about three times more likely than their scanning counterparts to overcharge;
vendors who railed to provide a receipt were about seven times more likely to overcharge;
and vendors who violated FI countersignature procedures were over five times more likely to
overcharge;
• Three-Variable Model: Small vendors, vendors who failed to provide a receipt, and
vendors who violated FI countersignature procedures. When the contribution of
individual variables contained in the above three-variable model (R2O.209) were examined,
small vendors appeared about three times more likely to overcharge than medium or large size
vendors; vendors who failed to provide a receipt were about 7.S times more likely to
overcharge; and vendors who violated FI countersignature procedures were about S.6 times
more likely to overcharge; and
• Four-Variable Model: Vendors who violated FI countersignature procedures, failed
to provide a receipt, small-sized vendors, and vendors who did not scan When the
contribution of individual variables contained in the above four-variable model (R2=0.213)
were examined, vendors who violated FI countersignature procedures were over five times
more likely to overcharge; vendors who did not provide a receipt were 6.7 times more likely
to overcharge; small-sized vendors were approximately 1.5 times more likely to overcharge
than medium or large-sized vendors; and vendors who did not scan were over two times
more likely to overcharge.
60
Close inspection of models suggested that some variables may have assumed a mediating (or partial
mediating) role. In other words, the effect of one variable on an outcome is lessened when another
variable is introduced For example, vendor size is assumed to effect overcharging. Often, the effect
of vendor size on overcharging is mediated by the use of scannmg equipment by the vendor. Vendor
size is often an indicator of the availability of scanning equipment Once the mediator variable, use of
scanning equipment, enters the equation vendor size no longer affects overcharging. This would be
regarded as an instance of complete mediation. Partial mediation is the case in which the association
from vendor size to overcharge is reduced in absolute size but is still different from zero when the use
ofscanning is controlled
Baron and Kenny (1986) and Judd and Kenny (1981) have discussed four steps in establishing
mediation;
Step 1: Show the initial variable (X) is correlated with the outcome (Y). Use Y (overcharge)
as the criterion variable in a regression equation and X (vendor size) as a predictor. This step
establishes that there is an effect that may be mediated
Step 2: Show the initial variable (vendor size) is correlated with the mediator (scanning). Use
M (scanning) as the criterion variable in die regression equation and X (vendor size) as a
predictor. This step essentially involves treating the mediator as if it were an oticome
variable.
Step 3: Show the mediatorM (scanning) affects the outcome variable Y (overcharge). Use Y
as the criterion variable in a regression equation and X (vendor size) andM (scanning) as
predictors. Essentially, die initial variable X (vendor size) must be controlled to establish die
effect of the mediator M (scanning) on the outcome Y (overcharge).
Step 4: Establish that M (scanning) completely mediates the X-Y (vendor size-overcharge)
relationship: die effect ofX (vendor size) on Y (overcharge) controlling forM should be zero.
The effects in both Steps 3 and 4 are estimated in the same regression equation.
61
The use of scanning equipment appears to mediate the contribution of vendor size. Vendor size does
appear to constitute a viable predictor ofscanning across all buy types. And as previously described,
scanning is a factor that tends to reduce the frequency and amount ofovercharge.
Similarly, providing a receipt may also, at least partially, mediate the influence of lack of scanning, and
vendor size in overcharge. Both lack ofscanning (R2=0.172) and vendor size (R2=0.165) may
facilitate prediction ofreceipt provision. Vendors who lack scanning equipment are uniformly (across
all buys) much more likely not to provide a receipt On average, medium-sized vendors are five times
less likely, and large vendors are almost eight times less likely, than small vendors not to provide a
receipt (see Table C-21 in Appendix C).
From a practical standpoint, it is easy to understand how these factors influence one another. Vendors
who lack scanning equipment or simply fail to scan are likely not to provide a receipt and may be
more inclined to overcharge. Similarly, small vendors who may lack scanning equipment appear more
inclined not to provide a receipt and, consequently, may also be inclined to overcharge. Again, it is
not possible to determine the extent to which overcharge and receipt provision reflects an intentional
trend to overcharge or simply reflect carelessness on the part of the cashier.
2. Repeat Offender Models
Data were organized to permit analysis ofovercharge that could distinguish WIC vendors who
repeatedly violate proper FI countersignature procedures, fail to provide a receipt, or do not use
scanning equipment from vendors who only occasionally engage in such behaviors (see Table C-
21-C-22 in Appendix Q.4
When examining overcharges in terms of repeated failure to provide receipts, it was interesting to note
that while the single-variable model describing WIC vendor's failure to provide a receipt accounted
It should be noted that a substantial attrition in N (18%) was observed in this context since missing data
in any one of the three completed buys effectively warranted elimination of that observation.
62
for the most variance (R2*.14); WIC vendors who failed to provide a ieceipt only once could not be
distinguished from vendors who always provided a receipt. More specifically, vendors who failed to
provide a receipt once were only about 12 times more likely to overcharge than vendors who always
provided a receipt, while those who did not provide a receipt twice were 2.5 times more likely to
overcharge than vendors who always provided a receipt Those that failed to provide a receipt all
three times were 12.8 times more likely to overcharge than vendors who always provided a receipt.
in terms of WIC vendors who had scanning equipment but repeatedly did not use the equipment (R2
=. 10)5 those who used scanning equipment once were 2.8 times more likely to overcharge than
vendors who always used scanning equipment, those who failed to use scanning equipment twice
were 4.7 times more likely to overcharge, and those who never used scanning equipment were 6.4
times more likely to overcharge.
When examining .he failure of cashiers to obtain the buyer's signature after the purchase price is
entered on the FI, vendors failing once (R2=07) seemed to be 1.7 times more likely to be associated
with overcharge, two failures to properly obtain a buyer's countersignature seemed to be 3.6 times
more likely to be associated with overcharge, and three failures to properly obtain a buyers
countersignature seemed to be 5.5 times more likely to be associated with overcharge.
When examining two-variable models, die results from combining die use of scanning with vendor size
was again suggestive of a mediation effect; that is, the vendor not scanning invariably diminished the
significance of the contribution of vendor size. The best three-variable model (R2=212) described
overcharge as a function of violating FI countersignature procedures, failure to provide a receipt and
small vendor size.
Il was neonsary to re-code this variable into two levels - vendor scanned and vendor did not I
63
The four-variable saturated model (R2=213) failed to meet requisite criteria for success, probably
because ofthe mediation ofthe "no scanning" variable on vendor size and/or the mediation of failure
to provide a receipt on scanning.
C. Vendor Proclivity to Overcharge
In previous studies, no attempt was made to distinguish vendors who overcharged as a function of
deliberate intent from vendors who overcharged due to random error. One of the factors that can be
examined in this study to determine the randomness ofovercharge is to look at the vendors in the
context of undercharging. It seems unlikely that a vendor would intentionally undercharge the WIC
Program for foods obtained with FIs. More likely, an error was made by the cashier transferring
information from the cash register to the FI. Because these errors are random, and a cashier would
be as likely to make an overcharge error as an undercharge error, it seemed appropriate to consider
differences between vendors who consistently overcharged, vendors who may have both overcharged
and undercharged, and vendors who did not overcharge or undercharge. Data were analyzed to
facilitate examination of overcharging by vendors who consistently overcharged and those who
occasionally overcharged. While we can not assume vendors who consistently overcharged are doing
so intentionally, this group is more likely to intentionally overcharge than those who occasionally do
so.
A little over 13 percent of the WIC vendors overcharged at least once but never undercharged