Evaluation of the expanded EBT demonstration in Maryland patterns of food stamp and cash welfare benefit redemption

              ML Q*«t 'vt>lH'A-OI -* ftW^t'L i/¥
USDA
0#pQftm#fw of
AgrtcuHn
Food ond
Contumor
S#fvicc
Offlc«of
Andy* and
Evaluation
Evaluation of the Expanded EBT
Demonstration in Man/land
Patterns of Food Stamp and Cash
Welfare Benefit Redemption
February 1997
^ /t/£ f?'-cm z3 <?■*- Qs
USDA
Deportment of
Agrtcufcra
Food ond
Consumer
Swvte*
Offlc#o(
Analysis ono
EVOkJCnOfl
Evaluation of the
Expanded EBT
Demonstration in
Maryland
Patterns of Food
Stamp and Cash
Welfare Benefit
Redemption
February 1997
Prepared for:
Margaret Andrews
U.S. Department of Agriculture
Food and Consumer service
3101 Park Center Drive
Alexandria, VA 22302
Prepared By:
Nancy Cole
Abt Associates Inc.
55 Wheeler Street
Cambridge, MA 02138
Contract No.
53-3198-1-019
£
TABLE OF CONTENTS
Executive Summary <
Chapter One: Introduction and Description of Data I
Report Organization 2
Description of Data 3
Food Stamp Program-Authorized Retailers 7
Chapter Two: Food Stamp Benefit Redemptions II
Redemption Behavior by Case Type, Location, and Allotment Size 16
Food Stamp Redemption by Store Type 22
Benefit Exhaustion 24
Benefit Carryover 31
Out-of-State Transactions 33
Chapter Three: Cash Benefit Redemptions 35
Caseload Composition 36
Timing and Frequency of Cash Redemptions 39
Location of Cash Redemptions 42
Redemption Behavior by Case Type, Location, and Allotment Size 48
Cash Benefit Exhaustion 48
Out-of-State Transactions 58
Appendix A: Description of Transactions Log File and Processing Methods
Appendix B: Comparison of Survey Responses with Transactions Behavior
Appendix C: Studies of Benefit Redemption
Appendix D: Cash Benefit Redemption Tables, By County
Prepared by Abt Associates Inc. s
EXECUTIVE SUMMARY
This report presents a description of the benefit redemption behavior of food stamp and
cash assistance clients participating in the statewide electronic benefit transfer (EBT)
demonstration in Maryland. The research is based on the redemption activity of approximately
130,000 food stamp recipients and 75,000 recipients of cash assistance during September 1993.
following full implementation of Maryland's EBT system.' The main findings of the research
are summarized below.
Food Stamp Recipients Use the EBT System Frequently
• Food stamp recipients average a bit over ten EBT transactions a month. The
average is even higher in Baltimore City (nearly 13 transactions per month, versus
fewer than eight transactions per month in both metro and non-metro counties).*
• The average number of transactions per case increases with allotment size, but at
a decreasing rate.
• Non-public assistance (NPA) food stamp cases use the system considerably less
frequently than public assistance (PA) food stamp cases, but this is due to
differences in average monthly benefits.
Food Stamp Recipients Use Their Benefits Very Quickly
• Statewide, nearly 23 percent of all monthly benefits are spent the day they are
disbursed. In Baltimore City, over 25 percent of benefits are spent immediately,
versus about 20 percent in both metro and non-metro counties.
• Statewide, about 71 percent of food stamp benefits are redeemed within one week
of disbursement (76 percent in Baltimore City versus about 66 percent elsewhere in
the state).
1. EBT was fully implemented statewide in July 1993.
2. "Metro" counties are those within the borders of Standard Metropolitan Statistical Areas (SMSAs),
excluding Baltimore City; "non-metro" counties are all other counties.
Prepared by Abt Associates Inc.
Executive Summon-
Average purchase amounts fall dramatically over the first week, and then tend to
level off until the end of the month. This is true in all store types except
convenience stores. Average purchase amounts in convenience stores are relatively
constant throughout the month.
The speed at which food stamp benefits are redeemed is nearly the same regardless
of the size of the recipient's allotment. To the extent a difference exists, it appears
that recipients with small monthly allotments are more likely to retain benefits at
the end of the month.
Nearly 80 percent of all food stamp recipients completely exhaust their benefits by
the end of the month. Less than 5 percent of recipients have more than 5 percent
of benefits remaining at the end of the month.
For the recipients who do not exhaust their benefits, the average amount carried
over to the next month is $24.
Most Food Stamp Benefits are Redeemed at Supermarkets
• Seventy-two percent of all food stamp benefits in Maryland are spent at supermar-kets,
even though supermarkets represent only 17 percent of program-authorized
stores in the state.
Even in Baltimore City, where supermarkets represent only 6 percent of authorized
stores, recipients spend 61 percent of their benefits at supermarkets.
The average value of a purchase in a supermarket is $48, versus $21 in specialty
stores (the store type having the next highest average purchase amount).
On any given day of the month, supermarkets' percentage of total daily food stamp
redemptions throughout the state is roughly constant. The same is true for the
other store types. Thus, there is no evidence that recipients make their big
purchases at the beginning of the month in supermarkets and then, during the rest
of the month, make smaller-value purchases at other store types.
Cash Assistance Benefits are Accessed Even More Rapidly Than Food Stamp Benefits
• Cash benefit recipients average 4.6 withdrawal transactions per month, or 1.7
transactions per $100 in benefits.
• Over 60 percent of all cash benefits are redeemed on the day of disbursement; 91
percent are redeemed by the end of the first week.
Prepared by Abt Associates Inc. jj
//
Executive Summon
Baltimore City recipients access their cash benefits more rapidly than recipients in
other parts of the state.
POS Terminals Are an Important Source of Access to Cash Program Funds
• Forty-seven percent of all cash program transactions (representing 27 percent of
total benefits accessed) are conducted at POS terminals. (The data do not allow one
to distinguish between cash withdrawals at POS and regular purchases using EBT
benefits.)
• POS terminals are an even more important point of access in Baltimore City where,
depending on program, 55 to 61 percent of all transactions (representing 33 to 45
percent of cash benefits redeemed) occur at POS terminals.
• Most cash program recipients use both ATMs and POS terminals to access their
benefits. Statewide, only 16 percent of cases use ATMs exclusively, and only 20
percent use POS terminals exclusively.
• In general, the relative use of ATMs increases as their relative availability
increases. A small number of recipients use out-of-state ATMs.
• Statewide, the average ATM withdrawal is about $114. The average POS
withdrawal/purchase is about $53.
• Twenty-four percent of cash assistance cases had benefits remaining in their EBT
accounts at the end of the month; the average balance for this group was $35.
Prepared by Abt Associates Inc. ij, IS
CHAPTER ONE
INTRODUCTION AND DESCRIPTION OF DATA
The purpose of the Food Stamp Program (FSP) is "to permit low-income households
to obtain a more nutritious diet through normal channels of trade".3 Yet prior to the EBT
demonstrations, information about where recipients redeem their benefits could only be tracked
by aggregate statistics showing the volume of redemptions at each type of program-authorized
retailer, or from small-sample surveys of individual recipients. Hence, it was not possible to
examine the general patterns of benefit redemption—such as the types of stores accessed, the
frequency of shopping trips, and the timing of benefit exhaustion—on a large scale at the micro
level.
With the introduction of electronic benefit transfer (EBT) systems for benefit
disbursement, detailed information becomes available about food stamp benefit redemption. For
example, instead of simply finding that 80 percent of all food stamp coupons are redeemed at
supermarkets, we can now examine how that figure varies across subgroups of the caseload,
how it varies by geographic area holding casemix constant, and how it depends on the number
and types of retailers in an area. This is because, in an EBT system, every purchase transaction
is electronically recorded within a central processing system. Hence, detailed information is
available at the case level and without the measurement error inherent in survey data.
This report examines benefit redemption data from the statewide EBT demonstration
in Maryland. EBT was fully implemented in Maryland by July 1993, to serve both the food
stamp population and recipients of cash benefit transfers (Aid to Families with Dependent
Children (AFDC) and Disability Assistance Loan Program (DALP)). The Maryland demonstra-tion
is not the first EBT demonstration, and indeed previous demonstrations have provided
information about redemption patterns in the FSP.4 The Maryland demonstration, however,
provides the first opportunity to analyze the redemption patterns of both the FSP caseload and
3. The Food Stamp Act of 1977 (7 U.S.C.).
4. See BarUett, Susan and Meg Hut (1987) and Phoenix Planning and Evaluation (1993) for evidence from
the Reading, Pennsylvania and Montgomery County, Ohio demonstrations, respectively. Appendix Table C
describes the samples of data used in those reports.
Prepared by Abt Associates Inc.
Chapter One: Introduction and Descriptions of Data
the cash assistance caseload. In addition, the State of Maryland encompasses a large caseload
of varied demographic groups in varied geographic areas. This large cross-section allows us to
examine the different redemption patterns in urban and non-urban areas, and in inner-cities—
areas cited by Congress to be "problem" areas due to "few supermarkets."5
In addition to examining the distribution of food stamp redemption patterns, the
Maryland EBT system data are valuable for examining the transaction processing demands (both
spatial and temporal) that a welfare caseload places on an EBT system; this, includes the
demands of both the FSP caseload and the cash benefit caseload. This report addresses both of
these analytical needs by providing a descriptive summary of EBT transactions activity.
Report Organization
Chapter Two examines five main aspects of redemption behavior in the Food Stamp
Program:
(1) Frequency and timing of food stamp redemptions throughout the disbursement
month;
(2) Average purchase amounts, overall and by time since disbursement;
(3) Distribution of purchases by store type;
(4) Timing of benefit exhaustion; and
(5) "Carryover" of benefits from one month to the next.
For each of these topics the report examines the distribution of caseload behavior, rather than
concentrating solely on average behavior.
Chapter Three is analogous to Chapter Two, but examines the cash redemption
behavior of AFDC and DALP recipients. This chapter examines:
(1) Characteristics of the caseload: the mix of the cash benefit caseload according
to the size of allotment, receipt of food stamp benefits in addition to cash
benefits, and timing of disbursements;
5. U.S. House of Representatives, hearing of the Committee on Agriculture, "Ensure Adequate Access to
Retail Food Stores by the Recipients of Food Stamps and to Maintain the Integrity of the Food Stamp
Program," November 4, 1993.
Prepared by Abt Associates Inc.
Chapter One: Introduction and Descriptions ofData
(2) Frequency and timing of cash withdrawal transactions throughout the disburse-ment
month;
(3) Average withdrawal amounts, overall and by time since disbursement; and
(4) Number of transactions and average withdrawal amounts, by location (i.e..
ATM versus POS device).
The following section describes the transactions data and the types of FSP-authorized retailers.
A detailed description of our processing methods appears in Appendix A.
In addition to analyzing the redemption behavior of the Maryland caseload, we are able
to match the actual transactions activity of a sample of recipients to survey d^ta that were
collected as part of the evaluation of the expanded EBT demonstration. Although a full analysis
of these matched data is beyond the scope of this report, we iiave done some simple
comparisons of the actual redemption behavior with reported shopping patterns. This analysis
of the reliability of survey responses is included as Appendix B.
Description of Data
We obtained a file from Deluxe Data Systems of all transactions processed through the
Maryland EBT system during the period from mid-August 1993 through October 1993. Deluxe
Data Systems provides the central processing services for the Maryland EBT system. For
reasons detailed in Appendix A, this report examines transactions activity in the September
disbursement month only. For each case, the disbursement month is measured as the period
beginning on the day the September allotment is received, and ending on the day prior o receipt
of the October allotment.
The entire Maryland EBT caseload of food stamp and cash benefit recipients consists
of approximately 170,000 cases. One calendar month of transactions activity contains approxi-mately
three million records. We processed these data to obtain a separate food stamp
redemption history and cash benefit withdrawal history for each case. Figure 1.1 shows the
distribution of the number of food stamp purchase transactions and cash benefit withdrawal
trar ctions per case during September 1993. Appendix A contains a detailed description of the
content of the transactions file and the methods that we used to process these data.
Prepared by Abt Associates Inc.
Figure 1.1
Distribution of EBT Transactions Per Case
Food Stamp Purchase Transactions
25% >*■
I
n%-
15%
10%
Cash Redemption Transactions
o■
8I
35%
30%
25%
20%
19%
10%
Prepared by Abi Associates Inc.
Chapter One: Introduction and Descriptions of Data
All of the analyses presented in this report correspond to the September 1993
disbursement month. The regular disbursement cycle is staggered over two three-day periods:
the first, second, and third days of the month for cash benefits; the fifth, sixth, and seventh
days for food stamp benefits. Because we analyze transaction activity in the disbursement
month, we essentially align the transaction activity of each case to correspond to a "days since
disbursement" time line; this smoothes over the calendar day variations that are due solely to
the three-day disbursement cycle.
In addition to the regular disbursement cycle, both food stamp and cash benefit
disbursements may occur throughout the month to new cases or on an emergency basis. New
cases receive prorated benefits according to the number of days remaining in the disbursement
cycle; cases receiving emergency benefits are observed to receive both a "regular" disbursement
and a supplementary disbursement. In order to simplify the analysis, we restrict the sample to
cases receiving only a "regular" disbursement in both September and October.6 This excludes:
newly-opened cases; cases that, in September, were about to close; and cases that received
supplementary benefits in addition to a regular disbursement.7 This simplification allows us
to examine variations in transaction behavior without having to control for caseload hetero-geneity—
arising from differences in number of disbursements and length of disbursement
month—that would otherwise be present in the data.
Table 1.1 describes the composition of the entire caseload in terms of the timing of
disbursement in September 1993. Ninety percent of the September food stamp caseload
received a single monthly disbursement on one of the regular disbursement dates (the fifth,
sixth, or seventh days). Of those regular cases, 97 percent redeemed at least some of their
benefits during the calendar month. In contrast, of the food stamp cases receiving an
6. That is, we select cases that received regular cash benefit disbursements on the first, second, or third days
of both months, or food stamp disbursements on the fifth, sixth, or seventh days of both months.
7. We could only identify ongoing cases based on the date of benefit disbursement Therefore, h is possible
mat we have included new cases that happened to enter the caseload at the beginning of the month.
Prepared by Abt Associates Inc. 5
Table 1.1
MARYLAND CASELOAD COMPOSITION
Cases Receiving a Disbursement in September 1993
Total Cases
Number Percent
75T
Cases wttk Transactions
in September
Number Row Percent
75) W
Analysis Sample
(Regular disbursement
in Sep. and Oct.)
Number Percent of (1)
75T IS
Total
Regular Disbursement Only*
Irregular Disbursement Only
Regular + Supplement
(T)
Food Stamp Cases
159,054 100.00% 133.039 96.22%
143.718 90.36 139.345 96.96
13.216 8.31 11.602 87.79
2.120 1.33 2.092 98.68
9
129.594 90.17%
Total
Cash Benefit Cases
104.180 100.00% 100.353 96.52%
76.335 73.27 74.599 97.73 64.294 84.23%
11.678 11.21 11.365 97.32 10.163 87.03%
7.235 6.94 6.504 89.90 - .
5.570 5.35 5.463 98.08 . .
3.362 3.23 2.622 77.99 - .
Regular Disbursement Only*
Regular + Child Support Bonus
Irregular Disbursement Only
Regular + Supplement
NPA-CS Only**
Source Maryland EBT Transactions Log from Deluxe Data Systems, September 1993 disbursement month
'Regular" Food Sump disbursements are issued on the 5th.6th. 7th of the month; "regular" cash benefits
disbursements are issued on the 1st, 2nd, 3rd of the month. Column (5) is a subset ofcolumn (3).
••NPA-CS cases are "Non-Public Assistance Child Support" cases. These are cases for which the collection of
child support is administered by the State and disbursement of funds is via EBT. In 1993. EBT disbursement of
NPA-CS payments was primarily limited to Baltimore City.
Prepared by Abi Associates Inc. 6
Chapter One: Introduction and Descriptions of Data
"irregular" disbursement, 88 percent redeemed benefits during the calendar month.8 Less than
2 percent of cases received supplementary benefits in addition to their regular disbursement.
Cash benefit cases are characterized by a lower percentage of "regular" disbursements
compared to food stamp cases. Much of this difference is due to the disbursement of the child
support bonus (98.4 percent of cases received the bonus on the 21st of the month: nearly all of
the remainder received the bonus on the first day).9 Cases receiving the bonus in addition to
a regular disbursement are considered "regular" cases for purposes of inclusion in our analysis
sample.
All tables in this report are based on either the entire caseload (column 1) or cases with
a "regular" disbursement (column 5). The sample used for analysis (column S) represents 81
percent of all food stamp cases, and 90 percent of all cases with a regular disbursement in
September (the remaining 10 percent of regular cases either did not receive an October
disbursement or did not redeem any benefits in September). The analysis sample of cash
benefit recipients (column 5, lower panel) represents 71 percent of all cash benefit cases, and
85 percent of all cases with a regular disbursement in September.
Food Stamp Program-Authorized Retailers
Food stamp benefits may be redeemed at a variety of locations for food items for home
preparation and consumption. Authorized retailers include supermarkets, small and medium
grocery stores, specialty food stores (e.g., produce and seafood stores), convenience stores, and
certain stores that sell a variety of merchandise, including food. In addition, elderly and
disabled recipients may redeem food stamps for home-delivered meals, and homeless recipients
may redeem food stamps for prepared meals at shelters.
8. "Irregular" benefits received at the end of the disbursement month might be more subject to "saving"
behavior—i.e., accumulation of benefits across disbursement periods. This is because the time between
disbursements is short (irregular disbursements to newly-opened cases are subsequently followed by "regular"
disbursements), and because the irregular disbursement may be prorated. In addition, we may observe what
appears to be "saving" behavior if recipients are slow in learning to use their EBT card. Hence, we make
the distinction between "regular" disbursement cases and "irregular" disbursement cases.
9. The maximum child support bonus is $50; 90 percent of all bonus disbursements in the September
calendar month were for the maximum amount.
Prepared by Abt Associates Inc.
Chapter One: Introduction and Descriptions of Data
We identified "store type" based on a listing of FSP-authorized retailers obtained from
the Food and Consumer Service's (FCS) Minneapolis Computer Service Center.10 Table 1.2
lists the types of authorized redemption locations in Maryland and the aggregate volume of food
stamp transactions and purchase amounts at each type of location during the September
disbursement month.11 Throughout this report we concentrate on the four main retailer types
(supermarkets, grocery stores, specialty stores, and convenience stores), and group all other
redemption locations in the "other" category. The one modification that we make to the FSP
"store type" categorization is that we group "combination grocery and gas" and "combination
grocery and merchandise" locations together with "convenience stores".12
10. Store type is self-reported by the retailer at the time of application to the FSP.
11. The redemption numbers are based on ail food stamp cases redeeming benefits in September, not just
those in the analysis sample.
12. This categorization is similar to that used in the evaluation's final report (and previous EBT evaluations).
In past reports, "grocery stores" and "specialty stores" have been grouped together as "grocery stores."
Prepared by Abt Associates Inc. g
Table 1.2
FSP-AUTHORIZED RETAILERS IN MARYLAND
Purchie Transaction Volume
Number of Locations Number of Transactions
Number Percent Number Percent
Dollars Redeemed
Amount Percent
Supermarket (SM) 547 16.9% 723.222 43.8* 20.219.786 71.8*
Small/Medium Grocery (GS) 800 24.7* 465.196 28.2% 2.920.515 10.4*
Specialty Food (SF) 314 9.7* 132.647 8.0* 2.654.407 9.4*
Convenience Stores:
Convenience Store (CS) 971 30.0* 184.915 11.2% 997.501 3.5*
Comb. Groc/Gas (CG) 42 1.3* 12.904 0.8% 49.723 0.2%
Comb. Groc/Merchandiie (CM) 40 1.2* 6.018 0.4% 48.710 0.2%
Other Stores:
Total 445 13.8* 124,715 7.6% 1.253.949 4.5%
Alcoholic Treatment (AT) 3 0.1* 81 0.0% 7.642 0.0%
Non-profit Coop (BQ 13 0.4* 795 0.0% 14,161 0.1%
Bread Route (BR) 6 0.2* 914 0.1% 17.693 0.1%
Comb. Groc/Bar (CB) 10 0.3* 2,053 0.1% 14.380 0.1%
Other Combination (CO) 139 4.3* 52.047 3.2% 510.264 1.8%
Comb. Groc/Resuurant (CR) 59 1.8* 22,363 1.4% 200.071 0.7%
Drug Addict Treatment (DT) 1 0.0* 9 0.0% 1.057 0.0%
Farmers Market (FM) 12 0.4* 3,614 0.2% 82.973 0.3%
Group Living Anaugment (GL) 1 0.0* 10 0.0% 923 0.0%
Health/Natural Food (HF) 29 0.9* 852 0.1% 15.779 0.1%
Homeless Meal Provider (HP) 1 0.0* 5 0.0% 300 0.0%
Military Commissary (MQ 8 0.2* 373 0.0% 27.377 0.1%
Non-profit Meal Delivery (ND) 3 0.1* 13 0.0% 240 0.0%
Non-profit Communal Dining (NP) 4 0.1* 5 0.0% 202 0.0%
Other Firm (OF) 60 1.9* 20.134 1.2% 143.042 0.3%
Other Route (OR) 26 0.8* 2.110 0.1% 53.849 0.2%
Produce Route (PR) 7 0.2* 110 0.0% 6.117 0.0%
Produce Stand (PS) 52 1.6* 18.527 1.1% 134.080 0.3%
Wholenler (WH) 11 0.3* 700 0.0% 23.800 0.1%
Unknown* 74 2.3* 1,737 0.1% 4.166 0.0%
3.233 100.0* 1.651.354 100.0% 28.148.755 100.0%
Source: FCS Minneapolis Computing Service Center and Deluxe Transactions Log File. Number of retailers includes
authorized retailers at which no redemptions occurred in September 1993. Eicludri ansactions made in 39
out-of-staie stores which accounted for 0.3% of transactions and 0.3% of food stamp dollars redeemed.
* Unknown store type is due to the presence of newly authorized retailers in the transactions database that could not
be matched to the master list of retailers from the Minneapolis Computer Service Center.
Prepared by Abt Associates Inc.
Prepared by Abt Associates Inc. 10
CHAPTER TWO
FOOD STAMP BENEFIT REDEMPTION PATTERNS
This chapter presents a description of food stamp recipients' redemption behavior. The
main purpose is to examine the heterogeneity within the caseload with respect to the frequency
of benefit redemption, average purchase amounts, location of benefit redemption (i.e.. types of
stores), and speed of benefit exhaustion. For example, the variation with respect to location of
benefit redemption (i.e., store types) provides some evidence of the variation in access to
different types of authorized food stamp retailers. In addition to the behavioral implications, the
simple patterns of benefit redemption—in terms of the volume and timing of redemption
transactions—provide information about operational requirements and peakload capacity
requirements that may be useful for the general planning purposes of future EBT systems.13
Food stamp recipients in Maryland made over 1.3 million EBT purchases in September
1993, spending over $23.5 million in benefits. Figures 2.1 through 2.3 display the main
characteristics of these food stamp redemptions: most redemption activity (in terms of both
transactions and dollars redeemed) occurs early in the disbursement month; the majority of both
transactions and purchases occur at supermarkets; and average purchase amounts at all store
types, except convenience stores, fall precipitously during the first week after disbursement and
then level off.14 Somewhat surprisingly, Figure 2.4 shows that the distribution of both
transactions and dollars redeemed across store types remains fairly constant throughout the
disbursement month. One would have expected the opposite if, for example, a majority of
recipients redeem the bulk of their benefits during a supermarket trip early in the disbursement
month and then shop at smaller, more accessible stores later in the month.
Table 2.1 provides a detailed description of redemption activity in Maryland in
September 1993. The distribution of the caseload and redemption characteristics are shown for
13. Throughout this chapter, the term "transaction" refers to redemption transactions and does mot include
food stamp refunds, balance inquiries, or rejected or reversed transactions.
14. Recall that by examining a disbursement-month time line, instead of the calendar month, we have
smoothed over day-of-the-week variations. Even on a calendar timeline, however, these data do not display
the Saturday peaks that were found in the Reading and Ohio demonstration data. This may be because the
disbursement dates (fifth-seventh) fell on Sunday-Tuesday in September 1993.
Pnpared by Abt Associates Inc. 11
200
150
100
SO
Figure 2.1
Total Volume of Food Stamp Redemption
By Day Since
Disbursement
liiumiiittto^ 8 15 22
Days Smce DaburMtnent
ifl^LiULjjo.o
$6.0
$5.0
S4.0
$3.0
$20
$10
□ Dolar* RMtewrwd
Prepared by Abt Associates Inc. 12
Figure 2.2
Share of Monthly Transactions and Purchases By Store Type
100%
80%
60%
40%
20%
0%-*
□ # Transactions I I S Purchases
Prepared by Abt Associates Inc. 13
Figure 2.3
Average Purchase Amount at Supermarkets
l»r
Avenge Purchase Amount at Other Store Types
Dm*
r i • SM
Prepared by Abt Associates l*c. 14
Figure 2.4
Percent of Food Stamp Transactions
By Store Type
100*
»%
60%
40%
20%
I I'S'; ■ ■ ' "22
Day* Since Disbursement
supennkt grocery specialty —•— conveuencc
Percent of Food Stamp Dollars Redeemed
By Store Type
100%
80%
* «*
!
0% r^ i i i i i TT^" ' 'A' ' I J. I I I I I I _L I
Days Since
w
grocery
Prepared by Abt Associates Inc. 15
Chapter Two: Food Stamp Benefit Redemption Patterns
each county and by the metropolitan/non-metropolitan breakdown of the caseload.1S Over 523
million of food stamp benefits were issued in September to the analysis sample; nearly half of
that was disbursed in Baltimore City, so statewide statistics are heavily dominated by the
characteristics of the most urban population of recipients. The table shows that, on average,
recipient households make a little over ten purchase transactions per month, or about seven
transactions per $100 of benefits. The average purchase amount over the entire month is about
$18. Recipients in Baltimore City, on average, make more transactions for smaller purchase
amounts than recipients in the rest of the state.
The tar right columns of Table 2.1 show the speed of benefit exhaustion; this
essentially translates the "volume of redemptions per day since disbursement" (Figure 2.1) unto
statistics that characterize case-level behavior. The average recipient spends 23 percent of his
or her allotment on the day of disbursement, and 70 percent by the end of the first week;
halfway through the month, the average recipient has exhausted nearly 90 percent of his or her
benefits. The pattern of benefit exhaustion does not vary greatly across counties. Below we
explore other sources of heterogeneity.
Redemption Behavior by Case Type, Location, and Allotment Size
Tables 2.2 through 2.4 show how redemption behavior varies with the mix of the
caseload. These tables contain information for subgroups defined by: (1) type of welfare case
(FSP only, FSP plus AFDC cash benefits, and FSP plus DALP cash benefits); (2) location; and
(3) size of monthly allotment. For "location", we categorize recipients according to whether
they reside in a "metro" county (i.e., a county within the boundaries of an SMSA, excluding
Baltimore City) or a "non-metro" county (i.e., outside an SMSA), or whether they reside in
Baltimore City.
Table 2.2 shows the distribution of the caseload by these subgroups and the mean
number of food stamp purchase transactions per subgroup. This table yields two findings.
First, the difference in transaction volume between Baltimore City and the remainder of the
state cannot be explained by case type (i.e., public assistance (PA) versus non-public assistance
IS. "Metro" counties are those within SMSA borders, excluding Baltimore City; "non-metro" counties are
all other counties.
Prepared by Abt Associates Inc. \ 6
TABLE 2.1
FOOD STAMP BENEFIT REDEMPTION IN MARYLAND
CASES WITH 'REGULAR' MONTHLY DISBURSEMENT
SEPTEMBER 1993
State Total
N
•r
Total Total N
Aftf Average
Average
ABotawt
129.594 $23,461,179 $181.04 $23,582,220 1.336.697
pw
10.31
P«
$100 allot
Avi Percent of Sp««by:
$17.64
■°aJ 2aL Day 14
8881
J2SL 21
7.02 22.90 70.80 97.19
Non-Metro Counties 11.031 1,802.639 163.12 1.814.173 83.938 7.60 6.22 21.68 20.24 67.12 86 96 97.15
Metro Counties 52.809 9,739.278 184.42 9.801.227 403.432 7.68 5.25 24.34 20.13 65.54 85.70 9608
Baltimore City 63.721 11.916.269 181.32 11.963,634 847.183 12.89 8.57 14.12 25.57 75.65 91.61 98.09
County:
Alkgany
AnneArandel
3.034
5.476
494.448
1.031.624
162.97
188.39
494,290
1.038.102
27.099
41.242
8.93
7.53
7.28
5.10
18.24
25.17
21.06
21.77
65.56
68.17
85.80
87.63
96.35
97.04
Baltimore
Calvert
10,524
820
1.917,315
142.401
182.19
173.66
1.926.408
142.632
87.305
5.674
8.30
6.92
5.58
5.25
22.07
25.14
20.11
20.04
65.30
66.08
85.63
85.76
95.79
95.36
Caroline
Carroll
577
1.016
86.739
170.003
150.33
167.33
87.574
171.794
4.326
6.701
7.50
6.60
6.81
5.40
20.24
25.64
17.92
18.27
66.93
60.08
88.74
81.84
99.80
95.35
Cecil
Charles
1.754
2.114
302.732
401.252
172.60
189.81
305.503
402.150
12.183
15.586
6.95
7.37
S.ll
4.76
25.08
25.80
12.00
20.90
61.48
67.10
8265
85.77
95.23
95.38
Dorchester
Frederick
1.157
1.841
179.754
311.482
155.36
169.19
180.383
313.585
8.989
13.340
7.77
7.25
6.83
5.67
20.07
23.51
26.37
19.54
71.21
61.28
89.74
82 46
97.79
96.13
Garret!
Harfbrd
950
2.489
161.342
426.229
169.83
171.25
161.287
430.034
7.575
18.612
7.97
7.48
5.78
5.59
21.29
23.11
20.39
21.26
62 98
66.86
82.47
86.31
94.59
96.91
Howard
Kent
1.285
339
230.688
41.819
179.52
123.36
232.522
41.914
9.459
2.197
7.36
6.48
542
7.99
24.58
19.08
20.32
17.16
64.48
65.57
85 32
87 65
96 39
98.75
Montgomery
Prince George's
7.617
13.575
1.354.531
2.753.906
177.83
202.87
1.366.193
2.775.707
54.726
100.640
7.18
7.41
5.20
4.31
24.96
27.58
16.33
20.90
60.23
68.06
82.78
87.19
95.02
96.42
Queen Anne's
St Marys
429
1.519
62.547
280.880
143.80
184.91
63.479
283,398
2.646
11.998
6.17
7.90
6.00
5.35
23.99
23.62
14.95
19.29
63.99
65.50
85.57
85.88
98 70
96.41
Somerset
Talbot
910
665
143.695
93.333
157.91
140.33
143.435
94.761
7.513
4.162
8.26
6.26
7.18
6.31
19.09
22.77
26.37
15.40
72.58
65.06
90.56
86.32
97.91
97.17
Washington
Wteomko
Worcester
SSL
2,589
2,212
442.851
366,068
171.05
165.49
444.329
368.909
22.422
17.802
8.66
8.03
33
60
19
20
82
72
22 93
23 98
66.66
71.29
968
65.721
146.276
11.916.269
151. II
181.32
147.009
11.963^654
7.193
847J83
7.43
12.89
46
57
20
14
44
12
20 711
25.57
6642
75.65
85.94
89 39
87 74
91.61
95 63
98.43
97.78
98.09
Notes: • The number of transactions per $100 allotment i calculated first for each case, and then averaged over cases.
The sample of 129.594 cases excludes 2.8% (3.700 cases) that received benefits in Seriemher hut made mi redcmptiim
The Total caseload does not match the sum over regions due to 13 cases with missing county code.
transact**!*.
o
Table 2.2
MEAN NUMBER OF FOOD STAMP TRANSACTIONS
RyCaaaTjfa,)
Meaa Naaaber of Traaaartioaa
Food Stamp Food Stamp Food Stamp Food Stamp Food Stamp Food Stamp
Case Type: Onlv AAFDC ADALP Only AAFDC ADALP
State Total 30.963 63.449 15.182 4.9 13.2 9.8
N—Metro Ciaartai
Total •VS37 4.129 385 5.9 18.4 7.6
By Allotment Size
SIO 483 3 2 1.7 1.3 2.5
$11-100 2.117 160 28 3.4 4.3 4.1
$101-200 2.001 1.040 330 7.1 8.2 7.2
$201-300 916 2.0S1 24 10.3 10.4 8.3
$301-400 23S 672 1 13.2 13.7 13.0
$401-300 57 176 • 15.2 15.1 .
$501-600 20 25 , 15.9 16.4 #
MM4 5 2 • 18.2 12.5 •
anci—tin
Total 23,9*7 26,126 2,782 5J M 1A
B* Allotment Size:
S10 1.194 6 2 1.6 12 1.5
SI MOO 10.081 540 142 3.5 4.1 5.6
$101-200 7.884 6.302 2.405 6.8 7.3 7.1
$201-300 3.460 13.744 202 9.3 9.3 11.2
$301-400 938 4.327 21 12.1 11.9 13.9
$401-500 248 1.034 6 14.9 14.1 12.8
$501-600 82 133 3 14.2 15.8 23.3
$601 + 20 34 1 17.8 20.1 21.0
- -i a,
Total 2*515 33,192 12,014 RJ 16J 18.5
By AllotmentSize
SIO 883 24 3 2.4 2.0 4.0
$11-100 9.476 838 239 5.4 7.0 7.8
$101-200 7.459 7.915 11.456 10.5 12.7 10.4
$201-300 2.071 16.573 304 14.9 16.0 16.2
$301-400 454 5.967 12 19.7 20.8 18.2
$401-500 133 1.593 , 23.1 25.8
$501-600 30 226 ■ 31.3 32.5
$601 + 9 56 • 29.1 35.4 •
NOOK: September dubtn
State Total does not match
• Tk« 4P %(\ " - _——_—.
Ike aan ow N
^mvidi'nguh
fjom due to 13 c
ts( 3* across al
IT' monthly on tavaeaaeiv Jee footnote on
lotmentof$2-S
Table 2.1.
aaes wah snam
* The $10 aJloanent group contains 72 caa kxatwns) that received anal .
Prepared by Abt Associates Inc. 18
Chapter Two: Food Stamp Benefit Redemption Patterns
(NPA) cues), or by allotment level. For each case type (i.e., each column), Baltimore City
recipients exhibit a greater mean number of transactions than recipients in other geographic
areas. Within each column, Baltimore City recipients exhibit a greater mean number of
transactions at each allotment level.
Second, the difference in transaction behavior across case types (the column
differences) can be explained by the different distributions of monthly allotments for PA and
NPA households For example, in all geographic areas, food stamp-AFDC cases exhibit a
mean number of transactions that is nearly twice that of food stamp-only cases, but this
difference does not exist when we make the comparison at each allotment level. In other
words, there is very little difference in the within-area transactions behavior of NPA and PA
cases once we control for allotment size. The bottom line is that EBT transaction levels depend
primarily on the distribution of the food stamp caseload with respect to allotment size and area
of residence.
Table 2.3 presents the breakdown of the mean number of transactions per $100 of
benefits and the average purchase amount. Again, the point is to compare redemption behavior
across regions controlling for case mix. These tables also provide information about the
expected use of an EBT system. For example, the average number of transactions per case
increases with allotment size, but at a decreasing rate. To see this. Table 2.2 shows the mean
number of transactions per case rising with allotment size, whereas Table 2.3 shows the mean
number of transactions per $100 allotment decreasing with allotment size. This occurs because
(as shown on Table 2.3) the average purchase amount increases with allotment size. (Note that
the average purchase amount does not vary across cas*. type once we control for allotment; this
must be true, because the mean number of transactions does not vary across case type once we
control for allotment)
One of the most striking characteristics of food stamp benefit redemption behavior in
Maryland is the differential use of supermarkets in Baltimore City relative to the rest of the
state. Table 2.4 shows that food stamp recipients in Baltimore City redeem from 50-65 percent
of their benefits at supermarkets, whereas residents in the remainder of the state redeem over
80 percent of their benefits at supermarkets. In addition, 10 percent of all food stamp recipients
in Baltimore City never shopped at a supermarket in September 1993, whereas this is true of
only 2.5 percent of recipients in the remainder of the state. These differences cannot be
Prepared by Abt Associates Inc. 19
Table 2.3
MEAN TRANSACTIONS PER S100 DISBURSEMENT AND AVERAGE PURCHASE AMOUNTS
By Case Type, Location, and Allotment Sixe"
Mean Transactions Per $100 Allotment
Food Sump Food Stamp Food Stamp
Averafe Purchase Amount
Food Stamp Food Stamp Food Stamp
Case Type: Only &AFDC &DALP Onlv &AFDC &DALP
Stale Total 8.52 5.41 8.69 16.61 19.13 11.74
Non-Metro Counties
Total 7.29 4.48 6.45 19.80 23.55 16.71
B\ Allotment Size:
$10 17.46 13.33 25.00 6.93 7.18 3.99
$11-100 8.25 7.04 6.81 14.71 15.91 16.92
$101-200 5.47 5.00 6.51 18.95 20.42 15.78
$201-300 4.27 4.27 3.78 23.68 23.58 27.57
$301-400 3.82 3.99 4.17 26.25 25.07 24.02
$401-300 3.47 3.51 . 28.99 28.57
$501-600 3.00 3.12 . 33.32 32.06
gm+ 2.78 2.03 • 36.21 49.67
Metro Counties
Total 6.59 3.92 6.26 21.12 26.55 16.52
By Allotment Size:
S10 1649 25.00 15.00 7.41 7.02 6.67
SI 1-100 7.76 6.17 7.07 15.47 17.81 14.85
$101-200 5.30 4.29 6.33 19.66 23.64 15.88
$201-300 3.89 3.85 5.10 26.08 26.21 20.06
$301-400 3.48 3.49 3.99 28.86 28.71 24.36
$401-500 3.37 3.24 2.92 29.78 30.89 32.51
$501-600 2.66 2.96 4.35 37.63 33.85 22.47
$601 + 2.58 2.99 3.15 37.61 34.33 32.18
Baltimore City
Total 11.16 6.7t 9.32 12.32 15.45 10.86
By Allotment Size:
$10 24.93 23.82 40.00 4.65 5.11 2.50
$11-100 13.36 10.17 9.36 9.12 10.43 11.33
$101-200 8.56 7.31 9.37 12.12 13.79 10.74
$201-300 6.25 6.51 7.57 16.15 15.48 13.21
$301-400 5.68 6.09 5.06 17.65 16.43 19.88
$401-500 5.28 5.95 , 19.20 16.78
$$01-600 5.78 6.09 16.95 16.44 Ml* 4.21 5.41 • 23.82 18.57
£^-5fP»«n^***»««na« month Oses wi* "regular" inorrtily disbiinemem See footnote on Table 2.1.
• The SI0 allotment group contains 72 cases ( 3* across all locations) dm received an allotment of $2-9
Prepared by Abt Associates Inc.
Table 2.4
FOOD STAMP REDEMPTIONS AT SUPERMARKETS
By Cam Type, Locatioa, awl Allotment Sue*
RSSSVKSX
At!
FractUofCmesNTVERRedWa,
Beaefltt at Swpen—rkets
Subgroup
Food Stamp
Only
Food Stamp
& AFDC
Food Stamp
&DALP
Food Stamp
Only
Food Stamp
&AFDC
Food Sump
&DALP
State Total 0.77 0.72 0.56
Non-Metro Counties
Total 0.85 0.85 0.82
By Allotment Size:
S10 0.82 1.00 0.53
SI 1-100 0.87 0.87 0.88
$101-200 0.84 0.85 0.81
$201-300 0.86 0.85 0.84
$301-400 0.86 0.83 0.91
$401-500 0.85 0.82 .
$501-600 0.87 0.84
$601 + 0.83 0.98
Metro Counties
Total 0.86 0.83 0.81
By Allotment Size:
S10 0.86 0.80 1.00
$11-100 0.86 0.85 0.87
$101-200 0.85 0.85 0.81
$201-300 0.86 0.84 0.79
$301-400 0.84 0.82 0.69
$401-500 0.81 0.80 0.60
$501-600 0.81 0.82 0.77
$601 + 0.79 0.79 0.59
Baltimore City
Total 0.65 0.62 0.50
By Allotment Size:
S10 0.65 0.53 0.50
$11-100 0.68 0.66 0.55
$101-200 0.61 0.64 0.49
$201-300 0.67 0.62 0.58
$301-400 0.66 0.59 0.46
$401-300 0.64 0.58
$501-600 0.70 0.51
$601 + 0.54 0.52
0.07
0.05
0.04
•.12
0.03
0.01
0.01
MS
0.15
0.05
0.16 0.00 0.50
0.06 0.04 0.04
0.03 0.01 0.05
0.01 0.01 0.00
0.01 0.00 0.00
0.00 0.01 ,
0.00 0.00 ,
0.00 0.00 t
0.03
0.11 0.17 0.00
0.06 0.03 0.02
0.02 0.01 0.03
0.01 0.01 0.02
0.00 0.01 0.10
0.01 0.00 0.00
0.00 0.00 0.00
0.00 0.00 0.00
•.It
0.27 0.42 0.33
0.14 0.12 0.21
0.11 0.05 0.11
0.03 0.05 0.06
0.02 0.04 0.25
0.03 0.02 .
0.00 0.06 .
0.00 0.04
Notes: September disbursement month. Cases with 'regular' monthly disbursement . See footnote on Table 2 1
* The SI0 allotment group contains 72 caws ( 3% across all locations) that received an allotment of $2-9.
Prepared by Abt Associates Inc. 21
Chapter Two: Food Stamp Benefit Redemption Patterns
explained by case mix or allotment size. Below we explore possible reasons for this observed
difference in redemption behavior.
Food Stamp Redemption By Store Type
Two hypotheses come to mind as to why food stamp recipients in Baltimore City are
less likely to shop at supermarkets than food stamp recipients living elsewhere in Maryland.
First, there may simplv be a relative scarcity of supermarkets within the city. Alternatively,
there may be relatively more non-supermarket retail outlets in Baltimore City, and these stores
may serve as "intervening opportunities" when clients go shopping for food.16
At first glance, it appears that there is a relative scarcity of supermarkets in Baltimore
City. As shown in Table 2.5, less than 6 percent of all retailers in Baltimore City are
supermarkets, whereas supermarkets represent 18 percent of retailers in non-metro counties and
26 percent of retailers in metro counties. The distribution of retailers, however, can be a
midsleading indicator of "access" to supermarkets. In fact, Baltimore City has 1.02
supermarkets per 10,000 residents, compared to 1.08 supermarkets per 10,000 residents in all
other metro counties and 1.59 in non-metro counties. Although the Baltimore City retailer mix
is substantially different than that in other areas of the state, it does not necessarily reflect a
situation of inadequate access to large supermarkets.
Rather than due to a lack of supermarkets, food stamp recipients' relatively low use of
supermarkets in Baltimore City appears to be due to a high availability of small and medium
grocery stores and specialty food stores. There are 7.31 grocery stores per 10,000 residents in
Baltimore City, versus 1.59 grocery stores per 10,000 residents in non-metro counties and only
0.53 grocery stores per 10,000 residents in metro counties. This relative availability correlates
with use. Compared to food stamp recipients outside Baltimore City, those in Baltimore City
spend four times the amount of food stamp dollars in small and medium grocery stores. The
relative number (and use) of specialty food stores is also much higher in Baltimore City than
elsewhere in the state.
16. Other hypotheses can be imagined as well (e.g., relative differences in food prices or service levels), but
we have no data to explore these other hypotheses.
Prepared by Abt Associates Inc. 22
Table 2.5
FOOD STAMP REDEMPTIONS BY STORE TYPE AND LOCATION
Store Type
Number of Locations Nnanber of
Number Percent Tranancoow
row Store Types
Dollars
Nm-Mttn CmmHu
18.1%
18.1%
8.0%
Supermarket
Small/medium Grocery
Specially Food
68
68
30
64.1%
10.5%
2.7%
85.2%
5.0%
2.9%
Convenience Store
Comb. Groc/Gas
Comb. Groc/Merchan
140
12
17
37.3%
3.2%
4.5%
18.3%
1.6%
0.7%
5.1%
0.4%
0.3%
Other Stores 39 10.7% 2.1% 1.2%
Total Stores 374 100.0%
MtnCmmius
25.7%
12.7%
8.0%
100.0% 100.0%
Supermarket
Small/medium Grocery
Specialty Food
405
201
127
69.8%
7.8%
4.6%
85.1%
3.2%
5.5%
Ctmretitmct Stares:
Convenience Store
Comb. Groc/Gas
Comb. Groc/Mercban
643
19
18
40.7%
1.2%
1.1%
14.3%
0.9%
0.3%
3.0%
0.2%
0.1%
Other Stores 161 10.5% 2.2% 3.0%
Total Stores 1.574 100.0%
BMmonGty
100.0% 100.0%
Supermarket
Small/medium Grocery
Specialty Food
74
531
157
5.6%
40.3%
11.9%
28.6%
40.2%
10.3%
55.1%
19.1%
14.9%
Cemttnituct Storti.
Convenience Store
Comb. Groc/Gas
Comb. Groc/Mercban
IIS
11
5
14.3%
0.8%
0.4%
8.9%
0.6%
0.4%
3.9%
0.1%
0.3%
Other Stores 319 26.6% 11.0% 6.6%
Total Stores 1.285 100.0% 100.0% 100.0%
Total 3033
recajNents.
Prepared by Abt Associates Inc. 23
Chapter Two: Food Stamp Benefit Redemption Patterns
Interestingly, although the perceived lack of supermarkets in inner cities has—almost
as a corollary—led to concerns about recipients' use of higher-priced convenience stores, food
stamp recipients in Baltimore City shop in convenience stores less frequently, and spend
relatively fewer benefit dollars, than recipients elsewhere in the state.
An additional dimension of shopping behavior is the number of different retailers
accessed by a household during the month. Table 2.6 shows that food stamp recipients shopped
at an average of nearly five different stores during September 1993; only 14 percent of cases
shopped exclusively at a single retailer. Food stamp recipients in Baltimore City shopped at
nearly twice as many different stores than recipients in other areas.
Access to food stores is typically defined by locationa! proximity, but another important
dimension of access may be hours of operation. In fact, Figure 2.5 shows that the retailers that
do not specialize in staple foods (convenience stores, combination grocery and gas, and other
combinations) are disproportionately accessed by food stamp recipients during "off-peak" hours.
Thus, a story that is consistent with the data is not one of inner-city residents shopping
at the comer convenience store (except maybe late at night), but of inner-city residents making
many small transactions at several different stores. Furthermore, at least on a city-wide scale,
access to supermarkets does not appear to be a problem. It is still possible, of course, that
supermarkets in Baltimore City are not located in or near food stamp recipients' neighborhoods.
If so, and if grocery stores and specialty stores charge higher prices than supermarkets, then
lack of ready access to supermarkets remains a problem for food stamp recipients.
Benefit Exhaustion
Food stamp benefits are disbursed on a monthly basis. This fact has several
implications for both the recipients of food stamp benefits and evaluators of the FSP. As was
seen in Table 2.1, recipients redeem their food stamp benefits quickly following disbursement.
The average recipient redeems more than half of his or her disbursement within the first week.
This evidence suggests that a monthly disbursement begets monthly spending of food stamp
benefits. Food stamp spending patterns do not, however, necessarily reflect recipients' overall
food shopping patterns. Available survey evidence, though dated, shows that food stamps
comprise lets than 50 percent of recipients' total food budget for 53 percent of food stamp
Prepared by Abt Associates Inc. 24
TABLE 2.6
NUMBER AND TYPES OF STORES ACCESSED FOR FOOD STAMP REDEMPTION
SEPTEMBER 1993
-Regular"
Food
Cases
ATgJJ
DWeieat
Showed
Fraction of cana saopp4ac at:
FrT
shopoai
ctkm of ci
if exchisrv
Grocery
Stores
nes
eh at:
One
Store
2-3
Stores
4-5
Stores
Six+
Stores
Super-markets
Com
Store?'
Stale Total 129.594 434 0.14 0.32 0.25 0.29 0.28 0.01 0.00
Non-Metro Counties
Metro Counties**
Baltimore C«y
11.031
32.809
63.721
3.03
3.50
5.62
0.24
0.18
0.09
0.42
0.39
0.24
0.23
0.26
0.25
0.11
0.16
0.43
0.44
0.43
0.13
0.01
0.01
0.02
0.08
0.00
0.00
Comnn:
AUegany
Anne Aruodel
3.034
3.476
3.53
3.55
0.21
0.17
0.36
0.40
0.26
0.26
0.18
0.17
0.34
0.38
0.02
0.01
0.01
0.00
Baltimore
divert
10.324
120
3.76
3.25
0.17
0.21
0.37
0.41
0.26
0.25
0.20
0.13
0.4^
0.26
0.00
0.06
0.00
0.00
Caroline
Carroll
377
1,016
2.19
2.64
0.26
0.30
0.43
0.44
0.23
0.20
0.08
0.06
0.49
0.60
0.01
0.00
0.01
0.01
Cecil
Charles
1.734
2.114
2.69
3.56
0.28
0.16
0.46
0.40
0.21
0.27
0.06
0.17
0.46
0.34
0.01
0.01
0.01
0.00
Dorchester
Frederick
1.137
1.841
3.23
2.69
0.24
0.28
0.38
0.47
0.26
0.19
0.14
0.06
0.39
0.72
0.04
0.00
0.00
0.00
Garren
Hirford
930
2.489
2.96
3.32
0.20
0.22
0.47
0.39
0.27
0.25
0.07
0.14
0.37
0.47
0.01
0.01
001
0.01
Howard
Kent
1.285
339
3.07
2.14
0.23
0.42
0.46
0.43
0.19
0.13
0.12
0.03
0J6
0.66
0.01
0.01
0.00
0.01
Montgomery
Prince George's
7.617
13.375
3.30
3.66
0.21
0.14
0.40
0.40
0.24
0.29
0.14
0.17
0.55
0.36
0.01
0.00
0.00
0.00
Queen Anne's
St Mary's
429
1.519
2.60
3.31
0.33
0.21
0.42
0.41
0.19
0.26
0.06
0.13
0.60
0.44
0.01
0.02
0.00
0.00
Somerset
Tslbot
910
663
3.17
2.52
0.23
0.30
0.41
0.47
0.24
0.19
0.12
0.04
0.36
0.69
0.02
0.01
0.01
0.00
Washington
wcomico
2.589
2.212
3.56
3.48
0.18
0.21
0.37
0.38
0.28
0.24
0.17
0.17
0.37
0.37
0.01
0.01
0.01
0.02
Worcester
Baltimore City
968
63.721
2.88
5.62
0.26
0.09
0.43
0.24
0.23
0.25
0.08
0.43
0.52
0.13
0.01
0.02
0.00
0.00
Notes: "Regular" cases are those that received a 'negular" mcmthry food tump disburseraM on or 7m of both
September i
* Convenience Stores include
** Metro counties are
asegones do not sum to total smce some cases had county mstsmg.
grocery and gas combination" and "combination grocery A merchandise- categories.
m SMSAs: Baltimore Cfcy is excluded by this group and shown separately.
Prepared by Abt Associates Inc. 25
Figure 2.5
PERCENT OF FOOD STAMP TRANSACTIONS
I
.2
I
100%
80%
60%
40%
20%
Prepared by Abt Associates Inc. 26
Chapter Two: Food Stomp Benefit Redemption Patterns
households; only 14 percent of food stamp households rely on food stamps for more than 90
percent of their food purchases.17
The fact that more than half of all food stamp benefits are redeemed in the week
following disbursement leads us to ask if nearly all recipients spend half of their benefits in the
first week, or conversely, if a significant number of recipients completely exhaust their benefits
in the tint week whereas the remainder spread their redemptions throughout the month. Figure
2.6 shows the cumulative percent of cases exhausting their benefits by days since disbursement,
according to various definitions of "exhaustion." On each day since disbursement, the
percentage of cases reaching exhaustion increases at a slightly decreasing rate. By the end of
the first week, only a little more than 20 percent of all cases have completely exhausted their
benefits (i.e., less than $1 in benefits remains), though 35 percent have exhausted at least 95
percent of their benefits.
Figure 2.7 shows cumulative exhaustion by allotment size. This figure shows that
exhaustion of benefits early in the month is disproportionately due to the exhaustion of benefits
by recipients with the smallest monthly allotments (less than $200). Only about 20 percent of
recipients in the three largest allotment categories have exhausted their benefits by the end of
week one, compared to 30 to 50 percent of recipients with smaller allotments.
The graph in Figure 2.7 tells us how quickly recipients spend all of their benefits, but
it does not address the question of how benefit redemption is spread throughout the month.
Figure 2.8 shows the distribution of the cumulative percent of benefits spent on each of several
days following disbursement This graph shows that a surprisingly large 58 percent of all NPA
food stamp cases do not redeem any benefits on the day of disbursement: only 32 percent of PA
cases do not redeem benefits on day one By the seventh day, over 35 percei.. jf both PA and
NPA cases have spent at least 90 percent of their benefits (consistent with Figure 2.6);
however, this graph additionally shows that a majority of cases have redeemed at least 50
percent of their benefits by the end of week one. This evidence is consistent with the
17. Ohk. Jaws C. ud Harold Fteebovt, The Food Stamp Program: Design, Tradeoffs. Policy. and Impacts.
Washington, DC: The Urban Institute Press, 1994. The cited statistics are based on tabulations of the 1979-
10 Low-Iacoaw fcpjjt— to the 1977-78 Nationwide Food Consumption Survey
Prepared by Abt Associates Inc. 27
Figure 2.6
Cumulative Percent of Cases "Exhausted'
Pood Stamp and Caah BanafltCasac (PA)
100%
lto<*1 «»<l !*>«•» m<m
Food Stamp Only C«OM(NPA)
100%
0%
-<sS5SS*13
J^r-£- —^
r"TJpSr
^r®
■^
1 • IS
r~i—i a i i i —r—i—i—■r -T-t
»<|l |*<1 l*<10%
frqpararf fry Abt Associates Inc. 28
Figure 2.7
Cumulative Percent of Cases "Exhausted", By Benefit Level
Exhausted to < SI of Benefits
100%
20%
$10 $11-100 -*- $101-300 -*- $201-300 -*— $301-400 -*- $01-800
Prepared by Abt Associates Inc. 29
EH
Figure 2.1
Speed of Food Stamp Benefit Eihwwtkwi By Days Since Disbursement
Percent of Cases According to Percent of Benefits Redeemed
Day of 30^.
INPA I PA I NTA
SOaya 70aya
I1
1
•'oHnaTflHI
•■«■■■■■
ri-wJBJBJBJki
•"•JMIMb
•i« H BJBj
<<•■!■■
ltoHBJf
""•PaW
"•■JBT «•*■
•Bn-11"1" •T « 1 an ah «•% ■
30
>CMN I ■NPACMM
Chapter 7Wo: Food Stamp Benefit Redemption Patterns
hypothesis that food stamp recipients make major food purchases on a monthly, rather than
weekly, basis.
If food stamp recipients shop on a monthly basis, one of the obvious implications is
that it may be very difficult to rely on survey methods to collect the same types of information
from the food stamp population that we have assembled here from the EBT transactions log.
For example, questions about "last week" will contain a high degree of sampling variance
according to variance in the timing of the interview relative to disbursement. A possible
solution is for surveys to use "last month" as a reference period, but this introduces an unknown
amount of recall error.
Appendix B compares survey responses from a sample of Maryland EBT food stamp
recipients to the actual redemption behavior of these respondent is logged on the EBT system.
Unfortunately, the survey was conducted over a period of several months (each respondent was
interviewed once during die period from May to October 1993), and the transactions log
references the September disbursement month. The comparison is therefore not ideal, but it
does shed some light on the issue of survey reliability. The appendix examines items such as:
• In a typical month how often do you shop?
• One week after you receive your food stamp benefits, how much do you have left?
• In what type of store do you spend the majority of your food stamp benefits?
The results show that, on average, there is little difference in the redemption behaviors of
groups that report very different behaviors.
Food stamp recipients redeem their benefits quickly following disbursement, yet a
significant number of cases do not exhaust completely. This may be because it is not worth the.
effort of using the EBT card to redeem small amounts when those funds may accumulate
without penalty. In this section we examine the magnitude of funds that remain unredeemed at
month's end.
Table 2.7 shows the percent of cases who "carry-over" any benefits and the mean
dollar amount carried over. The percent of cases with carr/over is measured in two ways.
First we measure the percent of cases with a positive food stamp balance greater than $1 on the
Prepared by sibt Associates Inc. 31
Table 2.7
FOOD STAMP BENEFITS CARRIED OVER TO NEXT DISBURSEMENT MONTH
By CMC Type, Location, and Allotment She*
Food Staaap Oar* Cases F-J 5tn-» A Cask AMau.ce
Fraction of Cases Ends*
Cases
Fraction of < Avg Avg
Caseload
Bajancj_>j Ead
CMMJ
PlMTTTT*' End
Defn#l Defn#2 Defn#! Defn#2 Balance
Stale Total S3,*** 0J87 •-232 24.35 79,334 •JM •.1*9 23.18
Non-Metro Counties
Toul CMS 0.457 •-274 21.9* 4.392 •J97 •-247 2142
By Allotment Size.
$10 616 0.435 0.242 6.85 5 0.400 0.400 2.21
$11-100 3.024 0.418 0.249 15.05 199 0.307 0.181 14.28
$101-200 2.034 0.458 0.283 22.89 1.403 0.420 0.239 18.42
$201-300 915 0.527 0.317 37.48 2.103 0.391 0.258 20.11
$301-400 234 0.658 0.393 38.91 676 0.392 0.253 30.19
$401-500 57 0.684 0.368 58.55 179 0.402 0.235 38.26
$501-600 20 0.600 0.450 38.54 25 0.440 0.240 36.27
$601 + 5 0.600 0.400 31.19 2 0.500 0.500 13.20
Metro Counties
Total 25.416 • 459 •J76 24.94 29.17B •.413 •-263 24.60
By Allotment Size:
S10 1.646 0.442 0.230 8.64 12 0.250 0.167 7.08
$11-100 10.S79 0.421 0.250 17.21 702 0.268 0.171 11.00
$101-200 8.089 0.473 0.284 26.56 8.S42 0.422 0.261 21.01
1201-300 3.510 0.519 0.326 39.87 14.028 0.420 0.270 24.68
$301-400 941 0.550 0.366 41.54 4.369 0.397 0.258 29.25
S401-S00 249 0.570 0.394 47.28 1.046 0.402 0.272 39.35
$501-600 12 0.561 0.354 56.92 136 0.415 0.324 39.93
$601 + 20 0.650 0.400 73.12 35 0.486 0.371 39.40
Baltimore City
Total 21.629 0.280 • 166 2434 45,554 •.197 •.136 21*9
By Allotment Size:
SI0 1.082 0.311 0.143 7.62 28 0.143 0.143 4.06
$11-100 10.038 0.273 0.160 15.33 1.084 0.157 0.101 12.32
$101-200 7.794 0.267 0.163 32.76 19.631 0.114 0.123 20.82
$201-300 2.093 0.317 0.197 38.29 16.942 0.216 0.149 20.61
$301-400 448 0.357 0.219 31.09 5.992 0.191 0.140 24.62
$401-500 134 0.351 0.231 52.51 1.394 0.2O8 0.150 33.06
$501-600 31 0.452 0.387 98.81 227 0.176 0.115 42.69
$601 + 9 0.111 0.111 79.18 56 0.179 0.107 19.06
Saae total exceeds OK nan over regions due » cases with nussmg county code.
Definition »\ - aafiaj balance greater dm SI.
Definition #2 ■ disbursement for September maws total redemptions in Seatembci
Average end balance is the average balance of food samp benefits on the day prior
for cases wan greater than Si remaaing (defh #1).
The S10 alloonent group contanv a small number of cases recervmg $2-9.
transactions.
SI.
to die October disbursement.
Prepared by Abt Associates Inc.
Chapter Two: Food Stamp Benefit Redemption Patterns
day prior to the October disbursement (this is "Defn #1" in the table). This is a "long-term"
measure, because the end balance reflects any carryover from the previous month. The second
measure (Defn #2) defines carryover as the difference between the current month's
disbursement and redemptions—thereby measuring only the carryover that would occur if all
cases started the month with a zero balance.18
Nearly 40 percent of NPA food stamp cases and 30 percent of PA food stamp cases
had more than $1 of benefits remaining in their account just prior to the October disbursement
of benefits. For both groups, the average carryover was about $24. Although this average is
pushed upward by those cases that did not access any benefits in September, the average
carryover for the remaining cases is still sizeable—about $20.
As might be expected, the average dollar amount of carryover generally increases for
recipients with larger monthly allotments (especially when groups with small numbers of
recipient are excluded). For any given allotment-size group, NPA food stamp recipients
generally have more benefits left over at the end of *he month than PA food stamp cases. (The
overall PA and NPA means are similar only because of different distributions in allotment size.)
Out-of-SUte Transactions
Every EBT system implemented to date has had to address the issue of food stamp
recipients within the system's implementation area who want to shop at stores outside the
system's boundaries. Usually, at least a few such stores have been eruipped with POS
terminals capable of handling EBT transactions.
Food stamp recipients in Maryland accessed their benefits at 39 out-of-state retailers
between August and October 1993.19 In the aggregate, the amount of cross-border shopping
was small. These 39 stores represented about 1.2 percent of all food stores capable of handling
EBT sales. Only 0.3 percent of all food stamp purchases in September 1993, however,
18. The latter—Iwas used to examine benefit exhaustion in the previous section.
19. In order to receive the necessary POS terminals, out-of-state retailers had to request and finance the
equipment. In contrast, aD FSP-autborized retailers within Maryland bad the option of receivinf limited
function fie., EBT-oely) terminals at no cost. (In-state food retailers could also enter into contracts with the
EBT system vendor for nrmhj-functioo terminals that could accept commercial POS transactions as well as
EBT transactions.)
Prepared by Abt Associates Inc. 33
Chapter t\n: Food Stamp Benefit Redemption Patterns
occurred in these stores; these purchases represented 0.5 percent of all food stamp redemptions
that month.
Out-of-state transactions have been included in all analyses within this chapter except
those that break out redemption activity by both store type and location (i.e.. Tables 2.4 and
2.5).
Prepared by Abt Associates Inc
CHAPTER THREE
CASH BENEFIT REDEMPTIONS
Prior to the expansion of the Maryland EBT system, cash assistance benefits were
disbursed to recipients via government-issued check. Since implementation of EBT, recipients
access their benefits by using an "Independence" card and PIN number to withdraw their
benefits at automatic teller machines (ATM) and point-of-sale (POS) machines.20 Maryland
cash benefit recipients may access their benefits at any ATM machine that is part of the MOST
network.21 Access at POS machines may be made via purchase transaction, cash-back
transaction, or combination purchase and cash-back transaction.
An obvious question is, "Why do we care about the pattern of benefit redemption for
cash benefits?" Unlike the FSP, which issues an in-kind benefit for a particular targeted
consumption outcome, cash assistance has "no strings attached" in terms of the use of benefits.
As mentioned in the Introduction, however, a large part of the reason for examining benefit
redemption is to gain information about the transactions demand patterns of the cash benefit
caseload.
There are two considerations with respect to transactions demand: first, by examining
the frequency and timing of redemption activity, we learn about the necessary capacity of an
EBT system serving a welfare caseload. Second, by examining the case-level variation in
transactions activity, we learn more about the cost considerations of transferring benefit issuance
to EBT.
The evaluation of the expansion of the Maryland EBT system found that EBT was not
cost-neutral for the cash-assistance programs.22 That is, the EBT issuance system is more
costly than the paper system of check-issuance that it supplanted. The main component of the
increased cost is the ATM fee for cash withdrawals; the driving force behind the cost of ATM
fees is the number of transactions made per recipient. Recall that in a check-issuance system,
20. A small number of recipient! receive their benefits via direct deposit to their bank account.
21. This includes MOST ATMs inside and outside of the state of Maryland.
22. Logan a aL, "The Evaluation of the Expanded EBT Demonstratioa in Maryland, Volume 2: System
Impacts oa Program Costs and Integrity," Abt Associates Inc., Cambridge, MA, May 1994.
Prepared by Abt Associates Inc. 35
Chapter Three: Cash Benefit Redemptions
the state incurs the cost of issuing checks but not the cost of check-cashing. In an EBT system,
the state incurs the fee cost for the ATM transactions that give recipients access to their cash
benefits. Hence, the mix of ATM versus POS transactions and the number of ATM transactions
per case have important cost implications for EBT issuance. Yet it is hard to know the extent
to which the Maryland cost-benefit results may be generalized to other regions of the country.
For example, do the number of transactions per case vary with the mix of the caseload? With
the availability of ATM machines? With the size of allotment? Do multiple transactions per
case represent distinctly separate transaction occasions, or nearly simultaneous withdrawals—
i.e., is there a reasonable chance for reducing the number of transactions per case through
better training of recipients in the use of ATM machines? This chapter addresses some of these
questions.
Caseload Composition
The EBT system serves nearly all beneficiaries of cash public assistance in Maryland.
For the most part, this includes recipients of AFDC and DALP.23 Only a very small number
of recipients of Public Assistance for Adults (PAA) were in the EBT system in September 1993,
because only a few local offices of the Maryland Department of Social Services placed their
PAA caseloads on EBT. PAA is a very small program relative to AFDC and DALP, however.
In addition to the public assistance programs, the NPA Child Support (NPACS) caseload of
Baltimore City was part of the EBT system in 1993.
Table 3.1 shows the composition of the Maryland cash assistance caseload in
September 1993. As seen in the table, the AFDC caseload accounts for 86 percent of the cash
assistance caseload and 91 percent of cash benefit disbursements outside of Baltimore City.24
The caseload composition is quite different in Baltimore City, with the AFDC caseload
accounting for 67 percent of all cases and 79 percent of all disbursements. The difference is
due to the much larger DALP caseload in Baltimore City. As with the food stamp caseload,
Baltimore City accounts for more than SO percent of the entire cash benefit caseload, which
23. The Disability Assistance Loan Program (DALP) was called General Public Assistance (GPA) prior to
December 1992.
24. In all tables we group the PAA cases together with the DALP cases, because there were only 78 PAA
cases on the system in September 1993.
Prepared by Abt Associates Inc. 36
Tabk3.1
THE CASH BENEFIT CASELOAD IN MARYLAND
September 1993
Non-Metro Metro Baltimore
State Total Counties Counties Cfcy
Total Caseload 104,180 *\517 39306 58,119
Ptitttt by Ptvgium
AFDC 75.63* 86.31* 86.88* 66.79*
AFDC-Unenployed Parents 0.83% 2.12* 1.50* 0.23*
Disability Auisunce Loan Program* 19.75* 9.81* 10.91* 26.88*
Child Support Bonus Only 0.56* 1.76* 0.71* 0.31*
NPA-Child Support 3.23* NA 0.01* 5.78*
Percent of AFDC cases with UJ4% 3036* 1736* 1339*
Child Support Bonus
Total Dnba-nnull 30,443318 1,9B3,941 12306300 16370369
Percertby Program
AFDC 85.03* 91.12* 91.28* 79.37*
AFDC-Unemployed Puma 1.23* 2.85* 2.10* 0.38*
Disability Assistance Loan Program 11.60* 5.73* 6.49* 16.19*
Child Support Bonus 0.10* 0.30* 0.12* 0.06*
NPA-Child Support 2.13* NA 0.01* 4.00*
Average Disbursement
AFDC 328.55 321.37 327.32 330.63
AFDC-Unemployed Parents 434.45 409.11 437.74 U5.89
Disability Assistance Loan Program 171.63 178.02 185.26 167.60
Child Support Bonus Only 51.37 52.35 5131 51.01
NPA-Child Support 192.80 NA 419.56 192.67
Percent receiving Food Stamps
Overall 8438% 78.38% 80.62* 88.22%
AFDC 85.63* 79.86* 81.75* 89.89*
AFDC-Unemployed Parents 93.18* 88.41* 93.40* 97.06*
Disability Assistance Loan Program 81.31* 67.14* 72.36* 84.38*
Child Support Bonus Only 48.19* 5652* 41.94* 53.30*
Notes: Table reflects entire caseload and is not limned to cases receiving "regular" disbursement!
Regions do not sum to state total because 38 cases are missing county code.
The average Child Support Bonus exceeds $50 due u the occurrence of iiaibiplf anjajpsj by some cues.
Disability Assistance Loan Program (DALP) includes 78 Pubbc Assiatsace fcr Adults (PAA) cases.
Prepared by Abt Associates Inc 37
Chapter Three: Cash Benefit Rtdemptions
means that aggregate redemption statistics are heavily dominated by the behavior of Baltimore
City recipients.
The NPACS cases of Baltimore City receive their payments via EBT, but these
payments are not disbursed on a regular schedule, so we do not include these cases in the
anaWsis of frequency and timing of transactions.23 A small percent of the caseload received
on., the child support bonus; these are probably cases that recently left the AFDC caseload.
Some important characteristics of the caseload are the average disbursement amount,
the percent of the AFDC caseload that receives the bonus child support payment, and the
percent of the caseload that receives food stamps. The average disbursement amount may
increase the number of transactions per case; receipt of the bonus will certainly increase the
number of transactions, because the bonus is not disbursed concurrent with the AFDC payment.
Receipt of food stamps may increase the percent of cash benefit withdrawals made at POS
machines, relative to ATMs, because food stamp recipients are familiar with the POS machines.
Table 3.1 shows that 16 percent of the AFDC caseload receives the bonus child support
payment; this varies from 31 percent of AFDC cases in non-metro counties to only 13 percent
of AFDC cases in Baltimore City. The percent of cash assistance cases receiving food stamps
varies from 78 percent in non-metro counties to 88 percent in Baltimore City.
All remaining tables and graphs presented in this chapter are based on the sample of
cash benefit cases with a "regular" monthly disbursement. These are cases that received a
disbursement on one of the "regular" disbursement days (the first, second, or third of the
month) in both September and October. As discussed earlier in this report, we limit our
analysis sample to these "regular" cases so as to get a clearer picture of redemption behavior
relative to the time since disbursement.26
25. NPACS recipients receive disbursements after the state receives a payment from the absent parent This
means that the timing of disbursements is highly variable and that multiple disbursements may occur each
month, depending on both the payment schedule and the compliance of the absent parent.
26. Consistent with the food stamp sample definition, we exclude cases that received a disbursement but
made no cash withdrawal transactions in September. These cases account for less than 2 percent of cases
receiving a regular disbursement.
Prepared by Abt Associates Inc. 38
Chapter Three: Cash Benefit Redemptions
Timing and Frequency of Cash Redemptions
Recipients of cash assistance may redeem, or access, their benefits in any of four
different ways in the Maryland EBT system:
(1) withdrawal of cash at any participating ATM (which includes all ATMs in the
MOST network);
(2) receipt of "cash-back" from any EBT-equipped retailer (food or non-food) who
agrees to provide cash to recipients using their EBT card;
(3) purchase of goods at any EBT-equipped retailer (food or non-food); and
(4) a combination purchase and cash-back at any retailer agreeing to provide cash-back.
The transactions log data do not allow one to distinguish among the latter three types of
transactions. This means that we cannot separately identify cash assistance recipients' EBT
withdrawals from their EBT purchases. Thus, even though we often refer to cash
"withdrawals" in this chapter, we note that this term is broadly defined to include all four types
of possible redemption transactions.
Figures 3.1 and 3.2 show the main characteristics of cash redemptions. An
overwhelming majority of transactions occur on the day of disbursement (this is spread over
three calendar days in the top panel and concentrated on "day one since disbursement" in the
lower panel); after falling dramatically, the number of transactions peaks again on the 21st of
the month when the child support bonus is disbursed. Figure 3.2 shows the average withdrawal
amount dropping sharply after die day of disbursement, hovering for a few days at $40 for
AFDC cases, and then staving near $20 for both AFDC and DALP cases. The AFDC average
withdrawal amount rises slightly on the 19th to 21st days since disbursement, when the child
support bonus is disbursed.
The lower panel of Figure 3.2 shows the percent of transactions that occur at ATMs
(versus PCS machines). Over the entire month, S3 percent of all cash benefit redemptions
occur at ATMs: 56 percent for AFDC cases, and 44 percent for DALP cases. The overall
percent of transactions at ATMs peaks at 67 percent on the day of disbursement, falls
Prepared by Abt Associates Inc.
3.1
VOLUME OF CASH BENEFIT WITHDRAWALS
AFDC 4 PALP CASES WITH •REGULAR' MONTHLY DBMJBSEMENT
SEPTEMBER 1993
•8.0
10
T
llllfckMfWWvwi/iW* is 22
Day
frpfWVm
•60
84.0
830
82.0
81.0
800
n
e.
140
120
100
30
rw.fi,ri
is
J1£&A■fl-n-Pfi n^ii,
8140
81X0
•10.0
88.0
88.0
' ICidiWj SWi
Prepared by Abt Associates Inc.
IJ
CASH BENEFIT WITHDRAWALS
AHX: * DALP CASES WITH "REGULAR-MONTHLY DISBURSEMENT
SEPTEMBER 1993
PvcMrt or Jmmmtmm at ATM*
100% r
40%
20%
Prepared by At* Associates Inc
Chapter Three: Cash Benefit Redemptions
throughout the month to 31 percent on day 18, and then rises again to 44 percent when the child
support bonus is disbursed.27
Table 3.2 presents the main characteristics of cash benefit redemption; these
characteristics are shown separately for AFDC and DALP cases, and by area of residence (i.e..
metro, non-metro, or Baltimore City).21 On average, recipients make 4.65 redemption
transactions per month, or 1.77 transactions per $100 of benefits.29 The average withdrawal
amount is just under $100, and is equal to about one-third of the average benefit amount. In the
aggregate, 61 percent of all benefits are redeemed on the day of disbursement, and 91 percent
are redeemed by the end of the first week in the disbursement month.
There is some variation in die above characteristics by case type and region. For
example, AFDC cases make more transactions per case and fewer transactions per $100 of
benefits; this reflects the fact that, on average, AFDC benefits are twice as large as DALP
benefits. AFDC and DALP cases exhaust benefits at about the same speed.30 Recipients in
Baltimore County make about one more transaction per case Jian recipients elsewhere, and
more transactions per $100 or benefits.
Location of Cash Redemptions
Table 3.3 shows the total number of cash benefit redemption locations in the state of
Maryland and by region; the percent of locations that are ATMs; and the distribution of
redemption activity at ATM versus PCS locations. The count of redemption (or withdrawal)
locations is equal to the number of FSP-authorized retailers (i.e., the number of POS locations)
27. The child support boaus is disbursed to nearly all cases on the 21st of the month Because AFDC
disbursements are staggered over a three-day period, however, the child support bonus appears itau>sut oa
a "days since disbursement" timeline.
28. Analogous tables with county-level information appear in Appendix D.
29. Recall that cash benefit recipients may withdraw cash at ATMs or use POS machines for purchases,
cash-back, or combination purchase and cash-back transactions. By "redemption" or "transaction" we refer
to any of transaction resulting in a debit from the account. Balance inquiry transactions are not included in
this analysis.
30. The child support boaus is not included ia the "total benefit amount" for lac purpose of calculating the
timing of benefit exhaustion.
Prepared by Abt Associates Inc. 42
1
Table 3 2
CASH BENEFIT REDEMPTION IN MARYLAND
SEPTEMBER 1993
i Total Average
Total
Withdrawal
Average
Withdrawal
AvgfTt
per
case
P"
$100 allot.
Aggregate rractiaa of 1
Day 1 Day2 Day3
teemed*,:
D»7
2
Total 'Regular' Cam 74.457 21.583.161 289.90 346.355 99.71 4.65 1.77 0.61 0.74 0.80 0.91
AFDCCawI with "Regular" Monthly m n
Non-Metro Counties
Metro Counties
Baltimore City
3.909
23.636
29.303
1.264.063
8.281.274
9.537.587
323.37
323.03
325.48
18.262
110.486
158.164
96 88
111.61
103.75
4.67
4.31
5.40
165
1.45
1.82
0.57
055
0.66
0.70
0.69
0.78
0.71
0.76
0.83
0.89
0.89
093
ft
Total AFDC 58.848 19.082.924 324.27 286.912 106.76 4.88 1.65 060 0.74 0.80 091
DALP Canet with "Regular" Monthly
Non-Metro Counties
Metro Counties
Baltimore City
467
3.167
11.954
75.744
535.750
1.883.734
162.19
169.17
157.58
1.414
9.956
47.959
77.82
83.41
70.12
3.04
3.16
4.02
195
1.91
2.55
0.46
046
0.68
0.61
0.60
081
0.70
0 68
0.83
0.83
086
0.93
Total DALP 13.588 2.495.229 160 07 59.329 73.05 382 2.41 063 0.76 0.81 092
Notes: Regions do not sum to total due to 21 cases with missing county code.
The average number of transactions per $100 allotment is calculated first per case and then averaged over cases.
Number of "withdrawal' transactions include POS transactions that may be purchase only, purchase and cash-back, or cash hat*, only transactions
&
Table 3.3
CASH BENEFIT REDEMPTIONS BY LOCATION
SEPTEMBER 1993
Tool
Withdraw,! Uc*»nsu»
Percent
Fractional Al
ATM POS
Ihw—ifA—i
Withdrawn fraaa:
A»gNtn
ATM
after at
tamMBha1 amti
POS
Fraction af cases
llllanl ill Banal
ATM POS
Average
Number ATM ATM POS ATM
5.037 34.2* 0.53 0.47 0.74 0.26 2 46 2.19 0 16 0.20 116 91 54.18
AFDC wsth -Refaar" Manthry
Nan-Metro Counties 459 183% 0.63 0.37 0 85 0.15 2.86 1.8* 023 009 129.05 31.54
Metro Counties 2.MI 45.2% 0.66 0.34 0 85 0.15 2.90 1.41 0.29 0.10 117.26 42.45
Bahimore City 1.660 20.2% 0.44 0.56 0.67 0.33 2.45 2.95 008 0.24 111.98 66.69
Total AFDC 5.037 34.2% 0.55 0.45 0.76 0.24 2.65 2.28 0.18 0.17 122.80 55.12
DALP Cases with "RegssSar" Monthly D
Non Metro Counties 459 18.3% 0.59 0.41 0.76 0.24 1.81 1.22 0.29 0.17 92.11 41 II
Metro Coaaam 2.881 45.2% 0.63 0.37 0.79 0.21 1.98 1.16 0.33 0.16 98.47 39.17
Baltimore Cky 1.660 20.2% 0.34 0.66 0.56 0.44 1.56 2.46 0.05 035 72.62 52.45
Total DALP 5.037 34.2% 0.41 0.59 0.61 0.39 1.63 2.21 0.11 0.30 77.42 49.95
• Number of withdrawal
Total aaaaber of location
location! ant ant reflected M
Withdrawals include POS
include ATM tocatioaa and FOB locations
37 ATM locations in Maryland for
taote. rony-iive aaaaorizt
■at may be purchase oaky, purchase
Note that we count locations and not terminals
■r could not determine county Oul-of-state retailer ami ATM
locations arc included is the count of retailers in Bahimore City
and cam back, or cash hack only transactions.
ft
Chapter Three: Cash Benefit Redemptions
plus the number of ATM locations. Note that we count locations and not terminals.31
Overall, 33 percent of cash withdrawal locations are ATMs, though this varies considerably by
region.
Fifty-three percent of all cash benefit transactions occur at ATMs and 74 percent of
dollars are withdrawn from ATMs. The average recipient makes nearly the same number of
transactions at ATMs and POS terminals, but this is driven largely by the number of POS
transactions by recipients in Baltimore City. Outside of Baltimore City, recipients make nearly
twice as many ATM transactions as POS transactions per month. Likewise, the overall percent
of recipients that exclusively use ATMs (16 percent) is not very different from the percent that
exclusively use POS machines (20 percent). Outside of Baltimore City, however, 29 percent
of recipients exclusively use ATMs, whereas only 10 percent exclusively use POS machines
(i.e., never use ATMs).
The relative use of ATM versus POS machines varies somewhat according to whether
cash benefit recipients also receive food stamps. This would be expected if food stamp
recipients find it efficient to withdraw their cash benefits when making food stamp transactions,
or if their cash transactions are influenced by their familiarity with the POS terminals. On the
other hand, food stamp disbursements occur four days after cash benefit disbursements, so that
for most recipients of both cash benefits and food stamps, there may be little overlap in the
timing of redemption of these two types of benefits. In fact, we compared cash recipients who
receive food stamps to those without food stamps, and found the following:
31. Wc did no* have an independent source of the number of ATM terminals or locations in the Male of
Maryland, and therefore relied on me EBT transactions log for this information Our count of ATM locations
is mas i count of ATMs that were actually accessed by the welfare caseload. We thus may have
underoounted the total number of ATM locations available to the caseload (we have no way of counting ATM
nt which no fransartioni occurred during the time period under study). In order to minimize the
t, wa used the entire transactions log that we received from Deluxe (transactions records from
Angus* 3, 1993 through October 31, 1993) to obtain a list of all ATM terminals accessed by me
during that three-month period. The number of ATM locations reflected in Tabk 3.3 is a count of i
slieet addresses for all ATM terminals.
Prepared by Abt Associates Inc. 45
Chapter Three: Cash Benefit Redemptions
Ciaifiid
Percent of Cash
Trwwaction&JLAIMS
AFDConly
AFDC and food stamps
64.5%
54.0%
DALPonly
DALP and food stamp
52.2%
42.0%
Percent of Dollars
Withdrawn at ATM?*
79.7%
75.4%
70.5%
59.7%
The chart shows differences in ATM utilization between cash-only and cash plus food stamp
cases. These differences, however, may be partly due to the fact that cash recipients who also
receive food stamps have higher average cash benefit allotments than those without food stamps.
One might think that the relative use of ATM and POS machines is related to the
relative availability of those machines. Table 3.3, however, shows that recipients in non-metro
counties have the same low relative availability of ATMs as recipients of Baltimore City; yet
non-metro recipients utilize ATMs in the same way that recipients of metro counties use ATMs
(making over 60 percent of their transactions at ATMs). We investigate this point further in
Figure 3.3, where we graph ATM utilization versus ATM availability for each Maryland
county.
In Figure 3.3 each "bubble" represents a county, and the size of each bubble represents
the size of the county caseload relative to the statewide total. This figure shows that the
aggregate data mask a generally positive relation between the percent of withdrawal locations
that are ATMs and the percent of cash transactions made at ATMs by recipients in the county.
There appears to be an upper bound on ATM utilization, however. Even in counties with a
very high percent of ATM locations, the percent of transactions at ATMs does not rise much
above 70 percent. This explains why non-metro and metro counties, in the aggregate, have
very different concentrations of ATM locations, but the same mean utilization rates.32 The
county-level variability in both ATM access and ATM utilization suggests that the projected cost
32. The apparent upper bound on ATM utilization in metro and non-metro counties cannot be explained by
the presence of POS machines at "check-cashing" establishments—authorized check cashers are present only
in Baltimore City.
Prepared by Abt Associates Inc. 46
«? I 100%
Figure 3.3
ATM Availability and ATM Utilization in Maryland Counties
30% 40% 50% 60% 70%
ATM-Percent of Withdrawal Locations
Metro County | | Non-Metro County
100%
Note: Each "bubbto" rapraMrti a county. Tha ralaWw ajaj oflha bubMa raltacte
Tha x-y coordinalM occur at tw canlar d tw bubbta.
tha ratatova lizi of tfw cash benefit cntload
rI
Chapter Three: Cash Benefit Redemptions
of future EBT systems must consider the existing infrastructure in determining relative
utilization rates and the resulting cost of ATM fees.
In addition to showing the relative use of ATM and POS locations. Table 3.3 shows
that ATMs and POS devices seem to be used for different purposes. The average withdrawal
amount at ATMs is more than twice the average withdrawal amount at POS locations; outside
of Baltimore City this difference is fourfold. Large cash withdrawals are made at ATMs,
whereas much smaller cash-back or purchase transactions are made at POS locations; this is
consistent with the pictures in Figure 3.2, which show that the percent of transactions at ATMs
falls throughout the month as the average withdrawal amount falls.
irior by Case Type, Location, and Allotment Si»
Table 3.4 shows how cash redemption behavior varies with the mix of the caseload.
The mean number of cash transactions rises with allotment size, whereas the average number
of transactions per $100 allotment falls with allotment sue. Comparisons between AFDC and
DALP cases are informative mainly for the $101-200 allotment group that contains the majority
of DALP cases; here we see that DALP cases make significantly more transactions than AFDC
cases in metro counties and Baltimore City, whereas there is no difference between DALP and
AFDC cases in non-metro counties. Table 3.S contains additional characteristics of cash
transactions by case type, location, and allotment size
The speed of exhaustion for cash benefits can be compared to that of food stamp
benefits by comparing Figure 3.4 to Figure 2.6. It is not surprising lo find that cash benefits
are exhausted at a much faster rate than food stamp benefits (as evidenced by the steeper slope
of the cash exhaustion curve). The differential rates of exhaustion may be explained by the fact
that cash benefits are likely to be used for items like rent payments at the beginning of the
month. In addition, recipients may withdraw their cash benefits from their EBT accounts prior
to actually using those benefits, whereas food stamp benefits cannot similarly be accessed and
"stored'' prior to use.
The differential rates of exhaustion for cash benefits and food stamp benefits is
consistent with the difference in mean transactions for these cases: cash recipients make an
Prepared by Abt Associates Inc. 48
Table 3.4
MEAN NUMBER OF CASH BENEFIT TRANSACTIONS
By CMType, Location aad Attotmat Sbe
oTTr
Meaa Traasactioas '
per 8160 ABotawt
AFDC PALP Su., Toui—wsr -n?—m-
N—fcn of
AFDC DALP AFDC -nr DALP ■nr
Noa-Metro Co—ties
Total
Bv Allotment Size
$1-100
$101-200
$201-300
$301-400
$401-300
$301-600
$601 +
3. 467
SO 37
493 396
1.550 23
1.230 3
391 2
149 1
14 5
4.67 3.04
2.S1 2.03
3.02 3.03
4.40 4.09
5.22 5.67
5.92 4.50
5.74 6.00
8.86 3.40
1.65 1.9S
5.38 2.44
1.99 1.93
1.62 1.96
1.47 1.80
1.33 1.08
1.09 1.03
1.39 0.49
Metre
Total
B\ Allotment Size
$1-100
$101-200
$201-300
$301-400
$401-500
$501-600
S601 +
yjgan 3.167
260 18
2.763 2.962
10.867 159
8.123 II
2.510 8
980 7
131 2
4.31 3.16
2.31 2.17
2.70 3.12
4.09 4.16
4.75 2.36
5.22 3.75
570 3.43
6.69 5.00
1.45 1.91
4.98 4.37
1.75 1.91
1.49 1.90
1.33 0.65
1.17 0.85
1.08 0.64
1 01 0.49
City
Total
*v Allotment Sue
$1-100
$101-200
$201-300
$301-400
$401-300
$501-600
$601 ♦
»Jt3 11,954
340 21
2.704 11.756
11.881 177
9.351 0
3.612 0
1.259 0
156 0
5.4
2 82 3.10
3 42 4.01
492 4.86
588 .
656 .
7.65 .
874 .
IKS
Regions do not sum to total one to 21 cann
DALP - Disability Assistance Loam Profraa
1-82
8 11 4 06
2.24 2 55
182 2 32
1.66 •
148 .
1 45 .
1 32 .
Prepared by At* Associates Imc.
Table 3.5
MEAN CHARACTERISTICS OF CASH
By CMC Type, Location, aw
TRANSACTIONS
Size
AvgFracttaaaf
at ATMs
Art Fraction of Cases
NEVER anac ATMs
State Total
AFDC -nnr PALP AFDC DALP -nr AFDC -rrr DALP
TST
Nca-MetroC«
Tatal
B\ Allotment Size
SI-100
$101-200
$201-300
$301-400
$401-500
$501-600
$601*
77.82 t.63 •J9
32.99 56.67 0.42 0.81
78.53 78.64 0.67 0.56
96.41 68 44 0.64 0.53
99.44 55.82 0.61 0.12
111.38 101.50 0.67 0.92
137.50 96.67 0.64 1.00
99.38 212.20 0.60 0.93
6.09 0.17
0.33 0.03
0.12 0.19
0.09 0.09
0.09 0.00
0.06 0.00
0.07 0.00
0.07 0.00
Metro)
Tatal
Bv Allotment Size
$1-100
$101 200
$201 300
$301-400
$401-500
$501-600
$60! +
11141 83.41 8.66 8.63
36.10 40.85 0.53 0.55
89.24 82.24 0.67 0.62
107.14 89.78 0.68 0.69
116.58 234.97 0.64 0.61
132.53 156.53 0.68 0.94
144.40 186.71 0.65 0.95
150.97 205.83 0.69 1.00
8.18 8.16
0.25 0.28
0.16 0.16
0 08 0.08
0.09 0.18
0 08 0.00
0 08 0.00
0.07 0.00
Ta
ffv
i§3.75
Sue
$1-100
$101-200
$201-300
$301-400
$401 500
$501-600
$601 ♦
-8.12 0.34
33.60 45.49 0.28 0.26
78.09 70.17 0.42 0.34
95.38 70.09 0.46 046
110.03 - 0.43 •
125.89 • 0.44 •
138.81 - 0.43 •
165 58 - 0.40 .
8.35
0 48 0.32
0.33 0.35
0.22 0.14
0 24 .
0.23 .
0.22 •
0.26 .
Prepared by Abt Associatei Inc 50
3.4
100%
20% ^T—r
1
('Regular' AFDC asd DALP can ovy)
_ . -^-
■ i l i i i -. ■ T i i i ■ r ; i : , i i
8 15 22
Day* Sine* Disbursement
Exhaust to < $100-e- Exhauatto<5% -©- Exhaust to < 10%
by Abt Associates lmc. 51
Table 3.6
CASH BENEFIT EXHAUSTION
SEPTEMBER 1993
Avg#
W/D
PerCi
Fraction of cases
Exhurttai WHh:
One Two
W/D W/D*
Average
A^A^—t
mWk*
Toul "Regular" Cases 4.65 0.14 0.14 289.90 136.30
AFDCC Me*wttk1 Hilhly DM
Non-Metro Countifs
Metro Counties
Baltimore City
4.67
4.31
5.40
0.06
0.09
0.15
0.18
0.17
0.11
323.37
323.03
323.48
168.74
181.29
161.03
Total AFDC 4.88 0.12 0.14 324.27 170.37
DALPi
Noo-Metro Counties 3.03 0.18 0.23 162.19 104.80
Metro Counties 3.14 0.18 0.23 169.17 114.08
Baltimore City 4.01 0.24 0.13 137.38 100.20
Toul DALP 3.81 0.23 0.13 160.07 103.15
Note: 'Withdrawals" include POS transactions that nay be
cash-back, or cash-back only transactions.
Exhaustion is defined as the withdrawal of the full disbursement
on*.
Prepared by Abi Associates Int. 52
Chapter Three Cash Benefit Redemption*
avenge of five transactions per month, whereas food stamp recipients make an avenge of ten
transactions per month. Table 3.6 shows that 28 percent of cash assistance recipients withdraw
all benefits with only one or two transactions. Recipients in Baltimore City are more likely to
withdraw all benefits in a single transaction than recipients in other areas of the state.
From the evidenct presented so far, it is hard to tell if cash benefit recipients use their
EBT accounts as bank accounts, leaving their benefits in a safe place until needed. Recipients
exhaust their benefits very quickly (90 percent of aggregate disbursements is withdrawn by the
end of week one), yet only 28 percent completely exhaust their benefits with one or two
transactions. There are several reasons why recipients may need to make several transactions
even if they exhaust their benefits within just one or two days afar disbursement. First. ATM
machines often impose limits on the dollar amount that may be withdrawn per transaction—thus
requiring recipients to make several transactions in order to withdraw their full benefit.
Second, ATM machines dispense only even dollar amounts, and typically only multiples of ten
dollars. This means that recipients must visit both ATM and POS locations in order to
withdraw their full benefit, and it increases the likelihood that they will carry a positive balance
over to the next disbursement month.
Table 3.7 shows the "carryover" behavior of the cash assistance caseload. Nearly 24
percent of AFDC cases have a positive balance at the end of the month. The end balance,
however, reflects the cumulative carryover from all past months; only half of all cases with a
positive balance redeemed less than the current month's disbursement. Among cases with an
end balance greater than $1, the mean balance is significant. Hence, carryover behavior cannot
be explained by the "even-dollar" withdrawal constraints of the ATM terminals.
Finally, in Figure 3.5, we examine the distribution of the caseload according to the
percent of benefits withdrawn by days since disbursement. Figure 3.5 shows that on the day
of disbursement 25 percent of AFDC cases have withdrawn their full benefit; an additional IS
percent have withdrawn between 91 and 99 percent of their benefits; and 25 percent of AFDC
cases have made no withdrawals. Hence, 65 percent of the caseload have either nearly
exhausted their benefits or made no transactions. Looking at the other graphs on this figure we
see that benefit withdrawal behavior is characterized by a bimodal distribution: on each day
shown, most recipients had either withdrawn at least 91 percent of their benefits or none of
their benefits.
Prepared by Abt Associates Inc. S3
Table 3.7
CASH ASSISTANCE BENEFITS CARRIED OVER TO NEXT DISBURSEMENT MONTH
By Cast Type,
AflW Cnac, DAUCW
Fractawaf< Fr
cJ BalancedV*9** - . -M
Detail
!§jajaa>J
Defntt
I A«tEad
Defn#l Detn#2 ■abate
State Tetal 68J17 •437 •.128 35.41 15.954 8.17B •- 3IJ8
■u ||-.,n a— «Maaa.4aaaa
TottJ 4,**4 •JOS •.154 4955 491 •J77 8.177 40.72
Bv Allotment Size
$1-100 66 0.303 0.121 16.44 413 0.341 0.140 33.13
$101-200 481 0.484 0.316 37.99 22 0.364 0.045 3.48
$201-300 1.363 0.323 0.158 43.91 4 0450 0.000 382.00
$301-400 1.298 0.266 0.127 62.03 2 0.500 0300 22.00
$401-500 6S3 0.255 0.107 60.37 1 1.000 1.000 9.00
$501400 174 0.201 0.109 39.74 7 0.857 0371 274.70
S601 + 27 0.000 0.000
M*Hr* CMUQH
Total 2*317 •J34 wm 35.42 34*2 •-398 •418 23.19
By AMoment Sat
$1-100 230 0.340 1.1*8 16.65 21 0.381 0.143 39.33
$101-200 2.868 0.556 0.351 34.09 3.047 0.392 0411 22.77
$201-300 10.231 0.338 0.152 30.99 165 0.479 0.182 21.97
$301-400 8.299 0.287 0.130 37.41 11 0.455 0473 1031
$401 500 3.521 0.281 0 108 46.33 8 0.375 0.125 12.02
$501-600 1.175 0.269 0.111 46.98 7 0.857 0.571 29.93
$601 + 161 0.244 0.161 33.83 2 1.000 0.000 • 00
BattaartOry
Totti 29.495 • 141 §.•79 31.17 12.194 •.ltl 8JS2 3939
«y Allotment Sue
$1-100 311 0.125 0.055 1164 23 0.174 0.043 41.47
$101-200 2.739 0.291 0.189 34.37 11.987 0.099 0.052 39.30
$201-300 11.261 0.148 0080 26.11 186 0420 0.081 47.76
$301-400 9.374 0.112 0.063 32.64
$401-300 4.428 0.111 0.056 35.56
$301-600 1.344 0.086 0.055 42.15
$501 + 226 0.075 0.040 127.87
Stale total exceeds the sun over regions due to cases wah missaig county code
Defianoafl - endng balance greater than $1
Definaion#2 ■ disbursemeai for September awaits total redemptions in
Average end balance is the average balance of food sump benefas on the day prior to a*
for cases wa*i greater dun $1 remaining (defh #1).
i$l.
Prepared by Abt Associates Inc. 54
Figwe3.5
Speed of Cub Benefit Edmmtai By Days Since Disbursement
ixrcent ot cases According to irrcent ot Benefits wiindrswn
Day of DMHNMNM SDiyi
nm
lATOC |OM*<
5Diyt 7D«yt
10%
lAFDC CMLPCMM I At DC Cm IDAIPO
^
Chapter Three: Cash Benefit Redemptions
The distribution of benefit withdrawal behavior shown in Figure 3.5 can be contrasted
with the analogous picture of food stamp benefit redemption shown in Figure 2.8. Food stamp
redemption is characterized by a gradual movement of cases from "no benefits redeemed'' to
"100 percent of benefits redeemed." In other words, on any day during the first week of the
disbursement month, a significant number of cases have redeemed a portion of their food stamp
allotment without redeeming their full allotment, whereas cash benefits are overwhelmingly
redeemed all at once.
Even though most cash benefit recipients withdraw their entire benefit amount all at
once (as seen above), the average recipient makes over four withdrawal transactions per month.
In Table 3.8 we investigate the extent to which the "number of transactions" may be a
misleading measure of transaction behavior.
Table 3.8 examines the incidence of ATM transactions that are "nearly simultaneous"—
that is, we counted transactions made by the same recipient and at the same terminal within a
five-minute window. It is not hard to imagine how these "five-minute events" might occur.
The automated "fast cash" feature of an ATM disburses a specified number of dollars, and
recipients might repeat this transaction until the desired cash is in hand. Alternatively, if the
ATM menu does not specify the dollar amount desired, multiple transactions may be made
(ATM menus typically branch into "fast cash"—i.e., a menu of dollar amounts; or custom
cash—an opportunity to specify the dollar amount). In addition, many ATMs have maximum
withdrawal amounts that are specific to the ATM owner (i.e., bank) and cannot be centrally
overridden (the typical maximums are $200 or $300). These withdrawal limits imply that
recipients who wish to withdraw their entire allotment must make multiple transactions to do so.
As seen in Table 3.8, 37 percent of cases had at least one "five-minute event" in which
they made back-to-back transactions. Most of these recipients did this only once, though 7
percent of these cases did it on two occasions. Three quarters of all "events" involved two
transactions within five minutes, and another 19 percent of events involved three transactions
within five minutes. This means that the mean number of transactions per case overstates the
number of "transaction occasions" per case. If recipients had consolidated the "multiple
withdrawals within a five-minute window" into single transactions, the mean number of ATM
transactions per case would have been 1.96 rather than the 2.46 shown in Table 3.3.
Prepared by Abt Associates Inc.
TABLE 3 8
CHARACTERISTICS OF CASH BENEFIT WITHDRAWALS IN THE MARYLAND EBT SYSTEM
SEPTEMBER 1999
MULTIPLE CASH BENEFIT WITHDRAWALS WITHIN 5-MINUTE WINDOWS
Total COM With Fraction of cues by
number of "events"
One Two Three +
Total
Number
ofS-uunute
Events
Fraction of "events" by
1 W/D within 5 minutes
Timing of "events" by
Days since disbursement
(fraction of events each day)
Fraction of "events
With:
Number 5-minulc Events Max W/D Max W/L>
of cases Number Percent Two Three Four Five + Day 1 Day 2 Days 3 4 -$200 -$300
Total "Regular* Cases 74.457 27,313 36.7% 0.92
N jn-Metro Counties 3.909 1.713 43.8% 0.91
Metro Counties 2S.636 7.959 31.0% 0.92
Ut
Baltimore CHy 29.303 13.438 45.9% 0.91
^
Total AFDC 38.848 23.121 39.3% 0.91
Non-Metro Counties 467 122 26.1% 098
Metro Counties 3.167 713 225% 0.93
Bahhnor-City 11.934 3.356 28 1* 0.97
Total DALP 15.588 4.192 26.9% 0.96
0.07 0.01 29.770 0.73 0.19 0.06 0.02
AFDC Cases wtth "Regular" Monthly Disbarstsmut
0.08 0.01 1.891 0.72 0 20 0.05 0.03
0.07 0.01 8.698 0.78 0.15 0 05 0.03
0.08 0.01 14.817 0.67 0.23 0.08 0.03
0.08 0.01 25.419 0.71 0.20 0.07 0.03
DALP Cases wMh "Regular" Monthly Msbursement
0.02 0.00 124 0 78 0.19 0.02 0.01
0 07 0.00 765 0.76 0.17 0.04 0.02
0.03 0.00 3.461 0 84 0.15 0.01 0.00
0 04 0 00 4.351 0.83 0.13 0.02 0.01
063 013 008
064
0.57
0.67
0.63
012
015
0 12
0.13
0.09
0.10
0.08
0.09
0.58 0.13 0.09
0 50 012 0.10
07U 0.12 0.06
0 66 0 12 0 07
029
030
027
038
034
003
0.02
0.00
0.01
0.09
0.17
017
0.05
0.10
002
0.01
0.00
0.00
Noses: A '5 minute event' to defined to be any 5-minute window containing more than 1 cash withdrawal transaction made at the same ATM location
Regions do not sum to tt ' due to cases with missing county code.
/?
Chapter Three: Cash Benefit Redemptions
Out-of-Sute Transactions
The MOST debit card network is a major ATM network operating throughout the mid-
Atlantic region. As noted at the beginning of this chapter, cash assistance recipients in
Maryland could access their EBT benefits at any ATM in the MOST network, even those
located in other states or the District of Columbia.
In fact, Maryland recipients made only limited use of ATMs outside Maryland. As
shown in Table 3.9, 2.2 percent of all recipients used out-of-state ATMs in September.33 The
transactions occurring at the out-of-state ATMs represented 1.5 percent of all EBT transactions
at ATMs that month and 1.6 percent of all EBT dollars withdrawn at ATMs. The cross-state
activity is largely due to the transactions behavior of recipients in bordering counties, and three
counties account for most of the cross-state activity.
Notably, the far right columns of Table 3.9 show that recipients with any out-of-state
transactions are not completely dependent on out-of-state processing capabilities. These
recipients, on average, make only slightly more than half of their cash ATM transactions across
state lines.
33. Table 3.9 is based on all cases receiving cash benefits in September, not just those receiving regular
disbursements.
Prepared by Abt Associates Inc. 58
Table 3 9
CASH BENEFIT WITHDRAWALS AT OUT-OF-STATE ATM LOCATIONS
wsth My Out-of-Stale
AIM.
Number Percent
All Counties 100.553
Non-Bordering Counties 13,087
Bordering Counties:*
Total 87,430
Montgomery 4,877
Prince George's 11.566
Worcester 4S3
2.162 2.2% 7.228 15%
83 0.6% 233 0.4%
2.079 2.4% 6.973 1.6%
309 6.3% 1.108 54%
1.488 12.9% 4.940 10.4%
17 3.8% 58 30%
460.500
18.480
442.020
61.225
321.490
5.640
16%
05%
17%
4 1%
8 7%
4.1%
066
068
066
062
0.63
0 73
073
072
0.73
066
069
0.84
Non-bordering counties: Anne Arundel. Calvert. Charles, Howard. St. Mary's. Somerset. Tilbot. Baltimore City
Bordering counties: Allegheny (WV.PA). Baltimore (PA). Caroline (DE). Carroll (PA), Cecil (PA.DE). Dorchester (DE). Frederick (PA.VA).
Garrett (PA.WV). Harford (PA). Montgomery (DC.VA). Prince George's (DC.VA). Queen Anne's (DE). Washington (PA.WV). Wicomico (DE).
Worcester (DE).
* Bordering counties with fewer than 3% of transactions occurring out-of-state are not shown separately.
Jl
APPENDIX A
DESCRIPTION OF TRANSACTIONS LOG FILE
AND PROCESSING METHODS
The primary source of data used in this report is the transactions log file of the
Maryland EOT system. The transactions log file was provided by the EOT vendor. Deluxe Data
Systems, and includes the withdrawal and redemption activity of the entire Maryland caseload
of EOT clients (this includes the caseloads of both the FSP and cash benefit programs).32
The file that we received contains almost all of the transactions affecting the accounts
of Maryland EOT clients for the period August 1993-October 1993. We used the month of
September for this study.33 The file contains one record for every transaction that is processed
by the system. There are five types of records: benefit authorizations, cash benefit
withdrawals, food stamp benefit redemptions, balance inquiries, and food stamp refunds (store
refunds that are added back to the EOT account). The transaction records for the September
calendar month are distributed approximately as follows:34
Type of Transaction Percent of Records
Benefit Authorization 9%
Food Stamp Purchase 54%
Food Stamp Refund < 1%
Cash Benefit Withdrawal 18%
Benefit Balance Inquiries 19%
32. The Maryland EBT System served the AFDC and DALP programs la addition, NPA-CS
Baltimore City were also part of the EBT System in 1993.
in
33. The August file was incomplete and hence could not be used (there were no records on the file with a
transaction dale prior to August 4, and the records of benefit authorization were incomplete). The October
file could not be used because analysis of the October disbursetmrnt monm retires data for the first few days
of November. This is because food stamp benefits are disbursed to ongoing cases on the fifth, sixth, and
seventh days of each month, thereby making the first week of November part of the October "disbursement
month."
34. Records documenting PIN changes and direct deposits to the NPACS accounts were not used in our
analysis.
Prepared by Abt Associates Inc. A-l 0
Appendix A: Description of Transactions Log File and Processin/c Methods
Each transaction record includes the following information: type of transaction, date
and time of transaction, amount of transaction, remaining balance, and location of redemption
or withdrawal (i.e., store ID and POS terminal ID, or ATM terminal ID). In addition, all
records except benefit authorizations include codes to identify system rejections and reversals.
A small percent of all records are rejected or reversed by the system. Rejections are
transactions that are rejected by the system due to insufficient funds, incorrect PIN. or other
problems with the account. Reversals are system generated transactions that reverse a previous
incorrect or incomplete transaction (thereby making the previous transaction a "reversed
transaction"). Appendix Table A.l shows the overall incidence of EBT system rejections and
reversals by reason for system interruption. The majority of system rejections are due to
insufficient funds; the majority of reversals are due to terminal errors. (Appendix Table A.2
presents this information by program, store type, and county.)
Data Processing
We processed the transactions data to obtain a separate food stamp redemption history
and cash benefit withdrawal history for each case. We initially processed all of the August-
October withdrawal/redemption records to get a complete history for each case over the entire
time period. This is because the disbursement months (i.e., time period between disbursements)
overlap calendar months, and because we wanted to ensure the accuracy of the starting balance
at the time of the September disbursement. Construction of the transaction histories required
that we "clean" the data in the following ways:
(1) Refunds: All food stamp "refunds" were removed from the file and an
adjustment was made to the relevant "purchase" to reflect the fact that it was
subsequently refunded (i.e., given that case A received a refund at retailer X.
we determined the most recent purchase made by case A at retailer X prior to
the refund, for an amount at least as great as the refund. We then subtracted the
refund amount from the purchase amount.) In cases in which the refund was for
the full amount of the purchase, both the refund record and the purchase record
were removed; hence, our count of redemptions will slightly undercount the
number of redemption transactions processed by the system.
(2) Reversals: All system reversals and "reversed transactions" were removed from
the file. The reversed transaction nearly always occurred immediately prior to
the reversal, and was identified as occurring at the same terminal for the
identical dollar amount.
Prepared by Abt Associates Inc. A-2
£1
Appendix A: Description of Transactions Log File ami Processing Methods
(3) Rejections: All system rejections were removed from the file.
We checked that this "cleaning" was performed properly by reconciling the transaction
history of each case with the records of benefit deposits and the remaining balance on the final
transaction.37 We were unable to reconcile the transaction history for approximately two
tenths of one percent of all cases (370 food stamp histories and 183 cash benefit histories).
These unreconciled cases were mostly due to reversals for which we could not find a matching
transaction, or reversals that were posted to the transactions log before the transaction that was
to be reversed. We subsequently dropped these cases from the sample.
A final step involved extracting that part of die case history that coincided with the
September disbursement month; this was achieved by merging the benefit disbursement
information to the withdrawal history. Operationally, we maintained a file of the caseload
transaction history for die month of September (with the benefit disbursement date) and a file
of the caseload activity for the days in October preceding the October disbursement, thus
allowing us to conduct both calendar month and disbursement month analyses. As discussed
in the text, we limit the analyses of redemption patterns to cases with a single "regular"
monthly disbursement in both September and October.
Supplementary Information
We supplemented the transactions log file with information about the store type and
county location of each FSP-authorized retailer in the state of Maryland. We obtained this
information from the FCS Minneapolis Computer Service Center. This information was merged
to the transactions log file via a crosswalk provided by Deluxe that mapped the Deluxe store ID
37. Thij process is complicated somewhat by the fret thai the remaining balance on each record reflects only
the "active pot" of money. The Deluxe accounting process is such thai each disbursement is treated as a
separate "pot." Even though a client has access to ail benefits thai have been disbursed to his account, rite
remaining balance oa the Deluxe record reflects only the "active pot"—i.e., the disbursement that is currently
being drawn rtowa and the next pot does not become active until the balance on the current pot thus bdow
$10. Hence, the "active pot" ■aitiiilnUi the true account balance whenever a balance of at least $10 is
carried over to the following month.
Prepared by Abt Associates Inc. A-3 M-
Appendix A Description of Transactions Log File and Processing Methods
to the FNS n>.M For a small number of retailers that did not appear in our FNS file, we
assigned county location based on the street address from the Deluxe cross-reference file.39
For these same stores we assigned store type based on the name of the store (if the store had
multiple locations and at least one appeared on the FNS file), or by consulting the Maryland
Department of Human Resources.
In addition, we identified the county location of ATM terminals based on the location
information that appears on the transaction records; the transaction records contain the street
address and city. Approximately 3,095 ATM terminals were accessed by the Maryland EBT
caseload during the entire period from mid-August 1993 through October 1993. The 3,095
terminals correspond to 2,800 different ATM locations (each ATM location may have multiple
terminals). Of the 3,095 ATM terminals, 1,899 are located in Maryland at approximately 1,740
different locations.40 Most of the Maryland ATM locations could easily be mapped to a
county based on the city designated on the transaction record. A significant number of ATMs
are located in cities that overlap county boundaries, however. In these cases, we searched the
Census TIGER database to determine county location based on street address. Approximately
40 ATMs could not be mapped to a county because the street address was either incomplete or
did not appear in the Census database.
38. Some anomalies were discovered and rectified. For example, all matches by ID were checked lo assure
that the store name and address in the two files matched and inconsistencies were resolved. In addition,
several FNS IDs matched to a pair of Deluxe store IDs; this appeared to be due to a renumbering of the
Deluxe IDs that occurred at a point in time.
39. Our FNS file was received in August 1993, but may not have included uott* that recently received
authorization. The Deluxe file contains the street address only, not the county. We used the street address
to search the Census TIGER database to determine the county. We consulted with Maryland DHR to
determine retailer locations that could not be found in the Census dntabnse.
40. The remaining ATMs are located in the District of Columbia and the following bordering and nearby
states: Delaware, Pennsylvania. Tennessee, Virginia, and West Virginia
Prepared by Abt Associates Inc. A-4
&
Appendix Table A. 1
EBT 90tm Bajirttd Ik—Hi—d System sWt—■
September 1993 Calendar Month
with any rejections
rfimrTrrr Percent* Number Percent'
Sumber mmi Rote of Rejections
Number of rejections and 113.435 7.25 70,377 14 50
percent of all transactions
Number and percent of 68.835 44.23 33.628 33.44
Insufficient Funds 62.548 55.14 43.380 61.64
Invalid PIN 37.771 33.30 12.989 18.46
PIN tries exceeded 220 0.19 1.045 1.48
PIN not selected 266 0.23 150 0.21
Expired Card 5.943 5.24 4.887 6.94
Lost/stolen Card 1.390 1.23 1.030 1.46
Cardholder not OB file 1.585 1.40 261 0.37
No benefit for request 3.590 3.16 6.594 9.3?
Opuoe of Decline 10 0.01 8 0.01
Bad FNS status for merchant 101 0.09 0.00
System malfunction 10 0.01 22 0.03
Bad card status 1 0.00 11 0.02
Number of reversals and 25.209 1.61 5.423 1.12
of all
of 18.987 12.20 4.389 4.36
error 19 0.08 1.515 27.94
970 3.85 945 17.43
0.00 6 0.11
24.097 95.59 2.922 53.88
1 0.00 1 0.02
IS 0.06 0.00
107 0.42 34 0.63
Notes * The rates of rejection and reversal are expressed as a percent of all valid transactions in the calendar month"
The percent of cases with any rejections/reversals is based on the number of cases with valid transactions in the
Prepared by AN Associates Inc. A-5 /U
•3-
sr i
Appendix Table A.2a
FOOD STAMP REJECTIONS AND REVERSALS, 1Y STORE TYPE
NuMbe
Toul
r of Stores Number c/
With Any Pmth—
Trans Tra—arti—■
Mnibir of Other TraBiirtlw,
Refunds Revemb Rejection*
Rale per Ituuaand Fraction of Rejection* for:
Authorized Pnrcbneei Invalid liwuff Expired
Store Type Refunds Reversals Rejections PIN hinds Card
Toul 3.272 2.998 1.655.915 2.299 25.251 113.076 1.4 15.2 683 0.33 055 005
Alcoholic Treatment (AT) 3 3 81 0 3 33 0 37.0 407.4 0.27 061 0.12
Non-profit Coop (BQ IS II 802 5 22 III 6.2 27.4 138.4 005 0.47 0.30
Bread Route (BR) 6 4 914 0 7 55 0.0 7.7 602 0.25 0.64 002
Comb. Groc/Bar (CB) 10 9 2.053 0 16 187 0.0 7.8 91.1 0.36 055 0.05
Comb. Groc/Gas (CG) 42 38 12.904 2 59 487 0.2 4.6 37.7 0.32 050 0.06
Comb Groc/Merchandise(CM) 40 38 6.018 21 89 649 3.5 14.8 107.8 0.35 054 005
Other Combination (CO) 139 135 52.047 54 723 4.783 1.0 139 91.9 0.37 0.52 0.06
Comb. Groc/Rcsuurant (CR) 59 52 22.363 144 930 3.692 6.4 41.6 165.1 0.38 0.50 005
!► Convenience Store (CS) 971 890 184.915 428 3.615 19.914 2.3 19.5 107.7 0.33 054 006
cVt Drug Addict Treatment (DT) 1 1 9 . . 00 00 00 .
Farmers Market (FM) 12 9 3.614 0 29 188 00 8.0 52.0 0.35 054 0.04
Group Living Arrangmrn (GL) 1 1 10 0 0 3 00 00 300.0 000 1 00 0.00
Small/Medium Grocery (GS) 802 769 465.615 327 3.549 26.874 0.7 7.6 57.7 0.37 0 52 0.05
Healih/Narvral Food (HF) 29 24 852 1 43 414 1.2 50.5 485.9 0.39 0.50 0.06
Homeless Meal Provider (HP) 1 1 5 . • . 0.0 00 0.0
Military Commissary (MQ 8 8 373 1 19 117 2.7 50.9 313.7 0.34 055 003
Other Firm (OF) 60 55 20.134 14 206 1.347 0.7 102 669 038 050 0.04
Other Route (OR) 30 24 2.240 1 19 389 0.4 8.5 173.7 0.06 064 0.10
Produce Route (PR) 7 7 110 0 5 193 00 455 1754.5 0.17 060 Oil
Produce Stand (PS) 53 41 18.527 47 388 1.991 2.5 209 107.5 0 38 051 004
Specialty Food (SF) 326 309 135.020 285 2.520 11.539 2.1 18.7 855 0 35 0 53 0.06
Supermarket (SM) 563 545 724.854 949 12.884 39.385 1.3 17.8 543 030 060 0 05
Wholesaler (WH) 11 10 700 14 75 296 200 107.1 422 9 029 058 0.07
Unknown UN 83 14 1.755 6 50 429 34 285 2444 0 22 052 004
W
I FOOD STAMP REJECTIONS AND REVERSALS, BY COUNTY
*X
Number of Stores
With Any
Total Trans
Number of
Purchase
Traosactsoaa
Number of Transactions,
By Type:
Refunds Reversals Rejections
Rate per thousand
AuthorUed Purchases
Fraction of
Invalid
PIN
Rejections
Insuff 1
Funds
Due to:
Expired
f County Refunds Reversals Rejections Card
ft
Tottl
§ Com*,:
Z Alkgany
5* Arme Anindel
3.272
93
176
2.998
84
170
1.655.915
37.381
60.830
2.299
37
108
25.251
268
1.351
113.076
1.499
5.090
1.4
1.0
1.8
15.2
7.2
22.2
68.3
40.1
83.7
0.15
0.33
0.28
0.28
0.38
0.59
0.03
0.04
0.06
:' Baltimore
Calvert
307
36
287
34
121.011
6.589
327
2
4.085
107
13.458
575
2.7
0.3
33.8
16.2
III 2
87.3
0 32
0.27
056
0.59
063
0.64
0.03
0.05
Caroline
Carroll
34
63
31
57
5.836
8.800
32
19
277
314
1.159
1.093
5.5
2.2
47.5
35.7
198.6
124.2
0.26
0.23
0.06
0.05
Cecil
Charles
50
56
44
52
13.533
18.850
23
18
524
386
2.006
1.292
1.7
1.0
38.7
20.5
148.2
68.3
0.30
0.30
0.58
0.59
0.05
0.04
Dorchester
^ Frederick
30
56
27
53
10.292
16.910
3
21
48
201
485
1.191
0.3
1.2
4.7
11.9
47.1
70.4
0.33
0.30
050
058
007
004
il Garrett
Harfon.
34
87
27
84
8.471
23.421
9
48
45
489
294
1.703
1.1
2.0
5.3
20.9
34.7
72.7
035
0.25
0.55
0.63
0.03
0.04
Howard
Kent
49
14
45
13
10.493
3.502
4
1
216
7
793
101
0.4
0.3
20.6
2.0
75.6
28.8
0.30
0.31
061
0.52
0.05
0 06
Montgomery
Prince George's
215
331
188
299
59.612
126.631
99
139
1.932
2.459
7.351
10.882
1.7
1.1
32.4
19.4
123.3
85.9
0 32
0.30
0 58
0.57
0 05
0.06
Queen Anne's
St Marys
18
54
17
50
2.935
15.829
1
7
23
289
295
986
0.3
0.4
7.8
18.3
1005
62.3
028
029
0.61
0.59
0.05
0.06
Somerset
Talbot
32
16
31
16
8.462
5.122
7
4
253
40
792
324
08
0.8
29.9
7.8
93 6
63.3
035
0 38
054
052
0.06
004
Washington
Wkomico
96
72
92
67
29.143
24.204
51
26
248
218
1.565
1.859
1.7
1.1
8.5
9.0
53.7
768
035
0.32
0.55
0.57
0.05
0.06
Worcester
Baltimore City
36
1,306
34
1.179
8.605
1.028.779
2
1.303
116
11.317
734
57.322
0.2
1.3
13.5
11.0
853
55.7
034
0.36
052
0.53
0.06
0.05
Source: Transactions Log file from Deluxe Data Systems, September 1993 calendar month. Fifteen counties with missing county code are not reflected in the table
u
Appendix Table A.2c
CASH-BENEFIT REJECTIONS * REVERSALS, BY STORE TYPE
8. Number of Stores Rate per thousand Fraction of Rejections Due to:
3"
Store Type
Total
Stores
With Any
Cash Trans
Number of
Withdrawals
Number of: Cash benefit transactions
Reversals Rejections
Invalid
PIN
Insuff 1
Funds
Expire
Card 1
No
f
E
Reversals Rejections Jenefit
Total 5.M5 3.017 3.0M 23.538
| Alcoholic Treaimenl (AT) 3 J 70 0 24 0 343 0.21 0.50 0.00 0.29 I Non profit Coop (BC) IS II 25 0 18 0 720 0.00 0.44 0.22 0.33 s Bread Route (BR) 6 4 18 1 3 56 167 0.67 0.33 0.00 0.00
3" Comb. Groc/Bar (CB) 10 9 618 7 45 11 73 0.18 064 0.02 009
Comb. Groc/Gas (CG) 42 M 1737 9 138 5 79 0.17 0 62 0.00 0.18
Comb. Groc/Merchandije(CM) 40 38 966 15 128 16 133 0.13 0.63 0.05 0.14
Other Combination (CO) 139 135 11427 107 1476 9 129 0.22 0.59 0.04 0.12
Comb. Groc/Restaurant (CR) 59 52 3420 131 1105 38 323 0.27 0.57 005 0.09
Convenience Store (CS) 971 891 20899 313 3772 IS 180 0.19 0.59 0.05 0.14
Drug Addict Treatment (DT) 1 1 9 3 3 333 333 0.67 0.33 0.00 0.00
Fanners Market (FM) 12 9 74 3 58 41 784 0.21 0.57 0.07 0.12
>i
Group Living Arrangmen (GL) 1 1 6 0 1 0 167 0.00 1.00 0.00 000
00 Small/Medium Grocery (GS) 802 770 69104 572 7642 8 III 0.21 0.60 0.05 0.13
Health/Natural Food (HF) 29 24 10 0 27 0 2700 0.22 0.48 0.04 0.22
Military Commissary (MC) 8 1 2 1 IS 500 7500 0.00 0.73 0.00 0.27
Other Firm (OF) 60 55 1632 26 305 16 187 0.21 0.62 0.03 0.11
Other Route (OR) 30 24 79 1 12 13 152 0.25 0.50 0.25 0.00
Produce Stand (PS) 53 41 477 19 314 40 658 0.19 0.59 0.03 0.16
Specialty Food (SF) 326 309 4968 247 1767 50 356 0.22 0.58 0.05 0.12
Supermarket (SM) 563 545 105982 1581 6148 15 58 025 0.57 0.06 0.09
Wholesaler (WH) 11 10 7 0 2 0 286 0.00 0.00 0.50 0.00
Non-Food Merchant 45 31 4805 26 447 5 93 0.29 0.50 0.09 0.09
Unknown UN 76 8 383 3 88 8 230 0.15 0.63 0.07 0.14
Source: Transactions Los File fro n Deluxe Data System!t, September 1993 calendar month 1 -ifteen stores with missing county code are not reflected in the table
Note: The 3.302 stores include 3,272 FSP-authorized retailers plus 45 nonfood merchants with POS terminals for cash benefit withdrawals.
W
Ii
x>
>1
NO
Appendix Table A.2d
CASH-BENEFIT REJECTIONS A REVERSALS, BY COUNTY
Number of Stores
Total With Any
Cash Trans
Number of
Withdrawals
Number of:
Rale per thousand
Cash benefit transactions
Reversals Rejections
Fraction of Rejections 1)
Invalid Insuff Expire
PIN Funds Card
ue to:
No
County Reversals Rejections Benent
Tofal
All eg any
Anne Anindel
3.302
95
177
3.016
85
171
JM.6ft9
2446
6604
5.0fJ2
21
124
53.511
185
600
14
9
19
104
76
91
22.07
1676
21.50
5ft 68
52.97
5800
3 02
595
4.17
24 32
13.17
Baliimore
Calvert
311
36
288
34
14735
660
453
12
2415
70
31
18
164
106
2501
17.14
59 09
4857
484
2.86
865
30.00
Caroline
Carroll
34
67
31
57
706
716
24
26
90
159
34
36
127
222
20 00
27.04
55 56
52.83
7.78
2.52
1222
10.69
Cecil
Charles
50
58
44
52
1549
1728
63
23
472
215
41
13
305
124
20 34
1721
58 26
58 14
5 08
2.79
II 86
1953
Dorchester
Frederick
30
56
27
53
1169
1579
4
20
102
157
3
13
87
99
20 59
20 38
59.80
56.05
2.94
5.10
14.71
12.74
Garret!
Harford
35
88
27
85
464
2645
5
35
39
240
11
13
22
0
84
91
64
51
25 64
23.75
13 16
22 22
4359
47.92
5921
33.33
5.13
5.42
263
0.00
2051
21.25
Howard
Kent
49
14
45
13
1182
355
26
0
76
18
2368
38.89
Montgomery
Prince George's
217
332
188
299
5531
15169
199
318
983
1795
36
21
178
118
21.67
21.28
60.22
58.27
631
5.52
926
12.26
Queen Anne's
St Mary's
19
54
17
50
314
1606
2
34
25
83
6
21
80
52
800
18.07
7600
5663
4.00
2.41
1200
22.89
Somerset
TaJbot
33
16
32
16
1262
411
16
16
120
78
13
39
95
190
18.33
29.49
60 00
51.28
4.17
6.41
17.50
12.82
Washington
Wicomico
97
73
93
67
2415
2598
17
27
150
2%
7
10
62
114
20.00
19.93
60.67
54.73
2.00
3.04
14.00
20.61
9
1588
90
15053
calendar month.
9
10
89
94
1889
22.07
51 II
59.18
8.89
5.06
20 00
11.35
Worcester 36 34 1014
Baltimore City 1325 1208 159831
Source: Transactions Log File from Deluxe Data Systems, September I Fifteen stores with missing county code aic not rellcck-d in the table
Note: The 3.302 stores include 3.272 FSP-authorized retailers plus 45 non-food merchants with POS terminals for cash bcncFit withdrawals.
6f
APPENDIX B
COMPARISON OF SURVEY RESPONSES
WITH TRANSACTIONS BEHAVIOR
The evaluation of the Expanded EBT Demonstration in Maryland included a survey of
a sample of EBT recipient households. The sample included approximately 1.200 food stamp
households that were interviewed after implementation of EBT. (The survey methods and
content of the instrument are documented in Kirlin tt al., 1994). The survey dataset from the
Maryland evaluation provides a rare opportunity to match records of reported behavior to a
database containing records of actual behavior.
This appendix presents the results of the comparison of reported an- actual food stamp
redemption behavior for the respondents of the Maryland Recipient Survey. Unfortunately,
although the survey interviews were conducted over a period of several months, our actual
transactions records are for September 1993. Reported transactions behavior, therefore, cannot
be expected to correlate perfectly with the actual transactions record even in the absence of
reporting error. For this reason, we focus the analysis on a selection of survey items that ask
about general behavior or behavior in a "typical month," rather than questions that specifically
ask about "last month." For example, survey respondents were asked "Do you ever save any
of your food stamp benefits in a particular month so that you can use them in a later month?"
and "In which type of store do you spend most of your food stamp benefits in a typical month?"
Tables B.l—B.3 show the results of this analysis. Each table presents the survey
questions on the right side of the table with the possible survey responses forming the column
headings. The frequency of survey responses is shown in brackets in each column. Respondents
are "grouped" according to their response to the survey item and the distribution of actual
transaction behavior is examined for each "group." In other words, we examine the actual
transactions behavior of survey respondents, according to their reported behavior.
Actual and reported measures of benefit exhaustion are compared in Table B. 1. Survey
respondents were asked "One week after you receive your food stamp benefits, do you usually
have ... (none/less than half/about half/more than half) remaining?" We examined the average
fraction of benefits remaining seven days following disbursement for each of the groups of
Prepared by Abt Associates Inc. B-l (H
Appendix Table B.l
Correspondence Between Survey Response and Actual Transactions Behavior
Benefit Exhaustion
Measure of Actual Transactions Benavior Survey Question
Fraction of Benefits Remaining on Day 7
Mean
25th percentile
Median
75th percentile
Q: 'One wee* after you receive your Food Stomp Benefits
do you usually hove?'
None Leu than 1/2 About 1/2 More than 1/2
[248] [332] [209] [115]
0.28 0.26 0.31 0.39
0.00 0.02 0.06 0.10
0.12 0.16 0.21 0.40
0.48 0.46 0.53 0.61
Fraction of cases starting the month with:
Any benefits carried over
More than SI carried over
More than S3 carried over
Q: 'Do you ever save any ofyour Food Stamp benefits Li a
particular month so that you cm use them in a later month?'
Yet No Don't Know
[190] [736]
0.67 0.57
0.32 0.20
0.16 0.08
m
Fraction of cases redeeming less than
current month's disbursement:
Any benefits left over
More than Si left over
More than S5 left over
0.41 0.40
0.19 0.14
0.11 0.06
Notes: Actual transactions information comes from September 1993. This month does not necessarily
correspond to the interview month. Number of households reporting survey response in brackets.
All percents are column percents.
Prepared by Abt Associates Inc. B-2 0
Appendix B: Comparison of Survey Responses with Transactions Behavior
respondents (grouped according to their response to the survey). We also present the quartiles
of the distribution for each group. As seen in the table, there is very little difference in the
actual fraction of benefits remaining after seven days for groups that reported very different
behavior. The group that reported "none" remaining and the group that reported "more than
hair actually differ by only 10 percentage points at the mean.
The bottom panel of Table B.l compares "saving" behavior. Respondents were asked
if they ever save benefits from one month to the next (i.e., do ever they fail to exhaust their
benefits?). The transactions log provides us with two points of observation for this "saving"
behavior. First, we look at whether respondents started the September 1993 disbursement
month with a balance remaining from a previous month; this provides a long-run view because
this excess may have been carried over in any month prior to September. Second, we look at
whether respondents ended the September disbursement month with any benefits remaining; this
measure detects the saving behavior in the September disbursement month only. The table
shows that reported "savers" were in fact more likely to display saving behavior, in both the
September month and prior to the September month. As expected, the differential between
savers and non-savers is greater according to the long run measure. According to our most
restrictive measure of "saving behavior" (defi