The extent of trafficking in the food stamp program an update

              '***'* 00lH-f\-o\ f\c[(6.2-:F~l>/2t/Upv*Te
THE EXTENT OF TRAFFICKING IN
THE FOOD STAMP PROGRAM:
AN UPDATE
Theodore F. Macaluso, Ph.D.
Office of Analysis, Nutrition, and Evaluation
Marcn 2000
Food and Nutrition Service
U.S. Department of Agriculture
Cs
The Extent of Trafficking in the Food Stamp Program
An Update
EXECUTIVE SUMMARY
Food stamps are intended for food. When individuals sell their benefits for cash it violates the
spirit and intent of the Food Stamp Program as well as the law. This practice, known as trafficking,
diverts food stamps away from their purpose. It reduces intended nutritional benefits and
undermines public perceptions of the integrity and utility of the program. A crucial question,
therefore, is the extent to which trafficking exists.
Several years ago, a method to calculate data-based estimates of the prevalence of trafficking was
developed by USDA. The Extent ofTrafficking in the Food Stamp Program used this method to
analyze over 11,000 completed undercover investigations of trafficking and generate an estimate
for calendar year 1993. This report duplicates the precise methodology of the earlier analysis with
more than 10,000 new investigations to generate an estimate for the 1996 - 1998 calendar year
period. We find that:
The amount of trafficking has decreased. Stores trafficked about $660 million per year for cash
from the government in the 1996 -1998 period, a 19 percent decline from the $815 million
trafficked in 1993.
The rate of trafficking has also decreased. The trafficking rate - which compares dollars
trafficked to benefits issued - declined 8 percent: from almost four cents of every dollar of food
stamp benefits issued to three-and-one-half cents of every dollar issued.
FNS concentrates its enforcement efforts on stores most likely to traffic. In addition, the
expansion of Electronic Benefit Transfer (EBT) - which had grown to half of all issuance during
this period - makes certain forms of trafficking harder to conduct and large-scale trafficking easier
to detect. For these reasons, we find the largest reduction in the trafficking rate among the
store categories most likely to traffic - privately-owned stores, especially small ones that do
not stock a full line of food.
When we repeat our analysis of where store violations occur the overall pattern remains
unchanged:
• Dramatic differences exist among store types: the percent of redemptions that are trafficked
ranged from nearly zero to over fifteen percent across store categories.
• The stores which redeem the overwhelming majority of food stamp benefits continue to
have very low trafficking rates.
* Theodore F. Macaluso, The Extent of Trafficking in the Food Stamp Program (Alexandria. VA: Food and
Nutrition Service. USDA; 1995).
Acknowledgments
The author wishes to express his appreciation to the many individuals who contributed to this
report. Richard Mantovani, Ph.D, Hoke Wilson and Tigran Markaryan at Macro International
successfully compiled and merged the data summarized here, faithfully reproduced the original
methodology, made thoughtful suggestions, and responded promptly to the author's numerous
requests for additional information and analyses.
Steven Carlson, Director of the Family Programs Staff in the Office of Analysis, Nutrition and
Evaluation (OANE), Food and Nutrition Service, provided guidance and commented
thoughtfully on drafts of the text. Ken Offerman, also of OANE, managed the contractual
support for the project, performed considerable legwork in tracking down data, and also
commented thoughtfully on drafts. Finally, the staff of the Benefit Redemption Division of the
Food Stamp Program provided many comments and corrections and helped to make this a
comprehensive - and better - report.
The Extent of Trafficking in the Food Stamp Program
An Update
United States Department of Agriculture
Food and Nutrition Service
Office of Analysis, Nutrition and Evaluation
March 2000
INTRODUCTION
Food stamps are intended for food. When individuals sell their benefits for cash it violates the
spirit and intent of the Food Stamp Program as well as the law. This practice, known as trafficking,
diverts food stamps away from their purpose. It reduces intended nutritional benefits and
undermines public perceptions of the integrity and utility of the program. A crucial question,
therefore, is the extent to which trafficking exists.
Several years ago, a method to calculate data-based estimates of the prevalence of trafficking was
developed by USDA. The Extent of Trafficking in the Food Stamp Program' used this method to
analyze over 11,000 completed undercover investigations of trafficking and generate an estimate
for calendar year 1993." The report found that:
• About $815 million was trafficked for cash from the government by food stores during 1993.
This amounted to just under four cents of every dollar of food stamp benefits issued.
• Significant differences across types of food retailers existed: supermarkets had very low
trafficking rates, non-supermarkets had substantially higher trafficking rates.
• The food stores which redeemed the overwhelming majority of food stamp benefits had very
low trafficking rates.
This report updates the earlier analysis with more than 10,000 new investigations to generate an
estimate for the 1996 - 1998 calendar year period. We continue to estimate three basic measures of
trafficking:
Page l
1. the amount oftrafficking (i.e., the total sum of dollars trafficked, which depends partly upon
the total sum of benefits issued and partly upon the next measure, the rate of trafficking);
2. the rate oftrafficking (the proportion of total benefits issued which were trafficked), and
3 the store violation rate (the proportion of all authorized stores that engage in trafficking).
While all three measures are important for different purposes, the second measure - the rate of
trafficking - is the one that provides an approximation of FNS' relative success in controlling
trafficking. The trafficking rate is independent of the size of the program (i.e., the total sum of
benefits issued) or the relative market share of different types of retailers (which is not reflected in
the store violation rate).
We undertook an update because there have been several significant developments which may
affect each of these measures of trafficking. These developments include the following:
• a 24 percent decline in food stamp caseload: from 10.8 million households per month in 1993
to 8.2 million in 1998. The caseload decline resulted in an 11.3 percent decline in total benefits
issued. This is likely to reduce the total dollar amount of trafficking (since total benefits issued
decreased), but is unlikely - by itself- to change the trafficking rate (i.e., the proportion of
benefits issued that are trafficked)."'
• a 16 percent decline in the number of food retailers authorized to acceptfood stamps: from
about 210,000 in 1993 to 177,000 in 1998. The decline in participating retailers may change
the store violation rate depending upon whether stores willing to traffic left the program at a
faster (or slower) rate than non-trafficking stores. However the influence of this factor on
changes in the rate of trafficking will depend upon two things: (i) whether trafficking-prone
stores that remain on the program changed their trafficking activity; and (ii) whether food stamp
participants choose to shop at trafficking-prone stores or not.
• a 50 percent change-overfrom paper food coupons to electronic benefit transfer (EBT). The
Personal Responsibility and Work Opportunities Reconciliation Act of 1996 mandates that all
states convert from paper food stamp coupons to electronic benefit issuance by 2002. By
September 1998 slightly more than half of all food stamp benefits were issued and redeemed
electronically. Under FBT certain forms of trafficking are harder to conduct and lar<»e-scale
trafficking is easier to detect. Therefore, we would expect its expansion to reduce the rate of
trafficking (i.e., the proportion of benefits issued that are trafficked).1'
The combined effect of these developments is hard to predict. Fortunately, one additional factor
that could affect results - the quality of FNS undercover investigations - appears to have remained
stable: there has been no meaningful change in the quantity or quality of FNS investigations.
The total number of investigations, the number in which any food stamp violation is disclosed
("positives") and the raw number in which trafficking is found have each remained relatively
constant from 1993 through 1998 (Chart 1).
Page 2
Chart 1
FNS Undercover Investigations: 1991 -1998
1
E
.Total
. Rwitive
. Trafficking
Year
APPROACH
This update uses the same methodology as the earlier report to ensure consistent comparisons. The
method focuses on authorized food retailers because all trafficking must eventually flow through a
food retailer authorized to participate in the Food Stamp Program. The reason is obvious, but
worth pointing out explicitly: authorizedfood retailers are the only ones who can redeem food
benefits for cashfrom the government.''
Because authorizedfood retailers are the only ones who can redeemfood benefits for cashfrom the
government, knowing the prevalence oftrafficking among retailers tells us the maximum amount of
dollars divertedfrom food benefits by trafficking for cash."'
The Food and Nutrition Service (FNS) maintains a staff of investigators who work undercover to
determine whether authorized food stores sell ineligible items or engage in trafficking. Stores
caught violating are fined or removed from the program and in some instances prosecuted.
Page 3
For the update, we followed the same approach used in the earlier report:"'
• First, we sorted a database of 10,354 completed investigations across five specific dimensions
that categorize store types and store locations.Vl"
• Second, for each specific category of store and location we compiled national data from
calendar years 1996 through 1998 on the total number of stores and the total food stamp
redemptions in that category.
• Third, we analyzed the investigation outcomes and calculated the weighted trafficking and store
violation rates within each category. '" We weighted the investigation data to accurately
represent the national figures." We calculated two of our three measures: the trafficking rate, a
redemption-based rate to reflect dollar diversions, and the store violation rate, a store-based rate
to identify the kinds of stores that contain the most violators.
• Finally, we multiplied the redemption-based trafficking rate against the total food stamp
redemptions in each category and summed across all categories to obtain the first of our three
measures: the amount of trafficking, which provides an estimate of dollars diverted from food
benefits by trafficking in the Food Stamp Program."'
FINDINGS
About $660 million per year was diverted from food benefits by trafficking between 1996 and
1998. This amounts to three-and-onc-half cents of every benefit dollar issued (Table 1).
Our methodology yields a cautious estimate that is likely to best represent the maximum dollars
diverted from food benefits per year by direct trafficking in 1996-1998.
Page 4
Table 1 - Trafficking Continues to be Low Among Supermarkets and Large Grocery
Stores But Substantially Higher Among Small Stores a».d Stores That Do Not Stock a Full
Line of Food.
Type of Store
1993 1996-1998
Store
Violation
Rate
Trafficking
Rate
Estimated
Trafficking
Amount
(S000)
Store
Violation
Rate
Trafficking
Rate
Estimated
Trafficking
Amount
($000) |
Supermarkets 4.2 1.7 $282,058 5.3 1.9 $279,163
Large Groceries 6.7 3.7 46.632 9.8 3.2 35.255
Subtotal 5.0 1.9 $328,690 6.7 2.0 $314,418
Small Groceries 12.8 15.7 177,809 14.4 15.8 154.109
Convenience 8.1 9.6 78.090 11.7 10.8 66.809
Specialty 17.6 14.2 117.004 10.7 8.1 55.782
Gas/Grocery 8.7 10.4 27,528 12.8 9.7 21.784
Other Types 10.2 12 4 82.605 16.2 9.4 43.892
Subtotal 10.7 13.0 $483,036 13.0 11.5 $342376
All Stores 9.4 3.8 $811,726 11.7 3.5 $656,794
Notes: The 1996-1998 data have been annualized - see endnote 7.
Trafficking violation rates are calculated separately for stores and redemptions. The store
violation rate is the percent of investigated stores caught trafficking weighted by the
national distribution ofstores. The trafficking rate is the percent of trafficked redemptions
in investigated stores, weighted by the national distribution of redemptions. The apparent
anomaly between the two rates - i.e.. the store-based rate was higher in 6 of 7 store types
while the redemption-based rate is lower both overall and in 4 of 7 store types - reflects the
fact that the two rates measure different aspects of trafficking.
Page 5
TRAFFICKING AND CHANGE IN BENEFITS ISSUED
Compared to 1993, the 1998 figure represents a 19 percent decline in the dollar amount of benefits
trafficked. As expected, we find a similarity among the changes in caseload, total redemptions, and
the amount of trafficking (Chart 2):
However, the decline in caseload and total redemptions is far from a complete explanation of
changes over this period of time: we also find an 8 percent decline in the rate of trafficking,
which is independent of benefits issued. The trafficking rate decreased from 3.8 percent of
benefits issued in 1993 to 3.5 percent of benefits issued in 1998 (Table 1).
Chart 2
Food Stamp Caseload and Dollar Amount of
Trafficking: 1993-1998
■ 1993
D1998
Caseload
(xl.OOOk)
Redemptions
(in bibns)
Trafficking
(x100k)
Page 6
TRAFFICKING AND CHANGE IN
THE AUTHORIZED RETAILER POPULATION
The 16 percent decline in number of authorized retailers also does not appear to explain the
improvement in the trafficking rate: we actually find an increase in the store violation rate
between 1993 and 1998 (Table 1 and Chart 3).
Chart 3
Authorized Food Stamp Retailers:
1993 and 1998
# of Stores
(x 10,000)
% of Stores
Trafficking
■ 1993
D1998
TRAFFICKING AND TYPE OF FOOD RETAILER
Part of the explanation for the improvement in the trafficking rate is to be found in two
critical facts:
(1) trafficking continues to vary by type of store;
(2) stores that redeem the most, traffic the least.
Tables 1 and 2 show that:
• Supermarkets and large grocery stores redeemed 84 percent of all benefit dollars but few of
those dollars are trafficked.
• In comparison to supermarkets and large grocery stores, trafficking rates among small stores
and stores that do not stock a full line of food are 4 to 8 times higher.
Page 7
Table 2 - Distribution and Market Shares of Authorized Food Stamp Ret .Hers.
Type of Store
1993
Percent of All
1996-1998
Percent of All
Stores Redemptions Stores Redemptions
Supermarkets 15.3 76.5 14.9 78.3
Large Groceries 6.9 6.0 7.0 5.8
Subtotal 22.2 82.5 21.9 84.1
Small Groceries 18.8 5.4 20.0 5.2
Convenience 27.7 3.8 26.8 3.3
Specialty 8.7 3.9 9.0 3.7
Gas/Grocery 10.3 1.2 11.9 1.2
Other Types 12.3 3.2 10.4 2.5
Subtotal 77.8 17.5 78.1 15.9
All Stores 100.0' 100.0b 100.0' 100.0-
Notes:
1 Based on a total of 200.568 authorized food retailers redeeming at any point during 1993.
b Based on a total of $21.1 billion.
1 Based on 237,824 unique food retailers redeeming at any point during the 1996-1998
period.,"
d Based on total of $56.16 billion over the three years.""
Page 8
Between 1993 and 1998 there was a modest increase in the relative market share of
supermarkets and large grocery stores - the stores least likely to traffic (Chart 4).
Chart 4
Change in Retailer Population and Market
Share: 1993-1998
2
15
1
05
0 —
-05
-1
-1.5
-2
Store
Misfit
Sh ire
■ Large Stores
DSma« Stores
Notes: Unlike earlier charts, in which each column was a
different year (l 993 or 1998), in this chart each column is
the difference between the two periods. The "large store"
category includes both supermarkets and large grocery
stores; "small stores" are everything else. Market share is
defined as the percentage of redemptions accounted for by
the given category of store.
Food retailers owned by public corporations (he., owned by a company whose stock trades
publicly) continue to have lower trafficking rates than privately-owned stores (Table 3). The
public corporation category includes many of the major national supermarket chains, many
convenience store chains, and many grocery marts associated with national gasoline retaileTS.'"v
• In 375 investigations of public corporations, FNS undercover investigators found trafficking
involved about four percent of publicly-owned stores.
• Among privately-o-.vned food retailers, FNS undercover investigators found trafficking in
almost thirteen percent of stores.
Page 9
Table 3 - Publicly-Owned Food Retailer* Display Low Trafficking Rates; Privatery-
(>wned Retailers, Especially Non-Supermarkets, Are Substantially More Likely
to Engage in Trafficking.
Type of Store
Trafficking
When Store is
Publicly-Owned
Trafficking
When Store is
Privately-Owned
Store
Violation
Rate
Trafficking
Rate
Store
Violation
Rate
T rafficking
Rate
1993 1998 1993 1998 1993 1998 1993 1998
Supermarkets 0.0 4.7* 0.0 3.0* 5.4 5.7 2.6 1.3
Large (iroceries 0.0 0.0 0.0 0.0 6.8 9.9 3.8 3.3
Other Types (small groceries,
convenience stores, gas/grocery,
specialty foods, etc.
1.7 4.3 1.8 4.6 12.0 14.0 15.1 12.3
All Stores 1.2 4.4 0.2 3,0 10.7 12.7 5.3 3.7
Notes: * See cndnotc"
Trafficking violation rates are calculated separately for stores and redemptions The
store violation rate is the percent of investigated stores caught trafficking weighted
by the national distribution ofstores. The trafficking rate is the percent of trafficked
redemptions in investigated stores, weighted by the national distribution of
redemptions.
Page 10
The store categories most likely to traffic continue to be small privately-owned stores and
privately-owned stores that do not stock a full-line of food (Table 4):
• Among these stores more than 1 of every 8 benefit dollars redeemed was trafficked.
• While these categories account for about 71 percent of all stores they account for only 14
percent of all redemptions.
Table 4 - Small Privately-Owned Stores Have the Highest Trafficking Rates But Redeem
Only 14 Percent of All Benefits Issued
Category of Store
... .
Trafficking
Rates
(Redemptions)
Percent of All
Stores
Percent of All Redemptions
1993 1998 1993 1998 1993 1998
Publicly-Owned Stores 0.2 t 12.8 12.8 28.0 30.0
Large Private Stores 2.7 1.5 17.2 16.5 56.2 55.8
Private - other stores 15.1 12.3 70.0 70.7 15.8 14.2
All stores 3.8 3.5 100.0 100.0 100.0 100.0
'See endnote 15.
Page 11
TRAFFICKING. FNS ENFORCEMENT AND EBT
FNS concentrates its enforcement efforts on stores most likely to traffic. In addition, the expansion
of Electronic Benefit Transfer (EBT) makes certain forms of trafficking harder to conduct and
large-scale trafficking easier to detect. For these reasons, it should not be surprising that we find
the largest reduction in the trafficking rate among the store categories most likely to traffic -
privately-owned stores, especially small ones that do not stock a full line of food (Chart 5).
Chart 5
Reductions In Trafficking Rate: 1993 -1998
■ 1993
□ 1998
All
stores
TRAFFICKING AND STORE LOCATION
The 1993 report examined the prevalence of trafficking by neighborhood and found that trafficking
is more frequent among stores located in the poorest of poor neighborhoods. The 1993 report also
found only a mild relationship between trafficking rates and a store's location in an urban
neighborhood. These two findings continued to be true in the 1996 - 1998 period.
Stores in the poorest of poor neighborhoods continue to be more likely to engage in
trafficking than stores located elsewhere, although the difference between rich and poor
neighborhoods has decreased somewhat (Table 5). Few recipients are likely to sell food stamp
benefits for less than they can buy in food, unless the need for cash is overwhelming. It is no
surprise, therefore, to find that the rate of trafficking (i.e., proportion of benefits trafficked)
continues to vary widely by the economic status of neighborhoods.
Page 12
Table 5 - Trafficking is More Frequent in the Poorest of Poor Neighborhoods.
Percent of
Households in Poverty
in Zip Code Where
Store is Located:
Trafficking Rates: Percent of All
Store
Violation
Rate
Trafficking
Rate
Stores
Redemp-tions
Otol0%
1993 1998 1993 1998 1993 1998 1993 1998
4.6 9.5 1.7 2.0 30.3 26.5 27.2 23.2
11 to 20% 8.7 10.7 4.1 3.1 3S.9 40.5 38.9 40.1
21 to 30% 13.0 13.2 3.8 3.3 20.1 20.5 20.1 21.6
over 30% 19.2 16.8 7.6 7.1 13.8 12.4 13.8 15.1
All Stores 9.4 11.7 3.8 3.5 100.0 100.0 100.0 100.0
Although some urban areas are widely perceived as having more crime than rural areas, we
found only a mild relationship between the trafficking rate and urbanicity. The Bureau of the
Census classifies zip codes by the urban/rural percentage of residents in the zip code. The
trafficking rates by urban/rural percentage in the zip code in which a store is located show a modest
increase in highly urban areas (Table 6).
Table 6 - The Trafficking Rate Is Slightly Higher In Highly Urban Areas.
Stores Located in Zip Codes
Where Percent Urban is:
Trafficking Rates:
Store
Violation
Rate Trafficking Rate
1993 1998 1993 1998
0 to 10% 6.1 12.9 3.5 2.4
11 to 50% 8.6 11.6 3.1 2.5
51 to 90% 7.1 10.9 2.8 3.0
90 to 100% 12.1 11.6 4.4 3.9
Page 13
rj
While trafficking rates remain low and do not vary sharply by urbanicity, between 1993 and
1998 we find a large increase in the store violation rate in rural and lower-urban areas (Chart
6). Table 5 indicates a similar increase in the store violation rate outside of the poorest areas. The
reason for these changes in store behavior is unknown.™
Chart 6
Change in Trafficking Patterns by Urbanicity:
1993 -1998
■ Store Violation Rate
D Traff icldng Rate
10%
Percent Urban
Stores in low trafficking areas continue to redeem the majority of food stamp benefits.
• Twelve percent of the nation's authorized food retailers are located in high poverty/high
trafficking areas, 88 percent are located in lower poverty/low trafficking areas.
• Eighty-five percent of redemptions flow through stores located in neighborhoods where less
than 30 percent of the population is below poverty.
Page 14
CONCLUSION AND
IMPLICATIONS FOR PROGRAM INTEGRITY
The rate of trafficking has decreased over this period. Although the data available are not
sufficient to determine causality, the direction and nature of the decrease are consistent with two
facts:
• The stores which redeem the majority of food stamp benefits continue to be stores with
the lowest trafficking rates. Overall, 84 percent of food stamp benefits are redeemed in
store categories with the lowest rates of trafficking.
• Electronic Benefit Transfer accounted for over half of all issuance during the measured
period.
EBT has expanded even more since these data were collected and it now represents over seventy
percent of all food stamp issuance.
Finally, during this period the store violation rate increased in rural and lower-poverty areas. While
this change should be monitored, its significance is muted by the fact that the proportion of benefits
trafficked in such areas (the rate of trafficking) is low.
Page 15
TECHNICAL DISCUSSION
When we look at additional considerations that bear on trafficking, we find two factors whi< h
would tend to increase our estimate and two others that would tend to decrease it. It is impu.tant to
discuss each of these additional considerations explicitly.
SOURCES OF UNDERESTIMATION
1. Our procedure underestimates two aspects of the trafficking problem. The first aspect leading
to Mm/erestimation is evasion trafficking:
• Among small retailers that are family-owned or where ownership is closely-held, some
violators do not redeem coupons for cash from the government (direct trafficking) but buy
food stock for resale from large stores with trafficked coupons (a form of tax evasion we
label "evasion trafficking"). Evasion trafficking is a gray area, since the practice does not
necessarily involve discounting: a small firm makes an illicit profit at the least risk of
detection if it accepts food stamps at full value for food from legitimate recipients, but uses
them (illegally) to buy food at supermarkets for resale.
• In our estimate we are most concerned about evasion trafficking when it is linked to
discounting (i.e., the firm buys food stamp benefits at a discount). We have no data to
estimate the extent of evasion trafficking by unauthorized food stores or restaurants.
However, evasion trafficking by authorized retailers is partially captured by our
estimating procedure, when the trafficking involves discounting. The data we use to
estimate direct trafficking adequately capture the rate at which all authorized stores engage
in discounting. What the data fail to do is account for redemptions that are unreported by
authorized discounting firms that buy food for resale with the coupons. If unreported
redemptions could be measured, then the evasion trafficking factor would increase the
national estimate of dollars diverted from food benefits by trafficking but would not change
the store-based violation rates useful for targeting future action.
• Engaging in evasion trafficking was relatively easy with food coupons but is substantially
more difficult under EBT."" Because the only ones to find evasion trafficking cost-effective
are small privately-owned stores who have not yet switched to EBT, the
potential impact of this factor is limited to a shrinking subset of the privately-owned
small-store component of our estimate.
Page 16
2. The second potential cause of underestimation is network trafficking:
• Some violating stores will traffic with strangers while others restrict their illegal activities to
people they know (which we label "network trafficking"). Investigators can and do catch
this type of trafficking, but it requires a harder investigation.
• As a result, some network trafficking is included in our estimate (because our
investigations include some cases where the network was penetrated and trafficking was
caught). But other instances of network trafficking are not included in our estimate
(because investigators were unable to penetrate the network and make the case). This source
of underestimation applies to all components of our model. If investigators could catch
all instances of network trafficking, the national estimate of trafficking diversions
would increase.""'
SOURCES OF OVERESTIMATION
1. However, our procedure also overestimates other aspects of the trafficking problem. A first
source of overestimation is the procedure used to determine legitimate food sales.
• With extremely rare exceptions, stores that engage in trafficking also sell food and we must
allocate some proportion of their total redemptions to legitimate food sales and the balance
to trafficking."" We purposefully used very low figures to estimate the percentage of
legitimate food sales by violating stores - this procedure serves our goal of assuring an
estimate of the maximum benefits diverted by trafficking. The estimate of trafficking
diversion would be lower to the extent that our method to estimate legitimate food sales was
more precise.
• This consideration is especially relevant to the large-store components of our model
(where most redemptions occur). We reviewed investigator reports in connection with
cases of supermarket trafficking." In supermarkets the percentage of total redemptions
our methodology attributes to trafficking (40%) is aboutfour times higher than
experienced FNS field investigators attribute to trafficking (10% or less) when
recommending sanctions or participating in other legal proceedings.
• To be consistent with the 1993 figures, we keep our method the same in this update report
- but it is likely that the percentage of a store's redemptions we attribute to trafficking
substantially overestimate trafficking, especially in supermarkets. Additional work is
being conducted to determine whether better estimates can be created.
Page 17
2. Another major source of overestimation is that investigations are a non-random sample of
stores.
• Our estimating procedure relies on investigations targeted to find fraud: our estimate
would decrease substantially if investigators had randomly selected average stores, rather
than selected suspicious stores on purpose.
• Of our four technical considerations, this is arguably the one with the largest impact
on our estimate and applies to all components of our model.
Page 18
ENDNOTES
1 Theodore F. Macaluso, The Detent of Trafficking in the Food Stamp Program (Alexandria,
VA: Food and Nutrition Service, USDA; 1995).
" Both the earlier report and this one intentionally use calendar, rather than fiscal, years for the
analysis. There are two reasons for this. First, it is necessary to combine investigations from
several years to achieve a sufficient number of cases for analysis, so the choice of a fiscal or
calendar metric is arbitrary. Second, the use of calendar year reinforces the fact that we are
providing estimates, rather than administrative data (which typically is presented on a fiscal year
basis).
'" There has been speculation that able-bodied adults without dependents (ABAWDS) are more
likely to traffic than other program participants. If this were true, then welfare reform time limits
on the duration of participation by ABAWDS might be expected to reduce the rate of trafficking.
However, the evidence available to USDA indicates that no one category of participant is either
more or less prone to traffic than any other category.
IV EBT also provides new ways to catch any trafficking that does occur. A new system, labeled
ALERT, analyzes EBT transaction data to catch some trafficking stores without the need for in-person
investigations. These cases are still relatively new and are not incorporated here. FNS is
working on developing a new trafficking measure to better reflect the impact of Electronic Benefit
Transfer. ALERT data will be included in the new measure.
v While food retailers constitute the overwhelming majority of authorized redeemers of food
stamp benefits, the Food Stamp Program has also authorized a few food wholesalers to accept
food stamp benefits. For simplicity, we refer to all authorized entities as retailers.
" Trafficked coupons are not always redeemed for cash from the government. Owners of small
authorized or unauthorized stores, restaurants, and the like can pretend o be recipients and illegally
use food stamps to buy food at supermarkets for resale in their stores. We label this "evasion
trafficking" (since it is a form of tax evasion) and discuss its impact on our estimate at the end of
this paper.
v" There is one trivial difference: the earlier report involved data on investigations started by January 1,
1991 and completed by March 1994 which were combined with redemption data from 1993 and presented
as a single result for calendar 1993: this update involves data on investigations completed between January
Page 19
1996 through December 1998 combined with redemptions from 1996 - 1998. which we annualize and
present as a single result for the 1996-1998 period. Because trafficking was less of a focus of investigators
in the 1980s than it is now. the earlier report involved a cut-off on the start of investigations to ensure that
the investigators' focus was on trafficking (rather than sale of ineligible items). Such a restriction is no
longer needed.
Vl" We obtained all investigations included in the FNS Store Investigation and Monitoring System
(SIMS) database for calendar years 1996 through 1998. A small fraction of these investigations
were of stores that could not be matched to zip codes in the redemption file and therefore were
not used in the analysis. Inspection of these dropped investigations indicated (1) that the
proportion of trafficking to non-trafficking outcomes in these investigations was similar to the
data used for the analysis and (2) the cases were distributed across the data in such a way that it
is implausible that they would change any substantive findings. The total number of SIMS
investigations and the number used in the analysis were as follows:
SIMS Analysis File
1996 3,709 3,690
1997 3,624 3,601
1998 3,095 3,063
Total 10,428 10,354
The five dimensions we employ consist of three that categorize stores (type of store, ownership,
and amount of food stamp business) and two that categorize the zip code in which each store was
located (degree of urbanization, percent of households in poverty). Specific definitions employed
are as follows:
Type of Store.
Store types on the FNS application form were collapsed to the following seven categories (to
ensure an adequate number of cases of each type):
Supermarket any store identifying itself to FNS as a supermarket or grocery with gross
sales over $2,000,000.
Large grocery any siore identifying itself to FNS as a supermarket or grocery with gross
sales between $500,000 and $2,000,000.
Small grocery any store identifying itself to FNS as a supermarket or grocery with gross
sales under $500,000.
Convenience
Specialty
any store identifying itself to FNS by this title, regardless of gross sales
any store identifying itself to FNS by this title, regardless of gross sales
They are almost always single product line stores such as meat markets, fish
markets, dairy stores, etc.
Page 20
Gas/Grocery
Other Types
any store identifying itself to FNS by this title, regardless of gross sales.
any store identifying itself to FNS by a title different than any of the
preceding, regardless of gross sales. Examples include produce stands,
general stores, combination grocery/bars, health/natural food stores, milk
and/or bread routes.
Ownership.
Ownership types on the FNS application form were collapsed to the following two categories (to
ensure an adequate number of cases of each type).
Public
Private
any store identifying itself to FNS as a public corporation (i.e., a retailer
whose stock trades publicly).
any store identifying itself to FNS as other than publicly-owned. This
includes private (i.e., closely-held) corporations as well as partnerships, sole
proprietorships, co-ops, etc.
("Franchise" is a separate category on the FNS application, not an ownership type: both public and
private ownership categories include stores that report themselves as franchises.)
Amount of Food Stamp Business.
Stores were categorized into deciles on the basis of food stamp redemptions. The purpose was
statistical, rather than analytical, to ensure that large disparities in redemptions by stores do not
distort results.
Urbanization.
Based on census data for the zip code in which the store is located. Four categories were employed:
0 to 10 percent urban population, 1! to 50 percent, 51 to 90 percent, and over 90 percent.
Poverty.
Based on census data for the zip code in which the store is located. Four categories were employed:
0 to 10 percent of residential population below poverty, 11 to 20 percent, 21 to 30 percent, and over
30 percent.
■ For calculating trafficking rates, the number of investigations in each store category are large
enough to give high confidence in the estimates (ranging from a low of 369 to a high of 3,665 by
store type).
' Statistically, the FNS investigation data base encompasses a sufficient number of cases to be used
as a post-stratified sample of the national "population" o. retailers. By categorizing the investigated
stores on the five dimensions described in note 8 and we ghting the stores, by category, to reflect
Page 21
the national population of retailers, by category, we are able to draw valid conclusions about the
national situation.
' The specific calculation was a two-stage one. The first stage combines the data on the trafficking
rates by type of store and store location with national redemption data to yield an estimate of the
gross redemptions by authorized food stores found trafficking. The second stage accounts for the
fact that some of the gross redemptions are legitimate food sales. To ensure consistency with the
earlier estimate, we continue to use the assumption that legitimate food sales account for 60 percent
of the gross redemptions among supermarkets and large grocery stores caught trafficking and treat
40 percent of their gross redemptions as trafficked. Among all other types of food stores, we
assume that only 10 percent of the gross redemptions are legitimate food sales among stores that do
not stock a full line of food (i.e., small grocery, convenience, specialty food, gas/grocery, and
"other" stores) and treat 90 percent of their gross redemptions as trafficked.
" We processed all stores received from FNS redemption files but used only the ones with a
match to zip code data in the analysis. Stores that had no redemptions were dropped from the
analysis (unless they had been investigated, in which case they were retained). For each specific
year the total number of authorized retailers received and total number in our analysis file are as
follows:
Received Analysis File
1996: 205.318; 202,850
1997: 196,408: 193,510
1998: 184.055. 180,857
For each specific year the sum of redemptions (total dollars) was:
Received Analysis File
1996: $21,713,774,005 $21,580,132,008
1997: $18,463,396,131 SI 8.322.710.580
1998: $16,433,240,311 $16,260,221,191
v We categorize stores according to how they categorized themselves in FNS authorization data.
Examples of public corporations are major supermarket chains, like Albertson's and Safeway and
gas-and-go mini-marts operated by companies like Texaco or Mobil. Many major supermarket
chains, such as the Publix chain in Florida, are private corporations. IGA stores which have the
appearance of a chain but are not public also fall under non-public ownership. Stores that most
readers consider "franchises" may fall under either the public or non-public heading, depending on
how they categorized themselves to FNS Southland's 7-Fleven chain are classified under public
corporations.
Page 22
"v In 1993 USDA investigators found no instances of trafficking at publicly-owned supermarkets.
Between 1995 and 1998, however, four cases of trafficking occurred in publicly-owned
supermarkets. Because there are relatively few investigations of supermarkets and because the
redemptions flowing through supermarkets are so large, these four cases have a large apparent
impact on trafficking rates. To be consistent, we report the trafficking rates exactly as computed in
the first trafficking report. However, an examination of the four cases indicates that the procedure-used
in the earlier report significantly overstate the amount of redemptions trafficked in
supermarkets. Relevant considerations include the following
• Only a very small number ofsupermarket cases delect trafficking in any one year. Combining
the data from the earlier report with this update, we found the following cases of trafficking in
publicly-owned supermarkets: 0 in 1993. 0 in 1994. 1 in 1995, 2 in 1996, 0 in 1997, 1 in 1998.
• Two ofthe four cases appear to involx t the actions ofa single clerk. In one of those cases, the
clerk was not even at the cash register when the transaction took place. Two of the four cases,
however, involved a lower-level manager at the store
• In three ofthe four cases, redemptions at the supermarket were in a pattern ofsignificant
decline; two ofthe three were being closed. It is possible that upper management gave
decreased attention to employee actions in such an atypical environment. (This speculation will
be evaluated as additional supermarket trafficking cases emerge over the next several years.)
• The percentage of redemptions attributed to trafficking in these four stores by the investigators
was substantially lower than the percentage we use in our calculations. In the first report when
trafficking was found at a supermarket or large grocery we attributed 40 percent of the total
redemptions in the store to trafficking. In these four instances of trafficking, investigators
estimated that 10 percent or less of total redemptions were trafficked.
• In light of the above, the true rate ofredemptions trafficked in supermarkets is likely to be
substantially below the 3 percent figure in Table 3
" The increase in store violation rates outs-Je of high poverty and highly urban areas may have
occurred for several reasons. For example, the results are possible if the decline in authorized
retailers differed by area. Alternatively, the results may reflect the expansion of F.BT. either if the
F.BT switch-over forces violators into nearby non-FBT areas (and those areas are less than 90
percent urban and/or the population in poverty is under 21%) or if rural or higher-income States are
implementing EBT at a slower rate It is also unclear at this stage whether the increase is occurring
among all non-urban stores or only those located along highways through rural a.-as. FNS is
developing a new trafficking measure to better reflect the impact of Flectronic Benefit Transfer.
These - and other - potential explanations will be analyzed as part of that effort.
" The store owner would need to have possession of multiple F.BT cards and make multiple
trips to supermarkets (a small-store owner using more than one card to pay for a large purchase
transaction would involve the supermarket in a violation that is readily detectable through the
Page 23
AI.HRT system; supermarkets are unlikely to accept that risk). Not only would the store owner
need to have several cards and use them at several places (or on different days). fa the practice
to be worth the risk of getting caught the balances left on the cards would need to be large (which
is not usually the case).
An additional potential consideration is the quality of the investigation. Even when retailers are
willing to traffic with strangers, investigators with greater experience and adequate time and
resources to establish a case are likely to catch more trafficking than investigators with less
experience, time and resources. We believe the overall quality of investigations in our samole is
high for two reasons. First, FNS investigative procedures provide adequate time and resources to
establish a case. Second, in the earlier report we only used cases from 1991 and later, to ensure that
investigators had at least two years of experience in establishing trafficking cases (or were hired
with the understanding that trafficking cases were highest priority), fn this report, most
investigators have at least six years of experience in establishing trafficking cases, which
strengthens our confidence in these estimates.
'" Cm rare occasions phantom stores - i.e.. fronts that take coupons but do not have a food business
- are found. This phenomenon is likely to decrease in the future for two reasons: (1) FNS has
expanded its staff resources to visit more stores H person; (2) EBT requires a visit from the F.BT
vendor to install terminals and the vendor will not install a terminal if they have questions about the
legitimacy of the business.
yx Seeendnote 15.
Page 24