This is the accessible text file for GAO report number GAO-14-641 entitled 'Supplemental Nutrition Assistance Program: Enhanced Detection Tools and Reporting Could Improve Efforts to Combat Recipient Fraud' which was released on August 21, 2014. This text file was formatted by the U.S. Government Accountability Office (GAO) to be accessible to users with visual impairments, as part of a longer term project to improve GAO products' accessibility. Every attempt has been made to maintain the structural and data integrity of the original printed product. Accessibility features, such as text descriptions of tables, consecutively numbered footnotes placed at the end of the file, and the text of agency comment letters, are provided but may not exactly duplicate the presentation or format of the printed version. The portable document format (PDF) file is an exact electronic replica of the printed version. We welcome your feedback. Please E-mail your comments regarding the contents or accessibility features of this document to Webmaster@gao.gov. This is a work of the U.S. government and is not subject to copyright protection in the United States. It may be reproduced and distributed in its entirety without further permission from GAO. Because this work may contain copyrighted images or other material, permission from the copyright holder may be necessary if you wish to reproduce this material separately. United States Government Accountability Office: GAO: Report to Ranking Member, Committee on the Budget, U.S. Senate: August 2014: Supplemental Nutrition Assistance Program: Enhanced Detection Tools and Reporting Could Improve Efforts to Combat Recipient Fraud: GAO-14-641: GAO Highlights: Highlights of GAO-14-641, a report to Ranking Member, Committee on the Budget, U.S. Senate: Why GAO Did This Study: In fiscal year 2013, SNAP, the nation's largest nutrition support program, provided about 47 million people with $76 billion in benefits. Fraud, including trafficking—-the misuse of program benefits to obtain non-food items-—has been a long-standing concern, and technology has provided additional opportunities to commit and combat such activities. State agencies are responsible for addressing SNAP recipient fraud under the guidance and monitoring of FNS. GAO was asked to review state and federal efforts to combat SNAP recipient fraud. GAO reviewed: (1) how selected state agencies combat SNAP recipient fraud; (2) the effectiveness of certain state fraud detection tools; and (3) how FNS oversees state anti-fraud efforts. GAO reviewed relevant federal laws, regulations, guidance, and documents; interviewed officials in 11 states; interviewed federal officials; tested fraud detection tools using fiscal year 2012 program data; and monitored websites for potential trafficking online. Although results are not generalizable to all states, the 11 states, selected based on various criteria including the size of their SNAP recipient household population and their payment error rates, served about a third of SNAP recipient households. What GAO Found: The 11 states GAO reviewed employed a range of detection tools, but experienced mixed success investigating and pursuing cases to combat potential Supplemental Nutrition Assistance Program (SNAP) recipient fraud. States reported using detection tools required or recommended by the Food and Nutrition Service (FNS), such as matching recipient data against prisoner and death files. However, most of selected states reported difficulties in conducting fraud investigations due to either reduced or maintained staff levels while SNAP recipient numbers greatly increased from fiscal year 2009 through 2013. Some state officials suggested changing the financial incentives structure to help support the costs of investigating potential SNAP fraud. For example, investigative agencies are not rewarded for cost-effective, anti-fraud efforts which prevent ineligible people from receiving benefits at all. GAO found limitations to the effectiveness of recommended replacement card data and website monitoring tools for fraud detection. FNS requires states to monitor SNAP households that request at least four cards per year, but selected states reported limited success detecting fraud this way. GAO's analysis found potential trafficking in 73 percent of households reviewed by focusing on SNAP households requesting cards in at least four monthly benefit periods. Benefits are allotted monthly, and a recipient selling their benefits and then requesting a new card would generally have one opportunity per month to do so. As a result, additional card requests in the same benefit period may not indicate increased risk of trafficking. Additionally, GAO found the FNS recommended e-commerce website monitoring tool to be less effective than manual searches in detecting posts indicative of SNAP trafficking. GAO found the recommended tool for monitoring social media to be impractical due to the volume of irrelevant data. Figure 1: Using Replacement Cards to Target Trafficking in Michigan, Fiscal Year 2012: [Refer to PDF for image: illustration] SNAP households identified with FNS approach: 8,190; SNAP households identified in GAO analysis: 4,935; Households reviewed with potential trafficking activity: 3,183; Households with most potential trafficking activity: 39. Source: GAO analysis of Supplemental Nutrition Assistance Program (SNAP) transaction data. GAO-14-641. [End of figure] FNS has increased its oversight of state anti-fraud activity in recent years by issuing new regulations and guidance, conducting state audits, and commissioning studies on recipient fraud since fiscal year 2011. Despite these efforts, FNS does not have consistent and reliable data on states' anti-fraud activities because its reporting guidance lacks specificity. For example, the guidance from FNS did not define the kinds of activities that should be counted as investigations, resulting in data that were not comparable across states. Additional oversight efforts, such as providing guidance to states for reporting consistent data, could improve FNS's ability to monitor states and obtain information about more efficient and effective ways to combat recipient fraud. What GAO Recommends: GAO recommends, among other things, that FNS reassess current financial incentives and detection tools and issue guidance to help states better detect fraud and report on their anti-fraud efforts. Agency officials agreed with our recommendations. View [hyperlink, http://www.gao.gov/products/GAO-14-641]. For more information, contact Kay E. Brown at (202) 512-7215 or brownke@gao.gov, or Seto J. Bagdoyan at (202) 512-6722 or bagdoyans@gao.gov. [End of section] Contents: Letter: Background: Selected States Employed a Range of Tools to Detect Fraud, but Conducted Investigations with Limited Staff and Pursued Cases with Mixed Success: FNS's Guidance and Tools Can Be Used to Detect Potential SNAP Trafficking but Effectiveness is Limited: FNS Increased Its Oversight of State Anti-Fraud Activities but Lacks Reliable Data on These Efforts: Conclusions: Recommendations for Executive Action: Agency Comments and Our Evaluation: Appendix I: Objectives, Scope and Methodology: Appendix II: Trafficking Flags in SNAP Households Receiving Excessive Replacement Cards: Appendix III: Total E-commerce and Social Media Postings Detected During Testing Periods: Appendix IV: Selected States' Experiences Using FNS's Recommended Automated Tool, and GAO 30-day Test Results: Appendix V: List of Food and Nutrition Service-Commissioned Studies: Appendix VI: GAO Contacts and Staff Acknowledgment: Related GAO Products: Tables: Table 1: Tools Used to Detect Potential Supplemental Nutrition Assistance Program (SNAP) Eligibility Fraud: Table 2: Tools Used to Detect Potential Supplemental Nutrition Assistance Program (SNAP) Trafficking: Table 3: SNAP Households Receiving Replacement Cards in Fiscal Year 2012: Table 4: SNAP Households Receiving Excessive Replacement Cards and Making Transactions Potentially Indicative of Trafficking in Fiscal Year 2012: Table 5: Examples of Selected Suspicious Transactions Made by One Household Resulting in Trafficking Flags, Fiscal Year 2012: Table 6: Key Food and Nutrition Service (FNS) Oversight Regulations, Guidance and Policies, Fiscal Years 2011 through 2014: Table 7: Data Ranges for Percentages: Table 8: State Selection Criteria: Table 9: Geographic Locations Monitored in 11 Selected States: Table 10: Number of Trafficking Flags Associated with Fiscal Year 2012 Transactions Made by Selected High Replacement Card Households in Three States: Table 11: Number of Michigan High Replacement Card Households Categorized by Count of Trafficking Flags and Replacement Card Benefit Periods, Fiscal Year 2012: Table 12: Number of Massachusetts High Replacement Card Households Categorized by Count of Trafficking Flags and Replacement Card Benefit Periods, Fiscal Year 2012: Table 13: Number of Nebraska High Replacement Card Households Categorized by Count of Trafficking Flags and Replacement Card Benefit Periods, Fiscal Year 2012: Table 14: E-commerce Postings Advertising Potential Sale of Food Stamp Benefits for Cash: Table 15: E-commerce Postings Advertising Potential Sale of Food Stamp Benefits for Services: Table 16: E-commerce Postings Advertising Potential Sale of Food Stamp Benefits for Goods: Table 17: Social Media Postings Soliciting Food Stamp Benefits: Figures: Figure 1: Number of Supplemental Nutrition Assistance Program (SNAP) households per Investigator in Selected States, Fiscal Years 2009 and 2013: Figure 2: Targeting Potential Supplemental Nutrition Assistance Program (SNAP) Benefit Trafficking Using Replacement Card and Transaction Data to Identify Higher Risk Households in Michigan, Fiscal Year 2012: Figure 3: Analysis of Automated Tool Recommended for Monitoring E- commerce Websites: Figure 4: Results of 30-day Monitoring of One Popular E-Commerce Website Using Automated and Manual Tools to Detect Posts Indicative of Food Stamp Trafficking and States' Experiences with Using Tools: Figure 5: FNS Recipient Integrity Review Components: Abbreviations: ALERT: Anti-Fraud Locator Using Electronic Benefits Transfer Retailer Transactions: ADH: Administrative Disqualification Hearing: eDRS: Electronic Disqualified Recipient System: EBT: electronic benefit transfer: FNS: Food and Nutrition Service: FY: fiscal year: OMB: Office of Management and Budget: PIN: Personal Identification Number: SLEB: state law enforcement bureau: SNAP: Supplemental Nutrition Assistance Program: Recovery Act: American Recovery and Reinvestment Act of 2009: USDA: U.S. Department of Agriculture: USDA OIG: U.S. Department of Agriculture Office of Inspector General: [End of section] United States Government Accountability Office: GAO: 441 G St. N.W. Washington, DC 20548: August 21, 2014: The Honorable Jeff Sessions: Ranking Member: Committee on the Budget: United States Senate: Dear Senator Sessions: In fiscal year 2013, the federal government provided more than $76 billion in benefits to help about 48 million people purchase food through the Supplemental Nutrition Assistance Program (SNAP). On average, recipient households received about $275 a month in assistance in that year. Since fiscal year 2009, SNAP has experienced an over 50-percent increase in distributed benefits and an over 40- percent increase in recipients. Such rapid program growth can increase the potential for fraud unless appropriate agency controls are in place to help minimize these risks. The Office of Management and Budget has designated SNAP as a high-error program due to the estimated dollar amount in improper payments for fiscal year 2013. [Footnote 1] Furthermore, program officials have had long-standing concerns that some recipients falsify information to improperly receive benefits, or misuse their benefits to solicit or obtain non- food goods, services and cash--a practice known as trafficking. Technology has provided new opportunities to commit as well as to combat such fraud. For example, e-commerce and social media websites have emerged as new venues for trafficking benefits. Conversely, monitoring recipient transaction data may provide clues to potential SNAP fraud. The state and federal governments share responsibility for addressing SNAP recipient fraud. State agencies are directly responsible for detecting, investigating, and prosecuting recipient fraud, and the U.S. Department of Agriculture's Food and Nutrition Service (FNS) is responsible for guiding and monitoring this state activity. FNS has traditionally focused on pursuing retailer fraud. We reported on these efforts in fiscal year 2007 and found that FNS was making progress in using electronic data to investigate trafficking.[Footnote 2] To enhance these efforts, FNS implemented most of our recommendations by taking additional steps to target and provide early oversight of stores most likely to traffic; developing a strategy to increase penalties for trafficking; and promoting state efforts to pursue recipients suspected of trafficking. Since then, the agency has increased attention to the recipient side of trafficking. For example, in fiscal year 2012, FNS recommended that states use certain tools, such as analyzing transaction data for those requesting multiple benefit card replacements and monitoring websites where traffickers may be attempting to buy or sell SNAP benefits. In light of this increased emphasis, you asked us to review federal and state efforts to combat SNAP recipient fraud. This report examines: (1) how selected state agencies combat SNAP recipient fraud; (2) the effectiveness of certain fraud detection tools recommended to states, including benefit card replacement data and e-commerce and social media website monitoring; and (3) FNS's oversight of state anti-fraud efforts. For all three reporting objectives, we focused on federal and state SNAP recipient anti-fraud work for fiscal years 2009 to 2014, a period after the program received additional funding through the American Recovery and Reinvestment Act of 2009 (Recovery Act).[Footnote 3] We reviewed relevant federal laws, regulations, program guidance and reports, and we interviewed FNS officials in headquarters and all seven regional offices to obtain information for all three objectives. For the first objective, we selected 11 states for our review-- Florida, Maine, Massachusetts, Michigan, Nebraska, New Jersey, North Carolina, Utah, Tennessee, Texas, and Wyoming--to achieve variation in geographic location, and a mix of high, medium and low SNAP payment error rates, percent of the total number of SNAP households nationwide, and proportion of recipients whom state officials reported as disqualified from the program due to non-compliance. For all 11 states, we interviewed knowledgeable state and local officials about their recipient anti-fraud work and obtained related documentation. For the second objective, we monitored a popular e-commerce website for 30 days and a popular social media website for 5 days, to determine how our selected states could use certain automated monitoring tools recommended by FNS to detect potential SNAP fraud. [Footnote 4] We also analyzed fiscal year 2012 replacement card and transaction data for households in three of the selected states-- Michigan, Massachusetts, and Nebraska--to assess the extent to which certain analyses could better uncover patterns of potential fraud. We selected these three states to provide information on a mix of high, medium and low states in terms of their percentage of the total number of SNAP households nationwide. Recipient households in the 11 states we reviewed represent about one-third of all SNAP program households; however, the information we report from these states is not generalizable to all states. For the third objective, we obtained and analyzed documents and reports relevant to FNS's program oversight, including their fiscal year 2013 assessments of state anti-fraud work for all 50 states and the District of Columbia. All of the data included in this report were assessed and determined to be sufficiently reliable for our purposes. We conducted this performance audit from April 2013 through July 2014 in accordance with generally accepted government auditing standards. Those standards require that we plan and perform the audit to obtain sufficient, appropriate evidence to provide a reasonable basis for our findings and conclusions based on our audit objectives. We believe that the evidence we obtained provides a reasonable basis for our findings and conclusions based on our audit objectives. Background: Federal and State Roles for Addressing SNAP Fraud: The goal of SNAP, formerly known as the federal Food Stamp Program, is to help low-income individuals and households obtain a more nutritious diet. It does so by supplementing their income with benefits to purchase allowable food items. The federal government pays the full cost of the benefits and shares the responsibility and costs of administering the program with the states. Specifically, FNS is responsible for promulgating program regulations and ensuring that states comply with these regulations by issuing guidance and monitoring their activity. FNS officials at headquarters are assisted in this oversight work by officials in seven regional offices. FNS also determines which retailers are eligible to accept SNAP benefits in exchange for food and investigates and resolves cases of retailer fraud. State officials, on the other hand, are responsible for determining the eligibility of individuals and households, calculating the amount of their monthly benefits and issuing such benefits on an electronic benefit transfer (EBT) card in accordance with program rules. States are also responsible for investigating possible violations by benefit recipients and pursuing and acting on those violations that are deemed intentional.[Footnote 5] Intentional program violations include acts of fraud, such as making false or misleading statements in order to obtain benefits and trafficking (i.e., using benefits in unallowable ways, such as by exchanging benefits for cash or non-food goods and services or attempting to do so).[Footnote 6] Recipients can traffic benefits by: * Selling benefits to retailers - recipients collaborate with retailers who exchange cash for SNAP benefits. For example, a retailer can allow a recipient to charge $100 on his or her EBT card and then pays the recipient $50 instead of providing food. * Selling EBT cards to another person - recipient exchanges the EBT card and the corresponding Personal Identification Number (PIN) [Footnote 7] for cash or non-food goods or services (e.g., rent or transportation). These sales can occur in person or by posting offers on social media and e-commerce sites. All of these trafficking activities may result in recipients having to give their EBT card and PINs to another person who may not return the card. Recipients can report sold EBT cards as lost or stolen to state agencies or EBT management contractors and receive new cards which can be used for future transactions, for example, when the benefits are replenished the next month. According to a September 2012 U.S. Department of Agriculture Office of Inspector General (USDA OIG) report, the magnitude of program abuse due to recipient fraud is unknown because states do not have uniform ways of compiling the data that would provide such information. Therefore, in the report, the USDA OIG recommended that FNS determine the feasibility of creating a uniform methodology for states to calculate their recipient fraud rate.[Footnote 8] As FNS seeks to address this recommendation, it is legally required to monitor its potential improper payments of SNAP benefits. The agency estimated an improper payment or error rate of the program at 3.4 percent, which represented an estimated $2.6 billion in wrongful payments, in fiscal year 2013.[Footnote 9] The percentage represents benefits distributed in error due to administrative as well as recipient errors, not all of which can be attributed to fraud. However, due to the large dollar amount involved in improper payments, the Office of Management and Budget (OMB) has placed SNAP on its list of high-error programs. [Footnote 10] Furthermore, after studying the cause of these errors, USDA officials stated that over 90 percent were due to verification errors. These types of errors occur when an agency fails to or is unable to verify recipient information--including earnings, income, assets, or work status--even though verifying information exists in third-party databases or other resources. Examples of verification errors include an agency not confirming a recipient's reported earnings or work status through existing databases, or the recipient failing to provide an agency with information on earnings. Given FNS's role of directly overseeing retailer eligibility and disqualification, federal officials have traditionally focused on retailer trafficking. In 1996, FNS was given legal authority to disqualify retailers by using EBT transaction data--which display suspicious patterns of benefit use--as its sole form of evidence. [Footnote 11] FNS maintains such transaction data within its Anti- Fraud Locator Using Electronic Benefits Transfer Retailer Transactions (ALERT) system. In our October 2006 report on potential retailer fraud, we found that federal officials had concerns about state efforts to address recipient trafficking, and recognized that retailer trafficking can only occur when willing recipients are involved. [Footnote 12] At the time of that report, federal officials told us that they were providing state officials with lists of recipients involved in their retailer trafficking cases, but many states were not acting on this information at the time because it was difficult and costly to prove individual trafficking cases. Furthermore, as we noted in our September 2010 report, the USDA OIG found that states were not analyzing their EBT data to detect misuse of benefits, largely because FNS did not require this.[Footnote 13] FNS has calculated a retailer trafficking rate, which was estimated to involve 1.3 percent of benefits issued from fiscal year 2009 to 2011--a total of $858 million.[Footnote 14] State Anti-Fraud Activity Requirements: States must adhere to several requirements for detecting SNAP recipient fraud, conducting investigations and providing due process prior to disqualifying program violators. For example, states are required to have fraud detection units covering areas in which 5,000 or more households participate in the program; however, those working on fraud investigations need not be dedicated to this work full-time or exclusively to SNAP cases. States must also conduct data matches at the time of application and at other times to determine whether the information provided for a potential recipient is for someone who is incarcerated, deceased, or disqualified from the program. State SNAP agencies are responsible for pursuing judgments against those who intentionally violate SNAP rules. These judgments can be pursued within the state agency through an Administrative Disqualification Hearing (ADH) or through the judicial system in a court determined to have jurisdiction over the case. When a state decides to administratively pursue disqualification of a recipient for intentional program violations, the state is responsible for conducting a series of actions, such as providing timely notification to the recipient that there will be an ADH, and for states that have waiver procedures, that the recipients may waive their right to a hearing. If it is determined through the hearing or criminal prosecution that a person has intentionally violated program rules or the person has waived the hearing, only the person involved in the case is disqualified and not the entire household, but the entire household is responsible for repaying the specific ill-gotten or misused benefit amount. States are generally allowed to retain 35 percent of the fraud-related, overpaid benefits they collect, and the rest is returned to the federal government. In fiscal year 2012, states reported to FNS that they collected about $74 million in fraud- related claims.[Footnote 15] To assist other states and FNS, states are also responsible for reporting on disqualifications and other fraud-related activities. Specifically, within 30 days of disqualifying a benefit recipient, state officials are to report this information to FNS through a database known as the Electronic Disqualified Recipient System (eDRS).[Footnote 16] This information allows other states to learn who has been disqualified elsewhere and to impose penalties, which vary based on the number and type of offense.[Footnote 17] Furthermore, states are required to report their fraud-related activity to FNS on their annual Program and Budget Summary Statements. This report, provided through the form FNS-366B, is to include the number of investigations and disqualifications, and the dollar amount of their fraud claims.[Footnote 18] Selected States Employed a Range of Tools to Detect Fraud, but Conducted Investigations with Limited Staff and Pursued Cases with Mixed Success: The 11 states we reviewed employed a range of detection tools, but experienced mixed success in combating SNAP fraud. Although some were able to leverage additional resources, officials in most states reported challenges with potential fraud because their staff remained limited while recipient numbers grew. Furthermore, pursuing cases through administrative hearings and the courts generally resulted in disqualifications but collecting overpayments was a challenge. For Fraud Detection, Selected States Employed Tools Such As Data Matching, Referrals, Analysis of Transaction Data, and Website Monitoring: Detection: Both standard and state-initiated tools enable states to detect eligibility and trafficking fraud: Source: GAO analysis of information provided by selected states. In the majority of states we reviewed, officials told us they were using well-known tools for detecting potential recipient eligibility fraud, such as data matching and referrals obtained through fraud reporting hotlines and websites. Specifically, all 11 states that we reviewed had fraud hotlines or websites, and all matched information about SNAP applicants and recipients against various data sources to detect those potentially improperly receiving benefits, as FNS recommended or required. (See table 1.) Table 1: Tools Used to Detect Potential Supplemental Nutrition Assistance Program (SNAP) Eligibility Fraud: Detection tool: eDRS match; Federal or state initiative: FNS required; Number of 11 states that reported using tool: 11; Description: The Food and Nutrition Service (FNS) requires states to check for disqualified individuals in the Electronic Disqualified Recipient System (eDRS) when certifying or recertifying them for SNAP (all 11 states). Detection tool: Prisoner match; Federal or state initiative: FNS required; Number of 11 states that reported using tool: 11; Description: States must routinely match applicants and recipients against a prisoner verification system to prevent receipt of SNAP benefits by incarcerated individuals (all 11 states). Detection tool: Death match; Federal or state initiative: FNS required; Number of 11 states that reported using tool: 11; Description: States must check SSA's Death Master File to prevent receipt of benefits by deceased individuals[A] (all 11 states). Detection tool: PARIS match[B]; Federal or state initiative: FNS recommended; Number of 11 states that reported using tool: 11; Description: States provide lists of SNAP recipients for the quarterly PARIS matching, and receive lists of recipients who are also on the SNAP rolls of another state (all 11 states). Detection tool: Wage matching; Federal or state initiative: State-initiated; Number of 11 states that reported using tool: 9; Description: 9 states reported data matching to detect unreported and underreported wages (Florida, Maine, Massachusetts, Michigan, North Carolina, New Jersey, Tennessee, Texas, Wyoming). Detection tool: Other data matching; Federal or state initiative: State-initiated; Number of 11 states that reported using tool: 6; Description: 6 states verify information provided by applicants/ recipients by matching with other data sources, such as local jails, schools, and lists of lottery winners (Florida, Maine, Michigan, North Carolina, Tennessee, Texas). Detection tool: Online data search services; Federal or state initiative: State-initiated; Number of 11 states that reported using tool: 2; Description: 2 states pay a private company for searches of numerous public and private databases including vital statistics, current wage and new hire data, child support, and residence information based on phone lines and motor vehicle registrations (Florida, Texas). Detection tool: Identity verification software; Federal or state initiative: State-initiated; Number of 11 states that reported using tool: 2; Description: 2 states use software that requires individuals to provide information confirming their identities when they set up or access an online SNAP account (Florida, Texas). Detection tool: Error-prone profile; Federal or state initiative: State-initiated; Number of 11 states that reported using tool: 1; Description: State provides case workers with a profile for applications that lists items to watch for that may indicate applicant fraud (Florida). Detection tool: Mapping software; Federal or state initiative: State-initiated; Number of 11 states that reported using tool: 1; Description: State uses locator software to identify individuals applying for SNAP from a computer in another state (Michigan). Detection tool: Public referrals; Federal or state initiative: State-initiated; Number of 11 states that reported using tool: 11; Description: States provide fraud hot lines or online fraud referral (all 11 states). Detection tool: Exchange of most recent available SNAP enrollment data; Federal or state initiative: FNS pilot; Number of 11 states that reported using tool: 1; Description: 5 states are participating in a pilot of the exchange of their most recent available SNAP enrollment data (Florida, Alabama, Georgia, Louisiana, Mississippi). Source: GAO analysis of data from selected states. GAO-14-641. [A] Death Master File database contains the complete name of the deceased, the Social Security Number, birth and death dates, and whether the death report was verified. [B] PARIS is the Public Assistance Reporting Information System administered by the Department of Health and Human Services. [End of table] Beyond the required and recommended data matches, Florida, Texas, Michigan, and one county in North Carolina use specialized searches that check numerous public and private data sources, such as school enrollment, vehicle registration, vital statistics, and credit reports and other data on out-of-state program participation and benefit use to detect potential fraud prior to providing benefits to potential recipients. Florida officials told us that this focus on preventive efforts was key to helping them manage recent constraints on their investigative budgets. Specifically, Florida officials mentioned that when their investigative staff was reduced because of budget cuts in 2005, they shifted the majority of their anti-fraud resources from post-eligibility fraud investigations to preventing ineligible individuals from receiving SNAP benefits. This shift has allowed the state to more cost-effectively manage its efforts to combat potential fraud by developing detection tools against eligibility fraud and improper benefit receipt, such as identification verification software and profiles that case workers can use to identify error-prone applications. To address trafficking, officials in the 11 states reported that they analyzed patterns of EBT transactions and monitored replacement card data and online postings, as recommended or required by FNS. (See table 2.) When reviewing EBT transactions, state officials attempt to uncover patterns that may indicate trafficking, much in line with what FNS has been doing for years to uncover retailer fraud. Officials in two states mentioned that, for some cases, this EBT data analysis is done only after receiving fraud referrals through the hotline and websites. For example, while Florida officials reported that they routinely review EBT transaction data for suspicious patterns, Texas officials reported that they only review transactions for individuals or households after they have been referred to them because of potential fraud. Table 2: Tools Used to Detect Potential Supplemental Nutrition Assistance Program (SNAP) Trafficking: Detection tool: Monitoring of excessive replacement EBT cards; Federal or state initiative: FNS required; Number of 11 states that reported using tool: 11; Description: The Food and Nutrition Service (FNS) requires states to track recipients who have requested 4 or more replacement electronic benefit transfer (EBT) cards in a 12-month period. States may send them letters explaining the proper use of EBT cards and/or warn them that their EBT transactions are being monitored. (all 11 states). Detection tool: Monitoring of online postings; Federal or state initiative: FNS recommended; Number of 11 states that reported using tool: 8; Description: 8 states use either automated feeds or manual monitoring to detect postings on social media and e-commerce websites by individuals seeking to sell SNAP benefits (Florida, Maine, Michigan, New Jersey, Tennessee, Texas, Utah, Wyoming). Detection tool: Analysis of EBT transactions data; Federal or state initiative: FNS recommended; Number of 11 states that reported using tool: 11; Description: States analyze EBT transaction data for patterns that may indicate trafficking. The data analysis efforts vary from state to state. Source: GAO analysis of data from selected states. GAO-14-641. [End of table] Most of the Selected States Reported Difficulty Conducting Fraud Investigations Due to Limited Staff and Growing Numbers of Recipients, but Some Leveraged Additional Resources: Investigation: States struggle to provide investigative resources that meet the expectations of the Food and Nutrition Service (FNS). Source: GAO analysis of information provided by selected states. The size and organization of the investigative units differed among the 11 states we reviewed, with wide variation in the number of staff available to investigate potential SNAP recipient fraud.[Footnote 19] For example, in 2013, Massachusetts and New Jersey had 498,580 and 432,270 recipient households, respectively, but Massachusetts, where SNAP was administered at the state level, had just 37 investigators, while county-administered New Jersey had nearly 300. Furthermore, the investigators in the 11 states we reviewed each had responsibilities unrelated to SNAP. Although officials in three states--Massachusetts, Tennessee, and Wyoming--reported that the majority of their investigations involved potential SNAP fraud, state investigators in all 11 states we reviewed were also responsible for pursuing fraud in other public assistance programs, such as Medicaid, Temporary Assistance for Needy Families, and child care and housing assistance programs. In North Carolina, fraud investigation was not the primary responsibility of some local officials who did this work; state officials reported that some counties opted to have caseworkers or program supervisors conduct fraud investigations. In general, state officials reported that limits on staffing levels are significant hindrances to their investigations of eligibility fraud and trafficking, with 8 of the 11 states we reviewed reporting inadequate staffing due to attrition, turnover, or lack of funding. Of the 10 states that were able to provide the information,[Footnote 20] the number of SNAP households per investigator increased in 8 states between fiscal years 2009 and 2013 by as much as 155 percent. In contrast, Maine and Michigan have increased their investigative staff, which decreased their household-to-investigator ratios in fiscal year 2013. (See figure 1.) Figure 1: Number of Supplemental Nutrition Assistance Program (SNAP) households per Investigator in Selected States, Fiscal Years 2009 and 2013: [Refer to PDF for image: horizontal bar graph] Number of households per investigator: State: Florida; 2009: 7,705; 2013: 19,635, State: Maine; 2009: 11,062; 2013: 7,669, State: Massachusetts; 2009: 9,884; 2013: 13,475, State: Michigan; 2009: 9,258; 2013: 8,833, State: Nebraska; 2009: 8,206; 2013: 11,340, State: New Jersey; 2009: 766; 2013: 1,470, State: Tennessee; 2009: 7,115; 2013: 8,829, State: Texas; 2009: 6,919; 2013: 14,434, State: Utah; 2009: 3,685; 2013: 7,216, State: Wyoming; 2009: 1,864; 2013: 4,053, Source: GAO analysis of information provided by selected states. GAO- 14-641. Notes: New Jersey provided calendar year information for investigators. Florida lost 27 investigators in late 2009, dropping from 130 to 103. North Carolina was unable to provide the number of investigators because some local offices do not have designated fraud investigators. Furthermore, all investigators in the selected state were responsible for pursuing fraud in other public assistance programs, and therefore, could be responsible for monitoring a larger population than is mentioned in the figure. [End of figure] In their effort to combat potential fraud, some states implemented a way to leverage their available investigative resources. Specifically, four of the states we reviewed--Florida, Massachusetts, Michigan and Nebraska--had implemented and two states--Maine and North Carolina-- were in the process of implementing state law enforcement bureau (SLEB) agreements. FNS has been supportive of states' efforts to establish these agreements between state SNAP agencies and federal, state, and local law enforcement agencies which enable state SNAP investigators to cooperate in various ways with local, state, and federal law enforcement agents, including those within the USDA OIG. For example, under these agreements, law enforcement agencies can notify the SNAP fraud unit when they arrest someone who possesses multiple EBT cards, and SNAP agencies can provide "dummy" EBT cards for state and local officers to use in undercover trafficking investigations. According to officials in one Florida county, this type of cooperation allowed local police officers to make 100 arrests in its first undercover operation of recipients who were allegedly trafficking SNAP benefits. Furthermore, some state and local officials in Michigan, Maine, and Florida told us that increasing awareness of SNAP trafficking among local law enforcement officials helps in resolving these matters when potential trafficking is uncovered in other police investigations. For example, while investigating drug- related crimes, officials in those states told us they have uncovered multiple EBT cards in the possession of one person. In light of their increased SNAP caseload, some officials suggested changing the incentive structure to help states address the costs of investigating potential SNAP fraud. According to GAO's Fraud Prevention Framework, investigations, although costly and resource- intensive, can help deter future fraud and ultimately save money. [Footnote 21] Officials in one state told us that it would help if FNS would provide additional financial incentives for states to prevent potential fraud at the time of application beyond what is currently provided for recovered funds.[Footnote 22] When fraud by a recipient is discovered, the state may generally retain 35 percent of the recovered overpayment, but when a state detects potential fraud by an applicant and denies the application, there are no payments to recover. Officials in four of the states we reviewed said that their anti-fraud efforts could be enhanced if the percentage of recovered overpayments that states may retain was increased, and officials in three states said that FNS should direct that states apply the retention money to anti-fraud efforts. Overall, state anti-fraud incentives have the potential to produce federal cost savings by encouraging state officials to prevent the benefits from being issued to ineligible people as well as deter fraud by more actively investigating and recovering funds.[Footnote 23] While Selected States Pursue Fraud through Administrative Hearings and the Courts, They Reported Difficulties with Prosecutions and Overpayment Recovery: Pursuing Claims: States pursue fraud suspects through administrative hearings or the courts, but prosecutions and overpayment recovery are both challenging. Source: GAO analysis of information provided by selected states. Officials in most of the 11 states we reviewed said that they have mainly pursued cases of eligibility fraud, such as the misrepresentation of household income or composition. In addition to testimony from witnesses, state investigators are able to build cases based on public records and employment statements to prove the misrepresentation. However, state officials reported that trafficking is more difficult to prove. Officials in North Carolina and a prosecutor in Michigan noted that trafficking cases involve two individuals breaking the law, and it can be difficult to get one to testify against the other. For example, the Michigan prosecutor told us about a case in which a landlord for a subsidized housing complex was receiving SNAP benefits in exchange for rent, and the tenants would not testify against this person because they thought she was doing them a favor by accepting the SNAP benefits as payment. State officials we interviewed also reported that the willingness of local prosecutors to pursue charges in court for SNAP fraud has varied across jurisdictions. Officials in eight states reported that a minimum dollar threshold of fraudulently-obtained benefits was required for prosecuting cases in court, ranging from $100 (in Tennessee) to $5,000 (in Texas). Prosecutors in some local jurisdictions were not willing to accept SNAP fraud cases at all. For example, prosecutors in one county in North Carolina told SNAP officials that they would not prosecute SNAP fraud cases because they need their resources for more serious criminal cases. Texas officials said that some local prosecutors in their state have also refused to prosecute SNAP cases due to workload concerns. Other prosecutors we interviewed said that to make efficient use of their limited resources, they have often sought plea deals that require the individual to repay the government rather than going to trial. Such plea deals may call for the individual to be arrested if the SNAP benefits are not repaid, and may also require that a person have a criminal record as a result of the plea. Furthermore, plea deals mitigate some of the unpredictability of trying a case before a jury. Prosecutors in Tennessee and Florida said that juries may be unwilling to convict individuals of SNAP fraud because they may be sympathetic to recipient claims that they do not understand government regulations or are compelled to commit fraud to support their families. SNAP officials in North Carolina said they were concerned about losing the deterrent effect of prosecutions due to the unwillingness of the judicial system to undertake SNAP recipient fraud cases. Recovering overpayments from individuals found to have committed fraud in either an administrative or a court proceeding has been a challenge, according to officials we interviewed in Florida and Michigan. Specifically, those officials reported that an individual who is disqualified may be required to repay an overpayment, but may not have enough income to do so. Furthermore, if the individual becomes eligible for SNAP benefits again after the period of disqualification is over, the state may garnish the future SNAP benefits to repay the recipient's prior debt. However, when an individual is permanently disqualified from the program, garnishment is not possible. To encourage people to repay the benefits, one local Michigan prosecutor has established a program that offers to erase the individual's criminal record if the individual makes full restitution through a repayment plan. The program helps collect restitution of fraud payments in all the county's welfare programs and has had an 80 percent success rate in collecting repayments, according to the local prosecutor. States' difficulty collecting overpayments compounds their concerns about having adequate resources for investigations because some states use recovered overpayments for this purpose. FNS's Guidance and Tools Can Be Used to Detect Potential SNAP Trafficking but Effectiveness is Limited: Selected states reported difficulties using FNS recommended replacement card data as a fraud detection tool, and our data analysis found that a more targeted approach may better identify potential fraud. Our testing found the recommended e-commerce monitoring tool less effective than manual searches in detecting postings indicative of potential trafficking, and we found the tool for monitoring social media to be impractical for states due to the volume of irrelevant data. Selected States Report Limited Success Using Replacement Card Data as a Detection Tool: Although FNS requires that states look at replacement card data as a potential indicator of trafficking, states reported difficulties using the data as a fraud detection tool. In 2012, FNS issued guidance to states based on a best practice used in North Carolina, encouraging states to review recipients who have requested four or more replacement EBT cards within 12 months because such behavior may indicate trafficking. In 2014, FNS finalized a rule that requires states to monitor replacement card data and send notices to those SNAP households requesting excessive replacement cards, defined as at least four cards in a 12-month period. All 11 states we reviewed reported tracking recipients who make excessive requests for replacement EBT cards and sending them warning letters, as required by FNS, but they have not had much success in detecting fraud through that method. At the time of our review, four states reported that they had not initiated any trafficking investigations as a result of this monitoring, and five states reported a low success rate for such investigations. One state had just started monitoring replacement card data. Only one of our selected states reported some success using the replacement card data to identify and pursue trafficking. Furthermore, although state officials recognized that some replacement card requests may be related to potential fraud, officials from 7 of the 11 states reported that the current detection approach specified by FNS often leads them to people who make legitimate requests for replacement cards for reasons such as unstable living situations or a misunderstanding of how to use the SNAP EBT card. North Carolina officials also mentioned that when they originally developed this approach currently required by FNS, it was not intended to detect trafficking. Rather, it was to help them manage the number of replacement card requests they received.[Footnote 24] FNS is aware of states' concerns about the effectiveness of this effort, but it continues to stress that monitoring these data is worthwhile. For example, FNS officials reported that they are also aware that many replacement card requests are legitimate but they feel that the monitoring of replacement card data has an important educational component, as it allows states to identify situations where a recipient requires education on how to use their SNAP EBT card. FNS officials also reported that states have seen a reduction in households continuing to request replacement cards related to these efforts. However, FNS's Western Regional officials reported that, given states' experiences with the current process, it may be better for states to be more selective in sending notices. Targeted Analysis of Excessive Replacement Cards Found Potential Recipient Trafficking: Our analysis found indicators of potential SNAP trafficking in households with excessive replacement cards, suggesting that states may be able to use replacement card data to help identify trafficking by taking a targeted approach to analyzing the data in conjunction with related transaction data. We identified 7,537 SNAP recipient households in three selected states--Michigan, Massachusetts and Nebraska--that both received replacement cards in four or more monthly benefit periods in fiscal year 2012 and made transactions considered to be potential signs of trafficking. Furthermore, as discussed below, we developed an approach for analyzing replacement card data that may provide states with a more targeted way to identify potential trafficking activity and reduce the number of households for further review by up to 40 percent. Given that states reported having limited resources for conducting investigations, a more targeted approach may enhance their ability to pursue SNAP households at higher risk of trafficking. Overall, our approach to analyzing replacement card data reduced the number of households for further review by 33 percent compared to the current FNS regulation. For the purposes of our analysis, we defined excessive replacement card households as those receiving replacement cards in four or more unique benefit periods in a year. Our approach took into account FNS's rule that defines excessive replacement cards as at least four requested in a year. However, we further refined our analysis to consider the monthly benefit period of replacement card requests. SNAP benefits are allotted on a monthly basis, and a recipient who is selling the benefits on their EBT card and then requesting a replacement card would generally have only one opportunity per month to do so. If a SNAP recipient is requesting a replacement card because they have just sold their EBT card and its associated SNAP benefits, it is unlikely that there would be more benefits to sell until the next benefit period. As a result, additional replacement card requests in the same benefit period may not indicate increased risk of trafficking. The current FNS regulation would include households for review that received at least four replacement cards at any time in the previous year, including households receiving four cards in the same monthly benefit period. Alternatively, the number of benefit periods with replacement cards may be a better indicator of trafficking risk than simply the number of requested replacement cards. By taking into account the benefit period of replacement card requests, we significantly decreased the number of households in the three selected states that may warrant further review of potential trafficking compared to all households requesting four or more replacement cards at any time during fiscal year 2012. For example, as shown in table 3, while there were 8,190 recipient households in Michigan that received four or more replacement cards in fiscal year 2012, our approach identified 4,935 households that received replacement cards in four or more benefit periods. Table 3: SNAP Households Receiving Replacement Cards in Fiscal Year 2012: State: Michigan; Total SNAP households[A]: 924,643; Number of households receiving: 4+ Cards: 8,190; Cards in 4+ monthly benefit periods: 4,935; Percent decrease in households using benefit periods: 39.74%. State: Massachusetts; Total SNAP households[A]: 479,830; Number of households receiving: 4+ Cards: 6,380; Cards in 4+ monthly benefit periods: 4,786; Percent decrease in households using benefit periods: 24.98%. State: Nebraska; Total SNAP households[A]: 77,066; Number of households receiving: 4+ Cards: 697; Cards in 4+ monthly benefit periods: 549; Percent decrease in households using benefit periods: 21.23%. State: Total; Total SNAP households[A]: 1,481,539; Number of households receiving: 4+ Cards: 15,267; Cards in 4+ monthly benefit periods: 10,270; Percent decrease in households using benefit periods: 32.73%. Source: GAO analysis of Supplemental Nutrition Assistance Program (SNAP) transaction data. GAO-14-641. [A] Average monthly participating households in fiscal year 2012. [End of table] For the 10,266 high replacement card households we reviewed, we found that 73 percent were conducting other suspicious activities based on criteria used by FNS and state SNAP officials. [Footnote 25] We reviewed fiscal year 2012 transaction data, analyzing transactions from the same benefit period when the household received a replacement card for indications of trafficking. Specifically, we analyzed the data for trafficking indicators based on suspicious transaction types already used by FNS and state SNAP officials, such as unusually large- dollar transactions or even-dollar transactions. We tested the transaction data for six different suspicious transaction types, resulting in 22,866 transactions flagged as potential trafficking indicators.[Footnote 26] As shown in table 4, we identified 7,537 households out of those we reviewed that made at least one suspicious transaction in the same benefit period that the household received a replacement card in fiscal year 2012. These 7,537 households made over $26 million in purchases with SNAP benefits during fiscal year 2012. Table 4: SNAP Households Receiving Excessive Replacement Cards and Making Transactions Potentially Indicative of Trafficking in Fiscal Year 2012: State: Michigan; SNAP households with: 4+ Replacement cards: 8,190; Replacement cards in 4+ benefit periods: 4,935; Suspicious transactions and cards in 4+ benefit periods: 3,183. State: Massachusetts; SNAP households with: 4+ Replacement cards: 6,380; Replacement cards in 4+ benefit periods: 4,786; Suspicious transactions and cards in 4+ benefit periods: 4,008. State: Nebraska; SNAP households with: 4+ Replacement cards: 697; Replacement cards in 4+ benefit periods: 549; Suspicious transactions and cards in 4+ benefit periods: 346. State: Total; SNAP households with: 4+ Replacement cards: 15,267; Replacement cards in 4+ benefit periods: 10,270; Suspicious transactions and cards in 4+ benefit periods: 7,537. Source: GAO analysis of Supplemental Nutrition Assistance Program (SNAP) transaction data. GAO-14-641. [End of table] Overall, 84 percent of high replacement card households in Massachusetts, 65 percent in Michigan, and 63 percent in Nebraska made at least one suspicious transaction indicating potential trafficking. For more detailed information on the number of flagged transactions made by selected households in each of the three states, see appendix II. Furthermore, we found that the likelihood of suspicious transactions generally increased with the number of benefit periods in which replacement cards were requested. For example, while 60 percent of Michigan households with replacement cards in four benefit periods made at least one suspicious transaction, 86 percent of households with replacement cards in seven benefit periods had made suspicious transactions, and 100 percent of households with replacement cards in 10 or 11 benefit periods had. In Nebraska, 100 percent of households with replacement cards in eight or more benefit periods also made suspicious transactions, indicating potential trafficking. While 84 percent of households had five or fewer trafficking flags, there were 262 households, or 3 percent, with 10 or more trafficking flags. The highest number of flags for a single household was 41. This household's flagged transactions showed suspicious large, even-dollar transactions, often at the same small grocery store. Table 5 provides examples of suspicious transactions made by this household in one benefit period. Table 5: Examples of Selected Suspicious Transactions Made by One Household Resulting in Trafficking Flags, Fiscal Year 2012: Transaction type: Benefits Issued; Date: Jan. 14, 2012. Transaction type: Purchase 1; Store: Store 1 - Supermarket; Date: Jan. 14, 2012; Time: 5:17:22 PM; Amount: $130.07. Transaction type: Purchase 2; Store: Store 2 - Small Grocery; Date: Jan. 14, 2012; Time: 8:43:12 PM; Amount: $3.85; Time from previous transaction: 3 hours 26 minutes. Transaction type: Purchase 3; Store: Store 2 - Small Grocery; Date: Jan. 14, 2012; Time: 8:44:13 PM; Amount: $240.00; Time from previous transaction: 61 seconds. Transaction type: Purchase 4; Store: Store 2 - Small Grocery; Date: Jan. 14, 2012; Time: 8:44:39 PM; Amount: $100.00; Time from previous transaction: 26 seconds. Transaction type: Purchase 5; Store: Store 2 - Small Grocery; Date: Jan. 14, 2012; Time: 8:45:06 PM; Amount: $140.00; Time from previous transaction: 27 seconds. Transaction type: Replacement card issued; Date: Jan. 18, 2012; Time: 10:05:52 AM. Source: GAO analysis of Supplemental Nutrition Assistance Program (SNAP) transaction data. GAO-14-641. [End of table] By comparing the number of benefit periods with replacement cards and the total number of transaction trafficking flags, we were able to better identify those households that may be at higher risk of trafficking. For example, as shown in figure 2, while there were 4,935 SNAP households in Michigan that received excessive replacement cards, we identified just 39 households that received excessive replacement cards and made transactions resulting in 10 or more trafficking flags. While state SNAP officials may not want to limit their investigations to such a small number of households, this type of analysis may help provide a starting point for identifying higher priority households for further review. Figure 2: Targeting Potential Supplemental Nutrition Assistance Program (SNAP) Benefit Trafficking Using Replacement Card and Transaction Data to Identify Higher Risk Households in Michigan, Fiscal Year 2012: [Refer to PDF for image: illustration] Michigan SNAP households receiving replacement cards in FY12: By taking a targeted approach to analyzing replacement card data in conjunction with related transaction data, we identified those households receiving excessive replacement cards that may be at higher risk of trafficking. SNAP households receiving 4+ replacement cards in FY12: 8,190; SNAP households with replacement cards in 4+ monthly benefit periods: 4,935; SNAP households with replacement cards in 4+ benefit periods that also made suspicious transactions indicating potential trafficking: 3,183; Higher risk households with excessive replacement cards and suspicious transactions resulting in 10+ trafficking flags: 39. Source: GAO analysis of Supplemental Nutrition Assistance Program (SNAP) transaction data. GAO-14-641. Note: FY = fiscal year. [End of figure] Recognizing the challenges with the current approach, FNS officials stated that they are working on how to better link excessive replacement card requests to potential trafficking. To inform these efforts, FNS has also commissioned a study focused on detecting indications of potential trafficking by those requesting excessive replacement cards. FNS officials feel it is too early provide additional guidance or draw conclusions about the effectiveness of current efforts, but officials intend to provide additional guidance to states once they have sufficient data to inform a trafficking detection methodology that can be used nationwide. Officials from Selected States Reported Difficulties Using Social Media and E-commerce Website Monitoring Tools: FNS provided states with guidance on installing free web-based software tools for monitoring certain e-commerce and social media websites for online sales of SNAP benefits, but some state officials from selected states reported problems with these detection tools. The tools employ Really Simple Syndication (RSS) technology, which is designed to keep track of frequently-updated content from multiple websites and automatically notify users of postings that contain key words. FNS stated that these tools could automate the searches that states would normally have to perform manually on these websites, but acknowledged that the tool for social media websites may not work well, given that these websites do not organize their posts geographically.[Footnote 27] Of the 11 states we reviewed, officials from only one selected state (Tennessee) reported that the tool worked well for identifying SNAP recipients attempting to sell their SNAP benefits online. Officials in three states--Michigan, Utah, and Florida--reported that they monitored social media websites manually because of the technical challenges they experienced with using the tools, including installation and operation.[Footnote 28] Additionally, officials in one state noted that the automated tools have placed an excessive demand on staff because they had to sift through the many false-positive leads that were generated. Officials from three of the states we reviewed reported that although they do not routinely monitor websites to detect fraud, they have found these websites to be useful sources of information about recipients they are already investigating. FNS officials acknowledge that there are limitations to the current monitoring tools, and stated that they provided these tools at the request of states to help with monitoring efforts as states had reported that manual monitoring was cumbersome and difficult given limited resources. FNS officials report that they are currently conducting a study of the effectiveness of the guidance to states and intend to make recommendations for improvements based on the results of the study. In addition to the guidance provided to states, FNS officials reported that they have contacted popular e-commerce and social media websites in the past regarding potential SNAP trafficking online, and continue to work with the websites on detecting and removing postings advertising the sale of SNAP benefits online. Effectiveness of Current Automated Monitoring Tools for Recipient Fraud Detection by States is Limited: We tested the automated detection tools recommended by FNS on selected geographical locations covering our selected states and found them to be of limited effectiveness for states' fraud detection efforts. A crucial element to an effective fraud prevention framework requires resources and tools to continually monitor and detect potential fraud. [Footnote 29] Our testing of the recommended automated tool for monitoring e-commerce websites found it did not detect most of the postings found through manual website searches. Furthermore, we found the automated tool for monitoring social media websites to be impractical for states' fraud detection efforts. E-commerce Monitoring: Although the recommended automated tool for monitoring e-commerce websites was intended to potentially replace the need for states to perform manual searches on these websites, our testing found that manual searches returned more postings indicative of potential SNAP trafficking than the automated tool, and that most of the postings detected through manual searches were not detected by the automated tool. We tested the recommended tool on one popular e-commerce website over 30 days, and monitored 19 geographical locations covering the 11 selected states.[Footnote 30] We spent an average time of about 30 minutes per day (10 hours total) monitoring for postings indicative of potential SNAP trafficking.[Footnote 31] Through our manual and automated searches, we detected a total of 1,185 postings containing one of our two key words of "EBT" or "food stamps." Out of these 1,185 postings, we detected 28 postings indicative of potential SNAP trafficking. They advertised the potential sale of food stamp benefits in exchange for cash, services, or goods (see figure 4). [Footnote 32] We refer to these types of postings as true positives, and to those postings that did not indicate trafficking as false positives. (See figure 3.) Figure 3: Analysis of Automated Tool Recommended for Monitoring E- commerce Websites: [Refer to PDF for image: illustrated flow chart] E-commerce posts identified and reviewed through automated and manual search tools: For 30 days, compared automated with manual search results from a popular e-commerce website; Based on searches using 2 key words, “EBT[A]” and “food stamps;” Monitored 19 selected geographical locations on popular e-commerce website, covering 11 selected states; Spent about 30 minutes a day, on average; Out of 1,180 posts reviewed, we detected 28 true positives, of which 21 were missed by the automated tool. True positives: Posts indicative of trafficking: November 29, 2013: “iPhone 5 black cracked screen works fine...purchase date of July 2013 so there is still Apple warranty… $350 or best offer...Trades for...EBT card...” December 2, 2013: “$65 food stamps for $30 cash: Really need cash money for the other half of my phone bill ...let you use my card and give the pin number...” December 18, 2013: “Interesting proposal: I am looking for a place to sleep ...if you can help me out (Oh, I May have a some cash to contribute as well as food stamps ...)” False positives: Posts not indicative of trafficking: November 21, 2013: “Large Banner: We Accept EBT Food Stamps 3x8 ft” November 30, 2013: “Huge savings on groceries! EBT accepted” November 30, 2013: “EBT isn't free--Sweatshirt” Source: GAO analysis of e-commerce posts. GAO-14-641. [A] States issue SNAP benefits through the use of Electronic Benefits Transfer (EBT) cards. [End of figure] Figure 4: Results of 30-day Monitoring of One Popular E-Commerce Website Using Automated and Manual Tools to Detect Posts Indicative of Food Stamp Trafficking and States' Experiences with Using Tools: [Refer to PDF for image: interactive graphic] Instructions: Rollover the locations GAO monitored to view selected states' online monitoring efforts. Print Version: Printable version of Figure 4 is available in Appendix IV. Source: GAO analysis of e-commerce posts identified and related state efforts. GAO-14-641. [End of figure] Side bar: RSS technology uses RSS readers that aggregate content from a website based on key-word queries, transmitting the content through RSS feeds. An RSS reader can be supported by an email system, like the RSS reader FNS recommended to the states. RSS readers automatically check for updated content and display results automatically as RSS feeds. For example, once the RSS reader finds a relevant posting, it will automatically send an RSS feed along with a link to the posting. The RSS feed is similar in appearance to a new email message. However, RSS readers use algorithms to automatically poll websites for updated postings pursuant to a specified maximum polling frequency set by the websites being polled. Source: GAO summary of literature review. [End of side bar] Overall, 21 of the 28 true positive postings were only detected through manual monitoring and did not appear in our RSS feeds during the 30-day testing period, potentially due to limitations associated with the RSS technology. Specifically, 10 of these 21 postings were listed on the e-commerce website under their respective geographical locations as "local" results. The remaining 11 postings were listed under their geographical locations, but as "nearby" results, and six of these postings were located in states other than the 11 states we monitored.[Footnote 33] According to the company that designed the automated tool, the RSS feeds do not currently transmit the "nearby" postings that would normally be found in manual searches. This may limit the number of potentially relevant postings that would be detected by the recommended tool and transmitted to SNAP officials for review. Additionally, the 10 manually detected postings listed as "local" results were not detected by the automated tool due to potential technical limitations associated with RSS technology. The FNS-recommended RSS reader we used on one popular e-commerce website was set to poll the website at most once per hour. According to the company that designed the recommended RSS reader, websites strictly enforce these frequency limits to help manage the demand on their servers. Setting such time intervals may prevent postings from appearing as RSS feeds in the user's RSS reader because such postings may be removed by the time the polling actually occurs. In addition, sometimes a website limits the amount of content transmitted through RSS feeds or the RSS capability can become inoperable or delayed, potentially causing the RSS reader to not detect a posting. Given these technical limitations and the results of our testing, we found that manual searches performed directly on the website were more effective at detecting postings indicative of SNAP trafficking. Accordingly, relying on the recommended automated tool may increase the risk of states missing opportunities to detect and deter individuals who are using the popular e-commerce website to facilitate SNAP trafficking. The 21 postings detected manually were enumerated under the selected geographic locations for 5 of 11 selected states: Massachusetts, Texas, North Carolina, Florida, and New Jersey.[Footnote 34] During our 30-day testing period, we did not detect any true positive postings for the selected geographic locations enumerated under the remaining six selected states: Michigan, Nebraska, Utah, Tennessee, Maine, or Wyoming. Although the automated tool for e-commerce websites delivered a total of seven postings indicative of potential SNAP trafficking during our 30-day testing period, the manual searches detected five of these seven postings. Below are three illustrative examples of the e-commerce postings that were detected through our manual searches of one popular e-commerce website. All three examples were identified solely through our manual searches. The automated tool did not detect these posts during our 30- day testing period. * On December 3, 2013, we detected an e-commerce posting advertising the potential sale of $400 in food stamp benefits in exchange for $240 in cash. This posting included the seller's telephone number. * On December 13, 2013, we detected an e-commerce posting advertising the potential sale of food stamp benefits in exchange for a place to live. This posting stated that the seller was "a single mother needing a room and...a place to live." The posting contained the seller's name and telephone number. * On December 18, 2013, we detected an e-commerce posting potentially advertising the sale of artwork valued up to $3,000 in exchange for food stamps. The posting read "Art for food. I am a local artist… looking to trade for EBT...my work ranges from $10 to $3000, so I'm open to a lot of offers." The posting provided the seller's contact information, including a website address to the artist's website. Social Media Monitoring: We also monitored one popular social media website for potential SNAP trafficking using the automated tool recommended by FNS, but found the tool to be inefficient for states' fraud detection efforts. For example, during our testing we found that the tool for monitoring social media websites had changed since FNS issued its guidance and no longer allowed a user to tailor the automated results to a specific social media website. Instead, we found that the automated tool would poll from more than a dozen websites, including social media and news websites, returning thousands of automated results not necessarily indicative of potential SNAP trafficking. We also found that the automated tool could not be tailored to a specific geographical location, potentially limiting a state's ability to effectively determine whether the postings detected are relevant to the state's jurisdiction. FNS officials are aware of this limitation, but still believe that the automated tool can help states detect social media postings indicative of SNAP trafficking. Further, we were unable to compare the automated tool to corresponding manual searches because, at the time of our testing, the popular social media website we chose to monitor did not support manual searches based on key words. Because of these technological limitations and the high volume of irrelevant postings delivered by the tool, we limited our testing of the automated tool to 5 days. Over 5 days using the automated tool to monitor one social media website, we reviewed a total of 3,367 social media postings; each posting contained one of our two key words ("EBT" and "food stamps"). [Footnote 35] We spent an average of 17 minutes per day (1 hour and 25 minutes total) reviewing the automated search results. However, we only detected four true positive postings that were potentially indicative of SNAP trafficking, and one of those four did not include potential location information.[Footnote 36] Although we were unable to compare the automated tool to manual approaches, we still found the automated tool for social media websites to be an impractical fraud detection tool for states, given the inability to limit monitoring to geographic areas within a state's jurisdiction and the inability to exclude websites, such as news websites, likely irrelevant to fraud detection activities. FNS Increased Its Oversight of State Anti-Fraud Activities but Lacks Reliable Data on These Efforts: FNS has recently issued regulations and guidance and conducted a national review of state anti-fraud activities as part of its increased oversight. Despite these efforts, FNS does not have consistent and reliable data on state anti-fraud activities, primarily because its guidance to the states on what data to report is unclear. FNS Issued Regulations and Guidance and Conducted a Nationwide Review as a Part of Its Increased Oversight: Since 2011, FNS increased its anti-fraud oversight activities, which included new regulations and guidance and a nationwide review of state agencies. Partially in response to public concerns, the Secretary for Food, Nutrition, and Consumer Services asked states to renew their efforts to combat SNAP recipient fraud, and since then FNS promulgated new regulations and provided additional guidance to direct states in these efforts.[Footnote 37] (See table 6 for details on key regulations, guidance and policy developments since 2011.) Table 6: Key Food and Nutrition Service (FNS) Oversight Regulations, Guidance and Policies, Fiscal Years 2011 through 2014: 2011: FNS issued guidance that encouraged states to use replacement card data. FNS encouraged states to use the Anti-Fraud Locator Using Electronic Benefits Transfer Retailer Transactions (ALERT)[A] data developed during retailer investigations to pursue individuals who trafficked with those retailers. 2012: FNS promulgated a final rule that implemented the requirement that states conduct death match verifications[B]. FNS issued guidance that encouraged states to use electronic benefit transfer (EBT) management reports[C]. FNS emphasized the importance of having states work with their contractors that manage EBT transaction data to develop effective methods for fraud detection. FNS issued guidance for detecting postings indicative of fraud on e- commerce sites. 2013: FNS issued guidance for detecting postings indicative of fraud on social media websites. FNS promulgated a final rule that included a new trafficking definition. This new definition included the following activities: the purchase with Supplemental Nutrition Assistance Program (SNAP) benefits of a product that has a container requiring a return deposit with the purpose of discarding the product and returning the container for cash and subsequently doing so; the purchase of a product with SNAP benefits with the intent of obtaining cash or other non-eligible items by reselling the product, and subsequently doing so; intentionally purchasing products originally purchased with SNAP benefits in exchange for cash or other non-eligible items; stealing SNAP benefits[D]. FNS promulgated an interim final rule that required states to monitor replacement card data[E]. 2014: FNS affirmed the interim final rule that required states to monitor replacement card data as a final rule[F]. Source: GAO Analysis of Federal Register notices, FNS guidance and policies. GAO-14-641. [A] The ALERT system receives daily transaction records from EBT processors and conducts analysis of patterns in the data, which indicate potential fraudulent activity by stores. [B] 77 Fed. Reg. 48,045 (Aug. 13, 2012). [C] The EBT reports provide each state with information on its EBT system, retailer transaction data, and the amount of SNAP benefits issued. [D] 78 Fed. Reg. 11,967 (Feb. 21, 2013). [E] 78 Fed. Reg. 51,649 (Aug. 21, 2013). [F] 79 Fed. Reg. 22,766 (Apr. 24, 2014). [End of table] In fiscal year 2013, for the first time, FNS examined states' compliance with federal requirements governing SNAP anti-fraud activities through Recipient Integrity Reviews. (See fig. 5 for an overview of the review components.) These assessments were conducted by FNS regional office staff and included interviews with state officials, observations of state hearing proceedings, and case file reviews in all 50 states and the District of Columbia. As part of these reviews, federal officials also analyzed information from program reports, including those from eDRS, which are used to track disqualified SNAP recipients, and the Program and Budget Summary (Form FNS-366B), which are used to report anti-fraud activities for all the states. Following these reviews, FNS regional officials issued state reports that included findings and, where appropriate, required corrective actions. FNS officials told us that timeframes for taking corrective actions varied by the problem, but they generally allow states a year to address them. FNS regional officials also acknowledged states' noteworthy initiatives or best practices in the state reports - such as Michigan's case management system which will improve the state's ability to track the status and outcomes of investigations, Washington's standardized training for investigators, and Indiana's out-of-state usage report aimed at identifying potential trafficking by listing households that made 100 percent of their EBT transactions in another state for three months. FNS officials also reported that they provide their regional staff the opportunity to discuss such best practices during monthly teleconferences. Additionally, FNS officials present information on best practices to states during national conferences. FNS began conducting fiscal year 2014 Recipient Integrity Reviews in November 2013 and intends to complete them in September 2014. Figure 5: FNS Recipient Integrity Review Components: [Refer to PDF for image: illustration] Investigations: FNS regional officials interview state officials about their investigations process and determine whether the states comply with federal regulations. Hearings: FNS regional officials assess state documentation, observe hearings, and complete case file reviews. Pursuing claims: FNS regional officials conducta case file review, assess state documentation, and observe interviews with fraud suspects. Reporting: FNS regional officials compare the state's reports with supporting documentation. Source: GAO analysis of Food and Nutrition Service (FNS) guidance. GAO- 14-641. [End of figure] In addition to its oversight efforts, FNS has 10 studies under way that are aimed at improving federal and state efforts to address potential recipient fraud. These studies represent a significant increase in its investment to learn more about recipient fraud; specifically, FNS designated about $3 million[Footnote 38] for this work in fiscal years 2013 and 2014, compared to none in prior years. Among other topics, these studies are to explore strategies for improving fraud detection. For example, the study titled Social Media Fraud Discovery is intended to assess the effectiveness of FNS's current fraud detection approach and make recommendations for improvements. There is also a series of work, known as the SNAP Recipient & Retailer Fraud Data Mining Studies, aimed at improving FNS and the states' ability to more effectively anticipate, discover and address fraudulent activity using predictive modeling. FNS expects to receive the results of these studies by September 2014. (Additional information on the 10 studies is provided in Appendix V.) FNS Lacks Reliable Data on State Anti-Fraud Activities: Although states are required to regularly submit information on their anti-fraud activities to FNS, we found that these data are not reliable for ensuring program integrity and assessing states' performance. Specifically, our review found that over half of the 2013 Recipient Integrity Review reports mentioned problems with the data states entered into eDRS, thereby affecting the information state officials used to ensure program integrity. Federal officials found that 30 states did not enter data within the federally-required timeframes, a problem that cut across each of the oversight regions. [Footnote 39] Federal officials also found that 15 states did not enter disqualification information for some cases at all. For 2 of these states, federal officials found that over 30 percent of the disqualifications mentioned in other federal reports were missing from their eDRS data. Furthermore, federal officials found that 10 states had entered data into the system inaccurately. Given the concerns with data quality, even though state officials are required to check eDRS to gather information on whether a program applicant has been disqualified in another state before issuing benefits, they are not allowed to deny an application based solely on the system's data. Federal regulations require that states gather additional verifying information about a disqualification before denying a claim based solely on information from eDRS.[Footnote 40] FNS regional officials told us that state's eDRS data problems stemmed from a variety of factors, including challenges with receiving timely information about administrative hearing and court decisions and transferring data to the system. To help address concerns with eDRS data quality, FNS officials are currently offering tools, guidance, and training to state and regional officials. Furthermore, states with related findings from the Recipient Integrity Reviews are expected to take corrective action, including improving communications with ADH and court officials to receive more timely information and enhance their procedures for validating data entered into the system. Through our review of the 2013 Recipient Integrity Review reports, we also found that FNS has a nationwide problem with receiving inaccurate data on state anti-fraud activities through the Program and Budget Summary Statement (Form FNS-366B),[Footnote 41] thereby potentially limiting its ability to provide oversight. We found that FNS regional officials could not reconcile the FNS-366B data reported with supporting documentation for 24 states, primarily due to data entry errors. Furthermore, some federal and state officials we interviewed recognized that there is not a consistent understanding of what should be reported on FNS-366B because the guidance from FNS is unclear. For example, on the form, FNS instructs states to report investigations for any case in which there is suspicion of an intentional program violation before and after eligibility determination. According to state and federal officials we interviewed, this information does not clearly establish a definition for what action constitutes an investigation and should then be reported on this form. Also, officials in three of the seven regional offices were not aware of FNS- sponsored training on what should be reported on this form.[Footnote 42] However, officials from the remaining four offices mentioned that FNS provided them training such as webinars and teleconferences on this form. As a result, various types of state efforts can be counted in the total number of investigations. After reviewing states' reports, we found examples of inconsistencies in what states reported as investigations on the FNS-366B. Specifically, in fiscal year 2009, one state had about 40,000 recipient households, but reported about 50,000 investigations. During the same year, another state that provided benefits to a significantly larger population (about 1 million recipient households) reported about 43,000 investigations. Officials from the state that serves a smaller population explained that they included activities such as manually reviewing paper files provided by the state's Department of Labor for each SNAP recipient with reported wages in the state; therefore, even if fraud was not suspected, this review was counted as an investigation. Officials from the state that serves a larger population said that they counted the number of times a potential fraud case was actively reviewed by investigators, including interviews with witnesses and researching of related client information. Given these differences, state officials said that FNS and states are not able to compare program integrity performance, because each state is not counting the same activities. In addition, by fiscal year 2012, the new head official in the state that serves a smaller population decided to use an automated system to review the wage data. Therefore, the query results identified cases indicative of a benefit overpayment, either from potential fraud or unintentional errors, were counted among the cases that needed to be investigated. As a result, for fiscal 2012, the state that serves a smaller population only reported conducting about 8,000 investigations, making this count of investigations not comparable to others for that state over time. Furthermore, these data inconsistencies could limit in FNS's ability to identify more effective and efficient practices for state anti-fraud efforts. For example, the lack of consistent data on investigations does not allow for studying the matters, such as the cost-benefit of investigations versus fraud claims established and/or collected across states, which could be of interest to FNS and states given states' concerns with managing investigative resources. Conclusions: Given the ongoing fiscal pressures that face our nation, the unprecedented increase in SNAP participation and spending in recent years has focused attention on the importance of ensuring that these publicly-funded benefits are used appropriately, and that both the federal government and state agencies have strong controls for detecting and addressing fraud. Although investigations can ultimately deter fraud and save agency resources, states we reviewed have faced the challenge of limited staff to manage a growing program and raised questions about whether federal incentive structures could be designed to better support their work. For example, even though GAO has found preventative efforts to be the most efficient and effective means to address fraud and may stop ineligible people from receiving benefits that may not be fully recovered, state officials said the current fraud-related incentive is focused on collecting overpayments. While federal officials would need to be mindful of the costs and benefits that any changes to the incentive structure would have for the overall program, absent additional incentives, states may not be taking advantage of opportunities to aggressively pursue recipient fraud. These investigative challenges have made efficient anti-fraud activities all the more critical. Although some states have questioned the efficacy of tools FNS requires or recommends for detecting SNAP benefit trafficking, some additions and refinements to the guidance for these tools could make them more effective. For example, a more targeted approach to reviewing requests for replacement benefit cards could substantially reduce the administrative burden by identifying recipients who are more likely to be misusing their benefits throughout the year. Furthermore, although FNS has tried to improve efficiency with monitoring online postings, the lack of relevant leads using the recommended tools cause others to question whether this monitoring could be done in a better way. Meanwhile, FNS is working to learn more about states' activities and better support anti-fraud work. For example, FNS has commissioned 10 studies intended to help the agency gain knowledge on how states can better detect potential recipient fraud. However, absent additional actions from FNS, such as guidance and training to the states on how and what data to report on their fraud-related activities, these data are not likely to be as useful as they should. Specifically, without performance data that are consistent across states, FNS will not be able to determine whether certain state anti-fraud efforts may be more efficient and effective than others. FNS will need accurate and comprehensive information at the state level if it is to move forward in building a stronger national infrastructure for program integrity. Recommendations for Executive Action: The Secretary of Agriculture should direct the Administrator of FNS to take the following four actions: * Explore ways that federal financial incentives can better support cost-effective state anti-fraud activities; * Establish additional guidance to help states analyze SNAP transaction data to better identify SNAP recipient households receiving replacement cards that are potentially engaging in trafficking, and assess whether the use of replacement card benefit periods may better focus this analysis on high-risk households potentially engaged in trafficking; * Reassess the effectiveness of the current guidance and tools recommended to states for monitoring e-commerce and social media websites, and use this information to enhance the effectiveness of the current guidance and tools; and: * Take steps, such as guidance and training, to enhance the consistency of what states report on their anti-fraud activities. Agency Comments and Our Evaluation: We requested comments on a draft of this product from USDA. On July 28, 2014, the Director of the SNAP Program Accountability and Administration Division provided us with the following oral comments. FNS agreed with our recommendations and reported that efforts were underway to address each of them. Specifically, FNS reported that, although the agency cannot change the state retention rate for overpayments without a change to federal laws, it plans to issue grants in this fiscal year to support state process improvements for detecting, investigating and prosecuting recipients engaged in trafficking. Furthermore, in the next fiscal year, FNS reported that it will issue grants to support states in building information technology to strengthen recipient integrity efforts, as authorized by the Agricultural Act of 2014. FNS also reported that its commissioned studies will help inform its efforts to assist states in developing better recipient fraud detection tools, including potentially issuing new related guidance. As of May 2014, the agency had already begun to receive study results. Lastly, in May 2014, FNS also formed a working group, consisting of program integrity staff from each of the regional offices, to revamp the Form FNS-366B. Among other things, FNS reported that this group is tasked with exploring ways to clearly define the data elements on this form and adding elements that will help FNS glean better information on recipient trafficking as well as the value and impact of state anti-fraud efforts. FNS also provided technical comments, which were incorporated into the report as appropriate. We are sending copies of this report to relevant congressional committees, the Secretary of Agriculture, the FNS Administrator and other relevant parties. This report will also be available at no charge on the GAO website at [hyperlink, http://www.gao.gov. If you or your staff have any questions about this report, please contact us at (202) 512-7215 or brownke@gao.gov, or (202) 512-6722 or bagdoyans@gao.gov . Contact points for our Offices of Congressional Relations and Public Affairs may be found on the last page of this report. GAO staff who made key contributions to this report are listed in appendix VI. Sincerely yours, Signed by: Kay E. Brown: Director: Education, Workforce and Income Security Issues: Signed by: Seto J. Bagdoyan: Acting Director: Forensic Audits and Investigative Service: [End of section] Appendix I: Objectives, Scope and Methodology: The objectives of this report were to review the following: (1) how selected state agencies combat SNAP recipient fraud; (2) the effectiveness of certain fraud detection tools recommended to states, including benefit card replacement data and e-commerce and social media website monitoring; and (3) FNS's oversight of state anti-fraud efforts. To address these objectives, we focused on federal and state SNAP recipient anti-fraud work for fiscal years 2009 to 2014, a period after the program received additional funding through the American Recovery and Reinvestment Act of 2009 (Recovery Act).[Footnote 43] We reviewed relevant federal laws, regulations, program guidance, and reports, and we interviewed FNS officials in headquarters and all seven regional offices to address all three objectives. Specifically, to determine how selected state agencies are pursuing SNAP recipient fraud, we reviewed 11 states, where we interviewed knowledgeable state and local officials about their recipient anti-fraud work and obtained related documentation. (See below for more information on the criteria we used to select states.) We also analyzed fiscal year 2012 replacement card and transaction data for households in three of the selected states to assess the extent to which certain analyses of replacement cards could better uncover patterns of potential fraud. (See below for more information about these tests and analyses.) Further, we tested automated tools and guidance that FNS recommended to states for monitoring popular e-commerce and social media websites for postings indicative of SNAP trafficking. Our test involved determining the extent to which our 11 selected states can use these tools for their fraud detection efforts. Lastly, to determine FNS's oversight of state anti-fraud efforts, we analyzed documents and reports relevant to FNS's program oversight, including their fiscal year 2013 assessments of state anti-fraud work--known as Recipient Integrity Review reports--for all 50 states and the District of Columbia. All the data included in this report were assessed and determined to be sufficiently reliable for our purposes. We conducted this performance audit from April 2013 through September 2014 in accordance with generally accepted government auditing standards. Those standards require that we plan and perform the audit to obtain sufficient, appropriate evidence to provide a reasonable basis for our findings and conclusions based on our audit objectives. We believe that the evidence we obtained provides a reasonable basis for our findings and conclusions based on our audit objectives. State Selection Criteria: To determine how selected state agencies are pursuing SNAP recipient fraud, we selected 11 states for our review--Florida, Maine, Massachusetts, Michigan, Nebraska, New Jersey, North Carolina, Utah, Tennessee, Texas, and Wyoming--based on geographic dispersion, SNAP payment error rates, percent of the total number of SNAP households nationwide, and the percent of recipients they reported as disqualified from the program due to non-compliance. For three of these criteria--the percent of the total number of households, the percent of the total number of disqualifications, and the payment error rates--we assigned the states to high, medium, and low categories under each set of data based on natural breaks in the data when the states were ranked from the lowest to the highest percent. As a result, states were designated based on data ranges shown in table 7. Table 7: Data Ranges for Percentages: Percentage: Percent of the total SNAP households; Low: less than .5%; Medium: .5 to less than 3%; High: 3% and higher. Percentage: Percent of the total number of disqualifications; Low: less than .5%; Medium: .5 to less than 2.5%; High: 2.5% and higher. Percentage: Payment error rates; Low: less than 3.82%; Medium: 3.82 to less than 5%; High: 5% and higher. Source: GAO of analysis of FNS data. GAO-14-641. [End of table] We selected the states to review for variety within our criteria. Table 8 provides information on how the selected states align with our criteria. Table 8: State Selection Criteria: State: Wyoming; Geographic region: West; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 9.63%; [yellow--high] Percent of SNAP households nationwide in fiscal year 2012[B]: 0.07%; [blue--low] Percent of recipients disqualified in fiscal year 2012[C]: 0.07%. [blue--low] State: Utah; Geographic region: West; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 4.19%; [green--medium] Percent of SNAP households nationwide in fiscal year 2012[B]: 0.51%; [green--medium] Percent of recipients disqualified in fiscal year 2012[C]: 1.13%. [green--medium] State: Nebraska; Geographic region: Midwest; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 4.5; [green--medium] Percent of SNAP households nationwide in fiscal year 2012[B]: 0.35%; [blue--low] Percent of recipients disqualified in fiscal year 2012[C]: 0.19%. [blue--low] State: Michigan; Geographic region: Midwest; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 3.12%; [blue--low] Percent of SNAP households nationwide in fiscal year 2012[B]: 4.14%; [yellow--high] Percent of recipients disqualified in fiscal year 2012[C]: 5.28%. [yellow--high] State: Maine; Geographic region: Northeast; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 3.28%; [blue--low] Percent of SNAP households nationwide in fiscal year 2012[B]: 0.59%; [blue--low] Percent of recipients disqualified in fiscal year 2012[C]: 0.12%. [blue--low] State: Massachusetts; Geographic region: Northeast; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 4.40%; [green--medium] Percent of SNAP households nationwide in fiscal year 2012[B]: 2.15%; [green--medium] Percent of recipients disqualified in fiscal year 2012[C]: 1.25%. [green--medium] State: New Jersey; Geographic region: Northeast; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 4.33%; [green--medium] Percent of SNAP households nationwide in fiscal year 2012[B]: 1.82%; [green--medium] Percent of recipients disqualified in fiscal year 2012[C]: 2.05%. [green--medium] State: Tennessee; Geographic region: South; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 5.46%; [yellow--high] Percent of SNAP households nationwide in fiscal year 2012[B]: 2.87%; [green--medium] Percent of recipients disqualified in fiscal year 2012[C]: 3.29%. [yellow--high] State: North Carolina; Geographic region: South; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 2.65%; [blue--low] Percent of SNAP households nationwide in fiscal year 2012[B]: 3.52%; [yellow--high] Percent of recipients disqualified in fiscal year 2012[C]: 3.98%. [yellow--high] State: Florida; Geographic region: South; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 0.87%; [blue--low] Percent of SNAP households nationwide in fiscal year 2012[B]: 8.18%; [yellow--high] Percent of recipients disqualified in fiscal year 2012[C]: 11.59%. [yellow--high] State: Texas; Geographic region: South; Fiscal year 2011 SNAP payment error rates (in percent)[A]: 3.48%; [blue--low] Percent of SNAP households nationwide in fiscal year 2012[B]: 7.46%; [yellow--high] Percent of recipients disqualified in fiscal year 2012[C]: 12.98%. [yellow--high] Source: GAO analysis of FNS and the United States Census Bureau data. GAO-14-641. Note: The low range of percentages is indicated in blue, the medium is green, and the high is yellow. [A] For the payment error rates, ranges were: less than 3.82%=low, 3.82 to less than 5%=medium, and 5% and higher=high. [B] For the percent of the total SNAP households, ranges were: less than .5%=low, .5 to less than 3%=medium, and 3% and higher=high. [C] For the percent of the total number of disqualifications, ranges were: less than .5%=low, .5 to less than 2.5%=medium, and 2.5% and higher=high. [D] The low range of percentages is indicated in blue, the medium is green, and the high is yellow. [End of table] We interviewed officials who oversee state activities in state fraud units in each of the 11 states. In some states, we also interviewed auditors and prosecutors who had knowledge of state activities. During each interview, we collected information on state policies and procedures for responding to and investigating fraud claims. We also gathered and reviewed information on how state authorities manage their investigations. We also discussed state anti-fraud efforts and common recipient fraud schemes that have been occurring in recent years. We conducted site visits in Michigan, North Carolina, and Florida and interviewed officials in the remaining eight states by telephone. The information we gathered for our report represents the conditions present at the time of the review. We cannot comment on any changes that may have occurred after our fieldwork was completed. Although the 11 states we reviewed administered SNAP for about one-third of the program's recipient households, the information we report from these states is not generalizable. Targeted Analysis of Excessive Replacement Card Data: To assess the effectiveness of replacement card data as a state fraud detection tool, we analyzed replacement card data for SNAP households in 3 of the 11 selected states--Michigan, Massachusetts, and Nebraska. We selected these states to include high, medium, and low percentage of the total number of SNAP households nationwide. We obtained replacement card data from the appropriate state agency overseeing SNAP in the three selected states, and analyzed fiscal year 2012 data to determine the number of households receiving four or more replacement cards in that year. We also analyzed the data to identify households receiving replacement cards in four or more monthly benefit periods, the approach we took to identifying households with excessive replacement cards. We then obtained fiscal year 2012 transaction data from FNS for those households that received excessive replacement cards. We analyzed the transaction data for suspicious transactions indicating potential trafficking that occurred during the same benefit period when a household received a replacement card. We tested the transaction data for six different suspicious transaction types that were reported to us as commonly used by FNS and state SNAP officials to identify potential trafficking. At the request of SNAP officials to maintain confidentiality over their fraud detection methods, we did not include descriptions of all six transaction tests in the report. We assessed the reliability of replacement card and transaction data used in analyses through review of related literature, interviews with knowledgeable officials, and electronic testing of the data, and found them to be sufficiently reliable for our purposes. Assessment of Tools Recommended by FNS for Monitoring Online SNAP Trafficking: We installed and used the automated tools recommended by FNS pursuant to the guidance FNS released to the states for monitoring popular e- commerce and social media websites for postings indicative of SNAP trafficking. We tested the automated tools and guidance to determine the extent to which our 11 selected states can use these tools for their fraud detection efforts. We also used GAO's Fraud Prevention Framework[Footnote 44] to assess the automated tools and guidance. Specifically, from November 22, 2013 to December 23, 2013, we spent 30 days testing the automated tool for monitoring e-commerce websites on one popular e-commerce website, comparing our automated search results against our manual search results from the same e-commerce website. Our automated and manual search queries were set to detect postings containing one of the key words "EBT" or "food stamps." Using both search approaches simultaneously, we monitored 19 selected geographic locations covering 11 selected states, and spent an average time of about 30 minutes a day monitoring for e-commerce postings indicative of potential SNAP trafficking. Then we compared our automated results with our manual results to determine the extent to which they were the same. We selected the 19 geographic locations to monitor (see table 9, below) to include the two highest population cities in each of the 11 states. For two states--Maine and Wyoming--the e-commerce website only allowed us to monitor postings statewide. Table 9: Geographic Locations Monitored in 11 Selected States: 11 selected states: Nebraska; Geographic Areas Monitored: Lincoln/Omaha. Tennessee; Geographic Areas Monitored: Memphis/Nashville. Texas; Geographic Areas Monitored: Houston/San Antonio. Utah; Geographic Areas Monitored: Salt Lake City/Provo. North Carolina; Geographic Areas Monitored: Raleigh/Charlotte. Massachusetts; Geographic Areas Monitored: Boston/Worcester. Michigan; Geographic Areas Monitored: Detroit/Grand Rapids. Florida; Geographic Areas Monitored: Jacksonville/Southern Florida. Maine; Geographic Areas Monitored: Maine. Wyoming; Geographic Areas Monitored: Wyoming. New Jersey; Geographic Areas Monitored: Northern New Jersey. Source: GAO analysis of United States Census Bureau data and e-commerce website geographic areas. GAO-14-641. [End of table] Additionally, from January 7, 2014 to January 13, 2014, we spent 5 days testing the automated tool and guidance that FNS recommended to states for monitoring social media websites on a popular social media using the same key words ("food stamps" and "EBT"). We spent an average time of about 17 minutes a day monitoring for social media postings indicative of potential SNAP trafficking. We were unable to compare the automated tool for social media websites to corresponding manual searches because, at the time of our testing, the popular social media website did not offer a capability to perform manual searches based key words, such as "EBT" and "food stamps." [End of section] Appendix II: Trafficking Flags in SNAP Households Receiving Excessive Replacement Cards: As discussed above, we analyzed transaction data for households enrolled in the Supplemental Nutrition Assistance Program (SNAP) who received excessive replacement cards[Footnote 45] in three selected states--Michigan, Massachusetts and Nebraska--in fiscal year 2012. We tested the transaction data for six different suspicious transaction types, potentially indicative of trafficking. Tables 10 through 13, below, provide detailed information on the findings of these tests. Table 10: Number of Trafficking Flags Associated with Fiscal Year 2012 Transactions Made by Selected High Replacement Card Households in Three States: State: Michigan; Flag 1: 1,567; Flag 2: 347; Flag 3: 2,077; Flag 4: 2,560; Flag 5: 51; Flag 6: 2,426. State: Massachusetts; Flag 1: 2,397; Flag 2: 1,530; Flag 3: 3,007; Flag 4: 3,002; Flag 5: 153; Flag 6: 5,261. State: Nebraska; Flag 1: 209; Flag 2: 37; Flag 3: 269; Flag 4: 177; Flag 5: 6; Flag 6: 274. Source: GAO analysis of Supplemental Nutrition Assistance Program (SNAP) transaction data. GAO-14-641. [End of table] Table 11: Number of Michigan High Replacement Card Households Categorized by Count of Trafficking Flags and Replacement Card Benefit Periods , Fiscal Year 2012: Number of benefit periods with replacement cards: 4; Total Trafficking Flags: 1-3: 1532; 4-6: 391; 7-9: 68; 10-19: 14; 20+: 0; Total Total: 2005; Percent of Households with 1+ Trafficking Flag: 60%. Number of benefit periods with replacement cards: 5; Total Trafficking Flags: 1-3: 499; 4-6: 167; 7-9: 31; 10-19: 7; 20+: 0; Total: 704; Percent of Households with 1+ Trafficking Flag: 70%. Number of benefit periods with replacement cards: 6; Total Trafficking Flags: 1-3: 183; 4-6: 68; 7-9: 28; 10-19: 6; 20+: 2; Total: 287; Percent of Households with 1+ Trafficking Flag: 79%. Number of benefit periods with replacement cards: 7; Total Trafficking Flags: 1-3: 67; 4-6: 38; 7-9: 10; 10-19: 5; 20+: 0; Total: 120; Percent of Households with 1+ Trafficking Flag: 86%. Number of benefit periods with replacement cards: 8; Total Trafficking Flags: 1-3: 24; 4-6: 10; 7-9: 9; 10-19: 4; 20+: 0; Total: 47; Percent of Households with 1+ Trafficking Flag: 90%. Number of benefit periods with replacement cards: 9; Total Trafficking Flags: 1-3: 5; 4-6: 6; 7-9: 2; 10-19: 1; 20+: 0; Total: 14; Percent of Households with 1+ Trafficking Flag: 70%. Number of benefit periods with replacement cards: 10; Total Trafficking Flags: 1-3: 3; 4-6: 0; 7-9: 1; 10-19: 0; 20+: 0; Total: 4; Percent of Households with 1+ Trafficking Flag: 100%. Number of benefit periods with replacement cards: 11; Total Trafficking Flags: 1-3: 0; 4-6: 1; 7-9: 1; 10-19: 0; 20+: 0; Total: 2; Percent of Households with 1+ Trafficking Flag: 100%. Number of benefit periods with replacement cards: Total; Total Trafficking Flags: 1-3: 2313; 4-6: 681; 7-9: 150; 10-19: 37; 20+: 2; Total: 3,183. Source: GAO analysis of Supplemental Nutrition Assistance Program (SNAP) transaction data. GAO-14-641. [End of table] Table 12: Number of Massachusetts High Replacement Card Households Categorized by Count of Trafficking Flags and Replacement Card Benefit Periods , Fiscal Year 2012: Number of benefit periods with replacement cards: 4; Total Trafficking Flags: 1-3: 1392; 4-6: 544; 7-9: 128; 10-19: 36; 20+: 0; Total: 2,100; Percent of Households with 1+ trafficking flags: 78%. Number of benefit periods with replacement cards: 5; Total Trafficking Flags: 1-3: 604; 4-6: 304; 7-9: 100; 10-19: 66; 20+: 2; Total: 1,076; Percent of Households with 1+ trafficking flags: 89%. Number of benefit periods with replacement cards: 6; Total Trafficking Flags: 1-3: 197; 4-6: 139; 7-9: 90; 10-19: 38; 20+: 4; Total: 468; Percent of Households with 1+ trafficking flags: 93%. Number of benefit periods with replacement cards: 7; Total Trafficking Flags: 1-3: 86; 4-6: 71; 7-9: 38; 0-19: 32; 20+: 2; Total: 229; Percent of Households with 1+ trafficking flags: 95%. Number of benefit periods with replacement cards: 8; Total Trafficking Flags: 1-3: 22; 4-6: 27; 7-9: 20; 10-19: 18; 20+: 0; Total: 87; Percent of Households with 1+ trafficking flags: 95%. Number of benefit periods with replacement cards: 9; Total Trafficking Flags: 1-3: 5; 4-6: 13; 7-9: 6; 10-19: 8; 20+: 1; Total: 33; Percent of Households with 1+ trafficking flags: 97%. Number of benefit periods with replacement cards: 10; Total Trafficking Flags: 1-3: 3; 4-6: 3; 7-9: 3; 10-19: 4; 20+: 0; Total: 13; Percent of Households with 1+ trafficking flags: 100%. Number of benefit periods with replacement cards: 11; Total Trafficking Flags: 1-3: 0; 4-6: 0; 7-9: 0; 10-19: 2; 20+: 0; Total: 2; Percent of Households with 1+ trafficking flags: 100%. Number of benefit periods with replacement cards: Total; Total Trafficking Flags: 1-3: 2,309; 4-6: 1,101; 7-9: 385; 10-19: 204; 20+: 9; Total: 4,008. Source: GAO analysis of Supplemental Nutrition Assistance Program (SNAP) transaction data. GAO-14-641. [End of table] Table 13: Number of Nebraska High Replacement Card Households Categorized by Count of Trafficking Flags and Replacement Card Benefit Periods, Fiscal Year 2012: Number of benefit periods with replacement cards: 4; Total Trafficking Flags: 1-3: 162; 4-6: 28; 7-9: 9; 10-19: 1; 20+: 0; Total: 200; Percent of Households with 1+ trafficking flags: 58%. Number of benefit periods with replacement cards: 5; Total Trafficking Flags: 1-3: 57; 4-6: 22; 7-9: 2; 10-19: 0; 20+: 0; Total: 81; Percent of Households with 1+ trafficking flags: 66%. Number of benefit periods with replacement cards: 6; Total Trafficking Flags: 1-3: 20; 4-6: 12; 7-9: 5; 10-19: 2; 20+: 0; Total: 39; Percent of Households with 1+ trafficking flags: 81%. Number of benefit periods with replacement cards: 7; Total Trafficking Flags: 1-3: 6; 4-6: 4; 7-9: 3; 10-19: 0; 20+: 0; Total: 13; Percent of Households with 1+ trafficking flags: 65%. Number of benefit periods with replacement cards: 8; Total Trafficking Flags: 1-3: 2; 4-6: 4; 7-9: 1; 10-19: 0; 20+: 0; Total: 7; Percent of Households with 1+ trafficking flags: 100%. Number of benefit periods with replacement cards: 9; Total Trafficking Flags: 1-3: 0; 4-6: 0; 7-9: 1; 10-19: 0; 20+: 0; Total: 1; Percent of Households with 1+ trafficking flags: 100%. Number of benefit periods with replacement cards: 10; Total Trafficking Flags: 1-3: 2; 4-6: 1; 7-9: 1; 10-19: 0; 20+: 0; Total: 4; Percent of Households with 1+ trafficking flags: 100%. Number of benefit periods with replacement cards: 12; Total Trafficking Flags: 1-3: 0; 4-6: 0; 7-9: 1; 10-19: 0; 20+: 0; Total: 1; Percent of Households with 1+ trafficking flags: 100%. Number of benefit periods with replacement cards: Total; Total Trafficking Flags: 1-3: 249; 4-6: 71; 7-9: 23; 10-19: 3; 20+: 0; Total: 346. Source: GAO analysis of Supplemental Nutrition Assistance Program (SNAP) transaction data. GAO-14-641. [End of table] [End of section] Appendix III: Total E-commerce and Social Media Postings Detected During Testing Periods[Footnote 46] As discussed above, during our 30-day testing period of the automated tool for e-commerce websites, we detected a total of 28 postings from one popular e-commerce website that advertised the potential sale of food stamp benefits in exchange for cash, services, and goods. During our 5 days of testing the automated tool for social media websites, we detected a total of 4 postings potentially soliciting food stamp benefits. The tables below summarize all the e-commerce and social media postings that we detected through the automated tools and manual searches. Table 14: E-commerce Postings Advertising Potential Sale of Food Stamp Benefits for Cash: Manual or auto detection: Manual; Date posted: 12/3/2013; Summary of posting: $400 EBT for $240 in cash; Geographic location monitored: Boston, MA[A]. Manual or auto detection: Manual; Date posted: 12/3/2013; Summary of posting: $200 EBT for $100 in cash; Geographic location monitored: Northern New Jersey[A]. Manual or auto detection: Manual; Date posted: 12/9/2013; Summary of posting: $350 EBT for $200 in cash; Geographic location monitored: Northern New Jersey[A]. Manual or auto detection: Manual; Date posted: 11/22/2013; Summary of posting: $150 in food stamps for $75 in cash; Geographic location monitored: San Antonio, TX. Manual or auto detection: Manual; Date posted: 12/15/2013; Summary of posting: $400 EBT for $280 in cash; Geographic location monitored: Northern New Jersey[A]. Manual or auto detection: Manual; Date posted: 12/15/2013; Summary of posting: $105 in food stamps for $50 in cash; Geographic location monitored: San Antonio, TX. Manual or auto detection: Auto; Date posted: 12/7/2013; Summary of posting: $190 EBT want $160 in cash; Geographic location monitored: Jacksonville, FL. Manual or auto detection: Auto; Date posted: 12/14/2013; Summary of posting: $228 in food stamps for $100 in cash; Geographic location monitored: Jacksonville, FL. Manual or auto detection: Auto and Manual; Date posted: 12/2/2013; Summary of posting: $65 EBT for $30 in cash; Geographic location monitored: Jacksonville, FL. Manual or auto detection: Auto and Manual; Date posted: 12/13/2013; Summary of posting: $180 EBT for $100 in cash; Geographic location monitored: Southern Florida. Source: GAO analysis of e-commerce website postings. GAO-14-641. [A] Posting from nearby geographic location outside the target state. [End of table] Table 15: E-commerce Postings Advertising Potential Sale of Food Stamp Benefits for Services: Manual or auto detection: Manual; Date posted: 11/23/2013; Summary of posting: Undisclosed amount of food stamps, for place to live[A]; Geographic location: Southern Florida. Manual or auto detection: Manual; Date posted: 11/26/2013; Summary of posting: Undisclosed amount of food stamps for place to live[A]; Geographic location: Southern Florida. Manual or auto detection: Manual; Date posted: 11/29/2013; Summary of posting: Undisclosed amount of food stamps for place to live[A]; Geographic location: Southern Florida. Manual or auto detection: Manual; Date posted: 12/5/2013; Summary of posting: Undisclosed amount of food stamps for place to live[A]; Geographic location: Southern Florida. Manual or auto detection: Manual; Date posted: 12/8/2013; Summary of posting: Undisclosed amount of food stamps for place to live[A]; Geographic location: Southern Florida. Manual or auto detection: Manual; Date posted: 12/11/2013; Summary of posting: Undisclosed amount of food stamps for place to live[A]; Geographic location: Southern Florida. Manual or auto detection: Manual; Date posted: 12/13/2013; Summary of posting: Undisclosed amount of food stamps for place to live[A]; Geographic location: Southern Florida. Manual or auto detection: Manual; Date posted: 12/17/2013; Summary of posting: Undisclosed amount of food stamps for place to live[A]; Geographic location: Southern Florida. Manual or auto detection: Manual; Date posted: 12/18/2013; Summary of posting: $500 proposal--some cash to contribute as well as food stamps for place to sleep or for a van[B]; Geographic location: Charlotte, NC. Manual or auto detection: Manual; Date posted: 12/18/2013; Summary of posting: $300 proposal--some cash to contribute as well as food stamps for place to sleep or for a van[B]; Geographic location: Charlotte, NC. Manual or auto detection: Manual; Date posted: 12/19/2013; Summary of posting: $1 proposal--some cash to contribute as well as food stamps for place to sleep or for a van[B]; Geographic location: Charlotte, NC. Manual or auto detection: Auto and Manual; Date posted: 12/17/2013; Summary of posting: 10 days of cooking and cleaning services in exchange for food stamps; Geographic location: Raleigh, NC. Source: GAO analysis of e-commerce website postings. GAO-14-641. [A] Eight postings from Southern Florida advertised an undisclosed amount of food stamps in exchange for services and appear to be a series of updates to the same posting. We counted these as unique postings in accordance with our methodology because the e-commerce website we monitored listed the postings under different dates. [B] Three postings from Charlotte, NC advertised an undisclosed amount of food stamps in exchange for services and had duplicative language. Per our methodology, these were counted as unique postings because their posting dates, times, or subject lines were distinct. [End of table] Table 16: E-commerce Postings Advertising Potential Sale of Food Stamp Benefits for Goods: Manual or auto detection: Manual; Date posted: 11/29/2013; Summary of posting: Phone for EBT benefits; Geographic location monitored: Northern New Jersey[A]. Manual or auto detection: Manual; Date posted: 12/18/2013; Summary of posting: Art for EBT: $10-$3000; Geographic location monitored: Worcester, MA[A]. Manual or auto detection: Manual; Date posted: 12/19/2013; Summary of posting: Catalytic converters for food stamps; Geographic location monitored: Houston, TX[B]. Manual or auto detection: Manual; Date posted: 12/23/2013; Summary of posting: Catalytic converters for food stamps; Geographic location monitored: Houston, TX[B]. Manual or auto detection: Auto and Manual; Date posted: 11/26/2013; Summary of posting: Phone for EBT benefits; Geographic location monitored: Southern Florida. Manual or auto detection: Auto and Manual; Date posted: 12/12/2013; Summary of posting: Food stamps for beer; Geographic location monitored: Charlotte, NC. Source: GAO analysis of e-commerce website postings. GAO-14-641. [A] Posting from nearby geographic location outside the target state. [B] Two postings from Houston, TX advertised an undisclosed amount of food stamps in exchange for goods and had duplicative language. Per our methodology, these were counted as unique postings because their posting dates, times, or subject lines were distinct. [End of table] Table 17: Social Media Postings Soliciting Food Stamp Benefits: Manual or auto detection: Auto; Date posted: 1/6/2014; Summary of posting: "Who got food stamps"; Potential location of posting: Rochester, NY. Manual or auto detection: Auto; Date posted: 1/6/2014; Summary of posting: "Who got food stamps"; Potential location of posting: Minneapolis, MN. Manual or auto detection: Auto; Date posted: 1/7/2014; Summary of posting: "Who got food stamps for sale"; Potential location of posting: Pennsylvania. Manual or auto detection: Auto; Date posted: 1/7/2014; Summary of posting: "Who selling food stamps"; Potential location of posting: Undisclosed. Source: GAO analysis of social media website postings. GAO-14-641. [End of table] [End of section] Appendix IV: Selected States' Experiences Using FNS's Recommended Automated Tool, and GAO 30-day Test Results: Geographic locations GAO monitored: Jacksonville and Southern Florida (FL); Total posts GAO reviewed: 123; Posts returned indicative of trafficking that GAO detected: Total posts detected: 13[A]; Total posts detected through manual searches only: 8; Total posts detected by recommended automated tool only: 2; FNS-recommended automated tool: [Empty]; Manual searches: [Check]; State experiences with using automated and manual tools: Automated tools difficult to use, provides unreliable data; As a result, monitoring websites only. Geographic locations GAO monitored: Boston and Worcester (MA); Total posts GAO reviewed: 245; Posts returned indicative of trafficking that GAO detected: Total posts detected: 2; Total posts detected through manual searches only: 2; Total posts detected by recommended automated tool only: 0; FNS-recommended automated tool: [Check]; Manual searches: [Empty]; State experiences with using automated and manual tools: Automated tools difficult to use, provides unreliable data, not using for social media websites. Automated tools not compatible with state's operating system; Using automated tools for ecommerce websites, but looking for other tools to use for monitoring. Geographic locations GAO monitored: Maine (ME); Total posts GAO reviewed: 25; Posts returned indicative of trafficking that GAO detected: Total posts detected: 0; Total posts detected through manual searches only: 0; Total posts detected by recommended automated tool only: 0; FNS-recommended automated tool: [Empty]; Manual searches: [Check]; State experiences with using automated and manual tools: Has not implemented automated tools yet; Manual monitoring websites only. Geographic locations GAO monitored: Detroit metro and Grand Rapids (MI); Total posts GAO reviewed: 89; Posts returned indicative of trafficking that GAO detected: Total posts detected: 0; Total posts detected through manual searches only: 0; Total posts detected by recommended automated tool only: 0; FNS-recommended automated tool: [Empty]; Manual searches: [Check]; State experiences with using automated and manual tools: Automated tools difficult to use, provides unreliable data; As a result, manual monitoring websites only. Geographic locations GAO monitored: Charlotte and Raleigh (NC); Total posts GAO reviewed: 153; Posts returned indicative of trafficking that GAO detected: Total posts detected: 5[B]; Total posts detected through manual searches only: 3; Total posts detected by recommended automated tool only: 0; FNS-recommended automated tool: [Empty]; Manual searches: [Empty]; State experiences with using automated and manual tools: Some counties have not implemented automated tools yet; Ad hoc monitoring of websites by some counties[C]. Geographic locations GAO monitored: Lincoln and Omaha (NE); Total posts GAO reviewed: 39; Posts returned indicative of trafficking that GAO detected: Total posts detected: 0; Total posts detected through manual searches only: 0; Total posts detected by recommended automated tool only: 0; FNS-recommended automated tool: [Empty]; Manual searches: [Empty]; State experiences with using automated and manual tools: Is not using automated or manual search methods. Geographic locations GAO monitored: Northern New Jersey (NJ); Total posts GAO reviewed: 367; Posts returned indicative of trafficking that GAO detected: Total posts detected: 4; Total posts detected through manual searches only: 4; Total posts detected by recommended automated tool only: 0; FNS-recommended automated tool: [Empty]; Manual searches: [Check]; State experiences with using automated and manual tools: Has not implemented automated tools yet; Manual monitoring websites only. Geographic locations GAO monitored: Memphis and Nashville (TN); Total posts GAO reviewed: 59; Posts returned indicative of trafficking that GAO detected: Total posts detected: 0; Total posts detected through manual searches only: 0; Total posts detected by recommended automated tool only: 0; FNS-recommended automated tool: [Check]; Manual searches: [Empty]; State experiences with using automated and manual tools: Automated tools are labor-intensive to use; However, still find automated tools to work well with some success. Geographic locations GAO monitored: Houston and San Antonio (TX); Total posts GAO reviewed: 75; Posts returned indicative of trafficking that GAO detected: Total posts detected through manual searches only: 4[D]; Total posts detected by recommended automated tool only: 0; FNS-recommended automated tool: [Empty]; Manual searches: [Empty]; State experiences with using automated and manual tools: Is not using automated or manual search approaches. Geographic locations GAO monitored: Salt Lake City and Provo (UT); Total posts GAO reviewed: 5; Posts returned indicative of trafficking that GAO detected: Total posts detected: 0; Total posts detected through manual searches only: 0; Total posts detected by recommended automated tool only: 0; FNS-recommended automated tool: [Empty]; Manual searches: [Check]; State experiences with using automated and manual tools: Automated tools not compatible with state's operating system; As a result, manual monitoring websites only. Geographic locations GAO monitored: Wyoming (WY); Total posts GAO reviewed: 0; Posts returned indicative of trafficking that GAO detected: Total posts detected: 0; Total posts detected through manual Total posts detected by recommended automated tool only: 0; FNS-recommended automated tool: [Empty]; Manual searches: [Check]; State experiences with using automated and manual tools: Has not implemented automated tools yet; Manual monitoring websites only. Geographic locations GAO monitored: Total; Total posts GAO reviewed: 1,180; Posts returned indicative of trafficking that GAO detected: Total posts detected: 28; Total posts detected through manual searches only: 21; Total posts detected by recommended automated tool only: 2. Source: GAO analysis of e-commerce website postings and data from selected states. GAO-14-641. [A] Three posts identified through the automated tool were also identified through manual searches. Eight postings from Southern Florida advertised an undisclosed amount of food stamps in exchange for services and appear to be a series of updates to the same posting. We counted these as unique postings in accordance with our methodology because the e-commerce website we monitored listed the postings under different dates. [B] Two posts identified through the automated tool were also identified through manual searches. Three postings from Charlotte, NC advertised an undisclosed amount of food stamps in exchange for services and had duplicative language. Per our methodology, these were counted as unique postings because their posting dates, times, or subject lines were distinct. [C] North Carolina's SNAP program is county-administered. [D] Two postings from Houston, TX advertised an undisclosed amount of food stamps in exchange for goods and had duplicative language. Per our methodology, these were counted as unique postings because their posting dates, times, or subject lines were distinct. [End of table] [End of section] Appendix V: List of Food and Nutrition Service-Commissioned Studies: 1. Indirect Trafficking Fraud Discovery: A study to identify a process to effectively detect indirect trafficking schemes. An example of an indirect trafficking scheme occurs when a SNAP recipient sells his or her EBT card to a third party, at a discount for cash, and the third party uses the EBT card to purchase eligible food. 2. Social Media Fraud Discovery: An evaluation on FNS's current fraud detection approach using social media in order to identify a more effective process. 3. Recipient Integrity Outcomes Metrics: An analysis to identify metrics that FNS can use to better monitor State performance and outcomes. 4. Multiple EBT Card Replacement: A study aimed at improving FNS's approach to using card replacement data to identify fraud. 5. Household Link Analysis: An analysis to further assess the relationship between client and retailers regarding trafficking schemes. 6. Household Demographic Fraud Discovery: An assessment of recipient benefit data to further refine models used to detect fraud. 7. Household Time and Distance (Geospatial Analysis): An analysis of recipient benefit data that focuses on geographical information; for example, recipients using their cards in Virginia and South Carolina within an hour. 8. Identify Clients Shopping in Geographic Areas Outside of their Normal Patterns (Geospatial Analysis): An analysis to identify an automated process to assess retailer and recipient EBT data. 9. Strengthen FNS Recipient Referrals from Disqualified Stores: A review of the existing recipient referral process and fraud detection models to develop a model for FNS to deploy that more effectively identifies suspicious recipients to states for investigation. 10. SNAP Recipient & Retailer Fraud Data Mining Studies: A series of recipient and retailer based analyses using predictive activity such as fraud discovery to increase FNS and states ability to detect fraudulent activity. [End of section] Appendix VI: GAO Contacts and Staff Acknowledgment: GAO Contacts: Kay E. Brown, 202-512-7215, brownke@gao.gov. Seto J. Bagdoyan, 202-512-6722, bagdoyans@gao.gov. Key Contributors: In addition to those mentioned above, the following staff members may significant contributions to this report: Kathryn Larin and Philip Reiff, Assistant Directors; Celina Davidson and Danielle Giese, Analysts-in-Charge; LaToya King, Flavio Martinez, Erik Shive and Jill Yost. Additionally, James Bennett, Holly Dye, Linda Miller, Maria McMullen and Almeta Spencer provided technical support; Shana Wallace provided methodological guidance; and Alexander Galuten and James Murphy provided legal counsel. [End of section] Related GAO Products: Standards for Internal Control in the Federal Government: 2013 Exposure Draft. [hyperlink, http://www.gao.gov/products/GAO-13-830SP]. Washington, D.C.: September 3, 2013. Supplemental Nutrition Assistance Program: Payment Errors and Trafficking Have Declines, but Challenges Remain. [hyperlink, http://www.gao.gov/products/GAO-10-956T]. Washington, D.C.: July 28, 2010. Food Stamp Trafficking: FNS Could Enhance Program Integrity by Better Targeting Stores Likely to Traffic and Increasing Penalties. [hyperlink, http://www.gao.gov/products/GAO-07-53]. Washington, D.C.: October 13, 2006. Individual Disaster Assistance Programs: Framework for Fraud, Prevention, Detection and Prosecution. [hyperlink, http://www.gao.gov/products/GAO-06-954T]. Washington, D.C.: July 12, 2006. Hurricanes Katrina and Rita Disaster Relief: Improper and Potentially Fraudulent Individual Assistance Payments Estimated to be between $600 Million and $1.4 Billion. [hyperlink, http://www.gao.gov/products/GAO-06-844T]. Washington, D.C.: June 14, 2006. Expedited Assistance for Victims of Hurricanes Katrina and Rita: FEMA's Control Weaknesses Exposed the Government to Significant Fraud and Abuse. [hyperlink, http://www.gao.gov/products/GAO-06-403T]. Washington, D.C.: February 13, 2006. Standards for Internal Control in the Federal Government. [hyperlink, http://www.gao.gov/products/GAO/AIMD-00-21.3.1] Washington, D.C.: November 1999. [End of section] Footnotes: [1] This dollar amount represents benefits distributed in error due to administrative as well as recipient errors, not all of which can be attributed to fraud. [2] Food Stamp Trafficking: FNS Could Enhance Program Integrity by Better Targeting Stores Likely to Traffic and Increasing Penalties, [hyperlink, http://www.gao.gov/products/GAO-07-53] (Washington, D.C.: Oct. 13, 2006). [3] Pub. L. No. 111-5, § 101, 123 Stat. 115, 120. [4] "E-commerce" websites allow users to advertise the sale of goods and services. "Social media" websites allow subscribers to exchange information and ideas with others who may or may not subscribe. [5] 7 C.F.R. § 273.16. [6] 7 U.S.C. § 2015(b), 7 C.F.R. § 273.16. Furthermore, under federal law, it is illegal for a person to knowingly use, transfer, acquire or possess SNAP benefits in any manner that is contrary to the laws and regulations that govern the SNAP program. 7 U.S.C. § 2024(b). The statute applies to program recipients and retailers as well as people not participating in the program. [7] Similar to a bank card, for security purposes, SNAP EBT cards require a PIN to access the benefits associated with the card. [8] USDA OIG, Analysis of FNS' Supplemental Nutrition Assistance Program Fraud Prevention and Detection Efforts. Audit Report 27002- 0011-13, (Washington, D.C.: Sept. 28, 2012). [9] It also represents the over-and underpayment of benefits, although the vast majority of improperly paid SNAP benefits are overpaid. [10] OMB designates a program as "high-error" based on improper payment information in agencies' annual Performance and Accountability Report and Agency Financial Report. Generally, a program is deemed susceptible to significant improper payments if the program has such payments greater than $10 million and over 2.5 percent of all payments made under that program, or if the program has more than $100 million in estimated improper payments. SNAP is ranked 7TH based on the dollar amount estimated in improper payments among the 13 programs on the high error list. [11] Pub. L. No. 104-193, § 841, 110 Stat. 2105, 2331. Prior to this law, FNS usually sent its investigators into stores numerous times over a period of months to attempt to traffic benefits as a way to gather evidence against a retailer. The use of EBT transaction evidence can help to reduce the resources needed for investigations by eliminating the need for multiple store visits for some cases. [12] Food Stamp Trafficking: FNS Could Enhance Program Integrity by Better Targeting Stores Likely to Traffic and Increasing Penalties, [hyperlink, http://www.gao.gov/products/GAO-07-53] (Washington, D.C.: Oct. 13, 2006). [13] USDA OIG, State Fraud Detection Efforts for the Supplement Nutrition Assistance Program - Use of EBT Management Reports, 27703-02- Hy(2) (Washington, D.C.: Sept. 10, 2010). [14] The USDA OIG also recommended that FNS improve the calculation of its retailer trafficking rate because it was based on a judgmentally- selected sample. For more information, see USDA OIG, Analysis of FNS' Supplemental Nutrition Assistance Program Fraud Prevention and Detection Efforts, Audit Report 27002-0011-13 (Washington, D.C.: Sept. 28, 2012). [15] Also, in fiscal year 2012, FNS reported that states established over $530 million in new claims against recipients and collected over $320 million. These claims represent those established due to intentional program violation (fraud), inadvertent household error, or agency error. [16] eDRS is also a data matching tool to help prevent improper payments and FNS requires that states check this system prior to providing benefits to an applicant. [17] According to program regulations, if a person is found to have trafficked $500 or more in benefits or used or received benefits in a transaction involving the sale of firearms, ammunition, or explosives, he or she will be disqualified permanently for the first offense. Also, a person who has fraudulently provided identification or residential information in order to gain duplicative benefits will be disqualified for 10 years for the first offense. Using or receiving benefits in transactions involving the sale of controlled substances will result in a 2 year penalty for the first violation and permanent disqualification for the second. Lesser offenses will result in 1 year disqualification for the first violation, 2 years for the second and permanent disqualification for the third. [18] These reports are submitted within 45 days of the end of the state's fiscal year. According to FNS, this means the agency receives the reports on August 15TH from most states. [19] In most of the 11 states we reviewed, the fraud units are housed in the human services department or a state division responsible for program integrity, and investigators may be concentrated in the state capital or located around the state. Investigators typically have backgrounds either in investigation or as case workers or both, and five states noted that their typical investigator has a 4-year college degree in a related field. The federal government generally reimburses states for 50 percent of the cost of their fraud investigative work. [20] Because counties in North Carolina vary in how they staff fraud investigations as mentioned earlier, officials were not able to provide the number of investigators working on SNAP. [21] GAO's Fraud Prevention Framework, developed during previous program audits, emphasizes that comprehensive controls are necessary to minimize fraud, waste, and abuse within any federal program. For more information, see Individual Disaster Assistance Programs: Framework for Fraud, Prevention, Detection and Prosecution, [hyperlink, http://www.gao.gov/products/GAO-06-954T] (Washington, D.C.: July 12, 2006); Hurricanes Katrina and Rita Disaster Relief: Improper and Potentially Fraudulent Individual Assistance Payments Estimated to be between $600 Million and $1.4 Billion, [hyperlink, http://www.gao.gov/products/GAO-06-844T] (Washington, D.C.: June 14, 2006); and Expedited Assistance for Victims of Hurricanes Katrina and Rita: FEMA's Control Weaknesses Exposed the Government to Significant Fraud and Abuse, [hyperlink, http://www.gao.gov/products/GAO-06-403T] (Washington, D.C.: Feb. 13, 2006). [22] According to the GAO Fraud Prevention Framework, prevention controls are the most efficient and effective ways to address fraud because payments can be difficult to recover. [23] For example, providing states with additional resources through increased retention rates may result in a net savings for SNAP if increased collections in payment recoveries outweigh the increased amount states receive in retentions. However, FNS officials noted the retention rate is set forth in federal law and that changing the retention rate would require legislative action. As early as December 2012, FNS officials reported encouraging states to use bonuses received from FNS for good performance and fraud claims collection retentions to support program integrity and anti-fraud efforts. [24] In addition to those in North Carolina, officials in Michigan and Massachusetts reported a reduction in replacement card requests after sending the notification letters. [25] We reviewed transaction data for 10,266 households receiving replacement cards in four or more benefit periods; we did not receive transaction data for 4 of the 10,270 total households we identified in our replacement card analysis. [26] A transaction may appear under more than one trafficking flag. For example, a transaction for $500.00 that occurred prior to a replacement card would be flagged as both a large-dollar transaction and an even-dollar transaction. There were 2,484 transactions in our analysis that appeared under more than one trafficking flag. [27] FNS is currently assessing the prevalence of SNAP trafficking occurring online and the effectiveness of these automated tools. [28] For more information on how the 11 selected states monitored online postings, see Figure 4. [29] GAO: Individual Disaster Assistance Programs: Framework for Fraud Prevention, Detection, and Prosecution, [hyperlink, http://www.gao.gov/products/GAO-06-954T] (Washington, D.C.: July 12, 2006). [30] We limited our searches to two key terms (i.e. "EBT" and "food stamps"). The use of other terms could have yielded additional or fewer posting results through our automated and manual searches. [31] We note that it could take additional or less time for states to conduct online monitoring based on the number of websites, geographical locations, and key terms that they actually decide to use. [32] See Appendix III for a summary of all e-commerce postings we detected on one popular e-commerce website. [33] A posting would be included under "local" results if the individual placing the posting online chose to advertise to that specific geographic location. However, searches of the website we reviewed also included postings from "nearby" areas to the specific geographic location selected by the individual placing the posting online. [34] Search results can include postings from nearby geographic locations that were outside the target states. Also, eight of the postings from Florida appear to be a series of updates to the same posting, and three of the postings from North Carolina contain duplicative language. However, we counted these as unique postings in accordance with our methodology because they were listed on the e- commerce website under different dates or subject lines. [35] While we limited our searches to these key terms, the use of other terms could have yielded additional or fewer posting results through our automated searches. [36] See Appendix III for a summary of all 4 social media postings potentially soliciting food stamp benefits. The 4 posts that we detected are not generalizable to the potential SNAP trafficking activity that may be occurring within or across all states or on the Internet. [37] Prior work by the USDA OIG found that FNS has not maintained strong internal controls for overseeing state fraud detection units. Specifically, in a 2010 report, the USDA OIG found that FNS had not conducted periodic reviews of state fraud detection activity to verify its effectiveness, and that FNS had not found such reviews necessary because the agency believed collecting data on states' activities through the FNS-366B to be sufficient for its monitoring purposes. However, the USDA OIG found that FNS did not have a system for ensuring the accuracy of state-reported data and cited data problems in the two states included in the study. For more information, see USDA OIG, State Fraud Detection Efforts for Supplemental Nutrition Assistance Program, 27703-02-Hy (Washington, D.C.: July 12, 2010). [38] This dollar amount includes funding for studying some retailer anti-fraud efforts as well. Given the structure of the contracts for these studies, FNS could not provide a separate amount for only recipient fraud. [39] According to federal regulations, states must report disqualifications to FNS no later than 30 days after the disqualification takes effect. 7 C.F.R. § 273.16(i). [40] Poor eDRS data would also affect states' ability to be aware of and impose penalties for repeat program violators. [41] As mentioned earlier, states are required to report their fraud- related activity to FNS on their annual Program and Budget Summary Statements. This report, provided through the form FNS-366B, is to include the number of investigations and disqualifications, and the dollar amount of their fraud claims. [42] According to federal internal controls, it is important that sufficient guidance and training is provided to ensure accuracy and reduce misunderstandings. For more information see, GAO: Standards for Internal Control in the Federal Government, [hyperlink, http://www.gao.gov/products/GAO/AIMD-00-21.3.1] (Washington, D.C.: November 1999) and OMB Circular A-123 Revised. [43] Pub. L. No. 111-5, § 101, 123 Stat. 115, 120. [44] GAO, Individual Disaster Assistance Programs: Framework for Fraud Prevention, Detection, and Prosecution, [hyperlink, http://www.gao.gov/products/GAO-06-954T] (Washington, D.C.: July 12, 2006). [45] For the purposes of our analysis, we defined excessive replacement cards as replacement cards in four or more monthly benefit periods. [46] We referred postings indicative of SNAP trafficking to the appropriate state and federal officials for further investigation. [End of section] GAO's Mission: The Government Accountability Office, the audit, evaluation, and investigative arm of Congress, exists to support Congress in meeting its constitutional responsibilities and to help improve the performance and accountability of the federal government for the American people. GAO examines the use of public funds; evaluates federal programs and policies; and provides analyses, recommendations, and other assistance to help Congress make informed oversight, policy, and funding decisions. GAO's commitment to good government is reflected in its core values of accountability, integrity, and reliability. Obtaining Copies of GAO Reports and Testimony: The fastest and easiest way to obtain copies of GAO documents at no cost is through GAO's website [hyperlink, http://www.gao.gov]. Each weekday afternoon, GAO posts on its website newly released reports, testimony, and correspondence. To have GAO e-mail you a list of newly posted products, go to [hyperlink, http://www.gao.gov] and select "E-mail Updates." Order by Phone: The price of each GAO publication reflects GAO's actual cost of production and distribution and depends on the number of pages in the publication and whether the publication is printed in color or black and white. Pricing and ordering information is posted on GAO's website, [hyperlink, http://www.gao.gov/ordering.htm]. Place orders by calling (202) 512-6000, toll free (866) 801-7077, or TDD (202) 512-2537. Orders may be paid for using American Express, Discover Card, MasterCard, Visa, check, or money order. Call for additional information. Connect with GAO: Connect with GAO on facebook, flickr, twitter, and YouTube. Subscribe to our RSS Feeds or E mail Updates. Listen to our Podcasts. Visit GAO on the web at [hyperlink, http://www.gao.gov]. To Report Fraud, Waste, and Abuse in Federal Programs: Contact: Website: [hyperlink, http://www.gao.gov/fraudnet/fraudnet.htm]; E-mail: fraudnet@gao.gov; Automated answering system: (800) 424-5454 or (202) 512-7470. Congressional Relations: Katherine Siggerud, Managing Director, siggerudk@gao.gov: (202) 512-4400: U.S. Government Accountability Office: 441 G Street NW, Room 7125: Washington, DC 20548. Public Affairs: Chuck Young, Managing Director, youngc1@gao.gov: (202) 512-4800: U.S. Government Accountability Office: 441 G Street NW, Room 7149: Washington, DC 20548. [End of document]