Skip to main content

Food Stamp Error Rates Hold At Record Low Levels In 2005

On June 23, the U.S. Department of Agriculture (USDA) released state and national food stamp error rates for federal fiscal year 2005 calculated through the food stamp quality control (QC) system.  The national overpayment error rate — the percentage of food stamp benefit dollars issued in excess of the amounts for which households are eligible — remained at record low levels of 4.5 percent for the second consecutive year.  The overpayment error rate has dropped by more than forty percent from the 1998 overpayment rate.  The underpayment error rate fell for the seventh consecutive year to the lowest level on record, 1.31 percent.  The combined payment error rate, which is calculated by summing (rather than netting) the overpayment and underpayment error rates, improved slightly over last year’s record low of 5.88 percent, to set a new all-time low of 5.84 percent.

The U. S. Government Accountability Office (GAO) statement of last year remains true with this year’s error rate, “[t]he payment error rate has fallen each year since 1999. This decline in the payment error rate has been widespread.”[1]  This year over 30 states had combined payment error rates that were lower in 2005, or essentially the same as in 2004.  Almost two-thirds of the states (31 states plus the Virgin Islands) had combined error rates below 6 percent, a level that until recently automatically qualified states for enhanced funding due to exemplary performance.  In 2005 no state had a combined payment error rate above ten percent; in 1998, when the national average was 10.69 percent, there were almost 30 states with combined payment error rates in double digits.

Most errors are relatively small, and very few represent fraud.  GAO reported that “[a]lmost two-thirds of the payment errors in the Food Stamp Program are caused by caseworkers, usually when they fail to act on new information or make mistakes when applying program rules.”  The program’s success in serving the working poor contributes in part to its error rate:  GAO reported that “managing cases with earnings contributes to payment error in part because caseworkers may find it difficult to keep up with frequent changes reported to them.”

This year, USDA identified three states that exceeded the QC threshold level in 2004 and in 2005.  Because of the rapid drop in national error rates, these three states’ error rates are close to or below the national average of the late 1990s.  Nonetheless, USDA determined that it is highly likely statistically that they exceeded the “threshold” of 105 percent of the national average.  All three of these states will therefore be required to spend their own money (without the usual federal match) to improve their administration of the Food Stamp Program — and will have to pay a fiscal penalty if their error rate for this year does not fall to the threshold level.  In addition, USDA identified five other states as being in “first year liability status.”  These are states where there is strong statistical likelihood that the state’s 2005 combined payment error rate exceeded the threshold level.  These states will be required to spend unmatched money on program improvements next year if USDA makes a similar finding based on each state’s 2006 combined payment error rate.

USDA also released states’ error rates for cases in which they denied or terminated bene­fits.  (The underpayment error rate includes only cases where states gave some benefits, but not as much as the household should have received under food stamp rules.  It does not include actions that completely denied food stamps to eligible low-income households.)  Nationally, in 6.91 percent of the instances in which households were denied food stamps or terminated from the Program, the action was found to be in error.  USDA did not attempt to calculate the amount of benefits that these improperly denied households would have received.  As a result, this “negative error rate” is not directly comparable to the overpayment and underpayment error rates, but is instead a less rigorous measure of whether the state followed the proper procedures before denying or terminating food stamps.  Nonetheless, improper denials and terminations, like under­payments, result in significant, if unintended, savings to the Program.

GAO reports that USDA and the states “have taken many approaches to increasing food stamp payment accuracy, … includ[ing] practices to improve accountability, perform risk assessments, implement changes based on such assessments, and monitor program performance.”  It found that these practices were “recognized as being effective in reducing payment errors.”

Although food stamp error rates have received little public attention in recent years, they do enter into discussions of the Program.  Sometimes these discussions fall victim to signifi­cant mistakes or mischaracterizations of the food stamp error rates.  To understand the error rates properly, several points should be kept in mind.


 What the New Food Stamp Error Rates Show

  • The combined error rate does not represent losses to the government.  USDA actually issues three separate payment error rates: the overpayment error rate, the underpayment error rate, and the combined payment error rate.  The over­payment error rate counts benefits issued to ineligible households as well as benefits issued to eli­gible households in excess of what federal rules provide.  The underpayment error rate measures errors in which eligible, participating households received fewer benefits than the Program’s rules direct.  As GAO notes, “[u]nderpayments represent unintentional financial savings to the federal government.”  The combined payment error rate is the result of summing (rather than netting) the overpayment and underpayment error rates.

Thus, for example, a state with a five percent overpayment error rate and a two percent underpayment error rate would be reported as having a combined error rate of seven percent.  The net loss to the federal government, however, from the errors in that state’s program (i.e., the bene­fits lost through overpayments minus those saved by underpayments) would be only three percent.[2]

As noted above, even this measure overstates the cost of errors to the Program.  If it were possible to quantify the amount of benefits eligible households lost due to improper denials and terminations, the net loss to the program would be less.  Indeed, it is possible that the combined savings from underpayments and improper denials is greater than the loss resulting from overpayments of benefits.  The media often pay the most attention to the com­bined error rate, presenting it as a reflection of the dimen­sion of excessive federal expenditures due to errors.  This is incorrect since the combined error rate in­cludes underpayments that save the Program money.

  • The decrease in error rates has been widespread.  In 2005, 15 states achieved their lowest combined payment error rates on record.  Some 48 states had lower error rates in 2005 than they did in 1998 (a year when error rates rose due in part to the com­plexity of implementing changes from the 1996 welfare law).  In 1998, only seven states achieved combined payment error rates below 6 percent, which is the level Congress historically has designated as representing exemplary administration.  In 2005, almost two-thirds of the states (32 states) had combined payment error rates below 6 percent.
  • The largest states have seen the biggest drop in error rates.  The decline in the national average error rates has been largely driven by declines in error rates for the states with the largest food stamp issuances.  Of the ten states with the largest food stamp issuance, five states have cut their error rate by about 60 percent, and two states have cut their error rate by about 40 percent, over the last several years.  Two of the largest states, Texas and Louisiana, have had error rates among the nation’s lowest for a decade.
  • The Food Stamp Program is requiring states to achieve much lower error rates than in past years.  Because error rates have fallen so much in recent years the performance standard for states has gotten much more stringent.  All of the states USDA identified this year as potentially subject to sanction next year had 2005 error rates that are below recent years’ national averages.  GAO found that “if error rates continue to decrease, this trend will continue to put pressure on states to improve because penalties are assessed using the state’s error rate as compared with the national average.”
  • The dollar amount of most errors is quite small.  A USDA study found that the overwhelming majority of food stamp overpayments go to eligible households and leave the recipient households still well below the poverty line.  It found, in 2003, that less than two percent of recipient households were completely ineli­gible for food stamps and that only two percent of food stamp benefits were in­correctly issued to these ineligible households.  In other words, almost 99 percent of food stamps are issued to eligible households.  The study also found that, for two-thirds of the households receiving overpayments, the extra food stamps im­proved its food purchasing power by less than ten percentage points of the federal poverty line.
  • Food stamp error rates compare favorably to those in other government programs for which data is available.  For example, the Internal Revenue Service estimates a noncompliance rate with federal personal income taxes of at least sixteen percent in 2001.  This represents about $300 billion lost to the federal government.[3]

The Difference between Overpayments and Fraud

Relatively few of these errors represent dishonesty or fraud on the part of recipients (e.g., recipients intentionally lying to eligibility workers to get more food stamps).  By its very nature, fraud is difficult to measure accurately.  The overwhelming majority of food stamp errors, however, appear to result from honest mistakes by recipients, eligibility wor­kers, data entry clerks, or com­puter programmers.  In recent years, states have reported that about half of the dollar value of overpayments and three-quarters of the dollar value of underpayments were their fault.  Most of the rest resulted from innocent errors by households.[4]  The Food Stamp Program has numerous anti-fraud measures in place, including sophisticated computer “matching” efforts to detect unreported earnings and assets, extensive requirements that households applying for or seeking to continue receiving food stamps prove their eligibility, and administrative and criminal enforcement mechanisms.

It also should be noted that an overpayment is counted in a state’s error rate whether or not the overpaid benefits are collected back from households.  In fiscal year 2003, states collected over $200 million in overissued benefits.

In addition, the error rates measure the accuracy with which benefits are issued, not whether food stamps are redeemed or spent properly.  Evidence from USDA research suggests that a very small fraction of food stamp benefits are improperly traded for cash, or “trafficked.”  USDA found that over the 1999 to 2002 period only two-and-a-half cents of every dollar issued in food stamps was trafficked.  The amount of benefits trafficked has fallen by over 50 percent since 1993.  Trafficking likely represents an even smaller proportion of benefits as the use of electronic benefit transfer (EBT) — or providing food stamps on cards that can be swiped at stores like credit or debit cards — has expanded to be nationwide.  One of the benefits of providing food stamp benefits through EBT is that it reduces the risks of trafficking by providing an electronic record of every transaction that is then scrutinized by sophisticated computers that USDA maintains.

What Factors Contributed to States’ Error Rates

Because the error rate measures how accurately states determine households’ benefit levels, it is an important measure of how well a state is managing program accuracy.  States with persistently high error rates often lack management focus or appropriate levels of oversight with respect to payment accuracy.  However, a state may experience an increased error rate for a number of other reasons that do not necessarily suggest a management failure.

  • State Fiscal Situation.  Since the beginning of the last economic downturn in March 2001, food stamp caseloads nationally have increased by about 50 percent.  Some of the states with the steepest increases in unemployment have also seen the largest increases in the number of people who receive food stamps.  For example, food stamp caseloads have increased between 2001 and 2005 by 69 percent in South Carolina and 65 percent in Michigan, two states where the unemployment rate has been among the highest in the country.  This is a strong indication that the Food Stamp Program is working — that it is responding to increases in need as unemployment rises.

These caseload increases are occurring, however, at the same time that states are facing limited resources for food stamp administrative costs.  Many are cutting back or freezing the number of eligibility workers who make food stamp eligibility determinations.  (Although food stamp benefits are 100 percent federally-funded, states provide about half of the administrative costs for determining eligibility and issuing benefits.)  Tight state budgets also can make it difficult for states to invest in computer upgrades, staff training, or other administrative activities that could help them improve their error rates.

The fact that over­all caseloads have been increasing and states have been under budget pressures makes the decline in error rates in most states over the past few years even more remarkable.

  • Increased share of working families receiving food stamps.   Families’ move­ment from welfare to work also has tended to increase error rates.  Households containing wage-earners historically have had higher error rates than those that rely solely on public assistance, SSI or Social Security.  This is because many low-wage workers have unpredictable or fluctuating employment that makes it difficult for states to predict their earnings into the future.  Between 1990 and 2004 the proportion of food stamp households with children that work rose from a quarter to almost half, while the share of food stamp families with cash welfare and no earnings fell from almost 60 percent to 23 percent.  The larger numbers of food stamp recipients that have been able to find work has likely increased both the over- and under-payment error rates above the levels that would otherwise have prevailed.  The fact that error rates are nonetheless declining means that improved state management and other factors have likely been in play to help offset this trend in the composition of food stamp households.
  • Changes in law or policy.  In the late 1990s, a significant part of states’ over-payments resulted from states’ difficulties in implementing complex provisions of the 1996 welfare law, notably the provision denying food stamp eligibility to the majority of legal immigrants.  On the other hand, changes that the Administration has made in policy and state options to simplify certain procedures in the delivery of food stamp benefits that were enacted in the 2002 farm bill — such as simpli­fied rules regarding what changes in circumstances clients must report in between visits to the welfare office and options to streamline what counts toward the income and asset limits — have likely had a significant role in helping to reduce errors in recent years.

 The Recent Changes to the Quality Control System

Prior to the 2002 reauthorization of the Food Stamp Program, a consensus emerged among states, advocacy groups, USDA, and other policy makers that the food stamp QC system exerted an inappropriate influence on state policy.  The prior system (which remained in effect through the 2002 error rates) subjected states with combined payment error rates above the national average to sanction.  This set up half the states to be viewed as failures each year.  As a result of this QC sanction system, states with high or rising error rates were under strong pressure from USDA to adopt policies that improve their error rates.  State officials, governors, and state legislatures take these sanctions very seriously.  Receiving a fiscal sanction can be per­ceived as a serious negative reflection on the state’s performance, even when the performance may be only modestly worse than average.

Some approaches that states may employ to reduce overpayments — improved staff training, giving eligibility workers more manageable caseloads, combating staff turnover, centralized change reporting functions, simplifying and better explaining households’ reporting obligations, etc. — also are likely to reduce underpayments and to improve needy families’ access to nutrition assistance.  Other approaches, however, such as requiring working recipients to take time off work more frequently to come into the food stamp office for inter­views, and increasing the amount of documentation a household must provide to verify their income and other circumstances, can have the effect of driving eligible families away from food stamps at the very time they may need these bene­­fits to support their transition from wel­fare to work.  This may have the effect of reducing states’ error rates by reducing participation by working poor families (a group with an above average error rate).  Unfortunately, it also undercuts efforts to make work more attractive than welfare and is likely to cause hardship for the families affected.

As a result of these concerns, the nutrition title of the 2002 farm bill included a major reform to the food stamp QC system’s sanction rules.  While retaining the program’s strong commitment to payment accuracy, the new system focuses penalties on the few states with consistently high error rates.  From a management perspective, this revised QC sanction system provides USDA with a broader range of options for how they respond to various payment accuracy concerns and how they assist states in improving their performance.  USDA is now better equipped to provide different interventions for different types of states as opposed to having only the blunt legal requirement to sanction all states with measured error rates above the national average each year, regardless of the cause.

States that have chronic, long-term, excessive payment accuracy problems still are subject to financial penalties and the new rules actually increase the likelihood that such states pay fiscal penalties.  However, many states experience short-term problems when, for example, they implement new computer systems, they implement a complex change in policy, or when their caseloads increase because of a downturn in the economy.  In these states, it is counter­productive to take away resources at the very time that the state needs more resources to cope with the problem.  Under the new system, states with short-term error rate problems have time to work to correct the problems before they are faced with a fiscal penalty.

Specifically, under the new rules a state is subject to fiscal sanction if, with high statistical likelihood, its combined payment error rate exceeds 105 percent of the national average for two consecutive years.  The new rules took effect for the 2003 error rates, and 11 states were identified as potentially subject to sanction based on that year’s error rates.  Four of the 11 states received sanctions based on 2004 error rates, but seven reduced their error rates sufficiently to avoid sanction.  Only two of those states (Rhode Island and Idaho) are in sanction status again, based on 2005 error rates.  One state (Ohio) received a sanction based on 2004 and 2005 error rates, and five states were put on warning that they could receive a sanction based on 2006 error rates if they do not improve their performance sufficiently.  If the old rules had been in effect for 2005 error rates, 24 states would have received fiscal sanctions, including states with combined payment error rates as low as 5.9 percent — a level that for many years would have en­titled states to bonuses for excellent administration.  The fact that error rates have dropped dramatically since 2002 demonstrates that states continue to take the food stamp QC system seriously and have made significant efforts to improve their performance.

This reform of the QC sanction system should lessen the pressure that states feel to adopt policies that impede access to the Food Stamp Program.  Indeed, USDA has recently released a study that finds that food stamp participation rates among those eligible for the Program rose from 56 percent in 2003 to 61 percent in 2004.  The QC system nonetheless remains the most sophisticated system for measuring payment accuracy in any major federal public benefit program and continues to be a critical tool for measuring and monitoring states’ stewardship of federal food stamp funds.

The new QC system also includes new performance bonuses that reward exemplary achievements in payment accuracy and service to eligible households.  Specifically, in addition to awarding bonus funds to states that achieve low or improved error rates, USDA also now rewards states with high or improved rates of serving eligible households and in doing so in a timely manner. 

End Notes

End Notes: 

[1] U.S. Government Accountability Office, Food Stamp Program: States Have Made Progress Reducing Payment  Errors, and Further Challenges Remain, GAO-05-245, May 2005.

[2] To be sure, these savings are not sought or desired by either federal or state agencies.  But in calculating the net cost to the federal government of errors, or the difference between the actual cost of the Program and what it would cost in the absence of errors, the value of benefits not provided due to underpayments must be subtracted.

[3]  Internal Revenue Service, IRS Updates Tax Gap Estimates (IR-2006-28, February 14, 2006), available at:,,id=154496,00.html.

[4] In fiscal year 2003, over 90 percent of the value of all overpayments states established were classified as non-fraud.  Some of these were innocent errors by households; others were mistakes the states themselves made.  The GAO report found that only about 5% of all errors were referred for suspected participant fraud investigation.