_ __ . ..__._. - __.... .--......._.-.--- --. -___p-_------ .I II I\,* I !VlO * FOOD STAMP PROGRAM The Household Definition Is Not a Major Source of Caseworker Errors E I II 141957 EE!3TRICTED-- Not to be released outside the General Accounting OfPice unless specif5caJly approved by the Office of Congressional Belations. Hm RELEASED ^l." __..,I-._.I._._--".._ ---- _...,_.. - -_ _.,. ___..-. ..- United States GAO General Accounting Office Washington, D.C. 20648 Resources, Community, and Economic Development Division B-217883 July 26,199O The Honorable Patrick J. Leahy Chairman, The Honorable Richard Lugar Ranking Minority Member, Committee on Agriculture, Nutrition, and Forestry United States Senate The Honorable Tom Harkin Chairman, Subcommitteeon Nutrition and Investigations, Committee on Agriculture, Nutrition, and Forestry United States Senate The Honorable Charles Hatcher Chairman, Subcommitteeon Domestic Marketing, Consumer Relations and Nutrition Committee on Agriculture House of Representatives This report is the first part of our responseto your request for informa- tion on the household definition used in the Food Stamp Program. It dis- cussesour evaluation of whether the current household definition used to compute food stamp benefits is a significant causeof caseworker errors. A secondreport, Food Stamp Program: Alternative Definitions of a Household for Food Stamp Eligibility (GAO/WED-90-137), will be issued shortly and will provide information on the historical evolution and complexity of the current household definition; whether the definition contributes to homelessness;and 11 alternative definitions and their potential effects on participation and benefit payments, homelessness, and program simplicity. The current definition of a household doesnot significantly contribute Results in Brief to caseworker errors. We estimate, on the basis of a statistically valid sample, that caseworkers,nationwide, made household definition errors in about 1 percent, or between 0.4 million and 1.2 million, of the 80 mil- lion food stamp issuancesin fiscal year 1988, the latest year for which complete data were available. As a result, between $23 million and $75 million of the $10.3 billion in food stamp benefits issued were either Page 1 GAO/RCED-90-183 Household Definition Caseworker Errors f \ n, 5217888 b overpaid or underpaid to recipients becauseof household definition errors.’ The Food Stamp Program, administered by the Food and Nutrition Ser- Background vice of the U.S. Department of Agriculture (USDA), is the nation’s largest food assistanceprogram, delivering about $11.1 billion in benefits to a monthly averageof 7.1 million households(18.6 million people) in fiscal year 1988. Since food stamp benefits are provided to householdsrather than to individuals, a key factor in determining applicants’ eligibility and benefits is how accurately caseworkersapply the household defini- tion. Under the current household definition, people who live together and who customarily purchase food and prepare meals together must generally form a single household. Several exceptions allow somefamily members and certain other people or groups to form separate food stamp householdswhile sharing housing. While the basis of the current household definitmn becameeffective in 1977, the exceptions, which stem mostly from the’omnibus Budget and Reconciliation Acts of 1981 and 1982 and the 1slEmStewart B. McKinney HomelessAssistanceAct, must be applied by caseworkersto determine the proper composition of households.(Seeapp. II for a list of major exceptions in the current household definition.) Becausefood stamp benefits are allocated according to the size and economicresourcesof a household, determining household composition-the people who should be included in or excluded from a household-is a key element in making accurate eligibility and benefit determinations. Other key ele- ments include correctly establishing the income, resources(assets),and deductions or expensesof household members. The Food Stamp Act of 1977 established the quality control review system used in the Food Stamp Program. Under this system, each state is required to conduct quality control reviews by selecting a statistically valid sample of its food stamp caseload.Each state’s quality control staff must review the casesin the sample to verify the accuracy of the state’s benefit and eligibility determinations. From this information, the state determines its error rate and reports it to the Service. ‘This dollar value represents the total, or absolute value, of all overissuances plus all underis- suances-not the net amount of overissuances less underissuances. The $10.3 billion amount is for the benefits issued to food stamp cases that were subject to the Food and Nutrition Service’s quality control review during fiscal year 1988. Page 2 GAO/RCED&lfJ2 Household Definition Caseworker Jbrora c IS217883 The results of each state’s quality control reviews are validated by the Service’s reviewers. The reviewers select a subsampleof the casesfrom each state’s quality control sample and determine if the state properly reviewed the casesand reported the results. The Service then notifies the state of its finding and applies a statistical procedure to calculate the official error rate. Either recipients or caseworkerscan causeincorrect eligibility determi- How Food Stamp nations resulting in over/under-issuancesof food stamp benefits. Recip- Eligibility Errors Are ient-causederrors occur when recipients do not provide information on Caused,Tracked, and changesin household membership or provide inaccurate information. Caseworker errors occur when caseworkersmake calculation errors or Categorized misapply provisions of the Food Stamp Act or other administrative provisions. At the federal level, the Food and Nutrition Service tracks errors which lead to improper food stamp benefit determination, and for analytical purposes, groups them into several categories.We have compressedthe Service’serror categoriesinto three major segments-financial, nonfi- nancial, and other-to simplify their presentation in this report. For fiscal year 1988-the latest year for which complete data were available-the Service estimated that mistakes, both recipient- and caseworker-caused,occurred in about 23.9 percent of all food stamp cases,’resulting in total erroneous payments of about $1 billion (about 10 percent of the food stamp benefits paid). To ascertain the causeof errors, the Service analyzes and reports on “variances”.3 Financial vari- ances-those made in the calculation of income, deductions, or resources-accounted for 87 percent of all variances. Nonfinancial vari- ances-which include household composition errors-accounted for 11 percent of the variances, while the “other” errors category accounted for the remaining 2 percent of the fiscal year 1988 variances. 2The Service estimated the food stamp overpayment error rate at 16.47 percent and the underis- suance error rate at 8.42 percent of its food stamp issuances. This error rate is known as the “case error rate.” Dollar overpayments and underissuances related to the case error rate are referred to as the “dollar error rate.” The Service estimated its dollar error rate for overpayments at 7.42 percent (about $763 million) and 2.63 percent (about $260 million) for underissuances. “A variance occurs when information verified by quality control reviewers is different from the infor- mation on which caseworkers made eligibility determinations or when food stamp policy has been misapplied. Because of the way the states report data, several variances can occur in a single case. The variance numbers which the Service reports reflect the relative frequency with which each vari- ance is identified. Page 3 GAO/RCED-90-183 Household Defhdtion Caseworker Error ‘T c * / B-217883 According to USDA quality control officials, the Service’sway of catego- rizing food stamp errors has been useful for its program management purposes. However, by design, the Service’sQuality Control Data Base doesnot track-and we could not determine directly from it-which errors resulted from caseworkerswho misapplied the household defini- tion. Therefore, for the purposes of this report, we created an error cate- gory which we called “household definition errors.” By definition, this category excludes all household composition errors causedby food stamp recipients. For example, if a food stamp recipient failed to report the birth of a child to the caseworker, a household composition error would exist becausethe food stamp household would contain one less person than it should. However, this error would not have been caused by a caseworker who misapplied the definition becausethe caseworker did not have accurate information on which to act. For us to classify an error as a household definition error, one of three conditions had to be met: the caseworker (1) did not acquire a piece of information that was critical to making a proper determination, e.g., he/she did not determine whether the people who lived together also purchased food and pre- pared meals together; (2) had all the information required to make a proper household determination but made an incorrect determination; or (3) had all the information required to make a proper household deter- mination but did not take any action, e.g.,the caseworker did not include a child in a food stamp household after being informed of the infant’s birth or adoption.4 The current definition of a food stamp household is complex and, Household Definition according to state officials, difficult to apply. However, caseworkers Is Not a Major Cause make few mistakes in applying the definition and thus it is not a major of Errors source of caseworker error. We selecteda probability sample” of about 2,400 issuanceswhich con- tained about 600 error casesdrawn from the Service’s fiscal year 1988 Quality Control Data Base.This data baseis used by the Service to iden- tify the sourcesof food stamp errors. After selecting our sample, we examined the quality control casefile to determine whether the caseworker had made an error in applying the household definition. The 4While household definition errors are usually found within the errors which the Food and Nutrition Service categorizes as household composition (nonfinancial) errors, we found a few household defini- tion errors listed in other error categories tracked by the Service. “A probability sample is a sample where each item in the universe has a known nonzero chance of selection. Page 4 GAO/RCED-90-183 Household Defiuition Caseworker Errora w * . R.217883 sampling technique we employed allowed us to form nationwide esti- mates of the number and value of incorrect food stamp issuanceswhich were causedby caseworkerswho misapplied the household definition.” From our examination of that sample, we estimated that, in fiscal year 1988, caseworkersmade errors in applying the household definition in about 1 percent of the food stamp issuances.We estimated that, nation- wide, between 0.4 million and 1.2 million household definition errors were made in the nearly 80 million fiscal year 1988 food stamp issu- ances.Thus, household definition errors amounted to about 2 to 6 per- cent of the 19.1 million food stamp issuanceerrors which the Service estimated were made from all sourcesthat year. The combined over/ under-issuancesmade as a result of these household definition errors amounted to between $23 million and $75 million of the $10.3 billion in food stamp benefits paid. (Seeapp. I for a detailed discussionof our methodology.) Generally, the state social service officials in the 12 states included in our sample believed that our estimate of the caseworker errors attribu- table to the household definition was representative of what occurred. They agreedthat the number was small and pointed out that caseworkerswere comfortable with applying the exceptions to the basic definition introduced by the Omnibus Budget and Reconciliation and McKinney Acts. Although someofficials would prefer a household defi- nition which was simpler to administer or was uniform for several social assistanceprograms, they pointed out that applying any household defi- nition which groups people together, and doesnot entitle each person to separate benefits, requires training and the use of judgment on the caseworkers’ part. Becauseof this, someamount of caseworker error will be associatedwith any household definition. According to these state service officials, changing the current definition would probably changethe household definition error rate. Whether the rate would increaseor decreasewould depend on the provisions of the new definition. Someofficials stressedthat constant changesin the food stamp household definition result in increasederrors until caseworkers are trained on and becomefamiliar with applying the new definition. For this reason, most officials said that they prefer to work with a stable household definition even if they do not agree with all of its provisions. “If we applied the same review procedure to all fiscal year 1988 issuances, the results would lie between the upper and lower bounds of our estimate about 19 of 20 times. Page 6 GAO/RCED-W-183 Household Definition Caseworker Errors * _ . B217883 USDA generally agreedwith our finding that caseworkersmake few Agency Comments errors in applying the current household definition. It suggesteda few minor technical changesto our draft report, which have been incorpo- rated into the text where appropriate. The full text of USDA'S comments is reproduced as appendix III of this report. Our evaluation is basedon our estimate of the frequency that caseworkersmisapply the household definition. This estimate was developed by examining a probability sample of about 2,400 issuances which contained about 500 error casesdrawn from the fiscal year 1988 Quality Control Data Base.This data basecontains information on all types of food stamp issuanceerrors. The errors are reported by the states and validated, compiled, and analyzed by the Service. These are the samedata which the Service relies on in assessingoverall food stamp issuanceerrors among the states. We also obtained the opinions of food stamp officials in 13 states regarding the impact of the house- hold definition on caseworker errors. (For more detailed information on our objectives, scope,and methodology, seeapp. I.) As arranged with your offices, unless you publicly announceits contents earlier, we plan no further distribution of this report until 7 days after the date of this letter. At that time, we will send copiesof this report to the appropriate House and Senatecommittees and subcommittees,inter- ested Membersof Congress,the Secretary of Agriculture, and other interested parties. We conducted this review between June 1989 and May 1990 in accor- dance with generally acceptedgovernment auditing standards. If you have any questions on the material in this report, please call me on (202) 275-5138.Major contributors are listed in appendix IV. Sincerely yours, John W. Harman Director, Food and Agriculture Issues Page 0 GAO/RCED-90-183 Household Definition Caseworker Errors Page 7 GAO/RCED-!jO-183 Household Definition Caseworker Errors Contents Letter 1 Appendix I 10 Objectives, Scope,and Methodology Appendix II 14 Major Exceptions to the Food Stamp Household Definition Appendix III 15 Comments From the US. Department of Agriculture Appendix IV 16 Major Contributors to This Report Table Table I. 1: Probability Sample of Food Stamp Error Cases 12 Abbreviations GAO General Accounting Office USDA US. Department of Agriculture Page 8 GAO/RCED-90-183 Household Definition Caseworker Errors , * b Page 9 GAO/RCED90-193 Household Definition Caseworker Errors ,c Appendix I * Objectives,Scope,‘md Methodology In responseto requests from the Chairman and Ranking Minority Member, SenateCommittee on Agriculture, Nutrition and Forestry; the Chairman, Subcommitteeon Nutrition and Investigations, SenateCom- mittee on Agriculture, Nutrition, and Forestry; and the Chairman, Sub- committee on Domestic Marketing, ConsumerRelations and Nutrition, HouseCommittee on Agriculture, and as modified in subsequentdiscus- sions with their offices, we agreedto provide a nationwide estimate of the food stamp error rate causedby caseworkerswho misapplied the household definition. To gain an understanding of the problems that caseworkersexperience in applying the household definition as well as the strategies that state officials have developed to reduce those problems, we talked to food stamp program and quality control officials in 13 states-Alabama, Ari- zona, Georgia, Illinois, Louisiana, Michigan, Missouri, North Carolina, New Hampshire, Pennsylvania, Texas, Vermont, and Washington. We also asked these officials for their assessmentsof the impact of the household definition on caseworker errors and asked them to comment orally on our estimate of the number and dollar value of issuanceerrors attributable to caseworkerswho misapplied the household definition. To determine the extent to which caseworker errors were attributable to the food stamp household definition, we reviewed casesextracted from the Service’sQuality Control Data Base.This data basecontains all of the different types of errors made by both caseworkersand participants but doesnot specifically identify errors that resulted from caseworkers who misapplied the household definition. Becausethe data base did not contain this information, after consulting with Service officials, we cre- ated an error category which we called “household definition errors.” For us to classify an error as a household definition error, one of three conditions had to be met: the caseworker (1) did not acquire a piece of information that was critical to making a proper determination, e.g., he/ she did not determine whether the people who lived together also pur- chasedfood and prepared meals together; (2) had all the information required to make a proper household determination but made an incor- rect determination; or (3) had all the information required to make a proper household determination but did not take any action, e.g.,the caseworker did not include a child in a food stamp household after being informed of the infant’s birth or adoption. To estimate the number of errors caseworkersmade in fiscal year 1988 benefit determinations due to misapplying the food stamp household Page 10 GAO/RCED-90-183 Household Definition Caseworker Errors Appendix I Objectives, Scope,and Methodology - definition, we drew a probability sample’ of casesfrom the Service’s fiscal year 1988 Quality Control Sample,the most recent year for which complete data were available for our review. The Quality Control Sample is a stratified sample of about 72,000 of the 80 million fiscal year 1988 food stamp issuances.The statistical-sampling technique which we used in selecting our sample allowed us to make estimates about the entire 80 million fiscal year 1988 food stamp issuanceson the basis of a sample of about 2,400 issuanceswhich contained about 500 error casesL+ To select our probability sample, we used a two-step processto draw cases.The Quality Control Sample is divided into 85 strata, someof which represent the food stamp issuancesof an entire state and others which represent only a portion of a state’s issuances.We selectedstrata for review after having assignedeach a probability of selection on the basis of the percentageof all fiscal year 1988 issuancescontained in that stratum. Thus, if a stratum contained 5 percent of the fiscal year 1988 issuances,it had a 5-percent probability of selection each time we drew a stratum for our sample. Onceselected,a stratum was replaced into the group of strata that made up the universe so that it had a chanceof being selectedmore than once. Second,we selecteda simple random sample of casesfrom each stratum chosen.The sample size was set so that 30 error caseswould be expected to be found. For example, if 20 percent of the state’s quality control caseshad errors, we would expect,a sample of 150 casesto con- tain about 30 error cases,and we chosea random sample of 150 cases.If 35 or fewer error caseswere in the stratum, we selectedall of the quality control casesrather than take a sample of them. We made 17 stratum selectionsusing this two-step method, and chosea sample of about 2,400 cases,of which about 500 were casesthat state quality control reviewers had determined contained errors, as shown in table I. 1. ‘A probability sample is a sample where each item in the universe has a known nonzcro chance of s&don. Page 11 GAO/RCED-90-183 Household Definition Caseworker Errors c * R Appendix I ObJectivea, Scope, and Methodology Table 1.1: Probability Sample of Food Stamp Error Cases Fiscal year 1999 issuance8 (in Stratum quality OAO samMe Stratum identifIcationa millions) control sample size All case8 Error Case8 Alabama ._.. 1.8 1,931 126 29 Arizona.i37j.- _ _._. I_. .,..._.-.___._... .-_.-_ 5 1,616 142 24 Georgia (02) 1.1 599 116 31 ljlinois (24) .3 376 96 31 Illinois (41) 1.3 634 129 38 Louisiana..^.-. _...... -.._---__ 2.7 1,248 112 34 Michigan (01) .-- 2.2 1,279 157 31 Michiaan’(Oi’j-- .._. ..____ 1.6 828 155 28 Missouri ._,.“.. “--.._ ._..-.-- ._ --- 1.5 2,649 272b 56b North Carolina 1.8 1,328 173 29” New Hamoshire .._._ _.‘_______... ___. _..- ._______ .l 503 193 28 Pennsylvania 4.3 1,323 178 32 Texas .- ^__(03) . . ..__- .._.-.-_ .-_- .4 99 99 20 Texas (05) .7 177 177 32 Texas (08) 1.2 177 121 29 Washmgton 1.4 2,600 156 26 Total 2,402 499 %tates with quality control samples composed of a number of strata are identified by a number in parentheses. This number is the stratum code number of the stratum we selected for review. bMissouri was selected twice (as the first and third strata selected) in our sampling procedure. There- fore, we selected two independent subsamples of 136 issuances each. The first sample contained 23 error cases and the second 33. ‘One North Carolina error case was unavailable for our review. We assumedin our review procedure that casesthat state reviewers found to be correct were actually correct. Although the number of error casescould be understated as a result, in our opinion, the amount of the understatement is likely to be small becausethe states’ quality control reviews were validated by the Service. Becausewe reviewed a probability sample of issuances,each estimate developed from the sample has a measurableprecision.2We express this precision using ranges formed by the lower and upper bounds of the 96 percent confidence interval. If we applied the samereview procedures to all fiscal year 1988 issuances,the results of such a review would lie Y ‘Our estimates were based on standard statistical formulas. (See, for example, William G. Cochran, Sampling Techniques, 3rd edition equations 11.31 and 11.36 and 2nd edition equations 11.37 and 11.39.) Page 12 GAO/RCED-90-183 Household Definition Caseworker Errors . # c u Appendix I Objectives, Scope, and Methodology between the lower and upper bounds of the confidence interval about 19 out of 20 times. Becausewe observedno household definition errors in 4 of the 17 strata selectedfor review, the ranges we formed from the con- fidence intervals for estimates related to such errors may be somewhat misstated. We are unable to statistically estimate the size of this mis- statement; however, we believe it to be small. Page 13 GAO/RCEBBO-183 Household Definition Caseworker Errors Appendix II Major Exceptionsto the Food Stamp HouseholdDefinition The household definition contained in the Food Stamp Act of 1977 gen- erally requires that people who live together and who customarily purchase food and prepare meals together be counted as a single house- hold. Related individuals such as spouses,parents and their children, and siblings who live together are consideredto be purchasing food and preparing meals together whether they do so or not. However, several exceptions allow certain personsto form separate food stamp house- holds. For example: . Parents having minor children and living with their parents or siblings can form separate householdswith their children if they purchase their food and prepare their meals separately from the relatives with whom they live. . Elderly or disabled people and their spousescan form separate house- holds from the relatives with whom they live if they purchase food and prepare meals separately. . Elderly disabled people who are unable to prepare their own meals can form householdswith their spousesseparate from relatives with whom they live if the gross income of the relatives doesnot exceed 166 percent of the poverty level. l Somehousehold members,such as college students who do not meet spe- cific eligibility requirements, illegal aliens, intentional program viola- tors, those refusing to provide their social security numbers, and those who have not complied with workfare requirements, are excluded by law from participating in the Food Stamp Program. l Roomersand live-in attendants are generally defined as nonmembersof the household with which they live. If otherwise qualified, they can form separate householdsapart from the families they live with. . Boarders-those residing with others and paying reasonablecompensa- tion for lodging and meals- are ineligible to participate in the Food Stamp Program independent of the household providing the board. They may participate as membersof the household they board with only at the request of that household. Page 14 GAO/RCED-90-183 Household Definition Caseworker Errors Appendix III CommentsFrom the U.S. Department of Agriculture United States Food and 3101 Park Center Drive Department of Nutrition Alexandria, VA 22302 Agriculture Service John W. Harman, Director Food and Agriculture U.S. General Accounting 411 G Street, N.W. Issues Office JUN 2 g 19~0 Washington, D.C. 20548 Dear Mr. Harman: We have reviewed your draft report entitled, Food Stamp Prow . Household Defbition Is Not a Maior Source Of CaSeW0rb.X Brors. RCED-90-183. In this study, State quality control files were examined to determine whether the current household definition is a significant cause of caseworker errors. The report confirms our information that few errors actually result from the household definition. On the whole, we have no problems with the report. My staff has already provided comments to yours on a few technical details. Thank you for giving us the opportunity to comment. Sincerely, Administritor Page 16 GAO/RCED-99-183 Household Definition Caseworker Errors Appendix IV Major Contributork to This Report Gerald E. Killian, Assistant Director Resources, Ned L. Smith, Assignment Manager Community, and Jerome T. Moriarty, Deputy Assignment Manager Economic Karen E. Bracey, Assistant Director Catherine T. Lojewski, Staff Evaluator Development Division, Washington, DC. John E. Stanfield, Senior Evaluator-in-Charge Atlanta Regional Johnnie E. Barnes, Site Senior Office Sherrill Y. Caldwell, Staff Evaluator Troy D. Thompson, Staff Evaluator (02321)O) Page 16 GAO/RCED-99-183 Household Definition Caseworker Errors -n- r First-Class Mail Postage & Fees Paid GAO I !l ! i
Food Stamp Program: The Household Definition Is Not a Major Source of Caseworker Errors
Published by the Government Accountability Office on 1990-07-26.
Below is a raw (and likely hideous) rendition of the original report. (PDF)