Regression Estimates in Federal Welfare Quality Control Programs
Hansen, Morris H., Tepping, Benjamin J., Journal of the American Statistical Association
The purpose of this article is to examine the statistical properties of the sampling and estimation procedures used in the Aid to Families with Dependent Children (AFDC) and Food Stamp quality control (QC) systems for estimating overpayment error rates and the statistical implications of the use of the overpayment error rates to assess liabilities of the states for overpayment errors that exceed what have been designated by the Congress as allowable tolerances. This article summarizes some of the results presented in reports on the statistical implications of the quality control programs in AFDC and in the Food Stamp program prior to changes in the Food Stamp program introduced by legislation in 1988. Along with other changes, the legislation changes the focus to the sum of the overpayment and underpayment error rates. We believe that the principal conclusions drawn in this article hold also under the 1988 legislation. Unless otherwise specified, however, all references in this article to the QC system refer strictly to pre-1988 conditions.
The Food and Nutrition Service (FNS), Department of Agriculture, is responsible for national administration and management of the Food Stamp program. The Office of Family Assistance (OFA), Department of Health and Human Services, has similar responsibility for the AFDC program. Each state is responsible for administration within the state and for day-to-day operation of these programs, including accepting applications for benefits, investigating and determining eligibility of applicants, determining the amount of allotments to applicants determined to be eligible, and issuing benefits. Ordinarily the Food Stamp program is administered within a state by the same state and local agencies that are responsible for administering the AFDC program. In the case of the Food Stamp program, the benefit costs are paid by the federal government. Only the costs of administration within a state are shared by the state and the federal government; usually, each pays approximately half. For AFDC all costs are shared.
The QC systems are prescribed by law and by agency regulations and are in operation in each of the states. In these programs, each state does an intensive review of a monthly sample of active cases. The state review is an intensive re-investigation of each active sample case to determine the eligibility of the case and the amount of the benefit or allotment for the case. With certain exceptions (the exceptions include cases that are defined to be out of scope for various reasons), active cases are households issued benefits for the sample month. Efforts are made by the federal agencies to have the states complete the reviews of substantially all of the cases selected for the state sample. Generally, the observed completion rates for in-scope cases are relatively high. In Food Stamp QC, for example, the completion rate is usually more than 95% and, quite often, is 98% or more.
The overpayment error rate is equal to the estimated average overpayment divided by the estimated average payment. The overpayments include both payments to eligible cases that are higher than what they should be and payments to ineligible cases. An underpayment error rate is also computed for cases receiving benefits, defined as the ratio of the average underpayment to the average payment. Case error rates are also estimated. These include the proportion of active cases that are ineligible, the proportion with overpayments, and the proportion with underpayments. In addition, samples of terminations and denials are reviewed, and the proportion of cases with improper terminations and the proportion of applications that are improperly denied are computed.
The federal regional offices review and approve the states' sampling plans and monitor the QC activities. In each program a subsample of the QC review sample of active cases for each state is drawn by the regional office and is given an intensive dependent re-review. …