Quality Control of Welfare Programs

Article excerpt


John E. Rolph, The RAND Corporation


James S. Hodges, The RAND Corporation

Roderick J. A. Little, University of California at Los Angeles

Robert L. Obenchain, Glaxo Inc.

This is the fifth special section to appear in JASA and the first one in the new review section format. This special section on quality control of welfare programs addresses a public policy problem with a substantial statistical content. In contrast, special sections in the earlier format were on general topics (survey research methods, statistical graphics, computational statistics, and biopharmaceutical statistics).


The potential stakes in quality control (QC) of federal-state family assistance programs are large. In 1987, $49 billion of federal monies went to the aid of families with dependent children, food stamp, and Medicaid programs, with tens of millions of people affected. The historical QC systems in these welfare programs have engendered controversy, litigation, and Congressional interest. This controversy spawned a recent National Academy of Sciences (NAS) Panel Study with substantial statistical content because many of the issues and the policy choices hinge on the answers to statistical questions. With these ingredients, QC in welfare programs raises issues that should spark the interest of the statistical community. JASA readers will be exposed here to the views of several major contributors to this policy debate.

The first question in the policy debate about QC systems in family assistance programs is whether their historical focus on measuring the accuracy of state decisions on eligibility and benefit levels is too narrow. Should such systems target more comprehensive performance measures? Should they be broadened into full-scale quality improvement programs? Specific statistical questions that have been raised about the historical QC systems include the choice between design-based and model-based inference, measurement issues, and how error rate estimates should be used to assess financial penalties. Some proponents of change suggest using alternative statistical methods, including Bayes and empirical Bayes estimators. Statisticians interested in applications will find a variety of challenging issues to grapple with.


Kramer's article opens the section by succinctly describing the QC systems for the federal-state family assistance programs, the context for the debate over QC in welfare programs, and the competing views on the major issues. Although the other articles can be understood without reading the Kramer article, the uninitiated reader will benefit by reading it first. …