be brought to bear on improving the adminis- He went on to say that: * once the issue of 'fiscal sanctions' was raised, it became the burning preoccupation of almost all involved federal and state officials. This has resulted in a misplaced and harmful emphasis on the value of fiscal sanctions and a corresponding neglect of positive assistance. These concerns are not new; when sanctions were first proposed, many concerns were raised about their impact on State-Federal relationships. During hearings before the Subcommittee on Oversight of the House Committee on Ways and Means in 1975, the Subcommittee Chairman quoted several States' officials who had such concerns before sanction regulations were finalized in 1973. For example, the Chairman of the North Dakota Legislative Council stated: "Proposed regulations [sanctions] would strike The Director, Institutions, Social and Rehabilitative Services, in Oklahoma warned that sanctions could do irreparable harm to the Federal-State partnership. A group of officials from New Mexico, West Virginia, New Jersey, and North Carolina said: "It would appear obvious that justice requires Of the States in our review, California and Indiana officials favored sanctions but said they could have some negative effects on error reduction in high error rate States. Officials from Maryland, New York, Hawaii, and Maine expressed opposition to sanctions. Several reasons given for not favoring sanctions were: --States will be encouraged to alter their programs so that various errors are no longer defined as errors. --QC error rates are not comparable because of QC review procedural differences and AFDC program differences. --The Federal-State working relationship is negatively affected. Some officials told us that sanctions can negatively affect the potential of QC for identifying errors. For example, one State official said sanctions would undermine the integrity of the QC process in that the reviewers may be less vigilant in identifying errors. QC ERROR RATES NOT SUFFICIENTLY Any basis for sanctions should equitably measure the administrative quality of a State's AFDC program in terms of the extent to which errors that affect payments are made. We question whether current QC system error rates reflect the same proportion of erroneous payments in all States because they do not take into account three kinds of differences unrelated to program administration. First, the statistical precision of QC error rates varies among States with some States having greater fluctuations in reported error rates because smaller samples of cases are reviewed. Second, the QC error rates can be affected by State program and policy differences. Finally, differences in the way QC reviews are made (discussed in ch. 3) can affect QC error rates regardless of how efficiently and effectively a State administers its program. Statistical precision of State QC error rates vary The statistical precision of the QC system error rates varies among the States. This raises the question of whether the QC error rates are comparable and adequate for administering sanctions. For The reported QC error rate is an estimate derived from a sample. Such estimates have ranges which vary due to several factors, the most important being sample size. example, if a State's reported error rate is 3 percent based on a QC review of 1,200 cases, 95 times out of 100 the State's "true" error rate will lie between 2.04 and 3.96 percent. However, if the reported error rate is based on a sample of only 150 cases, the State's "true" error rate will lie between 0.27 and 5.73 percent. Because error rate precision differs so much between large and small AFDC caseload States, the QC system does not appear to provide equitable criteria for sanctions. The lack of precision of error rates in small caseload States increases the likelihood that the decision to sanction will be based on chance. For example, a State with a 150-case sample and a 10 percent "true" error rate would have to reduce its true error rate by 4.03 percentage points for the reduction to be statistically significant. That is, the change in the reported error rate is due to an actual reduction in error rates and not just sampling variation. States with a 1,200-case sample and a 10-percent error rate would need an error rate change of only 1.42 percent to be statistically significant. State QC error rates do not clearly show the effects of State policy and program differences Federal regulations give States much latitude in determining the way they operate their AFDC programs and the elements of assistance in those programs. Some of the policies that States adopt can increase the potential for error. Also, some States provide benefits to categories of recipients that other States do not. QC error rates do not clearly show the effects of these differences. Therefore, QC error rates are not completely satisfactory for comparing the quality of States' program administration that is needed for equitably applying sanctions. The following are examples of program differences that affect States' error rates and their comparability. Treatment of basic need standard The basic need standard is one factor, along with recipient income and the State payment limit, that is considered in computing a recipient's grant amount. Traditionally, most States established the level of need based on the actual cost incurred by an applicant for such basic needs as food, clothing, and shelter. Currently, need is established using either consolidated or partially consolidated standards. For example, Maine uses a consolidated standard in which all basic needs--food, shelter, clothing, utilities, household supplies, personal care items, and recreation--are included in a standard allowance which varies according to family size. New York uses a partially consolidated need standard in which all basic needs except for shelter and utilities are included in the standard allowance, which varies according to family size, while shelter and utilities are added at cost up to a maximum amount. Consolidated need standards have generally been considered less error prone because eligibility workers need not keep track of changes in individual living expense items. Contrasting the error rates of Maine with a consolidated standard and New York with a partially consolidated standard indicates the effect of full consolidation on errors. In the January-June 1978 review period, Maine had only two errors related to basic needs in its 603-case sample. New York, however, had 101 such errors in its 1,238-case QC sample in the same period. A 1977 study by Touche, Ross and Company indicated that consolidation of standards has been a major factor in reducing AFDC error rates in the past 5 years. The firm found that, for 10 of 11 States examined that used consolidated standards, a subsequent decline occurred in payment error rates for the "basic needs" category after adopting consolidated standards. In eight of the States, the reduction was more than 70 percent. For example, Illinois reduced its payment error rate for basic needs by 93 percent, from 1.4 to 0.1 percent, and Louisiana experienced a reduction of 99 percent, from 1.5 to 0.01 percent. Effect of need/payment on error rates The effect of mistakes made in calculating aid payments on error rates can depend upon whether a State has a payment standard that equals or is less than its need standard. For example, some computation mistakes made by Indiana eligi bility workers are not considered errors, while the same mistakes by eligibility workers in New York are, and affect New York's error rate. The following cases demonstrate this. In New York, an eligibility worker calculates that a family of four is eligible for the maximum need allowance of $476. 1/ Since New York pays 100 percent of need, the family receives a $476 grant. Later, a QC reviewer finds that the eligibility worker had made a $15 mistake so that the correct need should be $461. As a result, the correct grant amount should also be $461, and not $476; therefore, a $15 error is reported in New York. In Indiana, an eligibility worker calculates that a family of four is eligible for a grant equal to the maximum adjusted need standard of $327. 2/ However, the most Indiana will pay to a family of four is $275 1/--therefore, the family receives a $275 grant. Later, a QC reviewer finds that the eligibility worker had made a $15 mistake so that the correct adjusted need should be $312. Since $312 is still higher than the $275 maximum grant, the family is still only eligible for $275. Because the payment is not affected by the $15 mistake, there is no error reported in Indiana. For the April-September 1978 review period, mistakes in 42 cases were not considered errors by Indiana QC because the mistakes did not affect maximum payments. Indiana had reported 73 cases in error in its 1,204-case sample for that period. Categories of potential recipients Under the AFDC program, States can at their option provide assistance to --children 18 to 21 years of age who are regularly attending a school, college, university, or course of vocational or technical training; 1/As of April 1, 1978. 2/This standard, against which income is applied, is 90 percent of the full need standard. |