Development of a UI payment accuracy and integrity measure

EMPLOYMENT AND TRAINING ADMINISTRATION ADVISORY SYSTEM
U.S. DEPARTMENT OF LABOR
Washington, D.C. 20210
CLASSIFICATION: UI
CORRESPONDENCE SYMBOL: OWS/OCTA/DPM
DATE: July 31, 2002
RESCISSIONS: None stated
EXPIRATION DATE: July 31, 2003
ADVISORY: UNEMPLOYMENT INSURANCE PROGRAM LETTER NO. 33-02
TO: ALL STATE WORKFORCE AGENCIES
FROM: GRACE A. KILBANE, Administrator, Office of Workforce Security
SUBJECT: Development of an Unemployment Insurance (UI) Payment Accuracy/Integrity Measure

1. Purpose. To solicit comments from State Workforce Agencies (SWAs) on a proposed operational definition of UI overpayments, and options for a new measure under the Government Performance and Results Act (GPRA) for assessing payment accuracy/integrity.

2. References. Unemployment Insurance Program Letter (UIPL) 27-02 (May 6, 2002); and The President's Management Agenda, Fiscal Year (FY) 2002 available on the Internet at http://www.whitehouse.gov/omb/budget/.

3. Background. Two initiatives - GPRA and the President's Management Agenda - provide the impetus for development of a UI accuracy measure.

GPRA seeks to shift the focus of government decision-making and accountability away from a preoccupation with the activities that are undertaken to a focus on the results of those activities. For FY 2002 and beyond, the Employment and Training Administration (ETA) has refined its UI goals to four, each with one indicator. It has also changed the focus from the number of states attaining a criterion to aggregate performance, to better measure system-wide service to customers. The four goals are: to improve payment timeliness; improve payment accuracy; facilitate reemployment of UI claimants; and improve the timeliness of new status determinations. ETA is using the timeliness of intrastate first payments as the indicator for payment timeliness, and the timeliness of new status determinations for the status determinations performance goal. Both are UI PERFORMS Tier I measures. It has promised to develop measures as indicators of the other two goals during FY 2002.

The President's Management Agenda for FY 2002 provides a strategy for improving the management and performance of the federal government with the emphasis on performance and results. One of five government-wide initiatives, "Improved Financial Performance," calls for establishing a "baseline of the extent of erroneous payments." Federal agencies are expected to include in 2004 budget submissions information on erroneous payment rates, including actual and target rates, where available, for all benefit and assistance programs over $2 billion, which include UI. Using this information, the Office of Management and Budget will work with agencies to establish goals to reduce erroneous payments for each program.

This UIPL discusses a process through which the UI Program will respond to both GPRA and the President's Management Agenda by developing an appropriate indicator for the goal of improving benefit payment accuracy.

4. Overview. Since 1988, ETA has published a measure based on the Benefit Accuracy Measurement (BAM) program that provides a comprehensive estimate of all overpayments in the UI system. Its standard is each state's own eligibility requirements as embodied in its law and policy. Although BAM provides statistically valid and representative estimates, the variation from one state to another reflects more the differing potential for error due to more or less stringent eligibility requirements than the quality of the states' benefit payment administration. Many of the errors are "technical" in nature. Many also do not tie well to the other source of information on overpayments, which are data on the amount of actual overpayments detected and recovered through ongoing Benefit Payment Control (BPC) operations. For example, BAM estimates include several categories of overpayments such as failure to meet work search requirements, that are not readily detectable, or cost-effective to detect, through BPC operations. They also include overpayments that the agency cannot recover and thus lie outside the scope of BPC.

For both GPRA and the President's Management Agenda, ETA seeks to develop a measure that ties better to the realities of program operations, including BPC operations, and reflects better the general understanding of "overpayments" than the standard BAM rate. This measure should thus reflect the underlying level of overpayments in the system while not being adversely affected by initiatives designed to improve integrity by increasing the detection and establishment of overpayments. For example, ETA is encouraging states to aggressively use data on new hires, from their own State Directory of New Hires and, if access can be obtained through legislation, from the National Directory of New Hires. These data would be used both to inform BAM estimates of claimants who have returned to work while continuing to claim benefits, and reduce the size of such overpayments by limiting their duration. The short-term result, suggested by states piloting the use of new hires data, could be the seemingly perverse effect of a rise in the overpayments BAM estimates, despite an increase in overall program integrity.

After conducting an analysis of both BAM and BPC data, ETA developed the three potential measures of accuracy and integrity described below. As the discussion indicates, none perfectly meets the ideal criteria for an indicator for this goal, and thus none is recommended at this time. This UIPL, and other consultation with states and stakeholders, are intended to help all involved develop a better understanding of the difficulties inherent in quantifying improvements in UI benefit payment accuracy. The comments will help guide development of a measure, either by offering refinements to one or more of the measures below, or by offering other options. To assist in the review, we identify what may be seen as advantages and disadvantages.

5. Proposed Measurement Options.

A. An "operational overpayment rate" derived from BAM data that includes all fraud and most non-fraud recoverable overpayments. Starting from the BAM "Annual Report" rate, and analyzing the various categories of errors, we developed a measure that includes only overpayments that states could/would recover, both fraud and non-fraud. Three major categories of errors are excluded: failure to meet work search requirements; failure to meet active employment service registration requirements; and base period wage errors. State laws and policies vary widely with respect to the first two categories, which thus contribute to high state-to-state variation in overall measured error rates. Work search violations are generally not cost-effective to detect either during the payment process or after payments have been made. The measure excludes base period wages errors (most due to employer reporting) because BPC can rarely detect them. The payment error rate using this new operational definition was 5.2% in calendar year (CY) 2001, and its five-year average was 4.9%. The proposed operational rate is in the same range as reported error rates for other entitlement programs (e.g., Food Stamps, 6.5%; SSI, 5.5%; and TANF, 6.1%.) For CY 2001, the standard published BAM overpayment rate was 8.2%.

Advantages

• It ties closely to program operations by excluding payments BAM considers errors but states' laws and policies preclude them from recovering, e.g., the agency judges the payment "technically" proper, or recovery would cause hardship.
• It excludes types of error that cannot be detected cost-effectively through program operations, and about which state laws vary considerably (e.g., work search and dependency violations). It thus levels the state error-rate playing field.
• It includes the types of errors policymakers and the public perceive as "overpayments," e.g., due to misrepresenting reasons for separation and continuing to claim benefits after returning to work.

Disadvantages

• As states incorporate data on new hires in BAM, improving error detection, states' error rates will likely rise.
• The recommended error rate is only about 60% of the familiar BAM "Annual Report" error rate that we have published for 13 years.

B. The ratio of fraud and non-fraud overpayments established for recovery (reported on the ETA 227 report of BPC activities) to dollars paid. Taking the ratio of reported overpayments established to benefits paid produces an accuracy rate that ties directly to state benefit payment operations. The rate is also based on actual overpayments, not an estimate. Progress toward achieving the GPRA goal would be articulated as an increase in this ratio of overpayments established. For 2001 this ratio was 3.0%. Its five-year average was also 3.0%.

Advantages

• It would give states an incentive to increase their efforts to detect and establish overpayments.
• The changes ETA is urging states to make-especially to use the new hires directories and better tools to detect fraudulent use of Social Security numbers-should raise overpayments established.

Disadvantages

• Overpayments established by BPC are not representative of all UI overpayments because they depend on states' ability to detect them, the resources they devote to BPC, and often cost-benefit determinations.
• Interpreting state-to-state differences in this ratio, and within-state changes in it over time, will be difficult, because the ratio reflects a combination of BPC effectiveness, the resources states devote to BPC, and also the underlying level of recoverable overpayments.
• The denominator of this measure will be sensitive to changes in the overall economy, and this may distort short-term movements in the ratio.

C. The ratio of estimated recoverable overpayments that are established for recovery through the BPC process to the estimated dollars of the "operational overpayment" rate from BAM. One way to clarify the meaning of the level of BPC establishments is to take them as a ratio to an estimate of the underlying level of recoverable overpayments that BPC can be expected to establish for recovery, instead of total payments. The estimate of dollars overpaid using the operational overpayment rate (Option A, above) is our best estimate of the underlying universe that BPC seeks to penetrate. Thus, the numerator of this ratio is the sum of fraud and non-fraud overpayments established, as reported quarterly on the ETA 227 report of BPC activities. The denominator uses the operational error rate described in Option A, above, i.e., the total fraud and non-fraud recoverable overpayments estimated from BAM data. The denominator is for a period 6 months earlier than the numerator, to allow for lags in detection through crossmatch activities, a major BPC detection tool. Progress toward achieving the GPRA goal would be articulated as an increase in this ratio. This ratio was 57% in 2001; for the past five years it averaged 61.5% and ranged from 57% to 65.4%.

Advantages

• It gives the right incentives-both to prevent overpayments when possible and to increase efforts to detect and recover those that cannot be prevented.
• The operational overpayment rate that is used to determine the denominator reflects the universe of overpayments that states are most likely to be able to prevent or detect.

Disadvantages

• Even though the goal focuses on improving overpayment detection as well as prevention, program evaluators may continue to focus on overpayment rates and criticize the program when overpayment rates rise as more states use new hires data.
• To some extent, the level of benefit payments affects the goal ratio. When benefit payments are very high, the BAM estimated error rate is applied to a larger total and produces a higher dollar estimate of overpayments. During high benefit payout times, the level of staff resources available for overpayment detection and recovery is likely to remain about the same-therefore the dollar level of detections and establishments may not rise at the same rate as the overpayment estimates.
• For both CY 2001 and for the CY 1999-2001 average, five states' ratios exceeded 100% and ranged from 101% to 136%. Although the causes of such high ratios remain to be examined in detail, they suggest a combination of very active BPC programs, and BAM programs which underestimate the true universe of recoverable overpayments detectable through BPC operations. It is known that the BAM methodology tends to underestimate errors due to excessive earnings during a week claimed. BAM cases must be completed within 90 days, which does not leave BAM time to use the wage record crossmatch. In 2001, ETA began instructing state BAM units to use the more quickly available new hires data to identify such violations.

6. Projected Timeline. After development of a measure, ETA will establish a baseline and a new GPRA goal from which to measure improvements in payment accuracy nationwide. The planned completion date for work on the goal is September 30, 2002.

7. Action Required. SWA administrators are requested to provide comments on options for payment accuracy/integrity measures presented in this UIPL, and to suggest improvements or additional options. Please submit comments by September 1, 2002. All comments should be mailed or faxed to:

• Grace A. Kilbane, Administrator Office of Workforce Security
  U.S. Department of Labor
  Room S4231
  200 Constitution Ave., NW
  Washington, D.C. 20210
  Attention: Andrew Spisak

Fax number: 202-693-3229

E-mail comments will be accepted at aspisak@doleta.gov.

8. Inquiries. Inquiries should be directed to your Regional Office.