Benefit Accuracy Measument Program (CY 1999)

The Benefit Accuracy Measurement (BAM) program (before 1996 called Benefits Quality Control) is designed to determine the accuracy of paid intrastate claims in three major unemployment compensation (UC) programs. Trained investigators reconstruct the unemployment insurance (UI) claims process for samples of weekly payments and verify the information on which the payments were based.

For claims that were overpaid or underpaid, BAM determines the amount of benefits the claimant should have received, the cause of and the party responsible for the payment error, the point in the UI claims process at which the error was detected, and actions taken by the agency and employer prior to the payment that is in error.

Coverage

BAM covers the three largest permanently authorized UC programs: State UI, Unemployment Compensation for Federal Employees (UCFE), and Unemployment Compensation for Ex-Service Members (UCX). These three programs account for over 92 percent of the UC benefits paid. BAM data are available for the 50 States, the District of Columbia, and Puerto Rico from January 1988 through the present. BAM does not estimate all underpayments, because it samples only paid UC claims. BAM plans to begin including denied UC claims in June 2001.

Sample Design

State BAM samples are randomly selected from the population of State UI, UCFE, and UCX payments issued by the State each week. BAM refers to this weekly sampling interval as a batch. Each batch begins at midnight Sunday and runs until 11:59 p.m. Saturday. BAM records the number of paid claims and the amount of benefits paid in the population from which the sample was selected so that the sample data can be weighted to make inferences concerning the population.

Sample Sizes

Before 1997, BAM sample sizes ranged from 400 to 1800 cases per year per State. Since 1997, allocated sample sizes range from 360 cases per year in the 10 States with the smallest UI workloads to 480 cases in the remainder of the States. Several States have chosen to select larger samples.

Database

The BAM database includes about 110 data elements for each sampled payment. Data for 15 of these elements are captured both before and after the BAM investigation, and 8 are completed only for payments containing errors. Aggregate data for each batch are collected for an additional 42 data elements, most of which are demographic characteristics of the sample and population.

Methodology Considerations

Estimates based on BAM data are subject to the usual sampling and non-sampling errors that can affect survey data. BAM has implemented several quality assurance procedures to minimize non-sampling errors, such as incomplete or improperly defined sampling frames, errors of interpretation, and data entry errors. Nonresponse bias is not significant. Nationally, BAM program staff gather sufficient information from claimants, employers and third parties to complete their investigations for over 99 percent of the paid claims that are sampled; sample case completion rates are 100 percent in most States. When the program began, all BAM verifications were done in person. Since 1993, investigators may use telephone, mail, e-mail and fax methods as well. Studies have shown that although such methods yield somewhat less information than in-person contacts, the overall payment accuracy rate estimates are not significantly affected.

The BAM program retroactively investigates the accuracy of the paid claim's monetary and separation determination, as well as all information relevant to the claimant's continued eligibility, such as the claimant's efforts to find suitable work, ability and availability for work, and earnings from casual employment or other income sources, such as pensions. BAM also records demographic, UI program, and labor market data on each claimant. BAM does not maintain longitudinal data on the claimant's benefit history subsequent to the compensated week sampled.

Although claimant characteristics can be inferred from the data, it is important to keep in mind that the BAM sampling frames consist of payments. Claimants have an increased chance of selection to the BAM samples the longer they remain in the UI system and are paid benefits. Estimates of claimant characteristics that are correlated with duration of receiving benefits are subject to bias unless they are adjusted to take into account the claimant's chance of sample selection.

Published Findings

The Department of Labor has published State BAM data and analyses annually since 1988. The Unemployment Insurance Benefits Quality Control Annual Report was published from 1988 to 1995; 1996 data are in the Unemployment Insurance Benefit Accuracy Measurement Annual Report. Since 1997 BAM data are part of the UI PERFORMS Annual Report, which also includes data from the Benefit Timeliness and Quality program and the Tax Performance System.

Calendar year 1999 BAM program data are summarized in the following table. Overpayment and underpayment rates are estimates and do not represent the actual percentages of errors detected in the population. These estimated rates are subject to sampling and nonsampling errors. The 95 percent confidence intervals for each State's overpayment and underpayment rates equal the estimated error rate plus and minus the value indicated in the column (+/-). The true error rate is expected to lie within this interval for 95 percent of the samples of the size selected by the State. The 25,766 UI payments for which data were collected represents a completion rate of 99.9 percent of the cases sampled in 1999.

A compendium of this data is available in the CY 1999 UI PERFORMS Annual Report. This report also includes data from the Benefit Timeliness and Quality and Tax Performance System programs. Printed versions of the report are available through the National Technical Information Service (NTIS) (Internet: www.ntis.gov). Orders through the NTIS may be placed by telephone (1-800-553-NTIS (6847) or 703-605-6000), fax (703-605-6900), e-mail (orders@ntis.gov), or mail (NTIS, 5285 Port Royal Road, Springfield, Virginia 22161). The printed version does not include optional State narratives which address the BAM results.

Readers are strongly cautioned that it may be misleading to compare one State's BAM overpayment and underpayment rates with the rates of other States. No two States' written laws, regulations, and policies specifying eligibility conditions are identical, and differences in these conditions influence the potential for error. States with stringent, complex provisions will tend to have higher overpayment rates than those with simpler, more straightforward provisions, for example. No performance standards for UI benefit accuracy have been established.

Contacts

To obtain further information about the BAM program and the use of its database, please contact:

* In millions of dollars