Preliminary Cost Benefit Analysis of Identifying Overpayments BPC Misses

Issue: In its Audit of the U.S. Department of Labor's Principal Financial Statements, the office of Inspector General makes the following recommendation:

Perform and document cost benefit analysis for those overpayments identified as not being detected by the BPC, in order to determine whether or not additional resources should be dedicated in these areas.

This analysis responds to that finding.

Overpayments not detected by the States' Benefit Payment Control (BPC) activities

Our analysis determined that the following types of overpayments were least likely to be detected, and thus established for recovery, by BPC. The table shows what the Benefit Accuracy measurement (BAM) program estimated for percentages of weeks paid and dollars (in thousands) paid in error for each cause for those causes in total and those overpayments that the agency could recover.

According to BAM, in CY 2001, approximately 7% of all weeks paid were overpaid all or in part, totaling nearly $900 million in overpayments, due to the main causes that BPC does not readily detect for recovery. Of that, about $560 million, or 62%, was deemed recoverable if set up for recovery.

If agencies are to begin recovering this half a billion in now-undetected overpayments, they first need a mechanism for detecting them over and above what BPC now does. We use as our guide how the BAM investigator detected the error. This information (called the point of detection) is one of the standard codes for every BAM case that contains an error.

BAM data show that three detection points account for nearly all of the overpayments: the base period and/or separating employer; the claimant interview; and work search employers. The claimant interview is key for work search violations. Either it leads to a direct identification of the violation ($128 million in CY 2001) or it is the means by which purported work search employers are identified, and the violation is detected through follow-up contacts with them. The BAM data are summarized in the following table.

Obtaining the Additional Information: Benefits and Costs

The BAM data suggest that a reasonable approach to obtaining the information now missed by BPC is a modified Eligibility Review Program (ERP) that would involve selecting claimants at a particular point in their claim series, and performing all or some of the follow-up investigations:

• Contacting their base period/separating employers in the way BAM does to obtain base period wage and separation information;
• Either calling the claimant in for a face-to-face interview or obtaining an interview over the telephone, and
• Following up on work search contacts if warranted by state law.

Whether this makes sense for the agency to do any or all of these things depends on benefit-cost considerations. Calculations using BAM data for CY 2001 and from the time requirements for these investigations calculated during the 1990 telephone pilot project follow.

The calculations assume that the agency will conduct each investigation at the 10th week of the claimant's benefit year. At this point, approximately 23% of claimants will have left the rolls for jobs [Mathematica Policy Research, Inc.: "Left Out of the Boom Economy: UI Recipients in the Late 1990s", ETA Occasional Paper 2002-03]. Using BAM data, we calculated the incidence of all recoverable overpayment errors and those that BPC could probably not detect, for each of the detection methods, as well as the average overpayment amount detected. To do this, we had to translate BAM's results from the perspective of incidence per week paid to incidence per individual claimant reviewed, and estimate what would be discovered at investigations conducted at the 10th week. Our calculations assumed those investigations would detect every issue affecting multiple weeks (Monetary Overpayments and Separations affect every week in the claims series in most states.) However, overpayments due to weekly eligibility issues such as work search are distributed roughly evenly over the whole claim series, so we divided the BAM estimate for each issue by the average number of weeks paid per claimant in FY 2001 (13.9, rounded to 14). Further, once detected we assumed that for overpayments due to monetary and separation reasons, the agency would be able to set up for recovery the 10 weeks overpaid to the point of detection; however, the overpayment set up for the weekly eligibility issue detected would be only the overpayment for the particular week.

Costs were calculated by assuming that each review would take as long as the BAM review segment estimated in the 1990 telephone pilot study (conducted by Abt Associates), and that state UI agency employee wages rates per hour in 2001 were $28.93 (based on the FY 2001 budget's average staff cost of $52,083 per year, and assuming 1800 hours per year worked). The costs of conducting the three kinds of investigations are not only the direct cost of the investigation (e.g., about 35 minutes for each contact with each claimant's base period/separating employers to examine wage and separation issues) but also some time to prepare, obtain agency documents, locate and establish contact with the proper person, close out the investigation, etc. Cost calculations are summarized in the following table.

* These activities were included at 50% the rate found in BAM due to differences between the BAM methodology and the procedures discussed in this study.

Using the assumptions described above, the following table summarizes the cost / benefit ratios:

Conclusion

This analysis suggests that if agencies conducted a modified ERP investigation in an arbitrarily selected week in the UI claimant's benefit year, the amount identified in overpayments would be insignificant compared to the costs associated with conducting such an audit. Following up with every claimant's base period and separating employers is the most cost-effective, and that only yields $0.21 per dollar spent if all causes-including the ones BPC regularly detects-are considered. When the scope of detected issues is restricted to those BPC does not detect, that benefit-cost ratio falls to $0.06 per $1. Benefit-cost ratios for claimant interview and for the work search verification are $0.02 per $1 or below.