Sampling From an Unreconciled Benefits Data Extract
While auditing the Social Benefits Payment Center, the auditor obtains a data extract of 48,000 monthly benefit payment records for 2024 from the legacy payment system. The system cannot produce control totals, and the extract has 5,760 records with duplicate recipient identifiers, some of which appear to be legitimate repeat payments. The ledger shows total benefits of 9.3 billion tenge, while the extract totals 9.1 billion tenge; the IT department says the difference is ‘timing.’ The auditor plans to select 120 records using random sampling and test eligibility documents. The audit manager is concerned that the sample will not support a conclusion on overpayment and that the data may be incomplete. The audit is scheduled to finish in three weeks, and the Center’s management wants the draft report before the weekend.
1. Assess whether the data extract is sufficiently reliable for sampling and describe the procedures you would perform before selecting the sample.
Read the worked answer
The extract is not sufficiently reliable yet. The missing control totals, the 200 million tenge gap, and the duplicate identifiers are red flags. Before sampling, reconcile the extract to the general ledger and treasury payment reports by month, benefit type, and region. Obtain a written explanation of ‘timing’ and test it. Check record counts and hash totals, review the query logic, and test the duplicate identifiers to separate legitimate repeat payments from true duplicates.
If the gap cannot be explained, do not project a precision-based estimate. Instead, use the data for risk targeting, select a judgmental or stratified sample, and disclose the limitation. Sampling from an unverified population can produce a conclusion that is not supportable.
2. Design a sampling approach that addresses the duplicate identifiers and the reconciliation gap, and explain how you would project the results.
Read the worked answer
Stratify by benefit type, amount, region, and duplicate status. Take a random sample from the verified population for eligibility testing and a targeted sample of all high-risk items: duplicates, payments near approval thresholds, and records in the reconciliation gap. For the random sample, calculate the sample error and project the exception rate to the population using a recognized method. For the targeted sample, report results separately and do not project them as a random estimate.
If duplicates are confirmed, quantify the overpayment and test whether it is recoverable. Treat the 200 million tenge gap as a separate population until reconciled; audit the reconciling items individually if their number is manageable.
3. How should the auditor document the data limitations in the working papers and report?
Read the worked answer
Document the source system, extraction date, query, record counts, control totals, reconciliation, explanations received, duplicate analysis, sample design, and limitations. In the report, state that the extract could not be fully reconciled, describe the effect on the audit conclusion, and avoid an unqualified projection. If management’s explanation is unsupported, report it as an information-system control finding. Keep the working papers so that another auditor can reproduce the sample.