data, we selected two probability (statistical) samples12 of purchase card activities—one covering the population of purchase card activity over $50 from July 1, 2005, through June 30, 2006, totaling almost $14 billion, and a second sample covering acquisitions over the micropurchase threshold during the same time period. In our statistical samples, we included purchase card transactions from federal agencies that are required to follow the FAR, including executive departments, independent establishments, and wholly owned federal government corporations as defined by the United States Code.13 We refer to these entities as executive agencies. We excluded transactions from the legislative and judicial branches, entities under treaty with the United States, and federal agencies with specific authority over their own purchase card programs.14 We tested these samples for proper authorization and independent receipt and acceptance. Where the purchase involved the acquisition of accountable or highly pilferable items that can easily be converted to personal use, such as cameras, laptops, cell phones, and iPods, we performed work to verify that the government could account for or had possession of these items.15
To identify additional examples of fraudulent, improper, and abusive purchase card activity, we data mined purchase card transactions from July 1, 2005, through September 30, 2006. This period included an additional 3 months of data subsequent to the period included in our statistical samples. We data mined using a number of criteria, including identifying vendors of goods or services prohibited by the agency or likely to be for personal use, split purchases, year-end purchases, and pilferable and accountable property purchases, among others. We also data mined
12In a probability sample (sometimes referred to as a statistical or random sample), each item in the population has a known, non-zero probability of being selected. The results of a probability sample can be generalized to the population from which the sample was taken.
135 U.S.C. §§ 101 and 104 and 31 U.S.C. § 9101 identify agencies required to follow the FAR.
14Because of limitations in the data, we were unable to remove all transactions related to entities outside the scope of our audit from the sample populations. If any transactions that should have been excluded were selected as part of either sample, we replaced them. See app. II for further details of our scope and methodology.
15Purchase card data provided by the banks did not always contain detailed information to allow for the complete identification of accountable and pilferable property. Thus, we were unable to statistically test property accountability. Consequently, our tests of property accountability were performed on a nonrepresentative selection of property that we identified when a transaction selected for statistical sampling or data mining contained accountable or pilferable property.
GAO-08-333 Governmentwide Purchase Cards