The sample included two categories of applications: those which automatically qualify due to "categorical eligibility (by participating in certain public-assistance programs or meeting an approved designation, such as foster children)", and those which qualify based on self-reported household size and income.
Of the six applications that indicated categorical eligibility, half were either completely ineligible (2) or qualified for reduced-price meals instead of free. These type of applications are not even subject to standard verification. The GAO recommended that the USDA conduct sample verification such as the one conducted by the GAO to help prevent this problem.
Of the remaining nineteen applications, the GAO found that nine were ineligible, and only two of the nine could have been verified as ineligible under standard USDA procedures. The GAO suggests that using computer matching with external income data (state payroll records) could help in weeding out participants who do not qualify despite the self-reported information that resulted in the initial approval.
A chart included in the report illustrates the relatively small window of income variation ($1,200) from the established thresholds that subjects an application to standard verification:
The report noted that "[t]he Office of Management and Budget (OMB) has designated the NSLP [National School Lunch Program] as 1 of 13 federal 'high-error' programs due to its large estimated improper payments—approximately $1.8 billion in fiscal year 2013," for an error rate of 15.7 percent. The School Breakfast Program (SBP) error rate was even higher at 25.3 percent for an estimated $831 million in improper payments in fiscal 2013.
As of this date, the USDA has not responded to the GAO regarding the recommendations made in the report.
Note: A version of this article first appeared at The Weekly Standard.