Quality control (QC) reviews of Food Stamp Program (FSP) cases show that error rates across States range from less than 5 percent to more than 25 percent when both overpayment and underpayment error are combined. This study uses QC data for 1997-2001 and a Markov probability framework to characterize year-to-year national and State error rates into variations due to errors occurring with first-month cases (those approved at initial certification); ongoing cases (those subject to the interim action process); and expiring cases (those subject to the recertification process). This information can be used in planning corrective actions by focusing attention on phases of the administrative process that are more responsible for errors. This study also explores the effect on payment accuracy and FSP participation of more frequent recertification of food stamp cases. A motivating concern is that the use of short certification periods (3 months or less) as a strategy to reduce case error may unintentionally reduce program participation.