We investigate the impact of the mixed attribute GAAP measurement model on accrual quality inferences. GAAP rules vary from an income statement “matching” focus to a balance sheet “fair-value” focus. Accrual properties are also affected by the business activity being measured and the activity's recurrence. Furthermore, accrual measurement is affected by managerial estimation error/manipulation. As a consequence, accruals are heterogeneous with predictably different statistical properties. We construct a dataset of more than 100,000 accrual items that Compustat aggregates into data item FOPO. We classify these accruals into 32 types over a 21-year span. We show that variation in persistence for the 32 accrual types is consistent with our predictions concerning GAAP rules. We provide a model of accruals and show that it is implausible for estimation error of reasonable magnitudes to explain this variation. Finally, we develop a measure of reporting consistency and show its impact on accrual quality inferences.