306 – Time Series Methods for Environmental Data
Revisions Revisited: Data-Driven Approaches for Detection in Quarterly Financial Report Macro-Level Data
Laura Bechtel
U.S. Census Bureau
Gregory Cepluch
U.S. Census Bureau
Melissa McDaniel
U.S. Census Bureau
The Quarterly Financial Report (QFR) program investigates statistical methods for identifying substantial macro-level revisions of the income statement and balance sheet data. Currently, macro-level relative revisions are identified as suspect if the absolute values are above a defined threshold, which is determined by subject matter expertise. In this paper, process control methodologies are explored to detect substantial revisions in the data, specifically p-charts and stair-step charts. The inputs necessary for these control charts are not readily available for revision estimates. As a result, a focus is placed upon estimating the control chart parameters. Once these parameters are developed, various evaluation diagnostics are employed to assess their validity. Finally, the performances of the new and existing revision identification methodologies are compared.