Investigation of Macro Editing Techniques in a Production Setting*Roberta Gibson, US Census Bureau
Katherine (Jenny) Thompson, U.S. Census Bureau
Keywords: macro editing, Hidiroglou-Berthelot edit, minimum volume ellipsoid
Even though economic programs at the U.S. Census Bureau dedicate a high proportion of their processing schedule to micro-data review and correction, program managers still must systematically review tabulations prior to publication to ensure that the published levels reflect changes due to the economy and are not attributable to uncorrected data collection errors. For many programs, these macro-review procedures are quite sophisticated. However, their effectiveness is largely dependent on subject-matter expertise.
This research investigates the usage of automated procedures that implement statistical algorithms to identify unusually large or small tabulations (macro editing), followed by the application of separate analysis tools that examine the contributing micro-data in the identified tabulations. Using empirical data from the Quarterly Financial Report (QFR), we evaluate the performance of two macro-editing algorithms and then examine the effectiveness of two different measures of influence to identify the micro-level contributors to the outlying estimates. We conclude with recommendations and ideas for future research.