The Synthetic Public-Use Release of the Longitudinal Business Database

• Activity Number: 588
• Type: Topic Contributed
• Date/Time: Thursday, August 6, 2009 : 10:30 AM to 12:20 PM
• Sponsor: Section on Government Statistics
• Author(s): Satkartar Kinney*+ and Jerome Reiter and Ron Jarmin and Arnold Reznek and Javier Miranda and John M. Abowd
• Companies: National Institute of Statistical Sciences and Duke University and U.S. Census Bureau and U.S. Census Bureau and U.S. Census Bureau and Cornell University
• Keywords: Data Confidentiality ; Synthetic Data ; Economic Data ; Multiple Imputation

Abstract

Longitudinal business data are widely desired by researchers, but difficult to make available to the public because of confidentiality constraints. We discuss the generation and use of the soon-to-be (or recently) released synthetic public use data sets of the U.S. Census Bureau's Longitudinal Business Database. The data sets contain actual units with most variables replaced with multiple imputations, i.e., values generated from probability distributions fit using the confidential data. This can protect confidentiality, since attributes are synthetic rather than real. And, when the models describe the data well, broad-scale inferences from the synthetic data sets will be inferentially valid.