Redesigning the Sample of the Company Organization Survey Using Predictive Modeling

• Activity Number: 247
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Survey Research Methods Section
• Author(s): Matthew Thompson*+ and Chrishelle Lawrence
• Companies: U.S. Census Bureau and U.S. Census Bureau
• Keywords: non-probability sample ; logistic regression ; establishment survey

Abstract

The Company Organization Survey is conducted annually between economic census years to collect data on multi-unit companies operating in the United States. The non-probability sample consists of certainty multi-unit companies, single-unit companies suspected of being multi-unit companies, and a targeted sample of multi-unit companies. Since organizational change is likely for larger companies, multi-unit companies with 250 or more employees are selected for the sample with certainty, and single-unit companies that meet certain criteria are selected. The targeted sample aims to maximize the number of multi-unit companies with less than 250 employees with organizational changes. Organizational change occurs when at least one of a company's units is opened, closed, moved, or sold. This research focuses on revising the sample for multi-unit companies. A logistic regression model was used to predict the likelihood of an organizational change. Results suggest that an improved sample could be obtained when multi-unit companies with more than 500 employees are selected with certainty, and companies with less than 500 employees are targeted by their likelihood of organizational change.