Fielding a Targeted Address Canvassing Operation: Alternative Approaches to Moving from Predictive Statistical Modeling to a Cost-Effective Address Canvassing Field Operation

• Activity Number: 648
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Government Statistics Section
• Author(s): John Luthor Boies*+ and Christine Gibson Tomaszewski
• Companies: U.S. Census Bureau and U.S. Census Bureau
• Keywords: Targeting Address Canvassing ; 2020 Census ; Statistical Modeling ; Address Listing ; Census Operations

Abstract

To date substantial effort researching how to use statistical modeling to reduce the costs associated with updating the master address file for the 2020 Census has yielded results suggesting that applying statistical modeling approaches to targeting geographic areas for selective address canvassing before the 2020 Census could generate significant cost reductions. However, the problem remains on how to move from good quality statistical models to an actual field operation. This paper details a range of alternative methods to operationalize statistical predictions into a cost effective field operation. The development of a cost/benefit continuum to array blocks into categories for different kinds of address updating, applying different geographic units to the problem, and linking workload determination to modeling outcomes will be explored.