Regency C
Innovative Approaches to Reduce Census Nonresponse Follow-Up (303941)
Yuliya Romanyuk, Statistics Canada*Monique Sidebottom, Statistics Canada
Keywords: machine-learning, admin data, Statistics Canada, occupancy, census, modeling, administrative data
A key objective for Statistics Canada is modernizing its collection methods and modelling techniques. Much effort is being put into the integration of administrative data and the use of machine-learning (ML) methods to address various aspects of data collection. The Canadian Census of Population is the largest and one of the most important surveys undertaken by Statistics Canada; as such, it is a prime candidate for maximizing the efficiencies resulting from the use of innovative approaches.
One aspect of the strategy to reduce census Non-Response Follow-Up (NRFU) aims at lowering response burden and collection costs by using admin data and ML methods to better focus the effort of NRFU staff. In particular, NRFU can be greatly reduced for dwellings believed to be unoccupied or nonexistent. The Dwelling Occupancy Model (DOM) was developed in order to predict the occupancy status of dwellings by using a variety of data sources. This presentation will give an overview of DOM development, the different ML methods that were assessed, some of the challenges encountered, as well as some preliminary results based on the 2019 Census Test.