469 – Official Statistics: Policy and Practice
A Discrete-Time Microsimulation Model of the 2016 Canadian Census Collection Operations: An Innovative Method for Responsive Design and Cost Control
Vincent Martin
Statistics Canada
The survey operations of national censuses require careful planning and monitoring. The problem of forecasting survey operations costs and response rates is one that has been discussed at great length already. However, models in the literature are generally not designed to update predictions over the course of the operations. For the 2016 Canadian Census, Statistics Canada tackled this problem through a microsimulation model that produced an updated forecast of subnational costs and response rates on a weekly basis that reflected the most recent collection data available. The model built to meet this operational requirement produces reasonable output at small geography levels and was used to identify and address collection issues early during field follow-up of non-response. In this talk, I will present the methodological framework that defines the model and some key results. I will also discuss the operational and methodological gaps filled by this model, the level of flexibility it enables for the planning of the 2021 Census, and future work to improve the model.