Abstract

Like most statistical agencies, Statistics Canada has adopted the international collection model for its survey measuring Canadian household expenditures. Until the 2019 reference year cycle, the survey program was based primarily on the Survey of Household Spending (SHS) which is a voluntary biennial survey with a sample size of approximately 17,500 households in the 10 provinces. The SHS combines two collection modes: an interview and an expenditure diary. The computer-assisted personal interview is used primarily to collect larger, less frequent expenditures. The reference periods are established according to the type of expenditure (1 month, 3 months, last payment, 4 weeks). The diary, in paper format, is used to collect frequent and smaller expenditures over a one-week period. Household income data are obtained from administrative sources and are added to the respondents' data following the collection process. The SHS data are used to update the Consumer Price Index basket of goods and services weights. They are also used by the System of National Accounts, as an input to derive the Gross Domestic Product, as well as by several federal and provincial government departments to develop social and economic policies and programs. Finally, various groups interested in better understanding issues related to the spending habits of Canadians use the SHS data.

The Covid-19 pandemic significantly altered the Agency's collection strategies, and personal interviews that were planned for the interview portion of the survey could not be implemented for the SHS 2021 cycle due to public health concerns. A collection strategy involving mailings, a self-administered electronic questionnaire in a new environment, telephone follow-ups, and a paper diary was implemented for this cycle. In addition, a variety of means were used to remind respondents to participate (letters, text messages, emails and face-to-face contacts). In addition to these changes to the collection strategy, new automated processing methods were either implemented or tested in parallel to reduce the survey burden on the Agency. Indeed, a machine learning algorithm was used for automated coding of items collected from our diary. In addition, an optical character recognition approach was optimized during the SHS 2021 collection cycle and will be implemented in 2023.

This presentation will provide an overview of the major new features implemented for the SHS 2021 and discuss their effects on collection performance. In addition, possible avenues for research and development will be described, particularly with respect to alternative data sources and the survey model in general.