83 – Analysis of Financial Markets: Stocks, Exchange Rates, and Monetary Policy
The Estimation Methodology of the 2011 National Household Survey
Francois Verret
Statistics Canada
Prior to 2011, the Canadian census of population was conducted with a mandatory long-form sent to 20% of the households and a mandatory short-form sent to the rest of the households. The 2011 Census was conducted with a mandatory short-form sent to the entire population. A voluntary survey called the National Household Survey (NHS) was created to collect the information that used to be collected with the long-form. To minimize the impact of non-response, a sample of 30% was selected for the NHS and after several weeks of collection, the follow-up efforts were concentrated on a random subsample of the remaining non-respondents. The design-weighted response rate was 77%, while for the last census long-form the response rate was 94%. This paper describes the estimation methodology used in the 2011 NHS. Design weights were first calculated. A non-response weight adjustment was done using census and linked administrative data and by converting a nearest-neighbour and whole-household imputation approach to a reweighting approach. Weight calibration to many census totals was performed. Variance was estimated with a multi-phase variance formula and Taylor linearization.