Manifold and Census 2011

Manifold has successfully incorporated the complete Census and National Housing Survey (NHS) 2011 into its current year updates of the population estimates and projections.

Census 2011 was conducted on May 10th, 2011 with a mandatory short form of Census Questionnaire and a voluntary National Household Survey (NHS), which replaced the historical mandatory long form of census questionnaire.  The mandatory short form consisted of 10 questions and was required by law to be filled by all Canadian households. The national response rate of the short form was 98.1%, the highest in the Census history.  Ontario and Prince Edward Island had the highest response rate at 98.3%, while Nunavut had the lowest response rate at 92.7%. Data collected from the mandatory form of Census 2011, i.e., population age group, family structure, household composition, dwelling types and languages were released completely by Statistics Canada (“StatCan”) at the DA (Dissemination Area) level. We were able to use these data to fully update the population, family, dwelling, household, home language, mother tongue categories in the SuperDemographics 2014. To capture the undercount of population in the Census, estimate current year and project future population statistics, we considered the StatCan’s post-Census survey,  analyzed and incorporated the historical trends, regional birth and mortality rates, migration and immigration statistics, real estate statistics and directory books. We used a bottom up technique to aggregate key data from the ground/street level to the standard Census geographies. At the same time, we also took a top down approach by comparing StatCan’s research and publications on population estimates and projections with our results in order to identify abnormalities and establish consistency. In this way we leveraged the strength of data both at street and municipal levels.

Instead of the mandatory long form questionnaire on ethnicity, immigration, labour force, income, dwelling value, religion…, government decided to change it into voluntary in Census 2011. Thus StatCan conducted the voluntary National Housing Survey (“NHS”) within four weeks of the May 2011 Census among 4.5 million Canadian households. Unfortunately the response rate was merely 68.6 percent, much lower than the traditional 93.5 percent when the long form was mandatory. Prince Edward Island had just 60.3% responders. Saskatchewan’s response rate was 63.8% although it has been one of the fastest growing regions in Canada for the last 10 years. Even worse, the variation in response rates by geography and by questionnaire is big and irregular. Unlike the previous Census in which response to the mandatory long form was also biased, but more in a systematic fashion and can be corrected statistically, the current voluntary NHS survey created a “random” variation which is much hard to fix for small geographies.

To ensure quality of the data, Statistics Canada has reduced the number of variables substantially and limited the publication to areas with sufficient high accuracy of data. 

Although Census/NHS 2011 has a lot of deficiencies, it is  the most comprehensive and current Canadian demographic data. No other data can replace the coherent structure in the Census data. For example, religion is surveyed by Statistics Canada only every 10 years, i.e., in the year 2001 and 2011. There were a lot of changes in these 10 years. For example, for Jewish community the high mortality among seniors caused a significant reduction of Jewish population. On the other hand, the Russian Jewish immigrants contributed to the growth of Jewish population. Similarly, the high birth rate and immigrants among Muslim population made Muslim the fastest growing religion in Canada.   Census 2011 captured these changes systematically and served as a new base for correction of the old projection and refinement of the future projection.  Mother Tongue and Home Language were moved from the 20% sample  of the long form in previous Census to the 100% sample of short form in Census 2011, which improved substantially the quality of language components in the Census data and offered a more accurate reflection of multi-culture landscape in Canada.  

Early 2014 Manifold conducted a systematic study of the Census and NHS 2011in the preparation of updating our population estimates and projections. We analyzed distribution of the variables. We filled the gaps in individual variables where StatCan suppressed the values due to low response rate. We adjusted the Census and NHS data and eliminated the round-off errors. We also validated and confirmed the consistence between our DA level estimates with StatCan’s CSD and FSA (Forward Sortation Area, i.e., first three digits of a postal code) level data. We  compared the adjusted Census 2011 with the historical Census to verify the trends and used it as the new based for our current year population estimates and projections which included the population growth from 2011 to 2014.