Innovative approaches used in the 2016 Canadian Census Session d: Innovations in census methodology and technology, and results of testing
Prepared by Marc Hamel (Statistics Canada) Presentation to the Expert Group Meeting on Population and Housing Census Geneva, 2016 Outline The Canadian Census The census collection methodologies Results from 2016 Response management Adopting corporate systems for the Census Tailored communication strategy Lessons Learned
2 Statistics Canada • Statistique Canada 26/10/2016 The Canadian Census A five years cycle : Canada conducts a census of population and dwellings every 5 years. The most recent one took place in May 2016 and expanded on methodological changes introduced for the 2011 Census. Overview: The wave collection methodology encourages response via internet while ensuring a high level of self-response. The basic approach was extended to a larger portion of dwellings (82%) in 2016. New for 2016 was the use of Statistics Canada’s common systems for the census program The communication strategy was also revamped. The census brand was modernized Specific strategies were developed for different key populations, including difficult to reach groups Collection efforts planned and coordinated based on predictive response model
3 Statistics Canada • Statistique Canada 26/10/2016 The census collection methodologies Various collection methodologies:
Mailout using the information available in an Address Register - 82% of dwellings in 2016 Questionnaire delivered to the dwellings by a census enumerator as part of a list and leave operation - 16% of dwellings Census collected directly at the door by a census enumerator - 2% in First Nations communities, and in remote and northern areas Income data obtained directly from the Canada Revenue Agency Long-form collected for 1 in 4 households
4 Statistics Canada • Statistique Canada 26/10/2016 The Wave collection methodology A little bit of history: The wave collection methodology used for the 2016 Canadian Census is largely based on the theory of Dilman, and was first introduced for the 2011 Census.
The process: Based on the results of tests conducted during the 2011 Census, the method was applied more uniformly for 2016, with a single treatment for mailout areas as opposed to two different treatments in 2011. It was left unchanged for list and leave areas. The choice of treatments in mailout areas in 2011 (letter only or paper questionnaire delivered in wave 1) was determined based on the propensity of the population to self-enumerate as demonstrated in the 2006 census and on the 2006 Census internet take-up rate by area.
5 Statistics Canada • Statistique Canada 26/10/2016 2016 Census Wave Methodology
Wave 1 Wave 2 Wave 3 Wave 4
Automated Call Letter Non-Response Reminder Non-Response Questionnaire Non-Response +Non-Response Letter Follow-up
out (82%) out May 2 May 11-13 May 18-27 Begins June 1
Paper Questionnaire Request System
Reminder / Non-Response Questionnaire All Thank you Non-Response Follow-up card
May 2-9 May 11-13 Begins May 20 List/Leave (16%) List/Leave
No Waves Canvasser (2%) Canvasser
6 Results from 2016 Response objectives for the 2016 census were surpassed: Overall collection response rate: 98.4% with fairly low variance in rates at the Census sub-division level (communities) Internet collection rate: 68.3% Overall collection self-response rate: 88.8% Collection response to the long-form: 97.8%
The higher quality of responses received by internet has been demonstrated in previous censuses. Lower rates of item non-response in 2016 compared to 2006 and 2011
7 Statistics Canada • Statistique Canada 26/10/2016 Wave 1 Receiv ed on May 2 Internet and Mail daily return rate (National) Wave 2 Wave 3 Voice Census Received Received 1,400,000 Broadcast Day between May between May On May 30-31 On May 11-16 19-26 9.0% 1,200,000 10 8.0%
1,000,000 7.0%
6.0% 800,000 5.0% Internet 600,000 Impact of 4.0% Mail Wave 2 Impact of 3.0% Self-Response 400,000 Wave 3 2.0% 200,000 1.0%
0 0.0%
04/05/2016 01/06/2016 04/06/2016 01/05/2016 02/05/2016 03/05/2016 05/05/2016 06/05/2016 07/05/2016 08/05/2016 09/05/2016 10/05/2016 11/05/2016 12/05/2016 13/05/2016 14/05/2016 15/05/2016 16/05/2016 17/05/2016 18/05/2016 19/05/2016 20/05/2016 21/05/2016 22/05/2016 23/05/2016 24/05/2016 25/05/2016 26/05/2016 27/05/2016 28/05/2016 29/05/2016 30/05/2016 31/05/2016 02/06/2016 03/06/2016 05/06/2016 06/06/2016 07/06/2016 08/06/2016 09/06/2016 10/06/2016 11/06/2016 12/06/2016 13/06/2016 14/06/2016 15/06/2016 Internet daily return rate (2011 vs 2016) 9% Census Day 8%
7%
6%
5% Impact of 4% Wave 2 Internet % 2016
3% Impact of Internet % 2011 Wave 3 2%
1%
0%
03/05/2016 02/05/2016 04/05/2016 05/05/2016 06/05/2016 07/05/2016 08/05/2016 09/05/2016 10/05/2016 11/05/2016 12/05/2016 13/05/2016 14/05/2016 15/05/2016 16/05/2016 17/05/2016 18/05/2016 19/05/2016 20/05/2016 21/05/2016 22/05/2016 23/05/2016 24/05/2016 25/05/2016 26/05/2016 27/05/2016 28/05/2016 29/05/2016 30/05/2016 31/05/2016 01/06/2016 02/06/2016 03/06/2016 04/06/2016 05/06/2016 06/06/2016 07/06/2016 08/06/2016 09/06/2016 10/06/2016 11/06/2016 12/06/2016 13/06/2016 14/06/2016 15/06/2016 Response management Strategy to manage the collection process Dynamic predictive microsimulation model improved from 2011 to optimize use of human and financial resources in pursuit of response objectives. Projected the end of collection by region based on a number of parameters such as: Self-response levels; Productivity of enumerators; Projected attrition of enumerators; Remaining number of cases to be resolved; Hours worked per day by enumerator by area and by collection supports units (telephone follow-up capacity). Also forecast collection costs until the end of collection Modeled every week during collection Estimate the number of work hours to be added in any given region/area to complete collection operations by the end of July at the targeted response level. 10 Statistics Canada • Statistique Canada 26/10/2016 Response management (cont’d) The results:
Because of actual and predicted favorable conditions during the 10 weeks of non-response follow-up, most areas in the country never moved below tier 2.
At tier 2, collection continued in any given geographic area until a response rate of 98.5% was reached for the short-form and 96% for the long-form.
The preliminary assessments show that the predictive power of the model was very accurate.
11 Statistics Canada • Statistique Canada 26/10/2016 Adopting corporate systems for the Census Corporate Business Architecture IT solutions and business processes developed for corporate use rather than for individual program specific requirements.
The Census Program successfully adopted such solutions for 2016, allowing economies of scale for the department and the census program itself. Common internet application (eQ) to replace the census-specific application used since 2006. Collection Management Portal (CMP) to replace the Field Management System used in 2011.
Overall, the adoption of corporate solutions for the Census was successful in 2016, and it is expected that their re-use in 2021 will allow for important efficiencies
12 Statistics Canada • Statistique Canada 26/10/2016 Tailored communication strategy The Integrated Communications Strategy (ICS) was designed as a social marketing campaign. Emphasis on research, segmentation, targeting and positioning. Targeting strategy consisted of full market coverage and use of differentiated marketing to reach difficult to enumerate audiences. Statistical analyses guided the design, implementation and monitoring of census communications activities. ICS employed behavioral economics and sociodemographic segmentation to identify hard-to-enumerate audiences. Some adjustments to these solutions will be necessary for 2021. Contributed to the overall increase in self-response compared to 2016.
13 Statistics Canada • Statistique Canada 26/10/2016 Preliminary results analysis
Cumulative Returns - All Segments 14,000,000
12,000,000
10,000,000
8,000,000
6,000,000
4,000,000
2,000,000
-
2011 2016 projected 2016
14 Statistics Canada • Statistique Canada 26/10/2016 Preliminary results analysis
Daily Returns - Low-income Tenant 120,000
100,000
80,000
60,000
40,000
20,000
0
02-May 05-May 28-May 01-May 03-May 04-May 06-May 07-May 08-May 09-May 10-May 11-May 12-May 13-May 14-May 15-May 16-May 17-May 18-May 19-May 20-May 21-May 22-May 23-May 24-May 25-May 26-May 27-May 29-May 30-May 2011 2016
15 Statistics Canada • Statistique Canada 26/10/2016 Preliminary results analysis
Daily Returns - Single Unattached Mobile 140,000
120,000
100,000
80,000
60,000
40,000
20,000
0
09-May 01-May 02-May 03-May 04-May 05-May 06-May 07-May 08-May 10-May 11-May 12-May 13-May 14-May 15-May 16-May 17-May 18-May 19-May 20-May 21-May 22-May 23-May 24-May 25-May 26-May 27-May 28-May 29-May 30-May 2011 2016
16 Statistics Canada • Statistique Canada 26/10/2016 Lessons Learned
What’s next for Statistics Canada? : The approaches used to increase internet and self-response in the Canadian census have again proven to be a success in 2016, following the successful introduction of many of these in 2011.
The focus of the evaluations from 2016 will be on the potential for expansion of the mailout universe which would increase the benefits of the methodology further for 2021.
Particular attention will be dedicated to improving the methods of contact with dwellings, anticipating that the efficiency of using traditional mail as the primary mode of contact will diminish in the next 5 years as more and more people rely on other means of communication, such as Internet.
17 Statistics Canada • Statistique Canada 26/10/2016 Fort McMurray
18 Statistics Canada • Statistique Canada 26/10/2016 Contact Information
. Marc Hamel . (613) 951-2495 . [email protected]
19 Statistics Canada • Statistique Canada 26/10/2016