Canadas 2006 Census Online Experience MSIS 2008 Luxembourg

Canada’s 2006 Census Online Experience MSIS 2008 Luxembourg, April 7 -9, 2008 Karen Doherty Director General Informatics Branch Statistics Canada

Table of Contents • • • Environmental Overview Canadian Census Program Preconditions for Online Option Census Online Application Features Results Conclusion 2

Canada’s Geography 3

Canada Overview • 10 M km² and 5½ time zones (8 hours non-stop by plane from St John’s Nfld to Victoria BC) • Two official languages, many other languages due to multi-cultural nature of the country • Very low unemployment rate especially in the booming west (Alberta and BC) • Large tracts of very sparsely populated land 4

Population by Province Population average/km² Canada=3. 5/km² US=31/km² UK=246/km² Canadian urban centres=245/km² 5

Population Distribution 6

Population Distribution 7

Census Districts 8

The Canadian Census Program • Legal requirement every 5 years • Census of Population and Census of Agriculture run simultaneously • Two main form types for Census of Pop: – Short form (8 questions) – 80% of households – Long form (53 questions) – 20% of households • Protecting confidentiality is paramount 9

Drivers for Change • Use of local enumerators raised issues around privacy • Canada Revenue Agency could no longer perform the data capture (Canadians largely submit their tax returns electronically) • Harder and harder to hire 22, 000 enumerators • Government policy is to offer all services on-line 10

New Methodology • Most radical change since 1971 • Questionnaires delivered by Canada Post to about 70% of households • Enumerators delivered the remaining 30% • Returns made by mail or via online option • 98% self enumerators, 2% enumerated by interview • All returns processed in one highly secure data centre • Development of the online option and processing applications contracted out 11

Preconditions for an Internet Response Option • Well designed collection methodology – Central repository of all households – Good tracking and communications mechanisms • • High level of Internet connectivity Public trust in online transactions Mature security infrastructure Experience with online surveys 12

Web Application Features • User friendly web application – – thin client application (no large download) intelligent skip patterns soft edits save and resume function • Support for both official languages • Accessibility options for the visually impaired • Support for a large range of browsers 13

Security Features • Specially designed PKI encryption solution delivered by the Canadian Government’s Secure Channel • Required a standard industry provided Java applet on the workstation PKI • Research indicated that about 6% of PCs would not have the Java applet pre-loaded • Login only required that the user enter the 16 digit access code provided on their Census form 14

15

16

Online Questionnaire 17

Results 22% of those who responded prior to non-response follow-up responded online 18. 5% in all responded online 18

Online Response Pattern May 06 - Planned VS Actual Census Day May 16, 2006 19

Census Day Logins Per Second Graceful deferral starts 20

Results Breakdown • Higher rate of use by larger households: – 5 or more occupants: 26% – 2 -4 occupants: 19. 5% – 1 occupant: 13. 5% • No difference in response rates between the short and long form respondents • Over 20% response rate in Alberta, BC, Ontario and the Yukon 21

Benefits • Better quality responses – Edit failure rates much lower for online questionnaires – Item non-response rates: • Short form – Paper forms have 102 x higher rate • Long form – Paper forms have 10 x higher rate • Reduced costs for paper-handling, mail-back postage and follow-up • Relieves the pressure to find 20, 000 increasingly scarce enumerators 22

Observations • Need to test the wide range of browsers • Need easy to use online questionnaires • Testing pointed to the need for a special questionnaire layout for the Help Line operators • Need strong partnership with network provider (in this case Secure Channel team) • Need a back up plan 23

Lessons Learned • Volume testing was harder and took longer than planned • Little industry expertise in modelling volume testing using PKI • Needed to monitor and limit access in Census Day – respondents were understanding • Still work to do to handle the Java applet 24

Onwards and Upwards • Target for 2006 was 15 -20% via Internet Target for 2011 is 40% via Internet • New mail out strategy based on results of test groups during 2006 Census and subsequent follow -up: – Majority of households will receive a letter asking them to respond online – They will be able to order a paper questionnaire via the Help Line if they can’t or choose not to respond online 25