STATISTICS CANADA 2006 CENSUS Internet Response Option

STATISTICS CANADA – 2006 CENSUS Internet Response Option Presentation at UN Statistical Commission February 28, 2007

Census in Canada • Legal requirement every 5 years, De Jure basis • Last Census May 16 th, 2006; 13 million dwellings • 80%-short form (8 questions); 20%-long form (53 questions + 8 dwelling questions ) • Confidentiality is paramount • Combined collection for Census of Population and Census of Agriculture • Development and integration of Internet application and processing systems was outsourced to private sector • Targeted 20% of initial response from the internet 2

Census 2006 – Drivers for change • • Privacy concerns – local enumerator issue Government On-Line (GOL) initiative Data Entry Capability/Manual Coding Recruitment and retention of staff 3

Radical Change • Most radical changes since 1970’s • Mail-out via Canada Post for 73% of households (Address register critical) • List leave for remainder • Unique dwelling identifier and internet access code for each questionnaire 4

Radical Change • Mail-back or Internet option (canvasser method for a few areas) • One Data Processing Centre (CATI or Field Follow-up initiated from it) –Intelligent Character Recognition (ICR) –Master Control List –Automated coding 5

Internet mode of collection – Nothing left on respondent’s computer (Zero footprint) – Log-on using internet access code printed on questionnaire – Innovative Security Solution using GOC “Secure Channel” (PKI “light”, secure “islands”) – One-time recyclable certificates – Multiple sessions for long form (user can assign own password) 6

How did we do - Results • Metrics – 22% of initial response was by Internet – Overall internet represented 18% of all responses • Graceful Deferral Contingency Procedure seldom used, except on Census night • Data Quality better for internet response 7

How did we do - Results • Internet available and used (on a gradually declining basis) until end of follow-up period (end of August) • “Push Strategy” experiment (no questionnaire sent, only a letter asking to respond via internet) yielded very good results • Successful use of a reminder letter to nonrespondents asking to reply by internet (no questionnaire sent) 8

Implementation Issues • • • Operational/development plan Sizing, technical, integration issues Testing strategies Support (technical and respondent) Security and public perception (independent validations) 9

Lessons Learned • Integrating with Secure Channel was harder and took longer than expected • Difficult to exercise volume tests that accurately reflected production • Being able to monitor and control the number of respondents accessing the system was critical • Not mailing paper questionnaires increases the internet take up rate • Testing every aspect and linkage is crucial • True partnership with contractors was critical success factor 10

Going forward - 2011 • Basic methodology will be retained • Aggressive assumption regarding Internet takeup Rate. Significant Potential Savings – Return postage – Paper handling and data capture – Follow-up for edit-failures • Security approach to be determined (availability of Secure Channel? Should we adopt a less customized security method, e. g. SSL? ) 11

Going forward - 2011 • Census Day versus Census Week to spread system load? • In-house, outsourced or hybrid development approach a critical decision – Integration – Competency profile 12

Going forward - 2011 • Expanding Mail-out (MO) opportunities - currently assessing an approach that will make better use of Mailout tools as well as encourage response by Internet as opposed to paper questionnaire ØWave 1 - Initial MO - 90% of all dwellings; § 80% of MO dwellings will receive an invitation to use Internet only - no paper questionnaire will be provided; § 20% of MO dwellings will receive paper questionnaire package (areas of low connectivity); § Approximately two weeks before Census Day. 13

Going forward - 2011 • Wave 2 - All dwellings will receive a mailed "pre-addressed reminder" to complete questionnaire via Internet or paper questionnaire received during Wave 1; Ø Mailed around Census Day. • Wave 3 - one week after Census Day determine mail-out universe that did not reply - A paper questionnaire package will be prepared and mail-out to each dwelling which did not reply; Ø Approximately two weeks allowed for package preparation and postal delivery. An additional week will be allowed for households to respond. • Wave 4 - Start Non- Response Follow-up (NRFU) approximately 4 weeks after census Day. 14

Going forward - 2011 2006 Collection Process Methodology Wait Time until 80% responses accounted Census Reference Day – May 10, 2011 NRFU WAVE 1 WAVE 2 Census Week May 2011 WAVE 3 ENRFU BEGINS WAVE 4 15 NRFU WAVE 5 ENDS 07/29/11 Deadline