Presentation on theme: "IMPACT OF ONLINE EDITS AND INTERNET FEATURES IN THE 2006 CANADIAN CENSUS Presented by Mike Bankier on behalf of: Danielle Laroche and Chantal Grondin Statistics."— Presentation transcript:
IMPACT OF ONLINE EDITS AND INTERNET FEATURES IN THE 2006 CANADIAN CENSUS Presented by Mike Bankier on behalf of: Danielle Laroche and Chantal Grondin Statistics Canada Work Session on Statistical Data Editing, April , Vienna
Outline Introduction Internet Features and Online Edits Impact of Online Edits and Internet Features Potential Enhancements for 2011 Census Internet Application Conclusion
Introduction Internet used as a new collection mode in the 2006 Census 18.3% of private dwellings reported on Internet Highest rate ever achieved by a Census in any country so far.
Internet Features and Online Edits To minimise mode effect, Internet questionnaire was very similar to paper one Edits were kept simple for the 2006 Census: – Performed on one question at a time – No consistency edits between questions 4 types of edits (or validation messages): – Non-response – Partial response – Invalid response – Amount verification On line help and information
Impact of Online Edits and Internet Features Much lower questionnaire follow-up rates for Internet (5.7%) compared to paper (39.1%) (long form) Less item non-response – Perception that an answer must be given to continue Less NR is good, but could become problematic in certain situations…
Impact of Online Edits and Internet Features Validation messages may encourage valid but incorrect answers when true answer is unknown Example 1: Day of birth – Compared people whose first answer reported to DOB was valid, to people who first received a validation message to that question, then gave a valid answer. – The proportion of day of birth 1 was much higher (7.3%) among those who had received the validation message compared to those who did not (3.6%).
Impact of Online Edits and Internet Features Example 2: Postal code of previous address – For people who lived in a different city 1 or 5 years ago, we compared the postal code of that previous address to the current residence postal code. – Internet reporters more likely to report same postal code before and after migration, compared to paper reporters (13.2% vs 9.5% for 5 year migration).
Impact of Online Edits and Internet Features When respondents are unsure on how to report their information An example : Number of hours worked – There is a higher proportion of 1 hour worked among Internet reporters than among paper reporters. – On paper, they must enter the number of hours in a text box, OR check the None box if havent worked. – On Internet, validation message for text box said that valid values had to be between 1 and 168. – Instead of using the None check box, some respondents seem to have entered 1 in the hour text box.
Impact of Online Edits and Internet Features Invalid answers due to multiple answers – The use of radio buttons in the Internet questionnaire made it impossible to select more than 1 answer when multiple answers were not allowed. – Although invalid rates are usually small even on paper, this was a gain for Internet questionnaires.
Impact of Online Edits and Internet Features Drop down menus – In general worked well, but can sometimes be tricky to use: When there is a need to scroll down after selecting an answer in the menu, one must click outside the menu for the response to be retained. Otherwise, the response might change to one that comes after it in the list. – No evidence of this kind of problem when no scrolling was needed after answer was selected from the menu. – Edit and imputation strategies were put in place to detect and remedy the problem.
Potential Enhancements for 2011 Census Internet Application A change will be considered to the online edit for the number of hours entered in the text box. Other changes might be done to remove/change validation messages for questions where we feel respondents might be prompted to enter a valid but incorrect answer. Other options for drop-down menus will be considered.
Conclusion Internet option offered many advantages and is considered a major enhancement. Follow-up rates are much lower for Internet questionnaires. Validation messages effective in obtaining answers for questions that could have been overlooked, and in having respondents correct errors they inadvertently committed. More research is needed to better understand respondents behaviour to validation messages.