1. GEOG 204 Introductory GIS for the Social Sciences
Social Area Analysis and the Elimination of the Long Form Census
26 October 2012
Neil Hanlon

2. OUTLINE Social Area Analysis (SAA) Census Long Form Elimination
The National Household Survey
Implications for SAA using Census Small Area Dissemination Data

3. Social Area Analysis (SAA)
Also known as urban or social ecology
Social and demographic features of small urban areas (e.g., Census tracts)
Assumption: areas are fairly homogeneous with respect to living conditions, social environment
Social conditions and resources vary from one neighbourhood to the next
Opportunity structures
Especially for “vulnerable” or mobility-limited groups (e.g., frail elderly, children, disabled)

4. Social Area Analysis (SAA)
E.g., health geography literature looking at relationship of “neighbourhood” characeteristics and health status
Living in “deprived” areas means fewer opportunities for health living (material conditions) and less likelihood of positive reinforcement of health lifestyle choices, security (social conditions)

8. Census Long Form Elimination
“Long Form” census had been around a long time (the 8 question short form was introduced in 1971)
Sections:
Demography; Labour markets & income; Transportation; Education; Language; Activity Limitations; Housing; Citizenship and immigration; Ethnicity/culture

9. Census Long Form Elimination
Mandatory – fill it out or else fine or jail time!
One in five households (20% random sample)
2010 – Federal Government decided to cancel it
Replaced with a voluntary household survey (National Household Survey)

10. National Household Survey
Sampling error – three factors: size of population, number of survey respondents, types of measures used
Larger # respondents lowers sampling error
Stats Canada anticipates higher overall sampling errors with move to NHS
i.e., they are not expecting to collect data from 20% of households!

11. National Household Survey
Non-sampling error – e.g., respondent mistakes, data collection mistakes, processing errors (but these are constant)
Biggest concern with a voluntary survey is non-response bias
Mandatory LFC had 96-98% responses, therefore non-response bias fairly small
Risk of non-response bias increases with decreasing response rates

12. Implications for SAA using Census Small Area Dissemination Data
Small area dissemination data most at risk from decision to cancel the mandatory Long Form Census
Voluntary surveys – uneven response rates, therefore both sampling AND non-sampling error
Increasing sample size does not resolve non-response bias (only random sampling will)

13. References
Canadian Centre for Policy Alternatives (no date) All the latest on the Census long-form debacle.
Ross, N, Tremblay, S, and Graham, K, 2004, Neighbourhood influences in health in Montreal, Canada. Social Science and Medicine 59:
Sheikh, MA, 2011, Good data and intelligent government. In A Gorbet and A Sharpe (eds) New Directions for Public Policy in Canada: Papers in Honour of Ian Stewart. Open access