Categories
Uncategorised

2011 NHS: Update on community-level data and data quality

It was just brought to our attention that Statistics Canada has quietly released the 2011 National Household Survey (NHS) data at the census tract (CT) level. As previously noted, only data at the census subdivision (CSD) level was made available at the time of the first major 2011 NHS release back in May. (The 2006 2B Census data for both levels of geography was released at the same time back in 2007.) No information has been provided as to whether/when the 2011 NHS data at the dissemination area (DA) level will be made available.

So good news for community-level data users. Sort of. Maybe… not.

Different geographies

Before jumping ahead, it may be worthwhile to provide a brief overview of the three major sub-provincial levels of geography.

Census subdivision (CSD)
Refers to a municipality or equivalent, such as an Indian reserve. Its size can vary widely in terms of both geography and population. The scope of coverage is all of Canada. There were 5253 CSDs in the 2011 Census. Of these, 1813 were suppressed from the 2011 National Household Survey, most because of low data quality owing to a high global non-response rate (GNR).

Census tract (CT)
Refers to a smaller geographic unit, typically covering a number of blocks or a territory bordered by major roadways. Its size is more consistent than a CSD, typically covering a population of roughly 2,500 to 8,000 people. The scope of coverage is limited to Canada’s major urban centres, also known as census metropolitan area (CMA) or census agglomeration (CA), where 2/3 of Canadians reside. There were a total of 5452 CTs in Canada’s 48 major urban centres. Of these, only 168 were suppressed from the 2011 National Household Survey, most because of low data quality owing to a high GNR.

Dissemination area (DA)
Refers to a small geographic unit, typically a neighbourhood covering a few blocks. It typically covers a population of roughly 400 to 700 people. The scope of coverage is all of Canada. There were 56,204 DAs in the 2011 Census. At the time of writing, no information regarding DA data suppression for the 2011 NHS was available.

Different data quality standards

Given the change from the mandatory long-form Census to the voluntary NHS, the number of suppressions due to non-response are far less than expected. Only a third of CSDs  were suppressed and practically no CTs were suppressed. Seems too good to be true.

Because it’s not (true). On closer inspection, it’s obvious why. Statistics Canada changed the maximum GNR threshold from 25 percent, used for previous long and short-form Census (including the 2011 Census), to 50 percent for the 2011 NHS.

Using the previous GNR threshold, data for most CSDs (4259, or 81 percent of them) and most CTs (3688, or 68 percent of them) would not have been fit to publish.

Recommendation

Short of simply refusing to use the data, analysts and researchers may wish to first review whether data for a specific area they wish to look at meets the previous data quality standard. This will not ensure the data is comparable, as Statistics Canada has made other changes that also effectively lowered data quality even further. But it’s the best one can hope for given the circumstances.

 

Leave a Reply

Your email address will not be published. Required fields are marked *