6. Weighting areas

Weighting areas (WA) are the geographical partitions that were used during the National Household Survey (NHS) weighting procedures. In the 2011 NHS, Canada was partitioned into 5,884 WAs. The weighting process attempted to achieve agreement between certain sample estimates and the corresponding census counts for each WA. Some changes were made to the design of the WAs in 2011 because the non-mandatory NHS resulted in different response rates for different parts of the country. Weighting areas were designed in two steps in 2011: the regular WA formation and the canvasser and reserve adjustment.

6.1 Regular WA formation

The first step in creating a set of WAs was to group together dissemination areas (DAs) while adhering to the following conditions:

1. A WA must respect the boundaries of census divisions (CD) (i.e., a WA can only be found in a single CD).
2. A WA should, where possible, contain between 300 and 699 households that responded to the NHS or were selected in the subsample (both subsample respondents and non-respondents). For simplicity in this chapter, these households will be known as respondent households.
3. A WA should, where possible, respect (in order of priority) census subdivision (CSD) boundaries and census tract (CT) boundaries.
4. A WA should, where possible, be made up of contiguous DAs (i.e., not be in two or more parts or contain any 'holes') and it should be as compact as possible.

The conditions above are very similar to those in 2006, with the main exception being condition (b). In 2006 the WA size was based on the number of households of the census that were subject to sampling, which in most cases was between 1,000 and 3,000 households. If WAs had been created in the same fashion in 2011, the number of respondent households would have varied widely between WAs. Calibration would not have been effective in some of the low-response WAs.

To remedy this problem in 2011, WAs were created based primarily on the number of respondent households, even though it meant that WAs in low-response areas would have to include a higher number of in-scope dwellings than in the past. The 2006 Census was sampled at approximately one in five, meaning that each WA with 1,000 to 3,000 in-scope households would have had approximately between 200 and 600 respondent households in the long-form sample. This interval was increased slightly in 2011 to include between 300 and 699 to allow for additional non-response and to give more census counts and NHS estimate agreement. Increasing the WA by more than this would have caused significantly more CSD boundary conflicts in condition (c).

The fourth condition was generally simple to apply in urban areas. However, the condition of having contiguous DAs had to be relaxed in areas where two or more disjoint pockets of land belonged to the same municipality. The condition was also relaxed if there were several small or low-responding municipalities in a CD that needed to be put into a single WA in order to satisfy condition (b). This most commonly occurred in areas with reserves and in the North.

The algorithm that was used in the first step of the WA formation process in 2011 was the same as in 2006. In both cases, many manual adjustments were done to abnormal WAs by splitting, joining, or realigning WAs to better fit the conditions described earlier. After the manual adjustments in the first step of 2011, 99.3% of all WAs had the desired number of households — which was the same rate as in 2006. Table 6.1.1 gives some statistics on how well the WAs fit together with CSDs and CTs. The descriptions of each category are provided after the table.

Table 6.1.1
Number of census subdivisions and census tracts that respect weighting area boundaries, first step, 2011 Census

In scenario 1, condition (c) was satisfied. This scenario occurred frequently for CSDs because there were many very small municipalities such as Indian reserve and villages that contributed less population than was required to create a WA. In scenario 2, condition (c) is not satisfied. In scenario 3, condition (c) was satisfied. In scenario 4, condition (c) is not satisfied.

6.2 WA adjustment for canvasser and reserve areas

The second step in creating WAs was new for 2011. In order to create more homogenous WAs, households collected by a canvasser were processed together. These households were removed from their original WAs and placed into a small number of special WAs known as canvasser WAs. In order to generate WAs with 300 to 699 respondent households, the restriction of being in the same CD and occasionally the same province or territory was lifted. Similarly, households on Indian reserves were also processed together with the same aboriginal tribe wherever possible. Households that resided on Indian reserves, excluding land that had been leased for non-aboriginal dwellings, were placed in a small number of special WAs called reserve WAs. Most reserve WAs also contained 300 to 699 respondent households and were permitted to cross CD boundaries and occasionally provincial or territorial boundaries. In fact, 15 of the 5,884 WAs crossed provincial and territorial boundaries and appeared in exactly two provinces or territories, and one WA appeared in British Columbia, Alberta, and the Northwest Territories at the same time. Table 6.2.1 shows the number of WAs that crossed CD boundaries, and particularly the number of CDs that they touched.

Table 6.2.1
Number of WAs associated with multiple census divisions

Table 6.2.1 shows that 5,793 of the 5,884 WAs remained in a single CD, while 58 WAs were involved in two CDs. The highest number of CDs touched by a single WA was 11.

All respondent households that were not placed in canvasser or reserve WAs remained in regular WAs. In most cases, these households remained in their original WA. However, because of the household extraction for canvasser and reserve WAs, some original WAs had a reduced number of households. If the original WA was reduced to a small number of respondent households (usually less than 200), its households were reassigned to nearby regular WAs and the original WA was eliminated. However, if the original WA retained a significant number of households (usually at least 200), then the WA was left alone, even if the number of households fell below the desired 300 to 699 range. Table 6.2.2 shows the breakdown of the number of WAs by WA type after step 2 was completed.

Table 6.2.2
Frequency of weighting area type

Table 6.2.3 shows the number of WAs by WA type and intervals of the number of respondent households. As specified in Section 6.1, most WAs should contain 300 to 699 respondent households. The table shows that 5,736 (97.5%) of the final 5,884 WAs were within the range of 300 to 699 in 2011.

Table 6.2.3
Distribution of WAs by the number of respondent households and the WA type

The set of occupied private dwellings and their corresponding households in the 2011 Census were considered in-scope for the 2011 NHS. However, due to the varying response rates in the NHS, the size of each WA was dependent on the number of respondent households in the NHS and not the number of in-scope households. After determining the WAs in the NHS, the number of in-scope census households was also determined. Table 6.2.4 shows the number of WAs by WA type as well as intervals of the number of in-scope households in the census.

Table 6.2.4
Distribution of WAs by the number of in-scope census households and WA type

The majority of WAs contained between 1,500 and 2,999 in-scope census households, which is in line with the 2006 Census in which WAs were designed to have between 1,000 and 3,000 households.