Discussion

View the most recent version.

1.0 Introduction
2.0 Changes to delineation methodology
3.0 2001 delineation and post-censal update
4.0 2006 delineation and calculation of adjusted counts
5.0 Conclusion

1.0 Introduction

An urban area has a population of at least 1,000 and no fewer than 400 persons per square kilometre. Part of the delineation process includes tests for spatial contiguity to ensure data confidentiality. As such, urban areas are reviewed and may be modified where appropriate, for example, the removal of interior holes. As a result, the land area of some urban areas may increase so that the population density of the entire urban area is less than 400 persons per square kilometre. The impact on the total population is minor since the added areas are usually sparsely populated, or contain no population at all.

The delineation of urban areas for the 2006 Census is largely impacted by:

• The change in the methodology used for delineating urban areas in 2001
• The delineation and post-censal update of 2001 urban areas.

In this paper, the impact of each of the above factors is reviewed with respect to the resulting changes to urban areas for 2006 and the calculation of associated 2001 modified population counts.

2.0 Changes to delineation methodology

In 2001, a number of geographic changes were implemented that affected the delineation of 2001 urban areas. Above all, for the first time, the delineation of urban areas was done with an automated process, the Generalized Area Delineation System (GArDS), that made it possible to use population counts and population density data from the current census. Prior to 2001, the process was manual and, to avoid a release delay of census information, population count and density data were based on the previous census.

Also in 2001, the block (an area equivalent to a city block bounded by intersecting streets) became the basic building block for forming urban areas. Prior to 2001, the geographic unit used for urban area delineation was the enumeration area.

Since 2001 urban area delineation must start with an assessment of urban areas from the previous census, all 2001 blocks that made up the 1996 urban areas had to be assigned a valid urban area code in order to be assessed in the 2001 urban area delineation process (see Appendix A for detailed delineation rules). This process was handled by GArDS and the resulting blocks were used as input for the 2001 delineation process. Since 2001 blocks did not necessarily respect 1996 enumeration areas, it was not possible to recreate 1996 urban areas based on 2001 blocks with 100% accuracy. In some cases, the block boundary extended beyond the enumeration area boundary. A decision was made to include the entire block as part of the urban area for data comparability since block splitting was not a viable option at the time.

Another important change for 2001 was the conversion of the geographic referencing system from North American Datum of 1927 (NAD27) to North American Datum of 1983 (NAD83). Spatial data based on NAD27 will not be coincident with the same spatial data based on NAD83. The conversion to NAD83 had an effect on the transfer of positional information from the 1996 urban area to the 2001 block. It is possible, therefore, that the block input used by GArDS to delineate 2001 urban areas did not exactly depict the 1996 urban area structure that was published.

3.0 2001 delineation and post-censal update

Notwithstanding the challenges described above in creating a good block-based input to the urban area delineation process, GArDS was used for the first time in 2001 to create urban areas. Initial results were deemed successful and the 2001 urban area boundaries were published. In subsequent months, a review of the boundaries revealed that a number of excessively large blocks had been added to some urban areas, mostly during the reconciliation process from the enumeration area to the block. Together, the reconciliation process and the 2001 delineation process increased the land area of 200 urban areas by more than 25% (see Table 3.1). Further analysis led to the conclusion that many of these urban areas were over-bounded.

In 2004, it was decided that many urban areas would be manually updated to rectify the substantial over-bounding of land area. To do so required the manual splitting of larger blocks into smaller blocks, allocating the population as the technician deemed appropriate and assigning valid urban area codes to those smaller blocks that would remain part of the urban area. Those blocks that were to be excluded from the urban area were assigned a rural code for the province in which it was located. This method of population allocation differs from the traditional method used when adjusting geographic boundaries, which allocates population and dwellings based on information obtained from census visitation records to the appropriate block.

At the same time that urban areas were being manually updated, the block, which is now called the dissemination block for 2006, was being delineated by GArDS for the 2006 Census. Since the 2006 dissemination block structure is largely dependent on the road network (as was the 2001 block structure), GArDS would create smaller blocks where road network had been added and would merge blocks where road network had been removed. GArDS was careful to maintain blocks that were defined by higher-level geographic boundaries. However, urban areas were not processed as a higher-level boundary and there were instances where blocks were added or removed from 2001 urban areas because they were merged with other blocks that were adjacent to, but outside of, the urban area boundaryy.

The manual update to over-bounding blocks and the automatic adjustment of blocks by GArDS that were associated with the 2001 urban area define the net effect of post-censal update to 2001 urban areas. In total, 412 (or 45%) of the 913 urban areas from the 2001 Census were updated post-censally to repair the over-bounding. A complete list of updated UAs and the corresponding area modifications are presented in Appendix B. This update also included the reinstatement of four urban areas that GArDS had merged with nearby urban areas in 2001: Fortune (UA 10 0300), Sainte-Croix (UA 24 0878), Châteauguay (UA 24 1177), and Dowling (UA 35 1084). Table 3.2 shows the counts of urban areas that experienced a land area reduction.

A review of the resulting urban area boundaries yielded satisfactory results and the areas were deemed final and ready to use for 2006 urban area delineation. Table 3.3 summarizes the land area change in urban areas by province between 2001 and 2006, showing the difference between the 2001 published land area and the 2001 post-censally modified land area.

Statistics Canada's land area measurements are unofficial, and are used only for calculating population density. The population density is calculated, using land area at four decimal places. The land area presented in these tables is calculated and stored in square kilometres at the block level, and then aggregated to the higher level geographic units. It is rounded to two decimal places. Land area data for the standard geographic areas reflect the boundaries in effect on the geographic reference date for the Census of Canada, which is January 1 of census year.

It is important to note that the land area between censuses differs, particularly between 1996 and 2001, because the methodology for calculating land area changed. Land area for 1996 was calculated manually whereas it was calculated using an automated approach for 2001. Table 3.4 shows the total and urban land areas for the provinces and territories in 1996, 2001 (modified) and 2006 censuses. Although there is a noticeable difference between the land area values in each census year, the percentage of land area that is urban remains relatively constant.

For more information on the calculation of land area at Statistics Canada, please refer to the Geography working paper, Automated Land Area Tests for the 2001 Census: Preliminary Results Using the 1996 Digital Cartographic Files, by Carolyn Weiss and Augustine Akuoko-Asibey (catalogue no. 92F0138MIE1998001).

The net effect on the population of the urban population counts was minimal. As shown in Table 3.5, the 2001 urban population of Canada was initially published at just over 23.9 million. Upon completion of the post-censal updates to all urban areas and the modification of the associated population counts, the urban population of Canada dropped by less than one-half of one percent (-0.2%). Four of the 10 provinces and all three territories experienced no change in their urban population counts. The largest decrease was experienced by Quebec, where the urban population decreased by -0.4%, or about 23,000 persons, after post-censal urban area updates. This decrease in the modified population count is consistent with the land area reductions following urban area updates in each province.

4.0 2006 delineation and calculation of adjusted counts

For the second consecutive census, 2006 urban areas were delineated using an automated approach. With improvements to the national road network and, hence, the dissemination block structure, GArDS delineated 2006 urban areas using current census population counts, urban areas as defined for the 2001 Census, and dissemination blocks as defined for the 2006 Census. The 2001 delineation experience provided a set of expected errors for the 2006 delineation process, including potential over-bounding and inappropriate urban area mergers. Thus, the GArDS process was modified where possible and minimal manual updates were required.

Evaluation of the resulting 2006 urban area structure yielded some results that required further investigation. First, an unexpectedly high number of urban areas were retired and the impact of post-censal land area modifications was required to ensure that areas were not inappropriately removed. Secondly, population decline was experienced in many of the urban areas that remained in the program for 2006. As such, a quality assessment of the calculation of previous census adjusted population counts (2001 urban area population counts adjusted to 2006 urban area boundaries) was required.

Thirty-one urban areas were retired for 2006; this count is higher than in past censuses. An assessment of these was necessary to determine if the land area adjustment was an important factor in the retirement of any of the urban areas. Of those 31 retired urban areas, only five (or 16%) had post-2001 land area reductions greater than 40%. The boundaries of these areas were reviewed and the modifications deemed valid. Using 2001 Census population counts adjusted to the updated 2001 urban area boundary, it was determined that two of these five retired urban areas had a modified population count below 1,000, which would have excluded them from the 2001 urban area program if they had been more precisely delineated in 2001. The other three had modified population counts greater than 1,000; these were, therefore, deemed to be retired due to 'standard' population decline (natural decrease, out-migration).

The remaining 26 (84%) retired urban areas had little to no post-2001 land area updates (22 UAs had 0% adjustment, one UA experienced a 2% increase in land area and three UAs had less than 10% reduction) and were considered to have experienced true population decline. Thus, they were appropriately removed from the program as they no longer met the population requirement. One of these, Moose Factory (former UA 35 1273), is an area that was identified for the 2006 Census as an incompletely enumerated Indian Reserve. Concerns were raised that the area may have been incorrectly removed from the 2006 urban area program. However, given the information that was available following census collection activities, it was determined that it no longer met the population criteria, thus the urban area was retired.

The high rate of urban area retirement, combined with the lack of new urban areas, reduces the total number of urban areas to 895 in 2006, down from 913 in 2001. A brief summary of the 2006 delineation results shows that Saskatchewan had the highest number of urban areas retired from the program (nine) followed by Quebec and Manitoba, each of which saw five of its urban areas retired in 2006 (see Table 4.1). A complete summary of the changes to urban areas for 2006 is presented in Appendix C.

The second result that required further investigation was the population decline in a higher than expected number of 2006 urban areas. Given the high number of urban areas that experienced post-censal land area adjustment, it was necessary to determine how much of the population decrease was due to shrinkage of the limits, as opposed to a decrease in the number of people within the area that remained part of the urban area. Such an assessment would provide an indication of the quality of the calculation of 2001 Census population counts adjusted to 2006 Census urban area boundaries, shown in Table 4.2.

To do this, the 2001 modified and the 2001 adjusted populations of the urban areas that experienced post-censal land area reduction were reviewed. Of the 412 UAs that experienced land area reduction, 219 (or 53%) had modified population decline between 2001 and 2006, compared to 208 (or 51%) that had an adjusted population decline. To assess the nature of the urban population decline, all urban areas were reviewed in two different ways: first at a dissemination block level, then according to urban area population size groups.

Dissemination block analysis reviewed all 2001 blocks that belonged to a 2001 urban area and related them to 2006 urban areas. The blocks were then categorized according to the criteria presented in Table 4.3.

Table 4.3 shows that about 98% of all 2001 blocks completely remained in an urban area following post-censal boundary update and 2006 urban area delineation. These blocks are considered to be the core blocks of all 2006 urban areas since they are the common intersect between 2001 and 2006 final urban areas. After linking these core blocks to 2006 dissemination blocks to review the adjusted population, it was determined that they contained 99% of the 2001 adjusted population for all 2006 urban areas, an excellent indication that the population decline was not due to post-censal boundary updates.

Table 4.4 shows that all provinces and territories, with the exception of Newfoundland and Labrador and Nunavut (at 90% and 73%, respectively) had core dissemination block adjusted population proportions above 98%.

Urban areas were then assessed by urban population size groups. According to the 2006 Census Dictionary, an "'urban population size group' refers to the classification used in standard tabulations where urban areas are distributed according to the following predetermined size groups, based on the current census population":

The final and adjusted population counts of all urban areas within each urban population size groups were summed for the 1996, 2001 and 2006 Censuses and the results displayed graphically. The expected results were smooth linear trends, regardless of direction (population growth or decline), although most groups were expected to experience population increase. The actual results, in general, indicated much variance, particularly between the final and adjusted counts in 2001 (see Figure 4.1 to Figure 4.8). Neither the final nor the adjusted population counts for each of the groups, with the exception of groups D, G and H, presented results in the expected trend. An important consideration at this point was it is likely for urban areas to switch size groups between censuses, particularly if the population of an urban area borders another size group.

Figure 4.1
Urban population of urban area size A group (1,000 to 2,499) - uncontrolled, 1996, 2001 (final and adjusted), and 2006

Figure 4.2
Urban population of urban area size B group (2,500 to 4,999) - uncontrolled, 1996, 2001 (final and adjusted), and 2006

Figure 4.3
Urban population of urban area size C group (5,000 to 9,999) - uncontrolled, 1996, 2001 (final and adjusted), and 2006

Figure 4.4
Urban population of urban area size D group (10,000 to 24,999) - uncontrolled, 1996, 2001 (final and adjusted), and 2006

Figure 4.5
Urban population of urban area size E group (25,000 to 49,999) - uncontrolled, 1996, 2001 (final and adjusted), and 2006

Figure 4.6
Urban population of urban area size F group (50,000 to 99,999) - uncontrolled, 1996, 2001 (final and adjusted), and 2006

Figure 4.7
Urban population of urban area size G group (100,000 to 499,999) - uncontrolled, 1996, 2001 (final and adjusted), and 2006

Figure 4.8
Urban population of urban area size H group (500,000 and over) - uncontrolled, 1996, 2001 (final and adjusted), and 2006

Once the size groups were controlled to include only urban areas that remained in the same size groups between censuses, expected population trends started to emerge. The smallest urban areas, those with a population less than 5,000 (groups A and B), showed population increases between 1996 and 2001, and population decreases between 2001 and 2006, regardless of whether the counts were final or adjusted. All other urban area size groups showed incremental population growth between all three censuses.

To assess whether the population decline in the two smallest urban size groups was due to true population decline or to post-censal boundary corrections, the urban areas were further controlled by assessing only those dissemination blocks that remained completely within both the 2001 urban areas following adjustment and the 2006 urban areas (i.e., blocks that were common, or represent the intersect of 2001 adjusted and 2006 final urban areas). Results showed that population of the small urban areas experienced continual decline between 2001 and 2006. Although the population counts of the controlled dissemination blocks were lower than those of the complete set of dissemination blocks belonging to the 2001 and 2006 urban areas, the population trends were consistent. This indicates that the population decline experienced by urban areas in these size groups was indeed due to standard population decline in those areas, rather than to urban area adjustment (see Figure 4.9 to Figure 4.16).

Figure 4.9
Urban population of urban area size A group (1,000 to 2,499) - controlled, 1996, 2001 (final and adjusted), and 2006

Figure 4.10
Urban population of urban area size B group (2,500 to 4,999) - controlled, 1996, 2001 (final and adjusted), and 2006

Figure 4.11
Urban population of urban area size C group (5,000 to 9,999) - controlled, 1996, 2001 (final and adjusted), and 2006

Figure 4.12
Urban population of urban area size D group (10,000 to 24,999) - controlled, 1996, 2001 (final and adjusted), and 2006

Figure 4.13
Urban population of urban area size E group (25,000 to 49,999) - controlled, 1996, 2001 (final and adjusted), and 2006

Figure 4.14
Urban population of urban area size F group (50,000 to 99,999) - controlled, 1996, 2001 (final and adjusted), and 2006

Figure 4.15
Urban population of urban area size G group (100,000 to 499,999) - controlled, 1996, 2001 (final and adjusted), and 2006

Figure 4.16
Urban population of urban area size H group (500,000 and over) - controlled, 1996, 2001 (final and adjusted), and 2006

5.0 Conclusion

Notwithstanding the population decline in the smaller urban population size groups, a continual decrease in the number of urban areas in Canada since 1996, and the post-censal boundary updates of 2001 urban areas, the population of urban areas, in general, is increasing (see Table 5.1). This trend is consistent with the expected results evaluated while assessing the quality of 2001 urban area population counts adjusted to 2006 urban area boundaries.

The change in the urban areas delineating methodology in 2001 from a manual enumeration area-based approach to an automated block-based approach has improved the efficiency and effectiveness of urban area delineation. Although 2001 urban areas were over-bounded, post-censal updates and more rigorous post-delineation evaluations have resulted in more reasonable urban area boundaries for 2006. Moreover, the quality of the 2006 final and 2001 adjusted population counts have been determined to be good.

To continue to improve the quality of our geographic areas, some improvements could be considered for 2011 urban area delineation that may contribute to the relevance of census output available at the urban area level. These include more stringent population regulations during delineation and the creation of applied regulations to take incompletely enumerated areas, such as Indian Reserves, into account. Since urban areas are delineated based on current census population counts prior to rounding and suppression, quantitative measures, such as confidence intervals, could be applied so that urban areas with an unrounded population count falling just below 1,000 (995 to 999) would be included. Now that data are available at the block level, methodology should be created to consider those areas facing data suppression before delineation. It should also be considered that urban area delineation could be based on rounded and suppressed counts. Now that delineation is automated and takes considerably less time, block-level rounding and suppression could be run prior to urban area delineation.