Data and Mapping Notes

The following are technical notes to assist the user in understanding how The Atlas of Canada mapped the Service Industries module. They have been organized into the following topics:

Service-employment Data and Mapping

• Employment Data
• Geographical Units
• Methodology: Regression Analysis
• Maps of Growth in Services

Commercial Land-use Data and Mapping

• Commercial Structure
• Data for Commercial Structure

Service-employment Data and Mapping

Employment Data

The employment data used in the maps come from the 1996 Census of Population (20% sample data) produced by Statistics Canada. (The maps of growth of service employment use similar measures for 1986.) Employment of an area is defined as the number of employed residents at the date of census (refer to the glossary or the 1996 Census Dictionary for more information on census-related terminology).

The service-employment data are aggregated into the two-digit categories of the standard industrial classification (SIC) for each census subdivision (municipality), and then aggregated into the set of urban and rural geographic units used for the maps. The various kinds of services and service combinations require different groups of the two-digit categories. There are nine standard industrial classification categories in wholesale; seven categories in retail; seven financial categories; seven categories in the commercial services, including business service; education; health; and three categories of government. The service maps differentiate between the private-sector services (wholesale, retail, finance, personal and leisure services) and the public-sector services (education, health and public administration). Each of these two broad groups is then further broken down into sectors for mapping.

The maps of change in employment took advantage of the fact that the standard industrial classifications were unchanged between 1986 and 1996, so the categories are comparable over time. Because the geographic units have changed, however, the data were first geo-coded to each census subdivision in each time period, and then the data for each time period were aggregated into the 1996 geographic units.

Geographical Units

Since the number of workers is measured at the place of residence and the maps are concerned with the place of work, it is necessary to aggregate the residential locations into a geographic unit that links workplace with residence. In this way, the number of persons reported as employees will approximate the number of workers employed by local businesses. The most appropriate geographic units to use are the urban areas defined by Statistics Canada. The census metropolitan areas include the largest cities (with at least 100 000 population in the central city), and the census agglomerations are the smaller centres with population of at least 10 000. As well, some of the census metropolitan areas and census agglomerations have been grouped together into consolidated census metropolitan areas, in which the component units are called primary census agglomerations and primary census metropolitan areas.

Tables 1-1 to 1-5 summarize the number of geographic units of various kinds and sizes that appear on these maps. The urban places total 159, with a small number of very large places and larger numbers of smaller centres - a typical distribution of city sizes. Altogether, urban places contribute 77.8% of the total population in 1996.

Table 1-1. Metropolitan Areas and Census Agglomerations Geographical Unit, 1996

Table 1-2. Consolidated Census Metropolitan Areas Geographical Unit, 1996¹

Table 1-3. Total Urban Places Geographical Unit, 1996

Table 1-4. Rural Residuals Geographical Unit, 1996

Table 1-5. Canada Total Geographical Unit, 1996

¹ Includes Hull, Quebec as a separate unit within the Ottawa-Hull consolidated census metropolitan area.

Census subdivisions not included in the urban areas were assigned to census divisions as residual census divisions. Those census division residuals with population less than 10 000 were combined with adjacent census division residuals. Twenty-eight out of 254 census division residuals were combined in this way, resulting in a total of 226 rural units. For the most part, these rural units are small in size with a maximum population of 87 000 (for example, Haldimand-Norfolk, Ontario).

In order to recognize the enormous size variation among the urban centres, which range in population from 10 000 to more than 4 million (Toronto, Ontario), the cities (and residual census divisions) are represented by symbols that are proportional to their population. To be precise, the area of the symbol is proportional to the population being represented. The cities are represented by circles and the residual census divisions by squares. The values for service specialization assigned to each location are derived from the regression equations for the cities, and the appropriate values are then calculated for each of the residual census divisions. Thus, the cities collectively define the specialization scale and the rural areas are assigned to the scale afterwards. After the ranges of values for each quintile have been determined, the residual census divisions are assigned to their legend categories according to their values.

Regression Analysis Procedure

In order to overcome the influence of market distribution on the location of services, it is necessary to develop a series of ratios or indices that compare the level of service activity to the size of the market. This could be done simply by calculating service activity on a per-capita basis, or per million dollars in market income. The resultant map would still be related to the city size distribution; however, because larger cities consistently have higher ratios of service activity, it is best to compare the observed level of service activity (the actual employment) to an expected level of service activity, as predicted by some combination of market measures.

The statistical procedure for generating the expected levels of service employment is called regression. It uses information for the entire set of cities to generate a kind of average or predicted value for each individual city, based on that city's market characteristics. In this case, the market measures used to predict service activity are employment, population and income per capita.

How is the level of employment estimated on the basis of population and income? The usual method is regression analysis, and a log-linear regression model should be applied because the distributions of city size and income per capita are log-normal. The model estimates the amount of commercial employment that should be generated in a community of given population and income level, and compares the estimate to the observed value. The difference (either positive or negative) is a measure of centrality.

The regression has the form:

log (employment) = A + B₁ log (population) + B₂ log (income per capita)

All three variables in the analysis (employment, population, income per capita) are converted to logarithms so that the intercept A is interpreted as a measure of scale (technically, the amount of employment when population and income per capita are zero). The regression coefficients B₁ and B₂ indicate the relationship between employment and the two independent variables. The coefficient R² is the coefficient of determination that indicates what proportion of the variance in employment is explained by the other two variables.

Consider the equation for total commercial employment (wholesale, retail, finance and commercial services):

log (employment) = -3.060 + 1.039 log (population) + 0.507 log (income/capita)

R² = 0.989

The fact that the regression coefficient for population (1.039) is greater than one indicates that the commercial employment increases disproportionately with population. A city that is ten times larger than another (plus one on a log scale) will have eleven times more service employment. In contrast, the regression parameter for income per capita is less than one, which suggests that an increase in income per capita is only partially captured as local consumption. The money may be taxed, invested elsewhere, spent in larger cities or used for travel. Finally, R² indicates that the equation predicts 98.9% of the variation in commercial employment, which is a substantial achievement. The variations mapped in this set of plates are fairly minor deviations within the overall patterns.

The equation provides an ‘all other things being equal’ prediction of local commercial employment, to which each urban area can be compared. The city with the largest residual (more jobs than expected), and therefore the greatest relative centrality, is Grande Prairie, an agricultural community serving a prosperous agricultural region in the Peace River district of Alberta. The residual from a log relationship is actually the logarithm of the ratio of the observed to expected values, and the value for Grande Prairie translates into a ‘surplus‘ of 36.0% of the predicted commercial employment. This amounts to about 2160 jobs more than the expected total of 6005; in other words, Grande Prairie as a market is 36% larger than it appears to be. At the other end of the list, the largest negative residual occurs in Kitimat, British Columbia, which generates 36.7% fewer jobs than predicted (1405 jobs instead of 2225). As a prospective store location, the Kitimat market is substantially smaller than it appears.

Maps of Growth in Services

Maps of growth pose different problems in analysis and mapping. The employment data include the number of workers in each service category and each geographic unit for the years 1986 and 1996. The difference in the employment totals (1996 value minus 1986 value) is called the absolute growth; the absolute growth divided by the 1986 value is called the growth rate.

The procedures used to map growth in service employment are quite simple. For each service sector or combination of sectors, the 1996 employment data for each location are compared to the 1986 employment for the same sector. The map of absolute growth shows where the growth has taken place across the country over the decade, among regions and within regions, as the growth in service employment has both responded to, and stimulated growth in, the urban markets. The maps of absolute growth use symbols (circles) for each city or residual census division (squares) that are proportional in area to the number of jobs added. Positive and negative growth are shown in different colours.

Given the enormous variation in the size of cities, however, a map of absolute growth in service employment is unable to communicate events within a particular city. Toronto is far larger than Orangeville (Ontario) and both are growing by different amounts, but is one city doing better than the other? The growth rates tell us how many jobs each city has added relative to its size, so that one can be compared against the other. The maps of relative growth begin with the set of symbols that represents the population of cities and residual census divisions, as in the maps of specialization, but the colours of the symbols represent quintiles of growth rates as defined for the cities.

Commercial Land-use Data and Mapping

Commercial structure can be defined as the geographic distribution of commercial activity within a metropolitan area. This includes the number, size and location of various kinds of commercial concentrations, such as downtown, shopping centres or pedestrian-oriented strips. The six elements of commercial structure that are shown in these maps are described and defined below.

Elements of Commercial Structure

• Downtown: This is the concentration of activities that serves the entire urban region and includes specialized retail, financial and business services, as well as public-sector facilities. These activities may be organized into subareas according to function and market. Typically, downtown is the oldest part of the city and the most accessible location overall. But every downtown is different!
• Shopping Centres: These are privately owned and operated facilities housing a number of stores, mostly retail, that are linked together by pedestrian flows within the mall. Centres are often isolated from other commercial facilities by parking lots and streets. Shopping centres vary widely in size, so that types of stores, store sizes and location vary with shopping-centre size.
• Pedestrian Strips: These are streets containing individually owned stores - a mixture of retail and service -linked together by pedestrian flows and closely integrated within a local market. Some are specialized by function (antiques) or to serve a special market. They tend to emerge in the older built-up areas of cities, or as former town centres in newer suburbs.
• Arterial Strips: These are stores located on an arterial road or highway that provides access for customers who come in cars. Customer linkages among the various stores are rare; customers come from the market area served by the road system. The stores share a common requirement for visibility and accessibility. Activities include auto sales and repair, fast food and small strip plazas.
• Industrial Zones: These are extensive areas zoned for industrial use that may also include wholesalers, big-box stores, and auto and other services. There are no internal links among stores and visibility gives way to low rent as a location advantage. Customers must seek out these facilities.
• Dispersed Stores: These activities remain after all the commercial polygons have been designated, and the stores assigned to various categories. Dispersed stores include traditional isolated activities, such as service stations and convenience stores, as well as more conventional store clusters that are simply too small to qualify as polygons, especially in smaller cities. Big-box stores may fall into this group.

Note that office buildings with concentrations of business and financial services can be part of any of these elements, but the offices shift the functional composition of the business location toward these sectors (and reduce the share of corporate outlets).

Data source: Simmons and Kamikihara, 2002a

Data for Commercial Structure

The process begins with more than 1 million business entries for Canada in an electronic telephone directory. The stores were classified by type of commercial activity (standard industrial classification), and 650 000 stores were extracted that belonged to the commercial service categories: retail, finance, business, leisure and personal services. (Wholesalers were excluded because of the confusion with manufacturing firms.) For each of the target cities, the stores were geo-coded (located on a map) within their respective cities. Given the geography of stores for each city, a geography of commercial locations was then imposed in the form of commercial polygons. A polygon is a commercial area composed of at least 25 stores, and/or 50 000 square feet of floor area in the case of malls. Analysts defined the boundaries of each polygon on the map by hand in order to best separate commercial and noncommercial land uses (Simmons and Kamikihara, 2002a).

Polygons have been identified for 81 of Canada's largest cities, including all census metropolitan areas and all census agglomerations with populations of more than 30 000. The maps of commercial land use show the proportion of stores in each city that belongs to each of the six types of commercial location. Each map shows circles proportional to the size of the census metropolitan area or census agglomeration. The cities are ranked according to each commercial location variable and assigned to quintiles, which are colour coded.

Back to previous page