Figures 3, 4 and 5 show the relative uniformity of the dispersions around the predicted trend lines and the residuals do not even hint at obvious thresholds to indicate the tipping of neighbourhoods into gentrification as discussed by Meen and Meen (2003) and Galster, Quercia and Cortes (2000). Since our definition of gentrification describes low-income neighbourhoods that are “invaded” by middle- and upper-income households, the census tracts in the lowest income quartile in 1981 for each CMA were selected from the set with positive principal component scores. Since the definition of “gentrification” applies to older neighbourhoods that were mostly renovated rather than redeveloped only the tracts that had in 2001 the largest proportion of older dwellings were selected. The largest proportion was defined as the tracts in the highest quartile proportion of pre-1946 stock for each CMA. The two screens eliminated tracts with middle and higher income households in 1981 and the tracts that had (or never had) a large proportion of pre-World War II stock by 2001. The principal component analysis and the screening process selected 249 tracts out of the 2,182 census tracts in the 10 CMAs as having, at least, the symptoms of gentrification.
The arbitrariness of the cut-offs used to select the gentrifying tracts is not a problem as the second phase of work draws on the knowledge of key informants to identify gentrifying tracts. The interviews were undertaken with Canada Mortgage and Housing Corporation (CMHC) housing market analysts in each regional office. CMHC market analysts, among other things, are responsible for the preparation of housing market reports containing rental vacancy rates, rents, housing starts and completions, and are thus knowledgeable about the macro changes and factors affecting the regional housing market. Furthermore, targeting key-informants who share the same job-title/description in each of the CMAs provided a degree of certainty regarding the common knowledge among the key informants. We probed for the factors driving the gentrification of particular neighbourhoods and asked for comments regarding the tracts we had identified through the quantitative methods. Overall, combining qualitative analysis of information from key-informants with an analysis of census data overcomes some of the limitations of relying solely on tract-level census data to identify gentrifying neighbourhoods as noted by Wyly and Hammel (1999, 727).
The interviews identified the census tracts that were clearly seen to be gentrifying in 2001 and reduced the count from the 249 identified by analyzing census data to 94, or 4.3 percent of all tracts.4 None of the tracts outside the set identified by the quantitative method were seen by the key informants as gentrifying. While the quantitative method identified tracts in each of the 10 CMAs, the key informants did not see Regina as having any gentrifying neighbourhoods. The definition of gentrification as implied by the key informants’ selections is more restrictive than the definition we proposed explicitly in the introduction. The estimates produced by the key informants understate the extent of gentrification that has taken place in the city as will be demonstrated later. In many cities, neighbourhoods changes in the 1970s and by 1981 their housing prices were too high to encourage further change (Ley, 1993).
The Classification of Census Tracts
The quantitative work followed by key informant interviews produced three categories of census tracts:
i) Gentrifying Census Tracts: This category includes census tracts that were selected by the screening of census variables (249 tracts) and identified by key-informants as likely gentrifying neighbourhoods. This category represents 94 of the 249 census tracts.
ii) Potential Gentrified Tracts: These census tracts were selected by the screening of census variables but not selected by key informants as gentrifying neighbourhoods in 2001. This category represents 155 of the 249 census tracts and may include tracts that had gentrified earlier.
iii) Other Census Tracts: This category includes the 1,933 tracts not selected by the threefold screening of variables from the 1981 and 2001 censuses. It includes tracts that had finished gentrifying before 1981. The extent and location of the tracts in Montreal, Toronto and Vancouver will be included and displayed in the maps.
Figures 9 and 10 show how the three classes on Montreal tracts fit into the CMA’s income and income change profiles.5 The vertical axis is the ratio of 2001 to 1981 average tract personal income and the horizontal axis is the 1981 average tract personal income in Figure 9 and the 2001 income in Figure 10. The plots distinguish the three classes of tracts and show them to be intermingled at the low end of the 1981 income scale. They have, as a result of the selection process, well above average increases in income. Figure 10 places the 2001 average tract personal income on the x-axis. Both the “potential” and the “gentrified” tracts appear to cascade over their neighbouring income groups as they move upward on the income scale. They do not move to the top of the income distribution but they clearly reduce the stock available to lower income people.