An approach for measuring urban deprivation change: the example of East Montre al and the Montre al Urban Community, 1986 ^ 96

Transcription

1 Environment and Planning A 2001, volume 33, pages 1901 ^ 1921 DOI: /a3450 An approach for measuring urban deprivation change: the example of East Montre al and the Montre al Urban Community, 1986 ^ 96 Peter Kitchen Department of Geography, University of Ottawa, 60 University Private, Simard Hall, Ottawa, Ontario, Canada K1N 6N5; pkitchen@uottawa.ca, pkitchen@magma.ca Received 23 July 2001 Abstract. The concept of urban deprivation has emerged as an important theme in urban geography and urban studies. In light of the significant problems and issues affecting cities in North America and Europe there is a growing need for researchers to acquire a better understanding of the process of urban social change. In this paper the author proposes a model of urban deprivation change and applies it with the use of census indicators to East Montre al, a section of the central city, and the Montre al Urban Community. The changing spatial structure and nature of urban deprivation is examined at the neighbourhood level in this study area over the period 1986 ^ 96. The author finds that urban deprivation is largely an economic problem, with declining conditions apparent throughout a large part of Montre al's central city and in several inner-suburban municipalities. The model, selection of indicators, and statistical techniques presented in this paper can be used as a guide for further research into urban deprivation change in other urban centres. 1 Introduction Within the fields of urban geography and urban studies increasing attention has been paid to the concept of urban deprivation. A number of researchers have attempted to conceptualize deprivation as it relates to disadvantaged people and areas. Townsend (1993, page 79) defines deprivation as a ``state of observable and demonstrable disadvantage relative to the local community or the wider society to which an individual, family or group belongs''. In the literature, the concept has been linked with poverty, unemployment, and other social, economic, and health-related problems. The majority of research on urban deprivation has focused on British and US cities. In particular, in the United Kingdom during the past ten years there has been growing interest in using census indicators and statistical techniques to examine the geography of urban deprivation in the British urban system as a whole and in individual cities. Major contributions have been made by Townsend (1993), Pacione (1995), and Fieldhouse and Tye (1996), emphasizing among other things the multidimensional nature of urban deprivation and its geographical character as well as its social and material components. By comparison, less research has been conducted on defining and applying this phenomenon in the Canadian context. In recent years, however, authors such as Broadway and Jesty (1998) and Ley and Smith (2000) have presented important studies on the incidence of urban deprivation in Canadian inner cities and the association between deprivation and immigrant groups in large urban centres. Many studies of urban deprivation in North America and the United Kingdom are conducted for a single point in timeöusually corresponding to the latest census year (Begg and Eversley, 1986; Bradford et al, 1995; Broadway, 1989; Langlois and Kitchen, 2001). Although this approach is effective in identifying important spatial patterns and their related issues, there is a need to explore more fully the process of urban social change occurring within cities by comparing indicators of deprivation over several points in time.

2 1902 P Kitchen The objective of this paper is to make a further contribution to the literature by proposing a model of urban deprivation change, and applying it, with the use of census indicators, to East Montre al, a section of the central city in one of Canada's most important metropolitan areas. At the same time, the changes occurring in the study area are compared with the situation in the larger Montre al Urban Community (MUC). Essentially, Montre al was selected as a study area because of the significant problems and changes that have occurred in the city over the past two decades. It has one of the highest rates of unemployment among Canadian metropolitan areas, one of the lowest median family incomes, and the largest share of low-income families (representing more than one third of all families in the central city in 1996). East Montre al contains a number of the city's poorest and most disadvantaged neighbourhoods and as an industrial areas was particularly hard hit by the effects of deindustrialization, economic change, and recessions during the 1980s and 1990s. As in other parts of the city, thousands of jobs were lost in East Montrëal with the closing of shipyards, railyards, shoe factories, and canneries. Figure 1 is a map of the MUC (also referred to as the Island of Montre al) showing its muncipal boundaries; figure 2 is a map of East Montrëal. (1) Langlois and I used data from a single census year (1996) to identify and measure dimensions of urban deprivation in the Montre al metropolitan area (Langlois and Kitchen, 2001). In the process, we devised a general deprivation index (GDI), which can be applied to study the spatial aspects of this phenomenon in other Canadian cities. In the present paper I take a different approach by examining the changing spatial structure of urban deprivation in East Montre al and the MUC between three time-points: 1986, 1991, and There is a significant and growing need for researchers and planners in Canada and other countries to study the complex and changing urban social geography of neighbourhoods and subneighbourhood areas over time. For The City of Montre al East Montre al Rest of City Figure 1. The Montre al Urban Community (the Island of Montre al), showing the City of Montre al, which includes the study areaöeast Montre al. (1) The municipal boundaries appearing in figure 1 are those of the old Montrëal Urban Community (MUC); these will `disappear' on 1 January 2002 with the merger into the new City of Montre al.

3 An approach for measuring urban deprivation change km Principal roads Generalized land use Residential Recreation and park Industrial Figure 2. East Montrëal. example, in the context of the study area several important questions need to be answered, including: What changes have taken place in these communities? Have these areas improved, declined, or remained stable over a period of time? What are the changes in the economic, social, and cultural circumstances of the residents living in these communities? How do these changes relate to levels of urban deprivation? The model of urban deprivation change described in this paper includes three hypotheses relating to the condition of urban areas over a period of time: improvement, stability, and decline. It will be used to study the changing geographic structure of urban deprivation at the neighbourhood (census-tract) level in the study area and the MUC. The model will be linked to indicators of urban deprivation drawn from the 1986, 1991, and 1996 censuses. The improvement hypothesis would be supported by a decrease in the concentration of urban deprivation in East Montre al between 1986 and 1996 and an improvement in the condition of the area's neighbourhoods. The stability hypothesis would be supported by little or no change in the concentration of urban deprivation over the study period and stable conditions in the neighbourhoods. The decline hypothesis would be supported by an increase in the concentration of urban deprivation in East Montre al and a worsening of conditions in its neighbourhoods. 2 A review of studies of urban deprivation change In this section I provide a brief review of several studies of urban deprivation that followed the longitudinal approach by analyzing data from several census years. Among the early work was Davidson's (1976) examination of social deprivation in Hull, England. The aim of the study was to assess changes in the spatial distribution of social deprivation in the city by analyzing 1966 and 1971 census indicators. As initial hypotheses, Davidson proposed three conceivable directions of change: (1) the status quo hypothesis, which would be supported by a situation where there were no significant change in the pattern of concentration or dispersion of deprivation in the city; (2) the convergence hypothesis, which would be supported by a reduction in the level of

4 1904 P Kitchen differentiation between areas of the city; and (3) the divergence hypothesis, which would be upheld by increasing spatial inequalities. Changes in the facets of deprivation were considered with the use of four indices: absence of inside WC (washroom facility), overcrowding, unemployment, and unskilled work.values for these indices were collected for enumeration districts in Hull for 1966 and Davidson concluded that, according to the four indices of deprivation, Hull did not experience an increase in spatial inequalities between 1966 and More recently, in 1994, the Policy Studies Institute in London, published Urban Trends 2, a study of the extent of change in Britain's most deprived areas between 1981 and It examined the thirty-six most deprived areas in Britain with use of data from the 1981 and 1991 censuses. These are the areas that the government listed as needing special attention and support. A number of dimensions were considered in measuring the changing conditions: (a) population structure, (b) employment and unemployment, (c) education and training, (d) housing, and (e) poverty, mortality, and crime. It was concluded that the analysis showed a mixture of progress and setback in the most deprived areas of Britain. On some measures, such as overcrowding and the level of housing amenities, conditions improved in deprived areas. However, the evidence revealed that levels of disadvantage are high in most deprived areas ``and that the gap in living standards between deprived areas and other kinds of areas is in some crucial respects no smaller'' (PSI, 1994, page 105). With respect to poverty and death rates, many deprived areas are worse than ever, compared with other places. Broadway (1992) examined differences in inner-city deprivation in seven Canadian cities: Toronto, Vancouver, Edmonton, Winnipeg, Regina, Halifax, and Saint John. Location quotients were used to examine changes in the concentration levels of inner-city deprivation between 1971 and 1986 with respect to several indicators, including population change, employment levels, unemployment, labour-force participation, and poverty. It was found that all of the inner cities had disproportionate concentrations of deprived groups. Broadway and Jesty (1998) examined changes in inner-city deprivation levels between 1981 and 1991 for the twenty-two largest cities in the Canadian urban system. Three variables were selected from the 1981 and 1991 censuses: the percentage of economically active persons unemployed, the percentage of persons aged 15 years and over with less than a grade 9 education, and the percentage of lowincome families. The data were collected for each census tract within the inner city for Canada's twenty-two largest metropolitan areas. The analysis of the data showed an absolute increase in levels of unemployment and low-income families among most Canadian inner cities during the study period. However, the study also revealed that there was little evidence to support a widening of disparities in inner-city deprivation levels within the urban system. Ley and Smith (1997; 2000) have contributed important work on the association between immigration and deprivation in Canadian cities, including Montrëal. In 1997 they examined the geographic concentration of urban poverty and levels of immigration in Canada's three largest citiesötoronto, Montre al, and Vancouver. They took a longitudinal approach, examining changes in spatial patterns at the census-tract level in the three cities between 1971 and They concluded that concentrated poverty does involve recent immigrants. However, they also revealed that the most extensive area of concentrated poverty was in the deindustrialized east-end districts of Montrëal that have not experienced international immigration. In their second paper, Ley and Smith (2000) studied the patterns of deprivation and their relations to immigrant distributions between 1971 and 1991 in the same three cities. Five indicators of deprivation were analyzed at the census-tract level: lack of a high-school diploma, and levels of male unemployment, government payments,

5 An approach for measuring urban deprivation change 1905 female-led families, and low-income families. They found that there was a positive but modest relationship between the five indicators and immigrant characteristics and that there was a suburbanization of deprivation over the twenty-year period, particularly in Toronto. Drouilly (1996) conducted a comprehensive time-series analysis of Montre al's changing demographic, social, and economic geography at the neighbourhood level between 1951 and The study included mapping a number of deprivation-related indicators such as levels of unemployment, labour-force participation, low-educational attainment, and low income; these were mapped over five census years: 1951, 1961, 1971, 1981, and A `spreading' of urban deprivation (according to several variables) was clearly demonstrated over the forty-year period from the city's core to other parts of the MUC and to the off-island suburbs. Also, the City of Montre al (2001) maintains a socioeconomic atlas on its website that illustrates changes in geographic patterns of certain deprivation indicators, including levels of unemployment, labour-force participation, and low education (less than grade 9), and changing income levels. Several of the maps allow for a cross-sectional comparison between the 1991 and 1996 censuses and clearly identify deprived areas in the central city, including Hochelaga ^ Maisonneuve, Parc-Extension, and Petite-Bourgogne. 3 A model of urban deprivation change In this section the model of urban deprivation change is described. It was applied to East Montre al and the MUC and was used to examine changes in the geographic structure and variation of urban deprivation at the neighbourhood (census-tract) level during the study period: 1986 ^ 96. The model was linked to indicators of urban deprivation drawn from the 1986, 1991, and 1996 censuses. The construction of an urban deprivation index is described in detail in the next section. The model is illustrated in figure 3, in which the multidirectional characteristics of urban social change are displayed. The model consists of three hypotheses and six outcomes. The model can be applied to and tested on areas such as municipalities, postal codes, census tracts, and enumeration districtsöat scales where sufficient data are available. The three hypotheses of the model referring to the condition of urban areas over a period of time and the six possible outcomes derived from these hypotheses are listed in the caption to figure 3 (see over). The characterization of `healthy' and `deprived' areas is assumed to be in relation to a benchmark or threshold. For example, the threshold for a healthy area can be conditions (based on social, economic, and housing indicators) that are found to be above the municipal, metropolitan, or regional average. Conversely, the threshold for a deprived area can be conditions that fall below the average. The assessment of these conditions and determining if areas are above or below a benchmark can be achieved, for example, by developing composite scores, based on a number of indicators, for areas such as census tracts. If a neighbourhood within the study area is found to have a higher composite score than the metropolitan area on variables such as levels of unemployment, low educational attainment, poverty, and old housing it can be considered deprived. If, however, a neighbourhood scores low on these measures and has a lower composite score it can be considered healthy. It is important for studies of deprivation to consider healthy and deprived areas because of the extensive variability that often exists within cities, particularly in North America. In some cases, deprived and healthy neighbourhoods are very near to one another, creating steep social gradients. The model first determines, through the analysis of census data, whether a neighbourhood is `deprived' or `healthy' at one point in time. It then gauges the extent and direction of change for each area over a period of time. First, as shown in figure 3,

6 1906 P Kitchen Healthy (3) Improvement Healthy (1) Stability Deprived (2) Deprived (4) Hypotheses of change: stability (deprived or healthy areas); improvement (deprived or healthy areas); decline (deprived or healthy areas). Possible outcomes derived from the hypotheses: (1) healthy (stability), healthy areas remain healthy; (2) deprived (stability), deprived areas reamin in this state; (3) healthy (improvement), healthy areas become healthier; (4) deprived (improvement), deprived areas experience improving conditions; (5) healthy (decline), healthy areas become less healthy; (6) deprived (decline), deprived areas suffer worsening conditions. Deprived (6) Decline Healthy (5) Figure 3. A model of urban deprivation change. a `healthy' or `deprived' neighbourhood can remain stable, meaning that has not changed from its previous stateöa healthy area remains healthy, and a deprived area remains deprived [outcomes (1) and (2)]. Second, a `healthy' or `deprived' neighbourhood can improve over a period of time [outcomes (3) and (4)]. Third, a `healthy' or `deprived' neighbourhood can decline over a period of time [outcomes (5) and (6)]. Therefore, for every neighbourhood, there are three hypotheses of change (stability, improvement, and decline) and six possible outcomes. 4 An urban deprivation index In this section the construction of the urban deprivation index used in conjunction with the model is described. Operationally, urban deprivation is defined as a set of social, economic, and housing conditions and problems experienced in an urban area that places some residents at a relative disadvantage compared with the population as a whole. The variables contained in the index were drawn from the 1986, 1991, and 1996 censuses corresponding to the study period (1986 ^ 96). The geographic unit of analysis was the census tract. Census tracts are small standardized units of measurement used to divide large urban communities. They are designed to be homogeneous with respect to population characteristics, economic status, and living conditions. Census-tract boundaries can change considerably between censuses, creating potential problems for spatiotemporal research. The problem of boundary changes is discussed below. The index consists of a set of fourteen variables relating to social, economic, and housing characteristics (table 1). It was constructed in consideration of a range of studies on urban deprivationöinternationally (Broadway, 1989; Fieldhouse and Tye, 1996; Pacione, 1995), nationally (Broadway, 1992; Broadway and Jesty, 1998; Ley and Smith, 1997; 2000), and for the specific case of Montrëal (Drouilly, 1996; City of Montrëal, 2001).

7 An approach for measuring urban deprivation change 1907 Table 1. An urban deprivation index. Variable Description PopChg Rate of population change (1981 to 1986, 1986 to 1991, 1991 to 1996) UnempMale Male unemployment rate UnempFem Female unemployment rate MalePart Male labour-force participation rate FemPart Female labour-force participation rate UnempYou Youth unemployment rate LoneFem Percentage of lone-female-parent families Grade9 Percentage of population with an education less than grade 9 NoHSDip Percentage of population with some high-school education but no diploma OldHouse Percentage of old housing (built before 1946) AvgDwel$ Average value of dwellings ($) LowIncFam Percentage of families living on low income MalePov Percentage of males living in poverty (total income less than $10 000) FemPov Percentage of females living in poverty (total income less than $10 000) Note: all dollar amounts are in Canadian dollars. With respect to other local studies of deprivation, the Montre al Island School Council (MISC, 2001) has developed to global index of disadvantaged families and has mapped the spatial distribution of these families in relation to the location of primary and secondary schools in the MUC. The index is based on a special tabulation of family poverty for each school district and is modified to take into account three additional variables: the percentage of families where the father is not working, the percentage of lone-female-parent families, and the percentage of households requiring major repairs. Se guin (1998) analyzed fourteen socioeconomic variables from the 1991 census in her study of the characteristics and geography of poor neighbourhoods in the Montrëal region. These included levels of male and female unemployment, labour-force participation and full-time employment, low educational attainment, lone-femaleparent families, total and recent immigrants, and median household income (Se guin, 1998, page 229). In addition, Renaud et al (1996) used 1991 census data to construct a similar index of poverty in their study of the socioeconomic geography of Montre al. All of this work provided insight into the development of the urban deprivation index for this paper, including the selection of appropriate variables and the spatial unit of analysis (the neighbourhood) as well as statistical methods and mapping techniques. The index was limited to 14 variables partly because a consistent set of indicators was required for the three census yearsö1986, 1991, and Other variables, such as recent immigration and housing needing major and minor repairs, were considered but were omitted because of changes in definitions and the unavailability of some indicators between censuses. There are no changes in the definitions of the fourteen variables over the three censuses. However, adjustments were made to the poverty indicator to reflect changes in income levels and the cost of living. From the list, it is clear that levels of unemployment, low educational attainment, low income, and poverty are direct measures of urban deprivation. As described, most of these have been included in other urban deprivation indices. Population decline was included as an indicator because a neighbourhood with a declining population can be viewed as a less desirable place for residents to live and for businesses to invest in. The problem of population decline, particularly in the central cities of many North American urban centres, has received considerable attention in urban social geography for several decades. In the literature, it has been linked to issues including the flight of middle-class residents to the

8 1908 P Kitchen suburbs, the retention of disadvantaged groups such as the poor, the departure of businesses and jobs, a shrinking tax base, and the aging of infrastructure (Bourne, 1980; Bradbury et al, 1982; Clark, 1982; Ley, 2000). However, it is important to note that in some cases population decline can be an indication of gentrification and social upgrading in a neighbourhood, including several communities in Montrëal's central city. The lone-female-parent variable is not a direct measure but is included because of the problems associated with these families, including poverty, unemployment, and dependence on social assistance. Lone-female-parent families are becoming increasingly visible in North American cities. Many of these families are living in poverty and are dependent on social assistance. Single mothers often lack access to employment opportunities and find it difficult to advance through education and training. Labour-force participation was included in the index because a low rate can reflect a lack of job opportunities or a loss of resolve in seeking work. With the economic restructuring of the central city during the 1980s and 1990s and a growing polarization of the workforce this variable is particularly useful as an indicator of urban deprivation. According to Broadway (1989, page 539), however, a low rate of participation may be the result of other factors: ``Other possible explanations for a low rate would include high student and elderly populations and cultural constraints on participation by women.'' The old-housing variable has been disputed as an appropriate measure of urban deprivation because it is proposed that many old housing units are as sound and attractive as new ones. This indicator is included in the study for several reasons. The age of the housing stock is a useful indicator of the vintage of the city government's capital stock and associated maintenance and replacement expenses. Also, numerous studies in the USA have found that the presence of older housing in central cities inhibits population growth and promotes out-migration. An increase in the proportion of old housing over a period of time could mean that an area has become less attractive for new development (Bradbury, 1984; Frey and Speare, 1988). Related to this, the average value of dwellings is used as an indication of the attractiveness and condition of housing in a neighbourhood. The index includes one low-income and two poverty variables: the percentage of families living on low-income, and the percentages of all males and females aged 15 years and over living in poverty. The low-income variable is included in each census and is based on Statistics Canada's low-income cut-off (LICO). The LICO standard is that families or individuals who spend 54.7% or more of their pretax income on food, clothing, or shelter are in financial difficulty. For example, the 1996 census stated that a family of four residing in an urban area with a population of or more was considered to be living on low-income if its income was less than Can $31753 in The low-income family variable is used as a direct measure of urban deprivation. Statistics Canada, however, does not recommend applying the LICO as a measure of poverty. In the absence of an official poverty line, the index sets a threshold of Can $ for 1995 (as presented in the 1996 census) as a measure of poverty for all males and females. The Can $ poverty threshold was selected in accordance with the 1996 census income categories. Also, this figure represents weekly earnings of about Can $200 and a minimum hourly wage. In order to conduct time-series analysis, this level was adjusted to reflect changes in inflation and the cost of living. The consumer price index (CPI) was used to calculate that Can $ in 1995 was equivalent to Can $8981 in 1990, and to Can $7245 in A custom tabulation was then obtained from Statistics Canada to determine the total number of males and females aged 15 years and over with income less than these levels for the three census periods.

9 An approach for measuring urban deprivation change Methods of analysis Two approaches can be followed to conduct a geographic analysis of urban social change: cross-sectional and vertical (analysis of change). The most frequently used is the cross-sectional approach, involving presenting separate `snapshots' of a study area at a certain scale and examining changes in structure from one point to another. Examples of studies employing this approach include King's (1966) use of factorial ecology to examine Canada's changing urban system between 1951 and 1961, Guay's (1978) analysis of the changing dimensions of urban social space in Montre al over three census periods (1951, 1961, and 1971), and Mata's (1996) study of labour force and ethnicity in Toronto with use of 1981 and 1991 census data. The method described in this paper follows the vertical (analysis-of-change) approach, which is used far less frequently in urban social geography. It is vertical in the sense that the changes in the nature and level of urban deprivation in each neighbourhood in East Montrëal and the MUC between 1986, 1991, and 1996 are examined and compared. This approach is essential for acquiring a more detailed understanding of the processes of change occurring in individual communities at the intraurban level. A good example of the vertical approach is Le Bourdais and Beaudry's (1988) study of the changing residential structure of Montre al, in which they combined social, economic, and housing variables from the 1971 and 1981 censuses to conduct a direct analysis of change over that decade. Another is Perle's (1983) examination of urban social change in Detroit, MI, in which he applied principal components analysis (PCA) to a change matrix containing forty-two socioeconomic variables at the censustract level from the 1960 and 1970 censuses. He stated that although the `vertical' or `analysis-of-change' approach has been suggested as a possibility for study it is rarely applied and less well understood. Perle (1983, page 309) states: ``the decomposition of change matrices represents an attempt to understand social change more directly than by comparing cross sections, and this approach will yield an alternative set of perspectives on social change''. The data analysis for this paper involved several steps. The first step was to establish a consistent census-tract configuration for 1986, 1991, and 1996 for the entire MUC, including the study area, and to construct the related data matrices. The EDIT function in ArcView (ESRI, was used to remove or add arcs on the 1996 coverage to make it geographically compatible with the 1986 and 1991 coverages. The attribute census data were then modified for the affected census tracts to create consistent datasets for the three time-points, based on the fourteen variables. This ensured that all of the census tracts (or areas representing several census tracts) could be compared over the study period. The three data matrices consisted of all the census tracts in the MUC, allowing the changes occurring in each neighbourhood in East Montrëal to be examined in relation to any other on the Island. The second step was to combine the data for the three time-periods (1986, 1991, and 1996) into a single matrix. A PCA was performed on the matrix to identify groups of intercorrelated variables and to assess the structure of change with respect to urban deprivation. PCA was selected over factor analysis as it is used often as a data-reduction technique and does not make any assumptions with respect to common and unique variance (Janson, 1980; Shaw and Wheeler, 1994). Moreover, it is appropriate under the assumption that the error and specific variance represent a small portion of the total variance in the original set of fourteen variables. PCA replaces a set of variables with a smaller number of components. The components are made up of intercorrelated variables representing as much of the original dataset as possible. The principal components were subjected to a varimax rotationöa common procedure in PCAöto improve their interpretability (Shaw and Wheeler, 1994). Component loadings

10 1910 P Kitchen are used to identify groups of intercorrelated variables in a dataset. It is from these loadings that a component score for each case in the original data matrix is calculated. Component scores are standardized by computing their z-values, which usually have a range of 3.0 to 3.0. This standardization permits the relative position of each case (census tract) on each component to be compared. It is important to point out that the combining of data from several censuses into the same matrix is not standard practice in factorial ecology. In this paper, however, it was attempted for two reasons. First, it allows for `vertical' analysis on a component structure that is unchanged between timepoints, creating standardization in the dataset, and, second, the analysis involves a relatively short time-period (10 years). The component loadings were interpreted and PCA was used to assign a component score to each census tract. Every tract has a standardized z-value on each component for the three time-points. For example, census tract 060, located in the study area close to downtown, has a score for components 1, 2, and 3 for 1986, 1991, and Several examples of the application of factor analysis and PCA in studies of urban deprivation include Braodway's (1989) comparative study of Canadian and US inner cities, Fieldhouse and Tye's (1996) examination of the ecological fallacy in studies of deprivation in Britain, and Renaud et al's (1996) study of Montre al's urban social space. The third step was to calculate change variables for all the census tracts across the principal components for three time-periods: 1986 ^ 96, 1986 ^ 91, and 1991 ^ 96. This is achieved, for example, in the first period simply by subtracting the 1986 score from the 1996 score. The process was repeated for the two other time-periods, creating a total of three change matrices. Analysis of these variables provided insight into the nature and extent of urban change in the neighbourhoods of East Montre al and the MUC. The change variables were related to the model of urban deprivation: negative scores (indicating an increase in levels of urban deprivation) reflect decline; scores close to the mean (0; no or little change) reflect stability; and positive scores (decrease in urban deprivation) reflect improvement. An index was devised to reflect the various categories of urban change based on the scores: for example, deprived areas becoming more deprived (decline), healthy areas becoming more healthy (improvement), and healthy areas staying the same (stability). The statistical parameters of this index of urban change are described in detail in the next section. Finally, the change scores were mapped with ArcView to illustrate the geographic pattern of urban deprivation change in the study area and the rest of the MUC. The results of the analysis were compared for the three periods to test the three main hypothesis. 6 An analysis of change: 1986 ^ The nature and extent of change As described, a PCA was performed on the data and a three-component solution was derived, accounting for nearly two thirds of the total variance. Table 2 shows the loadings for each of the fourteen variables. Component 1 is a composite measure labeled `socioeconomic deprivation', with its high loadings on male and female unemployment, youth unemployment, lone-female-parent families, male poverty, and low-income families. Component 2 is labeled `education and participation', with its high loadings on labour-force participation (particularly female) and the two education variables (achieving less than grade 9, and receiving no high-school education). Component 3 is referred to as `population decline and old housing', with its higher loadings on these two indicators. The fourth component accounted for about 7% of the total variance

11 An approach for measuring urban deprivation change 1911 Table 2. Principal components analysis (combined 1986, 1991, and 1996 data matrix): component loadings for the Montrëal Urban Community (varimax rotation). Variable Component a PopChg MalePart FemPart UnempMale UnempFem UnempYou Grade NoHSDip LoneFem OldHouse AvgDwel$ MalePov FemPov LowIncFam Percentage variance cumulative a Component 1, socioeconomic deprivation; component 2, education and participation; component 3, population decline and old housing. and its loadings could not be interpreted sufficiently. As result, the decision was made to retain only the first three components for further analysis. The next step was to examine the extent of urban change in the study area according to the three hypotheses. As outlined in the methodology, this was achieved by calculating change scores for each census tract for three time-periods. Table 3 displays the mean change scores for the MUC, for East Montre al, and for the `rest of the central city'. Negative scores signify declining conditions whereas positive scores Table 3. Mean change scores a in the Montrëal Urban Community (MUC), East Montrëal, and the `rest of the central city': 1986 ^ 96. Area Component b ^ 96 MUC East Montre al Rest of the central city ^ 91 MUC East Montre al Rest of the central city ^ 96 MUC East Montre al Rest of the central city a Negative scores signify declining conditions; positive scores signify improving conditions. b Component 1, socioeconomic deprivation; component 2, education and participation; component 3, population decline and old housing.

12 1912 P Kitchen point to improving conditions. What is immediately obvious in the table is that during the entire study period (1986 ^ 96) there was significant decline in the MUC on component 1, substantial improvement on component 2, and some improvement on component 3. With respect to `socioeconomic deprivation' (component 1), the average change score for the MUC was East Montrëal ( 0.495) recorded an average change score lower than the MUC and substantially less than the `rest of the central city' ( 0.822). The greatest decline occurred in three inner suburban municipalitiesömontrëal-nord ( 1.051), Saint-Laurent ( 1.032), and Saint-Lëonard ( 0.910). Surprisingly, on the `education and participation' dimension (component 2), East Montrëal had a high positive score (0.765), signifying considerable improvement. With respect to the `population decline and old housing' dimension (component 3), the study area had an average change score of This was close to those of the MUC and the `rest of the central city', pointing to some improvement during the first study periodöpopulation stability and a drop in percentage of old housing. An examination of the scores for the two subperiods (1986 ^ 91 and 1991 ^ 96) reveals several interesting trends. First, during the first five-year period, East Montrëal experienced virtually no change in its level of `socioeconomic deprivation' (0.059), whereas the `rest of the central city' and the MUC as a whole witnessed only slight declines ( and 0.121, respectively). As shown in table 3, overall, much of the decline in component 1 occurred during the second half of the study period. During this time, most of the areas listed had higher negative scores than they did in the 1986 ^ 91 period. In particular, East Montre al and the `rest of the central city' witnessed substantially worsening conditions of `socioeconomic deprivation' ( and 0.626, respectively). Table 3 clearly indicates that in the MUC much of the improvement on the `education and participation' dimension (component 2) occurred during the 1986 ^ 91 period. East Montrëal recorded a relatively high positive score (0.514). It can also be seen in table 3 that during the 1991 ^ 96 period there was a leveling off with respect to change in component 2. However, although the score for the MUC indicated virtually no change (0.072), the study area recorded the highest positive score (0.251). Similarly, on the `population decline and old housing' dimension (component 3) there was overall improvement during the first five-year period, followed by a flattening of average change scores in the second period. Keeping in mind that component 3 accounted for only 9.5% of the total variance and had only two significant variable loadings, this result basically points to the fact that there was a greater fall in the percentage of old housing in MUC neighbourhoods in the first five-year period than in the second. The overall drop in percentage of old housing is related, in part, to city policies promoting residential intensification by encouraging new housing construction on nonresidential sites. 6.2 Application of the model of urban deprivation change As described, the model consists of three hypotheses (improvement, decline, and stability) as well as six related outcomes. The statistical parameters of the model are outlined in table 4. The model first establishes the condition of the area in 1986 and 1991 based on component scores (the area being either deprived or healthy). The model then determines the direction of change over a period of time. Based on their change scores, census tracts are then grouped into a number of categories. As seen in the table, deprived tracts can experience severe decline [category (A)], decline [category (B)], stability [category (C)], or improvement [category (D)]. Healthy tracts can experience decline [category (E)], stability [category (F)], or improvement [category (G)].

13 An approach for measuring urban deprivation change 1913 Table 4. Statistical parameters of the model of urban deprivation: condition in 1986 and 1991, and change over time. Deprived area (negative component score): (A) severe decline (< 1.5 SD) (B) decline ( 0.5 to 1.5 SD) (C) stability ( 0.5 to 0.5 SD) (D) improvement (>0.5 SD) Healthy area (positive component score): (E) decline (< 0.5 SD) (F) stability ( 0.5 to 0.5 SD) (G) improvement (>0.5 SD) Note: SD, standard deviation. In table 5 (see over) the incidence of urban change according to component 1 (`socioeconomic deprivation') for the ten-year period is shown, as well as for the two subperiods. It indicates that between 1986 and 1996 a total of forty-eight deprived census tracts in East Montrëal suffered decline, including eight (5.1% of the total tracts in that area) experiencing severe decline. A further thirty-six (23.1%) remained stable, and twenty-one (13.5%) improved. Only fifty-one tracts in the study area were healthy in 1986 and of these thirty five (22.4%) suffered decline (became less healthy). In total, 53.1% of neighbourhoods in East Montrëal endured declining conditions [categories (A), (B), and (E)] with respect to `socioeconomic deprivation'. This means that, in a composite manner, they saw increases in male, female, and youth unemployment, in lone-female-parent families, in poverty, and in low-income families. Table 5 compares urban change in the study area to the remainder of the central city and conveys several interesting results. Incidence of severe decline [category (A)] occurred mostly in the `rest of the central city'; stability [category (C)] and improvement [category (D)] was much more evident in East Montrëal. In addition, cases of healthy tracts in decline [category (E)] were more visible in the `rest of the central city'. In total, about two thirds of all neighbourhoods in the `rest of the central city' (66.9%) witnessed decline on component 1 compared with 53.1% in the study area. Furthermore, only sixteen tracts in categories (A) and (B) were located outside the central city. However, the proportion of tracts in category (E) (healthy area in decline) was split almost evenly between the central city and the suburbs. A total of 46% of tracts in this category were located in suburban municipalities. Table 5 also displays the incidence of urban change on component 1 for the two subperiods. It is apparent that stability and improvement [categories (C), (D), (F), and (G)] were more evident in the first five-year period, the incidence of decline [categories (A), (B) and (E) being significantly higher in the second. Table 6 (see over) shows the incidence of urban change according to component 2 (`education and participation') for the ten-year period as well as for the two subperiods. It is immediately apparent that the overwhelming majority of census tracts in the MUC saw either stability or improvement on component 2 between 1986 and In fact, only six tracts recorded decline. The table also clearly shows that a large number of the study area's neighbourhoods were in category (D) (deprived areas experiencing improvement), representing 46.8% of the area's total tracts. A further thirty-three tracts were in category (G) (healthy areas becoming healthier). In total, nearly 68% of the census tracts in East Montre al experienced improvement on component 2. By comparison, the `rest of the central city' had far fewer of its tracts in category (D) (thirty five, or 25.2% of the area's total number of tracts) but about the same number in category (G) (healthy area experiencing improvement) and significantly more in

14 1914 P Kitchen Table 5. Incidence of urban change in the Montrëal Urban Community (MUC) for component 1 (`socioeconomic deprivation'). Area Number of deprived tracts a Number of healthy tracts a (A) (B) (C) (D) (E) (F) (G) 1986 ± 96 East Montre al percentage of total b percentage of area's tracts c Rest of the central city percentage of total b percentage of area's tracts c Total (including suburbs) ± 91 East Montre al percentage of total b percentage of area's tracts c Rest of the central city percentage of total b percentage of area's tracts c Total (including suburbs) ± 96 East Montre al percentage of total b percentage of area's tracts c Rest of the central city percentage of total b percentage of area's tracts c Total (including suburbs) a For a definition of categories (A) ± (G), see table 4. b Number of census tracts in the given area in the specified category as a percentage of the total number of tracts in the MUC in that category. c The number of tracts in the given area in the specified category as a percentage of the total number of tracts (deprived and healthy) in that area. category (F) (healthy area with stability). In total, 48.2% of its neighbourhoods witnessed improving conditions on component 2. All tracts outside the central city recorded either stability or improvement, with the largest number in category (F) (healthy area with stability). Overall, this improvement is primarily a reflection of a growing level of educational attainment and an increasing rate of labour-force participation by females in the Montrëal region, likely precipitated by economic restructuring and the creation of new jobs in the service sector and knowledge-intensive industries. As table 6 also reveals, the incidence of change on component 2 for the first subperiod (1986 ^ 91) closely mirrors that of the ten-year period, with the large majority of census tracts witnessing improvement or remaining stable. However, a different trend is evident in the second five-year period. Between 1991 and 1996 significantly fewer neighbourhoods in the MUC experienced improvement, and a higher number were listed as declining (forty-six tracts).

15 An approach for measuring urban deprivation change 1915 Table 6. Incidence of urban change in the Montrëal Urban Community (MUC) for component 2 (`education and participation'). Area Number of deprived tracts a Number of healthy tracts a (A) (B) (C) (D) (E) (F) (G) 1986 ± 96 East Montre al percentage of total b ± percentage of area's tracts c Rest of the central city percentage of total b ± ± percentage of area's tracts c Total (including suburbs) ± 91 East Montre al percentage of total b ± percentage of area's tracts c Rest of the central city percentage of total b ± ± percentage of area's tracts c Total (including suburbs) ± 96 East Montre al percentage of total b ± percentage of area's tracts c Rest of the central city percentage of total b ± percentage of area's tracts c Total (including suburbs) ± Figure not calculable. a For a definition of categories (A) ± (G), see table 4. b Number of census tracts in the given area in the specified category as a percentage of the total number of tracts in the MUC in that category. c The number of tracts in the given area in the specified category as a percentage of the total number of tracts (deprived and healthy) in that area. 6.3 The geography of urban deprivation change The geographic pattern of urban deprivation change in East Montre al and the MUC was put into focus by mapping the change scores for component 1. Figure 4 (see over) illustrates the spatial distribution of change with respect to `socioeconomic deprivation' (component 1) for the ten-year study period. The map shows that between 1986 and 1996 declining conditions were apparent throughout a large part of the Island. Deprived census tracts suffering decline [categories (A) and (B)] were visible primarily in East Montre al and the `rest of the central city'. Within the latter area, they were strongly present in Saint-Henri, a large part of the Sud-Ouest, Coª te-des-neiges, Notre-Dame-de-Graª ce, and in Parc Extension to the north. Cases of severe decline [category (A)] were dispersed between the study area and the `rest of the central city'. As seen in the map, healthy tracts witnessing decline [category (E)] were evident largely in suburban areas of the MUC. The inset in figure 4 provides an enlarged view of urban deprivation change within the boundaries of the study area of East Montrëal. It displays deprived tracts [categories

16 1916 P Kitchen Category of urban deprivation change, 1986 ^ 96: Deprived or distressed area in 1986 (A) Severe decline (<1.5 SD) (B) Decline ( 0.5 to 1.5 SD) (C) Stability ( 0.5 to 0.5 SD) (D) Improvement (>0.5 SD) Healthy area in 1986 (E) Decline (< 0.5 SD) (F) Stability ( 0.5 to 0.5 SD) (G) Improvement (>0.5 SD) no data Figure 4. Urban deprivation change in the Montrëal Urban Community at the census-tract level, 1986 ^ 96: component 1 `socioeconomic deprivation'. Inset: East Montrëal study area. Note: SD, standard deviation. (Based on Statistics Canada data, 1986; 1996.) (A), (B), (C), and (D)] in the western and southwestern portions of the study area, and healthy tracts [categories (E), (F), and (G)] in the northern and eastern sections. Figure 4 shows that there was a pronounced geographic concentration of severe decline [category (A)] and decline [category (B)] on `socioeconomic deprivation' in the area south of Sherbrooke Street between Saint-Laurent and Viau Boulevards (see figure 2 for a map of the area). This declining corridor includes the eastern part of downtown as well as the communities of Centre-sud, Terrasse Ontario, and Hochelaga ^ Maisonneuve. However, figure 4 also indicates that this section contained several improving neighbourhoods [category (D)], including the district surrounding Old Montre al. Just to the north, the Plateau Mont-Royal was characterized primarily by stability and improvement [categories (C) and (D)]. These two areas have been identified in the literature as communities benefiting from gentrification and social upgrading (Ley, 1996). Another clear pocket of decline [category (B)] was visible in St-Michel in the northern limit of the study area. Furthermore, in East Montrëal there was a number of healthy tracts suffering decline [category (E)] and these were visible in parts of Saint- Edouard, Villeray, and Rosemont in the north, and in Longue-Pointe in the east. In summary, through the application of the model of urban deprivation change it was found that between 1986 and 1996 most deprived census tracts suffering decline on component 1 (`socioeconomic deprivation) were, not surprisingly, located in the central city. However, the total incidence of decline (including healthy areas in decline) was far more evident in the `rest of the central city', with about 70% of its neighbourhoods in this state, compared with about 53% in East Montrëal. The model also found that decline was apparent primarily in the 1991 ^ 96 period. On component 2 (`education and participation' the vast majority of census tracts in the MUC