Review of the Roots of Youth Violence: Research Papers

Volume 4

A Methodology to Identify Communities in Ontario Where High or

Increasing Relative Disadvantage May Lead to Youth Violence

A Report Prepared for the Review of the Roots of Youth Violence

Desmond Ellis Ph.D.
La Marsh Research Centre on Violence
and Conflict Resolution
York University


Executive Summary

The primary objective of this proposal is to describe a methodology that will identify neighbourhoods in Ontario as priorities for interventions aimed at alleviating disadvantage. The proposal opens with a review of methodologies implemented in national and international studies on poverty. Methodologies implemented in them are reviewed and evaluated with respect to their study designs, sampling designs, definition/measurement of major concepts, data collection and data analysis. Findings from the review/evaluation provide the basis for implementing the methodology to be used for achieving the primary objective of the proposal. Salient features of this methodology are:


This proposal is presented in three segments. In the first one, research methods used in post-1986 studies on poverty are described, focusing on study designs, sampling, definition/measurement, data collection and data analysis. The second segment is devoted to a critical evaluation of the methods used in the studies reviewed in the first segment. Using the results of the evaluation, the third segment presents a proposal for a methodology to identify neighbourhoods in Ontario characterized by high or increasing relative disadvantage that may lead to youth violence.

The proposal describes five salient methodology features. First, is the implementation of a longitudinal design. Second, a population rather than a sample of DAs in Ontario is to be selected. Third, a multi-dimensional definition of disadvantage is to be applied, which includes economic, material, and social conditions and the creation of an IRD that summarizes disadvantage in five different domains (housing, family, education, employment and income). Fourth, information about levels of disadvantage in each of Ontario’s 19,177 DAs is to be collected from the Statistics Canada census for two census years, 2001 and 2005. Fifth, common factor analysis is to be used to create an IRD. This analytical procedure also yields a score indicating the level of disadvantage characterizing a DA. The scores are to be used to rank DAs according to their levels of disadvantage and to compare them with appropriate benchmarks.

Description of Research Methods in Previous Studies*

The studies described and evaluated in this proposal fall into three study design categories: cross-sectional, longitudinal and cross-national.

Cross-Sectional Design Studies

In his seminal publication, The Truly Disadvantaged: The Inner City, the Underclass and Public Policy (1987), William Julius Wilson described and explained the concentration of poverty in the inner-city census tracts (ghetto neighbourhoods) of Chicago and other major American cities. Ghetto neighbourhoods were characterized by “joblessness, teenage pregnancies, female-headed households, welfare dependency and serious crime” (p. 3). Ghetto neighbourhoods; that is, areas made up of adjacent census tracts; are also characterized by a high (30%) or a very high (40% or higher) proportion of families living below the poverty line. The poverty line used by Wilson was constructed by combining the yearly amount spent on purchasing enough cheap food to feed a family (of varying sizes) for a year with the proportion of annual family income devoted to buying food (Social Security Administration in 1964). Major findings reported by Wilson include the urbanization of poverty, the concentration of poverty in inner city areas (adjacent census tracts) and marked increases in the number of ghetto neighbourhoods between 1960 and 1980.

Encouraged and assisted by Wilson, Hajnal investigated “the nature of concentrated urban poverty in Canada and the United States” (1995). The Canadian part of his comparative, cross-sectional study included an examination of concentrated urban poverty in the census tracts of Canada’s 25 consolidated metropolitan areas (CMAs) in 1986. Using pre-tax LICO’s to determine poverty lines and a definition of a “severely disadvantaged neighbourhood” as a census tract with a 40% poverty rate, Hajnal found that poverty was geographically widespread across Canadian cities and concentrated in urban areas. He also found that male employment, the percentage of persons earning income from government transfers, the percentage of persons aged 15 and older with less than nine years of high school education, and the percentage of residences built 20 or more years earlier discriminated “all neighbourhoods” from “over 40% poor neighbourhoods.”

Hou and Miles (2004) investigated linkages among neighbourhood inequality, relative deprivation and the self-perceived health status of neighbourhood residents. They implemented a sample design that focused on CMAs. Census tracts were defined as “the basic neighbourhood unit” (p. 8). Median family income for individuals residing in a census tract/neighbourhood was calculated on the basis of their adult-equivalent adjusted family incomes.

Relative deprivation was not defined. It was conceived of as an intervening variable; that is, a variable mediating the impact of neighbourhood income inequality and health. Five domains of relative deprivation were selected. They were: education (percentage of adults with a university degree), age (percentage of persons aged over 65), family (percentage of single-parent families) immigrants (percentage of persons living in Canada for 10 years or less), and race (percentage of non-white persons) (p. 11).

Hou and Miles collected health data from 34,592 individuals who responded to Statistics Canada’s 1996–1997 National Population Health Survey. Covariate data were collected from residents of one CMA residing in 3,044 census tracts. Because of the ordinal level of measurement, the ordered logit model was used to analyze the data. Hierarchical lineal models were used to analyze neighbourhood-level observations that were not independent of each other. One of the major findings reported by these authors was that self-perceived health status of poor individuals improves when they live in the same neighbourhoods as richer, better-educated individuals do.

Using tax-filer data, the authors of Falling Behind: Our Growing Income Gap (Federation of Canadian Municipalities, 2007) examined the concentration of poverty in the Letter Carrier Walks (LCWs) of Calgary and Saskatoon and the Forward Sortation Areas (FSAs) of Toronto for the year 2000. One of their major findings had to do with the proportion of tax filers in low-income neighbourhoods who reported low incomes: 47% in Saskatoon and 43% in Toronto. They also found that the neighbourhoods in Toronto and Calgary were “more mixed;” that is, they included a greater variety of family types and income groups (p. 55).

Finally, they found that low-income neighbourhoods were more widely distributed in Toronto compared with Calgary and Saskatoon.

Longitudinal Design Studies

Hajnal’s cross-sectional study was described as “an important baseline study” by Canadian poverty researcher Kazemipur (2000:3). One reason for this attribution was the extant body of Canadian poverty literature, which Hajnal had found to be “almost completely devoid of any mention of concentrated urban poverty” (1995:499). Since 1995, a number of researchers have made significant contributions to the literature on the geographic distribution of poverty in Canada by conducting longitudinal investigations of this topic.

In their 1997 study, MacLachlan and Sawada investigated income inequality in 22 of the largest CMAs in Canada. To that end, they compared the distribution of household incomes across the census tracts of these cities for the years 1971, 1981 and 1991 (p. 1). Using average household income to measure the distribution of income in these census years, MacLachlan and Sawada found that household income inequality increased in the census tracts of all 22 cities over the 20 year period. They also found that intra-city differences in household income inequality were highest in five of the cities in their sample: Toronto, Montreal, Winnipeg, Calgary and Hamilton. A noteworthy methodological feature of their research was the use of the GINI Concentration Ratio to measure income inequality.

In 2000, Kazemipur published Ecology of Deprivation: Spatial Concentration of Poverty in Canada. He investigated the concentration of poverty in the census tracts of 46 Canadian cities for the years between 1986 and 1996. Using pre-tax LICOs to measure census tract poverty rates, Kazemipur followed Wilson and Hajnal by defining ghetto neighbourhoods as “census tracts with a poverty rate of at least 40%” (p. 410). One of his major findings was that the proportion of ghetto neighbourhoods had increased significantly in a few cities (e.g., Montreal and Winnipeg), increased in a greater number of others (including Toronto, Windsor, and Ottawa), and decreased in a few others (St. Catharines, Guelph, and Kitchener). He also found that Toronto and Hamilton were two of six cities in which 10% or more of the census tracts were ghetto neighbourhoods.

Using four indicators of deprivation/disadvantage associated with low income, Ley and Smith (2000) compared census tracts in three CMAs, Toronto, Montreal and Vancouver, at two points in time, 1971 and 1991. The indicators selected by Ley and Smith measure disadvantage in four domains. These are education (percentage of adults completing grade 9), unemployment (percentage male), family (percentage of female lone-parent households), and income (government transfer payments relative to total tract income) (p. 43).

One of their major findings was that one census tract in Vancouver and one in Toronto (Regent Park) had high ratings on all four indicators and also had a high concentration of social housing. Six census tracts in Toronto and five in Montreal had high ratings on three indicators. Supporting the concept of an “archipelago” (p. 39), one census tract that received elevated scores on all four indicators (Regent Park) was located adjacent to two census tracts that received elevated scores on three indicators. Poverty by Postal Code researchers found that the poverty rate for families in Regent Park was almost 72.8% and the family poverty rate was 59.1% in an adjacent census tract (p. 27).

Another finding supporting the archipelago concept was reported by Wilson (1987) for ghetto neighbourhoods in Chicago and other American cities. Ley and Smith also found that deprivation/disadvantage ratings for census tracts changed over time. Between 1971 and 1991, census tracts with high ratings on multiple indicators of disadvantage were transformed into census tracts with high ratings on none of the indicators or fewer of the indicators.

In 2000, Miles, Picot and Pyper (2000) published the results of their investigation of neighbourhood inequality in Canadian cities. Their longitudinal study design used data from four sequential census years, 1981, 1986, 1991 and 1995. The sample they selected was a focused one: eight of Canada’s largest CMAs, with populations of 500,000 plus. Neighbourhoods were defined “at the level of the census tract” (p. 3). Income inequality was implicitly defined as inequality in per capita income of individuals within census tracts. As many individuals lived in families, per capita incomes were adjusted for family size. GINI indices were used to describe income inequality in the eight major cities in their sample. Within and between cities, comparisons were analyzed using an economic segregation index.

Three of their findings are noteworthy. First, the relatively stable distribution of incomes in Canada during the 1981–1995 census years masked changes in income inequality among neighbourhoods in its major cities. Second, increasing unemployment was mainly responsible for increasing economic inequality in high-poverty neighbourhoods. Third, economic spatial segregation was mainly responsible for increases in neighbourhood inequality in Toronto, Montreal, Ottawa-Hull, and Quebec City, while increasing family income inequality was mainly responsible in Vancouver, Edmonton, Calgary, and Winnipeg.

In 2004, the United Way of Greater Toronto and the Canadian Council on Social Development published Poverty by Postal Code. Poverty by Postal Code researchers identified geographic concentrations of poverty across 522 City of Toronto census tracts over a three year period, 1981, 1991 and 2001. Pre-tax LICOs were used to measure poverty. City neighbourhoods were categorized as Low, Moderate, High or Very High poverty neighbourhoods, depending upon the proportion of census families (parent/s and children) in them who fell below the average 1981 Canadian census family poverty line. In “high poverty” neighbourhoods, 26.0% to 39.9 % of families fell below the poverty line (double the national average). In “very high poverty” neighbourhoods, 40% or more fell below the poverty line (three or more times greater than the national average). Another major finding reported by the authors of this report was that the number of “high poverty” neighbourhoods increased from 26 in 1981 to 97 in 2001. The number of “very high poverty” neighbourhoods was almost six times greater, having increased from four in 1981 to 23 in 2001 (p. 20). Another finding was that the spatial distribution of concentrated poverty; that is, poverty neighbourhoods; had changed from the inverted U pattern described by Ley and Smith (2000) to an O pattern in the former cities of Toronto and Etobicoke (p. 19).

The City of Toronto is subjected to an annual “Vital Signs Check-Up” by the Toronto Community Foundation (TCF). The study designs used for their reports are longitudinal and the samples selected are focused. Concepts are not defined. The data are subjected to a number of bivariate analyses. The methodology adopted by the authors of the annual reports are appropriate for communicating with members of the public, many of whom may not understand more complex multivariate statistical analyses. Compared with the TCF reports, the TCF website presents more detailed information on each of the vital signs measured annually. The nine City Vital Signs or domains of advantage/disadvantage are: income, safety/health, employment, education, housing, transportation, recreation, environment, and ethnicity. The Low Income Measure (LIM) was used to measure poverty.

One of the major findings was a 56.5% increase in child poverty in the Greater Toronto Area (Toronto plus 23 other municipalities) during the 25-year period 1980–2005 (TCF website, 2007). Increasing income inequality in the City of Toronto was another socially significant finding. Specifically, between 1980 and 2005, the median income gap between families in the top and bottom income groups increased from approximately 5% to 10.7% (TCF website 2007:6). Another important finding was an increase in the City poverty rate from 16.6% in 2000 to 24.7% in 2005. The poverty rate reported for the City of Toronto was found to be 17% higher than the comparable rate for Ontario and almost 14% higher than the rate for Canada.

Losing Ground: The Persistent Growth of Family Poverty in Canada’s Largest City (2007) was published by the United Way of Greater Toronto and used a longitudinal and comparative study design. Four geographical locations with varying populations were compared (Canada, Ontario, the City of Toronto and the CMA excluding the City of Toronto) across three points in time (1990, 2000 and 2005).

Poverty was measured using median income and Statistics Canada’s after-tax Low Income Measure (LIM). LIM is defined as “having an income less than half the median income of a family of the same size and age composition for all of Canada” (Statistics Canada, 2004). Aftertax LIM thresholds were calculated for families (parent/s with children aged 0 to 17). Families falling below this threshold were classified as “being in poverty” (p. 25). Income data were collected from tax-filer information.

One of the major findings reported by the authors of Losing Ground was that a higher proportion of Toronto’s families fell below the poverty line (28.8%) than did families in Canada (19.5), Ontario (19.7), and the Rest of Toronto CMA (16.3). A second major finding was that in 1990, one-third of the City of Toronto’s poorest families; that is, single-parent families; fell below the poverty line. The comparable proportion for 2005 was over half (51.6%). In 2005, the median income of single-parent families was $21,700. The comparable figure for two-parent families was twice as high ($53,300) (pp. 21–22).

In their longitudinal study (1920–2000), Saez and Veall (2005) used income tax data to investigate the proportion of total income; that is, income from all sources before taxes and excluding transfer payments and capital gains; accruing to the top 1% of income earners in Canada. Their findings revealed a U-shaped curve, in which the proportion of income earned was high (17%) during the 1920s, began going down during the World War II years, reached a low point of about 8% during the 1970s/1980s, and then increased to approximately 18% during the 2000s. The extremely high incomes earned by top company executives during the 1920s and 2000s, which made them “very rich,” were mainly responsible for the pattern of income inequality in Canada during this period. In the City of Toronto, Hulchanski (2007) found the very rich to be overrepresented in “City 1,” the City’s core.

In December 2007, Hulchanski published the results of a 30-year (1970–2000) study on changes in the spatial distribution of income inequality across all 527 City of Toronto census tracts. Income inequality was measured using average individual income from all sources. Changes in census tract (neighbourhood) income, 1970–2000, were measured. Inequality was measured using income ratio differences. The benchmark against which census tract/neighbourhood changes were assessed was average individual income for the CMA. The criterion used was a 20% change up or down the income ladder.

One of the major findings reported by Hulchanski was a change in the pattern of income inequality. Over the 30-year period, a City of mixed-income neighbourhoods was transformed into three homogenous-income cities. City 1 accounted for 20% of the population and 103 census tracts where average individual income increased by 20% or more. City 2 accounted for 43% of the population and 224 census tracts where average individual income decreased or increased by less than 20%. City 3 accounted for 36% of the City population and 192 census tracts where average individual income decreased by 20% or more. (p. 1) The pattern of income inequality had changed from an inverted U in 1970 to three concentric residential rings formed by a wide band of low-income census tracts (36%) circling a narrower band of middle-income tracts (43%) circling a small core of high and very high-income tracts (20%) in 2001. Another major finding was a significant decrease in the percentage of middle-income earners, from 66% in 1970 to 32% in 2000.


The “poverty studies” in Canada that examined the period 1970–2000 yield the following conclusions:

Cross-National Studies

In addition to being included as a unit of comparison in the national Urban Poverty Project study, Canada has been included in cross-national studies of poverty. The reliability of national and international comparisons of poverty depends on the degree to which key variables are defined and measured in the same way (Brady, 2003:716). The Luxemburg Income Study (LIS) made reliable comparisons across 25 countries possible because it included harmonized measures of income inequality and poverty (Brady, 2003b). Using LIS data and a Headcount definition of poverty (percentage of the population below a certain threshold, 50% of median income), Brady (2003:75) compared market-generated poverty, a measure that excludes taxes and government transfers, among 16 Western nations during different years in the 1990s. Belgium ranked highest (40%), Switzerland ranked lowest (24%), the US ranked twelfth (31%), the UK ranked fifth (37%) and Canada ranked thirteenth (30%).

LIS does not yield data on poverty dynamics. Cross-National Equivalent Files (CNEF) provided data on poverty dynamics in Canada, the United States, Great Britain and Germany, which is important for three reasons. First, the study of poverty dynamics, (entering into, remaining in and leaving poverty) required the collection of the same data from the same households over the same period of time. CNEF provided longitudinal household panel survey data. Second, income and other variables associated with it were measured in the same way. For example, LIMs were used to measure income inequality in all four countries. Third, the countries compared were similar with respect to levels of economic determinants and development (Valetta, 2005).

Valetta measured household income. Poverty dynamics, however, operate at the level of individuals in households. Individual-level data were produced by dividing total household income by the square root of the size of the household (p. 10). Using LIMs to measure poverty “from the 1980s to the 1990s,” Valetta found that the annual poverty rate for Canada and Great Britain were the highest, 19.5 and 19.4 respectively. The comparable rates for Germany and the United States were 16.2 and 18.3 respectively. Canada was also found to have the highest percentage of individuals “always in poverty.” “Always in poverty” rates for the four countries were: Canada (8.0%), United States (5.5%), Germany (3.6%) and Great Britain (3.1%). Different study designs, different time periods and different units of analysis may account for differences in the findings of Valetta and Brady.

Valetta also identified the factors that explain poverty dynamics in the four countries. In Canada, it was one factor: family structure. In the United States, the factors were educational attainment and unstable employment in poorly paid jobs. In Great Britain and Germany, the factors were government tax and income transfer policies (p. 15).

The cross-national findings reported by Valetta lead to two conclusions. First, poverty, as measured by LIM, is a more serious problem in Canada than it is in the United States, Great Britain or Germany. Second, family structure should be included in indices used to measure relative disadvantage in Canada.

Using a different cross-national, historical (1969–1997) and harmonized source of data on poverty (LIS), Brady (2003a) found that the presence of active left-leaning political institutions increased the contribution made by governments towards decreasing poverty through tax and income transfer policies. As a result, state-mediated poverty levels were found to be lower than market-mediated levels. Brady’s analysis of LIS data for Canada revealed significant differences in market-generated and state-mediated interval poverty in each of the years 1971–1997 (p. 740). Similar findings have been reported by other researchers using the same (LIS) source of data on relative poverty (Moller and associates, 2003).

Methodological differences characterizing the foregoing studies are summarized in Table 1. Not included in the table are noteworthy differences between Canadian and European studies on poverty. Compared with the European researchers, Canadian researchers tend to:


The methodology used by researchers to measure poverty in Canada is not as robust as the methodology used by European researchers to measure poverty/disadvantage in European societies such as England and Ireland.


The following evaluation of the studies reviewed above is presented with a view to identifying some methodological contributions that are worth reproducing and others that must be added in devising a robust methodology appropriate for achieving the objective of identifying communities in Ontario where high or increasing relative disadvantage may lead to youth violence.

Study Design

(a) Cross-Sectional Studies

Cross-sectional designs possess a number of strengths. First, compared with longitudinal studies, they can be completed relatively quickly and inexpensively. Second, cross-sectional studies are contemporary with respect to their procedures, measurements and analyses. Longitudinal (panel) studies, especially those covering longer periods of time, cannot make changes reflecting progress in theory, measurement and analysis without losing their main strength, which is the ability to apply the same measures to the same respondents over time. Third, cross-sectional studies are far less likely than longitudinal studies to lose subjects due to attrition.

Cross-sectional studies also have a number of weaknesses. First and foremost among these is their static nature. Societies and their constituent parts and places are constantly changing, but cross-sectional studies do not measure change. Instead, they offer a snapshot in which time is held constant. Second, although researchers are unlikely to completely solve the problem of causal ordering of the six indicators included in the IRD, the solution offered by cross-sectional researchers is likely to be less convincing than one offered by longitudinal researchers. Only the latter can identify the time-ordering of these variables. For example, although the former may calculate reverse correlations between the income and family indicators, the latter can actually determine whether children were below the poverty line before their families became single-parent families or whether single people became poor before they had children.The social policy implications are far clearer when such a determination can be made.

(b) Longitudinal Studies

Some of the weaknesses of longitudinal studies identified by advocates of cross-sectional studies (e.g., Hirschi and Selvin, 1967) were also identified in the preceding segment. First, they are more costly in terms of money and time. Second, in some cases, less-costly cross-sectional studies yield the same findings as more-costly longitudinal studies. An example that comes to mind is investigations of the association between age and crime (Loeber and Le Blanc,1990). Third, the loss of respondents through attrition is greater in longitudinal studies than it is in cross-sectional studies. Fourth, methodological advances cannot be implemented during the course of a longitudinal (panel) study because that would destroy the unique virtues of such studies, such as retaining original definitions, measures and analytic procedures during the entire course of the project.

For many if not most poverty researchers, the strengths of longitudinal study designs greatly outweigh their weaknesses. First, neighbourhoods and communities are constantly changing, and such designs measure change. Second, they can also be used to make cross-sectional comparisons. Third, a panel design can reveal poverty dynamics.

With specific reference to studies of poverty/disadvantage, cross-sectional studies cannot be advocated on the ground that they yield findings similar to those produced by longitudinal studies. Dissimilarity may be partly due to differences in study design and partly to differences in measurement, units of analysis, years studied, and methods of analysis. Attrition is a more serious problem, but the problem is restricted to panel studies where persons are the units of analysis. Through time, the number of panel members decreases for a variety of reasons, including that panel members move or lose interest in participating. Where places (cities, census tracts, DAs) rather than people are selected as units of analysis, creation of new places (particularly census tracts and DAs) can be a problem. However, this problem is not as great as attrition, because the “new place creation rate” is usually is far lower than the person attrition rate. Finally, the studies reviewed here indicate that methodological advances have been incorporated into longitudinal studies.

(c) Cross-national Studies

Canada is part of a wider global society. Cross-national studies that include Canada have a number of strengths. First, they yield findings comparing and rank-ordering Canada and other societies according to their levels of poverty/disadvantage. Second, rank-ordering before an international audience helps mobilize action aimed at reducing poverty/disadvantage in the societies studied. Third, they facilitate multi-societal collaboration aimed at advancing the measurement of poverty/disadvantage. Examples include the Luxemburg Income Study in Europe and the UN Human Development Index world wide.

The validity and reliability of findings from cross-national studies depend on the degree to which they use harmonized measures of poverty/disadvantage. Great progress has been made in Europe and some progress has been made in North America towards harmonizing and then using harmonized measures of relative poverty/disadvantage. In many other societies, there has been far less progress towards creating and using harmonized measures of relative disadvantage than there has been in measuring absolute disadvantage (United Nations, 2006).

The use of un-harmonized measures of relative disadvantage is a weakness of cross-national studies. Global harmonized measured of relative poverty/disadvantage are unlikely to be achieved, because harmonized measures may be appropriate only for societies that are structurally and culturally similar.

Historical, comparative cross-national household/individual panel study designs using harmonized definitions and measurement yield valid and reliable findings on poverty dynamics, but they are far more costly in terms of money and time than are longitudinal studies where panel members are not the units of analysis.

Sample Design [C/R from page 48]

Large populations are rarely studied by social scientists doing macro studies, for at least three reasons. First, studying large populations costs far more than studying samples selected from them.

Second, errors associated with generalizing from samples to populations can be statistically estimated. Third, large-population studies are more likely to yield unreliable findings due to non-sampling errors associated with non-response, coverage, and coding and entering data.

Notwithstanding these disadvantages, there are circumstances under which a population rather than a sample may quite appropriately be selected for study. One of the most compelling of these is where a population study better meets the objective of the researcher. Specifically, if the objective is to alleviate disadvantage and suffering in all the places where it is concentrated, then populations should be studied. Thus, the United Nations selects populations and not samples of nations because its objective is to alleviate Absolute (biogenic) Disadvantage in all nations in which it is found to be concentrated.

Census tracts for the City of Toronto and the GTA were selected for study by Toronto Community Foundation’s Vital Signs researchers. The use of focused samples is appropriate for achieving their objective of mobilizing support for meliorating adverse conditions in and around the City of Toronto. The salience of this objective may be reflected in a tendency to select samples and analyze data idiosyncratically rather than systematically. For example, without explanation, TCF researchers report income inequality comparisons (e.g., child poverty) for the City, but not for the GTA.

A population of City of Toronto census tracts, not postal codes, was studied by Poverty by Postal Code researchers because they were interested in describing the geographic distribution/concentration of poverty in geographical units smaller than FSAs. Money costs and non-sampling error costs were probably higher in this study than they would have been had a probabilistic sample been selected for study. However, given their objective, selecting a probability sample would have been inappropriate.

Probabilistic sampling designs were not used by the authors of any of the studies of poverty in Canada reviewed herein. Instead, “focused” sampling designs were implemented. Specifically, the authors selected census tracts in CMAs or CSDs in which they expected to find a range of variation in poverty and/or or census tracts in CMAs or CSDs in which poverty was likely to be concentrated.

The use of focused urban sampling designs is appropriate for “poverty/income inequality” researchers because poverty/income inequality is concentrated in cities. At the same time, focused sampling is inappropriate for studying rural areas and First Nations reserves, where poverty/income inequality may be even more highly concentrated.

Collectively, researchers studying poverty in Canada included census tracts in over 46 Canadian cities and selected CMAs and CSDs in their focused samples. All data available for census tracts is also available in the 2001 census for DAs (75% of them clustered in the 400–700 population range). The benefits of sampling a smaller geographical unit were mentioned by Hulchanski, the Poverty by Postal Code researchers, and Miles, Picot and Pyper, but they may have considered DAs too small or too many to sample. The ratio of DAs (8,140) to census tracts (530) in the City of Toronto, for example, is 15:1. In Kingston the ratio is 6:1 and in Ontario the ratio is 9:1 (Statistics Canada, 2007b). An alternative (and more likely) reason is that sampling DAs would not permit comparisons with Statistics Canada data entry points (census) earlier than 2001.

Neighbourhood poverty researchers using longitudinal study designs using 2001 and post-2001 census data may want to consider the relative costs and benefits of sampling DAs and census tracts. The former offer a closer approximation of the definition of neighbourhoods and provide a basis for implementing more focused policies of melioration.

Definitions and Measurement

Definition and measurement are interrelated. In the present context, definitions of poverty, relative disadvantage, and neighbourhood should precede their measurement. If researchers cannot define these concepts, they cannot measure them. Most if not all methodologists would agree with this statement.

Statistics Canada does not define poverty, but uses LICOs (Low Income Cutoffs) to measure income inequality. LICO is defined as “an income threshold below which a family will likely devote ... 20 per cent more of its income on the necessities of food, shelter and clothing than the average family” (Statistics Canada, 2004:7). LICOs refer to thresholds that vary according to the size and areas of residence of families.

Statistics Canada also uses LIM (Low Income Measure). LIM is defined as “a fixed percentage (50%) of median adjusted family income,” where “adjusted” means that the needs of families of different ages and sizes are taken into account (Statistics Canada, 2004:11). Families that fall below LICO and LIM thresholds are living “in straightened circumstances.”

LICOs and LIMs are widely used. Both measure relative income inequality. Both are more appropriate for measuring income inequality in advanced capitalist societies such as Canada than they are in underdeveloped countries where an absolute “basic needs” measure is more appropriate (United Nations, 2006). Both are grounded in the Canadian cultural context in the sense that Canadian values are used to define the needs, and the level of needs, of Canadian families. To this extent, LICOs and LIMs frame families “living in straightened circumstances” as a social condition (Brady, 2003b:722).

After-tax LICOs and LIMs share the strengths noted above. Compared with these measures, pretax LICOs yield slightly higher estimates of income inequality because the progressive taxes and government transfer payments that increase the income of poor recipients are not taken into account (Human Resources Development Canada, 2003:11; Statistics Canada, 2006). In this connection, the authors of Falling Behind report that in 1998, “families in the bottom income group received 29.8% of total transfers compared with 11.9% received by families in the highest quintile” (2007:9).

Researchers who choose to use pre-tax LICOs claim that federal and provincial taxes account for approximately 40% of government revenues. If the remaining 60% of mandatory contributions to the revenues of these governments (EI, CPP premiums, GST, property taxes) were also included in the adjustments made by Statistics Canada, then pre-and post-tax measures would yield similar low income thresholds (Ross et al, 2000). This has not been demonstrated in any of the studies reviewed herein. In the absence of such adjustments, pre-tax LICOs consistently yield higher income inequality thresholds. Findings reported by Human Resources Development Canada researchers indicate that 10.9% of “All Canadians” fall below the after-tax LICO threshold, 14.7% fall below the pre-tax LICO threshold and 11.15 fall below the LIM threshold. For “Female Lone-Parent Families,” the thresholds are 33.9%, 44.2% and 35.6% respectively (p. 11).

After-tax LICOs and LIMs yield more conservative and, some would claim, more accurate estimates of the economic wellbeing of families. At the same time, they do not share other strengths to an equal degree. The greater strengths of LIMs were identified by Wolfson and Evans (1990), and served as a rationale for using LIMs in national and cross-national studies of income inequality. Today, LIMs are routinely used in international comparisons of income inequality (Brady, 2003a).

LICOs and LIMs also share weaknesses. Both are based on surveys of income. These routinely “produce a greater concentration of families at both the upper and lower tails of the income distribution and hence higher values of standard inequality measures” (Miles, Picot and Pyper, 2000:26). Researchers analyzing Luxembourg Income Study data follow bottom-and top-coding procedures that increase the validity of inequality measures (Heisz, 2004). These procedures were used in only three of the Canadian studies reviewed: Heisz (2004), Myles, Picot and Pyper (2000) and Hou and Miles (2004). Researchers who do not increase the validity of their pre-tax LICOs by using the aforementioned LIS procedures are likely to find higher levels or degrees of income inequality than do researchers who increase the validity of their after-tax LICOs and LIMs by using LIS coding procedures.

Statistical indicators of economic wellbeing typically measure either central tendency (median or average income) or dispersion (income inequality). Most researchers use a measure of central tendency (La Free and Drass, p. 615). According to Brady (2003a), such measures ignore the depth of poverty. This criticism does not apply to researchers who report and take into account in their analyses the size of the gap below the threshold into which families, households or individuals fall. For example, LIM thresholds are based on families earning/receiving less than 50% of the median gross income of Canadian (or city, census tract) families. Analyses of LIM families could, however, include families whose income fell below the threshold by 5%, 10%, 15%, 20%, and 25% below the median gross income of the comparison unit.

A second criticism stated by Brady (2003a:717) has greater merit. Both LICOs and LIMs were constructed with administrative, political or other objectives in mind. Most ”poverty researchers” attached to universities and other organizations and associations probably use official measures of income inequality, rather than poverty, because they are harmonized measures available for rather long periods of time, they are available from a source (census) that includes many covariates, and their use facilitates comparison with studies conducted by others. Over time, their use is becoming conventional.

Canadian researchers, though fully aware of European contributions to policy-relevant theory and research on poverty, have contributed to making the use of measures of income inequality conventional. For example, Raphael first acknowledges that “Canadian efforts at defining poverty have been limited,” then describes the Townsend (1993) and Rainwater and Smeeding (2003) multi-dimensional definitions of poverty, then notes that Statistics Canada regards LICOs and LIMs as indicators of low income not poverty, and then uses pre-tax LICOs to measure poverty because their use for this purpose has become conventional (2007:37).

In sum, though Statistics Canada informs users about “the absence of an accepted definition of poverty,” researchers who use LICOs and LIMs routinely equate them with poverty (e.g., Losing Ground, p. 25).

Statistics Canada also states that the use of LICOs and LIMs is limited to studying “the characteristics of relatively worse-off families in Canada” (Statistics Canada,2005), yet researchers who use these measures routinely investigate the spatial distribution/concentration of only one characteristic: low income.

Statistics Canada researchers who created LICOs and LIMs, as well as the researchers who use those measures, may be very surprised to discover that Townsend’s (1979) definition of poverty, later refined by Gordon et al (2000), has been generally accepted in Europe for some time. In his account of Poverty in the United Kingdom, Townsend defines poverty in these terms: “Individuals, families and groups can be said to be in poverty if they lack the resources to obtain the types of diet, participate in the types of activities and have the living conditions and amenities which are customary, or at least widely encouraged or approved in the societies to which they belong” (p. 13). Defined in this way, poverty is most validly measured by the degree to which families and individuals possess resources of various kinds that either prevent them from experiencing deprivation or help them escape deprivation (Townsend, 1979:131-140). Disadvantage is defined as “suffering from social and/or economic deprivation” (Oxford English Dictionary, 2006). Therefore, it can be substituted for deprivation in Townsend’s definition.

Disadvantage has been found to vary with income held constant. In this specific connection, a number of researchers report findings indicating that “people who have low income are not the same as the population who are most materially deprived” (Capellari and Jenkins, 2006:2) Additional evidence is provided by Berthoud, Bryan and Bardasi, 2004; Bradshaw and Finch, 2003; and Callan, Nolan and Whelan,1993. Therefore, low income and its material and social consequences must be included in indices and scales measuring relative disadvantage.

One lesson to be learned from a review of European studies of poverty informed by Townsend’s definition is that poverty is a multi-dimensional concept, appropriately measured by indices or scales that combine a number of dimensions or domains (Noble and associates, 2006).

Like poverty, “neighbourhood” is frequently referred to but not explicitly defined by most of the researchers whose work was reviewed herein. For example, Wilson (1987) does not define neighbourhood but frequently refers to inner-city areas characterized by “high rates of joblessness, teenage pregnancies, female headed families and welfare dependency” as “ghetto neighbourhoods” (p. 3). As his unit of analysis was census tracts, the implication is that neighbourhoods are defined in terms of census tracts.

Fully aware of the fact that census tracts “by no means perfectly define how local residents would delimit their neighbourhoods” (p. 9), Poverty by Postal Code researchers used census tracts to define neighbourhoods without defining neighbourhood because “it was the only measure available” (p. 9). Authors of the Urban Poverty Project 2007 and Miles, Picot and Pyper (2000) also used census tracts to define neighbourhoods without defining neighbourhood.

Without explicitly defining neighbourhood, authors of Falling Behind: Our Growing Income Gap (2003) used Forward Sortation Areas (FSAs), designated by the first three characters of the postal code, to define “neighbourhoods with common characteristics” (p. 8). They reported that in urban areas such as Toronto, some FSAs have 10,000 people and others have close to 60,000. Variation in characteristics within a larger FSA may be greater than variation between FSAs. Given that low income and disadvantage can vary independently, the residents of even the smallest FSAs may not fully share both of these characteristics. In our research on youth street gangs in the GTA (2005), we found that in Victoria Village, one of the “the 13 unmet-needs communities” identified by the United Way of Greater Toronto (2006), the census tract rate for the characteristic “low income” was 8% of families, but the rate for one DA within this census tract was 41%.

After rejecting the City of Toronto’s “group of 3.7 census tracts, 17,600 persons” definition of neighbourhood because they were “too large to represent the lived experience of a neighbourhood,” Hulchanski selected individual census tracts because “they come closer to that experience” (2007:3). The nature of the lived experience he had in mind is not used to explicitly define neighbourhood.

Despite these limitations, the idea of commonality or homogeneity expressed by these researchers is central to the project of defining neighbourhood. The definition becomes useful when content is identified. Specifically, reference could be made to residential propinquity, homogeneity in values, norms, beliefs, perceptions, material possessions, and attachment to and identification with place.

Generally, the smaller the geographical unit, the greater the homogeneity among its residents. In Saskatchewan and Calgary, the other two cities studied by the authors of Falling Behind, neither census tracts with a population between 2,500 and 8,000 (Statistics Canada, 2007a), nor FSAs (population between 8,000 and 60,000 households), nor postal codes (population between zero and 10,000) were selected for study. Instead, the authors focused on letter carrier walks (LCWs) with a population between 500 and 2,500 persons. Progress towards the objective of identifying neighbourhoods in terms of their homogeneity was, however, undermined by aggregating them into larger geographical units: Calgary’s “planning areas” and Saskatoon’s “defined neighbourhood areas.” These two locations were constructed with administrative objectives in mind. Their relation to neighbourhoods is not self-evident and was not made evident by the authors of Falling Behind. Chicago had 75 official neighbourhoods, but researchers such as Hunter (1975) discovered “206 smaller, but meaningful, neighbourhoods embedded in the 75” (p. 10).

The Strong Neighbourhoods Task Force (2005) “based its research and recommendations” not on a reflexive definition of neighbourhood, but on the City of Toronto’s operational definition formulated for “planning and program implementation purposes” (p. 19). The City identified 140 neighbourhoods with populations ranging from 7,000 to 11,000. Hunter’s research suggests that three times as many (n=420) neighbourhoods would be identified by residents.

Kingston, Ontario is one of the very few cities in Canada that uses aggregations of “five to seven” DAs to define its neighbourhoods. The City defines neighbourhoods as, “areas of common social, physical and political attributes” (2007:11). Kingston’s 42 neighbourhoods are derived from 252 DAs spread across 40 census tracts (Personal communication, Planning Department, Kingston, January 5, 2008).

Research by Hunter and Suttles(1972) and Slovak (1986) remains the starting point for some of the most valid and useful contemporary work on defining neighbourhoods as areas smaller than census tracts. Hunter and Suttles found neighbourhoods to be embedded in “a pyramid of progressively more inclusive groupings” (1972:61). Slovak named and defined these groupings. The first is the “face block,“ which is one side of a city block. Statistics Canada defines a “block face” as “the whole residential block between two consecutive intersections” (Census Dictionary). Face blocks and block faces are grounded in residential propinquity and the shared use of local shops and other facilities. They help define neighbourhoods because they are the major source of friendship and acquaintance groups.

In Canada, the location that comes closest to measuring block faces is the DA. Face blocks and block faces are embedded in a larger grouping, the “nominal community;” that is, a place with a name and boundaries recognized by residents and strangers. In other words, residents of neighbourhoods share “cognitive maps” (Block, 1992; Kennedy and associates, 1996). In the cognitive maps of Jane-Finch residents, the nominal Jane-Finch community is defined as an area bounded by Highway 400 to the West, Black Creek to the East, Shepherd Avenue to the South, and Shoreham Drive or Steeles Avenue West to the North (Ellis and Sociology 4200 students, 2005). Much of the inter-gang violence, however, does not occur between Jane-Finch and other nominal communities, but rather between neighbourhoods constituted by archipelagos of DAs such as Downbottom, Driftwood, Shoreham, Tobermory and Eddystone that are embedded in the Jane-Finch nominal community.

Neighbourhoods (block faces) and nominal communities, in turn, are embedded within larger “communities of limited liability.” These include police and school districts, health regions, and political wards or constituencies. Resources controlled by government and other agencies, which are part of such communities, help create and change the quality of life experienced by neighbourhood residents.

Differences in the definition and measurement of neighbourhood are important because positive associations between relative disadvantage and health and safety outcomes, based on the analysis of data from larger areas such as provinces and CMAs, may be nullified when smaller areas such as census tracts are selected for analysis (Hipp, 2007; Land et al,1990; Miles, Picot and Pyper, 2000). Similarly, findings based on the statistical analysis of census tracts may be modified, qualified or even reversed when DAs are studied.

Absolute deprivation/disadvantage is present when families or households do not have access to economic resources that are sufficient to meet basic biogenic needs. Measures of central tendency, such as average and median income, are used to measure absolute deprivation. Relative deprivation/disadvantage is present when economic resources are, or are perceived to be, distributed unequally across families or households. Neither absolute nor relative disadvantage are conceptually defined in a way that includes the consequences of either absolute or relative disadvantage. Instead, in a majority of the studies of poverty in Canada, disadvantage tends to be equated with the presence of a single indicator of the single domain of income.

In some studies, for example Ley and Smith and Boardwalk, relative disadvantage is implicitly defined as multi-dimensional and multiple indicators of multiple domains are used to measure it. A definition of absolute or relative disadvantage as multi-dimensional is, however, more economical and useful when multiple indicators of multiple domains are combined in an index that yields a single measure of multiple domains.

Kazemipur (2000) was the only author of a Canadian study who referred to relative deprivation theory in the introductory segment of his article. One would therefore expect to find an index of deprivation/disadvantage in the method segment. Instead, deprivation/disadvantage was measured in a way that equated it with a single indicator: low income.

Hulchanski described his objective as determining “how the average socio-economic status of residents in each of 527 (city of Toronto) census tracts has changed over 30 years” (2007:1). Socio-economic status (SES) is invariably measured by combining and weighting social and economic indicators (e.g. Blishen Scale), yet Hulchanski uses one economic factor, average individual income, to measure the multidimensional concept of SES.

Like the European researchers whose publications were reviewed, Canadian researchers Hou and Miles, Ley and Smith, and Boardwalk conceived of relative disadvantage as a multidimensional concept. Unlike members of the former group, they identified multiple indicators but did not combine them in a single measure such as an index or scale. Thus, Ley and Smith simply examined ratings on the indicators present in census tracts and then ranked census tracts according to the number of high ratings on each of the four indicators. Those with high ratings on all four indicators were categorized as extremely disadvantaged. Hou and Miles (2004) identified five “correlates of neighbourhood income inequality” (p. 11) but provided neither a rationale for their selection nor an index that combined them into a single measure of relative deprivation.

In 2005, the Strong Neighbourhoods Task Force published a report on the “vitality” of neighbourhoods. The report asserted that “there is no typical “depressed” neighbourhood, nor a typical ‘strong’ one, and no single measure can accurately represent their overall health” (2005: 21). It is not clear from this statement whether they were referring to a single indicator, or to a single measure of neighbourhood vitality, or to depression based on an index including multiple indicators. At any rate, they did not combine indicators in an index or scale. Instead, they identifed five domains (economic, education, urban fabric, health and demographics), listed 11 indicators, and then (implicitly) defined a neighbourhood as “challenged” when it “measured 20% worse than the City average” on each of them (p. 21).

The selection of indicators to be included in an index or scale of relative disadvantage is, or should be, guided by theory, definitions, research findings, and the purpose the selection is meant to serve (Noble and associates, 2006). Guided by these criteria and by a reinterpretation of Robson, Bradford and Tye’s (1995) concept of “domain,” Noble and associates (2006) selected indicators from the domains of income, education, housing and employment (p. 201).

With the objective of social inclusion in mind, Percy-Smith (2000) selected 26 indicators from seven domains: education, social, political, neighbourhood, individual, spatial and group. The indicators for the social domain were: breakdown of traditional households, unwanted teenage pregnancies, homelessness, crime and disaffected youth. Apart from the problem of overlapping domains (spatial/group and social/neighbourhood) and the requirement of collecting survey data to measure a number of the subjective indicators, the indicators themselves were not combined in an index or scale, either for each domain or for all domains. The latter would yield an index of social exclusion if a principal components analysis revealed the existence of a single underlying factor. If all of the indicators identified by Percy-Smith are not highly correlated with one another, this outcome is unlikely.

Layte, Nolan and Whelan (2000) selected resources (low income) and multiple indicators of disadvantage/deprivation defined as “items generally regarded as necessities which families must do without.” Using household survey data, Lugo (2005) selected per capita household income, years of formal education and life expectancy at birth (2005, p. 28). Using survey data, Vranken (2002) selected low education, unemployment and “in arrears on payments.” Armand and Sen 1997 created a global Human Poverty Index derived from the proportion expected to die before age 40, illiteracy and economic deprivation.

These indicators were selected for the purpose of the targeted reduction of poverty. Indicators selected with crime, delinquency and youth violence reduction in mind were quite similar. For example, Krivo and Peterson (2006) created an Index of Concentrated Disadvantage based on low income, male joblessness and female-headed households (p. 4). To measure concentration, they created an Index of Isolation derived from inter-group (black/non-black) contact, and to measure community stability, they created an Index of Community stability based on homeowner occupancy and race-specific houses that are owner-occupied.

The Index of Concentrated Disadvantage included four domains and indicators (income, housing, family and employment) that are frequently used in research on crime generally and violent crime in particular. Indicators of these four domains were included in the IRD. The domains of education and income inequality within ethnic groups have also been included in similar indices (Table 2), yet Krivo and Peterson offer no rationale for excluding them. In addition, the Index of Concentrated Disadvantage was designed to identify conditions associated with homicide, and not conditions that may lead to youth violence.

European researchers who identify housing as a domain of disadvantage would also conceptualize their indicators of “community stability” as indicators of disadvantage (Noble and associates (2006)). Single-parent households and the high school dropout rate have been found to be so reliably and strongly associated with delinquency generally, and youth violence specifically, that “they must be included in any index of relative disadvantage that is created for the purpose of reducing these outcomes” (Loeber and Le Blanc, 1990). Table 2 reveals the similarity in disadvantage domains selected by some researchers studying income inequality and other researchers investigating linkages between relative disadvantage and crime, violent crime, youth violence, and delinquency.

The methods used to combine multiple indicators varied across researchers whose common objective was to construct a summary measure of relative disadvantage (index or scale) from multiple indicators of disadvantage. Some researchers constructed measures of multiple disadvantage/deprivation using multivariate probit regression analysis (Capellari and Jenkins, 2004). The benefits of using this rather complex method over methods that are equally valid and easier to use are not obvious.

For Noble, Wright and Dibben (2006), community or neighbourhood deprivation had the following compositional meaning: “An area is considered to be deprived if it has a large number or proportion of deprived people” (p. 170). They used the concept of domain to refer to “area-level dimensions of deprivation...which aggregate as a measure of multiple deprivation” (p. 173). Income, housing, education and family are domains identified by a number of researchers. Each one is measured using different indicators of disadvantage. Then, weights are attached to them based on (a) a review of the relevant literature, and (b) the results of consultation with experts, policymakers, and other stakeholders. Domain scores measure specific sources or types of disadvantage. Aggregated domain scores measure multiple disadvantage.

One of the major problems with attaching weights to indicators using this approach has to do with resolving contradictions or differences within and between “literature” and “stakeholder” sources.

Many if not most researchers use factor analysis (Curry and Spergel, 1988; McKay and Collard, 2003; Krivo and Peterson, 2000; Rosenfeld, Bray and Egley, 1999; Taylor and Covington, 1988). Although it has a number of weakness (Coombes and associates,1995; Senior, 2002), factor analysis has been effectively employed by researchers such as Callan and associates (1993) and McKay and Collard (2003) to identify a number of underlying domains of disadvantage and to attach weights to them. In the proposal herein, indicators selected for principal components analysis were selected on theoretical grounds.

Data Collection

In two of the Canadian studies reviewed, data relevant to identifying places where people living below the poverty line are concentrated were collected from taxpayers who file tax returns annually. As tax returns are filed annually, tax-filer data provide useful information about income distribution during inter-census years. However, census tract and DA income data are even more useful because, unlike tax-filer data, they include a relatively large number of covariates which permit multivariate and time-lagged analyses, and because from these analyses, indicators of relative disadvantage can be selected (Frenette, 2006).

In the remaining studies of income inequality in Canada, data were collected from the Statistics Canada census. In almost all cases, data from two or more census years were collected. In Europe, objective data on objective indicators of disadvantage were collected from the census, other government agencies, and universities (e.g., English Indices of Deprivation, 2005), and subjective data on what it means to be poor were collected from multi-year panel studies (e.g., Irish Deprivation Index, 2001).

Data Analysis

Descriptive statistics (means, medians, standard deviations, range and frequency distributions) and bivariate relational statistics (frequency distributions by geographical locations and time) were the modal methods of analyzing data in the studies reviewed. These were usually appropriate for the descriptive and policy objectives of the researchers.


Study Design

A longitudinal design with two data entry points, 2001 and 2006, will be implemented. Change characterizes the DAs (neighbourhoods) being studied and the direction, strength and patterning of change is captured in longitudinal study designs.


Like youth violence, disadvantage is not randomly distributed across geographical locations in Ontario. Instead, it is concentrated in large urban areas (CMAs) Approximately 88% of Ontario’s population resides in large urban areas. The remaining 12% reside in small urban areas (any urban area not part of CMA with a minimum population of I,000 persons) and rural areas (any area not falling into large urban or small urban places) (Statistics Canada, 2002). Within these three places, disadvantage is likely to be non-randomly distributed across DAs. If the requisite funding is available, and the primary objective is to identify DAs in large urban, small urban, and rural areas in Ontario characterized by high or increasing levels of relative disadvantage, then the population rather than a sample of DAs (n=19,177) should be studied. There are advantages to studying the population of DAs. First, it will yield information on levels of disadvantage in all of Ontario’s DAs. A sample will not produce this output. Second, to protect the privacy of families and individuals, Statistics Canada suppresses information on DAs with less than 500 persons, but in the present case, the use of archipelagos (contiguous DAs with similar IRD scores) will markedly decrease the number of DAs for which information on indicators is suppressed.

Third, the higher monetary costs usually associated with studying populations rather than samples drawn from them are, in the present case, quite reasonable. Statistics Canada personnel responsible for costing special orders for data collection roughly estimate that the cost of collecting data on five indicators for 19,177 DAs would fall somewhere between $20,000 and $25,000.

Fourth, non-sampling errors (see Sample Design, p. 21xx) are usually higher when populations rather than samples are studied, but in the present case, non-sampling errors are likely to be relatively low. Non-response rates are low because citizens are required to participate as census respondents and to answer all questions. Coverage is exhaustive because Canadian “reference persons” in all dwellings in Canada answer census questions about all persons who reside in them. Data entry errors are likely to be low because data on five indicators for 19,177 DAs was entered by well-trained, experienced Statistics Canada personnel.

Fifth, studying the population of DAs will enable researchers in any region of Ontario to investigate the impact of disadvantage on health, youth violence, in/out migration and other outcomes.

The population of DAs selected for study will permit the identification of “archipelagos,” defined as two or more contiguous very highly or highly disadvantaged DAs, and “islands,” defined as very highly or highly disadvantaged areas with no contiguous very highly or highly disadvantaged DAs. Scores on the IRD will be used to identify levels of relative disadvantage in the population or sample of DAs.

DAs were selected as the primary unit of analysis for several reasons. First, as they aggregate to census tracts findings on relative disadvantage, they can be compared with the results of income inequality studies in Ontario using census tract data. Second, compared with census tracts, archipelagos of DAs more closely approximate neighbourhoods generally and multi-disadvantaged neighbourhoods in particular (Ley and Smith, 2000; Mears and Bhati, 2006; Wilson, 1987).

Third, census data available for census tracts is also available for DAs. Fourth, stereotyping entire census tracts making up larger areas, such as “Jane and Finch” or Scarborough, is more easily avoided (Bursik and Grasmick (1993)).

Fifth, neighbours living in the same building, or in buildings close enough to each other to make their lifestyles known to each, other constitute the significant comparison group for assessment of relative advantage/disadvantage in wellbeing made by DA residents (Bursik and Grasmick, 1993; Lafree and Drass,1996; Messer, Raffalovich and Mcmillan, 2001).

Sixth, they provide a more specific focus for policies aimed at alleviating disadvantage and preventing youth violence (Noble and Associates, 2006). Seventh, they have been neglected as units of analysis by poverty researchers.

Definitions and Measurement

Following Townsend (1979), disadvantage is defined as a multi-dimensional concept that refers to living in circumstances or conditions that are associated with an impoverished quality of life and minimal participation in civil society. Disadvantage is relative to the experience of those residing in the same or other locations.

Index of Relative Disadvantage

The IRD is a multi-dimensional measure of relative disadvantage. Five domains of disadvantage are combined in the index. Each domain is measured by an indicator. Factor analysis is used to weight each of the indicators according to its importance and to combine them in such a way as to produce a single score measuring a factor called disadvantage. The higher the plus (or minus) score received by a DA compared with other DAs, the greater its level of relative disadvantage.

The construction of an IRD is a two-step process. The selection of domains is step one. The rationale for selecting domains is empirical and theoretical. The theoretical rationale is provided by relative disadvantage, social disorganization and social control theory (Agnew, 1985; Clark, 1964; Bursick and Grasmick, 1993; Clark, 1964; Hipp, 2007; Osgood and Chambers, 2000).

The empirical rationale is evident in Table 2. The domains selected for inclusion in the IRD are identified as domains of disadvantage by almost all of the authors of both the poverty and youth violence studies reviewed.

Five domains of disadvantage were selected. They are income, housing, education, family and employment. Findings presented in Table 2 indicate that income was included in 22 of the 28 studies, family was included in 21 of them, employment in 15, housing in 10, and employment in 15. Of equal if not greater importance is the finding that, collectively, the five domains were included in four of the most methodologically robust indices of deprivation/disadvantage reviewed (Sampson and Raudenbush,1997, Social Disadvantage Research Centre, 2007; Noble and Associates, 2004; Layte, Nolan and Whelan, 2000). Finally, articles using these domains of disadvantage were published in prestigious journals and included in some of the most highly regarded indices of deprivation/disadvantage (e.g., English Indices of Deprivation).

Each of the five domains is measured by an indicator. In social research, one important criterion used in selecting indicators is validity. An indicator may be said to be valid if the same or a very similar indicator is used by different researchers or by the most competent methodologists measuring the same domain of disadvantage. As discussed above, Low Income Measures (LIMs) are widely used to measure the domain of income in national and cross-national studies of poverty. Moreover, LIMs were used as an indicator of income disadvantage in some of the most methodologically robust European indices of deprivation/disadvantage. (e.g., Irish Index of Deprivation, Layte et al, 2000). Creators of the Irish Index selected LIMs because this indicator most clearly reveals the consequences of receiving an income below LIM thresholds for everyday living. Individuals and family members in this income group are excluded from experiencing a quality of life experienced by the average Irish family.

The indicator “percentage owner/occupied” is used as an indicator of the domain of housing. Indicators used by other researchers (e.g., Strong Neighbourhoods Task Force, creators of the Irish Index, the English Indices of Deprivation, and Hajnal) can be subsumed under this indicator because they are likely to be highly correlated with it. Unique to this indicator is the disjunction between home ownership promoted as a Canadian value and perceived/real barriers to meeting this standard. The disjunction is the source of real/perceived disadvantage (Bursik and Grasmick, 1993).

The indicator “percentage adults who failed to graduate from high school” is used to measure the domain of education. The same indicator is one of six indicators included in the English Indices of Deprivation – Education. Other researchers, for example Hajnal, Ley and Smith, and the Strong Neighbourhoods Task Force, use an indicator (percentage of adults aged 15 and older who have achieved less than grade nine education) that can be subsumed under the indicator included in the IRD. Unique to this indicator is the degree and duration of disadvantage of persons who fail to graduate from high school compared with those who complete high school but fail to “enter higher education” (English Indices of Deprivation), or “percentage of population with college or university qualifications” (Strong Neighbourhoods Task Force).

The indicator “percentage children aged 0 to 16 living in single, female-parent households” is used to measure the domain of family. This indicator is used to measure disadvantage among dependent children. Disadvantage is material, stemming from low income, and psycho-social, stemming from the inability of single mothers struggling to make ends meet to adequately care for and control their children. This indicator was included in the English Indices of Deprivation, and almost all of the researchers included in Table 2 have used the indicator “percentage single, female-parent households” to measure the domain of family.

The indicator “percentage males aged 25 and over who are unemployed” is used to measure the domain of employment. Disadvantage in this area has obvious material consequences. The stigma associated with unemployment is an important psychological consequence. These consequences may be expected to weigh most heavily on males who have reached an age where they are expected to support themselves and members of their own families. For this reason, the age of 25 was selected. Strong Neighbourhood Task Force researchers selected the same indicator. Male unemployment was selected, not only because it is used an indicator of the employment domain by all of the researchers included in Table 2, but also because it is highly correlated with low income and a high percentage of single, female-headed households (Wilson, 1987).

The domains and indicators are summarized below:

Income %LIM economic families
Housing %owner/occupied dwellings
Education %failed to graduate from high school
Family %children 16 and under living in single, female-headed households
Employment %males 25 and over who are unemployed

Step two involves combining the theorized, empirically validated indicators in an index. A purely mathematical/empirical procedure, common factor analysis, will be used for this purpose. The five domain indicators (variables) will be entered into the factor analytic program. This is the input. The program analyzes the association among all indicators (variables) and the association (correlation) between each indicator and the hypothesized underlying factor. The output will be one underlying factor, disadvantage, together with factor loadings (relative importance or weights) for each of the indicators (variables). Taken together, all five factors should explain a significant proportion of the variation in the factor we call disadvantage.

The expected (hypothetical) results of the factor analysis for all DAs in Ontario can be described as follows:

Indicators (variables) Factor (Disadvantage)
% LIM families .721
% children under 16 in female-headed households .668
% males over 25 unemployed .638
% non-high school graduates .616
% owner-occupied dwellings .594
Variance explained 74.9%

The coefficients in this table (.721, .668 etc.) are factor loadings or weights. These are specific to the population of DAs in Ontario. Data on the population of Ontario DAs (n=19,177) is used to create the Index of Relative Deprivation. In order to examine the stability of the factor loadings, similar output will be generated for all DAs in larger urban, small urban and rural areas. Major differences are not expected in the strength and direction of factor loadings for any of the five indicators. If differences are found, it will means that the pattern of relationships among the five variables is different in large urban, small urban and rural areas. It does not mean that variables with the highest factor loadings in each of these areas should be the focus of alleviating interventions. This is because the loading on this variable is a function of its association with, and the contribution of, the other four variables. Thus, policy-makers with a stake in education, housing, employment, family or income distribution cannot use factor loadings on variables in the IRD to justify the specific interventions they advocate. Instead, IRD scores can only be reduced by decreasing the values (percentages) of all five domain indicators (variables).

If the strength and direction of the factor loadings on the five variables are significantly different in the three different areas, the option of using area-specific IRDs to measure relative disadvantage may be exercised. If this were to be done, relevant benchmark large urban, small urban and rural IRD scores would be used to rank order DAs in them.

The concept of eigenvalue is used to refer to the amount of variation explained by the five variables. As there are five variables, the total variance to be explained is 5. Squaring each of the five coefficients in the table above results in a value 4. Dividing the latter by the former (4 by 5) yields an eigenvalue of .8 or 80%.

Finally, the factor analysis will yield IRD scores for each DA that may vary between, let us say, plus 3 and minus 3. Depending on the nature of the findings, plus or minus 3 may mean greater or lesser levels of disadvantage. IRD scores all DAs can be used to rank them according to their levels of disadvantage.

The objective of identifying communities characterized by relative disadvantage is shared by researchers such as Toronto Strong Neighbourhoods Task Force (2005) and the creators of the English Indices of Deprivation, researchers at the Social Disadvantage Research Centre (SDRC) at the University of Oxford in England (2007). Instead of creating a new, untested IRD, why not simply use the domains and indicators created by the former and the indices created by the latter?

Strong Neighbourhoods Task Force researchers identified five domains and 11 indicators of disadvantage “to determine the challenges faced by Toronto neighbourhoods” (p. 22). The challenges faced by Ontario’s rural and small urban areas may not be the same as those faced by the City of Toronto. Secondly, an index would have to be created and tested because the Task Force researchers did not create one. Third, because the rationale for the indicators selected by Task Force researchers was not stated, we do not know whether it is more compelling that the rationale presented for the indicators included in the IRD. Fourth, collecting data on 11 indicators from at least two different sources (the census and Ontario hospitals or health authorities) and then “fitting” the health data to DAs is almost certain to be more than twice as costly as collecting DA data on five indicators from one source (the Statistics Canada census).

The English Indices of Deprivation (SRDC, 2007) have a number of strengths. First, they measured multiple deprivation. Specifically, 33 indicators were used to measure seven different domains of deprivation. These were: Income, Employment, Health/Disability, Education, Housing and Crime. Second, they measured multiple deprivation at a level (“small area”) that more closely approximates the size of DAs than census tracts do (p. 4). Third, a relatively wide range of deprivations was measured. Fourth, individual domains were weighted and can be combined into an index of multiple deprivation.

Despite its evident strengths, there are good reasons for not using the English Indices of Deprivation in Ontario. First, the cost of collecting data on 33 indicators of seven domains from 16 different government, private and university sources is likely to be far more costly than collecting data on five indicators from one source.

Second, data on specific domain indicators may not be available, because data are not collected (e.g., barriers to owner-occupation, recipients of Job Seekers Allowance, Participants in the New Deal for 18–24s who are not in receipt of JSA, Participants in the New Deal for Lone Parents, Working Tax Credit Households), or because the data are collected by agencies unwilling to release the information for privacy or organizational reasons (e.g., hospital statistics on acute morbidity), or because the data are not included in the Statistics Canada census (e.g., road distances to services and facilities) and therefore require additional effort to collect and then “fit” to DAs. The English census is cited as the source for only three of the 33 indicators (household overcrowding, houses without central heating and workers with low qualifications). A majority of the indicators included in the English indices are not included in the Statistics Canada census.

Third, there is likely to be very little variation across discrimination areas in “households without central heating,” one of the indicators of the living environment deprivation domain, and all four indicators of the crime domain should not be included in the IRD (or the English Indices – Crime) because they measure crimes reported to the police and not actual rates of criminal victimization. The stated rationale for the crime domain indicators was “representing the risk of personal victimisation at small area level,” but the indicators used represent an unknown fraction of actual acts of criminal victimization (burglary, theft, criminal damage and violence).

In sum, the data to construct the IRD are far less costly to collect. The IRD includes four of the domains included in the English indices (income, employment, housing, education) and also uses the same or very similar indicators of them. The IRD does not include three of the domains of disadvantage that are included in the English indices for the following different reasons.

Health was excluded because the indicators are not included in the Statistics Canada census and the cost of collecting health data from other sources and then “fitting” them to DAs would be quite costly in terms of money and time. Still, the domain of health is important enough for inclusion in the IRD of any future index or scale of relative disadvantage that attempts should be made to persuade Statistics Canada to include questions measuring poor health, disability and early mortality, and acute morbidity in the census. These indicators measure the domain of health deprivation in the English Indices of Deprivation (p. 6). Alternatively, hospitals or regional health authorities may be persuaded to collect and record case information by DA.

Living environment was excluded because indicators of the “outdoor” living environment are not included in the Statistics Canada census, and one indicator of the “indoor” living environment (household central heating) is probably a constant in Ontario households. Crime was excluded because its indicators will yield an inaccurate measure of this domain.

Data Collection

Domain data measuring relative disadvantage at two points in time will be collected from the 2001 and 2006 census (Statistics Canada). All the information available for census tracts is also available for DAs.

Data on DA areas may be collected from sources other than the census, depending upon their availability, feasibility and applicability to DAs. Aggregate level data collected from the census are fully applicable to census tracts. They are also fully applicable to DAs with 500 plus individuals. Data are suppressed for DAs with smaller populations.

Researchers who may be interested in collecting data for the purpose of constructing an Index of Youth Violence, or in adding additional variables to the IRD and then conducting a principal components factor analysis to identify a smaller set of variables that may not include the ones selected for the IRD, are invited to peruse Appendix 1.

Data Analysis

As indicated earlier, common factor analysis will be used to create an IRD, which will yield a disadvantage score that will probably vary between plus 3 and minus 3. The interpretation of these numbers and signs will depend on whether plus or minus means more/less disadvantaged or more/less advantaged. In either case, all Ontario DAs can be ranked according to their plus or minus IRD scores. Different colours can be assigned to each of them and their locations can be identified on a map of Ontario showing the location of all 19,177 DAs. Computer software can be created for this purpose. For example a user would be able to click on the ten most disadvantaged DAs in 2001, then enter place-specific, post-2001 alleviating actions implemented, and then click on the same ten DAs for 2005. The strength and direction of change will be revealed.

The data can also be analyzed to compare levels of disadvantage in archipelagos or individual DAs with an Ontario benchmark; that is, the IRD score for the population of Ontario DAs. Similar comparisons can be made for large urban areas, small urban areas and rural areas using appropriate area benchmarks. These analyses will permit researchers to state that levels of disadvantage in a specific DA or archipelago of DAs is two, three, four, or more times higher or lower than the Ontario or specific-area benchmark.

Finally, reductions in the percentage places and proportion of the Ontario population suffering from high levels of disadvantage between 2001 and 2005 can be assessed by examining decreases in (a) the number of the most highly disadvantaged DAs, such as DAs with IRD scores of minus three or four, and (b) the proportion of the population residing in them during this five-year period.


This proposal describes the methodology to be used in identifying neighbourhoods in Ontario where high or increasing relative disadvantage may lead to youth violence. Methodology includes study design, sampling, definition/measurement, data collection and data analysis. The proposed study design is longitudinal because it is appropriate for measuring relative disadvantage at two periods of time (2001 and 2006). Neighbourhoods are defined in terms of DAs. Poverty is conceived of as disadvantage. Poverty/disadvantage is conceived of as a multi-dimensional variable for which the multi-dimensional definition presented is appropriate. Relative disadvantage among DAs in Ontario is measured using an IRD that includes five domains of disadvantage and one indicator of each of them. The domains are income, housing, education, family and employment. Instead of a sample, the population of Ontario DAs is selected for study, and relative disadvantage is measured in DAs in all three types of area: large urban, small urban and rural areas. Data on the indicators used in constructing the IRD were collected from the Statistics Canada census at two points in time, 2001 and 2006. Common factor analysis was used in creating the IRD. Data will be analyzed using disadvantage scores produced by factor analysis. These scores are used to (a) rank all DAs with respect to their levels of disadvantage; (b) rank DAs according to the degree to which they fall above and below Ontario, large urban, small urban, and rural benchmarks; (c) assess reductions (or increases) in disadvantage that may have occurred during the five-year period 2001–2006; (d) evaluate the impact of alleviating interventions on the proportion of the population suffering from high levels of disadvantage and on the number of highly disadvantaged DAs.


Agnew, R. (1985). A revised strain theory of delinquency. Social Forces, 64, 151–167.

Anand, S. and A. K. Sen. (1997). Concepts of Human Development and Poverty: A Multidimensional Perspective. New York: Human Development Papers, United Nations Development Programme.

Atkinson, A. B. (2003). Multidimensional deprivation: Contrasting social welfare and counting approaches. Journal of Economic Inequality, 1, 51–65.

Berthoud, R., M. Bryan and E. Bardasi. (2004). The relationship between income and material deprivation over time. Department for Work and Pensions Research Report 219, Corporate Document Services, Leeds, UK.

Block, C.R. (1992). Computer mapping as a tool in violence reduction. In Questions and Answers in Lethal and Non-lethal Violence. Proceeding of the First Annual Workshop of the Homicide Research Working Group. Ann Arbor, Michigan, June 14–16.

Bourguignon, F. and S. R. Chakravarty. (2003). The measurement of multidimensional poverty. Journal of Economic Inequality, 1, 25–29.

Bradshaw, J. and N. Finch. (2003). Overlap in dimensions of poverty. Journal of Social Policy, 32, 513–525.

Brady, D. (2003a). The Politics of Poverty: Left Political Institutions, the Welfare State and Poverty. Social Forces, 82, 557–588.

———. (2003b). Rethinking the Sociological Measurement of Poverty. Social Forces, 81(3), 715–752.

Broadway, M. (1992). Differences in inner-city deprivation: An analysis of seven Canadian cities. The Canadian Geographer, 36, 189–196.

Bursik, R. J. and H. Grasmick. (1993). Neighbourhoods and Crime: The Dimensions of Effective Community Control. New York: Lexington Books.

Callan, T., B. Nolan and C.T. Whelan. (1993). Resources, deprivation and the measurement of poverty. Journal of Social Policy, 22, 141–172.

Callan, T. B., B. J. Nolan, C. T. Whelan, and J Williams. (1996). Poverty in the 90s: Evidence from the 1994 Living in Ireland Survey. Dublin: Oak Tree Press.

Canadian Council on Social Development. (2007). Urban Poverty Project 2007. Online <>.

Cappellari, L. and S. Jenkins. (2006). Summarizing Multiple Deprivation Indicators. ISER Working Paper 2006–40. Colchester: University of Essex.

City of Kingston. (2007). Neighbourhood Profiles. Online: <>.

Clark, K. (1964). Dark Ghetto: Dilemmas of Social Power. New York: Harper Torch Book.

Coombes, M. G. S. Raybould, C. Wong and S. Openshaw. (1995). Toward an index of deprivation: A review of approaches. In Department of the Environment (Eds.) 1991 Deprivation Index: A Review of Approaches and a Matrix of Results. London: HMSO.

Crutchfield, R. D. (1989). Labour stratification and violent crime. Social Forces, 68, 90–118.

Curry, G. D. and I. A. Spergel. (1988). Gang homicide, delinquency and community. Criminology, 26, 381–405.

Ellis, D. (2005). Youth Street Gangs in the GTA. Honours sociology class project papers. York University.

Ellis, D. and Sociology 4200 students. (2006). Street Youth Gangs in the GTA: Neighbourhood Profiles. Department of Sociology, York University.

Ezcurra, R., P. Pascual and M. Rapun. (2007). The spatial distribution of income inequality in the European Union. Environment and Planning, 39, 869–890.

Federation of Canadian Municipalities (2007). Falling Behind: Our Growing Income Gap. Ottawa: FCM.

Francisco, J. and C. Chenier. (2007). A comparison of large urban, small urban and rural crime rates. Juristat, 27, 3. Catalogue no. 85-002-XIE. Canadian Centre for Justice Statistics. Statistics Canada. Ottawa

Frenette, M., D. A. Green, and K. Milligan. (2006). Revisiting recent trends in Canadian after-tax income inequality using census data. Analytical Studies Branch Research Paper Series. Statistics Canada. Ottawa.

Gordon, D. (2006). The concept and measurement of poverty. In C. Pantazis, D. Gordon and R. Levitas (Eds.), Poverty and Social Exclusion in Britain: The Millennium Survey (pp. 29–70). Bristol, UK: Policy Press.

Gordon, D., L. Adelman, K. Ashworth, J. Bradshaw, R Levitas, S. Middleton, C. Pantazis, D. Patsios, S. Payne, P. Townsend and J. Williams. (2000) Poverty and Social Exclusion in Britain. York: Joseph Rowntree Foundation.

Gordon, D. and C. Pantazis (Eds.). (1997). Breadline Britain in the 1990s. Aldershot: Ashgate.

Hajnal, Z.L. (1995). The Nature of Concentrated Urban Poverty in Canada and the United States. Canadian Journal of Sociology, 20, 497–528.

Heisz, A. (2007). Income Inequality and Redistribution in Canada: 1976–2004. Analytical Studies Branch Research Paper Series. Cat. no. 11F0019MIE – No. 298. Statistics Canada, Ottawa.

Heisz, A. and L. McLeod. (2004). Low Income in Census Metropolitan Areas, 1980–2000. Trends and Conditions in CMAs. Analytical Studies Branch. Cat. 89-613-MIE2004 –No. 001. Statistics Canada. Ottawa.

Hipp, J.R. (2007). Income inequality, race and place: does the distribution of race and class within neighbourhoods affect crime rates. Criminology, 45, 665–688.

Hirschi, T. and H. Selvin. (1967). Delinquency Research: An Appraisal of Analytic Methods. New York: The Free Press.

Hirschowitz, R., M. Orkin and P. Alberts. (2000). Key baseline statistics for poverty measurement. In R. Hirschowitz (Ed.) Measuring Poverty in South Africa (pp. 53–81). Pretoria: Statistics South Africa.

Hou, F. and J. Miles. (2004). Neighbourhood Inequality, Relative Deprivation and Self-perceived Health Status. Analystical Studies Branch Research Paper Series. Cat. no. 11F0019mIE – No. 228. Statistics Canada. Ottawa.

Hulchanski, David. (2007). The Three Cities within Toronto: Income Polarization Among Toronto Neighbourhoods, 1970–2000. Toronto: Centre for Urban and Community Studies, University of Toronto.

Human Resources Development Canada. (2003). Understanding the 2000 Low Income Statistics Based on the Market Basket Measure. Ottawa: Applied Research Branch Strategic Policy.

Hunter, A. (1985). Private, parochial and public social orders: The problem of crime and incivility in urban communities. In G.D. Suttles and M. Zald (Eds.) The Challenge of Social Control: Citizenship and Institution Building in Modern Society (pp. 230–242). Norwood, NJ: Ablex Publishing.

Hunter, A. and G. D. Suttles. (1972). The expanding community of limited liability. In G.D. Suttles (Ed.) The Social Construction of Communities (pp. 44–81). Chicago: University of Chicago Press.

Kazemipur, A. (2000). The ecology of deprivation: Spatial concentration of poverty in Canada. Canadian Journal of Regional Science, 3, 403–426.

Kazempiur, A. and S. S. Halli. (2000). Neighbourhood poverty in Canadian cities. Canadian Journal of Sociology, 25, 369–381.

Kennedy, D. M., A. A. Braga and A. M. Piehl. (1997). The (un)known universe: Mapping gangs and gang violence in Boston. In D. Weisburd and T. McEwen (Eds.)Crime Mapping and Crime Prevention (pp. 219–262). New York: Criminal Justice Press.

Kennedy, L. W., R. Silverman and D. R. Forde. (1991). Homicide in urban Canada: Testing the impact of income inequality and social disorganization. Canadian Journal of Criminology, 16, 397–410.

Krivo. L., and R.D. Peterson. (2000). The structural context of homicide: Accounting for racial differences. American Sociological Review, 65, 547–560.

La Free, G. and K. Drass. (1996). The effect of changes in intraracial income inequality and educational attainment on changes in arrest rates for African Americans and whites, 1957 to 1990. American Sociological Review, 61, 614–634.

Land, K., P. McCall and L. E. Cohen. (1990). Structural covariates of homicide rates: Are there any invariances across time and social space? American Journal of Sociology, 95, 922–963.

Lau, R. R. (1989). Individual and contextual differences on group identification. Social Psychological Quarterly, 52, 220–231.

Layte, R., B. Nolan and C. Whelan. (2000). Targeting poverty; Lessons from monitoring Ireland’s National Anti-Poverty Strategy. Journal of Social Policy, 29, 553–575.

Ley, D. and H. Smith. (2000). Relations between deprivation and immigrant groups in large Canadian cities. Urban Studies, 37, 37–62.

Loeber, R. and M. Le Blanc. (1990). Toward a Developmental Criminology. In M. Tonry and N. Morris (Eds.) Crime and Justice, Vol. 12 (pp. 375–473). Chicago: University of Chicago Press.

Lugo, M.A. (2005). Comparing Multidimensional Indices of Inequality: Methods and Application. Paper presented at the First meeting of the Society for the study of Economic inequality, Palma de Mallorca (July).

Ma, S.J. (2004). Just Listen to Me: Youth Voices on Violence. Toronto: Government of Ontario Office of Child and Family Service Advocacy. Online: <>.

McIntyre, D. and Associates (2000). Geographic Patterns of Deprivation and Health Inequities in South Africa: Informing Public Resource Allocation Strategies. Health Economics Unit, University of Cape Town, Cape Town, South Africa.

McKay, S. and S. Collard. (2003). Developing Deprivation Questions for the Family Resources Survey. IAD Research Division Working Paper, # 13. London: Department for Work and Pensions.

MacLachlan, I. and R. Sawada. (1997). Measure of income inequality and social polarization in Canadian metropolitan areas. Canadian Geographer, 41, 377–398.

Mears, D. P. and A. Bhati. (2006). No community is an island: The effects of resource deprivation on urban violence in spatially and socially proximate communities. Criminology, 44, 509-48.

Merton, R. K. (1957). Social Theory and Social Structure. New York: The Free Press.

Messner, S. F. (1982). Poverty, inequality and the urban homicide rate: Some unexpected findings. Criminology, 20 (1), 103–114.

Messner, S. F., L. E. Raffalovich and R. McMillan. (2001). Economic deprivation and changes in homicide arrest rates for white and Black youths, 1967–1998: A national time-series analysis. Criminology, 39, 591–613.

Miles, J., G. Picot and W. Pyper. (2000). Neighbourhood Inequality in Canadian Cities. Analytical Studies Branch Research Paper Series. Cat. no. 11F0019MPE – No. 160. Statistics Canada. Ottawa.

Moller, S., E. Huber, J. D. Stephens, D. Bradley and F. Nielsen. (2003). Determinants of relative poverty in advanced capitalist societies. American Sociological Review, 68, 22–51.

Moran, T. P. (2006). Statistical inference and patterns of inequality in the global north. Social Forces, 84, 1799–1818.

Morenoff, J. D., R. Sampson and S. W. Raudenbush. (2001). Neighbourhood inequality, collective efficacy and the spatial dynamics of urban violence. Criminology, 39(3), 517–558.

Noble and Associates. (2004). The English Indices of Deprivation 2004. London: Office of the Deputy Prime Minister, Neighbourhood Renewal Unit.

Noble, M., G. Wright, G. Smith and C. Dibbens. (2006). Measuring multiple deprivation at the small area level. Environment and Planning, 38, 169–185.

Nolan, B. and C. Whelan. (1996). Measuring poverty using income and deprivation indicators: Alternative approaches. Journal of European Social Policy, 6, 225–240.

Osgood, D. W. and J. M. Chambers. (2000). Social disorganization outside the metropolis: An analysis of rural youth violence. Criminology, 38(1), 81–116.

Percy-Smith, J. (2000). Introduction: The contours of social exclusion. In J. Percy-Smith (Ed.) Policy Responses to Social Exclusion: Toward Inclusion (pp. 2–21). Buckingham, UK: Open University Press.

Perry, B. (2002). The mismatch between income measures and direct outcome measures of poverty. Social Policy Journal of New Zealand, 19, 101–127.

Rainwater, L. and T. M. Smeeding. (2003). Poor Kids in a Rich Country; America’s Children in Comparative Perspective. New York: Russell Sage Foundation.

Raphael, D. (2007). Poverty and Policy in Canada. Toronto: Canadian Scholars Press.

Ringen, S. (1988). Direct and indirect measures of poverty. Journal of Social Policy,17, 351–365.

Robson, B., M. Bradford and R. Tye. (1995). A matrix of deprivation in English authorities,1991. In Department of the Environment (Eds.) 1991 Deprivation Index: A Review of Approaches and a Matrix of Results. London: The Stationery Office.

Rosenfeld, R., T. M. Bray and A. Egley. (1999). Facilitating violence: A comparison of gang-motivated, gang-affiliated and nongang youth homicides. Journal of Quantitative Criminology, 15, 495–516.

Ross, D. P., K. Scott and P. Smith. (2000). The Canadian Fact Book on Poverty 2000. Ottawa: Canadian Council on Social Development.

Ross, N. A., C. Hole, J. R. Dunn and M. Aye. (2004). Dimensions and dynamics of residential segregation by income in urban Canada, 1991–1996. The Canadian Cartographer, 48, 433–445.

Saez, E. and M. Veall. (2005). The evolution of high incomes in northern America: lessons from Canadian evidence. American Economic Review, 95(3), 831–849.

Sampson, R. J. and W. B. Groves. (1989). Community structure and crime: Testing social disorganization Theory. American Journal of Sociology, 94, 774–802.

Sampson, R., J. D. Morenoff and S. W. Raudenbush. (2005). Social anatomy of racial and ethnic disparities in violence. American Journal of Public Health, 95, 224–232.

Sampson, R. J., S. W. Raudenbush and F. Earls. (1997). Neighbourhoods and violent crime: A multilevel study of collective efficacy. Science, 277, 918–924.

Sampson, R. and W. J. Wilson. (1995). Toward a theory of race, crime and urban inequality. In J. Hagan and R. Peterson (Eds.) Crime and Inequality (pp. 37–54). Stanford: Stanford University Press.

Sanchez-Jankowski. M. (1991). Islands in the Street: Gangs and American Urban Society. Berkeley: University of California Press

Savaglio, E. (2007). Multidimensional inequality: A survey. Online: <>

Senior, M. (2002). Deprivation indicators. In P. Rees, D. Martin and P. Williamson (Eds.) The Census Data System (pp. 123–139). Chichester: John Wiley.

Sharkey, P. T. (2006). Navigating dangerous streets: The sources and consequences of street efficacy. American Sociological Review, 71, 826–827.

Slovak, J. (1986). Attachment in the nested community: Evidence from a case study. Urban Affairs Quarterly, 21, 575–579.

Social Disadvantage Research Centre. (2007). The English Indices of Deprivation 2007. University of Oxford. Oxford, England. Prepared for Department for Communities and Local Government.

Statistics Canada. 2002. Population and Dwelling Counts, for Canada, Provinces, Territories and Census Subdivisions (Municipalities), 2001 and 1996 Censuses, 100% Data (table). Population and Dwelling Counts. Statistics Canada Catalogue no. 93F0051XIE. Ottawa.

Statistics Canada. (2007a). 2006 Census Trends. Statistics Canada Catalogue no. 92-596-XWE.

Statistics Canada. (2007b). 2006 Census Dictionary. Statistics Canada Catalogue no. 92-566-XWE.

Stevens, S. S. (1951). Handbook of Experimental Psychology. New York: Wiley.

Strong Neighbourhoods Task Force. (2005). Strong Neighbourhoods: A Call to Action. Toronto.

Suttles, G. D. (1972). The Social Construction of Communities. Chicago: University of Chicago Press.

Taylor, R. R., and J. Covington. (1988). Neighbourhood changes in ecology and violence. Criminology, 26, 553–589.

Toronto Community Foundation. (2005). Toronto’s Vital Signs 2005. Toronto, Ontario.

Townsend, P. (1979). Poverty in the United Kingdom. Harmondsworth: Penguin.

———. (1993). The International Analysis of Poverty. Milton Keynes, UK: Harvester Wheatsheaf.

United Nations. (2003). Report of the World Summit on Sustainable Development. New York: United Nations, Department of General Assembly Affairs and Conference Services.

———. (2006). Human Development Report. New York: Oxford University Press.

United Way of Greater Toronto. (2007). Losing Ground: The Persistent Growth of Poverty in Canada’s Largest City. Toronto.

United Way of Greater Toronto and Canadian Council on Social Development. (2004). Poverty by Postal Code: The Geography of Neighbourhood Poverty, City of Toronto, 1981–2001. Toronto.

Valetta, R. (2005). The Ins and Outs of Poverty in Advanced Economies: Poverty Dynamics in Canada, Germany, Great Britain and the United States. Income Research Paper Series. Cat. no. 75F0002MIE2005001. Statistics Canada. Ottawa.

Ventakesh, S. A. (1996). The gang in the community. In Ronald Huff (Ed.) Gangs in America (2nd ed., pp. 241–255). Beverly Hills: Sage.

Vranken, J. (2002). Belgian Reports on Poverty. Paper presented at the conference reporting on the Income Distribution and Poverty – Perspectives from a German and a European Point of View. Berlin, (February).

Walks, R. A., and L. S. Bourne. (2006). Ghettos in Canadian cities? Racial segregation, enclaves and poverty concentration in Canadian urban areas. The Canadian Geographer, 50, 273–297.

Weisstein, E.W. (2007). Mathworld. Online: <>.

Wilson, W. J. (1987). The Truly Disadvantaged; The Inner City, the Underclass and Public Policy. Chicago: University of Chicago Press.

Wolfson, M. C. and J. M. Evans. (1990). Statistics Canada’s Low Income Cut-Offs: Methodological Concerns and Possibilities — A Discussion Paper. Analytical Studies Branch. Statistics Canada. Ottawa.

Wolfson, M. C. and B. B. Murphy. (1998). New views on inequality trends in Canada and the United States. Monthly Labor Review, 121(4), 3–23

Wright, J. P. and F. T. Cullen. (2001). Parental efficacy and delinquent behavior: Do control and support matter? Criminology, 39, 677–706.

Appendix 1:

Domains and indicators of relative disadvantage*







Appendix 2:


Census Geographic Units of Canada*

* Source: Statistics Canada Census Dictionary, 2006

Polygons formed by the intersection of streets.

Dissemination Areas
The smallest geographic areas for which census data are made available by Statistics Canada. The areas are composed of one or more neighbourhood blocks with a population of 400 to 700 persons. DAs respect the boundaries of census subdivisions and census tracts and therefore remain stable to the extent that census subdivisions and census tracts do. Each DA is assigned a four-digit code that is unique within a census division and a province or territory. In order to identify each DA uniquely in Canada, the two-digit province code and the two-digit CD code must precede the DA code.

Census Tract
Relatively stable geographic areas that usually have a population of 2,500 to 8,000. CTs are located in large urban centres that must have an urban core population of 50,000 or more

Census Subdivisions
Usually correspond with the municipalities of Canada.

Census Metropolitan Areas
A grouping of census subdivisions comprising a large urban area (urban core) and those surrounding “urban fringes” and “rural fringes” with which it is closely associated. To become a CMA, an urban area must register an urban core population of 100,000 at the previous census. CMA status is retained even if the core population later drops below 100,000.

Postal Code
Postal codes take the form ANA NAN (alphabetic character, numeric character, alphabetic character, numeric character, alphabetic character, numeric character). The first character represents a province or territory.

Forward Sortation Areas
Forward Sortation Areas are identified by the first three letters of the postal code and are associated with a postal facility from which mail delivery originates. The average number of households served by an FSA is approximately 8,000, but the number can range from zero to 10,000 households. Zero households exist because some postal codes contain only businesses (zero households).

Letter Carrier Walks
The average number of households to which mail is delivered by mail carriers based in local FSAs. Such walks may cover more than one postal code. The average number of households served by a postal code is approximately 19, but the number can vary between zero and 10,000.


Average Household Income
The weighted mean total income of households in the year preceding the census.

Economic Families
A group of two or more persons who live in the same dwelling and are related to each other by blood, marriage, common-law relationship, or adoption

Low Income*
(*Source: Low Income Cut-Offs for 2006 and Low Income Measures for 2006. Statistics Canada,Cat # 75F0002MIE -# 4.)
Refers to the position of an economic family or an unattached individual 15 by years of age and over in relation to Statistic Canada’s low-income cut-offs.

Low Income Measures
(a)LIM, the most commonly used low income measure, is a fixed percentage (50%) of median adjusted family income, where “adjusted” indicates that family needs are taken into account. Adjustment for family size reflects the fact that a family’s needs increase as the number of family members increases. Similarly, the LIM allows for the fact that it costs more to feed a family of five adults that it does to feed a family of two adults and two children. Procedures for calculating adjusted family size and adjusted family income are described in Statistics Canada publication 75F0002MIE-2005003.

(b)LICO is an income threshold below which a family will likely devote a larger share of its income to the necessities of food, shelter and clothing than the average family does. The approach is essentially to estimate an income threshold at which families are expected to spend 20% more than the average family on food, shelter and clothing. Such families are in “straightened circumstances.” LICO’s cut-offs vary by seven family sizes and five different populations of the area (urban vs. rural) of residence. Two types of LICOs, before-tax and after-tax, are available for use by researchers. Before-tax LICOs only partly reflect the entire redistributive impact of Canada’s tax/transfer system because they include the effects of transfers but not taxes. After-tax LICOs also take the effects of income taxes into account. Moreover, since the purchase of necessities is made with after-tax dollars, it is logical to use people’s after-tax income to draw conclusions about their overall economic wellbeing. For these reasons, Statistics Canada advocates the use of after-tax LICOs.

Procedures for calculating LICOs are described in the publication describing calculations for LIM adjustments.

Income Inequality*
The GINI coefficient is one of a number of measures developed to measure inequality in the distribution of income. More specifically, it shows the percentage of income received by a given percentage of the population. For example, households can be divided into fifths (quintiles) and ranked according to their aggregate household income. Then the percentage of Ontario’s or Canada’s income earned by households in the lowest to the highest quintiles can be calculated. If the top fifth receives over 80% and remaining fifths receive only 20% of Ontario’s/Canada’s income, then the GINI coefficient will come close to equaling zero. On the other hand, if 20% of households received 20%, 40% of households received 40%, 60% of households received 60% and 80% of households receive 80% of Ontario’s/Canada’s income, then a straight line (diagonal) would describe the relationship between percentage of income and percentage of households and the GINI coefficient would be 1.

A Lorenz curve is frequently used to illustrate inequality in income and wealth, and the GINI coefficient transforms the curve into a single variable. Specifically, the GINI coefficient is A/(A+B), where A is the area between the diagonal (perfectly equitable distribution), and B is the area under the curve.

Appendix 3:

Comparison of Previous Studies

Table 1: Methodology implemented in selected income inequality studies, 1986–2007
Authors and Dates Study Design Sample Measurement Indicators of Relative Disadvantage Data Collection Data Analysis
Long C.S1 Prob. Foc.2 Single Multiple3 Sec. Prim.4 Bi Multi5
Federation of Canadian Municipalities (2007) X X X X X X
United Way of Greater Toronto (2007) X X X X X
Canadian Council on Social Development (2007) X X X X X
Hulchanski (2007) X X X X X
Frenette et al. (2006) X X X X X
Walks and Bourne (2006) X X X X X
Brady (2005) X X X X X
Capellari and Jenkins (2004) X X X X X
Hou and Miles (2004) X X X X X
Heiss and McLeod (2004) X X X X X
United Way of Greater Toronto (2004) X X X X X
Ley and Smith (2000) X X X X
Miles, Picot and Pyper (2000) X X X X X
Kazemipur (2000) X X X X
Kazemipur and Halli (2000) X X X X X
Hajnal (1995) X X X X X
MacLachlan and Sawada (1997) X X X X
Wilson (1987) X X X X
1. longitudinal/ cross-sectional
2. probability/ focused
3. single/ multiple
4. secondary/ primary
5. bivariate/ multivariate

Table 2: Domains of disadvantage identified in studies of relative disadvantage
Authors and Date Income Housing Family Education Employment Ethnicity Welfare Dependency Violent Crime
SRDC (2007) X X X X
Sharkey (2006) X X X X
Noble and associates (2006) X X X X
Mears and Bhati (2006) X X X
International Youth Survey (2006) X
Belair and McNulty (2005) X X X
Valetta (2005) X X X X
Strong Neighbourhoods Task Force (2005) X X X X
Hou and Miles (2004) X X X
Heiss and McLeod (2004) X X
Kubrin and Hertig (2003) X X X X
Messer et al. (2001) X X X
Layle, Nolan and Whelan (2000) X X X
Ley and Smith (2000) X X X X X
Osgood and Chambers (2000) X X X
Bellair and Rosigno (2000) X X
Krivo and Peterson (1999) X X X X
Rosenfeld, Bray and Edgley (1999) X X X X X X
Ricketts and Sawhill (1998) X X X X
Sampson and Raudenbush (1997) X X X X
Harries (1997) X X
Wilson (1996) X X X X
Bursik and Grasmick (1993) X X
Broadway (1992) X X X X
Land et al. (1990) X X X
Hughes (1990) X X X X
Taylor and Covington (1988) X X X X X
Shihadeh and Ousey (1988) X X

Appendix 4:

Additional Variables for Further Study

Variables listed under the heading Relative Disadvantage may be of interest to researchers interested in creating their own indices of relative disadvantage. For researchers interested in youth violence, one of the outcomes to which relative disadvantage may lead, data are available from a variety of sources. These data can be subsumed under two headings. The first one is Available but not Obtainable. The data are not available because they are because they are protected by the Canadian Criminal Code, The Youth Criminal Justice Act or privacy laws.

The second heading is Available and Obtainable with Costs. Costs can take the form of additional time, money and effort (work/persuasion). Such costs can be incurred in collecting data of different kinds from the sources listed below.

Relative Disadvantage

Youth Violence

Using data collected from these sources will almost certainly involve “fitting costs” because they are unlikely to have been collected specifically for DAs. Fitting costs include paying research assistants to fit data collected for larger or other geographical units to DAs. For example, it took a considerable amount of unpaid time for 25 students in my Youth Street Gangs Course (2005) to fit police patrol area crime data to DAs in the 13 “unmet-needs” communities in the City of Toronto. In some cases, it may not be possible to fit data collected for larger geographical units to disseminations areas because they are aggregated in “health region” data collected by regional hospitals or “catchment data” collected by schools.

*The income measures and spatial terms referred to in this proposal are defined in the Glossary (pp. 61–62xx [this report]).

* *Identified in studies of the spatial distribution/concentration of ”poverty” and relative disadvantage.


Volume 1. Findings, Analysis and Conclusions

Volume 2. Executive Summary

Volume 3. Community Perspectives Report

Volume 4. Research Papers

Volume 5. Literature Reviews