EXPANDING THE BORDERS OF ENVIRONMENTAL INEQUALITY: AN EXAMINATION OF SOCIODEMOGRAPHIC PREDICTORS OF INEQUALITY IN ONTARIO, CANADA DARCY ROSE HAUSLIK

Transcription

1 EXPANDING THE BORDERS OF ENVIRONMENTAL INEQUALITY: AN EXAMINATION OF SOCIODEMOGRAPHIC PREDICTORS OF INEQUALITY IN ONTARIO, CANADA By DARCY ROSE HAUSLIK A thesis submitted in partial fulfillment of the requirements for the degree of MASTER OF ARTS IN SOCIOLOGY WASHINGTON STATE UNIVERSITY Department of Sociology MAY 2015 Copyright by DARCY ROSE HAUSLIK, 2015 All Rights Reserved

2 Copyright by DARCY ROSE HAUSLIK, 2015 All Rights Reserved

3 To the Faculty of Washington State University: The members of the Committee appointed to examine the thesis of DARCY ROSE HAUSLIK find it satisfactory and recommend that it be accepted. Jennifer Schwartz, Ph.D., Chair Gregory Hooks, Ph.D., Co-Chair Raoul S. Líevanos, Ph.D Emily Huddart Kennedy, Ph.D. ii

4 ACKNOWLEDGEMENT I would like to thank the many people who made this complete thesis possible. First, I d like to thank my committee for their valuable insights and guidance throughout this process. I must also thank Katherine Bittinger for answering all of my questions and demystifying the process. No acknowledgement section would be complete without including my mom who has been a tremendous source of inspiration and support. iii

5 EXPANDING THE BORDERS OF ENVIRONMENTAL INEQUALITY: AN EXAMINATION OF SOCIODEMOGRAPHIC PREDICTORS OF INEQUALITY IN ONTARIO, CANADA Abstract by Darcy Rose Hauslik, M.A. Washington State University May 2015 Chairs: Jennifer Schwartz and Gregory Hooks Quantitative research documenting the presence of environmental inequality in the United States is well established. While there is rising evidence that issues of environmental inequality do not respect national boundaries and spread far beyond the United States, quantitative analyses in different national contexts are limited. This project discusses and addresses this gap by applying theories of environmental inequality and justice popularized in the United States to explain the distribution of environmental burdens in Ontario, Canada. This thesis tests the extent to which socioeconomic class and race - the main indicators known to be associated with environmentally unequal outcomes in the United States are associated with environmentally unequal outcomes in the Canadian context. The results show clear support for the importance of socioeconomic status in predicting environmental hazards and mixed support for racial and ethnic indicators. These results are consistent with previous research in both the United States and Canada, and suggest that valuable insights about the iv

6 nature and extent of environmental inequality can be found through future international and comparative works. This project also provides insight to the role that scale plays in studying environmental inequality. v

7 TABLE OF CONTENTS ACKNOWLEDGEMENT... iii Page Abstract... iv LIST OF TABLES... viii LIST OF FIGURES... ix INTRODUCTION... 1 LITERATURE REVIEW... 5 Environmental Inequality... 5 Existing Theoretical Framework... 7 Environmental Justice and Inequality In Canada (De)Industrialization In Ontario Social Inequality in Canada Scale DATA AND METHODS Dependent Variables Data Limitations Independent Variables Income vi

8 Percent Renting Unemployment Rate Education Measures of race and nativity Proxy measurement for industrial zone Evolving Methods Unit-Hazard and spatial methods Improving spatial methods ANALYSIS Analytic Strategy LIMITATIONS DISCUSSION CONCLUSION REFERENCES APPENDIX vii

9 LIST OF TABLES Table 1. Descriptive Statistics Table 2. Independent Variables Table 3. Demographic Comparisons of Means Table 4. Logistic Regression Results viii

10 LIST OF FIGURES Figure 1. Unit-Hazard Incidence Figure 2. Spatial Methods Figure 3. Community-Centric Spatial Methods ix

11 INTRODUCTION Studies concerning environmental inequality problematize the assumption that environmental degradation affects everyone equally (Downey 2005) contending instead that environmental quality is fundamentally linked to social hierarchy. For the last quarter century quantitative analyses have shown the extent to which environmental burdens are disproportionately borne by the poor and people of color. Such studies are well established in the context of the United States (Been 1994; Ringquist 2005; Mohai and Saha 2007; Downey and Hawkins 2008) but have been sparingly utilized in different national or international contexts (Buzzelli and Jerrett 2003; Jerrett et al 2001; Mitchell and Dorling 2003). Movements for environmental justice have already spread beyond the borders of the United States, as have statements about the environmental inequality that necessitate such movements. Some academic research and international social movements take a global perspective as the processes of production, pollution, and waste disposal now span all regions of the world. Environmental justice scholars have expanded their concerns to include climate change (Roberts and Parks 2007) and the movement of toxics around the world (Pellow 2007) to name a few examples. However, though the Environmental Justice Movement has moved beyond its U.S. roots, systematic studies of environmental inequality prevalent in the U.S. are not used to the same extent outside of the U.S. Here, I discuss the potential for exporting the now classic questions of environmental inequality to other national contexts beginning with a case study of Ontario, Canada. Ontario serves as an important test site for understanding environmental inequality in other regional contexts as it shares a common industrial history with the Great Lakes region of the United 1

12 States and therefore a similar pattern of pollution and potential environmental burden but has not seen a comparable environmental justice movement nor the extensive studies of environmental inequality that are popular in the United States. In addition, I explore the potential reasons why studies of environmental inequality have not gained traction in other national contexts by highlighting the unique social movement roots that produced such academic works in the United States. I also discuss the particular history of racial inequality and segregation that led to disparate environmental outcomes in the United States in order to discuss the potential for applying these themes elsewhere globally. The lack of consistent analysis of environmental burdens in different scales and contexts is a clear gap in our understanding of environmental inequality. This gap is even more important in the current globalized context where hazards do not respect political boundaries. The goal of this project is to set the stage for a direct international comparison between both nations comprising the Great Lakes Region of North America. The former steel and automotive capital of the continent spans the United States and Canada and provides an integral test case for the advancement of environmental inequality on an international scale. The first step to comparison in this case is description. This project will be firmly situated in the large scale quantitative accounts of environmental inequality in answering the first, and principal research question: What characteristics of a community are associated with the presence of a hazardous waste generator in Ontario? To answer this first question I use Ontario census data from 2001 and a manifest of all generators of hazardous and large quantity waste. These data sets help to determine the 2

13 demographic differences between communities that have facilities within 500 meters and communities that do not. This research question needs to be answered completely before moving on to the second, more theoretically interesting question: How do these community characteristics compare to those associated with hazardous facility presence in the United States? This project only looks at producers of hazardous wastes in Ontario, making a direct comparison at this point impossible. Instead, this question will be answered based on the trends uncovered from this Ontario data and the trends already identified in the U.S. literature. Finally, this project questions scale in a way that is not usually done in studies of environmental inequality by asking: Do the demographic characteristics associated with the presence of a hazardous waste generator change depending on the operational and observational scale of interest? This specifically addresses concerns of the modifiable areal unit problem (MAUP) which states that different statistical relationships may present themselves at different scales and will highlight the extent to which environmental inequality studies should be concerned with the scale at which their relationship is articulated. Of particular concern to this study is the presence of environmental inequality in Toronto. Toronto is home to one-fifth of the Ontario population and boasts substantially higher numbers of visible minorities and new immigrants. This question poses what the differences might be if the larger region of Ontario is the observational scale or the specific city of Toronto is the observational scale while keeping the units of analysis constant. 3

14 I show that the dominant methods and theories associated with environmental inequality in the United States are useful in understanding the distribution of environmental inequality in Ontario. This opens the door for later direct regional comparisons between the Great Lakes region of the United States and Ontario. Ontario is an important test case as it shares a common economic history with the Great Lakes region and this economic history often manifests itself as the environmental problems of today. 1 1 This project can best be understood as a continuation of environmental inequality studies conducted at Washington State University. This study builds on the work of Katie Bittinger and Leah Thorpe who thoroughly analyzed the distribution of hazards in the United States (Thorpe 2014) and the application of EPA enforcement (Bittinger 2013) 4

15 LITERATURE REVIEW Environmental Inequality Scholarship on environmental inequality challenges earlier conceptions that environmental problems impact everyone equally (Pellow 2000; Mohai and Bryant 1992). Though the study of environmental inequality can be considered as analytically distinct from the Environmental Justice Movement the social movement which seeks to correct unequal outcomes the history of environmental inequality as an academic pursuit and the social movement can best be considered as part of a larger environmental justice praxis. The environmental justice praxis (Sze and London 2008) situates environmental justice as a field on the crossroads between scholarship, social movements and public policy. The success of the Environmental Justice Movement in the United States dictated the direction that scholarship regarding environmental inequality would take and the outcomes of environmental inequality research would shape the social movement. Therefore, this literature review, in keeping with this view of environmental justice as a praxis, describes both the roots of the Environmental Justice Movement and notable scholarship in the field of environmental inequality insofar as these histories both help to explain the larger circumstances of the United States and might help to understand why no similar movement or academic discourse is popular in Ontario. Pellow defined environmental inequality as a field of inquiry that focuses on [broad] dimensions of the intersection between environmental quality and social hierarchies. Environmental inequality addresses structural questions that focus on social inequality (the unequal distribution of power and resources in society) and environmental burdens (2000: 582). Quantitative research documenting the intersection of environmental inequality and social 5

16 hierarchies provides valuable insights into identifying the characteristics of those communities that are more likely to be affected by environmental burdens (Been 1994; Ringquist 2005; Mohai and Bryant 1992; Downey 2007). The most important factors that determine the distribution of environmental burdens in the United States are class and race. The majority of studies find that communities with high proportions of minority or low income residents, or some combination of the two, are significantly more likely to be affected by environmental burdens (e.g. Mohai and Bryant 1992; Hamilton 1995; Downey and Hawkins 2008; Saha and Mohai 2005) This challenge came out of a series of grassroots movements in the United States that took issue with disparate siting of hazardous waste facilities, zoning of toxic industries and exposure to workplace hazards within low-income communities of color in the 1980s. The most influential and highly publicized of these cases was in Warren County, North Carolina where a PCB landfill site was slated to be built in a low-income, predominately African American community. The local community rallied but was ultimately unsuccessful in thwarting construction. However, by framing this as an instance of environmental racism, this struggle was highly influential in shaping the debate around issues of disparate siting of environmental hazards (Bullard 1990). The case of Warren County, North Carolina attracted national attention and served as an impetus to carve out a voice for minority groups in the previously elite driven, narrowly construed field of environmentalism (Taylor 2000). This marks both the beginning of the environmental justice movement and the academic study of environmental inequality. These terms are both politically charged in their own right: whereas the term environmental racism focuses on the disproportionate impact of environmental hazards on communities of color, environmental justice is focused on ameliorating potentially life-threatening conditions or on 6

17 improving the overall quality of life for the poor and/or people of color (Pellow 2000: 582). The roots of environmental inequality literature in this politically and morally charged debate of environmental justice has had a significant effect on the type of scholarship done in this sociology subfield. The academic discourse in sociology (when not studying the dynamics of the environmental justice movement from a social movement perspective) has focused narrowly on issues related to the distribution of environmental hazards. The social sciences have established methods and developed an affinity for studies of inequality that make the narrow issue of distribution a prime avenue for advancing the goals of this social movement. Sociology was uniquely positioned to systematically account for the prevalence of the claims of environmental racism. Existing Theoretical Framework Initially, theories about environmental inequality were made in terms of simple racist or discriminatory intent (Bullard 1990). These claims of environmental racism were in line with the political path taken by the budding environmental justice social movement. As the environmental justice movement evolved, they began to look beyond simple issues of distribution to the structurally unjust and fundamentally undemocratic processes that silence minority voices in making environmental decisions (First People of Color Environmental Leadership Summit 1991). This evolution in the environmental justice movement pushed environmental inequality theory beyond simple models of environmental racism. Although what Saha and Mohai (2005) refer to as discrimination models are still popular explanations for unequal environmental outcome, these models now face competition from rational choice and sociopolitical explanations. Competition might be the most apt description 7

18 for how these models are addressed in the literature that features a dubious debate between whether class or race is the better predictor of environmental burden. Those who seek to advance the argument that race is more significantly linked to pollution (e.g., Mohai and Bryant 1992; Downey 2005) offer residential discrimination or racist intent as the mechanisms of environmental inequality. Those who argue class is the more significant predictor (e.g., Bowen et. al 1995) explain coincidence of race and environmental burden as a consequence of poverty within racial minority communities and not race itself. However, there is substantial warning in the literature about the dangers of this false debate and urging scholars to view race and income as interrelated factors (eg. Downey 1998). Researchers in the field headed this research agenda and set about explaining disparities in distribution via three major theoretical avenues: racial discrimination models, rational choice models, and sociopolitical models (Saha and Mohai 2005). Racial discrimination models, as already discussed, focus on intentional or institutionalized discrimination in siting decisions. Rational choice models emphasize the market efficiencies in siting facilities in low-income and high-minority neighborhoods which have lower land prices (Saha and Mohai 2005). Sociopolitical explanations, or what Pellow (2000) dubbed the perpetrator-victim scenario, argue that environmental inequalities occur when the poor or people of color are dumped on or exposed to hazards because they are less powerful than corporations and the state (p. 587). This lack of political power is also articulated as relative to other racial and income groups which can fight off environmental hazards in theories that can be broadly referred to as the path of least political resistance. These theories state that communities with high-minority and low-income populations were the least able to prevent the siting of hazardous facilities (Liu 2001),and would play host to hazards because they had the least access to social, political and economic capital. 8

19 Importantly, the intersection of these theoretical avenues as complimentary rather than adversarial explanations is also examined in historical analyses which focus on environmental inequality formation (Pellow 2000), spatial processes which result in environmental inequalities (Pulido 2000), and longitudinal analyses (Crowder and Downey 2010; Downey 2005). Underlying political economic theories of environmental inequality is a larger framework called the treadmill of production, which is a political economic theory that has had considerable influence in environmental sociology (Brulle and Pellow 2006). Environmental inequality concerns itself with two fundamental and related byproducts of modern industrial capitalism: environmental degradation and social stratification. The environmental degradation that results from capital intensification and accumulation was theorized originally by Schnaiberg (1980) as resulting in a treadmill of production that stipulates that environmentally adverse outcomes are the externalities of the expansionary logic of capital accumulation and unabated technological development. These environmentally adverse outcomes manifest as withdrawals (associated with the extraction of resources) and additions (typically in the form of pollution) to the environment that outpace and often outlast our abilities to curtail the wastes of production or sustainably manage the material resource inputs. Additions and withdrawals are distributed in society through the unquenchable search for profit under the treadmill of production logic (Hooks and Smith 2004). These forces rarely, if ever, lead to environmentally just outcomes. The basic social force that drives the treadmill is the intrinsic nature of competition and concentration of capital (Schnaiberg 1980). It is this fundamental nature of capitalism that allows us to expand our analysis beyond the border of the United States to other contexts that have similar capitalist economic organization. The same forces that result in pollution in the United 9

20 States function in different contexts and on different scales. In a globalized world it might be counterproductive to limit our analyses to national contexts at all. Environmental Justice and Inequality in Canada While there is a great deal of convergence between the United States and Canada, the Canadian literature on environmental inequality has taken a distinct path. As reflected in a special issue of Local Environment, there is a divide in opinion on how best to approach the subject in relation to how environmental justice is framed in the United States (Haluza-DeLay 2007). Many studies borrow heavily from U.S. literature and others caution against the crude exportation of the environmental justice frame across borders. To illustrate the necessity for an explicitly Canadian-grounded approach, Debbané and Keil (2004) document the differences in the processes of racialization, spatialization, and the political groundings between the United States, Canada, and South Africa. The differences between these processes result in a caution against the universalization of the term environmental justice. They argue instead that notions of environmental justice should be localized to account for diverse concerns and local constraints. Ultimately, Debbané and Keil (2004) argue that studies of environmental justice need not be restricted to particular sites but that every study needs to consciously and methodically take stock of the particular case of interest. The special issue of Local Environment hosts examples of comparative works between the United States and Canada that accomplish nuanced account of environmental inequality and justice studies. The introductory account situates the Canadian context as nearly inseparable from that of the United States owing to the entangled political, economic, academic spheres crossing these national boundaries (Haluza-Delay 2007). Specifically, three articles from this special issue 10

21 begin to compare and contrast environmental justice between the United States and Canada. Two of these articles speak to the multiple ways that environmental burdens affect indigenous communities which is a population of study that is highlighted in Canada but largely overlooked in U.S. studies. Mascarenhas (2007) discusses the ways that neoliberal reforms have disproportionately impacted indigenous communities by limiting sovereign water rights among the First Nations in southern Ontario. The framing of the environmental justice debate in terms of neoliberal threats to the tradition of social welfare policies highlights the different political contexts that shape environmental debates. Trainor and her coauthors (2007) also gets at issues of indigenous communities and political recognition in their direct comparison of the political disputes over climate impacts in the arctic. They compare the relative political power expressed by arctic and subarctic groups of Canada to those of the United States and concluded that the challenges faced by these two national groups are similar but that those in Canada are more able to effect change on the national level. These differences correspond with state policies of multiculturalism and group rights in Canada compared with the strong individual rights emphasis of the United States. Finally, Teelucksingh (2007) argues that because minority groups are not constructed in the same ways or spatially isolated to the same extent in Canadian cities, attempts to discuss environmental inequality in terms of environmental racism are not feasible in Canada. Instead, she draws from the body of literature in the United States to discuss a process of environmental racialization that better articulates the Canadian context. These articles all agree that transplanting environmental frameworks across national borders can lead to distortions and urge caution when borrowing from U.S. literature to examine the Canadian context. However, in pointing out these barriers and artfully negotiating the local 11

22 contexts that they described they provided models for thoughtful and nuanced scholarship which applies environmental inequality to a diverse range of populations in Canada. While these are informative cases illustrating the benefits and limitations of using theories of environmental inequality and justice outside of the United States, it is also important to note the state of environmental justice in Canada independent of the United States. The environmental justice movement within Canada specifically is less uniform in its form and functions than that of the United States. It is often subsumed within the social justice and human rights movement, and few Canadian publications have been published using an explicit environmental justice framework (Agyeman et al. 2009). Studies of environmental inequality are emerging, structured to analyze the proximity between vulnerable social groups and environmental burdens within Canada, and this literature includes several notable examples (e.g., Agyeman et al 2009). Several focus on the total suspended particulates, a measure of air pollution, in the city of Hamilton, Ontario (Buzzelli et al 2003; Jerrett et al 2001), chosen due to advanced air pollution monitoring systems located there. The results are consistent with U.S. findings: socioeconomic variables, and to a murkier extent, a race variable, were significant predictors of air pollution in Hamilton. As Hamilton is within the Great Lakes region, to which this project eventually hopes to generalize, these studies are important test cases to prove the theory that environmental inequalities operate in some uniform way across national contexts organized in a similar fashion, economically. Ontario, Canada provides an ideal case for discussing the potential expansion of studies of environmental inequality and justice beyond the borders of the United States. Environmental inequality describes the way that environmental burdens are distributed unevenly; this means that 12

23 it is necessary to understand both the histories of environmental degradation and social inequality to begin to understand the interplay between the two in the contemporary context. The industrial history of Ontario is fundamentally linked to the industrial trajectory of the Great Lakes Region of the United States and through this common history we are able to directly compare the histories of environmental degradation in both locations without much difficulty. The history of social inequality in Ontario, and Canada more generally, does differ from that of the United States in some fundamental ways. Both the industrial history and some of the prominent differences in the history of social inequality are considered at length. (De)Industrialization In Ontario By the turn of the 19 th century, Ontario s economy was defined by resource extraction in Northern Ontario particularly logging to support the paper needed for the booming newspaper industry in Northeastern United States and the Industrial Heartland in Southern Ontario. This region was colloquially referred to as the Golden Horseshoe. The industries prevalent in Southern Ontario will be familiar to anyone acquainted with industry in the former glory of the United States Rust Belt. Certain neighborhoods were particularly important to specific sectors of industrial production; Hamilton was known as steel town, Windsor became the home of automobile manufacturing, and Sarnia hosted some of the more toxic chemical processes. Rubber manufacturing and the production of agricultural machines were also important. On either side of the border, the Great Lakes Basin become known as the Industrial Heartland of their respective nations (White 1985). Both interconnected economies experienced the boon of the post-war years, though by the 1970s economic hardships threatened to devastate these economies. The industrial towns of 13

24 Ontario were often more diversified than their counterparts to the south and thus did not experience the hardships of deindustrialization in the same way. Though Ontario shares a common economic base with the Rust Belt, the Golden Horseshoe never gained the infamy associated with rusting (High 2003). This likely prepared Toronto and the rest of Ontario for a less dramatic transition from old to new economy jobs. The industrial legacy is integral to the environmental burden of concern in this study: generators of large quantities of waste in Ontario. With the difficulty in rezoning either to or from an industrial zone these histories of manufacturing and industry serve as the sites of environmental degradation today. Social Inequality in Canada While the industrial hazards might share common roots, there are some notable differences in what constitutes a vulnerable population and how these populations are arranged around industrial hazards. Environmental inequality in the U.S. focuses mostly on social class and race as predictors of environmentally unequal outcomes. As discussed previously, these groups face burdens on the basis of historical processes that render them less able to escape or halt these burdens via the operation of the treadmill of production or in some theories the prevalence of institutional racism. There is strong evidence for the operation of the Treadmill of Production and the facilitating logic of neoliberalism in determining socially and environmentally unequal outcomes. Buzelli and co-authors (2003) find clear evidence that social class is a powerful predictor of environmental degradation in the instance of air pollution in Hamilton, Ontario. Likewise, Teelucksingh and Poland (2011) identify some of the mechanisms by which neoliberalism in Ontario create environmentally unequal results with regards to energy sources. 14

25 They describe neoliberalism as manifesting as reduced opportunities for public participation in decision making: the repeal of land-use planning requirements intended to curb urban sprawl; and reduced budgets for provincial and local environmental and natural-resources agencies (189). These are enigmatic of the logic behind the Treadmill of Production: the market has the final say in resource distribution, good and bad. Mascarenhas (2007) specifically implicates neoliberal reforms of the 1990s as producing environmentally discriminatory results for Canada s indigenous populations. Socioeconomic class as a predictor of environmental justice seems well established on both sides of the border. Theories of race, however, are more problematic to export. The two nations have qualitatively different racial histories. This project is not an attempt to shoehorn a model of environmental inequality from the United States to the Canadian context, but an attempt to see to what extent this model might hold. The most glaring difference in the racial histories of the U.S. and Canada is the lack or near lack of slavery in Canada as opposed to the widespread enslavement of Africans and African American descendants that lasted centuries in the United States. In America, this history has led to widespread residential segregation, institutional barriers to economic incorporation, and generally lowered life chances. Residential segregation as a mechanism for predicting unequal environmental exposure is particularly important in studies of environmental inequality. Canada does not have the same widespread history of residential segregation (Fong 1999) though there is evidence that residential segregation is expanding in the Canadian cities (Keil, Ollevier and Tsang 2009). In addition to the spatial differences in patterns of segregation, Canada has a long-standing commitment both culturally and politically to multiculturalism as an ideology and 15

26 public policy, which reflects their commitment to immigration as a fundamentally important demographic and economic force (Berry 2013). This long-standing ideology of multiculturalism specifically multiculturalism constructed in opposition to the social inequality prevalent in the United States as a social identity may have stunted the exportation of an Environmental Justice Movement and the study of environmental inequality in the Canadian context (Keil, Ollevier and Tsang 2009). Though there is evidence of both increasing residential segregation and increasingly racialized patterns of poverty present in Canada, the official commitment to multiculturalism might be a veneer that promotes the treatment of environmental issues as narrowly construed technical issues (Keil, Ollevier and Tsang 2009). Studies of race and environmental inequality in the United States understand racism and racialized outcomes as part of dynamic sociospatial processes rather than the individual acts of racism that may be spatially expressed (Pulido 2000). This understanding of race as it relates to environmental inequality builds on the theoretical work of Omi and Winant (1994) that articulates the racial formation model or the process through which racial categories are constructed by state action justified by phenotypical differences between groups and then these racial categories are interpreted as static realities to distribute resources. Park and Pellow (2004) specifically articulate the racial formation projects - the explicitly political struggles between racial groups- to explain the different discursive justifications and material outcomes of environmental inequality in Silicon Valley. In studies of environmental inequality processes of racialization, or, how groups become imbued with racial meanings, are understood in spatial terms: Park and Pellow put it succinctly racism has shaped the very geography of American life (2004: 404). The primary 16

27 manifestation of the sociospatial process of racialization is residential segregation which stems from decades of state sanctioned policies including discriminatory lending practices, racially restrictive covenants, and red-lining (Pulido 2000). A large proportion of environmental inequality studies- including this one - point to the spatial distribution of environmental burdens relative to specific racialized groups. Such studies are predicated on the fact that racialized groups are to some extent spatially confined. Widespread residential segregation based on race does not characterize the landscape of Canadian cities (Debanné and Keil 2004; Fong 1996). This has led to the assumption that there are no comparable racialized dynamics of space in the Canadian context; that such problems are unique to the United States context (Debanné and Keil 2004). However, this assumption is more likely the outcome of the veneer of multiculturalism than an egalitarian reality. Teelucksingh (2007) reinterprets the processes of racialization in a way that is more apt to describe the Canadian context. Though her theory deviates significantly from more established concepts of racialization based strictly on race as expressed through phenotype as part of a political project, the outcome of how different groups are identified and deemed worthy of social risks and rewards as part of a hierarchy is consistent with theories of racialization in the United States. For Teelucksignh (2007) racialization is not reducible to race as any group can be racialized at some point. Racialization, as she argues in the Canadian context, is typically dependent on three factors: race, ethnicity, and nativity. These three factors work in concert to define certain spatially bounded communities as unworthy and ultimately vulnerable to environmental hazards in a way similar to those residentially segregated in the U.S. Though this conception of racialization would lead us to believe that there is a more nuanced process of racialization at 17

28 work in Canada there are clear cut parallels between the United States and Canadian racial histories. For example, a racially restrictive covenant in Hamilton, Ontario prevented the sale of land in question to Negroes or Asiatics, Bulgarians, Austrians, Russians, Serbs, Roumanians, Turks, Armenians, whether British subjects or not, [and] foreign-born Italians, Greeks or Jews (Re Byers and Morris 1931 cited in Mosher 1998:96). The idea that spatial manifestations of racialized differences is a distinctly American problem is in keeping with the self-perception of multiculturalism which, at least rhetorically, has come to define Canada as a nation. Though Canada s problems with racialization are most likely subtler than the glaring segregation present in the United States, the distribution of environmental hazards on the basis of race is a possibility worth testing in this project. In addition to the variable of race which is derived from the Canadian Census variable visible minority which measures the non-white or non-caucasian population. In keeping with the vital role of immigration in the national image and the loosely defined racialization of space in Canada a measure of nativity is treated similarly to this variable of race to determine if nativity functions in a manner similar to race in the United States with regards to the distribution of environmental hazards in Ontario. Scale Environmental inequality is fundamentally a spatial issue; the classic question of power is reframed as who gets what where? The methods of quantifying associations between environmental degradation and vulnerable communities rely on establishing spatial connections. The ever-evolving search for appropriate units of analysis and measurements of proximity are 18

29 discussed in more detail in the Data and Methods section. However, proximity is not the only spatial context that is present in environmental justice and inequality studies. Struggles to define the scale at which environmental inequality occurs and the scale at which environmental justice might be sought is also fundamentally important to this literature (Sze et al 2009). The local lived experiences of environmental burden are often divorced from the regional and even global forces that caused, and could potentially remedy, the suffering (Kurtz 2003). The politics of scale as the ability to define and curtail ecological, social, political, or institutional boundaries has become its own avenue of research for radical geographers (Swyngedouw 2004). This project engages with scale in a meaningful way by accepting that scale, in its diverse meanings, is socially produced though too often treated as static. Siu-Ngan Lam (2004) outlines the four meanings of scale in geography. The first, and only meaning not specifically addressed in this analysis, is scale as a cartographic ratio (i.e. one inch to one mile on a map). Second, scale can refer to the observational or spatial extent of the study, such as large-scale studies looking at larger areas and small-scale studies looking at relatively smaller areas. Third, the measurement scale, or resolution, operates like pixels in an image. This has been the main avenue of interaction with scale in mainstream environmental inequality studies. This project, as will be discussed later, uses the measurement scale of Dissemination Area (DA), a geographic unit associated with the Canadian Census, which captures units populated by residents. Lastly, the operational scale is the spatial extent at which processes operate; global warming operates on a global scale, flooding on a more concentrated scale, pollution, unfortunately, operates on many scales simultaneously. Observational scale is inherently related to operational scale and both are of concern in this paper. 19

30 The operational scale is complicated in environmental inequality. The distribution of hazards and their proximities to vulnerable communities operate simultaneously on multiple scales and often with different results. Some of the controversial race versus class debate comes down to the susceptibility of the race variables in this context to the classic modifiable areal unit problem (MAUP) in geography (Kurtz 2003; Sze et al 2009). MAUP states that different statistical relationships may present themselves at different scales and necessitates that researchers adequately engage in clear theoretical justifications for operational scale. This project engages with operational scale by beginning to parse out differences that occur within Toronto when compared to the larger Ontario region. The general dynamics of environmental inequality often manifest themselves differently within specific urban contexts, which is particularly true of Detroit, Michigan, US. As Downey (2005) illustrates, areas with high percentages of African Americans experienced lower rates of air pollution from manufacturing. This counterintuitive result is explained by spatial mismatch theory that predicts that African Americans will be spatially separated from potential economic opportunities. DATA AND METHODS I used the Canadian Census to gather information on the sociodemographic characteristics of communities in Ontario. The Canadian Census is undertaken by the national statistical agency Statistics Canada (StatCan). The sampling frame consists of all occupied dwellings in Canada. The census is administered via a physical questionnaire distributed to every doorstep of every occupied dwelling in Canada in the two weeks leading up to census day. Seemingly unoccupied dwellings are noted and census agents follow up at these addresses later. The census is generalizable to the non-homeless population of Canada. While there might be 20

31 systematic exposure of the homeless population to environmental hazards, this is beyond the scope of this project, which purposefully looks at residential settlement patterns. The Canadian census response rate is as close to 100% as is practicable. The independent variables of interest are fairly standard sociodemographic variables at the community level. Communities are conceptualized as people and institutions that hold particular cultural and political characteristics (Sampson 2012; Park and Burgess 1921). In this way, communities can be thought of as socially constructed and meaningful places (Lefebvre 1990). Measures of communities rather than aggregates might be theoretically relevant to questions of environmental inequality since the path of least resistance theories suggest that it is the relative lack of political power of communities that lead to disproportionate siting decisions. Unfortunately, the best operationalization of communities available on a large enough scale to do quantitative analysis is largely an artifact of census enumeration. This is the case in both the Canadian and United States censuses. The units of analysis chosen from the Canadian census are dissemination areas; these parallel the block groups of the United States which are the smallest census enumeration unit and should serve as a natural point of comparison for future research. In total there are over 16,000 dissemination areas in Ontario with an average population size of

32 Table 1: Means and Standard Deviation for Selected Demographic Measures (16,619 Dissemination Areas in the Province of Ontario, 2001 Canadian Census of Population) Median Household Income Percent that rents rather than owns Ontario Ontario without Toronto Toronto N= N=12,870 N=3,748 $60, 105 (26,039) 28% (28.44) Unemployment Rate 6% (5.02) Percent with Bachelor s Degree Percent Identifying as Visible Minority 19% (14.5) 17.5% (22.7) $60,333 (24,750) 24% (25.7%) 6.2% (4.96) 16.7% (12.7) 11.4% (17.2) Percent Foreign Born 31.8% (22.5) 24.7% (18) Note: Standard deviations in parentheses. Source: 2001 Canadian Census of Population, for details, see Statistics Canada (2003) $59,320 (30,037) 41% (32.9) 6.8% (5.2) 41% (32.9) 56% (19.6) 27% (44.3) The community-level sociodemographic variables of interest in this project are those that have been well established in studies of environmental inequality in the United States. These include variables of racial composition of neighborhoods, those of median income levels in communities, and a host of controls for political power and general neighborhood features (including urban/rural distinctions) (Been 1994; Ringquist 2005; Mohai and Bryant 1992; Downey 2007; Hamilton 2005). The key variables in this study are percent visible minority, median household income, percent renting, percent 1 st generation immigrant, and percent with a bachelor s degree or higher. 22

33 Dependent Variables The second data source used to get at the question of hazardous waste siting comes from the Ontario Ministry of the Environment (akin to the United States Environmental Protection Agency). This data source is the registration of all generators (and potential generators) of hazardous waste in Canada that include records of the quantity and characterization of waste produced by these generators. This is another attempt by a governmental agency to document all cases of a phenomenon in a geographic region but ultimately there are shortcomings. The sampling frame is an official listing of all registered manufacturers of hazardous waste in Canada as of The results are generalizable to industrial based waste generators in Canada but not to all sources of pollution or hazardous wastes. This data source was reduced further to look at large quantity generators (LQGs), a commonly researched polluter subgroup in the United States. LQGs are defined by the United States EPA as any facility that produces more than 1,000 kilograms of hazardous waste in any given month or as any facility that produces over 1 kilogram of acutely hazardous waste in any given month. In the US the EPA defines hazardous waste as a waste with properties that make it dangerous or capable of having a harmful effect on human health and the environment (EPA 2011). Acutely hazardous wastes are those substances that are capable of having a harmful effect on human health in much smaller doses (i.e. heavy metals). The definitions of hazardous and acutely hazardous wastes are similar (though not identical) between the United States and Canada, though some data manipulation was necessary to discuss LQGs in Canada. The registry of hazardous waste producing facilities is only available on a yearly basis while LQGs are designated on the basis of monthly production. Canadian generators were coded as LQGs if they produced 10,000 kilograms in the given year or 10 kilograms of acutely hazardous waste in the 23

34 given year. The logic behind this decision is that it was unlikely that facilities that produced this much waste in a year likely produced at least one tenth of it in a single month. LQGs were chosen as the outcome variable of interest for theoretical and practical reasons that increase both the validity and reliability of this and future comparative work. Theoretically, LQGs represent clear-cut cases of exposure that can serve as the basis for establishing environmental inequality. Practically, LQGs are of interest to the United States EPA meaning that these are a fruitful avenue for future comparisons because there is ample data available. Downey (2007) also suggests that while the particular hazard of interest does make a difference to analysis in environmental inequality, what is more important is consistently measuring the facilities. Facilities will be held constant for purposes of future comparison. Certain generators are present in the data set though they produced no waste in the time frame of interest. The list of waste handling facilities has been apparently appended over the years, where some facilities are reporting waste quantities of zero, which indicate that they have closed or changed their business practices in some profound way. These cases were dropped as they were not relevant to total waste produced in the given year of 2005, this year being chosen for two reasons. First, it represents a lag after the census data from 2001 that aligns better with the theory of siting as a result of least political resistance, indicating that the neighborhood would need to be a certain way in order to attract pollution hence the lag after the demographic information. Second, important changes were made in the recording of toxic waste facilities in the United States in this particular year, which could provide useful future comparisons. 24

35 Data Limitations A challenge to gathering data for this particular project is the availability of Canadian data. Canadian laws restricting data are different than those in the United States. In the U.S. there needs to be a reason to keep data private while in Canada there must be a good reason to make data publically available. In order to access data, affiliation with a Canadian university is often necessary and all data requests go through liaisons typically affiliated with university libraries. The census data set that I have available has a wealth of information, but the codebook was vague and information can only be checked on the tract level while my unit of analysis of interest is the block group level. Similar to the data availability issues discussed with the Canadian census, this data, though collected by the Canadian equivalent of the EPA, was made available for my study by a nonprofit organization that petitioned for this information. Because of this, I received access to the raw data and publically available documents that say what chemicals have what abbreviations but no official codebook. However, for initial stages of research that count facility presence, the geographic information is sufficient. In a cross national study of the movement of toxic wastes between the U.S. and Canada, it is likely that U.S. records would be used primarily as that information is more readily available and often more complete. Independent Variables The key independent variables were transformed into categorical variables. Categorical variables provide a more intuitive basis for interpreting results in logistic regression. Categorical variables will also prove useful in future comparisons as they are contextually based: as the categorical variables are relative to the other neighborhoods within a given context there is no 25

36 need to come up with absolute values of what constitutes a low income or largely minority neighborhood. The decisions of where to derive percentile cut points to form the discreet categories were reached based on empirical and theoretical considerations. Income Median income is a quintessential predictor of environmental inequality as it serves as a crude, yet efficient, measure of social class. Though not sufficient to describe all aspects of class, income is a crucial measurement in its own right. The median household income of communities forms the basis for the categories low, middle and high income. Communities that fall into the low-income category have a median household income below the Ontario poverty line for a family of four from 2001 ($35,471 Canadian Dollars). Those communities considered highincome have a median income of three times the poverty line ($106,413 Canadian). These thresholds also happen to closely approximate the standard deviation around the median income for the province low-income communities fall around one standard deviation below the mean and high income communities fall one standard deviation above. The average category encompasses those communities between these two thresholds. Percent Renting The percent of residents in a dissemination area that rent rather than own their home is another measurement of social class. This concept is also sometimes measured in the literature as home value or property value that is an indirect measurement of wealth. Renting vs. owning is also an appropriate, albeit crude, measurement of wealth in addition to sociopolitical power. Renters may have fewer ties to the community or less investment in the neighborhood, which could lead to outcomes consistent with theories that predict facility siting based on a lack of 26

37 community political power to rally against such decisions. The categories high and low renting were derived using the 75 th percentile (43%) as a cut point to indicate communities with a substantial renting populations. Communities where more than 43% of residents rented their home were coded high percent renters those with below 43% fell into the category low percent renters. Unemployment Rate Measures of occupation are often included in studies of environmental inequality. In this case, percent unemployed is used as another measurement of social class and collective political power in a community. The expected outcome is that those communities with higher rates of unemployment will be less likely to prevent any industry that might bring jobs, even if it might also bring pollution. Communities were coded as high unemployment rate if they fell at the 75 th percentile or above (8.6% or more unemployment) and low unemployment rate if they fell below the 75 th percentile. Education Education is another common measurement of social class and is particularly relevant to analyses of environmental inequality as it is often key in building social and political capital. Percent of the population with Bachelor s degrees or higher was chosen for this study. While the other variables so far have been measurements of disadvantage this is a measurement of advantage and thus the coding is reversed: those communities which fall below the 25 th percentile of percent with a bachelor degree were coded as lower education levels and those above the 25 th percentile were coded as higher education levels. Measures of race and nativity 27

38 Measures of race are central to studies of environmental inequality in the United States where some theories go so far as to attribute the entirety of unequal siting decisions to racist intent. While the particular racial history of Canada is different than that of the United States there is still a marked history of discrimination based on race and ethnicity specifically regarding labor market outcomes (Teelucksingh and Galabuzi 2005) and Ontario s legal and criminal justice systems (Mosher 1998), both of which are familiar lines of discrimination to an American audience. This history of racial and ethnic discrimination is particularly salient for native peoples and new immigrant groups. This analysis uses census designations of visible minority and firstgeneration immigrant to capture the concentration of minority populations within dissemination areas. The variable visible minority from the Canadian Census measures persons, other than Aboriginal peoples, who are non-caucasian in race or non-white in colour (StatCan 2011). The cut points to talk meaningfully about communities that are largely comprised of visible minorities and/or immigrant communities or the communities which lack this diversity were derived using a comparative approach advocated by Smith (1998) in regards to measuring racial integration. Sicotte (2014) successfully used this method in a study of environmental inequality in Philadelphia. The logic of a comparative approach for talking about diversity within communities is simple: it asks how diverse (or integrated) is this place in comparison to the larger context in which it is embedded. So, for example, if a county is 20% black and a community within that county reflects this proportion and is also 20% black that place is said to be diverse (or integrated) (Smith 1998). In this way the mean percentage of visible minority (17%) and percent foreign born (32%) serves as the mid-point for what we can consider substantially diverse. These means were divided by four and manipulated via the procedure 28

39 described in Appendix A to yield 7 categories. The two most extreme categories were collapsed for a total of 5 categories ranging from substantially visible minority or foreign-born to substantially white or native-born (See Table 2). The advantage to constructing cut points in this way is that the measures are relative to the local context and will thus be easy to compare across geopolitical borders. It also staves off a discussion of arbitrary absolutes as to what is diverse neighborhood. Proxy measurement for industrial zone Industrial zone is included in this analysis as an important, though crude, control variable to account for the basic geographic principle that things that are close to each other are closely related to each other. These categorical variables measure the average number of facilities that the neighbors within a 5 km zone buffer of each unit are exposed to. An area was coded as an industrial zone if the average number of neighbor exposure was one standard deviation above the mean. Low industrial zones were those that were one standard deviation below the mean and moderate industrial zones were those which fell between these two measures. 29

40 30 Table 2: Overview of Independent Variables included in Analyses (16,619 Dissemination Areas in the Province of Ontario, 2001 Canadian Census of Populations All Ontario DAs Ontario Without Toronto DAs Toronto DAs 16,619 DAs 12,870 DAs 3,748 DAs Variable Description Number Percent Number Percent Number Percent INCOME Low Median household income below poverty line (35,471) 2, % 1, % % Middle Median household income between low and high income (35, ,412) 13, % 10, % 2, % High Median household income greater than 3times poverty line (106,413) % % % PERCENTAGE RENTERS Low High UNEMPLOYMENT Low Below 75 th percentile (<43%) Above 75 th percentile (>43%) Below 75 th percentile (<8.6%) 12,460 75% 10, % 2, % 4,159 25% 2, % 1, % 12,486 75% 9, % 2, % High Above 75 th percentile (>8.6%) 4,133 25% 3, % 1, %

41 31 Table 2: Overview of Independent Variables included in Analyses (continued) All Ontario DAs Ontario Without Toronto DAs Toronto DAs 16,619 DAs 12,870 DAs 3,748 DAs Variable Description Number Percent Number Percent Number Percent EDUCATION LEVEL Low High Below25 th percentile Bach. degree (<8%) Above 25 th percentile Bach. Degree (>8) 4, % 3, % % 12, % 9, % 3, % VISIBLE MINORITY Substantially Predominately per cent visible minority per cent visible minority 2, % % 1, % 2, % 1, % 1, % Average Range Predominately White Substantially White per cent visible minority per cent visible minority per cent visible minority 1, % 1, % % 1, % 1, % % 8, % 8, % %

42 32 Table 2: Overview of Independent Variables included in Analyses (continued) All Ontario DAs Ontario Without Toronto DAs Toronto DAs 16,619 DAs 12,870 DAs 3,748 DAs Variable Description Number Percent Number Percent Description Percent FOREIGN BORN Substantially percent foreign born 2, % % 1, % Predominately percent foreign born 2, % 1, % 1, % Average Range percent foreign born 3, % 2, % % Predominately Native Born percent foreign born 2, % 2, % % Substantially Native Born percent foreign born 4, % 4, % 30.8% INDUSTRIAL ACTIVITY Low Moderate 1 standard deviation below mean (0-.089) within 1 standard deviation above and below mean ( ) 2,160 13% 2, % 6.16% 11, % 9, % 2, % Predominate 1 standard deviation above mean ( ) 2, % 1, % 1, % a Source: 2001 Canadian Census of Population, for details, see Statistics Canada 2001 (2003). b Source: Ministry of the Environment (2009).

43 Evolving Methods Unit-Hazard and spatial methods Large-scale quantitative studies documenting the spatial distribution of environmental burdens and resources have been popular for many years. Among these quantitative studies some approaches are more accurate than others. The Unit-Hazard Coincident approach, so named by Saha and Mohai (2007), was the classic approach for assessing demographic disparities in the distribution of hazards. This approach begins by selecting a geographic unit, usually zip code, county or census tracts, and then identifies those units that contain, within their geographic boundary, a hazard of interest. These are deemed host communities. The demographics of these communities are then compared with communities that do not host hazards. This approach assumes that those within host communities are burdened equally by the presence of the hazard and that they are burdened more than those communities that do not host these hazards. Neither of these assumptions hold true in all cases. As Figure 1 suggests, bordering geographic units might be more burdened than host communities and host communities rarely feel the effects of a hazard equally. This is especially true of the larger geographic units such as counties. Figure 1. Unit-Hazard Coincidence. Adapted from Saha and Mohai

44 The spatial methods advocated by Saha and Mohai (2007) are, compared to Unit-Hazard Coincidence, far more advanced and accurate. These methods begin by determining the location of environmental hazards, drawing a circle or buffer-zone of a determined distance around these hazards and aggregating the demographic characteristics of the geographic units of interest which are contained within this buffer. A unit is considered contained within this buffer if either the center point of the unit is within the buffer or 50% of the area is within the buffer. Figure 2 illustrates this method of determining an effected community. The aggregated or averaged demographic characteristics of the areas that were contained within the buffer are then compared to the demographic characteristics of aggregations not contained within the buffer. Figure 2. Spatial Methods. Adapted from Saha and Mohai Improving spatial methods Traditional spatial methods begin with a hazard of interest and then look outward from the hazard in the form of a buffer. A spatially-based method that takes community seriously would start with a community and then look outward to hazards. This process begins by taking the center of the geographic unit (which serves as an operationalization of community) and casting a buffer outward to see if any hazards of interest are within the set distance from the center of the geographic unit, shown in Figure 3. This is most fruitfully applied when the 34