493285 ASRXXX10.1177/0003122413493285American Sociological ReviewElliott and Frickel 2013 The Historical Nature of Cities: A Study of Urbanization and Hazardous Waste Accumulation American Sociological Review 78(4) 521–543 © American Sociological Association 2013 DOI: 10.1177/0003122413493285 http://asr.sagepub.com James R. Elliotta and Scott Frickelb Abstract Endemic uncertainties surrounding urban industrial waste raise important theoretical and methodological challenges for understanding the historical nature of cities. Our study advances a synthetic framework for engaging these challenges by extending theories of modern risk society and classic urban ecology to investigate the accumulation of industrial hazards over time and space. Data for our study come from a unique longitudinal dataset containing geospatial and organizational information on more than 2,800 hazardous manufacturing sites operating between 1956 and 2006 in Portland, Oregon. We pair these site data with historical data from the U.S. population census and the Oregon Department of Environmental Quality (DEQ) to examine the historical accumulation of hazardous parcels in relation to changing patterns of industrial land use, neighborhood composition, new residential development, and environmental regulation. Results indicate that historical accumulation of hazardous sites is scaling up in ways that exhibit little regard for shifting neighborhood demographics or existing regulatory policies as sites merge into larger, more contiguous industrialized areas of historically generated hazards, creating the environmental conditions of urban risk society. Keywords environment, industrial hazards, risk, urbanization U.S. manufacturers report depositing more than 3 billion pounds of hazardous waste into onsite lands every year; this figure includes only voluntary reporting from larger establishments operating in nonexempted subsectors and thus vastly understates the scope and complexity of the problem (U.S. EPA 2010). The environmental legacies of these industrial practices are particularly acute in older urban centers where, over the past century, manufacturing facilities that produce and release such wastes have left behind untold risks for contemporary residents and vexing problems for policymakers and urban planners. Endemic uncertainties associated with these historic industrial hazards raise important theoretical and methodological challenges for environmental social science, and as Harvey (1996:429) asserts, “we have as yet only scraped the surface.” One reason for the slow progress is that most research in this area continues to say little about how industrial hazards are generated and distributed through ongoing and largely endogenous processes of socio-ecological transformation (e.g., Braun 2005; Cronon 1991; Heynen, Kaika, and Swyngedouw 2006). Addressing this and a University of Oregon Washington State University b Corresponding Author: James Elliott, Department of Sociology, University of Oregon, Eugene, OR 97403-1291 E-mail: [email protected] 522 related questions requires synthesizing urban studies’ attention to the historically recursive development of cities with environmental studies’ attention to hazards that these processes produce and leave behind. The present article offers such a synthesis. We borrow and extend ideas from industrial ecology, environmental history, and geography, as well as sociological research on urbanization, risk, and environmental inequality to examine how hazardous industrial sites accumulate in cities. Conceptually, our framework begins with Beck’s (1992, 2009) theory of modern risk society. We reformulate Beck’s theory to pay greater analytic attention to industrial hazards, the actual sources of threat. We then turn to classic urban ecology’s concepts of invasion and succession to make theoretical sense of urbanization as a process of recursive socio-ecological transformation. Methodologically, we introduce a strategy for data collection and analysis that identifies a central mechanism of this transformation— industrial churning—and how it concentrates and diffuses hazardous industrial waste over time and space. To demonstrate this framework and methodology, we provide a historical analysis of Portland, Oregon, widely considered the most environmentally green city in the United States (Huber and Currie 2007). The centerpiece of this investigation is a unique longitudinal dataset containing geospatial information on more than 2,800 hazardous manufacturing sites in operation between 1956 and 2006. We pair these data with historical data from the U.S. population census and the Oregon Department of Environmental Quality (DEQ) to examine the historical accumulation of hazardous sites relative to changes in local industrial land use, neighborhood composition, residential development, and environmental regulation. Results indicate that historical accumulation of hazardous sites is substantial and occurs in ways that exhibit little regard for changing neighborhood composition and regulatory policies. These findings provide a basis for improving socio-ecological understanding of urbanization and linking it with sociological research American Sociological Review 78(4) on the making of modern risk society and its attendant urban dangers. THE HISTORICAL PROBLEM OF URBAN HAZARDOUS WASTE Industrial transformation of natural resources creates vast quantities of waste as well as wealth, particularly in cities. Elites in government and business have long framed these negative externalities as an inevitable price of material progress (Hays 1987), and in the process they have promoted an asymmetrical relationship between industries that produce hazardous wastes and urban residents who bear the environmental burdens. Over time, this unequal relationship has coalesced around an implicit urban-waste compact based on three basic principles. The first principle, evident since the late 1800s, is that municipal governments will assume primary responsibility for solid waste streams produced by local populations and related services, and industries will assume primary responsibility for hazardous waste streams they create through local production processes (Tarr 1996). This tacit agreement has given manufacturers significant control over decisions concerning how to treat and dispose of their discharges, including “a new spectrum of wastes whose quantities, toxicity, and [ecological] persistence took quantum leaps” after the 1930s (Colten and Skinner 1996:5). Still largely in effect, this historical arrangement dovetails with another long-standing principle—geographic isolation—that aims to separate residents from hazardous industrial wastes. Early on, this “out of sight, out of mind” solution to the accumulation of industrially produced hazards encouraged urban manufacturers to flush untold volumes of hazardous waste into local waterways, a practice supported by public perception that such wastes acted as bactericides that cleansed local waters of putrefying organic wastes from stockyards, slaughterhouses, and sewers (Hays 1987:76). Yet, as the volume, diversity, and toxicity of industrial wastes increased over time, public Elliott and Frickel concern about these practices also increased. By the 1950s, government regulators began to monitor and restrict release of industrial effluents into the air and water. In response, urban manufacturers began dumping more of their hazardous wastes onsite. This approach minimized growing legal liabilities associated with widespread public exposure to such hazards, safeguarded trade secrets by keeping information about new substances away from competitors, and reduced costs associated with more expensive off-site waste transit, treatment, and disposal (Colten and Skinner 1996). Indeed, a federal study found that between1950 and 1979 chemical manufacturers used landbased disposal methods for 94 percent of the waste they generated, dumping 80 percent of it at their own production facilities (U.S. Congress 1979; see also Dietrich 1981; Page 1997). Common methods for such onsite disposal include burying wastes in metal or fiber barrels, dumping them directly into open pits or lagoons, and injecting them into deep wells. These strategies reflect “an amalgam of science and engineering mixed with heavy doses of convenience and expediency” (Colten and Skinner 1996:46) that did little to ensure protection against migration of chemical wastes into local ground water and underground aquifer supplies. A third principle of the implicit urbanwaste compact is that any regulations will remain narrowly focused on specific waste streams rather than more general processes of historical accumulation. This approach reflects the fiscal realities of poorly funded municipalities and the fact that state and federal regulatory agencies consistently meet fierce, organized resistance from industry whenever they step in (Markowitz and Rosner 2002). Consequently, public efforts to regulate hazardous waste disposal did not occur consistently at the state level until the 1960s, and not until the mid-1970s at the federal level. The Comprehensive Environmental Response Compensation and Liability Act of 1980 (CERLCA, or Superfund) is a key example of federal legislation. Implemented in 1983, CERCLA targeted only the largest, most visible 523 sites of contamination for environmental remediation, created disincentives for owners to clean up contaminated property, and discouraged potential buyers from redeveloping abandoned properties (Leigh and Coffin 2000). The federal government later passed the Brownfield Redevelopment Financing Act of 1996, which responded to problems created by CERCLA by relaxing legal liabilities for redevelopment of contaminated sites. Within this policy context, local regulators have continued to rely heavily on industry’s voluntary compliance and to focus their limited resources on large, highly visible sites that raise the greatest public concern or obstacle to liability-free redevelopment, effectively ignoring unknown numbers of smaller hazardous sites that continue to accumulate. Consequently, the historical problem of hazardous industrial waste in urban areas persists, alongside new regulations that have unintentionally increased onsite, land-based disposal of hazardous industrial wastes. Indeed, a recent study shows that although emissions of hazardous wastes and other toxic releases decreased by 3 percent in North America during the late 1990s (largely through reductions in air emissions), disposal of hazardous wastes to land actually increased 25 percent (Fletcher 2003). Regulatory loopholes and exemptions have also allowed smaller-scale waste generators to skirt new treatment rules, encouraged others to dilute their hazardous wastes with nonhazardous wastes, and provided large-scale waste generating facilities incentives to cheat. How and to what extent these activities continue to occur undoubtedly varies from city to city, and decade to decade, but according to Colten (1990:154) there is little doubt “current environmental laws have not halted on-site or illegal urban releases” that continue to accrue and have negative consequences for urban residents. THEORETICAL FRAMEWORK Our theoretical framework for understanding the historical accumulation of land-based hazardous wastes begins with Beck’s (1992) 524 theory of “risk society,” which has emerged as a dominant perspective on the social production of toxic uncertainty (Auyero and Swistun 2008; Cable, Shriver, and Mix 2008) and a source of broad theoretical debate on modernization generally (e.g., Beck, Giddens, and Lash 1994; Rudel, Roberts, and Carmin 2011).1 Beck’s wide-ranging observations about the increasingly reflexive nature of risk make several claims pertinent to the present study. The first is that today’s risks differ from and compound those of the past because they emerge through large-scale technoscientific processes that result from conscious decisions by industrialists, engineers, and policymakers to downplay risks relative to economic growth and the broader promise of progress. Unlike early-modern risks associated primarily with material scarcity: “Today’s risks aren’t fated; we created them. They are human products (created at the intersection of technical knowledge and economic decision making) . . . protected by the state” (Beck 2009:25). As a result, today’s societies are increasingly characterized by manufactured risk associated with such unintended industrial side effects as nuclear radiation, chemical toxins, and other hazardous wastes. These manufactured dangers signal the decline of industrial society and the coincident emergence of contemporary risk society. Beck also claims that the scale and scope of these newer threats remain largely unknown despite advances in environmental science and risk assessment, and that this ambiguity permits powerful actors to evade accountability and dismiss critics as irrational. Consequently, less visible forms of manufactured risk must be rescued from state inattention by social movements (Szasz 1994), popular works such as Silent Spring or An Inconvenient Truth, and other forms of “subpolitics” that demonstrate society’s reflexive response to risk and associated decoupling of politics from government (Beck 2009:16). Beck’s more general point is that manufactured risks have become central features of modernity; as such, the dualism of society and nature should be recast to ask how society ought to handle American Sociological Review 78(4) the unintended consequences of its intended transformation of nature. From this perspective, the accumulation of land-based hazards generated by modern urbanization signals “a radical institutional crisis” (Beck 2009:92) that calls for societal reflection and political change, in which even the present study can be interpreted as a form of Beckian subpolitics— a reflexive response to manufactured risk. Engaging these aims, however, requires more guidance in moving from general theoretical claims about risk society to empirically grounded research than Beck’s framework readily offers. This ambiguity derives in part from heavy emphasis on reflexive risks and relative inattention to material hazards. For Beck, as for others (Jaeger et al. 2001), risk remains a probabilistic statement about the likelihood of future harm. Its essence lies in the social anticipation, expectation, and potential action that this probability produces, thus treating risks and social definitions of risk as analytically equivalent. This approach permits a reflexive account of unintended consequences of modernity, but it paradoxically limits understanding of the ecological basis of risk society. This limitation arises because material conditions from which different risks emerge and spread are not well specified, despite the theory’s premise of increasing diffusion and intensity of real and present ecological threats. Consequently, Beck’s framework tells us relatively little about the recursive transformation of nature from which today’s risk society emerges. This compromise is unnecessary. We contend it is possible to improve understanding of the ecological basis of risk society by extending analytic attention from risks to hazards. By hazards we mean the material substances (e.g., industrially produced toxins) and biophysical conditions (e.g., contaminated soils) that are the actual sources of threat. From this perspective, hazards precondition risk. Moreover, because they exist in actual time and space, whereas risks characterize future conditions and possibilities, hazards are empirically measurable in ways that risks are not. Shifting attention Elliott and Frickel from risks to hazards thus provides a way to recover the material, ecological basis of risk society while at the same time offering the potential for theoretical refinement through empirical analysis. In making this analytic shift, we theorize hazards not merely as discrete outcomes of modern industrial production but as historical and spatial processes of urban-ecological change that have become increasingly, if unevenly, unbounded in time and space, compounding uncertainties about the risks they represent (Beck 1992). Focusing on these urban change processes is where recent research on environmental inequalities becomes useful. Researchers in this growing field typically conceptualize and measure hazards as spatially located outcomes—places such as factories, weapons depots, or landfills that produce or store industrial hazards. Most studies then examine spatial relationships between such places and low-income and minority residents using cross-sectional data (for a review, see Mohai, Pellow, and Roberts 2009). However, a few studies using longitudinal data show that uneven spatial distributions of industrial hazards change over time, often with paradoxical consequences. For example, in a study of the Greater Boston area, Krieg (1995; see also 1998) found that as commercial and industrial activity moved from the inner city to outer-ring suburbs, it not only left hazardous wastes in de-industrializing, minority neighborhoods, but also brought new forms of pollution to industrializing, working-class (largely white) suburban communities in the region. Similarly, Downey (2005:1000, 2007) found that in Detroit, as polluting industries shifted from the inner city to select suburbs and “separated the region’s blacks from suburban manufacturing activity,” working-class whites (not inner-city blacks) became increasingly exposed to industrial emissions from operating facilities. In the same vein, Saha and Mohai (2005) found that growing public opposition to hazardous waste production and storage in Detroit has meant siting decisions for new industrial facilities have followed the path of least political resistance into low-income and 525 minority neighborhoods, increasing environmental inequality over time. These longitudinal studies show how the industrial hazards of risk society are not only spatially dynamic but aggregative as well. This is particularly true of industrial hazards released to land. Because such hazards typically take several generations or more to biograde, the historical problem they present is not limited to the uneven diffusion of current hazardous production, such as when inner-city factories close and relocate to suburban industrial parks. Instead, the historical problem they present extends to accumulation of hazards over time and space in ways that lastingly transform the areas where these hazards accrue. Making proper sense of these cumulative hazards requires that we move beyond existing longitudinal analyses to recover a deeper ecological appreciation for how processes of urban change recursively transform the physical environs where they take (and remake) place. One avenue for such recovery runs through classic urban ecology. Nearly a century ago, sociologists began to analyze the territorial organization of human behavior by importing concepts of “succession” and “invasion” from plant and animal ecology into the emergent study of urban society (Park 1936). For plant and animal ecologists, succession implied a complex process involving recursive changes to local biomes; invasion described a simpler, more linear process of serial replacement of one biotic community by another, which over time could contribute to succession. In extending these concepts to the nascent field of urban ecology, Park left no doubt that succession was the analytically richer of the two constructs. In his classic article, “Succession, an Ecological Concept” (1936:177), he explained that “changes, when they are recurrent, so that they fall into a temporal or spatial series—particularly if the series is of such a sort that the effect of each succeeding increment of change reinforces or carries forward the effects of the preceding—constitute what is described in this paper as succession.” In this way, Hernes (1976:535) explains, Park conceptualized succession as a recursive, 526 even dialectical, process by which “groups successively and simultaneously produce conditions for their own unmaking and for the generation of other groups. Their actions are simultaneously productive and counterproductive.” Turning attention to these endogenous processes of urban change, Park (1936:178) argued that the central focus of urban ecology should be “any sort of [territorial] change, in fact, which effects an existing division of labor or the relation of the population to the soil.” Subsequent research in urban ecology discarded these directives, replacing a broader focus on recursive urban change with narrower empirical studies of serial invasion of one type of land user by another (Aldrich 1975). Parallel tendencies are evident in more recent studies charting the transformation of wilderness or agricultural lands into newly suburbanized sprawl development (e.g., Bullard, Johnson, and Torres 2000), as well as in urban political economy more generally. These studies tend to linearize, or overlook, the recursive dynamics of urban change. Take, for example, Logan and Molotch’s (1987) prominent theory of cities as growth machines, which maintains urban ecology’s focus on urban lands as a legitimate object of sociological study (see also Molotch 1976). This theory draws much needed attention to how urban development is contested by social groups with different interests, power, and government access, demonstrating how cities are simultaneously material places and political processes. However, applications of the theory rarely take into account how these dynamics play out repeatedly over time in the same urban spaces, and thus how urbanization becomes not just a contested process in the present but also a historical process that continually builds on the remains of earlier land-use contests. Such theoretical extension, building from insights of classic urban ecology, highlights the fact that land-use contests do not start anew but rather unfold (again and again) on lands transformed by the hazardous wastes of prior growth machine activities. American Sociological Review 78(4) Recovering and extending this ecological appreciation for how processes of urban change recursively transform their physical environs draws attention to how ongoing shifts in urban economies, populations, and regulatory policies cumulatively shape current land-use decisions and practices as well as urban lands as they are reused. From this perspective, industrial wastes that accumulate onsite result from interrelated endogenous forces that play out across urban space repeatedly over time. This accumulation may occur when one manufacturer occupies and pollutes a single site for a given period of time, when different manufacturers reoccupy and repollute the same site sequentially, or when pollutants migrate through soils to adjacent lots (Litt and Burke 2002). With each subprocess, the result is the same. As hazardous manufacturers arrive and deposit wastes onsite, they distribute industrial hazards laterally across urban landscapes. Where such ecological succession occurs, the effects of accumulation will tend to scale up as industrialized sites merge into larger, more contiguous industrialized areas of historically generated hazards. We theorize that these temporal and spatial dynamics literally create the environmental conditions of urban risk society, which we now understand in empirical terms as the aggregative product of historically recursive redevelopment of urban lands—lands that bear the environmental as well as social imprint of earlier land-use transformations. In this reformulation of classic urban ecology, invasion refers to changes in land users, whereas succession refers to changes in the land itself, wrought by onsite disposal and subsequent accumulation of hazardous industrial waste. In this way our theoretical approach links the urban with the ecological; serial demographic changes (invasion) with lasting environmental changes (succession); and classic urban ecology with contemporary political ecology. To demonstrate these connections, we conduct a historical case study that shows how industrial churning, residential change, and regulatory policymaking come together in a specific place to produce Elliott and Frickel an increasingly hazardous urban landscape. We selected Portland, Oregon for this case study because it shows how the production of such hazardscapes extends beyond rustbelt cities emphasized in prior research (e.g., Bowen et al. 1995; Downey 2007) to older urban areas throughout the United States, including areas with reputations for progressive environmental regulation. HISTORICAL CASE STUDY: PORTLAND, OREGON Incorporated in 1851 along the convergent banks of the Willamette and Columbia Rivers, Portland has long promoted local economic growth by dredging navigation channels, infilling marshlands, and developing extensive dock works that have increased port capacity for moving the region’s agricultural and forest products and provided space for warehouses and small-scale manufacturers along the rivers (Abbott 1983, 2008; Lang 2010). Such developments continued largely unzoned and unregulated into the late 1930s, when the U.S. Army Corps of Engineers completed the Bonneville Dam 40 miles up the Columbia River (Robbins 2004). The dam included locks for continued shipping but also provided cheap, abundant hydroelectricity to power wartime booms in shipbuilding and metals manufacturing, which ushered in the city’s most significant period of heavy industrialization. Like other western cities of this period, wartime production of ships and light metals attracted thousands of workers, increasing Portland’s population by 22 percent during the 1940s and concentrating 40,000 new workers—25 percent of whom were African American—into Vanport, the nation’s largest wartime housing project, built on the city’s north floodplain. In this community as elsewhere, redlining and covenant agreements concentrated black workers in environmentally risky housing (Vanport was destroyed by a flood in 1948) and set the stage for patterns of residential segregation that mirrored trends in larger cities (Smith 2005). These agreements also conjoined with local topography 527 and urban planning decisions to produce locally distinct forms of environmental inequality (Hovey 1998; Stroud 1999). By the 1970s, these developments had generated significant environmental effects within the city limits and encouraged suburbanization that threatened valuable agriculture and timber lands in the surrounding region (Leo 1998). Such developments were not unique to Portland (see Rudel 1989), but the political response was. Pressured by local activists and empowered by new federal and state regulations, city officials instituted a series of initiatives to clean up Portland’s rivers, redevelop the largest, most visibly hazardous sites, and enhance local quality of life through affordable housing programs, mass transit, and downtown revitalization (Abbott 2001; Hovey 1998; Leo 1998). The cornerstone of these initiatives was a state-mandated Urban Growth Boundary (UGB) and metropolitan planning board that worked to ensure suburbanization spread slowly through highdensity development that was “forged of compromises” among state and local officials, farmers, downtown developers, industrial trade associations, and a variety of local advocacy groups (Leo 1998:376; see also Jun 2004). This new planning system offered an alternative to classic growth machine coalitions (Molotch 1976) and laid the foundation for Portland’s present-day green reputation. In addition to civic participation, land-use conversion remains central to this regional regulatory system. In Portland’s suburbs, conversion of rural lands to new urban uses, including high-tech manufacturing, is regulated (Walker and Hurley 2011); and in the city’s core, the same system facilitates the reuse of already urbanized space and infrastructure by new land users. This invasion through reuse includes the transformation of old manufacturing and warehousing facilities into new office spaces, retail outlets, and residential lofts, and the maintenance of industrial sanctuaries where new manufacturers can continue to locate and expand over time (Abbott 2008). Indeed, this continued manufacturing presence—especially in metals and 528 machinery—has helped the city maintain its industrial base and solidified industry’s footprint in older neighborhoods, some of which are now growing in population (Gibson and Abbott 2002). For example, in north Portland’s heavily industrialized neighborhoods, where hazardous manufacturing remains prominent, the residential population grew by 64 percent during the past two decades, compared with only 22 percent in the city as a whole (U.S. Census 2010). In these ways, reuse of urban lands brings new generations of workers and residents into established Portland neighborhoods through recursive processes of what classic urban ecologists called invasion. This invasion contributes to collective amnesia about previous uses and lingering hazards while enabling new hazardous manufacturers to churn through old and new sites, altering urban lands through onsite disposal of hazardous waste, or what we call succession. This deeper type of land-use conversion—which changes not only who does what where but also alters the biochemical conditions of the land itself— continues to accumulate within Portland’s urban core, even as city officials use federal and state resources to identify, test, and remediate select parcels. In this way, economic, demographic, and political forces combine to generate new hazardous sites while reusing and thus obscuring older ones from view. Revitalization of Portland’s urban core has thus not resolved the cumulative ecological problem of relict industrial waste. Instead, it has distributed it across an expanding and increasingly diverse array of residential groups and economic activities that continue to churn endogenously within city limits. In fact, a recent study by Eckerd (2011) found no correlation between gentrified urban neighborhoods and environmental cleanup of known hazardous waste sites—a finding consistent with Bullard and colleagues’ (2000) more general contention that anti-sprawl policies democratize exposure to environmental hazards by keeping wealthier residents in closer proximity to them, and with Rudel, O’Neill, and colleagues’ (2011) finding along similar lines in Northern New Jersey. American Sociological Review 78(4) Thus, even for a city that remains relatively industrialized and has institutionalized significant land-use regulations, Portland’s historical development is less exceptional than its popular reputation suggests. Like Detroit, Cleveland, and other rustbelt cities, Portland has a long history of industrial pollution (Eckard 2011; Lang 2010) and patterns of residential segregation and discriminatory housing policies (Smith 2005; Stroud 1999). Also, Portland’s UGB—a feature now shared with more than 180 U.S. cities (Uri and Bayer 2003)—did not stop suburban growth but instead urbanized it while creating renewed pressure for gentrified redevelopment of the urban core, ushering new residential populations into close proximity to historically accumulating industrial hazards. Early and proactive efforts to clean up the worst of these hazards and to reclaim underutilized industrial lots are distinctive, but like other older cities facing similar problems, Portland has relied largely on federal legislation—primarily CERCLA in the 1980s and the EPA brownfields program since the 1990s—to organize and fund these limited efforts. Under such constraints, local environmental agencies focus more on risk containment than on the deeper ecological problem we identify here— the historical accumulation of industrial hazards through successive waves of soil contamination and site reuse. DATA AND METHODS To identify the dominant industries depositing hazardous wastes to onsite lands in the City of Portland, we began with the (earliest available) 1988 Toxic Release Inventory (TRI) database assembled by the U.S. Environmental Protection Agency (U.S. EPA 2010).2 County-level reports from this source indicate that more than 18 million pounds of hazardous industrial wastes have been disposed onsite in Multnomah County (City of Portland) since 1988,3 originating from four prominent sectors. In order of volume, these sectors are primary metals, chemicals, transportation equipment (mostly shipbuilding), and fabricated metals. To these four sectors Elliott and Frickel we added petroleum refining and plastics/rubber production4 that, although not listed as significant sources of toxic releases during the 1980s when TRI data were first reported, were common onsite waste disposers with a strong local presence in Portland during the post-WWII era (Abbott 2001). Prior research by Noonan and Vidich (1992) confirms that the average probability of onsite contamination from these six industrial sectors is extremely high, ranging from an estimated 83 percent among sites of primary metals production to 95 percent among sites of petroleum refining and plastics manufacturing.5 On this basis, we assume the sites of hazardous manufacturing we analyze likely contain contaminated soils, regardless of the size of the operation in question. The broader point is that, although larger establishments tend to produce more waste (and thus may pose comparatively greater risk) than smaller establishments, the deeper underlying problem is aggregative and universal. It is aggregative because contaminated soils accumulate over time if not regularly monitored and remediated; it is universal because small and large firms alike produce waste and, intentionally or not, dispose of it onsite. Next, we identified local sites where these six industries operated during 1956 to 2006 using the Directory of Oregon Manufacturers, an underutilized source of information on industrial facilities well-suited for studies such as ours (Downey 2005; Krieg 1995, 1998). Beginning with the 1956 Directory and proceeding every other year through 2006, we located every address corresponding to facilities operating in each sector. To minimize inclusion of corporate offices rather than actual manufacturing facilities, we excluded all P.O. Box addresses as well as addresses identified as headquarters or home offices. This inventory net 2,851 unique sites occupied by 2,490 different establishments that remained in local operation for more than one year. For each establishment, we collected information on firm name, number of employees, and Standard Industrial Classification 529 (SIC) codes for products manufactured at that facility. By geocoding respective addresses and using GIS software, we can track respective sites as they enter and exit operation. We merged this facility-site data with two existing data sources. First, we used historical U.S. Census data collected at the tract level between 1950 and 2000, which includes information about the age and vacancy of local housing stocks and about residents’ changing demographic composition. We examined local areas’ changing physical and social dimensions within the city for all census years by standardizing all tracts to original 1950 boundaries and using constant city limits, as designated in 1950. This means we measured observed changes and estimated correlations with hazardous site accumulation for the same spatial units over the entire 50-year study period. In almost all cases, tract boundaries remained identical over time, simplifying the process. In most remaining cases, the 1950 tract simply split into two or more tracts that we could readily re-aggregate to original boundaries by summing data from constituent tracts. However, in three anomalous cases, two 1950 tracts first had to be merged into a single super-tract to permit standardization to original boundaries over time, resulting in 58 historically observed tracts rather than the 61 tracts originally identified in 1950. Data for these standardized tracts in 1950 and 1960 come from the Bogue Data Files (Bogue 1950) available electronically through ICPSR; data for standardized tracts in 1970, 1980, 1990, and 2000 come from Geolytics’ Neighborhood Change Database (Geolytics 2000). To estimate changes within tracts during specific time periods, we used simple linear interpolation because it introduces the fewest assumptions and is highly reliable for small area estimates (Chi 2009), with the baseline being the variable’s estimated value in the first year of the respective period (e.g., 1956 for period 1). Next, we paired our site data with the Oregon DEQ dataset on known and suspected hazardous waste sites. At time of retrieval, 530 this publicly available inventory of 716 sites included all contaminated and potentially contaminated sites identified since 1988 within Portland’s city limits (Oregon DEQ 2011). Importantly for our purposes, this list is historically cumulative, containing older sites of concern as well as those of more recent vintage. It is also broadly inclusive, including suspected (but unconfirmed) hazardous waste sites and sites that have undergone CERCLA Phase I environmental assessment. Among the latter, the list includes sites that have been remediated and sites awaiting cleanup. By examining this dataset alongside our own site dataset, we can assess the likelihood of different types of industrial parcels coming under regulatory review (see Guignet and Alberini 2010). To gain a sense of the reuse of hazardous industrial sites over time, we randomly sampled 120 addresses from our own site dataset and, in 2008, conducted in-person site surveys to document their current uses and surrounding environs. At each site, we took digital photographs and used detailed satellite maps to refine field observations and maximize data quality. For sites that appeared to be in commercial operation, we consulted company websites, local business directories, and the EPA’s TRI records to determine if onsite operations were considered hazardous. Together, these data collection efforts provide one of the most historically and spatially sensitive datasets developed to date on urbanecological changes associated with industrially produced hazards. Nonetheless, some limitations remain. For example, manufacturing directories contain no information about the physical size of industrial parcels, and they do not capture retail sites such as gas stations or dry cleaners that only store—and leak—hazardous materials onsite. Overshadowing such limitations, however, is the fact that our data collection strategy surmounts three major challenges that have hampered prior research. The first challenge is to historicize sample selection to avoid simply selecting sites based on current (and generally unwanted) land uses that remain visible and widely recognized at the time of study (e.g., American Sociological Review 78(4) Litt and Burke 2002). The second challenge is to refine analyses spatially to take account of both site-level dynamics and the local areas in which these sites are embedded simultaneously and interactively. The third challenge is to balance the local specificity offered by indepth case studies against national-level studies that offer generalizable findings but are largely inattentive to local conditions and historical processes. By using statistical analysis and spatial modeling techniques to conduct our local historical analysis, we forge a middle path between large-scale quantitative studies and fine-grained, site- and neighborhoodspecific analyses, thereby blending strengths of each while offering a possible route to future comparative research across cities. For analytic purposes, we subdivided our investigation of Portland into four regulatory periods. The first runs from 1956 to 1973, a period that experienced rapid industrial expansion and limited efforts by city, state, and federal officials to regulate hazardous waste disposal. The second period runs from 1974 to 1982, beginning with the advent of Portland’s UGB policy and ending with implementation of federal CERCLA legislation targeting derelict hazardous sites. The third period runs from 1983 to 1996, beginning with CERCLA implementation and ending just before implementation of the federal Brownfield Redevelopment Financing Act, which provided cities with funds to inventory derelict industrial lands and offer economic incentives for private-sector brownfields redevelopment. The final period runs from 1997 to 2006, during which time existing regulatory policies were increasingly relaxed at the federal level and burdens of enforcement further shifted to municipalities. To assess how industrial hazards accumulated spatially over the respective time periods, we used Moran’s I statistic and spatial regression techniques. Moran’s I is the most common measure of spatial autocorrelation and allowed us to assess spatial clustering across tracts over different historical periods (Cliff and Ord 1981; Moran 1950). For regression analyses, tests indicated significant spatial autocorrelation in our outcome variables; therefore, we Elliott and Frickel 531 Table 1. Active, New, and Relict Sites of Hazardous Manufacturing, by Subperiod and Size Active Subperiod and Size a Newly Emergent and Relict Total Relict Sites New Sites Exiting Cumulative Sites Entering Operation Sites Ever in Operating Operation for for Last Operation at Start of First Time Time (Avg. by End of Subperiod (Avg. per year) per year) Period All Sites 1956 to 1973: pre-UGB 1974 to 1982: UGB to CERCLA 1983 to 1996: CERCLA to Brownfields Act 1997 to 2006: Post-Brownfields Act 298 430 449 445 51.9 53.9 33.2 66.8 36.8 47.4 37.5 44.2 1,233 1,718 2,183 2,851 Large Sites (100+ employees) 1956 to 1973: pre-UGB 1974 to 1982: UGB to CERCLA 1983 to 1996: CERCLA to Brownfields Act 1997 to 2006: Post-Brownfields Act 24 58 70 66 2.7 3.9 2.6 3.0 .8 3.2 2.9 3.4 73 108 145 175 Small Sites (<10 employees) 1956 to 1973: pre-UGB 1974 to 1982: UGB to CERCLA 1983 to 1996: CERCLA to Brownfields Act 1997 to 2006: Post-Brownfields Act 105 205 242 200 28.6 31.7 17.8 35.6 23.1 27.6 20.8 16.5 620 905 1,159 1,510 a Establishment size is measured as the mean number of employees observed during all years of operation in hazardous manufacturing. estimated models using a spatially lagged dependent variable. This variable measures and statistically controls for spatial dependence among neighboring tracts, which if left unattended can violate assumptions of independence (Anselin and Bera 1998; Voss et al. 2006). To compute this spatial lag we first used Geographic Information Systems (GIS) software to construct a queen, first-order contiguity matrix, which identifies adjacent tracts in a movement similar to that of a queen in chess, with neighboring units selected based on shared borders radiating out from the observed unit, or tract, on all sides and diagonal corners. From these neighboring tracts, we computed an average value of the dependent variable as a spatial lag. Use of this lag as an independent variable assumes that spatial dependence in the outcome variable operates as a relatively short-distance spatial process whereby proximity increases interaction and similarity among neighboring tracts. RESULTS We present our results in five sections. The first examines accumulations of active and relict hazardous industrial sites over the past 50 years; the second investigates their shifting spatial configuration over the same period. The third section adds census data to examine local correlates of newly emergent and relict industrial sites, and the fourth adds DEQ data to determine which sites are most likely to receive regulatory attention. The final section summarizes site-survey research on what relict industrial sites tend to become with reuse. Historical Accumulation Table 1 reports the number of hazardous manufacturing sites operating in the City of Portland during 1956 to 2006, by subperiod and establishment size. Three patterns stand out. The first reveals the sheer accumulation 532 of such sites over time. In 2006, 445 hazardous manufacturing sites were in operation, up from 298 in 1956. During the intervening 50 years, more than 2,800 such sites entered and exited operation, leaving untold wastes behind. The second pattern indicates that this accumulation accelerated over recent years. Indeed, Table 1 shows that from 1997 to 2006, the number of new sites entering operation averaged 66 per year—the highest rate of any observed subperiod since 1956 and double the previous subperiod, 1983 to 1996, when CERCLA was first implemented. The third pattern shows that this historic accumulation was driven largely by the opening and closing of small and medium-sized manufacturers largely exempt from reporting hazardous waste releases. These smaller facilities are especially prevalent in chemicals, transportation equipment, and metals production (primary and fabricated). These findings affirm the value of close historical research for documenting the accumulation of hazardous sites in older urban areas. They also show that this accumulation has been driven largely by the churning of small- and medium-sized establishments whose openings and closings are largely ignored by regulatory agencies charged with overseeing hazardous waste generation and disposal. This evidence also affirms prior research showing that the organizational structure of hazardous manufacturing is a key predictor of the temporal accumulation of industrial hazards (Grant, Jones, and Bergesen 2002). Spatial Concentration Table 2 reports global Moran’s I statistics for new, active, relict, and total sites by historical subperiod. To compute these spatial statistics we used 2000 Census tract boundaries as the unit of analysis because they offer the most geographically refined boundaries available and because we are not yet examining historic census data. Again, three patterns stand out. First, none of the observed processes of hazardous site accumulation—entrance, exit, and American Sociological Review 78(4) total accumulation—are spatially random during any subperiod, as evidenced by pvalues consistently below .05 (based on 999 permutations). Second, spatial clustering of current and ever-operating hazardous sites increased over time, from a Moran’s I of .33 among active sites in 1956 to .42 in 2006; and from a Moran’s I of .40 among ever-active sites in 1956 to .44 in 2006. Third, this evidence of increasing spatial concentration of current and ever-operating sites occurred despite decreasing spatial concentration among newly operating and exited sites. How can this be? Evidence from Table 2 suggests that since the mid-1950s, smaller hazardous manufacturers have opened and closed operations on a growing number of sites throughout the city. This industrial churning has concentrated hazardous sites within historic areas of accumulation but also dispersed them into other areas of the city in relatively random fashion. In this way, hazardous manufacturers have pushed into newly industrializing neighborhoods at the same time that they have continued to infill established industrial zones, much the way ethnic enclaves solidify and reproduce themselves over time even as individual members relocate to other neighborhoods (Park, McKenzie, and Burgess [1925] 1984; Zunz 2000). Figure 1 illustrates this process with snapshots of new sites that emerged during each historical subperiod. These snapshots overlay a kernel-density map, in which darker areas outline relative hotspots of new sites. A map of discrete dots indicates where new sites emerged outside these hotspots. The result is a moving picture of hazards accumulation over space and time. The most striking feature of this dynamic image is that it exhibits very little change during the post-WWII era, despite significant and concurrent economic, demographic, and regulatory changes. Early hotspots of hazardous accumulation remain hotspots in more recent periods, clustering along transportation corridors defined by floodplains of the Willamette and Columbia Rivers, just as classic theories of industrial Elliott and Frickel 533 Table 2. Moran’s I Statistics of Spatial Autocorrelation, by Subperiod Active Subperiod All Sites 1956 to 1973: pre-UGB 1974 to 1982: UGB to CERCLA 1983 to 1996: CERCLA to Brownfields Act 1997 to 2006: Post-Brownfields Act Newly Emergent and Relict Total Cumulative Sites New Sites Relict Sites Sites Ever in Operating Entering Exiting Operation at Start of Operation for Operation by End of Subperiod First Time for Last Time Period .330* .336* .377* .417* .364* .413* .397* .337* .385* .326* .413* .319* .396* .430* .448* .441* Note: Unit of analysis is Census 2000 tracts. *p < .05 (two-tailed tests); based on 999 permutations with a Queen-One matrix. ecology might predict (Harris and Ullman 1945; Hoyt 1939). Similarly, new hazardous sites continued to emerge here and there during all four periods. This spatial stability over time raises questions about the shifting social and built environments nearby. Local Correlates of Active and Relict Accumulation We then used spatially informed panel regression models to examine the extent to which changes in local census tracts correlated with two subprocesses of hazards accumulation: new site operation (which changes the land itself via onsite waste disposal) and site exit (which opens this land to subsequent reuse). For these models, we used constant 1950 tract boundaries as our unit of analysis. To control for variation in tract size and thus spatial opportunities for new site operation and exit, we computed each dependent variable as the average spatial density per year of new entries (or exits). We did this for each subperiod, yielding four successive panels per tract. To measure changes in residential and industrial composition, we focused on two sets of factors. The first highlights changes in local demographic and residential populations over time and includes changes in average family income, rate of owner occupancy, and racial composition (percent white), in addition to changes in residential vacancy and average age of residential structures. The second set of factors focuses on industrial clustering, measured as the density of hazardous sites in operation in the tract at the start of the subperiod and our spatially lagged dependent variable of concurrent change in surrounding tracts during the same subperiod. To estimate these models, we followed established practice in spatial statistics and used maximum likelihood (ML) methods (cf. Blonigen et al. 2007). We also took advantage of the panel structure of our data to predict each dependent variable two ways: once using random effects and once using fixed effects. The random-effects model predicts the dependent variable allowing unobserved, time-invariant differences between tracts to play a role (e.g., location alongside the river or a major highway). The fixed-effects model purges these time-invariant factors to focus on changes within tracts over time, controlling for fixed, or stable, differences among them. Comparing results across both models illuminates the extent to which time variant and invariant factors influenced local accumulation and potential reuse of hazardous sites. Results appear in Table 3 and reveal several patterns. First, changes in median family income, owner occupancy, racial composition, and residential vacancy show no significant correlation with entry and exit of hazardous manufacturers in the local census tract over 534 American Sociological Review 78(4) Figure 1. New Sites of Hazardous Manufacturing in the City of Portland, by Subperiod time, regardless of model specification.6 In fact, the only census-based indicator to show a significant correlation is change in average age of residential structures. This factor exhibits no correlation with site entry, but it does correlate positively with site exit. This suggests areas with more extensive preservation and reuse of older structures are more likely to expel existing hazardous manufacturers, all else equal, but not prevent them from entering new sites elsewhere in the tract, thereby shifting rather than reducing ongoing accumulation in the area. Table 3 also affirms that existing industrial hotspots, as indicated by high densities of already-operating sites and high rates of new entry in surrounding areas, significantly influence site entrance and exit within the local tract. However, this influence varies greatly by model specification. In the random-effects model, the positive and statistically significant coefficients for respective indicators imply that, when time-invariant differences among tracts (e.g., physical location) are considered, areas located within established industrial zones gain more new sites, all else Elliott and Frickel 535 Table 3. Panel Regression Models Predicting Spatial Density of New Site Entrance and Exit within Portland Census Tracts, 1956 to 2006 Density of New Sites Independent Variables Temporal and Spatial Clustering Number of operating sites at start of subperiod Spatially lagged dependent variable Historical Subperiod 1956 to 1973 (reference) 1974 to 1982 1983 to 1996 1997 to 2006 Residential Change ∆ Mean family income, logged ∆ Owner occupancy (%) ∆ White (%) ∆ Unit vacancy (%) ∆ Mean age of residential units (years) RandomEffects Model FixedEffects Model Density of Exiting Sites RandomEffects Model FixedEffects Model .072*** (.009) .274*** (.071) –.019* (.008) –.510*** (.129) .075*** (.005) .246*** (.054) .051*** (.008) .242* (.095) .049 (.091) –.323*** –.042 (.068) –.258*** –.029 (.063) –.216** (.092) –.033 (.097) .150* (.066) –.114 (.070) .195* (.075) (.068) –.176* (.073) (.066) –.116 (.072) .064 (.036) –.771 (2.487) –1.543 (5.930) 4.215 (12.007) .062 (.078) .015 (.030) –1.597 (1.696) –.771 (4.316) 7.288 (8.503) –.072 (.070) .017 (.026) –3.394 (1.837) –.456 (4.391) 15.471 (8.841) .194** –.004 (.029) –2.557 (1.577) 2.537 (4.052) 17.768* (7.939) .181** (.059) (.066) .097 (.054) .109 (.035) .305 (.017) .114 (.070) –.085 (.167) 0 .256 (.012) 231 58 –65.6 10 231 58 –13.6 67 Constant Sigma_u Sigma_e Rho .209*** (.076) .141 (.064) .410 (.025) .106 (.093) N (tract-subperiod observations) N (tracts) Log Likelihood Degrees of Freedom 231 58 –133.0 10 .823*** (.186) 0 .272 (.012) 231 58 –26.7 67 Note: All models use Maximum Likelihood estimation and exclude one outlying observation (tract 50 in 1997 to 2006). Robust standard errors are in parentheses. *p < .05; **p < .01; ***p <.001 (two-tailed tests). 536 equal, reproducing hotspots evident in Figure 1. Yet, in the fixed-effects model that statistically controls for time-invariant differences among tracts, we find the opposite pattern: coefficients for temporal and spatial clustering become negative and statistically significant. This statistical shift between models affirms that concentration and dispersion have worked simultaneously to distribute hazardous sites throughout the city over recent decades. To better understand these dynamics, it is useful to look closer at the models predicting site exit, which creates and opens relict parcels for reuse. Here respective coefficients in Table 3 indicate strong, positive spatial and temporal autocorrelation, regardless of model specification. This suggests hotspots are not due to hazardous facilities that stay in operation for long periods of time; instead, hotspots attract facilities that continually enter and exit operation over relatively short periods across an expanding array of new parcels. Over time, this industrial churning produces an increasingly uniform hazardscape that becomes less confined to a few, discrete active sites and instead expands to a mix of new and relict sites that blur into one another, leaving fewer and fewer parcels untouched by industrial development. These dynamics raise considerable challenges for regulatory agencies charged with monitoring and remediating industrial hazards. Selective Regulation Table 4 examines the likelihood that Oregon’s DEQ reviewed a hazardous site in our dataset. By the end of 2010, the DEQ reported reviewing 716 contaminated and potentially contaminated sites within the City of Portland, including 170, or 6 percent, of the 2,851 sites in our dataset (Oregon DEQ 2011). To determine which manufacturing sites are most likely to invite (or avoid) regulatory review, we estimated three logistic regression models. The first two models focus on establishmentrelated characteristics, including relative size of operations (measured as the natural log of the mean number of employees observed for American Sociological Review 78(4) all active years of hazardous manufacturing), total years in hazardous operation, last year of hazardous operation, and major sector of last hazardous operation. Model 1 conducts this analysis for all sites within the contemporary city limits; Model 2 repeats this analysis for sites within the historic 1950 city limits; and Model 3 adds tract-level covariates, which include the estimated value of each variable in 1983, when DEQ started its review process, and its change between 1983 and 2010. By including start and change scores for tractlevel variables in the same model, we can assess the relative influence of each while controlling for regression to the mean among change scores over time. Results appear in Table 4 and reveal several patterns. First, use of contemporary or historic city boundaries yields the same basic finding. Sites occupied by larger facilities for longer periods of time are significantly more likely to receive regulatory attention from DEQ, all else equal. Indeed, in our database, the average site reviewed by DEQ employed 116 workers and operated for 20 years—an anomaly among post-WWII hazardous manufacturing sites. In fact, appropriate calculations from Model 2 indicate that, all else equal, a hazardous manufacturing site that once employed 100 workers is five times more likely to receive regulatory review than an otherwise similar site that employed only 10 workers. This finding complements research showing that regulatory monitoring of hazardous industrial facilities varies significantly by organizational structure (Grant et al. 2002; Grant et al. 2010) and size (Weil 1997). Second, results show that not all targeted sectors receive similar attention despite all sectors being top generators of priority chemicals nationally (see U.S. EPA 2011). Specifically, results show that sites once occupied by chemical and petroleum manufacturers are far more likely to receive regulatory attention than are those occupied by manufacturers of plastics, fabricated metals, and industrial machinery. Indeed, appropriate calculations from Model 2 indicate that, all else equal, an active or relict site of plastics manufacturing is about eight times less likely to receive review than an Elliott and Frickel 537 Table 4. Logistic Regression Models Predicting Log-Odds of Site Review by the Local Department of Environmental Quality Independent Variables Contemporary City Limits Hazardous Site Characteristics Size (mean number of employees during hazardous operation, logged) Duration (number of years in hazardous operation) Last year of hazardous operation (e.g., 1994) Last sector of hazardous operation Chemicals (reference) Petroleum Plastics and rubber Primary metals Fabricated metals Industrial machinery Transportation equipment Tract Characteristics at Start of CERCLA, 1983 Total site-years of hazardous industry 1956 to 1983, logged Historic City Limits a .474*** (.064) .046*** (.006) .008 (.007) .436*** (.092) .043*** (.008) –.002 (.010) .411*** (.091) .042*** (.008) .001 (.010) .465 (.374) –1.896*** (.407) –.419 (.325) –1.196*** (.246) –1.407*** (.243) –.876* (.397) .329 (.376) –1.716** (.588) –.451 (.392) –1.255** (.368) –.696*** (.375) –1.026 (.644) .326 (.381) –1.602** (.601) –.511 (.392) –1.218** (.381) –1.702*** (.392) –.964 (.670) .354* (.181) –.899 (.765) 1.680 (1.211) –.862 (1.279) 6.793 (9.049) .029 (.015) Mean family income, logged Owner occupancy (%) White (%) Unit vacancy (%) Mean age of residential units (years) ∆ in Tract Characteristics, 1983 to 2008 ∆ Site-years of hazardous industry, logged ∆ Mean family income, logged ∆ Owner occupancy ∆ White (%) ∆ Unit vacancy ∆ Mean age of residential units Constant N (all sites, past and present) N (tracts)a Log Likelihood Degrees of Freedom a –20.734 (14.188) 2,831 220 –501.8 9 .473 (19.008) 1,423 59 –261.9 9 –.327 (.440) –1.076 (.313) 2.643 (1.625) –.865 (1.497) 10.923** (4.242) .027 (.015) 2.326 (20.935) 1,417 58 –256.0 21 Historic city limits incorporate less physical space than contemporary city limits and tend to have spatially larger census tracts. Robust standard errors are in parentheses. *p < .05; **p < .01; ***p <.001 (two-tailed test). 538 otherwise similar site currently or formerly occupied by petroleum-related manufacturing. Third, Model 3 reveals two significant tract-level patterns. First, sites located in historic areas of hazardous industry—as indicated by a strong presence of hazardous manufacturers prior to 1983—are much more likely to receive DEQ attention than are similar sites located elsewhere in the city. The same is true for sites in areas with declining rates of residential occupancy, which is consistent with more general regulatory efforts to target areas with slumping housing markets in order to improve exchange values in historic manufacturing zones. Cumulatively, these findings indicate that even with some of the most aggressive regulatory policies in the nation, Oregon’s DEQ remains disproportionately focused on large chemical and petroleum facilities located in traditional industrial zones where housing demand has been historically low. Despite municipal brownfield inventories showing average lot sizes of “well below one acre” (Wernstedt et al. 2004:8; see also Miller et al. 2000a, 2000b), environmental regulation of potentially hazardous areas remains narrowly focused on large, visible sites—perhaps because they are deemed worst cases or because they represent attractive opportunities for large-scale redevelopment. Either way, such selective regulation systematically ignores thousands of relict sites once occupied by small and medium-sized manufacturers located inside and outside historic industrial zones (see Hawkins 1984). Site Reuse How have recursive reuse, or invasion, affected the hazardous sites identified in this study? One example is the 12-acre parcel formerly occupied by Allied Plating. Consistent with historical patterns, between 1957 and 1969, Allied Plating dumped wastes containing chromium, nickel, copper, lead, and cyanide into the Columbia River floodplain. After 1969, the company began dumping these wastes onsite into surface impoundments that, in 1987, were reviewed American Sociological Review 78(4) by DEQ and found hazardous to humans. The site was subsequently remediated by excavating 900 tons of contaminated soil and sludge and treating and disposing of it in an off-site landfill, using federal monies. The reused site now hosts a mix of fast food and retail establishments in addition to small manufacturers and storage lots. Yet, as our analyses indicate, most sites have received far less attention and federal assistance. Visits to 120 randomly selected sites from our dataset revealed that only 17 percent are still occupied by, and thus visibly associated with, hazardous manufacturing. By comparison, 21 percent are now occupied by private residences and public uses that include parks, government offices, and parking lots. The majority of sites (62 percent) converted to nonhazardous commercial enterprises, including offices, restaurants, and other retail outlets. Indeed, none of the sites we surveyed had become vacant lots indicative of brownfields awaiting redevelopment. Instead, all but one were actively absorbed and reused without the benefit of regulatory agency review or environmental assessment.7 CONCLUSIONS This study draws attention to the historical nature of cities associated with ongoing accumulation of hazardous wastes that are changing the ecological basis of urban society. To understand this transformation, we returned to classic urban ecology to conceptualize urbanization as an ongoing process of local land conversion (succession) and subsequent reuse (invasion). We then disaggregated this process into three recurrent subprocesses to illuminate how they intersect over time and space in ways that scale up to transform not just specific sites but entire urban areas ecologically and socially. The most prominent of these subprocesses is the churning of hazardous manufacturers across a growing number of urban parcels over time. This churning is a form of ecological succession that changes the biochemical nature of urban lands through onsite disposal of hazardous waste. A long-term historical Elliott and Frickel dynamic, such churning accelerated in Portland over recent decades as the number of small producers entering and exiting operation increased, expanding the cumulative land area they transformed inside and outside historic industrial zones. This ecological succession, in turn, intersects with two related subprocesses that reinforced its operation. The first subprocess is the invasion, or reuse, of industrial lands through ongoing and locally variable processes of commercial (re)development, demographic change, and historical preservation. Through such processes new arrivals not only inherit preexisting soil conditions but also unintentionally feed social and political conditions for ongoing hazards accumulation. These conditions emerge because newcomers typically know relatively little about prior land users (and what they deposited onsite), contributing to a collective amnesia that obscures hazardous legacies of past industrial practices while making room for new, smaller establishments that can more easily accommodate shifting local land uses. Findings from Portland indicate that such recursive reuse of industrial lands exposes not only marginalized renters and working-class homeowners to these lands’ buried hazards, but also urban dwellers from a wide array of sociodemographic backgrounds, including middle-class residents drawn to the urban core by historic buildings, architectural detail, and pedestrian friendly neighborhoods (Zukin 1987). The other relevant subprocess of urbanization is the selective regulation of industrial lands, which shapes their subsequent reuse and thus the broader dynamics of hazard accumulation. Historically, such regulation relied exclusively on polluters’ voluntary compliance, but since the 1980s state and federal regulators have begun to play a more active role. Yet, findings in Portland indicate that a selective—if well-meaning—focus on large, highly visible sites continues to leave thousands of smaller hazardous manufacturers and the sites they once occupied effectively unregulated. Consequently, increased public awareness and government regulation 539 of industrial wastes has not slowed their systemic and ongoing accumulation throughout older U.S. cities. A major implication of these findings is that beneath contemporaneous and highly unequal exposures to known hazardous sites there exists a much broader and more socially inclusive exposure to accumulated hazards of unknown scale and scope. For scholars, this means the manufactured risks of urbanization are more systemic, complex, and entrenched than most extant theories predict, thereby challenging currently dominant frameworks, data sources, and methodological debates over measurement specificity and drawing attention to hidden hazards. For residents, policy analysts, and social movements, our findings suggest that programs aimed at reducing urban environmental risks must go beyond targeting new sites of hazardous manufacturing or engaging in selective investigation of large, highly visible brownfield sites that threaten nearby use and exchange values. New institutional mechanisms must be designed specifically to identify less visible, perhaps forgotten, relict sites and spur the political will to investigate and remediate the dangers they collectively present. In light of these challenges, we see opportunities to extend the present study’s line of inquiry in two related directions. One direction is to deepen understanding of urban environmental change. Interested researchers can pursue this aim by incorporating data on site contamination or toxicity-weighted pollutant estimates that more directly address the relative health risks associated with working and residing in different proximities to accumulated hazards (see Sadd et al. 2011; Woodruff et al. 2009). Future research could also use public records to generate more detailed historical data on specific sites (e.g., Litt and Burke 2002). Such work will not be easy or straightforward and should avoid reductionist tendencies to simplify potential health risks to a specific type or level of chemical exposure from individual sites. Our research suggests the risks posed are more systemic; we found that an increasing range of accumulated 540 industrial hazards likely have diverse and possibly synergistic health effects in blighted and gentrifying neighborhoods alike. A second opportunity for future research is to focus on forms of subpolitics that are emerging to confront the historic accumulation of urban industrial hazards. Our study suggests that continued reliance on local struggles over new sites or particularly noxious brownfields may not only be shortsighted but may also miss broader and more endemic causes of hazards accumulation. Yet, broader trans-urban or even transnational networks organized around shared reflexive concerns about such accumulation—that is, a globally organized subpolitics of urban risk society—have yet to fully emerge (Pellow 2007). Such political responses to urban industrial wastes require further study, not simply as collective responses to known risk but as central elements of risk society’s reflexive engagement with accumulated hazards continuing to transform the historical nature of today’s cities. Funding This research benefited from funding from the National Science Foundation (Awards 0849826 and 0849823). Data used in this article will be made available to others on request beginning in 2014. Acknowledgments The authors thank Ali O. Ilhan and Mark Leymon for valuable research assistance. Greg Hooks, Paul Voss, Don Grant, and six anonymous ASR reviewers provided generous and constructive feedback on earlier versions of this article. Notes 1. Over time, researchers have developed many approaches to studying risk, variously conceptualized as objectively real, subjectively perceived, socially constructed, or socially amplified (see Slovic 2000). For purposes of this study, we find Beck’s broader institutional framework to be sociologically insightful and promising for further extension and refinement. 2. Relying on TRI data does not ensure that we capture all locally significant industries over the entire study period because economic shifts prior to 1988 could have altered the mix of relevant sectors. American Sociological Review 78(4) However, the database does provide a reliable sample of major polluting industries in Portland at the mid-point of our historical investigation. 3. Analyses exclude off-site releases and onsite fugitive, or point source, air emissions. 4. Petroleum refining falls under TRI reporting requirements because the sector is considered to manufacture fuels (e.g., gasoline), finished nonfuel products (e.g., solvents), and chemical industry feedstocks (e.g., propane); by contrast, chemical feedstocks are used to manufacture plastics. Hence petroleum refining and plastics manufacturing fall under different industrial classifications. 5. According to Leigh and Coffin (2000), researchers use Noonan and Vidich’s estimates because no comparable data exist (see also Amekudzi et al. 1998; Coffin 2003). Our review of the literature confirms this assessment. We also found that in Baltimore, at least, owners of small parcels were just as likely as owners of large parcels to seek public assistance for site remediation until regulations changed to favor larger sites, confirming that pollution has historically accrued across sites of all sizes (Guignet and Alberini 2010). 6. In addition to analyses in Table 3, we estimated supplemental models to test for interactions between all historical subperiods and census-tract measures. Results (available upon request) show no statistically significant patterns at the .05 level, indicating that conclusions drawn here are consistent across the entire 50-year period. 7. One sampled site did receive DEQ review in 2008 and had become a dog obedience school by the time of our site survey. From 1956 to 1989, the same site was occupied by a company employing 13 people (on average) that produced neon signs. According to DEQ reports, when interviewed, the former owner admitted to dumping waste oil and chemical solvents on lots adjacent to and across the street from the facility, prompting DEQ to recommend site evaluation. To date, no such evaluation has actually occurred (Oregon DEQ 2011). References Abbott, Carl. 1983. Portland: Planning, Politics, and Growth in a Twentieth-Century City. Lincoln: University of Nebraska Press. Abbott, Carl. 2001. Greater Portland: Urban Life and Landscape in the Pacific Northwest. Philadelphia: University of Pennsylvania Press. Abbott, Carl. 2008. Portland’s Working Rivers: The Heritage and Future of Portland’s Industrial Heartland. Portland, OR: Working Waterfront Coalition. Aldrich, Howard. 1975. “Ecological Succession in Racially Changing Neighborhoods: A Review of the Literature.” Urban Affairs Review 10:327–48. Amekudzi, Adjo, Paul Fishbeck, James Garrett Jr., Haris Kautsopoulos, Sue McNeil, and Mitchell Small. Elliott and Frickel 1998. “Techniques, Cases, and Issues: Computer Tools to Facilitate Brownfield Development.” Public Works Management & Policy 2:231–42. Anselin, Luc and A. Bera. 1998. “Spatial Dependence in Linear Regression Models with an Introduction to Spatial Econometrics.” Pp. 237–89 in Handbook of Applied Economic Statistics, edited by A. Ullah and D. E. Giles. New York: Marcel Dekker. Auyero, Javier and Debora Swistun. 2008. “The Social Production of Toxic Uncertainty.” American Sociological Review 73:357–79. Beck, Ulrich. 1992. Risk Society: Towards a New Modernity. London: Sage. Beck, Ulrich. 2009. World at Risk. Malden, MA: Polity. Beck, Ulrich, Anthony Giddens, and S. Lash, eds. 1994. Reflexive Modernization. Cambridge, UK: Polity Press. Blonigen, Bruce A., Ronald B. Davies, Glen R. Waddell, and Helen T. Naughton. 2007. “FDI in Space: Spatial Autoregressive Relationships in Foreign Direct Investment.” European Economic Review 51:1303– 1325. Bogue, Donald. 1950. Census Tract Data, 1950 and 1960. Elizabeth Mullen Bogue File. ICPSR02931-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2000. doi:10.3886/ ICPSR02931.v1 Bowen, William M., Mark J. Salling, Kingsley Haynes, and Ellen J. Cyran. 1995. “Toward Environmental Justice: Spatial Equity in Ohio and Cleveland.” Annals of the Association of American Geographers 85:641–63. Braun, Bruce. 2005. “Environmental Issues: Writing a More-Than-Human Urban Geography.” Progress in Human Geography 29:635–50. Bullard, Robert D., Glenn S. Johnson, and Angel O. Torres. 2000. Sprawl City: Race, Politics, and Planning in America. Washington, DC: Island Press. Cable, Sherry, Thomas E. Shriver, and Tamara L. Mix. 2009. “Risk Society and Contested Illness: The Case of Nuclear Weapons Workers.” American Sociological Review 73:380–401. Chi, Guangqing. 2009. “Can Knowledge Improve Population Forecasts at Subcounty Levels?” Demography 46:405–427. Cliff, A. D. and J. K. Ord. 1981. Spatial Processes: Models and Applications. London: Pion Limited. Coffin, Sarah L. 2003. “Closing the Brownfield Information Gap: Some Practical Methods for Identifying Brownfields.” Environmental Practice 5:34–39. Colten, Craig. 1990. “Historical Hazards: The Geography of Relict Industrial Wastes.” Professional Geographer 42:143–56. Colten, Craig E. and Peter N. Skinner. 1996. The Road to Love Canal: Managing Industrial Waste before EPA. Austin: University of Texas Press. Cronon, William. 1991. Nature’s Metropolis: Chicago and the Great West. New York: W. W. Norton. Dietrich, Gary N. 1981. “Ultimate Disposal of Hazardous Wastes.” Pp. 1–12 in Toxic and Hazardous Waste 541 Disposal, Vol. 3, edited by R. B. Pojasek. Ann Arbor, MI: Ann Arbor Science. Directory of Oregon Manufacturers. 1956 to 2006. Portland: Oregon Department of Economic Development. Downey, Liam. 2005. “The Unintended Significance of Race: Environmental Racial Inequality in Detroit.” Social Forces 83:971–1008. Downey, Liam. 2007. “U.S. Metropolitan-Variation in Environmental-Inequality Outcomes.” Urban Studies 44:953–77. Eckerd, Adam. 2011. “Cleaning Up Without Clearing Out? A Spatial Assessment of Environmental Gentrification.” Urban Affairs Review 47:31–59. Fletcher, Thomas H. 2003. From Love Canal to Environmental Justice: The Politics of Hazardous Waste on the Canada–US Border. New York: Broadview. Geolytics. 1970 to 2000. Neighborhood Change Database. East Brunswick, NJ: Geolytics, Inc. Gibson, K. and C. Abbott. 2002. “City Profile Portland, Oregon.” Cities 19:425–36. Grant, Don Sherman II, Andrew W. Jones, and Albert J. Bergesen. 2002. “Organizational Size and Pollution: The Case of the U.S. Chemical Industry.” American Sociological Review 67:389–407. Grant, Don, Mary Nell Trautner, Liam Downey, and Lisa Thiebaud. 2010. “Bringing the Polluters Back In: Environmental Inequality and the Organization of Chemical Production.” American Sociological Review 75:479–504. Guignet, Dennis and Anna Alberini. 2010. “Voluntary Cleanup Programs and Redevelopment Potential: Lessons from Baltimore, Maryland.” Cityscape 12:7–36. Harris, Chauncy D. and Edward L. Ullman. 1945. “The Nature of Cities.” Annals of the American Academy of Political and Social Science 242:7–17. Harvey, David. 1996. Justice, Nature and the Geography of Difference. Oxford, UK: Blackwell. Hawkins, Keith. 1984. Environment and Enforcement: Regulation and the Social Definition of Pollution. Oxford: Clarendon Press. Hays, Samuel P. 1987. Beauty, Health, and Permanence: Environmental Politics in the United States, 1955– 1985. New York: Cambridge University Press. Hernes, Gudmund. 1976. “Structural Change in Social Processes.” American Journal of Sociology 82:513– 47. Heynen, Nik, Maria Kaika, and Erik Swyngedouw. 2006. In the Nature of Cities: Urban Political Ecology and the Politics of Urban Metabolism. New York: Routledge. Hovey, Bradshaw. 1998. “Building the City, Structuring Change: Portland’s Implicit Utopian Project.” Utopian Studies 9:68–80. Hoyt, Homer. 1939. The Structure and Growth of Residential Neighborhoods in American Cities. Washington, DC: Federal Housing Administration. Huber, William T. and Timothy M. Curie. 2007. “The Urbanization of an Idea: Imagining Nature through Urban Growth Boundary Policy in Portland, Oregon.” Urban Geography 28:705–731. 542 Jaeger, Carlo, Ortwin Renn, Eugene A. Rosa, and Thomas Webler. 2001. Risk, Uncertainty, and Rational Action. London: Earthscan. Jun, Myung-Jin. 2004. “The Effects of Portland’s Urban Growth Boundary on Urban Development Patterns and Commuting.” Urban Studies 41:1333–48. Krieg, Eric J. 1995. “A Socio-Historical Interpretation of Toxic Waste Sites: The Case of Greater Boston.” American Journal of Economics and Sociology 54:1– 14. Krieg, Eric J. 1998. “The Two Faces of Toxic Waste: Trends in the Spread of Environmental Hazards.” Sociological Forum 13:3–20. Lang, William. 2010. “One City, Two Rivers: Columbia and Willamette Rivers in the Environmental History of Twentieth-Century Portland, Oregon.” Pp. 96–111 in Cities and Nature in the American West, edited by C. Miller. Reno and Las Vegas: University of Nevada Press. Leigh, Nancy Green and Sarah L. Coffin. 2000. “How Many Brownfields Are There? Building an Industrial Legacy Database.” Journal of Urban Technology 7:1–18. Leo, Christopher. 1998. “Regional Growth Management Regime: The Case of Portland, Oregon.” Journal of Urban Affairs 20:363–94. Litt, Jill S. and Thomas A. Burke. 2002. “Uncovering the Historic Environmental Hazards of Urban Brownfields.” Journal of Urban Health-Bulletin of the New York Academy of Medicine 79:464–81. Logan, John and Harvey Molotch. 1987. Urban Fortunes: The Political Economy of Place. Los Angeles: University of California Press. Markowitz, Gerald and David Rosner. 2002. Deceit and Denial: The Deadly Politics of Industrial Pollution. Berkeley: University of California Press. Miller, T., M. Greenberg, K. Lowrie, H. Mayer, A. Lambiase, R. Novis, D. Ioannides, S. Meideros, and A. Trovato. 2000a. Brownfields Redevelopment as a Tool for Smart Growth: Analysis of Nine New Jersey Municipalities. (Report 12 for the Office of State Planning.) New Brunswick, NJ: National Center for Neighborhood and Brownfields Redevelopment, Rutgers University. Miller, T., M. Greenberg, K. Lowrie, H. Mayer, A. Lambiase, R. Novis, D. Ioannides, S. Meideros, and A. Trovato. 2000b. Addendum: Brownfields Redevelopment as a Tool for Smart Growth: Analysis of Twelve New Jersey Municipalities. (Addendum to Report 12 for the Office of State Planning.) New Brunswick, NJ: National Center for Neighborhood and Brownfields Redevelopment, Rutgers University. Mohai, Paul, David N. Pellow, and Timmons Roberts. 2009. “Environmental Justice.” Annual Review of Environment and Resources 34:405–430. Molotch, Harvey. 1976. “The City as a Growth Machine.” American Journal of Sociology 82:309–330. Moran, Patrick A. P. 1950. “Notes on Continuous Stochastic Phenomena.” Biometrika 37:17–23. American Sociological Review 78(4) Noonan, Frank and Charles A. Vidich. 1992. “Decision Analysis for Utilizing Hazardous Waste Site Assessments in Real Estate Acquisition.” Risk Analysis 12:245–51. Oregon Department of Environmental Quality (DEQ). 2011. Environmental Cleanup Site Information. Retrieved September 27, 2011 (http://www.deq.state .or.us/lq/ecsi/ecsi.htm). Page, G. William. 1997. Contaminated Sites and Environmental Cleanup: International Approaches to Prevention, Remediation, and Reuse. San Diego, CA: Academic Press. Park, Robert E. 1936. “Succession, an Ecological Concept.” American Sociological Review 1:171–79. Park, Robert E., with R. D. McKenzie and Ernest Burgess. [1925] 1984. The City: Suggestions for the Study of Human Nature in the Urban Environment. Chicago: University of Chicago. Pellow, David N. 2007. Resisting Global Toxics: Transnational Movements for Environmental Justice. Cambridge, MA: MIT Press. Robbins, William G. 2004. Landscapes of Conflict: The Oregon Story, 1940–2000. Seattle: University of Washington Press. Rudel, Thomas K. 1989. Situations and Strategies in American Land-Use Planning. New York: Cambridge University Press. Rudel, Thomas K., Karen O’Neill, Paul Gottlieb, Melanie McDermott, and Colleen Hatfield. 2011. “From Middle to Upper Class Sprawl? Land Use Controls and Changing Patterns of Real Estate Development in Northern New Jersey.” Annals of the Association of American Geographers 101:609–624. Rudel, Thomas K., J. Timmons Roberts, and JoAnn Carmin. 2011. “Political Economy of the Environment.” Annual Review of Sociology 37:221–38. Sadd, James L., Manuel Pastor, Rachel Morello-Frosch, Justin Scoggins, and Bill Jesdale. 2011. “Playing It Safe: Assessing Cumulative Impact and Social Vulnerability through an Environmental Justice Screening Method in the South Coast Air Basin, California.” International Journal of Environmental Research and Public Health 8:1441–59. Saha, Robin and Paul Mohai. 2005. “Historical Context and Hazardous Waste Facility Siting: Understanding Temporal Patterns in Michigan.” Social Problems 52:618–48. Slovic, Paul. 2000. The Perception of Risk. London: Earthscan Publishers. Smith, Chad. 2005. “From Green to Red: The Intersection of Class and Race in Urban Environmental Inequality.” PhD dissertation, Washington State University, Pullman, Washington. Stroud, Ellen. 1999. “Troubled Waters in Ecotopia: Environmental Racism in Portland, Oregon.” Radical History Review 74:65–95. Szasz, Andy. 1994. EcoPopulism: Toxic Waste and the Movement for Environmental Justice. Minneapolis: University of Minnesota Press. Elliott and Frickel Tarr, Joel A. 1996. The Search for the Ultimate Sink: Urban Pollution in Historical Perspective. Akron, OH: University of Akron Press. Uri, Avin and Michael Bayer. 2003. “Right-Sizing Urban Growth Boundaries.” Planning February:22–26. U.S. Census Bureau. 2010. Census of Population and Housing, Summary File 1. Washington, DC: U.S. Census Bureau. U.S. Congress, House. 1979. Waste Disposal Survey: Report, 96th Congress, 1st Session, Committee Print 96-IFC 33, 1979, x, and Appendix E. U.S. Environmental Protection Agency (EPA). 2010. 2009 TRI National Analysis. Retrived December 30, 2010 (http://www.epa.gov/tri/tridata/tri09/national analysis/index.htm). U.S. Environmental Protection Agengy (EPA). 2011. Trends Report Highlights: Industrial Sector Analysis. Retrieved January 10, 2011 (http://www.epa.gov/ osw/hazard/wastemin/minimize/trend10/naics.htm). Voss, Paul R., David D. Long, Roger B. Hammer, and Samantha Friedman. 2006. “County Child Poverty Rates in the U.S.: A Spatial Regression Approach.” Population Research and Policy Review 25:369–91. Walker, Peter A. and Patrick T. Hurley. 2011. Planning Paradise: Politics and Visioning of Land Use in Oregon. Tucson: University of Arizona Press. Weil, David. 1997. “Implementing Employment Regulations: Insights on Determinants of Regulatory Performance.” Pp. 429–74 in Government Regulation of the Employment Relationship, edited by B. Kaufman. Madison, WI: Industrial Relations Research Association. Wernstedt, Kris, Lauren Heberle, Anna Alberini, and Peter Meyer. 2004. “The Brownfields Phenomenon: Much Ado about Something or the Timing of the Shrewd?” Discussion Paper 04–46. Washington, DC: Resources for the Future (http://www.rff.org/rff/ documents/rff-dp-04-46.pdf ). 543 Woodruff, Tracey J., Jennifer D. Parker, Lyndsey A. Darrow, Rémy Slama, Michelle L. Bell, Hyunok Choi, Svetlana Glinianaia, Katherine J. Hoggatt, Catherine J. Karr, Danelle T. Lobdell, and Michelle Wilhelm. 2009. “Methodological Issues in Studies of Air Pollution and Reproductive Health.” Environmental Research 1093:311–20. Zukin, Sharon. 1987. “Gentrification: Culture and Capital in the Urban Core.” Annual Review of Sociology 13:129–47. Zunz, Olivier. 2000. The Changing Face of Inequality: Urbanization, Industrial Development, and Immigrants in Detroit, 1880–1920. Chicago: University of Chicago Press. James R. Elliott is Associate Professor of Sociology at the University of Oregon. His current research focuses on urbanization, social inequality, and the environment. In addition to work on relict industrial waste, he is currently examining social and spatial dynamics of post-disaster recoveries and links between urbanization and carbon emissions at and from the local level. He is past recipient of multiple university-wide teaching awards and currently co-editor of Sociological Perspectives, the official journal of the Pacific Sociological Association. Scott Frickel is Associate Professor of Sociology and Boeing Distinguished Professor of Environmental Sociology at Washington State University. He predominantly works on environment, science, and the politics of knowledge. Current research projects include a relational sociology of interdisciplinarity, the sociology of ignorance, and the organization of expert activism. With Jim Elliott, he is developing a book manuscript on a comparative environmental sociology of cities. He is past recipient of the HackerMullins Student Paper Award, the Star-Nelkin Paper Award, and the Robert K. Merton Book Award, all from the ASA Section on Science, Knowledge, and Technology.
© Copyright 2026 Paperzz