This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 128.8.120.3

This content was downloaded on 26/01/2016 at 13:41

Please note that terms and conditions apply.

Linking ‘toxic outliers’ to environmental justice communities

View the table of contents for this issue, or go to the journal homepage for more

2016 Environ. Res. Lett. 11 015004

(http://iopscience.iop.org/1748-9326/11/1/015004)

Home Search Collections Journals About Contact us My IOPscience

Environ. Res. Lett. 11 (2016) 015004 doi:10.1088/1748-9326/11/1/015004

LETTER

Linking ‘toxic outliers’ to environmental justice communities

MaryBCollins1,2,4, IanMunoz2 and Joseph JaJa3

1 Department of Environmental Studies, StateUniversity ofNewYorkCollege of Environmental Science and Forestry, 1 ForestryDrive,106Marshall Hall, Syracuse, NY 13210,USA

2 National Socio-Environmental Synthesis Center, University ofMaryland, 1 Park Place, Suite 300Annapolis,MD21401,USA3 Institute for AdvancedComputer Studies, Department of Electrical andComputer Engineering, University ofMaryland, College Park,

MD20742,USA4 Author towhomany correspondence should be addressed.

E-mail:mbcollin@esf.edu, imunoz@sesync.org and joseph@umiacs.umd.edu

Keywords: environmental justice, disproportionality, quantitativemethods

AbstractSeveral key studies have found that a smallminority of producers, polluting at levels far exceedinggroup averages, generate themajority of overall exposure to industrial toxics. Frequently, suchpatterns go unnoticed and are understudied outside of the academic community. To our knowledge,no research to date has systematically described the scope and extent of extreme variations inindustrially based exposure estimates and sought to link inequities in harmproduced to inequities inexposure. In an analysis of all permitted industrial facilities across theUnited States, we show thatthere exists a class of hyper-polluters—theworst-of-the-worst—that disproportionately exposecommunities of color and low income populations to chemical releases. This study hopes tomovebeyond a traditional environmental justice research frame, bringing new computationalmethods andperspectives aimed at the empirical study of societal power dynamics. Ourfindings suggest thepossibility that substantial environmental gainsmay bemade through selective environmentalenforcement, rather than sweeping initiatives.

1. Introduction

As many have noted, society’s impact on the environ-ment is best characterized by its unevenness—bothwithin and between groups. For example, somesocieties pollute significantlymore than other societies(Chambers et al 2000); some groups within society usefar more resources than others (Baer 2009); and someindividuals produce much more environmental harm(Ash and Boyce 2011). Freudenburg coined the termdisproportionality to describe this pattern, defined as‘the strikingly unequal patterns of privileged access toenvironmental rights and resources’ that characterizemodern societies and economies (Freudenburg 2005,p 89). Although highly skewed patterns in pollutionintensity within industrial sectors were documented asearly as the mid-1990s (Streitwieser 1994), a clearunderstanding of why such highly reproducible andextreme unevenness exists (and persists) remainsunderstudied or simply overlooked by researchers andpractitioners. The research herein strives to lay

foundational groundwork related to the characteriza-tion and description of both highly unequal patternsof pollution production among US industries andlinks to disproportionate exposure among environ-mental justice (EJ) communities. If we are to makestrides against today’s most wicked environmentalproblems (Rittel and Webber 1973), we must knowmore aboutwhat gives rise to these observed patterns.

Disproportionality has two dimensions—dis-proportionality in the production of environmentalharm, polluter disproportionality, and disproportionalityin exposure, often discussed within the broader frame-work of EJ (Collins 2011). Of these two dimensions, thelatter has received the most attention, covered exten-sively by EJ scholars who explore the social structureslinking race and class to limited access to environmental‘goods’ and harmful exposure to environmental ‘bads’(Mohai et al 2009; see also: Boyce 1994). On the otherhand, polluter disproportionalities, or extreme varia-tions in polluter-based production, have received onlysporadic attention since the early 1990s. Although

OPEN ACCESS

RECEIVED

30May 2015

REVISED

4December 2015

ACCEPTED FOR PUBLICATION

9December 2015

PUBLISHED

26 January 2016

Original content from thisworkmay be used underthe terms of the CreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

© 2016 IOPPublishing Ltd

notable findings exist, which will be outlined later, thisbody of scholarship falls short of answering questionsrelated to why such unevenness may exist. To ourknowledge, the study herein is the first large-scale studyto show industrial polluter disproportionality patternsacross both a broad study area and a diverse group ofpolluters. Althoughwe recognize that a complete under-standing of why such patterns exist is likely dependenton awide range of factors (such as facility characteristics,environmental context, regulatory climate, etc), we findstriking evidence that extreme emitters are likelyimpacting EJ communities even more significantly thantypical EJ scholarshipmight predict.

We ask the following questions:

(I) Are producer disproportionalities present andconsistent across the study area?

(II) Are particular communities (low income and/orthose of color) disproportionately impactedby producers who generate a disproportionateamount of pollution?

Our findings suggest affirmative answers to boththese questions. Using public data and open-sourcesoftware, we assess industrially based exposure esti-mates and proximate socio-demographic character-istics on a polluter-by-polluter basis across thecontinental United States. We find a highly skeweddistribution of polluter-based harm generation withfewer than 10% of the nearly 16 000 study area facil-ities generating greater than 90% of estimatedexposure (question (I)). When describing the socio-demographic exposure profiles, we show thatalthough polluters are likely to disproportionatelyimpact poor and nonwhite communities, these dis-proportionalities become even more pronouncedwhen considering the smaller group of facilities whogenerate the majority of exposure risk (question (II)).We refer to this small group of disproportionate gen-erators as toxic outliers.

An implication from our study is that these twosides of disproportionality are connected in a ‘doubledisproportionality’ framework. This type of connec-tion has both applied and scholarly significance. First,double disproportionality would predict that indus-trial impacts overall, and in EJ communities specifi-cally, would decrease if toxic outliers could becompelled to reduce their emissions. Second, doubledisproportionality adds to our understanding of howsociety’s polluter-industrial complex works by expli-citly incorporating measurable power dynamics.Future studies should consider disaggregating pollu-ters rather than looking at polluters in the aggregate.

Our paper proceeds as follows: first, we brieflyreview the young, evolving body of work related topolluter disproportionality and make links to themore established body of EJ scholarship. Followingthis, we share methodological details, present analysis

results, and provide a brief discussion to contextualizeour findings both in terms of the existing science andwith respect to potential new research directions.

2. Polluter inequalities and EJ: a doubledisproportionality?

A disproportionality perspective challenges severalfundamental assumptions regarding the nature andmeaning of environmental harm production whilequestioning prevailing theoretical perspectives abouthuman-environmental relationships in a postindus-trial world. For example, in many social sciencedisciplines, it is common to assume that environmen-tal harm is proportional to economic well-being. Thisproportionate assumption is commonly representedin the formula for analyzing the effects of humanactivities on the environment: I=PAT, where ‘I’represents impacts, ‘P’ is population, ‘A’ is affluence,and ‘T’ is technology. IPAT emerged out of the Ehrlichand Holdren (1971)/Commoner (1972) debate in the1970s as a way to define the anthropogenic forces thatdrive environmental impacts. There have been cri-tiques of this formula, most recently in terms of thepossible benefits of technological advancement, butfew find problems with the assumption that all growthin population or affluence is problematic. Since thistime, there have been surprisingly few focused effortsto test this proposition (Freudenburg 2005). Theremainder of this section presents initial evidentiarysupport showing that, rather than problematizingincreases in population and affluence in the aggregate,it may be more accurate to view anthropogenicimpacts as being driven by just a few privileged actors.

More specifically, the disproportionality perspec-tive centers on the significance and consequences of‘the socially structured and strikingly disproportionatepatterns that characterize human access to the biophy-sical environment, both in terms of benefitingfrom ‘goods’ (resources and rights) and in terms ofavoiding ‘bads’ (wastes and responsibilities)’ (Freu-denburg 2005, p 90). Such inequalities, or ‘privilegedaccess’ (Freudenburg 2005, p 90) to the environmentalcommons, are often overlooked and unchallengedbecause many people assume that the harm is eco-nomically necessary for jobs, incomes, or the produc-tion of essential products. In fact, there is littleevidence to support a positive association betweendegree of environmental harm and economic good.

To test this empirically within societies andacross/within economic sectors, Freudenburg (2005,p 93) looked at the ‘differential access to the assim-ilative capacity of the biophysical environment’. Hefound that approximately 60%of all toxic emissions inthe United States resulted from the chemicals and pri-mary metals sectors, which together contributed lessthan 5% of the gross national product and only 1.4%of the nation’s jobs. Moving to intrasectoral analyses,

2

Environ. Res. Lett. 11 (2016) 015004

he also found striking disproportionalities, with themajor polluters being poor economic performers bothin terms of overall emissions/jobs ratios and in com-parison to peer companies in the same sector. At aneven finer level of analysis—facilities under commoncorporate ownership—overall levels of pollution weredriven by individual facilities that were emitting farmore than their ‘share’ of toxins. The revealed patternsof disproportionality hold or become more extremewhen one considers related factors of interest, such asthe relative toxicity of releases, the relationship toincome equality within nations, technological impera-tives, and other economic controls. As Freudenburgpoints out, early studies by others have shown thatarguments that attempt to explain or justify observeddisproportionalities are unsupported: such poorenvironmental performance does not seem to be dueto facilities being engaged in the production of ‘cri-tical’materials; more stringent regulation of the primepolluters would not tend to cause them to go out ofbusiness or to shift their operations to countries withweaker standards; and would not bring economicruin to the larger society. In a broader analysis ofthe significance of such findings, Freudenburg notesthat there have been few attempts by social scientiststo understand the social construction of environ-mental privileges as opposed to environmental pro-blems, or to analyze the societal and communicativemechanisms that support the perpetuation of relatedinequities.

Empirical work on disproportionality was takenone step further by Berry (2007) when she was able toshow that many policies do not recognize the fact thata few outliers tend to be responsible for a large fractionof environmental harms. This means that, at least insome cases, the disproportionality perspective holdsand implies that economic-environmental relation-ships will approximate a log-normal distribution(rather than standard Gaussian normal distribution),making it possible to achieve dramatic improvementsin environmental quality with low economiccosts. There is a clear need for future research to exam-ine the degree of disproportionality in other areasof economic–environmental relationships, but itis equally clear that the proportionality assumptioncan no longer be considered empirically credible(Freudenburg 2006).

The disproportionality perspective provides aninteresting overlay to some of themost prominent the-ories in environmental sociology. It directly challengesan important although generally unstated assumptionwithin the theoretical perspective of ecological moder-nization, that is, the assumption that all types of earlydevelopment efforts (economic and industrial) areenvironmentally harmful and that this initial harmthen decreases as society becomes modern and tech-nologically savvy. The disproportionality perspectivechallenges the universality of this idea by positing that

it is not that all development is environmentally harm-ful, but rather that a select few actors drive the major-ity of the environmental harm. Further, in regard tothe treadmill of production (Schnaiberg 1980)—another prominent theory—the disproportionalityperspective does not necessarily find capital invest-ment and capitalism to be a problem. Rather, it ques-tions a few capitalists rather than the system at large asdrivers of environmental degradation. As pointed outby Dunlap (2008, p 53), the disproportionality per-spective represents a much needed ‘finer-grained’approach for investigating the links between economicactivity and environmental degradation. It may alsolead to greater unification of the predominant theoriesin environmental sociology today, including the rele-vance of environmental state theory and the relation-ship of emissions levels to measures of ecologicalefficiency (see Fisher and Freudenburg 2004). Regard-ing the perpetuation of environmental inequalities,and in particular the tendency to focus on dis-proportionalities in environmental problems ratherthan environmental privileges, the social constructivistperspective also offers insight. As originally noted byTurk (1982, p 252) ‘it is likely that status-quo inequal-ities will be maintained ‘mainly by ideological power,secondarily by political and economic power, and onlyminimally and occasionally by the threat and use ofviolence’.

Empirical work along these lines is relatively rare,but certainly identifiable. Several studies have founddramatic disproportionalities in the production ofenvironmental harm both within a given pollutingsector and among the major polluters within commu-nities or regions (Bouwes et al 2001, Nowak et al 2006,Abel 2008, Berry 2007, Ash and Boyce 2011, Precheland Zheng 2012, Prechel and Touche 2014). Con-sistent with Freudenburg’s work, these studies havefound that a small minority of firms can drive overallpollution levels. Such findings also hold when con-sidering community-based health risk. For example,Bouwes et al (2001) found a significant correlationbetween minority status and cumulative risk in spe-cific spatial areas, with high risk facilities having aver-age risk scores 320 times greater than their lower riskcounterparts. In the same study, researchers foundthat the vast majority of facilities actually had very lowrisk scores, while a select few had scores up to thou-sands of times greater than a standardized mean. Astudy by Ash and Boyce (2011) was one of the first atthe intersection of EJ and corporate responsibility.They found that of the 100 worst polluters, the top tenimposed disproportionate impacts on disadvantagedcommunities. They found that minorities living incommunities surrounding these ten polluters werebearing more than half of the human health risk gen-erated in the region. In 2008, Abel conducted a casestudy in urban St. Louis, MO, showing that minorityand low-income residents live closer to industrial

3

Environ. Res. Lett. 11 (2016) 015004

polluters compared to their non-minority counter-parts, and that 20% of the region’s air pollution expo-sure risk generated over the last decade was spatiallyconcentrated among only six facilities. Nowak et al(2006) found that environmental behaviors interactedwith biophysical variables to explain variations inlevels of phosphorous loading in Lake Mendota, WI(see also Cabot andNowak 2005, Harlan et al 2008). Inrecent work, Freudenburg et al (2009)maintained thatthe damage done in Hurricane Katrina was not a ‘nat-ural disaster,’ but rather a case of a small number ofeconomic beneficiaries creating extreme environ-mental harm that left marginalized populations in theregion subject to catastrophe (see also: Bullard andWright 1987).

Explication of the disproportionality perspective isan important step toward shedding light on potentiallyoverlooked dynamics regarding how environmentaldomination of the powerless by the powerful happens.Grant et al (2002, 2010) suggest that research at theintersection of polluters and those who bear the bruntof pollution should analyze environmental actorsindividually (rather than in aggregate), and also treatthe relationship between organizations, organizationalstructures, and issues of environmental privilege in amore nuanced fashion. Herein, we look for pollutiondisproportionalities on a polluter-by-polluter basisacross the US and link such disproportionalities to EJimpacts.

3.Data andmethods

This research relies on two data sources—the USEnvironmental Protection Agency’s Risk ScreeningEnvironmental Indicators-Geographic Microdata(RSEI-GM) from2007 and theUSCensus of Populationand Households from 2000. Additional informationabout these data sources is provided below as well as abrief descriptionof themethods leading to our results.

3.1. RSEIAlthough we rely solely on the RSEI-GM, it is useful tofirst describe the EPAs aggregated version of RSEI,herein referred to as RSEI public-release.

3.1.1. RESI public-release dataExposure estimates to airborne toxics originating fromindustrial facilities across the US are generated by theUS EPA’s RSEI project. The RSEI project uses release-based data of more than 600 chemicals reportedannually by permitted industrial facilities in the US.These reports are mandated by EPA’s toxics releaseinventory (TRI) program. The TRI program tracks therelease of toxic chemicals, which may pose bothenvironmental and human health threats. As per TRIreporting requirements, facilities across industrialsectors must report how much of each chemical isreleased to the environment and/ormanaged through

recycling, energy recovery and treatment on an annualbasis5. Although a touted resource in itself, TRIinformation does have some shortcomings; of mostrelevance, TRI reports do not account for differencesin chemical toxicity (pound-for-pound, chemicals canvary widely) thereby making between chemical com-parisons difficult or simply impossible.

EPAs RSEI project addresses TRI-based chemical-to-chemical comparison limitations. On a release-by-release basis, RSEI estimates exposure, accounting forenvironmental fate and transport and chemical toxi-city. More specifically, for each chemical release, aGaussian-plume fate-and-transport model isemployed, which estimates how the chemical spreadsfrom its point of release to the surrounding geography.Geography, in the RSEI case, is delineated into a net-work of non-overlapping grid cells of 810 m2 in size.This network of grid cells extends over the entirety ofthe continental US. The modeling results in exposureestimates for each reported release for each grid cell.To produce such exposure estimates, EPA combinesdata on temperature and local wind patterns withfacility-specific information (stack heights, exit velo-cities) and chemical specifics (molecular weights,decomposition rates). Estimates represent an ambientexposure concentration. The RSEI project then over-lays the grid of toxicity-weighted exposure concentra-tions on population figures from the US Census.RSEI’s population weights account for age and sex,since the volume of air inhaled per unit body weight isknown to depend on these characteristics. Since themain purpose of the RSEI program is to aid in theprioritization of facilities for enforcement, only facil-ity-based estimates are publically released (publicrelease ‘RSEI scores,’ are a measure of human healthhazard aggregated over every release-grid cell impac-ted by each industrial facility). RSEI public-releasedata does not include information on a cell-by-cellbasis. EPA does provide such cell-by-cell estimates tothe research community by request.

3.1.2. Risk screening environmental indicators–geographicmicrodataOur measure of exposure relies on the RSEI-GMwhich provide disaggregated industrial toxicity expo-sure estimates (unlike RSEI public-release). The RSEI-GM raw data provide an exposure estimate on arelease-grid cell basis. This means that for everychemical release reported by every facility in a givenyear (nearly one billion releases in 2007, our study

5TRI does not include all facilities, rather facilities must meet all of

the following criteria: (1) the facility must be classified into a ‘TRI-covered’ North American Classification System (NAICS) code; (2)the facility must have ten ormore full-time employees (or employeeequivalents); and (3) the facility must manufacture, process, or use achemical mentioned by section 313 in quantities greater than thecurrently established threshold. section 313 chemicals are thosethought to cause cancer, chronic human health effects, significantnon-cancer adverse acute human effects, or significant adverseenvironmental effects.

4

Environ. Res. Lett. 11 (2016) 015004

year), RSEI-GM provides an exposure estimate for thenetwork of 810 m2 grid cells extending over thecontinental US. Exposure estimates account for envir-onmental conditions, facility characteristics and che-mical specifics, per the RSEI program modelmentioned in the previous section. Since our analysisis not chemical specific, we aggregate across allchemicals up to the 810 m2 grid cell-level resulting inan exposure estimate for each 810 m2 grid cell. Thefollowing formula represents how we accomplishedthis aggregation. We will use ToxicConcf,c,g to denotethe exposure estimate of chemical, c, generated byfacility, f, within grid cell, g. By adding over all facilitiesand all chemicals, we end up with a grid cell-basedexposure estimate

ToxicConc ToxicConc ,f c ggf c

, ,åå=

ToxicConc ToxicConc .f c gfg c

, ,åå=

In addition to aggregation at the grid cell-level,data can also be aggregated to the facility-level(ToxicConcf). This type of aggregation was used toassess producer disproportionalities, prior to theef-forts made to link extreme polluters to EJ demo-graphics (research question (I)).

Unlike the aggregated RSEI public-release data, wedo not employ the age and sex weighting and instead,link to race and class demographic characteristicswhen answering research question (II).

3.2. United States 2000 census of population andhouseholdsSocio-demographic information to support evalua-tion of question (II) comes from the US Census ofpopulation and households. The US constitutionmandates that a census of population and householdsbe conducted every 10 years. Besides its use indetermining the number of seats per state in the USHouse ofRepresentatives, census data are an extremelyimportant source of information for populationscientists. The census data are available at variousscales, from statistics at a large national scale, to thoseat the state and county levels; to those at extremely finegeographic areas such as blocks (approximately onecity-block) or block-groups (corresponding approxi-mately to neighborhoods). For our purposes, we usepopulation demographics available at the block (racevariable) and block-group (income variable) levels.The study year is 2000.

3.3. Study area stratified random sampling: linkingpollution to demographicsIn order to link RSEI-GM-based exposure estimatesand local socio-demographic characteristics (researchquestion (II)) we generated a large sample (over fourmillion points) across the study area. Our sample isstratified by population density at the US county-level.Our sample reaches every county in the United States

and includes chemical releases from every facility. Foreach of our 4, 172, 835 sample points, we note thefollowing characteristics: (1) the exposure estimate ofeach release impacting the grid cell in which the pointis located; (2) the number of facilities that contributereleases to the grid cell in which the point is located (3)the total population in the census block in which thepoint is located; (4) the total number of people whoreport their race as ‘white alone’ in the census block inwhich the point is located; (5) the total number ofhouseholds in the census block group in which thepoint is located; and (6) the total household income ofall households in the census block group in which thepoint is located. A schematic presenting a visualrepresentation of this procedure is shown in figure 1.The mean income associated with sample points is$64 581 and mean proportion reported ‘white alone’as their race is 82.5%. These figures are comparable toUS Census figures, indicating that our sample can beconsidered representative of theUS population.

3.4. Computing strategyThese computations are carried out on more than100 GB of exposure estimate data (RSEI-GM) andpopulation data (US Census). Conducting and visua-lizing our analyses proved to be a highly demandingtask, both in terms of processing power and datamanagement. Although other national scope studiesexist (Bullard et al 2007, Clark et al 2014), leveragingthe RSEI-GM data at this scale requires some strategiccomputing efforts. Our main platform is an open-source Ubuntu Linux environment. Our particularmachine had 24 GB RAM and 16 CPUs runningUbuntu 14.04 Trusty. To perform analyses, we used ahighly indexed PostgreSQL database with a PostGISspatial extension. As a point of reference, our analysisincludes over 1billion chemical releases spanning 8,080, 464 810 m2 grid cells originating from 15 758facilities. Using RStudio server, running R version 3.02and several helpful packages, we used explicitly parallelalgorithms to aggregate and compile sample statistics.

4. Results and discussion

As stated earlier, our central aim is to document theextent to which industrial hyper-polluters, or toxicoutliers, exist nationwide and whether they havedisproportionate effects on EJ communities. Towardthis end, we systematically evaluate the proportionalcontribution of all facilities included in our study(research question (I)) and the relationship betweenpollution generation extremes and local EJ variables(research question (II)).

4.1. Finding the toxic outliersResults provide strong support for research question(I): that toxic outliers exist. We examine the distribu-tion of facility-based exposure estimates to determine

5

Environ. Res. Lett. 11 (2016) 015004

Figure 1. Samplingmethodology schematic.

Figure 2.Disproportionality evaluation.

Figure 3. Stratified sample statistics: comparing release intensity groups regarding disproportionate demographic impacts.

6

Environ. Res. Lett. 11 (2016) 015004

the degree of variation and disproportionality. Acommon way to measure this is with the Ginicoefficient. A Gini coefficient of 0.96, our finding,indicates extreme distributional unevenness—provid-ing strong support for our first research question. Inpoint of fact, 90%of toxic concentration present in thestudy area is generated by only 809 (about 5%) offacilities. Results are presented in figure 2. These 809facilities spanmany sectors, which provides additionalsupport for earlier intrasectoral studies (Freuden-burg 2005). They are geographically distributed acrossthe study area.

4.2. Sampling: linking disproportionate exposure toEJ demographyTo link demographic characteristics and exposureestimates, we sampled across the study area, notingexposure estimates and population information atevery point. Figure 3 shows three probability densityfunction (PDF) plots. The left most plot includes allsample points, the middle plot includes only pointsthat account for themost severe 25%of those sampled,and the third includes only points that account for themost severe 10% of those sampled. Within each of theintensity groups, we present a race-income PDF. Ineach case, as we isolate the points with the highestexposure estimates, we find a greater density of lowincome households and nonwhite populations. MostEJ studies do not disaggregate in this way, therebymasking some of themost significant effects.

We also include a summary plot (figure 4) illus-trating how demographic impacts change when

you consider a small group of the most toxic pollutersas compared to the entire group of polluters. Inthis plot, all facilities are ranked fromhighest log expo-sure estimate to lowest log exposure estimate and weassess the proportion of sample points impactingcommunities with below average income and fewerthan average reporting ‘white alone’ as their race.To restate, as more facilities are considered, belowaverage impacts decrease, meaning that the smallgroup of the most toxic facilities are located in placeswhere residents tend to be lower income and people ofcolor.

5. Conclusion

We are interested in drawing attention to polluterdisproportionality in general and in exploring how itmight relate to EJ measures. These two sides ofenvironmental inequality are rarely drawn together inintegrated research designs, yet we find evidence oftheir connection. We contend that an explication ofthese relationships is important in gaining a deeperunderstanding of the theoretical underpinnings ofenvironmental injustice and in identifying potentialsolutions.

Although some researchers recognize the relation-ship between socially structured factors (i.e. privilege,power, etc) (see: Grant et al 2002, 2010) andfacility-based variations in industrial pollution(see: Streitwieser 1994, Grant et al 2002, Prechel andZheng 2012, Prechel and Touche 2014), meaningful

Figure 4. Stratified sample statistics: double disproportionality summary.

7

Environ. Res. Lett. 11 (2016) 015004

operationalization of such concepts in a broad quanti-tative framework is often difficult or outside of theresearch scope. Herein, we strive to link such sociallystructured factors to pollution production dis-proportionalities, suggesting that environmental pri-vilege is inherently intertwined with environmentaldisadvantage. For example, our results support thepossibility of Lerner-style sacrifice zones—or highmin-ority, low income neighborhoods where toxic outlierscan exist without the focus they might receive in otherlocations (Lerner 2010). If so, the ability of such facil-ities to impose health risk on populations with theleast capacity to resist would contribute to the persis-tence of these patterns and, potentially, help informour understanding of the societal machinery that sup-ports such unequal patterns.

Current EJ studies have been criticized as havingpotentially weak theoretical underpinnings (Pel-low 2000).Work is needed to reinforce this base and togive the EJ body of knowledge greater sophisticationand policy relevance. Other fruitful EJ research areasinclude stronger social class analyses; better integra-tion with social movement theory, environmentalsociology, history and ethnic studies; research thattakes advantage of alliances between community acti-vists and scholars; and studies focused on identifyingsolutions rather than quantifying problems (Pellowand Brulle 2005, Brulle and Pellow 2006). Sze andLondon (2008, p 1344) contend that research is nee-ded that ‘weaves together multi-leveled, multi-scalar,and multi-method analyses of historical, spatial, poli-tical, economic, and ecological factors’ and cite thedisproportionality perspective as one such approach.

At the intersection of power, pollution, and envir-onmental policy, Boyce (1994) puts forth two hypoth-eses that seem especially applicable: (1) thatenvironmental degradation depends on the balance ofpower where winners derive benefits and losers bearnet costs; and (2) that all else equal, greater inequalityin power and wealth leads to more environmentaldegradation. His analyses focus on which groups winand which lose, and why the winners are able toimpose the consequences of their activities on thelosers. The losers, in this context, could be future gen-erations or populations that are unaware of thedamage that they are absorbing, but Boyce is mostinterested in a third category of loser: those withoutsufficient power to prevent the winners from impos-ing the costs on them.He hypothesizes that in societieswith powerful winners and powerless losers, moreenvironmental degradation will occur because thewinners are likely to be unconcernedwith the effects oftheir actions on the losers.

More recent studies embody Boyce-style societalpower dynamics, linking inequity to issues of environ-mental quality. For example, in a historical analysis ofthe relationship between environmental policy andsocietal power in the US, Prechel (2012) found that

political mobilization of powerful organizations servesto weaken existing policy (and presumably leads topoorer environmental quality) and is inexorably inter-twined with economic agendas. In a synthesis, Cush-ing et al (2015) reviews existing evidence related tohealth disparities, concluding that social inequality inthe US is not only bad for the environment, but alsomay contribute to what, in some cases, are surprisinglypoor population health outcomes.

While the global implications of environmentalinequalities have been largely beyond the scope of ourwork, such issues obviously represent paramount con-cerns for the future. In relation to global climatechange and environmental politics, issues of inequal-ity in environmental and economic impacts are at theforefront. O’Brien and Leichenko (2000, p 221) havepointed to the dual factors of climate change and eco-nomic globalization as leading to ‘winners’ and‘losers’. In a reflection of the disproportionality per-spective, Robbins (1996) and others have emphasizedthe role of the transnational corporations as high emit-ters of greenhouse gases and as power brokers in globalenvironmental governance. At the same time, studieson the imposition of disproportionate environmentalburdens on marginalized populations are increasinglyfocusing on areas outside the US, such as in Asia,Africa, and Latin America. Such studies need toinclude not only the transnational toxics trade (seealso: Pellow 2007) but the disproportionate effects ofweather related ‘unnatural disasters’ (Freudenburget al 2009).

In an age where large amounts of environment dataare available, scholars are increasingly able to gainmoreinsight into how the winners win and who the losersreally are. It is our hope that a synthesis of availableknowledge coupled with innovativemethodologies willguide us toward a deeper understanding and potentialsolutions for the people underneath the statistics.

Acknowledgments

The authors would like to thank ERL reviewers for thethoughtful and careful review of this manuscript. Thiswork benefited by support from the University ofMaryland and National Science Foundation awardnumber DBI-1052875 to the National Socio-Environ-mental Synthesis Center (SESYNC).

References

Abel TD 2008 Skewed riskscapes and environmental injustice: acase study ofmetropolitan St. LouisEnviron.Manage. 42232–48

AshMandBoyce J K 2011Measuring corporate environmentaljustice performanceCorporate Social Responsibility andEnvironmentalManagement 18 61–79

Baer P 2009 Equity in climate–economy scenarios: the importanceof subnational income distribution Environ. Res. Lett. 4015007

8

Environ. Res. Lett. 11 (2016) 015004

Berry LM2007 Inequality in the creation of environmental harm:looking for answers fromwithin Equity and the Environment(Research in Social Problems and Public Policy vol 15) edRCWilkinson andWRFreudenburg (Bingley: EmeraldGroup) pp 239–65

BouwesNW,Hassur SMand ShapiroMD2001 Empowermentthrough risk-related information: EPA’s risk screeningenvironmental indicators projectPERI Working Papers 2

Boyce J K 1994 Inequality as a cause of environmental degradationEcological Econ. 11 169–78

Brulle R J and PellowDN2006 Environmental justice: humanhealth and environmental inequalitiesAnnu. Rev. PublicHealth 27 103–24

Bullard RD,Mohai P, Saha R andWright B 2007ToxicWastes andRace at Twenty: 1987–2007 UnitedChurch of Christ Report

Bullard RDandWright B 1987 Environmentalism and the politicsof equity: emergent trends in the black communityMid-Am.Rev. Soc. 12 21–38

Cabot P E andNowak P 2005 Planned versus actual outcomes as aresult of animal feeding operation decisions formanagingphosphorus J. Environ.Qual. 34 761

ChambersN, SimmonsC andWackernagelM2000 SharingNature’s Interest: Ecological Footprints as an Indicator ofSustainability (NewYork: Earthscan)

Clark L P,Millet DB andMarshall J D 2014National patterns inenvironmental injustice and inequality: outdoorNO2 airpollution in theUnited States PLOSONE 9 e94431

CollinsMB 2011Risk-based targeting: identifyingdisproportionalities in the sources and effects of industrialpollutionAm. J. Public Health 101 231–7

Commoner B 1972The environmental cost of economic growthChemistry in Britain 8 52

Cushing L,Morello-FroschR,WanderMandPastorM2015Thehaves, the have-nots, and the health of everyone: therelationship between social inequality and environmentalqualityAnnu. Rev. Public Health 36 193–209

Dunlap RE 2008Current trends inAmerican environmentalsociology Flows andNetworks of Environmental Sociology edTPKarjalainen et al (Oulu, FI: Thule Institute, University ofOulu) pp 49–68

Ehrlich PR andHoldren J P 1971 Impact of population growthScience 171 1212–7

FisherD and FreudenburgWR2004 Postindustrialization andenvironmental quality: an empirical analysis of theenvironmental state Soc. Forces 83 157–88

FreudenburgWR2005 Privileged access, privileged accounts:toward a socially structured theory of resources anddiscourses Soc. Forces 84 89–114

FreudenburgWR2006 Environmental degradation,disproportionality, and the double diversion: reachingout, reaching ahead, and reaching beyondRural Sociol. 713–32

FreudenburgWR,Gramling RB, Laska S and EriksonK2009Catastrophe in theMaking: the Engineering of Katrina and theDisasters of Tomorrow (Washington,DC: Island Press)

GrantD, TrautnerMN,Downey L andThiebaud L 2010 Bringingthe polluters back in: environmental inequality and theorganization of chemical productionAm. Sociological Rev. 75479–504

GrantD S, BergesenA J and Jones AW2002Organizational size andpollution: the case of theU S chemical industryAm.Sociological Rev. 67 389–407

Harlan S L, Brazel A J, Jenerette GD, JonesN S, Larsen L,Prashad L and StefanovWL2008 In the shade of affluence:the inequitable distribution of the urban heat islandResearchin Social Problems and Public Policy 15 173–202

Lerner S 2010 Sacrifice Zones: The Front Lines of Toxic ChemicalExposure in theUnited States (Cambridge,MA:MITPress)

Mohai P, PellowD andRoberts J T 2009 Environmental JusticeAnnu. Rev. Environ. Resour. 34 405–30

Nowak P, Bowen S andCabot P E 2006Disproportionality as aframework for linking social and biophysical systems Soc.Nat. Resour. 19 153–73

O’BrienKL and LeichenkoRM2000Double exposure: assessingthe impacts of climate changewithin the context of economicglobalizationGlob. Environ. Change 10 221–32

PellowDandBrulle R 2005 Power, justice, and the environment:toward critical environmental justice studies Power, Justice,and the Environment: ACritical Appraisal of theEnvironmental JusticeMovement (Cambridge,MA:MITPress) pp 1–19

PellowDN2000Environmental inequality formation towarda theoryofenvironmental injusticeAm.Behav. Sci.43581–601

PellowDN2007ResistingGlobal Toxics: TransnationalMovements for Environmental Justice (Cambridge,MA:MITPress)

PrechelH 2012Corporate power andUS economic andenvironmental policy, 1978–2008Camb. J. Reg. Economy Soc.5 357–75

PrechelH andToucheG2014The effects of organizationalcharacteristics and state environmental policies on sulfur-dioxide pollution inUS electrical energy corporations Soc.Sci. Q. 95 76–96

PrechelH andZheng L 2012Corporate characteristics, politicalembeddedness and environmental pollution by largeUScorporations Soc. Forces 90 947–70

RittelHW J andWebberMM1973Dilemnas in a general theory ofplanning Policy Sci. 4 155–69 (December 1969)

Robbins P 1996TNCs and global environmental change: a review oftheUNbenchmark corporate environmental surveyGlob.Environ. Change 6 235–44

Schnaiberg A 1980The Environment: FromSurplus to Scarcity(Oxford:OxfordUniversity Press)

StreitwieserML 1994Cross Sectional Variation in ToxicWasteReleases From theUSChemical IndustryBureau of theCensus

Sze J and London J 2008 Environmental justice at the crossroadsSociol. Compass 2 1331–54

TurkA 1982 Social control and social conflict Social Control: Viewsfrom the Social Sciences ed J PGibbs vol 51 (ThousandOaks,CA: Sage)

9

Environ. Res. Lett. 11 (2016) 015004