economic development quarterly - duke...

http://edq.sagepub.com

Economic Development Quarterly

DOI: 10.1177/089124249901300201 1999; 13; 107 Economic Development Quarterly

James L. Sadd, Manuel Pastor, Jr., J. Thomas Boer and Lori D. Snyder "Every Breath You Take... ": The Demographics of Toxic Air Releases in Southern California

http://edq.sagepub.com/cgi/content/abstract/13/2/107 The online version of this article can be found at:

Published by:

http://www.sagepublications.com

can be found at:Economic Development Quarterly Additional services and information for

http://edq.sagepub.com/cgi/alerts Email Alerts:

http://edq.sagepub.com/subscriptions Subscriptions:

http://www.sagepub.com/journalsReprints.navReprints:

http://www.sagepub.com/journalsPermissions.navPermissions:

http://edq.sagepub.com/cgi/content/refs/13/2/107SAGE Journals Online and HighWire Press platforms):

(this article cites 30 articles hosted on the Citations

© 1999 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at DUKE UNIV on June 2, 2008 http://edq.sagepub.comDownloaded from

http://edq.sagepub.com/cgi/alerts

http://edq.sagepub.com/subscriptions

http://www.sagepub.com/journalsReprints.nav

http://www.sagepub.com/journalsPermissions.nav

http://edq.sagepub.com/cgi/content/refs/13/2/107


FO RUM

“Every Breath You Take . . . ”:The Demographics of Toxic AirReleases in Southern California

James L. SaddOc ci den tal Col lege

Manuel Pas tor, Jr.Uni ver sity of Cali for nia, Santa Cruz

J. Thomas BoerU.S. En vi ron men tal Pro tec tion Agency

Lori D. Sny derAbt As so ci ates

In this arti cle, the authors inves ti gate the rela tion ship between eth nic ity and poten tialenvi ron mental haz ards in the met ro poli tan Los Ange les area. Using a vari ety of tech -niques, includ ing geo graphic infor ma tion sys tems (GIS) map ping, uni vari ate com -pari sons, and logit, ordered logit, and tobit regres sion analy sis, the authors find that,even con trol ling for other fac tors such as income and the extent of manu fac tur ingemploy ment and land use, minor ity resi dents tend to be dis pro por tion ately located inneigh bor hoods sur round ing toxic air emis sions. The results gen er ally sup port thepropo si tions of the pro po nents of “envi ron mental jus tice”; in the con clu sion, they con -sider what this might mean for urban land use and envi ron mental pol icy.

Over the past dec ade, the con cept of “envi ron mental jus tice” has cap tured the atten tion of activ istsand pol icy mak ers alike. Advo cates have argued that low- income and minor ity com mu ni ties in theUnited States host a dis pro por tion ate share of envi ron mental haz ards (Bullard, 1990, 1996; UnitedChurch of Christ [UCC], 1987), espe cially haz ard ous waste treat ment, stor age and dis posal facili -ties (TSDFs), and haz ard ous waste land fills. In 1992, the U.S. Envi ron men tal Pro tec tion Agency(USEPA) responded to these con cerns by estab lish ing an Office of Envi ron men tal Jus tice. In 1994,Presi dent Clin ton issued an execu tive order requir ing each fed eral agency to “make achiev ing envi -ron mental jus tice part of its mis sion by iden ti fy ing and address ing, as appro pri ate, dis pro por tion -ately high and adverse human health or envi ron mental effects of its pro grams, poli cies and activi tieson minor ity popu la tions and low- income popu la tions in the United States” (Execu tive Order12898, 1994).

James L. Sadd is an associate professor of environmentalscience at OccidentalCollege. A geologist withdegrees from the University of Texas at Austin and theUniversity of South Carolina,his research includes thegeochemistry of petroleumreservoirs, heavy metalpollution in coastalenvironments, and issues ofenvironmental justice.

Manuel Pastor, Jr. is the chair of Latin American and Latino Studies at the University ofCalifornia, Santa Cruz. Hehas a Ph.D. in economicsfrom the University ofMassachusetts, Amherst, andhe has written on urbanpoverty, communityempowerment, regionalrestructuring, andenvironmental racism.

J. Thomas Boer, a 1998graduate of The GeorgeWashington University LawSchool, is an attorney withthe U.S. EnvironmentalProtection Agency’s Office ofGeneral Counsel inWashington, D.C.

Lori D. Snyder has a master’s degree in economics fromYale University and currentlyworks as an environmentaleconomist for Abt Associates, Inc. Her research therefocuses on the impact ofregulation on industry andthe recreational benefitsresulting from changes inenvironmental quality.

ECO NOMIC DE VEL OP MENT QUAR TERLY, Vol. 13 No. 2, May 1999 107-123© 1999 Sage Pub li ca tions, Inc. 107

AUTHORS’ NOTE: The views and opin ions expressed in this arti cle are those of the authors and do not reflect those of their insti tu tions or employ ers. Thanks to the Cali for nia Endow ment and Occi den tal Col lege for pro -vid ing the fund ing and infra struc ture sup port for this proj ect, and to Envi ron men tal Data Resources, Inc. andthe South ern Cali for nia Asso cia tion of Gov ern ments for pro vid ing access to some of the data. Thanks also tosev eral anony mous refe rees for excel lent sug ges tions on how to improve the meth od ol ogy and organi za tionof this effort.



The pat tern of dis pro por tion ate prox im ity to haz ards by race partly reflects the impacts of pasteco nomic devel op ment. As many stud ies have shown (Ander ton, Ander son, Rossi et al., 1994;Szasz & Meuser, 1997), envi ron mental haz ards tend to be related to past and cur rent pat terns ofindus trial activ ity, and one rea son for the cor re la tion between race and risk is the ten dency ofminori ties to live in older indus trial areas. At the same time, both the real ity and per cep tion of envi -ron mental injus tice can impede future devel op ment in minor ity urban areas. Fed eral law holds pro -spec tive own ers liable for pre ex ist ing waste on prop erty they seek to pur chase, add ing to the costsof rede vel op ment (U.S. Depart ment of Hous ing and Urban Devel op ment [HUD], 1998). As theenvi ron mental jus tice move ment has gained stat ure and legiti macy in light of sup port ing research(Clark, 1997), many minor ity neigh bor hoods have become reluc tant to fur ther lower envi ron -mental stan dards in exchange for uncer tain prom ises of employ ment. Given this com bi na tion ofenvi ron mental, legal, and politi cal chal lenges, devel op ers may hesi tate to locate a new busi ness in a needy but pol luted urban area; as a result, pre ex ist ing pat terns of “envi ron mental racism” can alsogen er ate dif fer en tial future incomes by race, exac er bat ing the ine qual ity that plagues U.S. cit ies.

In this arti cle, we exam ine the pat terns of prox im ity to envi ron mental haz ards by eth nic ity andother vari ables in one of Ameri ca’s larg est and most une qual met ro poli tan areas: Los Ange les. Thefocus here is on the dis tri bu tion of air borne release of haz ard ous and toxic chemi cals from manu -fac tur ing facili ties, as recorded in the USEPA Toxic Release Inven tory (TRI). We inves ti gate fourpri mary ques tions: (a) What is the geo graphic dis tri bu tion of TRI facili ties in the south ern Cali for -nia region? (b) How does the geo graphic dis tri bu tion of TRI facili ties cor re late spa tially withdemo graphic, socio eco nomic, and land use fac tors stressed by envi ron mental jus tice pro po nentsand other research ers? (c) What is the level of sig nifi cance of these cor re la tions when con sid ered in the con text of a mul ti vari ate model? and (d) What is the rela tion ship between these demo graphicvari ables and rela tive degrees of tox ic ity or amounts of releases?

As it turns out, there is evi dence that minor ity areas are indeed more sub ject to haz ard ous airreleases than are their Anglo coun ter parts. This rela tion ship, ini tially explored through both map -ping and bivari ate analy sis, gen er ally holds up when we test a mul ti vari ate model that takes intoaccount resi dent income, local land use, popu la tion den sity, and other rele vant fac tors. Moreo ver,there is a posi tive rela tion ship between the degree of haz ard and per cent age minor ity, even in thecon text of a mul ti vari ate model. Thus, there may be rea son for pol icy mak ers to be con cerned about poten tial envi ron mental ineq ui ties.

The arti cle pro ceeds as fol lows. We start by review ing the TRI data set and not ing cer tain limi ta -tions to our study. We then turn to ana lyz ing the rela tion ship between eth nic ity and air releases andsub ject the data to uni vari ate sta tis ti cal tests as well as a logit model for loca tion. We then test fordegree and amount of tox ic ity, using ordered logit and tobit tech niques. We con clude by sum ma riz -ing the results and con sid er ing the impli ca tions for both envi ron mental and eco nomic devel op ment pol icy.

DATA AND METH OD OL OGY

Gen eral Is sues

Until recently, most aca demic research inves ti gat ing the degree of socio eco nomic ineq uity inexpo sure to envi ron mental haz ards has focused on the prox im ity to TSDFs. The exten sive per mit -ting process required for TSDF opera tion has allowed for case study analy sis of the poten tial role of insti tu tion al ized dis crimi na tion in siting deci sions (Andeola, 1994; Gelob ter, 1992; Hay nes, 1997;Pulido, Sidawi, & Vos, 1996). The docu men ta tion also has allowed for more gen eral sta tis ti calstud ies of national and regional pat terns with regard to eth nic ity and poten tial expo sure. Most ofthese lat ter stud ies have, indeed, found evi dence of dis pro por tion ate minor ity prox im ity to haz ards(Been, 1995; Boer, Pas tor, Sadd, & Sny der, 1997; Lav elle & Coyle, 1992; Mohai & Bry ant, 1992;UCC, 1987; U.S. Gen eral Account ing Office, 1983).1

Haz ard ous waste TSDFs and land fills, how ever, may not be the most sig nifi cant envi ron mentalhaz ards. For exam ple, TSDFs in south ern Cali for nia, the focus area of this study, man aged about

108 ECO NOMIC DE VEL OP MENT QUAR TERLY / May 1999

ECO NOMIC DE VEL OP MENTQUAR TERLY / May 1999Sadd et al. / EVERY BREATH YOUTAKE. . . there is evidence thatminority areas areindeed more subject tohazardous air releasesthan are their Anglocounterparts.



78,000 tons of haz ard ous waste in 1992, but actual releases of such waste should not occur atTSDFs that con duct nor mal opera tions in com pli ance with state law and the Resource Con ser va -tion and Recov ery Act (RCRA). Thus, being located near a TSDF is only poten tially dan ger ous. Incon trast, dur ing this same period, south ern Cali for nia sites listed in the USE PA’s TRI actu allyreleased a total of 15,206 tons of toxic sub stances into the air imme di ately adja cent to their facili -ties, 2,297 tons of which are classed as known or sus pected car cino gens. Moreo ver, in 1992-1993,there were more than 875 dis tinct facili ties in south ern Cali for nia that reported sub stance releasedata (USEPA, 1994) as com pared to 115 haz ard ous waste facili ties. In short, there are far moreTRI sites than TSDFs, and TRI air releases may result in more chronic expo sure for nearbyresi dents.2

This study focuses on such air releases, rely ing spe cifi cally on the 1992 Toxic Chemi calRelease Inven tory Sys tem (TRIS) data base for the south ern Cali for nia region.3 The data base liststhe release method, the quan tity, and the chemi cal(s) released for each TRI facil ity, and each ofthese meas ures is used as described below. Our defi ni tion of south ern Cali for nia fol lows that of thelocal asso cia tion of gov ern ments: We include Los Ange les, Orange, Ven tura, San Ber nardino, Riv -er side, and Impe rial coun ties.

There are vari ous lim its inher ent to the TRI data. First, there are widely rec og nized loca tionalinac cu ra cies asso ci ated with the geo graphic coor di nates of facili ties in the TRI report ing process(e.g., New York State Parks Man age ment and Research Insti tute, 1993; Wer ner, 1997). To cor rectfor this prob lem in our work, the geo graphic loca tion for each TRI facil ity was obtained by geo cod -ing listed addresses; dif fer en tially cor rected global posi tion ing sys tems sat el lite (GPS) data werethen used to locate sites in cases in which address match ing failed (4.6% of total sites). As a checkon the valid ity of this process, a ran domly selected sub set (7.5% of total) of the sites was vis ited tocon firm cor rect loca tion, using both street address and GPS. Despite this cau tion, loca tion errorsmay remain—but are proba bly small when com pared to tract size and ran domly dis trib uted whenthe site popu la tion as a whole is con sid ered.

Sec ond, the TRIS is not a com plete inven tory of all toxic air borne releases. Dur ing the 1992study year, the EPA required only manu fac tur ing facili ties with 10 or more employ ees—and thatrelease more than 25,000 pounds of listed chemi cals annu ally—to report to the TRI data base. TRIrelease reports also con sist only of self- reported esti mates, because the EPA does not man date thatfacil ity opera tors actu ally moni tor or meas ure their releases.4 On the other hand, there are sub stan -tial civil and crimi nal pen al ties under fed eral law for inten tion ally fail ing to report a TRI release orfor lying in the TRI report ing process, so it is rea son able to assume that the TRIS includes mostquali fy ing releases.

Third, many dan ger ous chemi cals are not sub ject to report ing in the TRI regu la tory regime andfall instead under other envi ron mental stat utes (USEPA, 1994). As a result, the TRI does not pro -vide full cov er age of toxic releases. Fur ther more, the tox ic ity and health risk asso ci ated with TRIair borne releases can vary depend ing on a number of fac tors, includ ing wind and local geog ra phy,sta bil ity and dis per sion char ac ter is tics of the chemi cal in the envi ron ment (with such fac tors assun light, heat, rain, and micro or gan isms con trib ut ing to deg ra da tion), bio con cen tra tion and bioa -vail abil ity, type of expo sure (inha la tion, der mal expo sure, or inges tion), body bur den, syn er gis ticimpact from expo sure to a com bi na tion of chemi cals, and varia tions in the vul ner abil ity of exposed indi vidu als due to age or physi cal con di tion.

There are also vari ous lim its to this sort of research as a whole. Most impor tant is our under ly ing assump tion—inher ent in most other stud ies on this topic (see, for exam ple, Pol lock & Vit tes,1995)—that resi den tial prox im ity is an impor tant fac tor in deter min ing expo sure to haz ard ous sub -stances and, there fore, risk.5 Sev eral epi de mi ol ogi cal stud ies have indeed dem on strated a sig nifi -cant rela tion ship between resi den tial prox im ity to urban toxic sub stance stor age and/or air releasefacili ties, and increased health risk and dis ease inci dence, espe cially among preg nant women andinfants (Berry & Bove, 1997; Croen, Shaw, San bon matsu, Sel vin, & Buf fler, 1997; Gold man, Pai gen,Mag nant, & High land, 1985; Guthe et al., 1992; Knox & Gil man, 1997; Nord strom, Beck man, &Nord strom, 1978).6 There are, how ever, other stud ies that do not find such a rela tion ship (Bell,Franks, Hil dreth, & Melius, 1991; Ozonoff, Col ten, & Cup ples, 1983; Shaw, Schul man, Frisch,Cum min, & Har ris, 1992).

Sadd et al. / EVERY BREATH YOU TAKE 109

There are also variouslimits to this sort ofresearch. . . . Mostimportant is ourunderlyingassumption . . . thatresidential proximity isan important factor indetermining exposure tohazardous substancesand, therefore, risk.



Although Berry and Bove (1997) pro vide detailed meth odo logi cal cri tiques of these non con -firm ing stud ies, our main response is two fold. First, even account ing for the less sup por tive stud ies, the over all evi dence sug gests that prox im ity to a TRI site is a prob able nega tive—a neu tral at best,but never a posi tive. As a result, a find ing that there is dis pro por tion ate prox im ity to TRI releasesmay sug gest, at a mini mum, that a for mal expo sure and risk assess ment should be done. Sec ond,hav ing a TRI facil ity sited nearby is per ceived as a risk, as has been shown in sur vey research (U.S.Coun cil on Envi ron men tal Qual ity, 1980) and mani fested by the usual “not in my back yard” chal -lenges to such siting (Bullard, 1990, 1996). A pat tern of dis pro por tion ate prox im ity to haz ards thus sug gests ineq uity in the dis tri bu tion of risk per cep tions—and there is some evi dence that these per -cep tions can even tu ally wind up nega tively influ enc ing eco nomic devel op ment pat terns (see HUD, 1998).

Data, Vari ables, and Meth ods

Although the TRIS data base includes facili ties engaged in dif fer ent types of release activ ity, wecon sider only TRI air releases in this study. These include both stack emis sions from con fined airstreams and fugi tive emis sions from sources such as equip ment leaks, evapo ra tive losses, andreleases from build ing ven ti la tion sys tems; together, these con sti tute 96% of all TRI releases intothe envi ron ment (land, water, or air) in the south ern Cali for nia study area. We also rank air releasesby their rela tive tox ic ity, con sid er ing sepa rately air releases that include USEPA 33/50 chemi calsand car cino gens. The 33/50 releases are iden ti fied by the USE PA’s 33/50 pro gram, a vol un tary pol -lu tion pre ven tion ini tia tive designed to reduce by half the releases and trans fers of 17 high- prioritytoxic sub stances dur ing the period from 1988 to 1995 (USEPA, 1994). Car cino gens are those TRIchemi cals that are either known or sus pected to cause can cer, as clas si fied by the Occu pa tionalSafety and Health Admin istra tion.7 Because nearly 90% of our 33/50 obser va tions are also classedas car cino genic releases, we decided to use the broader 33/50 cate gory and sepa rate geo graphicareas into those with no air releases, those with air releases, and the sub set of the lat ter with 33/50air releases. The results below do not change if we dis tin guish areas with 33/50 and car cino genreleases as sepa rate cate go ries.8

Using the 1992 U.S. Cen sus TIGER data files, each TRI site was geo coded to locate its host cen sustract, and tracts were then matched with demo graphic data drawn from the 1990 U.S. Cen sus Sum -mary Tape Files (STF-1 and STF-3) with 1992 land use data pro vided by the South ern Cali for niaAsso cia tion of Gov ern ments. Eleven dif fer ent demo graphic vari ables were con sid ered in our ini tialsta tis ti cal pro file: the per cent age of minor ity (non- Anglo9), Afri can Ameri can, and Latino resi -dents in each cen sus tract; tract val ues of mean per cap ita income; median house hold income;median house value (self- reported); median con tract rent; per cent age of resi dents employed in manu -fac tur ing; per cent age of tract used as resi den tial land and indus trial land; and popu la tion den sity.10

Most of these vari ables have been used in prior stud ies evalu at ing dis pro por tion ate prox im ity toenvi ron mental haz ards. The vari ous minor ity vari ables are, of course, the focal point of the envi -ron mental jus tice debate. The eco nomic vari ables are intro duced because many ana lysts have sug -gested that locally unde sir able land uses, such as toxic releases, tend to be located where landval ues and incomes are low est—partly because both par cel acqui si tion costs and com pen sa tion, ifany, to affected resi dents would likely be lower (Szasz & Meuser, 1997).

Con trol ling for the per cent age of resi dents employed in manu fac tur ing was first sug gested byAnder ton, Ander son, Oakes, and Fraser (1994) and Ander ton, Ander son, Rossi et al. (1994), and itreflects the sen si ble notion that indus trial and manu fac tur ing com pa nies (many with TSDFs ortoxic air releases) and their employee base tend to locate near one another. Popu la tion den sity isanother widely used vari able, reflect ing the gen er ally held notion that toxic releases and otherunde sir able out comes should occur where they will affect the least number of peo ple. The land usevari ables were intro duced into the debate by Boer et al. (1997), who found them to be very pow er -ful pre dic tors of TSDF loca tion and supe rior to popu la tion den sity. Gen er ally avail able only on aregional level, land use is used in this study to con trol for the gen eral char ac ter of the neigh bor hood(i.e., because area in resi den tial use is unavail able for manu fac tur ing facili ties that might have TRIreleases).




METH ODS AND RE SULTS

The essen tial strat egy of the analy sis is to com pare the demo graphic char ac ter is tics of tracts that con tain or are located near a TRI release site to tracts not proxi mate to a site. We begin with quali ta -tive geo graphic com pari son, con struct ing over lay maps with geo graphic infor ma tion sys tems soft -ware (GIS, Arc/Info). Uni vari ate analy sis is then used to test the sta tis ti cal sig nifi cance of theappar ent geo graphic pat terns. Finally, bino mial logit, ordered logit, and tobit regres sion analy sesare employed to evalu ate the rela tive sig nifi cance of demo graphic char ac ter is tics in pre dict ing theloca tion and degree of vari ous TRI releases.

Fig ure 1 is a map of the dis tri bu tion of all 1992 TRI air release sites. As can be seen, these sitesare con cen trated within the heav ily urban ized “met ro poli tan Los Ange les” area cen tered aroundsouth ern Los Ange les County. Fig ure 2 enlarges this area, chart ing both the TRI 33/50 and car cino -genic air release sites and the tracts in which the per cent age of Afri can Ameri can or Latino resi -dents exceeds the area mean. The vis ual pat tern strongly sug gests that these haz ard ous releases aredis pro por tion ately located in areas with high (rela tive to the mean) con cen tra tions of racial andeth nic minori ties and that few of the TRI air release sites are located out side of such neigh bor -hoods. In fact, 71% of all TRI air releases and 73% of all 33/50 and car cino gen releases are locatedwithin these dis pro por tion ately minor ity tracts; 89% of all TRI air releases are located within 1mile of these same tracts.

To evalu ate these pat terns sta tis ti cally, we cre ated sev eral subsam ples of cen sus tracts. Airincludes all tracts that con tain at least one TRI facil ity that reported a toxic release into the air in1992 (n = 318). Air/1.0 includes both these tracts and tracts with bounda ries located within 1 mileof such a facil ity (n = 1,490). A simi lar subsam ple denoted 3350 includes tracts that con tain TRIfacili ties that released EPA 33/50 chemi cals (n = 162). 3350/1.0 includes both these tracts andthose with bounda ries within 1 mile (n = 1,005).11


The visual patternstrongly suggests thesehazardous releases aredisproportionatelylocated in areas with high (relative to the mean)concentrations of racialand ethnic minorities . . .

Fig ure 1: Geo graphic Dis tri bu tion of All Toxic Re lease In ven tory Air Re leases in South ern Cali for nia Study Area



We then com pared the val ues of vari ous demo graphic, income, and land use vari ables for tractswith and with out the defined haz ard and range (Air, Air/1.0, etc.). The sig nifi cance of the dif fer -ences in means was cal cu lated using the t test, and the sig nifi cance of the dif fer ences in medi answas cal cu lated using the non para met ric Wil coxon pro ce dure.12 For con ven ience, we report for each vari able only the over all mean and the mean for tracts with TRI releases of dif fer ent sorts; ofcourse, the t tests appro pri ately refer to com pari sons with the (unre ported) means for non- TRItracts. Note also that the medi ans go unre ported for rea sons of space. The results are given in Tables 1and 2.

With few excep tions, the results of the uni vari ate analy sis are highly sig nifi cant and gen er allycon sis tent with the claim of ineq ui ta ble expo sure offered by envi ron mental jus tice advo cates.When com pared to the study area over all, tracts con tain ing a TRI air release facil ity (see Table 1) as well as a broader group includ ing those proxi mate to TRI air release facili ties (see Table 2) have, asa group, a resi den tial popu la tion that is sta tis ti cally poorer and more likely to work in the manu fac -tur ing indus try. Tracts with, or proxi mate to, air releases also have a higher per cent age of minori -ties in the popu la tion. These areas, as might be expected, are also com posed of a much higherper cent age of indus trial land use.13 Finally, the vari ous demo graphic and eco nomic dif fer encesgen er ally increase and remain highly sig nifi cant when we exam ine those neigh bor hoods that con -tain or are close to the more toxic 33/50 releases.

A close exami na tion of these results reveals some inter est ing pat terns. First, when Afri canAmeri cans are con sid ered sepa rately, the t tests are never sig nifi cant at the 5% level, and the per -cent age Afri can Ameri can is actu ally lower in the tracts that them selves con tain TRI releases.When we expand the analy sis to include all tracks within 1 mile of a toxic release, the per cent ageAfri can Ameri can is higher, but the t sta tis tic remains insig nifi cant. How ever, because Afri canAmeri cans are con cen trated in his tori cally Black South Cen tral and a few other areas,14 the non -para met ric Wil coxon test with regard to medi ans may be more reli able. This Z sta tis tic is sig nifi cant


Fig ure 2: En larged View of the Met ro poli tan Los An ge les Por tion of South ern Cali for nia, Lo ca tions of Car cino -gen and 33/50 Toxic Re lease In ven tory Air Re lease Sites, and Cen sus Tracts With a Pre domi nantly Mi nor ity Resi -den tial Popu la tion in the 1990 Cen sus



through out the in- tract and proxi mate com pari sons. Still, the posi tive asso cia tion of per cent ageminor ity and air releases seems to be driven mostly by the per cent age of Latino resi dents—a pat -tern that also emerges in the mul ti vari ate tests below.

A sec ond inter est ing result involves land use and popu la tion den sity. As expected, the in- tractsubsam ples exhibit much higher per cent ages of indus trial land than does the study area mean; theper cent age of resi den tial land and popu la tion den sity is cor re spond ingly lower (see Table 1). When the area within 1 mile of these releases is con sid ered, how ever, the ratio of indus trial to resi den tialland use is more simi lar to that of the study area, but popu la tion den sity is sig nifi cantly higher (seeTable 2). It is dis turb ing that peo ple liv ing within 1 mile of these releases are doing so under con di -tions even more crowded than in the rest of the study area.

Another way to look at the issue of dis pro por tion ate prox im ity is to con sider the per cent age ofvari ous groups liv ing in a tract with an air or 33/50 release. Fig ure 3 shows that 10.6% of Anglos insouth ern Cali for nia live in a tract with a TRI air release, and 5.4% live in a tract with a 33/50release; the “expo sure” rates for Lati nos are 17.7% for the air release tracts and 10% for the 33/50tracts, with Afri can Ameri can rates fal ling some where in the mid dle.15 In short, Lati nos are twice as likely as Anglos to be near an envi ron mental haz ard—a pat tern that we now dis en tan gle by simul -ta ne ously con sid er ing the effect of other key vari ables on TRI loca tion.


. . . the positiveassociation of percentageminority and air releasesseems to be driven mostly by the percentage ofLatino residents—apattern that also emerges in the multivariate testsbelow.

TA BLE 1

Uni vari ate Sta tis tics—Tracts Con tain ing Toxic Re lease In ven tory (TRI) Air Re leases

Mean of Mean ofTract Mean Tracts With Tracts With

Vari ables 5 County Air TRI t Value Z Sta tis tic 3350 TRI t Value Z Sta tis tic

Per cent age mi nor ity 48.9 60.7 8.33**** –7.95**** 62.7 6.94**** –6.27****Per cent age Af ri can Ameri can 8.2 7.7 –0.62 –2.38*** 8.1 –0.07 –2.08***Per cent age La tino 31.2 43.2 8.75**** –9.66**** 45.4 7.33**** –7.53****Per cap ita in come $17,445 $13,186 –12.27**** –7.33**** $12,764 –10.79**** –5.83****House hold in come $39,781 $35,065 –7.11**** –4.17**** $34,703 –5.66**** –3.24****Per cent age manu fac tur ing em ploy ment 19.5 24.9 11.88**** –11.40**** 25.9 9.75**** –9.36****Per cent age resi den tial land 58.8 39.9 –13.62**** –13.51**** 35.1 –11.62**** –11.12****Per cent age in dus trial land 9.8 31.5 18.73**** –20.69**** 37.7 15.18**** –16.31****Me dian house value $230,221 $189,806 –9.98**** –6.34**** $186,090 –8.01**** –4.90****Me dian rent $700 $663 –4.48**** –3.54**** $656 –3.86**** –3.13****Popu la tion den sity 8,964 6,144 –10.05**** –6.67**** 5,495 –9.19**** –6.40****

***p < .05. ****p < .01.

TA BLE 2

Uni vari ate Sta tis tics—Tracts Proxi mate to Toxic Re lease In ven tory (TRI) Air Re leases

Mean of Mean of Tract Mean Tracts With Tracts With

Vari ables 5 County Air TRI/1.0 t Value Z Sta tis tic 3350 TRI/1.0 t Value Z Sta tis tic

Per cent age mi nor ity 48.9 57.2 17.34**** –17.32**** 60.5 16.31**** –16.09****Per cent age Af ri can Ameri can 8.2 8.4 0.61 –5.18**** 8.5 0.77 –2.53***Per cent age La tino 31.2 38.4 18.34**** –17.24**** 41.9 16.71**** –16.53****Per cap ita in come $17,445 $14,742 –13.83**** –13.74**** $13,937 –14.74**** –13.24****House hold in come $39,781 $36,716 –9.66**** –8.05**** $35,907 –9.65**** –7.56****Per cent age manu fac tur ing em ploy ment 19.5 22.6 24.17**** –21.97**** 24.4 23.66**** –22.39****Per cent age resi den tial land 58.8 59.9 2.32*** –0.66 59.2 0.52 –1.00Per cent age in dus trial land 9.8 13.4 16.94**** –14.02**** 16.4 16.73**** –16.47****Me dian house value $230,221 $211,511 –9.6**** –8.26**** $202,501 –11.15**** –9.49****Me dian rent $700 $680 –6.55**** –6.54**** $673 –6.40**** –6.24****Popu la tion den sity 8,964 9,875 6.54**** –10.43**** 9,544 3.05**** –7.33****

***p < .05. ****p < .01.



MUL TI VARI ATE ANALY SIS

Bi no mial Lo git

By its nature, the uni vari ate analy sis above can not sepa rate the effects of each indi vid ual demo -graphic char ac ter is tic and deter mine its rela tive impor tance in pre dict ing facil ity loca tion. Thisposes par ticu lar prob lems in our research because many of the vari ables (e.g., per cent age minor ityand income) are them selves highly cor re lated, mak ing it more dif fi cult to dis tin guish whether siteloca tion is the result of income or eth nic dis crimi na tion or arises sim ply because of market- drivenoppor tu nity costs (Lam bert & Boer ner, 1994). If, for exam ple, the cor re la tions with race/eth nic itydis ap pear once we con trol for income, land use, and manu fac tur ing employ ment, the minor ityasso cia tion with TRI air release facili ties would be sim ply an unfor tu nate result of the strong cor re -la tion between per cent age minor ity and these other vari ables and would not nec es sar ily be linkedto any poten tial racial dis crimi na tion.

In an attempt to sepa rate these dif fer ent impacts and tease out the inde pend ent effect (if any) ofrace, we used sev eral dif fer ent mul ti vari ate regres sion tech niques. The first is a bino mial logit, con -ducted on sev eral dif fer ent depend ent vari ables. The vari ables Air and 3350 equal 1 if the tract con -tains a facil ity report ing a TRI air release or a USEPA 33/50 chemi cal air release, respec tively, and0 if it does not. The vari ables Air/1.0 and 3350/1.0 take a value of 1 if the tract bound ary is within a1-mile radius of a facil ity report ing the appro pri ate air release and 0 oth er wise. Both the logit tech -nique and this sort of tract clas si fi ca tion have become a stan dard in the quan ti ta tive envi ron mentaljus tice lit era ture (Ander ton, Ander son, Oakes et al., 1994; Ander ton, Ander son, Rossi et al., 1994;Been, 1995; Boer et al., 1997), but most of the pre vi ous work has focused on TSDFs.

In the regres sion analy sis, we elimi nate some of the demo graphic vari ables because of a highdegree of mul ti ple col line ar ity; for exam ple, house hold income and per cap ita income are obvi -ously highly cor re lated, and both are cor re lated with house val ues and rents.16 The result ing equa -tion to be tested is


Fig ure 3: Ex po sure by Group



Air (or the other de pend ents) = f [MI NOR ITY(+), PER CAPIN(+),PER CAP IN2(–), IND LAND(+), EMP MANU(+), POP DEN(–)],

where MI NOR ITY is the per cent age of non- Anglo resi dents, PER CAPIN is the tract per cap ita in -come, PER CAP IN2 is the square of per cap ita in come, IND LAND is the per cent age of tract areazoned for in dus try, EMP MANU is the per cent age of resi dents em ployed in manu fac tur ing, andPOP DEN is popu la tion den sity.

The signs in paren the ses indi cate the expected direc tion of influ ence for each vari able. Forexam ple, accord ing to the hypothe ses offered by envi ron mental jus tice advo cates, the per cent ageof non- Anglo resi dents (MINORITY) should increase the like li hood of liv ing near a facil ityreport ing a toxic air release. Although the expected effect of most of the vari ables is intui tivelyobvi ous, the speci fi ca tion of income deserves a spe cial dis cus sion. Fol low ing Been (1995), Boeret al. (1997), and Szasz and Meuser (1997), we assume that income has an “inverted U” rela tion -ship to the prob abil ity of a toxic release: The low est income areas have so lit tle eco nomic activ itythat they have lit tle like li hood of being near plants with toxic releases, whereas wealth ier areashave the politi cal and other resources to resist the place ment of dirt ier indus tries. As a result, theprob abil ity of expe ri enc ing toxic releases tends to “peak” some where in the mid dle of the incomedis tri bu tion—a pat tern cap tured by enter ing both per cap ita income and its squared value.

The results of these vari ous logit regres sions are reported in Tables 4 and 5; note that we reportWald sta tis tics rather than t sta tis tics for the sig nifi cance of the coef fi cients.17 As can be seen, the“in- tract” results (see Table 3) show only weak evi dence of racial ineq uity. Although MINORITYis signed as expected, it is sig nifi cant at only the .20 level for air releases and not sig nifi cant for the33/50 releases. The remain ing vari ables in the regres sion cap ture the land use, eco nomic, andemploy ment con di tions of the tract, and all are sig nifi cant pre dic tors at the 1% level. The results are simi lar if one con sid ers the per cent ages of Afri can Ameri can and Latino resi dents sepa rately (seeTable 3, col umns c and d). In both speci fi ca tions, the per cent age Afri can Ameri can is highly insig -nifi cant; per cent age Latino is sig nifi cant at the .10 level, but the truly sig nifi cant pre dic tors are theeco nomic, employ ment, and land use vari ables.

This pat tern changes when the 1-mile “buffer” zone sur round ing air releases is con sid ered (seeTable 4); recall that this group ing includes both the tracts that con tain an air release and tractswithin a 1-mile radius of that release. In this sam ple, MINORITY is posi tive and highly sig nifi cantat the 1% level for both air and 33/50 releases.18 When Latino and Afri can Ameri can eth nic ity arecon sid ered sepa rately, LATINPCT is sig nifi cant at the 1% level, but AFAMPCT is sig nifi cant onlyfor the 33/50 releases. Finally, note that popu la tion den sity is nega tively signed and sig nifi cantonly for USEPA 33/50 chemi cal air releases.19

Recall that the actual logit coef fi cients reflect the impact on the log of the odds, hence offer lit tleintui tive infor ma tion about the rela tive impor tance of each vari able in pre dict ing TRI loca tion.How ever, these regres sion coef fi cients can be manipu lated to deter mine the mar ginal effect of achange in an inde pend ent vari able on the prob abil ity that a TRI air release facil ity is located in ornearby. To do so, we first cali brated the model by set ting all vari ables at the study area mean. Wethen increased a sin gle vari able by 10%; the result ing change in prob abil ity is writ ten in the brack -ets in Tables 3 and 4.20 Thus, if the ran dom chance of liv ing in a TRI tract is, say, 0.30, and a 10%increase in per cent age minor ity raises the prob abil ity to 0.31, the effect is .01, or a 1% pointincrease.

The results of this sen si tiv ity analy sis for the vari ous in- tract subsam ples (see Table 3) show asimi lar pat tern regard less of the type of TRI release. When all air releases are con sid ered, the larg -est mar ginal effects come from chang ing the employ ment in manu fac tur ing (a 10% increase hereraises the prob abil ity by 1.2 per cent age points), fol lowed by popu la tion den sity (–1.0 per cent agepoints) and indus trial land use (0.75 per cent age points); the per cent age minor ity and per cap itaincome have much smaller mar ginal effects. When the sam ple is restricted to USEPA 33/50 chemi -cals and car cino gens, the rela tive rank ings on the inde pend ent vari ables change. Here, the larg estmar ginal effect comes from an increase in per cap ita income, fol lowed by popu la tion den sity.




Manu fac tur ing employ ment and indus trial land use pres ent almost equal mar ginal effects, andMINORITY has an (insig nifi cant) effect of less than 0.10%. The results sug gest that the racialcom po si tion of the tract plays a sec on dary role to the eco nomic char ac ter is tics in deter min ing theexact loca tion of toxic air releases.

When the model is used to pre dict tracts within 1 mile of a TRI release (see Table 4), dif fer encesemerge in the rela tive impor tance of race and income vari ables. Manu fac tur ing employ ment con -tin ues to have the larg est mar ginal effect in all subsam ples, with a 10% increase yield ing anincreased prob abil ity of 4 to 5 per cent age points. There is also, how ever, a sig nifi cant mar ginaleffect asso ci ated with MINORITY, rang ing from 1.5 to 1.8 per cent age points. Increases in per cap -ita income and indus trial land use have a lower impact on prob abil ity, and the effect of popu la tionden sity is quite small and sta tis ti cally insig nifi cant in the case of air releases. When AFAMPCTand LATINPCT are used, the mar ginal effect on prob abil ity for varia tion in per cent age Lati nos andincome is nearly equiva lent; varia tions in the per cent age of Afri can Ameri cans pro duce lit tle mar -ginal effect.

To sum, it seems that eco nomic vari ables are the pri mary deter mi nants of the tract- level loca tion of TRI air borne releases. When the radius of con sid era tion is increased, race may mat ter more, atthe mar gin, than either per cap ita income or land use.

In real ity, land use, employ ment, income, and eth nic com po si tion tend to move in tan dem. Tocap ture this and to check our model, we con structed two dif fer ent hypo theti cal tract pro files andthen used the logit coef fi cients to deter mine the prob abil ity that a tract fit ting this pro file wouldcon tain or be within 1 mile of an air release. Pro file A rep re sents a pri mar ily resi den tial cen sus tract(2.5%-7.5% indus trial land; popu la tion den sity of 1,000 to 10,000) where resi dents are pri mar ilyWhite (0-20% non- Anglo), mid dle to upper- middle class (per cap ita income $25,000 to $35,000),and employed in the white- collar pro fes sions (5% to 20% manu fac tur ing employ ment). Pro file B


. . . it seems thateconomic variables arethe primarydeterminants of thetract-level location ofTRI airborne releases.When the radius ofconsideration isincreased, race maymatter more, at themargin, than either percapita income or landuse.

TA BLE 3

Lo git Re sults on De ter mi nants of In- Tract Air Re leases

Dependent Air TRI 3350 TRI Air TRI 3350 TRIVari able (a) (b) (c) (d)

MI NOR ITY 0.007 0.006(2.088)* (.886){.0037} {.0007}

PER CAPIN 0.042 0.043 0.046 0.051(11.208)**** (6.886)**** (10.305)**** (7.685)****{–.0005} {–.0038} {.0049} {–.0033}

PER CAP IN2 –0.0002 –0.0003 –0.0002 –0.0003(15.178)**** (8.969)**** (14.874)**** (9.969)****

IND LAND 0.074 0.061 0.075 0.061(199.566)**** (119.633)**** (207.262)**** (125.562)****

{.0075} {.0013} {.0090} {.0018}EMP MANU 0.057 0.050 0.052 0.045

(25.617)**** (12.559)**** (20.077)**** (9.378)****{.0117} {.0021} {.0126} {.0027}

POP DEN –0.011 –0.011 –0.010 –0.011(34.287)**** (19.109)**** (33.552)**** (19.333)****{–.0104} {–.0022} {–.0119} {–.0031}

AFAMPCT –0.001 0.003(.017) (.134){.0008} {.0001}

LATINPCT 0.010 0.012(2.500)* (1.979)*{.0038} {.0011}

Us able obser va tions 2,567 2,567 2,567 2,567– 2 log like li hood 1243.7 787.8 1242.3 786.6Per cent age cases cor rect 90.7 94.9 90.4 94.7

NOTE: Wald Sta tis tics in paren the ses; mar ginal effects in brack ets (see dis cus sion in text).*p < .20. ****p < .01.



is pre domi nantly minor ity (80% to 100% non- Anglo), low income (per cap ita income $5,000 to$10,000), highly indus trial (20% to 30% indus trial land use), with higher blue- collar employ ment(30% to 45%) and popu la tion den sity (12,000 to 22,000).

Using the regres sion results in Table 4, col umn (a), we cal cu lated the esti mated like li hood ofeach tract pro file being located within a mile of a TRI air release. For Pro file A, this prob abil ity is awide band, rang ing from 14% to 58%; for Pro file B, the cal cu lated prob abil ity is a much tighter93% to 99%.21 A total of 24 actual tracts in the study area fit the descrip tion for Pro file A, of which 9are within a mile of a TRI air release; this actual inci dence rate of 37.5% is square in the mid dle ofthe esti mated band. Simi larly, 10 actual tracts fit the descrip tion for Pro file B, all of which arelocated within a mile of a TRI air release; this actual inci dence of 100% hit is quite close to our93%-to -99% esti mated band. The model proves equally accu rate for 33/50 at the 1-mile “buffer.”22

In short, the model does a rela tively good job of pre dict ing the prob abil ity of liv ing near a TRI airrelease.

Or dered Lo git

To fur ther explore the issue of dis pro por tion ate prox im ity to poten tial risks, we sub jected thedata to a tech nique called ordered logit. In an ordered logit, a dis crete depend ent vari able can takeon more than two pos si ble val ues. In our case, we con structed a vari able “ordered” accord ing to thelevel of assumed health haz ard; thus, the depend ent takes a value of 0 if the tract has no air release, a value of 1 if it has an air release of a com pound that is not a car cino gen or USEPA 33/50 chemi cal,and a value of 2 if it con tains a TRI site with a USEPA 33/50 or car cino gen air release. Althoughsuch rank ing strate gies are com mon in eco nomic research, these approaches to meas ur ing the


TA BLE 4

Lo git Re sults on De ter mi nants of Proxi mate Air Re leases

Dependent Air TRI/1.0 3350 TRI/1.0 Air TRI/1.0 3350 TRI/1.0Vari able (a) (b) (c) (d)

MI NOR ITY 0.014 0.015(24.542)**** (22.947)****

{.0153} {.0178}PER CAPIN 0.015 0.021 0.013 0.022

(12.405)**** (14.987)**** (8.180)**** (12.594)****{.0103} {.0142} {.0077} {.0141}

PER CAP IN2 –4.9E–05 –7.0E–05 –4.7E–05 –7.5E–05(14.735)**** (15.649)**** (11.493)**** (14.355)****

IND LAND 0.045 0.045 0.046 0.045(77.729)**** (103.292)**** (80.334)**** (106.350)****

{.0098} {.0107} {.0100} {.0108}EMP MANU 0.098 0.109 0.094 0.105

(187.222)**** (214.554)**** (167.527)**** (187.187)****{.0416} {.0525} {.0401} {.0505}

POP DEN 0.0003 –0.0018 0.0008 –0.0015(0.203) (6.844)**** (1.690) (4.895)***{.0007} {–.0044} {.0018} {–.0037}

AFAMPCT 0.001 0.007(0.092) (3.681)**{.0002} {.0013}

LATINPCT 0.015 0.016(15.310)**** (15.899)****

{.0107} {.0125}Us able obser va tions 2,567 2,567 2,567 2,567– 2 log like li hood 2803.2 2652.8 2810.9 2659.5Per cent age cases cor rect 71.2 75.9 71.7 75.6

NOTE: Wald sta tis tics in paren the ses; margi nal effects in brack ets (see dis cus sion in text).**p < .10. ***p < .05. ****p < .01.



degree or amount of poten tial expo sure have not widely been used in pre vi ous envi ron mental jus -tice stud ies (see also Glick man & Hersh, 1995).23 In any case, this essen tially con sti tutes a test ofwhether demo graphic vari ables are cor re lated with the degree of poten tial risk—at least to theextent that the USE PA’s 33/50 list is acknowl edged to con tain more dan ger ous chemi cals.

The pat tern that emerges from these tests (see Table 5) is gen er ally con sis tent with the resultsfound in the bino mial logit. Here, how ever, the per cent age of minor ity resi dents is posi tive and sig -nifi cant at both the tract and 1-mile level; when we break up the minor ity groups, we see once againthat these results are largely driven by LATINPCT. Also, although popu la tion den sity is nega tivelyand sig nifi cantly asso ci ated with the degree of in- tract haz ard, it is actu ally posi tively (albeit insig -nifi cantly) asso ci ated with the degree of haz ard at the 1-mile level. In any case, the over rid ing mes -sage from this set of results is that a tract with a higher per cent age of minor ity (but par ticu larlyLatino) resi dents is more likely to live near a facil ity with a release that poses a greater poten tialhealth haz ard.24

To bit Analy sis

We take a final cut at the degree of expo sure issue through a tobit analy sis. Here, the depend entvari able is pounds of either air or 33/50 release in any given tract. Because most tracts have noreleases, the sam ple is “cen sored” at zero. The tobit analy sis takes this into account and allows us to esti mate the rele vant rela tion ships, includ ing both the prob abil ity that releases will be posi tive ina tract and the expected change in releases, if any, for a change in the inde pend ent vari able (seeKen nedy, 1992). Like the ordered logit, this tech nique has not been used much in the lit era ture ondis pro por tion ate prox im ity to envi ron mental haz ards.25

The results of such an analy sis are shown in Table 6. As can be seen, all vari ables are signed asexpected, and most are quite sig nifi cant. MINORITY is posi tive and sig nifi cant at the .10 level forthe 33/50 releases, as is LATINPCT when the two major minor ity groups are bro ken out sepa rately. The over all pic ture that emerges from both this and the ear lier analy sis is that eth nic ity, income,indus trial land use, and prox im ity of manu fac tur ing employ ees all mat ter for both TRI loca tion and the degree of poten tial risk.


TA BLE 5

Or dered Lo gits: Tracts Or dered by No Re lease, Air Re lease, and 3350 Re lease

Dependent Or dered Or dered/1.0 Or dered Or dered/1.0Vari able (a) (c) (d) (f)

MI NOR ITY 0.008 0.014(1.683)** (5.343)****

PER CAPIN 0.045 0.014 0.049 0.012(3.744)**** (3.597)**** (3.556)**** (2.750)****

PER CAP IN2 –0.0002 –4.7E–04 –0.0003 –4.3E–05(–4.253)**** (–3.909)**** (–4.202)**** (–3.335)****

IND LAND 0.070 0.046 0.071 0.047(15.134)**** (10.784)**** (15.457)**** (11.014)****

EMP MANU 0.052 0.101 0.048 0.098(4.949)*** (15.896)**** (4.436)**** (15.027)****

POP DEN –0.010 –0.0001 –0.010 0.000(–5.897)**** (–0.127) (–5.796)**** (0.767)

AFAMPCT 0.0021 0.003(0.360) (1.141)

LATINPCT 0.010 0.013(1.604)* (3.669)****

Us able ob ser va tions 2,567 2,567 2,567 2,567Log like li hood –822.1 –2170.0 –822.1 –2177.4Pseudo R2 percentage 29.8 17.1 29.8 16.8

NOTE: Z sta tis tic in paren the ses.*p < .20. **p < .10. ***p < .05. ****p < .01.



CON CLU SION

One of the most suc cess ful community- organizing sto ries in south ern Cali for nia in recent yearsinvolved effects by resi dents of pri mar ily Afri can Ameri can South Cen tral and largely Latino EastLos Ange les to resist the place ment of a haz ard ous waste incin era tor in an area abut ting their twoneigh bor hoods. Driven by wor ries that such a facil ity would release dan ger ous sub stances intoneigh bor hood air, organ iz ers and com mu nity resi dents were able to coa lesce and then stop thefacil ity in 1991, partly by wav ing the ban ner of envi ron mental racism.26

The results of this study sug gest that these com mu nity groups may have been right to be con -cerned about whether an addi tional waste incin era tor would exac er bate pre ex ist ing pat terns of dis -pro por tion ate prox im ity. Racial dis pari ties clearly are evi dent in our sim ple uni vari ate tests. Usinga bino mial logit to con trol for mul ti ple causal vari ables, we find that indus trial land use, per cent ageof resi dents employed in manu fac tur ing, and popu la tion den sity all mat ter to TRI loca tion,whereas income takes on a U-shaped pat tern in which the poor est and rich est tracts expe ri ence theleast like li hood of hav ing a TRI. As for per cent age minor ity, this vari able is cor rectly signed butinsig nifi cant for the “hazard- in- tract” mul ti vari ate test; how ever, it is quite sig nifi cant for the“proximate- area” tests, with Lati nos most con sis tently affected. Moreo ver, both an ordered logittest, which ranks TRI air releases by the degree of haz ard, and a tobit test, which ranks air release by pounds of emis sion, sug gest that the degree of tox ic ity and per cent age minor ity are related, even ina mul ti vari ate set ting.

The results deserve some cave ats. The first is that a “snap shot” of the cur rent dis tri bu tion of TRIsites does not elu ci date whether TRI facili ties were located in minor ity areas or whether minori -ties moved in after the prox im ity to poten tial haz ards shifted prop erty val ues and neigh bor hooddesir abil ity. TRI facili ties are proba bly not ame na ble to such a his tori cal analy sis, becausereport ing is only recent; we are, how ever, cur rently con struct ing a his tory of TSDFs and con nect -ing this with 30 years of cen sus mate rial to get at this issue. We note, how ever, that the regres -sions in this arti cle con trol for income; as a result, any argu ment about “minor ity move-in” would need to explain why eth nic minori ties, even with the same eco nomic resources, live near poten -tial toxic haz ards.


. . . a “snapshot” . . . doesnot elucidate whetherTRI facilities werelocated in minority areasor whether minoritiesmoved in after theproximity to potentialhazards shifted propertyvalues and neighborhooddesirability.

TA BLE 6

To bit Re sults on Amount of Re lease

Dependent Lbs- Air Lbs- 3350 Lbs- Air Lbs- 3350Vari able (a) (c) (d) (f)

MI NOR ITY 599.877 414.910(1.243) (1.824)**

PER CAPIN 4522.522 1439.367 4284.962 1701.685(3.981)**** (2.656)**** (3.294)**** (2.728)****

PER CAP IN2 –23.3310 –7.8E+00 –22.8375 –8.8E+00(–4.351)**** (–2.908)**** (–3.942)**** (–3.033)****

IND LAND 6788.776 2080.314 6875.054 2089.835(14.847)**** (9.770)**** (15.163)**** (9.929)****

EMP MANU 4296.398 1040.420 4349.491 980.273(4.431)**** (3.362)**** (4.313)**** (2.144)***

POP DEN –866.676 –325.6477 –824.069 –313.982(–5.215)**** (–3.990)**** (–5.120)**** (–3.976)****

AFAMPCT 308.618 428.030(0.557) (1.730)**

LATINPCT 275.380 512.465(0.442) (1.774)**

Us able ob ser va tions 2,538 2,538 2,538 2,538Log like li hood –4553.5 –2234.7 –4554.1 –2234.3Pseudo R2 percentage 6.7 7.7 6.7 7.7

NOTE: Z sta tis tic in paren the ses.**p < .10. ***p < .05. ****p < .01.



The sec ond caveat is related. Although we have pre sented evi dence for ineq ui ta ble prox im ity,this can not sim ply be inter preted to show an orches trated effort to bur den poor or minor ity com mu -ni ties with haz ard ous emis sions. Unlike TSDFs, TRI facili ties can be (and are) located almost any -where, because the siting of manu fac tur ing opera tions is sub ject only to local zon ing restric tions.Thus, a regional pat tern of dis pro por tion ate prox im ity may not involve an insti tu tion al ized orinten tional effort to bur den spe cific com mu ni ties with spe cific haz ards but rather a gen eral prob -lem with how indus trial zon ing is deter mined in dif fer ent com mu ni ties.27 At the same time, weshould stress that our results for MINOR ITY show up even when we con trol for land use, sug gest -ing that addi tional fac tors may be at play.

In terms of pol icy impli ca tions, we should stress that the pat terns revealed here can have dis-e qual iz ing effects on eco nomic devel op ment: If haz ards are dis pro por tion ately dis trib uted, anddevel op ers seek to avoid politi cal chal lenges, eco nomic activ ity will be slower in affected areas.The fed eral gov ern ment and numer ous states have rec og nized the prob lem and launched vari ous“brown fields” ini tia tives, designed to ease devel op ment on sites sus pected of poten tial con tami na -tion.28 This approach can offer a win- win sce nario: Envi ron men tal stan dards can be upheld, minor itycom mu ni ties can gain jobs, and the appro pri ate sub si dies and aid in steer ing through the regu la torymaze can lower devel oper costs.29

In con sid er ing other pol icy options, par ticu larly those that may impose costs or slow devel op -ment, we are aware of the need for cau tion. Debate remains about the empiri cal evi dence for “envi -ron mental racism,” and we respect those who have chal lenged pro po nents of the con cept to gobeyond anec dotes and sim ple bivari ate cor re la tions. Still, the tech niques used in this study are rela -tively sophis ti cated, and although evi dence for dis pro por tion ate prox im ity to TRI air releases insouth ern Cali for nia is some what mixed, it is gen er ally con sis tent with a grow ing body of evi dencefor many regions of the United States.

We there fore would sug gest fol low ing the doc tor’s adage: “At least do no harm.” New TRI orsimi lar air releases should, at a mini mum, not con trib ute to wors en ing the pat tern of dis pro por tion -ate prox im ity. To imple ment this, for mal assess ment of envi ron mental/health impacts related tozon ing changes or siting deci sions could include an analy sis of whether the pro posed action aggra -vates exist ing ineq ui ties. Such a prac tice is already being intro duced into the air- permitting process in the state of Michi gan.30 When health may be threat ened with every breath you take, cau tion cutsboth ways: Although we may wish to avoid hast ily enact ing a new and inef fi cient bureauc racy forplan ning and regu la tion, we should also be work ing to pro tect the health of all resi dents.

NOTES

1. Ander ton, Ander son, Oakes and Fraser (1994); Ander ton, Ander son, Rossi et al. (1994); and Yan dle and Bur ton(1996) are the excep tions, but these stud ies have been criti cized for meth odo logi cal short com ings (Bark en bus, Peretz, &Rubin, 1996; Been, 1995; Boer, Pas tor, Sadd, & Sny der, 1997; Bullard, 1996; Mohai, 1996; Pol lock & Vit tes, 1995).

2. Toxic Release Inven tory (TRI) releases have been the focus of some stud ies (Bowen, Salling, Hay nes, & Cyran,1995; Burke, 1993; Glick man & Hersh, 1995; Pol lock & Vit tes, 1995; Ringquist,1997; Wer ner, 1997), and we draw on boththese and the treat ment, stor age, and dis posal facili ties (TSDFs) stud ies for appro pri ate meth ods.

3. The Toxic Chemi cal Release Inven tory Sys tem (TRIS) data base, avail able from the U.S. Envi ron men tal Pro tec tionAgency (USEPA/NTIS), iden ti fies all report ing facili ties that release toxic chemi cals into the air, water, land, under groundinjec tion, and off- site trans fers in report able quan ti ties under Super fund Amend ments and Reau thori za tion Act (SARA)Title III, Sec tion 313. Although more recent ver sions are now avail able, the 1992 gen era tion of the TRIS (date of gov ern -ment ver sion, Decem ber 31, 1992) was used, because it bal anced improved cov er age with a col lec tion date closer to those of the other data sets used in this analy sis, par ticu larly the 1992 land use data and the 1990 tract- level demo graph ics from thecen sus.

4. There may be a sys tem atic down ward bias in the reported releases, because there are few incen tives for a facil ity toover re port its chemi cal out put (Szasz & Meuser, 1997). This sug gests that the coef fi cients result ing from our vari ous regres -sion analy ses may be under stated.

5. Local resi dents are also sub ject to poten tial expo sure as a result of their use of local schools, parks, and com mer cialsites, and as they drive, walk, or social ize in the gen eral neigh bor hood. Trans port of haz ard ous sub stances within the neigh -bor hood and the risk of acci dents related to fire, earth quake, or acci den tal release also increase expo sure poten tial.

6. These results par al lel those obtained in tests of occu pa tional expo sure (Dan iell & Vaughn, 1988; Olsen, Hem minki, &Ahl borg, 1990; Tay lor, Stelma, & Law rence, 1989).




7. The Occu pa tional Safety and Health Administration clas si fi ca tion is based on car cino genic clas si fi ca tions fromthree sources: the National Toxi col ogy Pro gram (NTP), the Inter na tional Agency for Research on Can cer (IARC), and 29CFR 1910, Sub part Z, Toxic and Haz ard ous Sub stances.

8. In 1992, TRI facili ties reported air releases total ing 15,206 tons in the study area. Of this total, 15% were car cino genreleases, and 17% were chemi cals in the 33/50 pro gram.

9. Per cent age of non- White resi dents in a cen sus tract, cal cu lated by sub tract ing non- Hispanic Whites from total per -sons and divid ing by total per sons, with data taken from the 1990 Cen sus, Sum mary Tape File 1 (STF1).

10. Popu la tion den sity is cal cu lated by divid ing 1990 STF1 popu la tion data by the tract area, as given in the 1992 U.S.Cen sus TIGER data files. An edu ca tional vari able was sug gested by an anony mous refe ree, con cerned that pat terns of dis -pro por tion ate prox im ity to haz ards might sim ply “reflect the rela tively low human capi tal reserves of the resi dents.” Weunder stood this to reflect the notion that more edu cated indi vidu als would make wiser or less con strained choices regard ingresi den tial loca tion. Edu ca tional vari ables (per cent age col lege edu cated or per cent age with a high school diploma) are sig -nifi cant in uni vari ate con sid era tions. How ever, in the mul ti vari ate logit, ordered logit, and tobit regres sions, these vari ablesexhibit sign insta bil ity and mixed lev els of sig nifi cance and do not affect the sig nifi cance pat tern of the other vari ables.These results may be partly because of col line ar ity with other key vari ables such as income, employ ment in manu fac tur ing,and per cent age minor ity.

11. Results are simi lar if we use a half- mile “buffer” zone around release sites. This method of aggre gat ing cen sus tractsby dis tance buff er ing around point loca tions for the haz ard sites is supe rior to the meth ods used by Ander ton, Ander son,Oakes et al. (1994); Ander ton, Ander son, Rossi et al. (1994); Been (1995); and oth ers who aggre gate adjoin ing tracts ortracts sur round ing the tract that con tains the site; the lat ter process implic itly (and incor rectly) assumes that the site islocated in the cen ter of the affected tract. See Pol lock and Vit tes (1995), Bowen et al. (1995), Glick man and Hersh (1995),and Wer ner (1997) for alter na tive tech niques of spa tial aggre ga tion.

12. The t test can be per formed under the assump tion of equal or une qual vari ances in the two sam ples. We chose toassume the sam ple vari ances were une qual when the F sta tis tic from a Levine test was less than 0.05. The Wilcoxon- rankedsum test was also used in our pre vi ous work on Los Ange les County (Boer et al., 1997) and in Ander ton, Ander son, Oakes et al. (1994).

13. This pat tern is simi lar to that obtained in our analy sis of TSDFs in Los Ange les County (Boer et al., 1997).14. For exam ple, although about 9% of the total popu la tion of the study area is Afri can Ameri can, more than half of this

popu la tion lives in South Cen tral Los Ange les, an area con sist ing of less than one- half of 1% of the entire study area.15. That “expo sure” for Afri can Ameri cans exceeds that of Anglos is also due to the clus ter ing of Afri can Ameri cans in

the Los Ange les met ro poli tan region.16. We also drop the hous ing meas ures for tech ni cal rea sons: To deter mine actual resi den tial land value requires weight -

ing house val ues and rents—which, in turn, requires using means that are unfor tu nately dis torted because of the cen susprac tice of cap ping house val ues and rents at a cer tain level. We dis cuss this prob lem in detail in Boer et al. (1997); an addi -tional dif fi culty is con trol ling for hous ing size.

17. Although most pre vi ous lit era ture on the dis tri bu tion of envi ron mental haz ards tends to focus on the t meas ure, theWald is more appro pri ate in large sam ples esti mated using gen er al ized least squares tech niques (as with a logit). In any case, t sta tis tics cal cu lated in the usual fash ion fol low the same pat tern of sig nifi cance. Note that Wald sta tis tics, being based onthe chi- squared dis tri bu tion, are gen er ally larger in their numeric value than are t sta tis tics for the same level of sta tis ti calsig nifi cance.

18. This pat tern holds even if we reduce the radius to a tighter 0.5 miles. We do not report those results here to con servespace.

19. This is con sis tent with our pre vi ous work on TSDFs (Boer et al., 1997), in which popu la tion den sity per forms poorly in a logit when included with indus trial land use. As a result, we believe that much pre vi ous research is flawed, in that popu -la tion den sity is merely a proxy for land use; in fact, many of the resi den tial areas buff er ing haz ards are actu ally moredensely popu lated than else where in the region (see also Pol lock & Vit tes, 1995).

20. The rela tion ships between the inde pend ent and depend ent vari ables are non lin ear, and as a result, the mar ginal effect of an increase of 10% from the study area mean will dif fer from the mar ginal effect of an increase of 10% from zero or fromtwice the study area mean. For exam ple, an increase in the per cent age minor ity from 10% to 20% pro duces a mar ginal effect of 3.5%, whereas an increase from 80% to 90% yields an effect of 2.6%. The num bers in brack ets in Tables 3 and 4 reflectonly the per cent age change in prob abil ity when the vari able is increased 10% above the study area mean. Note fur ther thatinter pret ing the mar ginal effect of an increase in per cap ita income can be con fus ing. First, the mar ginal effect recorded forPERCAPIN cap tures the total impact of an incre mental increase in income and thus includes changes in PERCAPIN2. Sec -ond, because the func tional rela tion ship between per cap ita income and prob abil ity of a TRI release is an inverted U-shape,an incre mental increase in income can have either a posi tive or nega tive effect on prob abil ity, depend ing on the ini tialincome.

21. The lower bound prob abil ity is con structed using the low est per cent age minor ity, indus trial land use, and per cent age manu fac tur ing employ ment along with the high est per cap ita income level. The upper bound is con structed using the high -est per cent age minor ity, indus trial land use, and manu fac tur ing employ ment with the low est per cap ita income.

22. The pre dicted prob abil ity for Pro file A being within a mile of a 33/50 release is 6% to 34% with an actual inci denceof 21%, whereas the pre dicted prob abil ity for Pro file B is 83% to 95% with an actual inci dence of 100%.

23. Ringquist (1997) does try to take into account degree of tox ic ity by con sid er ing the number of facili ties in a trun -cated event count model. He also con sid ers the amount of pol lut ants with a weighted least squares model; we think the tobitto be more appro pri ate to the task. The unit of analy sis in that arti cle is the zip code, a scale some have rejected as being toobroad and non uni form (see Ander ton, Ander son, Oakes et al., 1994; Ander ton, Ander son, Rossi et al., 1994).




24. We do not attempt to cal cu late mar ginal effects in the case of mov ing from one sort of haz ard to another becausethere is less meth odo logi cal prece dent on how to cal cu late such dis con tinu ous jumps over a range of haz ard.

25. The only exam ple of a tobit analy sis on envi ron mental haz ards that we could find is Krie sel, Cent ner, and Kee ler(1996). These authors also use pounds of release from the TRI as their depend ent vari able; although they use a smaller geo -graphic unit (the cen sus block group) than do we, they attrib ute pounds of release to all block groups within a mile of the site, stretch ing the radius beyond our own approach in the tobit regres sions. In any case, with their sam ples drawn from Geor giaand Ohio, they also find that race mat ters in a mul ti vari ate set ting.

26. One of the key com mu nity groups involved—Citi zens of South Cen tral—had ear lier man aged to derail a trash- to- energy incin era tor in South Los Ange les. See Dolan (1991) and Koe nenn (1991).

27. The rela tion ship between indus trial zon ing, low- income hous ing, and envi ron mental haz ards is noted in Star key(1994). How ever, some authors con tend that efforts to bur den minor ity com mu ni ties may have taken place in some cases(Hay nes, 1997; Pulido, Sidawi, & Vos, 1996).

28. See Con nolly and Dad dario (1995), Jacob son (1996), and Peirce (1996). We should note that brown fields refer tosites in which the land is con tami nated, not in which air releases take place; we intro duce the brown fields con cept here as ananal ogy of how local pol lu tion can make rede vel op ment dif fi cult.

29. In the words of one Chi cago envi ron mental offi cial, brown fields rede vel op ment is a “very con struc tive way to workwith envi ron mental jus tice issues since so many of these sites are in minor ity areas” (see Ruben, 1995, p.13).

30. See “Judge Rules Michi gan” (1997).

REF ER ENCES

Andeola, F. O. (1994). Envi ron men tal haz ards, health, and racial ineq uity in haz ard ous waste dis tri bu tion. Envi ron ment and Behav ior, 26, 99-126.

Ander ton, D. L., Ander son, A. B., Oakes, J. M., & Fraser, M. R. (1994). Envi ron men tal equity: The demo graph ics of dump -ing. Demog ra phy, 31, 229-248.

Ander ton, D. L., Ander son, A. B., Rossi, P. H., Oakes, J. M., Fraser, M. R., Weber, E. W., & Cal abrese, E. J. (1994). Haz ard -ous waste facili ties: “Envi ron men tal equity” issues in met ro poli tan areas. Evalua tion Review, 18, 123-140.

Bark en bus, J. N., Peretz, J. H., & Rubin, J. (1996). More on the agenda. Social Sci ence Quar terly, 77, 516-519.Been, V. (1995). Ana lyz ing evi dence of envi ron mental jus tice. Jour nal of Land Use and Envi ron men tal Law, 11, 1-37.Bell, B. P., Franks, P., Hil dreth, N., & Melius, J. (1991). Meth yl ene chlo ride expo sure and birth weight in Mon roe County,

New York. Envi ron men tal Research, 55(1), 31-39.Berry, M., & Bove, F. (1997). Birth weight reduc tion asso ci ated with resi dence near a haz ard ous waste land fill. Envi ron -

men tal Health Per spec tives, 105(8), 856-861.Boer, J. T., Pas tor, M., Sadd, J. L., & Sny der, L. D. (1997). Is there envi ron mental racism? The demo graph ics of haz ard ous

waste in Los Ange les County. Social Sci ence Quar terly, 78(4), 793-810.Bowen, W. M., Salling, M. J., Hay nes, K. E., & Cyran, E. J. (1995). Toward envi ron mental jus tice: Spa tial equity in Ohio

and Cleve land. Annals of the Asso cia tion of Ameri can Geog ra phers, 85(4), 641-663.Bullard, R. (1990). Dump ing in Dixie: Race, class and envi ron mental qual ity. Boul der, CO: West view.Bullard, R. (1996). Envi ron men tal jus tice: It’s more than waste facil ity siting. Social Sci ence Quar terly, 77, 493-499.Burke, L. M. (1993, Octo ber). Race and envi ron mental equity: A geo graphic analy sis in Los Ange les. GeoInfo Sys tems,

44-50.Clark, L. (1997, Octo ber). Con tested rep re sen ta tions: The debate sur round ing evi dence of envi ron mental racism in the

United States. Pro ceed ings from the 1997 Envi ron men tal Jus tice Con fer ence, Global Eth ics for the 21st Cen tury, Uni -ver sity of Mel bourne, Mel bourne, Aus tra lia.

Con nolly, K., & Dad dario, D. (1995). How to find the green in your city’s brown fields. Ameri can City & County, 110(12),28-31.

Croen, L., Shaw, G., San bon matsu, L., Sel vin, S., & Buf fler, P. (1997). Mater nal resi den tial prox im ity to haz ard ous wastesites and risk for selected con geni tal mal for ma tions. Epi de mi ol ogy, 8(4), 347-354.

Dan iell, W. E., & Vaughn, T. L. (1988). Pater nal employ ment in solvent- related occu pa tions and adverse preg nancy out -comes. Brit ish Jour nal of Indus trial Medi cine, 45(3), 193-197.

Dolan, M. (1991, May 24). Gar den Grove Com pa ny’s bid for toxic waste incin era tor aban doned. Los Ange les Times, p. B11.

Exec. Order No. 12898, Fed eral actions to address envi ron mental jus tice in minor ity popu la tions and low- income popu la -tions (1994, Feb ru ary 11).

Gelob ter, M. R. (1992). Toward a model of “envi ron mental dis crimi na tion.” In B. Bry ant & P. Mohai (Eds.), Race and theinci dence of envi ron mental haz ards: A time for dis course (pp. 64-81). Boul der, CO: West view.

Glick man, T. S., & Hersh, R. (1995). Evalu at ing envi ron mental equity: The impacts of indus trial haz ards on selectedgroups in Alle gheny County, Penn syl va nia (Dis cus sion Paper No. 95-13). Wash ing ton, DC: Resources for the Future.Avail able: http://www.rff.org.dpa pers/abstract/9513.htm

Gold man, L. R., Pai gen, B., Mag nant, M. M., & High land, J. H. (1985). Low birth weight, pre ma tur ity, and birth defects inchil dren liv ing near the haz ard ous waste site. Haz ard ous Waste Haz ard ous Mate ri als, 2(2), 209-223.

Guthe, W. G., Tucker, R. K., Mur phy, E. A., Eng land, R., Ste ven son, E., & Luck hardt, J. C. (1992). Reas sess ment of leadexpo sure in New Jer sey using GIS tech nol ogy. Envi ron men tal Research, 59, 318-325.




Hay nes, R. C. (1997). The road to jus tice. Envi ron men tal Health Per spec tives, 105(9), 920-922.Jacob son, L. (1996, Novem ber). Out here, the rough is really rough. National Jour nal, 28(47), 2251-2252.“Judge rules Michi gan air per mit ting pro gram offers inade quate pro tec tion.” (1997, June 6). Inside EPA, p. 16.Ken nedy, P. (1992). A guide to econo met rics. Cam bridge, MA: MIT Press.Knox, E. G., & Gil man, E. A. (1997). Haz ard prox imi ties of child hood can cers in Great Brit ain from 1953-80. Jour nal of

Epi de mi ol ogy and Com mu nity Health, 51, 151-159.Koe nenn, C. (1991, Decem ber 23). South- Central stops an incin era tor. Los Ange les Times, p. E4.Krie sel, W., Cent ner, T. J., & Kee ler, A. G. (1996). Neigh bor hood expo sure to toxic releases: Are there racial ine quali ties?

Growth and Change, 27, 479-499.Lam bert, T., & Boer ner, C. (1994). Let ter to the edi tor. Envi ron ment, 36, 2-3.Lav elle, M., & Coyle, M. (1992, Sep tem ber 21). Une qual pro tec tion—the racial divide in envi ron mental law: A spe cial edi -

tion. National Law Jour nal, S2-S6.Mohai, P. (1996). Envi ron men tal jus tice or ana lytic jus tice? Reex am in ing his tori cal haz ard ous waste land fill siting pat terns

in met ro poli tan Texas. Social Sci ence Quar terly, 77, 500-507.Mohai, P., & Bry ant, B. (1992). Envi ron men tal racism: Review ing the evi dence. In B. Bry ant & P. Mohai (Eds.), Race and

the inci dence of envi ron mental haz ards: A time for dis course (pp. 163-176). Boul der, CO: West view.New York State Parks Man age ment and Research Insti tute. (1993). Loca tional veri fi ca tion and cor rec tion of 1989 New

York State TRI facili ties (Final report pre pared for the U.S. Envi ron men tal Pro tec tion Agency, Office of Toxic Sub -stances). New York: Author.

Nord strom, S., Beck man, L., & Nord strom, I. (1978). Occu pa tional and envi ron mental risks in and around a smelter innorth ern Swe den, I. Varia tions in birth weight. Here di tas, 88, 43-46.

Olsen, J., Hem minki, K., & Ahl borg, G. (1990). Low birth weight, con geni tal mal for ma tions, and spon ta ne ous abor tionsamong dry clean ing work ers in Scan di na via. Scan dana vian Jour nal of Work and Envi ron men tal Health, 16(3),163-168.

Ozonoff, D., Col ten, M. E., & Cup ples, T. (1983). Sil re sim area health study report of find ings. Bos ton: Mas sa chu settsDepart ment of Envi ron men tal Qual ity Engi neer ing and Depart ment of Pub lic Health.

Peirce, N. (1996). New life for lightly pol luted lands. National Jour nal, 28(37), 1970.Pol lock, P. H., & Vit tes, M. E. (1995). Who bears the bur dens of envi ron mental pol lu tion? Race, eth nic ity, and envi ron -

mental equity in Flor ida. Social Sci ence Quar terly, 76, 294-310.Pulido, L., Sidawi, S., & Vos, R. O. (1996). An archae ology of envi ron mental racism in Los Ange les. Urban Geog ra phy,

17(5), 419-439.Ringquist, E. J. (1997). Equity and the dis tri bu tion of envi ron mental risk: The case of TRI facili ties. Social Sci ence Quar -

terly, 78(4), 811-829.Ruben, B. (1995). Fields of dreams? Envi ron men tal Action Maga zine, 26(4), 12-17.Shaw, G. M., Schul man, J., Frisch, J. D., Cum min, S. K., & Har ris, J. A. (1992). Con geni tal mal for ma tions and birth weight

in areas with poten tial envi ron mental con tami na tion. Archives of Envi ron men tal Health, 47(2), 147-154.Star key, D. (1994). Envi ron men tal jus tice: Win, lose or draw? State Leg is la tures, 20(3), 27-30.Szasz, A., & Meuser, M. (1997). Envi ron men tal ine quali ties: Lit era ture review and pro pos als for new direc tions in research

and the ory. Cur rent Soci ol ogy, 45(3), 99-120.Tay lor, P. R., Stelma, J. M., & Law rence, C. E. (1989). The rela tion ship of poly chlori nated biphen yls to birth weight and

ges ta tional age in the off spring of occu pa tion ally exposed moth ers. Ameri can Jour nal of Epi de mi ol ogy, 129(2),395-406.

United Church of Christ, Com mis sion for Racial Jus tice (UCC). (1987). Toxic wastes and race in the United States: Anational report on the racial and socio- economic char ac ter is tics of com mu ni ties with haz ard ous waste sites. New York: Pub lic Data Access.

U.S. Coun cil on Envi ron men tal Qual ity, U.S. Depart ment of Agri cul ture, U.S. Depart ment of Energy, and U.S. Envi ron -men tal Pro tec tion Agency. (1980). Pub lic opin ion on envi ron mental issues: Results of a national pub lic opin ion sur vey.Wash ing ton, DC: Gov ern ment Print ing Office.

U.S. Depart ment of Hous ing and Urban Devel op ment (HUD). (1998). The effects of envi ron mental haz ards and regu la tionon urban rede vel op ment. Wash ing ton, DC: HUD Office of Pol icy Devel op ment and Research.

U.S. Envi ron men tal Pro tec tion Agency (USEPA). (1994). 1992 toxic release inven tory pub lic data release (U.S. Envi ron -men tal Pro tec tion Agency, Office of Pol lu tion Pre ven tion and Toxics Report No. 7408). Wash ing ton DC: Author.

U.S. Gen eral Account ing Office. (1983). Siting of haz ard ous waste land fills and their cor re la tion with racial and eco nomicstatus of sur round ing com mu ni ties. Wash ing ton, DC: Author.

Wer ner, R. J. (1997). Toxic releases and demog ra phy in Min nea po lis/St.Paul: A GIS explo ra tion. Pro ceed ings from the1997 Envi ron men tal Sys tems Research Insti tute Users Con fer ence, San Diego, Cali for nia. Avail able: http://www.esri.com/base/com mon/user conf/proc97/PROC97/TO150/PAP122/P122.HTM

Yan dle, T., & Bur ton, D. (1996). Reex am in ing envi ron mental jus tice: A sta tis ti cal analy sis of his tori cal haz ard ous wasteland fill siting pat terns in met ro poli tan Texas. Social Sci ence Quar terly, 77, 477-492.




economic development quarterly - duke...

Documents