- ANDERTON, D.L., ANDERSON, A.B., OAKES, J.M., and FRASER, M.R. (1994b). “Environmental Equity: The demographics of dumping.” Demography 31:229–248.
- ANDERTON, D.L., ANDERSON, A.B., ROSSI, P.H., OAKES, J.M., FRASER, M.R., WEBER, E.W., and CALABRESE, E.J. (1994a). “Hazardous waste facilities. ‘Environmental Equity’ issues in metropolitan areas.” Evaluation Rev. 18:123–140.
- ARC/INFO® (1991). Version 6. Environmental Systems Research Institute, Inc. Redlands, CA.
- BEEN, V. (1994). “Unpopular neighbors: Are dumps and landfills sited equitably?” Resources Spring:16–19.
- BULLARD, R.D. (1983). “Solid waste sites and the Black Houston community.” Sociol. Inq. 53(2/3):273–288.
- BULLARD, R.D. (1994a). “Environmental justice for all.” In: Unequal Protection: Environmental Justice and Communities of Color (R.D. Bullard, ed.). Sierra Club Books, San Francisco, CA. pp. 3–22.
- BULLARD, R.D. (1994b). Dumping in Dixie: Race, Class, and Environmental Quality. Westview, Boulder, CO.
- ENVIRONMENTAL PROTECTION AGENCY (EPA) (1987). Superfund: Looking Back, Looking Ahead. Environmental Protection Agency, Washington, DC.
- ENVIRONMENTAL PROTECTION AGENCY (EPA) (1993a). National Priorities List Boundaries [computer file]. Environmental Protection Agency, Washington, DC.
- ENVIRONMENTAL PROTECTION AGENCY (EPA) (1993b). U.S. Block Polygon Coverages [computer file]. Environmental Protection Agency, Washington, DC.
- GENERAL ACCOUNTING OFFICE (GAO) (1983). Siting of Hazardous Waste Landfills and Their Correlation with Racial and Economic Status of Surrounding Communities. GAO, Washington, DC.
- GLICKMAN, T.S. (1994). “Measuring environmental equity with geographical information systems.” Resources Summer(116):2–6.
- GREENBERG, M. (1993). “Proving environmental inequity in siting locally unwanted land uses.” Risk Summer:235–252.
- HIRD, J.A. (1993). “Environmental policy and equity: The case of superfund.” J. Policy Anal. Management 12(2):323–343.
- LAVELLE, M. (1992a). “The minorities equation.” Natl. Law J. September 21:S2.
- LAVELLE, M. (1992b). “Examining EPA’s scoring system.” Natl. Law J. September 21:S6.
- MASSEY, D.S. and DENTON, N.A. (1987). “Trends in the residential segregation of Blacks, Hispanics, and Asians: 1970–1980.” Am. Sociol. Rev. 52(December):802–825.
- MOHAI, P. and BRYANT, B. (1992). “Race, poverty and the distribution of environmental hazards: Reviewing the evidence.” Race, Poverty Environ. 6:3, 24–27.
- MYERS, D. (1992). Analysis with Local Census Data. Academic Press, Boston, MA. pp. 207–234.
- NATIONAL LAW JOURNAL (1992). “Methodology. Computing the patterns.” Natl. Law J. September 21:S4.
- REINHARDT, J.R. (1989). “Summary of Resource Conservation and Recovery Act legislation and regulation.” In: Standard Handbook of Hazardous Waste Treatment and Disposal (H.M. Freeman, ed.). McGraw-Hill, New York, NY. pp. 1.9–1.27.
- SAS INSTITUTE INC. (1988). SAS/STAT User’s Guide, Release 6.03 Edition. Cary, NC. pp. 549–640.
- STANFORD, R. and YANG, E.C. (1989). “Summary of CERCLA legislation and regulations and the EPA Superfund Program.” In: Standard Handbook of Hazardous Waste Treatment and Disposal (H.M. Freeman, ed.). McGraw-Hill, New York, NY. pp. 1.29–1.45.
- U.S. BUREAU OF THE CENSUS (1990a). TIGER: The Coast-to-Coast Digital Map Data Base. The Bureau, Washington, DC.
- U.S. BUREAU OF THE CENSUS (1990b). Public Law 94–171 [machine-readable data files]. The Bureau, Washington, DC.
- U.S. BUREAU OF THE CENSUS (1991a). TIGER/Line Census Files on CD-ROM [machine-readable data files]. The Bureau, Washington, DC.
- U.S. BUREAU OF THE CENSUS (1991b). Summary Tape File 1 Technical Documentation/prepared by the Bureau of the Census. The Bureau, Washington, DC. pp. B1–B21.
- U.S. CODE OF FEDERAL REGULATIONS (1994). 40 CFR Part 300. December 16.
- UNITED CHURCH OF CHRIST, COMMISSION FOR RACIAL JUSTICE (UCC) (1987). Toxic Wastes and Race in the United States: A National Report on the Racial and Socioeconomic Characteristics of Communities with Hazardous Waste Sites. United Church of Christ, New York, NY.
- WASHINGTON POST (1995). “EPA to cut sites from Superfund. Properties cleared of waste are open to development.” February 4, E1.
- ZIMMERMAN, R. (1993). “Social equity and environmental risk.” Risk Analysis 13(6):649–666.
- ZIMMERMAN, R. (1994). “Issues of classification in environmental equity: How we manage is how we measure.” Fordham Urban Law J. 21(3):633–669.
JANET L. HEITGERD
Division of Health Assessment and Consultation
JE ANNE R. BURG
Division of Health Studies
HENRY G. STRICKLAND
Division of Health Assessment and Consultation
Agency for Toxic Substances and Disease Registry
Public Health Service U.S. Department of Health and Human Services
Atlanta, Georgia
Demographic studies used to investigate whether minorities are more likely to live near hazardous waste sites have resulted in varying conclusions. Some reasons for these inconsistencies may be due to the design of studies used to collect and compare demographic information. In the research reported here, a Geographic Information Systems (GIS) approach to characterizing total population, by race and Hispanic origin, for areas within a mile of 1,200 National Priorities List (NPL) sites across the United States, was used. An intra-county statistical comparison was made between racial and Hispanic origin subpopulations living within one mile of a site and the subpopulations living in the same county, but more than one mile from the site. These results show that the percentage of the population reporting in a minority category is higher in areas nearer the NPL sites.
1. Address all correspondence to: Janet L. Heitgerd, ATSDR/DHAC/PRB, 1600 Clifton Road, MS E-32, Atlanta, GA 30333. Tel.:(404)639-0600. Fax:(404)639-0653.
2. Abbreviations: ANOVA, analysis of variance; ATSDR, Agency for Toxic Substances and Disease Registry; CERCLA, Comprehensive Environmental Response, Compensation, and Liability Act; CERCLIS, Comprehensive Environmental Response, Compensation, and Liability Information System; GAO, General Accounting Office; GIS, Geographic Information Systems; MCD, Minor Civil Division; NPL, National Priorities List; RCRA, Resource Conservation and Recovery Act; TIGER, Topologically Integrated Geographic Encoding and Referencing; TSDF, treatment, storage, and disposal facility; UCC, United Church of Christ; EPA, U.S. Environmental Protection Agency.
3. Key Words: demographics, environmental equity, GIS, NPL.
We believe that a GIS approach is most appropriate for obtaining site-specific information and should be used as a tool in future demographic studies of areas near environmental hazards.
INTRODUCTION
The results of an expanding body of research are being used to document and compare the demographic makeup of areas near hazardous waste sites. These results often are used to address issues of environmental equity (also referred to as “environmental justice”), specifically the unequal environmental and public health impacts of environmental hazards on minority communities. It is argued that minority communities are burdened by unfair public policies, unequal protection against environmental risks, lax enforcement of environmental regulations and laws, delayed cleanups, and discriminatory siting practices for hazardous waste facilities (Bullard, 1994a). The latter issue, discriminatory siting practices, has generated much demographic analysis of areas near hazardous waste sites.
The prevailing opinion, supported by a few well-known studies (e.g., Commission for Racial Justice, United Church of Christ [UCC], 1987; Mohai and Bryant, 1992; Bullard, 1994a,b), is that hazardous waste sites are disproportionately located in minority communities. Other results derived from empirical research (e.g., Hird, 1993; Anderton et al., 1994a,b) suggest, however, that such a conclusion might not be warranted. Some reasons for the differing conclusions may be found in the design of studies used to collect demographic information. In this article, we review current research on the demographics near hazardous waste sites and offer a method that uses Geographic Information Systems (GIS) to estimate better and, therefore, to characterize better site-specific populations. The proposed method is demonstrated for 1,200 National Priorities List (NPL) hazardous waste sites located across the United States.
HAZARDOUS WASTE SITE DEMOGRAPHIC RESEARCH
Review of Past Studies
The majority of the published empirical work regarding population characteristics of areas near hazardous waste sites focuses on areas with regulated hazardous waste treatment, storage, and disposal facilities (TSDFs), and a selected subset of uncontrolled NPL waste sites. Relevant studies cited in the literature are discussed in this section and evaluated in the next section.
Treatment, storage, and disposal facilities. TSDFs are regulated under the Resource Conservation and Recovery Act (RCRA) of 1976. RCRA is a preventive regulatory policy that sets standards for performance and permit requirements, and attempts “cradle-to-grave” control of hazardous waste management (Reinhardt, 1989). Demographic research on TSDFs usually examines operating commercial facilities; these comprise approximately 11% of the facilities regulated by RCRA (Anderton et al., 1994b). One of the earliest reports on TSDFs (General Accounting Office [GAO], 1983) is based on a selection of four off-site hazardous waste landfills, two of which are located within commercial TSDFs. In the GAO study, the communities surrounding the landfills are spatially defined as census subdivisions of political jurisdictions and townships. Demographic comparisons are made between these areas and (1) the remaining county areas located within approximately four miles, (2) the county or counties in which those areas are located, and (3) the state(s). Results of the GAO study show that the majority of populations of the communities near three of the four sites investigated is Black. Additionally, the adjacent county areas (i.e., within four miles) are similar to the site communities in their Black population percentage.
Been (1994) updates the GAO report by examining population data corresponding to the approximate time of the landfill sitings. According to Been (1994), “all four host communities studied by the GAO were predominantly African American at the time they were selected …” and had a higher percentage of Blacks than the states in which the sites are located. After the landfill sitings, the percentage of Blacks in the host communities decreased.
Employing a sample survey design, Mohai and Bryant (1992) compared the percentage of “people of color” at varying distances from 16 TSDFs (i.e, within 1 mile, 1 to 1.5 miles, and more than 1.5 miles) in a three-county area surrounding the city of Detroit. Their research indicates that race is more important than income as a predictor of proximity to commercial hazardous waste facilities (Mohai and Bryant, 1992). National studies on the relationship between race and location of TSDFs in the United States have shown mixed results (UCC, 1987; Anderton et al., 1994a,b). The UCC (1987) study used 1980 census ZIP code data to compare populations residing in areas with TSDFs to those that do not reside in such areas. The results of the report (UCC, 1987) “suggest the existence of clear patterns which show that communities with greater minority percentages of the population are more likely to be the sites of such facilities.” Anderton et al. (1994a,b), however, find less-clear patterns of racial inequity in siting of TSDFs. Using 1980 census tract data, they find instead that the amount of industrial activity in a tract is a better predictor than race of TSDF location. They conclude that race is significant insofar as minorities are more likely to be located in the census tracts bordering the industrialized census tracts that actually contain the TSDFs.
CERCLIS and NPL sites. Congress established the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA, also referred to as Superfund) in 1980 to provide “EPA [U.S. Environmental Protection Agency] with authority to remedy releases of hazardous substances into all environmental media … that are the result of past practices as well as current emergencies” (Stanford and Yang, 1989). The EPA maintains a database of CERCLA sites known as the Comprehensive Environmental Response, Compensation, and Liability Information System (CERCLIS) that lists approximately 13,000 uncontrolled toxic waste sites, i.e., those facilities found to have releases (EPA, 1976). Sites are listed in CERCLIS in response to reports by individuals of abandoned sites or illegal dumping (EPA, 1987). Sources of the wastes found at those sites include the chemical and petroleum industries, abandoned municipal landfills, and transportation or other accidental contaminant spills (EPA, 1987).
In a descriptive study of 18,164 CERCLA sites located within residential ZIP code areas, the UCC (1987) found some evidence of environmental inequity as it pertains to the location of hazardous waste sites. For example, “three out of every five Black and Hispanic Americans lived in communities with uncontrolled toxic waste sites” (UCC, 1987). Additionally, “more than half the population in the United States live in residential ZIP code areas with one or more … sites” (UCC, 1987).
Not all sites listed in CERCLIS will require response action by the EPA. For example, in early 1995, the EPA, in consultation with states, removed 25,000 sites from CERCLIS because they determined that the sites no longer required federal involvement (Washington Post, 1995). One of the main objectives of Superfund is to set up priorities for responding to and cleaning up uncontrolled toxic waste sites. Sites considered to present the most serious potential threat to public health and the environment are placed on the EPA’s NPL and are eligible for cleanup using Federal dollars from Superfund (EPA, 1987). Three criteria must be met for placing a site on the NPL (U.S. Code of Federal Regulations, 1994). The first is that a site scores sufficiently high on the EPA hazard ranking system; the second is that a state gives the site top priority. To fulfill the third criterion, three conditions must be met: the Agency for Toxic Substances and Disease Registry (ATSDR) issues a health advisory at the site, the EPA determines that the release poses a significant threat to public health, and Superfund is likely to be the most cost-effective response.
A number of controversies surround the placement of sites on the NPL list as well as the question of what happens to sites once they are placed on the list (Lavelle, 1992b). The National Law Journal (1992) conducted a study of 1,177 NPL sites. The study methodology included spatially defining communities so that they corresponded to the ZIP code location of the NPL sites. Analysis of the data involved dividing the ZIP code locations into quartiles on the basis of their percentage of Whites and comparing the length of time it took for their placement on the NPL and the levels of cleanup in the top (i.e., White communities) and bottom (i.e., minority communities) quartiles. It was demonstrated that the length of time it takes to place a site on the NPL and the type of cleanup method chosen differ according to whether the site is in a minority area (Lavelle, 1992a). Differences also were seen for low-versus high-income communities, but not to the same extent.
Hird (1993) compared a number of population and socioeconomic indicators for U.S. counties with and without an NPL site. The study results indicate that NPL sites are located in about 20% of U.S. counties. The characteristics of counties containing NPL sites include a strong manufacturing base, fewer new housing developments, and a larger percentage of wealthy, well-educated, or nonwhite populations than in counties without an NPL site. Further analysis of counties with a relatively large percentage of minorities (i.e., greater than the national average for counties) finds that the number of NPL sites in those counties “is below the national average” (Hird, 1993). In addition, by shifting the focus of the analysis to explaining differences in the number of NPL sites in a county, results indicate that the percentage of nonwhites is inversely related to the number of NPL sites in the county.
NPL population characteristics have also been studied using data available for geographic areas smaller than the county. This can result in more site-specific population information. Zimmerman (1993) compared population and socioeconomic information for the census place or minor civil division (MCD) location of NPL sites with the four census designated regions (i.e., Northeast, Midwest, South, and West) and the nation. The study results show that the percentages of Blacks and persons of Hispanic origin are greater for those census places or MCDs with an NPL site than for the nation. Other relevant findings include little difference from the comparison areas with regard to the percentage of the population below the poverty level, and, as an outcome of the EPA hazard ranking system that weights population proximity, NPL sites are located in relatively densely populated areas.
EVALUATION OF PAST RESEARCH ON TSDFS AND NPL SITES
Only tentative statements may currently be made about the relationship between community racial structure and proximity to hazardous waste sites, whether TSDFs or NPL sites. That tentativeness is due partly to the studies being restricted geographically, having differing spatial definitions of proximity to a hazardous waste site, and having varying choices of comparison areas. The ability to conduct national studies (e.g., UCC, 1987; National Law Journal, 1992; Hird, 1993; Anderton et al., 1994a,b) on the population dynamics around hazardous waste sites is dependent on and often hampered by available resources (e.g., constraints on budget, accessibility of national databases, or computer resources for creating and processing large data sets). Small area studies may produce notable research results (e.g., Bullard, 1983; Mohai and Bryant, 1992; Been, 1994), but have limited generalizability. The studies by Anderton et al. (1994a,b) are the most current when considering incorporating site-specific demographic information, but more work in this area is required.
Inconsistency prevails across studies regarding the spatial definition of proximity to hazardous waste sites. Popular choices in previous research are ZIP codes (UCC, 1987; National Law Journal, 1992), counties (Hird, 1993), other political jurisdictions (GAO, 1983; Zimmerman, 1993), and census tracts (Anderton et al., 1994a,b). Defining proximity sometimes may be a function mainly of the availability of data and may not result in the most appropriate selection of areas for site-specific demographics. This raises the possibility that observed differences are artifacts of the area boundaries. For example, it has been argued that ZIP codes, being generally representative of larger areas than census tracts, “might capture certain social aspects of the residential structure of cities while obscuring differences within these areas relevant to the question of environmental equity” (Anderton et al., 1994b). Zimmerman (1994) explores in considerable detail the issue of the spatial definition of areas near hazardous waste sites and points out a number of problems involved in the use of traditional boundaries. For example, it is not unusual to find hazardous waste sites located at the perimeter of a political jurisdiction or even a census-defined statistical area (e.g., census tracts and block groups). As Zimmerman (1994) notes, “This implies that adjacent jurisdictions [or census areas] could be more impacted than the one in which a hazardous waste site is located.”
Finally, the choice of comparison areas or comparison populations might affect the study results. Past researchers have used the nation (Zimmerman, 1993), states (GAO, 1983; Been, 1994), ZIP codes (UCC, 1987; National Law Journal, 1992), counties or county areas (GAO, 1983; Hird, 1993), and census tracts (Anderton et al., 1994a,b; Been, 1994) as comparison areas. In addition, as opposed to using all other similarly measured areas in the United States without a waste site (e.g., Hird, 1993; Anderton et al., 1994a,b), investigators sometimes restrict the comparison areas to being within a certain distance from the waste site (e.g., GAO, 1983; Mohai and Bryant, 1992; Anderton et al., 1994a,b).
Although population data are readily available for different political and statistical areas, the appropriate choice of a comparison group for a study depends to a great extent on how proximity of the potentially exposed population to the hazardous waste site is measured. Other considerations, however, may reasonably inform the decision. For example, Greenberg (1993) discusses the choice of a comparison group in terms of defining areas of burden (i.e., areas with any locally unwanted land uses) and benefit (i.e., states, the United States, or other user-defined primary and secondary areas of benefit). He finds that the choice of geographic comparison areas “can make an enormous difference” in outcomes. Indeed, Zimmerman (1994) argues that “The choice of standards of comparison continues to be one of the more subjective, discretionary areas of environmental equity research.”
In this research, we present a GIS method that can be used to assess site-specific demographic characteristics of areas around hazardous waste sites. We use a cross-sectional design to estimate the total population, including breakdowns by race and Hispanic origin, at the census block level for areas within one mile of 1,200 NPL sites across the nation. To evaluate the degree to which NPL sites are located in minority areas, we also analyze population information for U.S. counties that have boundaries within a mile of one or more NPL site. We compare data on racial and Hispanic origin for the county area (i.e., census blocks) within a mile of the remaining county area.
Our research is motivated by the need for more site-specific information on populations living near hazardous waste sites. In addition, our research addresses the problem areas discussed previously by (1) gathering demographic information on a national sample of NPL sites, (2) creating a spatial definition of proximity of potentially impacted populations that is site-specific and less restricted by traditional geographic boundaries than has previously been the case, and (3) using a comparison area that is related logically to our proximity measure. We discuss the methods in more detail in the following section.
METHODS
Data Sources
Data sources for the spatial processing are the 1990 Census block boundaries for the United States, excluding Alaska and Hawaii, and ARC/INFO®1 coverages representing 1,200 EPA NPL site boundaries (EPA, 1993a,b). The 1990 Census block boundary coverages were extracted from the U.S. Bureau of the Census’ 1990 Topologically Integrated Geographic Encoding and Referencing (TIGER)/Line files (U.S. Bureau of the Census, 1991a). TIGER/Line files are a digital database describing geographic areas in the United States (U.S. Bureau of the Census, 1990a). These data were then linked with information on total population, race, and Hispanic origin taken from the 1990 Public Law 94–171 census block data (U.S. Bureau of the Census, 1990b).
The EPA annually updates the NPL, and the 1,200 site boundaries used in this study are based on the 1992 listing. The boundaries define the approximate geographic location and extent of the hazardous substance release as delineated by the EPA and do not necessarily correspond to other (e.g., land use or land ownership) boundaries (U.S. Code of Federal Regulations, 1994). As more site information about the release is obtained, the site boundaries may become smaller or larger. In addition, the site boundaries may not be identical with the remedy area, which is determined by the Record of Decision (U.S. Code of Federal Regulations, 1994).
To locate the site in a ZIP code or census tract area, past research (e.g., UCC, 1987) on populations living near hazardous waste sites often relied on specific information on a single site point, either the address of the facility or the CERCLIS latitude and longitude values. One problem with this approach is that the releases from a site may not be confined to a single ZIP code or census tract boundary. The power of GIS to integrate the EPA-defined site boundaries and census enumeration polygons at the census block level mitigates this problem.
Identification of Populations
Site population. ARC/INFO® was used to delimit spatially the population living near the Superfund sites. The potentially impacted site or “community” area was defined spatially as the area within the NPL site boundary plus a one-mile buffer around that boundary. A one-mile buffer is useful for conveying information about the demographic structure of the area adjacent to the site and identifies those who might have a greater potential for exposure. It is important to note, however, that proximity, in itself, is not an adequate indicator of exposure to site-related contaminants. Whenever possible, proximity measures must be supplemented with environmental pathway (e.g., groundwater flow) information to evaluate exposed populations.
1ARC/INFO® is a GIS software product. Use of trade names is for identification only and does not imply endorsement by the Agency for Toxic Substances and Disease Registry, the Public Health Service, or the U.S. Department of Health and Human Services.
When combining census blocks with buffer distances, the periphery of the obtained boundary commonly intersects a census block. In such cases, applying an area weight factor to the variables of interest is arguably the best choice for collecting demographic data on any one site. The total population in a census block would be weighted to reflect the proportion of the block area that was actually in the buffered site area. This method assumes that the population is uniformly distributed throughout the block.
A problem occurs using weighted population data for calculating statistics on multiple NPL sites, however, if there are instances of area overlap between the site buffer areas. The overlap can result in the counting of a census block and some or all of its population more than once. It would be a complex process to separate out the actual block area (and thus, population) shared by two or more NPL sites. Also, comparison areas would be difficult to specify, for a census block may be both part of the area “near” a waste site and adjacent to another waste site. Therefore, we chose not to use what we call the area weighting method. The method we used in this paper instead was a total count method; that is, if any part of a block was located within the one-mile buffer, that block and its population were included in the site total.
The impact of not using the previously described area weighting method to create the population numbers is difficult to assess. To the extent that the spatial size of census blocks is inversely related to population density, however, it is likely that relatively more land area is accounted for in rural site areas, but that relatively more population is added in urban site areas. Therefore, we may overenumerate population subgroups (e.g., Asians and Hispanics) more likely to settle in urban areas (Massey and Denton, 1987). In addition, larger site areas, with a greater periphery, will be affected more often than smaller ones. Figure 1 illustrates, at a sample NPL site, the differences in population counts between the area weighting method and the total count method. As can be seen, the nonweighted, or total count method, identifies a larger number of people and arrives at a different rate of minority populations (for example, the percentage of Blacks is 35, according to the weighted, and 37 according to the nonweighted method).
There were 670 counties (22% of all counties in the contiguous United States) that had a part of their total area located within a mile of the 1,200 NPL sites. The areas within one mile comprised 184,191 census blocks, of which 10.6% (19,423) were within one mile of two or more NPL sites. The site values were derived by summing up all census blocks within the one-mile range for each county (taking care not to include any census block more than once). This shifted the focus of the analysis from the NPL sites per se to the counties within one mile of those sites while retaining block level observation data. In short, the site population and the associated demographic information used in the comparison analyses is the sum of the census block data for each block that lies in total or in part within one mile of the site of interest.
Comparison population. The comparison population is defined spatially as those living in one of the 670 impacted counties but more than one mile from the site (referred to as the comparison area). Population data for these areas are obtained by subtracting the site area data from the county totals. Using the remainder of the county as the comparison population will minimize regional, state, and county differences in the population composition.
FIGURE 1. Sample NPL site population information.
Statistical Analyses
Demographic variables. For the population comparisons, the dependent variables used were the percentage reporting their race on the 1990 census as Black; as American Indian, Eskimo, or Aleut; or as Asian or Pacific Islander; and the percentage responding positively to the census question of Hispanic origin. The responses to the question of Hispanic origin are distinct from and independent of the response to the race question. For further explanation of the census items denoting race and Hispanic origin, see Myers (1992) or refer to the 1990 Census technical documentation (U.S. Bureau of the Census, 1991b).
Statistical methods. The hypotheses of equal percentages (percentage one mile from site(s) compared with remainder of county for the 670 counties where the 1,200 NPL sites were located) of Blacks; American Indians, Eskimos, or Aleuts; Asians or Pacific Islanders; Whites; and persons of Hispanic origin between the site and comparison populations were tested.
A three-factor analysis of variance (ANOVA) model was used. The factors used were: NPL (two levels — within or outside one-mile buffer of NPL site), State (48 contiguous states), and County (670 counties). A mixed-effects model with the factor County nested within State was used. The factor County-within-State was considered the appropriate level of comparison; also, sites within states are expected to be more racially and ethnically homogeneous than sites in different states. The percentages of racial groups or groups of Hispanic origin were used as the dependent variable. The ANOVA distributional assumptions were tested; a transform of the data (arcsine for % whites, log10 for other dependent variables) resulted in a closer approximation to the assumptions. For computational purposes, the zero percentage values were replaced with the value 10-13. The analyses were executed using the general linear model procedure (mixed-effects model) (SAS, 1988). Interaction terms were included in the model and, if significant, retained, and the denominator for the F-test was modified accordingly.
Because racial and ethnic origin groups are not distributed equally throughout the United States, but are concentrated in certain states, we also analyzed a subset of states for each of the racial and ethnic origin groups. States were included in a specific analysis, if the population percentage for a select race or ethnic origin was equal to or greater than the population percentage for the nation. The ANOVAs were repeated for these subsets of states.
RESULTS
Population Summaries
The site population is defined as those living in the 670 counties and within one mile of the identified 1,200 NPL hazardous waste sites (referred to as the site areas). This includes portions of about 22% of all counties in the contiguous United States that contain over 60% of the nation’s total population (see Table 1). Figure 2 shows the locations of the impacted counties by state and EPA region. Table 1 presents summary data for the total U.S., site, and comparison populations. Appendix 1 provides descriptive information — number of sites and persons, together with racial, and Hispanic origin composition of site subpopulations — by state and EPA region.
Approximately 11 million people live within a one-mile buffer of one or more NPL sites in the United States. EPA Region IX, primarily California, accounts for almost one quarter of the total number of people, followed by EPA Regions II (Northeast), V (Midwest), and IV (Southeast). Although more NPL sites are found in New Jersey (EPA Region II) and Pennsylvania (EPA Region III), the California NPL sites are located in relatively more densely populated areas. California ranks highest in the categories of “average population per site” and “average population density.” The state ranks eighth in average site area per square mile. Fewer NPL sites are found in the Great Plains states (i.e., states in Regions VI, VII, and VIII), and those sites also account for relatively fewer people.
The racial and Hispanic origin distributions around the NPL sites are generally in line with regional population differences. As might be expected, the percentage of Blacks around NPL sites is largest in the Southeast (i.e., Region IV, excluding Kentucky) and, to some extent, the Southwest (e.g., Region VI, excluding New Mexico). Other areas with a relatively high percentage of Black populations near NPL sites are found in EPA Region III (Delaware, Maryland, and Virginia) and in Kansas, located in Region VII.
With the exception of Oklahoma, a relatively small percentage of American Indians, Eskimos, or Aleuts are located around NPL sites. There is also a relatively small percentage of Asians or Pacific Islanders located around NPL sites, except in California. As noted before, Alaska and Hawaii are excluded from the analysis.
The percentage of persons of Hispanic origin near NPL sites is high in Regions VI (New Mexico and Texas), VIII (Colorado), and IX (California). Large percentages are also found in New Jersey and Florida. These are the states that, in general, include relatively large numbers and percentages of people of Hispanic origin.
Population Comparisons
The racial composition and percentages of people of Hispanic origin within the site population (those living within one mile of the site) and the comparison population (those located in counties within one mile of an NPL site, but living in blocks farther than one mile from the site) were compared. A discussion of the statistical results follows.
Summary statistical data for the site and comparison populations are shown in Table 2 and the ANOVA results are shown in Table 3. The results of the ANOVA indicate that there is a significant difference in the mean percentage (log10 transformed) for each of the racial and Hispanic origin groups (p = < 0.001) between the population living within one mile of a site and the population living in the remainder of the county, when the 1,200 NPL sites are controlled for state and county of site. No statistical differences were found for percentage (arcsine transformed) of whites (p = 0.21). It is of interest that sites within counties did not differ for within states for the racial group American Indian, Eskimo, or Aleut or the Hispanic origin group, and was borderline significant for the racial groups Asian or Pacific Islander. These results are in keeping with what could have been predicted from the aggregated results seen in Table 2.
TABLE 1. Race and Hispanic Origin Summary Data for NPL Sites and Comparison Areas
| Race (%) | |||||||
| Total persons | White | Black | American Indian, Eskimo, or Aleut | Asian or Pacific Islander | Other race | Hispanic origin | |
| Total for site areas (N = 1,200) | 10,946,979 | 77.6 | 10.3 | 0.7 | 4.9 | 6.5 | 14.4 |
| Total for comparison areas | 141,381,926 | 79.7 | 12.1 | 0.6 | 3.3 | 4.4 | 9.6 |
| United States | 248,709,873 | 80.3 | 12.1 | 0.8 | 2.9 | 3.9 | 9.0 |
Note: Summary results are based on 1990 U.S. Bureau of Census information.
FIGURE 2. Counties located within a mile of one or more NPL sites.
TABLE 2. Race and Hispanic Origin Summary Statistics for Site and Comparison Areas
| Counties within one mile of an NPL site (N = 670) | ||||
| Site area | Comparison area | |||
| Demographic variables | Mean | Std | Mean | Std |
| Race (%) | ||||
| White | 86.4 | 18.6 | 87.7 | 13.2 |
| Black | 9.4 | 17.0 | 8.3 | 11.9 |
| American Indian, Eskimo, or Aleut | 1.1 | 5.4 | 1.1 | 4.6 |
| Asian or Pacific Islander | 1.1 | 2.2 | 1.2 | 2.1 |
| Other race | 1.9 | 5.0 | 1.7 | 3.8 |
| Hispanic origin (%) | 4.3 | 9.7 | 4.0 | 8.0 |
Note: Summary statistics for contiguous 670 counties based on 1990 U.S. Bureau of Census information.
In addition, the results are not appreciably changed by limiting the analysis to states with relatively greater concentrations of the racial and ethnic origin groups (see Table 4).
SUMMARY AND CONCLUSIONS
In this research, we used a GIS approach to obtain site-specific demographic information on 1,200 NPL sites in the contiguous United States. The NPL sites analyzed are located in and near the nation’s more populous counties. That the EPA ranking system for uncontrolled waste sites includes a factor for population proximity may in part explain why this is so. Results of the population analysis show a significantly higher percentage of minority populations near some of the NPL sites than in the comparison areas. If it is assumed that the NPL sites are representative of all uncontrolled hazardous waste facilities, the results support existing environmental inequity research that suggests that the location of hazardous waste facilities is more burdensome for minority communities. Our conclusion will gain more impact if similar results are obtained through an inclusion of other potentially relevant covariates (e.g., population density, socioeconomic status, area industry) in the model as well as through a varying of the buffer distance (e.g., one-half mile, two miles).
TABLE 3. Results of Racial and Hispanic Origin Percentages ANOVA Tests
| SS (Type III) | df | MS | F | p | |
| Race = Black (LOG10 transform %) | |||||
| Model | 56,520.6 | 717 | 78.8 | ||
| State | 10,610.8 | 47 | 225.8 | 4.75 | <.001 |
| County (State) | 37,213.3 | 622 | 59.8 | 1.26 | .002 |
| State*NPL | 3,006.6 | 47 | 64.0 | 1.35 | .065 |
| NPL | 3,157.4 | 1 | 3,157.4 | 49.35 | <.001 |
| Error | 29,553.0 | 622 | 47.5 | ||
| Race = American Indian, Eskimo, or Aleut (LOG10 transform %) | |||||
| Model | 46,719.3 | 717 | 65.2 | ||
| State | 7,023.7 | 47 | 149.4 | 1.78 | .025 |
| County (State) | 29,531.9 | 622 | 47.5 | 1.05 | .261 |
| State*NPL | 3,942.2 | 47 | 83.9 | 1.86 | .001 |
| NPL | 3,247.1 | 1 | 3,247.1 | 38.70 | <.001 |
| Error | 28,051.9 | 622 | 45.1 | ||
| Race = Asian or Pacific Islander (LOG10 transform %) | |||||
| Model | 58,671.9 | 717 | 81.8 | ||
| State | 7,032.8 | 47 | 149.6 | 2.44 | <.001 |
| County (State) | 38,032.1 | 622 | 61.1 | 1.14 | .048 |
| State*NPL | 4,340.5 | 47 | 92.3 | 1.73 | .002 |
| NPL | 5,530.7 | 1 | 5,530.7 | 59.88 | <.001 |
| Error | 33,280.6 | 622 | 53.5 | ||
| Race = White (ARCSINE transform %) | |||||
| Model | 102.55 | 717 | 0.14 | ||
| State | 49.14 | 47 | 1.04 | 12.38 | <.001 |
| County (State) | 52.53 | 622 | 0.08 | 3.18 | <.001 |
| State*NPL | 0.83 | 47 | 0.02 | 0.66 | .960 |
| NPL | 0.04 | 1 | 0.04 | 1.54 | .215 |
| Error | 16.54 | 622 | 0.03 | ||
| Hispanic origin (LOG10 transform %) | |||||
| Model | 42,532.2 | 717 | 59.3 | ||
| State | 6,782.3 | 47 | 144.3 | 3.22 | <.001 |
| County (State) | 27,897.2 | 622 | 44.8 | 1.08 | .181 |
| State*NPL | 3,190.0 | 47 | 67.9 | 1.63 | .006 |
| NPL | 2,561.5 | 1 | 2,561.5 | 37.74 | <.001 |
| Error | 25,926.4 | 622 | 41.7 | ||
Many complex social, political, environmental, and health issues are related to the location of hazardous waste sites. The evidence presented in this research points to a need for further inquiry into the demographics of NPL sites. For example, expanding our analysis to include land use characteristics (e.g., zoning) and additional census data (e.g., year householder moved into unit) at varying buffer distances would lead to better, more comprehensive explanations of the relationship between population and hazardous waste site location. It is also important to incorporate data from previous census years to examine longitudinal trends and develop a historical context for the placing of NPL sites. Findings such as ours, taken from a cross-sectional analysis of census data, do not necessarily help in evaluating historical siting decisions. Some knowledge of the sites and area histories, including longitudinal population patterns, is needed. Finally, we have examined proximity to uncontrolled waste sites without considering data which might help us to assess site exposures and, relatedly, health impacts. Therefore, we almost certainly have both included people who are not at risk of being exposed and excluded people who are at risk of being exposed to site-related contaminants. More work is needed to define and measure better the demographic characteristics of the “at-risk” populations. Glickman’s (1994) GIS study of TRI data for Allegheny County, Pennsylvania, illustrates the use of both proximity and risk-based exposure measures to analyze environmental equity. In summary, GIS capabilities for managing disparate site-specific spatial information (e.g., population, environmental, and health outcome data) provide a key to addressing environmental inequity issues in future research on NPL sites in the United States.
TABLE 4. Results of Racial and Hispanic Origin Percentages ANOVA Tests for States That Equal or Exceed the 1990 U.S. Percentage for That Group
| SS (Type III) | df | MS | F | p | |
| Race = Black (LOG10 transform %)a | |||||
| Model | 16,105.3 | 312 | 51.6 | ||
| State | 1,980.9 | 16 | 123.8 | 3.59 | <.001 |
| County (State) | 12,573.4 | 279 | 45.1 | 1.31 | .013 |
| State*NPL | 465.8 | 16 | 29.1 | 0.84 | .634 |
| NPL | 550.6 | 1 | 550.6 | 15.97 | <.001 |
| Error | 9,619.9 | 279 | 34.5 | ||
| Race = American Indian, Eskimo, or Aleut (LOG10 transform %)b | |||||
| Model | 11,035.6 | 275 | 40.1 | ||
| State | 1,409.2 | 19 | 74.2 | 2.29 | .002 |
| County (State) | 7,944.3 | 236 | 33.7 | 1.04 | .382 |
| State*NPL | 749.6 | 19 | 39.4 | 1.22 | .243 |
| NPL | 617.1 | 1 | 617.1 | 19.07 | <.001 |
| Error | 7,638.2 | 236 | 32.4 | ||
| Race = Asian or Pacific Islander (LOG10 transform %)c | |||||
| Model | 8,099.2 | 142 | 57.0 | ||
| State | 1,755.8 | 6 | 292.6 | 9.19 | <.001 |
| County (State) | 4,481.8 | 129 | 34.7 | 1.09 | .312 |
| State*NPL | 1,150.3 | 6 | 191.7 | 6.02 | <.001 |
| NPL | 606.9 | 1 | 606.9 | 19.05 | <.001 |
| Error | 4,109.4 | 129 | 31.9 | ||
| Hispanic origin (LOG10 transform %)d | |||||
| Model | 4,786.9 | 161 | 29.7 | ||
| State | 616.4 | 8 | 77.0 | 3.39 | .001 |
| County (State) | 3,703.2 | 144 | 25.7 | 1.13 | .230 |
| State*NPL | 230.1 | 8 | 28.8 | 1.27 | .266 |
| NPL | 227.8 | 1 | 227.8 | 10.02 | .002 |
| Error | 3,272.9 | 144 | 22.7 | ||
aStates that equal or exceed the 1990 U.S. percentage
black population (12%) are AL, AR, DE, FL, GA, IL, LA, MD, MI, MS, NC, NJ,
NY, SC, TN, TX, and VA.
bStates that equal or exceed the 1990 U.S. percentage American Indian,
Eskimo, or Aleut population (0.79%) are AZ, CA, CO, ID, KS, MI, MN, MT,
NC, ND, NE, NM, NV, OK, OR, SD, UT, WA, WI, and WY.
cStates that equal or exceed the 1990 U.S. percentage Asian or Pacific
Islander (4%) are CA, MD, NJ, NV, NY, VA, and WA.
dStates that equal or exceed the 1990 U.S. percentage Hispanic origin
population (9%) are AZ, CA, CO, FL, NJ, NM, NV, NY, and TX.
ACKNOWLEDGMENTS
The authors thank Ed Gregory, Cynthia Harris, Bob Williams, Heather Tosteson, Bob Spengler, Tim Aldrich, J. Wanzer Drane, and Lance Waller for their comments on this and earlier versions of the paper. Danika Holm helped in preparing the maps. Editorial assistance was provided by Anne A. Olin.
REFERENCES
APPENDIX 1. Percentage Distribution of Persons Located within One Mile of an NPL Site by State and EPA Region
| Race (%) | ||||||||
| Total sites | Total persons | White | Black | American Indian, Eskimo, or Aleut | Asian or Pacific Islander | Other racea | %Hispanic origin | |
| EPA Region I | ||||||||
| Connecticut | 13 | 97,132 | 92.4 | 5.0 | 0.2 | 1.4 | 1.0 | 3.1 |
| Massachusetts | 23 | 268,867 | 91.2 | 2.9 | 0.2 | 3.6 | 2.1 | 4.1 |
| Maine | 7 | 35,715 | 97.2 | 1.2 | 0.2 | 1.0 | 0.3 | 0.9 |
| New Hampshire | 16 | 74,760 | 96.8 | 1.6 | 0.2 | 1.0 | 0.4 | 1.1 |
| Rhode Island | 11 | 98,400 | 93.8 | 3.4 | 0.6 | 1.3 | 1.0 | 2.6 |
| Vermont | 8 | 31,469 | 97.4 | 0.7 | 0.3 | 1.2 | 0.3 | 1.3 |
| 78 | 606,343 | 93.2 | 3.0 | 0.3 | 2.3 | 1.3 | 3.0 | |
| EPA Region II | ||||||||
| New Jersey | 112 | 1,031,504 | 78.9 | 13.4 | 0.2 | 3.6 | 3.8 | 10.1 |
| New York | 83 | 730,489 | 87.2 | 5.4 | 0.3 | 4.5 | 2.6 | 7.7 |
| 195 | 1,761,993 | 82.4 | 10.1 | 0.3 | 4.0 | 3.3 | 9.1 | |
| EPA Region III | ||||||||
| Delaware | 21 | 92,993 | 77.0 | 20.7 | 0.4 | 1.1 | 0.9 | 2.2 |
| Maryland | 10 | 125,078 | 78.6 | 18.8 | 0.4 | 1.4 | 0.8 | 1.9 |
| Pennsylvania | 99 | 665,562 | 91.3 | 6.6 | 0.1 | 1.6 | 0.4 | 1.5 |
| Virginia | 21 | 92,092 | 68.5 | 29.4 | 0.3 | 1.4 | 0.4 | 1.2 |
| West Virginia | 5 | 20,208 | 96.9 | 2.2 | 0.1 | 0.5 | 0.3 | 0.6 |
| 156 | 995,933 | 86.4 | 11.5 | 0.2 | 1.5 | 0.5 | 1.6 | |
| EPA Region IV | ||||||||
| Alabama | 12 | 249,670 | 78.9 | 18.6 | 0.5 | 1.7 | 0.3 | 1.3 |
| Florida | 47 | 437,417 | 68.5 | 27.2 | 0.3 | 1.2 | 2.8 | 16.6 |
| Georgia | 14 | 76,656 | 54.7 | 43.4 | 0.3 | 0.9 | 0.8 | 1.8 |
| Kentucky | 16 | 27,881 | 98.1 | 1.4 | 0.1 | 0.3 | 0.1 | 0.4 |
| Mississippi | 3 | 22,272 | 27.7 | 71.9 | 0.1 | 0.2 | 0.1 | 0.7 |
| North Carolina | 23 | 158,563 | 72.8 | 22.8 | 0.6 | 1.5 | 2.2 | 4.2 |
| South Carolina | 23 | 119,120 | 81.2 | 17.4 | 0.2 | 1.0 | 0.2 | 0.9 |
| Tennessee | 13 | 60,009 | 57.7 | 41.3 | 0.2 | 0.7 | 0.1 | 0.5 |
| 151 | 1,151,588 | 71.1 | 25.8 | 0.4 | 1.2 | 1.5 | 11.2 | |
| EPA Region V | ||||||||
| Illinois | 35 | 219,227 | 84.1 | 9.6 | 0.3 | 1.8 | 4.3 | 8.9 |
| Indiana | 35 | 243,288 | 84.2 | 13.7 | 0.3 | 0.6 | 1.1 | 2.5 |
| Michigan | 80 | 407,061 | 84.2 | 12.0 | 0.8 | 0.9 | 2.1 | 3.8 |
| Minnesota | 43 | 346,968 | 91.5 | 3.7 | 1.3 | 3.0 | 0.5 | 1.3 |
| Ohio | 34 | 306,185 | 90.8 | 6.5 | 0.2 | 1.4 | 1.1 | 2.1 |
| Wisconsin | 40 | 206,821 | 94.8 | 1.6 | 0.8 | 2.3 | 0.6 | 1.6 |
| 267 | 1,729,550 | 88.1 | 8.0 | 0.7 | 1.6 | 1.6 | 3.2 | |
| EPA Region VI | ||||||||
| Arkansas | 12 | 27,122 | 78.9 | 18.6 | 0.5 | 1.7 | 0.3 | 1.3 |
| Louisiana | 11 | 21,837 | 75.3 | 23.5 | 0.2 | 0.6 | 0.4 | 2.2 |
| New Mexico | 11 | 35,688 | 62.8 | 6.1 | 4.0 | 1.5 | 25.7 | 58.6 |
| Oklahoma | 13 | 60,661 | 68.2 | 21.0 | 9.5 | 0.6 | 0.8 | 1.8 |
| Texas | 30 | 205,687 | 57.5 | 23.5 | 0.4 | 3.5 | 15.1 | 27.1 |
| 77 | 350,995 | 63.0 | 20.6 | 2.3 | 2.5 | 11.7 | 22.4 | |
| EPA Region VII | ||||||||
| Iowa | 23 | 105,847 | 94.0 | 3.8 | 0.3 | 1.1 | 0.8 | 2.0 |
| Kansas | 12 | 116,419 | 70.4 | 21.3 | 1.5 | 1.9 | 5.0 | 7.6 |
| Missouri | 23 | 147,877 | 90.9 | 7.5 | 0.5 | 0.9 | 0.3 | 0.9 |
| Nebraska | 6 | 43,661 | 98.4 | 0.5 | 0.3 | 0.3 | 0.4 | 1.0 |
| 64 | 413,804 | 86.7 | 9.7 | 0.7 | 1.2 | 1.8 | 3.1 | |
| EPA Region VIII | ||||||||
| Colorado | 18 | 401,262 | 71.3 | 11.9 | 1.3 | 2.2 | 13.3 | 27.2 |
| Montana | 10 | 32,752 | 97.1 | 0.2 | 1.8 | 0.4 | 0.5 | 1.7 |
| North Dakota | 3 | 21,109 | 91.3 | 4.8 | 0.7 | 1.6 | 1.5 | 2.6 |
| South Dakota | 3 | 23,561 | 90.9 | 4.3 | 2.0 | 2.0 | 0.9 | 2.6 |
| Utah | 11 | 116,747 | 86.9 | 1.7 | 1.5 | 3.7 | 6.2 | 11.9 |
| Wyoming | 3 | 31,612 | 91.1 | 3.4 | 0.6 | 1.6 | 3.3 | 7.6 |
| 48 | 627,043 | 78.0 | 8.4 | 1.3 | 2.3 | 10.0 | 20.3 | |
| EPA Region IX | ||||||||
| Arizona | 13 | 200,972 | 79.0 | 3.9 | 2.3 | 1.7 | 13.1 | 23.8 |
| California | 87 | 2,604,281 | 62.6 | 6.7 | 0.8 | 12.9 | 17.0 | 34.9 |
| Hawaii | — | |||||||
| Nevada | 1 | 12,189 | 93.5 | 0.4 | 2.1 | 1.2 | 2.8 | 7.2 |
| 101 | 2,817,442 | 63.9 | 6.5 | 0.9 | 12.0 | 16.7 | 34.0 | |
| EPA Region X | ||||||||
| Alaska | — | |||||||
| Idaho | 9 | 19,258 | 93.0 | 2.8 | 1.1 | 1.5 | 1.7 | 4.1 |
| Oregon | 8 | 30,947 | 79.5 | 13.7 | 1.7 | 2.6 | 2.6 | 4.9 |
| Washington | 46 | 442,083 | 83.7 | 7.6 | 1.6 | 5.0 | 2.0 | 4.4 |
| 63 | 492,288 | 83.8 | 7.8 | 1.6 | 4.8 | 2.0 | 4.4 | |
Source: 1990 U.S. Bureau of the Census.
aThe category “Other race” is included in the table only
to complete the race category and will not be analyzed. “Other race”
includes persons “providing write-in entries such as multiracial, multiethnic,
…, or a Spanish/Hispanic origin group …” (U.S. Bureau of the
Census, 1991b).
Agency for Toxic Substances and Disease Registry, 1825 Century Blvd, Atlanta, GA 30345
Contact CDC: 800-232-4636 / TTY: 888-232-6348
