Abstract
The proliferation of underutilized, derelict, and contaminated properties following a nationwide decline of industrial production has created a unique policy problem. Powerful liability schemes under Superfund made property owners, developers, and lenders hesitant to engage in transactions involving real estate that is contaminated or perceived as contaminated. Adoption of Federal and state policy to address these “Brownfields” in the 1990s and 2000s has attempted to promote redevelopment by limiting liability of involved parties and providing grant funding. This research hypothesizes that “environmental justice communities” have significantly lower likelihood of receiving benefit-maximizing redevelopment projects under both Federal and state-level voluntary cleanup programs. A multinomial logistic regression model considering past use of the site, socioeconomic status of the surrounding census tract and its composite urban sprawl score, and Republican control of the district containing the brownfield were used to assess the probability of a hierarchy of redevelopment outcomes.
Introduction
Urban areas throughout the United States are saddled with a prolific number of dilapidated properties, particularly in the cities of the Northeast and Midwest that once thrived on industrial manufacturing. These “brownfield” properties were developed long before the adoption of wide-ranging environmental regulations and served numerous purposes, including service stations, dry cleaners, warehouses, mines, and industrial facilities (Page & Berger, 2006). 1 Underutilization of these blighted properties is a source of frustration for citizens (Greenberg & Lewis, 2000) and government officials because they reduce surrounding property values, limit tax revenue for local municipalities, threaten public health, and contribute to urban sprawl (Alberini, Longo, Tonin, Trombetta, & Turvani, 2005; Greenberg & Hollander, 2006; Wernstedt, Blackman, Lyon, & Novak, 2013). The prevalence of at least 450,000 of these sites prompted the Environmental Protection Agency (EPA) to launch its Brownfields Program in 1994 which provides a variety of grant funding opportunities for both local governments and private property owners to aid in the redevelopment process. Voluntary cleanup programs (VCPs) arose at the same general time as the Federal program and are now operating in nearly every state (Schwarz, Depken, Hanning, & Peterson, 2009). These VCPs are far from uniform in their complexity, as most operate under a “risk based” scheme in which cleanup standards are dictated by the future use of the site.
Brownfield remediation and redevelopment is a multifaceted Federalist process of public and private collaboration. According to most observers, it is a successful policy that considers public health, environmental protection, and economic redevelopment concerns, which are addressed by Federal, state, and local governments as well as both private and quasigovernmental entities (Greenberg & Hollander, 2006). These concerns are particularly relevant to the pursuit of environmental justice (EJ) 2 and are associated with brownfield legislation in timing and ideology, but there is no causal link, only a widely recognized correlation between the number and proximity of brownfield proprieties to communities of lower socioeconomic status (Eckerd & Keeler, 2012) that many environmental and economic administrators earnestly considered. EJ became a required consideration for all Federal agency actions following the passage of Executive Order 12898 in 1994, one year after the inception of the Brownfields Pilot Program (Johnson, 1996), but has never become a policy goal for environmental redevelopment. Remediation, nonetheless, can offer considerable potential for promoting EJ in numerous cities throughout the country by reducing exposure to chemical contamination, improving blighted landscapes, and recovering real estate value for both citizens and local governments. Although these goals have been concerns of the Federal Brownfields Program since inception, the current implementation and future state and private programs are less concerned with EJ indicators as political priorities shifted with time and executive administrations.
Research Question
Although VCPs are somewhat intertwined with funding under the Federal Brownfield Program, the potential elimination of this funding is unlikely to scuttle the state-level programs, as the same concerns related to environmental risk that created the “brownfield problem” are now a standard component of property transfer proceedings for financial institutions and other corporate entities. Simply put, VCPs do not require public funding and ergo are not concerned with public good or other public assistance as they offer considerable mitigation of environmental risk for current property owners and prospective redevelopers. Reasonable EJ elements, however, are to be considered in these VCPs, as policy makers and scholars would be wise to scrutinize these state-level programs with the same vigor as was dedicated to evaluation of the Brownfields Pilot Program.
Twenty years have passed since the implementation of the EPA’s Brownfield Program and the signing of Executive Order 12898, and the inconsistency of existing assessments of the Program’s ability to promote EJ prompts this research to examine the equity of redevelopment outcomes. The EPA reports a host of statistics to highlight the Program’s success, most notably that every dollar allocated by the agency leverages US$17.79 from private parties, a ratio that has netted US$21.3 billion since the Program’s inception (U.S. EPA, 2014). The program is also financially advantageous to neighborhoods surrounding a brownfield, as cleanup alone can increase property values, within a mile of the property, between 5.1% and 12.8% (U.S. EPA, 2014). This return on investment is not equal for all redevelopment types, however, and is well described in existing hedonic price models. Measuring that redevelopment injustice is this project. With these results in mind, this research proposes to assess the EJ of brownfield redevelopment based on future use of the sites for greenspace, residential, commercial, or industrial purposes. This ranking of zoning types reflects both the greatest postremediation increases in surrounding property values in hedonic price models and the uses associated with the least environmental risk under the aforementioned risk-based cleanup standards. The zoning issue is not ignored in this study, but is not central as it has problems of generalizability in statistical modeling because industrial development is encouraged for development, employment, and tax base, and residential is discouraged because of urban sprawl and gentrification. A multinomial logistic regression model will estimate the likelihood of these redevelopment purposes based on the socioeconomic characteristics of the Census tract containing the brownfield as well as past use of the site. Brownfields are commonly cited as contributors to urban sprawl (Meyer, 2010), and the model will utilize urban sprawl composite scores by the widely used variable coding attained from the National Institutes of Health as a consideration of redevelopment outcomes. Brownfield properties enrolled in both the Federal Brownfield Program and state VCPs in Colorado and Ohio are examined to compare the equity of programs in two distinct regions. The states selected for this study have been chosen due to their status as “swing states” in recent national elections, and the analysis will consider the party affiliation of the legislator representing the district containing brownfield as a possible motivator of redevelopment outcomes.
Environmental Justice Literature
EJ has long grappled with the causation of environmental inequalities (Mohai, Pellow, & Roberts, 2009). For example, longitudinal analyses of studies in California and Michigan to address this question have found that polluting facilities are typically sited systematically in vulnerable communities and are not the product of residential sorting that occurs following decreases in real estate values surrounding a hazardous facility (Mohai et al., 2009; Saha & Mohai, 2005). Little evidence of discriminatory siting of “locally unwanted land uses” (LULUs) was found prior to the 1970s, but as awareness of environmental risk grew, affluent communities with greater sociopolitical status were better equipped to resist LULUs (Mohai et al., 2009; Saha & Mohai, 2005). This socioeconomic explanation is countered by an economics or market-based explanation that profit-maximization drives polluting industries to seek the cheapest land and areas with the largest pools of potential employees (Mohai et al., 2009; Ringquist, 2003).
Despite failed attempts to legislate EJ in Congress, Federal agencies were ordered to consider the issue in rulemaking with the signing of Executive Order 12898 in 1994. The EPA attempted to implement its own EJ policy based on the provisions of Title VI of the Civil Rights Act of 1964 that allowed any disproportionate impact to be considered regardless of intent. This public administration strategy was undermined by the Supreme Court holding in Alexander v. Sandoval (2001), which limited the applicability of Title VI (Mohai et al., 2009), and has resulted in a more implicit consideration of EJ in public policy. While formal EJ policy has stumbled, scholarly attention to the topic has provided extensive justification for continued emphasis. Several meta-analyses of this research have revealed that class, race, and ethnicity are significant predictors of proximity to known and prospective environmental hazards (Evans & Kantrowitz, 2002; Mohai et al., 2009; Ringquist, 2003).
Past findings and explanations of disproportionate environmental risk are overwhelming, but must theoretically question if future land use will continue or change the analysis of brownfield remediation and redevelopment. For example, many of the remediation efforts requiring the greatest investment are located within low-income or minority neighborhoods that exhibit the lowest market value and demand (McCarthy, 2009), and knowledge of funding opportunities is typically lacking in blighted or otherwise disadvantaged neighborhoods (Howland, 2007). This work is meritorious but never questions if we should expect future use plans for sites to be different in marginalized socioeconomic communities. EJ and brownfields are often the collocation of past industrial sites and poor minorities, and that may have a lot or little to do with cleanup and redevelopment. Brownfield redevelopment is often touted for its ability to revitalize communities by creating jobs, reestablishing the local tax base, and encouraging further redevelopment in the surrounding area (Bacot & O’Dell, 2006). There are, however, notable concerns as to whether or not the targeting and cleanup of brownfield sites is equitable. Early brownfield redevelopment policy supported targeting sites with minimal contamination and higher market demand, offering the highest potential return on investment (Howland, 2007). Critics of current practices question if the jobs and value created following redevelopment will benefit the original residents of the area surrounding the former brownfield (Howland, 2007). This criticism is further complicated by a distinct binary in brownfield markets: growing urban centers with high interest in development and declining urban centers with significantly less attraction of outside capital (Howland, 2007).
This work is theoretically needed as knowledge about the ability of brownfield redevelopment to promote EJ is sparse, particularly in regard to state programs. Analyses of the initial years of the Federal Brownfield Program agree that the program distributed grants to the most economically disadvantaged cities and counties and those with greater proportions of minorities. Examining the risk mitigation achieved by remediation suggests that the program has not consistently guaranteed timely targeting or completion of cleanup at brownfield sites. Furthermore, results are inconsistent depending on the spatial unit chosen and the analytical technique used. The majority of research on VCPs has focused primarily upon the real estate implications of redevelopment, and though EJ concerns are present in much of the literature, no known quantitative analysis has been completed. EJ is not a required component of most VCPs, but the use of Federal funding and Memorandums of Agreement with U.S. EPA for site closure validates EJ concerns.
This study will empirically test if brownfield redevelopment is correlated with EJ issues by looking at politically ambiguous states decades after the implementation of Federal and state programs by including novel factors like urban sprawl and conservative legislative districts (Eckerd & Heidelberg, 2015). This analysis will create a new EJ metric for economic development that can offer a clearer policy assessment of market-oriented state VCP programs that seem to have no explicit focus on EJ but offer some implicit concerns regardless of the fact that VCPs provide no Federal funding. Note this study is an attempt to assess brownfield practices as a social justice policy but is careful to not judge future use plans as being simply a coded benefit or burden to EJ communities as such statements may not be generalizable. For instance, in some communities, industrial redevelopment may be beneficial to EJ as it might make gentrification and displacement less likely, create new low-skill jobs, have a low likelihood of rent seeking revenue options like green spacing, and due to newer regulations, result in less contamination and pollution than historical industrial practices.
Challenges Associated With Brownfield Redevelopment
Redevelopment of contaminated property has the potential to be prolonged and problematic, as goals central to economic development are often incongruent with abatement of environmental risk. Very few contaminated properties are cleaned up completely because the cost of remediation increases disproportionately with the removal of pollutants at greater soil depths or within a wider distribution (Meyer, 2010). The reduction of environmental risk that is required for residential redevelopment is considerable, and the emphasis of economic goals may undermine the attainment of EJ by promoting industrial and commercial land use that maximize tax revenue streams for local governments. Howland (2007) asserts that many of the most heavily contaminated brownfields are found in urban centers characterized by large minority populations and high poverty rates. In addition, the diminishing returns associated with greater cleanup standards combined with existing low market value of real estate in those areas may substantially increase the likelihood of the same brownfields becoming industrial zones again.
Compounding remediation cost considerations is possible resistance from community members, who are often encouraged by the EPA. This is further compounded by state agencies that are asked to participate in the assessment and remediation process but whose input is not necessarily considered in redevelopment planning. A survey of over 300 representatives from firms who have experience in redevelopment found that developers tend to devalue community involvement due to the perception of increased costs or delays to the project, despite their general agreement that public involvement could enhance a project’s bottom line and overall quality (Alberini et al., 2005). Brownfield redevelopment policy is primarily focused upon the reduction of impediments to profitable real estate transactions despite the marketing of the policy as a means to promote “smart growth” and reduce environmental risk (Alberini et al., 2005). Reaching an objective standard to judge whether or not brownfield redevelopment is “sustainable” is challenging, but the sustainability and equity of the program are questionable if it perpetuates existing land use patterns instead of addressing greater EJ concerns.
Again, the purpose of our study is to create a general model for assessing if state and Federal Brownfield policies can, and under what situations does it, support communities at risk. This work is theoretically questioning the evaluation of justice in Brownfield Projects which is different than practical examinations like Daley and Layton (2004) who make a causal argument about redevelopment and if the local representative sits on the Superfund Committee or Eckerd and Heidelberg who specifically look at state policy and future use plans.
Can the EPA Brownfield Program Promote Environmental Justice?
Despite the evaluations of the program are somewhat sparse, the majority of the existing literature focuses on grant distribution. Eckerd and Keeler (2012) found that census tracts containing higher proportions of welfare recipients and Black residents were more likely to contain a brownfield receiving EPA funding, while higher levels of education decreased that likelihood. However, analysis of the progress of a site through assessment to the approval of a No Further Action (NFA) letter was not as flattering. A Cox regression estimating the likelihood of a site reaching each assessment milestone found that assessment periods took longer in tracts with greater proportions of Black and Hispanic residents but took less time in tracts with higher rates of citizens receiving public assistance (Eckerd & Keeler, 2012). A logistic regression on the same dataset illustrated that tracts with a greater proportion of residents possessing a bachelor’s degree significantly increased the likelihood of a brownfield receiving an NFA letter (Eckerd & Keeler, 2012). Other significant variables increasing the odds included the proportion of housing units built before 1940, multiple brownfields located within the tract, the Indiana Relative Chemical Hazard Score (IRCHS) for the site (a rough approximation of contamination), government ownership, past use for industrial purposes, and the creation of a future plan for the site (Eckerd & Keeler, 2012). The authors conclude that environmental equity concerns are still valid because minorities are subject to greater environmental risk posed by a combination of proximity and slower risk abatement (Eckerd & Keeler, 2012). This research is relevant to the formation of this study, as the findings expressed are notably different from those of the original research described below.
An earlier analysis of cities receiving grants from the Brownfields Pilot Program illustrated the disproportionate impact of brownfields by comparing the award-winning cities with those that were located in the same state and had a similar population but did not apply (Solitare & Greenberg, 2002). For instance, Pilot cities had significantly higher rates of non-White populations, poverty, unemployment, crime, and individuals with less than a high school education and significantly lower rates of individuals with a bachelor’s degree or higher, homeownership, self-employment, and median income (Solitare & Greenberg, 2002). While these findings help to illustrate the burden of brownfields within cities of low socioeconomic status, the failure to consider more focused spatial units, like a census tract, limits this study’s ability to make any conclusion about the function of the Pilot Brownfield program to promote EJ.
Greenberg and Hollander (2006) built upon the earlier analysis of Brownfield Pilot Program grant winners by performing an ordinary least squares regression on a similar dataset this time considering which of the 13 rounds of grant distribution the city was awarded. The probability of a previous award was significantly correlated with the grant awardee being a city rather than a county or state, with a higher proportion of African American residents and lower rates of homeownership (Greenberg & Hollander, 2006). The first 5 years of the Pilot Program saw 90% of its grants distributed to cities, a number that dropped to 77% in the remaining years prior to the implementation of the Brownfields Act (Greenberg & Hollander, 2006). Similarly, 57% of awardees in the first 5 years had homeownership rates of 50% or higher, but 83% of awardees in later rounds had similar homeownership rates (Greenberg & Hollander, 2006). Although the authors praised the Program’s early success, Greenberg and Hollander noted the potential for other confounding factors or steering variables necessary for robust local administrative capacity. Here again, focusing on government awardees limits the applicability of the research conclusions regarding EJ for populations living in the immediate vicinity of a brownfield.
Despite this limitation, the research did offer evidence that other factors may be at play when examining the distribution of EPA grants. Citing interviews with 20 representatives of cities that were awarded grants, Congressional representation was instrumental in promoting the Pilot Program (Greenberg & Hollander, 2006). Cuyahoga County, Ohio, was the first grant awardee (Coffin & Shepherd, 1998), and county officials attributed its early award to outreach from Congressman Lou Stokes, who was a member of several committees overseeing EPA activities (Greenberg & Hollander, 2006). A similar experience was reported in Bridgeport, Connecticut, whose economic development specialist was made aware of the Pilot Program by a staff member of the local Congressman (Greenberg & Hollander, 2006). These anecdotes suggest that larger cities had an advantage in the grant distribution due to informal networks and personal contacts with Federal officials (Greenberg & Hollander, 2006). Similar evidence of the brownfield program success was reported in a geographic information systems (GIS) buffer zone analysis of 389 cities in the Detroit Metropolitan Statistical Area. This analysis found a higher mean percentage of minorities within a half mile of brownfields that cleanup action had initiated before 1990 relative to those where cleanup had started after 1990 (Lee & Mohai, 2013). In fact, 26% of residents near an uninitiated cleanup were African American relative to 58% of residents near a completed cleanup, suggesting that communities of low socioeconomic status seemed to be prioritized for both cleanup and repurposing of land (Lee & Mohai, 2013). This research is somewhat limited in its generalizability due to its focus on a single area and on sites that were cleaned up well before the initiation of the Pilot Program.
Not all analyses of grant distribution are congruent, however. In a case study of brownfield sites in the City of Milwaukee, receiving financial assistance from government, a lower proportion of funded sites were found in low-income and minority neighborhoods despite 48 containing a greater number of sites (McCarthy, 2009). McCarthy cites this as an example of emphasis on “leveraged dollars” as being biased against delivering EJ goals because the sites in disadvantaged areas will not typically achieve the same investment (2009). This result is inconsistent with previous research, which suggests notable discrepancies in the application of redevelopment programs from state to state and city to city. Dull and Wernstedt utilized data on local governments that had applied for Federal Brownfield grants between 2003 and 2007 to determine the effects of a variety of socioeconomic and political indicators on both applications and awards of grants with a Heckman probit model staff member of the local Congressman (Greenberg & Hollander, 2006). These anecdotes suggest that larger cities had an advantage in the grant distribution due to informal networks and personal contacts with Federal officials (Greenberg & Hollander, 2006). Significant variables that increased the likelihood of a city government applying for a brownfield grant included the presence of other sites of environmental concern, higher rates of non-White and college-educated citizens, and higher property taxes. Higher rates of homeownership, poverty, and employment in the manufacturing sector decreased the likelihood of the city applying for a grant (Dull & Wernstedt, 2010). The presence of a Republican Governor at the time of the grant application increased the probability of an application, while a Republican-controlled legislature, divided legislature, presence of the local Congressman on the Superfund Subcommittee, or being located in a “swing state” reduced the likelihood (Dull & Wernstedt, 2010). Significant variables that increased the odds of an award included increased educational attainment, presence of the local Congressman on the Superfund Subcommittee, Republican legislature, and divided legislature (Dull & Wernstedt, 2010). The odds of an award decreased under a Republican Governor and for cities with higher rates of poverty and non-White citizens (Dull & Wernstedt, 2010). These findings illustrate that the Brownfields Program has seen modest success in assisting communities but lays out significant concerns regarding the ability of the program to promote environmental equity, particularly when political representation appears to shape outcomes. “To account for important social goals of brownfield cleanup such as targeting communities in need and addressing environmental justice” (Eckerd & Heidelberg, 2015, p. 256), we have used the same demographic variables as most scholars but have included some new theoretical factors to test if just redevelopment occurs in probusiness Republican areas and question when does voluntary Brownfield contribute to urban sprawl (Daley & Layton, 2004; Eckerd & Keeler, 2012; Howland, 2000). While strongly influenced by Eckerd and Keller’s scholarship, this analysis implies that research should evaluate brownfield success not on the policy of cleanup but on the influence of markets on redevelopment.
Evaluations of VCPs
The adoption of risk-based cleanup standards and the availability of loans, grants, and tax incentives for property owners enrolled in VCPs has been the topic of much debate but little quantitative analysis. Interest groups factored heavily in to the formulation and implementation of state VCP programs, with the vast majority of business interests supporting the programs while environmental groups opposed (Daley, 2007). The adoption of state VCP appears to be the result of frustration surrounding Superfund, as states containing sites that progressed slower through the remediation process were more likely to adopt VCPs earlier (Daley, 2007). The availability of state assistance to developers is also contentious, as some argue that subsidizing such projects overextends benefits to developers (Schwarz et al., 2009). Others argue that these incentives are necessary to spur redevelopment at sites with low market value or land planned for future residential use (Howland, 2007) or sites overlooked for their low acreage (Coffin, 2003). However, the market value of a brownfield for future residential redevelopment is higher than that of planned industrial or commercial purposes in some markets due to high demand for housing (De Sousa, 2000). Furthermore, cleanup costs for residential redevelopment may be exaggerated, as only 7% or 8% of total redevelopment costs are incurred from cleanup (Schopp, 2003). This finding is supported by Bacot and O’Dell’s (2006) analysis of brownfields in Charlotte, North Carolina’s South End, where average cleanup costs averaged less than half a percent and no more than 6% of all site development costs. The relatively low percentage of redevelopment costs that are typically incurred by remediation of contamination calls into question the ability of VCPs to adequately promote redevelopment that abates environmental risk.
Meritorious literature on VCPs is rare, but there is limited evidence describing the ability of VCPs to promote EJ by way of risk mitigation. By comparing brownfield properties in California’s CalSites (sites that may be Superfund eligible) with VCP sites, Schwarz et al. (2009) offered insight into the effects of risk-based assessment’s on redevelopment outcomes. Their model included past use, redeveloped use, and a hazard index ranking to characterize the contamination present at the site (Schwarz et al., 2009). More hazardous sites were less likely to be redeveloped residentially, but CalSites properties that had been residential were likely to be redeveloped as residential compared with previously residential VCP sites (Schwarz et al., 2009). Both industrial and commercial sites were significantly more likely to be redeveloped for the same purpose (Schwarz et al., 2009). These findings suggest that risk-based VCP’s are less likely to encourage brownfield redevelopment for residential purposes, a validation of some concerns that VCP’s are more focused on real estate matters than they are concerned with environmental risk mitigation.
Additional evidence that risk-based VCPs do not promote mitigation of environmental risk can be found in Alberini’s (2007) analysis of Colorado’s VCP. This effort identified variables that impacted the enrollment of a brownfield into Colorado’s VCP and found that size of the parcel and land use around it were significant, but sites with high development potential are ultimately more likely to participate. This suggests that the most contaminated sites may not be targeted in a timely manner. Furthermore, Alberini found that Colorado’s VCP had created a new “crop” of brownfields outside of those already on the EPA’s contaminated site registry and that the majority of participants had applied directly for a No Further Action Determination (NFAD), indicating that the Program had not encouraged a significant amount of remediation (Alberini, 2007). A similar effort examined sites participating in the VCP of Baltimore, Maryland. The primary goal of the research was to estimate the likelihood of a property being enrolled in the VCP with a probit model considering neighborhood characteristics. Similar to Colorado VCP participants, the vast majority of applicants request NFADs immediately, indicating that cleanup is not needed at most sites. The majority of sites enrolled were used for industrial or commercial purposes in the past and were likely to be used for those purposes in the future due to zoning restrictions (Guignet & Alberini, 2010). In the case of Colorado’s VCP, sites were more likely to be developed for residential use (Alberini, 2007). As such, this research focuses on the difference redevelopment makes in VCPs. These findings in Colorado and Maryland raise questions about the ability of VCPs to promote EJ, as the sizable proportion of property owners immediately requesting an NFA suggests that participation in such programs appears to be merely a step in a real estate transaction and not a mitigation of environmental risk. Equally concerning is the finding that sites in Baltimore were likely to be redeveloped for the same industrial and commercial purposes as they had been used for in the past, possibly leaving nearby residents with similar environmental risk and depressed property values. The tendency of VCP properties in Colorado to be converted to residential might be explained by high housing demand as it raises concern that redevelopment is contributing to gentrification.
Dependent Variable Justification
Hedonic price modeling is an econometric approach often used to illustrate the impact of environmental amenities and sources of environmental risk such as landfills, Superfund sites, and brownfields. This approach asserts that the price of a good results from the combination of its individual characteristics, which, in most studies, consider a home’s square footage, number of bedrooms, property acreage, and other aspects in relation to proximity to a pollution source (disamenity) or public good (amenity). By comparing the characteristics of homes and their values in relation to the proximity of a disamenity, a hedonic price model illustrates both the value lost attributed to the disamenity and the “marginal willingness to pay” of consumers to distance themselves from the disamenity (Mendelsohn & Olmstead, 2009). Kaufman and Cloutier (2006) utilized a similar model to analyze the Lincoln neighborhood in Kenosha, Wisconsin. The neighborhood once thrived upon the numerous industries located in the immediate downtown area (Kaufman & Cloutier, 2006). Demographics of the area are consistent with those of an EJ neighborhood, as poverty rates are high, income and educational attainment levels are low, and the majority of housing units are renter occupied (Kaufman & Cloutier, 2006). The hedonic pricing model considered the two brownfields, a large city park, and 890 residences within the neighborhood and found an inverse relationship between property value and proximity to a brownfield (Kaufman & Cloutier, 2006). The opposite was true for proximity to the city park, and a hypothetical model that replaced both brownfields with a city park estimated that the aggregate value of housing in the neighborhood could be increased by US$2.4 to US$7.1 million (Kaufman & Cloutier, 2006). This finding illustrates not only the deleterious effect of brownfields on residential housing values but also that redevelopment of the site for greenspace or recreational purposes substantially increases housing values.
De Sousa, Wu, and Westphal (2009) utilized hedonic price modeling to determine the impact of brownfields on nearby property values in Milwaukee and Minneapolis. Analysis of housing transaction data from 1996 was utilized to determine the detrimental impact of brownfield proximity on property values, finding a significant effect up to 4,000 ft away from undeveloped brownfields in Milwaukee but a less significant effect at a maximum of 2,500 ft away in Minneapolis (De Sousa et al., 2009). The impact of remediation on property values was determined by comparing housing transaction data from 1996 to transaction data from 2004. The model revealed an 11.4% net increase for housing prices after nearby site remediation in Milwaukee and a 2.7% net increase in Minneapolis (De Sousa et al., 2009). In the case of Milwaukee, this increase was only enough to recover lost value attributed to the proximity of undeveloped brownfields, while the increase observed in Minneapolis was additive, a difference possibly explained by Minneapolis’s strong housing market relative to that of Milwaukee (De Sousa et al., 2009). This study was particularly significant for its specification of property value increases by land use type. Creation of housing increased property values in Milwaukee and Minneapolis at 8.6% and 3.1%, respectively, and greenspace conversions increased housing values in Milwaukee and Minneapolis at 11.7% and 4.4%, respectively (De Sousa et al., 2009). The creation of new commercial areas resulted in an increase of 15.8% in Milwaukee, but only 4.6% in Minneapolis (De Sousa et al., 2009). Redevelopment for industrial purposes only regained 4.7% of an 11.5% decrease in property value in Milwaukee (De Sousa et al., 2009). The disparity in property value improvements observed in this model illustrate that redevelopment outcomes are unequal and can serve as one metric to consider the EJ of brownfield redevelopment. As such, our research will consider nonindustrial redevelopment outcomes as an indicator of EJ by analyzing both Federal and VCP brownfield properties in Colorado and Ohio.
Method
Given the inconsistencies and gaps in the existing literature, this research will examine the correlation of EJ of the Brownfields Program and state VCPs in Colorado and Ohio based on the future land use of the site. Brownfields are redeveloped for greenspace, residential, commercial, industrial, and mixed-use purposes, and hedonic price modeling suggests a hierarchy of these outcomes based on the value recovery and value increase of property near the site. Nonindustrial land use outcomes are advantageous to surrounding property values, and the cleanup standards imposed for these redevelopment types offer the greatest reduction of environmental risk for those living near the sites. Use of these general zoning categories also serves the purpose of addressing a shortcoming of the Eckerd and Keeler temporal analysis of redevelopment progression. In short, the complexity and duration of remediation can vary significantly based on myriad factors, most importantly, the contaminants in question at the site (or lack thereof), concentrations of the contaminants, affected media (soil, groundwater, soil vapor), and site size.
With these observations in mind, a multinomial logistic regression model will illustrate the probability of nonindustrial outcomes as explained by the past use of the site as well as the socioeconomic status and composite urban sprawl score of the census tract containing the site. Socioeconomic status is measured by three variables: percentage of citizens with a bachelor’s degree or higher, percentage of White citizens, and percentage of citizens receiving public assistance. These three variables are typical to EJ studies and were chosen to measure socioeconomic status in a broad manner. Other variables typical to EJ studies like median household income, percentage of renter-occupied housing units, and other variables such as median monthly rent were excluded due to multicollinearity in preliminary specification of the model. Similar to Dull and Wernstedt’s approach (2010), the political affiliation of the tract’s Congressional representative will be considered as a predictor of outcomes as a binary variable.
While not as intuitive as linear regression, use of the multinomial logit allows for similar analysis of the effect of both numerical and categorical covariates on a categorical or binary dependent variable. The functional form of the model takes the following general form:
Formally, the multinomial logit model specifies the probability that a discrete and unordered outcome is selected:
The coding is such that 1 corresponds to industrial use, 2 corresponds to commercial use, 3 corresponds with greenspace use, 4 corresponds with mixed use, and 5 corresponds with residential use. For the purpose of comparing nonindustrial redevelopment outcomes, industrial use serves as the baseline category in the specification of the model.
Maximization of the following log-likelihood function, where j is the number of choices, i represents an individual in the dataset, and n represents the total number of individuals, estimates the probabilities of each outcome:
The relative probability of choosing one outcome relative to the baseline category “1” is as follows:
Data Sources
All data on sites in the Federal Brownfield Program was attained from the EPA’s “Cleanups in My Community” website, a publicly accessible database of properties tracked under the Comprehensive Environmental Response, Compensation, and Liability Information System (CERCLIS). Raw datasets were generated for each state via the website’s filter and contained all information available regarding the site. Data on Colorado VCP sites were obtained via a Freedom of Information Act (FOIA) request submitted to the Colorado Department of Public Health and the Environment. The dataset contains sites that completed a cleanup under the program and those that applied directly for a NFA Letter. Data on sites in Ohio’s Voluntary Action Program were obtained from the Ohio EPA website and only covers sites that received a Covenant-Not-to-Sue from 2008 to 2012.
Demographic data were attained from the 2008-2013 American Community Survey 5-Year estimates at the tract level via Social Explorer. All datasets contained latitude and longitude coordinates that were fed into the Texas A&M University Geocoding Services reverse geocoder to identify each site’s census tract, which was then cross-referenced with the demographic variables and urban sprawl composite scores. Identification of the political representation of each site was also carried out with the latitude and longitude coordinates and cross-referenced with a table join of district data obtained from each state’s secretary of state office in the QGIS software package.
The composite urban sprawl score was attained from Measuring Urban Sprawl and Validating Sprawl Measures, a research effort executed with funding from the following: The National Cancer Institute, National Institutes of Health, the Ford Foundation, and Smart Growth America. This score was created via principal components analysis of numerous variables including gross population density, gross employment density, degree of job mixing, average walk score, percentage of small urban blocks, average block size, intersection density, and percentage of four-or-more-way intersections. The score was normalized to create an index with a mean of 100 and a standard deviation of 25 (Ewing & Hamidi, 2010). A higher score represents a more compact tract and a lower score represents a more sprawling tract. The IRCHS is used solely in the Colorado VCP model as a stand-in because the Colorado Department of Public Health and the Environment does not maintain records of past use at enrolled sites. The IRCHS is a composite index and rank of a variety of chemicals’ toxicity based on their threat to human health and persistence in air, water, and soil (Eckerd & Keeler, 2012). A table of scores of the most common contaminants listed in the dataset can be found in Table 1.
Common Brownfield Contaminants and Their IRCHS Scores.
Note. IRCHS = Indiana Relative Chemical Hazard Score.
Hypotheses and Theoretical Expectations
Generally, it is hypothesized that census tracts of high socioeconomic status as defined by higher percentages of White citizens, higher educational attainment, and lower rates of public assistance are more likely to contain brownfield redeveloped for nonindustrial purposes. This hypothesis is consistent with numerous EJ studies. More specifically, high socioeconomic status in the surrounding census tract is expected to increase the odds of greenspace and residential redevelopment and reduce the odds of commercial and mixed use relative to the industrial baseline category. These hypotheses are based on the results of the hedonic price models described above which attribute the greatest improvement in real estate value to greenspace, residential, and commercial redevelopment. Communities of relatively higher socioeconomic status have been shown to display a “marginal willingness to pay” to distance themselves from environmental disamenities. This effect is expected to occur for nonindustrial redevelopment outcomes as result of real estate demand or increased civic capacity.
Past industrial use is expected to increase the odds of industrial and commercial redevelopment and decrease the odds of mixed, residential, and greenspace redevelopment across all programs. This hypothesis is based on the history of land use planning and zoning throughout the United States that has been standard practice for nearly a century. It is also assumed that industrial use is associated with a greater past contamination of a brownfield property and a higher IRCHS. This score is essentially being used as a stand-in variable in place of the past industrial use variable that is not recorded by the Colorado Department of Public Health and the Environment in the VCP. Given the probusiness platform of the Republican Party, the presence of a Republican legislator at the time of assessment (Federal sites) or awarding of an NFA or Covenant Not to Sue (VCP sites) is expected to increase the odds of commercial and industrial redevelopment and reduce the odds of greenspace, residential, and mixed use. Please note the inclusion of this variable does not suggest that a member of Congress directly influences a lot of brownfield redevelopment choices, rather it suggests that conservative areas will implicitly be less likely to develop greenspace and have a higher likelihood of indirect business development.
A higher urban sprawl score (a less sprawling tract) is expected to increase the odds of commercial, mixed, and residential uses and decrease the odds of industrial and greenspace uses. The rationale for this hypothesis is that there is a greater potential for return on investment for redevelopment types in dense and connected urban areas than industrial and greenspace redevelopment because land is comparatively more cost effective and land intensive. The inverse of this hypothesis, however, must be acknowledged, as commercial, residential, and mixed land uses are primary drivers of urban sprawl. In the interest of a concise hypothesis, Table 2 provides a summary of the directionality expected for each coefficient. Table 3 provides a summary of expectations for the prevalence of redevelopment outcomes for each program.
Summary of Expectations for Multinomial Logistic Regression Odds Ratios.
A “+” indicates that the indpendent variable increases the likelihood of the outcome, a “-” decreases the likelihood.
Summary of Expectations for Redevelopment Outcome Prevalence by Program.
Note. “1” indicates most common, “5” indicates least common. VCP = Voluntary Cleanup Program. VAP = Voluntary Action Program.
As suggested by Alberini (2007), high demand for residential land in Colorado is expected to increase the overall likelihood that brownfields are redeveloped for such purposes relative to all other outcomes due to high potential return on investment, particularly at sites within the VCP. Commercial outcomes are expected to be the next most prevalent, followed by mixed use, industrial, and greenspace. Ohio’s long history as a leading state in manufacturing is expected to increase the overall likelihood of industrial development relative to all other outcomes. Greenspace redevelopment is expected to be the least likely in either state VCP but more likely at Federal sites due to a greater number of sites owned by government entities with less pressure to maximize return on investment.
Limitations
Several limitations in the research design must be noted. First, the datasets used in the development of the model for Federal sites only include data on properties with past and future use, which resulted in the analysis of about 15% of the total number of sites within each state. This share of sites may constitute a random sample of Federal brownfield properties, but it may also be subject to selection bias on the part of the EPA and entities reporting property information to CERCLIS. Requests to the EPA to clarify why such a large portion of properties lacked past and future use information were not answered, but it appears that submission of such data is at the discretion of property owners and grant recipients. A notable share of the Federal properties listed are owned by local governments, and cursory research on a handful of the properties possessing all relevant information suggests that they are high-profile redevelopment projects. Many of these can be found in EPA press releases and other documents. Similarly, many properties with missing information turn up little or no documentation by EPA or other entities. The possibility of selection bias in the datasets used to construct the model could skew or dilute the results. Second, missing data also proved to be somewhat problematic in the construction of the Republican legislator variable.
Finally, the redevelopment categories (industrial, commercial, greenspace, residential, and mixed use) used in the model are very general categories. These categories cannot capture the subtleties of redevelopment types completely. For instance, a residential subdivision is treated the same as public housing and a fast food restaurant is treated the same as high-value office space. These broad categories help to improve the generalizability of the study, as the risk-based cleanup standards imposed at brownfield properties follow a similar breakdown. Furthermore, these broad categories are focused upon reuse, the ultimate goal of a brownfield project, and are a metric that is unencumbered by the numerous site-specific factors that are not captured in large-sample datasets.
Results
Federal Brownfields in Colorado
Table 4 provides coefficients, standard errors, and odds ratios from a multinomial logistic regression of redevelopment outcomes at Federal Brownfield sites in Colorado. The sole variable attaining significance in the model was past industrial use, which reduced the likelihood of all nonindustrial redevelopment outcomes.
Multinomial Logistic Regression Summary of Redevelopment Outcomes at Federal Brownfields in Colorado.
Note. Standard errors in parentheses. OR = odds ratio; AIC = Akaike information criterion.
p < .05. **p < .01. ***p < .001 (one-tailed test).
Colorado VCP
The Colorado VCP model utilized the IRCHS in place of the past industrial use variable, as the Colorado Department of Public Health and the Environment does not track past use. This hazard score failed to achieve significance, and is a likely contributor to the relatively poor fit of the model. A greater number of citizens possessing a bachelor’s degree or higher in the surrounding census tract significantly increased the odds of commercial, mixed-use, or residential redevelopment. Table 5 describes the odds ratios and fit statistics of the model and Figure 1 illustrates the effect of educational attainment on the predicted probability of each redevelopment outcome.
Multinomial Logistic Regression ORs of Redevelopment Outcomes at Brownfields Enrolled in Colorado’s Voluntary Cleanup Program.
Note. Standard errors in parentheses. OR = odds ratio; AIC = Akaike information criterion.
p < .05. **p < .01. ***p < .001 (one-tailed test).
Average effect plot of educational attainment on redevelopment outcomes in Colorado’s VCP.
Federal Brownfields in Ohio
The odds of all nonindustrial outcomes were significantly reduced by past industrial use of the property. In addition, a greater number of White citizens within the surrounding census tract significantly reduced the likelihood of mixed-use redevelopment. Higher observed percentages of citizens possessing a bachelor’s degree or higher within the census tract containing a brownfield increased the odds of all nonindustrial redevelopment but was only significant for residential redevelopment. A greater percentage of welfare recipients significantly increased the likelihood of residential redevelopment as well. The presence of a brownfield in a Congressional district represented by a Republican lowered the odds of all nonindustrial redevelopment outcomes but was only significant for residential outcomes, the likelihood of which was reduced considerably. Table 6 provides odds ratios and fit statistics for the model. Figure 2, Figure 3, and Figure 4 illustrate the effect of each significant variable on the predicted probabilities of redevelopment outcomes.
Multinomial Logistic Regression Summary of Redevelopment Outcomes at Federal Brownfield Sites in Ohio.
Note. Standard errors in parentheses. OR = odds ratio; AIC = Akaike information criterion.
p < .05. **p < .01. ***p < .001 (one-tailed test).
Average effect plot of White percentage on redevelopment outcomes at Federal brownfields in Ohio.
Average effect plot of educational attainment on redevelopment outcomes at Federal brownfields in Ohio.
Average effect plot of welfare recipients on redevelopment outcomes at Federal brownfields in Ohio.
Just as observed at Federal Brownfield properties in Ohio, past industrial use was the strongest predictor of redevelopment outcomes for brownfields enrolled in Ohio’s Voluntary Action Program. Past industrial use significantly reduced the likelihood of commercial, mixed use, and residential redevelopment and significantly increased the odds of redevelopment for greenspace purposes. Increased percentages of White citizens in the surrounding census tract reduced the likelihood of residential redevelopment, while increased percentages of citizens possessing a bachelor’s degree or higher increased the likelihood of greenspace or mixed-use redevelopment. A greater number of citizens receiving welfare in the census tract reduced the likelihood of residential redevelopment. A higher census tract sprawl score increased the likelihood of all nonindustrial outcomes, but was only significant for mixed-use and residential outcomes. Table 7 provides odds ratios and fit statistics for the model. Figure 5, Figure 6, Figure 7, and Figure 8 illustrate the effect of the significant variables on the predicted probability of each redevelopment outcome.
Multinomial Logistic Regression ORs of Redevelopment Outcomes at Brownfields Enrolled in Ohio’s Voluntary Action Program.
Note. Standard errors in parentheses. OR = odds ratio; AIC = Akaike information criterion.
p < .05. **p < .01. ***p < .001 (one-tailed test).
Average effect plot of White percentage for redevelopment outcomes in Ohio’s VAP.
Average effect plot of educational attainment for redevelopment outcomes in Ohio’s VAP.
Average effect plot of welfare for redevelopment outcomes in Ohio’s VAP.
Average effect plot of urban sprawl for redevelopment outcomes in Ohio’s VAP.
Discussion
Generally, use of the multinomial logistic regression model to assess the environmental justice of brownfield redevelopment in Colorado and Ohio provided mixed answers to our hypotheses. All variables excluding the IRCHS score used in the Colorado VCP model attained significance, but several were limited to specific models or outcomes. Furthermore, some results seemed to be somewhat contradictory. Neither the census tract sprawl score nor representation of the tract by a Republican Congressman yielded notable results. An explanation of the results follows, as well as possible rationales for their directionality and significance. Finally, implications of the results and the potential for future research will be explored. Past use of properties for industrial purposes is the most consistent explanatory variable in the analysis. Furthermore, these variables attained the highest degree of significance.
The model utilized here does suggest two problems with brownfield redevelopment for those concerned with environmental justice. As noted throughout existing literature, communities of low socioeconomic status are considerably more likely to contain industrial properties that pose a variety of problems resulting from both past and present contamination. The results of this analysis further complicate this trend in that repurposing of former industrial sites for a nonindustrial purpose is significantly unlikely. This result is not surprising and many would argue to be rational under land use zoning practices. However, these long-standing zoning practices have led to an unequal distribution of environmental risk that is further perpetuated by the likely reintroduction of new pollution sources and the lowest remediation standards under brownfield redevelopment programs. Not surprisingly, redevelopment for industrial purposes also offers the least in property value improvement for the surrounding area as described in existing hedonic price models. This unequal distribution of risk and benefits deserves consideration as brownfield redevelopment policy evolves. Socioeconomic variables also provided mixed answers to our research questions. The percentage of White citizens and citizens receiving welfare benefits attained limited significance. However, educational attainment did prove to be a valuable predictor. Census tracts with a higher percentage of citizens within the census tract possessing a bachelor’s degree or greater increased the likelihood of several nonindustrial redevelopment purposes in all but the model for Federal properties in Colorado. These results will be explored in more detail below. The significance of this analysis is that it moves the specific knowledge of Brownfield assessment and general EJ evaluations to consider that both have socially preferred uses which favor the markets of commercial and mixed use over the redevelopment land uses of residential and greenspace.
Socioeconomic status was identified by the percentages of White citizens, citizens possessing a bachelor’s degree or higher, and citizens receiving public assistance. While each of the variables attained significance in at least one of the four models, none of the variables attained significance in all four of the models or in each redevelopment outcome. This inconsistency complicates interpretation and explanation, but the findings in this research are similar to those of Eckerd and Keeler (2012) who found that progression of a brownfield through the remediation process was slowed in census tracts with higher percentages of minorities. Surprisingly, however, progression was quickened in tracts with higher percentages of citizens receiving public assistance. These findings are not necessarily contradictory, and they seem to reflect the difficulty of assessing disadvantaged communities with limited demographic variables. The trend is significant as it demonstrates the importance of assessing the social justice of Brownfields. In regard to the demographics assessed in this research, neither Colorado’s nor Ohio’s Federal brownfields displayed any explicit trends in regard to a failure to promote environmental justice. Despite limited statistical significance, a greater degree of educational attainment increased the likelihood of all nonindustrial redevelopment in both models and attained relatively high significance for residential outcomes in Ohio. This finding likely reflects residential sorting in which individuals self-select to live in areas consistent with their own values and preferences. More simply, individuals with a high level of educational attainment working in “white-collar” professions would be less motivated to locate near industrial areas than an individual with lower educational attainment working in manufacturing.
The Federal Brownfield Program does appear to be successful in mitigating the racial disparities historically associated with brownfields. In both Colorado and Ohio, an increased percentage of White citizens reduced the likelihood of all nonindustrial outcomes. While the variable attained only limited statistical significance, it does suggest that the Program is adequately distributing environmental and economic benefits across racial demographics. A similar trend was observed for the percentage of citizens receiving welfare, but again, the variable attained only limited significance. As such, the hypothesis that socioeconomic status affects the likelihood of redevelopment outcomes is rejected for the Federal Brownfield properties examined in Colorado and Ohio. This is notably dissimilar than the findings by Eckerd and Keeler (2012), Lee and Mohai (2013), and McCarthy (2009), all of which found that various indicators of socioeconomic status were significant in the targeting, funding, and progression of brownfield properties through the redevelopment process.
Use of the composite urban sprawl score led to disparate results between the two states that did not attain statistical significance. In the case of Colorado, a higher sprawl score (less sprawling tract) reduced the odds of all nonindustrial redevelopment. The opposite was true in Ohio with the exception of greenspace redevelopment. This result defies easy interpretation but may reflect Colorado’s high population growth and a tendency to keep industry in older, more compact urban cores, while former industrial sites in Ohio’s urban cores are being redeveloped to adapt to a service-based economy quite different from the state’s industrial past. Representation by a Republican Congressman attained little statistical significance in the model. Republican representation decreased the likelihood of all nonindustrial outcomes in both states, excluding greenspace redevelopment in Colorado. However, the variable only attained significance for residential outcomes in Ohio. Because of the limited nature of the variable and its relative simplicity, the ability to draw conclusions is limited, particularly when considering the urban and rural divide between Republicans and Democrats.
The IRCHS, used in the Colorado VCP model as a substitute for past industrial use, reduced the odds of all nonindustrial redevelopment excluding greenspace but failed to attain significance. Past industrial use in Ohio significantly reduced the likelihood of all nonindustrial redevelopment outcomes. Similar to the Federal program results, educational attainment appears to be a fairly consistent predictor, increasing the likelihood of all nonindustrial redevelopment in Colorado excluding greenspace outcomes. The effect was statistically significant for mixed use and residential outcomes. Surprisingly, increasing educational attainment significantly decreased the likelihood of residential outcomes in Ohio’s VAP while increasing the likelihood of all other nonindustrial outcomes. Results for welfare were also mixed, and the only significant result was a profound decrease in the likelihood of residential redevelopment in Ohio’s VAP as the percentage of White citizens increased. Similarly, the percentage of White citizens achieved mixed results, and the lone outcome it attained significance in was reduced likelihood of residential redevelopment in Ohio’s VAP.
The relationship between urban sprawl and brownfield redevelopment was similar in both Federal and VCPs. In the case of Colorado, less sprawling tracts were less likely to receive nonindustrial outcomes. The opposite trend was true for Ohio’s VAP, which also contained the only outcome in which the tract urban sprawl variable attained significance, mixed-use redevelopment. Republican Congressional representation decreased the likelihood of all nonindustrial outcomes in Colorado, while the opposite was true in Ohio with the exception of commercial redevelopment.
Conclusion
Redevelopment of the thousands of brownfields throughout the country offer considerable potential to promote environmental justice. Advocates have scrutinized the Federal program since its inception in the mid-1990s, analyzing the program’s grant distribution and the progress of funded properties throughout the redevelopment process. Generally, the program has succeeded in allocating Federal funds to cities and regions hit hardest by a nationwide shift from manufacturing to a service-based economy. However, comparatively little attention has been paid to the VCPs that administer a considerably larger number of cleanups. Environmental justice advocates have raised concerns that the risk-based cleanup standards imposed under VCPs do not adequately protect human health. While this research does not take on the explicit risk standards imposed under such programs, it does consider the perception of risk that is established in existing hedonic price models of brownfield remediation. Not surprisingly, the greatest benefit to property values near a brownfield after remediation are residential and greenspace projects.
Notable disparities were expected between the Federal and state programs but the results across all four models were somewhat similar. Case selection for this analysis was swing states but future research must consider a larger number of states, and preferably those from differing regions. Past industrial use of a brownfield property was the most significant and consistent variable analyzed. It consistently reduced the odds of residential redevelopment across all programs and effectively lowered the likelihood to zero at Colorado Federal and Ohio VAP properties. With the exception of greenspace outcomes in Ohio’s VAP, past industrial use lowered the likelihood of all nonindustrial redevelopment. Given the nature of long-standing zoning practices, this finding is of some concern due to the widely acknowledged proximity of low socioeconomic status populations to sites that use, store, and produce hazardous materials. Although brownfield redevelopment has restored activity to idled properties, the lower cleanup standards associated with industrial redevelopment may not be satisfactorily mitigating risk for nearby populations. This is also problematic as such redevelopment simultaneously introduces new sources of environmental risk to the surrounding neighborhoods. Furthermore, industrial redevelopment of a brownfield property is least advantageous for the real estate values of surrounding property owners. Advocates of redevelopment like the U.S. Conference of Mayors cited the potential of various cleanup programs and new industry to create jobs, but as noted by Howland (2007), little is known about the extent to which jobs created by redevelopment go to individuals in the immediate area.
Of the three socioeconomic variables, educational attainment was the most consistently significant. Census tracts with lower educational attainment are generally less likely to receive the risk mitigation and property value increases afforded by nonindustrial redevelopment. This observation is probably the result of greater civic capacity observed in more educated areas with citizens that are more likely to participate in public comment periods. Another potential reason is the greater income and “willingness to pay” of college-educated individuals to distance themselves from sources of environmental risk. This result is best demonstrated by the median household income variable, which displayed such a high degree of colinearity with other socioeconomic variables that it could not be included in the model. This result may also be the result of high demand for housing that makes residential redevelopment highly lucrative, as observed by Alberini in Colorado (2007). In light of these arguments, it should be noted that the multinomial logit model did not include the percentage of citizen’s employment by sector. For example, a high degree of manufacturing employment in the area of a brownfield may increase pressure for industrial redevelopment. This variable could be significant, but was not considered due to sample size limitations. The multinomial logistic regression model used in this study was formulated to address three typical environmental justice variables as well as two novel variables, and the addition of another variable would not allow the algorithm to converge in the R software package.
Analysis of urban sprawl scores provided seemingly contradictory results, as the directionality of the likelihoods shifted between states. One possible explanation for this observation is the rapid population growth of Colorado’s metropolitan areas (U.S. Census Bureau, 2014a), pushing the boundaries of cities outward while manufacturing jobs in the urban core declined. The observations in Ohio are not as easily explained, however, but are likely the result of a greater number of former industrial properties being transformed into purposes more congruent with a service-based economy. The hypothesis regarding urban sprawl is rejected, primarily because graphing of predicted probabilities of outcomes across the length of the variable shows considerably more complex trends than are implied by the multinomial logit coefficients. Representation by a Republican legislator achieved significance only for mixed-use outcomes at Federal sites in Colorado and residential outcomes at Federal sites in Ohio. The odds of both outcomes were reduced by Republican representation. The limited significance of the variable in these two cases may be the result of limited sample size or the relative simplicity of the variable. For that reason, this objective research is hesitant to reach any conclusion regarding the variable.
A key difference between the two state programs may be the relative strength of their economies. As of 2010, Colorado’s per capita real gross domestic product (GDP) was US$46,757, 54 significantly higher than Ohio’s at US$36,936 (Avery, Siebenec, & Tate, 2011). A vital part of Colorado’s successful economy is its consistent growth, as it has reliably been one of the fastest growing states (U.S. Census Bureau, 2014b). The same cannot be said for Ohio, which has seen considerably slower growth rates (U.S. Census Bureau, 2014b). Furthermore, the two states have vastly different economies, with 12.8% of Ohio’s workforce employed in manufacturing, one of the highest rates in the country, compared with the relatively low 5.7% in Colorado (National Association of Manufacturers, 2014). This difference in manufacturing employment combined with increased growth could be a key contributor to the disparity in redevelopment outcomes and possibly the reason that industrial redevelopment was more popular at Ohio’s Federal and VAP properties. Comparison of these two distinctly different states makes the results of this research generalizable, as numerous other states can be compared. Table 8 describes the average population growth rate and per capita GDP of five thriving states and five states that have seen little growth.
Comparison of Population Growth and GDP Per Capita in 10 States.
Note. GDP = gross domestic product.
The average population growth percentage is the average of annual population growth rate in each state from 2001 to 2012 (U.S. Census Bureau, 2014b). High growth states would likely have better performing redevelopment programs due to increased demand for real estate, while low growth and declining states would likely struggle to repurpose already developed land. Similarly, these states also have a higher average GDP per capita, indicating greater overall economic health. Although this research model did not use any economic metrics, a stronger statewide economy would seem to improve the performance of economic redevelopment programs like brownfield cleanup and remediation. Future research should consider economic metrics as explanatory variables alongside the demographics used in this effort. Brownfield redevelopment policy is inherently complex; the language of both Federal and state programs is notably focused upon the reduction of liability for property owners and regaining of property tax revenue for local government, but the programs are administered by environmental protection agencies. This is not to suggest that such agencies are not up to the task, as the reduction of environmental risk is also a primary goal of redevelopment programs, but does illustrate the multidisciplinary nature of the problem (Daley & Layton, 2004).
Effective brownfield redevelopment policy must balance the often competing interests of economic development and environmental justice. This research presents a new model for policy makers and scholars to examine the equity of publicly funded brownfield redevelopment with variables representing the numerous and varied interests at play. This study attempts to establish how demographics of the census tract containing the brownfield affect the likelihood of obviously disparate redevelopment outcomes using socioeconomic variables common to environmental justice studies. Of the three variables used, increased educational attainment offered the most significant and consistent results, generally increasing the likelihood of nonindustrial outcomes offering the greatest community benefit. The limited and mixed results of the race and welfare variables suggest that benefit-maximizing redevelopment is not solely obtained by White, affluent communities. This is not to say that environmental justice concerns are invalid, however. Simply put, brownfield properties that had once been used for industrial purposes are significantly unlikely to be redeveloped for any other purpose, and the use of public funds to encourage the introduction or reintroduction of environmental risk sources may be considered questionable. A cursory review of the requirements and processes of most VCPs reveals that business interests are well-suited by the variety of financial incentives and liability reductions available. However, as the state agencies that administer these programs are recipients of Federal Brownfield Grants, additional scrutiny of VCPs is necessary. Similarly, it is important to distinguish Federal Brownfields and VCP sites from the more high-profile properties included in Superfund. While Superfund utilizes highly technical assessment techniques and scoring for inclusion of a new site, a Federal Brownfield can be nearly any property with known or suspected contamination.
A Federal Brownfield may be a site that narrowly missed the threshold for inclusion in Superfund, or it may merely be a site that a local government entity has decided is a priority and chosen to allocate Brownfield grant funds toward. Given that zoning decisions are the authority of local governments, the lack of zoning change observed in this model may be indicative of a flaw at the basis of the Brownfields Program from the perspective of environmental justice advocates. Focusing on redevelopment and land use change provides an understanding of what both the Brownfields Program and VCPs are actually delivering. As indicated by this study, the effectiveness of these programs at remedying disproportionate environmental risk remains unclear, but a similar model utilized by policy makers and regulators could be a simple method to monitor the programs. As uncertainties related to policy prioritization surround the Federal Brownfields Program, remediation and redevelopment of brownfield properties under VCPs could be a new focus point for advocates of environmental justice.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
