Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Jul 21.
Published in final edited form as: Rural Sociol. 2021 Feb 14;86(4):647–672. doi: 10.1111/ruso.12367

Proposed pipelines and environmental justice: Exploring the association between race, socioeconomic status, and pipeline proposals in the United States

J Strube a, B Thiede b, W Auch c
PMCID: PMC9302603  NIHMSID: NIHMS1815228  PMID: 35874273

Abstract

The current natural gas and oil boom in North America requires new pipelines which pose environmental risks from the wellhead to their destinations. The environmental justice literature suggests that minority populations, people with low socio-economic status, and rural communities are disproportionally exposed to risks associated with potentially harmful land uses. Using data from the 2015 five-year American Community Survey and pipeline route data compiled by The FracTracker Alliance, this study tests whether the above assumptions are true for proposed FERC permitted natural gas transmission pipelines in the United States for which planned routes have been made available. Using binary logistic regression, the study provides only limited, and in some cases contradictory, support for these hypotheses. Although a higher share of highly educated residents significantly decreases the likelihood of a pipeline proposal in a census tract, a higher poverty rate also significantly lowers this probability. Only the share of Black and Asian residents is significantly (negatively) associated with pipeline proposals. However, to test whether this holds true for built pipelines, reliable routing data are needed, which are considered confidential in the United States.

Keywords: pipelines, environmental justice, oil, natural gas, energy

1. Introduction

North America has experienced a natural gas and oil boom throughout the past decade (Wang et al.. 2012; Kinchy et al.. 2014), increasing the demand for new pipelines. At the same time, many older pipelines need to be replaced (Silver 2019). This expansion raises the question of whether the environmental risks of pipelines are equally distributed among the population, or if certain demographic groups are more likely to be exposed.

This article addresses this question for the United States by combining and analyzing census records and novel data on proposed pipeline routes. We test whether key demographic and socio-economic characteristics of a census tract—racial composition, educational attainment, and poverty—predict the probability of a pipeline being proposed in it. We also examine the differential associations between these predictors and the likelihood of proposed pipelines across the rural-urban continuum. We discuss our research within the context of the Environmental Justice literature. This body of work often finds that environmental risks are concentrated near disadvantaged communities. Whether this holds true for currently proposed pipelines, however, has not yet been tested.

2. Risks and opportunities of pipelines

There has been much controversy about the social, environmental, and economic impacts of the hydrocarbon boom. Seen by many as an opportunity for economic development, risks include environmental damages caused by oil and natural gas extraction, processing and consumption (O’Rourke and Connolly 2003), and social disturbances as a result of the short-term boom and bust economy (Schafft, Borlu, and Glenna 2013; Jacquet 2014; Fernando and Cooley 2016; Schafft et al.. 2017; Mayer, Olson-Hazboun, and Malin 2018). While less environmentally hazardous than transporting oil by trucks and trains (Clay et al. 2019), pipelines nonetheless pose significant risks to the surrounding environment, including landslides, leaks, and explosions (O’Rourke and Connolly 2003; Parfomak et al. 2013; PMHSA 2017). The US government classifiesi much data around pipelines—including routing—to protect pipelines from hacking and terror attacks which would be disastrous for the environment and economy (U.S. Government Accountability Office 2019). Spilled oil may contaminate water supplies and destroy ecosystems, compromise public health, and impact local economies (U.S. Bureau of Oceans and International Environmental and Scientific Affairs 2014).

The American public is divided on the subject of pipeline development. Supporters highlight construction jobs (Huber and Bowe 2014; Skinner and Sweeney 2012), economic growth, and increasing energy security (Energy Policy Research Foundation 2010; U.S. Chamber of Commerce Global Energy Institute 2017) as potential benefits. Opponents, on the other hand, focus on the long-term climatic effects as well as the local risks outlined above (Parfomak et al. 2013). Pipelines make exploiting oil and natural gas more profitable by reducing transport costs (Moore et al. 2011; Hoberg 2013). This contradicts the need to increase, rather than lower, the production costs for fossil fuels to slow global climate change (Le Billion and Kristoffersen 2019). For the proposed Keystone XL pipeline, the prospects of construction jobs close to the pipeline outweighed locally less tangible concerns about climate change, even among supporters of strong environmental protections on other issues (Gravelle and Lachapelle 2015). This complicates the idea of pipelines as Locally Unwanted Land Uses (LULUs), emphasizing the fact that for some pipelines are in fact locally desired.

In recent years, resistance against new pipelines has been mounted by various Indigenous communities. Prominent examples include blockades at Standing Rock in South Dakota and in the Wet’suwet’en First Nation territory in British Columbia. This resistance against pipelines must be understood in the context of the fossil fuel industry’s long history of dispossession of Indigenous lands (LaDuke 1999; O’Rourke and Connolly 2003). Of course, Indigenous communities are not the only vulnerable group affected by pipelines. For example, the Pacific Pipeline in Southern California transects almost exclusively neighborhoods with disproportionately high rates of minority populations (O’Rourke and Connolly 2003). These experiences prompt the question whether pipelines are disproportionally more likely to be constructed in areas with predominantly minority populations and residents of low socioeconomic status.

Examining developments in Canada, Scott (2013) argues that pipelines commit the economy to fossil capitalism, in which communities are unevenly burdened depending on their location within the network of extraction, transportation, refining, and consumption, for decades to come. According to this argument, each pipeline routing proposal integrates a bundle of spatial cost-benefit-calculations and decisions, with uneven effects for different social groups. Yet, whether the locations of (potential) environmental harms within the hydrocarbon industry correlate with the spatial distribution of race and class has not yet been studied in depth. We address this gap by questioning whether key demographic and socio-economic parameter predict the routing of proposed pipelines.

3. Environmental justice in extractive industries

Following the Environmental Justice literature, we hypothesize that racial and ethnic minorities and communities of poor socio-economic status are more likely to live in areas selected for new pipelines. To substantiate this expectation, we briefly review this literature as it relates to other environmentally harmful industries.

Critical geographers and environmental sociologists have long established that vulnerable groups in North America, such as racial minorities or low-income communities, are more likely to live near environmentally hazardous land uses (Bullard 1993; Gosine and Teelucksingh 2008). While some scholars find that race is most likely to predict this relationship (Bullard 1990; Mohai and Bryant 1992), others find low income to be the significant independent variable (Hamilton 1995; Kriesel, Centner, and Keeler 1996; Tarrant and Cordell 1999; Jerrett et al. 2001). Yet others suggest that both factors operate together (Glickman 1994; Foreman 1996). One proposed explanation for the greater vulnerability of racial minorities and low-income communities is that these groups often do not possess the social capital to resist unwanted developments (Bullard 1993). The availability of natural resources and the lack of social capital to resist unwanted land uses, concentrate many environmental harms in rural places (Bullard 1998; Ashwood and MacTavish 2016; Eaton and Kinchy 2016; Kelly-Reif and Wing 2016).

More recent literature, however, cautions against presuming that risky land uses are necessarily unwanted. In particular, politically conservative actors often support hazardous energy development, including in their own communities (Smith 2002; Gravelle and Lachapelle 2015; Clarke et al.. 2016). Political partisanship, proximity to targeted regions, and neighborly solidarity drive some individuals to embrace hydraulic fracturing as an economic opportunity—in some cases despite knowledge of the environmental risks (Dokshin 2016; Jerolmack and Walker 2018). Mayer and Malin (2019) show that residents of oil and gas dependent communities significantly oppose strong regulations of the industry when they perceive economic benefits for themselves and their community, while those who fear economic disadvantages endorse them. Evensen and Stedman (2018) find that local residents perceive hydraulic fracturing based on their ability to enjoy certain aspects of rural life, such as an aesthetically pleasing landscape or a specific community character based on agriculture.ii

4. Methods

4.1. Conceptual Model

Are proposed pipelines likely to reproduce the uneven distribution of environmental risk observed in other industries? With respect to distributive environmental justice, we posit that the population composition along a pipeline corridor should not be significantly different than that of the county surrounding it (if the pipeline was to be built at all). We assume that start and end points of proposed pipelines are fixed due to the location of hydrocarbon sources, processing industry, and major seaports in North America. The routing in between, however, is more flexible. While it may not be practicable to route a pipeline through a different county, leading a pipeline around a community within a county (operationalized as a census tract) is more feasible. Hence, for the purpose of our analysis, we define environmental injustice as a certain population’s significant, disproportional higher exposure to the risk of a pipeline spill or explosion, operationalized as the existence of a pipeline proposal within a census tract. Given our hypothesis that minority communities and those with lower socio-economic impact are more likely to live in areas selected for new pipelines, we thus expect a significant positive association between a tract’s share of these population and the likelihood of a pipeline proposal within the tract. We acknowledge that this operationalization does not consider the impact of the fossil industry to global climate change, which already disproportionally burdens disadvantaged populations regardless of pipeline routing (Roberts and Park 2006; Mohai et al. 2009).

4.2. Data

We construct a census tract-level dataset of proposed pipeline locations and population characteristics. We begin by compiling and vectorizing the routes of all known North American oil and gas pipeline proposals (Figure 1) (2015 to 2017) based on publications (predominately static image files) from the petro-industry and media reportsiii. The data are based on what pipeline companies voluntarily disclose or journalists uncover, since pipeline proposals and the routes of existing pipelines are considered confidential by both the companies and the US government. Many of these proposals present simplified representations of the proposed routes that do not necessarily represent the accurate and final plans. The data are also unlikely to be complete, since not all pipeline proposals and their routes are shared with the public. Although this limits the reliability of our findings, our dataset is the most complete dataset on pipeline proposals in the US that are publicly available. Considering the lack of research on the subject, the data allow for a satisfactory assessment of different social strata’s exposure to new pipeline development. However, pipeline operators and the US Pipeline and Hazardous Materials Safety Administration need to release reliable and complete data to engage the public in meaningful discussions about pipeline routing.

Figure 1.

Figure 1.

Open in a new tab

Current pipeline proposals in the United States.

Our dataset represents all proposals with routes publicly available in June 2017 and are cropped to the United States (see Figure 1). We excluded pipelines shorter than 300 km and excluded the first and last 100 km of each route. This accounts for the fact that the routing of a pipeline is fixed near the well and its destination, but in between it is open to local deliberations. Our analytic sample, then, includes 84 pipelines.

We extract tract level demographic and socioeconomic data on race, educational attainment, poverty, and shapefiles for the tract geography from the 2011–2015 five-year estimates of the American Community Survey (IPUMS NHGIS, Manson 2017)iv. We choose census tracts as the unit of analyses for three reasons. First, commonly, racial and socio-economic sub-populations are spatially clustered within counties. Second, pipeline planning often involves choices between alternative routes within the same county (e.g., one neighborhood versus another). Third, the risks associated with pipeline exposure are greater for populations residing in close proximity to the pipeline, which is better proxied by tract-level measures of exposure than county-level measures.

4.3. Measures

Our dependent variable is a binary indicator of whether a pipeline is proposed in a given census tract. We construct this variable by intersecting the routes of our analytical sample with all census tracts in ESRI ArcGIS. Tracts with a proposed pipeline are coded as 1 and all others as 0.

Our explanatory variables of interest are urban-metropolitan codes, racial composition, educational attainment, and the poverty headcount ratio (i.e., the poverty rate). We create an urban-metropolitan code matrix to combine the 2013 OMB metropolitan and non-metropolitan delineations with a measure of rurality based on population density. First, we divide census tracts in rural and urban. We characterize urban tracts by a population density of more than 1000 people per square mile; and code tracts with a lower density as rural. Finally, we cross-classify tracts using the urban/rural and metropolitan/non-metropolitan indicators, resulting in a four-category typology: rural non-metropolitan, rural metropolitan, urban non-metropolitan, and urban metropolitan. Since our analysis focuses on demographic differences between census tracts, it is important to use a scale that picks up intra-county differences of rurality (urbanity measure based on population density) as well as the relation of a location to metropolitan centers (Rural Continuum Codes, that are only available at the county level). Urban and rural areas tend to have different socio-economic and demographic characteristics (e.g. Berry and Hirschl 2017; Thiede, Lichter, and Slack 2018). Since low density tracts are generally larger, they would be more likely to host a pipeline. For these reasons we expect the association between socioeconomic and demographic characteristics of a census tract and pipeline proposals to vary across these different locations.

As demographic and socio-economic predictors, we use the mean share of different racial groups within a census tract, and educational attainment and poverty ratio as proxies for socio-economic status. We focus on race and socio-economic status as our variables of interest because of their frequent correlation with the siting of other environmentally hazardous developments (see review of the Environmental Justice literature above).

4.4. Data Analysis

We begin by producing descriptive statistics for the whole population and one for each category of urban-metropolitan codes. Within each category, we present descriptive statistic for the entire nation and for the restricted population of tracts in counties with pipeline proposals (Table 1).

Table 1.

Descriptive statistics of census tracts in the US (full and restricted sample by urban-metro-codes).

All tracts
 Low-density, non-metropolitan
 Low-density, metropolitan
 High-density, non-metropolitan
 High-density, metropolitan
Contiguous United States Counties with pipelines proposed Contiguous United States Counties with pipelines proposed Contiguous United States Counties with pipelines proposed Contiguous United States Counties with pipelines proposed Contiguous United States Counties with pipelines proposed
mean mean mean mean mean mean mean mean mean mean
Population (people) 4382 (2092) 4269.4 (2233) 3840 (1640) 3715 (1547) 4667 (2345) 4723 (2415) 3936 (1565) 3879 (1519) 4424 (2081) 4257 (2309)
Population density (people per square mile) 5345 (12031) 3591.454 (6738) 139 (207) 154 (224) 341 (289) 347 (284) 2529 (1713) 2829 (2125) 8473 (14507) 5949 (8122)
Non-hispanic White (in %) 62.9% (30.1) 68.8% (29.0) 81.1% (21.5) 85.4% (17.7) 78.2% (21.7) 83.9% (17.6) 71.3% (25.7) 76.0% (23.3) 52.7% (30.1) 57.4% (30.4)
Hispanic (in %) 15.8% (21.2) 12.3% (18.1) 6.6% (12.1) 5.8% (10.5) 9.2% (14.9) 6.9% (10.4) 12.9% (19.5) 10.9% (18.5) 20.5% (23.3) 16.5% (20.9)
Non-Hispanic Black (in%) 13.4% (21.8) 13.0% (22.3) 7.6% (15.7) 4.6% (11.7) 7.8% (14.5) 5.2% (11.4) 10.7% (20.3) 8.1% (16.4) 16.9% (24.4) 18.8% (26.2)
Asian (in %) 4.6% (8.8) 3.1% (5.7) 0.7% (2.0) 0.6% (1.3) 2.0% (4.4) 1.7% (3.1) 1.4% (3.4) 1.2% (2.0) 6.6% (10.4) 4.5% (6.8)
Native American (in %) 0.9% (4.6) 0.7% (3.4) 2.3% (9.9) 1.8% (7.1) 0.9% (4.7) 0.6% (2.9) 1.1% (3.0) 1.1% (2.7) 0.5% (1.6) 0.4% (1.1)
Some other race or more than one race (in %) 1.7% (2.7) 1.6% (2.4) 1.5% (2.3) 1.5% (2.2) 1.5% (2.2) 1.4% (1.9) 1.9% (2.9) 2.1% (2.9) 1.7% (2.9) 1.7% (2.6)
High school degree (in %) 28.4% (10.9) 30.4% (11.0) 36.7% (8.1) 37.9% (7.7) 30.7% (10.3) 32.3% (10.2) 33.3% (9.2) 34.0% (9.7) 25.4% (10.4) 27.3% (10.9)
Some college education (in %) 21.1% (6.2) 20.9% (6.1) 21.5% (5.1) 21.8% (4.8) 21.7% (5.7) 21.1% (5.5) 22.4% (6.0) 22.1% (5.6) 20.7% (6.6) 20.6% (6.6)
College degree (in %) 36.5% (18.6) 35.7% (18.2) 26.5% (10.4) 27.1% (9.5) 35.9% (16.3) 36.4% (15.6) 28.8% (12.8) 29.6% (13.4) 39.4% (20.1) 38.1% (20.3)
No degree (in %) 14.0% (11.1) 13.0% (10.2) 15.2% (8.2) 13.2% (7.4) 11.7% (8.8) 10.2% (7.3) 15.5% (9.7) 14.3% (9.2) 14.5% (12.3) 14.0% (11.6)
Poverty ratio (in %) 16.5% (12.5) 16.5% (13.2) 17.3% (9.0) 15.3% (8.0) 12.3% (9.3) 10.7% (8.5) 23.8% (13.0) 23.2% (13.5) 17.6% (13.8) 18.9% (15.0)
Number of tracts 72057 16561 10450 2649 15891 3937 1532 524 44184 9451
Open in a new tab

Next, we use binary logistic regression to predict the probability that a pipeline is proposed in a census tract given its characteristics as measured by our proposed explanatory variables. The model controls for population densityv and includes county fixed effects. County fixed-effects control for all county-level characteristics, using within-county variation to identify the association between sociodemographic characteristics and proposed pipeline location. Given this approach, we exclude counties with no proposed pipelines. This exclusion is consistent with our assumption that most routing decisions take place on the local scale (whereas the macro-scale routing is largely fixed due to the geography of oil and natural gas resources and processing). We cluster standard errors at the county level to account for the fact that tracts adjacent to tracts with pipeline proposals (that is tracts that are more likely to be in the same county) are more likely to also have a pipeline proposed, which would violate the assumption of independent observations.

Finally, we test for heterogeneity in our demographic and socioeconomic measures across the rural-urban continuum. Specifically, we estimate a series of binary logistic regression models that interact race, educational attainment, and poverty with population density, while controlling for all variables of our main specification.

5. The geography of proposed pipelines

Pipelines routes are determined by the geography of the current oil and natural gas boom. Gathering lines start at new wells and connect to existing pipelines, “value added” cracker or fractionation plants, or coastal industrial centers where these raw materials can be refined, used, or exported. Consequently, Figure 1 shows a concentration of proposed pipelines originating in the Bakken Formation of North Dakota and Montana, the Marcellus Shale of Pennsylvania, the Utica/Point Pleasant Shale of Ohio, the Woodford Shale of Oklahoma, and Barnett Shale of Texas. These are shale gas and tight oil plays that have only recently been exploited on a large scale due to the technological advancements in hydraulic fracturing and horizontal drilling. These pipelines are generally short and thus largely determined by the location of the oil and gas fields and their connecting lines or processing points, most of these pipelines have been excluded from our analytical sample. Furthermore, a number of proposed pipelines would enter the country in North Dakota and Montana to import oil from the Athabasca tar sands of Alberta, Canada. Most of the proposed pipelines are designated to end in the traditional industrial centers of the United States: the cities along the Great Lakes, the coastal areas of the North East, and the Gulf of Mexico. In between, the proposed pipelines largely follow the shortest route.

Apart from the urban industrial centers that have high percentages of Black working-class populations, the majority of pipelines are proposed in states with historically low numbers of ethnic minorities. Notably, the Southwest with its high percentage of Hispanic residents, and the Southern Black belt are not sites for new pipeline development, due to their lack of economically relevant natural gas and oil plays.

On a national scale, it would therefore be expected that areas with pipeline development have a higher share of White residents. However, since origin and destination as well as the general route in between are largely determined by natural and industrial geography, the more interesting question is the routing of pipelines around population centers along the way. Since non-White and low-income populations tend to be spatially concentrated (U.S. Environmental Protection Agency 1996; Lichter et al. 2007; Lichter et al. 2012), it is critical to assess whether these communities are targeted or avoided. Consequently, we look at demographic differences within counties and limits the main analysis to counties with proposed pipelines.

6. Results

Descriptive statistics of tracts by urban-metropolitan codes are presented in Table 1. The first two columns display the census tract statistics for our demographic variables of interest for the entire nation and our analytic sample of tracts in the 895 counties (724 counties in the restricted sample) in which new pipelines have been proposed. The next columns, then, break down these statistics by urban-metropolitan-codes. The 72057 tracts of the full population contain all tracts within the contiguous United States and, therefore, the estimated population characteristics are very similar to those aggregated on the national level and published by the US Census Bureau.vi Since census tracts are drawn to represent relatively homogeneous populations of roughly 4000 people, it does not surprise that this number does not vary much with location. But population density increases from 139 people per square mile in rural tracts in non-metropolitan counties to 8473 people per square mile in urban tracts in metropolitan counties. Consequently, the number of census tracts per given area increases with population density.

The restricted population of 16561 tracts is characterized by slightly higher population densities, which range from 154 people per square mile in rural tracts in non-metropolitan counties (2649 tracts, 16%) to 5949 people per square mile in urban tracts in metropolitan counties (9451 tracts, 57%). Unsurprisingly, we also observe substantively important socio-economic and demographic variance across urban-metropolitan codes. For example, low-density non-metropolitan tracts tend to have larger non-Hispanic White population shares and lower educational attainment than high-density tracts in metropolitan counties. These and similar descriptive patterns suggest that looking at differences along the urban-rural continuum is worthwhile.

We next estimate a binary logistic regression model to assess whether and how the probability that a pipeline is proposed in a census tract is associated with the tract’s location on the urban-metropolitan scale and socio-economic characteristics (Table 2). We first consider variation in proposed pipeline placement according to population density and metropolitan status. Net of population characteristics, pipeline construction tends to be proposed in tracts with low population densities, and this is true in both metropolitan and non-metropolitan counties. The odds that a pipeline is proposed in a rural, metropolitan tract decrease by a factor of 2.4 (OR=0.413) in low-density, metropolitan tracts, 8.2 (OR=0.122) for high-density, non-metropolitan ones, and 8.4 (OR=0.119) for high-density, metropolitan tracts compared to rural, non-metropolitan tracts. We next consider whether socio-economic characteristics predict proposed pipeline placement. Contrary to our expectations, tracts with large Hispanic (β = −1.738) and/or non-Hispanic Black (β = −0.860) populations are, on average, less likely to have pipelines proposed in them than other tracts. Our results also show that proposed pipeline placement is associated with tracts’ educational composition. Locations with higher shares of adults with at least a college degree are considerably less likely (β = −1.851) to have tracts proposed in them than other tracts. Finally, we find that a tract’s poverty rate is inversely associated with the odds that a pipeline has been proposed there. Each percentage point increase in the poverty rate is associated with a 5.4-times lower rate of proposed pipeline placement on average. This finding is also contrary to the expectation that pipelines will be targeted within socio-economically disadvantaged tracts within a county.

Table 2.

Binary logistic regression for pipeline proposal in tracts (analysis restricted to counties with pipeline proposals).

Model 1
β Std. Err.
Urban-Metro Codes
(reference: Low-density, non-metro)		 Low-density, metro −0.8848 (0.1048) ***
High-density, non-metro −2.1043 (0.24454) ***
High-density, metro −2.1295 (0.1685) ***
Race
(reference: Non-Hispanic White)		 Hispanic (in %) −1.7379 (0.4733) ***
Non-Hispanic Black (in %) −0.8602 (0.4319) **
Asian (in %) −1.6735 (1.5760)
Native American (in %) 0.82514 (0.7825)
Some other race or more than one race (in %) 1.47212 (1.5553)
Educational Attainment
(reference: High school degree)		 Some college education (in %) −0.6739 (0.8070)
College degree (in %) −1.8509 (0.4560) ***
No degree (in %) −1.2363 (0.8807)
Economics Poverty ratio −1.6914 (0.6822) **
Population density 0.0000 (0.0000)
County fixed effects Yes
n 16,540
Pseudo r-square 0.2239
Open in a new tab

Note: Standard errors clustered at the county level. 21 observations dropped from sample because no variation in outcome variable within their county.

*

p < .10

**

p < .05

***

p < .001

We expect that the associations between proposed pipeline locations and populations’ ethno-racial composition, educational attainment, and poverty may vary across urban-metropolitan-codes. We begin by examining the relationship between tracts’ racial and ethnic composition and the odds of being the site of a proposed pipeline (Table 3). Among low-density, non-metropolitan tracts, we find no statistically significant association between the Hispanic share of the population and proposed pipeline siting. However, both the non-Hispanic black (β = −1.237) and Asian (β = −24.015) population shares are negatively associated with proposed pipeline locations. The former association does not vary significantly between low-density, non-metropolitan tracts and other types of locations. However, the association between Asian population share is significantly different—and, on net, null—in all three other types of tracts (low-density, non-metro; low-density, metro; and high-density metro). We also find significant between-places differences in the associations between Hispanic and Native American population shares and pipeline location. Hispanic population share is not a significant predictor of pipeline location in low-density, non-metropolitan tracts but is significantly and inversely associated with pipeline location in low-density (β = −2.211) and high-density (β = −2.210) tracts in metropolitan areas. Likewise, the Native American share of tracts’ population is not a significant predictor of pipeline location in low-density, non-metropolitan tracts but is significantly and positively associated with pipeline placement in low-density metropolitan tracts (β = 2.110). The share of this population is also negatively associated with proposed pipelines in high-density metropolitan tracts (β = −20.001).

Table 3.

Interaction models estimating likelihood of pipeline proposal in tracts (analysis restricted to counties with pipeline proposals).

Model 2
β Std. Err.
Urban-Metro Codes
(reference: low-density, non-metro)		 Low-density, metro −0.8011 (0.1405) ***
High-density, non-metro −2.4582 (0.3220) ***
High-density, metro −2.1848 (0.2057) ***
Race
(reference: non-Hispanic White)		 Hispanic (in %) 0.8257 (0.6900)
Non-Hispanic Black (in %) −1.2372 (0.5694) *
Asian (in %) −24.0147 (7.5359) **
Native American (in %) 0.9821 (0.9513)
Some other race or more than one race (in %) 1.2065 (2.9820)
Race☓Urban-Metro Codes
(reference: Non-Hispanic White)		 (Reference: Hispanic☓low-density, non-metro)
Hispanic☓low-density, metro −3.0371 (0.9214) **
Hispanic☓high-density, non-metro −0.9642 (1.5406)
Hispanic☓high-density, metro −3.0355 (0.8160) **
(Reference: non-Hispanic Black☓low-density, non-metro)
Non-Hispanic Black☓low-density, metro −0.4174 (0.8347)
Non-Hispanic Black☓high-density, non-metro −0.4377 (1.6131)
Non-Hispanic Black☓high-density, metro 0.6529 (0.7192)
(Reference: Asian☓low-density, non-metro)
Asian☓low-density, metro 19.2107 (8.2901) **
Asian☓high-density, non-metro 31.0750 (10.4157) **
Asian☓high-density, metro 23.8419 (7.6569) **
(Reference: Native☓low-density, non-metro)
Native☓low-density, metro 1.1277 (1.3461)
Native☓high-density, non-metro 1.9190 (11.8935)
Native☓high-density, metro −20.9833 (7.8965) **
(Reference: other race☓low-density, non-metro)
Other race☓low-density, metro −1.2979 (3.9843)
Other race☓high-density, non-metro 5.7349 (7.6010)
Other race☓high-density, metro 2.3977 (3.7672)
n 16,540
Pseudo r-square 0.2295
Open in a new tab

Notes: Controlling for county, population density, Urban-Metro Codes, education, household income, poverty rate, unemployment rate, standard errors clustered at the county level. 21 observations dropped from sample because no variation in outcome variable within their county

*

p < .10

**

p < .05

***

p < .001

We next examine the relationship between educational attainment and a tract’s odds of being the site of a proposed pipeline across the four different tract categories (Table 4). In low-density, non-metropolitan tracts, the share of adults with a college degree or higher is inversely associated with the presence of a proposed pipeline. The magnitude of this association does not vary significantly between low-density, non-metropolitan tracts and high-density tracts in both metropolitan and non-metropolitan counties. However, this association is significantly less, in absolute terms, in high-density non-metropolitan tracts than in low-density tracts—and on net is a non-significant predictor of proposed pipeline location among the former. The share of adults with no high school education—a proxy for populations’ low educational attainment and socio-economic status—is not a significant predictor of pipeline location in low-density non-metropolitan tracts. However, we find marginally significant differences in this association between low-density non-metropolitan tracts and high-density tracts in metropolitan areas, and in the latter low educational attainment is significantly and negatively associated with pipeline location (net β = −2.626).vii

Table 4.

Interaction models estimating likelihood of pipeline proposal in tracts (analysis restricted to counties with pipeline proposals).

Model 3
β Std. Err.
Urban-Metro Codes
(reference: Low-density, non-metro)		 Low-density, metro −0.1909 (0.6894)
High-density, non-metro −4.1257 (1.5966) **
High-density, metro −0.4236 (0.7589)
Educational Attainment
(reference: High school degree)		 Some college education (in %) 2.3214 (1.1311) **
College degree (in %) −3.0099 (0.8140) ***
No degree (in %) −0.2106 (1.0995)
Educational Attainment☓Urban-Metro Codes
(reference: Highschool degree)		 (Reference: some college education☓low-density, non-metro)
Some college education☓low-density, metro −2.8142 (1.4742) *
Some college education☓high-density, non-metro 0.0913 (3.1409)
Some college education☓high-density, metro −6.8731 (2.1478) **
(Reference: college degree☓low-density, non-metro)
College degree☓low-density, metro 0.8572 (1.0559)
College degree☓high-density, non-metro 4.9701 (2.2317) **
College degree☓high-density, metro 0.4222 (1.0804)
(Reference: no degree☓low-density, non-metro)
No degree☓low-density, metro −2.7830 (1.7229)
No degree☓high-density, non-metro 3.9530 (3.8559)
No degree☓high-density, metro −2.4151 (1.4620) *
n 16,540
Pseudo r-square 0.2269
Open in a new tab

Notes: Controlling for county, population density, Urban-Metro Codes, race, household income, poverty rate, unemployment rate, standard errors clustered at the county level. 21 observations dropped from sample because no variation in outcome variable within their county

*

p < .10

**

p < .05

***

p < .001

Finally, we examine the relationship between a tract’s poverty rate and its odds of being the site of a proposed pipeline across the urban-metropolitan codes (Table 5). This association is negative in low-density, non-metropolitan tracts (β = −3.480), but the magnitude is significantly less in absolute terms in high-density tracts of metropolitan areas. The net association in the latter is null, as is also the case in high-density tracts of non-metropolitan counties. The magnitude of this association does not vary significantly according to whether low-density tracts are in a metropolitan or non-metropolitan county.

Table 5.

Interaction models estimating likelihood of pipeline proposal in tracts (analysis restricted to counties with pipeline proposals).

Model 4
β Std. Err.
Urban-Metro Codes
(reference: low-density, non-metro)		 Low-density, metro −1.04313 (0.1596) ***
High-density, non-metro −2.41295 (−0.5554) ***
High-density, metro −2.55065 (−0.1899) ***
Poverty Poverty ratio −3.48046 (−0.7499) ***
Poverty☓Urban-Metro Codes (Reference: poverty rate☓low-density, non-metro)
Poverty rate☓low-density, metro 0.781626 (0.9322)
Poverty rate☓high-density, non-metro 2.087211 (2.2362)
Poverty rate☓high-density, metro 3.008516 (1.0051) **
n 16,540
Pseudo r-square 0.2256
Open in a new tab

Notes: Controlling for county, population density, Urban-Metro Codes, race, education, household income, unemployment rate, standard errors clustered at the county level. 21 observations dropped from sample because no variation in outcome variable within their county

*

p < .10

**

p < .05

***

p < .001

7. Discussion

The hypothesis that pipelines are more likely to be proposed in areas with predominantly minority and other vulnerable populations does not hold for pipeline proposals whose routes are publicly available. Tracts with higher shares of Hispanic and Black residents and higher poverty rates are less likely to be the site of a proposed pipeline.viii So, do pipeline routings today contradict the thrust of the Environmental Justice literature? We do not know yet. The analyzed pipelines are proposed, not built. It would be instructive to compare these routes with the final, realized routes to see what groups have been more successful in fending off pipelines in their communities. Furthermore, we have assumed that pipelines are a Locally Unwanted Land Use (LULU), which is not a universally supported claim in the literature to date. If pipelines are in fact locally desired—or at least ignorable—for some groups, the findings of this study would demonstrate that socially advantaged groups have been more successful in attracting new pipelines.

The geography of the oil and natural gas boom in North America may be one explanation for the different siting of pipelines compared to other land uses with a legacy of environmental injustice. The sites for strip mining, mountain top removal, and deep mining depend on folded geologies with often dramatic slopes. Pipelines, on the other hand, are—if possible—routed to avoid those difficult terrains and thus predominantly traverse agricultural lands without a legacy of resource extraction and with a high degree of private landownership. While traditional mining communities have a proud history of unionized organizing and a developed awareness of the negative outcomes of extraction-oriented industries, collective action in agrarian landscapes is less common (cf. Eaton and Kinchy 2016). In those landscapes, less concerted resistance can be expected. Private landowners may even support pipelines if they benefit through royalties and right-of-way fees, as is the case for fracking sites (Dokshin 2016; Bugden and Stedman 2018). Non-Disclosure Agreements between landowners and pipeline companies that prevent the former from sharing their contracts and royalty figures with neighbors further dissuade collective action among landowners.

While we find that pipelines are not disproportionally proposed in census tracts with high minority populations, it does not suggest that various movements resisting pipelines have lost their footing. First, our findings suggest that the relationship between the existence of pipeline proposals in a tract and socioeconomic status is moderated by urbanicity of the tract. For instance, our research suggests that a higher share of Native American residents in sparsely populated areas—which is where most reservations are located—may in fact increase the odds of a pipeline being proposed in these tracts, whereas in more urban areas this relationship reverses. In addition, the potential negative impacts of pipelines on Native American communities may go far beyond the direct exposure of pipelines on places of residence of Native Americans and include their tribal lands, ceded territories, and sacred sites.

Second, racial groups or classes do not act as collectives. Because a certain demographic is not more affected than another on average, does not necessarily lead individuals or local communities to approve of pipeline development in their vicinity. Indeed, most formal planning takes place at scales above the neighborhoods that we consider in this tract-level analysis. This is particularly true where people perceive the local advantages of a new pipeline—such as temporary creation of construction jobs and income from leases and tax revenue—as less accessible or relevant to them. For instance, Gravelle and Lachapelle (2015) find that women, younger residents, and those with college education are less swayed by these prospects, partly because the expected new jobs would hardly benefit them. What is more, the attitudes towards the benefits of new pipelines may intersect with race, ethnicity, and socio-economic status, even though no study has probed into this question to this day.

Third, environmental justice goes beyond the distribution of proximate exposure to pollution and other hazards to include broader systemic issues. For instance, Whyte (2016) frames the Indigenous-led protests against the North Dakota Access Pipeline as part of a larger struggle for climate justice and decolonization. He argues that Indigenous people, through colonialist domination, are suffering disproportionally from climate change while contributing to it only marginally. The protesters at Standing Rock insisted that they were not just opposing the construction of the pipeline in their community, but pipelines in general. From this perspective, resistance against pipelines is not just about environmental justice, but about an alternative form of development independent from fossil energy.

This study contributes to the literature on pipeline development by applying a spatially specific environmental justice framework to it. It provides a model analysis that could be used in the future to assess the social equity of pipeline development, both proposed and existing. Federal and state lawmakers and agencies may find this approach useful as they develop framework legislation that distributes opportunities and risks of pipeline development in equal measures. Those who are involved in planning specific pipelines, including pipeline operators and local communities, may apply this toolset to find common ground for a conversation on the advantages and disadvantages of different proposed routes. The approach in this paper should not be used uncritically to optimize pipeline routes for the most neutral distribution of risk. The interpretation of risk is always dependent on personal values, interest, and subjective perception. Final decisions on routes or construction in general, thus, will always be contingent on group-specific interests, power, and democratic discourse and representation.

8. Limitations and potential for further research

Despite the novelty of this study, it is not without limitations. First, census tracts are vastly different in size across rural and urban areas. Consequently, they are spatially less precise in rural areas. If a pipeline is proposed in an urban tract, it will be routed close to the homes of all residents. This cannot be assumed in rural census tracts which may stretch over several miles. This problem can be addressed if the US Census Bureau were to release reliable census data in an even grid, so that all cells are the same size, unlike tracts.

Second, our analysis violates the assumption of independent observations because a pipeline’s proposed location in one tract is a function of its route through other places. We cluster standard errors at the county level to remedy this shortcoming, but this is an imperfect solution. Proximity does not end at county borders. A spatially explicit regression model that accounts for these relationships between adjacent tracts would be a welcomed improvement.

Third, this model assumes that the environmental risk of a leak is limited to the census tract of the pipeline. However, leaking pipelines also threaten the water quality downstream and air quality in all four cardinal directions, potentially in other census tracts. To improve the validity of the results, follow-up research should model risk zones based on the local hydrology and use the mean demographics of these in the model.

Fourth, regression analysis only explores the average association between variables, but not the pathways in which these associations operate. Though we observe that, for instance a higher share of well-educated residents raises the likelihood of a pipeline proposal in a tract, we can only speculate about the nature about these relationships. Further research is needed to explore these in depth.

The final limitation concerns the lack of reliable pipeline route data. We speculate that the limited and inconsistent availability of pipeline route data biases our results to underestimate the likelihood of vulnerable communities to be the site of new pipelines. This is because proposals that disproportionally expose such communities may be less likely to be shared publicly than less controversial ones. Pipeline operators and the government need to make the routes of pipeline proposals publicly available so impacted communities can consent (or dissent) and be involved in the planning. The published routes could be slightly modified using random algorithms to protect against sabotage or terror attacks, while still providing meaningful information about which communities would be exposed.

This paper operationalizes environmental (in)justice as the proportional distribution of the local risk of an oil or natural gas spill. Of course, pipelines support a much larger fossil economy, that produces winners and losers beyond the immediate vicinity to hydrocarbon infrastructure. Regardless of their location in relation to pipelines, marginalized populations are disproportionately burdened by climate change as a result of the transporting and burning of fossil energy (Oppenheimer et al. 2014). Further, some scholars (Malm 2016; Moore 2017) argue that the use of fossil energy has fueled social injustice by accelerating the reach and scope of capitalism. Since much of the resistance against pipelines is framed as opposition to the use of fossil energy in general, future inquiry into the social justice of pipelines need to consider non-local effects of pipelines on different social groups.

Acknowledgements:

We thank two anonymous students in the Rural Sociology graduate program at Pennsylvania State University who provided feedback on an early version of the manuscript.

Funding:

The compilation of pipeline proposals was supported by FracTracker Alliance, a non-profit that supports communities and policy makers to better understand the impacts of hydrocarbon development. FracTracker receives operational support from the Heinz Foundation, the Gund Foundation, the 11th Foundation, and the Hoover Foundation. The content is solely the responsibility of the authors and does not necessarily represent the views of FracTracker or any of the funding foundations. This research did not receive any specific grant from any funding agencies in the public, commercial, or not-for-profit sectors and no funders were involved in designing and conducting the research presented in the article. Thiede acknowledges assistance provided by the Population Research Institute at Penn State University, which is supported by an infrastructure grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development [P2CHD041025]. Thiede’s work was also supported by the USDA National Institute of Food and Agriculture and Multistate Research Project #PEN04623 [Accession #1013257].

10. Endnotes

i

Where data is available, it is commonly generalized so as to make it unusable for anything more granular than the national or regional scale.

ii

For the perspective of individuals from across the US demographic and political spectrum that live adjacent to, or in the path of, existing or proposed hydrocarbon infrastructures see The FracTracker Alliance’s (2019) “Energy Audio Stories” archive

iii

We use the Georeferencing and ‘Fit To Display’ tools in ArcMap 10.6 to create spatially explicit replicas of the pipeline proposals. If the geodetic reference system from the industry publication is unknown, we transform these image files so that known boundaries or landmarks match onto the base map in our reference system. We exclude proposals where the image files could not be projected onto our reference system or the spatial resolution of the original pipeline date was too coarse to match our tract level analysis.

iv

We exclude tracts with a population of zero from the full census.

v

We do not control for area which is highly correlated to population density.

vi

Differences are due to averaging unweighted tract means, whereas means weighted by tract population would produce estimates of the national population.

vii

The net association between population share without a high school degree and pipeline location is also statistically significant in low-density tracts in metropolitan counties (β = −2.994).

viii

We note that, surprisingly, the siting of a pipeline is also correlated with lower shares of residents with a college degree. We expected to see educational attainment, which is commonly used as a proxy for socioeconomic status, to correlate with the poverty rate.

Declaration of Interest: None

9. References

  1. Ashwood Loka and MacTavish Kate. 2016. “Tyranny of the majority and rural environmental injustice.” Journal of Rural Studies 47:271–277. [Google Scholar]
  2. Berry Paul G and Hirschl Thomas A. 2017. “Non-metro versus metro poverty in the transition to adulthood in the united states: (1980–2009).” Journal of Rural Studies 54:76–84. [Google Scholar]
  3. Bugden Dylan and Stedman Richard. 2018. “Rural Landowners, Energy Leasing, and Patterns of Risk and Inequality in the Shale Gas Industry.” Rural Sociology 84(3):459–488. [Google Scholar]
  4. Bullard Robert D. 1990. Dumping in Dixie: Race, class, and environmental quality. Boulder: Westview Press. [Google Scholar]
  5. Bullard Robert D. 1993. “Race and environmental justice in the United States.” Yale Journal of International Law 18:319–335. [Google Scholar]
  6. Bullard Robert D (1998) Anatomy of environmental racism and the environmental justice movement. In: Bullard RD (eds) Anatomy of Environmental Racism and the Environmental Justice Movement. Boston, MA: South End Press, pp.15–39. [Google Scholar]
  7. Clarke Christopher E., Bugden Dylan, Hart P. Sol, Stedman Richard C., Jacquet Jeffrey B., Evensen Darrick T.N. and Boudet Hilary. 2016. “How Geographic Distance and Political Ideology Interact to Influence Public Perception of Unconventional Oil/ Natural Gas Development.” Energy Policy 97:301–9. [Google Scholar]
  8. Clay Karen, Jha Akshaya, Muller Nicholas and Walsh Randall. 2019. “External costs of transporting petroleum products: Evidence from shipments of crude oil from North Dakota by pipelines and rail.” The Energy Journal 40(1):55–72. [Google Scholar]
  9. Dokshin Fedor A. 2016. “Whose backyard and what’s at issue? Spatial and ideological dynamics of local opposition to fracking in New York State, 2010 to 2013.” American Sociological Review 81(5):921–948. [Google Scholar]
  10. Eaton Emily and Kinchy Abby. 2016. “Quiet voices in the fracking debate: Ambivalence, nonmobilization, and individual action in two extractive communities (Saskatchewan and Pennsylvania).” Energy research & social science 20:22–30. [Google Scholar]
  11. Energy Policy Research Foundation. 2010. The Value of the Canadian Oil Sands (....to the United States). An Assessment of the Keystone Proposal to Expand Oil Sands Shipments to Gulf Coast Refiners. Washington, DC [Google Scholar]
  12. Evensen Darrick and Stedman Richard. 2018. “‘Fracking’: Promoter and destroyer of ‘the good life’.” Journal of Rural Studies 59:142–152. [Google Scholar]
  13. Fernando Felix N. and Cooley Dennis R.. 2016. “Socioeconomic System of the Oil Boom and Rural Community Development in Western North Dakota.” Rural Sociology 81(3):407–444. [Google Scholar]
  14. Foreman Christopher H. 1996. “A Winning Hand? The uncertain future of environmental justice.” The Brookings Review 14(2):22–25. [Google Scholar]
  15. Alliance FracTracker. 2019. “Audio Stories from People Living Near Oil & Gas Development.” Retrieved May 03, 2020 (https://www.fractracker.org/resources/oil-and-gas-101/audio-stories/#toggle-id-10)
  16. Glickman Theodore S. 1999. “Measuring environmental equity with geographical information systems.” Pp. 197–203 in The RFF Reader in Environmental and Resource Management, edited by Oates WE. Washington, DC: Resources for the Future [Google Scholar]
  17. Gosine Andil and Teelucksingh Cheryl. 2008. Environmental justice and racism in Canada: An introduction. Toronto: Emond Montgomery Publications Limited. [Google Scholar]
  18. Gravelle Timothy B and Lachapelle Erick. 2015. “Politics, proximity and the pipeline: Mapping public attitudes toward Keystone XL.” Energy Policy 83:99–108. [Google Scholar]
  19. Hamilton James T. 1995. “Testing for environmental racism: prejudice, profits, political power?” Journal of Policy Analysis and Management 14(1):107–132. [Google Scholar]
  20. Hoberg George. 2013. “The battle over oil sands access to tidewater: a political risk analysis of pipeline alternatives.” Canadian Public Policy 39(3):371–392. [Google Scholar]
  21. Huber Albert and Bowe Peter. 2014. “The Keystone Pipeline Would Create Thousands of Jobs.” Jersey City, NJ: Forbes. Retrieved May 03, 2020 (https://www.forbes.com/sites/realspin/2014/02/07/the-keystone-pipeline-would-create-thousands-of-jobs/) [Google Scholar]
  22. Jacquet Jeffrey B. 2014. “Review of risks to communities from shale energy development.” Environmental Science & Technology 48(15):8321–8333. [DOI] [PubMed] [Google Scholar]
  23. Jerolmack Colin and Walker Edward T. 2018. “Please in My Backyard: Quiet Mobilization in Support of Fracking in an Appalachian Community.” American Journal of Sociology 124(2):479–516. [Google Scholar]
  24. Jerrett Michael, Burnett Richard, Kanaroglou Pavlos, Eyles John, Finkelstein Norm, Giovis Chris and Brook Jeffrey. 2001. “A GIS - environmental justice analysis of particulate air pollution in Hamilton, Canada.” Environment and Planning A 33:955–973 [Google Scholar]
  25. Kelly-Reif Kaitlin and Wing Steve. 2016. “Urban-rural exploitation: An underappreciated dimension of environmental injustice.” Journal of Rural Studies 47:350–358. [Google Scholar]
  26. Kinchy Abby, Perry Simona, Rhubart Danielle, Stedman Richard, Brasier Kathryn and Jacquet Jeffrey. 2014. “New natural gas development and rural communities: Key issues and research priorities.” Pp. 260–78 in: Rural America in a Globalizing World: Problems and Prospects for the 2010s, edited by Bailey C, Jensen L and Ransom E. Morgantown: West Virginia University Press. [Google Scholar]
  27. Kriesel Warren, Centner Terrence J and Keeler Andrew G. 1996. “Neighborhood exposure to toxic releases: are there racial inequities?” Growth and Change 27(4):479–499. [Google Scholar]
  28. LaDuke Winona. 1999. All our relations: Native struggles for land and life. Cambridge, MA: South End Press. [Google Scholar]
  29. Le Billon Philippe and Kristoffersen Berit. 2019. “Just cuts for fossil fuels? supply-side carbon constraints and energy transition.” Environment and Planning A: Economy and Space: 308518. [Google Scholar]
  30. Lichter Daniel T., Parisi Domenico, Grice Steven M. and Taquino Michael C.. 2007. “National estimates of racial segregation in rural and small-town America.” Demography 44(3):563–581. [DOI] [PubMed] [Google Scholar]
  31. Lichter Daniel T., Parisi Domenico and Taquino Michael C.. 2012. “The geography of exclusion: Race, segregation, and concentrated poverty.” Social Problems 59(3):364–388. [Google Scholar]
  32. Malm Andreas. 2016. Fossil capital. London, UK: Verso. [Google Scholar]
  33. [dataset] Manson Steven, Schroeder Jonathan, Van Riper David and Ruggles Steven. IPUMS National Historical Geographic Information System: Version 12.0 [Database]. Minneapolis, MN: University of Minnesota [distributor], 2017 [Google Scholar]
  34. Mayer Adam, Olson-Hazboun Shawn K., Malin Stephanie. 2018. “Fracking Fortunes: Economic Well-being and Oil and Gas Development along the Urban-Rural Continuum.” Rural Sociology 83(3):532–567. [Google Scholar]
  35. Mayer Adam and Malin Stephanie. 2019. “How should unconventional oil and gas be regulated? The role of natural resource dependence and economic insecurity.” Journal of Rural Studies 65:79–89. [Google Scholar]
  36. Mohai Paul and Bryant Bunyan I.. 1992. “Environmental racism: reviewing the evidence.” Pp. 163.246 in Race and the incidence of environmental hazards. A time for discourse, edited by Bryant BI and Mohai P. Boulder, CO: Westview Press. [Google Scholar]
  37. Mohai Paul, Pellow David and Roberts J. Timmons. 2009. “Environmental Justice.” Annual Review of Environment and Resources 34(1):405–430. [Google Scholar]
  38. Moore Jason W. 2017. “The capitalocene, part I: On the nature and origins of our ecological crisis.” The Journal of Peasant Studies 44(3):594–630. [Google Scholar]
  39. Moore Michal C, Flaim Sam, Hackett David, Grissom Susan, Crisan Daria and Honarvar Afshin. 2011. “Catching the Brass Ring: Oil Market Diversification Potential for Canada.” SPP Research Papers 4(16):1–71. [Google Scholar]
  40. Oppenheimer Michael, Campos Maximiliano, Warren Rachel, Birkmann Joern, Luber George, O’Neill Brian and Takahashi Kiyoshi. 2014. “Emergent risks and key vulnerabilities.” Pp. 1039–1099 in Climate change 2014: impacts, adaptation, and vulnerability Working Group II Contribution to the IPCC 5th Assessment Report, edited by Intergovernmental Panel on Climate Change. Cambridge, UK and New York, NY: Cambridge University Press. [Google Scholar]
  41. O’Rourke Dara and Connolly Sarah. 2003. “Just oil? The distribution of environmental and social impacts of oil production and consumption.” Annual Review of Environment and Resources 28(1):587–617. [Google Scholar]
  42. Parfomak Paul W, Pirog Robert, Luther Linda and Vann Adam. 2013. Keystone XL pipeline project: Key issues. Washington, DC: Congressional Research Service. [Google Scholar]
  43. Pipeline and Hazardous Materials Safety Administration (PHMSA). 2017. “Pipeline Data and Statistics.” Retrieved May 03, 2020 (https://www.phmsa.dot.gov/data-and-statistics/pipeline/data-and-statistics-overview)
  44. Roberts J. Timmons and Parks Bradley. 2006. A climate of injustice: Global inequality, north-south politics, and climate policy. Cambridge, UK: MIT Press. [Google Scholar]
  45. Schafft Kai A., Borlu Yetkin and Glenna Leland. 2013. “Gas Development and Perceptions of Risk and Opportunity.” Rural Sociology 78:143–166. [Google Scholar]
  46. Schafft Kai A., McHenry-Sorber Erin, Hall Daniella and Burfoot-Rochford Ian. 2018. “Busted amidst the Boom: The Creation of New Insecurities and Inequalities within Pennsylvania’s Shale Gas Boomtowns.” Rural Sociology 83:503–531. [Google Scholar]
  47. Scott Dayna. 2013. “The networked infrastructure of fossil capitalism: implications of the new pipeline debates for environmental justice in Canada.” Revue générale de droit 43:11–66. [Google Scholar]
  48. Silver Jonathan. 2019. “Decaying infrastructures in the post-industrial city: An urban political ecology of the US pipeline crisis.” Environment and Planning E: Nature and Space: 251484861989051 [Google Scholar]
  49. Skinner Lara. Sweeney Sean and Goodman Ian. 2012. “Pipe Dreams? Jobs Gained, Jobs Lost by the Construction of Keystone XL.” Ithaca, NY: Cornell University Global Labor Institute. Retrieved May 03, 2020 (https://www.ilr.cornell.edu/sites/ilr.cornell.edu/files/GLI_KeystoneXL_012312_FIN.pdf) [Google Scholar]
  50. Smith Eric R. A. N. 2002. Energy, the Environment, and Public Opinion. New York: Rowman & Littlefield. [Google Scholar]
  51. Tarrant Michael A. and Cordell H. Ken. 1999. “Environmental justice and the spatial distribution of outdoor recreation sites: An application of geographic information systems.” Journal of Leisure Research 31(1):18–34. [Google Scholar]
  52. Thiede Brian C., Lichter Daniel T. and Slack Tim. 2018. “Working, but poor: The good life in rural America?” Journal of Rural Studies 59:183–193. [Google Scholar]
  53. U.S. Bureau of Oceans and International Environmental and Scientific Affairs. 2014. “Final Supplemental Environmental Impact Statement for the Keystone XL Project” Retrieved May 03, 2020 (https://2012-keystonepipeline-xl.state.gov/finalseis/index.htm)
  54. U.S. Chamber of Commerce Global Energy Institute. 2017. “Benefits of Keystone XL.” Retrieved May 03, 2020 (https://www.globalenergyinstitute.org/benefits-keystone-xl)
  55. U.S. Environmental Protection Agency. 1996. “Guidance for incorporating environmental justice concerns in EPA’s NEPA compliance analyses.” Washington, DC. [Google Scholar]
  56. U.S. Government Accountability Office. 2019. “Critical Infrastructure Protection. Actions Needed to Address Weaknesses in TSA’s Pipeline Security Program Management.” Testimony Before the Subcommittee on Energy, Committee on Energy and Commerce, House of Representatives. Washington, DC. [Google Scholar]
  57. Wang Qiang, Chen Xi, Jha Awadesh N. and Rogers Howard. 2014. “Natural gas from shale formation–the evolution, evidences and challenges of shale gas revolution in United States.” Renewable and Sustainable Energy Reviews 30:1–28. [Google Scholar]
  58. Whyte Kyle P. 2016. “Why the Native American pipeline resistance in North Dakota is about climate justice.” Boston, MA: The Conversation. Retrieved May 03, 2020 (https://theconversation.com/why-the-native-american-pipeline-resistance-in-north-dakota-is-about-climate-justice-64714) [Google Scholar]

RESOURCES