Jump to ContentJump to Main Navigation
The Political Economy of Environmental Justice$

H. Spencer Banzhaf

Print publication date: 2012

Print ISBN-13: 9780804780612

Published to Stanford Scholarship Online: June 2013

DOI: 10.11126/stanford/9780804780612.001.0001

Show Summary Details
Page of

PRINTED FROM STANFORD SCHOLARSHIP ONLINE (www.stanford.universitypressscholarship.com). (c) Copyright Stanford University Press, 2022. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in SSO for personal use. Subscriber: null; date: 29 June 2022

The Role of Demographic and Cost-Related Factors in Determining Where Plants Locate

The Role of Demographic and Cost-Related Factors in Determining Where Plants Locate

A Tale of Two Texas Cities

Chapter:
(p.199) 8 The Role of Demographic and Cost-Related Factors in Determining Where Plants Locate
Source:
The Political Economy of Environmental Justice
Author(s):

Ann Wolverton

Publisher:
Stanford University Press
DOI:10.11126/stanford/9780804780612.003.0008

Abstract and Keywords

This chapter presents a significant methodological advance on earlier environmental justice literature, looking at census tracts or other geographic entities as units of analysis for explaining the locational decision-making process. It compares results from two of the largest cities in Texas–Dallas-Fort Worth and Houston-Galveston. In particular, it directly models the behavior of firms, looking at their choice of where site a polluting facility. Given a firm's decision to establish a facility somewhere, it models the firm's choice of location among all the possible locations. The analysis choice patterns as a function of demographics, prices, transportation options, and so forth. It offers evidence that these choices are driven more by profits than by pure discrimination.

Keywords:   environmental justice, housing discrimination, housing profits, Dallas, Houston

Introduction

It is fairly common in the environmental justice literature to focus on the relationship between contemporaneous socioeconomic characteristics and site or plant location for the purposes of investigating disproportionate impacts. The studies that have examined whether socioeconomic factors contribute to location decisions at the time of siting often exclude variables recognized in the firm location literature as important determinants of location choice, for instance, the costs of land, labor, and transportation. In this chapter, I use a similar approach to Wolverton (2009): I examine plant location decisions at the time of siting but incorporate variables recognized as important in both the firm location and the environmental justice literatures into a single analysis. While most environmental justice studies that model location choice use a binary response model, I allow for multiple location alternatives to more closely approximate a firm's evaluation of potential substitute sites to the location chosen.

Unlike Wolverton (2009), this chapter examines the potential influence of geographic scope on the analytic results. Studies in the environmental justice literature report mixed results with regard to the relevance of race, ethnicity, poverty, and income to location decisions. Bryant and Mohai (1992) point out that one possible reason for such a mix of findings may be that the scope of the analysis differs so widely by study—some focus on a particular urban area or region, while others are national in scope.1 Results from city-specific analyses cannot be easily generalized to other geographic contexts. However, more aggregate studies—those on the state or national level—may mask the importance of socioeconomic factors in firm decision making. This chapter examines (p.200) factors related to a polluting plant's decision of where to locate within two large Texas cities, Dallas–Fort Worth and Houston, at the time of siting between 1978 and 1985, using a conditional logit framework. It then compares these results to those for the state of Texas while using consistent methodology and sets of variables.

The Environmental Justice Literature

Early studies that match site location to contemporaneous socioeconomic characteristics often rely on simple statistical techniques and tend to find strong evidence of a relationship between race and poverty variables and site location.2 Later studies that examine similar relationships often use more sophisticated techniques and therefore tend to be more careful in the interpretation of results. Scope varies widely across these studies—some focus on a particular urban area or region, while others are national in scope.3 Zimmerman (1993) finds that a greater percentage of

Zimmerman (1993) finds that a greater percentage of minorities lives near inactive hazardous waste sites that appear on the National Priority List but that the population living in poverty does not differ significantly from the national average. This trend is found to hold at the regional level as well. Baden, Noonan, and Turaga (2007) find that race and ethnicity are correlated with the presence of a Superfund site at the national level, but they conclude that this relationship is sensitive to changes in both geographic scope and scale (i.e., how the neighborhood is defined). On the other hand, Anderton et al. (1994) find only limited evidence of disproportionate numbers of hazardous waste facilities located in minority or poor neighborhoods. This result is also found to hold at a more disaggregated level when the country is divided into 10 regions.4

There are a handful of studies that examine the relationship between neighborhood characteristics and facility location decisions at the time of siting. These studies also find a mixed record with regard to the importance of socioeconomic variables to plant location decisions. Unlike studies that match site location to contemporaneous socioeconomic characteristics, however, they rarely examine how results change with the scope of the analysis. Been (1997) obtains mixed evidence that race played a role at the time of siting for active commercial hazardous waste treatment, storage, and disposal facilities (TSDFs) in the United States. While waste disposal sites are correlated with certain 1990 socioeconomic characteristics such as race and income, neither the percentage of the poor nor the percentage of African Americans in a neighborhood is significant at the time of siting. The percentage of Hispanics does remain (p.201) significant at the time of siting. Pastor, Sadd, and Hipp (2001) examine the location of TSDF sites in Los Angeles County, California, and find greater evidence of disproportionate siting in established Latino and African American communities than minority move-in after the TSDF establishment. Baden and Coursey (2002) examine the location of Superfund sites in Chicago, Illinois, and find that sites were disproportionately located in poor neighborhoods in the 1960s but not in the 1990s. However, they find little evidence for disproportionate exposure of African Americans either currently or at the time of siting. Jenkins, Maguire, and Morgan (2004) study compensation to communities in exchange for hosting municipal solid waste landfills. Controlling for tipping fees paid from the landfill to the community, they find that socioeconomic characteristics such as income and race do not matter at the city level but do appear to matter at the county level.5 Finally, Wolverton (2009) examines the siting decisions of Toxics Release Inventory (TRI) plants in the 1980s and 1990s in the state of Texas and finds that input-related cost factors are consistently more important than determinants related to the socioeconomic characteristics of the surrounding neighborhood. Race and ethnicity are not related to plant location decisions, while poverty appears to act as a deterrent.

The Firm Location Literature

In the economics literature, a firm is assumed to evaluate potential locations for a new plant based on the principle of profit maximization. In doing so, the firm takes into account many location-specific attributes related to production and transportation costs that may affect potential profits in each potential location. Production costs include costs related to relatively immobile inputs such as land, labor, and housing, and costs related to operation such as taxes, public utility fees, and environmental regulations. Transportation costs include freight rates, distance to input markets, and distance to output markets. Most studies of new plant location do not have measures for all production and transportation costs because of limitations in the data but do usually include measures of labor costs, land costs, transportation costs, energy costs, and level of taxation.6 It is also important to consider any offsetting location benefits from agglomeration economies such as a shared infrastructure or labor pool.

Some environmental justice studies include variables to serve as proxies for land and labor costs but rarely include other variables associated with firm location (e.g., Davidson and Anderton 2000). Kriesel, Centner, and Keeler (1996), while focusing on the incidence of emissions rather than on plant (p.202) location, is a notable exception within the environmental justice literature. Along with land and labor costs, they include proximity to an interstate highway and find that the inclusion of these factors renders race and poverty insignificant. This finding points to the importance of including such variables in any study of site location decisions. Wolverton (2009) includes measures of labor costs, land costs, distance to major highways and rail, and possible agglomeration economies. She also finds that traditional firm location variables dominate in importance and render race variables insignificant.

As discussed in Chapter 1 of this volume, Hamilton (1995) offers three additional hypotheses as to why a plant may locate in a poor or minority neighborhood. The first hypothesis stems from Coase (1960): a plant is established where residents' valuation of environmental quality, and therefore the potential compensation by the firm to the neighborhood residents, is lowest. Since local willingness to pay for environmental quality is positively correlated with income, firms will tend to locate plants in poorer neighborhoods to minimize the costs of compensation.7 The second hypothesis as to why plants may locate in poor or minority neighborhoods is that firms locate polluting plants where the likelihood of a community engaging in collective activities is relatively low. In this case, a firm owes less to the community in the form of compensation not because the neighborhood values the externality any less than other communities but because the transaction costs of collective action are high. Hamilton's final hypothesis is that firm owners or managers trade off profits in favor of discriminating against a particular demographic group by locating a heavily polluting plant in that community. Since it is easier and therefore less costly to discriminate in neighborhoods with a substantial minority population, plants tend to be located in these neighborhoods.

Empirical Model and Approach

I adopt the empirical model of firm location decisions first developed by Levinson (1996) and then adapted by Wolverton (2009) for the purposes of incorporating Hamilton's additional hypotheses related to firm location. Levinson (1996) assumes that each firm has an unobserved profit function for each possible location that depends on location-specific variables such as factor prices, fixed inputs (land, labor), and the stringency of environmental regulation. Wolverton (2009) includes the cost of discrimination in the form of foregone profits and the cost of required compensation, which is a function of the value placed on environmental amenities in the neighborhood and the propensity of (p.203) the neighborhood to engage in collective action. Based on the assumption that firms are motivated by a desire to maximize their profits, a firm then chooses to locate a plant in the neighborhood that yields the highest potential profit. An increase in the cost of a location—because of an increase in input prices, the cost of discrimination, or the level of compensation required—implies a decrease in profits. An increase in the availability of inputs implies an increase in profits.

Most environmental justice studies that model location choice use a binary response model.8 Allowing for multiple location alternatives seems more appropriate, since firms typically choose from a spectrum of competitive locations when deciding where to site a plant. Following Wolverton (2009), I use a conditional logit model to represent the choice of a particular location from a set of many neighborhoods. Assume that firm i faces J possible plant location alternatives and that these J choices are independently and identically distributed. The firm will choose location j when its profits πij are maximized in that particular location compared to all other possible choices. It is possible to write firm i's profits as follows:

(8.1)
πij=βzij+eij,

where zij is defined as the set of observed characteristics specific to location j and plant i. Assume that the error term eij has a Weibull distribution. If the firm's underlying production function is assumed to be Cobb-Douglas, then profits will be log-linear.

Conditional on the decision to open a new plant, the probability that firm i will choose a particular location k can be written as

(8.2)
Pr(ik)=eβzikj=1Jeβzij.

Because of the limited number of observations in the Dallas–Fort Worth and Houston areas, a firm is modeled as selecting a location for its plant from the actual location and nine randomly selected alternatives drawn from the full choice set for a given metropolitan area. This technique has been shown to yield consistent estimates and has the added advantage that the likelihood function is identical to that used for estimating a conditional logit with the full choice set (McFadden 1978).

(p.204) Data

In this chapter, I focus on location decisions in two urban areas of Texas: Dallas–Fort Worth and Houston.9 These are two of the largest cities in Texas, and both rank within the top 10 largest cities in the United States by population. These two metropolitan areas differ in a number of interesting ways that may influence plant location decisions. For instance, Houston has a much more concentrated industry profile than Dallas–Fort Worth, with the majority of its industry engaged in chemical manufacturing. Houston also does not limit land use through zoning restrictions, while Dallas–Fort Worth does. This implies that industry in the Houston area faces fewer constraints on its location decisions, all else equal, in comparison to those that choose to locate in Dallas–Fort Worth. This could increase the likelihood that facilities in Houston are located closer to existing residential areas.

I examine the location decisions of manufacturing plants sited between 1976 and 1985 in the Houston and Dallas–Fort Worth areas that reported to the TRI. Each TRI plant in Texas is matched to the census tract in which it is located. Any plant that appears in the TRI at least once is eligible for inclusion.10 The location of the plants is matched to the appropriate census tract and socioeconomic characteristics from the 1980 US Census of Population and Housing. A total of 106 census tracts have one or more plants locating in the Dallas–Fort Worth area, and 56 census tracts have one or more plants locating in the Houston-Galveston area during this time period.11 Data are also drawn from the US Census of Manufactures, the County and City Data Books, and several directories of manufacturers for the state of Texas.12

Variable Definitions

I use variables associated with each of the four relevant considerations for plant location decisions that were outlined previously: profit maximization,13 willingness to pay for environmental amenities, propensity to engage in collective action, and opportunities to discriminate.

I capture differences in the cost of land, labor, and transportation, all of which enter into a firm's calculation of potential profits, through the use of the average property value of owner-occupied housing in a neighborhood, PROPERTYj;14 the average wage of a production worker in manufacturing at the county level, WAGEj; and the average distance of a given neighborhood from the nearest railroad, RAILj.15 To control for potential differences in the (p.205) costs of environmental regulation, I also include the percentage of years for which a county was out of attainment for ozone and total suspended particulates over the years studied, NATTNj.16 This is a potentially relevant factor since it is arguably more difficult to locate a polluting plant in a county already out of attainment with existing regulations. Following Arora and Cason (1998), I also include the percentage of the population employed in manufacturing, MANUFij, to capture the potential trade-off between jobs and environment. To account for the role that zoning or agglomeration economies may play, the number of preexisting TRI facilities in the same census tract, OLDSITEj, is included. Finally, a variable measuring the percentage of the population living in an urbanized area, URBANj, is also included in the analysis. More urbanized areas may offer more immediate access to large labor pools, better infrastructure, and easy access to public services. However, they also tend to have higher taxes, more traffic, and more crime.17

The potential compensation a firm pays to a neighborhood depends on the neighborhood's willingness to pay for environmental quality and its propensity for collective action. Residents' willingness to pay for environmental amenities is most closely associated with income levels, INCOMEj. The percentage of households living below the poverty line, POVERTYj, is also included as a variable. If a firm compensates each member of the neighborhood, then the more densely populated a neighborhood, POPDENSj, the more costly it is to the firm and the less likely it will be to locate a plant in that neighborhood.18 Following Arora and Cason (1998), I also include variables that affect a population's “stake” in the neighborhood as well as their desire to free ride: the average number of children per household, CHILDj; the percentage of the population over the age of 65, AGE65j; and the percentage of households that are renters, RENTERj.19

Two variables are included to represent the possibility that firms seek out neighborhoods where it is easier to discriminate on the basis of race or ethnicity: the percentage of the population who are nonwhite, NONWHTj, and the percentage who are foreign-born, FOREIGNj.20

Multicollinearity

A few of the independent variables described in the previous section are highly correlated. For instance, use of property values in the same regression as income is potentially problematic since they have a correlation coefficient above 80 percent. Likewise, the percentage of the population living in poverty (p.206) is highly correlated with income and the percentage of the population who are nonwhite. Because the traditional environmental justice literature includes these variables indiscriminately, I include one specification that ignores these multicollinearity problems. However, I also explore an alternate specification: I use PBUILT70j, the percentage of housing in a neighborhood that was built prior to 1970, as a proxy for land value. This measure is expected to be a rather imperfect substitute since it is related to the housing stock and therefore more closely associated with property values than with land value, but it allows me to explore the robustness of the results. I also use NOPHONj, the percentage of households without a phone in their home, as a proxy for the poverty rate. This measure is fairly highly correlated with poverty (67 percent) but is far less correlated with the income and race variables.

Matching of All Plants Regardless of When They Are Established

The focus of much of the environmental justice literature is on the correlation between plant location and socioeconomic characteristics without accounting for the timing of the siting decision. To ensure our sample is consistent with previous studies, I match all plants from the TRI, for which establishment data are available, that have been sited in the Houston-Galveston and Dallas–Fort Worth areas earlier than 1986 to socioeconomic characteristics from the 1990 US Census. A total of 150 census tracts have one or more plants from the data set being established in the Dallas–Fort Worth area and 134 census tracts have one or more plants being established in the Houston-Galveston area prior to 1986.

I find that for both the Dallas–Fort Worth and the Houston-Galveston areas, the summary statistics show that neighborhoods with a plant generally have, on average, lower incomes (17 percent lower in Houston and 25 percent in Dallas), higher percentages of nonwhite (5 percentage points more in Dallas) and foreign-born (1–3 percentage points across cities), higher percentage living in poverty (2 percentage points), and a greater percentage of renters (2–4 percentage points).21 Such a finding is consistent with the correlations observed in the literature when all plants are matched to contemporaneous neighborhood characteristics regardless of the time of siting. Figures 8.1 to 8.4 further illustrate the broad correlation between the location of older TRI plants and two socioeconomic characteristics, percentage nonwhite and per capita income in 1990. This correlation is less evident for the subset of plants sited between 1976 and 1985. (p.207)

The Role of Demographic and Cost-Related Factors in Determining Where Plants LocateA Tale of Two Texas Cities

Figure 8.1 Map of number of TRI plants and percentage nonwhite in Dallas–Fort Worth, before 1976 and 1976–1986, respectively

Summary Statistics for 1980 Established Plants Only

For the time period studied, plants are concentrated in only a few two-digit Standard Industrial Classifications (SICs) in the two metropolitan statistical areas (MSAs). Dallas–Fort Worth appears somewhat more diverse than Houston-Galveston in this regard. In Houston-Galveston, almost half of the plants are in the chemicals and allied products (SIC 28) industry. Another 26 percent are in fabricated metal products (SIC 34), and 9 percent are in the rubber and (p.208)

The Role of Demographic and Cost-Related Factors in Determining Where Plants LocateA Tale of Two Texas Cities

Figure 8.2 Map of number of TRI plants and percentage nonwhite in Houston-Galveston, before 1976 and 1976–1986, respectively

miscellaneous plastics (SIC 30) industry. In Dallas–Fort Worth, most plants are spread across five main industries: 24 percent of manufacturing plants are in the chemicals and allied products (SIC 28) industry, 15 percent are in fabricated metal products (SIC 34), another 15 percent are in electronic and other (p.209)
The Role of Demographic and Cost-Related Factors in Determining Where Plants LocateA Tale of Two Texas Cities

Figure 8.3 Map of number of TRI plants and per capita income in Dallas–Fort Worth, before 1976 and 1976–1986, respectively

electrical equipment (SIC 36), 13 percent are in rubber and miscellaneous plastics (SIC 30), and 8 percent are in the industrial/commercial machinery and computer equipment (SIC 35) industry.

Several characteristics differ between Houston area tracts with a TRI plant established between 1976 and 1985 and Houston area tracts without a TRI plant established during this same time period (see Table 8.1). While the summary statistics generally adhere to expectations with regard to input-related costs, (p.210)

The Role of Demographic and Cost-Related Factors in Determining Where Plants LocateA Tale of Two Texas Cities

Figure 8.4 Number of TRI plants and per capita income in Houston-Galveston, before 1976 and 1976–1986, respectively

(p.211)

Table 8.1 Houston-Galveston and Dallas–Fort Worth summary statistics—plants established between 1976 and 1985 matched to 1980 characteristics

Houston-Galveston MSA

Dallas-Fort Worth MSA

Variable

Tracts with plant (n = 56)

Tracts without plant (n = 584)

Tracts with plant (n = 106)

Tracts without plant (n = 529)

Nonwhite

0.21

(0.28)

0.29

(0.31)

0.21

(0.25)

0.23

(0.30)

Foreign-born

0.06

(0.07)

0.07

(0.07)

0.04

(0.04)

0.05

(0.06)

Poverty

0.09

(0.09)

0.11

(0.09)

0.11

(0.10)

0.11

(0.11)

Average income

10,064.69

(3,944.61)

11,334.16

(4,961.67)

9,671.05

(2,949.16)

10.982.96

(6,008.98)

Population density

828.95

(1,068.45)

2,822.44

(2,594.61)

922.16

(1,065.02)

3,278.46

(2,964.74)

Renter

0.26

(0.19)

0.39

(0.24)

0.38

(0.26)

0.38

(0.25)

Percentage with children

0.47

(0.16)

0.41

(0.15)

0.39

(0.14)

0.38

(0.15)

Percentage over the age of 65

0.05

(0.04)

0.07

(0.05)

0.09

(0.07)

0.10

(0.07)

Manufacturing

0.21

(0.08)

0.18

(0.07)

0.27

(0.08)

0.22

(0.07)

Average wage

20,811.34

(1,800.33)

21,110.82

(2,044.82)

15,560.74

(1,773.37)

15,835.02

(1,694.78)

Average property value

36,131.34

(18,588.62)

44,978.37

(27,661.09)

31,163.21

(21,799.31)

45,760.67

(33,426.65)

Nonattainment status

0.84

(0.37)

0.88

(0.32)

0.71

(0.46)

0.80

(0.40)

Distance to rail

1.09

(1.35)

1.32

(1.45)

1.06

(1.43)

1.11

(0.99)

Built prior to 1970

0.49

(0.30)

0.64

(0.31)

0.58

(0.28)

0.71

(0.28)

Number of old TRI sites

4.52

(5.91)

0.23

(0.65)

2.87

(3.96)

0.25

(0.78)

Urban

0.71

(0.43)

0.84

(0.35)

0.66

(0.46)

0.84

(0.36)

Note: Each row shows the mean and standard deviation of the respective variable and location.

this is not always the case for socioeconomic characteristics. Tracts in which a plant locates tend to have a higher percentage employed in manufacturing, lower property values, and a greater number of preexisting TRI facilities. They also tend to be less urban, closer to a rail line, and have lower population densities. With regard to socioeconomic characteristics that are often the focus of (p.212) the environmental justice literature, tracts in which a plant locates tend to have lower incomes but fewer nonwhite households and less poverty. There is little difference in the percentage of foreign-born residents, on average. Contrary to expectations, they also tend to have fewer renters, fewer older homes, and more children. To examine whether socioeconomic characteristics in these communities show closer adherence to the environmental justice story in the subsequent decade, I also examine 1990 socioeconomic characteristics for tracts with and without a plant established in the 1980s. I find a story consistent with the summary statistics presented in Table 8.1: while average income is higher in census tracts without a plant, the percentages nonwhite, foreign, in poverty, and renters are all lower in neighborhoods where a plant was established a decade previously. Thus, a large part of the environmental justice story when contemporaneous socioeconomic characteristics are matched to plant location appears to be driven by the existence of older plants.

While many variables for Dallas–Fort Worth look similar to those for Houston, there are a number of differences across the two MSAs worth noting. In Dallas–Fort Worth, tracts in which plants locate between 1976 and 1985 appear to have similar percentages of minority populations to those without plants. This is not the case in Houston, where tracts in which plants locate have, on average, a noticeably lower percentage of minority populations. Likewise, Houston appears to have fewer households with children living in the neighborhoods in which plants locate, while there is little difference in the percentage of children living in neighborhoods with or without a new plant in Dallas–Fort Worth. Houston also appears to have a much higher percentage of renters residing in tracts without plants than in tracts in which they locate. In Dallas–Fort Worth, there is little difference in the proportion of residents who are renters. As is the case with the Houston area, 1990 socioeconomic characteristics for tracts with and without a plant established in the 1980s are consistent with the summary statistics presented in Table 8.1 for the main socioeconomic characteristics of interest.

Finally, note that the number of preexisting TRI sites is far greater in tracts in which plants locate than in tracts where they do not across the two time periods and the two cities. However, while tracts without plants appear to have a similar average incidence of preexisting sites across the two cities, Houston appears to have a noticeably greater average number of preexisting sites in tracts with plants than Dallas–Fort Worth.

(p.213) Results

Columns 2–5 of Table 8.2 report the results of conditional logit regressions for Dallas–Fort Worth and Houston-Galveston; the last two columns contain results for the entire state of Texas using the same set of variables.22 Two specifications are presented for each geographic region: the first uses alternate measures of poverty and property values—percentage of households without a phone and percentage of housing built prior to 1970—to better account for multicollinearity between these variables and the race and income variables; the second specification ignores the multicollinearity problem and presents the variables typically used in the environmental justice literature—race, income, poverty, and property values. It is worth noting that the fit of the three regressions varies—the best fit is for the Houston-Galveston regression (54 to 55 percent), followed by Texas as a whole (39 percent) and then Dallas–Forth Worth (28 to 29 percent).

Contrary to results cited in the environmental justice literature, there is little evidence to support the hypothesis that firms discriminate on the basis of race or ethnicity, controlling for other location-relevant factors. The results are remarkably consistent in this regard in spite of variation in the scope of the analysis across the three sets of regressions.23

With regard to income and poverty, variables related to a neighborhood's willingness to pay for environmental amenities, geographic scope appears to have some influence on the results. Neither income nor poverty (or its proxy) is significant for either of the two sets of MSA-level regressions. Income is also not significant at the state level. However, at the state level, depending on the specification, poverty is significant and negatively related to location choice. Note that the sign on the poverty variable is opposite of what has been posited in the environmental justice literature.

Of the variables associated with collective action by the community, only population density is consistently significant across the three geographic areas I examine. The more densely populated an area, the less likely it is that a polluting plant locates there. The percentages of renters and children are not significant for any of the three geographic areas. The percentage over the age of 65 is significant for Texas as a whole but for only one of the two specifications presented. It is not significant for either Dallas–Fort Worth or Houston-Galveston.

Variables traditionally considered in the firm location literature but often omitted from environmental justice studies—those associated with (p.214)

Table 8.2 Conditional logit regression results

Dallas-Fort Worth (for 58 plants)

Houston-Galveston (for 138 plants)

All of Texas (for 362 plants)

Variable

Model I

Model II

Model I

Model II

Model I

Model II

Nonwhite

0.11

(0.76)

0.48

(0.80)

0.28

(0.93)

−0.08

(1.14)

0.32

(0.37)

0.52

(0.39)

Foreign-born

−4.52

(3.46)

−3.21

(3.28)

0.57

(3.02)

0.15

(3.00)

−0.23

(1.09)

0.43

(1.00)

Income

−0.51

(0.62)

−0.04

(0.53)

−0.08

(0.24)

−0.73

(0.81)

−0.02

(0.05)

0.02

(0.05)

No phone

−1.35

(2.12)

1.91

(3.18)

−0.73

(1.00)

Poverty

−3.37

(2.67)

1.63

(4.72)

−2.58**

(1.07)

Population density

−0.18*

(0.11)

−0.23**

(0.97)

−0.47**

(0.22)

−0.55**

(0.24)

−0.28***

(0.05)

−0.28***

(0.05)

Renter

0.79

(0.87)

0.88

(0.91)

−0.16

(1.74)

0.11

(1.73)

0.42

(0.32)

0.18

(0.35)

Children

−0.26

(1.51)

0.33

(1.46)

3.30

(2.88)

3.25

(2.93)

0.87

(0.63)

0.83

(0.59)

Over 65 years

−0.68

(3.17)

−4.42

(2.71)

2.64

(10.05)

0.68

(8.57)

−2.44

(2.06)

−4.66**

(1.84)

Manufacturing

6.93***

(1.93)

6.45***

(1.88)

5.39**

(2.72)

5.90**

(2.79)

3.00***

(0.71)

2.82***

(0.68)

Wage

−0.78

(1.57)

−0.68

(1.52)

1.97

(2.15)

1.53

(2.09)

−0.63*

(0.38)

−0.96**

(0.39)

Distance to rail

−0.10

(0.10)

−0.13

(0.10)

−0.41**

(0.17)

−0.41**

(0.16)

−0.25***

(0.05)

−0.23***

(0.05)

Nonattainment

1.35**

(0.56)

1.17**

(0.59)

−0.61

(0.80)

−0.53

(0.80)

0.06

(0.25)

0.002

(0.02)

Built before 1970

−1.94***

(0.57)

−0.59

(1.19)

−1.13***

(0.33)

Property value

−0.05

(0.04)

0.62

(0.68)

−0.03

(0.03)

Urban

−1.81***

(0.54)

−1.86***

(0.58)

−0.04

(0.91)

0.03

(0.93)

−0.49**

(0.25)

−0.61**

(0.26)

Old TRI sites

0.19***

(0.06)

0.18***

(0.05)

0.49***

(0.16)

0.50***

(0.16)

0.22***

(0.04)

0.22***

(0.03)

Old × SIC 28

0.25*

(0.14)

0.20*

(0.12)

0.08

(0.06)

0.09

(0.06)

Old × SIC 30

0.22*

(0.13)

0.17

(0.11)

0.01

(0.08)

0.004

(0.08)

Old × SIC 34

0.49*

(0.30)

−0.51*

(0.29)

0.28**

(0.14)

0.28**

(0.14)

Log-likelihood

−171.52

−173.94

−63.40

−63.12

−832.77

−834.49

Pseudo R2

0.30

0.29

0.51

0.51

0.40

0.40

Note: Standard errors are shown in parentheses.

(*) Significant at 10% level;

(**) significant at 5% level;

(***) significant at 1% level.

(p.215) production and transportation costs—are significant. This finding is consistent with Wolverton (2009); what appears to matter most to a plant's location decision, regardless of geographic scope, are the variables emphasized in the firm location literature. Without these variables the pseudo R2 falls to 7 percent for Dallas–Fort Worth, 18 percent for Houston-Galveston, and 31 percent for Texas as a whole.

What is perhaps most interesting is that there are differences in the specific profit maximization variables that are significant across the three sets of regressions. County-level wage rate is significant and of the expected sign (negative) for Texas as a whole but is insignificant at the MSA level. Distance to a major railroad is significant and negatively related to plant location for Houston-Galveston and Texas, but it does not appear to matter to location decisions in the Dallas–Fort Worth area. County-level attainment status for traditional air pollutants is important to plant location decisions in the Dallas–Fort Worth area but not for Houston-Galveston or Texas as a whole.24 Property value is not significantly correlated with plant location for any of the geographic areas. However, the alternative measure of land value, average age of housing, is significant and negatively correlated with plant location in Dallas–Fort Worth and Texas as a whole. However, its sign indicates that it may be capturing factors other than property values, to the extent homes in older neighborhoods are less expensive. The percentage of the population residing in urban areas also follows this pattern: it is significant in Dallas–Fort Worth and Texas as a whole but not in Houston-Galveston.

Two variables are consistently significant across all geographic areas and specifications: the percentage employed in manufacturing and the presence of one or more preexisting TRI sites. The greater the percentage employed in manufacturing, the more likely it is that a plant will locate in that neighborhood. Likewise, once an older site is located in a particular neighborhood, it is more likely for an additional plant to locate there. This may be due to agglomeration economies or factors not controlled for in these regressions such as taxes and zoning.

Given the importance of preexisting TRI sites in the regressions, I also examine whether the significance of the main race and income variables changes when this variable is dropped. For the MSA-level regressions, percentage nonwhite and percentage foreign remain insignificant. However, income is now significant in three of the four MSA-level regressions. The percentage of the population living in poverty becomes significant at the 10 percent level for Dallas–Fort Worth but is still negatively related to plant location: the higher (p.216) the poverty rate, the less likely a plant is located in that neighborhood. It remains insignificant for Houston-Galveston. Finally, the percentage of households without a phone is now significant for both MSA-level regressions and continues to be positively related to plant location. When preexisting TRI sites are dropped from the regressions for all plants sited in Texas over this time period, the results remain unchanged for all race and income variables with one notable exception. Percentage foreign is now significant in both specifications at the 1 percent level. This result runs counter to what was found in Wolverton (2009) but points to the importance of preexisting sites to the main finding and highlights an area for continued future research.

Conclusion

In the environmental justice literature, evidence of disproportionate siting in poor or minority neighborhoods is decidedly mixed. Some allege this is because of the difference in whether the study looks at evidence at the national, state, or city level. Here, I compare results from two of the largest cities in Texas—Dallas–Fort Worth and Houston-Galveston—to results for the state overall to discern whether important demographic or other differences are evident at the city level that may be masked at a more aggregate level of analysis.

I examine whether results associated with four possible hypotheses for why plants may locate in poor or minority neighborhoods remain consistent across geographic scope. I find remarkably consistent results for most hypotheses. Variables associated with profit maximization appear to contribute most to the overall fit of the regressions, both at the city and the state level. Variables associated with possible discrimination or collective action on the part of the community appear to be largely insignificant across specifications and geographic scope, with the exception of population density. It is worth noting, however, that I include relatively indirect measures of collective action compared to other studies. Variables associated with willingness to pay for environmental amenities appear to be the exception: poverty is sometimes significant at the state level but is never significant at the level of the city. Thus, I find that in most cases, geographic scope does not play a major role in determining the importance of either demographic or cost variables in plant location decisions.

Notes

For their helpful comments and suggestions, I thank Spencer Banzhaf, two reviewers, and the participants of the 2008 Markets for Land and Pollution: Implications (p.217) for Environmental Justice workshop. I also thank Emma Roach for her superior GIS skills. The views expressed in this chapter are those of the author and do not necessarily represent those of the US Environmental Protection Agency. This chapter has not been subjected to EPA's review process and therefore does not represent official policy or views.

References

Bibliography references:

Anderton, Douglas, Andy Anderson, John Michael Oakes, and Michael Fraser. 1994. “Environmental Equity: The Demographics of Dumping.” Demography 31 (2): 229–48.

Arora, Seema, and Timothy N. Cason. 1998. “Do Community Characteristics Influence Environmental Outcomes? Evidence from the Toxics Release Inventory.” Journal of Applied Economics 1 (2): 413–53.

Baden, Brett M., and Don L. Coursey. 2002. “The Locality of Waste Sites within the City of Chicago: A Demographic, Social, and Economic Analysis.” Resource and Energy Economics 24 (1–2): 53–93.

(p.220) Baden, Brett M., Douglas S. Noonan, and Rama Mohana R. Turaga. 2007. “Scales of Justice: Is There a Geographic Bias in Environmental Equity Analysis?” Journal of Environmental Planning and Management 50 (2): 163–85.

Bartik, Timothy. 1985. “Business Location Decisions in the United States: Estimates of the Effects of Unionization, Taxes, and Other Characteristics of States.” Journal of Business and Economic Statistics 3 (1): 14–22.

Beckman, Martin, and Jacques Thisse. 1986. “The Location of Production Activities.” In Handbook of Regional and Urban Economics, vol. 1, edited by P. Nijkamp. Amsterdam: North-Holland.

Been, Vicki, with Francis Gupta. 1997. “Coming to the Nuisance or Going to the Barrios? A Longitudinal Analysis of Environmental Justice Claims.” Ecology Law Quarterly 24 (1): 1–56.

Boer, J. Tom, Manuel Pastor Jr., James L. Sadd, and Lori D. Snyder. 1997. “Is There Environmental Racism? The Demographics of Hazardous Waste in Los Angeles County.” Social Science Quarterly 78 (4): 793–810.

Bowen, William, Mark Salling, Kingsley Haynes, and Ellen Cyran. 1995. “Toward Environmental Justice: Spatial Equity in Ohio and Cleveland.” Annals of the Association of American Geographers 85 (4): 641–63.

Bryant, Bunyan, and Paul Mohai, eds. 1992. Race and the Incidence of Environmental Hazards. Boulder, CO: Westview Press.

Bullard, Robert D. 1983. “Solid Waste Sites and the Black Houston Community.” Sociological Inquiry 53 (2–3): 273–88.

Carlton, Dennis. 1983. “The Location and Employment Choices of New Firms: An Econometric Model with Discrete and Continuous Endogenous Variables.” Review of Economics and Statistics 65 (3): 440–49.

Coase, Ronald H. 1960. “The Problem of Social Cost.” Journal of Law and Economics 3 (1): 1–44.

Davidson, Pamela, and Douglas Anderton. 2000. “Demographics of Dumping II: A National Environmental Equity Survey and the Distribution of Hazardous Materials Handlers.” Demography 37 (4): 461–66.

Finney, Miles. 1994. “Property Tax Effects on Intrametropolitan Firm Location: Further Evidence.” Applied Economic Letters 1 (2): 29–31.

Hamilton, James T. 1995. “Testing for Environmental Racism: Prejudice, Profits, Political Power?” Journal of Policy Analysis and Management 14 (1): 107–32.

Harrington, James, and Barney Warf. 1995. Industrial Location: Principles, Practice, and Policy. New York: Routledge.

Hersh, Robert. 1995. “Race and Industrial Hazards: An Historical Geography of the Pittsburgh Region, 1900–1990.” Discussion Paper 95-18, Resources for the Future, Washington, DC.

(p.221) Jenkins, Robin R., Kelly B. Maguire, and Cynthia L. Morgan. 2004. “Host Community Compensation and Municipal Solid Waste Landfills.” Land Economics 80 (4): 513–28.

Krieg, Eric. 1995. “A Socio-Historical Interpretation of Toxic Waste Sites: The Case of Greater Boston.” American Journal of Economics and Sociology 54 (1): 1–14.

Kriesel, Warren, Terence J. Centner, and Andrew G. Keeler. 1996. “Neighborhood Exposure to Toxic Releases: Are There Racial Inequities?” Growth and Change 27 (4): 479–99.

Lambert, Thomas, and Christopher Boerner. 1995. Environmental Inequity: Economic Causes, Economic Solutions. St. Louis: Center for the Study of American Business.

Lee, Haeduck, and Michael Wasylenko. 1987. “A Comment on the Appropriate Estimation of Intrametropolitan Firm Location Models.” Land Economics 63 (3): 306–9.

Levinson, Arik. 1996. “Environmental Regulations and Manufacturers' Location Choices: Evidence from the Census of Manufactures.” Journal of Public Economics 62 (1–2): 5–29.

McConnell, Virginia, and Robert Schwab. 1990. “The Impact of Environmental Regulation on Industry Location Decisions: The Motor Vehicle Industry.” Land Economics 66 (1): 67–81.

McFadden, Daniel. 1978. “Modeling the Choice of Residential Location.” In Spatial Interaction Theory and Residential Location, edited by A. Karlquist. Amsterdam: North-Holland.

Pastor, Manuel, Jim Sadd, and John Hipp. 2001. “Which Came First? Toxic Facilities, Minority Move-In, and Environmental Justice.” Journal of Urban Affairs 23 (1): 1–21.

Ringquist, Evan. 2005. “Assessing Evidence of Environmental Inequities: A Meta-Analysis.” Journal of Policy Analysis and Management 24 (2): 223–47.

United Church of Christ. 1987. Toxic Wastes and Race in the United States: A National Report on the Racial and Socio-Economic Characteristics of Communities with Hazardous Waste Sites. New York: Public Data Access.

US General Accounting Office (GAO). 1983. Siting of Hazardous Waste Landfills and Their Correlation with Racial and Economic Status of Surrounding Communities. Washington, DC: GAO.

Wolverton, Ann. 2009. “Effects of Socio-Economic and Input-Related Factors on Polluting Plants' Location Decisions.” Berkeley Electronic Journal of Economic Analysis and Policy, Advances 9 (1). http://www.bepress.com/bejeap/vol9/iss1/art14.

Zimmerman, Rae. 1993. “Social Equity and Environmental Risk.” Risk Analysis 13 (6): 649–66. (p.222)

Notes:

(1.) Other reasons for differences in results include variation in neighborhood definition, empirical technique, control variables, and type of facility examined.

(2.) For instance, see Bullard (1983), US Government Accounting Office (GAO; 1983), and United Church of Christ (1987).

(3.) Ringquist (2005) presents the results of a meta-analysis with regard to environmental equity studies. He finds that studies that are national in scope tend to result in smaller estimates of race-based inequities than studies at a more disaggregated level. However, Ringquist lumps together studies at the time of siting with those that examine characteristics after the fact and does not include this as a relevant factor for explaining differences across studies included in his meta-analysis.

(4.) Bowen et al. (1995) do not examine site location, instead focusing on how releases of toxic chemicals vary with study scope. They find that releases and minority populations are highly spatially correlated at the state level but that this relationship disappears when the study scope is limited to the metropolitan area. The authors posit that a state-level analysis is less appropriate in this instance since both industry and minority populations are concentrated in the metropolitan area in their sample.

(5.) Lambert and Boerner (1995) examine site location at the time of establishment in the context of changing socioeconomic dynamics. They do not find large initial differences in the percentage of poor and minority residents between neighborhoods with and without waste sites. However, housing values grew less rapidly in neighborhoods with waste sites, and minority populations moved into these neighborhoods at a faster rate. Hersh (1995) conducts a historical analysis of the change in racial and industrial dynamics for firms reporting to the Toxics Release Inventory (TRI). He finds that, in general, industries and blue-collar neighborhoods located near each other for job-related reasons. Also, he notes that both white and rich residents took flight to cleaner parts of the city after firms located in a particular neighborhood and that there was an eventual movement of minorities into more polluted areas. Krieg (1995) finds that race is associated with the number of waste sites in areas with a long history of industrial activity and that class is more closely associated with the number of waste sites in areas with more recent industrial activity. Noonan (Chapter 7) examines how environmental quality is capitalized into property values when both residents and environmental quality are changing over time.

(6.) See Carlton (1983), Bartik (1985), Beckman and Thisse (1986), Lee and Wasylenko (1987), McConnell and Schwab (1990), Finney (1994), Harrington and Warf (1995), and Levinson (1996).

(p.218) (7.) Compensation can be thought of as both monetary and in-kind (e.g., free access to certain services, the building of a community park) forms of remuneration given by the firm to the community to offset the perceived risks of an increase in pollution because of the location of a new plant in the area.

(8.) See, for example, Pastor, Sadd, and Hipp (2001); Davidson and Anderton (2000); Been (1997); Boer et al. (1997); and Anderton et al. (1994).

(9.) The urban areas are based on the definitions of the metropolitan statistical areas used in 1980 by the US Census Bureau.

(10.) Plants that use more than 10,000 pounds or manufacture more than 25,000 pounds of the 329 listed toxic chemicals are required to report how much of each chemical is released into air, land, or water.

(11.) I do not include plant decisions that occur later in time—between 1986 and 1993—because the data set becomes too small to include a reasonable number of control variables and alternate locations before running out of degrees of freedom. I have data on only 32 plant locations in the Dallas–Fort Worth area and 26 plant locations in the Houston-Galveston area for this time period.

(12.) The establishment date for each plant is collected from the Bureau of Business Research Directory of Texas Manufacturers: Volume I (1990–1993), the Harris Texas Manufacturers Directory (1995), the Texas High Technology Directory (1995), and the Texas Manufacturers Register (1994).

(13.) A number of other variables are potentially important to location decisions, for instance, differences in energy costs and property taxes. Unfortunately, no information is available on the cost of electricity by location during the 1980s. Property tax rates by county in Texas are only available beginning in 1991. Because they fluctuate across time, it seems inappropriate to use 1991 tax rates as a proxy unless the relative difference in rates stays roughly constant across counties over time. That said, when the 1991 tax rate is included, it is insignificant.

(14.) Both property values and household income are adjusted to 1980 dollars. The consumer price index for the southern region of the United States is used to make this adjustment. Property values act as a proxy for land values (which are unavailable at the census tract level in 1980) faced by firms when making location decisions.

(15.) I also explore a variable measuring the average distance to a major highway. It was not significant in any of the regressions and did not change the sign or significance of other variables.

(16.) Shadbegian and Gray (Chapter 9) speak to potential differences in regulatory costs in the environmental justice context: they examine whether regulators focus more regulatory attention on plants in rich, white neighborhoods than on plants in poor, minority neighborhoods.

(17.) Average plant size is a significant explanatory variable in Wolverton (2009). In that article, I used an MSA-level definition. It is not included here because of the inability to access 1982 census data.

(p.219) (18.) Since census tracts vary in size, we include population density instead of population.

(19.) The percentage who voted in the presidential election was used in Wolverton (2009) to represent the propensity to engage in collective action. It was significant. However, this variable does not have enough variation at the county level to allow for inclusion here.

(20.) Because Hispanics are included in both percentage nonwhite and percentage white in the US Census, using percentage Hispanic directly in the regression is problematic. In Texas, the percentage foreign-born is strongly correlated with the percentage Hispanic.

(21.) The table of summary statistics when plants are matched to 1990 socioeconomic characteristics is presented in Table A.8.1 in the online appendix at http://www.sup.org/environmentaljustice.

(22.) The Texas-level regressions are similar to those in Wolverton (2009), including the actual location and 49 alternatives. Here, I improve on previously published results by using population density instead of just population, taking into account whether a county is in attainment, and including interaction dummies between preexisting TRI sites and SIC codes. As in Wolverton (2009), I also include geographic dummies for the Houston-Galveston and Dallas–Fort Worth areas, though neither is significant.

(23.) As an alternative, a count regression model is used to examine what variables are associated with the number of facilities located in a particular neighborhood. The results for the race, ethnicity, poverty, and income variables appear to be robust to the regression technique.

(24.) I also explored the significance of nonattainment status interacted with industry-related variables such as percentage manufacturing, county wage, or SIC dummy variables. None of these interaction terms were significant for plant location decisions in the Houston-Galveston area. In the Dallas–Fort Worth area, only one interaction term was significant: between nonattainment status and percentage manufacturing. However, when the interaction term is significant, nonattainment status alone is no longer significant.