Carroll_TMESthesis2017.pdf

Media

Part of Equitable Greenspace Access and Neighborhood Gentrification in San Francisco, Ca

extracted text

EQUITABLE GREENSPACE ACCESS AND NEIGHBORHOOD
GENTRIFICATION IN SAN FRANCISCO, CA

by
Terence Carroll

A Thesis
Submitted in partial fulfillment
of the requirements for the degree
Master of Environmental Studies
The Evergreen State College
June 2017

©2017 by Terence Carroll. All rights reserved.

This Thesis for the Master of Environmental Studies Degree
by
Terence Carroll

has been approved for
The Evergreen State College
by

________________________
John Withey, Ph. D., Member of the Faculty

________________________
Date

ABSTRACT
Equitable Greenspace Access and Neighborhood Gentrification in San Francisco, CA
Terence Carroll
Due to health and wellness benefits associated with city dwellers’ proximity to
parks and urban greenspace (UGS), equitable access to these spaces can be considered an
environmental justice concern. However, in attempting to address the greenspace
inequities that often exist within cities, planners may inadvertently contribute to
processes of gentrification and the displacement of current residents by encouraging real
estate speculation. Using San Francisco as a model, this thesis explores UGS distributions
and assesses their relationships with socioeconomic demographics and gentrification. It
also considers a set of investment strategies that may help to avoid gentrification and
displacement. The identification of several forms of economic, educational, and agegroup disparity in residents’ access to UGS in San Francisco were among the key
findings. Racial-ethnic minority groups in San Francisco experienced greater than
average greenspace access. Gentrified neighborhoods were also positively correlated with
greater access, supporting the idea that greenspace investments might contribute to real
estate speculation and so-called “green gentrification”. The results of this thesis’s
analyses did not support the strategy of favoring smaller, more distributed UGS
installations as a method of avoiding gentrification and displacement in San Francisco.
This thesis expands upon the existing greenspace equity literature by incorporating into
its definition of UGS several innovative forms of green infrastructure that have not been
well-studied in the past.

Table of Contents

Introduction ----------------------------------------------------------- 1
Literature Review ---------------------------------------------------- 6
Introduction ----------------------------------------------------------------- 6
Benefits of Urban Greenspace and Access to Nature ------------------8
Green Infrastructure’s Ecological Services ---------------------------- 11
Challenges in Achieving Greenspace Equity -------------------------- 13
Distributive Environmental Justice ------------------------------------- 16
Green Gentrification ------------------------------------------------------ 17
Just Green Enough -------------------------------------------------------- 21
Techniques for Measuring UGS Access ------------------------------- 24
Findings of Past Spatial Analyses -------------------------------------- 27
Conclusion ----------------------------------------------------------------- 32

Methods ------------------------------------------------------------- 33
Overview ------------------------------------------------------------------ 33
Study Area ---------------------------------------------------------------- 34
Measuring Access: Subcategories ------------------------------------- 35
Measuring Access: Final Models -------------------------------------- 36
Measuring Socioeconomic Status ------------------------------------- 37
Accessibility at the Block Group Scale ------------------------------- 38
Creating a Gentrification Index ---------------------------------------- 40
Data Sources and Data Preparation ------------------------------------ 42
Statistical Analysis ------------------------------------------------------- 44

iv

Results --------------------------------------------------------------- 45
Overview ------------------------------------------------------------------ 45
New Greenspace and Green Infrastructure --------------------------- 48
New Parks ---------------------------------------------------------------- 51
Total Greenspace and Green Infrastructure -------------------------- 53
Total Parks --------------------------------------------------------------- 56

Discussion ----------------------------------------------------------- 57
Spatial Patterns Influencing Results ---------------------------------- 57
Equity Analyses --------------------------------------------------------- 58
Gentrification and the Impact of Park Size --------------------------- 62
Strengths, Limitations, and Future Studies --------------------------- 65

Conclusion ---------------------------------------------------------- 70

Reference: Abbreviations --------------------------------------73
Bibliography ---------------------------------------------------- 74
Appendix: Maps ------------------------------------------------ 82

v

List of Figures

Figure 1: Conflicts to Achieving Triple Bottom Line Sustainability (p. 14)
Figure 2: Locations of New Urban Greenspace in San Francisco (p. 82)
Figure 3: Locations of Total Urban Greenspace in San Francisco (p. 82)
Figure 4: Distribution of Black/African American Population (p. 83)
Figure 5: Distribution of Asian Population (p. 83)
Figure 6: Distribution of Latino Population (p. 84)
Figure 7: Distribution of Youth Population (p. 84)
Figure 8: Distribution of Postgraduate Population (p. 85)
Figure 9: Median Household Income by Census Block Group (p. 85)
Figure 10: Median Home Value by Census Block Group (p. 86)
Figure 11: Gentrification Index Score by Census Block Group (p. 86)
Figure 12: Districts of San Francisco, CA (p. 87)

v

List of Tables

Table 1: New Greenspace Regression Analysis (p. 46)
Table 2: Total Greenspace Regression Analysis (p. 47)
Table 3: New Greenspace & Number of Residents (p. 48)
Table 4: New Greenspace & Dollar/Index Values (p. 50)
Table 5: Total Greenspace & Number of Residents (p. 54)
Table 6: Total Greenspace & Dollar/Index Values (p. 55)

vi

Acknowledgements

I would like to thank the entire MES faculty for their help throughout the thesis
writing process; particularly my thesis reader, John Withey, prospectus reader, Kathleen
Saul, and GIS instructor, Mike Ruth. Thanks also to Rhys Roth for the opportunity to
work with the Center for Sustainable Infrastructure throughout my MES career. This
experience got me interested in learning more about triple bottom line sustainability and
many of the other issues explored in this thesis.
Thanks to my thesis review group, Joey Burgess, Alex Case-Cohen, Ben
Harbaugh, and Stephanie Heiges, who were all extremely helpful in providing feedback
over the past year. Thanks to Wendy Loosle and Trace McKellips for conversations that
helped me land on a research topic. Cheers to everyone in the 2015 cohort!
This thesis would not have been possible without the support of my wife, Kelly
Carroll, and family; Chuck, Cindy, Rob, Emma, Ted, Tanya, Scott, Ashley, Larry &
Nancy Sampson, Dorothy Carroll, Anna Matters, and all the friends and family I have not
listed here.

vii

Introduction
Through art, literature, and traditional belief systems, cultures around the world
have described the beneficial and restorative effects of spending time in natural settings.
The idea that spending time outdoors, away from more human-impacted environments,
could be linked with mental and physical health benefits makes sense intuitively, but in
the past, had not been well studied. Today, researchers have amassed a substantial body
of evidence supporting the idea that time spent in “green” environments, like forests,
meadows, or parks, is not only associated with many specific health benefits, but also
with an improved overall quality of life (Boone et al., 2009; Chiesura, 2004; Kabisch &
Haase, 2014).
To measure these health impacts and other beneficial properties emerging from
contact with nature, researchers often compare individual metrics for health and wellness
with the amount of urban greenspace (UGS) city dwellers regularly encounter in their
daily lives. Urban greenspace can be defined in many ways. Most definitions at least
include parks, undeveloped open spaces, and nature reserves, but more expansive
definitions may also incorporate community gardens, green corridors, and certain forms
of green infrastructure (Comber, Brunsdon, & Green, 2008; Maas et al., 2006). Peoples’
access to greater amounts of UGS near to where they live, has been associated with their
improved mental, physical, and perhaps even spiritual wellbeing (Taylor, Kuo, &
Sullivan, 2001; Zhou & Kim, 2013).
Upon discovering the link between health and UGS access, researchers
hypothesized that these health benefits were simply byproducts of residents’ higher
1

activity levels when given the chance to spend more time outdoors. But subsequent
studies closely examining this link have shown that increased exercise is just one of
several key factors leading to improved overall health in affected individuals (Aytur et
al., 2008). Other contributing factors may include the strengthened community ties
people often experience when spending more time in communal spaces, or their overall
improved sense of emotional and psychological wellness stemming from an ability to
reflect or relax more effectively (Chiesura, 2004; Jennings et al., 2012).
Given these associations, how UGS is distributed spatially has emerged as an
environmental justice concern among community activists and city planners (Rutt &
Gulsrud, 2016; Heynen et al., 2006). Initially, the environmental justice movement arose
in response to situations where lower socioeconomic status (SES) groups such as
communities of color were disproportionately impacted by pollution, toxic waste or other
environmental burdens. Sometimes these groups are more impacted by these hazards due
to structural or institutional inequities, while at other times evidence may exist that they
have been targeted more directly by decision makers and those in positions of power.
Either way, the end results are similar; groups most impacted by pollution or toxic waste
often experience higher rates of mortality and various diseases (Pope & Dockery, 2006).
In recent years, the environmental justice literature’s focus on spatial disparities has
expanded into cataloguing and addressing peoples’ disproportionate access to healthboosting “environmental amenities” in addition to environmental burdens or toxins (Wen
et al., 2013).
In this vein, many studies over the past fifteen years have begun to examine the
spatial distribution of parks and greenspace in relation to neighborhood SES. The results
2

of these studies have varied widely depending on their geographic area of focus and their
methodologies. Some have identified economic disparities, others racial-ethnic disparities
(Zhou & Kim, 2013; Comber et al., 2008; Wolch et al., 2002), and still others have found
certain cities’ parks and UGS to be distributed quite equitably (Timperio et al., 2007). For
example, Boone et al.’s (2009) findings show that communities of color and lower
income people have access to significantly less acreage of UGS in Baltimore, MD
compared with white and higher-income residents. On the other hand, Wen et al. (2013)
conducted a nationwide study in the US and found that many cities’ communities of color
experience disproportionately higher access to parks and greenspace, though they
observed significant differences in this pattern depending on population density (i.e.
urban vs. rural).
While spatial UGS equity analyses of particular cities and regions have been
commonplace recently in the environmental justice and political ecology literature, few
of these studies have examined the role that shifting neighborhood demographics might
play in contributing to persistent or worsening inequities. For example, some scholars are
concerned that over-investment in new parks or greenspace might spur real estate
speculation in surrounding neighborhoods (Wolch et al., 2014). They fear that UGS
investments might be linked with neighborhood gentrification and the potential
displacement of the very residents that new parks and greenspace were meant to serve
(Dooling, 2009; Wolch et al., 2014). Checker (2011) describes observing this dynamic in
New York City’s Harlem neighborhood following the implementation of large-scale
“sustainable development” initiatives. The process has been referred to by various names

3

including “ecological gentrification”, “eco-gentrification”, and “green gentrification”
(Wolch et al., 2014). Green gentrification is the term I will adopt for this thesis.
When cities display a tendency toward green gentrification, Wolch et al. (2014)
have suggested combatting this trend using a set of strategies that they term “just green
enough”. They point toward several examples where community members or city
planners have implemented a version of the strategy successfully, such as a neighborhood
in Brooklyn, NY that chose to restore existing trail systems and clean up industrial areas
rather than investing in a centralized new park space that may have greatly impacted local
retailer’s rent and property values. Although activists and planners might choose to adopt
one of many different approaches to the “just green enough” strategy, the easiest to
measure the effects of using spatial analyses are situations like the one described above.
By favoring smaller, more discrete UGS investments scattered across multiple sites,
rather than larger, more expensive civic projects, the wealth-concentrating impacts often
associated with these bigger projects might be largely avoided (Wolch et al., 2014).
My main objective for this thesis is to explore UGS inequity in San Francisco,
and to assess how it might relate to green gentrification. A secondary objective is to
indirectly assess the applicability of the “just green enough” strategy in combatting green
gentrification trends in San Francisco, should they be identified. Addressing these
objectives will require answering three related questions. First, how equitably distributed
are parks and urban greenspace in San Francisco, CA based on various demographic
metrics for neighborhood SES? Then, what is the relationship between UGS access and
neighborhood gentrification in San Francisco, when considering the placement of UGS
installations, as well as changes in SES between 1990 and 2010? And finally, are
4

gentrified neighborhoods associated more with the total number of nearby urban
greenspaces or their total acreage? If gentrification is found to be more correlated with
total acreage, this would support an element of the “just green enough” hypothesis.
I have chosen to focus on San Francisco, because over the past few decades it has
experienced the rising property values and skyrocketing rental prices that are typically
associated with gentrification more so than many other cities in the Western US (Miciag,
2015). Its numerous examples of urban parks and greenspaces, and its adoption of several
innovative new forms of green infrastructure which I will include in this analysis, are two
more reasons that San Francisco makes an excellent model city for this project.
The smaller-scale examples of urban greenspace that I’ll include in this analysis
under the umbrella of a green infrastructure category are not typically factored in to most
UGS equity analyses. They represent one of this thesis’s unique contribution to the
existing literature, and include green roofs, green stormwater infrastructure, parklets, and
privately-owned public open spaces. The purpose of including these is to employ a more
expansive definition of what constitutes urban greenspace, and to further assess the
differential impacts of smaller-scale vs. larger-scale UGS as it relates to gentrification
and displacement.
Green gentrification and “just green enough” strategies have been explored to
some extent in the social science and political ecology literature, but have not been wellstudied in the past using quantitative data or spatial equity analyses. Analyzing changes
in neighborhood demographic composition over time will be an important test for the
existence of green gentrification in San Francisco. And, by quantitatively testing the

5

hypothesis that neighborhoods can be made “just green enough” through prioritization of
smaller UGS projects, I can test the effectiveness of one strategy for pushing back against
green gentrification-related displacement. My goal in exploring the challenges and
opportunities cities face when striving toward greater UGS equity is to contribute to this
important, emerging field of study.

Literature Review
Introduction
Due to the wide range of social and ecological benefits that people experience
through their association with greenspace, scientists consider it to be an important
environmental amenity (Boone et al., 2009). Thus, many would assert that lack of access
to urban greenspace (UGS) by residents of lower socioeconomic status (SES) represents
an environmental injustice akin to these groups shouldering disproportionate burdens in
their exposure to toxic waste or pollution (Perkins et al., 2004).
Attempts to address greenspace inequity are often unsuccessful. Some of the
challenges arise from urban developers’ and environmentalists’ ideas and assumptions
about how a project’s social dimensions ought to be addressed. City planners and
developers employ ideas that have arisen from the literature on sustainable development
to assess a green development project’s social, economic, and ecological sustainability
aspects, also referred to as the “triple bottom line” (Dale & Newman, 2009). These three
dimensions are intended to be viewed as equally important, but in effect this model may
inadequately address inequity due to internal contradictions inherent in the way social
sustainability is defined. For example, the concepts of “equity” and “livability”, both key
6

components in achieving social sustainability, may be at odds with one another, such that
increasing one causes a decline in the other (Dale & Newman, 2009; Godschalk, 2004).
The purpose of this literature review is to explore the many factors influencing
where greenspace projects are targeted in order to develop a deeper understanding of why
inequity so often persists. Why are some communities more susceptible than others to
real estate speculation and gentrification? How might residents of these communities
achieve greater greenspace equity without increasing their susceptibility to real estate
market trends? To answer these questions, I’ll discuss literature from a variety of fields.
First, I’ll expand on the many beneficial impacts of exposure to UGS to underline
why equitable access is such an important environmental justice issue. Then, I’ll discuss
triple bottom line analyses’ inadequacies, situate UGS equity within the broader
environmental justice literature, and discuss why spatial analysis tools are appropriate for
addressing distributional injustices. I’ll then explore the idea that investments in UGS and
green infrastructure may be directly linked with the processes of gentrification and
displacement of lower SES residents. And I’ll briefly discuss case studies where
implementation of “just green enough” strategies may have helped communities to avoid
being impacted by gentrification. Finally, I will provide an overview of past UGS equity
analyses’ measurement techniques and their findings. This will serve to highlight ways in
which my thesis expands upon the extant literature and provides some unique insights to
this burgeoning new field.

7

Benefits of Urban Greenspace and Access to Nature
Parks and urban greenspace are key components of the urban built environment,
offering a wide range of social and environmental services (Zhang et al., 2011). They’ve
been shown to improve residents’ overall quality of life as well as their physical and
mental health. Some evidence suggests that neighborhoods become safer and experience
lower crime rates following investments in UGS (Kuo & Sullivan, 2001; Kondo et al.,
2015). Green infrastructure installations and certain other forms of UGS can offer
additional ecological perks that may also directly benefit nearby residents, particularly
those on their periphery.
Proximity to parks correlates with increased physical activity among urban
residents and with their improved overall health (Zhou & Kim, 2013). At least one study
suggests that senior citizens who exercise by walking through parks or UGS experience
increased longevity (Takano et al., 2002). The likelihood of residents receiving their daily
recommended amount of physical exercise was found to triple among those living within
walking distance of a park compared with those living further away (Giles-Corti et al.,
2005). Childhood asthma in 4- to 5-year-olds was found to be less common in areas of
higher urban street tree density in New York City (Lovasi et al., 2008). A study exploring
the connection between UGS and obesity found that children living in areas with greater
access to greenspace generally had lower body mass index (BMI) scores (Bell, Wilson
and Lui, 2008).
Chiesura (2004) has argued that access to parks and greenspace provides a
particularly important metric for determining overall quality of life, especially in big

8

cities or other increasingly urbanized settings. Boone et al. (2009, p. 784) state “more
than a recreation space, parks serve the critical functions of providing public space and a
right to the city.” By this, they mean that there are few open spaces for the public to
congregate where they are not expected to engage in consumerism or to have their
activities restricted in other ways. This is especially important to consider with regard to
homeless or very low-income populations.
According to surveys conducted by Kabisch & Haase (2014), access to UGS is
directly associated with a self-reported improvement in quality of life among residents, as
measured through their increased ability to engage in exercise, passive recreation, social
activities, or to commune with nature. UGS provides people with space to experience
solitude, peace, and tranquility that might otherwise be lacking in a hectic urban setting
(Wolch, Byrne, & Newell, 2014). Other survey findings suggest that people associate
experiences of nature in an urban setting with positive feelings such as “happiness”,
“silence”, and “beauty” (Chiesura, 2004). Respondents considered these emotions to be
important to their overall sense of wellbeing, describing benefits such as “regeneration of
psychophysical equilibrium” or “stimulation of spiritual connection with the natural
world”. Additionally, parks or UGS represent some of the few existing areas left in the
city where urban residents might experience contact with plants, animals, and other forms
of biodiversity on a day-to-day basis (Wolch, Byrne, & Newell, 2014).
UGS promotes relaxation, meditation and social cohesion, all of which may
contribute to the improved health outcomes that are observed among residents (Zhou &
Kim, 2013). A wealth of evidence appears to confirm this; for example, an early study
testing this hypothesis found that hospital patients with a view of nature from their
9

windows recovered more quickly than those with a view of buildings (Ulrich, 1984).
Many subsequent investigations have strengthened the link between natural environments
and improved psychological health (Chiesura, 2004). One major study which interviewed
more than 4,500 patients found that those with greater amounts of UGS within a 3km
radius of their homes, were significantly less affected by stressful life events compared
with those who had more limited access to UGS (Van den Berg et al., 2010). Another
study found that children diagnosed with Attention Deficit Disorder (ADD) displayed
significantly fewer symptoms after engaging in activities where they had been exposed to
nature or “green” environments (Taylor, Kuo, & Sullivan, 2001).
Those living in proximity to greener environments may also experience less
mental fatigue and aggression, and this association could be responsible for the lower
crime rates observed in these areas (Kuo & Sullivan, 2001). Compared with those areas
lacking them, people perceive a greater sense of security and stronger social ties in
neighborhoods with parks, perhaps because more widely used public spaces result in
greater opportunities for social interaction (Boone et al., 2009).
In summary, a wealth of evidence has shown that those living near parks, urban
trees, and other green environments experience improved physical health, mental
wellbeing, and a greater sense of ownership in their communities and connectedness with
their neighbors. Exposure to greenspace may even provide residents with an enhanced
ability to cope with stress, contributing to an overall drop in neighborhood crime rates.

10

Green Infrastructure’s Ecological Services
Since this thesis incorporates green infrastructure systems as separate forms of
UGS, it is worth noting some of the unique benefits these systems may provide. Broadly
defined, green infrastructure refers to any system making use of soil and vegetation to
provide ecological services that protect or enhance environmental or public health (for
example bio-swales, constructed wetlands, rain gardens, urban farming, green alleys and
roofs, or even vegetated median strips can all serve these functions; Dunn, 2010). Even
green infrastructure designed primarily for ecological improvement may still provide
important social benefits. When unused grey infrastructure such as alleys, utility
corridors, or railways is converted to green infrastructure, people use these spaces more
frequently for socializing, walking, or exercising (Wolch, Byrne, & Newell, 2014). These
forms of green infrastructure, while unlikely to provide enough space for organized
recreational activities, may be equipped with “micro-gyms” which have been shown to
encourage residents’ physical activity (Wolch, Byrne, & Newell, 2014).
Green infrastructure may improve community self-image and foster community
pride. Dunn (2010) argues for specifically targeting the installation of green infrastructure
in lower-income neighborhoods due to such poverty-reducing benefits as helping to
ensure urban food security or providing “green collar” jobs. She points out that President
Obama’s economic stimulus package lead to a 31 percent increase in the number of
people hired to work “green collar” jobs between 2007 and 2009 (Dunn 2010). While
majority of these jobs likely involved R&D and manufacturing of components for
renewable energy technologies, future initiatives could be targeted to include the
installation and maintenance of green stormwater infrastructure and related technologies.
11

In this way, green infrastructure investment could provide opportunities to expand the
urban workforce and provide green job training, while improving the function of urban
environments’ ecosystem services.
Ecological services associated with green infrastructure include pollutant
removal, filtration of air, reduction in ambient temperature, attenuation of noise pollution,
infiltration of stormwater, reduction of floods, groundwater recharge, provision of habitat
for wildlife, and potentially even providing food for residents (Wolch, Byrne, & Newell,
2014; Boone et al., 2009). These benefits and services have been extensively documented
in the literature. For example, an urban greening initiative in Hangzhou, China was found
to reduce temperatures by 4 to 6 degrees in certain areas of the city, ameliorating the
urban heat island effect (Wenting, Yi, & Hengyu, 2012). The US EPA states that
increased temperatures from the urban heat island effect contribute to peak energy
demands as well as an increase in heat-related illness and mortality. The capacity to
reduce local temperature and buffer the effects of heat on nearby residents becomes
particularly evident and important during periods of very hot weather (Kabisch & Haase,
2014).
Researchers have found that due to the close association between air quality and
prevalence of trees, they can use leaf area index as an indicator of local air quality
(Jennings, Johnson Gaither, & Gragg, 2012). Studies suggest that new parks or
greenspace may also help to mitigate the human health impacts of contaminated soil
since many plant species can take up and sequester contaminants (Poggio & Vrščaj,
2009). Of course, in these cases it is important for residents to realize that toxins may
then accumulate in these plants and trees themselves and take all necessary precautions.
12

Urban planners may want to develop a stronger awareness of these ecosystem
services, ensuring that they are considered when planning for how open spaces will be
developed. Carefully targeting green infrastructure installations to ensure they are as well
distributed as possible will be important, since it is often unclear at what scale their
biophysical processes operate. Those living directly adjacent to them may be the only
beneficiaries of some ecological services, though other services may impact the city as a
whole, or even the larger global climate system (Kabisch & Haase, 2014). Ngom,
Gosselin, and Blaise (2016) contend that improving cities’ amount and quality of UGS
and green infrastructure may be a key component in combatting climate change.
Challenges in Achieving Greenspace Equity
“Sustainable development, if it is actually to be sustainable, should not be for some
but for all” – Dale & Newman, 2009
Unfortunately, many challenges stand in the way of achieving greenspace equity.
The inadequacy of the “triple bottom line” social sustainability model, the
unpredictability of demographic and economic fluctuations, and the stubborn persistence
of systemic classism and racism all play important roles. Exploring the relevance of each,
through case studies in the social science literature will illuminate this bigger picture that
might otherwise be missed by simply reviewing past quantitative analyses addressing
distributional inequities.
Campbell (1996) effectively illustrates some of the inherent contradictions in
sustainable development’s attempts to address social, ecological, and economic
sustainability simultaneously (Figure 1). He identifies three major tensions which he
13

represents as the lines of a triangle where each of its 3 points represents a sustainability
goal (economic, social, and ecological). The contradiction between ecological and
economic sustainability is termed the “resource conflict”, between ecological and social
is the “development conflict”, and between social and economic is the “property
conflict”. Taking the resource conflict as an example: ensuring a project is ecologically
sound often costs money, making it less economically sustainable, whereas favoring
financial stability may result in less ecologically sustainable outcomes. Building upon
these critiques, Godschalk (2004) goes even further, making the case that the definition
of social sustainability is contradictory in and of itself.
Figure 1: Conflicts to Achieving Triple Bottom Line Sustainability
The corners of this triangle represent social, economic and environmental sustainability, the italicized text
highlights the types of conflict that arise between these different forms of sustainability when attempting to
achieve all of them simultaneously. Adapted from Campbell (1996).

14

Social sustainability’s inherent tensions can be illustrated by exploring two
concepts that tend to be at odds with one another. The first, “soft infrastructure”,
contributes to community wellness and includes recreation or cultural centers as well as
human and social services (Dale & Newman, 2009; Godschalk, 2004). The second,
“livability elements” refer to things that improve residents’ quality of life such as
greenspace, street furniture, access to cafes, etc. Unfortunately, installing new livability
elements can potentially destroy or degrade soft infrastructure (Dale & Newman, 2009).
For example, expensive new parks or housing developments might result in a desire from
more affluent residents for facilities providing human and social services for homeless
and at-risk populations to be relocated.
City planners’ explicit or implicit goals for neighborhood greening often involve
raising property values and inviting commercial development rather than improving the
lives of a neighborhood’s current residents (Soper, 2004). This may help to explain their
failures to address social equity concerns. For example, expensive “green” housing
developments and shopping centers provide benefits to those who can afford to access
them, but often do nothing to improve low income residents’ lives (Godschalk, 2004).
Although conceptually, all three pillars are meant to be valued equally, in practice, the
ecological and economic aspects of sustainability often overshadow its social dimensions
(Dale & Newman, 2009). Sustainability in general often becomes conflated with mere
ecological sustainability, thus obscuring the social dimension and fueling many of the
more explicit criticisms of sustainable development’s three-pillar paradigm (Curran &
Hamilton, 2012).

15

Godchalk’s (2004) recommendation for addressing the social sustainability
pillar’s lack of clear focus and inherent internal tensions, is to incorporate “livability” as
its own separate dimension of the sustainable development paradigm, in other words as a
fourth pillar. He argues that equity and livability are clearly distinct, and should not be
lumped together under the banner of social sustainability. “Livable” communities
routinely exclude low SES residents, and those featuring green amenities usually demand
significant market premiums (Luke, 2005; Dale & Newman, 2009). A paradigm that
adequately incorporates the social dimension of community development by focusing on
both equity and livability would not (actively or passively) cater to only the higherincome segment of the population. But, in practice this is what we often observe
happening through the displacement of lower-income households (Dale & Newman,
2009).
Distributive Environmental Justice
Greenspace inequity in the United States may stem in part from differing
historical patterns of land development and park design philosophies, but underlying
ethnic-racial inequalities and structural discrimination play important roles as well (Byrne
& Wolch, 2009). The environmental justice (EJ) framework is useful in examining
distributional inequities. It embraces the idea that all people, regardless of SES, are
equally entitled to livable, clean, and pollutant-free environments (Wen et al., 2013).
Historically, EJ research has focused primarily on exploring the disproportionate burdens
faced by lower SES communities, but recent studies have also begun to analyze the
spatial inequities in their access to environmental amenities such as UGS (Wen et al.,
2013).
16

Different forms of environmental inequity often occur simultaneously, so it is
important to recognize that spatial analysis tools can only identify distributive injustices.
According to Kabisch and Haase (2014), distributive justice refers to the fair allocation of
amenities or burdens. Procedural justice refers to inclusion in planning and decisionmaking processes. And Interactional justice refers to safety from violence and overt
discrimination. Procedural and interactional injustices are usually undetectable using
spatial analysis techniques. Addressing them will require creativity and direct inclusion
of groups in the community that are most impacted. For example, implementing
participatory landscape development plans might be one effective strategy for combatting
procedural injustice (Jennings, Johnson Gaither, & Gragg, 2012).
Since urban land use changes affect all segments of the population and often
directly impact residents’ health and wellbeing, assessing a project’s potential EJ
implications is critical (Jennings, Gaither, & Gragg, 2012; Wolch, Byrne, & Newell,
2014). Yet despite a wealth of evidence from over three decades of research
demonstrating UGS’s social benefits, few public policy or planning strategies for
greenspace explicitly incorporate factors such as human health and wellness into their
decision-making processes (Jennings, Johnson Gaither, & Gragg, 2012).
Green Gentrification
Displacement of lower SES residents resulting from investments in green
livability elements has been referred to as “green gentrification” (Wolch et al., 2014). In
this section, I will describe some of the ways which green gentrification contributes

17

towards persistent UGS inequity. But first, it will be necessary to provide some
background information about gentrification more generally.
Ruth Glass coined the term gentrification in the 1960s to describe demographic
changes that she had observed in London neighborhoods following economic
revitalization (Dale & Newman, 2009). Merriam-Webster traces the word’s etymology to
the concept of gentry, referring to a privileged or ruling class. Some scholars prefer
Nelson’s (1988) definition. She defines gentrification as investment in urban
communities that leads to an inflow of higher-income or higher socio-economic status
(SES) residents. This definition does not presuppose that displacement of lower SES
residents. But others have argued that displacement is an integral component of the
gentrification process, occurring more often than not (Eckerd, 2011). In this thesis, I will
measure gentrification as a function of displacement. Therefore, instances where
displacement has not occurred will not be identified by my spatial methodologies.
Gentrification may impact neighborhoods positively in a number of ways. For
example, it can improve aesthetic appeal and safety of a community, upgrade its housing
stock, reduce sprawl, and lead to a decline in car use (Atkinson, 2004; Bromley et al.,
2005). However negative impacts on the community may include less diversity, more
expensive housing, and lower-income residents may often be displaced to the edge of
cities where public transit and other services are limited (Atkinson, 2004; Dale &
Newman, 2009).
Attempts to address UGS inequity can be thwarted by the processes of
gentrification. Improved public health and attractiveness of neighborhoods resulting from

18

improved UGS access may lead to higher demand and increased housing costs for
residents. This can potentially kick-start a positive feedback loop where commercial
developers compete with one another to buy up property and invest in shops and
apartment complexes (Bentley, Baker & Mason, 2012). When determining where to live,
residents seek out communities where public goods match their desires and ability to pay
and sort themselves out accordingly (Eckerd, 2011). A self-reinforcing mechanism leads
wealthier neighborhoods with access to higher quality greenspace to attract higherincome residents (Boone et al., 2010). This can lead to negative public health
consequences particularly with regard to mental health. Not only do residents often
experience continued UGS inequity after being displaced, but displacement itself causes
additional stress, as do the higher housing costs experienced by those who might avoid
being displaced (Bentley, Baker, & Mason, 2012).
Differential UGS quality and function has been shown to reinforce social
stratification (Ngom, Gosselin, and Blais 2016). Urban real-estate market forces
combined with environmental policies and institutional racism can be powerful drivers
for displacement. Gould and Lewis (2009) state that these factors working together
explain how the ‘greening’ of urban areas can be a euphemism for their ‘whitening’.
Some studies have specifically identified distinct racial trends that can occur during the
green gentrification process. For example, Essoka (2010) found that across four EPA
regions, brownfield revitalization projects were associated with widespread displacement
of Black and Latino populations.
Curran and Hamilton (2012) contend that green gentrification processes and
resultant displacement of long-term residents stem directly from urban environmental
19

policies that are inextricably linked with the desire for growth and economic
development. Rising property values in typically lower-income neighborhoods often lead
to higher property taxes and impact retail infrastructure as well, disrupting long-time
residents’ ability to afford basic goods and services (Zukin et al., 2009). Even the
addition of relatively minor greenspace embellishments can have the potential to
significantly increase property values and displace lower-income residents, according to a
study in Hangzhou, China (Chen, 2012).
Checker (2011) argues that high-end developers who initiate the processes of
green gentrification tend to ignore the political implications of building and marketing
sustainable communities. Their rhetoric is built upon the successes of the EJ movements,
but they appropriate its language to serve their own purposes. By tending to focus solely
on improved livability elements and environmental amenities, developers ignore
distributive justice concerns and the potential for low SES residents to be negatively
impacted by their decisions. Thus, urban sustainability and political ecology scholars
should explicitly distinguish between differing views of social sustainability to raise
awareness about attempts to co-opt EJ concepts to suit commercial interests (Checker,
2011). These attempts are particularly important to recognize because privileging
ecological sustainability over social sustainability through environmental policy can
profoundly impact lower SES residents, spatially, politically, and economically (Dooling,
2009).

20

Just Green Enough
Dale & Newman (2009) argue that planners must explicitly design projects based
around social equity by directly considering distributive justice concerns and other
political implications, otherwise their often-stated goals of improving both ecological
health as well as social equity cannot be realized together. Addressing distributive
environmental justice issues among low SES communities while avoiding the unwanted
consequences of their displacement requires a clear set of intentions and a careful
balancing act between the need to provide essential environmental amenities and the
desire to avoiding triggering real estate speculation stemming from improved livability
elements. For sustainable community development to address the social imperative,
development projects must actively plan how to ensure communities will be accessible to
all (Dale & Newman, 2009). Although, it is also important to keep in mind that
maintaining a community as working class should not equate with denying those
residents access to UGS (Curran & Hamilton, 2012).
One approach planners might take to address green gentrification involves
adopting a “just green enough” strategy. Implementation depends on city officials and
community stakeholders designing UGS in a way that is explicitly shaped by community
needs and concerns rather than designing only for ecological restoration or commercial
development purposes (Wolch, Byrne, & Newell, 2014). Replacing a traditional marketbased and/or ecological approach to greenspace design with a “just green enough”
strategy can be very challenging, so its success often hinges on community involvement
and activism (Wolch, Byrne, & Newell, 2014).

21

One potential method for implementing “just green enough” solutions involves
promoting greenspace projects smaller in scale and more evenly distributed throughout
neighborhoods, avoiding large-scale, more geographically concentrated civic projects
likely to be associated with green gentrification (Wolch, Byrne, & Newell, 2014).
However, use of this strategy may hinge on multiple appropriately spaced parcels of land
being available, or on creative implementation of smaller scale green infrastructure
projects. Schauman and Salisbury (1998) contend that smaller-scale projects have the
potential to more evenly distribute access to greenspace among urban residents rather
than creating centralized focal points that property developers might be drawn to.
Alternative “just green enough” strategies often require addressing procedural injustices
through grassroots activism and community involvement (Curran & Hamilton, 2012;
Jonas & While, 2007). Wolch, Byrne, & Newell (2014) warn that being just green
enough demands a calculated balancing act. Collaboration among local officials and
community groups is an important piece of this. Stakeholders’ willingness to contest
powerful developers’ interests and oftentimes the interests of mainstream environmental
advocates can also be critical. This usually requires grassroots activism as well as early
and continued involvement in the planning process by concerned members of the
community.
The following case studies illustrate how communities can address UGS inequity
while avoiding “green gentrification” in a variety of different ways.


Some local governments have enacted policies which allow them to take advantage of
underused lands and convert them to greenspace with the goal of addressing
environmental justice as a planning priority (Wolch, Byrne, & Newell, 2014). The
22

State of New York’s Brownfield Opportunity Area (BOA) program illustrates how
important intentionality can be to the production of new greenspace (Curran &
Hamilton, 2012). BOA does not intend to transform vacant toxic sites with new
development in mind, but rather to clean them up to foster ecological and public
health. Because they have these specific goals in mind, projects that the agency
supports are less likely to trigger neighborhood gentrification compared with other
brownfield redevelopment projects throughout the nation (Curran & Hamilton, 2012).


The Greenpoint neighborhood in Brooklyn illustrates how higher and lower income
community members might work together to address environmental inequity. Here
“gentrifiers” and working class people collaborated in demanding toxic waste cleanup
that allowed for continued industrial activity, saving blue collar jobs. The restored
area around a polluted creek was converted to new greenspace, but the community
sought to explicitly avoid the “parks, cafes, and a riverwalk” model that residents
feared would trigger speculative development leading to gentrification (Curran &
Hamilton, 2012).



In Toronto, nonprofit groups encouraged local city planners to move away from a
rewilding restoration approach and instead adopt a parks strategy preferred by the
community that focused more on community gardens and urban agriculture. The
refocusing connected the project with residents’ concerns about food security, human
health, and job creation (Newman 2011).



Based on suggestions by members of the homeless population in Seattle, Dooling
(2009) suggests a novel approach: homeless residents could participate in a green
stewardship program run by local parks departments to remove invasive species. The

23

program would allow them the right to continue using public park spaces, provide
them with a living wage and a chance to participate in the formal economy, but would
also encourage both social and ecological sustainability. Such an approach would
combat greenspace inequity and improve ecological health, although it is less
applicable to the planning and development phase of park installation (Dooling 2009).
Evidently, several “just green enough” strategies have been shown to potentially
work, some requiring more activism and community involvement than others. From a
city planning perspective, the most obvious and easily implementable strategy would
appear to be avoiding very large parks and civic projects, and attempting instead to
distribute urban greenspace more evenly throughout a neighborhood (installing greater
numbers of smaller, more discrete parcels). This approach to combatting green
gentrification represents the easiest method to analyze spatially. For my analysis, number
of UGS projects vs. their total area for a given census block group will be the only metric
I examine quantitatively in assessing support for the “just green enough” hypothesis.
Techniques for Measuring UGS Access
Initiatives attempting to ensure that greenspace is more accessible to all require
carefully targeting parks and UGS through analysis of spatial distribution of vulnerable
populations (Comber, Brundson, & Green, 2008). Measuring UGS accessibility for
different socioeconomic groups though not always straightforward, can be done using
many different Geographic Information Systems (GIS) techniques. This section describes
several of these techniques, while the following section touches on the diversity of
findings that researchers from around the world have arrived at so far.

24

Kabisch and Haase (2014) highlight the importance of city planners considering
age and culturally dependent needs with regard to park space, in addition to focusing on
overall amount of greenspace access by neighborhood. Many past studies have simply
measured acres per person of nearby park space, failing to consider the differing needs or
choices of distinct segments of the population (Agyemon, 2008; Abercromie et al., 2007).
For example, a middleclass family may have different needs than a working class single
mother (Boone et al., 2009). Ngom, Gosselin, and Blais (2016) concluded that the social
function and quality of greenspace are both extremely important, but are not always
measured. This raises the possibility that some neighborhoods may have inadequate
access to park features that encourage physical exercise. This could lead them to
experience a significant reduction in physical health benefits.
Park activities and preferences can differ significantly between ethnic groups
(Comber et al., 2008). Kabisch and Haase (2014) found that native Germans in Berlin
prefer grassy open spaces, playgrounds, and sports fields, while immigrant communities
prefer areas conducive to barbecuing and picnicking. So, besides considering the diverse
social and ecological functions of different park types, it may also be necessary to
develop an awareness of local populations’ preferences to encourage the most widespread
use of new UGS projects.
Abercrombie et al.’s (2008) findings show that low-income and racial/ethnic
minority groups in the US have lower levels of physical activity. Although many factors
likely exist that contribute to this problem, inequitable distribution of parks and
recreation facilities could help to explain these disproportionate activity levels. Jennings
et al. (2012) point out that a significant number of previous studies exploring racial/ethnic
25

groups’ access to UGS have failed to account for differing park configurations and
purposes preferred or needed by different groups. Ngom et al. (2016) contend that how
greenspace is defined strongly impacts planning strategies for improving equitable access
to it, but that consensus can be difficult to reach, since some would prefer to measure
social benefits such as public gathering spaces and recreation facilities while others focus
more on ecological services provided by natural areas.
How close must someone live to a UGS project for this greenspace to be
considered accessible to them? A quarter mile (or a roughly five-minute walk) represents
the typical standard threshold where people are most likely to walk to reach a park or
recreation area. People living greater than a half mile away tend to be much more likely
to drive to reach a park. So, many U.S. cities (Seattle, Phoenix, and Portland, to name a
few) have adopted a half mile or 800-meter distance metric to set park placement goals
(Boone et al., 2009). One technique for ensuring people live within a quarter- or half-mile
buffer from a park or UGS involves creating centroids (a point that identifies a geometric
object’s center of mass) for all park polygons and measuring buffers outward from there,
analyzing the proportional socio-economic makeup of overlapping census block groups
(Boone et al., 2009).
A more commonly used technique for measuring residents’ distance from UGS or
other such areas of interest involves creating centroids for each individual census block
or tract polygon and to create buffers around these, measuring number and area of park
spaces or UGS that overlaps (Boone et al., 2009). This technique has been referred to as
the “container approach” (Zhou & Kim, 2013). Census block groups, as the smallest
available geographic unit that incorporates SES data, tend to provide the most useful
26

information for these analyses (Zhou & Kim, 2013). However, some studies have
conducted analyses using the larger census tract areas, finding this scale to be useful for
its ease of use and generalizability (Heynen, Perkins, & Roy, 2006).
Network analysis represents a more sophisticated technique for measuring access,
incorporating an analysis of the actual accessibility by roads, trails, or walking paths by
which neighborhood residents are likely to travel to get to parks (Oh & Jeong, 2007). Past
network analyses have had to rely on road data only, but more sensitive analyses based
on Google Maps tools can now be performed specifically measuring walking path
distances (Zhou & Kim, 2013). Another more sophisticated data analysis tool
incorporates kernel-smoothing functions to transform point values into continuous
surface values, a technique that overcomes some of the imprecision of the simple polygon
buffering that the container approach utilizes (Zhou & Kim, 2013).
For this thesis, I’ve chosen to use a simplified technique where quarter mile
(400m) buffers are applied directly to census block groups rather than to their centroids.
While slightly less precise, this technique should be sufficient due to the density and
average size of census block groups in San Francisco.
Findings of Past Spatial Analyses
The following section illustrates the diversity of geospatial analyses that have
been performed in the past, and identifies gaps in this literature that this thesis will
attempt to fill. A diverse number of approaches to geospatial analyses of UGS
distribution have been taken in the past, and results have varied widely depending on
methodologies and locations chosen. Some have identified significant racial-ethnic or
27

class-based disparities (Boone et al., 2009; Comber et al., 2008; Wolch et al., 2005),
while others have largely identified no such patterns (Timperio et al., 2007). This section
will compare past studies findings and describe some gaps in the existing literature that
this thesis may help to address.
Comber, Brundsen, and Green (2008) quantified UGS access by different
religious-ethnic groups in Leicester, UK, believing their study to be the first UGS equity
analysis focusing specifically on religious groups. They identified significant inequities
experienced by minority groups, and found that while Leicester generally has adequate
amounts of greenspace, neighborhoods with large Indian, Hindu, or Sikh populations
experienced much more limited access. Subsequent case studies looking at other cities
have found similar patterns with regard to ethnic and religious minorities throughout
Europe and in parts of the US as well (Kabisch & Haase, 2014).
Boone et al. (2009) found that in Baltimore, a higher proportion of African
American residents lived within walking distance of a park, but white residents generally
enjoyed greater access to larger parks. Heckert (2013) discovered that non-white
residents and renters in Philadelphia had access to significantly less overall greenspace
when compared with white residents and homeowners. Heynen et al. (2006) found that
total tree canopy was positively correlated with median household income in Milwaukee,
WI, implying that investment in urban trees may disproportionately favor the wealthy
over those with greater socioeconomic need. A nationwide study of the US by GordonLarsen et al. (2006) identified a pattern of lower SES residents having consistently less
access to both free and for-profit recreational facilities.

28

Not all spatial analyses have identified inequities in UGS access. For example,
Timperio et al. (2007) found no association between resident’s SES and their access to
public open space in Melbourne, Austalia. Highlighting the inconsistency in equity
findings, Wen et al. (2013) point towards a few additional studies that find a positive
correlation between lower SES communities and access to UGS. One example is Cutts et
al. (2009) who found residents of Phoenix, AZ designated as lower SES and more at risk
for obesity, tended to have greater access to public parks and walkable space. Other
studies have identified some level of inequity but often paint a more nuanced picture. For
example, Miyake et al. (2011) found that nearly all residents in New York City (>95%)
may have adequate access to city parks when defined as living within walking distance of
at least one park, but digging deeper into the data, the authors recognized distinct racial
trends in total amount of accessible park space available to different groups.
Zhou and Kim’s (2013) analysis of park and greenspace access in several Illinois
cities discovered that in Bloomington and Rockford, number of African American
residents positively correlated with neighborhood parks. There appeared to be no such
correlation in either direction for any of the other cities they examined. These researchers
did however identify a negative correlation between amount of tree canopy and
proportion of African American residents in all cities but one, Bloomington. They did not
provide reasons for why this might be the case, but did point out that the larger than
average Asian population in Bloomington correlated with reduced tree canopy, so
perhaps African Americans represented a smaller proportion of the overall minority
group population in this particular area (Zhou and Kim 2013).

29

Wen et al. (2013) conducted one of the largest scale UGS equity map analyses in
the literature. They cross-referenced park and greenspace access with demographic data
from cities across the entire coterminous US, and analyzed distributions in relation to
Black, Hispanic, and low-income populations. They found that in urban/suburban
settings, poverty negatively correlated with distance from parks, but in more rural areas,
the opposite proved to be true. They found race-ethnicity represented important correlates
with park access as well, but they differ across urbanization levels (rural vs. urban), and
patterns are less straightforward than might have been assumed. This may be explained in
part by the tendency towards more rural areas to be less economically or racially
segregated, as described by the following study.
Saporito and Casey (2015) conducted another interesting nationwide analysis in
which they examined levels of racial and economic segregation by city and found that
more segregated cities, were associated with greater inequality in access to greenspace.
Their findings show that while lower income groups and minority populations generally
tend to live in neighborhoods with less vegetation, this general pattern is exacerbated and
becomes particularly pronounced where racial or economic segregation is occurring on a
wider scale.
Abercrombie et al.’s (2008) analysis of access to recreational facilities based on
SES demographics found low SES communities to be negatively correlated with
accessible recreational facility. However, the authors identified significant variations
between communities and regions. So once again this pattern appears to be very contextspecific and probably relates to local policies, political priorities, and resource
availability.
30

Only a few studies have examined how neighborhoods change over time with
regard to park placement and greenspace equity. Ngom, Gosselin, and Blais (2016)
conducted a temporal analysis investigating how greenspace access in Montreal has
changed between 1996 and 2011. They incorporated additional demographic statistics
that other studies have not examined such as age and gender breakdowns, and they
accounted for population density to explore the role that it might play. Their study has
shown that over those 15 years, Montreal’s UGS equity improved significantly, but some
inequities still existed. For example, wealthier neighborhoods still experienced overall
greater access to UGS as well as disproportionately greater access to parks and
recreational facilities.
Weems’ (2016) dissertation analyzed changes in greenspace access between 1990
and 2010 in Seattle, WA. She also explored trends in neighborhood gentrification over
this 20-year period, and identified a pattern of increased neighborhood gentrification
following greenspace investments. Interestingly, her results implied a potential causal
connection between the two, as she estimated that neighborhoods experiencing
gentrification often had received greenspace investments about ten years earlier. Weems’
(2016) method of identifying “gentrification” was to assess whether a census block
group’s median level of education, household income or home value had moved from
below the city’s overall median levels to above. She considered an area to have gentrified
if it met two out of these three metrics. Other studies have used a gentrification index
composed of two factors: percentage of adult population over 25 that holds a college
degree (BA or higher) and percentage of adult population working in professional or
managerial positions (Eckerd, 2011). My thesis will adopt a combination of these
31

techniques, and using ordination construct an index that incorporates median home value,
median household income, and percent college-educated residents.
Conclusion
Scientists have found broad support for the idea that UGS promotes human and
ecological health. However, despite wide recognition of these associations, distributing
UGS more equitably remains a serious challenge. The way that city planners and
commercial developers define social sustainability may contribute to this problem’s
persistence. By explicitly prioritizing social equity concerns so that they are at least held
to be equal with ecological and/or economic sustainability concerns, planners might
better address UGS inequity.
Systemic racism and discrimination also play important roles in perpetuating the
status quo. Assessing distributional inequities can be relatively straightforward, but
addressing procedural injustices will also be necessary to ensure low SES populations are
adequately served in the future. The environmental justice literature provides a useful
framework for exploring these issues since many past struggles relating to disparities in
the distribution of environmental burdens may be directly applicable.
Green gentrification of lower income neighborhoods can negatively impact efforts
to achieve greater UGS equity. Some communities have identified strategies to get
around this problem, but they will require further testing to assess their effectiveness and
applicability. I have only identified a few other quantitative geospatial analyses that
specifically addresses green gentrification, so my thesis will contribute to a new and
emerging field of study in this way. The equity analysis portion of this thesis will set
32

itself apart from other analyses by incorporating a specific focus on new forms of green
infrastructure which tend to be smaller in scale. These increasingly popular forms of UGS
may point toward the future potential for more widely-distributed and well-targeted
greenspace installations.

Methods
Overview
This methodology’s purpose is to assess whether individuals of lower
socioeconomic status (SES) experience disproportionate access to urban greenspace
(UGS) in San Francisco, and to assess whether the amount of newly installed and/or total
UGS in the City is linked with neighborhood gentrification. I divided UGS into several
categories, each to be analyzed separately. Total acreage of UGS as well as overall
number of separate installations were assessed to see if differences in their associations
existed, and if so to assess which was more strongly linked with neighborhood
gentrification.
My three related research questions were:
1 Is San Francisco’s new and total urban greenspace (UGS) correlated with
neighborhood socioeconomic status (SES)?
2 Is San Francisco’s newly installed or total UGS associated with gentrified census
block groups (CBGs) compared with non-gentrified CBGs?
3 Is gentrification more highly correlated with 1) total area of city parks or 2) total
number of city parks?

33

Study Area
I chose The City of San Francisco as my study area because it shares many
similarities with Seattle, WA, making it a good point of comparison with Weems’ (2016)
analysis that also examined trends in gentrification over time in relation to amount of
UGS. San Francisco and Seattle are both major metropolitan cities on the West Coast of
the United States. They both are known for their extensive park systems and have both
made substantial investments in neighborhood parks and other forms of green
infrastructure over the past several decades. Since 1990, San Francisco has installed 23
new municipal parks and a variety of green infrastructure projects (green corridors, green
alleys, green streets, etc.) In addition, San Francisco has experienced significant trends
towards gentrification across large swathes of the city (Maciag, 2015). If there is a
significant association between expansion in available UGS and gentrification, choosing
to examine a city with multiple recently gentrified neighborhoods could make this
connection easier to detect.
As of 2010, San Francisco had a population of 805,235 (US Census Bureau,
2010). White residents made up 48.5% of the total, with Asian residents accounting for
33.3%. Black residents accounted for another 6.1%, and those identifying as Hispanic or
Latino made up 15.1% of the overall total. Since 1990 the Black population has declined
significantly and the Asian population has grown. The City has experienced ballooning
property values, but income has not grown as quickly. From 2000-2010, the median value
of owner-occupied homes in San Francisco nearly doubled, rising from less than
$400,000 to $785,200. Over that same period, median household incomes rose from
about $55,000 to roughly $71,000. San Francisco residents are highly educated: nearly
34

1/3 of the population held a Bachelor’s degree in 2010, and a full 20% of the population
held Graduate or Professional degrees (US Census Bureau, 2010).
Measuring Access to UGS: Subcategories
UGS can be defined in several different ways. Rather than simply including
municipal parks and open spaces in my analysis, I incorporated all the following forms of
UGS as model subcategories. Each was included as components of models measuring
total or new urban greenspace. Subcategories denoted with an asterisk (*) also fall under
the umbrellas of the total or new green infrastructure models.


Parks – I obtained detailed information about San Francisco’s municipal park
system through the City’s Recreation and Parks Dept. I considered subdividing
parks into different use categories but did not end up using this data.



Open Spaces – This dataset includes state parks, nature reserves, public plazas,
and other public open spaces that do not fall into the category of municipal public
parks.



Parklets* – San Francisco Public Works (SFPW) describes a parklet as “… a
sidewalk extension that provides more space and amenities for people using the
street. Usually parklets are installed on parking lanes and use several parking
spaces” (SF Public Works).



Green Stormwater Infrastructure (GSI) Projects* – These include vegetated
smaller-scale greenspace installations such as greenways, green alleys, and green
streets which all have the potential to provide important ecological services (SF
Public Works).

35



Green Roofs* – These are privately installed, but licensed by San Francisco’s
Planning Department which intends “to make living roofs a more viable option
for existing and planned buildings” (San Francisco Planning Department). They
represent another form of green, vegetated infrastructure which can also provides
many important ecological services.



Privately Owned Public Open Spaces (POPOs)* –San Francisco’s Planning
Department describes these as “publicly accessible spaces in forms of plazas,
terraces, atriums, small parks, and even snippets that are provided and maintained
by private developers.”

Measuring Access to UGS: Final Models
Some of the above categories of UGS are typically much smaller than others.
Those UGS installation types that were large enough to be represented by polygons rather
than simple points included Parks and Open Spaces. Those represented by points only
include all subcategories for Green Infrastructure (Green Roofs, GSI, POPOs, and
Parklets). Since these subcategories did not contain data for their area, total UGS area
could not be accurately assessed and compared directly with the number of total UGS
installations. Therefore, categories which assessed total park area and total park number
were included, since this data was all readily available. I also separated newly-installed
(since 1990) green infrastructure and parks, and considered them to be their own
categories, to be assessed separately. The main reason for this was to explore whether
New or Total UGS was more associated with the Gentrification Index variable. This was
quite easy for GSI, parklets, and green roof categories as they were all considered newly
installed since 1990. When considering Parks and POPOs, it was necessary to refer to the
36

date which they were installed. Residents of a given CBG are considered to have access
to all parks or UGS that fall within 400m of that CBG’s perimeter. The final categories
included in my statistical analysis are listed here with their units of measurement:


New Urban Greenspace (UGS) – number of new installations since 1989



New Green Infrastructure (GI) – number of new installations since 1989



New Parks – number of new parks sited after 1987



New Park Area – area of new parks sited after 1987 (hectares)



Total Urban Greenspace (UGS) – number of total installations



Total Green Infrastructure (GI) – number of total installations



Total Parks – number of total parks



Total Park Area – area of all parks (km2)

Because park projects can take a number of years to plan and install, all parks
sited after 1987 where considered new since 1990. For other forms of green infrastructure
(namely, POPOs) which are smaller scale and can be installed more quickly, those sited
after 1989 were considered new.
Measuring Socioeconomic Status
Demographics measuring each CBG’s socioeconomic status (SES) were obtained
by referencing 1990 and 2010 census data for San Francisco County. CBG-level data was
chosen since it represents the smallest scale at which extensive demographic information
is documented for populations, so is most appropriate for parsing out patterns at a smaller
spatial scale. I measured SES and equity in terms of the following statistics (units of
measurement included):
37



Black/African American – number of residents



Asian – number of residents



Hispanic/Latino – number of residents



Youth – number of residents (19 or younger)



Graduate/Professional Degree Holders – number of residents (25+ years old)



Median Household Income – US dollars ($)



Median Home Value – US dollars ($)



Gentrification Index – Index score based on change in income, home value,
and education demographics between 1990 and 2010 (defined below).

Accessibility at the Block Group Scale
My intent in measuring accessibility was to analyze the total number and total
area of walkable parks within ½ mile or 800m from the average CBG resident. Previous
research suggests people are most likely to walk to a park in their neighborhood if it is
located within 400m of them, but they are still inclined to consider a park to be within
walking distance if it is up to 800m away (Boone et al., 2009). With this in mind, I placed
a 400m buffer around each CBG polygon using ESRI ArcMap 10.3’s Buffer tool. Based
on visual inspection and use of ArcMap’s Measurement tool, I concluded that most CBGs
in San Francisco are represented by polygons less than 400 meters in length on any given
side. Therefore, for the average San Francisco CBG, a resident should have access to any
park within 400m of the CBG periphery since they would generally be walking a total
distance of 800m or less.
38

While there are more nuanced methods for determining access to UGS in relation
to CBG polygons (such as the container approach or network analysis; see Oh & Jeong
(2007), Zhou & Kim (2013), or Kabisch & Haase (2013)), I decided that simple buffering
would be sufficient, since several other similar studies have made use of the technique
successfully (Weems, 2016: Wolch, Wilson, & Fehrenbach, 2002; Boone et al., 2009).
After buffering CBGs, my next step was to intersect the buffers with polygons or
points representing instances of UGS. This was accomplished using the Tabulate
Intersection (Analysis) tool in ArcMap. These procedures produced a single GIS layer
and associated data table that contained the following information appended to each 2010
CBG:


Demographic data for the 8 parameters being assessed in these models.



Number of overlapping UGS installation of each type and subtype including
overall total numbers.



Total acreage of intersecting Park Area (in addition to number of Total Parks).
A separate data table and layer for 1990 CBGs was also created to provide a

comparison point to analyze temporal trends related to gentrification. To assess the
number of New UGS installations, most categories of Green Infrastructure were all
considered new since 1990. All Open Spaces were assumed to be older than 1990 due to
a lack of reliable data about when they were actually installed. Date of installation for
Parks and POPOs was available, so for these two categories the number of specific new
installations (since 1990) was appended to each 2010 CBG polygon.

39

Creating a Gentrification Index
The portion of my analysis measuring changes in park access compared with
trends toward or away from gentrification is based in part on Weems’ (2016) analysis of
Seattle, but aims to improve upon this study’s methodology. Weems utilized a simplified
gentrification index based on median household income, median home value, and percent
of population with a bachelor’s degree or higher, where a CBG was considered gentrified
if 2 out of these 3 statistics moved from below average to above average compared with
the citywide averages. Using this methodology results in discrete categories (gentrified,
not gentrified, or reverse-gentrified) rather than a continuous scale.
My gentrification index uses the same 3 statistics but converts gentrification to a
continuous variable which may help to detect shifts in neighborhood SES on a finer scale.
First, the three relevant parameters were expressed in terms of median absolute
deviations since this is considered a more robust measurement of variability. Relative
variability was particularly important to measure here since the resultant index scale
represented a measurement of relative change.
I used detrended correspondence analysis (DCA) to create a demographic index
for both 1990 and 2010 data sets. Ordination groups similar locations (in this case,
CBGs) closer to each other on unit-less axes that account for multiple environmental
variables simultaneously. Those CBGs most similar to one another in terms of all 3
parameters are grouped nearer to each other, while those with more differences are
furthest away from each other. After running separate DCAs on 1990 and 2010 datasets,
it was important to confirm that both relative scores were “oriented” in the same

40

direction. In this case, higher income, home value and more-educated households
received a higher index score for both 1990 and 2010 data.
There are many different forms of ordination. DCA was chosen because it adjusts
for some of the irregularities that might be experienced when using traditional
correspondence analysis, correcting for edge effects, and flattening the normally
horseshoe shaped curve to be more linear. (Ter Braak & Prentice, 2004) Reducing edge
effects was an important consideration for my analysis because the areas of greatest
interest to me are those that fall on the extreme ends of this “gentrification” scale. DCA
ordination was performed in R (R Core Team, 2016) using the Vegan package (Oksanen
et al., 2016) and the “decorana” tool.
Unfortunately, there is not a one to one correspondence between 1990 CBGs and
2010 CBGs in San Francisco. This is because boundaries were redrawn in the intervening
years based on changes in population. Using the Intersect tool in ESRI’s ArcMap 10.3, I
identified the 1990 CBG polygon that overlapped the most with each 2010 polygon and
considered them to correspond with one another. The clear majority (89%) of 2010 CBGs
overlapped with greater than ½ the area of a 1990 CBG, but only about 53% of 2010
CBGs overlapped with greater than 2/3 the area of a specific 1990 CBG. Except for two
extreme outliers, the range of percentage overlap fell between 31% and 100%. The
median amount of overlap was 68% with a standard deviation of 13%.
UGS and demographic data about the 1990 dataset was joined in ArcMap with
info relating to the corresponding 2010 dataset based on CBG overlap. Demographic
index scores based on DCA analysis for both years were among the data fields joined

41

together in this process. Afterward, 1990 DCA index scores were subtracted from 2010
DCA index scores, resulting in a unit-less relative index score which measures the change
in demographic index between 1990 and 2010. Because this score simultaneously takes
into account changes in household income, home value, and education, I will consider it
to be a simplified measurement of each CBGs’ amount of gentrification.
Data Sources and Data Preparation
I downloaded the following TIGER/Line shapefiles from the US Census Bureau
(1990, 2010)


Census block group boundaries for 1990



Census block group boundaries for 2010

GIS Boundary Shapefiles were at Block Group level for years 1990 and 2010 (Based on
2000 TIGER/Line + and 2010 TIGER/Line + respectively). Each shapefile contained a
join ID and GEOID that allowed it to be joined with data tables from the National
Historic GIS Database (Minnesota Population Center, 2016) curated by the University of
Minnesota. Those data tables contained specific demographic information that had been
downloaded in various formats and related to both Census 1990 and Census 2010 for
CBGs in California. These formats and the years of data collection which they are
associated with are as follows:


1990 Census: STF 1 – 100% Data (by census block group for California) for
o Race (Total population)
o Age
o Median Value (Owner-occupied housing units)
42



1990 Census: STF 3 – Sample-Based Data (by census block group for
California) for
o Educational Attainment (Population 25 years and over)
o Median Household Income – past year i.e. 1989 (Households)



2010 American Community Survey: 5-Year Data [2006-2010, Block Groups
& Larger Areas] for
o Race (Total population)
o Age
o Median Value (Owner-occupied housing units)
o Educational Attainment (Population 25 years and over)
o Median Household Income – past 12 months (Households)

It should be noted that although information from the 1990 Census contains some
demographic data based on “100% data” and some that is sample-based, and although
2010 data is based on the 5-year long 2010 American Community Survey, in each case
the most complete data set available for that year and metric was used. Even though
methodologies may differ slightly, due to the sampling methods used by census
demographers, I will assume that the sample-based 1990 metrics are directly comparable
with their more exhaustively complete 2010 counterparts.
All UGS data were downloaded directly from San Francisco Open Data, but as
mentioned, were originally curated by a few different public agencies:


San Francisco Planning Dept. (POPOs and green roofs)



San Francisco Public Works (GSIs and parklets)

43



San Francisco Recreation and Parks Dept. (Open Spaces and parks)

All 6 forms of UGS data were downloaded as GIS shapefiles except for the parklets
dataset which arrived as a table and was converted to a GIS feature class using embedded
GPS data points. To obtain a UGS dataset containing only the parks that existed in 1990,
I requested a spreadsheet from San Francisco Recreation and Parks Dept. containing the
age of all city parks. Those sited before 1987 were deleted from a copy of the parks layer
resulting in a parks dataset containing only newly-installed parks that didn’t exist prior to
1988-1990. This allowed me to assess the placement of New Parks in relation to
gentrification. Similarly, a version of the POPOs dataset containing only new POPOs was
created to assess only the installations that were new since 1990 in the New UGS/GI
categories.
Statistical Analysis
Relationships between different categories of UGS and demographic data were
analyzed using multiple linear regression (MLR) with ordinary least squares. These tests
were conducted after importing all data into JMP 13.0. Data was assessed in terms of
CBG access to: Total UGS, New UGS, Total GI, New GI, Total Parks, New Parks, Total
Park Area, and New Park Area. Each of these 8 dependent variables were considered in a
separate MLR equation incorporating all 8 demographic statistics as independent
variables.

44

Results
Overview
Separate multiple linear regression equations were produced for each of the 4
metrics measuring access to new parks and greenspace (Table 1). Adjusted R2 values
ranged from 0.10 and 0.19 for New Parks and Area of New Parks models to 0.20 and
0.22 for New Green Infrastructure and New Urban Greenspace models, respectively.
Output for these models are reported in terms of x1,000 residents for variables measuring
number of residents per census block group (CBG) and x$100,000 for variables
measuring income and home values.
Metrics measuring access to total parks and greenspace were also analyzed using
four separate multiple linear regression models (Table 2). Adjusted R2 values were lower
for these models, ranging from only 0.04 and 0.08 for Area of Total Parks and Total
Green Infrastructure, up to only 0.10 and 0.12 for Total Parks and Total Urban
Greenspace, respectively. The shift in number of residents and in dollar values/index
scores that would be associated with each additional park, greenspace, or hectare of park
area have also been reported (Tables 3, 4, 5, and 6) to help better communicate the
magnitude of effect sizes being discussed.

45

Table 1: Regression Analyses for New Parks and Greenspaces
Each column in the following chart represents a separate multiple linear regression equation, with the
column heading representing the dependent variable.

New Urban
Greenspaces

New Green
Infrastructures

New Parks

Area (Km2) of
New Parks

3.2***
[1]

2.5**
[1]

0.78***
[0.2]

0.79***
[0.1]

1.0**
[0.4]

0.84*
[0.4]

0.15*
[0.08]

-0.12**
[.06]

1.7***
[0.7]

1.5**
[0.0007]

0.27**
[0.1]

-0.33***
[.08]

-7.0***
[1]

-6.3***
[1]

-0.77***
[0.2]

0.59***
[0.2]

3.3***
[1]

2.8***
[0.9]

0.46***
[0.2]

-.073
[0.1]

1.9***
[0. 5]

1.7***
[0.5]

-0.27***
[0.08]

0.19
[0.6]

(x$100,000)

0. 30***
[0.1]

0.32***
[0.1]

0.18
[0.02]

0.016
[0.01]

Gentrification Index
Score

0.36***
[.05]

0.36***
[.05]

-0.0022
[.008]

0.0042
[.006]

Intercept

3.3***
[0.9]

2.7***
[0.9]

0.61***
[0.2]

0.23**
[0.1]

Parameter
Black/Afr. American
(x1,000 residents)

Asian
(x1,000 residents)

Latino/Hispanic
(x1,000 residents)

Youth
(x1,000 residents)

Postgraduates
(x1000 residents)

Median Household
Income
(x$100,000)

Median Home Value

Standard errors bracketed. Asterisks indicate level of significance of the coefficient:
* <0.1, **<0.05, ***<0.01

New Urban Greenspace: F(8,528)=19.6, p<0.001, R2=0.22
New Green Infrastructure: F(8,528)=17.9, p<0.001, R2=0.20
New Parks: F(8,528)=8.3, p<0.001, R2=0.11
Area of New Parks: F(8,528)=16.6, p<0.001, R2=0.19

46

Table 2: Regression Analyses for Total Parks and Greenspaces:
Each column in the following chart represents a separate multiple linear regression equation, with the
column heading representing the dependent variable.

Total Urban
Greenspaces

Total Green
Infrastructures

Total Parks

Area (Km2) of
Total Parks

14***

2.5

2.1**

-0.086

[4]

[2]

[0.9]

[.06]

3.6*
[2]

1.8**
[0.8]

-0.76*
[0.4]

-0.023
[0.03]

-1
[3]

-0.15
[1]

0.48
[0.6]

-0.065*
[0.04]

-17***
[5]

-5.5***
[2]

-1.2
[1]

0.22***
[0.07]

12***
[4]

0.61
[2]

4.1***
[0.8]

0.13**
[0.05]

-4. 7**
[2]

-1.5*
[0.8]

-1.4***
[0.4]

-0.029
[0.03]

(x$100,000)

0.78*
[0.5]

0.63***
[0.2]

.012
[0.09]

-0.0068
[.006]

Gentrification Index
Score

1.1***
[0.2]

0.38***
[.08]

-0.033
[0.04]

-0.0068**
[0.003]

Intercept

14***
[4]

-0.45
[2]

4.3***
[0.8]

0.091*
[0.06]

Parameter
Black/Afr. American
(x1,000 residents)

Asian
(x1,000 residents)

Latino/Hispanic
(x1,000 residents)

Youth
(x1,000 residents)

Postgraduates
(x1,000 residents)

Med. Household
Income
(x$100,000)

Med. Home Value

Standard errors bracketed. Asterisks indicate level of significance of the coefficient:
* <0.1, **<0.05, ***<0.01

Total Urban Greenspace: F(8,528)= 9.2, p<0.001, R2=0.12
Total Green Infrastructure: F(8,528)=5.6, p<0.001, R2=0.06
Total Parks: F(8,528)=7.5, p<0.001, R2=0.09
Area of Total Parks: F(8,528)=2.9, p<0.004, R2=0.03

47

Newly Installed Greenspace and Green Infrastructure
The models assessing associations for New Urban Greenspace (UGS) and New
Green Infrastructure between 1990 and 2016 had the highest adjusted R2 values (Table
1). Each of the racial-ethnic groups included in these analyses (Black, Asian, and Latino
residents) was positively correlated with New UGS, New Green Infrastructure, and (to a
lesser extent) New Parks. Postgraduate resident populations (those who had obtained
graduate or professional degrees beyond a Bachelor’s) were also positively associated
with these three response variables. The only negative associations detected in these three
models was for number of youth residents (younger than age 19). Using Area of New
Parks as a response variable, most coefficients had the opposite sign compared with
corresponding coefficients for the other three models.

Table 3: New Greenspace & Number of Residents
Numbers indicate quantity of residents associated with each additional new installation/hectare.

Black

Asian

Latino

Youth

Postgrads

New Urban
Greenspace

310

1000

580

-140

300

New Green
Infrastructure

410

NS

690

-160

350

New Park

1290

NS

>2000

-1300

>2000

Ha of New Park
Area

13

-81

-30

17

NS

NS = not significant (p >0.05)

Estimates for the number of additional residents required in the model for the
average CBG to be associated with one additional greenspace or green infrastructure
installation ranged from about 300 for postgraduate and Black residents up to 1,000 for

48

Asian residents, or down to -140 for youth residents (Table 3). Considering that the mean
population for CBGs in San Francisco is ~1,363 residents, the 310 additional Black
residents or 140 fewer youth residents associated with one additional installation of
greenspace in the model represent +23% or -10% shifts in the total population of a CBG.
Variations between CBGs of a similar magnitude in the percentage of Black and Asian
residents can be observed by referring to maps of the distribution of current resident
populations (Appendix; Figures 4 and 7). The magnitude of effect for the number of
Asian residents is much smaller, considering that 1,000 additional residents are
associated with one additional New UGS installation. There are very few CBGs that
would even contain that many Asian residents (Appendix; Figure 5). The magnitude of
effect for number of Latino residents falls between that of Black and Asian residents in
both models. Though a quite large shift in this population would be needed for a CBG to
be associated with an additional installation of new UGS/GI, variations of this magnitude
can be observed between CBGs by referring to a map of the distribution of current Latino
resident populations (Appendix; Figure 6).

49

Table 4: New Greenspace & Dollar/Index Values
Numbers indicate index score or value ($) associated with each
additional new installation/hectare.

Med. Income
($)

Med. Home
Value ($)

Gentrification
Index*

New Urban
Greenspace

-52k

330k

2.9

New Green
Infrastructure

-61k

310k

2.8

New Park

-367k

NS

NS

Ha of New
Park Area

NS

NS

NS

*35 pt. relative scale, NS = not significant (p>0.05)
Med. Home Value is for owner-occupied homes.
Med. Income is per household

The shifts in median income associated with one additional installation of
UGS/GI were negative whereas the shifts in home value were positive (Table 4). Shifts in
gentrification index scores associated with additional greenspace installations were also
clearly positive. These shifts in index score values represent relative demographic
changes along a 35-point relative scale, created using DCA ordination, which
incorporated neighborhoods’ relative changes in home value, household income, and
education level between 1990 and 2010.
Changes associated with one installation of New UGS (in median income,
(-$52,000) and median home value (+$330,000)) represent roughly -73% and +42%
shifts in comparison to San Francisco’s overall median values, reflecting low effect sizes.
Some of the most statistically significant and largest magnitude changes associated with
additional new greenspaces are those associated with the gentrification index variable.

50

The 2.8- and 2.9-point shifts associated with New UGS and New Green Infrastructure,
respectively, represent only about +8% shifts along the 35-point relative scale.
Model results for New Urban Greenspace and New Green Infrastructure were
very similar. This was expected to some extent since New Green Infrastructure represents
a subset of New Urban Greenspace, and there were several dozen new green
infrastructure installations since 1990, whereas there were only about two dozen new city
parks during that same period (Appendix; Figure 2). Levels of statistical significance for
some variables differed somewhat between the two models, but for the most part
estimates and standard errors tracked quite closely with one another.
Newly Installed Parks
Models assessing the number and area of new parks produced fewer statistically
significant results, and associated R2 values were slightly lower. In the case of the New
Parks model, all variables’ correlations were found to move in the same directions as they
did with the New UGS and Green Infrastructure models, except for the variable
representing median household income. On the other hand, 3 out of 4 of the Area of New
Parks model’s statistically significant correlations were found to move in the opposite
direction of the New Parks model’s relevant correlations. This implies there is a negative
association between Asian as well as Latino resident populations and area of new park
space. There is also a positive correlation between area of new park space and youth
resident populations.
Only a small increase in the number of youth residents (17) or decrease in Latino
(30) or Asian (81) residents was associated with one additional hectare of park space in
51

the model (Table 3). In the case of Latino and Asian residents, these represent only -2%
and -6% shifts in comparison to the overall population of the average CBG in San
Francisco. The positive association identified in Black residents’ relationship to Area of
New Parks is even higher magnitude, representing only about a +1% shift at only 13
residents. For reference, the average newly installed park between 1990 and 2016 was
only about 0.59 hectares, so measuring results in terms of hectares here makes more
sense than leaving the data in terms of square kilometers.
Conversely, the magnitude of effect sizes for parameters in the New Parks model
was extremely low. The number of additional Latino residents and advanced degree
holders needed for a CBG to be associated with one additional park, exceeds the number
of total residents that are likely to be found in one CBG. So, these higher numbers are
reported simply as >2000 (Table 3). These values are so large, that the magnitude of
effect size for their variables to be associated with one additional new park is almost
unrealistically small. The magnitude of effect size for the positive association between
Black residents and new parks, and the negative association between youth residents and
new parks, are a bit higher (both ~1300). But these variables’ association with one
additional new park, would require a nearly 100% demographic shift in the population.
So, in practice, these correlations also display unrealistically low magnitude effect sizes.
None of the economic variables produced statistically significant results with
regard to New Parks and Area of New Parks, with one exception. Median Income was
negatively associated with New Parks, as it was with other forms of new greenspace. But,
in this case the magnitude of effect size was very low. A reduction in median income of
$367,000 would be required for a CBG to be associated with one additional new park.
52

That dollar amount is more than five times as large as the overall median household
income for San Francisco. Though considering the amount of wealth inequality in the
City, a shift of this magnitude is perhaps within the realm of possibility.
Total Greenspace and Green Infrastructure
Models using Total Urban Greenspace and Total Green Infrastructure as response
variables had lower adjusted R2 values compared to the models measuring associations
with new parks and urban greenspace (Table 2). For all variables in these two models
except for one, correlations moved in the same direction as they did for their
corresponding “new” models (i.e. New UGS and New Green Infrastructure). The variable
for median income was the one exception to this rule. It was found to be negatively
associated with both Total UGS and Total Green Infrastructure. Overall, these two
“Total” models differed from each other more so than their corresponding “New” models
did. This was expected considering there are relatively more older parks and open spaces
to balance out the number of Total Green Infrastructure installations compared with the
only 23 new parks and 0 new open spaces which allowed the New UGS model to be
overwhelmed by the much more prolific and easy-to-install New Green Infrastructure
installations (Appendix; Figures 2 and 3).

53

Table 5: Total Greenspace & Number of Residents
Numbers indicate quantity of residents associated with each additional total installation/hectare.

Black

Asian

Latino

Youth

Postgrads

Total Urban
Greenspace

71

NS

NS

-60

85

Total Green
Infrastructure

N/S

550

NS

-180

NS

Total Parks

470

NS

NS

NS

240

Ha of Total
Park Area

NS

NS

NS

46

77

NS = not significant (p>0.05)

The Total UGS model’s magnitude of effect size for Black, youth, and postgrad
residents is clearly larger than the effect sizes observed in any of the previous models
(Table 5), except for the Area of New Parks model that assessed number of associated
residents per hectare rather than per installation. The 71 additional Black residents, 85
additional advanced degree holders, and 60 fewer youth residents associated with one
additional urban greenspace all correspond with roughly +/- 5-6% variations in the total
population of the average CBG. For reference, variations between CBGs of percentages
of Black, youth, and postgraduate resident populations can be observed by referring to
maps of current resident distributions (Appendix; Figures 4, 7, and 8). When examining
the Total Green Infrastructure model, the number of Asian (550) and youth (180)
residents in a CBG associated in the model with an additional installation, is more in line
with what we might expect from previous models. The shift in youth population
particularly though, still represents a quite large magnitude effect, and corresponds with
about a 13% shift in the average CBG’s overall population.
54

Table 6: Total Greenspace & Dollar/Index Values
Numbers indicate index score/ value ($) shift associated with each
additional installation/hectare

Med. Income
($)

Med. Home
Value ($)

Gentrification
Index*

Total Urban
Greenspace

-21k

NS

0.9

Total Green
Infrastructure

NS

160k

2.7

Total Parks

-71k

NS

NS

Ha of Total
Park Area

NS

NS

-1.5

*35 pt. relative scale, NS = not significant (p>0.05)
Med. Home Value is for owner-occupied homes.
Med. Income is per household

The negative shift in median household income associated with one additional
installation of Total UGS (Table 6) represents the largest magnitude of effect observed
for this variable in any of the models. This $21,000 decrease is equivalent with a -29%
shift in the average CBG’s median household income. Likewise, the largest magnitude
effect associated with the home value variable can be observed in the Total Green
Infrastructure model. A $160,000 shift in home value is equivalent to only about a 20%
shift in the median value for the average CBG. Both of these variations are of a similar
magnitude as observed variations between CBGs among current resident populations
(Appendix; Figures 9 and 10). Finally, the two largest magnitude of effects observed so
far for the gentrification index variable are associated with Total UGS (0.9) and Total
Green Infrastructure (2.7). The shift in gentrification index score of 0.9 represents a < 3%
shift along the 35-point scale. Variation is again similar in magnitude to current variation
in index scores observed between different San Francisco CBGs (Appendix; Figure 11).

55

Total Parks
Models measuring associations with Total Parks and Area of Total Parks
produced the fewest statistically significant results and their R2 values were among the
lowest. Significant estimates for the Total Parks model all moved in the same direction as
estimates for the New Parks model, except for the variable representing number of Asian
residents, which was positively associated with New Parks, but negatively associated
with Total Parks. When considering only statistically significant results, and comparing
the Area of Total Parks model with the Area of New Parks model, none of the variables’
correlations have switched directions (Table 2).
The magnitude of effect sizes observed for the Total Parks model are in the mid
ranges compared with those models already discussed previously (Tables 5 and 6). Gains
associated with one additional total park, of 470 Black residents and 240 advanced degree
holders represent roughly +34% and +18% shifts in the overall population of the average
CBG. Gains associated with one additional hectare of total park area are of slightly lower
magnitude effect size compared with those associated with an additional hectare of new
park area, but are still very large in magnitude. Those 46 additional youth residents and
77 additional postgraduate residents represent only 3-6% shifts in the overall CBG
populations.
Referring to the economic variables associated with these two models (Table 6),
one additional park is likely to be associated with a $71,000 decrease in median income.
This represents about a 100% shift compared with the average CBG’s median income,
but again considering wealth inequality, this may not be as inordinately large of a shift as

56

it sounds. Many CBGs have median incomes of $35,000 or less, while a comparable
number have median incomes of $140,000 or significantly more (Appendix; Figure 9).
And finally, in this model the gentrification index variable is, for the first time, negatively
associated with a UGS metric; one additional hectare of total park area. The magnitude of
effect size is still quite large for the gentrification index variable even though its direction
of association has reversed compared with other models included in the analysis.

Discussion
Spatial Patterns Influencing Results
This analysis explored a series of relationships involving urban greenspace,
neighborhood demographics, and gentrification, examining their connections in detail
using San Francisco as a model. Comparing the location of San Francisco neighborhoods
(Appendix; Figure 12) with the location of parks and greenspace installations (Appendix;
Figures 2 and 3), and the spatial distribution of demographic statistics in the City
(Appendix; Figures 4 through 11), may provide additional insights to readers familiar
with these neighborhoods.
Some of the patterns that arose from this analysis’s statistical tests may be
disproportionately related to the location of a few very large urban greenspace
installations like Golden Gate Park. Other patterns may relate to the extremely
disproportionate adoption of smaller-scale green infrastructure installations in the
Northeastern corner of the City near the downtown area.

57

On the other hand, many of these observed patterns, especially those persisting
across most of the 8 models explored in this analysis, might relate more to underlying
economic trends or structural inequities. The techniques used here were not sophisticated
enough to parse out these intersecting and potentially cross-influencing causes. Truly
assessing the potential cause and effect relationships between park and greenspace
placement decisions, neighborhood demographic shifts, and gentrification, will require
much more study, and perhaps the development of innovative spatial analysis and other
assessment techniques.
Equity Analyses
Results show that when disproportionate urban greenspace (UGS) access occurs
among racial-ethnic groups in San Francisco, it tends to favor rather than disfavor the
minority groups included in the analysis. For example, a positive correlation was found
between Black/African American residents and every metric for UGS except one. The
one model where greenspace was not found to be positively correlated with Black
resident populations (Area of Total Parks) was not found to be statistically significant
(Table 2). Based on this, it appears that the African American residents in San Francisco
generally do not suffer from lack of access to parks and urban greenspace, but in fact
enjoy greater than average access to them. This may sound surprising. But, results like
these are not particularly unusual, despite the opposite pattern’s relative frequency. For
example, analyses of Bloomington and Rockford, Illinois, produced similarly favorable
results for African American residents (Zhou & Kim, 2013).

58

Asian and Latino residents were found to be either positively correlated or to
show no association in either direction with the metrics for UGS that I’ve assessed.
Although, in the case of both groups, there was one exception to this general rule—
Asian and Latino resident were both found to be negatively correlated with Area of New
Parks. In the case of Latino residents, it should also be noted that all four metrics that
measured “total” amounts of parks and urban greenspace showed no association in either
direction. And in the case of Asian residents, only one of the four “total” metrics (Total
Green Infrastructure) was found to be positively correlated while the others showed no
association in either direction. So, based on this, it will be important that city planners
ensure that future park sites be selected with these groups in mind. If not, the current
trend of new parks being less associated with Asian and Latino occupied neighborhoods
could lead to future greenspace inequities. Currently it appears that no such inequities yet
exist for Asian and Latino residents of San Francisco.
Some greenspace inequities were discovered when examining other
socioeconomic groups. Census block groups with more children (residents under the age
of 19) were found to disproportionately lack access to Total UGS as well as Total Green
Infrastructure. Additionally, this group was found to be negatively correlated with New
Parks. However, exceptions to this overall pattern were observed for both New and Total
Park Area categories which both displayed positive associations. Overall, these results
show that families with children are a group that suffers from disproportionately low
access to UGS in San Francisco. City planners may want to keep these people in mind
when siting the location of future parks, especially considering the mental and physical
health benefits that have been associated with children’s exposure to greener
59

environments (Lovasi et al., 2008; Taylor, Kuo, & Sullivan, 2001; Bell, Wilson & Lui,
2008).
Residents with fewer years of formal education were found to experience lower
levels of access to UGS in San Francisco compared with those who are more highly
educated. Residents with graduate or professional degrees were found to be positively
associated with every metric of UGS, with just two exceptions; Total Green
Infrastructure and New Park Area. But those two categories displayed no significant
association in either direction. Based on these findings, UGS installations may
disproportionately favor neighborhoods with more highly educated residents.
Those with higher household incomes were not found to be at an advantage in
their access to UGS. Rather, lower income neighborhoods instead appear to have a
significant advantage. Median Household Income was negatively correlated with three
out of four “new” metrics for greenspace, as well as with two of the four “total” metrics.
None of the 8 greenspace metrics were positively associated with household income. This
suggests that if San Francisco has policies in place to incorporate household income into
their assessments of where parks or greenspace should be sited, these policies appear to
be effective in ensuring that lower-income residents are adequately served.
Perhaps predictably (considering the supposed connection between property
values and parks that is at the root of green gentrification), higher Median Home Values
were positively correlated with several metrics for greenspace, including New UGS and
Green Infrastructure as well as Total UGS and Green Infrastructure. The apparent
disconnect between high-income residents’ and high home value residents’ access to

60

UGS observed here may stem from the fact that the Median Home Value metric only
incorporates home values of owner-occupied homes. Many high-income renters in a
particular CBG might cause home value and household income associations to diverge
widely from each other for that CBG. I would hypothesize that this explains the
differences observed between Median Home Value and Median Household Income
results. Based on these findings, residents owning lower value homes are likely to
experience less access to parks and UGS compared with residents owning higher value
homes.
Overall, the results from this equity analysis portion of my study support previous
findings from the literature. Past studies have produced a wide range of varying results
depending on geography, groups included in the study, and measurement techniques
used. Compared with many other cities and regions, San Francisco appears to distribute
its greenspace quite equitably between different racial-ethnic groups. Examples from the
literature of cities with less equitable racial-ethnic/park placement relationships include
Baltimore, MD (Boone et al., 2009) and Berlin, Germany (Kabisch & Haase, 2013). But
there are plenty of other examples of findings like the ones observed in this study where
few differences are detected between these groups, such as in Melbourne, Australia
(Timperio et al., 2007).
Some notable examples of social inequities include findings from Wen et al.
(2013) and Saporito and Casey (2015), both of whom conducted nationwide analyses that
parsed some of the differences observed between urban vs. suburban and segregated vs.
less segregated cities. Others have identified inequities regarding residents’ age groups
(Ngom et al., 2016; Cutts et al., 2009). As was already mentioned, the most surprising
61

result regarding any of the above variables remains the discrepancy between UGS’s
relationship with median household income vs. its relationship with median home value.
Again, this is probably the result of higher-income professionals that rent rather than own
their homes being included in a CBG’s measurement for Median Household Income but
not its measurement for Median Home Value (of owner occupied homes). There are
presumably many tech industry workers and other young white-collar professionals living
in San Francisco that fall into this category. The US Census Bureau’s 2015 American
Community Survey, shows that in San Francisco there were almost triple the number of
new renters compared with new homeowners making $150,000 or more between 2014
and 2015. And the City’s share of high-income renters is higher than any comparable city
in the US with nearly 25 percent earning $150,000 or more (Sisson, 2016; US Census
Bureau). All in all, none of the results found in this portion of my analysis were out of
line with previous studies.
Gentrification and the Impact of Park Size
The second part of my analysis pertains to the relationship between urban
greenspace and gentrification. I asked two related questions; are newly gentrified
neighborhoods more strongly associated with parks and urban greenspace compared with
less gentrified neighborhoods? And if a positive association exists between gentrification
and UGS, are gentrified neighborhoods more associated with amount of new greenspace
area or the total number of greenspace installations? The purpose of asking this follow-up
question is to assess whether support exists for an element of the “just green enough”
theory of greenspace investment which states that gentrification may be less associated

62

with smaller and more discrete instances of greenspace and more strongly associated with
large civic projects.
The results of this portion of my analysis show that more highly gentrified CBGs
are in fact strongly correlated with New as well as Total UGS. They are also positively
correlated with New and Total Green Infrastructure. These findings appear to support the
hypothesis that amount of nearby urban greenspace is directly associated with the amount
of gentrification a neighborhood experiences. Of course, it is important to remember that
a causal relationship cannot be established in either direction based on this data alone.
It is also important to note that one negative association was discovered between
gentrification and Area of Total Parks. It is interesting that metrics for Park Area often
display the opposite direction of correlation compared with other UGS metrics. This
appears to be true across several different demographic variables included in these
models. In addition to Gentrification Index, this pattern was also observed for variables
relating to the numbers of Latino, Asian, and Youth residents. Overall, this section’s
results appear to support the hypothesis that there is indeed a direct relationship between
neighborhood greenspace accessibility and the gentrification process. Identifying this
association provides preliminary evidence for the theory that increased investments in
greenspace may be one of the factors contributing to neighborhood gentrification in San
Francisco.
My follow-up question in this section addresses one element of the “just green
enough” theory of greenspace investment. To do this it was necessary to directly compare
amount vs. area of urban greenspace. Unfortunately, I was unable to obtain that

63

information for every form of UGS. Data directly comparing area vs. total number of
greenspaces was only available for municipal parks specifically. So, my assessment of
this question is based only on park-specific data. When comparing gentrification’s
relationships to these two variables, gentrified census block groups were found to be
negatively associated with Total Park Area, but were not associated in either direction
with Total Parks. These results fail to support the hypothesis that gentrification is more
strongly associated with park area than with total number of parks. Apparently the “just
green enough” theory does not apply here in the way that it was predicted to.
One other data point to consider in discussing the results for this section is the
positive correlations observed between green infrastructure and gentrified census block
groups. Because green infrastructure installations tend to be numerous, smaller-scale, and
widely distributed throughout some neighborhoods, and since they are associated with
gentrification whereas larger scale parks are not, this provides further supporting
evidence that the “just green enough” hypothesis being tested is not applicable in this
case.
As noted in the literature review section of this thesis, “just green enough”
strategies can be implemented in several different ways, some of which address
procedural rather than distributional justice (Jennings et al., 2012). My results support the
idea that simply distributing smaller scale green infrastructure installations more evenly
may be an insufficient strategy by itself to avoid neighborhoods from being affected by
displacement and other unwanted impacts of gentrification. This points toward the need
to implement “just green enough” strategies more creatively. Activists and community
members need to be heavily involved at every stage of the planning process, instead of
64

simply hoping that procedural or policy changes will solve the problem of lower income
residents being displaced from the neighborhoods where they live.
Few previous studies have specifically addressed the impacts of gentrification
through spatial analysis. The one that did, used a discrete yes/no system for identifying
gentrification rather than implementing a continuous scale based on ordination (Weems,
2016). That study, like this thesis, identified a connection between gentrification and
urban greenspace. The connection identified however, related to greenspace investment.
When simply looking at the placement of new UGS, Weems (2016) failed to identify the
same strong positive correlation in Seattle as was detected in San Francisco. So, the
findings of my analysis are certainly in line with this previous study, but have perhaps
identified a more significant association. This could be related to actual differences
between the two cities being analyzed. Or it could be simply due to my gentrification
index being significantly more granular and sensitive to relative change. In either case,
my findings support the idea that “green gentrification” may be one key factor in the
observed demographic changes of San Francisco neighborhoods.
Strengths, Limitations, and Future Studies
One limitation of this analysis was its simplified methodology for measuring UGS
access. It would be interesting to see how the study’s results might differ if network
analysis or other more advanced techniques were employed. Network analysis measures
actual walkable paths between parks and residential streets using Google Maps API,
rather than employing a simple buffer (Zhang & Wang, 2006).

65

Likely no major changes in the results would be observed, but particularly if city planners
have been siting parks and greenspaces based on some form of spatial equity analysis that
did incorporate a more advanced technique, then some subtle difference would perhaps
be detected. Considering that more advanced spatial analysis techniques are all relatively
new, having just been developed within the past ten years or so, it seems unlikely that this
would be the case.
The section of my analysis exploring the impact of park size may have been
strengthened if data were available for Total UGS Area. Results obtained would have
been more complete if all forms of UGS were directly compared for their area vs. total
number rather than relying only on park-specific data for this purpose. Datasets obtained
from San Francisco Open Data might have also been strengthened in other ways. For
example, The Open Spaces category, a subset of Total UGS, did not contain information
about when open spaces were sited or “installed”, so they were all assumed to date to
before 1990. If any of these open spaces were newly designated (perhaps because of an
old lot or brownfield being restored), they would not have been detected or included in
the New UGS category.
Another point of caution to keep in mind when interpreting these multiple linear
regression results is that their adjusted R2 values were rather low, particularly in the cases
of the Park Area and Total Green Infrastructure models (less than 0.1). These low values
imply there is significant variation occurring in the placements of parks and greenspace
that is not explained by the demographic variables included in these regression models.
Perhaps, prevalence of urban trees, proximity to water, or subtler cultural or geographical
differences between neighborhoods also play significant roles. Adjusted R2 values for the
66

remaining models, while still on the lower side (between 0.1 and 0.3) are in line with
those obtained by many past spatial analyses that employed similar methodologies
(Abercrombie et al., 2008; Zhou & Kim, 2013).
An additional limitation of this analysis was that entire new categories of green
infrastructure were newly implemented between 1990 and 2010, so rather than measuring
the expansion of existing forms of greenspace into new areas, new forms of greenspace
were being implemented and invested in for the first time in many cases. Because of the
innovative nature of some of these new forms of installations, and the ability of private
entities to implement public greenspaces with the City’s approval, green roofs, parklets,
and other forms of green infrastructure seem to have been disproportionately targeted to
wealthier commercial districts such as the downtown area. As green roofs and roadside
parklets become more established greenspace options, perhaps they will begin to crop up
more evenly throughout the City.
Confidence in this analysis’s results might be enhanced by the fact that xvariables in separate multivariate equations tended to behave similarly in relation to each
other. For example, as noted earlier, Total/New Park Area metrics often exhibited the
opposite direction of associations compared with trends observed for all other UGS
metrics for a given variable. But the fact that this pattern was observed somewhat
consistently (for 4 out of the 8 demographic variables analyzed) supports the idea that
there are real differential relationships between Total UGS/Green Infrastructure and Park
Area rather than random noise or variation. Another example of variables aligning in
similar patterns can be observed in the relationship between UGS and Green
Infrastructure metrics. Their associations tended to nearly always track with one another
67

in the same direction. Though in this case, that relationship is to be somewhat expected
considering Green Infrastructure represents one of the subsets of Total UGS.
Another reason for confidence in the results of this analysis is the diversity in
types of greenspace that were incorporated compared with some past studies that only
looked at parks or open spaces. San Francisco’s Open Data website contains more indepth information about the City’s multiple forms of urban greenspace than perhaps is
the case with many other cities. Besides San Francisco being particularly transparent with
their information compared with other municipalities, another factor limiting previous
studies may have been that this information was not readily accessible in a digitized
format at all in the past. By including multiple forms of green infrastructure, smaller
scale parklets, and open spaces, as well as municipal parks, the present study’s
methodology should paint a fuller picture of the actual amount of urban greenspace
experienced by residents of San Francisco’s neighborhoods.
Regarding San Francisco residents’ access to urban greenspace, there are many
questions that could be clarified by future studies. If detailed information could be
obtained about parks’ and urban greenspace’s locations and their exact dates of
installation, including information from additional census years (1980, 2000, etc.), this
might help to establish whether there is in fact a causal relationship between
gentrification and parks. As already mentioned above, future studies could also
incorporate more sophisticated park access measurement techniques such as network
analysis, and assess whether this makes a significant difference to their outcomes.

68

A valuable follow-up study could incorporate qualitative or mixed-methods
techniques such as interviews, and could directly record residents’ perception about
greenspace equity in their neighborhoods. Perhaps urban street trees or other city features
not measured by this analysis play a role in peoples’ perceptions about the amount of
greenspace they regularly encounter. Interviewers could also explore questions related to
procedural as well as distributive justice, painting a fuller picture of all the factors
contributing to this problem, rather than simply comparing two distributional snapshots in
time. Boone et al.’s (2009) study does just this in its exploration of Baltimore, comparing
resident’s perceptions with spatial analysis findings and then connecting these results to
different forms of institutionally reinforced discrimination.
Future work inspired by these findings, rather than exploring San Francisco in
greater depth could also simply adopt some of the technique that I’ve developed for
measuring gentrification. Future work might apply a version of the continuous
gentrification index developed here to other cities to assess its applicability in other cases.
My methodology adopts the basic principles of Weems’ (2016) simplified technique but
converts them into a continuous scale using ordination. This technique provides a finer
grained measurement of the actual change in neighborhood demographics compared with
the simple discrete method of considering a neighborhood to be gentrified if it has gone
from below average to above average economic/educational status.
More broadly, this study along with others that have inspired it, supports the idea
that complex demographic and socioeconomic changes, such as trends toward or away
from gentrification might be measured using spatial analysis techniques and assessed
using quantitative methods. There have been many papers written about green
69

gentrification and related theories in the social sciences, political ecology, and
environmental justice literature, but very few attempts to study this topic using GIS and
statistical analysis. Future studies might build off the few that have begun using these
techniques, and deepen their impact by supporting them with complementary qualitative
analyses. These future studies might also expand their geographic focus beyond one city
and assess the gentrification index technique’s applicability to smaller cities and/or those
that have experienced less of an intense overall trend towards gentrification (compared
with Seattle and San Francisco). Over time, studies of this nature could build a case for
how “just green enough” urban greening strategies might be successfully implemented to
avoid current residents’ displacement while also addressing environmental justice
concerns through assuring citywide greenspace equity.

Conclusion
With the availability and increasing sophistication of powerful new spatial
analysis tools, it will be increasingly easy to map out and analyze the inequities that may
exist in communities’ access to greenspace. Since lack of access to nature has emerged as
a serious public health and social justice concern, now is the perfect time for researchers
to ensure that they are leveraging these technologies to their full capacities. But mapping
out where inequities exist is just the first step in addressing them. Moving forward, it will
be important to recognize the dynamic nature of cities’ physical features and
demographic compositions and to explore factors impacting cities on these and other
levels.

70

This thesis’s findings that gentrification is associated with neighborhoods that
experience greater access to urban greenspace does not necessarily imply that installing
new greenspace will lead to gentrification. It is certainly possible that there is a causal
connection in the other direction, and that as a neighborhood’s socioeconomic status
(SES) increases, residents demand better parks and amenities. Perhaps there is a causal
connection in both directions leading to a positive feedback loop where greenspace leads
to higher SES, and higher SES leads to more greenspace (or there could be no causal link
at all). Whatever the case may be, understanding the internal dynamics that cause cities’
structures and residents to respond in different ways to physical and demographic
changes will be just as important as pinpointing where inequities exist.
Some of the factors leading to entrenched inequities might be well understood by
outside observers, while others will require impacted peoples’ participation and greater
communitywide engagement. For example, if a causal connection could be established
between park installations and gentrification, providing substantial support for the “green
gentrification” hypothesis, city planners and other decision makers could simply refer to
demographic and spatial data to help them make decisions about how to offset the
potential economic impacts of green gentrification. But in many cases, these issues must
be understood at the procedural level. A better understanding of how park policies, city
zoning, and anti-displacement measures work together to produce the results obtained
when assessing equity using spatial analysis software will be necessary to formulate a
fully informed plan moving forward.
If other studies continue to establish and support the ideas underpinning green
gentrification, then more work can be done assessing how to ensure cities are “just green
71

enough”. This thesis explored one potential technique for implementing a “just green
enough” strategy. And while that technique’s effectiveness was not supported by my
findings, there are countless other ways that the overarching strategy might still be
adopted.
Only a limited number of studies have dug into distributional equity concerns
beyond conducting basic equity maps exploring race and income. Those that have looked
substantially deeper and examined connections with procedural injustice like Boone et al.
(2009) have only begun to scratch the surface. A further integration between quantitative
equity map analyses and the rich environmental justice and social sciences literature will
be needed to chart a more equitable path into the future. Ensuring that everyone,
regardless of socioeconomic status, can enjoy the opportunity to experience and reinforce
a connection with nature, results in broadly shared social and environmental benefits for
all, in addition to improving the health and wellbeing of specific vulnerable communities
and individuals.

72

Abbreviations

DCA = Detrended Correspondence Analysis
EJ = Environmental Justice
GI = Green Infrastructure
GIS = Geographic Information Systems
GSI = Green Stormwater Infrastructure
MLR = Multiple Linear Regression
POPO = Privately Owned Public Open Space
SES = Socioeconomic Status
UGS = Urban Greenspace

73

Bibliography

Abercrombie, L. C., Sallis, J. F., Conway, T. L., Frank, L. D., Saelens, B. E., & Chapman, J.
E. (2008). Income and Racial Disparities in Access to Public Parks and Private
Recreation Facilities. American Journal of Preventive Medicine, 34(1), 9–15.
https://doi.org/10.1016/j.amepre.2007.09.030
Agyeman, J. (2001). Ethnic minorities in Britain: short change, systematic indifference and
sustainable development. Journal of Environmental Policy & Planning, 3(1), 15–30.
Atkinson, R. (2004). The evidence on the impact of gentrification: new lessons for the urban
renaissance? European Journal of Housing Policy, 4(1), 107–131.
https://doi.org/10.1080/1461671042000215479
Aytur, S. A., Rodriguez, D. A., Evenson, K. R., Catellier, D. J., & Rosamond, W. D. (2008).
The sociodemographics of land use planning: Relationships to physical activity,
accessibility, and equity. Health & Place, 14(3), 367–385.
https://doi.org/10.1016/j.healthplace.2007.08.004
Bell, J. F., Wilson, J. S., & Liu, G. C. (2008). Neighborhood Greenness and 2-Year Changes
in Body Mass Index of Children and Youth. American Journal of Preventive Medicine,
35(6), 547–553. https://doi.org/10.1016/j.amepre.2008.07.006
Bentley, R., Baker, E., & Mason, K. (2012). Cumulative exposure to poor housing
affordability and its association with mental health in men and women. Journal of
Epidemiology and Community Health (1979-), 66(9), 761–766.
Boone, C. G., Buckley, G. L., Grove, J. M., & Sister, C. (2009). Parks and People: An
Environmental Justice Inquiry in Baltimore, Maryland. Annals of the Association of
American Geographers, 99(4), 767–787. https://doi.org/10.1080/00045600903102949
Bromley, R. D. F., Thomas, C. J., & Tallon, A. R. (2005). City centre regeneration through
residential development: Contributing to sustainability. Urban Studies (Routledge),
42(13), 2407–2429. https://doi.org/10.1080/00420980500379537
Butlin, J. (1989). Our common future. By World Commission on Environment and
Development. (London, Oxford University Press, 1987, pp. 383\pounds 5.95.).
Byrne, J., & Wolch, J. (2009). Nature, race, and parks: past research and future directions for
geographic research. Progress in Human Geography, 33(6), 743-765.
74

Campbell, S. (1996). Green cities, growing cities, just cities? Journal of the American
Planning Association, 62(3), 296.
Checker, M. (2011). Wiped Out by the “Greenwave”: Environmental Gentrification and the
Paradoxical Politics of Urban Sustainability. City & Society, 23(2), 210–229.
https://doi.org/10.1111/j.1548-744X.2011.01063.x
Chen, C.-C. (2012). Understanding the Value of Amenities: A Study of the Land Value
Determination Process in Hangzhou, China. Honors Thesis, Duke University Durham.
Chiesura, A. (2004). The role of urban parks for the sustainable city. Landscape and Urban
Planning, 68(1), 129–138. https://doi.org/10.1016/j.landurbplan.2003.08.003
Comber, A., Brunsdon, C., & Green, E. (2008). Using a GIS-based network analysis to
determine urban greenspace accessibility for different ethnic and religious groups.
Landscape and Urban Planning, 86(1), 103–114.
Curran, W., & Hamilton, T. (2012). Just green enough: contesting environmental
gentrification in Greenpoint, Brooklyn. Local Environment, 17(9), 1027–1042.
https://doi.org/10.1080/13549839.2012.729569
Cutts, B. B., Darby, K. J., Boone, C. G., & Brewis, A. (2009). City structure, obesity, and
environmental justice: an integrated analysis of physical and social barriers to walkable
streets and park access. Social Science & Medicine, 69(9), 1314–1322.
Dale, A., & Newman, L. L. (2009). Sustainable development for some: green urban
development and affordability. Local Environment, 14(7), 669–681.
https://doi.org/10.1080/13549830903089283
Dooling, S. (2009). Ecological Gentrification: A Research Agenda Exploring Justice in the
City. International Journal of Urban & Regional Research, 33(3), 621–639.
https://doi.org/10.1111/j.1468-2427.2009.00860.x
Dunn, A. D. (2010). Siting Green Infrastructure: Legal and Policy Solutions to Alleviate
Urban Poverty and Promote Healthy Communities (SSRN Scholarly Paper No. ID
1517909). Rochester, NY: Social Science Research Network. Retrieved from
http://papers.ssrn.com/abstract=1517909
Eckerd, A. (2011). Cleaning up without clearing out? A spatial assessment of environmental
gentrification. Urban Affairs Review, 47(1), 31–59.

75

Essoka, J. D. (2010). The Gentrifying Effects of Brownfields Redevelopment. Western
Journal of Black Studies, 34(3), 299–315.
Giles-Corti, B., Broomhall, M. H., Knuiman, M., Collins, C., Douglas, K., Ng, K., ... &
Donovan, R. J. (2005). Increasing walking: how important is distance to, attractiveness,
and size of public open space? American journal of preventive medicine, 28(2), 169-176.
Godschalk, D. R. (2004). Land Use Planning Challenges. Journal of the American Planning
Association, 70(1), 5–13.
Gordon-Larsen, P., Nelson, M. C., Page, P., & Popkin, B. M. (2006). Inequality in the Built
Environment Underlies Key Health Disparities in Physical Activity and Obesity.
Pediatrics, 117(2), 417–424. https://doi.org/10.1542/peds.2005-0058
Gould, K. A., & Lewis, T. L. (2012). The environmental injustice of green gentrification. The
World in Brooklyn: Gentrification, Immigration, and Ethnic Politics in a Global City.
Plymouth: Lexington Books, 113–146.
Heckert, M. (2013). Access and Equity in Greenspace Provision: A Comparison of Methods to
Assess the Impacts of Greening Vacant Land. Transactions in GIS, 17(6), 808–827.
https://doi.org/10.1111/tgis.12000
Heynen, N., & Perkins, H. A. (2005). Scalar dialectics in green: urban private property and the
contradictions of the neoliberalization of nature. Capitalism Nature Socialism, 16(1), 99–
113.
Heynen, N., Perkins, H. A., & Roy, P. (2006). The Political Ecology of Uneven Urban Green
Space. Urban Affairs Review, 42(1), 3–25.
Jennings, V., Johnson Gaither, C., & Gragg, R. S. (2012). Promoting environmental justice
through urban green space access: A synopsis. Environmental Justice, 5(1), 1–7.
Jonas, A. E., & While, A. (2007). Greening the entrepreneurial city. The Sustainable
Development Paradox: Urban Political Economy in the United States and Europe, 123–
159.
Kabisch, N., & Haase, D. (2014). Green justice or just green? Provision of urban green spaces
in Berlin, Germany. Landscape and Urban Planning, 122, 129-139.
Kondo, M. C., Low, S. C., Henning, J., & Branas, C. C. (2015). The Impact of Green
Stormwater Infrastructure Installation on Surrounding Health and Safety. American
Journal of Public Health, 105(3), e114–e121. https://doi.org/10.2105/AJPH.2014.302314
76

Kuo, F. E., & Sullivan, W. C. (2001). Aggression and violence in the inner city effects of
environment via mental fatigue. Environment and Behavior, 33(4), 543–571.
Lovasi, G. S., Quinn, J. W., Neckerman, K. M., Perzanowski, M. S., & Rundle, A. (2008).
Short report: Children living in areas with more street trees have lower prevalence of
asthma. Journal of Epidemiology and Community Health (1979-), 62(7), 647–649.
Luke, T. W. (2005). Neither sustainable nor development: reconsidering sustainability in
development. Sustainable Development, 13(4), 228–238. https://doi.org/10.1002/sd.284
Maas, J., Verheij, R. A., Groenewegen, P. P., de Vries, S., & Spreeuwenberg, P. (2006). Green
space, urbanity, and health: how strong is the relation? Journal of Epidemiology and
Community Health (1979-), 60(7), 587–592.
Maciag, M. (2015, February). Gentrification in America Report. Governing: The States and
Localities. Published by e.Republic. Retrieved April 16, 2017, from
http://www.governing.com/gov-data/census/gentrification-in-cities-governing-report.html
Minnesota Population Center. National Historical Geographic Information System: Version
11.0 [Database]. Minneapolis: University of Minnesota. 2016.
http://doi.org/10.18128/D050.V11.0.
Miyake, K. K., Maroko, A. R., Grady, K. L., Maantay, J. A., & Arno, P. S. (2011). Not just a
walk in the park: Methodological improvements for determining environmental justice
implications of park access in New York City for the promotion of physical activity.
Cities and the Environment (CATE), 3(1), 8.
Nelson, Kathryn P. "Gentrification and distressed cities." Madison: University of
WisconsinPress Nolan v. California Coastal Commission (1988).
Newman, A. (2011). Inclusive urban ecological restoration in Toronto, Canada. In Human
Dimensions of Ecological Restoration (pp. 63-75). Island Press/Center for Resource
Economics.
Ngom, R., Gosselin, P., & Blais, C. (2016). Reduction of disparities in access to green spaces:
Their geographic insertion and recreational functions matter. Applied Geography, 66, 35–
51. https://doi.org/10.1016/j.apgeog.2015.11.008
Oh, K., & Jeong, S. (2007). Assessing the spatial distribution of urban parks using GIS.
Landscape and Urban Planning, 82(1–2), 25–32.
https://doi.org/10.1016/j.landurbplan.2007.01.014
77

Oksanen, J., Blanchet, F.G., Friendly, M., Kindt, R., Legendre, P., McGlinn, D., Minchin,
P.R., O'Hara, R.B., Simpson, G.L., Solymos, P., Stevens, M.H.H., Szoecs, E., & Wagner,
H. (2016). Vegan: Community Ecology Package. R package version 2.4-1.
https://CRAN.R-project.org/package=vegan
Perkins, H. A., Heynen, N., & Wilson, J. (2004). Inequitable access to urban reforestation: the
impact of urban political economy on housing tenure and urban forests. Cities, 21(4),
291–299. https://doi.org/10.1016/j.cities.2004.04.002
Poggio, L., & Vrščaj, B. (2009). A GIS-based human health risk assessment for urban green
space planning—An example from Grugliasco (Italy). Science of The Total Environment,
407(23), 5961–5970. https://doi.org/10.1016/j.scitotenv.2009.08.026
Pope, C. A., & Dockery, D. W. (2006). Health Effects of Fine Particulate Air Pollution: Lines
that Connect. Journal of the Air & Waste Management Association, 56(6), 709–742.
https://doi.org/10.1080/10473289.2006.10464485
Quastel, N. (2009). Political Ecologies of Gentrification. Urban Geography, 30(7), 694–725.
R Core Team (2016). R: A language and environment for statistical computing. R Foundation
for Statistical Computing, Vienna, Austria. http://www.R-project.org/.
Rutt, R. L., & Gulsrud, N. M. (2016). Green justice in the city: A new agenda for urban green
space research in Europe. Urban Forestry & Urban Greening, 19, 123–127.
https://doi.org/10.1016/j.ufug.2016.07.004
San Francisco City and County. (n.d.). Retrieved April 16, 2017, from
http://www.bayareacensus.ca.gov/counties/SanFranciscoCounty.htm
San Francisco OpenData. (n.d.). Retrieved April 16, 2017, from https://data.sfgov.org/
San Francisco Planning Department. (n.d.). Retrieved April 16, 2017, from http://sfplanning.org/
San Francisco Public Works. (n.d.). Retrieved April 16, 2017, from http://sfpublicworks.org/
San Francisco Recreation and Parks. (n.d.). Retrieved April 16, 2017, from
http://sfrecpark.org/
Saporito, S., & Casey, D. (2015). Are There Relationships Among Racial Segregation,
Economic Isolation, and Proximity to Green Space? Human Ecology Review, 21(2), 113–
131.

78

Schauman, S., & Salisbury, S. (1998). Restoring nature in the city: Puget Sound experiences.
Landscape and Urban Planning, 42(2–4), 287–295. https://doi.org/10.1016/S01692046(98)00093-0
Sisson, P. (2016, November 1). Staggering growth in high-income renters one reason rent is so
damn high. Retrieved May 2, 2017, from
https://www.curbed.com/2016/11/1/13489792/rental-real-estate-apartment-high-endrentals
Soper, K. (2004). Rethinking the “good life”: The consumer as citizen. Capitalism Nature
Socialism, 15(3), 111–116.
Takano, T., Nakamura, K., & Watanabe, M. (2002). Urban residential environments and
senior citizens’ longevity in megacity areas: the importance of walkable green spaces.
Journal of Epidemiology and Community Health, 56(12), 913–918.
https://doi.org/10.1136/jech.56.12.913
Taylor, A. F., Kuo, F. E., & Sullivan, W. C. (2001). Coping with add: The Surprising
Connection to Green Play Settings. Environment and Behavior, 33(1), 54–77.
https://doi.org/10.1177/00139160121972864
Ter Braak, C. J. F., & Prentice, I. C. (2004). A Theory of Gradient Analysis. In B.-A. in E.
Research (Ed.) (Vol. 34, pp. 235–282). Academic Press. https://doi.org/10.1016/S00652504(03)34003-6
Timperio, A., Ball, K., Salmon, J., Roberts, R., & Crawford, D. (2007). Is availability of
public open space equitable across areas? Health & Place, 13(2), 335–340.
https://doi.org/10.1016/j.healthplace.2006.02.003
Ulrich, R. (1984). View through a window may influence recovery. Science, 224(4647), 224–
225.
US Census Bureau. American Community Survey (ACS). Retrieved May 2, 2017, from
https://www.census.gov/programs-surveys/acs
US Census Bureau (1990, 2010). Census.gov. Retrieved April 16, 2017, from
https://census.gov/
US EPA, (n.d.). Heat Island Effect [Collections and Lists]. Retrieved April 15, 2017, from
https://www.epa.gov/heat-islands

79

Van den Berg, A. E., Maas, J., Verheij, R. A., & Groenewegen, P. P. (2010). Green space as a
buffer between stressful life events and health. Social Science & Medicine, 70(8), 1203–
1210. https://doi.org/10.1016/j.socscimed.2010.01.002
Weems, C. M. (2016). Examining the Spatial Distribution of Park Access and Trajectories of
Gentrification in Seattle, Washington 1990-2010. Ph.D. Dissertation, Oregon State
University.
Wen, M., Zhang, X., Harris, C. D., Holt, J. B., & Croft, J. B. (2013). Spatial disparities in the
distribution of parks and green spaces in the USA. Annals of Behavioral Medicine, 45(1),
18–27.
Wendel, H. E. W., Downs, J. A., & Mihelcic, J. R. (2011). Assessing Equitable Access to
Urban Green Space: The Role of Engineered Water Infrastructure. Environmental
Science & Technology, 45(16), 6728–6734. https://doi.org/10.1021/es103949f
Wenting, W., Yi, R., & Hengyu, Z. (2012). Investigation on Temperature dropping effect of
Urban Green Space in summer in Hangzhou. Energy Procedia, 14, 217–222.
https://doi.org/10.1016/j.egypro.2011.12.920
While, A., Jonas, A. E., & Gibbs, D. (2004). The environment and the entrepreneurial city:
searching for the urban “sustainability; fix”in Manchester and Leeds. International
Journal of Urban and Regional Research, 28(3), 549–569.
Wolch, J. R., Byrne, J., & Newell, J. P. (2014). Urban green space, public health, and
environmental justice: The challenge of making cities “just green enough.” Landscape
and Urban Planning, 125, 234–244. https://doi.org/10.1016/j.landurbplan.2014.01.017
Wolch, J., Wilson, J. P., & Fehrenbach, J. (2002). Parks and park funding in Los Angeles: An
equity-mapping analysis. Urban Geography, 26(1), 4–35.
Zhang, L., & Wang, H. (2006). Planning an ecological network of Xiamen Island (China)
using landscape metrics and network analysis. Landscape and Urban Planning, 78(4),
449–456.
Zhang, X., Lu, H., & Holt, J. B. (2011). Modeling spatial accessibility to parks: a national
study. International Journal of Health Geographics, 10(1), 1.
Zhou, X., & Kim, J. (2013). Social disparities in tree canopy and park accessibility: A case
study of six cities in Illinois using GIS and remote sensing. Urban Forestry & Urban
Greening, 12(1), 88–97. https://doi.org/10.1016/j.ufug.2012.11.004
80

Zukin, S., Trujillo, V., Frase, P., Jackson, D., Recuber, T., & Walker, A. (2009). New Retail
Capital and Neighborhood Change: Boutiques and Gentrification in New York City. City
& Community, 8(1), 47–64. https://doi.org/10.1111/j.1540-6040.2009.01269.x

81

Appendix: Maps
Figure 2: Location of New Urban Greenspaces in San Francisco (1990-2016)
Symbols not to scale, *GSI = green stormwater infrastructure, **POPOs = privately-owned public open
spaces (San Francisco Recreation & Parks; San Francisco Public Works; San Francisco Planning Dept.).

Figure 3: Location of Total Urban Greenspace in San Francisco
Point-based symbols not to scale, *GSI = green stormwater infrastructure, **POPOs = privately-owned
public open spaces (San Francisco Recreation & Parks; San Francisco Public Works; S.F. Planning Dept.).

82

Figure 4: Distribution of Black/African American Population in San Francisco
Polygons represent census block groups; shading indicates percentage Black population (US Census
Bureau, 2010).

Figure 5: Distribution of Asian Population in San Francisco
Polygons represent census block groups; shading indicates percentage Asian population (US Census
Bureau, 2010).

83

Figure 6: Distribution of Latino Population in San Francisco
Polygons represent census block groups; shading indicates percentage Latino population (US Census
Bureau, 2010).

Figure 7: Distribution of Youth Population in San Francisco
Polygons represent census block groups; shading indicates percentage youth population, age 19 or younger
(US Census Bureau, 2010).

84

Figure 8: Distribution of Postgraduate Population in San Francisco
Polygons represent census block groups; shading indicates percentage postgraduate population (US Census
Bureau, 2010).

Figure 9: Median Household Income by Census Block Group in San Francisco
Polygons represent census block groups; shading indicates median household income (US Census Bureau,
2010).

85

Figure 10: Median Home Value by Census Block Group in San Francisco
Polygons represent census block groups; shading indicates median home value of owner occupied homes.
Cross-hatch indicate no data was available for median value of owner-occupied homes (US Census Bureau,
2010).

Figure 11: Gentrification Index Score by Census Block Group in San Francisco
Polygons represent census block groups; shading indicates gentrification index score based on ordination of
census data for change in income, home value, and education by CBG (US Census Bureau, 1990 & 2010).

86

Figure 12: Districts of San Francisco, CA
This map was produced by Peter Fitzgerald using OpenStreetMap. Creative Commons Attribution-Share
Alike 2.0 Generic License (CC BY-SA 2.0). http://creativecommons.org/licenses/by-sa/2.0/legalcode

87