Content uploaded by Alessandro Rigolon
Author content
All content in this area was uploaded by Alessandro Rigolon on Feb 18, 2020
Content may be subject to copyright.
Contents lists available at ScienceDirect
Landscape and Urban Planning
journal homepage: www.elsevier.com/locate/landurbplan
Beyond proximity: Extending the “greening hypothesis” in the context of
vacant lot stewardship
Paul H. Gobster
a,⁎
, Alessandro Rigolon
b
, Sara Hadavi
c
, William P. Stewart
d
a
USDA Forest Service, Northern Research Station, Evanston, IL 60201, USA
b
Department of City and Metropolitan Planning, The University of Utah, Salt Lake City, UT 84112, USA
c
Department of Landscape Architecture and Regional & Community Planning, Kansas State University, Manhattan, KS 66506, USA
d
Department of Recreation, Sport and Tourism, University of Illinois at Urbana-Champaign, Champaign, IL 61820, USA
ABSTRACT
Research increasingly shows that greening activity can spur contagious or imitative behavior among nearby neighbors within residential landscapes. Krusky et al.
(2015) examined this phenomenon in the context of vacant lots and found support for a “greening hypothesis” that residential yards near vacant lots that were
converted to community gardens exhibited higher levels of care than yards near untended vacant lots. Although such activity implies a temporal, causal relationship,
research to date has only tested the spatial dimension of greening through correlational measures of proximity assessed at one point in time. We extend this work by
analyzing vacant lot greening as a function of time, space, scale of analysis, and other factors. We studied residential property owners (N= 321) who purchased
nearby city-owned vacant lots through the Chicago Large Lot Program. Improvements made in the condition and care of large lots in the year after purchase were
positively related to the proximity, condition and care of the individual’s previously owned property, and signs of use and care of the lot before purchase (blotting).
We also examined whether block-level indicators of care and disorder were associated with improvements made to lots purchased on the block. We found few
associations but discovered these same block-level indicators of care and disorder more strongly predicted the percent of large lots sold on that block, suggesting that
greening activity may be bidirectional. These findings expand understanding of the dynamics of vacant lot stewardship and have implications for building more
robust theories of urban greening.
1. Introduction
A landscape’s appearance plays an important role in people’s en-
vironmental preferences and can affect other ways in which they per-
ceive, engage with, and value landscapes (Gobster, Ribe, & Palmer,
2019). Residential landscapes with well-maintained homes and attrac-
tive yards, on tree-lined streets, near public green space can contribute
to improved neighborhood quality of life (Douglas, Russell, & Scott,
2019), increased property values (Crompton, 2004), and other positive
individual and social outcomes (Krekel, Kolbe, & Wüstemann, 2016;
Root, Silbernagel, & Litt, 2017). Although the cultural norms that un-
derlie these preferred landscape characteristics are most often studied
in suburban settings (Nassauer, Wang, & Dayrell, 2009; Uren, Dzidic, &
Bishop, 2015), recent research shows they also apply to the improve-
ment of urban neighborhoods diminished by high levels of vacancy. In a
study of vacant lot greening in Flint, Michigan (USA), Krusky et al.
(2015) found that residential yards located near vacant lots that had
been transformed into community produce (vegetable) gardens were
better maintained than yards near untended vacant lots, providing
support for their “greening hypothesis.” Noting the prevalence of such
patterns across their study area, the authors concluded that greening
initiatives can play a catalytic role in community revitalization efforts.
The proximal relationship that underlies the greening hypothesis
has been identified in other residential landscape contexts, including
similarity in front yard planting designs (e.g., Minor, Belaire, Davis,
Franco, & Lin, 2016; Zmyslony & Gagnon, 2000) and clustering of ea-
sement gardens (Hunter & Brown, 2012) within neighborhoods. These
studies underscore the potential of urban greening in neighborhood
improvement and more broadly demonstrate how nearby nature can
promote positive behavioral change (Norwood et al., 2019; Roberts,
McEachan, Margary, Conner, & Kellar, 2018). But though the process of
greening described by the investigators implies a temporal aspect,
studies to date have focused primarily on the spatial (proximal) re-
lationships between properties. Moreover, the studies also imply a
causal agent and direction of change, yet again they provide little in-
sight into how changes originate and spread across the landscape.
A vacant lot repurposing program in the City of Chicago provided
the opportunity to address these knowledge gaps. The Chicago
Large Lot Program sells city-owned vacant lots in high-vacancy
neighborhoods to nearby property owners for a nominal fee, with the
https://doi.org/10.1016/j.landurbplan.2020.103773
Received 2 October 2019; Received in revised form 6 February 2020; Accepted 7 February 2020
⁎
Corresponding author.
E-mail addresses: paul.gobster@usda.gov (P.H. Gobster), alessandro.rigolon@utah.edu (A. Rigolon), sarahadavi@ksu.edu (S. Hadavi),
wstewart@illinois.edu (W.P. Stewart).
Landscape and Urban Planning 197 (2020) 103773
0169-2046/ Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
T
goals of reducing municipal maintenance burdens, creating local wealth
through ownership, and stabilizing population loss (City of Chicago,
2014a). To help evaluate the program, we studied the greening activity
undertaken by new owners of “large lots” (i.e., lots purchased under the
Chicago Large Lot Program), and in a previous paper detailed the types
and magnitude of changes made in the condition (level of maintenance)
and care (signs of stewardship and occupancy) of lots purchased
(Gobster, Hadavi, Rigolon, & Stewart, 2020). Additional information
collected on owner- and block-level characteristics now allows us to
extend work on the greening hypothesis to examine how owner
proximity and other factors affect vacant lot stewardship.
2. Background and hypotheses for research
Our research examines changes in lot-level condition and care re-
lative to two other scales of analysis: the purchaser’s originally owned
property and the block in which the large lot was purchased. At these
two scales, our hypotheses fall into five conceptual categories described
in the subsections below. Fig. 1 provides a layout of concepts and hy-
potheses described in this section and Table 1 summarizes each of the
eleven hypotheses tested. Four of the five conceptual categories cover
hypotheses relating to large lot-owned property level relationships:
proximity, occupant type, behavioral antecedents of care, and blotting;
while the fifth covers hypotheses relating to large lot-block level re-
lationships.
2.1. Proximity and care
The benefits of nearby nature in urban settings are well docu-
mented, with investigations spanning more than four decades (e.g.,
Kaplan, 1973; Lewis, 1979). Kaplan and Kaplan (1989) found that vi-
sual and physical access to natural environments can play an important
role in environmental preference, residential satisfaction, and human
health and well-being; that even small green areas such as yards and
gardens can provide important benefits; and that it is often proximity
and not size that matters in determining frequency of use and beneficial
outcomes.
Kaplan and Kaplan (1989) also noted that some types of green
spaces and activities bring people in closer contact with nature, and
that of these, gardens and gardening can deepen connections with
nature for individuals and engage the nearby community. This ability to
generate broader engagement has since been described spatially by
urban ecologists, who have documented the “mimicry” in planting
designs and species selections across nearby front yards (Minor et al.,
2016; Zmyslony & Gagnon, 1998, 2000; for a counterexample see
Kirkpatrick, Daniels, & Davison, 2009) and “spatial contagion” of ea-
sement gardens within neighborhoods (Hunter & Brown, 2012). In
these studies, the proximal nature of activity is of central interest and
shows a decrease with distance—Minor et al. (2016) found patterns of
mimicry decline for front yards more than nine lots away (~67 m),
whereas Hunter and Brown (2012) found that the peak clustering of
Fig. 1. Conceptual diagram of the research. The ovals at the top of the diagram portray the scale relationships examined, while the boxes portray the concepts
studied. The numbered, darker boxes correspond to concepts tested by hypotheses and the numbered arrows are hypothesized interaction effects between concepts.
See text and Table 1 for further details.
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
2
easement gardens occurred within a 91 m radius of a given garden.
Proximity is also the central concept behind the greening hypothesis
examined by Krusky et al. (2015), who extended the work described
above into the important issue of urban vacancy. Flint, Michigan, like
many post-industrial cities in the US and globally, has suffered major
population loss resulting in a large number of vacant properties in
need of repurposing. Drawing on other research that greening can po-
sitively impact revitalization efforts, Krusky et al. (2015) posited
that well-maintained community produce gardens in high-vacancy
neighborhoods would lead nearby homeowners to maintain their own
yards at a higher level than residents who lived by similar open parcels
that remained vacant and untended. Referencing the 91 m proximity
threshold identified by Hunter and Brown (2012), Krusky et al. (2015)
studied two residential areas and measured the maintenance levels of
yards within 100 m and 50 m of 19 different produce gardens, com-
paring them with yards of similar proximities to a sample of vacant lots.
While the investigators found significant differences in yard main-
tenance between garden- and vacant lot-proximate lots in support of
their greening hypothesis, their data and research design prevented
them from determining causality as to whether “the greening effect
radiated from the produce gardens or if engaged residents created the
produce garden” (Krusky et al., 2015, p. 74). The authors suggested that
future research could consider ownership to determine the origin and
direction of the greening relationship.
Data from our large lot study provide such an opportunity because
the ownership and timing of greening activity are known. Our earlier
work assessed the maintenance condition and care of large lots before
and after purchase and found that significant changes were made, in-
cluding an increased percentage of lots with gardens and other “cues to
care” (Nassauer, 1995), reductions in parked vehicles, and improve-
ments in the condition of mature trees (Gobster et al., 2020). While that
analysis provided the temporal and directional evidence to support a
causal relationship between ownership and improvements in condition
and care, the particular requirements of the Large Lot Program also
provide a natural experiment of sorts to test the effects of proximity on
lot improvements. Unlike most vacant lot “side yard” programs in the
US (Ganning & Tighe, 2015), the Large Lot Program does not stipulate
that residents share a common property boundary with a vacant lot in
order to purchase it; they only need to own property on the block or
adjacent block. In practice, this distance spans the range identified by
the studies mentioned above, and thus we anticipated that improve-
ments in the condition and care of large lots would increase with
greater proximity between the large lots and the purchaser’s original
property (H1: “Proximity”).
2.2. Ownership and care
2.2.1. Occupant type
For many vacant residential lot resale programs, in order to qualify
for purchase of a city-owned lot at a low price, the applicant must be an
owner-occupant of the adjacent lot (e.g., Cuyahoga Land Bank
(Cleveland), n.d.; Land Bank of Kansas City, MO, n.d.; Philadelphia
Land Bank, n.d.). This requirement assumes that owner-occupancy
carries a heightened level of responsibility, tenure and fiscal stability,
and commitment to neighborhood improvement over absentee owners
or renters. Each of these reasons may be generally true (e.g., McCabe,
2013); however, they can limit the scope of programs, particularly in
areas where there is a high percentage of rental units (Ganning & Tighe,
2015). Other programs stipulate that the owned property need only be
an occupied residence (e.g., City of St. Louis, MO, n.d.; Detroit Land
Bank Authority, n.d.; New Orleans Redevelopment Authority, n.d.), and
although there is some evidence that absentee owners and renters may
be less likely to maintain their residences than owner-occupants
(Garvin, Branas, Keddem, Sellman, & Cannuscio, 2013; Goldstein,
Jensen, & Reiskin, 2001), it is unknown whether this also applies to
vacant lot greening. The Large Lot Program also requires an applicant
to own property on the block, but there are no restrictions on whether
they are an owner-occupant or even if their owned property is just
another vacant lot. This flexibility provides another range of conditions
to test, and because homeownership generally has positive effects for
neighborhoods (Aarland & Reid, 2019; Heidelberg & Eckerd, 2011), we
expected that large lots purchased by owner-occupants would show
bigger improvements in condition and care than if the purchaser’s
original property was in absentee ownership or vacant (H2.1: “Occu-
pant type”). Furthermore, we expected a positive sign to the interaction
term between owner-occupant and proximity to the large lot, indicating
that large lots purchased by owner-occupants who live in greatest
proximity would show the biggest improvements in condition and care
(H2.2: “Owner-occupant × proximity”).
2.2.2. Behavioral antecedents of owner care
Another ownership-related factor concerns the behavioral ante-
cedents of vacant lot care. A growing body of research has examined
how homeowners perceive and manage their yards and gardens (e.g.,
Table 1
Summary of hypotheses tested in the paper.
Number Concept Hypothesis
H1 Proximity The condition and care of large lots would increase with greater proximity between the large lots and the purchaser’s
original property.
H2.1 Occupant type Large lots purchased by owner-occupants would show bigger improvements in condition and care than if the purchaser’s
original property was in absentee ownership or vacant.
H2.2 Owner-occupant × proximity The interaction term between owner-occupant and proximity would have a positive sign, indicating that large lots
purchased by owner-occupants who live in greatest proximity would show the biggest improvements in condition and care.
H3.1 Prior stewardship Owners whose original property showed a high level of condition and care would extend higher levels of care to their
purchased large lots than owners whose original property showed a low level of care.
H3.2 Number of lots Owners who purchased a single large lot would extend higher levels of care to it than owners who purchased two lots.
H3.3 Area managed Owners whose combined property area was smaller would extend higher levels of care to their large lots than owners whose
combined property area was larger.
H3.4 Tax payments Owners whose large lot property taxes were fully paid would extend higher levels of care to their large lots than owners
who carried a large deficit in their tax payments.
H4.1 Blotting × owner-occupant The interaction term between blotting and owner-occupant would have a positive sign, indicating that large lots that were
both blotted and purchased by owner-occupants would show the biggest improvements in condition and care.
H4.2 Proximity × owner-occupant × blotting The interaction term between proximity, owner-occupant, and blotting together would have a positive sign, indicating a
synergistic effect of these three variables on improvement in large lot condition and care.
H5.1 Block-level care Large lots purchased on blocks showing high levels of care and environmental amenities and low levels of disorder would
show higher levels of condition and care than lots purchased on blocks with low levels of care and amenities and high levels
of disorder.
H5.2 Percent of large lots sold on block A higher proportion of available large lots would be sold on blocks showing high levels of care and amenities and low levels
of disorder than on blocks with low levels of care and amenities and high levels of disorder.
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
3
Cook, Hall, & Larson, 2012; Giner, Polsky, Pontius, & Runfola, 2013).
These studies suggest that behavioral changes by individual property
owners can enhance the delivery of ecosystem services (Goddard,
Dougill, & Benton, 2010; Larson et al., 2016). Among the factors that
predict environmental behavior (Harland, Staats, & Wilke, 1999; Kurz
& Baudains, 2012), prior stewardship behavior has particular relevance
to vacant lot management. Specifically, prior behavior can be a good
predictor of future behavior if it is closely aligned with future behavior,
frequently practiced, and similar with respect to cost and effort (Moore
& Boldero, 2017; Oulette & Wood, 1998). We expected that owners
whose original property showed a high level of condition and care
would extend higher levels of care to their purchased large lots than
owners whose original property showed a low level of care (H3.1:
“Prior stewardship”). With respect to cost and effort, we anticipated
that owners who purchased a single large lot, whose combined property
area was smaller, or whose large lot property taxes were fully paid
would extend higher levels of care to their large lots than owners who
purchased two lots, whose combined property area was larger, or who
carried a large deficit in their tax payments (respectively, H3.2:
“Number of lots”; H3.3: “Area managed”; H3.4: “Tax payments”).
2.2.3. “Blotting” interactions
In our initial coding of large lot condition and care, we observed
that nearly a third of the city-owned vacant large lots in our sample
showed signs of use and/or stewardship prior to purchase (Gobster
et al., 2020). This unsanctioned activity has been termed “blotting”
(Armborst, D’Oca, & Theodore, 2008) and includes such things as fen-
cing and mowing lots, parking cars, and planting flowers in vacant lots
owned by the city. We expected that blotting would affect observed
changes in large lot condition and care and the results of our previous
study confirmed our hypotheses. We found that although unblotted
large lots exhibited bigger changes in condition and care after purchase
than blotted ones, additional care was invested in blotted lots after they
were purchased and their overall level of care was higher than un-
blotted lots (Gobster et al., 2020). Our added data on ownership cannot
identify who used or stewarded the parcels prior to purchase, but it
does allow us to test whether occupant type and blotting have an in-
teractive effect on changes in large lot condition and care. We expected
that the interaction term between blotting and owner-occupant would
have a positive sign, indicating that large lots that were both blotted
before purchase and purchased by owner-occupants would show the
biggest improvements in condition and care (H4.1: “Blotting × owner-
occupant”). This interaction would support the assumption that blotting
takes place by those who own and occupy property on the block
(Armborst et al., 2008). Furthermore, we expected that the interaction
between proximity, owner-occupant, and blotting together would have
a positive sign, indicating a synergistic effect of these three variables on
improvement in large lot condition and care (H4.2: “Proxi-
mity × owner-occupant × blotting”).
2.3. Block-level effects
Krusky et al. (2015) investigated the greening hypothesis at two
scales of analysis, parcel-level measures of yard maintenance and
neighborhood-level measures of disorder, social capital, and participa-
tion. Such social-ecological frameworks contend that human behavior is
influenced at multiple levels, from intrapersonal to public policy
(McLeroy, Bibeau, Steckler, & Glanz, 1988). Along related lines,
Nassauer et al. (2009) found that neighborhood norms of residential
landscape design had dramatic effects on individual homeowner land-
scaping preferences. Following this work, we hypothesized that changes
made to large lots purchased on blocks exhibiting high levels of care
and environmental amenities and low levels of disorder would show
higher levels of condition and care than those purchased on blocks with
low levels of care and amenities and high levels of disorder (H5.1:
“Block-level care”).
Finally, while Krusky et al. (2015) assumed a causal, directional
path of greening extending from produce gardens to nearby yards,
they also discussed how relationships between landscape aesthetics and
resident behavior can be bidirectional. In the context of our work, this
suggests that blocks exhibiting higher levels of care may be seen as
more attractive locations in which to purchase a large lot than
blocks exhibiting lower levels of care. Other research has shown that
neighborhoods with attractive environmental features such as trees and
nearby public green space are associated with increased property values
(e.g., Staats & Swain, 2020; Donovan, Landry, & Winter, 2019) and
greater housing demand (e.g., Koprowska, Łaszkiewicz, & Kronenberg,
2020). To test this bidirectionality, we examined the degree to which
block-level characteristics would have a motivating influence on whe-
ther individuals chose to buy lots on that block. We expected that a
higher proportion of available large lots would be sold on blocks ex-
hibiting high levels of care and amenities and low levels of disorder
than on blocks with low levels of care and amenities and high levels of
disorder (H5.2: “Percent of lots sold”).
3. Methods
3.1. The research setting
The Chicago Large Lot Program began in 2014 as part of the im-
plementation of the Green, Healthy Neighborhoods Plan (City of
Chicago, 2014b). Under the program, qualified property owners in
targeted high-vacancy areas are able to purchase one or two city-owned
vacant lots on their block or the adjacent block for $1 each, with the
provision that they maintain the lots, pay the property taxes, and fence
the property if it is not immediately side-adjacent to their existing
owned property. The city’s Department of Planning and Development
expected that at least initially, most purchasers would use their large lot
as private or shared green space (City of Chicago, 2014a).
The first wave of large lot sales was offered in 2014 in Greater
Englewood and East Garfield Park, two community areas located on the
south and west sides with high land vacancy, a large proportion of
African American residents, and high poverty profiles (City of Chicago,
2014b). In Greater Englewood, 209 qualified applicants purchased 275
out of 4,062 available lots (7%) on 185 different blocks. In East Garfield
Park, 112 qualified applicants purchased 149 out of 418 available lots
(36%) on 67 different blocks. In total, our study examined 424 large
lots distributed across 321 owners and 252 blocks.
3.2. Variables and their measurement
Our variables derive from multiple data sources including large lot
and owner address information from the City of Chicago and Large Lot
Program website (LargeLot.org, n.d.), property boundary and lot size
information from the Cook County property tax portal (Cook County,
n.d.), and aerial and street-level imagery from Google, Bing, and our
own field photography. As illustrated in Fig. 2 and described in the
subsections below, these data sources were used to develop and mea-
sure variables at three different scales of analysis corresponding to the
purchased large lots (N = 424), previously owned property (N = 321),
and block where large lots were purchased (N = 252). Census block
group data would provide yet another scale of analysis to test for socio-
demographic effects such as undertaken by Krusky et al. (2015) but,
given the relatively high homogeneity of resident demographics in our
two study areas (Stewart et al., 2019), we did not include it in the
present study.
3.2.1. Large lot condition-care index and its temporal sequence
Our main dependent variable was lot-level condition and care, an
index of seven, binary-coded features developed in our earlier study
based upon a visual assessment of street-level images of the 424 large
lots. The seven items were selected for the index from a larger set of
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
4
variables (see Supplementary Appendix 1) as they dealt specifically
with lot condition and care and showed acceptable internal consistency
when added together to form the index. The items included: the visual
condition of lot features including pavement, shrubs and small trees,
mature trees, and fencing (0 = absent or not in good condition,
1 = good condition); and the presence of cues to care including gar-
dens, yard ornamentation, and social/recreational uses (0 = absent,
1 = present). To capture the temporal nature of changes made, the
variable was treated as a repeated measure and assessed by rating
images of the purchased lots taken just before (fall 2014) and one
growing season after (fall 2015) purchase (Cronbach’s alpha = 0.753
before purchase and 0.698 after purchase).
Because our tests of the proximity and ownership hypotheses and
their interactions with blotting (Hypotheses 1–4) used the owner
(N = 321) as the unit of analysis, we averaged the values of the con-
dition-care index for those owners (n = 103, 32%) who had purchased
two large lots.
3.2.2. Proximity
Proximity was measured as a four-level ordinal variable (1 = low or
farthest, 4 = high or closest) reflecting the degree of adjacency be-
tween the purchased large lot and the owner’s previously owned
property (Fig. 3). Large lots and owner lots were identified by their
property identification numbers (PINs) and located spatially using the
Cook County property tax portal. The highest level of proximity (i.e., 4)
was assigned to owner/large lot pairs that were immediately side ad-
jacent, with the next highest level to lots that were front, rear, or di-
agonally adjacent (i.e., 3). This distinction reflects more than just
physical distance, and in the latter cases entails having to cross a street
or alley to access the purchased property, which could present an added
burden if running a hose or carrying heavy materials to the large lot.
Lots two-to-four lots away from each other were assigned the next level
of proximity (i.e., 2) and lots five or more lots away the most distant
level (i.e., 1).
For analysis purposes, proximity values were averaged for owners
who had purchased two large lots. However, if their second large lot
was immediately side adjacent to their first large lot, we considered
both lots to be at the closest proximity level to the owner’s original
property, again reflecting burden level versus simple distance.
3.2.3. Occupant type
We used a two-stage process to code the occupancy status of the
large lot purchaser’s originally owned property on the block. First, we
checked the tax bill mailing address on the Cook County property tax
portal for correspondence with the owned property. Second, we ex-
amined the street-level image of the owned property to code whether
the lot was vacant or had a habitable structure. Given these values, we
coded occupant type as a categorical variable with three levels: owner-
occupant, owner-absentee, or owner-vacant.
3.2.4. Owner lot condition-care index
We used street-level imagery to assess the visible characteristics of
the large lot purchaser’s previously owned lot relating to condition and
care prior to purchase. In addition to the seven indicators mentioned
above for large lots, we also included indicators reflecting the condition
of turf and buildings. All measures were converted to binary variables
Fig. 2. Illustration of large lot-, owner-, and block-level units of analysis and corresponding variable sets used in the study. See Supplementary Appendix 1 for full
description of variables. (For interpretation of the colours in this figure legend, the reader is referred to the web version of this article.)
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
5
(0 = absent or not in good condition, 1 = good condition) and added
together to form a nine-item index of owner lot condition and care
(Cronbach’s alpha = 0.711).
3.2.5. Cost and effort measures
Using lot data from the Cook County property tax portal we calcu-
lated the total lot area owned by each large lot owner, including their
previously owned lot on the block plus newly purchased large lot(s).
Total lot area ranged from 250 – 3000 m
2
, with a median value of
800 m
2
. The typical Chicago lot is 290 m
2
.
Property tax data were also retrieved. We noted how much the last
tax assessment (2018, first half) was per lot, whether the owner had
paid, and from that calculated the amount of taxes unpaid (delinquent).
We multiplied this value for the number of lots purchased to arrive at a
total amount per owner ($0 - $3143 USD, median $0, SD $416). As a
measure of cost and effort, this variable reflects both the magnitude of
the investment and the commitment to paying it.
3.2.6. Blotting
Adapting definitions by Armborst et al. (2008) and Dewar,
Nassauer, and Dueweke (2013), we coded lots as blotted if, prior to
purchase, they were fenced and showed signs of regular mowing, or if
they were unfenced but were mown and showed at least one other sign
of care or occupancy such as parked cars or gardens. We used Google
Street View imagery taken prior to lot purchase to code this binary
variable.
3.2.7. Block-level measures
The block-level analysis used blocks as the unit of analysis, con-
sidering all blocks where at least one large lot was sold (N = 252).
Blocks were defined as the area between two street intersections and
included lots facing both sides of the street (see Fig. 2). Our definition
differs from the one used under which blocks were purchased (see
Fig. 3) because our primary concern in this analysis was the view of the
streetscape in front of where the owner’s lot is located and not in the
alley behind it. This definition conforms to similar types of analyses
(e.g., Maroko, Weiss Riley, Reed, & Malcolm, 2014; Mooney et al.,
2014) and also avoids overlap between blocks. Because blocks varied in
length (70–410 m, median 200 m) and number of lots (6–107, median
35), most of the block-level predictor variables were standardized for
comparison by dividing values by total block length or lot number.
Block-level variables were measured using Google Earth (aerial) and
Google Street View (street-level) imagery and linked to aerial imagery
from the Cook County property tax portal showing individual parcels.
In the assessment, a researcher examined the aerial images and made
“virtual walks” down the block in Street View and inventoried features
selected as predictors of our two dependent variables (see below). Most
of the images were taken in the summer-fall of 2014, representing
conditions when lots were selected for purchase.
Block-level predictors were adapted from previous studies that ex-
amined lot and block-level attributes as part of neighborhood quality of
life assessments. These assessments typically use aerial and street-level
images or field observation to inventory conditions relating to care
Fig. 3. Demonstration of proximity coding showing a hypothetical owner’s property (center, blue with dashed border), proximity values (ordinal scale with 1 = low
or farthest, 4 = high or closest) for available large lots (green), and other vacant lots on the block (yellow). Solid border around the two large lots at the top of the
photo shows coding if two lots are purchased directly side adjacent to one another (see text). (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
6
(Dewar et al., 2013), disorder (Mooney et al., 2014; Sampson &
Raudenbush, 2004), or amenities influencing concepts such as walk-
ability (Hajna, Dasgupta, Halparin, & Ross, 2013). Using this literature
for general guidance, we coded multiple indicators to test in our
models, anticipating some would perform better than others. Indicators
of care and environmental amenities included the number of street trees
(on the public easement between street and sidewalk) and big trees
(canopy trees in yards and on the public easement), the presence of
street calming features (e.g., speed humps, traffic circles), the presence
of public or semi-public open space (green or hardscape) on or im-
mediately adjacent to the block, and the number of lots exhibiting cues
to care (e.g., gardens, play equipment). Indicators of disorder included
the number of vacant lots (open and with vacant buildings), number of
lots exhibiting cues to neglect or mistreatment (e.g., eroded ground,
abandoned vehicles) and percent of the block with unwalkable (e.g.,
broken concrete) sidewalks (see Supplementary Appendix 1 for details).
These initial measures were used as-is or standardized using lot
number and block length information and a subset was selected for
modeling purposes (see next section). Two dependent variables were
used in the prediction models, the large lot condition-care index (for
Hypothesis 5.1) and the percent of large lots sold on the block (for
Hypothesis 5.2). For the former, values of the index were averaged for
the number of large lots sold on the block (1–8, mean 1.69, median 1).
For the latter, the number of large lots sold was divided by the number
of all large lots available for sale on that block (1–27, mean 4.2, median
3). Sales data came from the LargeLots.org website.
3.3. Analysis and modeling
We tested our hypotheses with different multivariate linear models
to address the owner and block levels of analysis. For the models where
large lot condition and care was the dependent variable, we used
hierarchical linear modeling (HLM) because we have repeated mea-
sures of our dependent variable before and after the purchase of large
lots and because of its flexible assumptions of normality (Raudenbush &
Bryk, 2002). To account for this temporal sequence, we repeated the
data matrix to create a binary variable, time, with the top half of the
matrix representing conditions before purchase (time = 0) and the
bottom half after purchase (time = 1). In the HLM models, we treated
time along with the other independent variables as fixed effects and
used owner ID (for owner-level models) and block ID (for block-level
models) as grouping factors (random-effect intercepts). Because all
owners purchased at least one lot, time represents the “treatment” in
our mixed-effects models. The models were estimated using maximum
likelihood ratio, and t-tests of independent variable significance used
Satterthwaite’s formula. Model fit was examined using the Akaike In-
formation Criterion (AIC) and the marginal R
2
(estimating the variance
explained by the fixed-effects variables).
Besides time, the other independent variables included in the
owner-level models were input as either continuous or binary variables.
Proximity, the owner condition-care index (prior stewardship), size of
combined lot areas (area managed) and tax payments unpaid were in-
cluded as continuous variables. Occupant type was recoded as two
binary variables, owner-occupant and owner-absentee, and number of
lots owned and blotting were also included as binary variables.
While we had specific hypotheses driving selection for each of the
independent variables in our owner-level models, the hypotheses for
our block-level models were more general and we had developed a
number of different measures of block-level care, amenities, and dis-
order (Supplementary Appendix 1). To help guide the selection of
particular independent variables, we examined bivariate correlations
and selected predictors that correlated highest across both dependent
variables and were not highly correlated with each other (see
Thompson, 1978).
For the block-level model where the percent of large lots sold was
the dependent variable, we used an OLS regression model, entering all
independents in a single step. Independent variables were checked for
multicollinearity with variance inflation factors (VIF) and model fit was
assessed using R
2
adj
. Both HLM and OLS models were estimated with
RStudio, using the packages lme4, lmerTest, and sjstats to fit the mixed-
effects models and estimate the marginal R
2
(R Core Team, 2013). All
model assumptions, including the normal distribution of residuals (see
Pinheiro & Bates, 2000), were tested and met.
3.3.1. Sensitivity analysis and post-hoc tests
For the owner-level models, we also conducted a sensitivity analysis
to evaluate whether results of the proximity and occupant type vari-
ables were sensitive to the ordinal and categorical nature of these two
variables, respectively, or whether simplified versions of these variables
would yield similar results (see Greenland, 1996). The analysis also
permitted us to simplify the post-hoc tests of the interaction effects
between these variables as described below.
For the sensitivity analysis, we recoded the proximity and occupant
type variables as binary and re-ran HLMs with those binary variables.
Specifically, the proximity variable was coded as 1 = close if the large
lot was immediately adjacent to the owner’s original lot (coded as 4 in
the main analysis, see Fig. 3), and 0 = far if otherwise. We coded oc-
cupant type as 1 = owner-occupant if the owner of the property was
living on the premises, and 0 = not owner-occupant if otherwise (i.e.,
absentee or vacant lot). Accordingly, we also built the interaction
variables described earlier (for Hypotheses 2.2, 4.1, and 4.2) using the
recoded binary proximity and occupant type variables.
For the post-hoc tests, we used Tukey-adjusted pvalues to test for
significance between paired estimated marginal means of the three
simplified interaction variables. We ran post-hoc pairwise comparisons
for the interaction terms in the sensitivity analysis because the binary
nature of the proximity, blotting, and occupant type variables made the
post-hoc tests easier to interpret (2x2 cells for Hypotheses 2.2 and 4.1,
and 2x2x2 cells for Hypothesis 4.2) than for the related ordinal
proximity (7 possible values) and categorical occupant type variables (3
possible values) used in the main analysis. The emmeans package in R
calculates Tukey adjusted estimated marginal means for pairwise
comparisons between cells of cases resulting from interaction terms (for
example, owner-occupant and blotted vs. owner-occupant and non-
blotted for Hypothesis 4.1).
4. Results
4.1. Owner-level models
Table 2 reports the unstandardized coefficients of the owner-level
HLM main and interaction effects models. As discussed in our earlier
paper (Gobster et al., 2020), large lot condition and care was sig-
nificantly associated with time and prior blotting, and these variables
remained important predictors in all models (see Tables A1–A4 in
Supplementary Appendix 2 for complete specifications of all models,
including pand tvalues and 95% confidence intervals). Model 1 in
Table 2 shows the main effects of the additional owner-related variables
examined in this paper, and here proximity and the level of care given
to the purchaser’s own lot were significantly associated with large lot
condition and care (p= 0.017 and p= 0.008) in support of the
proximity (H1) and prior stewardship (H3.1) hypotheses, respectively.
The occupant type binary variables owner-occupant and owner-ab-
sentee and the cost and effort related variables of lot number, size, and
tax payments showed no significant associations with large lot condi-
tion and care and thus the occupant type (H2.1), number of lots (H3.2),
area managed (H3.3), and tax payments (H3.4) hypotheses are not
supported.
We then introduced interaction terms in three separate models
(Models 2–4 in Table 2). The two-way interactions between proximity
and owner-occupant and blotting and owner-occupant were each sig-
nificant and had positive signs (p= 0.044 and p= 0.002), in support of
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
7
Hypotheses 2.2 and 4.1, respectively. Similarly, the three-way inter-
action between proximity, owner-occupant, and blotting was significant
and had a positive sign (p = 0.0005), in support of Hypothesis 4.2.
Each of these interaction terms improves model fit, with model 4 in-
corporating the three-way interaction providing the best fit (lowest AIC
value).
The sensitivity analysis shown in Table 3, in which the proximity
and occupant type variables were recoded as binary, confirmed the
results of the main models reported in Table 2. All the significant re-
gression coefficients in the four models presented in Table 2 (main
analysis) are also significant and have the same sign in Table 3 (sen-
sitivity analysis). This shows that results for proximity and occupant
type (and related interaction terms) are not sensitive to the ordinal and
categorical nature of these two variables. In other words, proximity is
significantly associated with large lot condition and care regardless of
whether the variable is coded as ordinal or binary (H1), while occupant
type shows no significant associations with large lot condition and care
for neither its categorical nor binary version (H2.1).
The results of the post-hoc tests for the three interaction terms in the
sensitivity analysis are summarized in Table 3 and described in more
detail in Table A5 of Supplementary Appendix 2 (the three interactions
were all significant). At least one pairwise comparison of the estimated
marginal means is significant for each interaction term (see Table A5).
Overall, the post-hoc tests suggest that blotting and to lesser extent
proximity are stronger determinants of the statistical significance and of
the positive sign of the three interaction terms than occupant type. In
addition, the post-hoc tests showed that occupant type serves as an
effect modifier of these variables in relation to large lot condition and
care. For the interaction between occupant type and proximity, large
lots located in greatest proximity to the original property where an
owner-occupant resided had higher estimated marginal means in con-
dition and care than large lots located farther from an original property
with an owner-occupant (p< 0.01), in further support of Hypothesis
2.2. For the interaction between occupant type and blotting, large lots
purchased by owner-occupants that were also blotted had significantly
higher condition and care than 1) large lots purchased by owner-oc-
cupants but were not blotted (p< 0.001), 2) without an owner-oc-
cupant that were also blotted (p< 0.05), and 3) without an owner-
occupant and were not blotted (p< 0.001), in further support of
Hypothesis 4.1. Finally, 14 out of the 28 pairwise comparisons were
significant in the post-hoc test for the three-way interaction between
proximity, occupant type, and blotting (H4.2; see Table A5). The largest
effect sizes were for pairwise comparisons including blotted and non-
blotted large lots (see the tratios in Table A5).
Table 2
Owner-level models predicting condition and care.
Variable/(concept) Hypothesis tested Model 1 (main) Model 2 Model 3 Model 4
Fixed effects
(Intercept) −0.410 −0.036 −0.306 −0.179
Time 0.623
***
0.624
***
0.624
***
0.624
***
Proximity H1 0.123* −0.033 0.128* 0.071
Owner-occupant (occupant type) H2.1 0.043 −0.557 −0.121 −0.091
Owner-absentee (occupant type) H2.1 −0.043 0.020 0.052 0.066
Owner condition-care Index (prior stewardship) H3.1 0.078
**
0.086
**
0.086
**
0.088
**
Number of lots owned H3.2 0.039 0.059 0.028 0.039
Size of combined lot areas (area managed) H3.3 0.006 0.003 0.001 −0.001
Tax payments unpaid H3.4 −0.001 −0.001 −0.001 −0.001
Blotting 1.156
***
1.160
***
0.461
^
0.560
**
Proximity × owner-occupant H2.2 0.220*
Blotting × owner-occupant H4.1 0.902
**
Proximity × owner-occupant × blotting H4.2 0.257
***
Random effects
Intercept variance (owner ID) 0.557 0.541 0.525 0.517
AIC 1709.8 1707.7 1702.6 1699.7
Marginal R
2
0.272 0.279 0.288 0.293
N = 321.
^
p< .10, * p< .05,
**
p< .01,
***
p< .001.
Table 3
Sensitivity analysis for owner-level models predicting condition and care using binary-coded proximity and occupant type variables.
Variable/(concept) Hypothesis tested Model 1 (main) Model 2 Model 3 Model 4
Fixed effects
(Intercept) −0.323 −0.107 −0.134 −0.144
Time 0.623
***
0.624
***
0.623
***
0.623
***
Proximity binary H1 0.343
**
−0.068 0.351
**
0.183
Occupant type binary H2.1 0.097 −0.214 −0.147 −0.024
Owner condition-care Index (prior stewardship) H3.1 0.078
***
0.087
**
0.086
**
0.083
**
Number of lots owned H3.2 0.055 0.069 0.045 0.056
Size of combined lot areas (area managed) H3.3 0.009 0.004 0.004 0.004
Tax payments unpaid H3.4 −0.001 −0.001 −0.001 −0.001
Blotting 1.152
***
1.153
***
0.476
^
0.889
***
Proximity × occupant type H2.2 0.570
* a
Blotting × occupant type H4.1 0.881
** b
Proximity × occupant type × blotting H4.2 0.647
** c
Random effects
Intercept variance (owner ID) 0.550 0.534 0.519 0.526
AIC 1705.6 1702.8 1698.6 1700.3
Marginal R
2
0.276 0.284 0.291 0.288
N = 321.
^
p< .10, * p< .05,
**
p< .01,
***
p< .001.
a
Tukey adjusted post-hoc test was significant for the pairwise comparison between owner-occupant, far
and owner-occupant, close (p< 0.01).
b
Tukey adjusted post-hoc test was significant for three pairwise comparisons out of six (p< 0.05 – see Table A6).
c
Tukey
adjusted post-hoc test was significant for 14 pairwise comparisons out of 28 (p< 0.05 – see Table A6).
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
8
4.2. Block-level models
Table 4 reports the HLM model results for large lot care as a func-
tion of time and block-level indicators of care and disorder. The pre-
sence/absence of public/semi-public green space was significant
(p = .013) and the average number of cues to care per lot approached
significance (p = .095). With much of the variance in the marginal R
2
(0.107) provided by the time variable (as emerged in step-by-step
models, data not shown), the other variables in the model do not pro-
vide statistically significant support the block-level care hypothesis
(H5.1) that high levels of block care and low levels of disorder lead
owners to improve the condition of their large lots.
Finally, the percent of large lots sold on a block (Table 5) was
strongly associated with the average number of cues to care per lot and
the percent of block with vacant lots (both p= .001), and less strongly
associated with the number of big trees per lot (p = .013) and the
presence of green space on the block (p = .084). Together, the in-
dicators of block-level care and disorder explain a much larger pro-
portion of variance of the percentage of large lots sold (Table 5) than
the block-level model of large lot care (Table 4), lending stronger
support for the percent of lots sold hypothesis (H5.2) that high levels of
block care and low levels of disorder lead to a higher proportion of
large lot sales.
5. Discussion
5.1. Beyond proximity
That simple acts of cleaning and greening by individuals can inspire
others around them to do the same, for themselves and for their com-
munity, is a powerful idea and one that underlies the greening hy-
pothesis and its variants. Although investigators have described this
process in relation to private yards (Minor et al., 2016; Zmyslony &
Gagnon, 1998, 2000), easement gardens (Hunter & Brown, 2012), and
community gardening on vacant lots (Krusky et al., 2015), work to date
has provided only correlational evidence of these proximal relation-
ships. The conditions for purchasing vacant property under the Chicago
Large Lot Program provided us with the opportunity to look beyond
proximity, and knowledge of ownership, activity over time, and other
factors that we were able to link to improvements in the condition and
care of large lot purchases serve to extend the greening hypothesis
along several important dimensions.
Causal inference is an essential part of the greening hypothesis, and
support for causality is enhanced when research designs incorporate a
temporal sequence that aligns with the hypothesized cause/effect re-
lationship. As Krusky and colleagues (2015) recommended for future
research, knowledge of ownership is a key determinant of the agency
and directionality of greening activity, and because we knew who the
large lot purchasers were, we were able to attribute changes made after
the time of purchase directly to them. While a significant portion of
large lots in our sample showed signs of blotting prior to purchase, our
pre-post research design showed that levels of large lot condition and
care increased after purchase for both blotted and unblotted lots, fur-
ther strengthening the idea that ownership matters.
But our findings also revealed that ownership matters in complex
ways. Although proximity to the newly purchased large lot from the
owner’s original property was significantly associated with the level of
care extended (H1: proximity), our data showed no difference whether
owner-occupancy led to higher care than absentee ownership (H2.1:
occupant type). Furthermore, while the HLM models incorporating in-
teraction terms that included ownership were significant, had positive
coefficient signs, and showed improvements in model fit (H2.2, 4.1,
4.2), the post-hoc tests indicated that most of their significance could be
attributed to proximity and blotting rather than occupant type. In ad-
dition, the post-hoc tests showed that occupant type served to modify
the effect of these variables and clarify the conditions under which
proximity and blotting influenced large lot condition and care. In the
case of proximity, being immediately adjacent to a purchased large lot
resulted in a significantly higher level of care if the owner lived there
versus if it was in absentee ownership or vacant (see post-hoc tests for
H2.2). Likewise, large lots that were blotted before purchase showed
bigger improvements in condition and care if they were purchased by
owner-occupants than if they were in absentee ownership or vacant (see
post-hoc tests for H4.1). Further knowledge of these synergistic
Table 4
Block-level HLM model of large lot condition-care (H5.1).
Estimate Conf. Int. Std. Error df t p
Fixed effect
(Intercept) −0.423 −1.033 – 0.188 0.310 312 −1.361 0.174
Time 0.656 0.518 – 0.793 0.070 249 9.363 0.000
Presence public/semi-public greenspace 0.324 0.069 – 0.578 0.129 249 2.502 0.013
Average number of cues to care per lot 1.209 −0.207 – 2.625 0.720 249 1.679 0.095
% block with vacant lots 0.201 −0.676 – 1.079 0.446 249 0.451 0.652
% block unwalkable −0.148 −0.364 – 0.067 0.110 249 −1.351 0.178
Random effects
Intercept variance (block ID) 0.716
AIC = 1484.5, marginal R
2
= 0.107.
Table 5
Block-level OLS model of percent of large lots sold on block (H5.2).
Estimate Conf. Int. Std. Error t p
(Intercept) 0.531 0.331 – 0.732 0.102 5.222 0.000
Presence street calming features 0.034 −0.044 – 0.112 0.040 0.845 0.399
Presence public/semi-public greenspace 0.070 −0.009 – 0.148 0.040 1.738 0.084
Average number of big trees per lot 0.126 0.011 – 0.240 0.058 2.166 0.031
Average number of cues to care per lot 0.731 0.290 – 1.172 0.224 3.267 0.001
% block with vacant lots −0.442 −0.712 – −0.172 0.137 −3.228 0.001
Average number of cues to neglect per lot 0.045 −0.415 – 0.505 0.233 0.192 0.848
% block unwalkable −0.013 −0.080 – 0.054 0.034 −0.393 0.695
F= 8.626 (7, 241), p< .0000, R
2
adj = 0.177.
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
9
relationships could have important implications for vacant lot resale
programs, helping to ensure broad participation that maximizes the
equitable transfer of properties to residents and minimizes the risk that
the lots will become poorly managed (Armborst et al., 2008; Ganning &
Tighe, 2015).
The condition and care of the owner’s previously owned lot turned
out to be an important predictor of large lot condition and care
(H3.1: prior care), surpassing proximity in significance in our main
effects model and maintaining a high level of significance across all
interaction models. This finding supports research showing that prior
pro-environmental behavior can be a good predictor of future such
behavior if it is similar and specific (Harland et al., 1999; Oulette &
Wood, 1998). Because most applicants of vacant lot reuse programs
intend to undertake some sort of greening activity, at least in the early
years of ownership (Gobster et al., 2020), application forms could ask
the extent to which they have had prior experience in different greening
activities. This screening could be used by planners to estimate man-
agement capacity so that an appropriate level of assistance might be
provided. This would both help ensure that new lot owners will be able
to achieve their plans and help realize the overall success of the pro-
gram in revitalizing communities.
Finally, our measures of cost and effort added little to the prediction
of large lot condition and care, with no differences found on the number
of lots purchased, total area of lots owned, or tax payments (Hs
3.2–3.4). It could be that these variables insufficiently captured the
types of costs and effort described by behavioral theory (Moore &
Boldero, 2017), or it may be that the burden created by the number and
size of lots and by the level of tax payments did not exceed the
threshold where it would show significant reductions in condition and
care. With respect to the Large Lot Program, this latter point could be
further explored for those owners who have purchased additional lots
since the initial offering. For example, an owner who bought two lots in
the initial offering is theoretically eligible to purchase two more lots for
each lot owned in a second offering, and in fact some residents have
now accumulated several lots in their neighborhood this way. If large
accumulations of property by individuals leads to reduced upkeep or
land holding for future resale, these behaviors could suppress the re-
vitalization potential of the program.
5.2. From lots to landscapes
In order to understand relationships at the owned property and
block levels, large lot values were averaged for the number of owners
(N= 321) and blocks (N = 252). Because the change in large lot
condition and care was our primary dependent variable of interest, this
averaging was necessary but may have introduced error into our
models, particularly at the block level when several large lots were
purchased on a block. When we examined percent of lots sold on a
block as a second block-level dependent variable, the number of sig-
nificant indicators of care and disorder increased substantially, as did
the percentage of variance explained, and in comparing the perfor-
mance of the two models (Tables 4 and 5), the issue of error is a
plausible explanation.
High-resolution ground and aerial imagery and parcel-level data are
ideal for the type of fine-scale analysis of landscape change needed for
evaluation of the Large Lot Program and has been used in numerous
other studies of neighborhood quality of life (e.g., Rundle et al., 2011;
Ye et al., 2019). Like the work of Krusky et al. (2015), these fine-scale
assessments are often integrated with data from larger geographic areas
such as census tracts to draw more general conclusions consistent with
social-ecological health models (McLeroy et al., 1988) and to make
comparisons between neighborhoods and cities (e.g., Bader & Ailshire,
2014). Given the rapid distance decay of contagious and imitative be-
havior found in previous studies, we suggest that the block is the ap-
propriate scale of analysis to detect the adoption and diffusion of
greening practices. Although our model only weakly supported the
hypothesis that blocks with high levels of care and environmental
amenities like public green spaces spurs owners to improve their newly
purchased large lots (H5.1: block-level care), we believe further work is
warranted, including better ways to account for changes across a longer
time span.
The performance of our second block-level model using the percent
of large lots sold on the block is encouraging. Although there is still a
great deal of variance left unexplained, the predictor terms attained
respectable levels of significance and support our contention that high
levels of block care and the presence of environmental amenities would
lead to a higher proportion of available large lots being purchased
(H5.2: percent of lots sold). The finding is also encouraging in that
block-level care may affect owner actions both before and after pur-
chase, supporting the idea raised by Krusky et al. (2015) that greening
behavior may be bidirectional in nature. Together, both of our block-
level models deepen understanding of how actions undertaken by
property owners on individual lots contribute to the greater good of the
neighborhood, and how the neighborhood landscape can affect the
norms and behavior of individual lot owners. In each of these ways, the
neighborhood landscape, as seen by visitors from the street and ex-
perienced by residents within homes and yards, is a critical scale of
concern for research and planning (Sullivan, 2001).
6. Conclusions
In the context of the greening hypothesis initiated by Krusky et al.
(2015), our work supports a causal, directional, and temporal re-
lationship between several physical and behavioral factors and in-
creased vacant lot care. As a primary characteristic of nearby nature,
proximity expresses itself in numerous ways across the urban land-
scape, and figures prominently not only in patterns of residential
greening but also with respect to public access to greenspace
(Crompton, 2004), the provision of ecosystem services (Stessens, Khan,
Huysmans, & Canters, 2017), environmental justice concerns (Rigolon,
2017), and other issues. But by itself, proximity usually provides only
correlational evidence of attraction or repulsion between people and
green space, and without further knowledge about why proximity
matters, it is difficult to develop standards or guidelines to ensure that
people’s greenspace needs are met. Because our study design captured
urban landscape change across time and at different scales of measure,
we were able to document the impact of a vacant lot re-purposing
program and its potential for a catalytic effect on neighborhood re-
vitalization. It shows that prior stewardship as an individual variable
and occupant type as an effect modifier of proximity strengthen our
ability to predict changes in vacant lot condition and care. Continued
research efforts, using temporal and ownership data, examining pat-
terns across scales of analysis, and incorporating other social and en-
vironmental information, will strengthen our understanding of prox-
imal relationships in the landscape and help build more robust theories
of urban greening.
Acknowledgements
This work was supported in part through USDA Forest Service
Northern Research Station Cooperative Research Agreement 15-JV-
11242309-075 with the University of Illinois-Urbana-Champaign. The
authors thank Rose Grenen for research assistance, and Sonya Sachdeva
and Michelle Kondo and two anonymous reviewers for their helpful
comments in improving this paper.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.landurbplan.2020.103773.
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
10
References
Aarland, K., & Reid, C. K. (2019). Homeownership and residential stability: Does tenure
really make a difference? International Journal of Housing Policy, 19(2), 165–191.
https://doi.org/10.1080/19491247.2017.1397927.
Armborst, T., D’Oca, D., & Theodore, G. (2008). Improve your lot!. In K. Pallagst (Ed.).
Cities growing smaller (pp. 45–64). Cleveland, OH: Cleveland Urban Design
Collaborative, Kent State University.
Bader, M. D. M., & Ailshire, J. A. (2014). Creating measures of theoretically relevant
neighborhood attributes at multiple spatial scales. Sociological Methodology, 44,
322–368. https://doi.org/10.1177/0081175013516749.
City of Chicago. (2014a). Large lot program. Chicago: City of Chicago Department of
Planning and Development. https://www.cityofchicago.org/content/dam/city/
depts/dol/rulesandregs/DPDLARGELOTSPROGRAM.pdf.
City of Chicago. (2014b). Green healthy neighborhoods. Chicago: City of Chicago
Department of Planning and Development. https://www.cityofchicago.org/city/en/
depts/dcd/supp_info/green‐healthyneighborhoods.html.
City of St. Louis, MO. (n.d.). Mow to own program. https://www.stlouis-mo.gov/gov-
ernment/departments/sldc/real-estate/mow-to-own-program.cfm.
Cook County, IL. (n.d.). Cook County property tax portal. http://www.cookcountypro-
pertyinfo.com/.
Cook, E. M., Hall, S. J., & Larson, K. L. (2012). Residential landscapes as social-ecological
systems: A synthesis of multi-scalar interactions between people and their home
environment. Urban Ecosystems, 15, 19–52. https://doi.org/10.1007/s11252-011-
0197-0.
Crompton, J. L. (2004). The proximate principle: The impact of parks, open space and water
feature on residential property values and the property tax base (2nd ed). Asburn, VA:
National Recreation and Park Association.
Cuyahoga Land Bank. (n.d.). Side yard program. http://www.cuyahogalandbank.org/
sideYard.php.
Detroit Land Bank Authority. (n.d.). Side lot disposition policy. https://buildingdetroit.
org/our-policies/.
Dewar, M., Nassauer, J., & Dueweke, E. (2013). Cues to care database‐codebook.
Unpublished codebook for Detroit Sustainability Indicators Integrated Assessment.
Graham Environmental Sustainability Institute, University of Michigan, Ann
Arbor, MI.
Donovan, G. H., Landry, S., & Winter, C. (2019). Urban trees, house price, and re-
development pressure in Tampa, Florida. Urban Forestry and Urban Greening, 38,
330–336. https://doi.org/10.1016/j.ufug.2019.01.014.
Douglas, O., Russell, P., & Scott, M. (2019). Positive perceptions of green and open space
as predictors of neighbourhood quality of life: Implications for urban planning across
the city region. Journal of Environmental Planning and Management, 62(4), 626–646.
https://doi.org/10.1080/09640568.2018.1439573.
Ganning, J. P., & Tighe, J. R. (2015). Assessing the feasibility of side yard programs as a
solution to land vacancy in U.S. shrinking cities. Urban Affairs Review, 51(5),
708–725. https://doi.org/10.1177/1078087414555435.
Garvin, E., Branas, C., Keddem, S., Sellman, J., & Cannuscio, C. (2013). More than just an
eyesore: Local insights and solutions on vacant land and urban health. Journal of
Urban Health, 90, 412–426. https://doi.org/10.1007/s11524-012-9782-7.
Giner, N. M., Polsky, C., Pontius, R. G., Jr., & Runfola, D. M. (2013). Understanding the
social determinants of lawn landscapes: A fine-resolution spatial statistical analysis in
suburban Boston, Massachusetts, USA. Landscape and Urban Planning, 111, 25–33.
https://doi.org/10.1016/j.landurbplan.2012.12.006.
Gobster, P. H., Hadavi, S., Rigolon, A., & Stewart, W. P. (2020). Measuring landscape
change, lot by lot: Greening activity in response to a vacant land reuse program.
Landscape and Urban Planning, 196, 103729. https://doi.org/10.1016/j.landurbplan.
2019.103729.
Gobster, P. H., Ribe, R. G., & Palmer, J. F. (2019). Themes and trends in visual assessment
research: Introduction to the Landscape and Urban Planning special collection on the
visual assessment of landscapes. Landscape and Urban Planning, 191, 103635. https://
doi.org/10.1016/j.landurbplan.2019.103635.
Goddard, M. A., Dougill, A. J., & Benton, T. G. (2010). Scaling up from gardens:
Biodiversity conservation in urban environments. Trends in Ecology and Evolution,
25(2), 90–98. https://doi.org/10.1016/j.tree.2009.07.016.
Goldstein, J., Jensen, M., & Reiskin, E. (2001). Urban vacant land redevelopment: Challenges
and progress (Working paper WP01JG1). Cambridge, MA: Lincoln Institute of Land
Policy.
Greenland, S. (1996). Basic methods for sensitivity analysis of biases. International Journal
of Epidemiology, 25(6), 1107–1116. https://doi.org/10.1093/ije/25.6.1107.
Hajna, S., Dasgupta, K., Halparin, M., & Ross, N. A. (2013). Neighborhood walkability:
Field validation of geographic information system measures. American Journal of
Preventive Medicine, 44(6), e55–e59. https://doi.org/10.1016/j.amepre.2013.01.033.
Harland, P., Staats, H., & Wilke, H. A. M. (1999). Explaining proenvironmental intention
and behavior by personal norms and the theory of planned behavior. Journal of
Applied Social Psychology, 29, 12, 2505–2528. https://doi.org/10.1111/j.1559-1816.
1999.tb00123.x.
Heidelberg, R. L., & Eckerd, A. (2011). A challenge to the ownership society: Does home
ownership alone improve relative neighborhood quality? Innovation Journal, 16(1), 8.
Hunter, M. C. R., & Brown, D. G. (2012). Spatial contagion: Gardening along the street in
residential neighborhoods. Landscape and Urban Planning, 105, 407–416. https://doi.
org/10.1016/j.landurbplan.2012.01.013.
Kaplan, R. (1973). Some psychological benefits of gardening. Environment and Behavior, 5,
145–152. https://doi.org/10.1177/001391657300500202.
Kaplan, R., & Kaplan, S. (1989). The experience of nature: A psychological perspective. New
York: Cambridge University Press.
Kirkpatrick, J., Daniels, G., & Davison, A. (2009). An antipodean test of spatial contagion
in front garden character. Landscape and Urban Planning, 93, 103–110. https://doi.
org/1010.1016/j.landurbplan.2009.06.009.
Koprowska, K., Łaszkiewicz, E., & Kronenberg, J. (2020). Is urban sprawl linked to green
space availability? Ecological Indicators, 108, 105723. https://doi.org/10.1016/j.
ecolind.2019.105723.
Krekel, C., Kolbe, J., & Wüstemann, H. (2016). The greener, the happier? The effect of
urban land use on residential well-being. Ecological Economics, 121, 117–127 https://
doi.org/1010.1016/j.ecolecon.2015.11.005.
Krusky, A. M., Heinze, J. E., Reischl, T. M., Aiyer, S. M., Franzen, S. P., & Zimmerman, M.
A. (2015). The effects of produce gardens on neighborhoods: A test of the greening
hypothesis in a post-industrial city. Landscape and Urban Planning, 136, 68–75.
https://doi.org/10.1016/j.landurbplan.2014.12.003.
Kurz, T., & Baudains, C. (2012). Biodiversity in the front yard: An investigation of
landscape preference in a domestic urban context. Environment and Behavior, 44,
166–196. https://doi.org/10.1177/0013916510385542.
Land Bank of Kansas City, Missouri. (n.d.). Side lot program. http://www.kcmolandbank.
org/side-lot-program.html.
LargeLot.org. (n.d.). Large lots: The Chicago large lot program web portal. https://lar-
gelots.org.
Larson, K. L., Nelson, K. C., Samples, S. R., ... Trammell, T. L. E. (2016). Ecosystem ser-
vices in managing residential landscapes: Priorities, value dimensions, and cross-re-
gional patterns. Urban Ecosystems, 19, 95–113. https://doi.org/10.1007/s11252-015-
0477-1.
Lewis, C. A. (1979). Healing in the urban environment: A person/plant viewpoint. Journal
of the American Planning Association, 45, 300–338.
Maroko, A. R., Weiss Riley, R., Reed, M., & Malcolm, M. (2014). Direct observation of
neighborhood stressors and environmental justice in the South Bronx, New York City.
Population and Environment, 35, 477–496. https://doi.org/10.1007/s11111-013-
0197-5.
McCabe, B. J. (2013). Are homeowners better citizens? Homeownership and community
participation in the United States. Social Forces, 91(3), 929–954. https://doi.org/10.
1093/sf/sos185.
McLeroy, K. R., Bibeau, D., Steckler, A., & Glanz, K. (1988). An ecological perspective on
health promotion programs. Health Education Quarterly, 15, 351–377.
Minor, E., Belaire, A., Davis, A., Franco, M., & Lin, M. (2016). Socioeconomics and
neighbor mimicry drive yard and neighborhood vegetation patterns. In R. A. Francis,
J. D. Millington, & J.A. Chadwick (Eds.), Urban landscape ecology: Science, policy and
practice (pp. 56-74). London: Earthscan.
Mooney, S. J., Bader, M. D. M., Lovasi, G. S., Neckerman, K. M., Teitler, J. O., & Rundle,
A. G. (2014). Validity of an ecometric neighborhood physical disorder measure
constructed by virtual street audit. American Journal of Epidemiology, 180, 626–635.
https://doi.org/10.1093/aje/kwu180.
Moore, H. E., & Boldero, J. (2017). Designing interventions that last: A classification of
environmental behaviors in relation to the activities, costs, and effort involved for
adoption and maintenance. Frontiers in Psychology, 8, 1874. https://doi.org/10.3389/
fpsyg.2017.01874.
Nassauer, J. I. (1995). Messy ecosystems, orderly frames. Landscape Journal, 14(2),
161–170. https://doi.org/10.3368/lj.14.2.161.
Nassauer, J. I., Wang, Z., & Dayrell, E. (2009). What will the neighbors think? Cultural
norms and ecological design. Landscape and Urban Planning, 92(3–4), 282–292.
https://doi.org/10.1016/j.landurbplan.2009.05.010.
New Orleans Redevelopment Authority. (n.d.). Lot next door program. http://www.
noraworks.org/lotnextdoor.
Norwood, M.F., Lakhani, A., Fullagar, S., Maujean, A., Downes, M., Byrne, J.,… Kendall,
E. (2019). A narrative and systematic review of the behavioural, cognitive and
emotional effects of passive nature exposure on young people: Evidence for pre-
scribing change. Landscape and Urban Planning, 189, 71-79. https://doi.org/10.1016/
j.landurbplan.2019.04.007.
Oulette, J. A., & Wood, W. (1998). Habit and intention in everyday life: The multiple
processes by which past behavior predicts future behavior. Psychological Bulletin, 124,
54–74. https://doi.org/10.1037/0033-2909.124.1.54.
Philadelphia Land Bank. (n.d.). Side yards. http://www.philadelphialandbank.org/
Home/SideYards.
Pinheiro, J. C., & Bates, D. M. (2000). Mixed-effects models in S and S-Plus. New York, NY:
Springer.
R Core Team (2013). R: A language and environment for statistical computing. Vienna,
Austria: R Foundation for Statistical Computing.
Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data
analysis methods. Thousand Oaks, CA: Sage.
Rigolon, A. (2017). Parks and young people: An environmental justice study of park
proximity, acreage, and quality in Denver, Colorado. Landscape and Urban Planning,
165, 73–83. https://doi.org/10.1016/j.landurbplan.2017.05.007.
Roberts, H., McEachan, R., Margary, T., Conner, M., & Kellar, I. (2018). Identifying ef-
fective behavior change techniques in built environment interventions to increase use
of green space: A systematic review. Environment and Behavior, 50(1), 28–55. https://
doi.org/10.1177/0013916516681391.
Root, E. D., Silbernagel, K., & Litt, J. S. (2017). Unpacking healthy landscapes: Empirical
assessment of neighborhood aesthetic ratings in an urban setting. Landscape and
Urban Planning, 168, 38–47. https://doi.org/10.1016/j.landurbplan.2017.09.028.
Rundle, A. G., Bader, M. D. M., Richards, C. A., Neckerman, K. M., & Teitler, J. O. (2011).
Using Google Street View to audit neighborhood environments. American Journal of
Preventive Medicine, 40, 94–100. https://doi.org/10.1016/j.amepre.2010.09.034.
Sampson, R. J., & Raudenbush, S. W. (2004). Seeing disorder: Neighborhood stigma and
the social construction of “broken windows”. Social Psychology Quarterly, 67(4),
319–342. https://doi.org/10.1177/019027250406700401.
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
11
Staats, H., & Swain, R. (2020). Cars, trees, and house prices: Evaluation of the residential
environment as a function of numbers of cars and trees in the street. Urban Forestry
and Urban Greening, 47, 126554. https://doi.org/10.1016/j.ufug.2019.126554.
Stessens, P., Khan, A. Z., Huysmans, M., & Canters, F. (2017). Analysing urban green
space accessibility and quality: A GIS-based model as spatial decision support for
urban ecosystem services in Brussels. Ecosystem Services, 28, 328–340. https://doi.
org/10.1016/j.ecoser.2017.10.016.
Stewart, W. P., Gobster, P. H., Rigolon, A., Strauser, J., Williams, D. A., & van Riper, C. J.
(2019). Resident-led beautification of vacant lots that connects place to community.
Landscape and Urban Planning, 185, 200–209. https://doi.org/10.1016/j.landurbplan.
2019.02.011.
Sullivan, W. C. (2001). Does the neighborhood landscape matter? Landscape Journal, 20,
198–201. https://doi.org/10.3368/lj.20.2.198.
Thompson, M. L. (1978). Selection of variables in multiple regression: Part I. A review
and evaluation. International Statistical Review / Revue Internationale de Statistique,
46(1), 1.https://doi.org/10.2307/1402505.
Uren, H. V., Dzidic, B. J., & Bishop, B. J. (2015). Exploring social and cultural norms to
promote ecologically sensitive residential garden design. Landscape and Urban
Planning, 137, 76–84. https://doi.org/10.1016/j.landurbplan.2014.12.008.
Ye, Y., Richards, D., Lu, Y., Song, X., Zhuang, Y., Zeng, W., & Zhong, T. (2019). Measuring
daily accessed street greenery: A human-scale approach for informing better urban
planning practices. Landscape and Urban Planning, 191, 103434. https://doi.org/10.
1016/j.landurbplan.2018.08.028.
Zmyslony, J., & Gagnon, D. (1998). Residential management of urban front-yard land-
scape: A random process? Landscape and Urban Planning, 40, 295–307. https://doi.
org/10.1016/S0169-2046(97)00090-X.
Zmyslony, J., & Gagnon, D. (2000). Path analysis of spatial predictors of front-yard
landscape in an anthropogenic environment. Landscape Ecology, 15, 357–371.
https://doi.org/10.1023/A:1008160131014.
P.H. Gobster, et al. Landscape and Urban Planning 197 (2020) 103773
12
