Environmental Science and Policy 112 (2020) 293–304

Contents lists available at ScienceDirect

Environmental Science and Policy

journal homepage: www.elsevier.com/locate/envsci

Property price effects of green interventions in cities: A meta-analysis and T

implications for gentrification⋆
M. Bockarjovaa,*, W.J.W. Botzena,b, M.H. van Schiec, M.J. Koetseb
a
Utrecht University School of Economics (U.S.E), Utrecht University, Utrecht, the Netherlands
b
Institute for Environmental Studies, Vrije Universiteit, Amsterdam, the Netherlands
c
PBL, Netherlands Environmental Assessment Agency, The Hague, the Netherlands

ARTICLE INFO ABSTRACT

Keywords: Although green interventions, like nature-based solutions, contribute to more sustainable urban environments
Cities and provide ecosystem services to urban populations, some impacts are not well understood. This particularly
Ecosystem services applies to social impacts in the domain of environmental justice, including (green) gentrification. Gentrification
Hedonic pricing refers to a process in which green urban renewal raises property prices, which results in an influx of affluent
Nature-based solutions
people, displacing poorer residents. Our study conducts a meta-analysis based on 37 primary hedonic pricing
Benefit transfer
Property value
studies, to estimate value transfer functions that can assess the effects of nature types on property prices in
various urban settings. Urban nature has positive impacts on house value in the areas surrounding it, which
depend on population density, distance to, and the type of, urban nature. We illustrate how the estimated benefit
transfer function can be applied to natural interventions in a Dutch city, and visualize the obtained effects using
mapping. These maps show the distance decay of the cumulative effects of urban nature interventions on the
house value at the city and the neighbourhood levels. Our application estimated increases in local property
values up to a maximum of 20 % compared with properties not affected by the interventions, with value
equivalent of 62,650 USD, at average prevailing price level in a particular area in Utrecht. When new nature is
being planned in urban areas our mapping approach can be used for guiding assessments of potential undesirable
effects on property values that may lead to green gentrification, and for identifying where additional policies
may be needed to contribute to environmental justice.

1. Introduction which for example focus on ecosystem services, ecosystem-based

adaptation and mitigation, green and blue infrastructure, as well as
Population projections indicate that trends of increased urbaniza­ nature-based solutions1 . An advantage of urban green interventions
tion will continue (UN, 2018), which will increase pressure on the such as nature-based solutions is that they often provide multiple co-
urban environment. This highlights the importance of creating sus­ benefits (Raymond et al., 2017). For example, a park cools the city,
tainable urban living environments which are healthy, attractive and captures precipitation, limits air pollution, and contributes to biodi­
resilient to climate change (Estrada et al., 2017; Gill et al., 2007), but versity and recreation. Moreover, introducing nature to a city can make
also to take an environmental justice and sustainable development the city aesthetically more attractive and increase social cohesion.
perspective to urban development. This calls for moving towards gen­ However, there is insufficient knowledge on the way that different
erating, improving and maintaining social, economic and environ­ types of nature may affect other social domains such as gentrification.
mental justice by both scientists and practitioners. The latter refers to a process in which green urban renewal through the
A whole array of approaches is used to address urban sustainability, provision of ecosystem services creates added value on the property

⋆
Paper prepared for the special issue “Advancing urban ecosystem service implementation and assessment considering different dimensions of environmental
justice” in the journal Environmental Science and Policy edited by: Francesc Baró, Nadja Kabisch, Johannes Langemeyer, Edyta Łaszkiewicz.
⁎
Corresponding author.
E-mail address: [email protected] (M. Bockarjova).
1
Nature-based solutions are interventions aimed as positive responses to societal challenges that involve the innovative application of knowledge about nature,
inspired and supported by nature. These interventions can have the potential to simultaneously meet environmental, social and economic objectives (European
Commission, 2015). Examples are urban parks, forests, rivers, and lakes and combinations of so-called grey and green infrastructure, like sustainable drainage
systems, green roofs and green walls.

https://doi.org/10.1016/j.envsci.2020.06.024
Received 6 September 2019; Received in revised form 26 June 2020; Accepted 29 June 2020
Available online 08 July 2020
1462-9011/ © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

market as the effects of being located close to nature are capitalized in distinct advantage that transfer errors that occur at the local scales may
house value, which raises property and rental prices, and results in an average out at larger scales, potentially improving the accuracy and
influx of affluent people, displacing original poorer residents credibility of value transfer results. Further research on the validity of
(Anguelovski et al., 2018). Hence, social injustices may be produced this claim is warranted. Moreover, a meta-analysis can give insights
through ecosystem services-based urban policies and planning (Lange­ into overall patterns of results found in the literature of primary he­
meyer, this issue). This possibility of green gentrification needs to be donic pricing studies (Bateman and Jones, 2003), and enhance our
considered in urban planning since the most needy can be deprived of understanding of how different nature types, or green interventions,
ecosystem services and their benefits that urban nature offers influence property price developments.
(Hochstenbach, 2017; Hochstenbach and Musterd, 2018). Gentrifica­ Our study extends a previous meta-analysis of hedonic house price
tion is often associated with the United States, but green gentrification relations with urban open space by Brander and Koetse (2011) in three
in particular is not a foreign concept in Europe or elsewhere, with many ways. First, our study extends the types of urban nature (forest, park,
prominent cities seeing the departure of low-income residents in areas green space, undeveloped land, and agricultural land) to also include
with improved green conditions (Harris, 2008; Gould and Lewis, 2016; blue nature, such as lakes, rivers and canals for a more detailed ex­
Cole et al., 2019; Hochstenbach, 2017; Anguelovski et al., 2018). amination of how house prices relate with different urban nature types.
A disturbing development for urban planners and dwellers is that Second, including more (recent) studies increases the statistical power
gentrification is enhanced the past decades, with an increased percen­ of the analyses. Moreover, this update increases the number of included
tage of lower income neighborhoods being displaced in 2000s com­ countries and regions, which allows estimating regional value transfer
pared to the 1990s (Maciag, 2015). Being caused by a variety of un­ functions. Third, we illustrate the application of the derived value
derlying processes, gentrification has recently been associated with the transfer function to actual nature-based solution projects from a re­
increased suburbanization of low-to-middle-income earners in Europe, cently developed database called the Urban Nature Atlas (www.
and increased concentration of high-income earners in urban areas naturvation.eu/atlas), and derive implications for gentrification.
(Hochstenbach and Musterd, 2018). As an unintended consequence of Quantitative assessment tools are thus complemented in our approach
green urban interventions, green gentrification signals latent social by visualised analytical tools, such as mapping, to plot price effects of
processes that urban planners need to acknowledge, monitor and proximity to new green projects on the housing markets in cities. By
manage to ensure that benefits of urban nature and ecosystem services gaining a better understanding of the drivers of increased house prices,
provided by it can be universally promoted. Therefore, potential gen­ the issues associated with gentrification can be acknowledged, better
trification consequences of green renewal or other kinds of green in­ monitored and addressed. In this way, this paper presents the quanti­
terventions in residential areas need to be carefully considered before, fication and assessment of ecosystem services and their impacts on
during and after their introduction. Consideration should be given as to urban populations by urban nature through interdisciplinary methods
how environmental managers might focus activity and investment to applied to understand distributive justice (Baró et al., 2020), with a
balance environmental opportunities with the ongoing priorities of particular focus on property markets as signal of potential green gen­
delivering socially inclusive, ecologically rich and climate change-re­ trification.
silient green spaces (Bell et al., 2017). The remainder of this paper is structured as follows. Section 2 de­
The hedonic pricing method is often used to measure the effect of scribes the database and statistical methods. Section 3 presents the
environmental amenities on house prices. This method estimates the results of the value functions and discusses how they compare with
direct use value of nature to property owners as is embedded in house previous studies. Section 4 illustrates the applications of the derived
prices (for method beckground see Champ et al., 2003; for some recent value transfer functions to nature-based solution projects from the
applications see Sohn et al., 2020; Czembrowski and Kronenberg, 2016; Urban Nature Atlas. Section 5 concludes.
Schläpfer et al., 2015). This analysis provides insights into the value
local property owners attach to ecosystem services provided by nature 2. Data and methods
in cities as is reflected in a willingness-to-pay as a mark-up for prop­
erties that are located close to urban nature sites. Insights into the The data used for the meta-model estimation contains observations
benefits of nature-based solutions as well as the potential gentrification that are value points obtained from primary hedonic pricing studies,
consequences of introducing nature to an area need to be considered by evaluated at various distances from urban nature, which results in
policy makers and city planners (Lafortezza et al., 2018), for which a multiple value point observations per study. The meta-analysis pre­
better understanding of how house prices relate with different types of sented in this paper builds on the earlier meta-analysis of primary he­
urban nature is useful. donic price studies conducted by Brander and Koetse (2011) that ana­
In this study we perform a meta-analysis with the following objec­ lyzed the effects of urban green space on property prices. For reasons of
tives: to estimate relationships between property prices and different consistency and comparability we have followed the same procedure
types of urban nature; and to illustrate the applicability of the derived for literature search as Brander and Koetse (2007) and Brander and
value transfer function. Our study can provide first steps towards a Koetse (2011). In searching for studies specific key words were used
better understanding and future modelling of potential implications of which included three main components: valuation method, location,
gentrification through house market dynamics, to limit adverse social and the type of nature or ecosystem service. The resulting literature
impacts from the environmental justice and sustainable development search yielded a collection of papers that included 37 new hedonic
perspectives on urban development. Our meta-analysis is based on 37 pricing studies that analyze the effect of proximity to both green and
previously published hedonic pricing studies which estimated re­ blue open spaces on house prices in urban areas. These articles were
lationships between urban nature and house prices for specific areas. published between 2000 and 2017 and were not included in the ori­
Even though conducting a primary hedonic pricing study in a particular ginal meta-analysis. Detailed description of the database is found in the
area can give more reliable estimates for this particular area, such a Supplementary Material S1.
study is data intensive and time consuming. In case a detailed site- The meta-analysis presented in this study uses a multi-level model,
specific valuation study is not feasible, applying benefit transfer method in which value observations are at the first level and the primary study
and value transfer functions in particular can serve as a useful alter­ is the second level (see Bateman and Jones, 2003; Schmidt and Hunter,
native. Value transfer may also be useful for scaling up previous value 2004; Brander and Koetse, 2011). The idea behind this approach is that
estimates from local/regional to national or even continental levels, for there are characteristics in the context or in the methodological ap­
example to estimate the impact of certain interventions at a European proach that determine whether observations are clustered at some le­
scale. We claim that such an application of value transfer may have the vels. This means that clustered observations reveal systematic patterns

294
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

between each other according to a specific characteristic, and not with introduced. The reason is that it allows us to interpret the constant as
other observations in the sample. Observations included in the model the capitalized value of urban nature in property prices for the re­
are weighted with the inverse square of sample size used in the primary ference urban nature type category and at the average values of in­
study to account for the quality of primary estimates. dependent variables, i.e. the effect at the average evaluation point and
The estimated model is: over the average distance, for a location with average income and
density levels. Table 1 gives the coding of the variables and their means
yij = + SX S + MX M + TX T + DX D + µj +
ij ij ij ij ij (1) for the overall sample, as well as for subsamples for Europe and North
America. We note here that the difference between the variables
where i stands for the first level (observation) and takes values from 1 “evaluation point” and “distance” lies in the difference between spatial
to 803, and j stands for the second level (primary study) and takes and methodological variables. In particular, “evaluation point” reflects
values from 1 to 37. The dependent variable yij is based on the esti­ the average distance at which the valuation was performed in a primary
mated coefficients extracted from the primary valuation studies and
study, and is thus a spatial variable. At the same time, “distance” re­
transformed to our marginal effect size, i.e. the percentage change in
flects the average distance over which the valuation was performed in a
house price due to a decrease in distance to nature of 100 m. The error
primary study, is connected to the scale of effect measurement, and is
term is split into two components, µj and ij : the error term at the study
thus a methodological variable.
level and the error term at the observation level, respectively. Both
We estimate Eq. (1) for subsamples for Europe and North America in
error terms are assumed to have zero mean and to be uncorrelated,
order to identify different patterns that are specific to each of the value
meaning that they have constant variances σμ and σε. Model specifi­
functions of how urban nature affects property prices in these two re­
cation with regions at the second level has been estimated, but per­
gions. Such effect differences may arise due to various factors governing
formed worse in terms of model fit, and is therefore not presented here.
local preference for urban nature, for example its socio-cultural aspects
Regional dummies for Asia, Europe and North America were also tested
such as favoring specific types of urban nature, general abundance of
in order to control for the level of regional effect differences, but proved
nature in and surrounding the cities, and income level. These aspects
to be statistically insignificant, and were thus not included in the final
may influence both the level of willingness to pay for urban nature via
model. In the case of the primary study at the second level, it can be
property prices, and the relative significance of various factors con­
expected that value observations that come from the same study might
tributing to the value of urban property.
be closer to each other than values from other studies due to some in­
trinsic unobserved determinants which cannot be captured by the in­
3. Results of the hedonic pricing meta-analysis
dependent variables included in the model, such as for example similar
data sets or preference for a particular research method, specific to each
In this section we describe the results of the hedonic pricing meta-
primary study.
analyses (Table 2). Model 1 is estimated for the whole sample and in­
The independent variables are separated into four groups. The
cludes primary hedonic pricing studies from all over the globe, Model 2
vectors S , M , T and D , contain the estimated model coefficients for
is based on European studies, and Model 3 is based on North American
the variables included in XijS , XijM , XijT and , respectively.2 The vector XijS
(essentially, U.S.) studies. Our dependent variable is a relative in­
includes study and location variables, such as estimation point, as well
dicator, meaning that in absolute monetary terms the value of nature or
as GDP per capita and urban population density. We have used GDP per
its characteristics may differ dependent on prevailing house prices.
capita instead of for example purchasing power per capita as it is
Model 1 is estimated based on 803 value observations from 37
consistent with the common practice and the previously conducted
primary studies, Model 2 based on 159 value observations from 8 pri­
meta-analysis on the effect of urban green space on house prices
mary studies, and Model 3 on 562 value observations from 25 primary
(Brander and Koetse, 2011). It is also a common metric to correct for
studies. Variances at both levels are statistically significantly greater
price level in the estimated meta-model, where average effect values
than zero for all three models, which means that clustering of errors at
from individual primary studies are analyzed. The vector XijM includes
methodological variables, such a distance over which the effect was the primary study level contributes significantly to the explanation of
obtained in a primary study, and model specification. The vector XijT total variance. Due to its hierarchical structure, a multi-level model
contains variables that identify the type of urban nature or biome. The does not have a straightforward absolute fit indicator except for the log-
vector XijD contains the dummy representing landscape diversity of likelihood statistic. In this case the models cannot be compared in terms
urban nature. This is a new variable that is added to this meta-model, of model fit because they are estimated on different sets of data.
which measures whether an urban nature site contains two or more The constant in the three models measures the percentage change in
natural landscapes, such as green parkscape and waterscape. This house price due to moving 100 m closer to the peri-urban nature (re­
variable will capture the additional green premium, or price mark-up, ference group of urban nature type), when the effect is measured at
that is attributed to the landscape diversity of urban nature average or reference values of independent variables. That is, in Model
(Łaszkiewicz et al., 2019). 1 the constant measures the effect of a house price change of 0.220 %
Note, that in the way multiscape is defined in this study, it does not due to moving 100 m closer to peri-urban nature without a multiscape
reflect the quality level (including maintenance, aesthetic quality, etc.) feature, at a distance of 162 m (average evaluation point), measuring
of urban green and blue areas, their quantity or abundance of other the change over a distance of 154 m (average distance), at an average
green and blue areas in the vicinity of valued nature. GDP per capita (USD 43,011), for an average population density (810
For continuous independent variables such as evaluation point, inhabitants per km2), using a non-linear functional form. Respective
distance, income and population density, the centered logarithm is average values of independent variables for the interpretation of models
2 and 3 can be found in Table 1.
2
Among the location independent variables, evaluation point is ne­
Property prices are likely to be influenced by other variables, like the gative and statistically significant in all models 1–3, meaning that
availability of public transport, which could not be included in our analysis
houses at locations further from urban nature collect lower green pre­
because they were lacking in the primary valuation studies. Omitted variables
mium compared to houses located closer to nature, and indicating de­
would only bias our main coefficients of interest, which are the nature type
variables, in case they significantly influence the relation between nature and creasing returns to distance as the distance from urban nature increases.
property prices. For instance, this would occur if the availability of public Coefficients of income per capita are both negative (model 1) and po­
transport would change the relationship between property prices and the pre­ sitive (models 2 and 3) but in all cases statistically insignificant against
sence of nature. There is no reason to expect that this is the case systematically, expectations and earlier findings on the positive and statistically sig­
so we argue that our estimates are robust to these omitted variables. nificant relation between GDP per capita and WTP for urban nature

295
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

Table 1
Variable description and sample statistics (st.dev. in parentheses).
Variable Description Overal sample Europe-only North America-only
mean mean mean

N observations 803 159 562

Dependent variable
% change % change in property price due to 100 m decrease in distance to urban nature 0.924 1.497 0.823
(1.857) (2.317) (1.74)
minimum ; maximum −7.251; 13.550 −7.251; 13.118 −5.960; 13.55
Study and location variables:
Asia 1= primary study conducted in Asia, 0=otherwise 0.022 0 0
Europe 1= primary study conducted in Europe, 0=otherwise 0.198 1 0
North America 1= primary study conducted in North America, 0=otherwise 0.700 0 1
Evaluation point* Distance at which primary measurement took place 162 178 157
minimum ; maximum 1; 7116 1; 7116 1; 2912
GDP * GDP per capita in 2016 US dollars 43,012 42,321 45,958
minimum ; maximum 4,235; 55,747 13,145; 55,747 25,837; 52,983
Population density * Population density in number of people per km2 810 480 1006
minimum ; maximum 23; 6014 23; 4371 138; 6014
Methodological variables
Distance* Distance over which the effect was valued in primary study 154 143 168
minimum ; maximum 0.30; 2400 1; 1000 0.30; 2400
Linear 1= linear functional form used in primary study, 0=otherwise 0.087 0.321 0.034
Double-log 1= double-log functional form used in primary study, 0= otherwise 0.621 0.302 0.696
Semi-log 1= semi-log functional form used in primary study, 0= otherwise 0.267 0.377 0.235
Box-Cox 1= box-cox functional form used in primary study, 0= otherwise 0.025 0 0.036
Type of urban nature / biome:
Forest 1=effect of an urban forest is captured in primary study, 0=otherwise 0.136 0.252 0.101
Park 1= effect of an urban park is captured in primary study, 0=otherwise 0.328 0.289 0.329
Other urban green space 1= effect of other open green space is captured in primary study (such as 0.146 0.195 0.125
neighbourhood green spaces, pocket parks, green corridors), 0=otherwise
Blue 1= effect of urban blue nature is captured in primary study (such as lake, ponds, 0.255 0.075 0.326
rivers, streams, canals, urban sea coasts, wetland), 0=otherwise
Peri-urban 1= effect of peri-urban nature bordering on urban areas is captured in primary study 0.136 0.189 0.119
(such as undeveloped land, agricultural land, golf course), 0=otherwise
Lansdcape diversity
Multiscape 1=studied nature in the primary study resembles multiple landscape types (such as 0.412 0.572 0.331
green and blue), 0=otherwise

Note: * We do not report the standard deviation as these are average values at estimation, i.e. average ln(x).

(Brander and Koetse, 2011). The coefficient of the population density is the European sub-sample to avoid multicollinearity. The North
positive, which is in line with expectations from previous studies, but is American model reveals statistically significant and positive coefficients
not statistically different from zero in either of the three estimated only for urban blue nature, which compared to peri-urban nature re­
models. Current findings thus indicate independence of the relative sults in the effect size of 0.59 %. The multiscape dummy in model 3 has
weight of the green premium relative to distance in the composition of a positive coefficient (p-value < 0.063), again suggesting positive re­
property price throughout various levels of income and urban density. turns to diversity of urban nature.
A linear model specification in a primary hedonic pricing study We observe differences in the effects of different types of urban
(methodological variable) significantly adds to the average level of the nature on house prices in North America and Europe; urban blue nature
effect of urban nature on property prices in models 1 and 3, and acts as is appreciated on both continents, while urban forest and parks are
a methodological control variable. Distance over which the effect is particularly appreciated in Europe. We note that the American and the
estimated is not statistically different from zero at conventional sig­ European sub-samples differ; the population density and urban income
nificance levels in either of the models. This variable acts as a metho­ per capita have higher average levels in American cities compared to
dological control variable of decreasing returns to distance as the European ones (Table 1), which would be expected to result in stronger
measurement range increases in primary studies. effects of the presence of nature on house prices through scarcity and
All models in Table 2 feature a specification that includes 4 dum­ income mechanisms. However, our models show no statistically sig­
mies for urban biomes, or types of urban nature, such as urban forest, nificant association between urban density respectively income, and the
park, other green urban spaces, blue urban nature, in addition to peri- effect of proximity to urban nature on house prices. This suggests that
urban nature (reference category). These variables signify the different observed differences in estimated effects could be attributed to differ­
ways that types of urban green space are capitalized in house prices. ences in preferences for nature between American and European urban
Globally (model 1), urban parks (0.78 %) and blue nature (0.57 %) add residents.
statistically significantly to the value of urban property compared to
peri-urban nature, resulting in total in 1.00 % higher property values
4. Value function application for the case of Utrecht: European
for a park and 0.79 % for blue nature, ceteris paribus. The multiscape
Urban Nature Atlas and implications for gentrification
dummy in model 1 is a proxy for the diversity of urban nature, which
increases house prices by an additional 0.5 %.
In this section we illustrate the application of our value transfer
The European model reveals statistically significant positive coeffi­
functions. These functions can be used to estimate the influence of
cients for urban forest, urban park and blue nature, resulting in the
planned nature in cities on property prices. This way city planners can
expected increase in the house price due to moving 100 m closer to an
obtain insights into impacts of nature development projects on the
urban forest or park in a European city of 1.82 %, 2.06 % and 0.47 %,
prices and affordability of properties. Although our analysis does not
respectively. The multiscape variable was excluded from the model for
capture rental prices directly, it can be expected that increased property

296
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

Table 2 4.1. Value transfer function

Hedonic pricing meta-regression estimation results. The dependent variable is
the % change in house price due to 100 m decrease in distance to nature To illustrate the application of our estimated value transfer func­
(coefficients are reported with their t-statstics in parentheses). tion, we have made use of the Urban Nature Atlas (the NATURVATION
MODEL 1 MODEL 2 MODEL 3 project: www.naturvation.eu/atlas). It includes an extensive, albeit not
Whole sample Europe only North America exhaustive, selection of green urban initiatives and interventions across
(global) only Europe that have been completed, are being planned or are in im­
Constant 0.220 0.127 0.241
plementation in varying urban conditions, with wide differences in
(0.75) (0.16) (0.88) socioeconomic, ecological and geographic circumstances. The Urban
Location variables Nature Atlas includes intervention-specific information, such as the
Evaluation point (ln) −0.549*** −0.263** −0.581*** type of urban nature and landscape, ecosystem services provided by it,
(-4.28) (-2.03) (-3.04)
the type of intervention and institutional context, urban sustainability
GDP per capita (ln) −0.204 0.898 0.230
(-0.42) (0.58) (0.25) challenges, budget amount, etc. (Almassy et al., 2017). We illustrate the
Population density (ln) 0.170 0.096 0.232 application of our estimated value transfer functions for a Dutch city,
(1.03) (0.20) (1.24) Utrecht. with 10 green urban initiatives (described in the Online Sup­
Methodological variables
plementary Material S2), although the value functions are more broadly
Linear functional form 0.683*** 1.585 0.761***
(2.72) (1.57) (2.64)
applicable to other cities since they are based on global primary va­
Distance (ln) −0.031 −0.698* −0.024 luation studies. Utrecht was chosen as an illustrative example since it
(-0.47) (-1.82) (-0.39) experiences environmental pressures as well as ongoing population
Type of urban nature growth due to its central position in the country and rich cultural
(reference group: peri-urban nature)
heritage which attracts new inhabitants. Moreover, affordability of
Forest 0.526 1.821** 0.373
(0.93) (2.50) (1.18) housing for low-income households and gentrification is a concern of
Park 0.781*** 2.056*** 0.548 city planners in Utrecht (Utrecht Municipality, 2019).
(3.11) (5.67) (1.49) For the application, the global value transfer function based on es­
Other urban green space 0.504 −0.035 0.739* timated meta-regression model 1 is reported below (Eq. (2), results
(1.35) (-0.05) (1.66)
Blue nature 0.572** 0.471** 0.591**
from Table 2). LN stands for a natural logarithm, D stands for a dummy
(2.25) (2.48) (2.35) variable that takes a value of 1 if true, and 0 otherwise. We also recall
Lansdcape diversity that all continuous explanatory variables are centered logarithms of
Multiscape 0.500** 0.680* respective average values of independent variables as found in Table 1.
(1.97) (1.86)
Random variables
% change in house price due to 100m decrease in distance to nature
Estimate variance 1.096** 3.719** 0.882**
(study) (GLOBAL) = 0.220 – 0.549*(LN(Evaluation point) LN(162))–
(0.35) (2.84) (0.57)
Residual variance 2.046** 2.757** 1.776**
0.204*(LN(GDPpc ln c) LN(43.011))
(0.46) (0.61) (0.59) + 0.170*(LN(Population density) LN(810))
Estimation statistics
−2LL −1467 −318 −986 + 0.683*D(Linear model)– 0.031*(LN(Distance)
AIC 2960 660 1992
LN(154)) + 0.526*D(Forest)
VPC 0.349 0.574 0.332
N observations 803 159 562 + 0.781*D(Urban park)
N studies 37 8 25
+ 0.504*D(Other urban green) + 0.572*D(Blue)
*. **. *** indicate statistical significance at the 10 %. 5 %. and 1 % level. + 0.500*D(Multiscape) (2)
respectively.
Table 3 presents relevant variables for applying the value transfer
values are in the end passed on to renters in the form of higher renter function to nature interventions in Utrecht. Inner-city variability in
prices, although this depends on local conditions, such as price reg­ income can be quite substantial. In this application we have used city
ulations or quotas in rental markets. Price increases as a result of the average values for population density and per capita income, conform
development of nature in cities may under certain circumstances result
in gentrification, if lower-income households cannot afford the high Table 3
prices and are replaced over time by new residents with higher in­ Key explanatory variables for the city of Utrecht.
comes. Our value transfer functions can be used to raise awareness of Utrecht. Netherlands
this issue and serve as a first quick scan by showing whether high
a
property price increases can be expected in an area as a result of Average GDP per capita (USD) 53.558
Average house price (2017, USD) a,b 341.588
planned nature development, which may trigger a gentrification pro­ a,c
Average value of housing property (2017, USD) 260.850
cess. However, it should be realized that the occurrence of gentrifica­ Average population density (pers/km2) 3658
tion is a complex process that, among others, depend on the socio- Range of the evaluation point (m) 100−4000
economic conditions in an area, how the development process is con­
a
ducted, and what other policies are in place in a city that may limit Assumed prevailing exchange rate of 1 EUR = 1.11 USD.
b
gentrification processes, like social housing. If the quick scan applica­ Average house price is determined on the basis of all sold housing prop­
erties in a given year, within a municipality. Source: Statistics Netherlands
tion of the value transfer function shows the potential of substantial
(CBS, https://opendata.cbs.nl/statline/#/CBS/nl/dataset/83625NED/table?
property price increases, then additional examination should be con­
ts=1567154700379).
ducted to explore whether indeed gentrification will be problematic c
Average value of house price (WOZ. in Dutch) is the value that is de­
and how such adverse effects can be limited. Moreover, apart from termined by each municipality for all housing property within its adminis­
gentrification, our value functions show the value of urban green and trative borders for the purposes of taxation. It is based primarily on the market
blue through house prices, irrespective of the owner and use of the value of housing property (Waarderingskamer, 2010, 2017). Source: Statistics
house. Netherlands (CBS. https://opendata.cbs.nl/statline/#/CBS/nl/dataset/37610/
table?fromstatweb).

297
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

Fig. 1. City-level application: estimated extent of spatial impacts of multiple nature-based interventions on house value in Utrecht using model 1 (global value
transfer). Panel A: maximum effect for the overlapping values; panel B: the sum of effects for the overlapping values.
*Average housing values at the neighbourhood level in Utrecht for the base level of 2017, (CBS, 2017).

298
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

Fig. 2. Distance decay functions for the 10 urban nature interventions in the city of Utrecht. plotting the estimated cumulative effect of urban nature-based solutions
(nbs) on house value using model 1 (global value transfer), with the marked 1 percent threshold value.

the derivation of the value function in the meta-function. Information transfer (model 3) should be interpreted in the same way as the results
on the value of housing property on the neighborhood level was ob­ of the estimated meta-models in Section 3. For example, for interven­
tained from the Statistics Netherlands (Table 3). tion 6 (Maxima Park), houses that are found 300 m away from the park
would expectedly increase in value by 1.41 % if they would be found
200 m away from the park (see Table S3, column 3 in the Supplemen­
4.2. Effect visualisation
tary Material). However, marginal effect representation on a map is not
straightforward as it does not reflect the total effect of an intervention
For visualization purposes, we have mapped the data on the esti­
on property value at a specific location, which is often used for esti­
mated effects of the selected interventions on the housing market. To
mations of welfare changes. Alternatively, in such cases cumulative
our knowledge, this is a first use of mapping to such application where
effect is able to reflect such a total effect as it sums marginal effects
socio-economic impacts are related to the location of urban nature.
throughout the range at which a positive marginal effect of urban
Earlier uses of mapping in relation to urban green included a variety of
nature is estimated.
health effects of urban green and green interventions (Reid et al., 2009;
The maps on Fig. 1 thus portray cumulative effects of urban nature
Dadvand et al., 2012; Norton et al., 2015; Flacke et al., 2016), mental
interventions on property values for 10 interventions in the city of
health (Gascon et al., 2015), physical activity (Lwin and Murayama,
Utrecht, which included urban parks, small green patch parks and blue
2011; Brown et al., 2014), social need (McPhearson et al., 2013), crime
nature (canal renovation and sustainable drainage system). We notice
and violence (Gorham et al., 2009; Wolfe and Mennis, 2012). By
that different interventions have expectedly different effects on the
combining our classical representation of results, as shown in previous
house value, in particular in terms of the range at which this effect is
tables, with a visual application on a map, we can make a more in­
positive, as well as the size of the effect. For example, for urban parks
tegrated representation of the application results, at a city level. Such
like interventions 6 and 8, the positive effect stretches up to 3900 m
visualisation to our knowledge has not been done in previous meta-
from the intervention and reaches a maximum cumulative effect of 20
analysis applications. The background layer of the map on Figs. 1 and 3
% for properties found directly at the park compared to those found at
contains different shades of grey on building footprints to provide a
3900 m distance. This is essentially the effect of the park on the prop­
better visualization of urban fabric, and the average values of housing
erty value compared to the property where this effect is not positive, or
property at the neighbourhood level in Utrecht for the base level of
is not present anymore. For urban blue nature like interventions 1 and
2017, which are readily available from the Statistical Office (CBS,
2, the positive effect stretches up to 1000 m with the maximum cu­
2017). Cumulative effects (as explained below in this section) are dis­
mulative effect of 4.77 %, and for other urban green nature with a
played as aggregates on a a per-pixel basis, leading to a spatially con­
multiscape like intervention 4, the positive effect stretches up to
tinuous representation. Highways act as additional orientation land­
2300 m with the maximum cumulative effect of 11.64 %. Fig. 2 re­
marks in the city.
sembles the distance decay functions estimated for Utrecht interven­
It is possible that nature interventions in urban areas appear close to
tions, by type of intervention. Besides, Supplementary material contains
each other, and that their effects can therefore overlap. If this occurs,
Table S3 where application results for the interventions 3 through 6 are
assessments of an individual impact of each intervention may not be the
presented and include both marginal and cumulative estimated effects.
best way of understanding, presenting and analysing its impacts as they
The resulting differences in house value depend on the prevailing
are not able to show those overlapping impacts of nature on a neigh­
house value in the area, which differ throughout the city. For the case of
bourhood or a city level. First, we shall describe and interpret in­
the Maxima Park (intervention 6), the maximum cumulative effect of
dividual effects (marginal and cumulative) of a specific intervention on
20.05 % would translate into the expected difference in value of 62,656
the house value, which will be followed by a reflection on the value of
USD for the average house value of 312,400 USD in the area. At the
overlapping effects of multiple interventions on the house value on the
same time, for the City Island Ring Park (intervention 4), the maximum
neighbourhood level.
cumulative effect of 11.64 % would translate into the expected
Marginal effects resulting from the application of the global value

299
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

Fig. 3. Case-specific application: estimated effects of the City Island park ring (4) and the Maxima Park (6) with their respective shapes, plotting the cumulative effect
of urban green on house value using model 1 (global value transfer). Panel A: maximum effect for the overlapping values; panel B: the sum of effects for the
overlapping values.
*Average housing values at the neighbourhood level in Utrecht for the base level of 2017, (CBS, 2017).

300
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

difference in value of 22,535 USD for the average house value of 3, panel A), while addition of all overlapping effects would serve as its
193,600 USD in the area. For a green neighborhood square (interven­ upper bound (Figs. 1 and 3, panel B). For the same example, the effect
tion 3), the maximum cumulative effect of 4.11 % would translate into of the City Island Ring park (effect ring 6 of intervention 4 in Supple­
the expected difference in value of 8736 USD for the average house mentary material S3) in the Hoge Weide area would alone correspond
value of 212,300 USD in the area. to the expected increase of house value of 5.58 %, or 14,500 USD; the
As we pointed above, the effect ranges and effect sizes differ for additional effect of the Maxima Park (effect ring 15 of intervention 6,
various green and blue interventions as we have applied them to the 10 idem) would add extra 5.67 % to the expected increase of house value
cases in the city of Utrecht. Fig. 2 shows the distance decay functions and total 33,380 USD, at average prevailing value in the area of
for all these cases, marking the maximum cumulative effects and the 289,300 USD per property.
spread of their respective ranges. We note that due to the functional
form of the global value transfer function, some depicted distance decay 5. Discussion
functions have a long tail when approaching zero. For example, the
estimated positive cumulative effect of interventions 6 and 8 (urban 5.1. Value transfer function and its limitations
parks) stretches as far as 3900 m, which may seem unrealistic. Marking
the 1 % threshold for the cumulative effect, below which the effect on In this section we address main limitations associated with the re­
the house value may essentially be seen as negligible, may provide a so- sults of our estimated value transfer function and its application. In
called effective range of cumulative effects. In such a case, the effective general, a meta-analysis can only take into account variables that are
range of the cumulative effect for urban green and blue interventions common to all primary valuation studies that are included in the ana­
drops to 2900 m for the urban parks (nr. 6 and 8), and to 500 m for lysis. As such, a first possible limitation of our study is that is that the
urban blue nature (nr. 1 and 2). Such an approach to interpretation of set of explanatory variables included in our model is restricted. For
value transfer function application results may offer more realistic in­ example, relevant excluded variables could include additional con­
dications of actual value changes at the city level. textual or spatial variables, such as area sizes of studied nature areas or
However, as Fig. 1 shows, the cumulative effects of each particular availability of other nature areas that may be used as substitutes for the
intervention overlap in many instances, even for the selection of 10 nature areas under study. These variables were not available for all
interventions which are taken as illustrations in this particular case, primary studies in our sample and can therefore not be included in the
with other interventions in a city. This means that the overall resulting presented meta-models. Systematic inclusion of such additional in­
effect on the house value may be even higher than the effect of a single formation in future primary valuation studies should not only enrich
intervention. Besides, some of the interventions cover a bigger area, the primary hedonic pricing studies, but will strengthen the evidence
which means their point representation on a map is not realistic. To base for meta-analyses and thus improve the benefit transfer applica­
explore this, we have zoomed into the intervention 4, the City Island tions by better informing practitioners and policy-makers on the ex­
Ring park, where a continuous park ring is being created along the pected effect of green policies in urban areas.
canals surrounding the City Island of Utrecht. Fig. 3 shows the shape of Another limitation of hedonic pricing studies is the lack of in­
the City Island and its park (in green), and its effect sizes throughout the formation from transactions data on specific ecosystem services that
surrounding area. In this particular case, the effects stretch both within homeowners are willing to pay for when they pay a higher price for a
the City Island, and outside the City Island. We have also depicted in­ property that is closely located to nature. This means that without ad­
terventions 3 and 5 to illustrate the overlap of a major intervention 4 ditional data from each primary study it is not possible to explicitly
with minor interventions at a neighborhood scale. For example, houses pinpoint and hence control for ecosystem services in our analysis. We
at the third ring of intervention 4 (the City Island Ring park), within the added Supplementary Tables S4.1 and S4.2 which show the distribution
City Island would expectedly increase in value by 8.54 %. However, the of ecosystem services that were mentioned in primary valuation studies,
presence of intervention 5 (Food for Good garden) may add extra value as assumedly being related to the type of nature examined in that study.
to those houses as they overlap with the first ring of intervention 5. These tables show that a broad range of ecosystem services were as­
Adding both overlapping effects together at this location would result in sociated with the specified nature types including: (1) provisioning
the expected increase in property value of 12.66 %, or 25,153 USD for services, like raw materials from forests, (2) regulating services, like
the average house value in the area (Fig. 3, panel B). Another example flood and climate regulation from blue nature, and (3) habitat services,
of overlapping effects at this local scale is the cumulative effect of the like biodiversity from parks. The five most frequently reported services
major urban park in Utrecht, Maxima Park (intervention 6), which we are: recreation, aesthetic appreciation, noise reduction, flood regulation
have also depicted with its actual shape. In this way, the non-trivial and water purification.
effects of the Maxima Park overlap with the non-trivial effects of the Although our case study in Utrecht is merely meant as an illustra­
City Island Ring Park, which was not evident at the city level (Fig. 1A, tion to demonstrate general applicability of the method, it also reveals a
the two interventions do not seem to overlap). As Fig. 3B shows, there is trade-off that this method poses in terms of scale. While a city level
an extensive area where the cumulative effects of interventions 4 and 6 application offers a quick scan of effects of multiple nature interven­
overlap in the non-trivial range. For example, in the area where the tions, an application at a local level may offer more precise estimates
yellow effect rings of both interventions overlap, the properties located especially if the actual shape of interventions can be depicted. Here, the
in this area may experience a double effect on their value, adding up to practitioner has to realise that obtained effect estimates are originating
10–11 % in total. It is important to notice here that theory does not from averaged affects and studies collected all over the world, without
provide a rule of thumb concerning value transfer function applications specific reference to urban morphology or the housing market.
and overlapping effects of multiple urban green and blue areas as in the Therefore, it is important to stress that our approach can be used for
cases discussed above. Essentially, it is natural to assume that vicinity of applications in other EU cities, and more widely for assessing the values
multiple urban nature areas would add more to the property value in a of urban green and blue at national or even continental scales. We
particular location than vicinity of a single piece of nature. However argue that such scaling up may have the added advantage that transfer
simple addition of individual cumulative effects of multiple interven­ errors that occur at local scales may average out at larger scales, which
tions should be overestimating the true effect, due to the decreasing would improve the accuracy and credibility of value transfer results. As
returns to scale and, thus, not only due to the distance decay, but also far as we could find, empirical work on this particular issue is lacking,
due to the amenity satiation effect. We further discuss this point in and future studies are needed to look into the validity of this claim.
Section 5. We thus suggest that the highest of the overlapping effects Our estimates of the (dollar) value changes may be considered quite
would serve as a lower bound for the total effect estimation (Figs. 1 and conservative. This is due to the fact that while the indicator of average

301
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

property value is related to house sales prices, it lags behind the on­ financing a purchase of a 322,680 USD house. Moreover, rental prices
going developments in the housing market which may bring about and house prices are connected to the value of house property. This
greater changes in the spot sales prices. When new nature is being implies that increasing house prices are reflected in higher rental prices
planned in urban areas our mapping approach can be used for assessing that may become unaffordable for lower-income households, in parti­
potential undesirable property price effects that may lead to green cular for younger adults who often have to rely on the financial support
gentrification, and for identifying where additional policies may be from their parents even after moving out of their parents’ home
needed to limit such effects, like social housing. (Hochstenbach, 2018; Amsterdam Municipality, 2019).
Simple addition of multiple overlapping effects may lead to over- or It should be noted that the development of urban nature is often not
underestimation of values. As table S1 reveals, 18 out of 37 primary the main cause of problems with affordability of housing for low-in­
studies used in our meta-analysis included multiple nature types in the come households (Amsterdam Municipality, 2019), but the effects of
vicinity of housing properties, all of which were assumed to simulta­ nature on property value and property prices can exacerbate such
neously influence the purchasing price. This means that if primary problems. This does not mean that additional nature should not be
studies have modelled the availability of other natural areas in the developed, since it brings various benefits that are clearly valued by
study site, they should have identified unique effects of each of those people as is reflected in their willingness to pay higher prices for
multiple types of nature on the house price. This implies that effect sizes property close to nature. It does mean that possibly undesirable effects,
of home properties obtained in the meta-analysis can be attributed to a like green gentrification, should be assessed when new nature is being
particular type of urban nature, independent of the presence of other developed in urban areas, and additional policies may be needed to
types of urban nature in the vicinity of property of interest. Because we limit such effects. Mapping approach such as multi-layering of data can
have not explicitly modelled substitution in our meta-analysis due to be even more helpful for policy and practice offering an integrated
data restrictions, value transfer should lead to value underestimation approach to the solution of urban challenges. For example, combining
(overestimation) when the policy site has a lower (higher) availability city maps of nature intervention impact on the house value with ad­
of substitutes compared to the average availability of substitutes of ditional information, such as physical and mental health status of city
study sites in our meta-analysis sample. More empirical work on the residents or prevailing environmental conditions (Flacke et al., 2016;
effects of substitute availability is needed, especially in the context of Gascon et al., 2015) as well as crime and violence levels (Gorham et al.,
meta-analysis and value transfer. 2009; Wolfe and Mennis, 2012) may better guide city planners on the
Finally, our example serves purely illustrative purposes, and appli­ decisions concerning the best areas for green interventions as well as
cations for other locations can be made depending on valuation needs signal problematic areas for potential green gentrification. Possible
and data availability. While value transfer based on the European and solutions could be the development of social housing in existing green
the North American functions (models 2 and 3, respectively) can be neighbourhoods in order to promote social and environmental inclusion
derived in a similar way, we note that the European function based on and keep neighbourhoods accessible for low-income families. More­
model 2 should be treated with particular caution for benefit transfer over, a more even provision of green spaces throughout the city would
purposes as it appears to overestimate the effects of most types of ensure equal benefits for all inhabitants.
nature on house prices, judging by the fact that the value function
suggests that value effects of urban nature spread over unrealistically 6. Conclusion
long distances. In that respect the predictive validity of the global value
transfer function is substantially higher. Therefore, for all applications, Ecosystem services provided by nature interventions, such as
the global value transfer (model 1) is suggested as the preferred func­ nature-based solutions, present an opportunity for cities to offset pro­
tion at the moment. blems from increased urbanization and climatic changes. By combining
natural characteristics with grey infrastructure. urban nature creates
5.2. Using the value transfer functions to identify potential gentrification benefits for citizens and visitors, and results in a more aesthetically
issues pleasing city. However, the economic and social impacts of natural
interventions remain insufficiently considered in urban ecosystem ser­
Our application of the value transfer functions illustrates how dif­ vices assessments (Marshall and Gonzalez-Meler, 2016). Nature-based
ferent green interventions have a variety of effects on the housing solutions often impact the housing market with rising house and rental
market, and under which circumstances large effects on house value prices. As a result, and most often unintentionally, lower income po­
can occur. Neighbourhoods in the vicinity of multiple green interven­ pulations become displaced by better-off inhabitants, and thus become
tions may experience higher percentage increases in house value due to deprived of the advantages of urban nature that are of particular benefit
the cumulative effects. Clearly, the absolute amount of house value to them (Lovell et al., 2018). Such processes of green gentrification
changes depends on local conditions, such as prevailing house value in need close monitoring that can be facilitated by an assessment of the
the area, which may vary substantially within a city. Our mapping impact of (newly) developed green and blue areas on housing markets.
analysis can be used to visualise in which areas high percentage in­ This study conducted a meta-analysis to estimate value transfer
creases in house value can be expected due to the creation of additional functions that can be used for assessing the impacts of urban nature on
nature in cities, and show whether this occurs in areas where house house prices at different distances from the nature site. Compared to a
value are already high. This information can be used to visualise the previous study (Brander and Koetse, 2011) our updated meta-analysis
quantified information on the expected effects on the housing market, includes more types of urban nature and more observations from dif­
and serve as a quick scan for identifying areas where green gentrifica­ ferent countries, which among others allows for estimating regional
tion may occur (Hochstenbach. 2018), although gentrification is a benefit transfer functions. The results show that urban nature has an
complex process for which additional examination would be needed to impact on house prices in the areas surrounding it, and that the mag­
determine its severity and possible solutions. For instance, in some nitude of this effect decreases as house distance from nature increases,
neighbourhoods the development of nature can increase value of revealing conventional distance decay relationship. Furthermore, the
housing where the average house value is already high. In such a results show that the impact on property prices differ by type of nature
neighbourhood, a gentrification process can already be ongoing and intervention. In particular, homebuyers value the presence of a park or
improved green infrastructure may further accelerate it. As in the ex­ blue nature in the vicinity of their property more the presence of other
ample above of Hoge Weide area above, a yearly gross income of ap­ types of urban nature. This effect can be explained by the high direct
proximately 71,700 USD, which is about twice modal income in the use value (Hein et al., 2006) created by many ecosystem services of
Netherlands, would be necessary to be able to obtain a mortgage for urban parks and blue nature such as aesthetics, recreational

302
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

opportunities and local climate regulation. References

In our application of the European value transfer function to green
interventions in a Dutch city of Utrecht, we have used maps to overlay Almassy, D., Pinter, L., Rocha, S., Naumann, S., Davis, M., 2017. Urban Nature Atlas: A
the effects of these interventions on the housing market with the Database of Nature-Based Solutions across 100 European Cities. H2020 Naturvation
Project Report. (Accessed on 20 Aug 2019). https://naturvation.eu/result/results-
average value of housing property in the city. This information can be urban-nature-atlas.
used for identifying areas where green gentrification may occur. These Amsterdam Municipality, 2019. Programmaplan Huisvesting Kwetsbare Groepen 2019 -
maps show the distance decay of the cumulative effects of urban nature 2022. Gemeente Amsterdam. 14 juli 2019. www.amsterdam.nl/woonagenda (in
Dutch).
interventions on the house value at the city and the neighbourhood Anguelovski, I., Connolly, James J.T., Masip, Laia, Pearsall, Hamil, 2018. Assessing green
levels. We estimated increases in local property values up to a max­ gentrification in historically disenfranchised neighborhoods: a longitudinal and
imum of 20 % compared with properties not affected by the interven­ spatial analysis of Barcelona. Urban Geogr. 39 (3), 458–491. https://doi.org/10.
1080/02723638.2017.1349987.
tions, with value equivalent of 62,650 USD, at average prevailing price Baró, F., Kabisch, Nadja, Langemeyer, Johannes, 2020. Edyta Łaszkiewicz (____).
level in a particular area in Utrecht. Besides, overlapping effects of Introduction to the special issue on environmental justice. Environ. Sci. Policy cur­
nature may in specific neighbourhoods double the increases in property rent SI.
Bateman, Ian J., Jones, Andrew P., 2003. Contrasting conventional with multi-level
values from 5 to 6% to 10–11 %. We recognize that simple addition of
modeling approaches to meta-analysis: expectation consistency in UK woodland re­
individual cumulative effects of multiple interventions should be creation values. Land Economics 79 (2), 235–258. https://doi.org/10.2307/
overestimating the true effect, and thus suggest that the highest of the 3146869.
overlapping effects could serve as a lower bound for the total effect Bell, S.L., Westley, Michael, Lovell, Rebecca, Wheeler, Benedict W., 2017. Everyday green
space and experienced well-being: the significance of wildlife encounters. Landscape
estimation, while addition of all overlapping effects would serve as its Res. 43 (1), 8–19. https://doi.org/10.1080/01426397.2016.1267721.
upper bound. Further research should shed light into the dynamics of Brander, L.M., Koetse, M.J., 2007. The Value of Open Space: Meta-Analyses of Contingent
multiple overlapping effects of urban nature. Valuation and Hedonic Pricing Results. IVM Working Paper 07/03. Institute for
Environmental Studies, VU University, Amsterdam.
Our analysis showed that the presence of urban nature has a distinct Brander, L.M., Koetse, M.J., 2011. The value of urban open space: meta-analyses of
positive impact on housing prices in the areas surrounding it. This in­ contingent valuation and hedonic pricing results. J. Environ. Manage. 92 (10),
sight is useful for monitoring the societal dimension of green inter­ 2763–2773. https://doi.org/10.1016/j.jenvman.2011.06.019.
Brown, G., Schebella, M.F., Weber, D., 2014. Using participatory GIS to measure physical
ventions, tailoring policy and helping stakeholders build understanding activity and urban park benefits. Landscape Urban Plann. 121, 34–44. https://doi.
of the environmental, economic, and social impacts of green urban org/10.1016/j.landurbplan.2013.09.006.
interventions. Future research on nature-based solutions in cities and CBS, 2017. Neighbourhood Level Statistics 2004–2018. (Accessed on 20 Aug 2019).
https://www.cbs.nl/nl-nl/dossier/nederland-regionaal/wijk-en-buurtstatistieken.
their impacts on house prices can focus on nature interventions and the Champ, P.A., Boyle, K.J., Brown, T.C., 2003. A Primer on Nonmarket Valuation. The
process of gentrification using integrated urban justice assessments, Economics of Non-Market Goods and Resources Vol. 3https://doi.org/10.1007/
because planning and implementation of nature interventions require 97894-007-0826-6.
Cole, H.V.S., Triguero-Masa, Margarita, Connollya, J.T., Anguelovskia, Isabelle, 2019.
integration of multiple data and balancing stakeholder priorities and
Determining the health benefits of green space: does gentrification matter? Health
interests. Doing this could provide more direct insights into the po­ Place 57, 1–11. https://doi.org/10.1016/j.healthplace.2019.02.001.
tential societal impacts that nature-based solutions can create, in ad­ Czembrowski, P., Kronenberg, J., 2016. Hedonic pricing and different urban green space
dition to assessing benefits that they bring to cities on an ecological or types and sizes: insights into the discussion on valuing ecosystem services. Landscape
Urban Plann. 146, 11–19. https://doi.org/10.1016/j.landurbplan.2015.10.005.
aesthetic level. Dadvand, P., de Nazelle Figueras, A.F., Basagaña, X., Su, J., Amoly, E., Jerrett, M.,
Vrijheid, M., Sunyer, J., Nieuwenhuijsen, M.J., 2012. Green space, health inequality
and pregnancy. Environ. Int. 40, 110–115. https://doi.org/10.1016/j.envint.2011.
CRediT authorship contribution statement 07.004.
Estrada, F., Botzen, W.J.W., Tol, R., 2017. A global economic assessment of city policies to
reduce climate change impacts. Nat. Clim. Change 7, 403–406. https://doi.org/10.
M. Bockarjova: Conceptualization, Methodology, Software, 1038/nclimate3301.
Validation, Formal analysis, Investigation, Data curation, Writing - European Commission, 2015. Towards an EU Research and Innovation Policy Agenda for
Nature-Based Solutions & Re-Naturing Cities. Report of the Directorate-General for
original draft, Writing - review & editing. W.J.W. Botzen:
Research and Innovation. EU, Brussels. https://doi.org/10.2777/765301.
Conceptualization, Methodology, Validation, Writing - original draft, Flacke, J., Schüle, S.A., Köckler, H., Bolte, G., 2016. Mapping environmental inequalities
Writing - review & editing, Supervision, Project administration, relevant for health for informing urban planning interventions—a case study in the
City of Dortmund, Germany. Int. J. Environ. Res. Public Health 13 (7), 711. https://
Funding acquisition. M.H. van Schie: Software, Writing - review &
doi.org/10.3390/ijerph13070711.
editing, Visualization. M.J. Koetse: Conceptualization, Methodology, Gascon, M., Triguero-Mas, M., Martínez, D., Dadvand, P., Forns, J., Plasència, A.,
Validation, Writing - review & editing. Nieuwenhuijsen, M.J., 2015. Mental health benefits of long-term exposure to re­
sidential green and blue spaces: a systematic review. Int. J. Environ. Res. Public
Health 12 (4), 4354–4379. https://doi.org/10.3390/ijerph120404354.
Gill, S.E., Handley, J.F., Ennos, A.R., Pauleit, S., 2007. Adapting cities for climate change:
Declaration of Competing Interest the role of the green infrastructure. Built Environ. 33 (1), 115–133. https://doi.org/
10.2148/benv.33.1.115.
Gorham, M.R., Waliczek, T.M., Snelgrove, A., Zajicek, J.M., 2009. The impact of com­
The authors declare that they have no known competing financial munity gardens on numbers of property crimes in urban Houston. HortTechnology 19
interests or personal relationships that could have appeared to influ­ (2), 291–296. https://doi.org/10.21273/HORTSCI.19.2.291.
ence the work reported in this paper. Gould, K.A., Lewis, T.L., 2016. Green Gentrification: Urban Sustainability and the
Struggle for Environmental Justice. Routledge, London. https://doi.org/10.4324/
9781315687322.
Harris, A., 2008. From London to Mumbai and back again: gentrification and public
Acknowledgments policy in comparative perspective. Urban Stud. 45 (12), 2407–2428. https://doi.org/
10.1177/0042098008097100.
This research received funding from the Horizon 2020 Framework Hein, L., van Koppen, K., de Groot, R.S., van Ierland, E.C., 2006. Spatial scales, stake­
holders and the valuation of ecosystem services. Ecol. Econ. 57 (2), 209–228. https://
Programme of the European Union, NATURVATION project (grant doi.org/10.1016/j.ecolecon.2005.04.005.
number 730243). We are also indebted to Nikola Lipovac and Alberto Hochstenbach, C., 2017. Inequality in the Gentrifying European City. PhD thesis.
Oeo for their assistance in database compilation. University of Amsterdam ISBN 978-94-91602-95-5.
Hochstenbach, C., 2018. Spatializing the intergenerational transmission of inequalities:
parental wealth, residential segregation, and urban inequality. Environ. Plann. A:
Econ. Space 50 (3), 689–708. https://doi.org/10.1177/0308518X17749831.
Appendix A. Supplementary data Hochstenbach, C., Musterd, S., 2018. Gentrification and the suburbanization of poverty:
changing urban geographies through boom and bust periods. Urban Geogr. 39 (1),
Supplementary material related to this article can be found, in the 26–53. https://doi.org/10.1080/02723638.2016.1276718.
Lafortezza, R., Chen, Jiquan, van den Bosch, CecilKonijnendijk, Randrup, T.B., 2018.
online version, at doi:https://doi.org/10.1016/j.envsci.2020.06.024.

303
M. Bockarjova, et al. Environmental Science and Policy 112 (2020) 293–304

Nature-based solutions for resilient landscapes and cities. Environ. Res. 165, Reid, C.E., O’Neill, M.S., Gronlund, C.J., Brines, S.J., Brown, D.G., Diez-Roux, A.V.,
431–441. https://doi.org/10.1016/j.envres.2017.11.038. Schwartz, J., 2009. Mapping Community determinants of heat vulnerability. Environ.
Łaszkiewicz, E., Czembrowski, P., Kronenberg, J., 2019. Can proximity to urban green Health Perspect. 117 (11), 1730–1736. https://doi.org/10.1289/ehp.0900683.
spaces be considered a luxury? Classifying a non-tradable good with the use of he­ Schläpfer, F., Waltert, F., Segura, L., Kienast, F., 2015. Valuation of landscape amenities: a
donic pricing method. Ecol. Econ. 161, 237–247. https://doi.org/10.1016/j. hedonic pricing analysis of housing rents in urban, suburban and peri-urban
ecolecon.2019.03.025. Switzerland. Landscape Urban Plann. 141, 24–40. https://doi.org/10.1016/j.
Lovell, R., Depledge, M., Maxwell, S., 2018. Health and the Natural Environment: A landurbplan.2015.04.007.
Review of Evidence, Policy, Practice and Opportunities for the Future. Defra Report. Schmidt, F.L., Hunter, J.E., 2004. Methods of Meta-Analysis: Correcting Error and Bias in
Project Code BE0109. Department for Environment, Food and Rural Affairs, UK. Research Findings, second edition. Sage publicationshttps://doi.org/10.4135/
http://randd.defra.gov.uk. 9781483398105.
Lwin, K.K., Murayama, Y., 2011. Modelling of urban green space walkability: eco-friendly Sohn, W., Kim, H.W., Kim, J.H., Li, M.H., 2020. The capitalized amenity of green infra­
walk score calculator. Comput. Environ. Urban Syst. 35 (5), 408–420. https://doi. structure in single-family housing values: an application of the spatial hedonic pricing
org/10.1016/j.compenvurbsys.2011.05.002. method. Urban For. Urban Green., 126643. https://doi.org/10.1016/j.ufug.2020.
Maciag, M., 2015. “Gentrification in America Report,” Governing the States and 126643.
Localities. . February http://www.governing.com/gov-data/census/gentrification-in- UN, 2018. UN World Urbanisation Prospects 2018. (Accessed on 20 Aug 2019). https://
cities-governing-report.html. population.un.org/wup/.
Marshall, K.A., Gonzalez-Meler, M.A., 2016. Can ecosystem services be part of the solu­ Utrecht Municipality, 2019. Woonvisie: Utrecht Beter in Balans. Gemeente Utrecht. juli
tion to environmental justice? Ecosyst. Serv. 22, 202–203. https://doi.org/10.1016/j. 2019. https://omgevingsvisie.utrecht.nl/thematisch-beleid/wonen/ (in Dutch).
ecoser.2016.10.008. Waarderingskamer, 2010. Market Value as a Basis for Determination of Real Estate Value.
McPhearson, T., Kremer, P., Hamstead, Z.A., 2013. Mapping ecosystem services in New Netherlands Value Assessment Chamber, the Hague. https://www.
York City: applying a social–ecological approach in urban vacant land. Ecosyst. Serv. waarderingskamer.nl/fileadmin/publieksportaal/documents/public/klopt-mijn-woz-
5, 11–26. https://doi.org/10.1016/j.ecoser.2013.06.005. waarde/Brochure_Marktwaarde_als_waarderingsgrondslag.pdf.
Norton, B.A., Coutts, A.M., Livesley, S.J., Harris, R.J., Hunter, A.M., Williams, N.S.G., Waarderingskamer, 2017. Memo on Model-Based Determination of Real Estate Value. 21
2015. Planning for cooler cities: a framework to prioritise green infrastructure to August 2017. Netherlands Value Assessment Chamber, the Hague. https://www.
mitigate high temperatures in urban landscapes. Landscape Urban Plann. 134, waarderingskamer.nl/fileadmin/publieksportaal/documents/public/klopt-mijn-woz-
127–138. https://doi.org/10.1016/j.landurbplan.2014.10.018. waarde/Notitie_modelmatige_waardebepaling_2017-08-21.pdf.
Raymond, C.M., Frantzeskaki, N., Kabisch, N., Berry, P., Breil, M., Nita, M.R., Calfapietra, Wolfe, M.K., Mennis, J., 2012. Does vegetation encourage or suppress urban crime?
C., 2017. A framework for assessing and implementing the co-benefits of nature- Evidence from Philadelphia, PA. Landscape Urban Plann. 108 (2–4), 112–122.
based solutions in urban areas. Environ. Sci. Policy 77, 15–24. https://doi.org/10. https://doi.org/10.1016/j.landurbplan.2012.08.006.
1016/j.envsci.2017.07.008.

304