# RECORD 43d34a3cb07ebe23b958a009b21ab082b8ff0225
# TITLE: Canals vs. Streams: To What Extent Do Water Quality and Proximity Affect Real Estate Values? A Hedonic Approach Analysis
# YEAR: 2016.0  VENUE: nan  DOI: 10.3390/w8120577
# AUTHORS: Federico Bonetti; S. Corsi; L. Orsi; I. De Noni
# PAGES: 17  CANDIDATE TABLE PAGES (1-based): [8, 11, 14]
# ABSTRACT: The presence of and proximity to water influences human settlement decisions, because water has been able to characterize the environment and affect human health since prehistoric times. Many scholars have studied the effects of environmental amenities on the real estate market; however, only a few of them have explored the effects of water. This study stresses the role of water with a focus on the dichotomy between artificial and natural waterways (canals vs. streams), which involves managerial and qualitative differences. A hedonic approach, based on a large web scraped and geo-referenced dataset of 10,530 observations, is used to analyze the effect of water proximity and quality on residential housing prices in the province of Milan (Italy). Our main results show a significant but conflicting impact of water proximity on real estate value: reducing the canal distance, we observe an increase in the residential housing sale price, in urban more than peri-urban environments; in contrast, reducing the stream distance, our results show decreases in property value. Moreover, a significant moderating effect of water quality on distance from streams is identifiable.


===== PAGE 1 (score 14, numeric-paren 0, table-words 1, var-words 12) =====

water
Article
Canals vs. Streams: To What Extent Do Water Quality
and Proximity Affect Real Estate Values? A Hedonic
Approach Analysis
Federico Bonetti, Stefano Corsi, Luigi Orsi * and Ivan De Noni
Department of Economics, Management and Quantitative Methods, University of Milan, via Conservatorio 7,
20122 Milan, Italy; federico.bonetti2@studenti.unimi.it (F.B.); stefano.corsi@unimi.it (S.C.);
ivan.denoni@unimi.it (I.D.N.)
* Correspondence: luigi.orsi@unimi.it; Tel.: +39-2-5032-1542
Academic Editor: Richard C. Smardon
Received: 6 September 2016; Accepted: 30 November 2016; Published: 6 December 2016
Abstract: The presence of and proximity to water inﬂuences human settlement decisions, because
water has been able to characterize the environment and affect human health since prehistoric
times. Many scholars have studied the effects of environmental amenities on the real estate market;
however, only a few of them have explored the effects of water. This study stresses the role of
water with a focus on the dichotomy between artiﬁcial and natural waterways (canals vs. streams),
which involves managerial and qualitative differences. A hedonic approach, based on a large web
scraped and geo-referenced dataset of 10,530 observations, is used to analyze the effect of water
proximity and quality on residential housing prices in the province of Milan (Italy). Our main results
show a signiﬁcant but conﬂicting impact of water proximity on real estate value: reducing the canal
distance, we observe an increase in the residential housing sale price, in urban more than peri-urban
environments; in contrast, reducing the stream distance, our results show decreases in property value.
Moreover, a signiﬁcant moderating effect of water quality on distance from streams is identiﬁable.
Keywords: water quality and proximity; real estate value; hedonic model; environmental impact
1. Introduction
In rural areas, i.e., outside the city, water is a critical element because it is strongly related to
agricultural utilization. In contrast, in fragmented and highly urbanized areas, water is able to produce
several positive amenities, such as preserving natural ecosystems [1], mitigating the impacts caused by
human activity, such as air pollution and heat islands [2–5] and providing aesthetically pleasing open
spaces and recreational sites [6–8]. However, in spite of these positive amenities, some researchers
have speciﬁcally focused only on the aptitude of water to leverage the real estate market by inﬂuencing
property values.
In this light, the effect of surface water bodies on residential housing prices has been speciﬁcally
related to different features such as water proximity, security and quality. Walsh et al. (2011) [9]
identiﬁed the signiﬁcance of water body proximity on the value of residential properties in Orange
County, Florida. Positive amenities from proximity to water are also reported by Anderson and West
(2006) [10]. Tapsuwan et al. (2009) [11] argue that the value of houses increases by reducing the
distance to the wetland. Cohen et al. (2014) [12] show conﬂicting ﬁndings for proximity to wetlands
but conﬁrm the positive effect of water bodies’ proximity. Cho et al. (2006) [13] suggest the inﬂuence of
proximity to creeks and streams in Tennessee. Nelson et al. (2005) [14] highlight the positive effect of
proximity to artiﬁcial water bodies. Other studies focus on water views rather than spatial proximity
and suggest surface water is able to increase the waterfront property value [15].
Water 2016, 8, 577; doi:10.3390/w8120577
www.mdpi.com/journal/water


===== PAGE 8 (score 133, numeric-paren 126, table-words 1, var-words 5) =====

Water 2016, 8, 577
8 of 17
Table 3. Correlation matrix.
Variables
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
1 price
1
2 dist_stream
−0.03 ***
1
3 quality_stream
0.06 ***
0.00
1
4 dist_canal
−0.18 ***
−0.07 ***
−0.04 ***
1
5 quality_canal
0.02 *
0.00
−0.17 ***
−0.01
1
6 energ_clas
−0.07 ***
−0.05 ***
−0.01
0.01
−0.03 **
1
7 sup_mq
0.74 ***
0.01
0.04 ***
−0.03 ***
0.02 *
−0.08 ***
1
8 cond_fees
0.42 ***
−0.03 **
0.02 *
−0.05 ***
0.02 *
0.01
0.29 ***
1
9 kitch_type
0.24 ***
0.01
0.03 **
0.03 ***
0.00
0.04 ***
0.42 ***
0.12 ***
1
10 elevat
0.13 ***
−0.09 ***
0.02 *
−0.04 ***
0.01
−0.14 ***
−0.01
0.11 ***
0.07 ***
1
11 balcony
0.04 ***
−0.02 *
−0.01
0.07 ***
0.01
0.02 *
0.12 ***
0.05 ***
0.21 ***
0.08 ***
1
12 box
0.08 ***
0.10 ***
−0.01
0.12 ***
0.02 *
−0.23 ***
0.29 ***
0.02 *
0.18 ***
0.01
0.08 ***
1
13 furnish_type
0.04 ***
−0.02 *
0.06 ***
−0.05 ***
0.01
0.04 ***
−0.05 ***
−0.01
−0.10 ***
−0.01
−0.08 ***
−0.08 ***
1
14 prop_status_1
−0.01
0.00
0.00
−0.01
0.01
−0.02 *
0.00
0.00
−0.01
0.01
0.02 *
0.00
−0.01
1
15 prop_status_2
0.01
0.02 *
0.04 ***
−0.01
0.01
0.01
0.00
0.00
0.00
−0.02 *
−0.01
−0.01
0.01
−0.31 ***
1
16 prop_status_3
−0.02
−0.03 **
−0.04 ***
0.02 *
−0.02 *
0.01
0.00
0.00
0.01
0.01
0.00
0.01
0.00
−0.26 ***
−0.70 ***
1
17 prop_status_4
0.03 **
0.01
0.00
−0.01
0.00
−0.01
0.01
0.00
0.00
0.01
0.00
0.00
0.00
−0.10 ***
−0.26 ***
−0.21 ***
1
18 dist_sub_st
−0.22 ***
0.31 ***
−0.02 *
0.24 ***
−0.01
−0.10 ***
0.00
−0.12 ***
0.02
−0.18 ***
0.05 ***
0.32 ***
−0.10 ***
0.00
−0.01
0.01
−0.01
1
19 dist_tr_st
−0.15 ***
0.16 ***
−0.05 ***
0.05 ***
−0.01
−0.08 ***
0.02 *
−0.07 ***
0.04 ***
−0.10 ***
0.06 ***
0.26 ***
−0.03 **
0.01
−0.02
0.02
−0.01
0.26 ***
1
20 dist_mall
−0.17 ***
0.23 ***
−0.04 ***
0.11 ***
0.01
−0.09 ***
0.01
−0.09 ***
0.00
−0.18 ***
0.02 *
0.25 ***
−0.06 ***
0.01
−0.02 *
0.01
0.00
0.42 ***
0.45 ***
1
21 green_area
−0.02 *
−0.01
−0.02 *
0.05 ***
0.00
0.02 *
0.02 *
0.03 ***
0.06 ***
0.06 ***
0.04 ***
0.03 **
−0.01
0.00
−0.02 *
0.02 *
0.01
−0.10 ***
0.04 ***
−0.01
1
Notes: The signiﬁcance levels are *** p < 0.001, ** p < 0.01, * p < 0.1.


===== PAGE 9 (score 13, numeric-paren 0, table-words 1, var-words 11) =====

Water 2016, 8, 577
9 of 17
Territorial Variables
The territorial variables (all territorial and environmental variables calculated are based on data
provided by the GeoPortal of the Lombardy region) are the shortest distance, i.e., “as the crow ﬂies”,
from subway stations (dist_sub_st is the distance in meters from the subway station or the nearest
urban rail station; it is the main line of urban displacement), from train stations (dist_tr_st is the
minimum distance in meters from the train station) and shopping malls (dist_mal) is the minimum
distance in meters from the shopping center. The green area (we used the classiﬁcation “Parchi
e giardini” of the Lombardy region’s Dusaf 4.0) (green_area_300 m) represents a radius of 300 m
in diameter from the center of the building in the analysis. This variable allows us to assess how
much green area in square meters is available within a walking distance of 300 m from the property.
Finally, the geographical location of the property using the 96 zip codes (zip_codes = dummy variables
for all the zip codes in the Milan province) to take account of spatial heterogeneity correlated with
cross-jurisdictional variations in the housing market in the overall area.
The value of territorial variables is calculated by punctuated information bases indicating,
for example, the location of Metro stations, the location of shopping malls, etc.
3.2.2. Explanatory Variables
Environmental Variables
Our paper focuses on the evaluation of the impact of water proximity and quality on
housing prices.
The environmental variables’ distance from streams (dist_stream represents distance in meters
from the nearest streams) and distance from canals (dist_canal represents distance in meters from the
nearest canal) represent the shortest distance, i.e., “as the crow ﬂies”, from streams and canals to assess
the impact of a water stream proximity on housing prices.
Stream quality (quality_stream) and canal quality (quality_canal) measure the quality of water on
a 5-point Likert scale using the ARPA Lombardy water quality report for the period 2009–2014 [28].
This 5-point Likert scale indicator is calculated by combining biological quality elements, general
physical–chemical elements and speciﬁc water pollutants (1 = bad water quality, 2 = poor, 3 = sufﬁcient,
4 = good and 5 = excellent water quality). Finally, we have linked together each single property with
the quality of the nearest stream and canal and with the nearest ARPA detection point.
In addition, to evaluate any potential moderating effect of water quality and proximity,
we
introduce
two
interaction
variables:
stream
distance
multiplied
by
stream
quality
(dist_stream*quality_stream) and canal distance multiplied by canal quality (dist_canal*quality_canal).
To achieve this goal, we standardized the two variables prior to calculating their interaction terms,
in order to avoid unnecessary multi-collinearity.
3.2.3. Dependent Variable
Utilizing the website of a consenting large Italian real estate agency, we obtained data for each
property sale in the province of Milan. The house price information (price) is a bid price because the
proposed purchase price is not based on any real negotiation of sale.
3.3. Hedonic Price Model
After we analyzed the literature, we chose the hedonic price model for our analysis [27,30,33–35].
The hedonic price model consists of a semi-logarithmic least squares hedonic model that is ascribable
to the following general formula [34]:
LnPi = β0 + β1Si + β2Ni + β3Qi + "i
(1)


===== PAGE 11 (score 160, numeric-paren 129, table-words 6, var-words 19) =====

Water 2016, 8, 577
11 of 17
(Model 5). We used the address variable to divide the database by the position of the property, based on
whether the property was in the city of Milan or in another town of the province.
Table 4. Regression models results.
Dependent Variable:
House Price (log)
OLS and Spatial Regression Models
Model 1 Base
Model OLS
Model 2 Full
Model OLS
Model 3 Full Model
Spatial-Lag (SL)
Model 4 SL Full
Model Milano
Model 5 SL Full
Model Province
Intercept
11.996 (0.021) ***
12.024 (0.022) ***
7.323 (0.082) ***
11.414 (0.059) ***
12.282 (0.026) ***
Rho (spatial lag)
0.406 (0.006) ***
0.080 (0.004) ***
0.023 (0.002) ***
Explanatory variables
dist_stream
0.012 (0.005) *
0.012 (0.003) ***
0.058 (0.005) ***
0.024 (0.004) ***
quality_stream
0.005 (0.003) *
0.008 (0.003) **
0.011 (0.005) *
0.010 (0.004) **
dist_stream*quality_stream
−0.004 (0.003) *
−0.005 (0.003) *
−0.001 (0.005)
−0.006 (0.003) *
dist_canal
−0.020 (0.005) ***
−0.074 (0.003) ***
−0.071 (0.006) ***
−0.003 (0.003)
quality_canal
−0.001 (0.003)
0.002 (0.003)
−0.007 (0.005)
0.000 (0.004)
dist_canal*quality_canal
−0.001 (0.003)
−0.003 (0.003)
−0.013 (0.005) *
−0.003 (0.003)
Control variables
energ_clas
−0.042 (0.003) ***
−0.042 (0.003) ***
−0.046 (0.003) ***
−0.030 (0.005) ***
−0.048 (0.004) ***
sup_mq
0.405 (0.003) ***
0.405 (0.003) ***
0.439 (0.004) ***
0.531 (0.006) ***
0.350 (0.004) ***
cond_fees
0.006 (0.003) *
0.006 (0.003) *
0.019 (0.003) ***
0.010 (0.005) *
0.025 (0.004) ***
kitch_type
0.083 (0.006) ***
0.083 (0.006) ***
0.064 (0.007) ***
0.103 (0.012) ***
0.052 (0.008) ***
elevat
0.115 (0.006) ***
0.115 (0.006) ***
0.154 (0.007) ***
0.206 (0.013) ***
0.077 (0.008) ***
balcony
0.011 (0.006) *
0.012 (0.006) *
−0.019 (0.007) **
0.008 (0.011)
−0.043 (0.008) ***
garage
0.169 (0.006) ***
0.168 (0.006) ***
0.121 (0.007) ***
0.070 (0.012) ***
0.205 (0.008) ***
furnish_type
−0.006 (0.006)
−0.006 (0.006)
0.017 (0.007) *
0.036 (0.011) **
−0.023 (0.008) **
prop_status_2
−0.079 (0.009) ***
−0.079 (0.009) ***
−0.045 (0.011) ***
0.015 (0.020)
−0.100 (0.012) ***
prop_status_3
−0.213 (0.010) ***
−0.213 (0.010) ***
−0.175 (0.012) ***
−0.153 (0.021) ***
−0.221 (0.012) ***
prop_status_4
−0.310 (0.013) ***
−0.310 (0.013) ***
−0.248 (0.016) ***
−0.205 (0.024) ***
−0.400 (0.021) ***
year_dummy_2012
−0.013 (0.009)
−0.013 (0.009)
0.004(0.011)
0.012 (0.017)
−0.020 (0.012) *
year_dummy_2013
−0.013 (0.008)
−0.013 (0.008)
−0.018 (0.010) *
−0.024 (0.016)
−0.026 (0.011) *
year_dummy_2014
−0.005 (0.007)
−0.007 (0.007)
−0.009 (0.008)
0.000 (0.013)
−0.009 (0.009)
year_dummy_2015
−0.030 (0.009) **
−0.030 (0.009) **
−0.038 (0.011) **
−0.054 (0.016) **
−0.015 (0.014)
dist_sub_st
−0.115 (0.007) ***
−0.120 (0.007) ***
−0.119 (0.004) ***
0.005 (0.007)
−0.096 (0.004) ***
dist_tr_st
−0.028 (0.005) ***
−0.029 (0.006) ***
−0.029 (0.004) ***
−0.053 (0.005) ***
−0.026 (0.004) ***
dist_mall
−0.001 (0.005)
−0.003 (0.005)
0.018 (0.004) ***
−0.066 (0.005) ***
0.003 (0.004)
green_area
0.004 (0.003)
0.004 (0.003)
−0.016 (0.003) ***
−0.017 (0.005) **
0.020 (0.003) ***
zip_codes
Yes
Yes
Yes
Yes
Yes
No. of observations
10,530
10,530
10,530
5318
5212
F-stat
562.5 ***
535.9 ***
adj. R-squared
0.858
0.861
LogLik
−623.86
−608.94.13
−421.38
−573.05
−291.49
AIC
1099.8
1277.1
393.9
Max VIF
2.755
2.859
2.859
2.810
2.396
Mean VIF
1.298
1.321
1.321
1.289
1.278
Notes: The coefﬁcients are mean centering standardized values. The standard errors are in parentheses.
The signiﬁcance levels are *** p < 0.001, ** p < 0.01, * p < 0.1. Zip codes are not reported due to spatial constraints.
Model 1 estimates the coefﬁcients of the control variables. The model results allow us to state
that energy class, subway stations’ distance and railway stations’ distance have a negative and
highly signiﬁcant (p < 0.001) inﬂuence on the house sale price. In contrast, the house surface area,
the habitable living space, kitchen, presence of an elevator, balcony and/or garage, status of the
property and the condominium fees have a positive and signiﬁcant impact. Speciﬁcally, the positive
effect of condominium fees may be related to more luxury services or improving conditions of the
property. Finally, the distance from the mall, the house furnishings and the presence of green areas
within a walking distance of 300 m are not statistically signiﬁcant (p > 0.10) on house prices. The zip
codes variable (not reported in Table 4 due to spatial constraints) shows higher and signiﬁcant values
for housing prices in the zip codes close to Milan city center, while in certain suburbs of Milan and,
especially, in the peripheral province areas, it shows a negative or no signiﬁcant impact on house
prices. Finally, the year dummy variables introduced to control for a potential time trend show a slight
negative trend with respect to the reference year (2011) but no signiﬁcant statistical differences among
years except for 2015. We also tested different time periods (semesters and quarters) and found no
signiﬁcant time trends.
The overall ﬁt of the models increases compared to the baseline, as Models 2, 3 and 4 ﬁt our
data better and have more explanatory power than Model 1. Moreover, the coefﬁcients and signs of


===== PAGE 12 (score 13, numeric-paren 0, table-words 1, var-words 11) =====

Water 2016, 8, 577
12 of 17
the control variables remain quite stable along the different models, showing robust results and that
multi-collinearity is not a particular problem in these regressions.
In Model 2, we include all the explanatory variables related to water proximity and quality and
their interaction effect. The results show a signiﬁcant (p < 0.10) and positive effect of the streams’
distance on house prices. The more we move away from the rivers, the higher the price of houses.
As expected, the quality of the water streams positively and signiﬁcantly affects housing prices and
mitigates the negative effect of the streams’ distance. This latter result is conﬁrmed by the negative and
signiﬁcant sign of the interaction effect (dist_stream*quality_stream). In contrast, we ﬁnd a positive
and highly signiﬁcant impact of canal proximity on property value, while the quality of the canals and
the interaction terms with the distance are non-signiﬁcant.
Based on the results of Moran I and Lagrange Multiplier (LM) tests, in Model 3, a spatial
dependence technique is introduced. Spatial lag is used to estimate the regression coefﬁcients.
However, AIC and log-likelihood tests suggest that spatial-lag and also spatial-error models are
better than OLS models, because of the spatial dependence structure in our data (see Moran and LM
tests in the previous section).
Model 3 (spatial lag model) highlights that the residential housing price is a spatial dependence
phenomenon (lag = 0.405, p-value < 0.001). The results are very close to the linear model. No changes
in terms of sign and signiﬁcance are identiﬁable across the coefﬁcients of control variables between the
linear and spatial lags models; the unique exceptions are the presence of a balcony, which changes the
sign; the presence of green areas; the types of furnishings; and the distance from malls, which become
signiﬁcant. The exploratory variables concerning water quality and proximity keep the same sign and
signiﬁcance. However, in the spatial model, coefﬁcients are more robust because they take spatial bias
into account.
The different responses of explanatory variables that indicate the distance from natural water
bodies and canals can be attributed to the fact that proximity to the ﬁrst is considered a negative
function of poor water quality. However, the inﬂuence of canals is positive in relation to the better
water quality and to the increase in environmental conditions in heavily anthropic areas.
This theory is strongly supported by studies carried out by the public authorities appointed to
monitor the water quality. The results come from a study carried out by ARPA of the Lombardy region
on the natural streams basin of the rivers Lambro and Olona for the period from 2009 to 2014.
The results of this work indicate that a state of stress exists for the entire basin with an alteration
to the rivers’ cleaning capacity. In addition, more than 50% of the areas analyzed reveal contamination
by heavy metals, and a complete analysis of the areas shows a bad or poor water quality, especially in
reference to the presence of bad odors and dirty water [28].
In contrast, the canals that derive from rivers such as the Ticino and Adda have the highest
average water quality compared to other waters in the province [41]. The high water quality is a result
of actions taken by the Ticino and Adda parks authorities. This fact, combined with the maintenance
that is performed on the canals periodically by land reclamation authorities to ensure the proper
hydraulic features, explains the results from the hedonic model application.
To support this thesis, we refer to the “Carta Provinciale delle Vocazioni Ittiche” [42] that,
in addition to reporting the status of the ﬁsh and fauna of the natural streams and canals, also lists
the chemical and physical conditions of the water. These are synthesized by an indicator called the
environmental status of watercourses (SECA) that allows us to deﬁne the status of water through the
pollution level, the status of the biological communities, the ecosystems and the presence of dangerous
substances and bio accumulation.
There is chemical, physical and biological quality impairment of the main natural streams of the
province, such as Seveso, Lambro and Olona, that worsens along the route of the rivers. However,
for rivers such as the Ticino and Adda, the judgement on chemical and physical quality is deﬁned
as good; this judgement is applied, for example, to the Naviglio Grande, the Pavese, and the Muzza
Canal that derive from waters that are in good condition.


===== PAGE 14 (score 37, numeric-paren 18, table-words 2, var-words 15) =====

Water 2016, 8, 577
14 of 17
For dummy variables, the value is not tied to the percentage increase but to the presence of the
characteristic that the variable describes.
Table 5. Price change according to the average of the value of real estate in the province of Milan.
Variable
% Variation
Price Change According to the Average of the Value of
Real Estate in the Province of Milan on 7 July 2015 in †
sup_mq
0.440%
Increase in value
† 1,311.39 ***
elevat
0.154%
Increase in value
† 459.38 ***
box
0.121%
Increase in value
† 360.88 ***
kitch_type
0.064%
Increase in value
† 190.82 ***
cond_fees
0.019%
Increase in value
† 56.64 ***
furnish_type
0.018%
Increase in value
† 53.66 ***
dist_mall
0.017%
Increase in value
† 50.68 *
dist_stream
0.012%
Increase in value
† 35.77 ***
quality_stream
0.008%
Increase in value
† 23.85 *
quality_canal
0.002%
Increase in value
† 5.96
green_area
−0.016%
Decrease in value
−† 47.69 ***
balcony
−0.019%
Decrease in value
−† 56.63 **
dist_tr_st
−0.029%
Decrease in value
−† 86.43 ***
energ_clas
−0.045%
Decrease in value
−† 134.10 ***
prop_status_2 (renewed)
−0.046%
Decrease in value
−† 137.08 ***
dist_canal
−0.074%
Decrease in value
−† 220.49 ***
dist_sub_st
−0.119%
Decrease in value
−† 354.49 ***
prop_status_3 (habitable)
−0.175%
Decrease in value
−† 521.16 ***
prop_status_4 (to be renewed)
−0.248%
Decrease in value
−† 738.29 ***
Notes: Standardized coefﬁcients of spatial lagged model (model 3) are used. The signiﬁcance of change is
related to signiﬁcance of coefﬁcients. The signiﬁcance levels are *** p < 0.001, ** p < 0.01, * p < 0.1.
The results show that the structural features are strongly able to affect the change of sale price.
The most signiﬁcantly inﬂuencing factors are the size of property, the status (with respect to new
residential houses), the presence of an elevator and of garage, the type of kitchen and the energy class.
Signiﬁcant are also some territorial variables, such as distance from the subway and train station.
Of low relevance are the presence of green areas, the types of furnishings, the distance from shopping
malls and the presence of a balcony. Looking at water, the most inﬂuencing variable is the distance
from a canal. In contrast, the distance from a stream and its quality have low effect on the change of
sale price.
5. Conclusions
This study applies a hedonic approach in order to investigate the role of water body proximity
and water quality as leveraging residential property values in the province of Milan, by implementing
a spatial modelling of the real estate market on a database consisting of 10,802 data. Moreover,
it contributes to a better understanding of the effect of water on property prices by distinguishing
between natural and artiﬁcial water bodies.
The results show that proximity to streams in the province of Milan is a negative factor leading to
a decrease in the value of real estate. Conversely, the proximity to artiﬁcial canals is seen as a positive
feature that inﬂuences the housing market by increasing property values. Models 2 and 3 in Table 4
further suggest that this effect may be moderated by the water quality. The quality of canal water
does not inﬂuence the effect of proximity. This is likely because the water quality of canals, which is
inclined to be generally better than that of streams, does not show large differences between canals
(see Figure 1). In contrast, the impact of proximity to streams with a good ecological status of water
is the opposite if compared to streams with bad water quality. In other words, proximity to canals is
always positively perceived whilst the effect of proximity to streams depends on the water quality.
The conﬂicting effect of proximity to natural waterways suggests, as previously supposed, that the
ﬂooding risk is considered to be low when compared to quality of water. This is likely to be due to low


===== PAGE 15 (score 8, numeric-paren 0, table-words 0, var-words 8) =====

Water 2016, 8, 577
15 of 17
probability and low impact of past overﬂows in this area. Moreover, in the last decades, several water
security practices and programmes have been implemented by local municipalities to minimize urban
and peripheral environmental vulnerability.
The comparison between urban and peripheral environments highlights, on the one hand, that the
proximity to canals signiﬁcantly affects property values only in the city. In addition, even if canal
water quality does not have a direct effect, it indirectly increases the willingness to pay for a property
near a clean, not smelly and unpolluted canal. On the other hand, the impact of proximity to streams
is negative in both environments. Moreover, even though the stream water quality is expected to be
better in the periphery, it positively and directly affects property values inside as well as outside the
city. This suggests that in the peri-urban environment, the difference in quality may be balanced by a
higher risk of ﬂooding.
Despite the risks, the wish of residents for water quality and proximity is strong and fast increasing.
Today, Milan is trying to change itself, trying to recover its history as a city of water. In the last century,
the industrial development has led the city to invest in urban infrastructures and facilities in order
to support economic and social growth. The uncontrolled cementing over the waterways was a
consequence of the urbanization process. In the last decade, increasing environmental attention has
led to a rediscovery of waterways inside and outside the city. This process has recently had a strong
acceleration in preparation for Expo 2015. As a consequence, the value of water in the real estate
market is likely to increase further.
Some limitations need to be stated. First, the lack of an Italian ofﬁcial database concerning the
information of residential housing prices over time makes it impossible to implement panel-based
techniques and makes cross-sectional modelling the best option. Moreover, since the hedonic model
is a preference methodology, where buyers and sellers are in equilibrium based upon the sales price
realized, the use of bid price rather than real sale price may lead to an overestimation of the property
prices, introducing a potential model bias. Second, due to differences between urban and peri-urban
environments in the Milan province, we introduced another potential bias in the achievement of
equilibrium in a speciﬁc housing market related to the general model results. Third, the lack of data on
ﬂooding risk limits the general insights related to the effect of water security, which we are only able
to indirectly suppose and discuss. Fourth, this research could be further improved by looking at the
change in water quality or by implementing a qualitative analysis of the municipal protection plan,
programmes and practices to identify the perception of hydrogeological risks in the local area.
Author Contributions:
Stefano Corsi and Ivan De Noni contributed to the introduction section (1).
Federico Bonetti contributed to the background context (2). Ivan De Noni and Luigi Orsi contributed to the
methodology (3). Stefano Corsi, Ivan De Noni and Luigi Orsi contributed to the results and discussion (4).
Federico Bonetti, Stefano Corsi, Ivan De Noni and Luigi Orsi contributed to the conclusions (5).
Conﬂicts of Interest: With the submission of this manuscript we would like to undertake that: (1) All authors of
this research paper have directly participated in the planning, execution, or analysis of this study; (2) All authors
of this paper have read and approved the ﬁnal version submitted; (3) The contents of this manuscript have not
been copyrighted or published previously; (4) The contents of this manuscript are not now under consideration
for publication elsewhere; (5) The contents of this manuscript will not be copyrighted, submitted, or published
elsewhere, while acceptance by the Journal is under consideration; (6) There are no directly related manuscripts
or abstracts, published or unpublished, by any authors of this paper; (7) My Institute’s (University of Milan)
representative is fully aware of this submission.
References
1.
Sadler, J.; Bates, A.; Hale, J.; James, P. Bringing cities alive: The importance of urban green spaces for people
and biodiversity. In Urban Ecology; Cambridge University Press: Cambridge, UK, 2010; pp. 230–260.
2.
Bowler, D.E.; Buyung-Ali, L.; Knight, T.M.; Pullin, A.S. Urban greening to cool towns and cities: A systematic
review of the empirical evidence. Landsc. Urban Plan. 2010, 97, 147–155. [CrossRef]
3.
Depietri, Y.; Renaud, F.G.; Kallis, G. Heat waves and ﬂoods in urban areas: A policy-oriented review of
ecosystem services. Sustain. Sci. 2012, 7, 95–107. [CrossRef]