1# RECORD 43d34a3cb07ebe23b958a009b21ab082b8ff02252# TITLE: Canals vs. Streams: To What Extent Do Water Quality and Proximity Affect Real Estate Values? A Hedonic Approach Analysis3# YEAR: 2016.0 VENUE: nan DOI: 10.3390/w81205774# AUTHORS: Federico Bonetti; S. Corsi; L. Orsi; I. De Noni5# PAGES: 17 CANDIDATE TABLE PAGES (1-based): [8, 11, 14]6# 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.789===== PAGE 1 (score 14, numeric-paren 0, table-words 1, var-words 12) =====1011water12Article13Canals vs. Streams: To What Extent Do Water Quality14and Proximity Affect Real Estate Values? A Hedonic15Approach Analysis16Federico Bonetti, Stefano Corsi, Luigi Orsi * and Ivan De Noni17Department of Economics, Management and Quantitative Methods, University of Milan, via Conservatorio 7,1820122 Milan, Italy; federico.bonetti2@studenti.unimi.it (F.B.); stefano.corsi@unimi.it (S.C.);19ivan.denoni@unimi.it (I.D.N.)20* Correspondence: luigi.orsi@unimi.it; Tel.: +39-2-5032-154221Academic Editor: Richard C. Smardon22Received: 6 September 2016; Accepted: 30 November 2016; Published: 6 December 201623Abstract: The presence of and proximity to water influences human settlement decisions, because24water has been able to characterize the environment and affect human health since prehistoric25times. Many scholars have studied the effects of environmental amenities on the real estate market;26however, only a few of them have explored the effects of water. This study stresses the role of27water with a focus on the dichotomy between artificial and natural waterways (canals vs. streams),28which involves managerial and qualitative differences. A hedonic approach, based on a large web29scraped and geo-referenced dataset of 10,530 observations, is used to analyze the effect of water30proximity and quality on residential housing prices in the province of Milan (Italy). Our main results31show a significant but conflicting impact of water proximity on real estate value: reducing the canal32distance, we observe an increase in the residential housing sale price, in urban more than peri-urban33environments; in contrast, reducing the stream distance, our results show decreases in property value.34Moreover, a significant moderating effect of water quality on distance from streams is identifiable.35Keywords: water quality and proximity; real estate value; hedonic model; environmental impact361. Introduction37In rural areas, i.e., outside the city, water is a critical element because it is strongly related to38agricultural utilization. In contrast, in fragmented and highly urbanized areas, water is able to produce39several positive amenities, such as preserving natural ecosystems [1], mitigating the impacts caused by40human activity, such as air pollution and heat islands [2–5] and providing aesthetically pleasing open41spaces and recreational sites [6–8]. However, in spite of these positive amenities, some researchers42have specifically focused only on the aptitude of water to leverage the real estate market by influencing43property values.44In this light, the effect of surface water bodies on residential housing prices has been specifically45related to different features such as water proximity, security and quality. Walsh et al. (2011) [9]46identified the significance of water body proximity on the value of residential properties in Orange47County, Florida. Positive amenities from proximity to water are also reported by Anderson and West48(2006) [10]. Tapsuwan et al. (2009) [11] argue that the value of houses increases by reducing the49distance to the wetland. Cohen et al. (2014) [12] show conflicting findings for proximity to wetlands50but confirm the positive effect of water bodies’ proximity. Cho et al. (2006) [13] suggest the influence of51proximity to creeks and streams in Tennessee. Nelson et al. (2005) [14] highlight the positive effect of52proximity to artificial water bodies. Other studies focus on water views rather than spatial proximity53and suggest surface water is able to increase the waterfront property value [15].54Water 2016, 8, 577; doi:10.3390/w812057755www.mdpi.com/journal/water565758===== PAGE 8 (score 133, numeric-paren 126, table-words 1, var-words 5) =====5960Water 2016, 8, 577618 of 1762Table 3. Correlation matrix.63Variables641652663674685696707718729731074117512761377147815791680178118821983208421851 price861872 dist_stream88−0.03 ***891903 quality_stream910.06 ***920.00931944 dist_canal95−0.18 ***96−0.07 ***97−0.04 ***981995 quality_canal1000.02 *1010.00102−0.17 ***103−0.0110411056 energ_clas106−0.07 ***107−0.05 ***108−0.011090.01110−0.03 **11111127 sup_mq1130.74 ***1140.011150.04 ***116−0.03 ***1170.02 *118−0.08 ***11911208 cond_fees1210.42 ***122−0.03 **1230.02 *124−0.05 ***1250.02 *1260.011270.29 ***12811299 kitch_type1300.24 ***1310.011320.03 **1330.03 ***1340.001350.04 ***1360.42 ***1370.12 ***138113910 elevat1400.13 ***141−0.09 ***1420.02 *143−0.04 ***1440.01145−0.14 ***146−0.011470.11 ***1480.07 ***149115011 balcony1510.04 ***152−0.02 *153−0.011540.07 ***1550.011560.02 *1570.12 ***1580.05 ***1590.21 ***1600.08 ***161116212 box1630.08 ***1640.10 ***165−0.011660.12 ***1670.02 *168−0.23 ***1690.29 ***1700.02 *1710.18 ***1720.011730.08 ***174117513 furnish_type1760.04 ***177−0.02 *1780.06 ***179−0.05 ***1800.011810.04 ***182−0.05 ***183−0.01184−0.10 ***185−0.01186−0.08 ***187−0.08 ***188118914 prop_status_1190−0.011910.001920.00193−0.011940.01195−0.02 *1960.001970.00198−0.011990.012000.02 *2010.00202−0.01203120415 prop_status_22050.012060.02 *2070.04 ***208−0.012090.012100.012110.002120.002130.00214−0.02 *215−0.01216−0.012170.01218−0.31 ***219122016 prop_status_3221−0.02222−0.03 **223−0.04 ***2240.02 *225−0.02 *2260.012270.002280.002290.012300.012310.002320.012330.00234−0.26 ***235−0.70 ***236123717 prop_status_42380.03 **2390.012400.00241−0.012420.00243−0.012440.012450.002460.002470.012480.002490.002500.00251−0.10 ***252−0.26 ***253−0.21 ***254125518 dist_sub_st256−0.22 ***2570.31 ***258−0.02 *2590.24 ***260−0.01261−0.10 ***2620.00263−0.12 ***2640.02265−0.18 ***2660.05 ***2670.32 ***268−0.10 ***2690.00270−0.012710.01272−0.01273127419 dist_tr_st275−0.15 ***2760.16 ***277−0.05 ***2780.05 ***279−0.01280−0.08 ***2810.02 *282−0.07 ***2830.04 ***284−0.10 ***2850.06 ***2860.26 ***287−0.03 **2880.01289−0.022900.02291−0.012920.26 ***293129420 dist_mall295−0.17 ***2960.23 ***297−0.04 ***2980.11 ***2990.01300−0.09 ***3010.01302−0.09 ***3030.00304−0.18 ***3050.02 *3060.25 ***307−0.06 ***3080.01309−0.02 *3100.013110.003120.42 ***3130.45 ***314131521 green_area316−0.02 *317−0.01318−0.02 *3190.05 ***3200.003210.02 *3220.02 *3230.03 ***3240.06 ***3250.06 ***3260.04 ***3270.03 **328−0.013290.00330−0.02 *3310.02 *3320.01333−0.10 ***3340.04 ***335−0.013361337Notes: The significance levels are *** p < 0.001, ** p < 0.01, * p < 0.1.338339340===== PAGE 9 (score 13, numeric-paren 0, table-words 1, var-words 11) =====341342Water 2016, 8, 5773439 of 17344Territorial Variables345The territorial variables (all territorial and environmental variables calculated are based on data346provided by the GeoPortal of the Lombardy region) are the shortest distance, i.e., “as the crow flies”,347from subway stations (dist_sub_st is the distance in meters from the subway station or the nearest348urban rail station; it is the main line of urban displacement), from train stations (dist_tr_st is the349minimum distance in meters from the train station) and shopping malls (dist_mal) is the minimum350distance in meters from the shopping center. The green area (we used the classification “Parchi351e giardini” of the Lombardy region’s Dusaf 4.0) (green_area_300 m) represents a radius of 300 m352in diameter from the center of the building in the analysis. This variable allows us to assess how353much green area in square meters is available within a walking distance of 300 m from the property.354Finally, the geographical location of the property using the 96 zip codes (zip_codes = dummy variables355for all the zip codes in the Milan province) to take account of spatial heterogeneity correlated with356cross-jurisdictional variations in the housing market in the overall area.357The value of territorial variables is calculated by punctuated information bases indicating,358for example, the location of Metro stations, the location of shopping malls, etc.3593.2.2. Explanatory Variables360Environmental Variables361Our paper focuses on the evaluation of the impact of water proximity and quality on362housing prices.363The environmental variables’ distance from streams (dist_stream represents distance in meters364from the nearest streams) and distance from canals (dist_canal represents distance in meters from the365nearest canal) represent the shortest distance, i.e., “as the crow flies”, from streams and canals to assess366the impact of a water stream proximity on housing prices.367Stream quality (quality_stream) and canal quality (quality_canal) measure the quality of water on368a 5-point Likert scale using the ARPA Lombardy water quality report for the period 2009–2014 [28].369This 5-point Likert scale indicator is calculated by combining biological quality elements, general370physical–chemical elements and specific water pollutants (1 = bad water quality, 2 = poor, 3 = sufficient,3714 = good and 5 = excellent water quality). Finally, we have linked together each single property with372the quality of the nearest stream and canal and with the nearest ARPA detection point.373In addition, to evaluate any potential moderating effect of water quality and proximity,374we375introduce376two377interaction378variables:379stream380distance381multiplied382by383stream384quality385(dist_stream*quality_stream) and canal distance multiplied by canal quality (dist_canal*quality_canal).386To achieve this goal, we standardized the two variables prior to calculating their interaction terms,387in order to avoid unnecessary multi-collinearity.3883.2.3. Dependent Variable389Utilizing the website of a consenting large Italian real estate agency, we obtained data for each390property sale in the province of Milan. The house price information (price) is a bid price because the391proposed purchase price is not based on any real negotiation of sale.3923.3. Hedonic Price Model393After we analyzed the literature, we chose the hedonic price model for our analysis [27,30,33–35].394The hedonic price model consists of a semi-logarithmic least squares hedonic model that is ascribable395to the following general formula [34]:396LnPi = β0 + β1Si + β2Ni + β3Qi + "i397(1)398399400===== PAGE 11 (score 160, numeric-paren 129, table-words 6, var-words 19) =====401402Water 2016, 8, 57740311 of 17404(Model 5). We used the address variable to divide the database by the position of the property, based on405whether the property was in the city of Milan or in another town of the province.406Table 4. Regression models results.407Dependent Variable:408House Price (log)409OLS and Spatial Regression Models410Model 1 Base411Model OLS412Model 2 Full413Model OLS414Model 3 Full Model415Spatial-Lag (SL)416Model 4 SL Full417Model Milano418Model 5 SL Full419Model Province420Intercept42111.996 (0.021) ***42212.024 (0.022) ***4237.323 (0.082) ***42411.414 (0.059) ***42512.282 (0.026) ***426Rho (spatial lag)4270.406 (0.006) ***4280.080 (0.004) ***4290.023 (0.002) ***430Explanatory variables431dist_stream4320.012 (0.005) *4330.012 (0.003) ***4340.058 (0.005) ***4350.024 (0.004) ***436quality_stream4370.005 (0.003) *4380.008 (0.003) **4390.011 (0.005) *4400.010 (0.004) **441dist_stream*quality_stream442−0.004 (0.003) *443−0.005 (0.003) *444−0.001 (0.005)445−0.006 (0.003) *446dist_canal447−0.020 (0.005) ***448−0.074 (0.003) ***449−0.071 (0.006) ***450−0.003 (0.003)451quality_canal452−0.001 (0.003)4530.002 (0.003)454−0.007 (0.005)4550.000 (0.004)456dist_canal*quality_canal457−0.001 (0.003)458−0.003 (0.003)459−0.013 (0.005) *460−0.003 (0.003)461Control variables462energ_clas463−0.042 (0.003) ***464−0.042 (0.003) ***465−0.046 (0.003) ***466−0.030 (0.005) ***467−0.048 (0.004) ***468sup_mq4690.405 (0.003) ***4700.405 (0.003) ***4710.439 (0.004) ***4720.531 (0.006) ***4730.350 (0.004) ***474cond_fees4750.006 (0.003) *4760.006 (0.003) *4770.019 (0.003) ***4780.010 (0.005) *4790.025 (0.004) ***480kitch_type4810.083 (0.006) ***4820.083 (0.006) ***4830.064 (0.007) ***4840.103 (0.012) ***4850.052 (0.008) ***486elevat4870.115 (0.006) ***4880.115 (0.006) ***4890.154 (0.007) ***4900.206 (0.013) ***4910.077 (0.008) ***492balcony4930.011 (0.006) *4940.012 (0.006) *495−0.019 (0.007) **4960.008 (0.011)497−0.043 (0.008) ***498garage4990.169 (0.006) ***5000.168 (0.006) ***5010.121 (0.007) ***5020.070 (0.012) ***5030.205 (0.008) ***504furnish_type505−0.006 (0.006)506−0.006 (0.006)5070.017 (0.007) *5080.036 (0.011) **509−0.023 (0.008) **510prop_status_2511−0.079 (0.009) ***512−0.079 (0.009) ***513−0.045 (0.011) ***5140.015 (0.020)515−0.100 (0.012) ***516prop_status_3517−0.213 (0.010) ***518−0.213 (0.010) ***519−0.175 (0.012) ***520−0.153 (0.021) ***521−0.221 (0.012) ***522prop_status_4523−0.310 (0.013) ***524−0.310 (0.013) ***525−0.248 (0.016) ***526−0.205 (0.024) ***527−0.400 (0.021) ***528year_dummy_2012529−0.013 (0.009)530−0.013 (0.009)5310.004(0.011)5320.012 (0.017)533−0.020 (0.012) *534year_dummy_2013535−0.013 (0.008)536−0.013 (0.008)537−0.018 (0.010) *538−0.024 (0.016)539−0.026 (0.011) *540year_dummy_2014541−0.005 (0.007)542−0.007 (0.007)543−0.009 (0.008)5440.000 (0.013)545−0.009 (0.009)546year_dummy_2015547−0.030 (0.009) **548−0.030 (0.009) **549−0.038 (0.011) **550−0.054 (0.016) **551−0.015 (0.014)552dist_sub_st553−0.115 (0.007) ***554−0.120 (0.007) ***555−0.119 (0.004) ***5560.005 (0.007)557−0.096 (0.004) ***558dist_tr_st559−0.028 (0.005) ***560−0.029 (0.006) ***561−0.029 (0.004) ***562−0.053 (0.005) ***563−0.026 (0.004) ***564dist_mall565−0.001 (0.005)566−0.003 (0.005)5670.018 (0.004) ***568−0.066 (0.005) ***5690.003 (0.004)570green_area5710.004 (0.003)5720.004 (0.003)573−0.016 (0.003) ***574−0.017 (0.005) **5750.020 (0.003) ***576zip_codes577Yes578Yes579Yes580Yes581Yes582No. of observations58310,53058410,53058510,53058653185875212588F-stat589562.5 ***590535.9 ***591adj. R-squared5920.8585930.861594LogLik595−623.86596−608.94.13597−421.38598−573.05599−291.49600AIC6011099.86021277.1603393.9604Max VIF6052.7556062.8596072.8596082.8106092.396610Mean VIF6111.2986121.3216131.3216141.2896151.278616Notes: The coefficients are mean centering standardized values. The standard errors are in parentheses.617The significance levels are *** p < 0.001, ** p < 0.01, * p < 0.1. Zip codes are not reported due to spatial constraints.618Model 1 estimates the coefficients of the control variables. The model results allow us to state619that energy class, subway stations’ distance and railway stations’ distance have a negative and620highly significant (p < 0.001) influence on the house sale price. In contrast, the house surface area,621the habitable living space, kitchen, presence of an elevator, balcony and/or garage, status of the622property and the condominium fees have a positive and significant impact. Specifically, the positive623effect of condominium fees may be related to more luxury services or improving conditions of the624property. Finally, the distance from the mall, the house furnishings and the presence of green areas625within a walking distance of 300 m are not statistically significant (p > 0.10) on house prices. The zip626codes variable (not reported in Table 4 due to spatial constraints) shows higher and significant values627for housing prices in the zip codes close to Milan city center, while in certain suburbs of Milan and,628especially, in the peripheral province areas, it shows a negative or no significant impact on house629prices. Finally, the year dummy variables introduced to control for a potential time trend show a slight630negative trend with respect to the reference year (2011) but no significant statistical differences among631years except for 2015. We also tested different time periods (semesters and quarters) and found no632significant time trends.633The overall fit of the models increases compared to the baseline, as Models 2, 3 and 4 fit our634data better and have more explanatory power than Model 1. Moreover, the coefficients and signs of635636637===== PAGE 12 (score 13, numeric-paren 0, table-words 1, var-words 11) =====638639Water 2016, 8, 57764012 of 17641the control variables remain quite stable along the different models, showing robust results and that642multi-collinearity is not a particular problem in these regressions.643In Model 2, we include all the explanatory variables related to water proximity and quality and644their interaction effect. The results show a significant (p < 0.10) and positive effect of the streams’645distance on house prices. The more we move away from the rivers, the higher the price of houses.646As expected, the quality of the water streams positively and significantly affects housing prices and647mitigates the negative effect of the streams’ distance. This latter result is confirmed by the negative and648significant sign of the interaction effect (dist_stream*quality_stream). In contrast, we find a positive649and highly significant impact of canal proximity on property value, while the quality of the canals and650the interaction terms with the distance are non-significant.651Based on the results of Moran I and Lagrange Multiplier (LM) tests, in Model 3, a spatial652dependence technique is introduced. Spatial lag is used to estimate the regression coefficients.653However, AIC and log-likelihood tests suggest that spatial-lag and also spatial-error models are654better than OLS models, because of the spatial dependence structure in our data (see Moran and LM655tests in the previous section).656Model 3 (spatial lag model) highlights that the residential housing price is a spatial dependence657phenomenon (lag = 0.405, p-value < 0.001). The results are very close to the linear model. No changes658in terms of sign and significance are identifiable across the coefficients of control variables between the659linear and spatial lags models; the unique exceptions are the presence of a balcony, which changes the660sign; the presence of green areas; the types of furnishings; and the distance from malls, which become661significant. The exploratory variables concerning water quality and proximity keep the same sign and662significance. However, in the spatial model, coefficients are more robust because they take spatial bias663into account.664The different responses of explanatory variables that indicate the distance from natural water665bodies and canals can be attributed to the fact that proximity to the first is considered a negative666function of poor water quality. However, the influence of canals is positive in relation to the better667water quality and to the increase in environmental conditions in heavily anthropic areas.668This theory is strongly supported by studies carried out by the public authorities appointed to669monitor the water quality. The results come from a study carried out by ARPA of the Lombardy region670on the natural streams basin of the rivers Lambro and Olona for the period from 2009 to 2014.671The results of this work indicate that a state of stress exists for the entire basin with an alteration672to the rivers’ cleaning capacity. In addition, more than 50% of the areas analyzed reveal contamination673by heavy metals, and a complete analysis of the areas shows a bad or poor water quality, especially in674reference to the presence of bad odors and dirty water [28].675In contrast, the canals that derive from rivers such as the Ticino and Adda have the highest676average water quality compared to other waters in the province [41]. The high water quality is a result677of actions taken by the Ticino and Adda parks authorities. This fact, combined with the maintenance678that is performed on the canals periodically by land reclamation authorities to ensure the proper679hydraulic features, explains the results from the hedonic model application.680To support this thesis, we refer to the “Carta Provinciale delle Vocazioni Ittiche” [42] that,681in addition to reporting the status of the fish and fauna of the natural streams and canals, also lists682the chemical and physical conditions of the water. These are synthesized by an indicator called the683environmental status of watercourses (SECA) that allows us to define the status of water through the684pollution level, the status of the biological communities, the ecosystems and the presence of dangerous685substances and bio accumulation.686There is chemical, physical and biological quality impairment of the main natural streams of the687province, such as Seveso, Lambro and Olona, that worsens along the route of the rivers. However,688for rivers such as the Ticino and Adda, the judgement on chemical and physical quality is defined689as good; this judgement is applied, for example, to the Naviglio Grande, the Pavese, and the Muzza690Canal that derive from waters that are in good condition.691692693===== PAGE 14 (score 37, numeric-paren 18, table-words 2, var-words 15) =====694695Water 2016, 8, 57769614 of 17697For dummy variables, the value is not tied to the percentage increase but to the presence of the698characteristic that the variable describes.699Table 5. Price change according to the average of the value of real estate in the province of Milan.700Variable701% Variation702Price Change According to the Average of the Value of703Real Estate in the Province of Milan on 7 July 2015 in †704sup_mq7050.440%706Increase in value707† 1,311.39 ***708elevat7090.154%710Increase in value711† 459.38 ***712box7130.121%714Increase in value715† 360.88 ***716kitch_type7170.064%718Increase in value719† 190.82 ***720cond_fees7210.019%722Increase in value723† 56.64 ***724furnish_type7250.018%726Increase in value727† 53.66 ***728dist_mall7290.017%730Increase in value731† 50.68 *732dist_stream7330.012%734Increase in value735† 35.77 ***736quality_stream7370.008%738Increase in value739† 23.85 *740quality_canal7410.002%742Increase in value743† 5.96744green_area745−0.016%746Decrease in value747−† 47.69 ***748balcony749−0.019%750Decrease in value751−† 56.63 **752dist_tr_st753−0.029%754Decrease in value755−† 86.43 ***756energ_clas757−0.045%758Decrease in value759−† 134.10 ***760prop_status_2 (renewed)761−0.046%762Decrease in value763−† 137.08 ***764dist_canal765−0.074%766Decrease in value767−† 220.49 ***768dist_sub_st769−0.119%770Decrease in value771−† 354.49 ***772prop_status_3 (habitable)773−0.175%774Decrease in value775−† 521.16 ***776prop_status_4 (to be renewed)777−0.248%778Decrease in value779−† 738.29 ***780Notes: Standardized coefficients of spatial lagged model (model 3) are used. The significance of change is781related to significance of coefficients. The significance levels are *** p < 0.001, ** p < 0.01, * p < 0.1.782The results show that the structural features are strongly able to affect the change of sale price.783The most significantly influencing factors are the size of property, the status (with respect to new784residential houses), the presence of an elevator and of garage, the type of kitchen and the energy class.785Significant are also some territorial variables, such as distance from the subway and train station.786Of low relevance are the presence of green areas, the types of furnishings, the distance from shopping787malls and the presence of a balcony. Looking at water, the most influencing variable is the distance788from a canal. In contrast, the distance from a stream and its quality have low effect on the change of789sale price.7905. Conclusions791This study applies a hedonic approach in order to investigate the role of water body proximity792and water quality as leveraging residential property values in the province of Milan, by implementing793a spatial modelling of the real estate market on a database consisting of 10,802 data. Moreover,794it contributes to a better understanding of the effect of water on property prices by distinguishing795between natural and artificial water bodies.796The results show that proximity to streams in the province of Milan is a negative factor leading to797a decrease in the value of real estate. Conversely, the proximity to artificial canals is seen as a positive798feature that influences the housing market by increasing property values. Models 2 and 3 in Table 4799further suggest that this effect may be moderated by the water quality. The quality of canal water800does not influence the effect of proximity. This is likely because the water quality of canals, which is801inclined to be generally better than that of streams, does not show large differences between canals802(see Figure 1). In contrast, the impact of proximity to streams with a good ecological status of water803is the opposite if compared to streams with bad water quality. In other words, proximity to canals is804always positively perceived whilst the effect of proximity to streams depends on the water quality.805The conflicting effect of proximity to natural waterways suggests, as previously supposed, that the806flooding risk is considered to be low when compared to quality of water. This is likely to be due to low807808809===== PAGE 15 (score 8, numeric-paren 0, table-words 0, var-words 8) =====810811Water 2016, 8, 57781215 of 17813probability and low impact of past overflows in this area. Moreover, in the last decades, several water814security practices and programmes have been implemented by local municipalities to minimize urban815and peripheral environmental vulnerability.816The comparison between urban and peripheral environments highlights, on the one hand, that the817proximity to canals significantly affects property values only in the city. In addition, even if canal818water quality does not have a direct effect, it indirectly increases the willingness to pay for a property819near a clean, not smelly and unpolluted canal. On the other hand, the impact of proximity to streams820is negative in both environments. Moreover, even though the stream water quality is expected to be821better in the periphery, it positively and directly affects property values inside as well as outside the822city. This suggests that in the peri-urban environment, the difference in quality may be balanced by a823higher risk of flooding.824Despite the risks, the wish of residents for water quality and proximity is strong and fast increasing.825Today, Milan is trying to change itself, trying to recover its history as a city of water. In the last century,826the industrial development has led the city to invest in urban infrastructures and facilities in order827to support economic and social growth. The uncontrolled cementing over the waterways was a828consequence of the urbanization process. In the last decade, increasing environmental attention has829led to a rediscovery of waterways inside and outside the city. This process has recently had a strong830acceleration in preparation for Expo 2015. As a consequence, the value of water in the real estate831market is likely to increase further.832Some limitations need to be stated. First, the lack of an Italian official database concerning the833information of residential housing prices over time makes it impossible to implement panel-based834techniques and makes cross-sectional modelling the best option. Moreover, since the hedonic model835is a preference methodology, where buyers and sellers are in equilibrium based upon the sales price836realized, the use of bid price rather than real sale price may lead to an overestimation of the property837prices, introducing a potential model bias. Second, due to differences between urban and peri-urban838environments in the Milan province, we introduced another potential bias in the achievement of839equilibrium in a specific housing market related to the general model results. Third, the lack of data on840flooding risk limits the general insights related to the effect of water security, which we are only able841to indirectly suppose and discuss. Fourth, this research could be further improved by looking at the842change in water quality or by implementing a qualitative analysis of the municipal protection plan,843programmes and practices to identify the perception of hydrogeological risks in the local area.844Author Contributions:845Stefano Corsi and Ivan De Noni contributed to the introduction section (1).846Federico Bonetti contributed to the background context (2). Ivan De Noni and Luigi Orsi contributed to the847methodology (3). Stefano Corsi, Ivan De Noni and Luigi Orsi contributed to the results and discussion (4).848Federico Bonetti, Stefano Corsi, Ivan De Noni and Luigi Orsi contributed to the conclusions (5).849Conflicts of Interest: With the submission of this manuscript we would like to undertake that: (1) All authors of850this research paper have directly participated in the planning, execution, or analysis of this study; (2) All authors851of this paper have read and approved the final version submitted; (3) The contents of this manuscript have not852been copyrighted or published previously; (4) The contents of this manuscript are not now under consideration853for publication elsewhere; (5) The contents of this manuscript will not be copyrighted, submitted, or published854elsewhere, while acceptance by the Journal is under consideration; (6) There are no directly related manuscripts855or abstracts, published or unpublished, by any authors of this paper; (7) My Institute’s (University of Milan)856representative is fully aware of this submission.857References8581.859Sadler, J.; Bates, A.; Hale, J.; James, P. Bringing cities alive: The importance of urban green spaces for people860and biodiversity. In Urban Ecology; Cambridge University Press: Cambridge, UK, 2010; pp. 230–260.8612.862Bowler, D.E.; Buyung-Ali, L.; Knight, T.M.; Pullin, A.S. Urban greening to cool towns and cities: A systematic863review of the empirical evidence. Landsc. Urban Plan. 2010, 97, 147–155. [CrossRef]8643.865Depietri, Y.; Renaud, F.G.; Kallis, G. Heat waves and floods in urban areas: A policy-oriented review of866ecosystem services. Sustain. Sci. 2012, 7, 95–107. [CrossRef]867