1# RECORD d956659c01d343fed17ef54a0b1e248a34a329952# TITLE: THE EFFECT OF URBAN GREEN SPACES ON HOUSE PRICES IN WARSAW3# YEAR: 2018.0 VENUE: International Journal of Strategic Property Management DOI: 10.3846/ijspm.2018.52204# AUTHORS: R. Trojanek; Michał Głuszak; Justyna Tanas5# PAGES: 14 CANDIDATE TABLE PAGES (1-based): [7, 9, 10, 11]6# ABSTRACT: In the paper, we analysed the impact of proximity to urban green areas on apartment prices in Warsaw. The data-set contained in 43 075 geo-coded apartment transactions for the years 2010 to 2015. In this research, the hedonic method was used in Ordinary Least Squares (OLS), Weighted Least Squares (WLS) and Median Quantile Regression (Median QR) models. We found substantial evidence that proximity to an urban green area is positively linked with apartment prices. On an average presence of a green area within 100 meters from an apartment increases the price of a dwelling by 2,8% to 3,1%. The effect of park/forest proximity on house prices is more significant for newer apartments than those built before 1989. We found that proximity to a park or a forest is particularly important (and has a higher implicit price as a result) in the case of buildings constructed after 1989. The impact of an urban green was particularly high in the case of a post-transformation housing estate. Close vicinity (less than 100 m distance) to an urban green increased the sales prices of apartments in new residential buildings by 8,0–8,6%, depending on a model.789===== PAGE 1 (score 16, numeric-paren 0, table-words 0, var-words 16) =====1011Copyright © 2018 The Author(s). Published by VGTU Press12This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unre-13stricted use, distribution, and reproduction in any medium, provided the original author and source are credited.14*Corresponding author. E-mail: r.trojanek@ue.poznan.pl15International Journal of Strategic Property Management16ISSN: 1648-715X / eISSN: 1648-9179172018 Volume 22 Issue 5: 358–37118https://doi.org/10.3846/ijspm.2018.5220 19THE EFFECT OF URBAN GREEN SPACES ON HOUSE PRICES IN WARSAW20Radoslaw TROJANEK1,*, Michal GLUSZAK2, Justyna TANAS3211 Department of Microeconomics, Poznan University of Economics and Business, Al. Niepodleglosci 10, 2261-875 Poznan, Poland232 Department of Real Estate and Investment Economics, Cracow University of Economics, 24ul. Rakowicka 27 bld. B, 31-510 Krakow, Poland253 Department of Investment and Real Estate, University of Gdansk, ul. Bazynskiego 8, 80-309 Gdansk, Poland26Received 17 July 2017; accepted 18 November 201727Abstract. In the paper, we analysed the impact of proximity to urban green areas on apartment prices in Warsaw. The data-28set contained in 43 075 geo-coded apartment transactions for the years 2010 to 2015. In this research, the hedonic method 29was used in Ordinary Least Squares (OLS), Weighted Least Squares (WLS) and Median Quantile Regression (Median QR) 30models. We found substantial evidence that proximity to an urban green area is positively linked with apartment prices. 31On an average presence of a green area within 100 meters from an apartment increases the price of a dwelling by 2,8% to 323,1%. The effect of park/forest proximity on house prices is more significant for newer apartments than those built before 331989. We found that proximity to a park or a forest is particularly important (and has a higher implicit price as a result) 34in the case of buildings constructed after 1989. The impact of an urban green was particularly high in the case of a post-35transformation housing estate. Close vicinity (less than 100 m distance) to an urban green increased the sales prices of 36apartments in new residential buildings by 8,0–8,6%, depending on a model.37Keywords: hedonic methods, parks, urban green, property prices.38Introduction39The market value of a property depends mainly on its 40physical characteristics, out of which the most important 41is location. The location choices depend mostly on indi-42vidual preferences but generally can be related to the dis-43tance from work, schools, and hospitals, an accessibility 44of public transport, positive or negative neighbourhood 45effects both about built and natural environment. Among, 46various factors affecting housing choices, and thus house 47prices, environmental attributes rank very high in the 48hierarchy of importance (Źróbek, M. Trojanek, Źróbek-49Sokolnik, & R. Trojanek, 2015; Gluszak & Marona, 2017).50Green spaces not only provide a pleasant and natural 51environment but also improve the quality of life in urban 52areas and undertake essential environmental functions 53(B. Zhang, Xie, C. Zhang, & J. Zhang, 2012; Streimikiene, 542014). A considerable range of benefits provided by urban 55green areas has been studied and reported in the literature. 56The list includes, but is not limited to (Konijnendijk, Anner-57stedt, Maruthaveeran, & Nielsen, 2013): human health and 58wellbeing, social cohesion, tourism, biodiversity, air qual-59ity and carbon sequestration, water management and the 60role of parks in the cooling of urban areas. Moreover, green 61spaces have a significant economic effect, most notably an 62impact on neighbouring property values. The latter effect 63was reported in a great number of research papers (for ex-64ample Espey & Owusu-Edusei, 2001; Jim & Chen, 2010).65The topic of urban green in CEE (Central and Eastern 66Europe) cities has been neglected in the discussion on the 67post-1989 transformation of the major metropolitan areas. 68Like many cities in the region, Warsaw experienced dis-69continuity of urban development (Staniszkis, 2012), and 70as a metropolitan area entity consists of two different sys-71tems. First of them is a heritage of a socialist period, and 72it is key characteristics are relatively low housing quality, 73but low development densities (building coverage ratios), 74greenness, and well-planned accessible public infrastruc-75ture. The second urban system is an effect of spontaneous, 76chaotic and dynamic development (Kusiak, 2012), that 77was not adequately regulated, due to malfunctions in ur-78ban planning. The higher quality of housing often comes 79at a price of high development density, lack of urban green 808182===== PAGE 7 (score 38, numeric-paren 0, table-words 3, var-words 22) =====838436485R. Trojanek et al. The effect of urban green spaces on house prices in Warsaw86construct models that will not be biased because of a low-87quality dataset). The research is based on transaction data. 88The sample is based on residential property sales from the 891st quarter of 2010 to the 4th quarter of 2015. The data was 90obtained from Board of Geodesy and Municipal Cadastre 91in Warsaw. All non-market transactions (e.g. debt enforce-92ment sales) were removed from the final dataset. Notarial 93contracts contain following information: transaction date, 94sales price, area of a dwelling, the floor on which a dwell-95ing is located, and the area of any auxiliary premises (e.g. 96a garage/parking spot in an indoor parking lot or a cellar/97storage). On the other hand, notarial contracts do not in-98clude information on various price determinants such as 99construction technology, architecture, building quality. 100Using cadastre data, additional information was gathered 101on the height of buildings and the year of construction. Fi-102nally, using Street View application on maps.google.com, 103the missing data concerning the height, and most impor-104tantly building technology was added. Locations of trans-105actions (exact addresses) were geocoded with google maps 106API. Based on WMS servers of the Warsaw City Hall, the 107vector layer of parks was prepared. In this research, we ex-108cluded apartments built before 1939. The reason for which 109was the fact that in the case of such apartments, the techni-110cal condition of a building is a significant determinant of 111their value (our base does not include this factor), which 112might burden the obtained results. Moreover, during the 113Second World War Warsaw was almost completely demol-114ished, so the share of such transactions is very small (should 115not burden the results). The final dataset contained 43 075 116geo-coded apartments sold between 1q 2010 and 4q 2015 117(Figure 1). Taking into account Dubin (1988) and Malpezzi 118(2008) suggestions three major categories of characteristics 119of dwellings may be distinguished: (1) structural attributes 120S, (2) neighbourhood related services and features N, (3) 121location and accessibility L. Additionally it is important to 122control for time (T).123The choice of qualitative and quantitative data was 124limited by the availability of information gathered. Table 2 125presents variables used in the study.126To conclude we controlled for various characteristics 127typically used in hedonic pricing models for residential 128market: location (administrative district), construction 129technology, floor, time of construction, area of an apart-130ment, presence of a garage, presence of a basement, height 131of building, and distance to a city centre, a kindergarten, a 132primary school, a subway and an urban green areas. The 133summary statistics of variables are presented in Table 3.134The results of the hedonic models’ estimation are pre-135sented in Section 3.136Table 3. Summary statistics of variables used in the research (source: own elaboration)137Category138Variable139Mean140Standard 141deviation142Category143Variable144Mean145Standard 146deviation147T14820101490.161500.37151S152area15352.8615424.2315520111560.161570.37158area21593381.061604089.2416120121620.141630.35164basement1650.431660.5016720131680.171690.38170cons19501710.181720.3817320141740.171750.38176cons19601770.161780.3617920151800.191810.39182cons19701830.111840.31185L186d11870.061880.24189cons19801900.041910.20192d21930.101940.30195cons19901960.061970.23198d31990.062000.23201cons20002020.402030.49204d42050.152060.36207cons20102080.062090.24210d52110.052120.23213floor12140.132150.34216d62170.102180.30219floor22200.512210.50222d72230.032240.17225floor32260.362270.48228d82290.012300.09231garage2320.422330.66234d92350.092360.28237height2381.672390.67240d102410.052420.22243technology2441.732450.44246d112470.042480.20249N250kindergarten2510.282520.19253d122540.072550.25256school2570.452580.29259d132600.012610.11262subway2632.782642.48265d142660.012670.09268duga1002690.062700.24271d152720.032730.18274duga2002750.112760.31277d162780.022790.13280duga3002810.122820.32283d172840.102850.30286duga4002870.112880.32289d182900.022910.14292duga5002930.102940.30295dcc2966.312973.37298duga5012990.503000.50301302303===== PAGE 8 (score 14, numeric-paren 0, table-words 3, var-words 8) =====304305International Journal of Strategic Property Management, 2018, 22(5): 358–3713063653072.4. Hedonic price models308In a great number of researchers, the aim of which is to es-309timate the implicit price for each characteristic of a good, 310the hedonic method is used. Since Lancaster’s (1966) 311and Rosen’s (1974) seminal work, the hedonic method is 312widely employed in housing studies. Hedonic price mod-313els focus on the utility derived from individual character-314istics of a good. The core of the hedonic methodology is 315the assumption that the price of any heterogeneous good 316(apartment in this study) is the function of its attributes 317(localisation, area, quality of an apartment). As housing is 318heterogeneous good, it is difficult to even indicate a full 319list of crucial attributes. In our study, we examine the im-320plicit value of one of neighbourhood related environmen-321tal features, namely the proximity to the urban green area 322(duga). We hypothesise that transaction price is a function 323of a distance to the urban green area (duga), while con-324trolling for other relevant structural (S), neighbourhood 325(N), and location attributes (L), as well as time (T). In 326general, the price is a hedonic function:327 328 (329, , , , )330price331f332duga S N L T333=334. 335(1)336In our research, we use a typical log-linear hedonic 337model specification, commonly used in the literature 338(Malpezzi, 2008). Regress apartment’s price (natural loga-339rithm of sales price) on a set of independent variables. 340Our baseline model is given by:341034213431003442345200346334730034843494003505351500352ln353.354k k355m356m357i i358j359j360price361duga362duga363duga364duga365duga366S367N368L369T370= β +β371+β372+β373+374β375+β376+ ∑θ377+ ∑ω378+379∑γ380+ ∑τ381+ ε382 383(2)384In this research, we used several variants of hedon-385ic regression, namely standard Ordinary Least Squares 386(OLS), robust Weighted Least Squares (WLS) and Median 387Quantile Regression (Median QR) models. We use differ-388ent estimation techniques to ensure the results are robust 389and reliable. The housing literature is quite consistent on 390treating housing as heterogeneous in several dimensions. 391From econometric (or data analysis) perspective, this het-392erogeneity can create heteroscedasticity in the residuals, 393while estimating the price function using standard OLS. 394Therefore, we decided to address the problem using differ-395ent analytical approaches. Firstly, a robust model, employ-396ing OLS with heteroscedasticity-correction (WLS) was es-397timated. Secondly, we decided to employ median quantile 398regression. The quantile regression relies on minimisation 399of weighted absolute deviations, and during the process, 400conditional quantiles (percentile) functions are estimated 401(Koenker & Bassett, 1978; Koenker & Hallock, 2001). The 402quantile regression involves weighing, both symmetric (for 403the median, quantile = 0.5), and asymmetric (for all other 404quantiles). In this research we decided to use symmetric 405weighting (quantile-0.5). Within the quantile regression 406approach, there is no limitation imposed on explaining the 407mean of the dependent variable. The quantile regression is 408more flexible, as it can be employed to explain the implicit 409prices of housing attributes at any point of the distribution 410of the dependent variable, thus for low-priced, medium-411priced, and high-priced properties (J. Zietz, E. Zietz, & 412Sirmans, 2008).413The mechanism to perform the quantile regression 414is similar to ordinary regression. The difference in the 415mechanism to perform the quantile regression to ordinary 416regression lies in the way of searching for the margin of 417sums of squared residuals; the quantile regression looks for 418the margin of weighted sums of absolute residuals. There 419are several theoretical advantages of a hedonic quantile 420regression reported in the literature. The technique can be 421particularly useful in case of heteroscedasticity, outliers, 422and unobserved heterogeneity found in the empirical data 423on housing transactions (Liao & Wang, 2012).424Due to the high number of independent variables 425available, multicollinearity may be a serious concern. 426Multicollinearity leads to unstable coefficients and inflated 427standard errors. The Variance Inflation Factors (VIFs) was 428used to detect it. The VIF values in models do not exceed 4296.8 which is in line with the most conservative rules of 430thumb that the mean of the VIFs should not be consider-431ably larger than 10.4323. Results433We estimated two groups of regression models. The first 434group of models, later referred to as a general model, is 435based on the whole sample of housing transactions con-436ducted in Warsaw (apartments built after 1950). The sec-437ond group of models was based on the same sample. How-438ever, new variables were introduced to track the influence 439of a construction period (different housing estate) on 440proximity to the urban green area. The estimation results 441for the general model is presented in Table 4.442As a robustness check, alongside baseline OLS model 443we used WLS and quantile regression. The results are rela-444tively similar, albeit we found several differences regarding 445statistical significance for selected parameters. The esti-446mates discussed below refer to quantile regression model.447In our study nominal prices were used (we did not 448correct for inflation) and we observed that within the pe-449riod under study (2010–2015), the time had a significant 450impact on transaction prices. It is worth mentioning that 451housing prices in the biggest cities in Poland increased by 452about 100% between 2006 and 2007 (Trojanek, 2012). At 453the end of 2007, the subsequent decreasing phase in the 454house price cycle began resulting from this abnormal price 455increase and the beginning of the financial crisis.456The regression coefficients of the locational variables 457(districts) can be interpreted as district values. In this 458research, the base variable was d9 corresponding to the 459Downtown district. Statistically significant negative coef-460ficients of regression of other districts confirmed the fact 461that consumers for apartments in the city centre are will-462ing to pay more (Trojanek, 2015).463In general, the relations between physical character-464istics of apartments and sales prices were consistent with 465expectations. Additionally, we found that the increase 466467468===== PAGE 9 (score 45, numeric-paren 0, table-words 6, var-words 6) =====469470366471R. Trojanek et al. The effect of urban green spaces on house prices in Warsaw472Table 4. Estimation results – general mode (dependent variable is natural logarithm of sale price) (source: own elaboration)473OLS474WLS475Median QR476Variable477Coefficient478Probability479Coefficient480Probability481Coefficient482Probability483const48411.95934850.000148611.86824870.000148811.90524890.0001490Y2011491–0.00714920.0130493–0.00714940.0027495–0.00884960.0011497Y2012498–0.08054990.0001500–0.08095010.0001502–0.07885030.0001504Y2013505–0.12905060.0001507–0.12825080.0001509–0.12615100.0001511Y2014512–0.09875130.0001514–0.09955150.0001516–0.09815170.0001518Y2015519–0.09075200.0001521–0.09515220.0001523–0.09225240.0001525d1526–0.15725270.0001528–0.14905290.0001530–0.13585310.0001532d2533–0.21745340.0001535–0.23475360.0001537–0.21155380.0001539d3540–0.14795410.0001542–0.15535430.0001544–0.13935450.0001546d4547–0.07465480.0001549–0.08025500.0001551–0.06875520.0001553d5554–0.15535550.0001556–0.13965570.0001558–0.12915590.0001560d6561–0.17085620.0001563–0.16905640.0001565–0.15785660.0001567d7568–0.28605690.0001570–0.28375710.0001572–0.27055730.0001574d8575–0.11265760.0001577–0.11545780.0001579–0.09875800.0001581d10582–0.22205830.0001584–0.21425850.0001586–0.20465870.0001588d11589–0.13095900.0001591–0.12585920.0001593–0.10215940.0001595d12596–0.01025970.2429598–0.02445990.00186000.00136010.8743602d13603–0.10696040.0001605–0.09936060.0001607–0.08706080.0001609d146100.01776110.23036120.01826130.17676140.04786150.0006616d15617–0.15046180.0001619–0.11866200.0001621–0.12216220.0001623d16624–0.21116250.0001626–0.19116270.0001628–0.17336290.0001630d17631–0.19536320.0001633–0.18596340.0001635–0.17586360.0001637d18638–0.06516390.0001640–0.07976410.0001642–0.06856430.0001644area6450.02276460.00016470.02566480.00016490.02426500.0001651area2652–0.00016530.0001654–0.00016550.0001656–0.00016570.0001658basement6590.02946600.00016610.02046620.00016630.02166640.0001665cons19606660.01476670.00016680.01546690.00016700.01806710.0001672cons19706730.03416740.00016750.02456760.00016770.03076780.0001679cons19806800.06086810.00016820.03716830.00016840.04156850.0001686cons19906870.13546880.00016890.13106900.00016910.13666920.0001693cons20006940.22736950.00016960.20796970.00016980.21396990.0001700cons20107010.22137020.00017030.22107040.00017050.20597060.0001707floor27080.02697090.00017100.02567110.00017120.02697130.0001714floor37150.02847160.00017170.02887180.00017190.03027200.0001721garage7220.02917230.00017240.02507250.00017260.02677270.0001728height729–0.00617300.0001731–0.00377320.0063733–0.00267340.0757735technology7360.04977370.00017380.04827390.00017400.04847410.0001742dcc743–0.02687440.0001745–0.02177460.0001747–0.02337480.0001749kindergarten750–0.01197510.0131752–0.00297530.5049754–0.00817550.0718756school757–0.03527580.0001759–0.02887600.0001761–0.03077620.0001763subway764–0.01607650.0001766–0.01887670.0001768–0.01867690.0001770duga1007710.04527720.00017730.02767740.00017750.03047760.0001777duga2007780.01457790.00017800.01387810.00017820.00777830.0042784duga3007850.01207860.00017870.00697880.00447890.00657900.0126791duga4007920.00747930.00747940.00217950.37887960.00287970.2807798duga5007990.00568000.05108010.00158020.53728030.00298040.2778805R-squared8060.87028070.8495808–809N810430758114307581243075813814815===== PAGE 10 (score 39, numeric-paren 0, table-words 6, var-words 11) =====816817International Journal of Strategic Property Management, 2018, 22(5): 358–371818367819in distance from the city centre negatively affects the 820value of the apartment. The negative relation between 821the distance and a property price was also observed in 822case of a kindergarten, a primary school and a subway 823station (Trojanek & Gluszak, 2017). What is important, 824estimation results confirm the findings from earlier stud-825ies in Poland which suggest that the proximity to an ur-826ban green area has a positive impact on property values 827(Zygmunt & Gluszak, 2015; Czembrowski & Kronenberg, 8282016; Trojanek, 2016).829The results suggest that proximity to urban green has 830a significant positive impact on apartments’ prices up 831within 400 metres distance, and the reported price pre-832miums were the highest within 100 meters distance band. 833The effect is statistically insignificant beyond this distance 834threshold. These findings are also in line with the existing 835body of evidence, which suggests that urban green affects 836property process within 500–600 meters radius, and the 837values of adjacent properties are affected the most (Lut-838zenhiser & Netusil, 2001; McCord et al., 2014).839On average the distance to urban green areas has a 840significant nonlinear impact on residential prices. Direct 841proximity to a park or an urban forest (up to 100-meter 842distance) increased apartment sale prices by 2.8%–3.1% 843(depending on a model) compared with apartments lo-844cated outside 500 distance band. The premium declined 845with distance – apartments located in the second distance 846band (100–200 meters, thus relatively close, but not in di-847rect proximity to the urban green) were still sold for a 848higher price than comparable apartments situated further 849from the green area, but the difference was only by about 8501.0%. The premium had somewhat steep distance decay 851function – proximity to the urban green did not affect 852significantly sales prices of otherwise comparable apart-853ments located outside 400 meters distance band from ur-854ban green. The result can be intuitively explained – it is 855worth to pay for proximity to the park/forest only if it is 856easily accessible (a short walk from the building entrance, 857plus sometimes a relaxing view from the apartment). If 858the trip to the park/forest demands longer walk/drive the 859direct benefits from urban green will diminish.860To discuss the impact of an urban green on a housing 861value of different housing estates, we estimated an alterna-862tive model with interaction terms. We interacted distance 863to an urban green (duga) with period1 and period2 dum-864my variables. Variable period1 takes value 1 if the dwelling 865was built during the years 1950–1989, otherwise it takes 0. 866Analogously, period2 is equal to 1 if the dwelling was built 867after 1989, and 0 otherwise. The models were estimated 868again, and the results were presented in Table 5.869Table 5. Hedonic regression estimation results (dependent variable is natural logarithm of sale price) (source: own elaboration)870OLS871WLS872Median QR873Variable874Coefficient875Probability876Coefficient877Probability878Coefficient879Probability880const88111.96858820.000188311.87468840.000188511.90748860.0001887Y2011888–0.00688890.0174890–0.00708910.0031892–0.00798930.0024894Y2012895–0.08058960.0001897–0.08108980.0001899–0.07969000.0001901Y2013902–0.12919030.0001904–0.12839050.0001906–0.12749070.0001908Y2014909–0.09899100.0001911–0.09969120.0001913–0.09849140.0001915Y2015916–0.09059170.0001918–0.09519190.0001920–0.09139210.0001922d1923–0.16029240.0001925–0.15219260.0001927–0.13879280.0001929d2930–0.21989310.0001932–0.23639330.0001934–0.21189350.0001936d3937–0.14519380.0001939–0.15339400.0001941–0.13549420.0001943d4944–0.07479450.0001946–0.08019470.0001948–0.06739490.0001950d5951–0.15479520.0001953–0.13969540.0001955–0.12849560.0001957d6958–0.17539590.0001960–0.17129610.0001962–0.15999630.0001964d7965–0.29009660.0001967–0.28699680.0001969–0.27389700.0001971d8972–0.12539730.0001974–0.12309750.0001976–0.10839770.0001978d10979–0.22599800.0001981–0.21809820.0001983–0.20569840.0001985d11986–0.13459870.0001988–0.12959890.0001990–0.10519910.0001992d12993–0.00419940.6345995–0.01889960.01619970.00699980.3836999d131000–0.113610010.00011002–0.106210030.00011004–0.090210050.00011006d1410070.000810080.958910090.006710100.631410110.036810120.00631013d151014–0.144910150.00011016–0.115310170.00011018–0.116110190.00011020d161021–0.215910220.00011023–0.196110240.00011025–0.177110260.00011027d171028–0.197210290.00011030–0.188110310.00011032–0.176310330.00011034d181035–0.062010360.00011037–0.074910380.00011039–0.063110400.00011041area10420.022710430.000110440.025710450.000110460.024310470.00011048area21049–0.000110500.00011051–0.000110520.00011053–0.000110540.00011055basement10560.028710570.000110580.019810590.000110600.021510610.00011062cons196010630.014410640.000210650.014310660.000110670.018110680.0001106910701071===== PAGE 11 (score 35, numeric-paren 0, table-words 5, var-words 10) =====107210733681074R. Trojanek et al. The effect of urban green spaces on house prices in Warsaw1075The results of the second group models estimation 1076(with variables which should distinguish the influence of 1077proximity to an urban green area on different housing es-1078tate) provide interesting results. On an average proximity 1079to an urban green areas did not have a significant impact 1080on sale prices (neither in OLS, not WLS and Median QR 1081model) in case of apartments built in years 1950–1989. 1082Even immediate proximity (an urban green area located 1083within 100 meters from the apartment) did not influence 1084house prices. No significant (at alpha = 0.05) difference 1085was observed for other distance bands.1086The results differed significantly in case of building 1087constructed after 1989. The impact of an urban green was 1088particularly strong in case of post-transformation housing 1089stock. Close vicinity (less than 100 m distance) to urban 1090green increased the sales prices of apartments in new resi-1091dential buildings by 8.0–8.6%, depending on a model. It 1092did not have a significant impact in case of apartments 1093located in socialist housing estates. The empirical results 1094provide arguments in support of the research hypothesis. 1095Apparently, relative lack of green areas around buildings 1096and dense development typical for post-transformation 1097residential development increases the hedonic value of 1098public urban green areas. In the Figure 3 duga’s regression 1099coefficients of quantile models are presented.1100OLS1101WLS1102Median QR1103Variable1104Coefficient1105Probability1106Coefficient1107Probability1108Coefficient1109Probability1110cons197011110.034211120.000111130.023811140.000111150.029711160.00011117cons198011180.056411190.000111200.032811210.000111220.040811230.00011124cons199011250.119711260.000111270.119011280.000111290.123711300.00011131cons200011320.211611330.000111340.196011350.000111360.202611370.00011138cons201011390.207411400.000111410.209811420.000111430.194911440.00011145floor211460.026811470.000111480.025411490.000111500.026711510.00011152floor311530.028111540.000111550.028811560.000111570.030511580.00011159garage11600.029311610.000111620.025611630.000111640.026511650.00011166height1167–0.005911680.00021169–0.003811700.00461171–0.002311720.10291173technology11740.049511750.000111760.047611770.000111780.049211790.00011180dcc1181–0.027011820.00011183–0.021911840.00011185–0.023711860.00011187kindergarten1188–0.011711890.01451190–0.002411910.56901192–0.008211930.05961194school1195–0.035411960.00011197–0.029011980.00011199–0.029412000.00011201subway1202–0.014612030.00011204–0.017612050.00011206–0.017512070.00011208period1 x duga1001209–0.003712100.45871211–0.005712120.181412130.002212140.63031215period1 x duga2001216–0.001512170.696512180.002012190.55451220–0.003412210.32841222period1 x duga30012230.000912240.80701225–0.000112260.96851227–0.001712280.61981229period1 x duga4001230–0.000912310.80151232–0.004612330.16721234–0.004312350.19401236period1 x duga50012370.004212380.296312390.001512400.661412410.000212420.94841243period2 x duga10012440.099312450.000112460.076812470.000112480.082612490.00011250period2 x duga20012510.031012520.000112530.027312540.000112550.020812560.00011257period2 x duga30012580.022012590.000112600.013612610.000212620.014712630.00011264period2 x duga40012650.013912660.000812670.008412680.019112690.011112700.00321271period2 x duga50012720.004012730.315212740.000312750.933912760.005112770.16651278R-squared12790.870912800.84981281–1282N1283430751284430751285430751286End of Table 51287–0.0212880.0012890.0212900.0412910.0612920.0812930.1012940129510012962001297300129840012995001300Coef.1301Distance to urban green area1302All13031950‐198913041990 and afer1305Figure 3. Regression coefficients of distance to urban green 1306areas (quantile models) (source: own elaboration)1307We believe that this finding can have interesting policy 1308implications. It is easy to dismiss positive effects of urban 1309green areas when the city is relatively green. Increasing de-1310mand for residential land and chaotic housing development 1311resulted in a decrease of green areas at the urban fringe, 1312where most of the new housing projects have been locat-1313ed. The relative shortage of green areas has increased the 1314implicit price of proximity to parks and urban forests in 131513161317===== PAGE 12 (score 24, numeric-paren 0, table-words 0, var-words 23) =====13181319International Journal of Strategic Property Management, 2018, 22(5): 358–37113203691321Warsaw. The residential value increase around parks and 1322value can be captured in property taxes, but currently, this 1323option is not exploited (in Poland nowadays property taxes 1324are based on the size of a dwelling, not on its value – the 1325location does not influence the tax amount). Currently, the 1326tax system in Poland does not allow to link taxation with 1327property value directly. Although the law allows for tax 1328rate differentiation (based on location and other factors), 1329in practice tax rates in most of the cities in Poland (Warsaw 1330included) are set uniformly at a highest possible level for 1331residential property (thus no differentiation is possible if we 1332want to increase revenues). The increase in revenues from 1333property tax revenues could be used to establish and main-1334tain urban green and recreational areas, that would gener-1335ate positive externalities and help to mitigate the negative 1336effects of air pollution, typical to major cities in Poland.1337Conclusions1338We found strong evidence that proximity to a park is posi-1339tively linked with apartment prices what is in line with 1340previous studies. On the average presence of an urban 1341green area within 100 meters from apartment increases 1342the price of the dwelling by 2.8% to 3.1%. However, the 1343effect of park/forest proximity on house prices is more 1344significant for newer apartments than those built before 13451989. We found that proximity to a park is particularly 1346important (and has a higher implicit price as a result) 1347in the case of buildings constructed after 1989. Modern 1348residential buildings are located smaller lots and built ac-1349cordingly to the density maximisation principle (both re-1350garding floor area ration and building coverage ratio). As 1351a result, green areas within gated communities are often 1352limited and can be treated like a club good, available only 1353to residents. In that context, proximity to a publicly ac-1354cessible green area outside the gates, only within a short 1355walking distance is highly beneficial and translates into a 1356significantly higher price of subject apartments.1357We did not find a significant effect of urban green prox-1358imity on prices of apartments located in housing estates 1359built during the years 1950–1989. This somewhat surpris-1360ing result may stem from the fact that contrary to modern 1361residential buildings there is an abundance of green ar-1362eas (children playgrounds, recreation facilities, conveni-1363ent walking paths) within housing estates for public use. 1364Presence of additional public park in proximity may not 1365result in significantly higher residential satisfaction, and as 1366a consequence will not be capitalised in apartment price.1367The latter result may shed new light on previous find-1368ings, notably Trojanek (2016) and Czembrowski and Kro-1369nenberg (2016), that overlooked potential differences in 1370hedonic prices of urban green areas for different types 1371of housing estates. The results may be interesting for (i) 1372housing developers, providing incentives for better loca-1373tion choices regarding residential projects, as well as for 1374(ii) urban planners, providing arguments to protect and 1375maintain existing parks and green belts in the city.1376References1377Anderson, L. M., & Cordell, H. K. (1998). Influence of trees on 1378residential property values in Athens, Gerogia (U.S.A.): a sur-1379vey based on actual sales prices. Landscape and Urban Planning, 138015(1-2), 153-164. https://doi.org/10.1016/0169-2046(88)90023-01381Anderson, S. T. (2000). The effect of open space on single-family, 1382residential home property values. Macalester Journal of Eco-1383nomics, 10, 261-270.1384Bark, R. H., Osgood, D. E., Colby, B. G., & Halper, E. B. (2011). 1385How do homebuyers value different types of green space?. 1386Journal of Agricultural and Resource Economics, 36(2), 395-1387415. http://jareonline.org.1388Bertaud, A., & Bertaud, M. A. (2000). The spatial development 1389of Warsaw Metropolitan area. Comments on Warsaw Devel-1390opment Strategy until the Year 2010 prepared for the World 1391Bank, 11 June 2000.1392Biao, Z., Gaodi, X., Bin, X., & Canqiang, Z. (2012). The effects of 1393public green spaces on residential property value in Beijing. 1394Journal of Resources and Ecology, 3(3), 243-252. 1395https://doi.org/10.5814/j.issn.1674-764x.2012.03.0071396Bolitzer, B., & Netusil, N. R. (2000). The impact of open spaces 1397on property values in Portland, Oregon. Journal of Environ-1398mental Management, 59(3), 185-193. 1399https://doi.org/10.1006/jema.2000.03511400Conway, D., Li, C. Q., Wolch, J., Kahle, C., & Jerrett, M. (2010). 1401A spatial autocorrelation approach for examining the effects 1402of urban greenspace on residential property values. Journal of 1403Real Estate Finance and Economics, 41(2), 150-169. 1404https://doi.org/10.1007/s11146-008-9159-61405Crompton, J. L. (2001). The impact of parks on property values: a 1406review of the empirical evidence. Journal of Leisure Research, 140733(1), 1-31. https://doi.org/10.1080/136067105003480601408Crompton, J. L. (2005). The impact of parks on property values: 1409empirical evidence from the past two decades in the United 1410States. Managing Leisure, 10(4), 203-218. 1411https://doi.org/10.1080/136067105003480601412Czembrowski, P., & Kronenberg, J. (2016). Hedonic pricing and 1413different urban green space types and sizes: insights into the 1414discussion on valuing ecosystem services. Landscape and Ur-1415ban Planning, 146, 11-19. 1416https://doi.org/10.1016/j.landurbplan.2015.10.0051417Des Rosiers, F., Thériault, M., Kestens, Y., & Villeneuve, P. (2002). 1418Landscaping and house values: an empirical investigation. 1419Journal of Real Estate Research, 23(1/2), 139.1420Degórska, B. (2012). Spatial growth of urbanised land within the 1421Warsaw Metropolitan Area in the first decade of the 21st cen-1422tury. Geographia Polonica, 85(3), 77-95. 1423https://doi.org/10.7163/GPol.2012.3.191424Do, A. Q., & Grudnitski, G. (1995). Golf-courses and residential 1425house prices – an empirical-examination. Journal of Real Es-1426tate Finance and Economics, 10(3), 261-270. 1427https://doi.org/10.1007/BF010969411428Dombrow, J., Rodriguez, M., & Sirmans, C. F. (2000). The market 1429value of mature trees in single-family housing markets. Ap-1430praisal Journal, 68(1), 39-43.1431Doss, C. R., & Taff, S. J. (1996). The influence of wetland type 1432and wetland proximity on residential property values. Journal 1433of Agricultural and Resource Economics, 21(1), 120-129.1434Dubin, R. A. (1988). Estimation of regression coefficients in the 1435presence of spatially autocorrelated error terms. The Review 1436of Economics and Statistics, 70(3), 466. 1437https://doi.org/10.2307/19267851438