1# RECORD dde22a8458111f4d97568d09ce7efc13ea6e96aa2# TITLE: Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw3# YEAR: 2020.0 VENUE: European Journal of Transport and Infrastructure Research DOI: 10.18757/ejtir.2020.20.2.39794# AUTHORS: Marcin Torzewski5# PAGES: 19 CANDIDATE TABLE PAGES (1-based): [6, 9, 11, 15, 16, 17, 18]6# ABSTRACT: This paper identifies the relation between metro proximity and the prices of residential properties in Warsaw. The analysis is based on 2006-2013 AMRON-SARFiN data. The results show that metro has a positive impact on property prices if it is located at most 1.5 km from the property. The greatest increase in price of 8-13 per cent is observed when metro is located 400-800 meters from the property. The results indicate that when the metro transportation is poor, it causes a large increase in property prices, which is not comparable to the impact of other transports, such as bus or tram.789===== PAGE 1 (score 10, numeric-paren 0, table-words 0, var-words 10) =====1011Issue 20(2), 2020, pp. 41-59 12https://doi.org/10.18757/ejtir.2020.20.2.3979 13 14 15 16EJTIR 17 18 ISSN: 1567-7141 19http://ejtir.tudelft.nl/ 20 21Public transport accessibility and the prices of nearby 22properties: the case of the first metro line in Warsaw 23 24Marcin Torzewski1 25Faculty of Management, University of Warsaw, Poland 26This paper identifies the relation between metro proximity and the prices of residential properties 27in Warsaw. The analysis is based on 2006-2013 AMRON-SARFiN data. The results show that metro 28has a positive impact on property prices if it is located at most 1.5 km from the property. The 29greatest increase in price of 13 per cent is observed when metro is located up to 800 meters from 30the property. The results indicate that when the metro transportation is poor, it causes a large 31increase in property prices, which is not comparable to the impact of other transports, such as bus 32or tram. 33 34Keywords: hedonic pricing model, metro proximity, property prices, public transport, spatial models, 35Warsaw housing market. 361. Introduction 37The subway / metro transportation plays a crucial role in the urbanization and urban development 38processes because it depends on its own specific infrastructure that is mostly located underground. 39This distinct feature of subway gives it a relative advantage over other means of transport, such as 40bus, trolleybus or tram. By its underground location metro allows its users to reach downtown city 41areas that struggle with high traffic limiting the number of vehicles and reducing the travel time. 42Due to the high number of users it also attracts the location of businesses and services, which seek 43for potential clients. 44Given the characteristics of the metro transportation network, its accessibility should be capitalized 45by the value of the surrounding properties. There exists an extensive number of studies that look 46at the relation between metro accessibility and the prices of nearby properties. Most of these studies 47focus on the cities with a highly developed railway network infrastructure, such as cities in the 48USA (Damm et al., 1980, Armstrong and Rodríguez, 2006), South Korea (Bae et al., 2003), Taiwan 49(Lin and Hwang, 2004), the UK (Du and Mulley, 2007), Portugal (Martínez and Viegas, 2009), Italy 50(Pagliara and Papa, 2011 and Gallo, 2018), Greece (Efthymiou and Antoniou, 2013) or China (Zhang 51 521 A: Szturmowa 1/3, 02-678 Warsaw, Poland. T: +48 791 676 646 E: m.torzewski@hotmail.com 535455===== PAGE 6 (score 37, numeric-paren 0, table-words 6, var-words 15) =====5657EJTIR 20(2), 2020, pp.41-59 58 5946 60Torzewski 61Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 62 63Table 1. 64Summary statistics of property prices per square meter and distance related 65variables 66Variable 67Mean 68Std. Deviation 69Variable 70Mean Std. Deviation 71Price per m2 (PLN) 727968.957 732815.194 74Distance (km): Tram 1.727 752.065 76Distance (km): Center 6.442 773.532 78Tram: 0 - 200 m 790.176 800.380 81Centrum: 0 - 3 km 820.187 830.390 84Tram: 200 - 400 m 850.183 860.386 87Centrum: 3 - 5 km 880.209 890.407 90Tram: 400 - 800 m 910.179 920.383 93Centrum: 5- 7 km 940.188 950.391 96Tram: > 800 m 970.463 980.499 99Centrum: 7 - 10 km 1000.234 1010.423 102Distance (km): Bus 1030.179 1040.124 105Centrum: > 10 km 1060.182 1070.386 108Bus: 0 -100 m 1090.276 1100.447 111Distance (km): Metro 1123.934 1133.413 114Bus: 100 - 250 m 1150.514 1160.500 117Metro: 0 - 400 m 1180.075 1190.263 120Bus: 250 - 400 m 1210.156 1220.363 123Metro: 400 - 800 m 1240.133 1250.339 126Bus: > 400 m 1270.054 1280.225 129Metro: 800 - 1500 m 1300.103 1310.304 132Number of 133observations 1345576 135Metro: > 1500 m 1360.689 1370.463 138 139Source: Own calculations based on AMRON-SARFiN data and geographical coordinates of the properties 140recoded from their address and geographical coordinates of metro, bus and tram stops in Warsaw. 141 142Table 2. 143Summary statistics of physical characteristics of properties 144Variable 145Mean Std. Deviation 146Variable 147Mean Std. Deviation 148Square meters 14962.514 31.000 150Floor 0 1510.158 1520.365 153Square meters: 0-35 1540.128 1550.334 156Floor 1 1570.166 1580.372 159Square meters: 35-60 1600.467 1610.499 162Floor 2 1630.200 1640.400 165Square meters: 60-90 1660.262 1670.440 168Floor 3 1690.178 1700.383 171Square meters: >90 1720.144 1730.351 174Floor 4 1750.125 1760.331 177Parking 1780.232 1790.422 180Floor 5-6 1810.105 1820.307 183Primary market 1840.256 1850.436 186Floor 7-9 1870.047 1880.211 189Secondary market 1900.744 1910.436 192Floor >=10 1930.020 1940.140 195Age of the building 19627.606 29.685 197Private ownership 1980.448 1990.497 200Age of the building: 0-2 2010.287 2020.453 203Perpetual ownership 2040.548 2050.498 206Age of the building: 2-10 2070.191 2080.393 209Other type of ownership 0.004 2100.064 211Age of the building: 10-50 2120.304 2130.460 214Number of 215observations 2165576 217Age of the building: >50 2180.218 2190.413 220Source: Own calculations based on AMRON-SARFiN data 221Table 3. 222Summary statistics of neighborhood related variables 223Variable 224Mean 225Std. Dev. 226Living condition index 2270.494 2280.155 229Population density (thousands/m2) 2304.390 2312.762 232Share of green areas 23312.633 23416.642 235Share of recreational areas 2366.128 2376.821 238Crimes per 1000 inhabitants 23932.867 24015.822 241Children per 1000 children in kindergartens 242757.333 243138.412 244Number of districts 24518 246Source: Own calculations based on GUS (2012) and MOBP (2013) reports. 247The next set of control variable denoted as neighborhoodm includes characteristics of the surrounding 248measured at the district level. More specifically these measures account for population density, 249share of green areas, share of recreation areas, number of crimes per 1000 inhabitants, number of 250children attending kindergarten per 1000 children, living condition index. The data are extracted 251from GUS (2012) and MOBP (2013) reports. The summary statistics for the neighborhood related 252variables are shown in Table 3. The estimated equation additionally controls for time trend, 253measured by year and quarters fixed effects (denoted by time). 254255256===== PAGE 7 (score 16, numeric-paren 0, table-words 5, var-words 6) =====257258EJTIR 20(2), 2020, pp.41-59 259 26047 261Torzewski 262Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 263 264In the analysis, we also test the hypothesis concerning the interaction effect between various public 265transportation modes. We thus ask whether the importance of the metro accessibility weakens 266when other means of transportation are located closer than metro. To do that we define a dummy 267variable denoted as MOT which takes the value of 1 if other means of transport (bus or tram) are 268closer to the property than the nearest metro station and the relative distance between them is more 269than 250 meters; in other case the variable takes a value of 0. Put it differently, the variable equals 270to 1 if there is bus or tram available that is closer than metro and the distance to metro compared 271to bus or tram is quite far (i.e. more than 250 meters). This variable is then interacted with variables 272measuring the distance to the nearest metro station. The estimated coefficients on the interactions 273reveal whether the effect of metro proximity is affected by the relative distance of metro to other 274means of transportation. It is expected that when the other means of transportation are located 275closer than metro (i.e. the variable MOT takes the value of 1), the role of metro in shaping the 276property price will be lower. In other words, it is expected that the estimated coefficients on the 277interaction terms will prove to be negative. 2783.2 Alternative methods: SEM and SAR models 279Hedonic pricing models, though widely used in applied research on property prices, suffer from 280several important limitations. These mostly relate to heteroscedasticity, collinearity between the 281variables, and the spatial dependence/autocorrelation. 282While the first problem – heteroscedasticity – is a common problem embedded in ordinary least 283squares estimation, it can be overcame with the use of logarithmic form and robust standard errors. 284Similarly, in order to avoid collinearity between variables, the VIF test can be performed and one 285of the linearly dependent variables can be dropped from the model. The third issue – the spatial 286dependence – is, however, more serious and therefore requires more attention. 287More specifically, the problem of spatial dependence relates to the correlation of values within a 288given group of close objects, which may lead to the misspecification and estimation bias (Anselin 289and Arribas-Bel, 2012). It is worth noting that despite its significance, this issue has not been 290properly addressed in the number of previous studies on the topic (Damm et al., 1980; Bae et al., 2912003, Li et al., 2017). Spatial autocorrelation is closely linked to the so-called first law of geography 292(Tobler, 1970), which says that objects which are near to each other are more similar (correlated) 293than those that are distant. In the context of real estate, this means that observation units 294(properties) in close proximity should exhibit similar effects (prices) to neighboring units. In this 295case, the dependence can be viewed as error dependence that violates the crucial assumption of 296OLS of the uncorrelation of the error terms (LeSage and Pace, 2009). 297To deal with this issue, we re-estimate the models by applying: 1) spatial error model (SEM), and 2) 298spatial lag model (SAR); (Anselin and Arribas-Bel, 2012). 299𝐴 spatial error model (SEM) introduces the spatial autocorrelation into the error term, as: 300 301ln(𝑝𝑟𝑖𝑐𝑒)𝑖= 𝛼0 + 𝛼1𝑚𝑒𝑡𝑟𝑜𝑖+ 𝛼2𝑏𝑢𝑠𝑖+ 𝛼3𝑡𝑟𝑎𝑚𝑖+ 𝛼4𝑐𝑒𝑛𝑡𝑒𝑟𝑖+302 ∑303𝛼𝑘(𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙)𝑖,𝑘304𝑙305𝑘=5306+ ∑307𝛼𝑚(𝑛𝑒𝑖𝑔ℎ𝑏𝑜𝑟ℎ𝑜𝑜𝑑)𝑖,𝑚308𝑛309𝑚=𝑙+1310+ 𝛼𝑛+1(𝑡𝑖𝑚𝑒)𝑖+ 𝜀𝑖 311(2) 312 313𝜀𝑖= 𝜆𝑾𝜀𝑖+ 𝑢𝑖 314 315where 𝜆 is an autoregressive coefficient, 𝑾 is a spatial weight matrix and 𝑢 is an iid. error term. 316A spatial lag model (SAR), on the other hand, regresses the dependent variable (property price) on 317the set of independent variables, as well as on the property price reweighted by the spatial matrix. 318Spatial dependence is included here as an additional regressor, that is a spatially lagged dependent 319variable: 320 321ln(𝑝𝑟𝑖𝑐𝑒)𝑖= 𝛼0 + 𝜌𝑾∗ln(𝑝𝑟𝑖𝑐𝑒)𝑖+ 𝛼1𝑚𝑒𝑡𝑟𝑜𝑖+ 𝛼2𝑏𝑢𝑠𝑖+ 𝛼3𝑡𝑟𝑎𝑚𝑖+ 𝛼4𝑐𝑒𝑛𝑡𝑒𝑟𝑖+322 ∑323𝛼𝑘(𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙)𝑖,𝑘324𝑙325𝑘=5326+ ∑327𝛼𝑚(𝑛𝑒𝑖𝑔ℎ𝑏𝑜𝑟ℎ𝑜𝑜𝑑)𝑖,𝑚+ 𝛼𝑛+1(𝑡𝑖𝑚𝑒)𝑖+ 𝜀𝑖328𝑛329𝑚=𝑙+1330 331(3) 332 333334335===== PAGE 9 (score 106, numeric-paren 74, table-words 3, var-words 15) =====336337EJTIR 20(2), 2020, pp.41-59 338 33949 340Torzewski 341Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 342 3433-5 per cent in case it is located within 200 meters. The negative impact of a close proximity of bus 344and tram stops on the property selling price might be related to the noise caused by both bus and 345tram, which as opposed to metro (which in Warsaw is entirely located underground) operate on 346regular roads. As expected, close proximity to the city center of up to 3 km is associated with the 347highest property selling prices; such properties are even by 26 per cent more expensive than similar 348properties located at the city’s peripheries (more than 10 km away from the city center). 349Table 4. 350Estimation results using OLS, SEM and SAR methods 351Method: 352OLS 353SEM 354SAR 355Variable: 356coef/se 357coef/se 358coef/se 359Metro: 0 - 400 m 3600.131*** 3610.131*** 3620.084*** 363 364(0.025) 365(0.020) 366(0.014) 367Metro: 400 - 800 m 3680.132*** 3690.133*** 3700.088*** 371 372(0.026) 373(0.017) 374(0.012) 375Metro: 800 - 1500 m 3760.108*** 3770.110*** 3780.069*** 379 380(0.021) 381(0.017) 382(0.013) 383Metro: >1500 m 384(base) 385(base) 386(base) 387Bus: 0 -100 m 388-0.044** 389-0.027 390-0.031** 391 392(0.023) 393(0.020) 394(0.016) 395Bus: 100 - 250 m 396-0.019 397-0.001 398-0.007 399 400(0.020) 401(0.019) 402(0.015) 403Bus: 250 - 400 m 404-0.006 4050.007 4060.003 407 408(0.018) 409(0.019) 410(0.016) 411Bus: >400 m 412(base) 413 (base) 414 (base) 415Tram: 0 - 200 m 416-0.049*** 417-0.032* 418-0.030*** 419 420(0.018) 421(0.016) 422(0.012) 423Tram: 200 - 400 m 424-0.006 4250.003 4260.002 427 428(0.018) 429(0.015) 430(0.011) 431Tram: 400 - 800 m 4320.045** 4330.044*** 4340.035*** 435 436(0.019) 437(0.014) 438(0.010) 439Tram: >800 m 440(base) 441(base) 442(base) 443Centrum: 0 - 3 km 4440.250*** 4450.251*** 4460.173*** 447 448(0.051) 449(0.028) 450(0.021) 451Centrum: 3 - 5 km 4520.212*** 4530.209*** 4540.152*** 455 456(0.047) 457(0.023) 458(0.017) 459Centrum: 5- 7 km 4600.126*** 4610.118*** 4620.092*** 463 464(0.033) 465(0.020) 466(0.014) 467Centrum: 7 - 10 km 4680.103*** 4690.104*** 4700.078*** 471 472(0.034) 473(0.016) 474(0.011) 475Centrum: >10 km 476(base) 477(base) 478(base) 479Constant 4808.348*** 4818.346*** 4825.431*** 483 484(0.129) 485(0.050) 486(0.142) 487No. of observations 4885 576 489Notes: *** p<0.01, ** p<0.05, * p<0.1. The models control for: usable area, floor location, ownership, age of the 490building, parking place, type of the market, district characteristics (living conditions index, population density, 491share of green and recreational areas, crime rate and childcare availability), and year fixed effects. 492Table 5. 493Direct, indirect and total effects derived for SAR model 494 495Direct 496Indirect 497Total 498Metro: 0 - 400 m 4990.086 5000.041 5010.127 502Metro: 400 - 800 m 5030.090 5040.043 5050.133 506Metro: 800 - 1500 m 5070.071 5080.033 5090.104 510Bus: 0 -100 m 511-0.032 512-0.015 513-0.047 514Bus: 100 - 250 m 515-0.007 516-0.003 517-0.011 518Bus: 250 - 400 m 5190.003 5200.001 5210.004 522Tram: 0 - 200 m 523-0.031 524-0.015 525-0.046 526Tram: 200 - 400 m 5270.002 5280.001 5290.003 530Tram: 400 - 800 m 5310.036 5320.017 5330.052 534535536===== PAGE 10 (score 16, numeric-paren 0, table-words 4, var-words 7) =====537538EJTIR 20(2), 2020, pp.41-59 539 54050 541Torzewski 542Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 543 544Centrum: 0 - 3 km 5450.176 5460.083 5470.260 548Centrum: 3 - 5 km 5490.155 5500.074 5510.229 552Centrum: 5- 7 km 5530.094 5540.044 5550.138 556Centrum: 7 - 10 km 5570.079 5580.038 5590.117 560 561The coefficients obtained on variables measuring physical characteristics of the properties are in 562line with the theoretical predictions and expectations (see Appendix Table A.1). First, the most 563expensive properties are the smallest ones (0-35 square meters). Second, the floor location matters 564for the price: compared to the price of the properties located on the ground floor, the price of 565properties located up to the 7th floor are higher. Perpetual ownership is also associated with a 566higher selling price of the property (of around 3 to 5 per cent). Compared to properties located in 567old buildings (more than 50 years old), the properties located in new buildings aged up to 2 years 568are by around 12-18 per cent more expensive, and properties located in buildings aged 2-10 by 569even 16-24 per cent more expensive. We also observe that the selling price per square meter is by 570around 7-10 per cent lower at the primary market than at the secondary one. Importantly, the 571coefficients obtained on spatial parameters in SEM and SAR models (see Appendix Table A.1) are 572significant and reveal an existence of the spatial dependency between the neighboring properties. 573Finally, we assess how the proximity to other means of transportation (bus or tram) affects the 574uncovered relationship between metro accessibility and the prices of nearby properties. To do that 575we interact variables reflecting metro proximity with the variable MOT measuring the relative 576distance between metro and other means of transportation. The main results on the interaction 577terms are presented in Table 6 (Appendix Table A.3 gives the full set of coefficients). The MOT 578variable is interacted with variables indicating that metro is within 0-400 and 400-800 meters but 579not with the variable indicating that it is within 800-1500 meters. This is because of the definition 580of MOT variable, which takes a value of 1 if other means of transport (bus or tram) are closer to the 581property than the nearest metro station and the relative distance between them is more than 250 582meters. If metro is located within 800-1500 meters from the property it is very unlikely that there 583are no other transports that are closer to the property; in other words in this case the MOT variable 584nearly always takes a value of 1. 585The results obtained from the model with the interactions show that if the nearest metro station is 586very close to the property – within 0-400 meters – the fact that other means of transport are closer 587relative to metro does not matter for the property price, as indicated by the insignificant and close 588to zero interaction term between metro 0-400 and MOT variables in all three models. Even if other 589means of transport are closer than metro, the fact that metro is within 0-400 meters from the 590property increases its selling price by 13 per cent. As for the metro proximity of 400-800 meters the 591results indicate that when the nearest metro is within 400-800 meters and other means of transport 592are even further away or are in a close proximity to metro (variable MOT takes a value of 0), the 593presence of metro has a high and significant impact on the price of the property of 14 per cent to 594even 18 per cent (depending on the model used). On the other hand, if the nearest metro is within 595400-800 meters and the bus or tram is closer than metro (and at the same time metro is quite far 596away compared to bus or tram, i.e. more than 250 meters), the importance of the presence of metro 597for the property price is lowered by around 1-5 per cent. This result is, however, significant at 5% 598significance level only in the case of OLS model; for other two models we also find negative 599coefficients but they are not statistically significant. The OLS results thus show that if the nearest 600metro is relatively far (400-800m) and there are other transports that are much closer, we may not 601expect as high increase in property prices caused by metro as in the case when there are no other 602transports that are closer than metro or other transports are nearly as far as the metro is (i.e. the 603distance between metro and tram/bus is less than 250 meters). 604 605 606607608===== PAGE 11 (score 42, numeric-paren 25, table-words 4, var-words 5) =====609610EJTIR 20(2), 2020, pp.41-59 611 61251 613Torzewski 614Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 615 616Table 6. 617Estimation results with the interactions: OLS, SEM and SAR methods 618Method: 619OLS 620SEM 621SAR 622Variable: 623coef/se 624coef/se 625coef/se 626Metro: 0 - 400 m 6270.130*** 6280.132*** 6290.084*** 630 631(0.026) 632(0.021) 633(0.015) 634Metro: 0 - 400 m x MOT 6350.001 636-0.006 637-0.002 638 639(0.025) 640(0.038) 641(0.036) 642Metro: 400 - 800 m 6430.181*** 6440.143*** 6450.115*** 646 647(0.034) 648(0.032) 649(0.028) 650Metro: 400 - 800 m x MOT 651-0.055** 652-0.011 653-0.030 654 655(0.025) 656(0.032) 657(0.013) 658Metro: 800 - 1500 m 6590.108*** 6600.110*** 6610.069*** 662 663(0.021) 664(0.017) 665(0.029) 666No. of observations 6675 576 668Notes: the same as in Table 4; Variable MOT is defined as a dummy variable equal to 1 if the nearest metro station 669is further away than 250 meters from other transport (bus or tram) and 0 if the nearest metro station is closer than 670250 meters from other transport (bus or tram). 671 672Table 7. 673Direct, indirect and total effects derived for SAR model with interactions 674 675Direct 676Indirect 677Total 678Metro: 0 - 400 m 6790.086 6800.041 6810.127 682Metro: 0 - 400 m x MOT 683-0.002 684-0.001 685-0.003 686Metro: 400 - 800 m 6870.117 6880.055 6890.172 690Metro: 400 - 800 m x MOT 691-0.030 692-0.014 693-0.045 694Metro: 800 - 1500 m 6950.071 6960.033 6970.104 6985. Conclusion 699This paper provides quantitative evidence on how the proximity to metro affects selling prices of 700properties in Warsaw. The relation was assessed using AMRON-SARFiN 2006-2013 data and three 701estimation methods: OLS regression with clustered standard errors, SEM and SAR models. 702We found that metro proximity matters for the property prices and properties that are located close 703to the metro stations are valued more by the market. The positive impact of metro on prices of 704nearby properties – that are located up to 800 meters from the metro - is estimated to be 13 per cent. 705The positive effect is revealed in all three models and the two spatial models – SEM and SAR – 706additionally show spatial dependency of the neighboring properties. The effect is considerably 707higher than the findings received from meta-analyses of existing studies on the topic, which report 708virtually zero effect. As opposed to previous studies that analyzed the situation in cities with well-709developed metro infrastructure, the current study shows how metro affects prices of nearby 710properties in Warsaw, in which the metro system is far less developed. The high positive effect of 711metro availability on property prices that has been uncovered suggests that the impact of metro is 712much larger when it is not easily accessible and is not considered as the main public transportation 713mode. Furthermore, the fact that housing affordability in Warsaw is low may also lead to a greater 714differentiation in property prices, with relatively more expensive properties having highly valued 715characteristics, such as easy access to metro system. 716At the same time, the obtained results are somehow lower than the ones provided for Warsaw by 717Bazyl (2009). This difference may stem from the fact that in the present study we use transaction 718prices, while the analysis in Bazyl (2009) is based on asking prices. In consequence, higher positive 719effects of metro found with the use of asking prices than actual transaction prices suggests that 720metro is perceived as highly valuable and it is expected to be capitalized in the prices of residential 721properties even by a higher value. 722723724===== PAGE 12 (score 10, numeric-paren 0, table-words 0, var-words 10) =====725726EJTIR 20(2), 2020, pp.41-59 727 72852 729Torzewski 730Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 731 732Finally, because Warsaw public transportation system may be characterized by poor metro 733infrastructure but rich system of other public transports, we also analyze how the availability of 734other means of transport affects the uncovered high positive impact of metro. The results obtained 735from all three models suggest that the fact that other transports are closer than metro is not 736important for the high increase in the property selling price associated with metro if metro is close 737enough to the property (within 0-400 meters). As for the metro proximity of 400-800 meters from 738the property, the results suggest that other means of transport that are closer than metro lower the 739high positive impact of metro. The significant finding is, however, found only for OLS model. 740Several comments concerning the data and the analysis are called for. Given that the available data 741covered property transactions during 2006-2013 period, the present analysis examined the impact 742of metro proximity on property prices in Warsaw focusing on the first metro line. Since 2015 743Warsaw metro network has been extended and a second metro line is now in operation. The results 744presented in this paper could be re-assessed by additionally including data from 2015 onwards to 745see whether the uncovered relations hold. An interesting future extension would be the 746comparison of the impacts of newly opened and the existing metro stations on property prices. 747Other possible extension could include the analysis of selected districts to reveal the regional 748heterogeneity in the estimated effect of metro proximity on property prices. 749Acknowledgments 750This work was supported by the grant No. 2016/21/N/HS4/00080 from the National Science Centre, 751Poland. The author would like to thank the participants of 2019 Spatial Statistics conference. 752References 753 754Al-Mosaind, M., K. Dueker, and J. Strathman (1994). Light Rail Transit Stations and Property Values: A 755Hedonic Price Approach. Transportation Research Record, 1400, 90–94. 756Anselin, L. (1988). Spatial econometrics: Methods and models. Kluwer Academic Publishers, Dordrecht. 757Anselin, L., and D. Arribas-Bel (2012). Spatial fixed effects and spatial dependence in a single cross-758section. Papers in Regional Science, 92(1), 3-17. 759Armstrong, R.J., and Rodríguez, D.A. (2006). An Evaluation of the Accessibility Benefits of Commuter 760Rail in Eastern Massachusetts using Spatial Hedonic Price Functions. Transportation, 33, 21–43. 761Bae, C.C., Jun, M., and Park, H. (2003). The impact of Seoul’s subway Line 5 on residential property 762values Transport Policy, 10, 85–94. 763Bajic, V. (1983). The effects of a new subway line on housing prices in metropolitan Toronto. Urban 764Studies, 2, 147–158. 765Bazyl, M. (2009). Hedonic price model for Warsaw housing market. Working Papers, Department of 766Applied Econometrics, Warsaw School of Economics, 8, 1–16. 767Borkowska, M., Rozwadowska, M., Śleszyński, J., and T. Żylicz (2001). Environmental Amenities on the 768Housing Market in Warsaw, Hedonic Price Method Research. Ekonomia, 3, 70-82. 769Bowes, D. R., and Ihlanfeldt, K. R. (2001). Identifying the Impacts of Rail Transit Stations on Residential 770Property Values. Journal of Urban Economics, 50(1), 1–25. 771Cervero, R., and Kang, C. D. (2011). Bus rapid transit impacts on land uses and land values in Seoul, 772Korea, Transport Policy, 18(1), 102–116. 773774775===== PAGE 15 (score 164, numeric-paren 139, table-words 1, var-words 7) =====776777EJTIR 20(2), 2020, pp.41-59 778 77955 780Torzewski 781Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 782 783Appendix 784Table A.1. 785 Full estimation output for the main model using OLS, SEM and SAR methods 786 Method: 787 788OLS 789SEM 790SAR 791 Variable: 792 793coef/se 794coef/se 795coef/se 796Distance variables Metro: 0 - 400 m 7970.131*** 7980.131*** 7990.084*** 800 801(0.025) 802(0.020) 803(0.014) 804Metro: 400 - 800 m 8050.132*** 8060.133*** 8070.088*** 808 809(0.026) 810(0.017) 811(0.012) 812Metro: 800 - 1500 m 8130.108*** 8140.110*** 8150.069*** 816 817(0.021) 818(0.017) 819(0.013) 820Centrum: 0 - 3 km 8210.250*** 8220.251*** 8230.173*** 824 825(0.051) 826(0.028) 827(0.021) 828Centrum: 3 - 5 km 8290.212*** 8300.209*** 8310.152*** 832 833(0.047) 834(0.023) 835(0.017) 836Centrum: 5- 7 km 8370.126*** 8380.118*** 8390.092*** 840 841(0.033) 842(0.020) 843(0.014) 844Centrum: 7 - 10 km 8450.103*** 8460.104*** 8470.078*** 848 849(0.034) 850(0.016) 851(0.011) 852Tram: 0 - 200 m 853-0.049*** 854-0.032* 855-0.030*** 856 857(0.018) 858(0.016) 859(0.012) 860Tram: 200 - 400 m 861-0.006 8620.003 8630.002 864 865(0.018) 866(0.015) 867(0.011) 868Tram: 400 - 800 m 8690.045** 8700.044*** 8710.035*** 872 873(0.019) 874(0.014) 875(0.010) 876Bus: 0 -100 m 877-0.044** 878-0.027 879-0.031** 880 881(0.023) 882(0.020) 883(0.016) 884Bus: 100 - 250 m 885-0.019 886-0.001 887-0.007 888 889(0.020) 890(0.019) 891(0.015) 892Bus: 250 - 400 m 893-0.006 8940.007 8950.003 896 897(0.018) 898(0.019) 899(0.016) 900Physical 901characteristics 902Square meters: 0-35 9030.054*** 9040.055*** 9050.052*** 906 907(0.011) 908(0.010) 909(0.010) 910Square meters: 60-90 9110.000 912-0.009 913-0.005 914 915(0.010) 916(0.008) 917(0.008) 918Square meters: >90 9190.011 920-0.002 9210.003 922 923(0.043) 924(0.011) 925(0.010) 926Floor 1 9270.047*** 9280.038*** 9290.038*** 930 931(0.018) 932(0.011) 933(0.011) 934Floor 2 9350.036* 9360.020* 9370.025** 938 939(0.022) 940(0.011) 941(0.011) 942Floor 3 9430.034* 9440.019* 9450.024** 946 947(0.020) 948(0.011) 949(0.011) 950Floor 4 9510.041* 9520.033*** 9530.034*** 954 955(0.024) 956(0.012) 957(0.012) 958Floor 5-6 9590.044** 9600.034*** 9610.035*** 962 963(0.022) 964(0.013) 965(0.013) 966Floor 7-9 9670.024 9680.024 9690.021 970 971(0.024) 972(0.017) 973(0.017) 974Floor >=10 9750.053 9760.068*** 9770.056** 978 979(0.033) 980(0.024) 981(0.024) 982Perpetual ownership 9830.033*** 9840.033*** 9850.032*** 986 987(0.012) 988(0.009) 989(0.008) 990Other type of ownership 991-0.001 9920.019 9930.018 994 995(0.044) 996(0.052) 997(0.050) 998Parking 9990.001 10000.007 10010.005 1002 1003(0.011) 1004(0.008) 1005(0.008) 1006Age of the building: 0-2 10070.129*** 10080.135*** 10090.119*** 1010 1011(0.020) 1012(0.014) 1013(0.013) 1014Age of the building: 2-10 10150.177*** 10160.171*** 10170.157*** 101810191020===== PAGE 16 (score 85, numeric-paren 52, table-words 2, var-words 14) =====10211022EJTIR 20(2), 2020, pp.41-59 1023 102456 1025Torzewski 1026Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 1027 1028 1029(0.020) 1030(0.013) 1031(0.013) 1032Age of the building: 10-50 1033-0.014 10340.002 1035-0.005 1036 1037(0.016) 1038(0.010) 1039(0.010) 1040Primary market 1041-0.082*** 1042-0.069*** 1043-0.065*** 1044 1045(0.019) 1046(0.011) 1047(0.011) 1048Neighborhood 1049characteristics 1050Living condition index 10510.169 10520.219*** 10530.122*** 1054 1055(0.110) 1056(0.065) 1057(0.047) 1058Population density (thousands/m2) 1059-0.009 1060-0.013*** 1061-0.008*** 1062 1063(0.006) 1064(0.004) 1065(0.003) 1066Share of green areas 1067-0.001 1068-0.001*** 1069-0.001** 1070 1071(0.001) 1072(0.001) 1073(0.000) 1074Share of recreational areas 10750.006*** 10760.006*** 10770.005*** 1078 1079(0.002) 1080(0.001) 1081(0.001) 1082Crimes per 1000 inhabitants 1083-0.000 1084-0.001** 1085-0.001** 1086 1087(0.001) 1088(0.001) 1089(0.000) 1090Children per 1000 children in kindergartens 10910.000 10920.000 10930.000* 1094 1095(0.000) 1096(0.000) 1097(0.000) 1098Constant 10998.348*** 11008.346*** 11015.431*** 1102 1103(0.129) 1104(0.050) 1105(0.142) 1106Lambda 1107 11080.366*** 1109 1110 1111 1112(0.017) 1113 1114Rho 1115 1116 11170.335*** 1118 1119 1120 1121(0.016) 1122Number of observations 11235576 1124 1125Table A.2. 1126 Direct, indirect and total effects derived for full estimation output for SAR 1127model 1128 1129 1130Direct 1131Indirect 1132Total 1133Distance variables 1134Metro: 0 - 400 m 11350.086 11360.041 11370.127 1138Metro: 400 - 800 m 11390.090 11400.043 11410.133 1142Metro: 800 - 1500 m 11430.071 11440.033 11450.104 1146Centrum: 0 - 3 km 11470.176 11480.083 11490.260 1150Centrum: 3 - 5 km 11510.155 11520.074 11530.229 1154Centrum: 5- 7 km 11550.094 11560.044 11570.138 1158Centrum: 7 - 10 km 11590.079 11600.038 11610.117 1162Tram: 0 - 200 m 1163-0.031 1164-0.015 1165-0.046 1166Tram: 200 - 400 m 11670.002 11680.001 11690.003 1170Tram: 400 - 800 m 11710.036 11720.017 11730.052 1174Bus: 0 -100 m 1175-0.032 1176-0.015 1177-0.047 1178Bus: 100 - 250 m 1179-0.007 1180-0.003 1181-0.011 1182Bus: 250 - 400 m 11830.003 11840.001 11850.004 1186Physical 1187characteristics 1188Square meters: 0-35 11890.053 11900.025 11910.079 1192Square meters: 60-90 1193-0.005 1194-0.002 1195-0.007 1196Square meters: >90 11970.003 11980.002 11990.005 1200Floor 1 12010.039 12020.018 12030.057 1204Floor 2 12050.026 12060.012 12070.038 1208Floor 3 12090.025 12100.012 12110.037 1212Floor 4 12130.035 12140.016 12150.051 1216Floor 5-6 12170.036 12180.017 12190.053 1220Floor 7-9 12210.021 12220.010 12230.032 1224Floor >=10 12250.058 12260.027 12270.085 1228Perpetual ownership 12290.033 12300.015 12310.048 1232Other type of ownership 12330.018 12340.009 12350.027 1236Parking 12370.005 12380.002 12390.008 1240Age of the building: 0-2 12410.122 12420.058 12430.179 1244Age of the building: 2-10 12450.161 12460.076 12470.237 1248Age of the building: 10-50 1249-0.005 1250-0.002 1251-0.008 1252Primary market 1253-0.066 1254-0.031 1255-0.098 125612571258===== PAGE 17 (score 146, numeric-paren 117, table-words 1, var-words 12) =====12591260EJTIR 20(2), 2020, pp.41-59 1261 126257 1263Torzewski 1264Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 1265 1266Neighborhood 1267characteristics 1268Living condition index 12690.125 12700.059 12710.184 1272Population density (thousands/m2) 1273-0.008 1274-0.004 1275-0.012 1276Share of green areas 1277-0.001 12780.000 1279-0.001 1280Share of recreational areas 12810.005 12820.002 12830.008 1284Crimes per 1000 inhabitants 1285-0.001 12860.000 1287-0.001 1288Children per 1000 children in kindergartens 0.000 12890.000 12900.000 1291 1292Table A.3. 1293Full estimation output for the model with interaction terms using OLS, SEM and 1294SAR methods 1295 Method: 1296 1297OLS 1298SEM 1299SAR 1300 Variable: 1301 1302coef/se 1303coef/se 1304coef/se 1305Distance 1306variables 1307Metro: 0 - 400 m 13080.130*** 13090.132*** 13100.084*** 1311 1312(0.026) 1313(0.021) 1314(0.015) 1315Metro: 0 - 400 m x MOT 13160.001 1317-0.006 1318-0.002 1319 1320(0.025) 1321(0.038) 1322(0.036) 1323Metro: 400 - 800 m 13240.181*** 13250.143*** 13260.115*** 1327 1328(0.034) 1329(0.032) 1330(0.028) 1331Metro: 400 - 800 m x MOT 1332-0.055** 1333-0.011 1334-0.030 1335 1336(0.025) 1337(0.032) 1338(0.029) 1339Metro: 800 - 1500 m 13400.108*** 13410.110*** 13420.069*** 1343 1344(0.021) 1345(0.017) 1346(0.013) 1347Centrum: 0 - 3 km 13480.251*** 13490.251*** 13500.173*** 1351 1352(0.051) 1353(0.028) 1354(0.021) 1355Centrum: 3 - 5 km 13560.213*** 13570.209*** 13580.153*** 1359 1360(0.046) 1361(0.023) 1362(0.017) 1363Centrum: 5- 7 km 13640.126*** 13650.118*** 13660.092*** 1367 1368(0.033) 1369(0.020) 1370(0.014) 1371Centrum: 7 - 10 km 13720.104*** 13730.104*** 13740.078*** 1375 1376(0.033) 1377(0.016) 1378(0.011) 1379Tram: 0 - 200 m 1380-0.048*** 1381-0.032* 1382-0.030** 1383 1384(0.018) 1385(0.016) 1386(0.012) 1387Tram: 200 - 400 m 1388-0.006 13890.003 13900.002 1391 1392(0.018) 1393(0.015) 1394(0.011) 1395Tram: 400 - 800 m 13960.045** 13970.044*** 13980.035*** 1399 1400(0.019) 1401(0.014) 1402(0.010) 1403Bus: 0 -100 m 1404-0.042* 1405-0.026 1406-0.030* 1407 1408(0.021) 1409(0.020) 1410(0.016) 1411Bus: 100 - 250 m 1412-0.016 1413-0.001 1414-0.006 1415 1416(0.019) 1417(0.019) 1418(0.015) 1419Bus: 250 - 400 m 1420-0.006 14210.007 14220.003 1423 1424(0.018) 1425(0.019) 1426(0.016) 1427Physical 1428characteristics 1429Square meters: 0-35 14300.054*** 14310.055*** 14320.052*** 1433 1434(0.011) 1435(0.010) 1436(0.010) 1437Square meters: 60-90 14380.001 1439-0.009 1440-0.005 1441 1442(0.010) 1443(0.008) 1444(0.008) 1445Square meters: >90 14460.011 1447-0.002 14480.003 1449 1450(0.043) 1451(0.011) 1452(0.010) 1453Floor 1 14540.047*** 14550.038*** 14560.038*** 1457 1458(0.018) 1459(0.011) 1460(0.011) 1461Floor 2 14620.036* 14630.020* 14640.025** 1465 1466(0.022) 1467(0.011) 1468(0.011) 1469Floor 3 14700.033* 14710.019* 14720.024** 1473 1474(0.020) 1475(0.011) 1476(0.011) 1477Floor 4 14780.041* 14790.032*** 14800.034*** 1481 1482(0.024) 1483(0.012) 1484(0.012) 1485Floor 5-6 14860.044** 14870.034*** 14880.035*** 1489 1490(0.022) 1491(0.013) 1492(0.013) 1493Floor 7-9 14940.024 14950.024 14960.021 149714981499===== PAGE 18 (score 103, numeric-paren 80, table-words 1, var-words 12) =====15001501EJTIR 20(2), 2020, pp.41-59 1502 150358 1504Torzewski 1505Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 1506 1507 1508(0.024) 1509(0.017) 1510(0.017) 1511Floor >=10 15120.053 15130.068*** 15140.056** 1515 1516(0.033) 1517(0.024) 1518(0.024) 1519Perpetual ownership 15200.033*** 15210.033*** 15220.032*** 1523 1524(0.012) 1525(0.009) 1526(0.008) 1527Other type of ownership 15280.001 15290.019 15300.019 1531 1532(0.045) 1533(0.052) 1534(0.050) 1535Parking 15360.002 15370.007 15380.005 1539 1540(0.011) 1541(0.008) 1542(0.008) 1543Age of the building: 0-2 15440.129*** 15450.135*** 15460.119*** 1547 1548(0.020) 1549(0.014) 1550(0.013) 1551Age of the building: 2-10 15520.177*** 15530.171*** 15540.157*** 1555 1556(0.020) 1557(0.013) 1558(0.013) 1559Age of the building: 10-50 1560-0.014 15610.002 1562-0.005 1563 1564(0.016) 1565(0.010) 1566(0.010) 1567Primary market 1568-0.082*** 1569-0.069*** 1570-0.065*** 1571 1572(0.019) 1573(0.011) 1574(0.011) 1575Neighborhood 1576characteristics 1577Living condition index 15780.169 15790.219*** 15800.122*** 1581 1582(0.110) 1583(0.065) 1584(0.047) 1585Population density (thousands/m2) 1586-0.009 1587-0.013*** 1588-0.008*** 1589 1590(0.006) 1591(0.004) 1592(0.003) 1593Share of green areas 1594-0.001 1595-0.001*** 1596-0.001** 1597 1598(0.001) 1599(0.001) 1600(0.000) 1601Share of recreational areas 16020.006*** 16030.006*** 16040.005*** 1605 1606(0.002) 1607(0.001) 1608(0.001) 1609Crimes per 1000 inhabitants 1610-0.000 1611-0.001** 1612-0.001** 1613 1614(0.001) 1615(0.001) 1616(0.000) 1617Children per 1000 children in kindergartens 16180.000 16190.000 16200.000* 1621 1622(0.000) 1623(0.000) 1624(0.000) 1625Constant 16268.344*** 16278.345 16285.435*** 1629 1630(0.129) 1631(0.050) 1632(0.143) 1633Lambda 1634 16350.366*** 1636 1637 1638 1639(0.017) 1640 1641Rho 1642 1643 16440.335*** 1645 1646 1647 1648(0.016) 1649Number of observations 16505576 1651 1652 1653 165416551656===== PAGE 19 (score 27, numeric-paren 0, table-words 1, var-words 14) =====16571658EJTIR 20(2), 2020, pp.41-59 1659 166059 1661Torzewski 1662Public transport accessibility and the prices of nearby properties: the case of the first metro line in Warsaw 1663 1664Table A.4. 1665Direct, indirect and total effects derived for full estimation output for SAR model 1666with interactions 1667 1668 1669Direct 1670Indirect 1671Total 1672Distance variables 1673Metro: 0 - 400 m 16740.086 16750.041 16760.127 1677Metro: 0 - 400 m x MOT 1678-0.002 1679-0.001 1680-0.003 1681Metro: 400 - 800 m 16820.117 16830.055 16840.172 1685Metro: 400 - 800 m x MOT 1686-0.030 1687-0.014 1688-0.045 1689Metro: 800 - 1500 m 16900.071 16910.033 16920.104 1693Centrum: 0 - 3 km 16940.177 16950.084 16960.261 1697Centrum: 3 - 5 km 16980.156 16990.074 17000.230 1701Centrum: 5- 7 km 17020.094 17030.044 17040.139 1705Centrum: 7 - 10 km 17060.080 17070.038 17080.118 1709Tram: 0 - 200 m 1710-0.030 1711-0.014 1712-0.045 1713Tram: 200 - 400 m 17140.002 17150.001 17160.003 1717Tram: 400 - 800 m 17180.036 17190.017 17200.053 1721Bus: 0 -100 m 1722-0.030 1723-0.014 1724-0.045 1725Bus: 100 - 250 m 1726-0.006 1727-0.003 1728-0.009 1729Bus: 250 - 400 m 17300.003 17310.001 17320.004 1733Physical 1734characteristics 1735Square meters: 0-35 17360.053 17370.025 17380.079 1739Square meters: 60-90 1740-0.005 1741-0.002 1742-0.007 1743Square meters: >90 17440.003 17450.002 17460.005 1747Floor 1 17480.039 17490.018 17500.057 1751Floor 2 17520.026 17530.012 17540.038 1755Floor 3 17560.025 17570.012 17580.036 1759Floor 4 17600.034 17610.016 17620.051 1763Floor 5-6 17640.036 17650.017 17660.052 1767Floor 7-9 17680.021 17690.010 17700.032 1771Floor >=10 17720.058 17730.027 17740.085 1775Perpetual ownership 17760.033 17770.016 17780.048 1779Other type of ownership 17800.019 17810.009 17820.029 1783Parking 17840.005 17850.003 17860.008 1787Age of the building: 0-2 17880.122 17890.058 17900.179 1791Age of the building: 2-10 17920.161 17930.076 17940.236 1795Age of the building: 10-50 1796-0.005 1797-0.002 1798-0.008 1799Primary market 1800-0.066 1801-0.031 1802-0.098 1803Neighborhood 1804characteristics 1805Living condition index 18060.125 18070.059 18080.184 1809Population density (thousands/m2) 1810-0.008 1811-0.004 1812-0.012 1813Share of green areas 1814-0.001 18150.000 1816-0.001 1817Share of recreational areas 18180.005 18190.002 18200.007 1821Crimes per 1000 inhabitants 1822-0.001 18230.000 1824-0.001 1825Children per 1000 children in kindergartens 0.000 18260.000 18270.000 1828 1829 1830 1831 1832