1# RECORD b827a50a5b3fe735ba3475c0514ffecf35e91d7e2# TITLE: If you build it, they will change: Evaluating the impact of commuter rail stations on real estate values and neighborhood composition in the Rotterdam–The Hague metropolitan area, the Netherlands3# YEAR: 2021.0 VENUE: Journal of Transport and Land Use DOI: 10.5198/jtlu.2021.17954# AUTHORS: Amir Forouhar; D. van Lierop5# PAGES: 25 CANDIDATE TABLE PAGES (1-based): [10, 13, 15]6# ABSTRACT: In many metropolitan regions, transit-oriented developments are built to motivate the use of sustainable travel by promoting urban growth within walking distances of public transport stations. Changes in residential property values are a common way to assess the success of transit-oriented developments. However, studies that focus on property values alone have reported mixed effects. This paper attempts to evaluate the land value impact around commuter rail stations by analyzing the change in property values within the context of the transformation of socio-spatial neighborhood attributes. The study sets out to estimate the effect of Randstad Rail stations using real estate transaction data of residential properties and neighborhood socio-spatial attributes in the Rotterdam–the Hague metropolitan area of the Netherlands covering a period from 1985 to 2018. Adopting a quasi-experimental design, the effect is estimated for properties within different catchment zones around three commuter rail stations using a Difference-in-Differences Model and Multivariate Analysis of Variance. The results demonstrate the overall negative effect of the Randstad Rail on the value of residential properties at a distance equal to or less than 400 meters from the selected rail stations in the range of -18.8% to -11.5%. In contrast, a positive effect is observed for the residential properties located within a radius of 400 to 800 meters from the rail stations, which is estimated to be +15% to +33.2%. The findings also indicate a considerable socio-spatial transformation in the neighborhood composition after the opening of the rail stations in terms of neighborhood population density, land-use density, housing characteristics, and car ownership, which significantly affect the magnitude and direction of the impact.789===== PAGE 1 (score 12, numeric-paren 0, table-words 0, var-words 12) =====10111 12Introduction13In many regions, the introduction of urban rail transit has been associated with reductions in air pol-14lution, noise, traffic accidents, and road congestion (Anderson, 2014; Baum-Snow, Kahn, & Voith, 15Article history:16Received: May 11, 202017Received in revised form: 18January 15, 2021 19Accepted: May 2, 202120Available online: September 8, 21202122Copyright 2021 Amir Forouhar & Dea van Lierop23http://dx.doi.org/10.5198/jtlu.2021.179524ISSN: 1938-7849 | Licensed under the Creative Commons Attribution – Noncommercial License 4.0 25The Journal of Transport and Land Use is the official journal of the World Society for Transport and Land Use (WSTLUR) 26and is published and sponsored by the University of Minnesota Center for Transportation Studies. 27T J T L U http://jtlu.org28V. 14 N. 1 [2021] pp. 949–97329Amir Forouhar30Art University of Isfahan31a.forouhar@aui.ac.ir32Dea van Lierop33Utrecht University34D.S.vanlierop@uu.nl35If you build it, they will change: Evaluating the impact of 36commuter rail stations on real estate values and neighborhood 37composition in the Rotterdam–The Hague metropolitan area, 38the Netherlands39Abstract: In many metropolitan regions, transit-oriented developments 40are built to motivate the use of sustainable travel by promoting urban 41growth within walking distances of public transport stations. Changes 42in residential property values are a common way to assess the success of 43transit-oriented developments. However, studies that focus on property 44values alone have reported mixed effects. This paper attempts to evaluate 45the land value impact around commuter rail stations by analyzing the 46change in property values within the context of the transformation of 47socio-spatial neighborhood attributes. The study sets out to estimate 48the effect of Randstad Rail stations using real estate transaction data of 49residential properties and neighborhood socio-spatial attributes in the 50Rotterdam–the Hague metropolitan area of the Netherlands covering 51a period from 1985 to 2018. Adopting a quasi-experimental design, 52the effect is estimated for properties within different catchment zones 53around three commuter rail stations using a Difference-in-Differences 54Model and Multivariate Analysis of Variance. The results demonstrate 55the overall negative effect of the Randstad Rail on the value of residential 56properties at a distance equal to or less than 400 meters from the selected 57rail stations in the range of -18.8% to -11.5%. In contrast, a positive 58effect is observed for the residential properties located within a radius 59of 400 to 800 meters from the rail stations, which is estimated to be 60+15% to +33.2%. The findings also indicate a considerable socio-spatial 61transformation in the neighborhood composition after the opening of 62the rail stations in terms of neighborhood population density, land-use 63density, housing characteristics, and car ownership, which significantly 64affect the magnitude and direction of the impact. 65Keywords: Urban rail transit, transit-oriented development, property 66value, socio-spatial transformation, Difference-in-Differences Model, 67Randstad Rail, the Netherlands686970===== PAGE 10 (score 23, numeric-paren 0, table-words 2, var-words 15) =====717295873JOURNAL OF TRANSPORT AND LAND USE 14.174to match the CBS dataset with the housing sales data for each station at a given transaction year. The 75second limitation was due to the changes in the variables collected by the CBS and changes to variable 76definitions over time. For example, since 2006, the classification of companies and institutions was 77significantly changed, and thus, the number of companies and institutions in 2007 was no longer com-78parable with the previous data. In addition, some additional variables were also added to CBS reports 79such as the type of households, income of inhabitants, and social security conditions. Thus, we excluded 80the variables which were inconsistent in the CBS datasets between 1985 and 2018. The resulting dataset 81was exported into IBM SPSS Software (version 25) for each station and catchment zone. The ultimate 82dataset contains 12492, 8619, and 9961 valid transactions respectively for Barkel Westpolder, Pijnacker 83Zuid, and Nootdorp stations, covering a period from 1985 to 2018. The dataset encompasses informa-84tion related to housing structural characteristics, distance-based locational factors, land-use conditions, 85population factors, and transportation features which are further described in Table 1. 86Table 1. Descriptive statistics of contextual variables87Category88Variable Name89Variable 90Type91Minimum92Maximum93Mean94Std. 95Deviation96Source97Structural Factors98Transaction price (euro/m2)99scale10016610158991022085.32103695.61104Brainbay105Construction date106nominal107before 1500108from 2001 109onwards110-111-112Brainbay113Insulation114nominal115without 116insulation117fully insulated118-119-120Brainbay121Surface area (m2)122scale12370124450125139.9412633.56127Brainbay128Number of balconies129scale13001312132.08133.29134Brainbay135Number of floors136scale137113851392.79140.55141Brainbay142Number of toilets 143scale144114541461.38147.62148Brainbay149Garden area (m2)150scale15115152192515378.8515478,17155Brainbay156Garage157dummy158without 159garage160with garage161-162-163Brainbay164Location Factors165Distance to nearest CBD (km)166scale1671.8231685.3421692.732170.453171CBS172Distance to nearest arterial 173road (Km)174scale175.1431762.331177.972178.260179CBS180Distance to built-up area181dummy182inside built-183up area184outside built-185up area186-187-188Brainbay189Land-use Factors190Available land (ha)191scale19210.001931159.0019453.77195162.1196CBS197Land-use density (per km2)198scale199129.002002518.00201877.41202300.94203CBS204Number of houses 205scale20610.002077765.00208457.20209713.28210CBS211Number of business locations 212scale2131.00214870.0021533.5121656.04217CBS218Population Factors219Population density (per km2) 220scale22119.0022210752.002234920.102243360.88225CBS226Number of inhabitants 227scale22820.0022915940.002301319.862311935.99232CBS233Number of households 234scale23510.002367830.00237459.62238714.43239CBS240Number of immigrants241scale24232.502436376.00244255.41245422.93246CBS247Transportation 248Factors249Number of cars 250scale25115.002525350.00253568.81254807.73255CBS256Number of cars per household257scale258.702591.602601.23261.14262CBS263264265===== PAGE 11 (score 23, numeric-paren 0, table-words 3, var-words 17) =====266267959268Evaluating the impact of commuter rail stations on real estate values and neighborhood composition2696 270Results 271The results obtained from the DID model are reported in Table 2 for the selected stations and differ-272ent catchment zones. The table not only demonstrates the impact of opening the selected rail stations 273on property values, but also specifies the effect of housing structural characteristics, location factors, 274land-use conditions, population factors, and transportation features on the results. The DID models for 275residential properties explain between 0.651 and 0.828 percent of the variation in property values for 276the selected stations. Broadly speaking, the DID estimator demonstrates a negative effect of Randstad-277Rail on the value of residential properties located close to the stations at a distance equal to or less than 278400 meters from the rail stations, while a significant positive impact on the residential properties located 279within a radius of 400 to 800 meters from the rail stations. The results of the DID models indicate that 280the treatment effects of the RandstadRail on housing values within a 400-meter radius from the selected 281stations are estimated to be -18.8%, -11.5%, and -15.3% respectively for Barkel Westpolder, Pijnacker 282Zuid, and Nootdorp stations. In other words, the average selling price of treated properties declined 283after the opening of the selected stations, compared to the control properties which were located within 284a radius of 1000 to 1600 meters of the rail stations. On the contrary, the DID estimator demonstrates a 285significant positive effect for properties located in the second catchment zone (400 to 800 meters), com-286pared to the control properties. The treatment effects on the housing values within a radius of 400 to 287800 meters from the rail stations indicate +33.2%, +15%, and +25.8% for Barkel Westpolder, Pijnacker 288Zuid, and Nootdorp stations, respectively. 289The DID models also yield that the attributes of the dwelling units and the neighborhood com-290position in terms of accessibility, land-use, population, and transportation features also played a con-291siderable role in changing housing prices in both catchment zones around the rail stations. The models 292reveal that an increase in housing prices is often associated with improvements in housing facilities such 293as insulation type, the number of toilets and balconies, and the presence of a garage. Furthermore, an 294increase in house size (surface area and garden area) and a decrease in the building age of the properties 295have a positive relationship with the dependent variable. In the case of the Barkel Westpolder station, 296the total surface area, insulation type, and construction date explain the highest percentage of change 297in housing prices near to the station, compared to other structural factors. However, the number of 298balconies does not show a significant relationship. The surface area also has the most explanatory power 299in the cases of Pijnacker Zuid and Nootdorp stations, as it explains 37.1 and 31.3 percent of the change 300in housing prices close to the stations, respectively.301 In addition, the models demonstrate that all selected locational variables have the expected signs. 302This means that when the distance from the nearest CBD and arterial roads increases, the housing value 303declines when holding all other variables constant. Distance to the nearest CBD explains 54.2, 23.1, and 30420.5 percent of the changes in housing prices close to Barkel Westpolder, Pijnacker Zuid, and Nootdorp 305stations, respectively. The location of residential properties in built-up areas can also positively uplift the 306housing prices, compared to the properties located in non-urban areas.307Land availability is also estimated to have a strong relationship with the housing values in the DID 308models. This is mostly an important factor in the context of the Netherlands in which land reclamation 309has a long history (Hoeksema, 2007). The models reveal that the land availability explains between 21.3 310and 61.8 percent of the changes in housing prices close to the selected stations, when holding all other 311variables constant. Moreover, land-use density—which refers to the number of houses, offices, schools, 312and shopping facilities per square kilometers—as well as the number of business locations are positively 313correlated with the price in the catchment areas around the rail stations. For example, in the case of 314Nootdorp station, the land-use density explains 30.9 percent of the change in housing prices in the first 315316317===== PAGE 13 (score 24, numeric-paren 6, table-words 2, var-words 14) =====318319961320Evaluating the impact of commuter rail stations on real estate values and neighborhood composition321Table 2. Estimating the effect of RandstadRail on residential property values using the difference-in-differences model322Berkel Westpolder323Pijnacker Zuid324Nootdorp325Catchment 326Zone 1327Catchment 328Zone 2329Catchment 330Zone 1331Catchment 332Zone 2333Catchment 334Zone 1335Catchment 336Zone 2337Category338Overall R Square339.828340.760341.698342.724343.664344.651345F-test3461101.552***347803.288***348413.835***349526.631***350320.725***351331.419***352Structural Factors353Construction Date354.114***355.118***356.147***357.054***358.144***359.080***360Insulation 361.136***362.158***363.136***364.156***365.014366.048***367Surface area (m2)368.529***369.006370.371***371.078***372.313***373.036**374Number of toilets375.092***376.036***377.097***378.032***379.054***380.019381Garden area (m2)382.004**383.006***384.008385.005**386.099***387.091***388Number of balconies389.003390.050***391.094***392.056***393.048***394.025**395Garage 396.107***397.141***398.243***399.176***400.016401.192***402Location Factors403Distance to nearest 404CBD (Km)405-.542***406-.583***407-.231***408-.050***409-.205***410-.222***411Distance to nearest 412arterial road (Km)413-.060***414-.061***415-.067***416-.095***417-.022***418-.011***419Distance to the built-420up area421-.551***422-.615***423-.195***424-.025***425-.241***426-.233***427Land-use Factors428Available Land (ha)429.638***430.441***431.064***432.209***433.213***434.037435Land-use density (per 436km2)437.485***438-.060439.404***440.095***441.309***442.346***443Number of houses444-.195***445-.099**446-.048***447-.123**448-.030***449-.096**450Number of business 451locations452.136***453.112***454.099***455.077***456.093***457.167***458Population 459Factors460Population density 461(per km2)462.142***463.158***464.294***465.160***466.010***467.041***468Number of Immi-469grants 470.058***471.035***472.132***473.097***474.038**475.127***476Transportation 477Factors478Number of cars479-.583***480-.488***481-.213***482-.088**483-.274***484-.095***485Time486Month487.019***488.001489.027***490030***491.091***492.090***493Year494.072***495.287***496.150***497.163***498.276***499.422***500Difference-in-501Differences 502Estimation503Interaction 504(Time*Group)505-.188***506.332***507-.115***508.150***509-.153**510.258***511 512Legend: *p = .10; **p = .05; ***p = .01. Values represent Standardized Coefficients Beta for each variable. All entered inde-513pendent variables satisfy a VIF of less than 2.5.514The results obtained from the Multivariate Analysis of Variance (MANOVA) suggest that the 515neighborhoods located near the selected stations (in both the treated catchment zones) experienced a 516considerable socio/spatial transformation after the opening of the rail stations compared to the control 517neighborhoods (Table 3). The findings indicate that the residential properties closest to the rail stations 518tended to become denser after the opening of the rail stations, since the average surface area of the 519residential properties decreased in the range of -6.29 to -75.25 square meters, and the number of floors 520significantly increased around the selected stations, compared to the control properties. For example, 521while the average surface area of the residential properties located in the control zone of Barkel Westpol-522523524===== PAGE 14 (score 13, numeric-paren 0, table-words 0, var-words 10) =====525526962527JOURNAL OF TRANSPORT AND LAND USE 14.1528der station increased from 136.86 to 144.10 square meters following the opening of the rail station, the 529treated properties close to the rail station experienced a decrease from 159.73 to 137.70 square meters. 530Regarding the changes in the number of floors, the changes near the selected station were more severe 531compared to the control neighborhoods. For instance, the residential properties located in the control 532zone of Nootdorp station experienced an increase in their number of floors from 2.53 to 2.92, however 533the treated properties experienced a more dramatic increase from 2.26 to 2.93 following the opening of 534the rail station.535 The results of the MANOVA also demonstrate a considerable increase in the number of houses, 536offices, schools, and shopping facilities (land-use density) in both the treated catchment zones following 537the opening of the selected rail stations, in such a way that the land-use density increased by 352.30, 538302.04, and 676.97 per square kilometer, respectively within a 400-meter radius from Barkel Westpol-539der, Pijnacker Zuid, and Nootdorp stations. Furthermore, in the neighborhoods close to the selected 540stations, the total number of houses increased by 500.81, 795.42, and 1466.21, and the number of busi-541ness locations rose by 100.89, 39.18, and 166.85, respectively for Barkel Westpolder, Pijnacker Zuid, 542and Nootdorp stations. The model also yields that the increase in land-use density significantly hap-543pened more intensely close to the rail stations compared to the control neighborhoods. As the land-use 544density increased by 243.86, 147.85, and 232.3 in the control neighborhoods of Barkel Westpolder, Pi-545jnacker Zuid, and Nootdorp stations respectively, the catchment zones close to the stations experienced 546a sharper increase in the land-use density by 596.16, 449.88, and 909.27. This means, that as expected, 547the areas near the new stations gained greater increases in land-use density. 548Population density also increased in both the treated catchment zones following the opening of the 549selected stations in the range of 1738 to 4638 inhabitants per square kilometer. For example, while the 550population density in the control zone of Pijnacker Zuid station increased from 6902.20 to 7512.66 per 551square kilometers following the opening of the rail station, the treated neighborhoods closer to the rail 552station experienced a sharper increase, changing from 2246.88 to 5185.00 per square kilometer. Subse-553quently, a significant increase in the number of inhabitants and households is observed around all three 554stations. In addition, the percentage of immigrant populations had a significant positive correlation 555with the introduction of the selected rail stations in the range of 3.05 to 18.17 percent. In this respect, 556the neighborhoods located within a 400-meter radius from the stations experienced a stronger wave of 557immigration, compared to the control neighborhoods and even the second catchment zones around 558the selected stations. Considering the change in the transportation features also revealed that opening 559the rail stations significantly decreased the number of cars per household in the range of 0.022 to 0.188 560in both the treated catchment zones around the selected stations even though the total number of cars 561increased considerably in both treated neighborhoods up to 2046. It should be mentioned that while 562the number of cars in the control neighborhoods around the selected stations often dropped during the 563period of the study, the treated neighborhoods close to the stations experienced a considerable increase 564in the number of cars, most likely due to the increases in population density. 565566567===== PAGE 15 (score 85, numeric-paren 68, table-words 1, var-words 8) =====568569963570Evaluating the impact of commuter rail stations on real estate values and neighborhood composition571Table 3. Socio-spatial transformation of the neighborhoods around the rail stations using Multivariate Analysis of Variance572Berkel Westpolder573Pijnacker Zuid574Nootdorp575Category576Catchment 577Zone 1578Catchment 579Zone 2580Catchment 581Zone 1582Catchment 583Zone 2584Catchment 585Zone 1586Catchment 587Zone 2588Partial Eta Squared5890.8675900.8655910.8755920.8905930.8135940.699595F-test5961252.142***5971310.622***5983109.525***5992165.279***6001985.660***6014256.494***602Structural 603Factors604Surface area (m2)605-29.273***606-24.543***607-6.292***6080.514 609-75.250***610-20.218***611Number of floors6120.238**6130.169***6140.290***6150.030 ***6160.289***617.133***618Land-use 619Factors620Land-use density (per 621km2)622352.299***623277.253***624302.036***62569.991***626676.974***627286.010***628Number of houses629500.811***630468.106***631795.419***63287.286***6331466.213***634729.485***635Number of business 636locations 637100.888***638-11.263 63939.176***640-38.242***641166.854***64217.804***643Population 644Factors645Population density (per 646km2)6474638.663***6483789.933***6492327.668***6501738.397***6512611.191***6522505.587***653Number of inhabitants6541674.402***6551505.673***6562287.514***657216.215**6584509.106***6592448.001***660Number of households661523.827***662480.425***663790.390***664113.011**6651458.585***666729.036***667Percentage of immi-668grants66918.168***6703.054***6718.578***6724. 017***67313.503***6743.330***675Transportation 676Factors 677Number of cars678602.496***679556.758***680934.350***681156.388**6822046.924***6831016.806***684Number of cars per 685household686-0.087***687-0.087***688-0.188***689-0.188***690-0.022 **691-0.116***692Legend: *p = .10; **p = .05; ***p = .01. Values represent Unstandardized Coefficients Beta for each variable.6937 694Discussion695The results of this study contribute to the existing literature by providing a possible explanation of how 696property values can be differently affected by the presence of commuter rail stations through the use of 697a context-sensitive approach that connects the change in property values to the socio-spatial transforma-698tion of neighborhoods around the rail stations. The results reveal that the introduction of the Rands-699tadRail brought a negative effect on the value of residential properties at a distance equal to or less than 700400 meters from the rail stations. However, the results indicate a significant positive effect for properties 701located within a radius of 400 to 800 meters around the stations. The findings are consistent with the 702literature in which housing prices located relatively close to rail stations are affected negatively, but those 703at an intermediate distance and which are beyond the externality effects, benefit from the transporta-704tion access provided by the rail stations. For example, in the context of the Netherlands, Debrezion 705et al. (2011) found that the neighborhoods located immediately next to new stations were affected 706negatively due to the externality effects as the houses located adjacent to the commuter rail stations (less 707than 250 meters) sold for lower prices compared to houses located only a few hundred meters from the 708station. Furthermore, the properties located between 250 and 500 meters of a station were on average 709about 25% more expensive than dwellings located at a distance of 15 kilometers or farther. Similarly, in 710Dubai, the United Arab Emirates, Mohammad et al. (2017) estimated a negative effect between -9% 711and -17.7% on the value of residential prices within 0.5 kilometer of the stations, but a positive impact 712of about 13% for properties located within 701 to 900 meters of the rail stations. AlQuhtani and Anjo-713mani (2019) also concluded that the residential properties that were closer to rail stations in the Dallas 714715716===== PAGE 16 (score 20, numeric-paren 0, table-words 0, var-words 20) =====717718964719JOURNAL OF TRANSPORT AND LAND USE 14.1720Fort-Worth metropolitan area experienced lower changes in housing value compared to the properties 721located farther away from stations. Since the present study captured the negative impacts close to the 722selected stations nearly a decade after the opening of the stations, the results contradict the findings by 723Pan (2019), Lee and Sener (2017), Cervero and Duncan (2002), and Chen et al. (1998) who reported 724that accessibility effects finally negate nuisance as a transit system matures or a significant positive effect 725would be expected to appear after a reasonable timespan.726The literature insinuates that factors such as noise, vibration, and pollution, which are associated 727with being too close to a rail line or station, particularly in the case of heavy or commuter rail, may be 728responsible for observed negative effects (AlQuhtani & Anjomani, 2019; Bowes & Ihlanfeldt, 2001; 729Debrezion et al., 2011; Forouhar & Hasankhani, 2018; Martínez & Viegas, 2009; Mohammad et al., 7302017). These are also important factors in the context of the Netherlands in which commuter rail lines 731and stations are often located above ground and near residential neighborhoods. A study on the impact 732of the nuisance and vibration caused by high-speed railways on housing prices in the Netherlands indi-733cated that the noise produced by Dutch railways negatively affected property prices located adjacent to 734the rail lines at up to 12 percent (Theebe, 2004). However, the results of the current study indicate that 735the nuisance effects could simply be a part of the reason behind the mixed estimated effects, and further 736explanation could lie in the transformation of the surrounding neighborhoods. The results of the cur-737rent study showed that the opening of the rail stations was often accompanied by a significant increase 738in population density as well as an increase in the percentage of immigrants residing in the adjacent 739neighborhoods. This observation is in line with empirical studies conducted by Wang and Chen (2018), 740Papa et al. (2008), and Transecon Consortium (2003) who reported that high-speed rail significantly 741contributes to a shift of residents from the city center to peripheral areas. It is worth mentioning that in 742the current study, the increases in population density and percentage of immigrants were higher in the 743immediate station area (≤400 meters) compared to the 400-800 meters zone. The observed increase in 744the percentage of immigrants could be related to the fact that newcomers may seek home locations that 745are within walking distance to transit stations to have good access to potential jobs, retail, education, and 746other opportunities. At the same time, the number of business locations considerably increased close to 747the selected rail stations compared to the farther neighborhoods. This finding is in line with those of Pan 748(2019) and Bowes and Ihlanfeldt (2001), who reported that increases in population density and im-749migration around rail stations could expand the retail activities and attract businesses and firms to locate 750around the station areas. However, Bowes and Ihlanfeldt (2001) explained that the population changes 751around rail stations may give rise to socio-cultural tensions like crime rates which would not be desir-752able for local residents (Bowes & Ihlanfeldt, 2001). A study by Forouhar and Hasankhani (2018) also 753reported that the local identity and privacy of the immediate neighborhoods can become profoundly 754damaged following the opening of a rail station due to increased population density and the presence 755of strangers. 756On the other hand, the current study also captured a significant transformation in the character-757istics of houses and built-up areas close to the case stations which could be related to the negative ex-758ternalities in the immediate neighborhoods. The findings showed that the residential properties tended 759to become denser after the opening of the rail stations, since the average surface area of the residential 760properties decreased and the number of floors in residential buildings increased around the selected sta-761tions, compared to the control properties. The results also demonstrated a considerable increase in the 762number of houses, offices, schools, and shopping facilities (land-use density) in both the treated catch-763ment zones following the opening of the selected rail stations. Thus, the overall findings imply that there 764is an increasing tendency toward a denser spatial structure around the selected rail stations. The results 765are consistent with the findings by Kasraian et al., (2016) who reported that rail infrastructure prompted 766further urban growth in the Randstad, the Netherlands, suggesting that urbanization is more likely to 767