1# RECORD 7059eb726206d6598f1ef88ae9a43a35a98c69372# TITLE: The relationship of single-family detached house prices with neighborhood walkability and disaster risk3# YEAR: 2023.0 VENUE: PLoS ONE DOI: 10.1371/journal.pone.02926574# AUTHORS: Hongjik Kim; H. Baba; C. Shimizu; Kimihiro Hino5# PAGES: 23 CANDIDATE TABLE PAGES (1-based): [6, 7, 10, 11, 12]6# ABSTRACT: People’s preferences regarding their neighborhood environment can vary depending on their socioeconomic status and the cities where they live. This study aims to discern the relationship between neighborhood environment factors and single-family detached house sales by sale price and by central and noncentral cities. We analyzed sale prices in the Tokyo Metropolitan Area from 2015 to 2020. The neighborhood environment was assessed using flood/sediment risk and neighborhood walkability measured by net residential density, intersection density, and facility density (walking opportunity). Flood and sediment risk is a major concern that restricts the available land and is included as a negative aspect of the neighborhood environment, taking the topographic features into consideration. A comparison of the results showed that the preference for neighborhood walkability varies by socioeconomic status as well as by target cities. For most facility types, the number of walking opportunities within walking distance from houses was found to be positively related to the sale price of single-family detached houses in all quantiles. The relationship of house price with population and intersection density was found to vary depending on the price level, with a negative relationship with the sale price of relatively more expensive houses being exhibited. People who considered buying houses with relatively higher sale prices were found to devalue houses located in flood/sediment-hazardous areas more. However, it was also found that the negative relationship was slightly mitigated in the highest quantile of sale prices for houses in areas with a moderate flood risk (maximum flooding depth: 3–5 m). Plains near rivers with amenities offer high walkability but pose a flood risk, resulting in a trade-off between flood risk and neighborhood walkability. The findings suggest the use of indices representing diverse preferences in accordance with the target socioeconomic status when policymakers assess the neighborhood environment.789===== PAGE 1 (score 12, numeric-paren 0, table-words 0, var-words 12) =====1011RESEARCH ARTICLE12The relationship of single-family detached13house prices with neighborhood walkability14and disaster risk15Hongjik KimID1*, Hiroki BabaID2, Chihiro Shimizu2, Kimihiro HinoID1161 Department of Urban Engineering, Graduate School of Engineering, The University of Tokyo, Tokyo,17Japan, 2 Center for the Promotion of Social Data Science Education and Research, Hitotsubashi University,18Tokyo, Japan19* hongjik.kim@ua.t.u-tokyo.ac.jp20Abstract21People’s preferences regarding their neighborhood environment can vary depending on22their socioeconomic status and the cities where they live. This study aims to discern the rela-23tionship between neighborhood environment factors and single-family detached house24sales by sale price and by central and noncentral cities. We analyzed sale prices in the25Tokyo Metropolitan Area from 2015 to 2020. The neighborhood environment was assessed26using flood/sediment risk and neighborhood walkability measured by net residential density,27intersection density, and facility density (walking opportunity). Flood and sediment risk is a28major concern that restricts the available land and is included as a negative aspect of the29neighborhood environment, taking the topographic features into consideration. A compari-30son of the results showed that the preference for neighborhood walkability varies by socio-31economic status as well as by target cities. For most facility types, the number of walking32opportunities within walking distance from houses was found to be positively related to the33sale price of single-family detached houses in all quantiles. The relationship of house price34with population and intersection density was found to vary depending on the price level, with35a negative relationship with the sale price of relatively more expensive houses being exhib-36ited. People who considered buying houses with relatively higher sale prices were found to37devalue houses located in flood/sediment-hazardous areas more. However, it was also38found that the negative relationship was slightly mitigated in the highest quantile of sale39prices for houses in areas with a moderate flood risk (maximum flooding depth: 3–5 m).40Plains near rivers with amenities offer high walkability but pose a flood risk, resulting in a41trade-off between flood risk and neighborhood walkability. The findings suggest the use of42indices representing diverse preferences in accordance with the target socioeconomic sta-43tus when policymakers assess the neighborhood environment.44Introduction45Urban agglomeration has proceeded through the gathering of firms and workers in cities,46alongside urban growth [1–3]. Urban agglomeration creates clusters of amenities that hinge47PLOS ONE48PLOS ONE | https://doi.org/10.1371/journal.pone.029265749October 11, 2023501 / 2351a111111111152a111111111153a111111111154a111111111155a111111111156OPEN ACCESS57Citation: Kim H, Baba H, Shimizu C, Hino K (2023)58The relationship of single-family detached house59prices with neighborhood walkability and disaster60risk. PLoS ONE 18(10): e0292657. https://doi.org/6110.1371/journal.pone.029265762Editor: Bing Xue, Institute for Advanced63Sustainability Studies, GERMANY64Received: June 10, 202265Accepted: September 26, 202366Published: October 11, 202367Copyright: © 2023 Kim et al. This is an open68access article distributed under the terms of the69Creative Commons Attribution License, which70permits unrestricted use, distribution, and71reproduction in any medium, provided the original72author and source are credited.73Data Availability Statement: The data used in the74study are owned by third party organizations. The75contact information for the organizations to which76data requests may be sent is as follows: Sales price77data: SUUMO research center (which is an institute78of the Recruit Co., Ltd.; https://www.suumo-79research.com/) Shibaura 3 Chome, 12-7-12,80Minato City, Tokyo, 108-0023, Japan Telepoint81Pack! and Advanced Digital Road Map Database822015: Center for Spatial Information Science, The83University of Tokyo (https://joras.csis.u-tokyo.ac.84jp/); each also can be purchased by the vendors85(Telepoint Pack! from Zenrin Co., Ltd.: https://868788===== PAGE 6 (score 75, numeric-paren 57, table-words 2, var-words 12) =====8990of the maximum projected rainfall [64]. Furthermore, the ministry has designated hazardous91sediment areas based on surveys of topography, geology, and land use [64]; the provision of92hazard information (e.g., hazard maps) is mandatory before transactions take place [65].93Variable descriptions94Table 1 shows the characteristics of single-family detached houses sold between 2015 and952020. The sale price was recorded for each single-family detached house sold. The characteris-96tics of each house, such as years/months of transactions, floor space, land area, age of house,97road width in front of the house, and main window orientation, were also recorded and served98as control variables. Years/months of transactions were included on a quarterly basis to99account for the impact of temporal factors on sale price (e.g., inflation). The maximum flood-100ing depth of the area where each house was located was assessed and the results were classified101Table 1. Characteristicsa of single-family detached houses sold between 2015 and 2020.102Target area103Central cities104Other cities105Sale price (JPY10,000)1064,529 ± 3,2211075,314 ± 3,7971083,487 ± 1,764109Year of transactions (month is omitted)11020151119,082 (12.7)1125,357 (13.2)1133,725 (12.2)11420161158,841 (12.4)1165,010 (12.3)1173,831 (12.5)11820171199,032 (12.7)1205,089 (12.5)1213,943 (12.9)122201812313,603 (19.1)1247,640 (18.8)1255,963 (19.5)126201912715,438 (21.7)1288,845 (21.8)1296,593 (21.5)130202013115,293 (21.5)1328,720 (21.4)1336,573 (21.5)134Floor space (m2)135106.47 ± 38.42136108.22 ± 41.90137104.14 ± 33.10138Land area (m2)139113.22 ± 52.19140107.79 ± 54.18141120.44 ± 48.49142Road width in front of houses (m)1435.38 ± 2.481445.30 ± 2.421455.48 ± 2.55146Age of houses (months)147179.9 ± 101.07148187.4 ± 101.97149170.0 ± 99.00150Main window orientation151Facing south15212,626 (17.7)1536,949 (17.1)1545,677 (18.5)155Other15658,663 (82.3)15733,712 (82.9)15824,951 (81.5)159Travel time to Tokyo Station from the nearest station in daytime (min)16044.72 ± 13.0916139.98 ± 12.1816251.00 ± 11.51163Maximum flooding depth (m)1640.016556,440 (79.2)16632,653 (80.3)16723,787 (77.7)168(0.0, 3.0)1699,287 (13.0)1704,410 (10.8)1714,877 (15.9)172[3.0, 5.0)1734,692 (6.6)1743,176 (7.8)1751,516 (4.9)176[5.0, inf]177870 (1.2)178422 (1.0)179448 (1.5)180Sediment-hazardous area181Houses inside the area1822,107 (3.0)1831,256 (3.1)184851 (2.8)185Houses outside the area18669,182 (97.0)18739,405 (96.9)18829,777 (97.2)189Land use zones190Exclusively low-rise residential zone19129,194 (41.0)19216,469 (40.5)19312,725 (41.5)194Mid/high-rise-oriented residential zone19515,112 (21.2)1968,492 (20.9)1976,620 (21.6)198Residential zone19914,341 (20.1)2008,925 (21.9)2015,416 (17.7)202Commercial zone2033,614 (5.1)2042,523 (6.2)2051,091 (35.6)206No land use zone designated2073,868 (5.4)2081,377 (3.4)2092,491 (8.1)210Sample number21171,28921240,66121330,628214a The mean and standard deviation (SD) are displayed for the continuous variables (and the number of samples for the binary and categorical variables).215https://doi.org/10.1371/journal.pone.0292657.t001216PLOS ONE217The relationship of single-family detached house prices with neighborhood walkability and disaster risk218PLOS ONE | https://doi.org/10.1371/journal.pone.0292657219October 11, 20232206 / 23221222223===== PAGE 7 (score 33, numeric-paren 0, table-words 2, var-words 19) =====224225into four groups [66] according to the risk of damage to the floors, ranging from no risk (0 m226of flooding depth) to high risk (5 m and higher of flooding depth, where the water level on the227second floor becomes hazardous). In the case of a sediment disaster, we identified whether228each house was inside a hazardous sediment area; this was defined as a binary variable. Land229use zones and travel time to Tokyo Station from the nearest station were included to account230for designated zoning restrictions on building height and building types for the area where231each house is located and the distance to the central business district, respectively.232The neighborhood walkability—net residential density, intersection density, walking time233to the nearest station, and facility density—was assessed for each single-family detached house234(Table 2). Net residential density, intersection density, and facility density are popular indica-235tors used to measure neighborhood walkability in many countries [67–71]. An area with high236net residential density is considered to be one that is densely developed. Net residential density237was measured at the level of the census district where each single-family detached house was238located. Facility density was measured as the density of each facility type within a catchment239area of 1,200 m from each house along a road network. The distance of 1,200 m corresponds240to a 15-minute walk, which is the average walking distance to neighborhood destinations [72].241The density of neighborhood facilities enabled us to assess the number of destinations that res-242idents can travel to within a walkable distance. Witten et al. [71], for example, considered eight243types of facilities (education, transport, recreation, social and cultural, food retail, financial,244health, and other retail). Similarly to the classification by Witten et al. [71] (but with more245Table 2. Neighborhood characteristicsa of single-family detached houses sold between 2015 and 2020.246Target area247Central cities248Other cities249Net residential densityb2501,918 ± 1,8872512,250 ± 2,1372521,477 ± 1,374253Intersection density (four-way)25453.5 ± 25.4625561.50 ± 26.9425642.94 ± 18.68257Walking time to the nearest station (min)25813.04 ± 5.7525912.29 ± 5.6626014.03 ± 5.72261Facility density262Bars26313.46 ± 22.1526416.64 ± 26.622659.23 ± 13.03266Cafe´s2672.73 ± 4.672683.54 ± 5.652691.65 ± 2.51270Restaurants27125.27 ± 32.4927231.71 ± 39.3927316.72 ± 16.39274Hobby classes2754.24 ± 4.552764.98 ± 4.912773.27 ± 3.81278Cultural institutions2793.58 ± 4.342804.23 ± 5.162812.71 ± 2.68282Parks2830.15 ± 0.342840.21 ± 0.382850.09 ± 0.27286Gyms and fitness clubs2872.75 ± 2.612883.19 ± 2.952892.16 ± 1.93290Convenience stores2914.31 ± 3.262925.27 ± 3.772933.04 ± 1.75294Schools2952.13 ± 1.322962.55 ± 1.452971.57 ± 0.86298Childcare facilities2994.73 ± 2.673005.64 ± 2.803013.52 ± 1.89302Retail shops3035.28 ± 10.303046.52 ± 12.013053.64 ± 7.12306Welfare facilities for older adults3078.32 ± 4.723089.99 ± 4.843096.11 ± 3.48310Hospitals3110.24 ± 0.613120.30 ± 0.663130.16 ± 0.51314Pharmacies3156.15 ± 4.453167.35 ± 4.773174.56 ± 3.38318Bus stops3197.39 ± 2.943207.35 ± 2.543217.43 ± 3.40322Stations3230.47 ± 0.643240.59 ± 0.723250.30 ± 0.45326Sample number32771,28932840,66132930,628330a The mean and standard deviation (SD) are displayed for the neighborhood characteristics. All the densities are estimated per square kilometer.331b Net residential density is estimated for each census tract where the targeted samples are located (4,126 census tracts); that is, buffered areas based on a 1,200 m network332distance away from houses are used in the case of intersection and facility density.333https://doi.org/10.1371/journal.pone.0292657.t002334PLOS ONE335The relationship of single-family detached house prices with neighborhood walkability and disaster risk336PLOS ONE | https://doi.org/10.1371/journal.pone.0292657337October 11, 20233387 / 23339340341===== PAGE 8 (score 22, numeric-paren 0, table-words 0, var-words 22) =====342343detailed facility types), we considered 16 types of facilities, including cafe´s, restaurants, gyms,344hobby classes, schools, hospitals, train stations, bus stops, and cultural institutions (museums,345libraries, community centers, etc.). These facilities provide leisure, recreation, and welfare346opportunities (e.g., schools can also be perceived as recreation facilities by making their play-347grounds available as public places for recreation activities), which are considered attractive348components of the neighborhood environment and are major destinations in daily living [56,34971]. Thus, a high density of neighborhood facilities indicates more neighborhood walking350opportunities.351Stations were also assessed using the walking time to the nearest station from each house352based on the average walking speed. Because we included both the distance and the density of353stations, it was possible to consider not only the proximity to the nearest station from resi-354dents’ homes but also their options to access a station (e.g., when there are many stations, peo-355ple who visit a destination within their neighborhood can more easily access the nearest356station from that point). In addition, the proximity to the nearest station enabled us to con-357sider each house’s proximity to a local central area (i.e., the Tokyo Metropolitan Area was358developed alongside railways, and areas near stations were considered to be local central359areas). With regard to intersection density, we selected four-way intersections (20.5% of all360intersections) because they represent good neighborhood walkability and frequent use of pub-361lic transportation better than three-way intersections [73–75]; their density was measured in362the same way as facility density. High intersection density indicates that the road network near363each house is more likely to be a grid pattern.364Analytic method365A conditional quantile regression model was used to investigate the relationships between the366neighborhood characteristics and the conditional quantiles of the sale price rather than the367mean value [76–78]. Given that the quantiles were conditional, the model could consider the368sale price after controlling for other factors, such as the floor space and age of the house, which369enabled us to test the relationships between sale price and neighborhood characteristics. As370previously mentioned, assuming a budget constraint, those who can pay more for a house with371similar features to others are more likely to be high-income earners. In other words, the differ-372ence by conditional quantiles in the relationship between the sale price and the neighborhood373characteristics indicates the difference by income level in home buyers’ revealed preference for374the neighborhood characteristics. The sale price was included in the model as a natural loga-375rithmic scale owing to the large variance. The net residential density, intersection density, and376facility density were transformed into a nonnegative form, ln (1 + density), which is a special377case of Box–Cox transformation; this enabled us to escape a negative infinity issue (i.e., zero378values) and transform the right-skewed distribution to make it proximal to a normal one [79].379The years/quarters of transactions, floor space, land area, age of the houses, road width in380front of the houses, main window orientation, land use zones, and travel time to Tokyo Station381from the nearest station were adjusted. The models were estimated in accordance with condi-382tional quantiles ranging from the 0.05 to the 0.95 quantile at 0.05-point intervals. The 0.50383quantile corresponds to the median.384We estimated two models as follows:385Model 1: Regression model considering differences by centrality386Model 2: Regression model considering differences by income level387The analysis was conducted in three steps. We initially estimated the conditional quantile388regression model (Model 1) to capture the relationships regarding the median of the sale price,389PLOS ONE390The relationship of single-family detached house prices with neighborhood walkability and disaster risk391PLOS ONE | https://doi.org/10.1371/journal.pone.0292657392October 11, 20233938 / 23394395396===== PAGE 10 (score 22, numeric-paren 0, table-words 2, var-words 15) =====397398to 0.95 at 0.05-point intervals) of the sale price and their relationships with neighborhood399characteristics. Based on a comparison of each neighborhood characteristic, we discerned the400relationship between neighborhood environment and conditional quantiles of single-family401detached house prices.402Results403Results of regression models by centrality404Table 3 summarizes the results of the conditional quantile regression models (for the median405sale price, which is the 0.50 conditional quantile; Model 1) for all samples and subareas (cen-406tral and other cities). Column 3(a) shows the results for all samples from the target area.407High net residential density was negatively related to the median sale price of single-family408detached houses (β = -0.003). Variables related to flood and sediment risk also showed nega-409tive relationships (β = -0.122, β = -0.188, and β = -0.219 for flood risk by maximum flooding410depth; β = -0.026 for sediment risk). In terms of facility density, the density of most facility411types was positively related to the median sale price, except for bars and cultural institutions412(β = -0.077 and β = -0.060, respectively). A shorter time distance to the nearest station was413negatively related to the median sale price (β = -0.011). The intersection density was found414to be insignificant.415Columns 3(b) and (c) show the results for the central and other cities. Most of the vari-416ables were related to the sale price, and the results were similar to those for the total sample.417However, there were also differences in the results for each subarea. A high intersection den-418sity was insignificant in the case of central cities but negatively related to the sale prices of419houses in the case of other cities (β = -0.099). A high density of restaurants was positively420related to the sale price in the case of central cities (β = 0.048) but was insignificant in the421case of other cities. In contrast, a high density of bus stops was positively related to the sale422price in the case of other cities (β = 0.122) but was insignificant in the case of central cities. A423high density of pharmacies was negatively related to the sale price in the case of central cities424(β = -0.016) but positively in the case of other cities (β = 0.037). Flood and sediment risk425were negatively related in the case of central cities (β = -0.118, β = -0.248, and β = -0.213 for426low to high levels of flooding risk, respectively; β = -0.078 for sediment risk), but moderate427flood risk (maximum flooding depth ranging from 3.0 to 5.0 m) and sediment risk were428insignificant in the case of other cities.429Results of regression models by price range430Table 4 summarizes the results of the conditional quantile regression models regarding repre-431sentative conditional quantiles, which are the (a) 0.15, (b) 0.50 (median), and (c) 0.85 quantiles432(Model 2). High net residential density was negatively related to sale price in the 0.85 quantile433(β = -0.009), a similar result to that of the model for the median sale price (β = -0.003). How-434ever, it was insignificant in the 0.15 quantile. High intersection density was negatively related435to sale price in the 0.85 quantile (β = -0.010); however, it was insignificant in the 0.15 and 0.50436quantiles. The facility density and time distance to the nearest station showed a similar trend437(with some differences) to the model for the median sale price, except for restaurants (negative438relationship of restaurant density with sale price in the 0.15 quantile, β = -0.003) and pharma-439cies (negative relationship of pharmacy density in the 0.85 quantile, β = -0.029). Single-family440detached houses in flood- and sediment-hazardous areas had lower sale prices than those in441nonhazardous areas.442PLOS ONE443The relationship of single-family detached house prices with neighborhood walkability and disaster risk444PLOS ONE | https://doi.org/10.1371/journal.pone.0292657445October 11, 202344610 / 23447448449===== PAGE 11 (score 111, numeric-paren 69, table-words 2, var-words 23) =====450451Results of regression models for each neighborhood characteristic by452conditional quantiles453Figs 3 and 4 present the estimates of the conditional quantile regression models for the condi-454tional quantiles of the sale price, ranging from the 0.05 to the 0.95 quantile at 0.05-point inter-455vals. Fig 3(a)–3(d) show the coefficients for flood and sediment risks. As mentioned in the456brief comparison above, being located in a hazardous area was negatively related to the sale457Table 3. Comparison of the resultsa,b of the conditional quantile regression models (50% quantile) between central and other cities.458(a) Target area459(b) Central cities460(c) Other cities461Difference in the direction of the coefficientd462(a) ! (b) ! (c)463Higher net residential densityc464-0.003 (0.001) ***465-0.010 (0.001) ***466-0.004 (0.001) **467−! −! −468Higher intersection densityc469-0.006 (0.003)4700.001 (0.004)471-0.099 (0.004) ***472 ! ! −473Shorter time distance to the nearest station474-0.013 (0.000) ***475-0.015 (0.000) ***476-0.009 (0.000) ***477−! −! −478Higher facility densityc479Bars480-0.077 (0.003) ***481-0.086 (0.004) ***482-0.049 (0.003) ***483−! −! −484Cafe´s4850.104 (0.003) ***4860.095 (0.004) ***4870.104 (0.004) ***488+ ! + ! +489Restaurants4900.014 (0.004) ***4910.048 (0.006) ***492-0.006 (0.004)493+ ! + ! 494Hobby classes4950.032 (0.003) ***4960.074 (0.004) ***4970.018 (0.004) ***498+ ! + ! +499Cultural institutions500-0.060 (0.003) ***501-0.050 (0.004) ***502-0.029 (0.004) ***503−! −! −504Parks5050.052 (0.005) ***5060.023 (0.005) ***5070.043 (0.011) ***508+ ! + ! +509Gyms and fitness clubs5100.063 (0.003) ***5110.070 (0.003) ***5120.045 (0.003) ***513+ ! + ! +514Convenience stores5150.092 (0.004) ***5160.067 (0.006) ***5170.050 (0.005) ***518+ ! + ! +519Schools5200.144 (0.004) ***5210.099 (0.005) ***5220.114 (0.005) ***523+ ! + ! +524Childcare facilities5250.063 (0.003) ***5260.052 (0.005) ***5270.059 (0.004) ***528+ ! + ! +529Retail shops5300.036 (0.003) ***5310.042 (0.003) ***5320.015 (0.003) ***533+ ! + ! +534Welfare facilities for older adults5350.020 (0.003) ***5360.012 (0.004) **5370.005 (0.003) ***538+ ! + ! +539Hospitals5400.028 (0.005) ***5410.017 (0.006) **5420.021 (0.006) ***543+ ! + ! +544Pharmacies5450.010 (0.004) **546-0.016 (0.005) ***5470.037 (0.004) ***548+ ! −! +549Bus stops5500.029 (0.003) ***551-0.004 (0.005)5520.122 (0.004) ***553+ ! ! +554Stations5550.056 (0.005) ***5560.009 (0.006)5570.000 (0.006)558+ ! ! 559Maximum flooding depth (ref: 0.0 m)560(0.0, 3.0)561-0.122 (0.003) ***562-0.118 (0.005) ***563-0.028 (0.004) ***564−! −! −565[3.0, 5.0)566-0.188 (0.005) ***567-0.248 (0.005) ***568-0.006 (0.005)569−! −! 570[5.0, inf]571-0.219 (0.007) ***572-0.213 (0.007) ***573-0.085 (0.010) ***574−! −! −575Located inside sediment-hazardous areas576-0.026 (0.005) ***577-0.078 (0.007) ***578-0.008 (0.009)579−! −! 580AICe5813507.2365821773.203583-8107.784584Pseudo R25850.5095860.5435870.463588Sample number58971,28959040,66159130,628592a The years/quarters of transactions, floor space, land area, road width in front of the house, age of the house, main window orientation, travel time to Tokyo Station,593and land use zone were adjusted.594b The values in parentheses are standard errors.595c Density was transformed into a nonnegative natural logarithmic scale.596* p < 0.05,597** p < 0.01,598*** p < 0.001.599d Signals (+, −, and ) correspond to positive, negative, and neutral signals, respectively; the direction (positive, negative, or neutral) was based on a criterion at the600significance level of 0.05.601e Akaike information criterion.602https://doi.org/10.1371/journal.pone.0292657.t003603PLOS ONE604The relationship of single-family detached house prices with neighborhood walkability and disaster risk605PLOS ONE | https://doi.org/10.1371/journal.pone.0292657606October 11, 202360711 / 23608609610===== PAGE 12 (score 117, numeric-paren 69, table-words 5, var-words 23) =====611612price of single-family detached houses in all the quantiles. The higher the quantile, the greater613the negative relationship with the sale price for houses in hazardous areas. Interestingly, when614the maximum flooding depth was between 3 and 5 m, the coefficient increased slightly in the615highest quantile. The coefficient of net residential density was positive in the lower quantiles;616however, it became negative in the 0.25 quantile and decreased further in the higher quantiles617(Fig 3[e]). In the case of intersection density, its coefficient was negative in the 0.70 quantile618and higher but insignificant in the other quantiles (Fig 3[f]).619Table 4. Comparison of the resultsa,b of the conditional quantile regression models by price range (N = 71,289).620(a) 15%621(b) 50%622(c) 85%623Difference in the direction of the coefficientd624(a) ! (b) ! (c)625Higher net residential densityc6260.001 (0.001)627-0.003 (0.001) ***628-0.009 (0.001) ***629 ! −! −630Higher intersection densityc631-0.006 (0.003)632-0.006 (0.003)633-0.010 (0.004) **634 ! ! −635Shorter time distance to the nearest station636-0.010 (0.000) ***637-0.013 (0.000) ***638-0.014 (0.000) ***639−! −! −640Higher facility densityc641Bars642-0.068 (0.002) ***643-0.077 (0.003) ***644-0.087 (0.003) ***645−! −! −646Cafe´s6470.082 (0.003) ***6480.104 (0.003) ***6490.148 (0.004) ***650+ ! + ! +651Restaurants652-0.001 (0.003)6530.014 (0.004) ***6540.019 (0.005) ***655 ! + ! +656Hobby classes6570.028 (0.003) ***6580.032 (0.003) ***6590.045 (0.003) ***660+ ! + ! +661Cultural institutions662-0.040 (0.003) ***663-0.060 (0.003) ***664-0.081 (0.004) ***665−! −! −666Parks6670.057 (0.004) ***6680.052 (0.005) ***6690.047 (0.006) ***670+ ! + ! +671Gyms and fitness clubs6720.043 (0.003) ***6730.063 (0.003) ***6740.082 (0.003) ***675+ ! + ! +676Convenience stores6770.108 (0.004) ***6780.092 (0.004) ***6790.094 (0.005) ***680+ ! + ! +681Schools6820.167 (0.004) ***6830.144 (0.004) ***6840.142 (0.004) ***685+ ! + ! +686Childcare facilities6870.048 (0.003) ***6880.063 (0.003) ***6890.059 (0.004) ***690+ ! + ! +691Retail shops6920.033 (0.003) ***6930.036 (0.003) ***6940.035 (0.003) ***695+ ! + ! +696Welfare facilities for older adults6970.028 (0.003) ***6980.020 (0.003) ***6990.015 (0.003) ***700+ ! + ! +701Hospitals7020.018 (0.004) ***7030.028 (0.005) ***7040.073 (0.006) ***705+ ! + ! +706Pharmacies7070.029 (0.004) ***7080.010 (0.004) **709-0.029 (0.005) ***710+ ! + ! −711Bus stops7120.061 (0.004) ***7130.029 (0.003) ***7140.030 (0.004) ***715+ ! + ! +716Stations7170.068 (0.005) ***7180.056 (0.005) ***7190.062 (0.006) ***720+ ! + ! +721Maximum flooding depth (ref: 0.0 m)722(0.0, 3.0)723-0.078 (0.004) ***724-0.122 (0.003) ***725-0.134 (0.004) ***726−! −! −727[3.0, 5.0)728-0.114 (0.003) ***729-0.188 (0.005) ***730-0.189 (0.006) ***731−! −! −732[5.0, inf]733-0.113 (0.005) ***734-0.219 (0.007) ***735-0.278 (0.016) ***736−! −! −737Located inside sediment-hazardous areas738-0.004 (0.004)739-0.026 (0.005) ***740-0.041 (0.007) ***741 ! −! −742AICe74322298.377443507.23674525155.46746Pseudo R27470.4377480.5097490.531750a The years/quarters of transactions, floor space, land area, road width in front of the house, age of the house, main window orientation, travel time to Tokyo Station,751and land use zone were adjusted.752b The values in brackets are standard errors.753c Density was transformed into a nonnegative natural logarithmic scale.754* p < 0.05,755** p < 0.01,756*** p < 0.001.757d Signals (+, −, and ) correspond to positive, negative, and neutral signals, respectively; the direction (positive, negative, or neutral) was based on a criterion at the758significance level of 0.05.759e Akaike information criterion.760https://doi.org/10.1371/journal.pone.0292657.t004761PLOS ONE762The relationship of single-family detached house prices with neighborhood walkability and disaster risk763PLOS ONE | https://doi.org/10.1371/journal.pone.0292657764October 11, 202376512 / 23766767768===== PAGE 13 (score 16, numeric-paren 0, table-words 0, var-words 16) =====769770Fig 4 shows the coefficients of facility density by facility type. Most facility types were posi-771tively related to the sale price, but there were also certain further trends or greater, lesser, or772similar positive relationships in higher quantiles. It is also true that some facility types indi-773cated a negative relationship in either some or all quantiles. The density of bars and cultural774institutions showed a negative relationship with the sale price in all quantiles and a greater775negative relationship in higher quantiles. In the case of restaurants, their density was negatively776related to the sale price in lower quantiles but became positive in the 0.25 quantile and higher.777Fig 3. Estimates of the coefficients of flood/sediment risk, net residential density, intersection density, and percentage of778older adults for sale price quantiles ranging from 0.05 to 0.95 at 0.05 intervals. The estimates are from the conditional779quantile regression model. The years/quarters of transactions, floor space, land area, road width in front of the house, age of the780house, main window orientation, travel time to Tokyo Station, and land use zone were adjusted. Conditional quantiles are781displayed on the x-axis, and coefficients are presented on the y-axis. The black dots show the estimated coefficients. The gray area782represents the 95% confidence intervals.783https://doi.org/10.1371/journal.pone.0292657.g003784PLOS ONE785The relationship of single-family detached house prices with neighborhood walkability and disaster risk786PLOS ONE | https://doi.org/10.1371/journal.pone.0292657787October 11, 202378813 / 23789