1# RECORD 5242a8699dbfd95d3c698daaa18f7bab00d59c972# TITLE: THE IMPACT OF LUXURY HOUSING ON NEIGHBORHOOD HOUSING PRICES: AN APPLICATION OF THE SPATIAL DIFFERENCE-IN-DIFFERENCES METHOD3# YEAR: 2020.0 VENUE: nan DOI: 10.3846/ijspm.2020.136494# AUTHORS: Chih-Min Liang; Chun-Chang Lee; You-Hsin Lin; Zheng Yu; Wen-Chih Yeh5# PAGES: 18 CANDIDATE TABLE PAGES (1-based): [9, 10, 11, 12]6# ABSTRACT: This study investigated the spatial spillover effects of luxury housing during and after construction, in regards to increases in housing prices in neighboring areas as well as the spatial dependence of neighboring housing. This study focused on already completed luxury housing in Taipei, Taiwan. First, the nearest-neighbor matching approach of propensity score matching was used to overcome the problem of data heterogeneity. The difference-in-differences (DD) method and spatial econometrics were used for analysis. The empirical results indicated that the spatial error model had the best goodness of fit. This indicated that housing prices increased by 13.0% during construction of luxury housing nearby. This indicated that housing prices increased by 5.8% after the construction of luxury housing nearby. The empirical results showed that the ongoing and completed construction of luxury housing had spillover effects on housing prices. The effect of ongoing construction of luxury housing was particularly large in scope, indicating its role as a predictor of psychological reaction in the market.789===== PAGE 1 (score 12, numeric-paren 1, table-words 0, var-words 11) =====1011This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unre-12stricted use, distribution, and reproduction in any medium, provided the original author and source are credited.13Copyright © 2020 The Author(s). Published by Vilnius Gediminas Technical University14THE IMPACT OF LUXURY HOUSING ON NEIGHBORHOOD15HOUSING PRICES: AN APPLICATION OF THE SPATIAL16DIFFERENCE-IN-DIFFERENCES METHOD17Chih-Min LIANG1, Chun-Chang LEE2,*, You-Hsin LIN2, Zheng YU2, Wen-Chih YEH2,**181 Department of Public Finance and Tax Administration, National Taipei University of Business, Taipei, Taiwan192 Department of Real Estate Management, National Pingtung University, Pingtung, Taiwan20Received 09 September 2019; accepted 07 June 202021Abstract. This study investigated the spatial spillover effects of luxury housing during and after construction, in regards22to increases in housing prices in neighboring areas as well as the spatial dependence of neighboring housing. This study23focused on already completed luxury housing in Taipei, Taiwan. First, the nearest-neighbor matching approach of propen-24sity score matching was used to overcome the problem of data heterogeneity. The difference-in-differences (DD) method25and spatial econometrics were used for analysis. The empirical results indicated that the spatial error model had the best26goodness of fit. This indicated that housing prices increased by 13.0% during construction of luxury housing nearby. This27indicated that housing prices increased by 5.8% after the construction of luxury housing nearby. The empirical results28showed that the ongoing and completed construction of luxury housing had spillover effects on housing prices. The effect29of ongoing construction of luxury housing was particularly large in scope, indicating its role as a predictor of psychological30reaction in the market.31Keywords: luxury housing, propensity score matching, difference-in-differences, spatial dependence, spatial lag model,32spatial error model.33Introduction34With the improvement of living standards in Taiwan, 35housing now constitutes the highest expenditure with re-36spect to the basic necessities of life. In addition to consid-37ering affordability, transportation convenience, and loca-38tion when choosing housing, people have recently started 39pursuing high-standard, high-quality, high-profile, and 40high-priced housing.41In recent years, citizens of Taiwan have paid more 42attention to quality of life. This has led to the develop-43ment of a new type of housing, referred to in the market 44as luxury housing, derived from existing types of hous-45ing such as condominiums, studios, multi-story houses, 46and high-rise suites. Lin and Jou (2005) described typical 47luxury housing as housing with rare spatial resources such 48as beautiful landscapes and good locations, and character-49ized by a high total price, a large living space, good facili-50ties, and high prestige. According to the 15th provision of 51the “Property Price and Present Value Appraisal Criteria” 52as established by the Taipei City Revenue Service, build-53ings that are made from reinforced concrete or belong to 54a higher construction class are used for residential pur-55poses. Properties with a total value of over NT$80 mil-56lion, and a total registered area of ownership over 80 ping 57(1 ping is equal to 3.306 square meter) or a per apartment 58price of over NT$1 million per ping (excluding parking 59spaces), are defined as high-level housing or luxury hous-60ing. According to Wang (2008), luxury housing is a small 61market segment targeting upper class consumers; Wang 62suggested that such housing is showy by nature, and its 63target consumers are relatively homogeneous, which pro-64vides a cluster effect. Lin (2004) maintained that luxury 65housing has shaped a particular lifestyle for upper class 66society, and that it is not only local landmark, but also an 67indicator of wealth and ostentation.68Past studies on luxury housing have discussed histor-69ical development, purchase decision-making, integrated 70marketing, and positioning design of luxury housing 71(e.g. Lin & Jou, 2005; Ho, 2011; Wu, 2008; Chen, 2011). 72However, there is a lack of research regarding the effect 73of luxury housing on neighborhood housing prices.74International Journal of Strategic Property Management75ISSN: 1648-715X / eISSN: 1648-9179762020 Volume 24 Issue 6: 456–47377https://doi.org/10.3846/ijspm.2020.1364978*Corresponding author. E-mail: lcc@mail.nptu.edu.tw79**Department of Real Estate Management, HungKuo Delin University of Technology, New Taipei, Taiwan808182===== PAGE 9 (score 54, numeric-paren 20, table-words 9, var-words 5) =====838446485C.-M. Liang et al. The impact of luxury housing on neighborhood housing prices: an application...86groups of samples, and check if self-selection bias was 87effectively reduced. As shown in Table 2, before match-88ing, there were 352 samples in the experimental group 89and 6,940 samples in the control group, with significant 90differences being observed in terms of the number of 91rooms, number of living rooms, housing type, residence 92floor, distance to the nearest MRT station, distance to the 93nearest senior high school, and distance to the nearest 94junior high school. After matching, there were 352 sam-95ples in the experimental group and 4,772 samples in the 96control group, and in this case, significant differences in 97the variables were not observed, with the exception of 98the number of rooms and housing type. As shown in Ta-99ble 3, before matching, there were 542 samples in the ex-100perimental group and 9,953 samples in the control group, 101with significant differences being observed in terms of the 102number of rooms, number of living rooms, housing type, 103residence floor, distance to the nearest MRT station, dis-104tance to the nearest senior high school, and distance to the 105nearest junior high school. After matching, there were 541 106samples in the experimental group and 7,106 samples in 107the control group, and in this case, significant differences 108in the variables were not observed, with the exception of 109housing type.110Table 2. Differences in mean values between the experimental group and the control group during construction111Variable112Before matching113After matching114Treated group115(N = 352)116Control group117(N= 6,940)118t-value119Treated group120(N = 352)121Control group122(N= 4,772)123t-value124AREA12530.50112631.614127−1.28012830.50112930.2651300.180131AGE13220.90313320.0911341.28013520.90313621.322137−0.460138ROOM1392.2931402.492141−3.940***1422.2931432.452144−2.060**145LIVROOM1461.5911471.773148−7.010***1491.5911501.625151−0.880152BATH1531.4461541.451155−0.1501561.4461571.477158−0.680159MRT160290.630161554.350162−11.160***163290.630164275.7401651.320166HISC167402.650168676.610169−12.160***170402.650171389.0601720.910173JUNI174475.090175529.020176−3.400***177475.090178472.0801790.180180PRIM181390.430182377.6901831.160184390.430185387.7701860.220187Type188Frequency189Frequency190χ2191Frequency192Frequency193χ2194MULTISTORIED 195BUILDINGS1962741974,67419816.91***1992742003,4312015.78**202APARTMENT HOUSES203782042,266205782061,341207FLOOR208Frequency209Frequency210χ2211Frequency212Frequency213χ2214FLOOR1215102163922175.07**218102192002201.52221NONFLOOR12223422236,5482243422254,572226Note: ***, **, and * indicate that the coefficient reaches the 1%, 5%, and 10% level of significance, respectively; significantly different from 0.227Table 3. Differences in mean values between the experimental group and the control group after construction228Variable229Before matching230After matching231Treated group232(N = 542)233Control group234(N = 9,953)235t-value236Treated group237(N = 541)238Control group239(N = 7,106)240t-value241AREA24234.82424333.6922441.730*24534.81024635.007247−0.220248AGE24926.32925024.1572514.300***25226.28625326.618254−0.520255ROOM2562.6022572.668258−1.690*2592.6042602.649261−0.740262LIVROOM2631.8402641.7992651.990**2661.8392671.847268−0.240269BATH2701.6302711.6112720.7402731.6292741.647275−0.450276MRT277302.180278628.310279−15.890***280302.340281302.0402820.030283HISC284507.980285725.530286−11.090***287507.920288514.520289−0.340290JUNI291415.650292551.230293−9.770***294416.190295408.9002960.580297PRIM298437.370299390.4703004.890***301437.040302427.3403030.900304Type305Frequency306Frequency307χ2308Frequency309Frequency310χ2311MULTISTORIED 312BUILDINGS3134563146,067315117.37***3164553174,88031856.73***319APARTMENT HOUSES320863213,886322863232,226324FLOOR325Frequency326Frequency327χ2328Frequency329Frequency330χ2331FLOOR1332323337893342.92*335323364783370.53338NONFLOOR13395103409,1643415093426,628343Note: ***, **, and * indicate that the coefficient reaches the 1%, 5%, and 10% level of significance, respectively; significantly different from 0.344345346===== PAGE 10 (score 25, numeric-paren 0, table-words 4, var-words 7) =====347348International Journal of Strategic Property Management, 24(6): 456–4733494653503.4. Descriptive statistics of the data351The acquisition of a construction license was set as the event 352(Table 4). The average total price of housing (PRICE) was 353NT$1,439.89, with the standard deviation of NT$1,113.18. 354The average housing area (AREA) was 30.93 ping, with the 355standard deviation of 16.42 ping. The average house age 356(AGE) was 20.13 years, with the standard deviation of 11.82 357years. The mean value of the residence floor (FLOOR) was 3580.041, indicating that4.1% of house residents lived on the 1st 359floor and 95.9% lived on other floors. The mean value of the 360housing type (TYPE) was 0.723, indicating that multisto-361ried buildings and apartment houses accounted for 72.3% 362and 27.7% of the sample, respectively. The average number 363of rooms (ROOM) was 2.40, with the standard deviation 364of 0.95. The average number of living rooms (LIVROOM) 365was 1.70, with the standard deviation of 0.49. The average 366number of bathrooms (BATH) was 1.44, with the standard 367deviation of 0.58. The average distance to the nearest MRT 368station (MRT) was 357 meter, with the standard deviation 369of 200.04 meter. The average distance to nearest senior high 370school (HISC) was 495.42 meters, with the standard devia-371tion of 243.69 meters. The average distance to nearest junior 372high school (JUNI) was 486.83 meters, with the standard 373deviation of 232.19 meters. The average distance to nearest 374primary school (PRIM) was 373.09 meters, with the stand-375ard deviation of 173.44 meters.376The acquisition of a usage license was set as the event 377(Table 5). The average total price of housing (PRICE) was 378NT$1,892.83, with the standard deviation of NT$1,153.54. 379The average housing area (AREA) was 33.73 ping, with 380Table 4. Descriptive statistics: During construction – Construction license acquisition381Variable382Mean383Standard deviation384Min.385Max.386PRICE (ten thousand NT$)3871,439.893881,113.183894439023,100391AREA (ping)39230.9339316.423942.14395174.43396AGE (year)39720.1339811.823990.1040053.80401ROOM4022.404030.9540414055406LIVROOM4071.704080.4940914105411BATH4121.444130.5841414155416TMRT417357.00418200.044193.404201397.44421HISC422495.42423243.6942413.114251561.37426JUNI427486.83428232.1942917.564301613.65431PRIM432373.09433173.4443416.424351148.42436Frequency437Percent438Cumulative percent439TYPE440MULTISTORIED BUILDINGS4413,70544272.3144372.31444APARTMENT HOUSES4451,41944627.69447100.00448FLOOR449FLOOR14502104514.104524.10453NONFLOOR14544,91445595.90456100.00457Table 5. Descriptive statistics: After construction – Usage license acquisition458Variable459Mean460Standard deviation461Min.462Max.463PRICE (ten thousand NT$)4641,892.834651,153.5446621746733300468AREA (ping)46933.7347015.064717.85472312473AGE (year)47424.6147511.514760.147755478ROOM4792.624800.9048114829483LIVROOM4841.804850.4748614878488BATH4891.604900.5949114928493MRT494415.77495229.404962.464971572.23498HISC499607.72500395.4450114.365023617.10503JUNI504469.33505239.2850615.795071561.37508PRIM509388.40510184.4551119.1585121334.04513Frequency514Percent515Cumulative percent516TYPE517MULTISTORIED BUILDINGS5185,33551969.7752069.77521APARTMENT HOUSES5222,31252330.23524100.00525FLOOR526FLOOR15275105286.675296.67530NONFLOOR15317,13753293.33533100.00534535536===== PAGE 11 (score 51, numeric-paren 10, table-words 13, var-words 10) =====537538466539C.-M. Liang et al. The impact of luxury housing on neighborhood housing prices: an application...540the standard deviation of 15.06 ping. The average house 541age (AREA) was 24.61 years, with the standard devia-542tion of 11.51 years. The mean value of the residence floor 543(FLOOR) was 0.068, indicating that 6.8% of house resi-544dents lived on the 1st floor and 93.2% lived on other floors. 545The mean value of the housing type (TYPE) was 0.698, in-546dicating that multistoried buildings and apartment houses 547accounted for 69.8% and 30.2% of the sample, respectively. 548The average number of rooms (ROOM) was 2.62, with the 549standard deviation of 0.90. The average number of living 550rooms (LIVROOM) was 1.80, with the standard devia-551tion of 0.47. The average number of bathrooms (BATH) 552was 1.60, with the standard deviation of 0.59. The average 553distance to the nearest MRT station (MRT) was 415.77 554meter, with the standard deviation of 229.40 meter. The 555average distance to nearest senior high school (HISC) was 556607.72 meters, with the standard deviation of 395.44 me-557ters. The average distance to nearest junior high school 558(JUNI) was 469.33 meters, with the standard deviation of 559239.28 meters. The average distance to nearest primary 560school (PRIM) was 388.40 meters, with the standard de-561viation of 184.45 meters.5624. Empirical results563This study investigated the effect of luxury housing on 564housing prices in neighboring areas during (indicated 565by a construction license) and after (indicated by a us-566age license) its construction. The spatial concepts were 567integrated into the regression model in order to solve the 568issues of spatial heterogeneity and spatial autocorrelation 569in the traditional OLS model. As shown in Table 6, Mo-570ran’s I reached the 1% level of significance both during 571and after construction, indicating the presence of positive 572autocorrelation and spatial clustering in housing prices in 573areas neighboring with luxury housing. Housing prices in 574administrative districts showed positive autocorrelation. 575The spatial dependence test was further used to determine 576which spatial model had the best explanatory power. Both 577LM-lag and LM-error reached the 1% level of significance. 578Both Robust LM-lag and Robust LM-error reached the 1% 579level of significance. Thus, the residuals differed in terms 580of heterogeneity. The results of the spatial lag model and 581spatial error model were further compared.582(1) During construction583Table 7 shows the results of evaluating the period after 584construction using the three models (OLS, SLM, SEM). 585The Breusch-Pagan test was used to determine heterosce-586dasticity of residuals. Its values in the three models were 5871483.72, 1436.74, and 1523.74, reaching the 1% level of 588significance. The value in the spatial error model was 589the highest, indicating heteroscedasticity of its residuals. 590R2 was equal to 0.797, 0.807, and 0.813; the spatial er-591ror model showed the highest explanatory power. Smaller 592AIC and SC values indicate better goodness of fit. AIC was 593equal to 1948.61, 1708.89, and 1642.35 and SC was equal 594to 2125.23, 1892.06, and 1818.97. The larger log likelihood 595values indicate better goodness of fit. The log likelihood 596values of the models were −947.30, −826.45, and −794.17, 597indicating that the spatial lag model and spatial error 598model had better goodness of fit. The results described 599above showed that the spatial error model was better than 600the traditional OLS regression model and spatial lag mod-601el. The evaluation results show that the spatial error model 602had the best goodness of fit. The results of the spatial error 603model are discussed further.604With regard to property attributes, the building trans-605fer area coefficient was 0.022, reaching the 1% level of sig-606nificance, which indicated that the housing price increased 607by 2.2% with each additional ping of the building transfer 608area. The house age coefficient was −0.013, reaching the 6091% level of significance. The coefficient of the square of 610house age was 0.001, reaching the 1% level of significance, 611which indicated prices of older housing were lower and 612such decrease in prices was decreased. The residence floor 613coefficient was 0.129, reaching the 1% level of significance, 614which indicated that the prices of housing located on the 615first floor were 12.9% higher than that on other floors. The 616coefficient of the number of rooms was 0.086, reaching the 6171% level of significance, which indicated that the hous-618ing price increased by 8.6% with each additional room. 619The coefficient of the number of living rooms was 0.173, 620reaching the 1% level of significance, which indicated that 621the housing price increased by 17.3% with each additional 622living room. The coefficient of the number of bathrooms 623was 0.083, reaching the 1% level of significance, which 624indicated that the housing price increased by 8.3% with 625each additional bathroom. The housing type coefficient 626Table 6. Spatial dependence during and after construction of luxury housing627Test628During construction629After construction630MI/DF631Z(I) value632MI/DF633Z(I) value634Moran’s I6350.1381***63621.6266370.2078***63834.996639LM-lag6401641289.411***6421643848.676***644LM-error645164657.076***647164895.030***649Robust LM-lag6501651435.721***65216531183.005***654Robust LM-error6551656203.386***6571658429.358***659Note: MI/DF indicate Moran’s I and degrees of freedom. ***, **, and * indicate that the coefficient reaches the 1%, 5%, and 10% level of significance, 660respectively; significantly different from 0.661662663===== PAGE 12 (score 231, numeric-paren 186, table-words 10, var-words 2) =====664665International Journal of Strategic Property Management, 24(6): 456–473666467667Table 7. Empirical results for the period during construction and after construction of luxury housing668During construction (N = 5124)669After construction (N = 7647)670OLS671Spatial lag 672model673Spatial error 674model675OLS676Spatial lag 677model678Spatial error 679model680Independent variable681coefficient682(S.E.)683coefficient684(S.E.)685coefficient686(S.E.)687coefficient688(S.E.)689coefficient690(S.E.)691coefficient692(S.E.)693Intercept6945.260***695(0.035)6963.914***697(0.091)6985.267***699(0.047)7006.010***701(0.029)7023.958***703(0.082)7046.055***705(0.042)706AREA7070.023***708(0.001)7090.022***710(0.001)7110.022***712(0.001)7130.018***714(0.001)7150.017***716(0.001)7170.017***718(0.001)719AGE720−0.013***721(0.001)722−0.013***723(0.001)724−0.013***725(0.001)726−0.004***727(0.001)728−0.004***729(0.001)730−0.004***731(0.001)732AGES7330.001***734(0.001)7350.001***736(0.001)7370.001***738(0.001)7390.001***740(0.001)7410.001***742(0.001)7430.001***744(0.001)745FLOOR7460.124***747(0.021)7480.126***749(0.020)7500.129***751(0.020)7520.221***753(0.013)7540.219***755(0.012)7560.221***757(0.012)758ROOM7590.085***760(0.007)7610.085***762(0.007)7630.086***764(0.007)7650.082***766(0.005)7670.078***768(0.005)7690.081***770(0.005)771LIVROOM7720.167***773(0.012)7740.166***775(0.011)7760.173***777(0.011)7780.097***779(0.008)7800.088***781(0.008)7820.092***783(0.008)784BATH7850.087***786(0.010)7870.085***788(0.010)7890.083***790(0.010)7910.018**792(0.007)7930.017***794(0.007)7950.019***796(0.007)797TYPE7980.068***799(0.012)8000.073***801(0.012)8020.069***803(0.012)8040.272***805(0.008)8060.255***807(0.008)8080.289***809(0.008)810Location fixed effects811yes812yes813yes814yes815yes816yes817TIME8180.427***819(0.009)8200.427***821(0.009)8220.430***823(0.009)8240.218***825(0.008)8260.211***827(0.007)8280.228***829(0.007)830TREAT8310.021832(0.024)8330.012834(0.023)8350.003836(0.029)8370.050**838(0.021)8390.014840(0.020)8410.027842(0.026)843DD8440.146***845(0.033)8460.132***847(0.032)8480.130***849(0.032)8500.050*851(0.026)8520.053**853(0.024)8540.058**855(0.025)856MRT857−0.001***858(0.001)859−0.001***860(0.001)861−0.001***862(0.001)863−0.001***864(0.001)865−0.001***866(0.001)867−0.001***868(0.001)869HISC8700.001871(0.001)8720.001873(0.001)8740.001875(0.001)8760.001877(0.001)878−0.001879(0.001)880−0.001881(0.001)882JUNI8830.001884(0.001)885−0.001886(0.001)887−0.001888(0.001)8890.001890(0.001)891−0.001892(0.001)8930.001894(0.001)895PRIM896−0.001897(0.001)898v0.001899(0.001)900−0.001901(0.001)902−0.001***903(0.001)904−0.001***905(0.001)906−0.001***907(0.001)908Spatial Lag(ρ)909−9100.205***911(0.0130)912−913−9140.298***915(0.011)916−917Lambda(λ)918−919−9200.406***921(0.022)922−923−9240.472***925(0.016)926Breusch-Pagan9271483.72***9281436.74***9291523.74***9308491.93***9317766.88***9328929.17***933Log Likelihood934−947.30935−826.45936−794.17937−1140.42938−803.25939−729.85940SC9412125.239421892.069431818.979442522.289451856.879461701.13947AIC9481948.619491708.899501642.359512334.859521662.499531513.69954R29550.7979560.8079570.8139580.7199590.7469600.757961Note: The estimated coefficient value are non-standardized. ***, **, and * indicate that the coefficient reaches the 1%, 5%, and 10% level of significance, 962respectively; significantly different from 0.963964965===== PAGE 13 (score 29, numeric-paren 0, table-words 7, var-words 11) =====966967468968C.-M. Liang et al. The impact of luxury housing on neighborhood housing prices: an application...969evaluation of data related to the period during construc-970tion. Thus, these results are not further described here. 971With regard to neighborhood characteristics, the coef-972ficient of the distance to the nearest MRT station was 973−0.001, indicating a 1% level of significance, and the coef-974ficient of the distance to the nearest primary school was 975–0.001, indicating a 1% level of significance.976With regard to DD properties, the estimated coef-977ficients of the construction license acquisition time for 978the three models were 0.218, 0.211, and 0.228, reaching 979the 1% level of significance, which indicated that housing 980prices were 21.8%, 21.1%, and 22.8% higher after acqui-981sition of a construction license by luxury housing than 982before. The estimated coefficients of location in the neigh-983boring area were 0.050, 0.014, and 0.027; the coefficient in 984the OLS model reached the 5% level of significance and 985the coefficients in the spatial lag model and spatial error 986model have not reached the 10% level of significance. 987The estimated coefficients of the interaction variable of 988construction license acquisition time and location in the 989neighboring area were 0.050, 0.053, and 0.058, reaching 990the level of significance. The DD estimated coefficients 991were 5.0%, 5.3%, and 5.8%.992In the spatial error model, the spatial error coefficient 993(λ) was 0.472, reaching the 1% level of significance, which 994indicated that interference factors in error terms caused 995spatial autocorrelation. In the spatial lag model, the spatial 996lag coefficient (ρ) was 0.298, reaching the 1% level of sig-997nificance, which indicated that there was spatial depend-998ence between neighboring areas and housing prices were 999affected by neighboring housing prices.10005. Discussion1001The empirical results regarding basic property attributes 1002indicated that the building transfer area had a significant 1003positive effect on housing prices, which corresponded to 1004the findings of Frew and Judd (2003), Martins-Filho and 1005Bin (2005), and Abbott and Klaiber (2011). This indi-1006cated that prices of housing with larger area are higher. 1007House age was found to have a significant and negative 1008effect on housing prices, which corresponded to the re-1009sults reported by Glasock et al. (2000), Osland (2010), and 1010Bin and Landry (2013). With time, housing material and 1011facilities become older, which reduces the housing price. 1012The square of house age was found to significantly and 1013positively affect housing prices, which corresponded to the 1014results reported by Malpezzi et al. (1987), Geltner et al. 1015(2010), and Bin and Landry (2013). The house age variable 1016can have non-linear changes and, thus, the square of house 1017age was added that was found to have a significant positive 1018effect on housing prices. The number of rooms was found 1019to significantly and positively affect housing prices, which 1020corresponded to the results reported by McMillen and 1021Redfearn (2010) and Paredes and Iturra (2013). A larger 1022number of rooms was found to result in higher housing 1023prices. The numbers of living rooms and bathrooms were 1024found to significantly and positively affect housing prices, 1025was 0.069, reaching the 1% level of significance, which 1026indicated that the housing prices in multistoried buildings 1027and mansions were 6.9% higher than those in apartment 1028houses. With regard to neighborhood characteristics, the 1029coefficient of the distance to the nearest MRT station was 1030−0.001, indicating a 1% level of significance. The three 1031coefficients of the distance to the nearest school did not 1032attain the level of significance.1033With regard to DD properties, the estimated coef-1034ficients of the construction license acquisition time for 1035the three models were 0.427, 0.427, and 0.430, reaching 1036the 1% level of significance, which indicated that hous-1037ing prices were 42.7%, 42.7%, and 43.0% higher after the 1038acquisition of a construction license by luxury housing 1039than before. The estimated coefficients of location in the 1040neighboring area were 0.021, 0.012, and 0.003, and did 1041not reach the 1% level of significance, indicating that there 1042was no significant difference between housing prices in 1043areas neighboring and not neighboring with luxury hous-1044ing. The estimated coefficients of the interaction variable 1045of construction license acquisition time and location in 1046the neighboring area were 0.146, 0.132, and 0.130, reach-1047ing the 1% level of significance. The DD estimated coef-1048ficients were 14.6%, 13.2%, and 13.0%. This indicated that 1049the average difference in housing prices in areas neigh-1050boring and not neighboring luxury housing was 14.6%, 105113.2%, and 13.0% during construction of luxury housing.1052In the spatial error model, the spatial error coefficient 1053(λ) was 0.406, reaching the 1% level of significance, which 1054indicated that interference factors in error terms caused 1055spatial autocorrelation. In the spatial lag model, the spa-1056tial lag coefficient (ρ) was 0.205, reaching the 1% level 1057of significance, which indicated that housing prices were 1058affected by neighboring housing prices.1059(2) After construction1060The results of three models regarding the period after the 1061completion of luxury housing construction are discussed 1062further. The Breusch-Pagan test was used to determine 1063heteroscedasticity of residuals. Its values in three models 1064were 8491.93, 7766.88, and 8929.17, reaching the 1% level 1065of significance. The value in the spatial error model was 1066the highest, indicating heteroscedasticity of its residuals. 1067R2 was equal to 0.719, 0.746, and 0.757; the spatial error 1068model showed the highest explanatory power. AIC was 1069equal to 2334.85, 1662.49, and 1513.69 and SC was equal 1070to 2522.28, 1856.87, and 1701.13. The log likelihood val-1071ues of the models were −1140.42, −803.25, and −729.85, 1072indicating that the bias could not be solved in the OLS 1073regression model but could be solved in the spatial lag 1074model and spatial error model. The results showed that 1075the spatial error model was better than the traditional 1076OLS regression model and spatial lag model. The evalu-1077ation results show that the spatial error model had the 1078best goodness of fit. The results of the spatial error model 1079are discussed further. With regard to property attributes, 1080the results showed that the estimated coefficients and their 1081significance levels were similar to those observed in the 1082