1# RECORD e1b44941a0d59a0ba33dc85660a7305568feae7f2# TITLE: Assessing the economic premium of additional elevator: PSM hedonic analysis in Beijing, China3# YEAR: 2024.0 VENUE: International Journal of Strategic Property Management DOI: 10.3846/ijspm.2024.226064# AUTHORS: Zhao-Wen Zhang; Yihua Mao; Yueyao Shui; Ru-Yun Deng5# PAGES: 12 CANDIDATE TABLE PAGES (1-based): [4, 6, 7, 9]6# ABSTRACT: Although many Chinese cities have made remarkable progress in promoting age-friendly housings, few empirical studies related to the premium price of housings with additional elevators installed have been reported. In this study, we constructed propensity score matching–hedonic price models as well as quantile–hedonic models to assess the impact of adding elevators on housing prices. This study concludes: (1) The impact of additional elevators on housing prices varies depending on the floor level. (2) Single-facing and older housings are more likely to be negatively affected by the addition of elevators, and this interaction increases with floor level. (3) Adding an elevator reduces the price of low-priced housings on low floors and raises the price of high-priced housings on low floors, while for middle-floor and high-floor housings, the premium increases with the price of the housing. (4) Compared with the commercial community housings, the elevator premium of the affordable community housings is higher. Our findings provide an important reference for local governments to formulate appropriate charging and compensation mechanisms for adding elevators to multistory housings to promote retrofitting of aging communities.789===== PAGE 1 (score 12, numeric-paren 0, table-words 0, var-words 12) =====1011This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), 12which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.13Copyright © 2024 The Author(s). Published by Vilnius Gediminas Technical University14ISSN: 1648-715X / eISSN: 1648-917915INTERNATIONAL JOURNAL 16OF STRATEGIC PROPERTY 17MANAGEMENT18INTERNATIONAL JOURNAL 19OF STRATEGIC PROPERTY 20MANAGEMENT21202422Volume 2823Issue 624Pages 381–39225https://doi.org/10.3846/ijspm.2024.2260626ASSESSING THE ECONOMIC PREMIUM OF ADDITIONAL ELEVATOR: 27PSM HEDONIC ANALYSIS IN BEIJING, CHINA28Zhao ZHANG, Yihua MAO*, #, Yueyao SHUI, Ruyu DENG29College of Civil Engineering and Architecture, Zhejiang University, Hangzhou, China30Article History: 31Abstract. Although many Chinese cities have made remarkable progress in promoting age-friendly housings, 32few empirical studies related to the premium price of housings with additional elevators installed have been 33reported. In this study, we constructed propensity score matching–hedonic price models as well as quantile–34hedonic models to assess the impact of adding elevators on housing prices. This study concludes: (1) The im-35pact of additional elevators on housing prices varies depending on the floor level. (2) Single-facing and older 36housings are more likely to be negatively affected by the addition of elevators, and this interaction increases 37with floor level. (3) Adding an elevator reduces the price of low-priced housings on low floors and raises the 38price of high-priced housings on low floors, while for middle-floor and high-floor housings, the premium in-39creases with the price of the housing. (4) Compared with the commercial community housings, the elevator 40premium of the affordable community housings is higher. Our findings provide an important reference for 41local governments to formulate appropriate charging and compensation mechanisms for adding elevators to 42multistory housings to promote retrofitting of aging communities.43 ■received 11 March 202444 ■accepted 3 November 202445Keywords: age-friendly retrofit, additional elevator, housing price, floor level, affordable community.46 * Corresponding author. E-mail: 12012142@zju.edu.cn47 # Center for Balance Architecture of Zhejiang University, Hangzhou, China48 # Binhai Industrial Technology Research Institute of Zhejiang University, Tianjin, China491. Introduction 50Compared with young and middle-aged people, the mo-51bility of the elderly is limited, and the lack of elevators in 52multistory buildings has become a major health obstacle 53for the elderly (Chen et al., 2023). Therefore, elevator in-54stallation is an important part of age-friendly community 55retrofits (Yang & Li, 2023). The Chinese government has put 56enormous effort into promoting and installing elevators in 57multistory buildings, and “installing elevators” has been in-58cluded in the State Council’s government work report for 59five consecutive years since 2018. Though various regional 60governments in China have made great efforts to develop 61the project of adding elevators to existing residential build-62ings, the actual process remains far from the policy goal. 63The implementation of the work of installing elevators in 64old neighborhoods has been hindered by reasons related 65to fund-raising, cost-sharing, approval procedures, opera-66tion, and supervision, among others (Ning, 2014). Howev-67er, according to the interviews conducted by this research 68group in Beijing with residents, communities, streets, and 69staff of the government departments in charge, the great-70est difficulty in installing elevators is the development of 71the charging and compensation mechanism for installing 72elevators, which cannot be separated from the discussion 73on the premium for installing elevators (Liu & Sun, 2019).74The hedonic price model suggests that the price of a 75housing is determined by a set of features of the housing, 76which combine to form a package of features that affect 77the price of the housing (Tsai, 2022). Compared with mul-78tistory housings without elevators, those with elevators are 79favored by most prospective buyers due to their conveni-80ent travel characteristics, and these buyers show a higher 81willingness to pay (Dai et al., 2022). The economic benefits 82brought by this willingness to pay make the elevator con-83figuration of multistory housings one of the features de-84termining the price of housings, which pushes up the price 85of multistory elevator houses (Ma et al., 2022). In recent 86years, many scholars at home and abroad have explored 87the impact that the addition of elevators to multistory 88houses has brought to the lives of residents and hous-89ing prices based on the perspectives of case studies (Liu 90& Sun, 2019) and qualitative discussions (Zhou & Tang, 912019). However, few studies have quantitatively analyzed 92the economic premium of the addition of elevators based 93on the hedonic price model.94To that end, we sought to answer the following ques-95tions: First, will the provision of additional elevators change 96the floor structure of housing prices? In other words, com-97pared with high-floor flats, can middle-floor and low-floor 98flats without elevators be sold at higher prices due to their 99100101===== PAGE 4 (score 26, numeric-paren 0, table-words 2, var-words 19) =====102103384104Z. Zhang et al. Assessing the economic premium of additional elevator: PSM hedonic analysis in Beijing, China105the variables such as transportation facilities, education-106al facilities, living facilities, landscape and parks, and the 107straight-line distance to Tiananmen Square. Finally, the 108data were screened according to the research purpose to 109remove the samples with crippled or unreliable informa-110tion. Finally, 17,952 valid samples were obtained, of which 111the number of samples without elevator was 17,095, and 112the number of samples with additional elevator was 857. 113Table 1 summarizes the definitions of all the variables as 114well as the results of descriptive statistics.1153.3. Research design and method116First, we constructed a hedonic price model to explore 117the effect of the floor on the price of housings without 118elevators versus housings with additional elevators. After 119a series of preliminary tests, we found that compared with 120other forms of models, the logarithmic hedonic price mod-121el could better explain the relationship between housing 122prices and explanatory variables with the best model ac-123curacy (Malpezzi, 2002). This model form has been devel-124oped and tested over a long time, proven to be reasonable 125and reliable (Wen et al., 2022). The first model was de-126signed as follows:127012811292130ln131i132i133j134k135P136LF137HF138X139= α + β140+ β141+142γ143+ λ + δ + ε144∑145. 146 (1)147retrofitting of elevators in old neighborhoods, Beijing has 148the advantages of early retrofitting work, strong govern-149ment support, and a deep public understanding of the 150policy, which lays a solid foundation for our study, which 151is why we chose Beijing as a case city for the study.1523.2. Data and variable description153The research data described herein were acquired from 154Lianjia’s second-hand real estate transaction website. As 155an intermediary organization with over 50% market share 156of second-hand housing in Beijing, Lianjia’s second-hand 157housing transaction records are highly comprehensive and 158representative. The transaction time is limited to March 1592017 to March 2021 after the first additional elevators 160started to be put into use. The database contains infor-161mation such as housing price, area, age, decoration status, 162floor, and orientation. According to the policy related to 163the additional elevator and the original elevator in Beijing, 164the data of the total floor height of 4–7 floors were re-165tained, and based on the related literature, 1–2 floors were 166defined as “low floor”, 3–4 floors were defined as “middle 167floor”, and 5–7 floors were defined as “high floor” (Ma 168et al., 2022). Subsequently, the POI geographic informa-169tion data were used to mine the residential neighborhood 170characteristics and location characteristics to determine 171Table 1. Variable definition and statistic description172Variable173Description174Min.175Max.176Mean177S.D.178Housing price (P)179Total transaction price of a housing (CNY)180505,00018126,000,0001824,162,2901831,773,343184Area (AR)185Building area of a housing (m2)18613.51187338.8618877.9718930.62190Age (AG)191Age of building (years)19221937119423.211959.00196Floor (F)197Floor level of a housing198119972003.122012.43202Low floor (LF)203Whether the housing is located on a low floor204(1 = yes; 0 = no)205020612070.162080.37209High floor (HF)210Whether the housing is located on a high floor211(1 = yes; 0 = no)212021312140.362150.48216Elevator (E)217Whether the housing with an installed elevator218(1 = yes; 0 = no)219022012210.052220.21223South (S)224Whether the housing with a south-facing orientation 225(1 = yes; 0 = no)226022712280.702290.46230Decoration (D)231Whether the housing well decorated232(1 = yes; 0 = no)233023412350.412360.49237Bedroom number (BN)238Number of bedrooms239024062412.122420.71243Living room number (LN)244Number of living rooms245024642471.142480.39249Educational facilities (EF)250Primary school, secondary school, and senior high 251school within 1 km of a housing, each one takes 2521 score and total 3 scores253025432552.432560.80257Living facilities (LF)258Shopping malls, 3A hospitals, post offices and banks 259within 1 km of a housing, each one takes 1 score 260and total 4 scores261026242633.412640.73265Traffic facilities (TF)266Bus stops and subway stations within 1 km of a 267housing, each one takes 1 score and total 2 scores268026922701.622710.50272Landscapes & parks (LP)273Landscapes and parks within 1 km of a housing, 274each one takes 1 score and total 2 scores275027622771.382780.76279Distance to distance 280(D_TAM)281Straight–line distance from a housing to Tiananmen (m)282289.02283111,815.4628415,335.102858,461.19286287288===== PAGE 5 (score 15, numeric-paren 0, table-words 4, var-words 7) =====289290International Journal of Strategic Property Management, 2024, 28(6), 381–392291385292In the model, P is the price of the housing, LF is wheth-293er the housing is located on a low floor, HF is whether the 294housing is located on a high floor, Xi is the other control 295variables, where the discrete variables are in their original 296form, and the continuous variables were introduced into 297the model using logarithmic form, α0, β1, β2, and γi are the 298coefficients of the characteristic variables to be estimated, 299λj is the year fixed effect, δk is the district fixed effect, and 300ε is the error term.301Unlike urban public goods with complete non-exclu-302sivity, the impact of urban semi-public goods with par-303tial non-exclusivity, such as elevators, and electric vehicle 304charging piles (Gao et al., 2022) on the price of housings is 305more complicated. The benefits of these urban semi-public 306goods can only be enjoyed by residents living within spe-307cific boundaries and thus are usually defined as dichoto-308mous variables (Tsai, 2022). If these dichotomous variables 309are directly added into the OLS hedonic price model, the 310systematic differences between dwellings equipped with 311urban semi-public goods and those not equipped with 312urban semi-public goods are often overlooked, result-313ing in a selectivity bias in the sample. Consequently, a 314PSM model with whether or not to equip an additional 315elevator as the core explanatory variable was constructed 316to remove the effect on the model caused by endogeneity 317from sample self-selection (Wang & Li, 2022). To enhance 318the robustness of the empirical results, we chose three 319matching methods: K-nearest neighbor matching (KNM), 320radius matching (RM), and kernel matching (KM). Further-321more, we retained the maximum loss data of the samples 322under the three different matching methods and con-323structed the model based on the matched data.32403251326ln327i328i329j330k331P332E333X334= α + β335+336γ337+ λ + δ + ε338∑339; 340(2)341034213432344ln345i346i347j348k349P350E351E352F353X354= α + β355+ β356×357+358γ359+ λ + δ + ε360∑361. 362(3)363E is an additional elevator variable in the model, and 364the interaction term E × F for floor and additional elevator 365were added to examine the interaction effect of floor and 366additional elevator on housing prices, and the other vari-367ables are the same as in Model 1. 368In addition, to examine the interaction effect of orien-369tation and additional elevator on housing prices, the sin-370gle orientation feature variable and the interaction term 371E × Single for single orientation and additional elevator 372were added to the model, and the final model form was 373obtained as follows:3740375137623773378ln379.380i381i382j383k384P385E386Single387E388Single389X390= α + β391+ β392+ β393×394+395γ396+ λ + δ + ε397∑398 399(4)400Single is the single orientation variable in the model, 401E × Single is the interaction term between the two, and the 402other variables are the same as in Model 1.403Likewise, the interaction term E × lnAG for age and 404additional elevator were added to the model to examine 405the interaction effect of age and additional elevator on 406housing prices, and the final model form was obtained 407as follows:4080409141024113412ln413ln414ln415.416i417i418j419k420P421E422AG423E424AG425X426= α + β427+ β428+ β429×430+431γ432+ λ + δ + ε433∑434 435(5)436The ordinary least squares (OLS)-based regression 437method ignores the real estate market differentiation ef-438fect of the additional elevator premium. To capture the 439change in the additional elevator premium in the distribu-440tion of housing prices in the overall market, we introduced 441a quantile regression model to assess the heterogeneous 442impact of the additional elevator on the housing prices at 443different price points in the form of the following model:4440445ln446( )447( )448( )449j450i451i452j453k454P455E456X457= α458τ +459β460τ461+462γ τ463+ λ + δ + ε464∑465∑466, 467(6)468where τ is the quantile point, and the other variables are 469consistent with the baseline model.4704. Results and discussion4714.1. Impact of the floor level on housing 472prices 473Table 2 presents the results of OLS regression analysis for 474the sample without an elevator and the sample with an 475additional elevator. We compared the results of the re-476gressions for the two sets of samples and analyzed them. 477Based on our analysis, we conclude that for the sample 478without elevator, the impact coefficient of low floors on 479housing prices is 0.004 and non-significant, and the impact 480coefficient of high floors on housing prices is −0.063 and 481significant at the 1% level, indicating no significant differ-482ences between housing prices of middle floors and those 483of low floors, and that the housing prices of high floors 484have decreased by 6.3% compared with middle floors. This 485suggests that for housings that are not equipped with el-486evators, residents are more concerned about the positive 487externalities of the convenience of travel on the middle 488and low floors than the negative externalities due to their 489uncomfortable environmental factors. They show a higher 490willingness to pay for middle and low-floor housings com-491pared with high-floor housings. While in the analysis of 492Model 2 on the sample of adding elevators, the coefficient 493of the impact of low floors on the price of housings is 494−0.055 and significant at the 5% significant level, the coef-495ficient of the impact of high floors on the price of hous-496ings is 0.014 and non-significant, the price of housings in 497the middle floors and high floors is not significantly differ-498ent, and the price of housings in the low floors compared 499with those in the middle floors decreases by 5.5%. Most 500additional elevator housings were built in the 20th century 501in older communities, where drainage and community en-502vironment are poorer, and residents living there are more 503likely to feel the noise, mosquitoes, humidity, and air pol-504lution caused by low floor (Maloutas & Botton, 2023). In 505this case, they express a higher willingness to pay for the 506additional elevator high-floor housings that compensate 507508509===== PAGE 6 (score 79, numeric-paren 41, table-words 12, var-words 6) =====510511386512Z. Zhang et al. Assessing the economic premium of additional elevator: PSM hedonic analysis in Beijing, China513for the negative externalities of accessibility. Comparing 514the results of the two sets of regressions reveals a differ-515ence in the effect of the floor on the price of dwellings in 516the no-elevator sample and the additional-elevator sam-517ple. Therefore, H1 is partially verified.5184.2. Impact of additional elevators and floor 519level on housing prices520Subsequently, we applied the PSM–hedonic model to 521analyze the differential premium of adding an elevator to 522housings on different floors. After using the PSM method 523to deal with endogeneity due to sample self-selection, the 524maximum loss data of the samples under the three differ-525ent matching methods were retained, and based on the 526matched data, the economic premium generated by the 527addition of elevators on different floors of housings was 528analyzed. Based on our results (Table 3), we conclude that 529the effect of adding an elevator is significant at the 1% lev-530el for the full sample as well as for the middle- and high-531floor samples, while the effect on the low-floor sample is 532not significant. Specifically, the estimated coefficient value 533for adding an elevator in the full sample is 0.054, a result 534that suggests that the addition of an elevator housing is 5355.4% more expensive than a housing without an elevator. 536Furthermore, residents living in housings on middle and 537high floors are willing to pay more to live in a housing with 538an elevator, and the magnitude of the effect is in the order 539Table 2. Empirical results of without elevator sample and installed elevator sample540Variable541Without elevator sample542Installed elevator sample543Coef.544p-value545Coef.546p-value547LF5480.0045490.277550–0.055**5510.048552HF553–0.063***5540.0005550.0145560.477557lnAR5580.776***5590.0005600.827***5610.000562lnAG563–0.040***5640.000565–0.034*5660.083567S5680.032***5690.0035700.0275710.282572D5730.052***5740.0005750.064***5760.001577BN5780.034***5790.0005800.058***5810.001582LN5830.014**5840.0155850.0205860.371587EF5880.007***5890.0005900.006***5910.000592LF5930.018***5940.0005950.033***5960.007597TF5980.071***5990.0006000.055**6010.011602LP6030.023***6040.0006050.0036060.826607lnD_TAM608–0.159***6090.000610–0.158***6110.000612Constant61313.617***6140.00061513.714***6160.000617Year FE618Yes619Yes620District FE621Yes622Yes623Adj. R26240.8206250.713626Observations62717095628857629Notes: ***, **, and * indicate that the estimates are significant at the 1%, 5%, and 10% levels.630Table 3. Results of basic regression and interaction analysis of the effect of floor level and additional elevator on housing prices631Variable632Whole sample633LF sample634MF sample635HF sample636Model 2637Model 3638Model 2639Model 3640Model 2641Model 3642Model 2643Model 3644E6450.054***646–0.023***6470.0246480.0116490.070***650–0.007***6510.119***652–0.061***653E × F6540.025***6550.008***6560.019***6570.033***658Constant65913.533***66013.462***66113.120***66213.099***66313.608***66413.573***66513.813***66613.721***667Controls668Yes669Yes670Yes671Yes672Yes673Yes674Yes675Yes676Year FE677Yes678Yes679Yes680Yes681Yes682Yes683Yes684Yes685District FE686Yes687Yes688Yes689Yes690Yes691Yes692Yes693Yes694Adj. R26950.8146960.8146970.7946980.7946990.8187000.8187010.8247020.825703Observations70417750705261770686437076124708Notes: ***, **, and * indicate that the estimates are significant at the 1%, 5%, and 10% levels.709710711===== PAGE 7 (score 72, numeric-paren 36, table-words 13, var-words 6) =====712713International Journal of Strategic Property Management, 2024, 28(6), 381–392714387715of high floors (0.119) > middle floors (0.070). This suggests 716that the positive externality of convenience of travel from 717additional elevators gradually increases as the floor rises, 718and H1 is verified. However, residents living on the lower 719floors are unwilling to pay more for living in a housing 720equipped with an original elevator. This is ascribable to 721the fact that the additional elevator does not reduce their 722travel costs and even negatively affects their lives due to 723the impact on the ventilation and lighting of the housing. 724In addition, in order to further analyse the change in 725the economic premium of adding elevators with the floor, 726we added an interaction term E × F to Model 2 and plotted 727Figure 1. The positive coefficient of E × F shows that as 728the floor increases, the coefficient of adding elevators 729will increase, further illustrating the floor differences in 730the premium of adding elevators. Besides, we further 731confirmed the nonlinear relationship between floor and 732the coefficient of adding elevators through a regression 733with low, medium and high floor sub-samples. Specifically, 734the coefficient of adding elevators does not increase 735uniformly with the floor level. It increases slightly at low 736floor level, while increases dramatically at medium and 737high floor level.7384.3. The moderating effect of orientation and 739age on the premium for additional elevators740The addition of an elevator, while convenient for residents 741to get around, also negatively impacts the safety, silence, 742lighting and ventilation in the housing. This negative 743impact is more pronounced for single-facing and older 744housings than for newer and multi-facing housings. On 745the other hand, the floor level is an important factor influ-746encing the silence and lighting of a house, and the effect 747of adding an elevator on the silence and lighting of houses 748on different floors may also be different (Jia et al., 2014). 749To explore the impact of orientation and housing age on 750the premium of adding elevators for housings on differ-751ent floors, we introduced the single orientation variable 752into the model as well as the interaction term E × Single 753and E × lnAG for hedonic price analysis. The results of the 754analysis are summarized in Table 4. The direction and sig-755nificance of the effect of single orientation on the price of 756dwellings is not stable in the regressions for the full sam-757ple as well as for the low-, middle- and high-floor samples. 758However, the coefficient on the interaction term E × Single 759and E × lnAG are significantly negative in the full sample 760as well as in the regressions for the low-, middle- and 761high-floor samples. This suggests a moderating effect of 762orientation and housing age on the premium for adding 763an elevator. In contrast, the benchmark model only cap-764tures the average premium for adding an elevator. With 765this, H2, H3 was tested.766To compare the difference between the premium 767for adding an elevator to a single-facing housing and a 768multi-facing housing on different floors, we plotted a line 769graph of the premium for adding an elevator (Figure 2). 770The coefficients of the model’s add-elevator variable are 771the coefficients of the premium for adding an elevator to 772a multi-facing housing. The coefficient on the interaction 773term is the coefficient on the housing price loss of add-774ing an elevator to a single-facing housing compared with 775a multi-facing housing. The coefficient on the premium 776Figure 1. Relationship between additional elevator economic 777premium coefficients and floor level778Table 4. Interaction analysis of the effect of orientation, age and additional elevator on housing prices779Variable780Whole sample781LF sample782MF sample783HF sample784Model 4785Model 5786Model 4787Model 5788Model 4789Model 5790Model 4791Model 5792E7930.090***7940.133***7950.045**7960.040*7970.094***7980.103***7990.156***8000.217***801lnAG802–0.065***803–0.062***804–0.055***805–0.059***806–0.083***807–0.080***808–0.083***809–0.071***810Single811–0.000812–0.0168130.029***814–0.025***815E × Single816–0.123***817–0.091***818–0.125***819–0.151***820E × lnAG821–0.015***822–0.005**823–0.013***824–0.039***825Constant82613.594***82713.535***82813.175***82913.140***83013.590***83113.551***83213.928***83313.782***834Controls835Yes836Yes837Yes838Yes839Yes840Yes841Yes842Yes843Year FE844Yes845Yes846Yes847Yes848Yes849Yes850Yes851Yes852District FE853Yes854Yes855Yes856Yes857Yes858Yes859Yes860Yes861Adj. R28620.8158630.8158640.7948650.7948660.8208670.8188680.8268690.824870Observations87117750872261787386438746124875Notes: ***, **, and * indicate that the estimates are significant at the 1%, 5%, and 10% levels.876877878===== PAGE 8 (score 28, numeric-paren 0, table-words 9, var-words 10) =====879880388881Z. Zhang et al. Assessing the economic premium of additional elevator: PSM hedonic analysis in Beijing, China882for adding an elevator to a single-facing housing is thus 883the sum of the coefficients on the add elevator variable 884and the coefficients on the interaction term in the model. 885For the full sample, the premium coefficient of adding an 886elevator for multi-directional housings is 0.090, whereas 887the premium coefficient of adding an elevator for single-888directional housings is −0.033. That is to say, adding an 889elevator for single-directional housings not only does 890not increase the price of housings but also decreases the 891price of housings by 3.3%, which reflects a strong nega-892tive externality of adding an elevator for single-directional 893housings. Regarding floor heterogeneity, the coefficient 894of housing price loss for adding an elevator to a single-895facing house increases as the floor rises. This is related 896to the fact that higher floors have better light and views. 897Compared with the middle and lower floors with average 898light and views, the shading of an additional elevator has 899the greatest impact on the lives of residents of the higher 900floors who would otherwise have a wider view. For multi-901facing housings, adding an elevator will increase the price 902of housings on the middle and upper floors. For single-903facing housings, adding elevators will only increase the 904price of housings on the higher floors, reducing the price 905of housings on the middle and lower floors. This is the 906result of the trade-off between the negative externality of 907adding elevators affecting lighting and ventilation and the 908positive externality of reducing travel costs.909Similarly, in order to compare the difference in 910the premium for adding an elevator to newer and 911older housings on different floors, we plot Figure 3. 912The negative coefficient on E × lnAG indicates that the 913premium coefficient for adding an elevator decreases 914as the age of the housing increases. Specifically, the 915coefficient of premium for adding an elevator is 0.093 for 916newer housings (with lnAG one standard deviation below 917the mean) and 0.081 for older housings (with lnAG one 918standard deviation above the mean) in the full sample, 919suggesting that negative externalities such as safety 920hazards and noise associated with the addition of elevators 921have a stronger impact on older housings with a lower 922level of construction. In terms of floor heterogeneity, the 923coefficient of E × lnAG increases with the rise of floors, 924which is due to in the additional elevator buildings, due 925to the initial design did not reserve the space for the 926winch, and thus the winch is often installed on the top 927floor, causing great noise nuisance to the residents of 928the upper floors. For older housings with thinner walls, 929the intolerable noise caused by the winch has a stronger 930impact on the lives of residents of the upper floors.9314.4. The socially heterogeneous impact of 932additional elevators on housing prices933From the quantile regression results in Table 5 and Figure 4, 934we deduce that the impact of the variable of additional el-935evator on the price of housings at different quantile points 936is different. We only report the coefficients of adding an 937elevator due to space considerations. In the first column, 938the impact of additional elevator on the price of housings 939at Q0.1 is not significant, and the semi-elasticity coeffi-940cient of additional elevator on the price of housings at 941Q0.9 reaches the maximum value of 0.102. Furthermore, 942the coefficient shows a gradual increase from the low- to 943the high-quantile points, which indicates that buyers of 944high-priced housings will pay more attention to the el-945evator configuration of the housing, while purchasers of 946low-priced housings do not pay much attention to the ad-947ditional elevator, H4 was tested. As stated earlier, buyers of 948high-priced housings generally have higher incomes; they 949have higher demand for life-enhancing amenities, such as 950additional elevators as well as for the environment of the 951community in which their housings are located. The plot 952ratio of the high-end community is generally lower, and 953the negative externalities of the additional elevator affect-954ing ventilation and lighting are limited, which tend to pro-955duce a relatively high premium for the additional elevator. 956For low-priced housings, the relevant facilities and the sur-957rounding environment are poor, and the community floor 958area ratio is high. At this time, the addition of elevators will 959largely affect the ventilation and lighting, thus affecting 960buyers’ willingness to pay. Therefore, the quantile regres-961sion results show the above trend.962Next, to analyze whether there is a difference in the 963premium for adding an elevator for low-, middle- and 964Figure 2. Additional elevator economic premium coefficients 965in single and multiple orientation samples966Figure 3. Additional elevator economic premium coefficients 967in older and newer samples968969970===== PAGE 9 (score 77, numeric-paren 50, table-words 6, var-words 9) =====971972International Journal of Strategic Property Management, 2024, 28(6), 381–392973389974high-floor housings at different quartiles, we re-estab-975lished the quartile regression model based on the market 976segments of low, middle, and high floors. For low-floor 977housings, the premium coefficient of adding an elevator 978is significantly negative between Q0.1 and Q0.2, signifi-979cantly positive between Q0.8 and Q0.9, and non-signifi-980cant between Q0.3 and Q0.7, suggesting that adding an 981elevator reduces the price of low-priced housings on low 982floors and raises the price of high-priced housings on low 983floors. This is attributable to the weaker positive exter-984nality of the additional elevator on low floors to facilitate 985travel. As the price of the housing rises, the grade of the 986community improves, and the floor area ratio increases, 987the negative externality of the additional elevator on low 988floors that affects the ventilation and lighting is weakened, 989its net externality shifts from negative to positive, and the 990premium coefficient also gradually increases. As for the 991middle- and high-floor housings, except for the premium 992coefficient of adding elevators on middle floors, which is 993not significant at Q0.1 and Q0.2, all other coefficients are 994significant, and the coefficients exhibit a trend of gradual 995increase. The elevator premium increases with the rise of 996the price of housings, similar to the results of the whole 997sample regression. 998Column (5) and (6) encapsulates the estimation results 999based on OLS regression and quantile regression using the 1000sample data of commercial and protected housing, which 1001show that the coefficients of additional elevator premi-1002ums for commercial and protected housing are 0.047 and 10030.052, respectively, and both of them are significant at the 10041% level. That is to say, under other conditions, the prices of 1005commercial housing and affordable housing with elevators 1006are respectively 4.7% and 5.2% higher than that of housing 1007without elevators, and the premium for affordable hous-1008ing with elevators is slightly higher than that of commer-1009cial housing. The strong demand for additional elevators 1010from mobility-impaired residents in the affordable housing 1011community outweighs the depreciation of the investment 1012value of additional elevators due to poor liquidity in the 1013affordable housing market. Figure 5 show the changes in 1014the coefficients of the premiums for adding elevators at 1015different quartiles, H5 was tested. The results demonstrate 1016that the premium coefficients of adding elevators for com-1017mercial housing housings show a gradual increase from the 1018lower to the higher quartile points, which is not significantly 1019different from the results of the previous analysis for the 1020full sample. However, for affordable housing, the premium 1021coefficient of adding elevators shows a trend of increasing 1022Table 5. Results of quantile regression1023Quantile1024(1)1025(2)1026(3)1027(4)1028(5)1029(6)1030Whole sample1031LF sample1032MF sample1033HF sample1034AH sample1035CH sample1036OLS10370.054***10380.02410390.070***10400.119***10410.052***10420.047***1043Q.0.11044–0.0101045–0.099***1046–0.00910470.064***10480.019***1049–0.0151050Q.0.210510.031***1052–0.027*10530.01410540.083***10550.050***10560.027***1057Q.0.310580.038***1059–0.00910600.024**10610.091***10620.054***10630.032***1064Q.0.410650.053***1066–0.00310670.031***10680.125***10690.058***10700.056***1071Q.0.510720.087***10730.01310740.058***10750.148***10760.059***10770.080***1078Q.0.610790.091***10800.01610810.087***10820.155***10830.080***10840.095***1085Q.0.710860.098***10870.04210880.090***10890.156***10900.082***10910.099***1092Q.0.810930.102***10940.050*10950.091***10960.158***10970.073***10980.103***1099Q.0.911000.102***11010.055***11020.093***11030.159***11040.062***11050.108***1106Observations110717750110826171109864311106124111159961112117541113Notes: ***, **, and * indicate that the estimates are significant at the 1%, 5%.1114Figure 4. Quantile regression results of the whole sample 1115and floor-level sub-samples1116Figure 5. Quantile regression results of commercial housing 1117and affordable housing samples111811191120===== PAGE 10 (score 12, numeric-paren 0, table-words 0, var-words 12) =====112111223901123Z. Zhang et al. Assessing the economic premium of additional elevator: PSM hedonic analysis in Beijing, China1124and then decreasing from the low to the high tertile, which 1125can be analyzed from the two aspects of the use value 1126and investment value of residential housing. For low- and 1127middle-priced housings, buyers are more concerned about 1128the use value of housings, and the strong demand for addi-1129tional elevators from mobility-impaired residents in the af-1130fordable housing community triggers a higher willingness to 1131pay. For middle- and high-priced housings, buyers are more 1132concerned about the investment value of housings, and the 1133poor liquidity in the housing market leads to the loss of the 1134investment value of the additional elevator, which has never 1135been noticed by scholars before (Ma et al., 2018).11365. Conclusions and policy implications1137With the gradual aging of the population, age-friendly 1138multistory houses equipped with elevators are increas-1139ingly favored by home-buyers. In this study, we used the 1140PSM–hedonic and quantile–hedonic models to delve into 1141people’s willingness to pay for an additional elevator. In 1142contrast to existing studies, while analysing the change in 1143the premium for additional elevators with the floor, we 1144also discussed in depth the possible negative impact of 1145additional elevators on residents’ lives through the mod-1146erating effect of orientation and age of the housing (Liu 1147& Sun, 2019). Additionally, we comprehensively examined 1148the differences in the premiums for adding an elevator 1149in two housing markets, commercial community housings 1150and affordable community housings, which has never ap-1151peared in previous studies (Ning, 2014). 1152The above findings reveal an important reference 1153for promoting the project of adding elevators to exist-1154ing housings in old neighborhoods. First, in carrying out 1155renovation projects to add elevators to old residential ar-1156eas, when faced with the demands and conflicts of owners 1157on different floors regarding damage and benefits, com-1158pensation and contributions, etc., grassroots staff such 1159as street offices and neighborhood committees can refer 1160to the quantitative results in this article, use quantitative 1161data as the basis for communication work, and alleviate 1162communication conflicts. This provides a basis for ap-1163propriately charging and subsidising residents on differ-1164ent floors. Second, when setting up a funding model for 1165elevator retrofitting, the differences in price due to floor 1166level, orientation, age, housing price and community type 1167should be taken into account to determine a reasonable 1168model for additional elevator installation. Specifically, 1169for single-oriented buildings and buildings that are too 1170old, the relevant departments need to carefully assess 1171the lighting, sound insulation and safety of the building 1172before adding an elevator to avoid negative impacts. For 1173low-value communities and affordable housing communi-1174ties, the relevant departments should do a good job of 1175subsidising and publicising the work to ensure that the 1176benefits of adding elevators are enjoyed by residents in 1177need as much as possible. Finally, the measurement of 1178the benefits of retrofitting elevators in old communities 1179and the construction of a funding mechanism provide a 1180reference for the implementation of other improvement 1181and upgrading urban renewal projects. For other types 1182of urban renewal projects, local authorities can also first 1183measure their economic benefits, guide the beneficiaries 1184of the project to make reasonable contributions, solve the 1185financing difficulties of urban renewal projects, alleviate 1186the financial pressure on local governments, and promote 1187the stable and long-term development of urban renewal.1188However, our study still has some limitations. First, as 1189the same housing is not usually traded repeatedly for a 1190short period of time, we could not get the transaction prices 1191before and after the addition of lifts to the same housing, 1192and we did not use methods such as DID to explore the 1193changes in the price of housings before and after the ad-1194dition of elevators, which is an important direction for our 1195future research. Second, due to the limitation of technical 1196means, we could inaccurately obtain the lighting time of 1197each housing and only used the orientation variable to ex-1198plore the negative externality of adding elevators, which 1199concluded that the study was not precise and comprehen-1200sive enough. Therefore, using urban 3D modeling and other 1201means to fully understand the impact of additional elevators 1202on residential lighting is a research topic worthy of attention 1203by future scholars. Finally, due to privacy protection reasons, 1204we do not have access to personal data of home-buyers. If 1205data on the personal characteristics or demand of home-1206buyers is available, it may be possible to further enrich 1207understanding of the impact of additional elevators on 1208housing prices from the perspective of demand differences.1209Acknowledgements1210The authors would like to thank the editor and the anon-1211ymous reviewers for the positive and constructive com-1212ments and suggestions.1213Funding 1214This work was supported by the Center for Balance 1215Architecture, 1216Zhejiang 1217University 1218under 1219Grant 1220No. 20203512-28C; The National Natural Science Founda-1221tion of China under Grant No. 72174178; and National Key 1222Research and Development Program of China under Grant 1223No. 2022YFC3601604.1224Author contributions 1225Zhao Zhang was responsible for formal analysis, writing 1226original draft and writing review & editing. Yihua Mao 1227was responsible for conceptualization, methodology and 1228funding acquisition. Yueyao Shui was responsible for 1229investigation. Ruyu Deng was responsible for data curation.1230Disclosure statement 1231The authors declare that they do not have any competing fi-1232nancial, professional, or personal interests from other parties.1233