# RECORD e4f33aca0968c7c930129cd6f3fbf32507e5ac89 # TITLE: THE EFFECTS OF A SHOPPING MALL ON HOUSING PRICES: A CASE STUDY IN HANGZHOU # YEAR: 2018.0 VENUE: International Journal of Strategic Property Management DOI: 10.3846/ijspm.2019.6360 # AUTHORS: Ling Zhang; Jiantao Zhou; E. Hui; Hai-Zhen Wen # PAGES: 16 CANDIDATE TABLE PAGES (1-based): [7, 8, 9, 11, 14, 15, 16] # ABSTRACT: There are few studies on the externalities of shopping malls affecting the housing market. This study aims to discuss two issues: (1) What is the intensity of the impact of a shopping mall? (2) When does the external influence of a shopping mall begin to reveal itself? The West Intime Shopping Mall in Hangzhou offers a unique situation to research the questions. By dividing the study area into nine blocks, using hedonic price theory, and the price gradient approach with housing price data from 2011 to 2015, we found that in the space dimension, the mall exerted a significantly positive effect on the housing prices of nearby blocks. With the increase in distance from the mall, the positive effect decreased. There were more significantly positive effects in blocks far away from the city center. In the time dimension, the effects of West Intime did not reveal themselves until the mall had started to operate and gradually matured over time, implying that the mall did not have the obvious expected impact on housing prices before the mall had begun operating. ===== PAGE 1 (score 11, numeric-paren 0, table-words 0, var-words 11) ===== Copyright © 2019 The Author(s). Published by VGTU Press This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unre- stricted use, distribution, and reproduction in any medium, provided the original author and source are credited. *Corresponding author. E-mail: wenhaizhen@263.net International Journal of Strategic Property Management ISSN: 1648-715X / eISSN: 1648-9179 2019 Volume 23 Issue 1: 65–80 https://doi.org/10.3846/ijspm.2019.6360 THE EFFECTS OF A SHOPPING MALL ON HOUSING PRICES: A CASE STUDY IN HANGZHOU Ling ZHANG1, Jiantao ZHOU1, Eddie C. M. HUI2, Haizhen WEN1,* 1 Center for Real Estate Studying, Department of Civil Engineering, Zhejiang University, China 2 Department of Building and Real Estate, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, China Received 4 December 2017; accepted 30 August 2018 Abstract. There are few studies on the externalities of shopping malls affecting the housing market. This study aims to discuss two issues: (1) What is the intensity of the impact of a shopping mall? (2) When does the external influence of a shopping mall begin to reveal itself? The West Intime Shopping Mall in Hangzhou offers a unique situation to research the questions. By dividing the study area into nine blocks, using hedonic price theory, and the price gradient approach with housing price data from 2011 to 2015, we found that in the space dimension, the mall exerted a significantly positive effect on the housing prices of nearby blocks. With the increase in distance from the mall, the positive effect decreased. There were more significantly positive effects in blocks far away from the city center. In the time dimension, the effects of West Intime did not reveal themselves until the mall had started to operate and gradually matured over time, implying that the mall did not have the obvious expected impact on housing prices before the mall had begun operating. Keywords: shopping mall, housing price, price index, price gradient, hedonic models. Introduction With the suburbanization and economic growth occur- ring since the 1960s in the United States, shopping centers and malls have been spreading throughout cities, first in the suburbs, then in the downtown areas (Carter, 2009). Shopping malls have dramatically reshaped the retailing industry, providing the impetus for regional economic de- velopment and the improvement of living standards. Now, shopping malls are major centers of retail activity in the United States and around the world (Larsen, Shelton, & Wright, 2015). In recent years, alongside urban develop- ment in China, shopping malls have grown quickly with urbanization, versatility, and convenience in transporta- tion. Most malls deliver convenience to local residents and have strong influences on neighboring areas. Large-scale shopping malls have become attractive places to shop, vis- it, and spend time. Most malls have also increased their leisure activity components (Fasli, Riza, & Erbilen, 2016). There have been many types of studies focusing on the designs, functions, rents, and satisfactoriness of shop- ping malls (Carter, 2009). Some studies have discussed the role of shopping centers in societies and urban spaces, but most have not covered the externalities of shopping cent- ers. Shopping malls can deliver commercial convenience to the surrounding residences, indirectly increasing the values of these properties in the housing market. There- fore, it makes sense to explore the effects of shopping malls on housing prices. Compared to infrastructure, schools, and other public facilities, shopping malls are so complex that they may possess distinctive externalities. Sirpal (1994) and Des Ro- siers, Lagana, Thériault, and Beaudoin (1996) discovered positive and negative effects at different distances from shopping malls. The externalities are directly related to the internalities, such as the scale, design and anchor tenants, of a mall. Moreover, the influence of a mall may change at different stages over time. Promotional advertising be- fore a mall’s construction may fuel a rise in housing prices while noise from the construction works could worsen the living conditions of residents and lower the housing pric- es. After a mall’s completion, its external influence usually differs with the maturity of its operation and the establish- ment of word of mouth. Such externalities, particularly in terms of their scope and times of impacts on the housing ===== PAGE 7 (score 24, numeric-paren 0, table-words 3, var-words 8) ===== International Journal of Strategic Property Management, 2019, 23(1): 65–80 71 ral environment, living facilities, educational facilities, and sports facilities. In China, parents pay much attention to children’s education. The school district significantly influ- ences housing prices (Feng & Lu, 2013; Wen, Y. Zhang, & L. Zhang, 2014b), so this paper introduces the quality of schools and the distances to schools to characterize the impact of education and to measure school quality in the same way as did the research of Wen et al. (2014b) on Hangzhou. Commercial facilities also have a certain de- gree of impact on housing prices. Small commercial facili- ties can be considered together with living facilities due to their small influence while the impact of shopping malls is examined separately. The sample data are the housing transaction prices from the beginning of 2011 to the end of 2015, which were ob- tained from a real estate agency in Hangzhou. Some rel- evant information, such as living facilities, sports facilities, and bus route, was not provided by the agency but obtained through a field survey conducted in 2010, 2012 and 2014. Our institute has conducted follow-up surveys on the resi- dential buildings in Hangzhou every two to three years since 2008 to score the neighborhood and location characteristics of the community. A questionnaire has been designed to measure variables of interior environment, surrounding en- vironment, sports facility, property management, quality of primary schools, and quality of junior high schools. Mainly two aspects are considered in the choice of the research period. First, the real estate around West Intime developed later than that in the main city area. Although the average price of second-hand houses in the study area increased from 2008 to 2011, the housing transactions be- fore 2011 of Zones 6, 7, 8 and 9 are absent, so the pre- construction period cannot form a price gradient contrast. Second, data on the marketing activities of the mall before construction are scarce. In addition, there was little adver- tising about West Intime before the construction. Therefore, we chose observations after 2011 for the sake of the equi- librium of sample spatial distribution. To ensure the quality and quantity of the data, we have adopted the price data of the second-hand houses rather than the newly-built hous- es. According to the statistical data, from 2011 to 2015 the transaction volume of newly-built houses was about 10,000 units in study area, which is very close to the volume of the second-hand houses. However, these newly-built units are concentrated in 16 communities and mostly distributed in the area north of Zone 6. There are few newly-built hous- es in Zones 1 to 5. The trading units of the second-hand houses are evenly distributed in 71 communities, which can better reflect the housing prices in the study area. To distin- guish the changing impacts of the different stages, the data are divided into two periods: the construction period (P1: January 2011 – August 2013) and the operating period (P2: September 2013 – December 2015). The number of second- hand transactions in this area was 838 in 2011, which dou- bled in 2012 to 1,659, and continued to grow to 1,943 in 2013. The increasing volume of second-hand transactions and the significant growth in the construction of newly- built houses north of West Intime reflect the signal effect of shopping malls on the housing market. The descriptive statistics of the variables are shown in Table 4. Table 4. Descriptive statistics Period 1 (January 2011 – August 2013, N = 3939) Period 2 (September 2013 – December 2015, N = 5315) Min Max Mean Std.dev. Min Max Mean Std.dev. Sale price (¥/m2) 8228.00 37623.17 20374.56 4214.88 8203.45 40851.23 20195.14 4613.46 Size (m2) 23.82 264.60 90.09 31.105 23.71 330.98 94.79 31.259 Floor/Level No. 1 27 6.21 5.103 1 27 6.44 5.136 Total floors 4 29 11.56 7.445 4 29 12.14 7.516 Building’s age (year) 1 25 11.42 6.446 3 27 12.69 5.834 Distance to CBD (km) 4.39 10.79 6.60 1.672 4.39 10.79 6.69 1.626 Bus route 2 36 20.15 7.390 2 36 19.57 7.146 Distance to West Intime (km) 0.51 6.50 3.53 1.269 0.51 6.50 3.48 1.300 Distance to Hangzhou Incity (km) 0.46 7.40 3.00 1.346 0.46 7.40 2.96 1.277 Distance to Xicheng Square (km) 0.65 7.80 3.24 1.933 0.65 7.80 3.27 1.809 External environment 1 5 3.14 0.670 1 5 3.24 0.666 Inner environment 1 5 3.36 0.861 1 5 3.53 0.859 Living facilities 1 5 4.05 1.251 1 5 3.96 1.240 Sports facilities 1 5 2.66 1.081 1 5 2.81 1.092 Property management 1 5 3.03 0.852 1 5 3.18 0.796 Quality of primary school 1 3 2.02 0.847 1 3 1.98 0.862 Quality of junior high school 1 2 1.69 0.462 1 2 1.64 0.480 Distance to primary school (km) 0.10 2.06 0.59 0.369 0.10 2.06 0.59 0.366 Distance to junior high school (km) 0.19 3.28 1.42 0.838 0.19 3.28 1.45 0.841 ===== PAGE 8 (score 103, numeric-paren 57, table-words 14, var-words 5) ===== 72 L. Zhang et al. The effects of a shopping mall on housing prices: a case study in Hangzhou 4. Analysis of results The regression of the three models produced good results, which are discussed below. 4.1. The results of simple hedonic price modelling approaches To estimate the impact of West Intime on the surrounding housing prices, we take advantage of the road distance to the mall to capture the price elasticity of the surround- ing housing. The SPSS was used to estimate Equation (1) and analyze the housing characteristics. The results of the regression model are shown in Table 5. Economic theory places few restrictions on the form of the hedonic price function (Hui et al., 2007; Feng & Lu, 2013). By comparing the regression results of commonly adopted forms such as linear, semi-logarithmic (as shown in Table A2) and logarithmic, we found that logarithmic forms fit slightly better than the others. The correlation between independents was smaller and coefficient inter- pretation was more significant. The adjusted R2 of Model 1 is 0.642, which means that the independent variables could explain 64.2% of the dependent variable. Almost all the 21 variables are significant at the 1% level. To test for the spatial independence of the estimated error terms, referring to Fik et al. (2003), the residuals were regressed on the independent variables and location coordinates x, y and their polynomial expression. The coefficient esti- mates of distance independent variables are all insignifi- cant. While the location variable x2 is significant in Model 1, there is no significant location variable in Model 2 or Model 3, in which the spatial dummy variables are in- cluded (see Table 5). Due to the lack of location coordi- nates for housing units, and in order to solve the possible spatial correlation problem, we estimated the coordinates for house units by random calculations with GIS, based on the coordinates and shapes of the communities. Then, we conducted spatial econometric models using the con- structed spatial weights as a robustness test. The results are shown in Appendix 2. The effect of the structure characteristics all obtain the expected results. The coefficient of Total floors is Table 5. Results of regression of the three models Independent variables Model 1 Model 2 Model 3 Coefficient T – statistics Coefficient T – statistics Coefficient T – statistics Constant 10.431*** 231.554 10.947*** 140.912 10.961*** 141.590 Ln Size –0.093*** –19.351 –0.112*** –24.940 –0.115*** –25.243 Ln Building’s age 0.017*** 3.326 0.011*** 5.166 0.012*** 5.332 Ln Floor/Level No. 0.010*** 4.000 –0.073*** –15.675 –0.074*** –15.837 Ln. Total floors –0.033*** –7.127 –0.047*** –8.464 –0.046*** –8.473 Ln Distance to CBD –0.235*** –17.770 –0.338*** –11.650 –0.342*** –11.665 Ln Distance to West Intime –0.109*** –20.037 / / / / Ln Distance to Xicheng Square –0.048*** –9.958 –0.039*** –4.025 –0.039*** –3.947 Bus route –0.002*** –3.998 –0.004*** –7.507 –0.003*** –7.344 External environment 0.048*** 11.484 0.070*** 15.541 0.069*** 15.134 Inner environment –0.022*** –6.345 –0.031*** –7.815 –0.029*** –7.247 Living facilities –0.007*** –3.089 –0.031*** –12.806 –0.031*** –12.824 Sports facilities 0.006** 2.587 0.022*** 9.695 0.022*** 9.451 Property management 0.084*** 19.028 0.071*** 16.670 0.072*** 16.702 Quality of primary school 0.084*** 34.341 0.063*** 21.108 0.063*** 20.962 Quality of junior high school 0.041*** 5.725 0.005 0.386 0.005 0.340 Distance to primary school –0.032*** –9.596 –0.028*** –7.140 –0.028*** –7.027 Distance to junior high school 0.033*** 11.329 –0.047*** –10.440 –0.046*** –10.061 Y2 –0.071*** –12.109 / / –0.059*** –10.879 Y3 –0.007 –1.189 / / 0.004 0.715 Y4 –0.055*** –8.861 / / –0.057*** –8.248 Y5 –0.059*** –9.797 / / –0.054*** –7.991 Adjusted R2 0.642 0.701 0.696 Residuals regressed on location variables x2 –0.002*** –2.111 x*y  –0.001   –1.515 x2 0.000   –1.085 x*y  0.000   –1.272 x2 0.000   –0.952 x*y  0.000   –1.147 Notes: ***, **, and * represent significance at the 1%, 5% and 10% levels, respectively. The dependent variable is lnP. To test for the spatial independence of the estimated error terms, the residuals were regressed on the location variables, x, y coordinate and x2, y2, x*y. ===== PAGE 9 (score 65, numeric-paren 39, table-words 7, var-words 7) ===== International Journal of Strategic Property Management, 2019, 23(1): 65–80 73 –0.033, which is significant and negative. This result is consistent with the reality that sub-high-rise buildings with lower heights are more expensive due to less shar- ing of public areas and to more comfort. The coefficient of Distance to CBD is –0.235, which means that hous- ing prices will decrease by 0.235% if the distance from a house to the city center increases by 1%, indicating that Wulin Square definitely has much influence on the housing market. The quality of primary and junior high schools nearby can significantly increase housing prices and Distance to primary school has significant negative effect. Contrary to expectations, the coefficient of Bus route is negative, this phenomenon may be attributed to the fact that increasing the number of bus routes may produce negative effects, such as noise and environmen- tal pollution (Wen, Xiao, & Zhang, 2017). As for Neigh- borhood Characteristics, neither Inner environment nor Living facilities is as expected, which may be because the sample is concentrated in a limited area. The variables measured in the same way did not perform well either in housing submarket around the Grand Canal (Wen et al., 2017), but they were positively significant in house price regression with the sample of entire city (Wen et  al., 2014a). We should improve the measure of these vari- ables in future studies. The distances to the shopping malls are the main re- search objects. West Intime and Hangzhou Incity are too close to each other, so there would be collinear problems if the two distance variables were both introduced into the model. Considering the smaller influence of Hang- zhou Incity, Distance to Hangzhou Incity is not introduced into the model. The coefficients of Distance to West Intime and Distance to Xicheng Square are both significant and negative as were expected, but Distance to West Intime is a little larger with a value of –0.109. In the case of ex- cluding the impacts of other characteristic variables, as the road distance to West Intime increases 1%, housing prices decrease by 0.109%. The result indicates that for the study area, West Intime has a significant impact on hous- ing prices and is more influential than Xicheng Square. In this model, the influence of West Intime on housing prices is characterized initially and roughly. The following two models provide further analysis. 4.2. The results of improved hedonic price modeling approaches Similarly, the least-square method was used to estimate Model 2. The adjusted R2 increases to 0.701, indicating that compared to Model 1, this model is more explanatory. Almost all the characteristic variables are significant at the 1% level. The coefficient of Distance to Xicheng Square is also significant and similar to the result of Model 1. By controlling the influence of Xicheng Square, the impact of West Intime on housing prices can be captured by spatio- temporal interactions, whose coefficients are shown in Table 6. There are 44 spatio-temporal interaction coeffi- cients, of which 39 are significant. In the space dimension, the housing price indices of the five zones in the southwestern region (Zones 1–5) are relatively higher than the four zones in the northwestern region (Zones 6–9). With the purpose of observing the influence of West Intime on the different zones in more detail, we further calculated the price gradients among the various zones and observed the changes. This helped to analyze the temporal and spatial influences of West Intime. Zone 1 is the farthest south, and so, is farthest away from West Intime and closest to the city center. The impact received from West Intime is relatively minimal. Therefore, Zone 1 was set as the reference area and the price indices of the other zones were subtracted from the price gradient of each zone. For each zone, the price gradient of Zone i (i = 2, …, 9) is equal to the difference between the price indices of Zone i and Zone 1. The results are shown in Figures 2 and 3. Figure 2 shows the price gradients for Zones 2–5 based on Zone 1. The results of 2011–2013 represent the price gradients of Period One while the results of 2014–2015 rep- resent Period Two. The results are summarized as follows: 1) In Period One, the price gradients of the four zones were basically stabilized with few changes. Even in 2012, which was the peak of the construction pe- riod, the price gradient of Zone 5 was reduced by 0.027. The noise generated by the project construc- tion might have had a negative impact on housing prices (Henneberry, 1998). Table 6. The value of the coefficient ,i j σ (Model 2) Z1 Z2 Z3 Z4 Z5 Z6 Z7 Z8 Z9 Y1 / 0.042*** 0.121*** 0.167*** 0.121*** –0.108*** –0.039 0.105*** –0.082*** Y2 –0.043*** 0.002 0.069*** 0.126*** 0.051** –0.209*** –0.197*** 0.025 –0.148*** Y3 0.009 0.078*** 0.125*** 0.200*** 0.126*** –0.132*** –0.103*** 0.068** –0.063** Y4 –0.048*** 0.027** 0.099*** 0.164*** 0.145*** –0.141*** –0.131*** –0.058** –0.070** Y5 –0.071*** 0.006 0.119*** 0.159*** 0.217*** –0.083*** –0.127*** –0.071*** –0.069** Note: ***, **, and* represent significance at the 1%, 5% and 10% levels, respectively. ===== PAGE 10 (score 12, numeric-paren 0, table-words 2, var-words 7) ===== 74 L. Zhang et al. The effects of a shopping mall on housing prices: a case study in Hangzhou 2) In Period Two, the price growth rate of Zone 5 was very significant. Zones 3 and 4, which were slightly farther away, were also affected by the positive effect of the shopping mall, leading to the increase in the price gradients based on Zone 1. The price gradi- ent of Zone 2 increased slightly compared with the price gradients of the other zones. The results show that a significant effect on the residential prices in the south area after the open of the mall and the “value-added” range decreased with the distance. Zones 6–9 are located away from the city center and West Intime is the only large-scale commercial facility in the surrounding area. So, theoretically, these four zones are better able to reflect the spatio-temporal impact of West In- time on the surrounding housing prices. Figure 3 shows the price gradients of the four zones based on Zone 1: 1) In Period One, the price gradients of the four zones fluctuated less. Zones 6 and 7 had a certain degree of price decline in 2012 and rebounding in 2013, largely due to the impact of the neighboring construction. 2) In Period Two, the price gradient of Zone 6 in- creased significantly and the value was near zero in 2015, showing that the price of Zone 6 had begun to catch up with Zone 1. Similarly to Zone 6, the price gradient of Zone 7 also showed a clear upward trend with a smaller rise. The changes in the price gradient of Zone 8 was confusing while there were still some visible positive changes in Zone 9. In this model, the value-added effect can be clearly seen, especially in Zones 6–9, and the influence decreases from Zone 5 to the other zones on both sides, but this method of price gradient calculation still could not elimi- nate the time fixed effect completely. 4.3. The results of extended price gradient approaches The explanatory power of Model 3 is 69.6%, which is very close to that of Model 2. The confidence levels of the char- acteristic variables are almost the same as before. Only one spatial-temporal interaction coefficient θ was not signifi- cant. Therefore, the model can explain the existing data well. Then, the price gradient was calculated by λi + θi,p. The results are shown in Table 7. 0.000 0.050 0.100 0.150 0.200 0.250 0.300 0.350 2 0 1 1 2 0 1 2 2 0 1 3 2 0 1 4 2 0 1 5 σi,j-σ1,j Year Z2-Z1 Z3-Z1 Z4-Z1 Z5-Z1 Figure 2. Price gradients of Zones 2–5 based on Zone 1 –0.200 –0.100 0.000 0.100 0.20 0 0.300 0.400 2 0 1 1 2 0 1 2 2 0 1 3 2 0 1 4 2 0 1 5 σi,j-σ1,j Year Z5-Z1 Z6-Z1 Z7-Z1 Z8-Z1 Z9-Z1 Figure 3. Price gradients of Zones 6–9 based on Zone 1 Note: As a reference, the price gradient for Zone 5 is also shown. ===== PAGE 11 (score 28, numeric-paren 15, table-words 1, var-words 6) ===== International Journal of Strategic Property Management, 2019, 23(1): 65–80 75 Figure 4 exhibits the price gradients of nine zones in two periods. Zone 1 is still set as the benchmark. In the premise of eliminating the time fixed effect and assuming that the housing prices in Zone 1 had not changed during the construction and operating periods, the price gradi- ents of the other zones were obtained based on Zone 1. Compared to the previous two methods, this treatment was better able to remove the influence of other factors on the time dimension. The findings are as follows: 1) The price gradients of Period 2 are shown to be sig- nificantly higher than those of Period 1 (except for Zone 8). In the area south of West Intime (from Z2–Z5), the price gradient differences between the two periods are 0.006, 0.043, 0.031, and 0.127, re- spectively. Z3 and Z4 almost have the same value added effect caused by West Intime while Z2 has the smallest appreciation. Zone 5 has become the new peak in housing prices and the price increases 13.5% in Period 2. The value-added effect can be found to decrease with distance, but it is not a com- pletely linear relationship. 2) In the north area of West Intime (from Z6–Z9), the price gradient differences between the two periods are 0.090, 0.029, –0.093, and 0.057, respectively. Generally the price gradient difference between the two periods in each zone is negatively correlated with the distance from West Intime. The price gra- dient difference of Zone 6 is the largest while Zone 7 is the smallest except the negative appreciation in Zone 8. The time fixed effect has largely been elimi- nated, so the price gradient difference is caused by the influence of West Intime. The above three models can be good proof that West Intime has had a positive impact on the surrounding hous- ing prices. Model 1 identifies the impact of West Intime on housing prices as being inversely proportional to the distance. The two price gradient models allow insight into the spatial price trend of West Intime’s surrounding area. We found that the impacts on the surrounding housing prices in Zones 5 and 6 were the largest. In contrast, the impacts on Zones 2 and 8 were relatively weak. In terms of time, examining the housing prices in the two periods shows that the increase in housing prices during the oper- ating period was very large while during the construction period, housing prices changed little and even appeared to have experienced a certain degree of negative effects. Conclusions Using the price gradient method and hedonic price the- ory, this study analyzed the influence of a shopping mall on the surrounding housing prices in terms of time and space. First, the results of the hedonic model show that West Intime has a significant impact on housing prices, which decays with distance. Then, in the price gradient method, location and time dummy variables were intro- duced to capture changes in housing prices at different times in various locations. The results indicate that almost all the zones in the study area have had their prices influ- enced to some extent if there was a shopping mall near- by. Especially in the zones closely surrounding the mall, housing prices had obviously increased by more than 10 percent upon the mall’s opening. The spatial and tempo- ral interaction terms used in Model 2 and Model 3 can Table 7. The value of λi + θi,p (Model 3) Z1 Z2 Z3 Z4 Z5 Z6 Z7 Z8 Z9 li / 0.063*** 0.120*** 0.183*** 0.112*** –0.137*** –0.105*** 0.087*** –0.078*** θi,p / 0.006 0.043*** 0.031*** 0.127*** 0.090*** 0.029*** –0.093*** 0.057*** T1(λi + θi,p) 0.000 0.063 0.120 0.183 0.112 –0.137 –0.105 0.087 –0.078 T2(λi + θi,p) 0.000 0.068 0.164 0.214 0.239 –0.047 –0.076 –0.005 –0.022 Note: ***, **, and* represent significance at the 1%, 5% and 10% levels, respectively. –0.20 –0.15 –0.10 –0.05 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Z1 Z2 Z3 Z4 Z5 Z6 Z7 Z8 Z9 λi+ θi,p T1 T2 Figure 4. Price gradients of Zones 1–9 ===== PAGE 12 (score 14, numeric-paren 0, table-words 0, var-words 13) ===== 76 L. Zhang et al. The effects of a shopping mall on housing prices: a case study in Hangzhou effectively separate the impact of the shopping mall on the surrounding house prices during the construction and operation phases. Through the block division based on the main roads, we can clearly observe the reaction of residen- tial groups at different locations relative to the shopping mall, which is rarely discussed in the previous literature. To the best of our knowledge, this paper is one of the few studies concerning the impacts of a shopping mall at different periods of time. Before the completion of the construction, the expected effect of West Intime is very small. The surrounding housing prices were subject to minimal influence, which is different from the expectation effect found in some subway or transportation studies. The reason may be that the influence of a shopping mall large- ly depends on its own anchor tenants and operational ca- pacity. Before the completion of the construction of West Intime, people knew little about the mall, because it had not been promoted much. Therefore, there was no obvious influence on the housing price. Nevertheless, during the construction of the mall, the transaction of second-hand houses and the construction of new houses in the affect- ed area increased significantly, showing the signal effect of the shopping mall. Upon completion in Period 2, the mall’s good operating conditions and reputation greatly enhanced its radiative influence in the area. This paper provides some implications for city plan- ners and governments. The model has evaluated the range of a mall’s influence on housing prices and offers a good reference for avoiding the imbalanced distribution of shopping malls in different districts and the scarcity of retail shops. Meanwhile, the effects of shopping malls on housing prices can be taken into consideration for land use planning by the government. Due to the vital influence of shopping malls on the surrounding housing prices, the introduction of malls and commercial facili- ties is a good practice to leading the development of a new city area. The limitations of this study include the short span of the data. We cannot study the impact of the shopping mall on housing prices before construction. Therefore, it is impossible to explore if the shopping mall had had an effect before its construction. The data only covers the two years after the opening of the mall, so we cannot observe changes in housing prices for any longer term. Neverthe- less, due to the brand effect and good management, the mall performed well and entered a stable operating period during the first two years. Further research is needed to explore how the impact pattern differs at different scales and operational levels of shopping malls, and to observe the interaction effect of two or more shopping malls on housing market. Acknowledgements The authors would like to thank the editor and the anon- ymous reviewers for the positive and constructive com- ments and suggestions. Funding This work was supported by the China Scholarship Coun- cil under Grant No. 201706325013; and Zhejiang Provin- cial Natural Science Foundation of China under Grant No. LY18G030002. References Ahlfeldt, G. M., & Wendland, N. (2011). Fifty years of urban accessibility: the impact of the urban railway network on the land gradient in Berlin 1890–1936. Regional Science and Ur- ban Economics, 41(2), 77-88. https://doi.org/10.1016/j.regsciurbeco.2010.10.001 Alonso, W. (1964). Location and land use: toward a general theory of land rent. Cambridge: Harvard University Press. https://doi.org/10.4159/harvard.9780674730854 Berry, B. J., & Garrison, W. L. (1958). Recent developments of central place theory. Papers in Regional Science, 4(1), 107-120. https://doi.org/10.1111/j.1435-5597.1958.tb01625.x Brueckner, J. K. (1993). Inter-store externalities and space alloca- tion in shopping centers. The Journal of Real Estate Finance and Economics, 7(1), 5-16. https://doi.org/10.1007/BF01096932 Carter, C. (2009). Review articles: what we know about shopping centers. Journal of Real Estate Literature, 17(2), 163-180. Carter, C. C., & Haloupek, W. J. (2002). Dispersion of stores of the same type in shopping malls: theory and preliminary evi- dence. Journal of Property Research, 19(4), 291-311. https://doi.org/10.1080/0959991022000013550 Chau, K. W., & Chin, T. L. (2003). A critical review of literature on the hedonic price model. International Journal for Housing Science and Its Applications, 27(2), 145-165. Chau, K. W., & Ng, F. F. (1998). The effects of improvement in public transportation capacity on residential price gradient in Hong Kong. Journal of Property Valuation and Investment, 16(4), 397-410. https://doi.org/10.1108/14635789810228204 Christaller, W. (1933). Die zentralen Orte in Süddeutschland: eine ökonomisch-geographische Untersuchung über die Gesetzmäs- sigkeit der Verbreitung und Entwicklung der Siedlungen mit städtischen Funktionen. University Microfilms. Clark, D. E., & Herrin, W. E. (2000). The impact of public school attributes on home sale prices in California.  Growth and Change, 31(3), 385-407. https://doi.org/10.1111/0017-4815.00134 Clark, W. A. (1968). Consumer travel patterns and the concept of range 1. Annals of the Association of American Geogra- phers, 58(2), 386-396. https://doi.org/10.1111/j.1467-8306.1968.tb00651.x Colwell, P. F., Gujral, S. S., & Coley, C. (1985). The impact of a shopping center on the value of surrounding properties. Real Estate Issues, 10(1), 35-39. Coulson, N. E. (1991). Really useful tests of the monocentric model. Land Economics, 67(3), 299-307. https://doi.org/10.2307/3146425 De Palma, A., Ginsburgh, V., Papageorgiou, Y. Y., & Thisse, J. F. (1985). The principle of minimum differentiation holds under sufficient heterogeneity. Econometrica: Journal of the Econo- metric Society, 767-781. https://doi.org/10.2307/1912653 Des Rosiers, F., Lagana, A., Thériault, M., & Beaudoin, M. (1996). Shopping centres and house values: an empirical investiga- tion. Journal of Property Valuation and Investment, 14(4), 41-62. https://doi.org/10.1108/14635789610153461 ===== PAGE 14 (score 86, numeric-paren 48, table-words 8, var-words 10) ===== 78 L. Zhang et al. The effects of a shopping mall on housing prices: a case study in Hangzhou Wen, H., Zhang, Y., & Zhang, L. (2014b). Do educational facili- ties affect housing price? An empirical study in Hangzhou, China. Habitat International, 42, 155-163. https://doi.org/10.1016/j.habitatint.2013.12.004 Yiu, C., & Tam, C. (2004). A review of recent empirical stud- ies on property price gradients. Journal of Real Estate Litera- ture, 12(3), 305-322. Appendix 1 With the purpose of testing the robustness of the em- pirical results, this study used the traditional partition- ing method with distance rings to verify whether the re- sults are consistent. The study area was divided into eight zones (0−1 km, 1−1.5 km, 1.5−2 km, 2−2.5 km, 2.5−3 km, 3−3.5 km, 3.5−4 km, and 4−5 km) according to their dis- tance to West Intime, as shown in Figure A1. Owing to the lack of communities, Zones 1 and 8 span 1 km while other zones span 0.5 km. In the east, we set a boundary line along Xueyuan Road, which is located in the middle line of West Intime and the CBD. The area to the east is closer to the city center. The three models were applied again. The results are shown in Table A1. The three models all behave well and the adjusted R2 are above 0.55. Almost all independent Yiu, C. Y., & Wong, S. K. (2005). The effects of expected trans- port improvements on housing prices. Urban Studies, 42(1), 113-125. https://doi.org/10.1080/0042098042000309720 Yu, T. H., Cho, S. H., & Kim, S. G. (2012). Assessing the resi- dential property tax revenue impact of a shopping center. The Journal of Real Estate Finance and Economics, 45(3), 604-621. https://doi.org/10.1007/s11146-010-9292-x Figure A1. Partition method of distance ring Table A1. Results of regression of the three models Independent variables Model 1 Model 2 Model 3 Coefficient T – statistics Coefficient Coefficient T – statistics Coefficient Constant 10.718*** 270.712 11.123*** 233.887 11.177*** 244.944 Ln Size –0.090*** –21.711 –0.085*** –20.923 –0.085*** –20.976 Ln Floor/Level No. –0.033*** –8.278 –0.036*** –9.047 –0.036*** –9.120 Ln Total floors 0.009*** 3.961 0.009*** 4.567 0.010*** 4.599 Ln Building’s age –0.001 –1.162 –0.003*** –3.579 –0.004*** –4.426 Ln Distance to CBD –0.431*** –39.491 –0.538*** –46.694 –0.542*** –46.896 Ln Distance to West Intime –0.053*** –15.442 Ln Distance to Xicheng Square –0.016*** –4.077 –0.036*** –7.412 –0.038*** –7.848 Bus route –0.003*** –9.935 –0.004*** –14.169 –0.004*** –14.187 Surrounding environment 0.090*** 25.051 0.051*** 12.635 0.051*** 12.472 Inner environment –0.019*** –6.597 –0.003 –1.110 –0.003 –1.160 Living facilities –0.006** –2.113 0.004 1.347 0.003 1.120 Sports facilities 0.035*** 17.221 0.031*** 15.402 0.030*** 15.225 Property management 0.063*** 22.996 0.061*** 22.294 0.062*** 22.615 Quality of primary school 0.054*** 25.602 0.059*** 25.861 0.059*** 25.830 Quality of junior high school 0.003 0.513 –0.009 –1.262 –0.014 –1.872 Ln Distance to primary school –0.005** –2.339 0.009*** 4.219 0.009*** 4.206 Ln Distance to junior high school 0.038*** 16.232 –0.002 –0.812 –0.002 –0.738 Y2 –0.063*** –12.570 / / –0.062*** –13.046 Y3 0.008 1.579 / / 0.011 2.386 Y4 –0.039*** –7.633 / / –0.023*** –3.310 Y5 –0.048*** –10.295 / / –0.031*** –4.675 Adjusted R2 0.557 0.596 0.592 Note: ***, **, and* represent significance at the 1%, 5% and 10% levels, respectively. ===== PAGE 15 (score 116, numeric-paren 94, table-words 4, var-words 3) ===== International Journal of Strategic Property Management, 2019, 23(1): 65–80 79 variables are significant and the values of the coefficients are hardly different from the original partitioning meth- od. In Model 1, the coefficient of Distance to West Intime is – 0.053, which shows the negative price gradient and validates the influence of West Intime. In Model 2, house prices in Zones 1 and 2 increased after 2014, while those in other zones declined over two years, excepting for Zone 82. In Model 3, as shown in Figure A2, there was a clear rise in Zones 1 and 2 during Period 2, as prices in the other regions fell. The results of the three models support the robustness of the findings in the text. 2 The high house price in Zone 8 may be because most of the samples are concentrated in the southeast of the ring, which is closer to the city center. The uneven distribution of samples and the difference in house prices in all directions may affect the accuracy of the results. This is the main reason why we have not chosen this division method in the text. Appendix 2 Using the location coordinates and the shape of the com- munities, as well as the number of samples therein, we randomly generated the coordinates for each house unit by GIS. Spatial weights could then be established in order to conduct a spatial test and spatial regression. The re- gression results show that the spatial econometric model SEM performs better (see Table A2 and A3). However, the pattern of price gradients of Zones 1−9 with Model 2 and Model 3 are very close to the results in the text. Due to limited space, the results are not shown here and can be provided upon request. –0.25 –0.20 –0.15 –0.10 –0.05 0.00 0.05 Z1 Z2 Z3 Z4 Z5 Z6 Z7 Z8 λi + θi,p T1 T2 Figure A2. Price gradients of Zones 1–8 Table A2. Results of regression of different models Coefficient Model 1 Model 2 Model 3 Linear Linear – log Log – linear SAR SEM SEM SEM Constant 22205.649*** 30108.698*** 9.995*** 2.972*** 10.566*** 10.837*** 10.806*** Size –22.814*** –1740.935*** 0.001*** –0.096*** –0.122*** –0.122*** –0.124*** Floor/Level No. 20.424** 190.989** –0.001*** 0.011*** 0.012*** 0.012*** 0.013*** Total floors –32.672*** –652.257*** –0.002*** –0.052*** –0.102*** –0.099*** –0.100*** Building’s age –30.326*** 235.921** –0.001 –0.013*** 0.028*** –0.036*** –0.030*** Distance to CBD –789.131*** –4095.293*** –0.046*** –0.046*** –0.219*** –0.291*** –0.282*** Distance to West Intime –305.480*** –1893.405*** –0.018*** –0.037*** –0.136*** / / Distance to Xicheng Square –327.487*** –959.592*** –0.015*** –0.008* –0.030* –0.019 –0.013 External environment 718.103*** 730.509*** 0.047*** 0.015*** 0.068*** 0.076*** 0.073*** Inner environment –284.909*** –366.970*** –0.016*** –0.002 –0.005 –0.018* –0.014 Living facilities –181.152*** –92.520* –0.013*** –0.003 –0.004 –0.025*** –0.026*** Sports facilities 300.827*** 131.780*** 0.014*** 0.004* 0.008 0.024*** 0.023*** Property management 1170.150*** 1656.332*** 0.061*** 0.030*** 0.081*** 0.060*** 0.061*** Bus route –14.807* –18.432** –0.001*** –0.001*** –0.005*** –0.005*** –0.005*** Distance to primary school –763.224*** –548.397*** –0.045*** –0.014*** –0.026** –0.019** –0.017* ===== PAGE 16 (score 73, numeric-paren 54, table-words 4, var-words 3) ===== 80 L. Zhang et al. The effects of a shopping mall on housing prices: a case study in Hangzhou Coefficient Model 1 Model 2 Model 3 Linear Linear – log Log – linear SAR SEM SEM SEM Distance to junior high school 835.816*** 613.014*** 0.044*** 0.008*** 0.023** –0.047*** –0.046*** Quality of primary school 1534.241*** 1790.623*** 0.071*** 0.026*** 0.093*** 0.067*** 0.067*** Quality of junior high school 139.638 186.550 0.038*** 0.008 0.056** 0.037 0.038 Y2 –1277.337*** –1370.727*** –0.067*** –0.068*** –0.074*** / –0.064*** Y3 60.759 –84.845 0.000 –0.002 –0.015*** / –0.004 Y4 –825.525*** –1015.551*** –0.046*** –0.046*** –0.061*** / –0.066*** Y5 –745.043*** –984.428*** –0.048*** –0.047*** –0.067*** / –0.066*** W_lnprice / / / 0.755*** / / / LAMBDA / / / / 0.803*** 0.685*** 0.681*** Adjusted R2 0.591 0.587 0.644 0.709 0.722 0.734 0.728 Note: ***, **, and * represent significance at the 1%, 5% and 10% levels, respectively. Table A3. Diagnose test for OLS regression of Model 1−3 Model 1 Model 2 Model 3 Moran᾽s I (resudals) 0.223 (0.000) 0.118 (0.000) 0.114 (0.000) Lagrange Multiplier (lag) 5855.282 (0.000) 1512.888 (0.000) 1434.977 (0.000) Robust LM (lag) 169.693 (0.000) 3.189 (0.074) 2.301 (0.129) Lagrange Multiplier (error) 8528.628 (0.000) 2386.997 (0.000) 2232.014 (0.000) Robust LM (error) 2843.040 (0.000) 877.298 (0.000) 799.337 (0.000) Note: the number in the parentheses are p value. End of Table A2