1# RECORD 58a9ab2513035056fdc6d56cafc04ae9d61a09552# TITLE: Willing to pay more for high-quality schools?3# YEAR: 2018.0 VENUE: nan DOI: 10.14246/irspsd.6.1_454# AUTHORS: Lin-Chuan Yang; Bo Wang; Yunyi Zhang; Zi-Wei Ye; Yu-Zhuo Wang; Peng-Fei Li5# PAGES: 18 CANDIDATE TABLE PAGES (1-based): [9, 10, 12]6# ABSTRACT: : Principally, the enrolment of elementary school in China is solely based on residential location. Due to the scarcity of prestigious schools, the household registration (or hukou in Chinese) and the territorial-based school admission policy, a feasible approach for parents to provide the children with good education is to purchase the house in the attendance zone of a high-quality school (or xuequfang in Chinese). The supply-demand imbalance gives rise to the xuequfang phenomenon (much higher prices of xuequfang relative to non xuequfang) . Based on 1250 housing samples in 286 multi- or high-storey residential districts, this paper firstly develops two basic and four Box-Cox transformed hedonic price models to estimate the effect of high-quality schools on residential property prices. The consistency of six models greatly enhances the credibility of this study. Moreover, complementary, the propensity score matching technique is used to estimate the treatment effect. The two methods consistently suggest that residential property values are affected by top-tier schools. They reveal that xuequfang exhibit values that are between 9.3% and 12.1% higher than non- xuequfang , ceteris paribus . The negative influences of the xuequfang phenomenon and several countermeasures (gradually reforming household registration system , optimizing resource distribution to balance the quality of education, highlighting family education and children’s all-round development) are discussed.789===== PAGE 1 (score 5, numeric-paren 0, table-words 0, var-words 5) =====101145 12International review for spatial planning and sustainable development, Vol.6 No.1 (2018), 45-62 13ISSN: 2187-3666 (online) 14DOI: http://dx.doi.org/10.14246/irspsd.6.1_45 15Copyright@SPSD Press from 2010, SPSD Press, Kanazawa 16 17Willing to pay more for high-quality schools? 18A hedonic pricing and propensity score matching approach 19Linchuan Yang¹, Bo Wang2*, Yunyi Zhang3, Ziwei Ye4, Yuzhuo Wang5, 20Pengfei Li6 21 221 Faculty of Architecture, the University of Hong Kong 232 Department of Geography, the University of Hong Kong 243 Shanghai Digital Intelligence System Technology Co.Ltd 254 Taubman College of Architecture and Urban Planning, University of Michigan 265 China Academy of Urban Planning and Design 276 School of Architecture and Civil Engineering, Xiamen University, Xiamen 28* Corresponding Author, Email: wangbo.nju@gmail.com 29Received: June 11, 2016; Accepted: June 23, 2017 30 31Keywords: Box-Cox transformation; Hedonic price model, School attendance zone, 32Property price, Propensity score matching, Xuequfang 33Abstract: 34Principally, the enrolment of elementary school in China is solely based on 35residential location. Due to the scarcity of prestigious schools, the household 36registration (or hukou in Chinese) and the territorial-based school admission 37policy, a feasible approach for parents to provide the children with good 38education is to purchase the house in the attendance zone of a high-quality 39school (or xuequfang in Chinese). The supply-demand imbalance gives rise to 40the xuequfang phenomenon (much higher prices of xuequfang relative to non-41xuequfang). Based on 1250 housing samples in 286 multi- or high-storey 42residential districts, this paper firstly develops two basic and four Box-Cox 43transformed hedonic price models to estimate the effect of high-quality 44schools on residential property prices. The consistency of six models greatly 45enhances the credibility of this study. Moreover, complementary, the 46propensity score matching technique is used to estimate the treatment effect. 47The two methods consistently suggest that residential property values are 48affected by top-tier schools. They reveal that xuequfang exhibit values that are 49between 9.3% and 12.1% higher than non-xuequfang, ceteris paribus. The 50negative 51influences 52of 53the 54xuequfang 55phenomenon 56and 57several 58countermeasures (gradually reforming household registration system, 59optimizing resource distribution to balance the quality of education, 60highlighting family education and children’s all-round development) are 61discussed. 621. 63INTRODUCTION 64Education, work and retirement, are three stages of our life. Education is 65the foundation of a nation and is closely related to the overall quality of 66citizens as well as the long-term development of a nation. The importance of 67education to individuals and a state has been universally recognized. 68“Residential property is a multidimensional commodity, characterized by 69durability and structural inflexibility as well as spatial fixity” ((So, Tse, & 70Ganesan, 1997), p.40). It is distinguished by a number of characteristics 717273===== PAGE 9 (score 19, numeric-paren 0, table-words 2, var-words 13) =====7475Yang et al. 765377 78Until 2015, there were eight provincial-level demonstration elementary 79schools in Xiamen Island (Fig.1). They are Yanwu Elementary School, 80Minli Elementary School, Experimental Elementary School, Second 81Experimental Elementary School, Datong Elementary School, Affiliated 82elementary school of Foreign Language School, Bindong Elementary School 83and Binlang Elementary School. It is noteworthy that all of them are situated 84in the west of Siming District, which reveals the apparent spatial variations 85and imbalance. 863.3 87Variables 88Table 1 shows the definitions and descriptive statistics of twelve 89independent variables. Except for School, all other eleven independent 90variables are theoretically important variables that need to be statistically 91controlled for by being included in our regression models. It should be noted 92that a variable widely used in the western literature, number of bedrooms, 93was not included in our regression models, in the interest of model 94parsimony, because its effect would overlap with floor area, resulting in the 95presence of multi-collinearity (So, Tse, & Ganesan, 1997). Moreover, job 96accessibility is not introduced due to the unavailability of data. 97Table 1. Variable Definitions and Descriptive Statistics 98Variable 99Description 100Expected 101sign1 102Mean 103SD 104Area 105Gross floor area (m2) 106+ 107105.69 43.02 108Age 1092015 minus the year of occupation permit 110(number of years) 111- 11212.13 1134.81 114Local 115environment 116Environment quality of residential 117complexes the property, rated on a 5-point 118Likert item from “very bad” (1) to “very 119good” (5) 120+ 1213.08 1220.93 123School 124Dummy variable, 1 if the property is 125xuequfang, 0 otherwise 126+ 1270.13 1280.33 129Kindergarten 130Dummy variable, 1 if there are at least one 131kindergarten within 500 meters, 0 otherwise 132+ 1330.83 1340.37 135Middle school 136Dummy variable, 1 if there are at least one 137middle school within 1000 meters, 0 138otherwise 139+ 1400.82 1410.38 142Xiamen 143University 144Dummy variable, 1 if Xiamen University is 145within 1000 meters, 0 otherwise 146+ 1470.02 1480.13 149BRT 150Dummy variable, 1 if there are at least one 151BRT (Bus Rapid Transit) station within 800 152meters, 0 otherwise 153+ 1540.23 1550.42 156Distance to 157hospital 158Euclidean distance to the nearest Class 2A or 1593A hospital (km) 160? 1610.81 1620.77 163Distance to 164cultural/sports 165center 166Euclidean distance to the nearest 167cultural/sports center (km) 168- 1691.11 1700.66 171Distance to 172city center 173Euclidean distance to Zhongshan Road, the 174city center of Xiamen Island (km) 175- 1766.84 1772.57 178Distance to 179lake 180Euclidean distance to the nearest lake (km) 181- 1821.08 1830.68 184Note: a + and − represent increasing and decreasing impacts on housing prices respectively; ? 185indicates a priori undetermined sign. 186187188===== PAGE 10 (score 30, numeric-paren 1, table-words 8, var-words 7) =====18919054 191IRSPSD International, Vol.6 No.1 (2018), 45-62192 1934. 194RESULTS 1954.1 196Initial hedonic price regression 197EVIEWS 9 is used to build a semi-log and a double-log model using the 198ordinary least squares (OLS) method. The dependent variable is modified by 199a natural log transformation in both models whereas independent variables 200are modified by a natural log transformation only in the double-log model. 201We do not transform five dummy variables as they are not definitely greater 202than 0. Table 2 provides the modeling results. The goodness-of-fit values of 203both models (0.801, 0.850) are higher than that based on the same data 204source in Yang et al. (2015) (0.789). This indicates that the two models were 205able to predict higher proportions of variability in their outcomes compared 206to the model in Yang et al. (2015). Moreover, with a higher model fit, the 207double-log model outperforms the semi-log model, which indicates that it 208captures more variations in housing prices. All but two variables are 209significant at the 5% level in both models. Kindergarten and Middle school 210are not significant at the 5% level, which implies that they are not associated 211with housing prices. 212Table 2. Regression results of semi-log and double-log models 213Variable 214Semi-log model 215Double-log model 216coefficient 217t-value 218coefficient 219t-value 220Constant 2214.579a 22288.67 2231.091a 22413.89 225Area 2260.010a 22759.16 2281.008a 22970.70 230Age 231-0.011a 232-7.65 233-0.153a 234-11.26 235Local environment 2360.068a 2378.36 2380.177a 2397.94 240School 2410.089a 2424.00 2430.108a 2445.45 245Kindergarten 2460.009c 2470.47 248-0.011c 249-0.68 250Middle school 251-0.001c 252-0.06 2530.032c 2541.91 255Xiamen University 2560.206a 2573.45 2580.144a 2593.04 260BRT 2610.077a 2624.72 2630.100a 2646.85 265Distance to hospital 2660.047a 2674.41 2680.018b 2692.13 270Distance to cultural/sports center 271-0.057a 272-5.04 273-0.049a 274-5.25 275Distance to city center 276-0.021a 277-6.04 278-0.077a 279-6.36 280Distance to lake 281-0.072a 282-6.06 283-0.051a 284-5.95 285Performance statistics 286Adjusted R2 2870.801 2880.850 289Log-likelihood 29066.00 291238.69 292AIC 293-0.09 294-0.37 295Note: a significant at the 1% level. b significant at the 5% level. c not significant at the 5% 296level. 297With no exception, the signs of all variables’ coefficients agree with a 298priori expectation. The coefficients associated with Area, Local 299environment, School, Xiamen University, BRT are positive. This implies 300that properties with larger size, having better local environment and BRT 301access, within attendance zone of a high-quality school and closer to Xiamen 302University, exhibit higher values. The coefficients of Age, Distance to 303cultural/sports center, Distance to city center, Distance to lake, have the 304305306===== PAGE 11 (score 19, numeric-paren 0, table-words 4, var-words 11) =====307308Yang et al. 30955310 311hypothesized negative signs. It indicates that houses at the younger age, with 312better accessibility to the city center, culture/sports center and lake, are more 313expensive, all else being fixed. 314The coefficient associated with Distance to hospital is positive, which 315indicates that proximity to Class 2A and 3A hospitals negatively affects 316housing prices. This indicates the nuisance effect of being located near 317hospitals, which seems somewhat surprising, but consistent with the findings 318in Li et al. (2013) and Peng, B. et al. (2015). There are two possible 319explanations. Firstly, different age cohorts have differing needs and 320preferences of services, for example health care services (Wong et al., 2018). 321Normally, young adults seldom go to hospital, so they may not consider 322proximity to hospitals when making residential location decisions, even tend 323to live far away from hospitals. Notwithstanding, the elderly are in general 324more likely to go to hospital (Yang, 2016; Szeto et al., 2017; Wong et al., 3252017) and may tend to live close to hospitals for better access due to higher 326visiting frequency (Li et al., 2013). In addition, it is crowded around high-327level comprehensive hospitals generally. There are a huge number of people 328and vehicles within these regions, generating noises and air pollution, 329thereby reducing nearby residents’ quality of life (Peng, B. et al., 2015). 330The coefficients of the double-log regression function correspond to 331average attribute elasticities. Gross floor area is found to be the dominating 332elemental attribute, which has the greatest influence on housing prices as 333hypothesized. In addition, being located in the attendant zone of a high-334quality school has nearly the same positive impact on housing prices as 335access to BRT (0.108 vs 0.100). Furthermore, among four location variables, 336Distance to city center exhibits the largest impact on housing prices, which is 337consistent with the bid-rent theory proposed by Alonso (1964). 3384.2 339Box-Cox transformation 340For further checking the robustness of variables, four Box-Cox functions 341are estimated. All functional forms represent relatively simple applications 342of the Box-Cox transformation: the simple left-hand-side Box-Cox model; 343the simple right-hand-side Box-Cox model; the simple both-side Box-Cox 344model; the separate both-side Box-Cox model. It should be noted that we do 345not transform five dummy independent variables as they are not definitely 346positive. 347Table 3 shows the results and reveals that the most flexible form, the 348separate both-side Box-Cox model, outperforms other three Box-Cox 349models. It has the lowest AIC (Akaike information criterion) value and the 350greatest adjusted R2 and Log-likelihood. 351The levels of statistical significance of all twelve independent variables 352are consistent in all six model specifications (two pre-specified and four 353Box-Cox transformed models). Kindergarten and Middle school are not 354associated with housing prices while other variables significantly affect 355housing prices (significant at the 5% level). This implies that our hedonic 356price models (both pre-specified and Box-Cox transformed) can explain 357price variations well and the variables included do affect housing prices. In 358addition, the signs of all coefficients are in line with expected. 359Table 3. Results of four Box-Cox transformed functional forms 360Variables 361Simple LHS 362Box-Cox 363model: 364Simple RHS 365Box-Cox 366model: 367Simple both-368side Box-Cox 369model: 370Separate both-371side Box-Cox 372model: 373374375===== PAGE 12 (score 85, numeric-paren 52, table-words 7, var-words 7) =====37637756 378IRSPSD International, Vol.6 No.1 (2018), 45-62379 380coefficient (t-381value) 382coefficient (t-383value) 384coefficient (t-385value) 386coefficient (t-387value) 388Constant 3899.784 (34.54)a 39058.907 (5.20)a 3911.137 (12.87)a 3921.618 (19.41)a 393Area 3940.055 (63.53)a 3951.429 (59.11)a 3961.035 (70.56)a 3970.833 (70.52)a 398Age 399-0.074 (-9.30)a 400-3.044 (-40110.17)a 402-0.169 (-40311.32)a 404-0.151 (-40511.20)a 406Local environment 4070.434 (9.36)a 40816.629 (8.05)a 4090.203 (8.00)a 4100.195 (8.01)a 411School 4120.616 (5.14)a 41336.359 (5.90)a 4140.128 (5.48)c 4150.129 (5.46)c 416Kindergarten 417-0.005 (-0.05)c 418-4.621 (-0.88)c 419-0.013 (-0.70)c 420-0.013 (-0.67)c 421Middle school 4220.029a (0.28)c 4234.011 (0.76)c 4240.036 (1.84)a 4250.033 (1.70)a 426Xiamen University 4270.990 (3.09)b 42842.604 (2.53)b 4290.170 (3.04)a 4300.172 (3.03)a 431BRT 4320.429 (4.93)a 43319.596 (4.38)a 4340.117 (6.81)a 4350.116 (6.72)a 436Distance to hospital 4370.252 (4.47)a 43811.195 (4.06)a 4390.022 (2.16)b 4400.023 (2.21)b 441Distance to 442cultural/sports center 443-0.294 (-4.71)a 444-12.988 (-4454.16)a 446-0.059 (-5.31)a 447-0.062 (-5.39)a 448Distance to city 449center 450-0.134 (-7.26)a 451-5.477 (-7.12)a 452-0.088 (-6.37)a 453-0.084 (-6.28)a 454Distance to lake 455-0.345 (-5.42)a 456-10.734 (-4573.32)a 458-0.061 (-5.99)a 459-0.063 (-5.99)a 460Left-hand side 461parameter 4620.317 (11.41)a 463- 464- 4650.032 (1.02)c 466Right-hand side 467parameter 468- 4691.128 (19.03)a 470- 4710.080 (1.45)c 472Both-hand side 473parameter 474- 475- 4760.030 (0.95)c 477- 478Performance statistics· 479Adjusted R2 4800.826 4810.806 4820.851 4830.851 484Log-likelihood 485-1999.85 486-6864.68 48739.83 48826.60 489AIC 4903.26 49111.15 492-0.04 493-0.02 494Note: a significant at the 1% level. b significant at the 5% level. c not significant at the 5% 495level. 4964.3 497Propensity score matching 498The “MatchIt” package in R constructs (Ho et al., 2011), is used for 499treatment effects estimation. The model employed to estimate propensity 500score, defined as the probability of receiving treatment, conditional on the 501covariates, is a binary logistic regression model. The explanatory variables 502used for logistics regression are the same as those used for the hedonic price 503function estimates. There are 158 treatment and 1092 control observations in 504our data set. The matching method we use is “optimal” matching, which 505seeks out the matched observations with the smallest total absolute distance 506within matched units. This method (i.e., “optimal” matching) is, in nature, 507strikingly different from “greedy” matching, which starts with some treated 508units and matches the closest control unit that has not yet been matched, 509without trying to minimize a global distance measure, so it efficiently avoids 510a problem of “greedy” matching, variant to the order in which units are 511matched (Ho et al., 2007). 512After matching, we compare the difference between the means for each 513explanatory variable in full samples and matched samples and find that 514matching does greatly enhance the balance of covariates. The estimation of 515the treatment effect reveals that the residential properties being located 516within high-quality school attendance zones (xuequfang) are 12.1% higher 517than the properties located outside of these areas (non-xuequfang), all else 518keeping fixed. 519520521===== PAGE 13 (score 4, numeric-paren 0, table-words 1, var-words 2) =====522523Yang et al. 52457525 5264.4 527High-quality school value premiums 528The following interpretation of the value premiums offered by high-529quality schools is based on the semi- and double-log hedonic price models 530developed and propensity score matching. The coefficient of School in the 531semi- and double-log model is 0.089 and 0.108, respectively. This indicates 532that residential properties being located within high-quality school 533attendance zones (xuequfang) exhibit values that are between 9.3% (=e0.089-5341) and 11.4% (=e0.108-1) higher than the properties located outside of these 535areas (non-xuequfang), ceteris paribus. The value premium estimation 536provided by propensity score matching (12.1%) is very consistent with the 537hedonic price regression results. 538The magnitude of the price premium offered by high-quality school of 539Xiamen Island is a little higher than that estimated in Beijing (8.1%) (Hu, 540Zheng, & Wang, 2014) and Shanghai (8.7%, 9.7%) (Shi & Wang, 2014; 541Peng, B. et al., 2015), but lower than that in Tianjin (l4.7%) (Wang, Z., Mei, 542& Wang, 2014) and Hong Kong (39.4%, 27.3%) (Jayantha & Lam, 2015). 5435. 544DISCUSSIONS 545According to the pedagogy theory, four factors - genetics, environment, 546school and subjective initiative - can affect the physical and mental 547development of people. Among them, only environment and school are 548space-relevant. The former involves loads of factors, with family 549environment accounting for a major proportion, while school is a very 550important factor which parents can do something for the physical and mental 551development of their children. 552The demarcation of school attendance zones is based on many factors 553such as school distribution, population distribution and natural boundary. It 554is, originally, to ensure the accessibility and security of school-age children. 555However, since elementary schools show exceptional heterogeneity (obvious 556disparities, tremendous diversity) in some aspects such as test performance 557and entering-school rate, parents’ expectation would deviate from the reality 558in terms of school choices. That is, assuming that a child with local hukou 559registered in the attendance zone of school A wants to attend school B, it is 560very difficult, if not impossible, according to strict policy of entrance from 561attendance zones alone, however. So, in most cases, parents need to buy a 562residence in the attendance zone of school B. As a result, the imbalance of 563inadequate supply and growing demand leads to the higher prices of 564xuequfang. For example, in Nanjing, the supply-demand ratio of xuequfang 565reaches 10 (Chen & Tang, 2009). Obviously, the advent of the xuequfang 566phenomenon is in stark contrast to original goals of school attendance zone 567demarcation. 568According to Chen and Tang (2009), there are a series of negative 569effects of the xuequfang phenomenon: (1) unfair educational resource 570allocation. Coupled with the advancement of the nation, education is no 571longer regarded as a privilege. Good educational resources are expected to 572be unfairly allocated and driven by the market mechanism. The introduction 573of territorial-based school admission policy which links educational 574resources with residential properties, distorts the market. As a result, the 575better the nearby educational resources are, the higher value the housing 576exhibits. This turns out that the rich can still obtain high-quality educational 577