Research in Transportation Economics 90 (2021) 100844

Available online 21 April 20200739-8859/© 2020 Elsevier Ltd. All rights reserved.

Subway boosts housing values, for whom: A quasi-experimental analysis

Chuanhao Tian a, b, Ying Peng a, Haizhen Wen b, c, *, Wenze Yue a, Li Fang d

a School of Public Administration, Zhejiang University, Hangzhou, 310058, China b Center for Real Estate Studying, Zhejiang University, Hangzhou, 310058, China c Department of Civil Engineering, Zhejiang University, Hangzhou, 310058, China d Department of Urban and Regional Planning, Florida State University, Tallahassee, 32306, USA

A R T I C L E I N F O

Keywords: Property price Subway Difference-in-differences Temporal dynamics Heterogeneous sub-market

A B S T R A C T

This study investigates the property price premium brought by the opening of a subway to illustrate temporal dynamics and heterogeneous mechanism of property value effects. The estimation of changes caused by subways on property value can aid in assessing the benefits of public transit investments well. On the basis of residential property transaction records in Hangzhou, China in 2009–2013, hedonic models in a difference-in-differences framework are applied to handle certain endogeneity problems of estimation by eliminating unobserved fac-tors. Results show that treatment groups located within 1000 m of a subway have an average price increase of 444 yuan per m2 from the opening of Line1. High-cost houses constantly gain significant increment and their price premium regress during the research period. However, for their counterparts in low-cost communities, the insignificant price effects are negative for a short time, and then become positive. The generalized results are robust when subway radius is adjusted.

1. Introduction

Subway lines play an important role in the process of urbanization. Traffic convenience determines the capacity and attractiveness of a city. Cities with a developed transportation system can cover additional re-sources, and citizens tend to have access to favorable opportunities. Therefore, city governments tend to build subway lines to boost the social and economic development.

However, not all of people in cities need subway lines to the same extent. In accordance with their features and affordability, they can have heterogeneous preference towards subway lines (Yang, Chen, & Le, 2016). Even during a short time period, the preference of a same group of people can be unstable and varies regularly (Agostini & Palmucci, 2010). An investigation of this heterogeneous preference can help policy makers to recognize people in urgent need for subway and optimize the allocation.

A way to study people’s preference towards subway lines is provided by Lancaster (1966). Since there is no explicit market for public infra-structure, the value of subway lines for citizens is embedded into the purchasing costs of housing. People should pay more for a recent subway line. This implied price shows people’s willingness to pay for subway

lines, as well as their deferential preference. A research on this hetero-geneous price premium can be a theoretical foundation for land value capture to keep a fiscally sustainable urbanization.

Previous studies utilized a cross-sectional comparison to probe the price effects (Wang. 2017, Teng, Yan, & Zhou, 2014; Cervero & Duncan, 2002). As an outcome of endogeneity, the estimation in those papers is subject to biases (Ko & Cao, 2013; Pagliara & Papa, 2011). Furthermore, the price effects reflecting people’s preference would vary over time and among sub-markets, which has not been investigated at the same time. The present study aims to fill those gaps and has two contributions to the present understanding of property value effects. On the one hand, this study applies a hedonic price model within a difference-in-differences (DID) framework to control unobserved factors and estimate property value effects precisely. On the other hand, this study elucidates the dynamics of a heterogeneous effect across different sub-markets, trun-cated based on the average property prices of the community. This improvement is helpful in determining the kind of property that has the highest price premium.

To achieve the mentioned goals, this study selects Hangzhou, China as the case study. Hangzhou is a regional central city in China, wit-nessing massive infrastructure construction in recent years. The subway

* Corresponding author. Center for Real Estate Studying, Zhejiang University, Hangzhou, 310058, China. E-mail addresses: [email protected] (C. Tian), [email protected] (Y. Peng), [email protected] (H. Wen), [email protected] (W. Yue), [email protected]

(L. Fang).

Contents lists available at ScienceDirect

Research in Transportation Economics

journal homepage: http://www.elsevier.com/locate/retrec

https://doi.org/10.1016/j.retrec.2020.100844 Received 2 April 2019; Received in revised form 9 February 2020; Accepted 6 April 2020

Research in Transportation Economics 90 (2021) 100844

2

has sprung up from nothing in the city center as well as the suburb. The change in Hangzhou reveals patterns about the citywide relation be-tween property prices and local transportation operation in developing areas. Consequently, experience in Hangzhou can be a favorable example for urbanizing cities worldwide wherein large-scale subway constructions will occur.

The structure of this paper is as follows. Part 2 reviews previous literature about the property value effect caused by subways and the improvement that requires accurate dynamics. Part 3constructs the main model for estimating the price effects. Part 4 describes the quasi- experimental setting in Hangzhou and the corresponding variables. Part 5 presents the empirical findings on the dynamics of property value effects. The final part provides the discussions and conclusions drawn from this study.

2. Literature

A subway line offers significant advantages for most real estate. Savings in commuting costs and promotion of accessibility are the main advantages (Ahlfeldt, 2013; Alonso, 1964; Hess & Almeida, 2007; Mills, 1967). The superiority can also partly contribute to subways’ image effect (Pan, 2013), that is, a subway line makes neighborhoods appear modern and dynamic. Properties with such advantages have higher values and this relation can be called property value effects caused by the subway on houses.

The added value that a subway line can contribute is typically dis-cussed using the hedonic price model. According to the hedonic price model, the price of a property is derived from its feature rather than the goods itself (Lancaster, 1966). By calculating the price that can be contributed to the distance feature, the extent of property value effects can be estimated. Using this strategy, Bae, Jun, and Park (2003) inves-tigate the effect of a new subway line on property price in Seoul, Korea in four separate years. He finds that property earns a 0.3% price increase from being 100 m near the subway station in the opening year. Similar negative relations between the distance to subway and property value are found in Beijing, CN (Wang, 2017), Tianjin, CN (Teng et al., 2014), Washington DC, US, Atlanta, US (Cervero & Landis, 1993), Utah, US (Li et al., 2016) and San Diego, US (Cervero & Duncan, 2002). Several literature works adopt a binary variable to represent the houses’ prox-imity to the subway (Dai, Bai, & Xu, 2016; Pan, 2013). Properties adjacent to the subway line will exhibit lower values than comparable properties located slightly further away (Pagliara & Papa, 2011; Ko & Cao, 2013). Influence of environmental negative externalities is verified and accounted for certain insignificant results, such as expected con-gestions and crime (Gatzlaff & Smith, 1993; Landis, Guhathakurta, & Zhang, 1994).

A universal concern of these studies is that the cross-sectional esti-mation might be biased because of endogeneity (Bae et al., 2003; Riekkinen, Hiironen, & Tuominen, 2015; Ko & Cao, 2013). On the one hand, the project placement is an endogenous decision of a local gov-ernment (Donaldson, 2018; McDonald, 2010). It means, the local gov-ernment tends to build the subway in places of low property prices and small population to reduce the cost and develop regional economy (Knaap, Ding, & Hopkins, 2001; Quan, Liu, & Chen, 2006; Yang et al., 2016). On the other hand, omitted variables which should be included into hedonic model can result in false estimations, such as neighborhood features, including race, neighbors’ income, and educational back-ground (Bayer & Mcmillan, 2007). Additionally, the potential for reaching opportunities by subway (P�aez, Scott, & Morency, 2012), walking impedance (Higgins, 2019), direct distance (Wen, Gui, Tian, Xiao, & Fang, 2018; Xiao, Hui, & Wen, 2019) and commuting time (Alonso, 1964) have been used as the measurement of accessibility. The lack of a definite proxy for distance also leads to omitted variables (Higgins & Kanaroglou, 2016). Furthermore, an increase in property value can be contributed to economic development rather than a sub-way’s opening (Agostini & Palmucci, 2010; Zhang, Liu, & Hang, 2016).

An overlook of the above mentioned problems can lead to insignificant or opposite results.

Several researchers use the DID methods to address endogeneity problems and improve the estimation accuracy of property value effects (Rosenthal, 2003; Black, 1999; Ahlfeldt, 2013; Dub�e, Th�eriault, & Ros-iers, 2013). Heckman (1978) proposes a DID estimator to handle auto-correlation, whereas Ashenfelter and Card (1985) first suggest a DID framework and apply this framework to policy evaluation. The core of this method is the removal of unobserved factors through longitudinal and latitudinal differences. Out of a similar thinking, Li, Wei, et al. (2016) and Li, Yang, et al. (2016) develop a first-differenced hedonic price framework to compare the property price for move-in date and survey date. Furthermore, Gibbons and Machin (2005) show that the magnitude of quasi-experimental coefficients is significantly larger than those of cross-sectional estimates. Im and Hong (2018) compare a before–after estimate of property prices between homes within and beyond 500m from a subway in Daegu, South Korea. The results illus-trate that the treatment group could earn a premium of US$96.3 per square meter mainly when these homes are within 500 m of the new line and beyond 5 km of the existing lines.

However an empirical gap is that most of them do not reveal varia-tion over time and among different sub-markets simultaneously. The price premium is an implication of the buyers’ willingness to pay for the subway and this preference could be temporally adjusted. See analysis in Dijon, FR (Dub�e, Legros, & Devaux, 2018), Laval, CA, Hangzhou, CN (Wen et al., 2018), and Warsaw, PL (Trojanek & Gluszak, 2017). Agostini and Palmucci (2010) claim that, if consumers have rational and fluctuant expectations, the future utility will be capitalized into the present price, and the effect of a subway on property price can vary during different times. When information is gradually released following the basic project, the excessive price premium will fall off. Based on this thought, they show that the magnitude of property value effects after opening tends to regress to a conservative status. Pilgram and West (2018) and Diao, Leonard, and Sing (2017) apply similar DID model and also demonstrate that significant price effects diminish over time. This temporal variation mentioned above can explain some insignificant generalized effects (Gatzlaff & Smith, 1993; Lawhon, Nilsson, Silver, Ernstson, & Lwasa, 2018). On the contrary, Dub�e, Des Rosiers, Th�eriault, and Dib (2011) show that the price premium keeps going up after over 12 years and explain that market needs time to react. Therefore, investigation of temporal variation is crucial to understand the supply of mass transit.

Furthermore, a potential buyer that purchases different kinds of properties may have varying preference or requirements for public goods (Pilgram & West, 2018). The housing market may be segmented into several sub-markets and property prices are not affected evenly by a subway (Devaux, Dub�e, & Apparicio, 2017; Gatzlaff & Smith, 1993; Lawhon et al., 2018). Therefore, property value effects can vary with features of houses. Because property price can be a perplexing symbol of the owners’ social status (Wen et al., 2018; Zhang & Yi, 2017), and residents tend to sort and live with people sharing similar preference (Bayer & Timmins, 2005), community level features are improved in-dicators for specifying sub-markets. Interestingly, while Gatzlaff and Smith (1993) show that the price premium in high-income communities is larger, Nelson and McCleskey (1990) find an opposite result that property price decrease in high income neighborhoods. One explanation for the mixed results is a failure to account for heterogeneous attributes of the residence (Bowes & Ihlanfeldt, 2001), including the distance to the city center and the efficiency of connection. Another explanation can be that a generalized estimation in the sub-markets do not consider the temporal variation (Debrezion, Pels, & Rietveld, 2007). So an estimation with combination of temporal variation in sub-markets is important to demonstrate the property value effects and the underlying mechanism.

In summary, previous studies identify price effects of subway con-struction with unobserved factors controlled. Some of them address price adjustment (Agostini & Palmucci, 2010; Diao et al., 2017) and

C. Tian et al.

Research in Transportation Economics 90 (2021) 100844

3

some others show the differential effects (Pilgram & West, 2018). However, an estimation of this differential effects over time and across sub-markets simultaneously is still needed to explain the mixed results and investigate the heterogeneous preference. Therefore, this study will utilize a similar DID model, and address the differential impact among different communities, as a comparison to Wen et al.’s work (2018). Then, this study will show the price effects in sub-markets over time, as an improvement of Pilgram and West’s (2018) work. To achieve the combination of DID and investigation of heterogeneity, the empirical part of the present study presents the following objectives:(1) to esti-mate the average property value effects of subway lines within a quasi-experimental framework,(2) to investigate the variation of this price premium over time, (3) to present the differential price effects among sub-market truncated based on the average price on a commu-nity level (APC) index and show the temporal fluctuations of these price premium in different sub-markets, and (4) to summarize the dynamics of property value effects from the subway’s operation.

3. Methodology

The present study constructs regression models to specify property value effects caused by subway’s opening (Line 1 in Hangzhou). The application of the DID framework can be helpful in evaluating the effect of Line 1 when addressing endogeneity (Abadie, 2005; Grafova, Freedman, & Lurie, 2014; Jim�enez & Perdiguero, 2017). Properties around subway stations are divided into treatment and control groups on the basis of whether they are located in the subway catchment. Thus, the direct distance from community centroid to the nearest subway station is less than 1000 m (Chen, 2011; Zhang & Jiang, 2014), the properties in this community are viewed as the treatment group. Moreover, properties in communities outside 1000 m are treated as the control group, thereby indicating that the distance to a subway is large, so that they are not influenced by the subway’s opening. A comparison of the property price of the control and treatment groups before and after the opening of Line1 shows the extent of the subway’s opening affecting house prices.

The statistic varies less as different subsets of houses are draw in the simple linear model (Kuminoff, Parmeter, & Pope, 2010) and this study aims to investigate the price premium rather than the price elasticity. So the form of price function in this paper is set linear. Another potential concern is that the spatial dependency of property price might result in estimation bias (Bae et al., 2003). To tackle this issue, apart from con-structing a spatial model (Dub�e, Legros, & Th�eriault, 2014; Higgins, 2019), Kuminoff et al. (2010) provide an alternative option to add spatial fixed effects in the non-spatial model. Therefore district fixed effects are included in the model. Hence, the basic DID model for eval-uating the effect is expressed below (Bogart & Cromwell, 2000; Dub�e et al., 2013):

priceit ¼αTreatit þ γTreatit Tit þ δTit þ βXit þ ut þ λi þ eit (1)

Legend i in the equation represents each observation. Legend t rep-resents each period. Tit equals 1 indicates that the observation has been performed after the opening of Line1.When Tit equals 0, the transaction has been performed before the opening of Line1. Other observable fea-tures, such as Xit, are included. ut represents time fixed effects, λi refers to district fixed effects, and eit is the error term. For the interaction item (Treatit*Tit), the magnitude of the coefficient (γ) equals the house price difference between the treatment group and control group. This price difference can be interpreted as the average added value brought by the subway’s opening.

When the time period after the subway’s operation is divided into several phases, temporal adjustment can be probed. Siq in equation (2) refers that this transaction happens during the qth season after the subway’s opening. Parameters γiq represent the price effects during this season. It also means, after Line 1 is operated for q season, the properties in treatment group gain a price premium of γq.

Table 1 Definitions of variables in this study.

Category Variable Definition Theoretical basis

transaction information

property price Average price of a house (yuan/m2 )

Variable of interest

time to the opening of Line1

How many months did the trade occur before (� ) or after(þ) the opening?

Variable of interest

dwelling size Total area of a house (m2 )

Witte, Sumka, and Erekson (1979)

dwelling feature

building height How many floors does the building have?

Conroy, Narwold, and Sandy (2013)

housing height On which floor is the house located?

Conroy et al. (2013)

Age Age of house (years) Xu, Zhang, and Zheng (2018).

location feature distance to Xizi Lake

Shortest walking distance from the community center to the Xizi Lake (km)

Gibbs, Halstead, and Boyle (2002)

distance to Qianjiang New Town

Shortest walking distance to the Qianjiang New Town (km)

Wen, Jin, & Zhang, 2017a,b

distance to Qiantang River

Shortest walking distance to the Qiantang River (km)

Gibbs et al. (2002)

distance to CBD Shortest walking distance to the CBD (km)

Alonso (1964)

distance to Line1 Shortest walking distance to the nearest subway station of Line1 (km)

Variable of interest

neighborhood feature

property management quality

Property management quality of a community divided into five degrees (scores): very good (5), good (4), common (3), poor (2), and very poor (1)

Wen et al., 2018,2019

environment quality

Quality of external environment divided into five degrees (scores): very good (5), good (4), common(3), poor (2), and very poor (1)

Wen et al., 2018,2019

neighborhood quality

Quality of internal environment divided into five degrees (scores): very good (5), good (4), common(3), poor (2), and very poor (1)

Wen et al., 2018,2019

infrastructure quality

Supermarkets, food markets, hospitals, banks, and post offices within 1 km of the community. Each item receives a score of 1 out of a total score of 5.

Li, Wei, et al. (2016) and Li, Yang, et al. (2016).

University University within 1 km of the community: yes (1) and no(0)

Wen et al., 2017a,2019

educational convenience

Kindergarten schools, primary schools, junior high schools, and senior high schools within 1 km of the community. Each item receives a score of 1 out of a total score of 4.

Oates (1969)

C. Tian et al.

Research in Transportation Economics 90 (2021) 100844

4

priceit ¼αTreatit þX4

q¼1γqTreatit Siq þ

X4

q¼1δq Siq þ βXit þ ut þ λi þ et (2)

Furthermore, this study claims that properties in communities with different average price can undergo heterogeneous property value ef-fects. Therefore, this study divides the properties into four sub-markets on the basis of the APC. The communities are quartered by the position of the APC (i.e., the cheapest quartile of the communities denoted as very cheap communities, the second cheapest quartile as cheap, the third cheapest quartile as expensive, and the least cheapest quartile as very expensive). When the analyzed sample of regression model (1) covers different sub-markets, then the differential price effects can be discovered. When the sample of temporal model (2) covers different sub- markets, the dynamics of how price effects vary over time and across sub-markets can be probed.

4. Data and variables

This study uses the opening of subway Line 1 in Hangzhou as the quasi-experimental setting. Hangzhou is the capital and political, eco-nomic, cultural, traffic, and educational center of Zhejiang Province, located in the southeast of Mainland China. This second-tier city in China has experienced a rapid development and infrastructure con-struction after the 2000s. The local subway Line 1 in Hangzhou consists of 27 stations with two parts, namely, the western part located in main districts and eastern part in suburb districts. The construction of Line1 started in March 2007, and the western part opened in November 2012. The eastern part opened much later and has not been explored.

Each house is taken as the basic analysis unit. The relevant data comes from three main sources. The housing price data are gathered from the Hangzhou Real Estate Administration. The neighborhood characteristic data, such as environment quality, educational conve-nience, infrastructure quality, etc., are obtained through a field survey involving 660 housing communities. Then, an electronic map is used to measure location features. Using the measuring distance function, we can obtain the distance from the housing community to the nearest subway station, Xizi Lake, Qianjiang New Town, and Qiantang River.

The proximity variable is also coded with the electronic map. We use a dummy variable to evaluate a university within 1 km from the com-munity as 1 and that outside 1 km as 0. Table 1 summarizes the defi-nitions of the above variables mentioned.

This study emphasizes the property value effect of subway Line 1. Only observations within 2000 m (twice the effect radius) of Line1 (Pagliara & Papa, 2011) are included in the empirical analysis. When the observed property is too far from Line1, factors other than the subway line may have a significant effect on the housing price (Cervero & Duncan, 2002). Furthermore, properties affected by subway lines other than Line 1 are disregarded. The object of previous studies is frequently one subway line. This choice of object’s quantity mainly results from two reasons. First, the subway system has not been developed well during the time. Second, the specification can be easy and precise when one subway line is focused on, and the disturbing effect from other subway lines is becoming constant. In addition, the effects from the different subway lines cannot be added literally (Jiao, 2016), and, oc-casionally, the effect is significant only when no interference exists (Im & Hong, 2018). In 2009–2013, two subway lines, namely, Lines 1 and 2, exist in Hangzhou. Therefore, observations influenced by Line2 are

Fig. 1. Location of subway Line 1 and distribution of observations.

Fig. 2. Average property price over time.

C. Tian et al.

Research in Transportation Economics 90 (2021) 100844

5

removed. Moreover, only 17360 records are retained, and their distri-bution in the city is demonstrated in Fig. 1. Table 2 summarizes the statistical description of those observations (see Fig. 2).

The average property price of the observations in each month is presented in Fig. 3. The treatment group and control group share a common trend. An apparent decline has been observed for all properties prior to the opening given the policy to limit property purchasing in 2011. The price steadily increases after the subway opening. After a comparison with the control group, whether the treatment group significantly earns a price premium after the opening remains unclear. Whether this effect is significant and whether this gap originates from the subway can be observed through the succeeding regression.

5. Empirical results

Table 3 displays the regression result based on cross-sectional com-parison. These models investigate the relation between the house price and distance to Line 1 before and after the subway’s opening separately. This cross-sectional comparison is universally adopted by previous

studies. The result shows that, before and after opening, the property price is lower in the treatment group than in the control group. It means that the property price is low when the property is close to a subway station. This relationship illustrates the existence of self-selection that subway station is tend to be built at places with lower property price (Yang et al., 2016), which causes endogeneity that leads to estimation bias and false causality. When compared horizontally, the coefficients changes from negative to positive, thereby the price gap is reversed. This comparison confirms that the opening of a subway line thus the prox-imity to the subway line increases the price of the treatment group. Whether the change in price is statistically significant or not remains to be investigated.

Modification to address endogeneity is implemented by applying the DID framework. The results of Equation (1) are summarized in Table 4. A robust standard error is applied to calculate the significance of

Table 2 Summary statistics of variables.

Variable Obs Mean Std. Dev.

Min Max

property price 17360 18963.18 5314.29 694.93 82788.89 time to the opening of

Line 1 17360 � 17.69 18.99 � 46 13

Treat 17360 0.63 0.48 0 1 Dwelling size 17360 76.01 36.00 15.55 378.01 distance to Line1 17360 899.38 455.77 14.99 1999.52 building height 17360 11.06 8.07 3 35 housing height 17360 5.99 5.15 1 33 distance to CBD 17360 4.08 2.62 0.54 12.14 distance to Xizi Lake 17360 4.11 2.25 0.16 13.1 distance to Qianjiang

New Town 17360 5.15 1.97 0.72 11.03

distance to Qiantang River

17360 4.22 2.43 0.1 11.22

property management quality

17360 2.62 1.32 1 5

environment quality 17360 3.25 0.70 2 5 neighborhood quality 17360 3.27 0.91 1 5 infrastructure quality 17360 4.24 1.13 0 5 university 17360 0.20 0.40 0 1 educational

convenience 17360 2.92 0.56 1 4

Age 17360 13.55 7.09 1 50

Fig. 3. Temporal dynamics of the property value effects for all the samples (yuan/m2).

Table 3 Relation between property price and distance to the nearest Line1 station based on cross-sectional comparison.

Property price (1)before opening (2) after opening

Time phases coefficients Standard errors

coefficients Standard errors

Treat (distance to line 1 < 1 km)

¡504.8*** 88.4 317.7* 143.3

dwelling size 19.7*** 2.3 8.1* 3.4 building height 106.4*** 15.8 ¡56.7*** 13.4 housing height � 5.1 11.3 17.9 11.3 distance to CBD 270.2** 84.3 20.5 165 distance to

XIZILake ¡1172.5*** 91.6 2098.2*** 266.1

distance to Qianjiang Newtown

165.1 86.2 1573.4*** 241.8

distance to Qiantang River

26.3 81.9 1780.3*** 304.7

university ¡424.4** 156.9 � 257.4 145.9 Age 46.0*** 11.9 ¡105.2*** 13.8 property management quality ¼ 1 Referred Referred ¼ 2 1280.2*** 134.9 67.2 217.4 ¼ 3 1082.3*** 139 685.5** 245.7 ¼ 4 2782.1*** 231.8 1218.0*** 289.4 ¼ 5 3659.0*** 351.2 3036.0*** 426.9 environment quality ¼ 1 No obs No obs ¼ 2 Referred Referred ¼ 3 189.8 128.1 ¡1116.9*** 173.3 ¼ 4 243 153.2 236 219.6 ¼ 5 3943.2*** 428.5 ¡1773.5*** 422 neighborhood quality ¼ 1 Referred Referred ¼ 2 1121.4*** 141.1 No obs ¼ 3 1308.0*** 154.8 1052.0*** 212.9 ¼ 4 1066.9*** 210.5 1289.5*** 259.3 ¼ 5 443.2 308.7 1989.7*** 389.2 infrastructure quality ¼ 0 Referred Referred ¼ 1 ¡1655.0*** 407.1 ¡1482.3*** 370.1 ¼ 2 2342.0*** 465.9 2277.1*** 448.1 ¼ 3 3039.0*** 468.6 248.5 480.5 ¼ 4 954.4* 427.3 363.6 419.3 ¼ 5 1025.2* 440.1 � 31.7 423.5 educational quality ¼ 1 Referred Referred ¼ 2 2098.2*** 266.1 1441.8 1011.5 ¼ 3 1573.4*** 241.8 1451.1 1011.7 ¼ 4 1780.3*** 304.7 1708.5 1035.5

Month fixed effects YES YES District fixed

effects YES YES

N 12542 4818 R2 0.622 0.513

*p < 0.05, **p < 0.01, ***p < 0.001.

C. Tian et al.

Research in Transportation Economics 90 (2021) 100844

6

coefficients. The result of Model 2 demonstrates that the treatment group earns a 444 premium from the opening when other variables are controlled.

Considering that properties with a specific distance to subway station might experience various negative externalities (Lieske, Nouwelant, Han, & Pettit, 2019), this study introduces distance fixed effects into the present models. For example, properties located within 200 m to subway

line can be bothered by noise, and properties with a distance of 700 m to subway tend to experience congestion (Mohammad, Graham, & Melo, 2015; Wen et al., 2018). Numerous dummy variables (D200, D300, D400, D500, D600, D700, D800, D900, and D1000) that indicate the distance range are included to capture the non-linear impact of the proximity to subway line. Variable D200 equals 1 represents that the distance to Line 1 is less than 200 m. Variable D300 represents that the

Table 4 Baseline regression with a DID framework.

Property price (1) (2)

Coefficients Standard errors

coefficients Standard errors

Treat*T (γ) 369.7** 132.8 444.1*** 133.4 Treat (α) ¡361.6*** 81.6 8.2 165.6 T (δ) � 17.5 430.6 67.6 426.2 L200 Referred Referred L300 2259.1*** 288.2 L400 � 313 166.7 L500 � 36.2 195 L600 ¡1386.7*** 168.7 L700 ¡1177.8*** 162.9 L800 231.5 187.8 L900 ¡817.1*** 175.2 L1000 � 51.7 182.2 dwelling size 18.0*** 1.9 18.6*** 1.9 building height 65.4*** 12.9 50.9*** 11.3 housing height � 0.9 8.8 5.4 8.2 distance to CBD 245.8*** 72.9 486.4*** 75.2 distance to

XIZILake ¡1282.5*** 77.6 ¡1374.9*** 79.3

distance to Qianjiang Newtown

217.1** 66.3 136.9* 67.1

distance to Qiantang River

¡169.8** 63.6 33.8 66.7

university ¡344.1** 116 � 72.1 113.7 Age 13.2 10.3 � 5.4 8.7 property management quality ¼ 1 Referred Referred ¼ 2 1169.8*** 102.9 1073.1*** 107.2 ¼ 3 889.0*** 113.8 1073.9*** 114.5 ¼ 4 2506.1*** 175.4 3054.2*** 180.8 ¼ 5 3533.6*** 275.3 3854.4*** 288 environment quality ¼ 1 No obs No obs ¼ 2 Referred Referred ¼ 3 � 196.9 100.6 165 105.6 ¼ 4 91.2 124.1 393.0** 125.4 ¼ 5 2038.7*** 316.8 1817.3*** 320.2 neighborhood quality ¼ 1 Referred Referred ¼ 2 903.7*** 134.5 700.5*** 131.6 ¼ 3 1214.8*** 143.7 828.9*** 132.8 ¼ 4 1053.7*** 180.9 324.6 174.1 ¼ 5 580.4* 252.5 � 179.3 261 infrastructure quality ¼ 0 Referred Referred ¼ 1 ¡1709.2*** 274.4 ¡2526.0*** 277.3 ¼ 2 2404.7*** 330.8 2597.4*** 329 ¼ 3 2331.3*** 345.4 1789.3*** 338.1 ¼ 4 786.0* 306.7 291.8 308 ¼ 5 705.6* 316.2 216.8 314 educational quality ¼ 1 Referred Referred ¼ 2 2024.8*** 231.5 1845.5*** 229.7 ¼ 3 1597.3*** 216.6 1827.7*** 210.2 ¼ 4 1910.9*** 256 2389.5*** 249.4

Month Fixed Effects

YES YES

District fixed effects

YES YES

N 17360 17360 R2 0.603 0.615

*p < 0.05, **p < 0.01, ***p < 0.001.

Table 5 Changes in property value effects over time.

(1)

coefficients Standard errors

Treat (α) � 4.10 165.60 S1 (δ1) 10108.3*** 382.80 S2 (δ2) 10400.3*** 316.20 S3 (δ3) 10459.9*** 362.80 S4 (δ4) 10331.1*** 384.50 Treat*s1 (γ1) 111.00 249.90 Treat*s2 (γ2) 650.7*** 189.90 Treat*s3 (γ3) 235.10 249.30 Treat*s4 (γ4) 622.1* 242.10 L200 Referred L300 2264.8*** 288.10 L400 � 303.40 166.60 L500 � 30.10 195.00 L600 ¡1375.7*** 168.70 L700 ¡1165.3*** 162.80 L800 244.20 187.70 L900 ¡811.1*** 175.20 L1000 � 40.60 182.10 dwelling size 18.6*** 1.90 building height 50.9*** 11.30 housing height 5.50 8.20 distance to CBD 484.2*** 75.30 distance to XIZILake ¡1374.9*** 79.30 distance to Qianjiang Newtown 140.3* 67.10 distance to Qiantang River 29.90 66.80 university � 73.50 113.70 Age � 5.20 8.70 property management quality ¼ 1 Referred ¼ 2 1072.9*** 107.30 ¼ 3 1075.2*** 114.50 ¼ 4 3054.9*** 180.80 ¼ 5 3854.7*** 288.20 environment quality ¼ 1 No obs ¼ 2 Referred ¼ 3 168.40 105.60 ¼ 4 397.0** 125.50 ¼ 5 1817.6*** 320.00 neighborhood quality ¼ 1 Referred ¼ 2 701.7*** 131.70 ¼ 3 831.8*** 132.90 ¼ 4 326.90 174.20 ¼ 5 � 176.20 261.10 infrastructure quality ¼ 0 Referred ¼ 1 ¡2526.5*** 277.50 ¼ 2 2597.9*** 329.20 ¼ 3 1796.5*** 338.30 ¼ 4 296.90 308.10 ¼ 5 223.10 314.20 educational quality ¼ 1 Referred ¼ 2 1851.6*** 229.70 ¼ 3 1830.1*** 210.10 ¼ 4 2403.5*** 249.20

Month fixed effects YES District fixed effects YES

N 17360 R2 0.605

*p < 0.05, **p < 0.01, ***p < 0.001.

C. Tian et al.

Research in Transportation Economics 90 (2021) 100844

7

distance is more than 200 m but less than 300 m. The difference between Models 3 and 4 verifies that considering the distance effect can increase R square from 0.603 to 0.605 and contribute to an accurate estimation. Therefore, when other variables and fixed effects are controlled, the opening of Line 1 causes an average price increase of 444 for all influ-enced properties.

The above regression explores the effect for the whole market, and this part probes into the effect over time. The results are presented in Table 5. The property value effect is quite unstable. The confidence in-terval depicted in Fig. 4 can be interpreted as the distribution of the price effects over time. Although the magnitude of γ for different periods fluctuated, all the coefficients are not statistically significantly different.

Thus, the property value effects remain at a stable level after the sub-way’s opening. However, these property value effects appear to be sta-tistically significant after three months of opening (see Table 6).

The heterogeneity of this effect across sub-markets is discussed in the succeeding part. All properties are divided into four sub-groups on the basis of their APCs. The implication is that the property purchasers originate from different sub-markets and may have different preference for a subway. Therefore, the properties may be influenced by the opening of Line 1 to a diverse extent.

The property value effects differ depending on the average price of the community where the property is located. For properties sitting in the cheapest communities (first quartile communities), the price gap

Table 6 Estimation of the heterogeneous property value effects in different communities.

Property price (1) Very cheap communities (2) Cheap communities (3) Expensive communities (4) Very expensive communities

coefficients Standard errors coefficients Standard errors coefficients Standard errors coefficients Standard errors

Treat*T (γ) � 72.3 185.3 � 193.2 209.9 � 149.8 186 968.9** 364.9 Treat (α) 2873.4*** 759.6 1750.4*** 405.4 � 361.6 317.5 � 90.8 394.5 T (δ) 799.5 736 � 31.8 486.3 � 345 520.1 530.7 1011.6 L200 Referred Referred Referred Referred L300 � 1298.8 1048.7 No obs 150.1 460.4 2598.7*** 389.6 L400 � 2210.9* 890.8 � 2504.0*** 432.9 356.8 345.9 619.5 346.1 L500 � 3013.3** 951 � 2971.0*** 493.5 288 373.1 743.7 384.6 L600 � 1178.5 736.8 � 1029.5* 419.4 836.3** 260.2 393.2 441.8 L700 � 2225.4* 887.2 � 1702.0*** 457 1474.2** 487.9 � 2095.8*** 394 L800 � 1236 789.4 � 2433.5*** 444.9 � 304.2 458.2 1906.2*** 571.2 L900 � 2315.0** 773.7 � 813.3* 372.6 190.8 379.3 608.5 379.8 L1000 � 2259.6** 747 � 1667.6*** 421.1 1503.4** 517.6 � 599.6 703.5 dwelling size � 7.0*** 1.6 � 12.6*** 2.1 2.3 1.8 43.7*** 3.8 building height � 53.4*** 14.1 � 36.9* 14.6 � 53.6*** 9 134.3*** 20.7 housing height 9.3 9.3 24.6* 11.2 31.5*** 8.1 � 5 20.2 distance to CBD � 888.3** 297.9 � 209.9 231.6 � 88.1 218.1 � 348.7 340.4 distance to XIZILake 911.7** 289.3 � 181.3 243.2 � 366.7* 167.3 � 2205.9*** 294.8 distance to Qianjiang Newtown � 1320.4*** 240.2 496.8 295.7 � 223.3 197 1134.9*** 321.5 distance to Qiantang River 233.8 211.7 � 413.5 288.6 � 283.3 243.9 � 1546.3*** 463.2 university 896.9* 403.5 � 143.7 133.9 � 272 255.1 � 754.2 394.4 age � 49.4** 15.7 47.7* 22 � 72.2*** 21.5 91.0** 29.5 property management quality ¼ 1 Referred Referred Referred Referred ¼ 2 233 339.3 � 516.6* 234.9 � 592.6* 235.3 1993.9*** 378.8 ¼ 3 90 289 119.2 253 � 594.7* 284.1 1178.3** 391.9 ¼ 4 263.8 491.2 2287.0*** 332.9 1459.9*** 386.3 1552.3** 479.1 ¼ 5 279.4 612.8 1051.5* 450 788.1 563.3 2954.9*** 792.3 environment quality ¼ 1 ¼ 2 Referred Referred Referred Referred ¼ 3 � 667.9* 307.9 212.8 208.9 47.4 275.8 � 909.3* 366.7 ¼ 4 1178.3** 358.5 138.2 263.4 515.6 303.3 � 815.2 638.3 ¼ 5 � 388.7 711.1 No obs 636.6 546 1168.8 703 neighborhood quality ¼ 1 Referred Referred Referred Referred ¼ 2 687.6 370.8 1597.1** 583.6 566.4 567.3 No obs ¼ 3 175.1 268.1 1526.5* 605.9 249.2 565.1 � 257.2 478.6 ¼ 4 641.9 496.5 353.4 666.2 598.8 583.9 464.9 547.3 ¼ 5 � 611 622.3 2516.5* 1003.8 1736.8** 670 � 930.1 689.1 infrastructure quality ¼ 0 Referred Referred Referred Referred ¼ 1 361 632.8 No obs � 3581.5* 1658 No obs ¼ 2 � 251.9 705.6 No obs � 354.4 1499.5 2686 3655.1 ¼ 3 � 3307.3** 1172.4 2302.4* 985 � 1363.3 1314.1 922.9 3612 ¼ 4 � 3496.1*** 874.8 2848.6** 1041.1 � 993 1499.8 2275.3 3669.9 ¼ 5 � 2821.8** 963.9 3085.9** 1031.9 � 723.6 1441.3 51.7 3635.9 educational quality ¼ 1 Referred Referred Referred Referred ¼ 2 399.5 468.6 350.6 388.3 � 224.8 2001.9 No obs ¼ 3 399.9 340.7 � 682.7 384.9 765.4 1885.2 2883.3*** 368.3 ¼ 4 349.1 1076.9 No obs 1342.1 1930.3 1584.1*** 434.2

Month fixed effects YES YES YES YES District fixed effects YES YES YES YES

N 4398 4309 4346 4307 R2 0.786 0.735 0.628 0.442

*p < 0.05, **p < 0.01, ***p < 0.001.

C. Tian et al.

Research in Transportation Economics 90 (2021) 100844

8

between the treatment and control groups is � 72. Only properties in the most expensive quartile of communities experience a significant price increase at approximately 968 yuan. It means that prices of those property rocket by 968 yuan per m2 after the opening of Line 1. By contrast, properties in other community are faced with a price bump, though those decreases are not significant.

The final part investigate the heterogeneity over time and between the sub-markets at the meantime. Observations from different sub- markets are included in each regression and the results of these re-gressions are summarized in Table 7. The results show that after three months after opening, properties sitting in the most expensive commu-nities earn the most premium at around 1442 yuan per m2, and added value decreases over time. For properties in the second most expensive communities, the changing pattern of price effects is similar. The pre-mium increases at first and then declines, though the decline comes later and the premium is smaller and insignificant. Properties in the other two quartile of communities share a same dynamics of price effects, which is totally different from the expensive ones. Their prices decrease in the first season after the opening and keep going up, though the price pre-mium is not statistically significant.

Based on Table 7, more conclusion about the differential preference towards housing features can be drawn. For instance, the proximity to Xizi Lake, a famous sightseeing in Hangzhou, has a negative price impact upon expensive properties and a positive price impact upon cheap properties. While the property size have a positive impact upon price of expensive houses, the cheap property is negatively influenced. Though purchasers of expensive and cheap properties have a different taste to-wards sightseeing as well as size, they share a similar preference towards property management. They are willing to pay more for the better management.

In a word, Fig. 4 in the early part demonstrates that price effects in the entire property market is slightly stable after the subway’s opening. When different sub-markets are considered, Table 7 reflects that prop-erties in varying sub-markets react differently and unsteadily. Not only the subway’s operation, but also many other features bring about dif-ferential price premium.

The average price effects from the opening of Line 1 can reach 444 yuan per m2 when distance threshold equals 1000 m. To illustrate the robustness of the analysis, different radii used in literatures (Gu & Jia, 2008; Liang, 2007; Mohammad et al., 2015; Pilgram & West, 2018; Trojanek & Gluszak, 2017; Zhang & Jiang, 2014) are adopted in the following part, which can range from 500 to 1500 m. The coefficients and its confident interval are reorganized in Fig. 4. It can be told that the property value effect is always significant and the empirical analysis is robust. Thus, the opening of Line 1 results in a house price premium.

6. Discussions

In this study, properties located within 1000 m of a subway are used as the treatment group, whereas other properties are the control group. With other factors being constant, the properties in the treatment group earn a price premium significantly by approximately 444 yuan per m2

from a subway’s opening (nearly 2.3% of the average housing price). Nevertheless Wen et al. (2018) show the property price within 1000 m of subway stations is around 3% more expensive than those beyond 1000 m and they become around 6.5% more expensive after the subway operation. And it can be calculated that a premium at 3.5% is brought by the subway’s operation. While Wen’s work overlooks the temporal factor contributing to the overall premium, including the increasing population and economic boom in Hangzhou during the study period, and might lead to omitted variable bias in the cross-sectional compari-son, the smaller estimation within DID framework in this study should be more accurate. Additionally, the premium estimated with Wen’s strategy in Table 1 is around 800 yuan per m2, larger than the ultimate estimation at 444 yuan per m2, illustrating that the omitted variables result in an upward bias.

To date, this study reflects research works on a solution to endoge-neity. Moreover, this study unveils further heterogeneity of property value effects, which have been frequently overlooked in previous studies. This study finds that the property that is located in the most expensive communities experience a price increase of nearly 1,000 yuan per m2. By contrast, estimates for properties in other community are negative and none of them are significant. This phenomenon is consis-tent with Pilgram and West’s (2018) finding.

When the heterogeneous effects is investigated with temporal vari-ation, more interesting findings are revealed. Although some houses in expensive communities have positive generalized premium, the pre-mium fall off to a insignificant level in the late research period. For other houses which witness no significant generalized price effects, the pre-mium keep insignificant all the time. This conclusion adds to the pre-vious studies that consumer’s’ willingness to par for subway is going down, although the generalized estimation is significant positive. This finding is rational. Line 1 is the first subway line in Hangzhou and it is understandable that people have positive expectation. However, the connection is imperfect which would hinder the efficient use of subway and lower the need (Sasaki, 1990). Therefore, market recesses as a re-action to the subway’s opening.

Further, although not all the coefficients are significant, their magnitude demonstrate an opposite changing pattern, reflecting that consumers could have a totally different preference towards subway lines (Higgins & Kanaroglou, 2018). In the early stage, potential pur-chasers of expensive houses show a special preference towards subway lines though buyers of cheap houses have no preference. As time goes by, enthusiasm of expensive housing’s buyers towards subway lines de-clines, while buyers of cheap houses like subway lines more and more. This information demonstrates that provision of public transport could be optimized in accordance with the differential preference. Uniform provision might lead to an unbalance between supply and demand. This argument has been constantly overlooked in Pilgram and West’s studies (2018). The ignorance of discussion above is why this study lays emphasis on combining a time dimension with a sub-market dimension.

This accessibility increment value also has further implication for transport funding from two perspectives. First, the added-value brought by subway operation would amount to 2.3% of the property price. Hangzhou is going through more than five subway construction in the following five year, which imposes great financial burden on govern-ments. Therefore, it’s important to develop an alternative funding strategy and build up a land value capture mechanism. Second, the empirical results illustrate that only a few part of properties gain from public investment. So whichever methods the government is going to use to capture the added-value, properties from various sub-markets should be treated differently. More specifically, betterment tax should

Fig. 4. Effect specification with different radii (yuan/m2).

C. Tian et al.

Research in Transportation Economics 90 (2021) 100844

9

be more levied upon property in expensive communities to capture the added-value (Medda, 2012).

7. Conclusions

Cities are expanding, and accessibility becomes more important than before. Potential housing purchasers prefer properties within a radius of

a subway line. They are willing to pay an additional amount to purchase a house, and the property obtains a price premium. Thus, a subway’s opening is reflected in property value, whose dynamics is investigated in this study.

On the basis of a DID framework, this study empirically analyzes the property value effects. According to the results, three main conclusions are drawn. First, the problem of endogeneity and estimation bias is

Table 7 Regression to estimate property value effects in different communities over time.

Property price (1) Very cheap communities (2) Cheap communities (3) Expensive communities (4) Very expensive communities

coefficients Standard errors coefficients Standard errors coefficients Standard errors coefficients Standard errors

Treat (α) 2855.8*** � 758.1 1680.7*** 404.6 � 338.6 316.7 � 143.3 397.1 S1 (δ1) 8713.0*** 767.6 8751.8*** 525.1 8665.3*** 658.6 11416.7*** 1011.5 S2 (δ2) 8675.5*** 593.2 9502.5*** 585.7 9294.3*** � 624.2 11091.8*** 972.5 S3 (δ3) 8701.0*** 607.6 8959.1*** 701.6 9184.9*** 618.2 10409.1*** 1028.3 S4 (δ4) 8754.1*** 633.7 8014.7*** 548.2 9708.0*** 616.7 11086.1*** 977.4 Treat*s1 (γ1) � 403.8 358.2 � 574.7 340.1 � 393 341.9 788.7 628.5 Treat*s2 (γ2) � 104.4 247.6 � 568 316.9 26.9 269.1 1442.6** 513.1 Treat*s3 (γ3) 12.8 287 253.6 521.7 126.4 335 997.5 659.9 Treat*s4 (γ4) 79 330.3 654.6 346.4 � 528 317.9 887.3 627 L200 Referred Referred Referred Referred L300 � 1293.9 1046.3 No obs 143.4 460.1 2607.8*** 389.5 L400 � 2137.1* 889.3 � 2461.5*** 432.6 343.4 345.2 634.7 345.7 L500 � 2993.8** 953.1 � 2906.1*** 492.9 281.4 373.5 768.8* 384.1 L600 � 1155 735.1 � 996.1* 420.2 831.7** 260.4 412.3 441.8 L700 � 2272.0* 886.4 � 1653.3*** 457.3 1477.2** 487.4 � 2078.9*** 393.9 L800 � 1210.6 787.5 � 2367.9*** 444 � 320.3 456.8 1968.1*** 571 L900 � 2296.1** 773 � 787.4* 373.5 176.6 379.4 620.8 380.3 L1000 � 2237.0** 745.4 � 1613.9*** 420.8 1509.5** 517.1 � 576.7 704.2 dwelling size � 7.0*** 1.6 � 12.6*** 2.1 2.3 1.8 43.5*** 3.8 building height � 53.3*** 14.1 � 37.9** 14.4 � 54.1*** 9.1 134.5*** 20.8 housing height 9.3 9.3 25.1* 11.2 31.7*** 8.1 � 5 20.2 distance to CBD � 881.4** 298.2 � 170.3 230.9 � 86.1 217.4 � 337.5 340 distance to XIZILake 906.5** 289.2 � 228 242.4 � 364.9* 166.8 ¡2213.2*** 294.8 distance to Qianjiang Newtown ¡1322.8*** 240.1 463.8 295.6 � 225.4 197 1128.9*** 321.5 distance to Qiantang River 238.1 211.8 � 374.7 288.4 � 282.1 243.5 � 1533.0*** 462.9 university 894.6* 402.5 � 130.6 133.5 � 274.2 254.9 � 764.8 395.1 age � 49.5** 15.8 46.8* 22.1 � 72.2*** 21.5 90.4** 29.5 property management quality ¼ 1 Referred Referred Referred Referred ¼ 2 243.8 338.3 ¡493.3* 234.4 ¡603.8* 235 1996.8*** 379.6 ¼ 3 88.2 289.6 152.9 253.5 ¡600.9* 284.3 1166.0** 392 ¼ 4 290.7 490.8 2282.4*** 332.5 1443.6*** 387.5 1562.3** 478.6 ¼ 5 311.9 613.8 1082.2* 449.3 771.8 562.4 2973.5*** 791.9 environment quality ¼ 1 No obs No obs No obs No obs ¼ 2 Referred Referred Referred Referred ¼ 3 ¡673.5* 308.1 225.5 209 46.1 275.4 ¡906.4* 366.7 ¼ 4 1177.3** 358.2 172.1 263.6 514.9 302.5 � 825.2 639 ¼ 5 � 390.3 710.3 No obs 642.2 542.2 1154.4 703.1 neighborhood quality ¼ 1 Referred Referred Referred Referred ¼ 2 683.1 370.7 1614.8** 582.3 590.3 565.1 No obs ¼ 3 173.1 267.8 1550.5* 604.9 269.4 563.8 � 257.1 477.6 ¼ 4 616 496.1 375.2 664.7 625.4 583.1 455.2 545.9 ¼ 5 � 644.8 622.9 2559.9* 1003.8 1765.1** 669.6 � 943.5 687 infrastructure quality ¼ 0 Referred Referred Referred Referred ¼ 1 356.9 636.3 No obs ¡3551.2* 1653.1 No obs ¼ 2 � 253.6 704.1 No obs ¡350.8 1496.5 2675.8 3655.2 ¼ 3 ¡3324.1** 1172.7 2345.4* 984.6 � 1363.2 1312.9 936 3611.4 ¼ 4 ¡3507.4*** 876.1 2873.0** 1040.1 � 974.3 1496.7 2254.1 3669.3 ¼ 5 ¡2840.2** 964.7 3123.2** 1031.5 � 710.4 1438.8 32.4 3635.2 educational quality ¼ 1 Referred Referred Referred Referred ¼ 2 406.4 468.8 335.3 388.3 � 227.1 1997.4 No obs ¼ 3 404.8 341.8 � 697.8 385.4 762.3 1881.5 2850.6*** 368.6 ¼ 4 320.1 1077.6 No obs 1334.9 1928.6 1558.2*** 434.9

Month fixed effects YES YES YES YES District fixed effects YES YES YES YES

N 4398 4309 4346 4307 R2 0.779 0.734 0.621 0.436

*p < 0.05, **p < 0.01, ***p < 0.001.

C. Tian et al.

Research in Transportation Economics 90 (2021) 100844

10

revealed using the cross-sectional comparison in previous studies. A subway line tends to be constructed in a place where initial property prices are low. Our results also illustrate that the property value in treatment group has rocketed significantly, which can be put down to several factor, including the larger population. Our model shows that after the unobserved factors are controlled, the added value from the opening of a subway line is 444 yuan per m2, which explains about 2.3% of the price increase during the study period.

Second, this study depicts temporal patterns of property value effects that require several months for the affected property price to reflect its significant change. After such a reaction period, the market remains in a stable state. Static cross-sectional estimation is insufficient to illustrate this phenomenon. The evaluation of a price control policy must consider temporal changes.

Third, the value effects are larger and more obvious for high-priced properties than others. The effects may become negative for properties in the cheap community in the early stage and the positive effects appear to decline for properties in the expensive communities later. This effect is an implication of a differential preference towards public goods. In these sub-markets, the temporal patterns during the first three months and after three months are various. Buyers of expensive houses tend to bid for proximity to subway lines much earlier than that of cheap houses. But the enthusiasm of former ones lasts a short time and buyers of cheap houses begin to have a favor towards subway lines. This dif-ferential preference is also reflected in the results of educational resource, housing size, etc.

These conclusions can be a reference in policy-making field. First, public construction stimulates property markets. Thus, a cost–benefit analysis of public investment must take such benefits into consideration. Second, properties from varying sub-markets obtain various price pre-miums and differential mechanism should be imposed to capture the land value. Finally, consumers have a differential preference towards public goods over time, Local government can optimize the resource allocation through providing heterogeneous public services.

However, the possible explanations for empirical results require added empirical evidence. The micro mechanism of the heterogeneous property value effects can be revealed by incorporating a purchaser’s census and market analysis. Additionally, based on geographic infor-mation of each housing, the future research can construct a spatial model to obtain more robust estimation.

Funding

This study was funded by the National Social Science Foundation of China (No: 17BJY224), the National Natural Science Foundation of China (No:71974169) and Zhejiang Provincial Natural Science Foun-dation of China (No: LY18G030002). All errors remain are our own.

CRediT authorship contribution statement

Chuanhao Tian: Conceptualization, Supervision, Funding acquisi-tion. Ying Peng: Investigation, Formal analysis, Writing – original draft. Haizhen Wen: Methodology, Supervision, Funding acquisition. Wenze Yue: Conceptualization, Methodology. Li Fang: Investigation, Writing – review & editing.

References

Abadie, A. (2005). Semiparametric difference-in-differences estimators. The Review of Economic Studies, 72(1), 1–19.

Agostini, C. A., & Palmucci, G. A. (2010). The anticipated capitalisation effect of a new metro line on housing prices. Fiscal Studies, 29(2), 233–256.

Ahlfeldt, G. M. (2013). If we build, will they pay? Predicting property price effects of transport innovations. Environment & Planning A, 45(8), 1977–1994.

Alonso, W. (1964). Location and land use: Toward a general theory of land rent. Cambridge, Massachusetts: Harvard University Press.

Ashenfelter, O., & Card, D. (1985). Using the longitudinal structure of earnings to estimate the effect of training programs. The Review of Economics and Statistics, 67(4), 648–660.

Bae, C. H. C., Jun, M. J., & Park, H. (2003). The impact of Seoul’s subway Line 5 on residential property values. Transport Policy, 10(2), 85–94.

Bayer, P., & Mcmillan, R. (2007). A unified framework for measuring preferences for schools and neighborhoods. Journal of Political Economy, 115(4), 588–638.

Bayer, P., & Timmins, C. (2005). On the equilibrium properties of locational sorting models. Journal of Urban Economics, 57(3), 462–477.

Black, S. E. (1999). Do better schools matter? Parental valuation of elementary education. Quarterly Journal of Economics, 114(2), 577–599.

Bogart, W. T., & Cromwell, B. A. (2000). How much is a neighborhood school worth? Journal of Urban Economics, 47(2), 280–305.

Bowes, D. R., & Ihlanfeldt, K. R. (2001). Identifying the impacts of rail transit stations on residential property values. Journal of Urban Economics, 50(1), 1–25.

Cervero, R., & Duncan, M. (2002). Land value impacts of rail transit services in san Diego county. Report prepared for National Association of Realtors Urban Land Institute.

Cervero, R., & Landis, J. (1993). Assessing the impacts of urban rail transit on local real estate markets using quasi-experimental comparisons. Transportation Research Part A -Policy & Practice, 27(1), 13–22.

Chen, X. (2011). The proximity impact of metro station on price of surrounding residential property in Shenzhen. Master thesis. Hong Kong Polytechnic University.

Conroy, S., Narwold, A., & Sandy, J. (2013). The value of a floor: Valuing floor level in high-rise condominiums in San Diego. International Journal of Housing Markets and Analysis, 6(2), 197–208.

Dai, X., Bai, X., & Xu, M. (2016). The influence of Beijing rail transfer stations on surrounding housing prices. Habitat International, 55, 79–88.

Debrezion, G., Pels, E., & Rietveld, P. (2007). The impact of railway stations on residential and commercial property value: A meta-analysis. The Journal of Real Estate Finance and Economics, 35(2), 161–180.

Devaux, N., Dub�e, J., & Apparicio, P. (2017). Anticipation and post-construction impact of a metro extension on residential values: The case of Laval (Canada), 1995-2013. Journal of Transport Geography, 62, 8–19.

Diao, M., Leonard, D., & Sing, T. F. (2017). Spatial-difference-in-differences models for impact of new mass rapid transit line on private housing values. Regional Science and Urban Economics, 67, 64–77.

Donaldson, D. (2018). Railroads of the raj: Estimating the impact of transportation infrastructure. The American Economic Review, 108(4–5), 899–934.

Dub�e, J., Des Rosiers, F., Th�eriault, M., & Dib, P. (2011). Economic impact of a supply change in mass transit in urban areas: A Canadian example. Transportation Research Part A: Policy and Practice, 45(1), 46–62.

Dub�e, J., Legros, D., & Devaux, N. (2018). From bus to tramway: Is there an economic impact of substituting a rapid mass transit system? An empirical investigation accounting for anticipation effect. Transportation Research Part A: Policy and Practice, 110, 73–87.

Dub�e, J., Legros, D., & Th�eriault, M. (2014). A spatial Difference-in-Differences estimator to evaluate the effect of change in public mass transit systems on house prices. Transportation Research Part B, 64(64), 24–40.

Dub�e, J., Th�eriault, M., & Rosiers, F. D. (2013). Commuter rail accessibility and house values: The case of the Montreal South Shore, Canada, 1992–2009. Transportation Research Part A: Policy and Practice, 54(August), 49–66.

Gatzlaff, D. H., & Smith, M. T. (1993). The impact of the miami metro rail on the value of residences near station locations. Land Economics, 69(1), 54–66.

Gibbons, S., & Machin, S. (2005). Valuing rail access using transport innovations. Journal of Urban Economics, 57(1), 148–169.

Gibbs, J. P., Halstead, J. M., & Boyle, K. J. (2002). An hedonic analysis of the effects of lake water clarity on New Hampshire lakefront properties. Agricultural & Resource Economics Review, 31(1), 39–46.

Grafova, I. B., Freedman, V. A., & Lurie, N. (2014). The difference-in-differences method: Assessing the selection bias in the effects of neighborhood environment on health. Economics and Human Biology, 13(1), 20–33.

Gu, J., & Jia, S. (2008). The effects of expected transport improvement on housing prices and price spatial distribution- Hangzhou-based evidence of planning the mass transit railway. Economic Geography, 28(6), 1020-1024þ1034.

Heckman, J. J. (1978). Dummy endogenous variables in a simultaneous equation system. Econometrica, 46(4), 931–959.

Hess, D. B., & Almeida, T. M. (2007). Impact of proximity to light rail rapid transit on station-area property values in buffalo, New York. Urban Studies, 44(5), 1041–1068.

Higgins, C. D. (2019). A 4D spatio-temporal approach to modelling land value uplift from rapid transit in high density and topographically-rich cities. Landscape and Urban Planning, 185, 68–82.

Higgins, C. D., & Kanaroglou, P. S. (2016). Forty years of modelling rapid transit’s land value uplift in North America: Moving beyond the tip of the iceberg. Transport Reviews, 36(5), 610–634.

Higgins, C. D., & Kanaroglou, P. S. (2018). Rapid transit, transit-oriented development, and the contextual sensitivity of land value uplift in Toronto. Urban Studies, 55(10), 2197–2225.

Im, J., & Hong, S. H. (2018). Impact of a new subway line on housing values in Daegu, Korea: Distance from existing lines. Urban Studies, 55(15), 3318–3335.

Jiao, Y. (2016). Analysis of the effect of xi’an metro line 2 on housing price along the line on different time point. In Conference: 2nd international symposium on social science.

Jim�enez, J. L., & Perdiguero. (2017). J. Difference-In-differences. In A. Marciano, & G. Ramello (Eds.), Encyclopedia of law and economics. New York, NY: Springer.

Knaap, G. J., Ding, C., & Hopkins, L. D. (2001). Do plans matter? The effects of light rail plans on land values in station areas. Journal of Planning Education and Research, 21 (1), 32–39.

C. Tian et al.

Research in Transportation Economics 90 (2021) 100844

11

Ko, K., & Cao, X. (2013). The impact of hiawatha light rail on commercial and industrial property values in minneapolis. Journal of Public Transportation, 16(1), 47–66.

Kuminoff, N. V., Parmeter, C. F., & Pope, J. C. (2010). Which hedonic models can we trust to recover the marginal willingness to pay for environmental amenities? Journal of Environmental Economics and Management, 60(3), 145–160.

Lancaster, K. J. (1966). A new approach to consumer theory. Journal of Political Economy, 74(2), 132–157.

Landis, J. D., Guhathakurta, S., & Zhang, M. (1994). Capitalization of transit investments into single-family home prices: A comparative analysis of five California rail transit systems. In University of California transportation center working papers (Vol. 42, p. 860), 8.

Lawhon, M., Nilsson, D., Silver, J., Ernstson, H., & Lwasa, S. (2018). Thinking through heterogeneous infrastructure configurations. Urban Studies, 55(4), 720–732.

Liang, Q. (2007). Impact of URT on real estate value: The case of beijing metro line 13. China Civil Engineering Journal, 40(4), 98–103.

Li, H., Wei, Y. D., & Yu, Z. (2016). Amenity, accessibility and housing values in metropolitan USA: A study of salt lake county, Utah. Cities, 59, 113–125.

Li, S., Yang, J., & Qin, P. (2016). Wheels of fortune expansion and property values in Beijing. Journal of Regional Science, 56(5), 792–813.

Lieske, S. N., Nouwelant, R. V. D., Han, J. H., & Pettit, C. (2019). A novel hedonic price modelling approach for estimating the impact of transportation infrastructure on property prices. Urban Studies, 56(11), 1–21.

McDonald, N. C. (2010). School siting. Journal of the American Planning Association, 76 (2), 184–198.

Medda, F. (2012). Land value capture finance for transport accessibility: A review. Journal of Transport Geography, 25(3), 154–161.

Mills, E. S. (1967). An aggregative model of resource allocation in a metropolitan area. The American Economic Review, 57, 197–210.

Mohammad, S. I., Graham, D. J., & Melo, P. C. (2015). The effect of the Dubai metro on the value of residential and commercial properties. Journal of Transport & Land Use, 10(1), 263–290.

Nelson, A. C., & McCleskey, S. (1990). Improving the effects of elevated transit stations on neighborhoods. Transportation Research Record. Journal of the Transportation Research Board, 1266, 173–180.

Oates, W. E. (1969). The effects of property taxes and local public spending on property values: An empirical study of tax capitalization and the tiebout hypothesis. Journal of Political Economy, 77(6), 957–971.

P�aez, A., Scott, D. M., & Morency, C. (2012). Measuring accessibility: Positive and normative implementations of various accessibility indicators. Journal of Transport Geography, 25, 141–153.

Pagliara, F., & Papa, E. (2011). Urban rail systems investments: An analysis of the impacts on property values and residents’ location. Journal of Transport Geography, 19(2), 200–211.

Pan, Q. (2013). The impacts of an urban light rail system on residential property values: A case study of the houston METRO rail transit line. Transportation Planning and Technology, 36(2), 145–169.

Pilgram, C. A., & West, S. E. (2018). Fading premiums: The effect of light rail on residential property values in Minneapolis, Minnesota. Regional Science and Urban Economics, 69, 1–10.

Quan, Y., Liu, Y., & Chen, J. (2006). Some thoughts on urban traffic sustainable development strategy and its several key issues, presented at the annual meeting of China’s urban transportation planning.

Riekkinen, K., Hiironen, J., & Tuominen, H. (2015). The impact of a new subway line on property values in helsinki metropolitan area. FIG Working Week.

Rosenthal, L. (2003). The value of secondary school quality. Oxford Bulletin of Economics & Statistics, 65(3), 329–355.

Sasaki, K. (1990). Income class, modal choice, and urban spatial structure. Journal of Urban Economics, 27, 322–343.

Teng, S., Yan, J., & Zhou, S. (2014). Research on impact of urban rail transit on house price along the line: A case study on No. 1 line of Tianjin subway. In The 2nd international symposium on rail transit comprehensive development (ISRTCD) proceedings (pp. 227–236). Springer Berlin Heidelberg.

Trojanek, R., & Gluszak, M. (2017). Spatial and time effect of subway on property prices. Journal of Housing and the Built Environment, 1–26.

Wang, X. J. (2017). Subway capitalization effect in Beijing: Theory and evidence on the variation of the subway proximity premium. Papers in Regional Science, 96(3), 495–518.

Wen, H., Gui, Z., Tian, C., Xiao, Y., & Fang, L. (2018). Subway opening, traffic accessibility, and housing prices: A quantile hedonic analysis in hangzhou, China. Sustainability, 10, 2253–2276.

Wen, H., Jin, Y., & Zhang, L. (2017@a). Spatial heterogeneity in implicit housing prices: Evidence from hangzhou, China. International Journal of Strategic Property Management, 21(3), 15–28.

Wen, H., Xiao, Y., & Hui, E. C. M. (2019). Quantile effect of educational facilities on housing price: Do homebuyers of higher-priced housing pay more for educational resources? Cities, 90(7), 100–112.

Wen, H., Xiao, Y., & Zhang, L. (2017b). School district, education quality, and housing price: Evidence from a natural experiment in Hangzhou, China. Cities, 66(6), 72–80.

Witte, A. D., Sumka, H. J., & Erekson, H. (1979). An estimate of a structural hedonic price model of the housing market: An application of rosen’s theory of implicit markets. Econometrica, 47(5), 1151–1173.

Xiao, Y., Hui, E. C. M., & Wen, H. (2019). Effects of floor level and landscape proximity on housing price: A hedonic analysis in hangzhou, China. Habitat International, 87(5), 11–26.

Xu, Y., Zhang, Q., & Zheng, S. (2018). House age, price and rent: Implications from land- structure decomposition. The Journal of Real Estate Finance and Economics, 56(2), 303–324.

Yang, J., Chen, J., & Le, X. (2016). Density-oriented versus development-oriented transit investment: Decoding metro station location selection in Shenzhen. Transport Policy, 51, 93–102.

Zhang, X., & Jiang, Y. (2014). An empirical study of the impact of metro station proximity on property value in the case of Nanjing, China. Asian Development Policy Review, 2(4), 61–71.

Zhang, X., Liu, X., & Hang, J. (2016). Do urban rail transit facilities affect housing prices? Evidence from China. Sustainability, 8(4), 380.

Zhang, L., & Yi, Y. (2017). Quantile house price indices in Beijing. Regional Science and Urban Economics, 63, 85–96.

C. Tian et al.