# RECORD 290a4bf33ce230c96c2113bb929c287ce4514c65 # TITLE: Borrowed sizes: A hedonic price approach to the value of netowrk structure in public transport systems # YEAR: 2021.0 VENUE: nan DOI: 10.5198/jtlu.2021.1664 # AUTHORS: Helena Bohman; Desirée Nilsson # PAGES: 17 CANDIDATE TABLE PAGES (1-based): [11, 12] # ABSTRACT: Property prices are known to be higher in places with high accessibility, such as in proximity to train stations and especially to commuter rail, than in places without this access. This study provides a better understanding of how regional accessibility, through the structure of railway networks, can influence local agglomeration economies by providing accessibility to large labor markets. Previous literature has shown a positive impact of proximity to railway stations on housing prices, and our study adds to the literature by analyzing the impact of network structure. We argue that public transport systems can support the benefits of city networks in line with Alonso’s concept of borrowed sizes (1973). Using network theory to measure accessibility provided by the network, we show that stations that provide accessibility to large labor markets across the region are perceived as more attractive by households. Cities in proximity to other cities are strengthened through their public transport links, which allow agglomeration benefits to be exploited by residents. ===== PAGE 1 (score 10, numeric-paren 0, table-words 0, var-words 10) ===== 1 Introduction Infrastructure networks are important for providing accessibility to household and firms. Households need access to jobs, public services and leisure activities. Firms need access to labor, suppliers and cus- tomers. The benefits of being close to other actors, often referred to as agglomeration economies, have been extensively studied in the field of urban and regional economics (Melo, Graham, & Noland, 2009). Investing in infrastructure such as roads and railways has therefore been a common way to increase accessibility and enhance economic activity. Railways offer high-capacity transportation but incur high fixed costs, and there is mixed evidence of the effects of such investments (Debrezion, Pels, & Rietveld, 2007; Higgins & Kanaroglou, 2016; Mohammad, Graham, Melo, & Anderson, 2013). Several explanations to these differences have been mentioned, such as different types of communica- tions and different local context (Lieske, van den Nouwelant, Han, & Pettit, 2019). One reason for the differences is the extent to which investments contribute to achieving agglomeration economies (Chat- Article history: Received: September 6, 2019 Received in revised form: April 7, 2020 Accepted: September 28, 2020 Available online: January 15, 2021 Copyright 2021 Helena Bohman & Désirée Nilsson http://dx.doi.org/10.5198/jtlu.2021.1664 ISSN: 1938-7849 | Licensed under the Creative Commons Attribution – Noncommercial License 4.0 The Journal of Transport and Land Use is the official journal of the World Society for Transport and Land Use (WSTLUR) and is published and sponsored by the University of Minnesota Center for Transportation Studies. T J  T  L U http://jtlu.org V. 14 N. 1 [2021] pp. 87–103 Helena Bohman Malmö University and K2–The Swedish Knowledge Centre for Public Transport helena.bohman@mau.se Désirée Nilsson Malmö University and K2–The Swedish Knowledge Centre for Public Transport desiree.nilsson@mau.se Borrowed sizes: A hedonic price approach to the value of network structure in public transport systems Abstract: Property prices are known to be higher in places with high accessibility, such as in proximity to train stations and especially to commuter rail, than in places without this access. This study provides a better understanding of how regional accessibility, through the structure of railway networks, can influence local agglomeration economies by providing accessibility to large labor markets. Previous literature has shown a positive impact of proximity to railway stations on housing prices, and our study adds to the literature by analyzing the impact of network structure. We argue that public transport systems can support the benefits of city networks in line with Alonso’s concept of borrowed sizes (1973). Using network theory to measure accessibility provided by the network, we show that stations that provide accessibility to large labor markets across the region are perceived as more attractive by households. Cities in proximity to other cities are strengthened through their public transport links, which allow agglomeration benefits to be exploited by residents. Keywords: Hedonic price model, network theory, centrality measures, labor market accessibility, commuter rail ===== PAGE 11 (score 27, numeric-paren 0, table-words 4, var-words 13) ===== 97 Borrowed sizes: A hedonic price approach to the value of network structure in public transport systems The descriptive statistics of the data in Table 3 provide some key insights into the region of Scania. About two percent of the single-family houses sold in 2016 had direct access to a beach, and the average house had a size of 117 square meters and was about 25 years old. As mentioned before, there are large differences in the geography of the region. There are also large differences in the average number of pas- sengers per day travelling from the different stations in the railway network. Table 3. Descriptive statistics (variables in natural logs) Variable Mean Median Std. Dev. Min Max PRICE 14.508 14.631 0.77 0 18.414 SIZE 4.764 4.796 0.38 1.386 6.659 AGE 3.64 3.892 1.172 0 5.756 DEED 0.963 1 0.189 0 1 BEACH 0.018 0 0.134 0 1 INC 12.412 12.411 0.149 11.853 13.082 POP 7.117 7.209 0.845 2.565 9.077 LOC_LABOR 2.667 2.33 1.191 1.087 5.087 STATION_DIST 1.547 1.609 1.079 -3.671 3.47 PASSENGERS 7.534 7.294 1.419 4.29 10.682 CLOSENESS (GRAVITY) 7.812 7.936 1.342 5.142 10.702 DEGREE 4.295 4.484 0.527 2.877 4.936 BETWEENNESS (Öresund trains) 3.183 3.466 0.659 1.099 3.989 BETWEENNESS (Pågatåg trains) 0.025 0 0.044 0 0.178 Table 4 shows the results from the HPM. Columns (i) to (vi) show the results as station character- istics are successively supplemented. Estimates using the base ordinary least squares model provided sig- nificant results for robust Lagrange multiplier testing for both error and lag. However, the test statistics provided higher value for a spatial error model to be performed (Osland, 2010). A spatial error model (SEM) was applied to all regression models using generalized method of moments (GMM) estimations. The weight matrix was calculated using Euclidian distances, with a default bandwidth estimated using GeoDa Space software (Anselin, Syabri, & Kho, 2006). Standard errors were estimated based on Kele- jian and Prucha (2010). Using SEM means that the estimated parameters will be efficient since lambda accounts for any misspecification as well as spatial autocorrelation in the models. The ordinary variables used in HPM turn out as expected and are stable across the different model specifications used. Income and population in the local small area have a positive effect on price, as well as the number of employments in the municipality where people live. The number of jobs measure local agglomeration effects and show an elasticity of 0.05-0.06. The variable distance to station are as expected negative, with an elasticity of around -0.01. Station and network characteristics present interesting results. PASSENGERS show positive pa- rameters over all specifications, although the parameters fall somewhat as more variables are added to the model. Adding CLOSENESS to the specification (column (ii)) improves the PseudoR2 to 0.479. The estimated coefficient represents a positive impact on sales prices of single-family houses. Thus, the significant result of this variable reflects that it measures the importance of accessibility to workplaces in other nodes of the network. In accordance with Osland and Thorsen (2008) our results show that ===== PAGE 12 (score 107, numeric-paren 79, table-words 2, var-words 7) ===== 98 JOURNAL OF TRANSPORT AND LAND USE 14.1 accessibility to workplaces has a positive impact on housing prices. As the variable is a weighted mea- sure of employments in the vicinity of destination stations, and distance is measured using the rail infrastructure, it also indicates the perceived utility of using public transport for work. This suggests that household value both the local agglomeration as well as the agglomerations provided through the network. Adding the variable DEGREE, which measures direct access to other stations, the estimated parameter also turns out significant and positive with an elasticity of around 0.09. Thus, the number of stations that you can reach without changing trains has a significant impact on the price of single-family houses. The relevance of not having to change trains in order to get to destinations in the network seem to enhance the importance of closeness to labor market in other nodes of the network. Table 4. Results from HPM regressions using spatial error model. Z values in parentheses, *** p<0.01, ** p<0.05, * p<0.1. N=18,273, all continuous variables in log form (i) (ii) (iii) (iv) (v) (vi) CONSTANT 3.139 (6.732)*** 2.654 (5.967)*** 2.329 (5.248)*** 2.293 (5.036)*** 2.240 (4.932)*** 2.223 (4.848)*** SIZE 0.787 (68.211)*** 0.787 (67.918)*** 0.789 (68.015)*** 0.789 (68.157)*** 0.789 (68.033)*** 0.789 (68.160)*** AGE -0.021 (-5.367)*** -0.021 (-5.353)*** -0.021 (-5.484)*** -0.021 (-5.467)*** -0.021 (-5.497)*** -0.021 (-5.488)*** DEED 0.367 (18.589)*** 0.367 (18.559)*** 0.366 (18.538)*** 0.366 (18.536)*** 0.366 (18.563)*** 0.366 (18.559)*** BEACH 0.413 (13.019)*** 0.413 (13.015)*** 0.415 (12.998)*** 0.415 (12.986)*** 0.415 (13.000)*** 0.415 (12.988)*** INC 0.532 (15.052)*** 0.512 (14.107)*** 0.525 (14.443)*** 0.523 (14.466)*** 0.528 (14.482)*** 0.526 (4.534)*** POP 0.043 (5.114)*** 0.041 (4.783)*** 0.042 (4.932)*** 0.042 (4.938)*** 0.043 (4.955)*** 0.042 (4.973)*** LOC_LABOR 0.062 (5.739)*** 0.055 (5.187)*** 0.060 (5.687)*** 0.061 (5.686)*** 0.061 (5.758)*** 0.062 (5.732)*** STATION_DIST -0.016 (-2.431)** -0.013 (-2.118)** -0.013 (-2.124)** -0.013 (-2.076)** -0.013 (-2.121)** -0.013 (-2.082)** PASSENGERS 0.034 (6.644)*** 0.029 (5.886)*** 0.019 (3.516)*** 0.026 (3.703)*** 0.020 (3.566)*** 0.026 (3.563)*** CLOSENESS (Gravity) 0.187 (2.991)*** 0.172 (2.733)*** 0.175 (2.855)*** 0.184 (2.937)*** 0.181 (2.864)*** DEGREE 0.089 (3.843)*** 0.087 (3.789)*** 0.092 (3.977)*** 0.090 (3.933)*** BETWEENNESS (Öresund train) -0.303 (-1.173) -0.291 (-0.988) BETWEENNESS (Pågatåg train) -0.742 (-0.797) -0.260 (-0.239) Lambda 0.935 (95.694)*** 0.928 (89.860)*** 0.921 (83.961)*** 0.921 (83.558)*** 0.920 (83.278)*** 0.921 (83.099)*** AIC -17706.67 -19279.98 -19626.16 -19751.39 -19702.72 -19758.01 Pseudo R2 0.449 0.479 0.509 0.510 0.512 0.512 ===== PAGE 13 (score 11, numeric-paren 0, table-words 1, var-words 9) ===== 99 Borrowed sizes: A hedonic price approach to the value of network structure in public transport systems The two variables representing BETWEENNESS capture transfer possibilities in two types of re- gional public transport networks. The Öresund train network stops at fewer stations allowing for direct and faster mobility between nodes, while the Pågatåg network covers more stations in the region provid- ing commuter trains to also smaller places. Both betweenness measures turn out insignificant. In the context of regional train networks, transfer possibilities in nodes does not seem to impact the perceived value of single-family houses. This article addresses the importance of regional public transport system and access to agglomera- tion economies in regional economic development. Most previous studies that consider the importance of accessibility using hedonic price model performs it in a more local context. The regional setting in this study incorporate more heterogeneity compared to local settings, which explain the lower fit of the model. AIC statistics are provided as a complement to the pseudo r2, and indicate that the variables added contribute to the model specification, and improve as the network variables are added. 5 Discussion This study aims to provide a better understanding of regional development by focusing on how network structure and node characteristics of public transport network influence willingness to pay for housing. By introducing measures based on network theory, we can analyze not only local agglomerations, but also how the network can help exploit agglomerations accessible through the network. Thus, by using measures that capture network characteristics we can better understand how regional accessibility can influence local agglomeration economies. Four characteristics have been explored by using three centrality measures and the size of stations in terms of passengers travelling. These variables are not highly correlated among themselves, which in- dicates that they capture different aspects of the network. Among the control variables in the base speci- fication, number of employments in home municipality is included. Including network characteristics provide measures of access to regional agglomerations in local markets. Although the explanatory power of the models are lower than most HPM estimates, the findings are robust across the specifications. Results show that the network variables improve the fit of the model and thus contribute to the variables used in the base model. In the words of Alonso (1973), local markets can benefit from borrowed ag- glomerations accessible through the public transport network. The first network variable introduced was the closeness measure, applied as a gravity specification, and it captures accessibility to labor markets, as measured through the indirect effects offered by the re- gional rail system. The estimate of this variable implies an elasticity of 0.18 - 0.20, all else equal. Using our definition of this measure, the stations with the lowest gravity measures provide a labor market of slightly less than 20,000 employed people. The station with the highest corresponding number provides accessibility to approximately 214,000 employments. As an example of how important this aspect is in economic terms, one can compare a house close to the station providing the least accessibility (low close- ness score) to one demonstrating the highest accessibility (high closeness score). A house by the station providing the largest accessibility would then sell for nearly three times the price of an identical house by the station of the lowest accessibility. This effect is in addition to the effect of the local labor market which is already controlled for. In other words, although the exact figures should be interpreted with caution, the findings do suggest an important contribution of the network. The second centrality measure (Degree) captures the possibility of reaching stations without chang- ing trains. We interpret this as a trip quality indicator; measuring to what extent a station offers uninter- rupted trips across the network. As expected, this feature has a significant impact on property values with an elasticity of around 0.09. The positive impact of direct connections is not surprising. Given that