What is the effect of the redevelopment of a train station on surrounding residential property prices? : an analysis across the Netherlands

(1)

Master Thesis:

MSc Business economics - Finance & Real Estate Finance

What is the effect of the redevelopment of a train station on surrounding

residential property prices? An analysis across the Netherlands

Student name: P.C. Ten Holter Student number: 11132701

Master: Business economics - Finance & Real estate finance (dualtrack) Thesis supervisor: Dhr. Prof. Dr. J.B.S. Conijn

(2)

Abstract

In the Netherlands train stations are constantly being renovated to cope with the expected growth of travelers by train. In this research the effect of redevelopments of train stations on surrounding residential property prices is tested. Fourteen municipalities where train station redevelopments have been conducted are examined. Train station renovations of Alphen aan de Rijn, Amsterdam Bijlmer-Arena, Apeldoorn, Boxtel, Den Bosch, Den Haag Hollands Spoor, Enschede, Hilversum, Leeuwarden, Leiden, Lelystad, Nijmegen, Rotterdam and Zutphen are used in the research. The renovations are not examined individually but a general effect of a train station redevelopment is tested. In order to test the differences in transaction prices a dataset of residential property prices between 1995 and 2015 in these municipalities is used. It is found in this research that houses within a radius of 800 meters from train stations generally are sold with a small discount of around 0.836%. The effect of the redevelopment of a train station on residential property prices compared to the same group within 800 meters before the renovation is positive. In the period during the renovation a premium on transaction prices of 1.67% is given in general. Residential properties get a premium of around 4.62% on the transaction price after renovation.

Statement of originality

This document is written by Pieter Christian Ten Holter who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

(3)

1. Introduction

More than 1 million people a day are travelling through one of the 410 stations that are located across the Netherlands (NS, 2015). This amount grows with approximately 3% per year and the prediction is that this growth will continue in the future (Rijksoverheid, 2016). Because many existing train stations are too small to cope with enormous amount of people traveling, train stations have to be adjusted for this. This means that one of the consequences of this extensive travelling is that development of new train stations and renovations on existing train stations is constantly being planned and executed. Currently, the Dutch government invests over 1 billion euros in an enormous project only concerning the six biggest train stations in the Netherlands. These projects, called the ‘Nieuwe Sleutelprojecten (NSP),’ are partially being made for the use of a newly constructed high-speed rail to realize a better international railway network. In order to realize this, a handful of important Dutch railway stations are being renovated and extended to cope with the additional expected increase of travelers. Currently, some of these train stations are already finished while others are still being finished at the moment. However, the stations that are on the list of this enormous Dutch railway renovation project are not the only railway stations that are being renovated at the time. Again, to deal with the general growth of travelers by train anywhere in the Netherlands, renovation projects of train stations are constantly being planned and carried out.

Renovations of train stations often come with newly developed or extensively renovated station buildings. This development can be a contribution to the appearance of this part of the city. One great example of a big improvement to a city is the recently developed train station of Rotterdam. The design of the newly constructed so called ‘railway cathedral’ has won the architectural price for the best Dutch building in 2015 because of its beauty (Volkskrant, 2015). Moreover, the focus in these projects not only lies on a better accessibility of a city and its surroundings but also on sustainability of the buildings, comfort and amenities, accessibility for handicapped people and an increase of the safety in stations (Prorail, 2015). The function of a railway station therefore switches from a place where trains arrive or pass by, to an important central place connecting several areas, which provides services and amenities to fulfill travelers demand. With this new purpose of many railway stations the value for this central place within a city has been changed. It can be seen as an important central motor of a city where a lot of economic activity takes place. The improvement of the safety of stations is another factor that makes a station a more pleasant environment to spend time. These improvements combined make the surrounding area of a railway station a very interesting place to do business because of the great accessibility of this area but also because of the surrounding facilities. Therefore, these improvements of existing railway stations probably not only increase the function of doing business in this area but it could also increase the function of living in this area. This increase can be measured in an upward change in property value of the surrounding area of a renovated train station. This is exactly what is going to be researched in order to test this prognosis.

However, despite the positive prospects within cities due to these extensive renovations of train stations there are some divided opinions in the Netherlands about the positive impact for the city center. Wouter Jan Verheul (2015), a researcher of the Technical University Delft argues that the composition of the amenities in the train stations changes the local and cultural economy of these places. This is because he claims that although in the newly developed station areas there is

(5)

often more space created for retail business, the traditional local shops tend to disappear. Numerous multinational shop businesses such as Starbucks and Burger King are opening new braches in train station buildings due to large-scale contacts they signed with the National Dutch railway company (NS) for several different stations. Another drawback for the preservation of the unified city center is the recently introduced public transport chip-card in the Netherlands. The public transport chip-card enables the NS to close the gates of train stations. One of the consequences is that people only can reach the other side of the station by checking in and out. However, for people not carrying a public transport chip-card it is impossible to reach the other side of the station by trespassing the station hall. As a result of this the city is split into two parts according to Verheul (2015). Even though, this is not a direct consequence of the renovation of a train station, it might impact the value of the surrounding property values of one side due to the decrease accessibility of that side. In many cases the commercial city center is located at one side of a train station in the Netherlands so this effect has to be taken into account.

Another negative aspect of the projects for the train stations that are plotted in the whole of the Netherlands is that there has been a lot of delay and additional costs with the construction of the train stations (De Gelderlander, 2016). Some of the projects that recently were completed were considered under budgeted in the first place. This miscalculation made projects last longer and this goes together with more nuisance. For instance, in Arnhem the design of the building was too hard to develop with the original budgeted amount of money. Since no construction company was willing to build the design with the available budgeted amount of money the project was slightly delayed in the first place. Thereafter, due to complications in the construction of the building of the station it delayed even more. Up till now there are still problems with the recently built station, its roof is still leaking despite the completion and opening of the station was already more than a half year ago (De Gelderlander, 2016). Another example of a station exceeded the budgeted amount is Den Haag Central Station, that turned out to be more expensive because the NS did not wanted to stop the rail traffic in and around the station. So far, the only project that remained in the budgeted amount and timeframe in the NSP is Rotterdam central station; ‘a project that can be referred a role model for the development of stations within the whole country’ according to the State Secretary of infrastructure and environment, Wilma Mansveld (NOS, 2015).

As can be seen, the prospects for the development of train stations can be considered divergent. On the one hand the renovation of a train station makes the central part of a city better. Often the renovation of train stations ensures that improvements such as safety, better accessibility and an increase of amenities are realized. Also, it often means that the construction of a new building is realized as well, this may improve the streetscape of that particular area; a possible side effect that may influence the price of surrounding properties. Unfortunately it also can take some negative aspects into account like the separation of the city center or severely delayed renovations with a lot of construction noise. This may result in a negative influence on the surrounding property value.

However, the fact is, that despite the continual constructions on existing train stations across the Netherlands, there is nothing conducted yet in this particular area of research. The reason why this event is going to be researched in this thesis is because it is very relevant since these enormous renovation projects seem to be quite ordinary in the Netherlands. Therefore, the main

(6)

focus of this research is to measure the impact of a renovation or renewal of an existing train station on surrounding property values.

1.1 Research question and hypotheses

The research question that can be conducted out of this introduction is:

‘What is the effect of the redevelopment of a train station on surrounding residential property prices? An analysis across the Netherlands’

Consequently, the aim of the research is to investigate if the renovation of an existing train station has a direct noticeable effect on residential property value in the neighborhood. Here the average overall effect is measured for areas around train stations in the Netherlands that have been renovated. Thus, train station renovations are not tested individually in the main research but as a group.

In combination with the research question and the aim of the research the following hypotheses are compiled:

Hypothesis 1: The presence of a train station has a positive impact on the value of

surrounding residential property if no construction takes place yet

Hypothesis 2: The redevelopment of a train station has a negative effect on the

surrounding residential property prices compared with before the renovation

Hypothesis 3: The completion of a renovated train station has a positive effect on the

surrounding residential property prices compared with before the renovation

It is expected that the presence of a train station nearby could have several benefits and therefore it would give a premium to the houses located nearby. Primarily, the presence of a train station will increase the accessibility of the neighborhood it is located in. Furthermore, train stations often contain amenities such as restaurants and shops. This can be seen as a great advantage due to the convenience it offers. Mainly because of convenience advantages like these it is expected that this will give a premium to the houses located in this area. Another possible outcome is that nuisance caused by the users of train stations affect the surrounding housing prices in a negative way. However it is expected that the positive effects of train stations mentioned, will overrule the negative effects.

Secondly, it is assumed that construction projects like the redevelopment of a train station generally bring a lot of nuisance and therefore will provoke a discount on residential property prices compared to the prices before the renovation. Possible examples of nuisance are obstructions such as roadblocks or the construction noise caused by the project. Because the renovation of a train station regularly means that surrounding projects are also taken care of (see table II) it is expected that this will bring a lot of discomfort into the area that results in a less attractive area to live at the moment of redevelopment. Therefore it is expected that the redevelopment of a train station has a negative effect on the surrounding residential property. Lastly it is expected that the completion of the renovation of the train stations will have a positive effect on the housing prices compared to the period before the renovation. Since the

(7)

completion of a train station vanishes most discomfort caused by the constructions, the nuisance is expected to be less. It also is likely that renovated train stations can improve this part of the city because it often leads to an increase in amenities and comfort. Also the attractiveness of the area is likely to improve because improved train station can contribute the appearance of this area of the city. Furthermore, it is the idea that the importance of redeveloped train stations increases because they are often adjusted for the growth in travelers by train.

As described, the overall effect is going to be measured over the selected train stations. The purpose of the research is to see if there is a general effect of the development of a train stations in the Netherlands. The hypotheses are set up in this way because it is interesting to test for the different phases of the renovation. Therefore it will be examined if the presence of a train station has a positive impact on the surrounding property prices at all. This is going to be tested by exclusively using the time period before any train station renovations. By comparing the transaction prices in the areas around the train stations with the areas located outside it is possible to test this. Secondly, comparing the transaction prices of the nearby residential property before the planning and construction with the transaction prices during the construction will be the way in order to test the second hypothesis. A comparable approach is used to test the third hypothesis; only a different time frame is chosen, the time frame after the renovation. In the second and third hypothesis the benchmark is the period before the renovation and the test periods are during and after this renovation.

It has to be taken into account that it is possible that in the very near distance of a train station an extension or renewal of the train station has a negative effect due to noise disturbance. This might be nuisance caused by travelers, trains or both. It can be that because of this nuisance the very nearby houses are increasing less, or even decreasing in value while houses a little bit more far away are better off with the renewal of the station. Although this could be the case, the expectation is that the net effect of the renovation of a train station is positive. It is expected that the positive effects of the train stations will overcome the negative effects as explained before. The continuation of the thesis is constructed the following way. Firstly, related literature concerning the effect of railways and train stations on surrounding residential property values is discussed to have a clear view about the pros and cons of a train station. Thereafter, the methodology and data description and are being evaluated. In this research the whole process of the development of a couple of train stations inside the Netherlands is captured. At first it is examined what the actual effect of the existence of a train station has on the surrounding property value. This is, when there has been nothing changed yet in terms of renovation of the existing train station. This effect is going to be estimated using a hedonic price model. The areas where a recent train station development has been executed are going to be observed if there is already a difference in residential property value nearby and further away located from the station. Although, this cannot be interpret as a causal effect. Following, different events in time are taken to compute the change in residential value where first the event of planning and building is going to be captured. Next, the event of completion is going to be captured to see what the difference is between the value during the construction and the value after the construction. After this, the two outcomes are going to be compared with each other to see the net value increase or decrease.

(8)

The hypotheses will be tested with an empirical research using the NVM database. In this dataset the transaction prices and housing characteristics of fourteen municipalities in the Netherlands of the time period from 1995 to 2015 are captured. Every municipality used in this research contains a train station that has been renovated in the time period from 1995 until 2015. The data about the redevelopments of the train stations combined with the data of the NVM are used in order to test the research question. Since the Dutch railway company: de Nederlandse Spoorwegen (NS) has no data of redevelopments saved in a database, dates of completion of renovations and dates of the start of a renovation are found on several websites. These dates are essential to measure the difference of the periods of construction in a station area. This research is going to be performed with the data analysis statistical software: Stata. The findings will be reported in the paragraph after the methodology is discussed. Furthermore a conclusion is created in order to sum up and summarize what the overall effect is.

2. Related literature

Within the existing literature the impact of renovations of train stations on residential properties is not really being dealt with yet. As can be seen in this part of the thesis there has been conducted a lot of research about the relationship of railway stations and its surrounding property value. Despite the fact that the focus in this research primarily lays on the development of train stations, other railway developments will be discussed in the related literature as well. The central place theory and the hedonic price model are discussed in this part as well.

2.1 Central place theory

The central place theory introduced by Walter Christaller (1933/1966) based on the pattern of cities within South Germany is a way to look at the system of cities. Using central place theories, it is possible to determine the value of a certain piece of land by making a trade-off between the transportation costs to get there and the rental costs of that location. The central place theory is further is elaborated by Alonzo (1964) and Mills (1972). The key concept of the theory is that the land value is higher in areas of interest. If we look at a city in a simple way, we only see one city center. This is the place where the economic heart of the city is located. Thus, if we work with the concept of a monocentric, we see that a city has one center where all economic activities take place (Geltner and Mills, 2013). Consequently, this area is the place to be for economic purposes. For people who live in the center of the city it is clear that the transportation costs to get there are at the minimum level, probably for free because the center is at walking distance. Therefore, central places are areas where the land has the highest value because no transportation costs are needed to get there. This means that at the spot where the transportation costs are the lowest to get to the central place of a city, (which can be a city center or a central business district) the costs that people pay for the location are at their maximum point. In the central place theory this tradeoff is leading for the calculation of the property prices.

In the Netherlands, one of the easiest ways to get to the central place of a certain city is by using the train. Most central train stations of cities in the Netherlands are located in the very center of that particular city. Because many cities in the Netherlands contain several train stations only at the larger stations of a city will be observed in this thesis. The reason for this decision is twofold. On the one hand many people make more use of a bigger station. This makes it more interesting

(9)

to look at bigger stations because the probability of a large price difference in property within the station area compared to the property prices outside of the area is larger. Property prices around train stations in the Netherlands that are used intensively do have a higher premium than houses around a quiet train station (Debrezion et al, 2006). Furthermore, the renovations of large train stations are more often extensive. The expectation is that this makes the event more powerful. If that is the case, the event will probably be good to measure a change.

2.2 Railway stations and residential property

Debrezion et al (2006) conducted a study where it has been researched what the presence of a train station does with the surrounding residential property prices in the Netherlands. Because in the Netherlands there exist many train stations with different magnitudes it is important to make a good distinction in the stations. In order to do this they first created an index that reflects the rail service quality of a certain station. This rail service quality index (RSQI) includes certain characteristics of railway stations concerning the availability of that particular train station. Indicators such as the daily frequency of trains, the intercity status of the station and the number of destinations from that station are included. Another indicator is added to this index: the generalized journey time (GJT_ij). The GJT_ij is a function that computes the generalized travel time between stations i and j. Waiting time, in vehicle time and also transfer time are included in GJTij. The lower the GJTij is, the better it is for the RSQI of a station. The higher the RSQI-rate, the more services and better connections a train station has. The main focus of this article is to investigate the effect of railway accessibility on house prices. They focused mainly on two characteristics of the stations: the nearest station and the most frequently used station in the neighborhood. At first, in order to test this they used the utility function of a traveler, derived by using different transportation ways to get to the stations (bicycle, car, walking or public transportation) and the availability of parking their transportation vehicle if they use a car or a bike. Also the RSQI index is incorporated into the empirical research used. With the empirical research by using a hedonic price model they found that the house price is higher around train stations and that especially the frequency and usage of a railway station has a large impact of the house prices (Debrezion et al, 2006). It shows that a doubling of the frequency of train service at a certain station drives up the price with a mean of 2.5% in the area around the station. Hence, in this research it is significantly showed that the existence of a train station influences the price of the surrounding housing prices however this is depends on the magnitude and importance of the train station. This means that some stations are competing with each other due to the difference in rail service quality.

Similar research, only in a more undeveloped country has been conducted by Celik and Yankaya (2006). In this paper the effect of retail transit investments on real estate price in Turkey is studied. By using both a linear and exponential hedonic price model, they measured what the direct effect the placement of rail transit investment has on residential property in Izmir, Turkey. An interesting instrument they have added in their research is the capitalization of the value of travel time by using walking distance. They found that with every meter away from a subway station the price per square meter decreased in the whole area researched. This finding was highly significant in all equations in the research. They conclude that in the whole area, every meter away from a station the price of residential property decreases with 4.76 dollar. Consequently, the main finding in the paper was that the walking distance to the station was

(10)

negatively correlated with the transaction price of the property; this indicates that the effect of a railway station has a positive effect on residential property values, which is a similar outcome like in the paper of Debrezion et al (2006).

Despite the conclusions of the two papers discussed above, Andersson et al. (2010) found other outcomes in their research that was performed with data from Taiwan. Their results did not directly imply a land value increase of property that was placed near metro stations. In the article several explanations are given for this. One reason could be that the price to make use of the high-speed line in Taiwan is too high for citizens. Costs for ticket for the month from Monday till Friday were around 70% of the median monthly salary of the Taiwanese people in 2008. In comparison with the same tickets between cities in Western Europe this percentage is around 10% of the median monthly salary. Therefore a reason could be that it is not very attractive to live near a station because they will not use it anyway. The result of this development is that there is more traffic noise while the positive sides of having a station around are not that essential because it is too expensive to make use of it. Still, the authors claim that this does not directly imply that the placement of the high-speed line has no effect at all.

Comparable conclusions are made in research about the improvements of the railway network in Hong Kong to create a modern and efficient railway system (Chau & Ng, 1998). The electrification in 1983, a big modernization in the railway network, was the event that has been used to measure if it changed the surrounding residential value. A really difficult task in this research was that it was difficult to control for other factors that possibly affected the price gradient along the line of the railway where the research was done. Because of this problem they used a combination of a hedonic price model and they looked at comparable houses to estimate a price change. They further looked to two different railway stations and the residential price change around these stations. Hereby one station was very close to the central business district while the other was much further away. Eventually they found that because of the better connection between the different city parts the price gradient was negatively affected by the improvement of the rail. Although this seems contradictory in a way, the central place theory, mentioned before, supports this finding because the costs to travel are reduced. In this case the factor travel time, a component of the travel ‘costs’ is reduced due to the development of the railway. A drawback of this research is that they did not look at places more distant from the stations. Only one building is chosen to compare the transaction prices before and after the renovation. They did not looked at differences in values of other areas.

In the following research different areas are being compared performed with data from investments in an urban railway network in Naples. Also, it mentions the image effect of the presence of a modern and dynamic station at an area into account (Pagliara & Papa, 2011). The focus of this research is the land-use impacts by looking at residential value changes due to the development of newly built stations. The treatment group of their research is the collection of residential property located within 500 meters from the railway station. The control group lies outside of this radius of 500 meter. A comparable methodology is used in this thesis as research approach, however an area within 800 meters around new stations is chosen as the treatment group and the control group is the batch of houses outside this area. This is because in the study of Pagliara and Papa it is about a metro line while in this research train stations are chosen. The main outcome of this study was that residential property increased in value around new stations.

(11)

The drawback of all these studies mentioned above is that it only has one moment in time where the difference in prices has been measured. This is not entirely fair because it is quite obvious that when the construction or development of a railway network is executed this takes a lot of noise and other nuisance with it.

Fortunately, there is a study that captures the whole process of the construction of a railway network. The whole process is broken down into different phases. This is a research that has been executed about whether the placement of a new light rail system has had an effect on transaction prices on single-family homes in North Carolina (Yan et al., 2012). The rail network itself already existed because the line was used by freight trains in the past. Yet the line had to be adjusted for the construction of the light rail system. The total timeline of the research was 11 years, from 1997 till 2008, and it is divided in four different time periods. The periods are as follows: the period of pre-planning of the light rail system, planning, construction and operation. Firstly, the planning of the rail system had a negative effect on the house prices, the transaction prices of the homes were the lowest of all periods, however thereafter the three remaining time period the adjusted house prices became higher. Thus, the house prices seemed to react positively on the new light rail system in the end. These findings imply that the presence of a railway station indeed increases the surrounding residential property values.

2.3 Side effects of renovations of railway stations

To extend this literature, some possible side effects of rail transit stations on residential property values have been identified as well. There are some positive effects that may give residential property that are located near a rail station a premium. Effects such as retail activity, direct access advantage are effects that are considered positive. However, the existence of a train station might entail negative external effects as well. Noise and pollution effects and neighborhood crime can lead to a discount on the price of a property. These long-term effects of the existence of a train station are examined in a study of Bowes and Ihlanfeldt (2001). They used a hedonic price model based on data of transaction prices in single-family homes in Atlanta in the USA. In the model locational characteristics such as crime and retail activity are taken into account. Also median neighborhood income information and the distance to the central business district is used. They state that it is possibility that within the literature there exist many differences in outcomes (something that is also shown in this literature review) because of the lack of taking into account the simultaneous and competing effects that come with a railway station. Some interesting results are that people are willing to pay more for a house near a rail station if it is located more outside of the central business district. Negative effects of railway stations are mainly found near the central business district. Also it seems that the existence of a parking lot next to the station increases the crime in that area. Furthermore it shows that retail activity increases when a station is placed in a neighborhood with a lower median income. In summary, Bowes and Ihlanfeldt (2001) showed that the total effects are heavily dependent on which neighborhood the railway station is placed.

In addition to these findings about nuisance, Theebe (2004) wrote a paper about traffic noise and the impact on housing prices in the Netherlands. The noise of airplanes, trains and automobiles is mainly examined in the research. It is shown in a map in the paper that in Amsterdam the places with the most noise are, logically, places directly neighboring freeways, rails and airports. There is corrected for possible positive effects in terms on accessibility for properties located

(12)

near a train station or a highway ramp by including the distance to the nearest one. With data analysis it is found out that there is a relationship between distance to a train station/highway ramp and property prices. This effect is until 3 kilometer at maximum though. With the use of a spatial regression model the research is obtained. It is found that traffic noise has a significant impact on value of houses in the Netherlands. This noise expressed in decibel has a negative effect on residential property. Noise levels above 65 decibel are negatively capitalized into house prices. The premium can run up till a discount of 12% due to this noise, according to Theebe (2004). For houses that are located in an area with traffic noises between 41 and 65 decibel, the exact noise level is not important. Houses that are located in a quiet neighborhood (below 40 decibel) can be sold at a premium with a maximum of 6.5%. The finding of this paper implies that the construction but also the presence of a train station can have a negative effect on house prices due to the noise that is caused.

Now it is researched in the existing literature what different studies have shown about the existence of railway stations and effects on the surrounding property prices. In this study though it is also important what the effect of a renovation has on surrounding houses because this is going to be the case in this research. This could be illustrated with a recent research that has been done by Shill et al. (2002). Here it was investigated what the effect of the revitalization of a neighborhood was on the value of the properties in various neighborhoods in New York. The project analyzed in this paper is the ‘Ten-Year plan’ of New York. This project encompasses the largest subsidized housing construction plan in the United States in history. The main focus in the research is whether there is a positive spillover effect in this project. The thought is that projects like these have a positive impact on for example distressed neighborhoods in terms of a spillover effect. Schill et al. (2002) used a hedonic price model to explain the transaction price in terms of house and neighborhood characteristics. In order to measure the spillover effect the properties that are located near place where the neighborhood was revitalized was compared with properties that were not. The conclusion in the research is that a revitalization project of a neighborhood has indeed positive spillover effects. Houses increased in value obviously related with the revitalization project. This could mean that when a renovation is taken place in a certain area, the nearby-located area gains from this because of a spillover effect. This is relevant for this research because when a train station is renovated in the Netherlands, the surrounding area is often developed and this can lead to a positive spillover effect as well.

It is relevant as well to examine whether the presence or development of shops have a positive impact on the surrounding property value. As is described in the introduction of this thesis many train stations gain more retail after a development. In a research of Sirpal (1994) it is examined if the presence a shopping center has an effect on the surrounding residential property value. Also it the size of the shopping center is taken into account. Three different sized shopping centers located in Gainesville, Florida are used for this research. Data of surrounding properties is collected and examined up to approximately 1 kilometer (3000 feet) outside the radius of the shopping center. This because with this distance it is possible to compare different distances within this radius and thereby comparing if there are differences in prices within the radii of the three sized shopping centers. Sirpal (1994) used a hedonic price model and excluded highly correlated variables. For example, every neighborhood surrounding a shopping center has a similar quality and therefore the variable quality could be excluded in the test. Hereafter multiple

(13)

regressions were tested using different forms of models. The outcome of this research is that the presence of a shopping center has a positive effect on the surrounding residential property. Also the size of the shopping center has a positive coefficient. This means, if a shopping center is bigger this would have a larger positive impact on the house prices surrounded according to Sirpal (1994). The outcome of this research can be seen as a positive effect of the renovation of a train station since the renovation of a train station usually involves an increase in retail activity in this area.

3. Data

3.1 Train stations & renovations

In the empirical analysis of this thesis roughly two main datasets are used. The first dataset concerns the list of train stations that have been extensively renovated in a period from 1998 and 2014. This is a timeframe where it is possible to measure if it has any effect on the transaction prices of the surrounding residential real estate of the three stages that are going to be caught in this thesis; the period before, during and after the renovation. Train stations of the national Dutch railway company (NS) are all located in the Netherlands. The cities Alphen aan de Rijn, Amsterdam, Apeldoorn, Boxtel, Den Bosch, Den Haag, Enschede, Hilversum, Leeuwarden, Leiden, Lelystad, Nijmegen, Rotterdam and Zutphen are incorporated in this research. In total 14 train stations are chosen which are distributed throughout the Netherlands and all are located in different cities (see figure I).

Figure I

(14)

The train stations in 12 of the 14 cities are the biggest and most centrally located stations. The reason why the most central stations are selected is because it is important to measure the renovations of essential train stations for a city. This is partly chosen because of the research conducted by Debrezion et al. (2006). It has become clear that the train stations with a high RSQI and consequently an important status had more influence on residential property than the train stations that have a less important status. In the Netherlands many cities only have one essential station. Some of the bigger cities have several stations of this kind. Despite the fact that the stations of Den Haag and Amsterdam are not the biggest or most centrally located stations, these train stations are very important. In Amsterdam, train station Amsterdam Bijlmer-Arena is chosen which is the most central station of the area of Amsterdam South-East. The station is accessible with public transport such as the tram, metro and bus, next to obviously by train. It contains 10 shops and restaurants (NS.nl) and the amount of people who were getting in and out were 22,684 per day in 2015 (Treinreiziger.nl). The station in Den Haag is the oldest station of the city. It contains a tram and bus network and it is located near the center and the academy of Den Haag. The amount of amenities such as restaurants and shops are 10 as well (NS.nl) and the amount of people who were getting in and out per day were 34,946 in 2015 (Treinreiziger.nl). A better-elaborated table with information about the characteristics of all of the train stations that were chosen for this research is provided in the table below, table I.

Table I

Train station characteristics (in 2016)

*Shops and restaurants

**People getting in and out per day in 2015

Information about the renovations of the train stations is gathered in table II. The fourteen train stations are listed in alphabetical order from the top down. The stations used in this research are those that had renovations that were drastic and had a long construction period. The main reason for this is that it is not important in this research to measure small renovations for regular maintenance purposes for example. Projects of interest are radical renovations of train stations that changed the function or appearance of the particular area of the city. Therefore, the minimum requirement for a renovation is that the construction period of the renovation is at least a year. Very often it is the case that not only the building is renovated or even redeveloped but that it is part of project concerning the whole area of the station. Please note that in many cases the development of a station area is finalized in phases. Consequently, many projects are taking a couple of years because of the fact that with redeveloping an area with the importance

Train station Amenities* Railways Bus Tram Metro Passengers** Routes _coordinateX _coordinateY

Alphen aan de Rijn 3 4 Yes No No 9,725 3 105030 459861

Amsterdam Bijlmer-Arena 10 6 Yes No Yes 22,684 6 124976 480592

Apeldoorn 4 3 Yes No No 14,628 6 194769 469201

Boxtel 2 4 Yes No No 5,917 2 150274 399577

Den Bosch 22 5 Yes No No 43,172 12 148529 411321

Den Haag Hollands Spoor 10 5 Yes Yes No 34,946 9 81950 454048

Enschede 6 4 Yes No No 18,508 6 257701 471512 Hilversum 8 5 Yes No No 24,105 7 140918 470952 Leeuwarden 11 6 Yes No No 9,682 7 182113 578977 Leiden 20 6 Yes No No 71,100 12 93130 464584 Lelystad 5 4 Yes No No 13,369 6 160829 502267 Nijmegen 13 4 Yes No No 43,195 9 187126 428382

Rotterdam 26 13 Yes Yes Yes 85,246 23 91886 437707

(15)

of a train station, the area still has to be available for people in order to be able to travel to another place by public transportation. Hence, it is not possible to construct everything at the same time.

The construction date that is given in table II is the date of the projects’ beginning. This is the phase that the actual construction starts and the planning phase has been completed. In many cases the construction date is an estimated date. This is because in many cases it is not documented what the exact date was that the actual renovation began. In some descriptions on the Internet about renovations it is mentioned when the construction began roughly. In general, the time of the year is mentioned when it comes to the starting period of a certain construction. Exceptionally an exact date is mentioned for this. This is especially the case in the smaller renovations that are listed in the table. Therefore, when it is mentioned that in the renovation took place from a certain month, a certain part of the year or a certain period mostly the 15th_day is chosen. If the mid-year is mentioned the month June is used.

The date of the completion of the total project is important in order to test the real difference. However sometimes it is not entirely clear what the end date is of the whole project. For example, it is also possible the constructions surrounding the train station are not completely finished yet. In these cases, when available, the official opening date of the train station is used. The opening dates of the stations used were quite easier to gather compared to the beginning dates of the construction. However, in some cases the exact date is still estimated in this research. Though, this does not really matter because an increase or decrease in the transaction price of the surrounding real estate will not rely on a certain day or even on a certain week. The expectation is that after several months or even a year a possible effect will become clear. In de column years it is calculated how many years the renovation lasted. This is given in years and after the decimal point in parts of a year. In all cases the building of the train station has been at least renovated or adjusted. Most of the time with a large renovation it concerns a whole area around the train station or other surrounding projects next to the station. These projects have been mentioned in the fifth column. When obtaining this data one of the outcomes was that there is a lot variety between construction intensity, length of the construction period and the facilities that have been changed within a renovation. A simple example is the time difference of the renovation of Rotterdam Central station, which took 9.1 years, and the station of Leeuwarden, that only took approximately 1 year to finish. This is a total time difference of 8.1 years. Naturally, a large difference between these renovations is the can be explained as the magnitude of the renovations of the different stations. Often this magnitude of a renovation offers a big extra portion in terms of extra facilities, additional railways and travelers’ expectations. These factors can influence the surrounding residential property prices. Consequently, the difference in changes have to be incorporated into the research in order to measure the magnitude of a renovation. All these columns and information combined has been used to estimate a certain grade for the magnitude of the renovation, the renovation magnitude grade (RMG). Each component of a renovation in table II has been weighted and a certain factor is specified. All components are multiplied with its factor and added up for each train station. With this simple method it is possible to give a certain grade for the magnitude of a renovation. Although this method is quite subjective, the RMGs are inspected and considered plausible. The detailed method of constructing the RMG is further elaborated in Appendix A.

(16)

In essence, how bigger the RMG is, how bigger is the magnitude of the renovation. Therefore the renovation of Rotterdam has the largest RMG because it is the biggest renovation in the sample of stations used. These grades can be seen in the last column of table II.

Table II

Train station renovation projects (1998-2014)

*Mostly estimated

** Renovation Magnitude Grade

The characteristics of the train stations that are elaborated in table I and table II are conducted from various websites. In the first place, the plan was to obtain information directly from the national Dutch railway company (NS) itself. With the help of somebody who works there it was possible to get in touch with the director of the department of real estate development. Unfortunately they could not provide any information requested. It was essential to gain some information about the actual construction date and the date of completion of the project. Other issues such as the change of amenities before and after the renovation, the total renovation costs, the amount of added railways if applicable or a list of the extra renovations within a project would have been helpful to obtain from the NS in order to determine the magnitude of the renovation project. The main reason that it was not possible to obtain such information from the NS was that it was almost not possible to find it in the databases. This was because they did not store the information of the projects in an actual database. Another reason they gave was that some of the information was considered confidential. Besides, for some projects it was hard to identify the magnitude because different organizations, such as a municipality or the company responsible for the railways (Prorail), were working together on a large project in and around the train stations. The last reason mentioned by the NS was that in the Netherlands many stations are in constant renovation. However, large renovations do have a certain beginning period and an end date. In many cases there even is an official opening date for a certain new renovated or placed station so this was not a very plausible argument in my opinion. For these reasons the information about the NS stations (Table I) and the information about the renovations of the NS stations (Table II) had to be obtained from several websites.

The data in order to construct table I is gathered mainly on the website of the NS itself,

www.NS.nl/stationsinformatie. For information about amenities, amount of railways, routes and

other public transportation the NS website is used. A list of the number of passengers who are

Station Construction date* Opening date Years Surrounding project RMG**

Alphen aan de Rijn 15-10-06 28-08-12 5.9 Building (new), tunnel, bicycle parking, bus station, square 6.9 Amsterdam Bijlmer-Arena 15-06-00 17-11-07 7.4 Building (new), metro station, bus station, railway arc, square 8.3

Apeldoorn 15-06-04 18-01-08 3.6 Building (new), tunnel, bus station, square 4.9

Boxtel 15-06-98 29-09-00 2.3 Building (new), pedestrian overpass 3.0

Den Bosch 15-06-11 24-04-14 2.9 Building (adjustments), platforms, railways, bus station 3.2

Den Haag Hollands Spoor 15-04-10 15-12-11 1.7 Building (renovation) 1.6

Enschede 15-04-13 19-08-14 1.3 Building (adjustments), platforms, (bus station floor) 2.1

Hilversum 15-06-06 03-09-09 3.2 Building (adjustments), tunnel, platforms 3.4

Leeuwarden 15-01-99 15-01-00 1.0 Building (extended) 1.1

Leiden 15-02-08 15-02-10 2.0 Building (extended), platforms 2.2

Lelystad 15-05-08 18-09-13 5.3 Building (adjustments), platforms, railways 5.2

Nijmegen 27-02-02 31-12-04 2.8 Building (adjustments), station square 3.0

Rotterdam 26-01-05 13-03-14 9.1 Building (new), metro station, bus station, parking, square 9.7 Zutphen 15-04-05 24-12-07 2.7 Building (new), cycle parking, bus station, square 4.2

(17)

getting in and out of the train per day on a certain train station in 2014 is available at

www.treinreiziger.nl. Finally the X and Y coordinates of the stations are acquired from the website: www.gpscoordinaten.nl

The data for table II was more complicated to attain because it is not really easy to gather the data for the opening of a renovated train station. Especially information about the projects in smaller cities or projects that had a short duration was hard to acquire. The information about opening dates, construction dates, surrounding projects are gathered from several websites. These websites are from newspapers, municipalities, construction companies, train hobbyists and sites and magazines affiliated with public transportation (see table BI in Appendix B).

3.2 Transaction prices

The database used for the transaction prices of residential property in the Netherlands is obtained from the Dutch Association of Realtors (NVM). The database of the NVM consists transaction prices of around 70% of total house transactions in the Netherlands (Mirabeau, 2005). This database is used in this research for the transaction prices of residential property before and after the renovations of train stations. The association ‘Amsterdam School of Real Estate (ASRE)’ has a long-term relationship with the NVM, which makes it possible for their students and students of the University of Amsterdam to obtain this data.

Since it is not possible to collect all of the existing data in the Netherlands from the NVM, a part of the total data of the transaction prices in the Netherlands is requested for this research. The transactions are used for the measurement of the change in transaction prices before and after the treatment effect: the renovation of a train station. The period from 1995 until 2016 is acquired. The large time frame is chosen because the renovations of the stations are conducted in different time areas with diverse lengths of renovations. Data from the following cities in the Netherlands is acquired: Alphen aan de Rijn, Amsterdam, Apeldoorn, Arnhem, Boxtel, Delft, Den Bosch, Den Haag, Deventer, Enschede, Hilversum, Leeuwarden, Leiden, Lelystad, Nijmegen, Rotterdam, Utrecht, Zutphen and Zwolle (see table CI in Appendix C). These particular cities are chosen because of recent large train renovations. For around 600,000 observations we know the transaction prices and the exact transaction dates from the chosen period. Besides, many house characteristics are specified in the database. House characteristics such as the size of the loft, garden and house in square meter, the construction year, house type, number of floors, number of rooms, the presence of a parking place, basement, swimming pool or a garden, but also the conditions of sale for instance. Furthermore, the exact locations of the houses observed are known, which is necessary for the determination of the straight-line distance from the nearest train station.

Despite the fact that there is requested data of the transaction prices of 19 municipalities in the Netherlands unfortunately only 14 municipalities could be used. Some of the train stations of interest were still in development or finished too late to determine a possible ‘after renovation’ effect. Therefore all transactions of municipalities: Arnhem, Delft, Deventer, Utrecht en Zwolle are dropped in order to create a database that could be used for this research1_.

(18)

3.3 Data construction and descriptive statistics

To construct the data that is shown in table III and table IV at first the data had to be constructed that outliers and wrong values are eliminated from the data2_{. Thereafter several} variables were constructed and outliers were dropped3_{. The motives to drop some of these} values are found in an article written by Dröes and Koster (2014). They conducted a research with a similar model in the Netherlands about wind turbines and its effect on surrounding housing prices. Because it is a recent research and the Netherlands as a whole is taken for research the similarities are quite large and therefore possible to look at their methods of constructing the data. Thereafter, variables are made to check if there is a garden, parking place, the quality of the property, the house type, the construction period and other specific housing characteristics. Most variables created in the dataset are dummy variables. For example, if houses have a garden the value of the variable garden denotes 1 and if it has no garden the value of the variable is 0. Also, for every house in the NVM dataset the x and y coordinates (PC6_X and PC6_Y) are given. The x and y coordinates of the train stations are found on the website:

www.gpscoordinaten.nl. With this information and the Pythagorean theorem the straight-line

distance between houses and train stations can be determined. Thereafter the minimum distance from any station is calculated and a ‘location dummy’ is constructed.

As described in the related literature the range that Pagliara and Papa (2011) used for the control group was 500 meter from a metro station. Because this research contains train stations instead of metro stations a somewhat wider range is used; train stations are more important due to the fact that it is possible to travel between cities. Therefore the expectation is that a wider range benefits from the presence of a train station. However, to actually validate this choice it is important that this range is tested in an empirically with the data available in this research. In order to do this a regression is used to test what ranges seem appropriate to use in this specific research. Distance bands of 250 meters are created in order to see if there is an effect. These distance bands constructed are from 0-250 meter, 250-500 meter to the band of 2000-2250 meter from a train station. With this regression it is decided what range is appropriate to choose as a range and selected as a treatment group. The range of houses that are located within 800 meters from a train station is considered as a valid treatment group because the housing prices are getting the biggest premium until this range. Thereafter the premium diminishes. Consequently, this means that the location dummy can be constructed. The location dummy has value 1 if it is within 800 meter from a renovated train station and it has value 0 is it is not. The control group of the research is located outside of this range of 800 meters but with a maximum of 4500 meters. This maximum range is chosen because from the article of Debrezion et al. (2007) it can be concluded that on average in the Netherlands from a range of 4200 meter the effect of the presence of a train station is not likely anymore. In this research this range is rounded to a range of 4500 meter outside of a train station. In the article it is stated that people tend to take public transport instead of a bicycle from this point to the train station. It is therefore plausible that taking houses into the regression from that point cause disturbances.

2_{If certain values are unknown or wrong they have to be dropped in order to create a reliable dataset. Possible errors are an} unknown or wrong construction year/lot size/house size/number of rooms/garden/house type/PC6_X/PC6_Y

3_{Values that are dropped are: houses constructed before 1700, all different sales conditions than buyers costs, number of rooms} higher than 30, lot size bigger than 3000 m2_{, house size bigger than 250 m}2_{, price per meter lower than 500 euro and higher than} 5000 euro, transaction price higher than 1,000,000 euro.

(19)

Furthermore, the construction and opening date of a station in a certain municipality is linked to a specific municipality code. This is possible because all train station renovations are executed in different municipalities. Also, the RMG is linked in the same way with the municipalities in order to indicate the magnitude of a renovation in a specific municipality. This variable is used later in the regressions. With this variable it is possible to test whether the magnitude of renovation has any effect on the housing prices. After the creation of the construction date variable and the opening date variable it is possible to make dummies for the different time periods that are going to be measured in this thesis: the pre-construction phase, during construction phase and the post-construction phase.

In table III it can be seen that the average transaction price in the dataset used is around 217,955 euros. The price per meter is around 1708.41, which is the transaction price divided by the house size in m2_{. The treatment group, the group with the presence of a train station within 800 meter,} of this dataset is around 8% of the observations. However, only 1% of the observations are sold within 800 meters from a train station during renovation and 2% after renovation. This group is compared with the group of transacted property that lies outside the area of 800 meters but within 4500 meters of a train station.

Table III

Descriptive statistics: residential transactions (2005-2015)

Variable Mean St. Dev. Min Max

Transaction price (in EUR) 217,955 119,050 27,227 1,000,000

Price per meter (in EUR) 1708.41 635.87 500 5,000

Nearest distance to renovated train station 2273.13 1075.04 24.76 4499.47

Train station ≤ 800 m 0.08 0.26

Train station ≤ 800 m during renovation 0.01 0.12 Train station ≤ 800 m after renovation 0.02 0.14

Lot size in m2 _215.99 _198.92 ₂₅ _3,000 House size in m2 _125.73 _35.30 ₂₅ ₂₅₀ Rooms 4.72 1.22 1 26 Quality –good- 0.86 0.34 Garden 0.74 0.44 Parking 0.36 0.48 Apartment 0.08 0.28 Row house 0.46 0.50 Corner house 0.19 0.39 Semi-detached row 0.03 0.16 Semi-detached 0.16 0.36 Detached 0.08 0.27 Monument 0.01 0.07 Monumental 0.01 0.08 Construction year < 1900 0.02 0.14 Construction year 1900 - 1919 0.08 0.27 Construction year 1920 - 1944 0.24 0.43 Construction year 1945 - 1959 0.07 0.26 Construction year 1960 - 1969 0.10 0.29 Construction year 1970 - 1979 0.17 0.38 Construction year 1980 - 1989 0.15 0.36 Construction year 1990 - 1999 0.13 0.33 Construction year ≥ 2000 0.04 0.20 Year of transaction 2004.38 5.46 1995 2015 Number of observations 123,220

(20)

The treatment group versus the control group is broken down in table IV after the renovation of train stations. It can be seen that the price per meter starts a little bit higher in the area of a station but the maximum price per meter is lower than outside of this area. This is also the case with the transaction prices. It is also notable that the average price per meter within the control group is higher than the price per meter outside this group. Another interesting point is that the average construction period within the treatment group is a lot earlier than in the control group. Around 66% of the housing transactions observed, are created in the first three construction periods (from 1700 to 1944). In the treatment group only around 30% of the housing transactions are constructed in this period. The consequence of this is that the amount of monumental or monument houses are twice the percentage within the treatment group than outside. This makes sense while many train stations are located near the historic center of a city. This also has the consequence that in the control group many relative new houses are sold. This could be a reason for the higher range of transaction prices, although we have to take the vintage effect into account. This may offset the effect of age of a property over time due to preferences and design of a certain time period, such as buildings from the golden age in the Netherlands (Goodman & Thibodeau, 1995).

(21)

Table IV

Descriptive statistics: treatment group vs. control group after renovation

Train station ≤ 800 m away from property after renovation Train station > 800 m away from property after renovation

Variable Mean St. Dev. Min Max Mean St. Dev. Min Max

Transaction price (in EUR) 228,302.00 107,451.00 56,722.00 950,000.00 244,365.00 124,597.00 46,521.00 1,000,000.00 Price per square meter (in EUR) 1,907.68 549.09 552.43 4,659.09 1,894.41 625.46 500.00 5,000.00 Nearest distance to renovated train station 565.40 163.29 24.76 799.39 2,361.04 976.79 800.02 4,499.47

Lot size m2 _174.01 _140.28 _26.00 _2,655.00 _235.24 _190.37 _25.00 _2,894.00 House size in m2 _118.82 _36.36 _40.00 _250.00 _126.31 _34.05 _33.00 _250.00 Rooms 4.70 1.38 1.00 14.00 4.96 1.18 1.00 14.00 Quality -good- 0.84 0.37 0.85 0.36 Garden 0.91 0.28 0.93 0.26 Parking 0.24 0.43 0.40 0.49 Apartment 0.03 0.17 0.01 0.08 Row house 0.52 0.50 0.49 0.50 Corner house 0.22 0.41 0.21 0.41 Semi-Detached row 0.20 0.12 0.03 0.17 Semi-Detached 0.16 0.37 0.17 0.38 Detached 0.06 0.24 0.10 0.30 Monument 0.02 0.13 0.01 0.07 Monumental 0.02 0.14 0.01 0.07 Construction year < 1900 0.06 0.23 0.02 0.12 Construction year 1900 - 1919 0.22 0.42 0.06 0.24 Construction year 1920 - 1944 0.38 0.49 0.22 0.42 Construction year 1945 - 1959 0.06 0.24 0.08 0.27 Construction year 1960 - 1969 0.01 0.12 0.11 0.31 Construction year 1970 - 1979 0.03 0.16 0.19 0.39 Construction year 1980 - 1989 0.07 0.26 0.15 0.35 Construction year 1990 - 1999 0.10 0.30 0.10 0.31 Construction year ≥ 2000 0.06 0.23 0.07 0.26 Year of transaction 2010.27 4.07 2,000.00 2,015.00 2,010.58 3.97 2000,00 2015,00 Number of observations 2,578 28,817

(22)

4. Model and methodology

In the related literature discussed it can be seen that the model that is commonly used to estimate prices of residential property is the hedonic price model. In this thesis the model that is chosen to estimate the residential property prices is the hedonic price model as well. Therefore the hedonic price model is described in this research in order to clarify its application and to give a basic idea how it is constructed. Thereafter the methodology of this research is discussed.

4.1 Hedonic price model

Hedonic models are used to appraise the value of certain objects. With a hedonic price model objects are valued based on their characteristics. It can be seen as a method to estimate the value of individual characteristics of a property (Malpezzi, 2003). The hedonic price model is a way to estimate the value of a house by dividing it into several measurable parts in a way that it is possible to predict and compare prices of different houses located in the same or different areas (Malpezzi, 2003). Houses entail different characteristics such as number of rooms, quality of the house, the size, if it has a garden etc. The price of a property is determined by neighborhood characteristics as well. The quality of the neighborhood, accessibility to amenities and value of other homes in the neighborhood are determining the value of the house as well. Thus, the hedonic pricing model is used to measure the magnitude of each factor price affecting the price of a property.

Lancaster (1966) and Rosen (1974) can be seen as the founders of the hedonic price model used on residential real estate. Lancaster (1966) wrote about a new approach to measure the utility one obtains from goods. In this approach it is assumed that the utility that is acquired from goods are derived from its characteristics. It is assumed that a good gives utility to somebody, has more than one characteristic and that one characteristic belongs to more than one good. His new approach was not only meant for houses but for more goods. Rosen (1974) elaborated the hedonic pricing model by especially observing the characteristics that are attributing to the price of the product. Moreover, the emphasis in this research is about the market equilibrium of demand and supply of a certain product. Hereby the composition of the characteristics is important in determining the price and the value of a product. It is really important that misspecification of the model is avoided (Cropper, 1988). This means that too many variables are leading to a wrong output, especially if the variables are irrelevant. Therefore it is important that characteristics that are truly essential for the pricing of the house are all included, otherwise this could lead to omitted variable bias. Important in a hedonic price model is to estimate the contribution of a characteristic of a product. Therefore, the characteristics that contribute to an essential increase in the marginal price are important to include according to Cropper (1988). These attributes increase the utility on consumer’s side and therefore cause this growth in the marginal price of a product.

Despite the fact that after the introduction of the hedonic price model the method is commonly used in investigations about the housing markets, there are is some criticism on the model by various researchers. Chin and Chau (2003) wrote a literature review to notify the drawbacks and provide a checklist before using the hedonic price model. One of the drawbacks they mention is that the model is very data consuming and that one needs transaction prices of the area

What is the effect of the redevelopment of a train station on surrounding residential property prices? : an analysis across the Netherlands

Master Thesis:

MSc Business economics - Finance & Real Estate Finance