Graves are coming closer : the effect of cemeteries on transaction prices of residential properties in the Netherlands

(1)

Graves are coming closer: the effect of cemeteries on transaction prices of

residential properties in the Netherlands

Sarah Fathi (10060618) MSc Thesis

MSc Business Economics

Specialization Finance & Real Estate Finance Thesis Supervisor: Dhr. dr. M. I. Dröes August 2016

(2)

Abstract

The aim of this thesis is (1) to find the effect cemeteries have on house prices, (2) to find a causal effect by combining the change in cemetery size with the change in regulation and (3) to quantify these effects. The dataset covers 77 municipalities between 1990 and 2015. This dataset is constructed by combining four types of data from three different sources. The first source is Statistics Netherlands (CBS) where Land Use data (for the changes in cemetery size and their location) and Statline data (for data about addresses) are from. The second data source is the NVM (for house prices and the house characteristics) and the third source is Ruimtelijkeplannen.nl via which the websites of the municipalities that are part of the sample provide data about their cemetery regulation. The main method used to analyze this dataset is a hedonic difference-in-differences model with time and location fixed effects. In order to do this a treatment and control group are defined (within 500 m from a cemetery versus further away). In addition to this, four different scenarios are distinguished, being when a cemetery increases, decreases, opens and closes. Results show that when only time fixed effects are added to the regression (model 1), one of the scenarios is significant. Only when a cemetery closes the effect is positive and significant, the transaction price of properties within 500 m of this closing cemetery are 11.6% higher than before the closing, relative to the control group over that same period. However, when location fixed effects are added (model 2, the main model of this thesis), the effect changes and now the opening of a new cemetery has a significant negative effect of 3.0% on average, relative to the control group. In addition to this there are also two variations of model 2, named model 2A and 2B, which both cover all scenarios. However, in model 2A regulation is added to the regression, if the change in size can be assigned exogenously to a change in regulation, then the causal effect can be captured. This model provides additional evidence that regulation most likely has an impact and it can be stated that regulation might contribute to extra openings, but there is no evidence for a causal effect. Model 2B focusses on distances and the main result that is found is that the opening of a cemetery only has a significant interaction effect within 200 m and between 400 and 500 m from a house, beyond that the effect is insignificant and decreasing in magnitude. Model 3 is the last model and concentrates on density, it is found that address density has a significant positive effect, while cemetery density has a negative effect that is insignificant. Concluding, the main variable of interest of this thesis is the interaction effect of model 2, which states that if a cemetery opens, it has a negative effect of 3.0% on the transaction prices. Additionally, following from model 2A, this effect can be amplified by regulation. This result is important for municipal councils, since they have to decide when and where to extend or place a cemetery.

Keywords: House prices, externalities, hedonic model, difference-in-differences analysis, cemeteries, distance, ArcGis

Statement of Originality

This document is written by Sarah Fathi 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. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

1. Introduction

Last year, 139,223 people died (CBS, 2015) and 38.6% of them were buried (Nederlandse Vereniging van Crematoria, 2015). That means that 53,741 people needed a place in a cemetery, due to ageing this number will increase significantly in the coming years. The part of the population that is 65 years or older will increase to 4,7 million in 2050, which will be 26% of the total population. This is a substantial increase compared to the percentage of 16% in 2013 (National Institute for Public Health and Environment (RIVM), 2014). What also changes is the ratio of people who are 80 years and older, which was 25% of the 16% mentioned earlier, but will increase to a third of the 26% in 2050 (RIVM, 2014). This means that the demand for cemetery space will increase, which leads to the question what kind of impact a cemetery has on its surroundings? This thesis will examine the effect cemeteries have on house prices.

Apart from aging, there is another reason why cemeteries are worth investigating and that comes from previous literature regarding open spaces. For example, Anderson and West (2006) compare in their study different types of open spaces, e.g. parks, golf courses, rivers, lakes and cemeteries. Only cemeteries have a negative effect on transaction prices. Anderson et al. (2006) however place a footnote by saying that several other factors, such as the neighborhood density and income can influence this effect, but in general people are willing to pay extra to be further away from a cemetery, even though it is an open space. This latter will be discussed more extensively in the literature part of this thesis.

Lastly, culture can also be important. For example, in Asia the pseudoscience Feng Shui is key. First it is essential when buying or renting a home, since living close to a cemetery can harm your Feng Shui and this is actually reflected in the values of properties and the rental values (Teather and Chow, 2000). On the other hand, Feng Shui is also used to determine the best position of a grave within a cemetery, leading to the fact that those graves are the most expensive (Teather et al., 2000). Also, in China it is bad luck to have a direct view on a cemetery, this is for example reflected in the residential property prices in Hong Kong, where Tse and Love (2000) found a negative and significant effect of the view on prices. However, cultures differ and so does the perspective on cemeteries. On the other side of the world in New Orleans for example, where many historical cemeteries were constructed in a certain style that mimicked the actual streets and social surroundings (Miller and Rivera, 2006). Each year several activities are planned on these cemeteries, such as Jazz festivals to provide for maintenance and even a running competition. Residents of New Orleans have stated that the cemeteries are part of their landscape and that use them as is a center of the community (Miller et al., 2006). This is a total different approach to cemeteries than in Asia. How is this for the Netherlands? A search on the internet provides links to all sorts of forums where people

(7)

2 express their concerns (about groundwater pollution and traffic noise), but also that they just do not want to live near the death. Some are willing to do so, but only if the price is right. One might conclude that the general Dutch sentiment towards cemeteries in relation to homes is negative. This opinion can also be seen in practice, for example the Dutch consumer movement Vereniging Eigen Huis (2016), states that a direct view on a cemetery has a negative effect on the WOZ value of a house. In addition to that, sometimes when a municipality makes a decision involving a change in the size of a cemetery, residents demonstrate. For example, in Blaricum, where 1,439 signatures were offered to the municipal council to prevent the creation of a new cemetery. In the end this demonstration helped and the plans were cancelled (RTVNH, 2010). This might mean that people care, but does it also affect the willingness to pay? Are people willing to pay more in order to live further away from a cemetery? Therefore, the research question arises:

Do cemeteries have a significant effect on the transaction prices of residential properties?

This thesis can identify this effect particularly well because of two reasons. First, by means of the law/regulation regarding cemeteries. The law on cemeteries is ever changing. Cemeteries built before 1991, had to have at least 50 meters between the boundary of the cemetery and nearby houses (van der Putten c.s., 1999). From March 7th_{1991 until January 1}st_{2010 municipalities and individuals wanting to} extend an existing cemetery, or build a new one, where obliged to get approval from the Ministry for Public Housing, Spatial Planning and Environmental Management (VROM) (Art 34 and 40 lid 3, Wlb). However, as of from 2010, the articles 34 and 40 lid 3 were removed from the Wet op de Lijkbezorging (Wlb). This made the municipal council/municipality responsible for the decisions regarding cemeteries. So following those changes in the law, there are no regulations left regarding the minimal distance between housing and a cemetery, such that when a cemetery expands, it could come very close to the border of your house. Second, the dataset that is used to examine the effect. For this thesis there will be made use of, inter alia, the NVM database, which is a unique database with a significant amount of housing characteristics about properties in the Netherlands. This is the most complete database available and will therefore improve the quality of the results found in this thesis. Another data source that will be used is the CBS, which provides information about for example the size of the cemeteries. This also improves the quality of this research, because of the fact that there are not only increases and decreases in size, but also openings and closings of cemeteries. This makes that the research will capture every possible effect there might be. Concluding, the effect of cemeteries will be examined by means of cemetery regulation in the Netherlands, but mostly by the changes in size. This kind of research has not been done before. If one

(8)

3 looks at the bigger picture, the results of this thesis might be useful for the municipal council, since they have to decide when and where to extend or create a cemetery. However, not only the size of the cemeteries, also the distance to where the effect holds is important for their zoning regulation. Cemeteries in the Netherlands are privately owned as well as owned by municipalities. Regardless the ownership, if a cemetery needs to be extended the zoning plan has to be amended. This has an effect on homeowners, if there is a significant negative effect, it could be that when people want to buy a house and they see that a nearby cemetery is expanding, they might think twice of buying the property. Since 2008 there is a new planning claim settlement (planschaderegeling) in the Spatial Planning Act, where it is stated that if the value of real estate decreases due to changes in the zoning plan, it can be seen as a type of loss for which compensation can be requested (Art 6.1 WRO). However, there is a lower bound to the claim that can be requested because there is an excess risk of 2% of the value right before the loss (in this case the change in the size of the cemetery) arose (Art 6.1 WRO). This settlement option is important in order to quantify the effects found in this thesis which is a 3.0% decrease in price if a cemetery opens in the proximity of a property, as well as the amplifying effect a change in regulation can have on the magnitude of this decrease in price.

The remainder of this thesis is structured as follows, in chapter 2 the literature related to the effect of externalities and open spaces on house prices is debated, chapter 3 covers the data, this is followed by chapter 4 where the methodology is discussed, then chapter 5 includes the results, in chapter 6 the robustness checks and limitations are examined and in lastly chapter 7 the conclusion, discussion and an additional idea for further research will be given.

2. Literature

This literature section will be structured as follows, first, in section 2.1, several studies estimating the effect of externalities on the house prices will be discussed. This is followed by section 2.2 which covers the literature regarding the effect of open spaces (including cemeteries) on house prices.

2.1 Effect of externalities on house prices

Within the field of residential real estate, hedonic models are used to study rental housing (Sirmans and Benjamin, 1991; Jud and Winkler, 1991), commercial properties (Mejia and Benjamin, 2002; Des Rosiers, Thériault and Ménétrier, 2005) and single-family homes (Des Rosiers and Thériault, 2006). Most of the hedonic literature focuses on the latter, as will be the case for this thesis, however multi-family homes are also included in the sample. Hedonic models can be used in understanding causal dimensions, meaning

(9)

4 that the models are often used in explaining the effect of an externality on the price of a house. In the rest of this section a brief overview of some applications in the literature will be given.

Do, Wilbur and Short (1994), use a hedonic model to test whether churches near single family houses in Chula Vista (California), have an effect on the value of those houses. They divide the distance into several ranges: up to 50 ft., between 50 and 75 ft. etc. (up to 850 ft., since that is the maximal distance for which they find a negative effect). A semi logarithmic transformation of the hedonic model with a variable for distance and a quadratic transformation of distance are used. They find a convex relationship between the distance and the selling price; the further away a church is, the lower the negative effect of the church is. For example, properties on 50 ft. from the church sell at a discount of $4,151, while properties at 500 ft. away sell at a discount of $802.

Theebe (2004) uses a hedonic model to estimate the effect of traffic noise of planes, trains and cars on properties in 5 cities within (the western part of) the Netherlands. He uses a noise dummy for different ranges of noise in order to be able to allow for the relationship, between the prices of properties and the level of noise, to be nonlinear. He concludes that, for the whole sample, if the noise levels are above 65 dB, the properties will sell at a discount of 12% maximal. Between 41 and 60 dB, there is no effect, while if the noise is below 41dB, then there is even a premium up to 6.5%.

Dröes and Koster (2014) examine the effect of wind turbines on house prices in the Netherlands. They calculate the distance from a sold house to the nearest wind turbine and use a difference-in-differences method the find a possible effect of the placement of that turbine on the value of the house. To control for the differences between the treatment and the control group, in terms of house characteristics, a hedonic model is used. They find that when a turbine is built within 2 km from the property, there is a negative effect of on average 1.4%. Beyond the 2 km there is no effect at all and the highest effect, of -2.6% is measured when a wind turbine is between 500 and 750 m from the property. Kim, Phipps and Anselin (2003) study the effect of air quality improvement on house values in the metropolitan area of Seoul. They do this by using of a hedonic price model and find that the level of SO2 (Sulfur dioxide) pollution has a significant effect on the house prices, while NO2 (Nitrogen dioxide) has no effect. Following from this, they conclude that, when the concentration SO2 is reduced by 4%, this can have a positive effect of on average 1.4% on the house values.

2.2 Effects of open spaces on house prices

This thesis relates to studies that have been done regarding the effects of open spaces on house prices. All in all, there is no clear consensus when it comes to the effects of open spaces on house prices. Anderson et al. (2006) do a hedonic transaction price analysis for houses in Minneapolis to examine what the effect

(10)

5 of open spaces on house prices is. They find that there is a small fall in sales prices when there is a cemetery near. In addition, they conclude that if the population density of a location is high, an open space nearby has a beneficial effect on the value of a house. Bolitzer and Netusil (2000), on the other hand, find no price effect at all. They compare the impact of different sorts of open spaces (private/public parks, golf courses and cemeteries) on the house prices in Portland. They do this by means of a hedonic regression with a linear- as well as a log-linear functional form. Lutzenhiser and Netusil (2001) do a similar study regarding the house prices in Portland. However, they extend their research by looking at the effect of the size of an open space, such as a cemetery. They conclude that in order for a house to reach (on average) its maximum value, the optimal size of a cemetery would be -8 acres. But regarding the price effect, they find no statistical evidence that cemeteries have an effect on the sales price of a home. This is in contrast to Anderson et al. (2006) and in line with Bolitzer et al. (2000).

The research of Larsen and Coleman (2010) is similar to the previous papers, but specifically focuses on cemeteries. They do a hedonic regression and measure the effect of cemeteries with variables for distance and view. However, where the other studies grouped the cemeteries together and only looked at the average effect, this study examines four cemeteries separately and looks at the effect each one has. Regarding distance, they find varying results, for two of the four cemeteries distance has a significant positive effect on the house value, meaning, the closer the cemetery the lower the value of the house. In addition to this, they investigate what price effect a direct view on the cemetery might have. Again results vary, there is no significant effect for two cemeteries, for one a negative effect is measured and for one there is even a positive effect on house prices.

Since the existing literature holds no conclusive answer regarding the effect of cemeteries on residential properties, I would like to investigate the effect for separate cemeteries in the Netherlands. Plus, add to the existing literature by the means of the scale of my research, since I use 77 municipalities which are good for far more than 77 cemeteries. This is significantly more than the 4 cemeteries Larsen et al. (2010) examine and additionally this thesis contributes regarding the combination of regulation change and change in cemetery size.

3. Data

The data collection for this thesis consists out of four parts: Land Use data of the CBS, NVM data, zoning plan information from the municipalities and Statline data of the CBS (combined with the Land Use data), respectively discussed in part 3.1, 3.2, 3.3 and 3.4. The first step is collecting information about the cemeteries in the Netherlands between 1996 and 2012. A map with all the cemeteries in the Netherlands

(11)

6 in 2010 can be found in figure 1 below. All those cemeteries are sorted per municipality. In 1996 there were 625 municipalities, while in 2012 the amount was 415 (CBS, 2016). Due to this decrease, the determination of changes in the size of the cemeteries per municipality were not as straightforward as it would have been without the decrease. The method used to do this will be explained in section 3.1.1 and examples that are associated with this method are given in 3.1.1.1. In section 3.1.2 the sample selection will be discussed. In section 3.2 of the NVM data, first the sample size (in terms of the amount of houses per municipality) will be discussed in section 3.2.1. This is followed by a description of the data in section 3.2.2. In section 3.2.3, 3.2.4 and 3.2.5 the descriptive statistics, correlation matrix and the outlier analysis of the NVM data are shown respectively. Furthermore, each municipality in the sample will be checked for local regulation changes (regarding cemeteries), this will be presented in section 3.3. If the change in regulation is different per municipality, I will include this in the analysis, because if the regulation changes it means that there is an exogenous shock and then I’m able to do a before and after analysis. So I will combine this change in regulation with a change in the size of the cemeteries, because if I can assign the change in size exogenously to a change in regulation, then I can capture the causal effect. Lastly section 3.4 consists the Statline Data of the CBS, which is combined with, the in section 3.1 used, information about the cemeteries from the Land Use data. In 3.4.1 and 3.4.2 the construction of the two additional density variables is discussed.

Figure 1 – Cemeteries in the Netherlands

Cemeteries in the Netherlands (2010), as can be seen there are relatively few cemeteries in Flevoland, while for example in Limburg there is a substantial amount of cemeteries.

(12)

7

3.1 Land use data CBS

3.1.1 Method

Rough Land Use data regarding the change in the size of the cemeteries are collected from the CBS (2016). For each municipality between 1996 and 2012 data are collected for both the total size of the cemeteries as well as the total size of the municipality itself (a summary of this can be seen in Appendix 4). The reason for this is that over the years there were several municipal reclassifications, meaning that municipalities merged. This could imply for example that there is no change in the size of a cemetery, but only a change in size of the municipality. That it is important to take these mergers into account is shown in the next section using several examples.

3.1.1.1 Examples important municipal reclassifications Example 1: Reclassification

Between 2000 and 2003 the municipality of Bergen (NH) merged with the municipalities of Egmond en Schoorl, and continued under the name Bergen (NH). As can be seen in the first table (under Example 1) in Appendix 1, in the time between the data collection in 2000 and 2003, the total surface used for cemeteries increased significantly. However, it was not able to see this directly from the table of the CBS. Another similar example can also be found in Appendix 1.

Example 2: Reclassification and name change (later on)

On top of that, another issue with the table of the CBS is that in some cases the name of the municipality has changed. Such as for example is the case with Rijssen-Holten, of which the name was Rijssen after the merge, but later changed in Rijssen-Holten. As can be seen in the table (under Example 2) in Appendix 1, there was an enlargement of the size of the cemeteries between 2000 and 2003, but also a further enlargement between 2003 and 2006.

Example 3: No actual change in size

The cemetery size of Aalten seems to increase between 2003 and 2006, but actually it is also a merge and is therefore removed from the table in the Appendix 4, because there is no change in size. This can be seen under Example 3 in Appendix 1.

Example 4: Reclassification and direct name change

It is also possible that a municipality changes its name directly after the merge, therefore it is also impossible to see the change in size of the cemetery directly in the CBS table. Examples can be found in Appendix 2.

(13)

8

Example 5: Reclassification and reduction in cemetery size

Not all changes in size are increases, it could also be a decrease in size. Examples can be found in Appendix 3.

3.1.2 Sample selection

The total amount of municipalities with changes in cemetery size can be found in Appendix 4. However, not all the municipalities of this table will be part of this thesis. The sample construction is as follows, first the municipalities with the most significant changes (+++ and ---) are selected, this is based on the table in Appendix 4. These are the municipalities of which the cemeteries changed at least 4 ha in size. Second, the municipalities with the highest cemetery surface compared to the total surface are added. These are the municipalities for which this ratio is between 0.8% and 2.0%. The third addition are the municipalities with the highest population density, meaning a density between 4,000/km2_{and 6,000/km}2_{. The fourth} addition are those municipalities with the largest population, between 84,000 and 120,000. And those with the largest total surface (between 30,000 ha and 80,000 ha). Lastly there is one municipality which did not have a cemetery at the beginning (in 1996), but there was a cemetery created overtime. This leads to a sample of 77 municipalities, of which the changes in size can be seen in the following table. So this table consists out of, not only municipalities with the most significant changes in size, but also some of the largest in terms of size, population and density. Hence, if there is an effect it is most likely to be found with this diverse sample.

(14)

9 Table 1 – Sample with changes in cemetery size

Municipality 1996 – 2000 2000 – 2003 2003 – 2006 2006 – 2008 2008 – 2010 2010 – 2012 Almere + --- ++ +++ Amersfoort + Amsterdam + - + -- + Apeldoorn + +++ -- - ++ Arnhem + + Barendrecht + +++ + Bergen NH +++# + Bergen op Zoom ++ + - - +++ + Beverwijk --- - Breda -

Capelle a/d IJssel ++ -- - - +

De Bilt +++ -- ++ ++ Delft + - Den Bosch - + Den Haag ++ ++ Diemen + + Dongeradeel + -- +++ - Dordrecht - Ede + +++ Eindhoven ++ +++ - -- Emmen +++ +++ + Enschede +++ -- Gouda ++ -- Groningen +++ --- + - - Haaksbergen + --- ++ Haarlem + Haarlemmermeer +++ - Hardenberg ++ +++# - Haren +++ + Heemstede + Heerde + - + +++

Krimpen a/d IJssel + ++

Landsmeer + Leiden ++ - ++ Lelystad -- Leusden ++ - +++ Maassluis + Maastricht + + --- Midden-Drenthe ++ - Nijmegen + + Noordoostpolder + ++ Ouder-Amstel + + + Rheden --- -- - --- Rhenen +++ -- ++ Rijswijk ZH +

(15)

10 Rotterdam ++ + + + Rozendaal ++* -- ++ Schiedam -- ++ - Schouwen-Duiveland + + + Sluis + Steenwijkerland ++ Terneuzen - + - Terschelling ++ -- Texel - + Tilburg +++ -- ++ Utrecht +++ + Veenendaal + -- + - Velsen --- --- +++ Venray + --- - +++ Vlissingen + +++ + Voorschoten ++ --- Waterland + - + Zaanstad +++ Zoetermeer + + -- Zutphen +++# Zwolle ++ - ++

+ (-) : increase (decrease) of 1ha ++ (--) : increase (decrease) of 2 or 3 ha +++ (---) : increase (decrease) of 4 or more ha

# : this change is not completely due to the change in size of the cemeteries, but due to one of the effects mentioned in section 3.1.1.1

* : meaning that this municipality did not have a cemetery before

There were no changes in cemetery size in the municipalities of Amstelveen, De Ronde Venen, Dronten, Eemsmond, Harlingen, Hollands Kroon, Oostzaan, Purmerend, Súdwest-Fryslân, Vlieland and Wormerland

3.2 NVM data

Based on the sample selection in the previous section, the Dutch Association of Real Estate Agents and Real Estate Experts (NVM) provided data regarding the transaction prices, XY coordinates and house characteristics of properties in the municipalities of the sample.

3.2.1 Sample size

The variable HouseID is used to identify how many different houses were sold in the period between 1990 and 2015. This can be seen in table 2 and is created by dropping the HouseID duplicates1_{. The reason for} not taking the TransactionID to identify these numbers is because one house can be sold multiple times and is therefore not representative for the number of houses.

(16)

11 Table 2 – Number of houses in sold between 1990 and 2015

Municipality Frequency Percentage

Almere 26,115 2,85% Amersfoort 20,996 2,29% Amstelveen 13,309 1,45% Amsterdam 96,634 10,55% Apeldoorn 23,581 2,58% Arnhem 21,02 2,30% Barendrecht 4,751 0,52% Bergen NH 4,627 0,51% Bergen op Zoom 5,568 0,61% Beverwijk 5,248 0,57% Breda 25,226 2,75%

Capelle a/d IJssel 8,071 0,88%

De Bilt 6,872 0,75% De Ronde Venen 5,615 0,61% Delft 9,824 1,07% Den Bosch 17,612 1,92% Den Haag 61,419 6,71% Diemen 2,83 0,31% Dongeradeel 1,788 0,20% Dordrecht 16,823 1,84% Dronten 5,329 0,58% Ede 14,004 1,53% Eemsmond 1,558 0,17% Eindhoven 26,102 2,85% Emmen 13,713 1,50% Enschede 17,543 1,92% Gouda 11,095 1,21% Groningen 29,028 3,17% Haaksbergen 2,18 0,24% Haarlem 26,86 2,93% Haarlemmermeer 18,964 2,07% Hardenberg 6,347 0,69% Haren 3,954 0,43% Harlingen 1,512 0,17% Heemstede 4,924 0,54% Heerde 2,353 0,26% Hollands Kroon 3,396 0,37%

Krimpen a/d IJssel 2,239 0,24%

Landsmeer 1,323 0,14% Leiden 15,711 1,72% Lelystad 10,146 1,11% Leusden 4,174 0,46% Maassluis 2,869 0,31% Maastricht 7,751 0,85% Midden-Drenthe 4,122 0,45% Nijmegen 19,781 2,16% Noordoostpolder 5,796 0,63% Oostzaan 1,144 0,12% Ouder-Amstel 1,637 0,18% Purmerend 10,708 1,17%

(17)

12 Rheden 6,47 0,71% Rhenen 1,771 0,19% Rijswijk ZH 6,906 0,75% Rotterdam 54,228 5,92% Rozendaal 285 0,03% Schiedam 9,932 1,08% Schouwen-Duiveland 2,234 0,24% Sluis 2,072 0,23% Steenwijkerland 3,874 0,42% Súdwest-Fryslân 9,436 1,03% Terneuzen 5,452 0,60% Terschelling 26 0,00% Texel 787 0,09% Tilburg 24,868 2,72% Utrecht 46,078 5,03% Veenendaal 6,759 0,74% Velsen 9,137 1,00% Venray 2,019 0,22% Vlieland 12 0,00% Vlissingen 5,238 0,57% Voorschoten 3,615 0,39% Waterland 2,148 0,23% Wormerland 1,705 0,19% Zaanstad 17,752 1,94% Zoetermeer 15,365 1,68% Zutphen 6,862 0,75% Zwolle 16,533 1,81%

Total number of houses in sample 915,756 100%

Source: NVM (2016)

3.2.2 Data description Transaction price

The dependent variable of this thesis is the logarithm of transaction price, Ln(TransactionPrice). The reason for this transformation is to get the percentage difference and not the absolute difference in value (similar to e.g. Bolitzer et al., 2000; Dröes et al., 2014).

Distance

In the first instance (meaning for the first 3 regressions, being models 1, 2 and 2A, that will be discussed in section 4, the methodology section) the distance that will be taken is 500 m, which follows from the distance of around 1,500 ft.2_{that is used in the articles of Bolitzer et al. (2000); Lutzenhiser et al. (2001);} Larsen et al. (2010). The Spatial Analysis Tool from the program ArcGis is used to to calculate the distance between the border of the cemetery (by using coordinates from the Land use data of the CBS (available

2

(18)

13 from 1996 up to 2010)) and the houses (of which the coordinates are available via the NVM). In figure 2 below the average distance over time is shown. Since there are no data available from before 1996 and after 2010, the distances of properties sold before and after this period are based on the data of the nearest available year (e.g. a house that is sold in 2011 has a distance to where the border of the cemetery was in 2010, etc.). Therefore, for this figure, the main focus should lie between 1996 and 2010. It can be seen that there is an increase until approximately 1999, this is followed by a small decrease and then a substantial increase in 2003. After 2008 this average distance decreased to about the same distance as before the increase in 2003. This course is not what one would expect given the theory, the average distance is likely to decrease over time, there where is doesn’t according to the data. It is however for example likely that in one year there are simply more houses sold that are further away from a cemetery than the following year. In the methodology, the aim is to adjust for such effects by filtering out the time effects.

Figure 2 – Average distance to cemeteries

The average distance (in m) from a house to a cemetery in all 77 municipalities, determined per year (from 1990 to 2015)

(19)

14 Table 3 control variables – house characteristics

Variable Description

LotSize3 _{A continuous variable of the size (in m}2_{) of the lot}

property i is located on.

HouseSize3_* _{A continuous variable of the size of property i in m}2

(exclude in size m3₎

Garden A continuous variable of the size (in m2_{) of the garden}

of property i. HouseType

Subtypes

A dummy variable which indicates the type of property.

SimpleHouse

_-

_{Mobile homes, simple homes and recreational} houses

WaterHouse

_-

Water houses TerracedHouse

_-

Terraced houses

CanalHouse

_-

_{Canal houses and mansions} DetachedHouse

_-

_{Farmhouses, bungalows and villas}

CountryHouse

_-

_{Manors and estates} Apartment

_-

_{All sorts of apartments}

MaintenanceIn* A dummy variable that reflects the level of maintenance

done on the inside of a property, varying from Bad (1) to Excellent (9).

NrFloors A continuous variable that reflects the number of floors

a property has.

NrRooms A continuous variable that reflects the number of

rooms a property has. All the units in a house are counted as rooms, except for the bathrooms.

Attic A dummy variable of whether or not the property has

an attic.

Elevator A dummy variable of whether or not the property has

an elevator.

Parking** A dummy variable that reflects the sort of parking a

property has, varying from No Parking (0) to a Garage for multiple Cars (8).

Garage** A dummy variable that reflects the sort of garage a

property has, varying from No Garage (0) to a Detached Wooden Garage (4).

Monument A dummy variable that is equal to 1 if property i is a

monument.

Pool A dummy variable that is equal to 1 if property i has a

pool.

Age4 _{A continuous variable of the age of property i in years.}

* The choice for these variables will be made clear in the outlier analysis (section 3.2.5)

** No extremely high correlation between Parking and Garage (not higher than 0.8, so they are both are left in the regression), this can also be seen in section 3.2.4.

3_{Following Anderson et al. (2006), who changed the functional form of both the lot size and house size to a natural logarithm.}

This is done because of the law of diminishing returns.

(20)

15

3.2.3 Descriptive statistics

Table 4 – Descriptive statistics

VARIABLES N Mean SD Min Max

TransactionPrice 687,787 215,067 127,303 25,000 1,000,000 LotSize 687,787 266.3 494.9 25 9,991 HouseSize 687,787 128.4 41.68 25 2,164 Garden 687,787 58.11 77.89 0 998 HouseType 687,787 6.459 4.292 2 27 MaintenanceIn 687,787 7.036 1.136 1 9 NrFloors 687,787 2.691 0.652 1 10 NrRooms 687,787 4.729 1.268 0 30 Attic 687,787 0.292 0.455 0 1 Parking 687,787 1.473 2.094 0 8 Garage 687,787 0.0577 0.233 0 1 Monument 687,787 0.00592 0.0767 0 1 Pool 687,787 0.00206 0.0454 0 1 Age 687,787 39.85 34.95 0 578 Distance 687,787 1,448 1,288 0 26,818 ConstructionYear 687,787 1964 34.85 1430 2015

Table 4 above shows the descriptive statistics of the properties in the sample after the outliers are removed (the outlier analysis will be discussed in section 3.2.5). From the table above it can be seen that between 1990 and 2015 the properties that have been sold, had transaction prices between €25,000 and €1 million, the mean of the transaction price is €215,067, but the standard deviation of €127,303 is quite large. Furthermore, what is notable is that the number of rooms varies from 0 (which is probably a studio) to 30 and that the properties in the 77 municipalities in this sample are on average approximately 39 years old, with several extreme cases of more than 500 years. The distance between a property and the nearest cemetery varies from 0 (so on the border of a cemetery) to more than 26 km (those properties are all located in either Lelystad, Almere or Dronten, since the number of cemeteries in Flevoland is relatively low compared to the rest of the Netherlands. A figure of the average distance over time is shown in section 3.2.2.

3.2.4 Correlation

As can be seen from the first column in the correlation matrix in table 5 below, there are no extremely high correlations (0.8 or higher) between TransactionPrice (the dependent variable before log transformation) and independent variables. Almost all variables are significant and positively correlated with TransactionPrice. Except for Elevator and Attic (which are negatively correlated), meaning that having an elevator or attic has a negative effect on the price. Apart from looking at the relation between the

(21)

16 dependent and independent variables, it is also important to look at the correlations between the independent variables, since there is a high chance of multicollinearity if two independent variables are highly correlated. This is the case for HouseSizeM3 and HouseSize, which was to be expected since the size of a house (in m2_{) and the volume of a house (in m}3_{) heavily depend on each other. The same holds for} MaintenanceIn and MaintenanceOut, since it seems unlikely that a lot of maintenance will be done on the inside, while the outside is neglected or the other way around. For this thesis I will exclude HouseSizeM3 and MaintenanceOut from the regressions that will be further specified in section 4.

(22)

17 Table 5 – Correlation matrix

Transaction Price

LotSize HouseSize Garden House

SizeM3 House Type Elevator Maintenance In Maintenance Out NrFloors NrRooms TransactionPrice 1.000 LotSize 0.239*** 1.000 HouseSize 0.613*** (0.255)*** 1.000 Garden 0.350*** 0.098*** 0.192*** 1.000 HouseSizeM3 0.614*** 0.301*** 0.907*** 0.206*** 1.000 HouseType 0.091*** 0.094*** (0.042)*** (0.102)*** (0.041)*** 1.000 Elevator (0.023)*** 0.017*** (0.113)*** (0.087*** (0.118)*** 0.503*** 1.000 MaintenanceIn 0.139*** (0.019)*** 0.095*** (0.004)*** 0.070*** 0.089*** 0.038*** 1.000 MaintenanceOut 0.141*** (0.020)*** 0.101*** (0.006)*** 0.077*** 0.087*** 0.038*** 0.808*** 1.000 NrFloors 0.162*** (0.099)*** 0.339*** 0.085*** 0.302*** (0.473)*** (0.293)*** 0.028*** 0.036*** 1.000*** NrRooms 0.427*** 0.113*** 0.619*** 0.171*** 0.600*** (0.181)*** (0.164)*** (0.008)*** 0.010*** 0.483*** 1.000 Attic (0.050)*** (0.028)*** 0.011*** 0.064*** 0.004*** (0.158)*** (0.079)*** (0.005)*** (0.003)* 0.355*** (0.011)*** Parking 0.279*** 0.267*** 0.360*** 0.199*** 0.369*** 0.004*** (0.042)*** 0.094*** 0.093*** 0.048*** 0.180*** Garage 0.169*** 0.073*** 0.245*** 0.044*** 0.247*** 0.042*** (0.009)*** 0.071*** 0.071*** 0.017*** 0.087*** Monument 0.083*** 0.004*** 0.074*** (0.006)*** 0.086*** 0.022*** 0.006*** (0.013)*** (0.008)*** 0.029*** 0.037*** Pool 0.094*** 0.065*** 0.082*** 0.048*** 0.086*** 0.018*** 0.003** 0.015*** 0.014*** (0.005)*** 0.035*** Age 0.109*** 0.029*** (0.008)*** 0.014*** 0.046*** 0.005*** (0.052)*** (0.228)*** (0.233)*** (0.030)*** 0.025*** Distance 0.023*** 0.020*** 0.023*** 0.042*** (0.002) 0.005*** 0.007*** 0.070*** 0.070*** 0.001 (0.023)*** *** p<0.01, ** p<0.05, * p<0.1

Attic Parking Garage Monument Pool Age Distance

Attic 1.000 Parking 0.004*** 1.000 Garage (0.057)*** 0.350*** 1.000 Monument (0.016)*** (0.028)*** (0.010)*** 1.000 Pool (0.006)*** 0.049*** 0.039*** (0.002) 1.000 Age (0.060)*** (0.150)*** (0.113)*** 0.285*** (0.005)*** 1.000 Distance 0.032*** 0.005*** 0.038*** (0.016)*** 0.008*** (0.229)*** 1.000 *** p<0.01

(23)

18

3.2.5 Outlier analysis

In order to improve the quality of the data several unknown observations of the relevant variables are deleted, meaning that for HouseSize, LotSize, Garden, and ConstructionYear all variables that reported -1 were deleted, as were observations with 9999 for ConstructionYear. In addition to this, observations with unknown X and Y coordinates are deleted and the same holds for HouseSize, HouseSizeM3 and LotSize with a value of 999, 9999 or 99999, which are also treated as unknown. On top of that there are also several observations which are extreme and can be treated as outliers. As can be seen in the first scatter plot between TransactionPrice and PC4 (zip codes) in Appendix 5, the transaction prices per zip code can vary significantly. Part of this variation can be assigned to the overall difference in average price in the Netherlands. Meaning that on the left side of the graph the zip codes of the western part of the Netherlands are shown (1000 – 1999 covers most of North Holland and 2000-3999 are South Holland and Utrecht). The average transaction price in the western part of the Netherlands is higher than for example in the north east of the Netherlands (zip codes 9000 – 9999). But apart from that, this scatter plot helps to identify the real extremes. This makes that the observations with a price above €4 million are deleted from the sample. However, there are more, less extreme, outliers, to determine those Dröes et. al. (2014) is followed. They choose to deleted all observations with a transaction price below €25,000 and above €1 million, since the houses with a price above a million belong to a different segment than the average Dutch house. Next to this, it can be seen from the scatter plot of NrFloors and the HouseType (scatter plot 2 in Appendix 5), that there is an observation with 12 floors, which is a canal house in Utrecht. It is assumed that those houses are not 12 floors high and therefore this observation is deleted from the sample. The same holds for the other extreme observation that can be seen on the right hand side of scatter plot 2, this is a 2-bedroom flat for which it is unlikely to have 7 floors. Scatter plot 3 (Appendix 5) shows that there are several outliers in the data regarding the number of rooms, one observation that will be deleted is for example a simple house with 64 rooms which had a transaction price of €145,000. Other observations include houses with 40 rooms or more, while the total house size is relatively low in combination with a low price, such as a property of 95 m2_{with 40 rooms and a transaction price of €110,000. Following this,} all observations of 30 rooms or more are removed from the sample. Before deleting the outliers, the number of observations was 691,158. After this process 687,787 observations remain, making that 0.49% of the observations are dropped as a result of the selections that are made.

(24)

19

3.3 Regulation

As stated in the introduction of this thesis, the law changed twice since 1991. Before 1991 the minimum distance between housing and a cemetery was 50 meters. After 1991 this changed, municipalities and individuals had to receive permission from the Ministry for Public Housing Spatial Planning and Environmental Management (VROM) regarding the minimal distance (Art 34 and 40 lid 3, Wlb). During that period the National Institute for Public Health and Environment (RIVM) advised this distance to be between 20 and 30 meters, however National Organization for Cemeteries (LOB) argued to set no minimal distance at all (RIVM, 2002). After 2010, municipalities were free to determine their own policy regarding those distances. In the table below several examples of regulation decisions over the years per municipality are shown.

Table 6 – Regulation per municipality

Municipality Regulation Cemetery & Year

Almere No minimal distance, but with hedge of 5,5 m high with a depth of 2 m

Cemetery Almere-Haven, 2005 Amersfoort No minimal distance, but the property boundary may

be 2 m

Roman Catholic Cemetery O.L. Vrouwe, 2011

Amstelveen No minimal distance; the maximum height of the property boundary is 2 m

Sint-Urbanus Cemetery, 2015

Amsterdam Minimal distance is 10 m Noorderbegraafplaats, 2012

Apeldoorn Minimal distance is 10 m if the surrounding area is residential and 0 m if the surrounding area is “mixed”, meaning that there are, apart from houses, also shops, cafés/restaurants and small businesses

Oude Beekbergerweg cemetery, 2012

Arnhem No minimal distance; the maximum height of the property boundary is 3 m

Cemetery Huissensedijk, 2012 Barendrecht No minimal distance, only a green zone between the

cemetery and the residential area

Cemetery Scheldestraat, 2012 Bergen N.H. No minimal distance; the maximum height of the

property boundary is 2 m

General Cemetery Egmond aan Zee, 2013

Bergen op Zoom Minimal distance between 11 and 18 m Cemetery Borgvliet, 2010

Beverwijk Minimal distance is 10 m Cemetery Duinrust, 2013

Breda Minimal distance is 5 m Cemetery Muiderslotlaan, 2011

Capelle a/d IJssel Minimal distance is 50 m Cemetery Schollevaar, 2013

De Bilt No minimal distance; the maximum height of the property boundary is 2 m

Cemetery Brandenburg, 2014 De Ronde Venen No minimal distance; the maximum height of the

Roman Catholic Cemetery, 2014

Delft Minimal distance 10 m Cemetery Ypenhof, 2013

Den Bosch No minimal distance; the maximum height of the property boundary is 2 m

(25)

20

Den Haag No minimal distance, but there is a green buffer zone and the maximum height of the property boundary is 3 m

Cemetery Oud Eik en Duinen, 2009

Diemen No minimal distance and the property boundary may be 2 m high.

Cemetery Rustoord, 2013 Dongeradeel No minimal distance; the property boundary may be

1 m high

Cemetery Bonifatiuspolder, 2010 Dordrecht No minimal distance; a hedge of 2 m is allowed Cemetery Essenhof, 2014

Dronten Minimal distance is 10 m Cemetery De Wissel, 2011

Ede Minimal distance 20 to 30 m, with a property

boundary and the graves need to be at least 1 m from the border

Natural cemetery Heidepol, 2010

Eemsmond No minimal distance; the maximum height of the property boundary is 2 m

Cemetery J.F. Kennedylaan, 2016 Eindhoven No minimal distance, but the property boundary can

be 3 m high

Sint Catharina Cemetery, 2010 Emmen Minimal distance is between 20 and 30 m; in addition

to this a visual separation as well as a hydrogeological separation is necessary

Cemetery Klazienaveen, 2010

Enschede Minimal distance is 10 m R.C. Cemetery, Oosterbegraafplaats

and Israelite Cemetery, 2014 Gouda No minimal distance; a property boundary on the

road side may be 1 m and on other sides (e.g. houses) it may be 2 m

Cemetery Bosweg, 2013

Groningen No minimal distance, cemeteries are placed in category 1, meaning that they can be situated next to a residential area

Cemetery Moesstraat, 2010

Haaksbergen No minimal distance Cemetery Goorsestraat, Israelite

Cemetery and General Cemetery Enschedestraat, 2015

Haarlem No minimal distance Cemetery Kerkplein, 2012

Haarlemmermeer Minimal distance is 10 m Cemetery Iepenhof, 2016

Hardenberg No minimal distance; the property boundary may be 1 m

Cemetery Dedemsvaart, 2013 Haren No minimal distance; the maximum height of the

Cemetery Eshof, 2013 Harlingen No minimal distance; the property boundary may be

1 m high

Cemetery Begraafplaatslaan, 2016 Heemstede No minimal distance; the maximum height of the

property boundary is 1 m or 2 m, if there it is part of a hedge

Cemetery Sint Bavokerk, 2011

Heerde No minimal distance; the property boundary can be 2.5 m high

Oude Begraafplaats, 2009 Hollands Kroon No minimal distance; the maximum height of the

General Cemetery, 2012 Krimpen a/d IJssel No minimal distance, the property boundary may be

3 m

Cemetery Waalhoven, 2013

Landsmeer Minimal distance is 10 m Cemetery Zuideinde, 2013

Leiden No minimal distance; the property border may be 2 m maximum

Cemetery Rhijnhof, 2013

(26)

21

Leusden No minimal distance; a property boundary of 2 m is allowed

Cemetery Rusthof, 2010

Maassluis The minimal distance is between 10 and 50 m General Cemetery Maassluis, 2010 Maastricht Distance after expansion will be 120 m, which is

classified by the municipality as ‘sufficient’

Cemetery Tongersehof, 2015 Midden-Drenthe No minimal distance and the maximum height of the

Cemetery Torenlaan, 2013 Nijmegen No minimal distance and the property boundary may

be 2 m high

Cemetery Jonkerbos, 2008 Noordoostpolder No minimal distance; the maximum height of the

Cemetery ‘t Westerhuis, 2011 Oostzaan No minimal distance; the maximum height of the

Cemetery Dominee Nanne Zwiepsingel, 2011

Ouder-Amstel No minimal distance; the property boundary can be 2 m at most

Jewish cemetery Beth Haim, 2013

Purmerend No minimal distance Pumerend Cemetery, 2010

Rheden No minimal distance; the maximum height of the property boundary is 1 m

Cemetery Pinkelseweg, 2016 Rhenen No minimal distance; the maximum height of the

Cemetery Rhenen-Stad, 2012

Rijswijk Z.H. The minimal distance is 10 m Cemetery Sir Winston

Churchilllaan, 2013

Rotterdam The minimal distance is 14 m Cemetery Rozenburg, 2012

Rozendaal No minimal distance Cemetery Rosendael, 2010

Schiedam No minimal distance; the maximum height of the property boundary is 2 m

Roman Catholic Cemetery, 2011

Schouwen-Duiveland

No minimal distance; the maximum height of the property boundary is 2 m

Cemetery Oude Lagezoom, 2016 Sluis No minimal distance; the maximum height of the

Cemetery Schoondijke, 2014

Steenwijkerland The minimal distance is 10 m Cemetery Meppelerweg, 2013

Súdwest-Fryslân The minimal distance is 10 m Cemetery Franekerstraat, 2008

Terneuzen The minimal distance is 10 m Cemetery Bellamystraat, 2008

Terschelling No minimal distance; the maximum height of the property boundary is 2 m

Cemetery Schoolstraat, 2012 Texel No minimal distance; the maximum height of the

Cemetery Kogerstraat, 2015 Tilburg Minimal distance is 10 m, cemeteries are placed into

environmental category 1, which has a corresponding guideline distance of 10 m

Cemetery Van Hogendorpstraat, 2008

Utrecht No minimal distance, but the property boundary may be 3 m

Kovelswade Cemetery, 2010 Veenendaal Minimal distance is 20 m, and this area needs to be

filled with additional greenery

Cemetery Munnikenhof, 2013 Velsen No minimal distance; the maximum height of the

property boundary

Westerbegraafplaats, 2013 Venray The minimal distance is 5 m; the maximum height of

the property boundary is 2 m

Cemetery Boschuizen, 2010 Vlieland No minimal distance; the maximum height of the

(27)

22

Vlissingen Cemeteries are classified in category 1, which means that the minimal distance is 10 m if the cemetery is located in a residential area and 0 m if the area also includes businesses, busy roads and schools

Noorderbegraafplaats, 2010

Voorschoten No minimal distance; the maximum height of the property boundary is 2 m

General Cemetery Rosenburgh, 2013

Waterland No minimal distance; the maximum height of the property boundary is 2 m

Cemetery Kloosterstraat, 2014 Wormerland No minimal distance; the maximum height of the

Cemetery Kerkstraat, 2015 Zaanstad No minimal distance; the maximum height of the

Roman Catholic Cemetery Maria Magdalena, 2016

Zoetermeer The minimal distance is 10 m Cemetery Meerbloemhof, 2014

Zutphen No minimal distance, however there needs to be a green border between the houses and the cemetery. Additionally, the border of cemetery, as determines in a zoning plan, cannot be changed by more than 5 m.

Oosterbegraafplaats, 2009

Zwolle No minimal distance; the maximum height of the property boundary is 2 m

Cemetery Stinsweg, 2011

Sources: Belangenvereniging Almere Haven (2005); Ruimtelijkeplannen.nl (2008, 2009, 2010, 2011, 2012, 2013, 2014, 2015 and 2016)

As can be seen from the examples in the table above, some of the municipalities adopt a minimal distance of 10 meters. This guideline comes from the handbook “Companies and Environmental Zoning” of the Dutch Municipalities Association (VNG, 2016). The information regarding the regulation is added to the final model (model 2) in section 4.4.

3.4 Statline data CBS combined with CBS Land Use data

As stated by Anderson et al. (2006) the effect of open spaces (including cemeteries) can be affected by characteristics of the neighborhood, such as population density. However, an assumption of the difference-in-differences estimation, the main method in this thesis, is that the control and treatment group should be alike (further limitations of the difference-in-differences method will be discussed in the limitations in section 6). Therefore, Anderson et al. (2006) will be followed and in order to examine the effect of density an OLS regression will be added to this thesis and can be found as model 3 in section 4. There are two variables of density that are created, one is the address density (number of addresses per km2_{) and the other one is the cemetery density (total km}2_{of cemetery surface per km}2_{of total surface),} which will both be discussed in the next two sections.

(28)

23

3.4.1 Address density

In order to construct the address density, the number of addresses per zip code area (PC4) are needed. The reason for taking the data per zip code, is to be in line with the data of the NVM which also contains PC4 data and will be used to merge the datasets on. For the collection of the address data, CBS Statline is used, after which this data is added to ArcGis where the size of each zip code area is calculated. This leads to the number of addresses per km2_{. In the following graph is this density visualized per range of zip codes} at the beginning of the sample period (1990) and at the end of this period (2015). There is a clear difference in address density, the most significant difference is between 2000-2999 (small part of North-Holland and most of South-Holland) and 8000-8999 (parts of Friesland, Drenthe, Flevoland and Overijssel), but this was to be expected, since there are less people living in the northern and eastern parts of the Netherlands compared to the west. There is a relatively small difference in the density per zip code region in 1990 compared to 2015, however this is a graph made with data from the sample, not from the whole of the Netherlands and might in that way provide a slightly distorted picture.

Graph 1 – Average address density per zip code region 1990 versus 2015

3.4.2 Cemetery density

The second density variable is created with the Land Use cemetery data, this data is also added to ArcGis, where the total size per cemetery is calculated, as well as the total number of cemeteries per zip code area. This leads to the total size of the area used for cemeteries per zip code. From this follows the

(29)

24 cemetery density, which is specified as the total surface of cemetery area per km2_{. In the graph below this} density is visualized. What stands out the most is the low density in the region 8000 – 8999, which is due to the fact that part of Flevoland falls within this region and in Flevoland are on average relatively few cemeteries compared to the rest of the Netherlands (this can also be seen in figure 1 in the beginning of section 3). Additionally, there are a substantial amount of regions where the density was higher in 1990 than it was in 2015, again this is due to the fact that this is a graph made from a sample and does not present the national averages.

Graph 2 – Average cemetery density per zip code region 1990 versus 2015

4. Methodology

4.1 Hedonic price model

The hedonic price model is perfectly suited to help understand the causal dimensions of determining property value (Rosen, 1974). In essence, the hedonic relation arises because of heterogeneity. The hedonic price model explains the difference in transaction values by the difference in house characteristics (Rosen, 1974), it measures the contribution of each individual characteristic to the overall property value (Malpezzi, 2002). In this chapter several versions of the hedonic price model are constructed, in section 4.2 model 1 is introduced, which is a difference-in-differences model with time fixed effects and house characteristics as control variables. Model 2 is presented in section 4.3, this is model 1 extended with

(30)

25 location fixed effects. In section 4.4 model 2 and the data regarding the regulation are brought together into a difference-in-differences model (2A). Section 4.5 discusses model (2B), which is also an adjusted version of model 2 where the distances for the treatment are changed. Section 4.6 covers the last model (3), which differs from the previous models and consists of an OLS regression with two density variables. Lastly section 4.7 covers the hypotheses of this thesis.

4.2 Model 1: Hedonic difference-in-differences model with time fixed effects

In order to estimate the effect of the change in cemetery size, a difference-in-differences method will be used. The size change is in this case the treatment, but since the effect of this on house values alone is not enough evidence, a control group is needed. The control group should not be influenced by the change in size, but should be affected by all the other changes that can occur around the time of the change. For part of this thesis (model 1, 2 and 2A) the control group will consist out of the properties within 500 m from cemeteries that did not change in size, so which were not affected by the treatment. In order to estimate the effect of the change of size, the model should be extended with control variables to control for composition effects (Dröes et al., 2014). There are no guidelines in economic theory regarding the specification of a model, making the valuation an empirical process. The aim is to find the most suitable model, meaning that the predicted value of the model needs to deviate as little as possible from the observed transaction prices (Francke, 2015). The difference-in-differences model is combined with a hedonic model. To specify a hedonic model, three steps need to be taken: (1) the dependent variable has to be chosen. In this thesis that is the transaction price, which in many cases is transformed by taking a natural logarithm of this price and that is also what is done here (as stated in section 3.2.2). (2) Then the independent variables need to be selected. Selections of those variables are made in section 3. (3) Lastly, the correct functional form for the independent variables needs to be chosen. To do this partial scatter plots of the dependent variable against an independent variable can be made, these scatter plots can also be used to check for outliers (see section 3.2.5) (Francke, 2015). Following this the subsequent model is constructed:

𝐿𝑛𝑃𝑟𝑖𝑐𝑒𝑖,𝑡= 𝛿1𝑖,𝑡+ 𝜑1𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑖+ 𝜃1 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑖∗ 𝑃𝑜𝑠𝑡𝑡+ 𝛾1+𝑋𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡+ 𝜏1𝑡+ 𝜀1𝑖,𝑡 (1) i = property

t = time in years

4.2.1 Main variables

LnPriceit – Natural logarithm of the transaction price of property i at time t.

Treatmenti – Dummy variable, which is 1 if the property i is within 500 m from a cemetery. Postt – Dummy variable, which is 1 if the observation is done in the period after a

(31)

26 ControlVariables – All the House Characteristics defined in section 3.2.2.

This model incudes time fixed effects and therefore the variable Postt is not in the model.

4.2.2 Main difference-in-differences variables coefficient interpretation

δ1 – Average price of the properties in the control group, before the cemetery size changed. ϕ1 – Difference in average outcome between properties in the treatment group and in the control group.

ϴ1 – Difference-in-Differences/Average Treatment Effect: the shift in average price of the treatment group (before – after treatment) compared to the shift in average price of the control group (before – after treatment) (Stock and Watson, 2012).

Within model 1, 4 sub regressions will be done in order to cover all scenarios regarding the changes in the size of the cemeteries. In order to do this 4 different Postt dummies will be created, respectively for decrease, increase, closing and opening of cemeteries. This leads to 4 sub models from which the results can be seen in chapter 5.

4.3 Model 2: Model 1 with location fixed effects

It is possible that there are differences between neighborhoods/locations. For example, in Amsterdam people might be willing to live closer to a cemetery, since houses are more scarce, than in Noord-Oost Groningen, where there is more choice between houses further away. In order to control for this, location fixed effect are added. This addition makes that the Treatmenti dummy drops out. The model is the following:

𝐿𝑛𝑃𝑟𝑖𝑐𝑒𝑖,𝑡= 𝛿2𝑖,𝑡+ 𝜃2 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑖∗ 𝑃𝑜𝑠𝑡𝑡+ 𝛾1+𝑋𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡+ 𝜏2𝑡+ 𝛼1𝑖+ 𝜀2𝑖,𝑡 (2)

4.4 Model 2A: The effect of regulation

This model is a version of model 2, only Regulationit is combined with the interaction between Treatmenti and Postt and is added to the regression. In other words, a more precise treatment group is added, since this group consists out of observations that are within 500 m of a cemetery that changed in size at some point in time and is located in a municipality that changed its regulation (by change is meant that the minimal distance between housing and a cemetery is reduced to zero). By comparing this additional restricted treatment group to the “original” treatment group without regulation, the impact of the regulation can be measured. Adding this to the model gives the following:

𝐿𝑛𝑃𝑟𝑖𝑐𝑒𝑖,𝑡= 𝛿3𝑖,𝑡+ 𝜃3 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑖∗ 𝑃𝑜𝑠𝑡𝑡+ 𝜓1 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑖∗ 𝑃𝑜𝑠𝑡𝑡∗ 𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛𝑖,𝑡

+ 𝛾1+𝑋𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡+ 𝜏3𝑡+ 𝛼2𝑖+ 𝜀3𝑖,𝑡 (2𝐴)

4.5 Model 2B: The effect of distance

For this model, the definition of the treatment group of model 2 is altered. Several ranges of distances are used to see what the effect might be on the outcome. This is related to the statement made in the

(32)

27 introduction regarding the zoning regulation of the municipal council for which it is important to what distance a certain effect holds. The distances that are tested are < 200 m, 200-300 m, 300-400 m, 400-500 m, 500-600 m and 600-700 m. It is assumed that there is no effect beyond 700 m, making that the control group will consist out of observations that are located at more than 700 m from a cemetery that did not change in size. This leads to the following model:

𝐿𝑛𝑃𝑟𝑖𝑐𝑒𝑖,𝑡= 𝛿4𝑖,𝑡+ ∑ 𝜃𝑟𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑖𝑟 𝑟

∗ 𝑃𝑜𝑠𝑡𝑡+ 𝛾1+𝑋𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡+ 𝜏4𝑡+ 𝛼3𝑖+ 𝜀4𝑖,𝑡 (2𝐵)

Where r corresponds to the radius around the cemetery, for this model several dummies are created that each cover a different radius of 100 m (apart from the radius < 200 m). This gives the opportunity to see to what distance effects remain significant or become significant.

4.6 Model 3: The effect of density

This model deviates from the previous models and is a simple OLS regression including the same control variables as the previous models. In addition to that there are two density variables added to see what effect the density in terms of addresses and cemeteries on the transaction price is. Both the density variables are transformed by taking a natural logarithm, in order to be able to properly interpret the results. This gives the following model:

𝐿𝑛𝑃𝑟𝑖𝑐𝑒𝑖,𝑡= 𝜆1𝑖,𝑡+ 𝛾1+𝑋𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠𝑖,𝑡+ 𝜈𝐿𝑛𝐴𝑑𝑑𝑟𝑒𝑠𝑠𝐷𝑒𝑛𝑠𝑖𝑡𝑦𝑖,𝑡+ 𝜙𝐿𝑛𝐶𝑒𝑚𝑒𝑡𝑒𝑟𝑦𝐷𝑒𝑛𝑠𝑖𝑡𝑦𝑖,𝑡+ 𝜔1𝑖,𝑡 (3)

4.7 Hypotheses

For this thesis the effect of cemeteries will be measure by the change in size, the effect of this change will be measured within a 500 m radius of the cemetery. For this thesis the first hypothesis will be:

Does the increase, decrease, opening or closing of a cemetery has an effect on residential transaction prices.

H0: ϴ2 = 0 H1: ϴ2 ≠ 0

ϴ2 captures the average treatment effect, is the main variable of interest in this thesis and it is part of model 2 (section 4.3). This will be tested via a difference-in-differences regression with a treatment group of observations within 500 of the cemetery and a control group of observations that are located further away. Since both Anderson et al. (2006) and Larsen et al. (2010) find a significant negative effect of

Graves are coming closer : the effect of cemeteries on transaction prices of residential properties in the Netherlands