• No results found

Mosques in Amsterdam : the effect on neighboring house prices

N/A
N/A
Protected

Academic year: 2021

Share "Mosques in Amsterdam : the effect on neighboring house prices"

Copied!
37
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

MOSQUES

IN

AMSTERDAM

THE

EFFECT

ON

NEIGHBORING

HOUSE

PRICES

The Islamic community and their mosques are part of a controversial public debate in Amsterdam. It took the initiators of the Wester mosque for approximately twenty years of lawsuits and negotiations before they were able to start building their mosque. Local residents argued that this mosque in their neighborhood would affect the transaction price of their houses. This paper investigates the external effect of a mosque on surrounding residential values in Amsterdam. A unique house price dataset covering the period 1985-2015 is used to determine the location, as well as several house characteristics. The data about the mosques is gathered by the author of this study by doing interviews. The results indicate that house prices within a 150 meter radius of a mosque, after the mosque is opened for public, decrease by about 2.3 percent on average.

2016

Niels Leijten (6088481) MSc Business Economics, Real Estate Finance Supervisor Marc Francke

(2)

1

Statement of Originality

This document is written by student Niels Leijten 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)

2

Table of content

1. Introduction………3 1.1. Historical Context…..………..……..……4 2. Literature Review………6 3. Methodology………..…...10 4. Data……….………..…………..……13 5. Results………..………..20

5.1. Average Treatment Effect………20

5.2. The Impact Radius……….………..23

5.3. Anticipation and Adjustment effects………..24

5.4. Sensitivity Analysis……….………..………26 5.5. Heterogeneity of mosques…..……….………….28 5.6. Counterfactual Analysis………30 6. Conclusion……….31 7. References………33 8. Appendix….……….……….35

(4)

3

1. Introduction

Establishment of a mosque in residential neighborhoods has become very controversial these days, but the externality effects of a mosque are not yet scientifically proven. This study is the first to address the issue whether a neighborhood mosque positively or negatively affects the value of residential house prices. It focusses solely on the potential effect that might takes place in the capital of the Netherlands; Amsterdam.

The existing literature has not solely studied the effect of mosques on residential real estate yet. Do et. al. did study the effect of churches on housing value in Chula Vista, California. They found that proximity to a neighborhood church has a significant negative impact on surrounding housing values. They showed that the negative impact decreases as distance from a church increases. Brandt et. al. (2014) found in their study in Hamburg, Germany that the externalities near churches do not differ from houses of worship of other religions. This suggests that the effect of a church should be similar as the effect of a mosque. Based on these studies is the hypothesis for this study formulated as follows:

A mosque has a negative effect on the surrounding house prices in Amsterdam, but the impact decreases as distance from a mosque increases.

The aim of this study is to shed light on the social and economic debate of locations of mosques. The main question is whether or not mosques influence residential transaction prices in the neighborhood. These findings have major impact in the discussion of constructing new mosques within a residential neighborhood. Furthermore, this study assists residential real estate appraisers and urban planners by carrying out their job in Amsterdam.

The hypothesis tested, that neighborhood mosques do in fact produce externalities, is based on two considerations. First, mosques exhibit some of the same characteristics found by other externality studies to produce measurable effects on nearby residential property values. These include potentially negative effects such as noise from services, bells, people arriving and departing, and traffic and parking problems (Harlem, 1991), architectural incompatibility with residential structures in the neighborhood and the loss of view

(5)

4

amenities (Brogan, 1991), furnishing meals and lodging to the needy (Niebuhr, 1991). These same characteristics also include potentially positive effects such as acting as a symbol of morality and order and a hub of activity that draws people together for worship and socials (Drummond, 1992).

The second basis of the hypothesis is the increased public debate over the siting of mosques in neighborhoods. This debate provides circumstantial evidence of a perception among residential property owners that mosques do, in fact, have an effect on residential property values. Using a hedonic price model and sales data from the NVM, this research takes a step towards resolving this debate by estimating the magnitude and direction of the effect imposed by neighborhood mosques.

The hypothesis is tested with a standard hedonic pricing model, using a sample of 86,185 sales transactions in Amsterdam over a period of 30 years.

1.1 Historical Context

Thousands of guest workers came to the Netherlands during the second half of the 20th century. Dutch companies welcomed them with open arms; the more foreign workers, the better. The economic growth after the Second World War ensured plenty of work, but local people refused to do this. Education programs had been a sounding success which resulted in the fact that local people were no longer willing to do unskilled work. Thus, during the economic prosperity, companies recruited foreign people who were eager to fulfill these vacancies. Initially, these foreign people were meant to stay in the Netherlands temporarily but that turned out to be different. The growing group of immigrant workers brought their families to the Netherlands and built up a new life. Their kids went to Dutch schools and they adopted the Dutch culture, which made it more difficult to go back to their countries of origin. Not just worker migrants and family reunification led to a growth in the Muslim population. From the late eighties, the Muslims population increased also as a result of a growing flow of refugees and asylum seekers from the Middle East.

(6)

5

These immigrants sought space to practice their religion which resulted in a growing number of mosques. Moreover, a mosque is not just a place to fulfill religious obligations and education; it is also a place where community members may provide advice and information.

According to research from the municipality of Amsterdam, 34.8 % of the population in the city has a non-western nationality nowadays (OIS, 2015). These non-western nationalities include mainly Moroccan (9.0%), Turkish (5.2%), Antillean (1.5%), Surinamese (8.1%) and remaining miscellaneous nationalities (11.0%). These people with different ethnicities brought their own cultures, including their religions and houses of worship, to the city. Figure 1 shows that 13% of the population in Amsterdam regards themselves as Muslim nowadays. Since the beginning of the 90s is the Islam the biggest religion is Amsterdam (O&S, 2014).

Figure 1: Muslim population in Amsterdam, Source: Schippers, 2014

The first mosque in Amsterdam, the Turkish Fatih mosque, was built in 1981. From then on grew the amount of mosques in Amsterdam quite rapidly. Momentarily are 48 permitted mosques situated in the municipality of Amsterdam. These mosques are usually organized by ethnicity. The majority of the mosques belong to the Turkish (14) and Moroccan (22) community. 0 2 4 6 8 10 12 14 16 1971 1981 1982 1983 1984 1986 1987 1988 1989 1990 1991 1992 1993 2000 2002 2004 2006 2008 2010 2012 2014 P o p u la ti o n A m st erd am (% )

Muslims in Amsterdam (%)

(7)

6

The direct cause of writing this study is the construction of the Turkish Wester mosque in Amsterdam. Previous research has shown that mosques in the Netherlands are frequently located in marginal spaces. They face resistance when transformed into a visible, constructed, house of worship. A Turkish association suggested in the early nineties to build a clearly visible huge mosque in the western part of Amsterdam. This controversial mosque met heavy opposition from the municipality and local residents. One of the main topics of the debate was whether or not a mosque affects the transaction prices of neighboring residential real estate. Eventually, it took twenty years of lawsuits and negotiations to open the doors of this newly built mosque. This study will investigate scientifically whether or not the arguments of the local residents in the west side of Amsterdam were true and righteous.

The next section provides a discussion about the existing related literature. Section 3 contains the methodology, which is followed by a description of the data in Section 4. Section 5 reports the result of this study and Section 6 gives a concluding remark.

2. Literature Review

The following section will evaluate on the existing related literature to the subject. First, the importance of externalities on residential price valuation will be explained. Subsequently, the existing literature and the methods used will be elaborated. Finally, some statistical points of attention will be described.

Kain and Quigley (1970) investigated the effects of the physical and environmental quality of dwellings on their prices. Obviously, the purchase of a dwelling includes more than just the acquisition of amenities contained in the building. Housing is more accurately and usefully defined to include the internal and external attributes of a dwelling. It is therefore worth to consider how all these factors combine to influence relative prices (Carrol et. al., 1996). The market price of a housing unit can be determined by the buyers’ evaluations of the housing units’ bundle of inherent attributes, such as locational, structural, or neighborhood attributes (Freeman, 1979). Externalities can arise when a surrounding specific land use

(8)

7

affects neighboring properties and causes a change in their values. Several studies found significant effects on property values from a variety of such external sources. This study belongs to the broader line of research on the effects of these externalities on house prices. Numerous studies have already reported significant effects of location-specific externalities on surrounding housing values.

Nearly all research done within this research line of externalities on housing prices used a standard hedonic regression model. Some examples of investigated externalities are the effects of neighborhood schools (Clotfelter, 1975), golf courses (Do and Grudnitski, 1995), and neighborhood parks (Weicher and Zerbst, 1973) on residential values. These amenities all have a positive effect on the neighboring house prices. Wind turbines (Dröes & Koster, 2014), brownfields (Kaufman and Cloutier, 2006), and oil and natural gas facilities (Boxall et. al., 2004) do have a negative effect on neighboring real estate prices.

The hedonic pricing model (Rosen, 1974) is ideally suited in this regard, since it attempts to identify the price affects associated with each of the many dimensions impacting property values. This hedonic pricing model technique has commonly been applied to urban housing markets (Follain and Jimenez, 1985). The implicit values of a wide range of environmental amenities and dis-amenities have been estimated using this technique (Harrison and Rubinfeld, 1978). Control group methods, such as the hedonic pricing model, attempt to explicitly control for variations in values determining factors by selecting comparison properties that are comparable in every respect except for the adjacent adverse land use. The notion of externalities states that external factors to a property can have either positive or negative effects on its value (Do et. al., 1994). Due to its physical immobility, real estate tends to be affected by externalities even more strongly than most other economic goods, services, or commodities.

The author of this study is not aware of existing studies solely based on the influences of mosques on neighboring house prices. Therefore, the effects of houses of worship in

general will be of main concern in this section. This includes all kinds of worship houses such as churches, mosques, synagogues, and temples. Location of houses of worship within residential neighborhoods is an important issue of concern and controversy for urban planners. The question arises whether the externalities exhibited by these houses of

(9)

8

worship are positive or negative. Until now, existing empirical studies are inconclusive about the result, which indicates that this paper is absolutely relevant.

The study by Do, Wilbur, and Short (1994) was the first which addresses the issue whether a neighborhood church positively or negatively affects the value of nearby single-family properties in the US. They tested there hypothesis with a traditional hedonic pricing model, controlled for non-church effects, using a sample of 469 sale transactions of single family homes in Chula Vista, California. They included a distance variable to measure the direction and magnitude of the externality effect. Do, Wilbur, and Short used a semi logarithmic model with a quadratic transformation of the distance variable. There results indicate that the effect of churches on nearby residential property values is negative. Selling prices tend to be higher, as the distance from a neighborhood church increases. However, the magnitude of the effect depends on both the house price level and the magnitude of distance. For an average priced house in their study, the discount ranges from approximately three percent next to a church to a little more than one half of a percent at 500 feet. Beyond 500 feet the discount declines from less than one half percent to zero percent at 850 feet. They stated that houses of worship exhibit some of the same characteristics as studied in previous research. These characteristics include potentially negative effects such as noise from services, bells, people arriving and departing, and traffic and parking problems (Harlem, 1991). Furthermore, they concluded that the hardened public debate regarding the Islam and mosques could be a logical explanation for the negative influence on real estate prices. Brogan (1991) added architectural incompatibility of church building with residential structures and the loss of view amenities to the list of negative externalities.

The article by Carrol, Clauretie, and Jensen (1996) extends the analysis concerning the impact of neighborhood churches on residential property values by investing 5,000 residential property transactions in Henderson, Nevada. Carrol, Clauretie and Jensen used exactly the same methodology as Do, Wilbur, and Short did. They found that real property values decrease, at a decreasing rate, as distance from neighborhood church increases. Even though they used the same methodology, their results suggest the opposite as reported by Do, Wilbur, and Short. Carroll et. al. even bolstered their findings by showing that distance

(10)

9

from the site of a future church has little or no impact on residential property values, whereas distance from an existing church is associated with lower property values. Their evidence indicates that neighborhood churches are amenities that enhance the value of neighborhood residential property. Finally, they demonstrated that larger churches tend to have a greater positive impact on residential property values. They tested their findings with a standard hedonic regression between January 1986 and December 1990. They argued that the enhancement of residential properties due to neighborhood churches arises because they create positive externalities such as a symbol of morality and the church also can serve as a hub of commercial and social activities that draws people together.

Although they used the same models, the findings of Do et. al. and Carroll et. al. sharply contradict each other. Whereas Do et. al. found that neighborhood churches are nuisances that reduce property values over relatively short distances, Carroll et. al. found that neighborhood churches are amenities that enhance sales prices over much larger distances. Perhaps, this contrast exists due to the different datasets used. Do et. al. used a small size of their sample and they were restricted to properties at a very short distance from churches. This might have distorted their findings. However, it can also be possible that both studies reflect the influence of churches on neighborhood properties accurately within their own communities.

Brandt, Maennig, and Richter (2008) used the hedonic pricing model to analyze the impact of houses of worship on the prices of adjacent condominiums in Hamburg, Germany. They also addressed the question whether this impact differs between Islamic and Christian religious centers. This was one of the first studies on the subject conducted outside the US. They used 4,832 listing prices for condominiums in Hamburg and each address was allocated to one of the 938 statistical districts of Hamburg. The actual transaction prices were not available. Their hedonic pricing model considered spatial lag terms because they assumed that listing prices also depend on the prices of the properties previously put up for sale in the neighborhood (Ahlfeldt and Maennig, 2010). Positive externalities of houses of worship within a radius of 1,000 meter were identified. Compared with properties beyond this threshold, price premiums of 4.6 percent are obtained for condominiums at distances of 100-200 meter to the next house of worship. Furthermore, they investigated whether or not

(11)

10

the effects of abandoned church buildings are different than consecrated buildings. They found that churches should be preserved as church buildings because they continue to have positive externalities on adjacent residential property prices even after they have been deconsecrated. Their results also show that the effect of neighboring mosques does not differ from those of worship houses of other religions such as churches, synagogues, or Buddhist temples. Though, it should be noted that this study by Brandt et. al. was conducted in a metropolis known for its liberalism and open-mindedness. The results might differ in a more conservative region. Furthermore, a drawback of their approach is the use of listing prices instead of actual transaction prices. Knight (2002) as well as Merlo and Ortalo-Magné (2004), showed that the difference between offer and transaction prices is greater the longer a property is on the market.

Subsequent to the former studies, Babawale (2011) extended the research on the impact of neighborhood churches on residential property values in the Lagos metropolis, Nigeria. He tested also with a standard hedonic pricing model using a sample of 450 rented apartments across the area. His results followed the more common belief that neighborhood churches, particularly the large ones, impact negatively on the values of nearby residential properties. Nigeria is reputed as one of the most religious nations in the world, which might have affected the results. Babawale conducted his research by selecting only three churches to investigate. This could potentially have led to a selection bias. He gathered information about rental values of the apartments and residential property attributes with 450 questionnaires.

3. Methodology

Research into the variables that impact property prices is essential because the purchase of a residential property is both an investment decision as well as a consumption decision. The hedonic price approach, derived from Lancaster’s (1966) consumer theory and Rosen’s (1974) theoretical model, is perfectly suited to do so. Numerous studies have utilized the hedonic price model technique to examine the relationship between attribute preference

(12)

11

and the price of properties. The hedonic price approach allows estimating the individual effects of each characteristic on housing prices, holding all other factors constant. This paper focusses mainly on the average treatment effect using a difference-in-difference model.

1) log⁡(Yit) = β1Treati⁡x⁡Postt+⁡β2Treati+⁡ θt+ εit

Let Yit⁡ be the price of the residential property i in year⁡t. The Treati dummy variable indicates whether subject i belongs to the treatment group within the specified distance (d) of the mosque, or the control group. The Postt⁡dummy variable distinguishes whether the property is sold before or after a mosque is built. Postt equals one if the transaction of a residential property took place after a mosque opened for public in the neighborhood. Thus, the interaction term Treati⁡x⁡Postt is an indicator variable that equals one in the years after a mosque is built within the specified distance (d) of the residential property. It turns out that this specified distance is a 150 meter radius of a mosque. The results section elaborates this issue in more detail and shows also the effects beyond 150 meter. The interaction term Treati⁡x⁡Postt is the most important variable; it measures the causal effect of a mosque and the property value. This makes the β1 parameter the parameter of interest which reflects the average treatment effect. Furthermore, the monthly time fixed effects are captured by θt and εit is an identically and independently distributed error term.

Subsequently to the precious model, the following estimation controls for differences in the housing characteristics of the control and treatment group:

2) log⁡(Yit) = β1Treati⁡x⁡Postt+⁡β2Treati+ β3xit+⁡θt+ εit

xit is a standard set of housing characteristics including the log size of the house, the number of rooms, several house type dummies, construction year dummies, and indicator dummies for a garage, a garden the maintenance quality, a swimming pool and cultural heritage.

When mosques are randomly distributed over space and when the effect of mosques on house prices is immediate and permanent, β1 will capture the causal effect of mosques on house prices. However, mosques are typically not randomly distributed across space.

(13)

12

Mosques are probably disproportionally located in areas with a higher multicultural demography. This could potentially lead to a selection bias. However, this is captured by the treatment group dummy variable. The following model is estimated to filter out additional area-specific effects to the extent that these factors are time invariant:

3) Yit = β1Treati⁡x⁡Postt+ β3xit+⁡θt+ ϑj+ εit

In this regression reflects ϑj the area fixed effects for location j. Because Amsterdam is divided into 325 areas (buurten), this regression essentially deals with all unobserved time-invariant spatial attributes that may cause the construction of mosques and may be correlated with εit (Van Ommeren & Wentink, 2012). Note that the control versus treatment group dummy is excluded in this specification since it is highly collinear with the area fixed effects. The benefit if this particular approach is that it adequately controls for neighborhood and (average) housing differences, but it does not lose most of the data as a result of differencing (repeat sales), which would result in sample selection bias (see Gatzlaff and Haurin, 1998).

Equation (3) is also estimated using a restricted sample to account for any other additional unobserved variables, such as local time trends (e.g. changes in local building restrictions). 4) Yit = β1Treati⁡x⁡Postt+ β3xit+⁡ θt+ ϑj+ εit

(<500 meter of (future) mosque)

In essence, houses within the treatment area (150 meter) are compared with houses outside the treatment area but within a short distance (500 meter) of a current or future mosque. Although this will severely decrease the sample size, it does provide a more convincing story whether the placement of a mosque has a causal effect on house prices. Moreover, it addresses the problem of unobserved time trends. The results section elaborates on several other models including different control groups as a robustness check. It may be expected that the effect of mosques becomes less pronounced when properties are located further away from a mosque. Therefore, several regressions are done with different treatment areas. Eventually, the cut-off value (d) is determined by examining the statistical significance of the treatment intervals.

(14)

13

One can expect that the effect differs before and after a mosque is constructed. New mosques are usually announced some years before a mosque is accessible. It is likely that house prices already incorporate this information, which implies that β1 may be an underestimate of the causal effect if the anticipation effects are not taken into account. It might also be the case that housing markets need time to adjust to the new equilibrium. Therefore, some additional models are constructed to account for anticipation and adjustment effects. The results of these regressions are elaborated in the Result section.

4. Data

This section elaborates on the data collection. The analysis in this study is based on two main datasets. The first dataset contains the exact location of all mosques in Amsterdam from 1881 until 2015. This dataset is constructed by the author of this study. For each mosque in Amsterdam, the location, the religion, the building type and the year in which the mosque opened for public are gathered. The exact date of opening and the announcement date are not known. This dataset is made by taking interviews and collecting data from the internet. The interviews were done with spokespersons and directors of the underlying foundations of the mosques. Surprisingly, the municipality is neither aware of the location nor the characteristics of the mosques in Amsterdam. As a result, the data is party derived from a limited number of (sometimes incomplete) websites, which might affect the results. Figure 2 shows the spatial distribution of the mosques across Amsterdam. As can be seen in figure 2, most mosques are predominantly clustered in the western and eastern part of Amsterdam, while none of them is located in the center. This is not surprising since most Muslims live in these parts. Less than 5 percent of the population in the center regards themselves as Muslim while for instance 33 percent of the population in the western part of Amsterdam claims to be Muslim (OIS, 2006). It makes sense that mosques are located close by their visitors.

(15)

14 Figure 2: Spatial distribution mosques

(16)

15 0 10 20 30 40 50 1980 1985 1990 1995 2000 2005 2010 2015 N u m b er o f m o sq u es

Number of mosques in Amsterdam

0 200 400 600 800 1000 1200 1400 1985 1989 1993 1997 2001 2005 2009 2013 D is ta n ce in met er s

Distance to nearest mosque

Religion of mosques Marokkaans (44 %)

Turks (32 %) Surinaams/Pakistaans (12 %) Overig (4 %) Indonesisch (2 %) Marokkaans/Pakistaans (2 %) Surinaams (2 %) Marokkaans/Turks (1 %)

Type of building woonhuis (30 %)

school (20 %) moskee (18 %) bedrijfspand (12 %) kerk (10 %) buurthuis (6 %) gymzaal (2 %) winkel (2 %)

Figure 3 shows that the number of mosques in Amsterdam has been increasing steadily since the opening of the first mosque took place. It shows that the first mosque was built in 1981 and that the number of mosques is increasing gradually in accordance with the growth of the Muslim society in Amsterdam. Nowadays, 48 mosques are located in Amsterdam.

Figure 3: Number of mosques in Amsterdam over time Figure 4: Distance to nearest mosque over time

(17)

16

The more mosques are located in Amsterdam, the smaller the straight line distance to the nearest mosque will be. Figure 4 shows that the distance to the nearest mosque declined from 1276 meters in 1985 to 730 meters in 2015. The increase in distance to the nearest mosque in 1996 can be explained by the high number of property transactions which were relatively far away from a mosque.

Figure 5 shows the distribution of the type of building of mosques. The pie chart indicates that 30 percent of the mosques in the dataset are located in residential dwellings, 20 percent of all mosques are located in a (former) school and only 18 percent is located in a building which is originally meant to be a mosque. Mosques are usually also subdivided by ethnicity. This distinction is mainly based on the origin of the initial directors of the mosque. The majority of all mosques are Moroccan (44%) and Turkish (32%). This is in line with the ethnicity of the majority of the immigrants living in Amsterdam. Although there is this distinction, all mosques are normally for everyone accessible.

The second dataset covers about 70 percent of all housing transactions in Amsterdam from 1985 until 2015. The essential characteristics of these residential properties are obtained from the Dutch Association of Realtors (NVM). This dataset includes the exact location, transaction prices, size in square meters, the number of rooms, maintenance quality, construction year and the presence of a central heater, garden, swimming pool and several other characteristics for all observations. Because the exact location of each property is known, it is possible to calculate the exact straight-line distance of each property to the nearest mosque. The straight-line distance between the property 𝑖 and the nearest mosque is calculated with online geolocation tools. This distance determines whether or not a property belongs to the treatment group or the control group.

Table 1 presents descriptive statistics for the housing transaction dataset. Transactions with prices above € 1 million or below € 25.000 are excluded. Also, the maximum number of rooms within one property is set at 10. Furthermore, transactions that refer to properties smaller than 25 m2 or larger than 300 m2 are excluded as well. These restrictions refer to less than one percent of the total observations. Table 1 shows that the average transaction price in this dataset is €220.941 and the average size of a property sold is 82.88 m2. The average distance to the nearest mosque is about 740 meters.

(18)

17

When a (future) mosque is located within a range of 150 meter of the property, it is considered to be part of the treatment group. Circa 4 percent of the observations (about 3500 observations) are considered to be in the treatment group and 3,37 percent of the housing transactions (about 2911 observations) are located within a 150 meter radius of a mosque after it has been constructed.

Variable Mean SD Min Max

Price (€) 220941 134817 25094 1000000

Price per m2 2725,48 1138,72 136,70 12187,50 Distance to nearest mosque (m) 740,07 590,78 2 5970 Mosque < 150 meter 0,0406 Size in m2 82,88 35,27 25 300 Rooms 3,15 1,16 1 10 Apartment 0,88 0,32 Terraced 0,10 0,31 Detached 0,01 0,09 Semi-detached 0,01 0,08 Garage 0,04 0,19 Garden 0,22 0,42

Listed (as cultural heritage) 0,03 0,18

Swimmingpool 0,00 0,02

Maintenance quality - good 0,87 0,33 Construction year 1500-1905 0,15 0,35 Construction year 1906-1930 0,30 0,46 Construction year 1931-1944 0,10 0,30 Construction year 1945-1959 0,06 0,23 Construction year 1960-1970 0,08 0,27 Construction year 1971-1980 0,05 0,23 Construction year 1981-1990 0,11 0,32 Construction year 1991-2000 0,10 0,30 Construction year > 2000 0,05 0,21 Year of observation 2006 6,84 1985 2015 Number of Observations

Table (1) - Descriptive Statistics: Housing Transactions

(19)

18

Variable Mean SD Min Max Mean SD Min Max

Price (€) 216323 115984 27227 950000 221136 135553 25094 1000000

Price per m2 2911,38 1119,28 157,56 7656,25 2717,61 1138,88 136,70 12187,50

Distance to nearest mosque (m) 95,69 39,37 2 150 767,33 587,72 151 5970

Size in m2 77,46 32,84 25 285 83,11 35,35 25 300 Rooms 3,06 1,12 1 10 3,15 1,16 1 10 Apartment 0,92 0,27 0,88 0,33 Terraced 0,07 0,26 0,11 0,31 Detached 0,01 0,08 0,01 0,09 Semi-detached 0,00 0,03 0,01 0,08 Garage 0,01 0,11 0,04 0,20 Garden 0,20 0,40 0,22 0,42

Listed (as cultural heritage) 0,02 0,13 0,04 0,18

Swimmingpool 0,00 0,00 0,00 0,02

Maintenance quality - good 0,86 0,34 0,87 0,33

Construction year 1500-1905 0,15 0,36 0,15 0,35 Construction year 1906-1930 0,40 0,49 0,29 0,46 Construction year 1931-1944 0,07 0,25 0,10 0,30 Construction year 1945-1959 0,05 0,21 0,06 0,23 Construction year 1960-1970 0,07 0,25 0,08 0,27 Construction year 1971-1980 0,03 0,18 0,06 0,23 Construction year 1981-1990 0,10 0,30 0,11 0,32 Construction year 1991-2000 0,06 0,24 0,10 0,31 Construction year > 2000 0,07 0,25 0,05 0,21 Year of observation 2007 6,49 1985 2015 2005 6,85 1985 2015 Number of Observations

Table (2) - Descriptive Statistics: Treatment versus Control group

3499 82686

(20)

19 0 1 2 3 4 5 6 7 8 150 meter 250 meter 500 meter 750 meter 1000 meter 1250 meter 1500 meter 1750 meter 2000 meter N u m b e r o f m o sq u e s

Mosques within range

Table 2 reports the descriptive statistics for those observations within versus outside a 150 meter radius of a (future) mosque (treatment versus control group). As expected, mosques are placed in areas with a relatively lower house price (selection effect). In the regression analysis, because the treatment group indicator is included, there is controlled for these differences. More in general, any differences in (average) house prices across locations are captured by the area fixed effects. (See appendix). There are 325 unique areas in this dataset used with, on average, 265 observations per area.

Moreover, given the other differences in housing characteristics between treatment and control group, it is important to control also for housing characteristics in the empirical analysis. There are, for example, relatively more houses built between 1906 and 1930 in the treatment group than the control group. To avoid that the estimated treatment effect is biased, this paper explicitly controls for the differences in composition of the control group and the treatment group.

Figure (7) indicates how many mosques are located within a certain range of a property sold in 2015. It means for instance that, on average, 0.60 mosques are located within a 500 meter radius of a property while there are 6.7 mosques in the range of 2000 meter.

(21)

20

5. Results

The results section is organized in the following manner. First, the baseline estimates for the average treatment effect are presented. Second, the exact radius of the treatment effects is elaborated and examined whether anticipation and adjustment effects are important. The next part checks robustness of the results regarding different identifying assumptions. Then, a closer look at further heterogeneity on the treatment effect is presented. Finally, a counterfactual analysis is presented about the total loss in house value as a result of the opening of a mosque in Amsterdam.

5.1 Average treatment effect

Table 3 contains the regression estimates based on equation (1) to (4). Column 1 shows the regression estimates of equation (1), the standard difference-in-difference model. The results in column 1 suggest that the placement of a mosque increases house prices within a 150 meter radius by about 6.9 percent on average in comparison to the control group. Column 2 adds housing characteristics as additional control variables. The treatment effect increases to 8.9 percent. Both estimates are counterintuitive due to omitted variable bias. The results also suggest that mosques are placed in locations where house prices are on average 8 percent lower. This selection effect (together with the treatment effect) could explain why many homeowners might think that a mosque has a very strong effect on house prices. This effect, however, is not necessarily causal. The results in column 2 further show that most of the hedonic characteristics are statistically significant. Interestingly, according to the estimates, apartments are on average more expensive than terraced houses, the reference category. This is an indication that the results of the simple hedonic model might be biased. In particular, in this model there are most likely omitted time (in)variant determinants of house prices that are correlated with the location of mosques. Column 3 adds area fixed effects to control for this. House prices, according to this model, suddenly decrease with 2.0 percent after a mosque is open for public. Still, there might be

(22)

21

unobserved changes, such as changes in zoning regulations, that affect the estimates. Consequently, house prices are examined of the treatment group versus a local control group (< 500 meter of a current, previous, or future mosque). The results are reported in column 4. Houses prices decrease by 2.3 percent. This is the preferred estimate although it is most likely a conservative estimate. When there are effects of mosques beyond 150 meter, the estimated coefficient will be an underestimate. Also, when anticipation and adjustment effects are important, the average treatment effect will also be an underestimate. Moreover, the results in column (3) and (4) do show that apartments are cheaper than terraced houses.

Nevertheless, column 1-4 show estimates of the average treatment effects. The following paragraphs will elaborate more on the geographical extent of the effect and the potential anticipation and adjustment effects.

(23)

22 1 2 3 4 Mosque <150 meter 0.0690*** 0.0893*** -0.0201*** -0.0232*** (0.0167) (0.00930) (0.00345) (0.00341) Mosque in 2016 <150 meter -0.0986*** -0.0803*** (0.0150) (0.00837)

House Size (Log) 0.867*** 0.740*** 0.698***

(0.00422) (0.00344) (0.00526) Rooms 0.00611*** 0.0235*** 0.0294*** (0.00129) (0.00101) (0.00154) Garage 0.136*** 0.0955*** 0.0918*** (0.00511) (0.00386) (0.00774) Garden 0.0525*** 0.0560*** 0.0567*** (0.00221) (0.00152) (0.00231) Monumental Heritage 0.183*** 0.0451*** 0.0407*** (0.00530) (0.00433) (0.00952) Swimmingpool 0.277*** 0.132*** -0.106** (0.0499) (0.0337) (0.0475) Maintenance quality 0.148*** 0.149*** 0.149*** (0.00259) (0.00189) (0.00282) Construction Year 1500-1905 0.171*** -0.0861*** -0.102*** (0.00490) (0.00379) (0.00551) Construction Year 1906-1930 0.0428*** -0.0945*** -0.119*** (0.00469) (0.00357) (0.00508) Construction Year 1931-1944 -0.0330*** -0.103*** -0.133*** (0.00509) (0.00392) (0.00567) Construction Year 1945-1959 -0.186*** -0.190*** -0.203*** (0.00574) (0.00457) (0.00693) Construction Year 1960-1970 -0.316*** -0.211*** -0.229*** (0.00540) (0.00493) (0.00713) Construction Year 1971-1980 -0.336*** -0.145*** -0.192*** (0.00630) (0.00555) (0.00803) Construction Year 1981-1990 -0.200*** -0.126*** -0.188*** (0.00526) (0.00399) (0.00586) Construction Year 1991-2000 -0.00361 -0.0399*** -0.0581*** (0.00503) (0.00365) (0.00557) Apartment 0.105*** -0.0721*** -0.0885*** (0.00351) (0.00345) (0.00659) Detached 0.229*** 0.254*** 0.285*** (0.0128) (0.0122) (0.0317) Semi-detached 0.0949*** 0.132*** 0.0500* (0.0114) (0.00978) (0.0270)

Housing Characteristics (16) No Yes Yes Yes

Month fixed effects Yes Yes Yes Yes

Area fixed effects No No Yes Yes

N 86185 86185 86185 33820

Adjusted R2 0.414 0.839 0.929 0.932

t statistics in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01 Clasical DID Housing

Characteristics Area fixed effects

Control group 150-500 meter

Table (3) - Baseline Regression Results: The effect of mosques on house prices

(24)

23

5.2 The impact radius

Up to now, a cut-off value of 150 meter was used to determine the treatment effect. To verify whether this is a valid choice, the treatment effect depends on distance in the following table. In essence, a set of dummy variables is included for several different intervals. The results are compared relative to those observations outside the treatment interval. Figure 4 reports the results. The table is divided into four rows. Each row shows the average treatment effect for a specific distance interval from the nearest. The first row shows the results for intervals of 100 meter, the second row shows the results for 75 meter intervals and the third row takes intervals of 50 meter into account.

Examining the 100 meter intervals shows that the price effect become close to zero and insignificant after 100 meter. This suggests a cut off distance of 100 meter. Separating this range into smaller ones of 75 meter shows a significant negative effect until 150 meter, followed by a significant slightly positive effect of 0,96 percent. Separating the intervals again into smaller ones of 50 meter shows again a cut off value of 150 meter since the effects become more or less zero after 150 meters.

Intervals of 100 meter 1 2 3 4

0-100 101-200 201-300 301-400 Average Treatment Effect (ATE) -0.0212*** 0.000721 0.00812*** 0.00218

(0.00493) (0.00282) (0.00228) (0.00210)

Intervals of 75 meter 1 2 3 4

0-75 76-150 151-225 226-300 Average Treatment Effect (ATE) -0.0176*** -0.0173*** 0.00958*** 0.00661***

(0.00579) (0.00378) (0.00265) (0.00242)

Intervals of 50 meter 1 2 3 4 5 6

0-50 51-100 101-150 151-200 201-250 251-300 Average Treatment Effect (ATE) -0.0313*** -0.0108** -0.0155*** 0.00836*** -0.00152 0.0159***

(0.00838) (0.00515) (0.00429) (0.00312) (0.00284) (0.00298)

Number of Observations 87671 87671 87671 87671 87671 87671

R squared 0.932 0.932 0.932 0.932 0.932 0.932

Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01

(25)

24

The figure clearly shows that there is a significant negative effect until 150 meters for each interval used. The results show, on average, a gradually decreasing negative effect of a mosque starting at -3.1%.

The results do not imply that there is no effect after 150 meter from a mosque, but the effect is most likely too small. On average, it would not be possible to reject the null hypothesis that there is no effect. As such, it is decided to use the 150 meter radius as the relevant treatment area throughout this study. However, section 5.4 will show some results using different sizes of the treatment and control area.

5.3 Anticipation and adjustment effects

House prices may already decrease before a mosque becomes operational. It might also be the case that housing markets slowly incorporate new information. This could lead to adjustment effect in prices after the construction of a mosque. Therefore, a model is estimated which decomposes the treatment effect before and after the construction of a mosque. Figure X, X and X report the results, in percentages.

Figure 8: anticipation and adjustment effect (1) -4,0% -3,5% -3,0% -2,5% -2,0% -1,5% -1,0% -0,5% 0,0% -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 Ef fe ct o f m o sq u e

(26)

25 -8,0% -6,0% -4,0% -2,0% 0,0% 2,0% 4,0% 6,0% -10 to -6 -5 to -1 0 1 to 5 6 to 10 Ef fe ct o f m o sq u e

Year before/after opening of the nearest mosque

Anticipation and adjustment effect (2)

-8,0% -6,0% -4,0% -2,0% 0,0% 2,0% 4,0% 6,0% -9 to -7 -6 to -4 -3 to -1 0 1 to 3 4 to 6 7 to 9 Ef fe ct o f m o sq u e

Year before/after opening of the nearest mosque

Anticipation and adjustment effect (3) The results in Figure 8 indicate that anticipation effects are not very important. The effect, seven years before a mosque opens is for instance 2.4 percent, but the effect is still 2.4 percent seven years after the mosque opened. In order to estimate these results, Figure 8 uses a reference group of houses that are further away than 150 meters away of a mosque and houses that are sold before the year of interest. The vertical line represents 95 percent confidence intervals. Seven years before the opening of a mosque, the reference consists of all houses outside the 150 meter range and all houses sold eight or more years before the nearest mosque opened. Seven years after the opening of a mosque consists the reference category of all houses outside the 150 meter range and all houses sold six or less year after the opening of a mosque. The effect might be biased due to an unbalanced distribution of reference and treated observations.

Figure 9.1: Anticipation and adjustment effect (2) Figure 9.2: Anticipation and adjustment effect (3)

The years before/after the opening of a mosque are clustered in Figure 9.1 and Figure 9.2 in order to elaborate on the anticipation and adjustment effects.

Figure 9.1 and 9.2 both show that there is an anticipation effect, although this is not significant in all cases. The dots represent the conditional averages for a given year before/after the opening of the nearest mosque within 150 meter radius of the property and the vertical line represents, again, the 95 percent confidence intervals. The reference categories are different though. The reference category consists in these figures of all houses outside the 150 meter radius and all houses which are not sold in the specified period of time.

(27)

26

In figure 9.1 are the effects estimated for clustered periods of 5 years. The figure shows that the effect on house prices is 2 percent in the five years preceding the opening of a mosque, while the effect is only 0.8 percent five years after a mosque is opened.

Figure 9.2 clustered only 3 years together in order to specify the effects. These results show, just as the results in figure X a clear effect of the opening of a mosque but these results are also not always significant on a 95 percent level. The results in Figure Y reflect a 1.9 percent effect in the three preceding years before the opening of a mosque. The effect is 2.8 percent when a house is sold in the same year as the mosques opens, but the effect decreases again to 1.5 percent in the next three years.

Hence, there seems to be some evidence that house prices incorporate this information three years before the opening of a mosque. The effects seem to be levelled out three years after the mosque was opened. To identify this effect to the full extent, further research would be needed using information on when a mosque was publically announced.

5.4 Sensitivity Analysis

In order to identify the causal effect of mosques on house prices, area fixed effects are included. Also, the control group consisted of properties within a 500 meter range of a mosque. The identifying assumption is that unobservable time-varying factors are not correlated to the treatment effect. Several robustness checks are done to investigate the validity of this assumption in more detail. The results of these checks are reported in table 5. First, there may be unobserved time trends that are not captured by the month fixed effect but are correlated with the treatment effect. Because the data spans a long time period (30 years), one might expect that most unobservable factors are changing over time. In order to take care of this issue, for each area are decade time fixed effects added (1985-1995, 1996-2005 and 2006-2015). The results indicate that the effect of mosques on house prices is -2.9 percent and it is still statistically significant at a 1 percent level. The second column shows the result when year fixed effects are added for each cluster of areas. This result is very similar to the result in column 1.

(28)

27

1 2 3 4 5 6 7

Mosque <150 meter -0.0287*** -0.0286*** 0.00289 -0.0124* -0.00644*** -0.0243*** -0.0208***

(0.00358) (0.00355) (0.0105) (0.00641) (0.00208) (0.00340) (0.00343)

Housing Characteristics Yes Yes Yes Yes Yes Yes Yes

Month fixed effects Yes Yes Yes Yes Yes Yes Yes

Area fixed effects Yes Yes Yes Yes Yes Yes Yes

Area x decade fixed effects Yes No No No No No No

Buco x year fixed effects No Yes No No No No No

Obs. <150 meter of mosque Yes Yes Yes Yes Yes Yes Yes

Obs. <500 meter of mosque Yes Yes No No No No No

Obs. <1000 meter of mosque No No No Yes Yes Yes Yes

Religion of the mosque No No No No No No No

Building type of the mosque No No No No No No No

Number of observations 34368 34368 3550 26718 68404 68404 81070

adjusted R2 0.943 0.949 0.930 0.939 0.935 0.933 0.931

Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01

Control group 150-1000 meter

Control group 150-1750 meter

Table (5) - Robustness: Identification and control groups

(Dependent variable: The Logarithm of house price)

Area x decade fixed effects Buco x year fixed effects Only obs. < 150 Treat < 150 Control 500-750 Treat < 500 Control 500-1000

(29)

28

1 2 3

Mosque < 150 meter -0.0229*** -0.0229*** -0.0229*** (0.00342) (0.00343) (0.00341)

Housing Characteristics Yes Yes Yes

Month fixed effects Yes Yes Yes

Area fixed effects Yes Yes Yes

Religion Yes No Yes

Buidling type No Yes Yes

Number of observations 33820 33820 33820

Adjusted R2 0.932 0.932 0.932

Religion Building type Religion & Building

Table (6) - Heterogeneity: The effect of mosques on house prices

(Dependent variable: the logarithm of house price)

Column (3) reports the results when all observations are part of the treatment group. This means that it takes into account the houses within the 150 meter range. The dataset becomes much smaller and the result is not significant either due to a lack of a reference group.

As mentioned before, the treatment effect might be underestimated if there is still an effect in the control group area. As such, the results in column 4-7 experimented with different distributions instead of using the 150-500 meter radius as the control group. Further away from a mosque it is more reasonable to assume that the treatment effect is zero. Column (4) used a control group of houses that are more than 500 meter away, but less than 750 meter away from a mosque. The treatment effect becomes 1.24 percent. Alternatively, the

regression in column (5) used a larger treatment area (500 meter) which made the effect declining to 0.6 percent. The remaining regressions in column (6) and (7) used the original treatment area, but they used slightly different reference groups. There effects differ from 2 percent to 2.4 percent. These results show that the initial findings are reasonably robust in terms of sign and magnitude.

5.5 Heterogeneity in the effect of mosques

There is considerable heterogeneity in the treatment effect, for example, because mosques differ across locations and several other characteristics. Table 6 reports the results of several alternative specifications in which interaction effects are included to examine the heterogeneity in the treatment effect.

(30)

29

1 2 3 4

One mosque < 150 meter -0.0203*** -0.0206*** -0.0206*** -0.0198*** (0.00347) (0.00344) (0.00346) (0.00345) Two mosques 0.0107*** 0.0130*** 0.00604 0.00477

(0.00301) (0.00326) (0.00420) (0.00626) Three mosques -0.000496 -0.000352 0.00429 0.0157***

(0.00392) (0.00376) (0.00511) (0.00602) Four or more mosques 0.0247*** -0.00499 0.0221*** 0.0277***

(0.00584) (0.00392) (0.00516) (0.00159) radius 500 meter radius 1000 meter radius 1500 meter radius 2000 meter

Table (7) - Heterogeneity: The effect of multiple mosques

(Dependent variable: the logarithm of house price)

First, the results in column (1) to (3) all show that the original religion and the type of building of a mosque barely influences the effect on surrounding house prices. The average treatment effects are constant and significant at a 99 percent level. Data about the visibility of mosques and the square meters of the surface might be interesting to add to the regression but this data is unfortunately not available.

Second, the treatment effect may depend on the number of mosques that are open for public. For example, several mosques in the same street are likely to have a much stronger price effect than one single mosque. As a result, several dummy variables are added to the regression. These results are reported in table 7. Column (1) shows the results taking into account a radius of 500 meter distance from a property; column (2) included a 1000 meter radius, column (3) included 1500 meter radius and column (4) used a radius of 2000 meter. The effect of the nearest mosque within a radius of 150 meter does not seem to depend on the number of additional mosques since the average treatment effect remains more or less the same. In contrast to that, the effect of the second mosque within a 1000 meter range is positive.

(31)

30

NVM Data 3499 82686 86185

Full owner-occupied market (100%) 4999 118123 123121

Full market (inc. Rental Market) 15621 369134 384754

Average house prices (€) 216323 221136 220941

Total Loss (€, in millions) 1972

Average loss per house (€) 22883

Average loss per mosque (€, in millions) 41

Houses < 150 meter

Houses > 150

meter Total

Table (8) - Counterfactual Analysis

5.6 Counterfactual analysis

This subsection carries out a counterfactual analysis to gain a better understanding of the quantitative implications of the empirical results. The results should be interpreted with caution because several simplifying assumptions are made to estimate the total external costs of a mosque.

First, the

The database of the transactions covers only 70 percent of all transactions. To obtain an estimate for the full owner-occupied market, the result of this study has to be multiplied by 100/70. The assumption is made that this is the same across Amsterdam and stable over time. Second, about 32 percent of the properties in Amsterdam are owner-occupied. To get an estimate for the full market, including rental housing, 68 percent is added to the total owner-occupied market estimate (Amsterdam in cijfers, 2015). The effect of a mosque on rental houses is assumed to be identical to the effect of mosques on owner-occupied houses as well.

Table (8) contains a back of the envelope calculation about the total loss in house values as a result of the mosques that is located in Amsterdam. The total loss in house value is, according to this simplified calculation, approximately 1.9 billion if the average treatment effect of -2.32 percent is multiplied by the average house price. In the end, this table suggests a loss of 41.09 million euro per mosque. Although it is clear that these calculations are subject to hard simplifying assumptions, it does provide a picture of the magnitude of

(32)

31

the price effect of mosque. The total loss in house value in Amsterdam as a result of mosques is absolutely substantial and runs into the millions of euros.

6. Conclusion

This paper has investigated the effect of mosques on house prices in Amsterdam. It is important to realize that most mosques are built in multicultural areas. However, this does not mean that these mosques have no external economic effect. The Islamic religion and mosques in particular, are very controversial in Amsterdam. Nowadays, 48 mosques are located in Amsterdam and this study investigated whether or not these mosques have an economic impact on the residential transaction prices.

The results show a pronounced negative external effect of the opening of a mosque on house prices of 2.3 percent. Though, the negative effect depends on the straight-line distance from the property to the nearest mosque. The results also indicate that the effect of mosques on house prices is extremely local. Future research should focus mainly on the heterogeneity of the effect of mosques but there must be sufficient data available to do so. Existing literature by Brandt et. al. already found evidence that the effects of churches on residential house prices are similar to the effect of places of worship of other religions. In fact, this suggests that the negative effect on surrounding house prices is not caused by religious factors. It is assumed that the effect occurs due to several externalities which are inherent to a house of worship. These externalities are for instance noise from arriving and departing people, traffic and parking problems and the loss of view amenities. Thus, the negative effect found in this study is not necessarily caused by the mosque itself but by the externalities that come with the construction of a mosque. Reasoning for the negative effect is also the hardening debate about the Muslim society in Amsterdam. People might get aversion against mosques, since their reputation is deteriorating rapidly.

Do, Wilbur, and Short found in their study a negative impact of a church of three percent on surrounding house prices. This result is very similar to the results of this study (2.3 percent).

(33)

32

They found, just like this study, a declining effect when distance to a church became bigger. While this study used a treatment area of 150 meter, they found a significant effect up to 260 meter distances from a church.

Carroll et. al. found the opposite result in their study. A possible explanation for this contradiction could be that both studies reflect the effects accurately within their own communities, but that the aversion differs for each community.

Further research could try to distinguish the mosques in Amsterdam in more detail and focus on heterogeneity effects. Mosques differ a lot with respect to size, capacity and visibility for the surrounding neighborhood. Bigger, more visible mosques do have most likely more impact than public prayer rooms located into residential living rooms. Unfortunately, the data was not available to make this distinction in this study. Furthermore, it would be interesting to investigate empirically the effects of houses of worship of different religions although it might be hard to find sufficient data since the database of the municipality is not up to date. Finally, a comparison between some big cities would be interesting. Is the effect in Amsterdam the same as in Berlin, Paris or Madrid? Finally, after all, the Wester mosque has opened the doors for a public despite the heavy opposition from the local residents. According to the results of this research were the arguments of the opposition about a potential decline in transaction price justifiable. Houses within a 150 meter range of the mosque, on average, declined with 2.3 percent.

(34)

33

7. References

Ahlfeldt, G.M., Maennig, W., (2010) Homevoters vs leasevoters: A spatial analysis of airport effects, Journal of Urban Economics, Vol 87, pp85-99

Babawale, G.K., (2011) The Impact of Neighborhood Churches on House Prices., University of Lagos, Journal of Sustainable Development, Vol 4 No. 1

Brandt, S. Maennig, W., Richter, F., (2014) Do Houses of Worship Affect Housing Prices? Evidence from Germany., University of Hamburg, Vol 45 No. 4 pp. 549-570

Bolitzer, B., N.R., Netsusil, (2000) The impact of open spaces on property values in Portland, Oregon. Journal of Environmental Management , Vol 59.

Boxall, P. C., Chan, W. H., & McMillan, l. (2005). The Impact of Oil and Natural Gas Facilities on Rural Residential Property Values: A Spatial Hedonic Analysis. Resource and Energy

Economic 27(4)248-269.

Carrol, T.M., Clauretie, T.M., Jensen, J., 1996 Living Next to Godliness: Residential Property Values and Churches, Journal of Real Estate Finance and Economics, 12: 319-330

Clotfelter, C.T., (1975) The Effect of School Desegregation on House Prices, The MIT Press; The Review of Economics and Statistics, Vol 57 No. 4 pp. 446-451

Downes, T.A., Zabel, J.E., (2002) The impact of school characteristics on house prices: Chicago 1987-1991, Journal of Urban Economics, Vol 52 pp. 1-25.

Do, A.Q., Grudnitski. G., (1995), Gold Courses and Residential House Prices: An Empirical Examination, Journal of Real Estate Finance and Economics, pp. 261-270

Dröes, M.I., Koster, R.A., (2014) Renewable Energy and Negative Externalities: The Effect of Wind Turbines on House Prices, Tinbergen Institute Discussion Paper

Follain, J.R., Jimenez, E., (1985) Estimating willingness to pay for housing attributes, Regional Science and Urban Economics, Vol 24 No. 5 pp. 577-599

Harrison, D., Rubinsfeld, D.L.,(1978) Hedonic housing prices and the demand for clean air, Journal of environmental economics and management, Vol 5 No. 1 pp 81-102

Iroham, C.O., Oloyode, S.A., 2010 Location of Worship Centers and its Effect on Residential Property Values Paper Delivered at the First National Conference, Department of Urban and Regional Planning

(35)

34

Kaufman D. A. & Cloutier N. R. (2006). The Impact of Small Brownfields and Greenspaces on Residential Property Values. Journal of Real Estate and Economics, 33: 19–30.

Onderzoek, Informatie en Statistiek gemeente Amsterdam, 2015 Stadsdelen in cijfers , Chapter 1.5, pp. 27-31.

Quang Do, Wilbur, R.W. and Short, J.L., 1994 An Empirical Examination of the Externalities of Neighborhood Churches on Housing Values, Journal of Real Estate Finance and Economics, 9: 127-136

Rosen S. (1974). Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition. Journal of Political Economics, 82:34-55.

Sociaal Cultureel Planbureau., (20120) Moslim in Nederland

Schippers, D., (2014) Religie in Amsterdam, Bureau research and statistics Amsterdam, Chapter 2.6 pp. 17.

Seiler, M. J., Bond, M. T., & Seiter, V. L. (2001). The Impact of World Class Great Lakes Water Views on Residential Property Value. Appraisal Journal. 69 (3) 287-295.

Weicher, J.C., Zerbst, R.H., (1973) The Externalities of Neighborhood Parks: An Empirical Investigation, University of Wisconsin Press, Vol 49, No. 1 pp. 99-105

(36)

35

8. Appendix

Table (|) - Distribution of the observations over the areas.

Area Freq. Percent Cum. Area Freq. Percent Cum. Area Freq. Percent Cum.

A00a 203 0.24 0.24 E40c 229 0.27 36.36 M34d 1 0.00 75.26

A00b 13 0.02 0.25 E41a 302 0.35 36.71 M35a 27 0.03 75.29

A00e 22 0.03 0.28 E41b 354 0.41 37.12 M35b 142 0.16 75.45

A01b 182 0.21 0.49 E41c 361 0.42 37.54 M35c 181 0.21 75.66

A01c 82 0.10 0.58 E42a 311 0.36 37.90 M35e 136 0.16 75.82

A01d 223 0.26 0.84 E42b 397 0.46 38.36 M35f 342 0.40 76.22 A01e 182 0.21 1.05 E42c 568 0.66 39.02 M51b 87 0.10 76.32 A01f 46 0.05 1.11 E42d 418 0.49 39.50 M51c 71 0.08 76.40 A01g 20 0.02 1.13 E43a 1,169 1.36 40.86 M55b 1 0.00 76.40 A01h 21 0.02 1.15 E43b 773 0.90 41.76 M55c 74 0.09 76.49 A02a 734 0.85 2.00 E75a 275 0.32 42.08 M55d 385 0.45 76.93 A02b 595 0.69 2.70 E75b 209 0.24 42.32 M55f 3 0.00 76.94 A02c 131 0.15 2.85 E75c 128 0.15 42.47 M55g 282 0.33 77.27 A02d 27 0.03 2.88 E75d 82 0.10 42.56 M55h 148 0.17 77.44

A03a 73 0.08 2.96 F76a 890 1.03 43.60 M56a 601 0.70 78.13

A03b 9 0.01 2.97 F76b 336 0.39 43.99 M56b 561 0.65 78.79 A03c 53 0.06 3.04 F77a 237 0.27 44.26 M56c 1,323 1.54 80.32 A03d 180 0.21 3.24 F77b 821 0.95 45.21 M56d 65 0.08 80.40 A03e 54 0.06 3.31 F77c 615 0.71 45.93 M56f 188 0.22 80.61 A03f 24 0.03 3.33 F77d 99 0.11 46.04 M56h 62 0.07 80.69 A03g 61 0.07 3.41 F77f 61 0.07 46.11 M57a 175 0.20 80.89 A04b 112 0.13 3.54 F78c 1 0.00 46.11 M57b 1 0.00 80.89 A04c 218 0.25 3.79 F78d 57 0.07 46.18 N60a 165 0.19 81.08 A04d 215 0.25 4.04 F78e 62 0.07 46.25 N60b 107 0.12 81.21 A04e 4 0.00 4.04 F81b 118 0.14 46.39 N60c 303 0.35 81.56

A04f 22 0.03 4.07 F81e 182 0.21 46.60 N61a 286 0.33 81.89

A04g 155 0.18 4.25 F85a 379 0.44 47.04 N61b 386 0.45 82.34

A04h 70 0.08 4.33 F85b 172 0.20 47.24 N61c 137 0.16 82.50

A04i 90 0.10 4.43 F85c 122 0.14 47.38 N61d 11 0.01 82.51

A05a 120 0.14 4.57 F86a 192 0.22 47.60 N62a 90 0.10 82.61

A05b 295 0.34 4.92 F86b 342 0.40 48.00 N62b 194 0.23 82.84

A05c 481 0.56 5.47 F86d 198 0.23 48.23 N63a 132 0.15 82.99

A05d 584 0.68 6.15 F87a 482 0.56 48.79 N64a 33 0.04 83.03

A05f 19 0.02 6.17 F87b 16 0.02 48.81 N64b 21 0.02 83.05 A06a 760 0.88 7.05 F87c 74 0.09 48.89 N64c 238 0.28 83.33 A06c 806 0.94 7.99 F87d 144 0.17 49.06 N64d 59 0.07 83.40 A06d 199 0.23 8.22 F88f 28 0.03 49.09 N65a 148 0.17 83.57 A06e 156 0.18 8.40 F88g 78 0.09 49.18 N65b 390 0.45 84.02 A06f 87 0.10 8.50 F88h 6 0.01 49.19 N65c 338 0.39 84.41 A06g 350 0.41 8.91 F89a 245 0.28 49.47 N65d 23 0.03 84.44 A06h 10 0.01 8.92 F89b 412 0.48 49.95 N66b 63 0.07 84.51 A06i 48 0.06 8.98 F89c 26 0.03 49.98 N66c 266 0.31 84.82 A06k 449 0.52 9.50 K23e 42 0.05 50.03 N66d 437 0.51 85.33 A06l 650 0.75 10.25 K24a 43 0.05 50.08 N66f 241 0.28 85.61 A07a 49 0.06 10.31 K24b 62 0.07 50.15 N66g 391 0.45 86.06

A07b 3 0.00 10.31 K24e 263 0.31 50.46 N67a 375 0.44 86.50

A07c 31 0.04 10.35 K25b 261 0.30 50.76 N67b 160 0.19 86.68 A07e 166 0.19 10.54 K26b 18 0.02 50.78 N68a 432 0.50 87.19 A07f 48 0.06 10.60 K26c 70 0.08 50.86 N68b 240 0.28 87.46 A07g 170 0.20 10.79 K44a 391 0.45 51.32 N68c 481 0.56 88.02 A07h 87 0.10 10.89 K44b 1,45 1.68 53.00 N68d 71 0.08 88.10 A08a 87 0.10 10.99 K44c 1,556 1.81 54.81 N68e 51 0.06 88.16 A08b 126 0.15 11.14 K44d 958 1.11 55.92 N68f 160 0.19 88.35

(37)

36

Area Freq. Percent Cum. Area Freq. Percent Cum. Area Freq. Percent Cum.

A08d 446 0.52 11.66 K44e 436 0.51 56.42 N69a 38 0.04 88.39 A08e 98 0.11 11.77 K44f 1 0.00 56.42 N69c 168 0.19 88.59 A09b 142 0.16 11.94 K45a 833 0.97 57.39 N69j 106 0.12 88.71 A09c 408 0.47 12.41 K45b 322 0.37 57.76 N69k 451 0.52 89.23 A09d 267 0.31 12.72 K46a 37 0.04 57.81 N69l 188 0.22 89.45 A09e 275 0.32 13.04 K46b 534 0.62 58.43 N69m 616 0.71 90.17 A09f 116 0.13 13.17 K46c 198 0.23 58.66 N70a 298 0.35 90.51 A09h 139 0.16 13.34 K47d 51 0.06 58.72 N70b 261 0.30 90.82 A09i 514 0.60 13.93 K48a 47 0.05 58.77 N70c 45 0.05 90.87 B10d 3 0.00 13.94 K48b 15 0.02 58.79 N70d 338 0.39 91.26 E12a 1 0.00 13.94 K48c 30 0.03 58.82 N70e 59 0.07 91.33 E12b 191 0.22 14.16 K48d 61 0.07 58.89 N70f 47 0.05 91.38 E13a 360 0.42 14.58 K48e 1 0.00 58.89 N71c 2 0.00 91.39 E13b 302 0.35 14.93 K48f 131 0.15 59.05 N72a 18 0.02 91.41 E13c 140 0.16 15.09 K49a 18 0.02 59.07 N73a 4 0.00 91.41 E13d 268 0.31 15.40 K49d 71 0.08 59.15 N73b 71 0.08 91.49 E13e 196 0.23 15.63 K49e 171 0.20 59.35 N73d 76 0.09 91.58 E13f 141 0.16 15.79 K49f 331 0.38 59.73 N73f 46 0.05 91.64 E14a 233 0.27 16.06 K52b 110 0.13 59.86 N73g 14 0.02 91.65 E14b 79 0.09 16.15 K52c 155 0.18 60.04 N73h 46 0.05 91.71 E14c 166 0.19 16.35 K52g 8 0.01 60.05 N74a 13 0.02 91.72 E14d 1,023 1.19 17.53 K52h 2 0.00 60.05 N74c 78 0.09 91.81 E14e 921 1.07 18.60 K53a 26 0.03 60.08 T93a 559 0.65 92.46 E14f 472 0.55 19.15 K59a 239 0.28 60.36 T93b 613 0.71 93.17 E15a 185 0.21 19.36 K90a 408 0.47 60.83 T93d 90 0.10 93.27 E15c 7 0.01 19.37 K90c 651 0.76 61.59 T93f 85 0.10 93.37 E15d 162 0.19 19.56 K91a 73 0.08 61.67 T93g 221 0.26 93.63 E16a 755 0.88 20.44 K91b 218 0.25 61.92 T93i 33 0.04 93.67 E16b 756 0.88 21.31 K91c 506 0.59 62.51 T93j 36 0.04 93.71 E16c 151 0.18 21.49 M28a 538 0.62 63.14 T94a 8 0.01 93.72 E17a 528 0.61 22.10 M28c 643 0.75 63.88 T94b 196 0.23 93.95 E18a 411 0.48 22.58 M28d 486 0.56 64.45 T94c 18 0.02 93.97 E18b 703 0.82 23.39 M29a 574 0.67 65.11 T94d 185 0.21 94.18 E19b 258 0.30 23.69 M29b 445 0.52 65.63 T94e 26 0.03 94.21 E19c 413 0.48 24.17 M29c 9 0.01 65.64 T94h 21 0.02 94.24 E20b 1 0.00 24.17 M30a 760 0.88 66.52 T94i 57 0.07 94.30 E20c 663 0.77 24.94 M30b 584 0.68 67.20 T94j 1,391 1.61 95.92 E21a 1,748 2.03 26.97 M31a 1,023 1.19 68.39 T94k 301 0.35 96.27 E21b 554 0.64 27.61 M31b 982 1.14 69.52 T94n 33 0.04 96.30 E22a 20 0.02 27.64 M31c 371 0.43 69.96 T95a 669 0.78 97.08 E22b 11 0.01 27.65 M32a 678 0.79 70.74 T95b 1 0.00 97.08 E36a 12 0.01 27.66 M32b 490 0.57 71.31 T95c 5 0.01 97.09 E37a 62 0.07 27.73 M32c 56 0.06 71.38 T96b 148 0.17 97.26 E37c 1,072 1.24 28.98 M33a 114 0.13 71.51 T96c 212 0.25 97.51 E37d 762 0.88 29.86 M33b 408 0.47 71.98 T96d 215 0.25 97.75 E37e 289 0.34 30.20 M33c 407 0.47 72.45 T96e 299 0.35 98.10 E37f 626 0.73 30.92 M33d 698 0.81 73.26 T96f 97 0.11 98.21 E37g 579 0.67 31.60 M33e 406 0.47 73.73 T96g 168 0.19 98.41 E38c 1,393 1.62 33.21 M33f 320 0.37 74.11 T97a 278 0.32 98.73 E38d 414 0.48 33.69 M33g 359 0.42 74.52 T97b 137 0.16 98.89 E39a 142 0.16 33.86 M33h 626 0.73 75.25 T97c 595 0.69 99.58 E39c 30 0.03 33.89 M33i 1 0.00 75.25 T97d 177 0.21 99.79 E39d 139 0.16 34.05 M33j 1 0.00 75.25 T98a 181 0.21 100.00 E40a 1,05 1.22 35.27 M33k 1 0.00 75.25 T98b 3 0.00 100.00 E40b 708 0.82 36.09 M34a 3 0.00 75.26

Referenties

GERELATEERDE DOCUMENTEN

indistinguishable from the thorn during the Middle Ages. This resulted in words like ye and yat, which I have changed into the and that. In addition, I also modernised the use of

The cost optimization has the strengths of an energy system coverage, evaluates the effect of overall parameters like biomass potential and competition between

We have shown in the above how journalistic entrepreneurs throughout the world focus on making a difference, having an impact. We have also shown that both the forms of making

Bubbles rising in ultra clean water attain larger velocities that correspond to a mobile (stress free) boundary condition at the bubble surface whereas the presence of

Non-parent couples can act independently as they do not have the constraint of children to look after (Duxbury &amp; Higgins, 2001). Therefore it seems that the

Numerical analyses for single mode condition of the high-contrast waveguides are performed using the software Lumerical MODE.. The calculated optimal parameters for the structure

Deze hebben te maken met: tweedeling van de studentenpopulatie, kosten, beperkte uitstralingseffect naar reguliere programma’s, motivatie van honoursstudenten ook

AC acquisition cost AR area cost rate CC component cost MC material cost MH machine hour rate P, p process steps PC production costs PR machine state PQ