The economiceExternalities of prostitution : the case of the Amsterdam Red Light District

(1)

Master Thesis VU / UvA: Entrepreneurship

The Economic Externalities of Prostitution:

The case of the Amsterdam Red Light District

Joris Z. van den Berg

*

Supervised by: Rafael P. Ribas

Date of Review: July 1, 2016

Abstract

Since the late twentieth century, prostitution in Red Light Districts (RLDs) has become increasingly intertwined with the tourist and entertainment industry. Besides perceived negative social implications, one may argue that RLDs create viable economic opportunities. However, there are not many studies that quantify the effect of a RLD on the local economy. With the use of a density discontinuity design, this thesis shows that the Amsterdam RLD is an economic hotspot that attracts businesses and employment. Moreover, with recent policies intended to reduce its size, in particular the implementation of Project 1012, which closed 27% of prostitution windows, the attractiveness towards businesses and employment has increased in the RLD area. Especially, around closed prostitution windows business and employee density seems to increase. Therefore, although Amsterdam’s RLDs are located right in economic hotspots, reducing prostitution in the area has created business opportunities and employment. This is, however, not to say that governments should completely ban prostitution as this may decrease its attractiveness to tourists. Moreover, as prostitutes often relocate to other areas economic gains in the RLD may be offset by economic losses elsewhere.

(2)

(3)

Foreword

My entire student life took place in Amsterdam. I have always been fascinated by the city and did not want to leave university without having researched it once to some degree. When the opportunity passed by to research the Amsterdam Red Light District in a way that is relevant and able to contribute to theory and practice, I immediately got excited. The Amsterdam Red Light District has always fascinated me. I find it very intriguing that on the one hand, prostitution is increasingly scrutinized and is often related to crime, drugs and other negative social impacts. While on the other hand, the Red Light District may be one of Amsterdam’s best selling points towards tourists, which in turn fuels the economy. To get more grip on organized crime and increase the attractiveness of the area, the municipality implemented Project 1012, which aims to decrease the Red Light District with almost 50 percent. However, only a few researchers have looked at what implications the shrinking of the Red Light District has on social and economic issues. Especially economic externalities have been neglected in academic literature. I am very fulfilled that with this thesis I shed more light on this issue and contribute the development of this research domain.

Writing this thesis has been a very interesting and challenging job. I have learned a great new deal about the city, its RLD but also on how to conduct a well-structured quantitative research. In the end, I am very satisfied with the end product. Next to my own efforts, several other people have aided me in the process. To these people I would like to show my gratitude. Firstly, I want to thank Rafael, my thesis supervisor, for the possibility of doing this research and for his guidance throughout the process. Secondly, I want to thank Ivo, from the municipality of Amsterdam that was willing to provide the necessary data to execute this study.

Hopefully you will read this thesis with the same amount of joy as I had writing it. For any comments or questions on this thesis you can contact me at jorisberg@gmail.com.

Sincerely,

(4)

Introduction

Prostitution has been part of urban and social life since the first cities were founded. There are no cities without prostitution and for centuries prostitution has been regarded as a ‘necessary evil.’ In most parts of the world, prostitution is illegal, though to some extent tolerated (Aalbers and Sabat, 2012). Due to its controversial nature, prostitution is often spatially concentrated in ‘Red Light Districts’ (RLDs), which often accommodate other adult activities, such as strip clubs and adult theatres (Ashworth et al., 1988). In many countries debates are raging whether prostitution should be regulated and decriminalized in order to reduce social nuisances (Aalbers and Deinema, 2012). Multiple studies have addressed this issue by looking at the social impacts of regulating and shrinking RLDs. However, in this debate economic externalities are often overlooked. Only a handful of researchers suggestively argue that RLDs may create viable (legal) economic opportunities (Ashworth et al., 1988; Brents and Saunders, 2010 and Aalbers and Sabat, 2012). To make a holistic decision on how to regulate RLDs it is crucial to have an insight in its economic externalities. Overlooking economic effects of the RLD may have direct as well as indirect consequences for the economy. However, up to my best knowledge the economic externalities of RLDs and how these are influenced by policies has yet to be empirically researched.

This thesis aims to gain a more comprehensive insight in the interplay between RLDs and the economy and proposes to quantify the influence of a RLD on local businesses and employment. To do so, I exploit the case of the Amsterdam RLD. Amsterdam holds one of the most famous RLDs in the world: De Wallen. Furthermore, this city provides a unique setting for the identification of direct economic externalities. Firstly, the borders of a RLD are often arbitrary. In Amsterdam canals border some parts of the RLD. The area around the canals are good counterfactuals as the main difference across the canal is the RLD. Secondly, as Amsterdam has recently started to shrink its RLD, it is possible to estimate the difference before and after the closing down of prostitution windows. Therefore, although using the canals still produces some bias, the case of Amsterdam does allow to estimate the effect of prostitution on local economic activity as prostitution decreases. By estimating the business and employee density and plotting the result on a map, this thesis shows how businesses and employees are spread out in Amsterdam and its RLDs. Afterwards, with the use of the McCrary test, this thesis investigates how the density of businesses and employees develops across the border of the RLD.

Prior studies show some initial suggestive evidence that RLDs hold the potential to open up new economic opportunities, where for instance restaurants, shops and governments are able to benefit (Brents and Saunders, 2010). However, these opportunities are argued to be particularly in the sex and tourist industry as both are increasingly connected and reinforcing each other (Burtenshaw et al., 1981; Aalbers and Sabat, 2012). Indeed, Frondizi and Porcher (2012) note that prostitution may attract businesses that complement the sex industry but repel more traditional types of businesses. Ryder (2004) and Aalbers and Sabat (2012) also support this notion and show that due to nuisance and high

(7)

rents traditional businesses are often not attracted to locate inside a RLD. In essence, the available literature on economic externalities of RLDs indicate that RLDs can be attractive for businesses and employment but more so to sex and tourist industries than to other industries. However, how the closing down of RLDs influences this attractiveness has hardly been discussed.

At present, the municipality of Amsterdam experiments with shrinking the RLD. In 2007, the municipality launched Project 1012, to make the RLD a more attractive place for businesses, tourists and residents. The program aims to reduce organized crime and improve the image of the area by closing down businesses with little economic value. Initially, the municipality wanted to close down 47 percent of the prostitution windows. Due to several delays and protests by sex workers, this number has often been adjusted (Aalbers and Deinema, 2012). At the time of writing, approximately 27 percent of the prostitution windows have been closed (VRP2012, 2016).

The results of this thesis, give some initial support to the notion that businesses and employees rather locate inside vis-à-vis outside the Amsterdam RLD. Moreover, as prostitution exits the RLD its attractiveness towards businesses and employees increases. Especially, since the implementation of Project 1012, it is observed that the attractiveness of the RLD towards businesses and employees has increased. In essence, these findings indicate that the RLD is in an economic hotspot, where the market size of prostitution negatively affects the attractiveness towards businesses and employment.

However, despite that shrinking the RLD appears to be an effective policy to increase local economic activity, it does not necessarily imply that the municipality should close down the RLD completely. The touristic nature of the RLD is likely of important influence to the attractiveness of the RLD. Arguably, there is a tipping point at which the size of the prostitution market will no longer negatively influence the local economy. Losing its touristic appeal can have detrimental effects for the entire city of Amsterdam. Furthermore, the literature indicates that prostitutes rather relocate than exit the business. Therefore, the municipality has to consider how it will tackle possible negative economic effects in other areas as prostitution displaces.

The rest of this thesis is divided into five parts. Section 1 provides an overview of prior academic literature and shows how the Amsterdam RLD has developed. Section 2 relates to what data has been used on how this was obtained. Section 3 discusses the empirical methods that are used in this study. Section 4 presents the findings of this study for businesses and employees in the RLD. Section 5 discusses the main findings and their implications and limitations. Finally, a conclusion elaborates on the main message of this thesis.

(8)

1 Literature Review

Despite its long history, social significance and perceived connections with social, political and economic domains, prostitution has been a somewhat neglected area in academic literature. In this paragraph I relate to the key insights that academic literature provides on prostitution and in turn show where researchers have refrained from providing empirical evidence.

Firstly, I provide an overview of how prostitution is driven by supply and demand and how prostitution is embedded in the legal and illegal economy. Secondly, I focus on how laws and regulations influence prostitution. After a short connecting paragraph, I zoom in on the case of Amsterdam and show how the Amsterdam RLDs developed into a global sex industry and to what extent it is regulated.

1.1 The economics of prostitution

Prostitution, also famous for being the oldest profession, has always been around (Giusta et al., 2009). Here, for the sake of clarity, prostitution is defined as the act where an individual receives a payment (whether pecuniary or not) for offering sexual services (Nagle, 1997). In essence, economists argue that prostitution can be seen as a commercial activity that involves consumer choice and comparative shopping. To this extent prostitution can be regarded as a commodity in which supply and demand determines its price (Hubbard, 2003; Giusta et al., 2009).

Economists have only very recently started to research prostitution. Research in this relatively new domain has focussed on modelling supply and demand. Here, several different methods have been used to explain how price equilibriums emerge and what factors can influence this equilibrium (Giusta et al. (2009). Research shows that demand of men for prostitutes is moderated by the perceived risk of sexual transmitted diseases (Cameron and Collins, 2003); prices are affected by the duration of the act, the age of the prostitute and the location (Peters, 2004) and that the use of condoms can lower incomes for prostitutes (Rao et al., 2003). Moreover, several researchers argue it is actually demand that partly influences supply, and not the other way around (Flight and Hulshof, 2009; Kővári, and Pruyt, 2012). According to Flight and Hulshof (2009) only a small part of the clients of prostitutes constitute the majority of the visits, as these clients are frequent visitors. In this sense, cutting supply will not necessarily cut demand. In general, the ‘microeconomics of prostitution’ gives an interesting insight in how supply and demand meet. However, it does not adequately address its entanglement with other economic functions. Hence, the ‘macroeconomics of prostitution’, which looks at the economic externalities of prostitution, has been fairly neglected in research.

(9)

In general, the externalities of prostitution have enjoyed limited coverage in academic literature. According to Ashworth et al., (1988) this is partly due to the position of prostitution on the edges of morality that caused a general ignorance on how to address prostitution. Ashworth et al. (1988), is one of the first researchers that raised the notion that prostitution areas contribute to the cognitive picture residents and visitors have of the city and that it is related to economic and social functions of urban life. Prostitution thrived in the last decades of the twentieth century, with the emergence of the global sex industry (Brents and Saunders, 2010). In turn, an increasing amount of researchers regard prostitution as a significant urban activity that is related to economic and social functions (Ashworth et al., 1988; Brents and Saunders, 2010; Aalbers and Sabat, 2012). Despite that still only a small amount of research is undertaken on the externalities of prostitution, some interesting (suggestive) connections are shown.

Links with the illegal economy

Hubbard (1998) argues that people generally link prostitution to wider problems of criminality, deprivation and environmental pollution. Here, prostitution is often seen as a cause rather than a symptom of problems that are apparent in a certain area. Based on moral anxiety it is prostitution that is often seen as the ‘scapegoat’ (Hubbard, 1998). However, though prostitution may not necessarily be the cause of organized crime, it does become apparent that both are strongly connected.

Prostitution is illegal in most parts of the world, which makes it deeply intertwined with the global illegal economy. This connection is strengthened by the fact that prostitution is often associated with human trafficking. If both the market for prostitution and human trafficking are added to each other they comprise one of the largest global illegal economies (Global Risks, 2011). Kővári and Pruyt (2012) argue that not only prostitution and human trafficking are connected, but also argue that all illegal economies are strongly interlinked and show this with the use of a conceptual model (see Figure 1). Since the turn of the century, organized crime got deeply involved in the Dutch prostitution industry. Official figures show that due to organized crime the majority of the prostitutes are enslaved and coerced to become a sex worker (GRTP, 2009; Kővári and Pruyt, 2012). In turn, the huge profits from prostitution are used to reinvest or sustain other illegal markets. Thus, in essence, due to human trafficking and prostitution there is more money for other illegal practices (Kővári and Pruyt, 2012).

What becomes apparent in the research of Kővári and Pruyt (2012), is that prostitution is big business that is deeply rooted in the illegal economy and in fact sustains this connection by financing other illegal activities. Despite the social and ethical problems that are raised by this connection, not interfering with prostitution can also be seen as a foregone economic opportunity for the legal economy. Especially governments forego an incredible sum of tax money (VAT as well as income tax). Despite the significance of prostitution for the illegal economy, it is particularly interesting how prostitution interacts with the legal economy.

(10)

Figure 1: Conceptual model of how prostitution is intertwined with other (illegal) industries

Source: Kővári and Pruyt, 2012

Links with the legal economy

Due to several conflicting forces prostitution typically tends to concentrate in specific parts of the city (Aalbers and Sabat, 2012). In order to maximize their profits, prostitutes need be accessible to their clientele. Concentrating in certain areas makes it easier for clients to find them. Besides accessibility there also needs to be the opportunity to provide the ‘service.’ Moreover, because prostitution is a marginalized and heavily scrutinized activity the location of it is subject to multiple external pressures and constraints (Ashworth, 1988). In the end, the boundaries of a prostitution district are determined by the interplay between clients, prostitutes, protestors, police and politicians (Hubbard and Sanders, 2003). How these forces interact gives rise to three different configurations of prostitution areas, see Figure 2. When a prostitution district is heavily concentrated and to some degree visible to the public, it is typified as a Red Light District (RLD). Here, it is irrelevant to what extent prostitution is intertwined with the entertainment industry, as long as prostitution is clearly visible in the area we can speak of a RLD (Aalbers and Deinema, 2012). To this extent, multifunctional and monofunctional districts can both be identified as RLDs.

(11)

Figure 2: Locational model of urban prostitution areas

Source: Ashworth et al., 1988

It is especially RLDs that have gained attention on how prostitution interacts with the legal economy. For instance, Brents and Saunders (2010), show that as RLDs develop into a sex industry it becomes intertwined with the mainstream economic market and opens up new opportunities for restaurants, shops and governments (VAT). Aalbers and Sabat (2012) add some nuance to this notion by showing that in general businesses do not want to locate inside a RLD due to noise, nuisance and suggestive relations with crime, drugs and homeless people. On the other hand, Aalbers and Sabat (2012) show that there are clear connections between the sex industry and tourism. A tourist may buy souvenirs, eat at a restaurant and may be interested in other tourist attractions. Several other academic works also highlight that the sex and tourist industries are merging and reinforcing each other (e.g. Burtenshaw et al., 1981; Brents and Hausbeck, 2009). In essence, these researchers provide suggestive evidence (as their claims are not grounded in empirical research) that RLDs mainly attract businesses in the sex and tourist industry.

Frondizi and Porcher (2012) strengthen this observation. The authors show how prostitution thrived and pushed legitimate businesses out of the district during the Belle Époque in Paris (1870-1914). In essence, the formal economy switched to informal activities, as these activities were more profitable. This study implies that prostitution may attract businesses that can complement the industry but repel more traditional types of businesses (e.g. shops and craftsmen). Despite its illegality, a boom in prostitution did go hand in hand with higher profits for shopkeepers and tenants that diverted their economic activity to prostitution. This finding thus supports the notion that prostitution can increase

(12)

revenues for businesses, as long as this business is connected to prostitution. Ryder (2004) gives an additional explanation for this phenomenon and argues that adult entertainment activities push out other activities, not so much due to a perceived negative social image but more so because of the inability of other types of businesses to pay the (high) rents. The adult entertainment industry is able to pay these (high) rents, as it is a highly profitable industry (Ryder, 2004). In turn, as other types of businesses do not have the financial means and profits they tend to locate elsewhere.

Finally, some researchers provide suggestive reasoning that prostitution also holds the potential to boost the local economy that is not directly linked to sex and tourism (Lee, 1994; Edlund and Korn, 2001; Van Wijk et al., 2010). Prostitution is a lucrative business, not only for pimps and exploiters but also for prostitutes (Edlund and Korn, 2002. Prostitutes often spend a high proportion of their income on clothes and appliances (Van Wijk et al., 2010). In essence, the money earned (despite its illegalness) in the prostitution industry has the potential to fuel the economy in other areas.

1.2 Laws and regulations

The RLD has always been placed on the edges of morality, at which there is a constant debate about its social impact. On the one hand, Aalbers and Sabat (2012) argue that RLDs are unpopular areas for businesses and residents, mainly due to their perceived relations with crime, drugs and homeless people. Though on the other hand, RLDs can actually be of social value for minorities, as RLDs can be seen as a zone of tolerance where the ‘undesirables’ that live on the edges of morality can seek refuge and feel ‘at home’ (Ward, 1975; Hekma, 2007).

In most parts of the world, policymaking has been preoccupied with how it can and should handle the social impacts of prostitution, such as human trafficking; protection of prostitutes; ties between criminality and drugs; immigration; and the displacement of prostitutes (Daalder, 2007). Multiple researchers have investigated to what extent the law can control prostitution and to what extent prohibiting and displacing prostitution has positive social outcomes. In general, the studies are consistent and support the view that regulation can influence the geographic location of prostitution. However, it has little influence in actually prohibiting prostitution. Interestingly, this effect is noted in different locations (Canada, US, UK, Netherlands) and over time (Shumsky and Springer, 1981; Hubbard, 1998; Kohm and Selwood, 2003; Van Wijk et al., 2010). Van Wijk et al. (2010) provides an insightful example of what usually happens when a RLD is closed down or shrinks. Here, the authors explicitly show that prostitutes that engaged in window prostitution generally do not transfer to other sectors (not even within prostitution). This is particularly due to the independence and profitability of window prostitution. In essence, these researchers show that closing down a RLD does not necessarily decrease prostitution, it rather replaces it.

(13)

Hubbard and Sanders (2003) extend this notion that multiple forces determine the boundaries of RLDs. In this sense, closing down a RLD by the government may work counterproductive, as it is only one of the forces that set the boundaries. Moreover, Flight and Hulshof (2009) show that only a small part of the clients of prostitutes constitute the majority of visits to prostitutes. This observation implies that cutting the supply of prostitution will not necessarily cut the visits to prostitutes by the same amount. Moreover, Flight and Hulshof (2009) show that when demand cannot be satisfied in one area, customers will go to another area. In turn, Flight and Hulshof (2009) conclude that demand is only to a small extent determined by supply, but demand partly influences supply. Kővári and Pruyt (2012), also argue that the most important element of an effective policy appears to be the ability to reduce demand. However, this should not be done by illegalizing prostitution, besides the accompanying costs of prosecuting, it is expected to only further strengthen ties of prostitution with the illegal economy, increase human trafficking and further worsen labour conditions for prostitutes (Kővári and Pruyt, 2012). Van Wijk et al. (2010) shows that indeed labour conditions can deteriorate as the supply of prostitution is lowered while demand stays the same. As there is less available space for prostitutes to work, rent prices increase. This rise leads to higher competition and in turn lower prices. Moreover, prostitutes have to work longer hours and have less job security. In essence, the power of exploiters of brothels and prostitutes rises at the expense of labour and social position of the prostitute.

Instead of illegalizing or shrinking RLDs, several governments have tried to regulate the industry to get more grip on organized crime. However, it appears that screening and monitoring brothel owners often do not provide sufficient transparency to expose human trafficking and other crimes. It is doubtful that regulation can drive out organized crime, though combating organized crime may be even harder if prostitution is illegal (Huisman, 2014). In essence, research has not been able to present the ‘Holy Grail’ to governments on how to combat prostitution and its negative impacts.

In sum, the first two sections indicate that there is a handful of researchers that have given suggestive evidence on the economic externalities of RLDs. The main findings are that prostitution, tourist and entertainment industries have become intertwined and are suggested to reinforce each other. Despite their suggestive links to the illegal economy, it also becomes apparent that other (legal) businesses tend to move out. Moreover, despite that the fact that closing down RLDs rather dislocates prostitution instead of decreasing prostitution, it does influence the composition of the RLD. The economic effects of shrinking the RLD have to the best of my knowledge never been analysed nor quantified. To gain an insight in whether RLDs actually attract or repel (legal) businesses and how this changes over time as the RLD shrinks, this study will adopt a case study point of view.

(14)

1.3 The case of Amsterdam

Amsterdam currently has three RLDs: de Wallen, one of the most famous RLDs in the world;

Singelgebied, also located inside the city centre; and Ruysdealkade, a smaller district outside the city

centre. Because of the size and location, this study will focus on the focus on the RLDs in the city centre. Figure 3 shows the location of both RLDs, here both districts are marked red and a purple line marks the border of the city centre.

Figure 3: Map of Amsterdam highlighting the RLDs and city centre border

Computed in R

1.3.1 Description of RLDs

In both RLDs the main source of prostitution is window prostitution. though both districts have a significant different character.

De Wallen

De Wallen is one of the most famous RLDs in the world (Aalbers and Sabat, 2012). Despite having the largest availability of window prostitutes in Amsterdam (75 percent), De Wallen has become a real tourist attraction (Flight and Hulshof, 2009; Van Wijk et al., 2010). It has become an entertainment area,

(15)

or as some researchers argue, a real ‘theme park’ (Nijman, 1999; Van Straaten, 2000). The ‘theme park’ comprises of bars, coffee shops (places to smoke marihuana), hotels, restaurants, peep shows, strip clubs, porn cinemas and sex shops. In conclusion, De Wallen has developed into a place where tourists can come to see rather than purchase prostitution (Aalbers and Sabat, 2012). An impression of what the De Wallen looks like is shown in Figure 4.

The prostitutes that work on De Wallen are female, though also some males (trans-genders) work here. The prostitutes at De Wallen are relatively young, with a mean age of 26 years. Nearly two-thirds of the prostitutes come from Eastern Europe and only 17 percent of the prostitutes is of Dutch decent (Van Wijk et al., 2010).

Figure 4: Impression of De Wallen RLD

Source: Napnieuws, 2016 (left); Amsterdam, 2016 (right)

Singelgebied

The Singelgebied is a much smaller area that only hosts about 20 per cent of the prostitution windows in Amsterdam (Flight and Hulshof, 2009). In contrast to De Wallen, it is not a touristic oriented district in the sense that there are no sex clubs and peep shows. Also this district is more intertwined with a residential and business function. In essence, it is a smaller and quieter district that is focussed on prostitution instead of on the entire ‘theme park’ (Aalbers and Sabat, 2012). An impression of what the Singelgebied looks like is shown in Figure 5.

The prostitutes that work in the Singelgebied have a significantly different profile than the prostitutes working at De Wallen. Here, also the majority of the women are from Eastern Europe but they are generally older and more independent (Van Wijk et al., 2010).

(16)

Figure 5: Impression of Singel RLD

Source: Brante, 2016

1.3.2 Development of De Wallen

In Amsterdam, despite is position at the edges of morality, prostitution has always been regarded as a social necessity and part of city life. In turn, for the larger part in history, Amsterdam had a tolerant perspective towards prostitution, regarding it as a ‘necessary evil’ (Brants, 1998). Being worried about its image abroad with respect to criminal activities that are often associated with prostitution (such as gambling, violence, corruption and human trafficking), tolerance did not came at the expense of imposing rules and regulations. The first licensed brothels were opened nearly 600 years ago. Ever since, due to occupancies by foreign powers and religious takeovers their legality has changed from time to time. Nevertheless, brothels never exited the scenery (Brants, 1998; Van de Pol, 2000).

Emergence of De Wallen

At the start of the twentieth century the majority of Dutch parliament was hold by religious parties. After a long abolitionist movement, the government passed the ‘Morality Laws’ in 1911. These laws criminalized the exploitation of prostitutes in brothels or through pimping. Prostituting itself was not officially criminalized (De Vries, 1997). However, these laws were never enacted. In practice, Amsterdam adopted a gedoogbeleid (pragmatic tolerance) towards prostitution. Though it was by law possible to be convicted, in practice as long as prostitution sticked to the designated RLDs and public order was not threatened, authorities turned a blind eye (Brants, 1998; Outshoorn, 2012).

In the 1930s, driven by the ban on brothels, the first window prostitutes made their appearance (Goodyear, 2009). Window prostitution became more popular ever since and started to gradually attract tourists to the Dutch RLDs. Surprisingly, at the time, De Wallen was one of the safer parts of Amsterdam. Even though the attitudes towards prostitution were mixed De Wallen were considered as

(17)

just another non-threatening neighbourhood. In the 1960s the prostitution windows did not make the neighbourhood but they were part of it. It was in fact a very heterogeneous neighbourhood, with schools, grocers and shops (Sabat, 2012).

The rise of sex tourism

The 1970s marked the beginning of a new era in prostitution. As laws on pornography were relaxed the RLDs proliferated and started to develop into a sex industry. Global forces started to shift local circumstances in the RLD (Wonders and Michalowski, 2001). On the one hand, organizational changes allowed the growth of sex tourism as an industry. On the other hand, sex tourists and sex workers gained an increasingly globalized character. Together these forces provided a shift from sexual services to a sex industry. Many new options opened up in order to reap economic benefits, such as opening strip clubs, live nude shows and adult cinemas. Moreover, due to increased mobility and communications, De Wallen attracted a more global customer character (Paasonen, 2007; Brents and Saunders, 2010). The boom in the sex industry and tourism went hand in hand with a boom in commercial activity, as these tourists also needed to sleep and eat (Wonders and Michalowski, 2001). In essence, increased globalization has shaped the growth and character of sex tourism. Sex tourism was increasingly shaped by migration and the alignment to the needs and desires of the consumer (McDowell, 2009). In sum, globalization and migration have increased the production of sex tourism by commoditizing the body, this attracted a large stream of consumers that wanted to buy, see and experience these ‘services’ (Wonders and Michalowski, 2001).

Fuelled by globalization and the emergence of the sex industry, De Wallen developed into an adult theme park. Numerous coffee shops, sex shops, adult cinemas, condomeries, peep shows, smart shops and souvenir shops started to develop around the window prostitutes (Nijman, 1999). English started to dominate commercial signs and advertisements, and just like their customers most window prostitutes were of non-Dutch descent (Dignum, 1990). This image is quite contradictory to the image of De Wallen in the 1960s, where at the time it was a district that served multiple functions and where tourism was non-existent (Heinemeijer, 1968; Nijman, 1999). Interestingly, the RLD located in the Singelgebied did not follow the same trajectory and remained its low-profile image. Looking back at the typology put forward by Ashworth et al. (1988) it becomes apparent that over time De Wallen has transformed from a multifunctional district to a monofunctional entertainment area, while the Singelgebeid remained a multifunctional district.

At De Wallen, prostitution, or better said the sex industry, became booming business. As the industry proliferated, criminality became more intertwined with the district. Drugs began entering the scene, and as the RLD attracted immigrants human trafficking became an increasing problem (Outshoorn, 2012). The high profits made brothel owners and pimps invest in new sort of business, like

(18)

restaurants and casino’s. In essence, these businesses were only a cover up for the complex criminal network that was evolving under the blind eye of the authorities. After the 1970s, the situation at De Wallen was no longer sustainable and it developed into a no-go area for the police (Brants, 1998).

Closing down De Wallen

As brothels were still considered illegal by law, the authorities were unable to regulate the business. The situation at De Wallen started to grow out of control. The need to change this and start cleaning up and organizing the sex industry started to enter the political realm in the 1990s (Aalbers and Deinema, 2012; Zuckerwise, 2012). However, it was not until 2000 that the ban on brothels in the Netherlands was lifted. The law aimed to acknowledge prostitution and regard it as a ‘normal’ profession. The RLDs were allowed to exist but had to oblige to certain rules that were intended to improve the working conditions of prostitutes and decrease criminality in the region (Sabat, 2012). However, criminal activities continued to dominate the sex business. To gain more control on the situation the BIBOB (Bevordering Integriteitsberoordelingen door het Openbaar Bestuur) act went into effect in 2003. The act enables the municipality to revoke or refuse contracts, subsidies and permits if it conceives that the organization has links with crime. In total 30 permits were revoked or refused under the BIBOB act. The most successful case was in 2007. Here, the municipality gained control over eighteen premises by settling with a large brothel exploiter (Bureau BIBOB, 2009; Huisman and Nelen, 2014).

However, by 2007 it was decided that prior efforts were insufficient to significantly reduce organized crime in the RLD. It was argued by Van Traa (2007) that the RLD remained an attractive area for criminogenic activity due to its concentration of brothels, coffee shops, smart shops, gambling houses and bars. In turn, the authorities decided to change the reputation of the RLD and change the functional use of the area. It led to the initiation of Project 1012 (Huisman and Nelen, 2014). Project 1012 is based on two main pillars, reducing organized crime and improving economic activity in the area. Its goal is to turn the RLD into a more attractive place for businesses, residents and tourists. To achieve this Project 1012 aims to close about half of the prostitution windows (VRP1012, 2016). At the time of writing, Project 1012 is still running, however due to many delays and conflicts with sex workers its aims and methods have changed over time (Aalbers and Deinema, 2012).

Despite whether the above-mentioned policies are actually effective, prostitution windows have been and are being closed. Quite frankly, it remains unclear what these exact numbers are and when the actual number of windows started to decrease. For instance, Van Wijk et al. (2010) argues that De Wallen increased rapidly after the brothel ban in 2000, from 200 to 400 prostitution windows. Other sources show that De Wallen only decreased in size after the brothel ban (Wageneaar et al., 2013). Even municipality sources seem to be inconsistent, here the OIS (Research institute of municipality of Amsterdam) argues that in 2007 there were 470 active prostitution windows, while the official report of

(19)

Project 1012 argues that there were 408 active prostitution windows in 2007 (OIS, 2007; VRP1012, 2016). Taken together, there exist many different records and perspectives on the amount of prostitution windows and how this has changed over time. For the sake of clarity, in this study I use a dataset that was available at the University of Amsterdam, its graphical representation is shown in Figure 6.

Figure 6: Graphical representation of closing down the Amsterdam RLD

Source: University of Amsterdam. Computed in R

In general, research concludes that the closing down of the RLD, has neither been able to decrease negative social impacts nor been able to tackle organized crime (e.g. Van Wijk et al., 2010; Rekenkamer, 2011; Aalbers and Sabat, 2012). On the contrary, it has been shown that criminogenic activity has in fact increased in the district (Rekenkamer, 2011).

In essence, most researchers have looked at the social impacts and reflected on regulation policies in general, but have refrained to look at the economic impacts. In other words, the economic effect is often overlooked. However, it may be the case that closing down the RLD attracts more legitimate businesses and creates positive economic effects that can offset negative social effects. To gain a deeper understanding in this phenomenon this study addresses whether the economic condition of the RLD changes as prostitution windows are closed and the RLD shrinks in size.

(20)

2 Data

In order to gain a deeper insight into the Amsterdam RLD it is necessary to gather data and conduct empirical research. In this section, I firstly discuss what data is needed to conduct the study, how this data is collected and what modifications to the dataset were needed. Secondly, I give an overview of the sample and the variables that are used in this study. Finally, I end this section with the descriptive statistics of the data that is used.

2.1 Data collection

The ability to make valid claims about the results of this study is grounded in the initial data that is used in the empirical analysis. Hence, to give a valid insight in the economic developments of the RLD it is necessary that the dataset matches three conditions. Firstly, the dataset has to provide data on interesting business variables (such as amount of employees, turnover and location) on all sorts of businesses, not predominantly cover small or large businesses. Secondly, for the sake of comparison the dataset needs to include not only the RLD but also other parts of Amsterdam. Thirdly, to be able to look at effects over time, it is necessary to have data on multiple years.

To retrieve this data I reached out to the municipality and several other private agencies that focus on collecting and constructing (large) datasets. It appears that many agencies track business activity in some way or another. The municipality of Amsterdam appeared to have to most comprehensive dataset at hand that comprised multiple business variables and multiple years. The dataset provides information on all main business characteristics except business turnover. In turn, I reached out to other organizations, however, none of them were able to share information on business turnover. There are two main reasons for this:

(1) All Dutch organizations have to provide their results on sales to the Belastingdienst (Dutch tax authority). Due to privacy reasons one cannot retrieve data from the Belastingdienst, all information that the public is aloud to see is made available through the municipality. Unfortunately, the municipality does not retrieve business turnover from the Belastingdienst. (2) Some private agencies do keep track of business turnover but use either Kamer van Koophandel

(Dutch chamber of commerce) results or their own research. Only incorporated organizations have to provide their results to the Kamer van Koophandel and when agencies do their own research they usually focuses on larger companies. In turn, these datasets have a bias towards larger organizations. Moreover, they usually charge high prices in order to acquire this data. In sum to the above, it appeared not feasible to get data on business turnover for the city of Amsterdam, and I decided to focus this study on business and employee concentration instead. The retrieved dataset

(21)

of the municipality of Amsterdam needed several adjustments. Firstly, the dataset provided the X and Y coordinates of most data points. These coordinates were given in RD notation, however when one wants to plot these data points on a map, a different notation, WGS 84 is needed. With the use of ‘JaVaWa RTWtool’, I converted the RD coordinates to WGS 84 coordinates. Secondly, of some data points there were not coordinates given. To solve this problem I use the ‘Excel Geocoding Tool’. This program converts an address to a latitude and longitude value. The result of both steps is a dataset in which every data point is represented by coordinates in WGS 84 notation, hence a latitude and longitude value. Thirdly, the municipality is unable to provide a list that shows the total amount of employees that work at every business. However, the municipality is able to provide the amount of part time and full time employees. Therefore, I added an extra column that adds part time and full time employees, which in turn represents the total amount of employees that work in a specific business.

To give an insight in how the closing down of prostitution windows affects businesses, it is also necessary to have data on the location of prostitution windows and where and when they were closed. Due to prior studies on the RLD by the University of Amsterdam, the university was able to make this dataset available to me. It was necessary to make one manipulation to this data set. I added a column to every data point that shows in which year prostitution ended at that specific location.

2.2 Variables

For the business dataset that was retrieved from the municipality the sample comprised information on the entire city of Amsterdam for the years 2000 until 2015. The variables that are present in this sample are shown in Table 1. The prostitution windows database comprised information on where prostitution windows were closed in the entire city of Amsterdam between 2000 and 2016. The variables in this sample are shown in Table 1.

Besides these independent variables, shape files are used. These are geographical shapes that are used in the data analysis to mark the eastern and western RLDs and the main island in the city centre of Amsterdam. These variables are also shown in Table 1. As the exact location of a data point (business) is known as well as the location of the RLD and the city centre, it is possible to calculate the distance between a data point and the border to the Eastern RLD (RLD1), the Western RLD (RLD2), and the border of the main island in the city centre (NAT). The distance is calculated for both businesses and employees. To calculate the distance for employees a dataset is used that weights every data point in the business database by its employee count. Therefore, as the only difference between the distance for employees and business is the frequency of a data points, the distance variables for both are calculated in a similar manner. However, their respective means and standard deviations differ, therefore for the sake of clarity, they are separately listed. Moreover, based on the distance it is possible to create a

(22)

dummy variable that shows whether a data point resides inside RLD1, RLD2 or NAT. In sum, all dependent variables are summarized in Table 1.

Table 1: Variable list of dataset Variable Description

Business variables

Date started business Date start of business in d/m/y

Date terminated business Date end of business d/m/y

SBI Code that shows what sector the business is in

Description business Description of type of business

TOTF Amount of full time employees

TOTP Amount of part time employees

TOTE Amount of total employees

LAT Latitude coordinate of business address

LONG Longitude coordinate of business address

Windows variables

LAT Latitude of window

LONG Longitude of window

CLOSED COUNT TOTAL

Year of closing Amount of windows

Total amount of open windows

Shape files

RLD_Line Line to represent Red light District (De Wallen)

RLD2_Line Line to represent Second Red Light District (Singelgebied)

NAT_Line Line to represent Natural borders

Dependent variables

RLD Dummy variable, inside / outside RLD

NAT Dummy variable, inside / outside RLD natural border

RLD2 BusinessDRLD BusinessDRLD2 BusinessDNAT

Dummy variable, inside / outside RLD2

Distance of Business to RLD (towards RLD_line) Distance of Business to RLD2 (towards RLD2_line) Distance of Business to NAT (towards NAT_line)

EmployeeDRLD Distance of Employee to RLD (towards RLD_line)

EmployeeDRLD2 Distance of Employee to RLD2 (towards RLD2_line)

EmployeeDNAT Distance of Employee to NAT (towards NAT_line)

2.3 Descriptive statistics

For the sake of feasibility it has been determined to analyse three years of the provided datasets. To give a clear insight into the development over time is has been chosen to look at the years 2000, 2007 and 2014. The descriptive statistics of these datasets and their used variables are shown in Table 2. Moreover, to give a clear insight in how the regions are defined Figure 7 provides a map that shows the location of the Eastern RLD (RLD1), the Western RLD (RLD2), the main island in the city centre (NAT) and the outside of the main island (outside NAT).

(23)

Table 2: Descriptive statistics

Figure 7: Definition of regions

2000 2007 2014

Mean Std. Dev Mean Std. Dev Mean Std. Dev

TOTF 8.0714 0.4512 7.0239 0.3082 3.8188 0.1246 TOTP 0.8896 0.0541 0.9384 0.0584 0.6375 0.0168 TOTE 8.9610 0.4747 7.9623 0.3314 4.4562 0.1340 RLD 0.0114 0.0008 0.0102 0.0005 0.0095 0.0003 RLD2 0.0029 0.0004 0.0025 0.0003 0.0023 0.0001 NAT 0.0530 0.0016 0.0461 0.0011 0.0369 0.0005 BusinessDRLD 3.0488 0.0161 3.1427 0.0114 3.1603 0.0064 BusinessDRLD2 3.2178 0.0164 3.3113 0.0117 3.3223 0.0066 BusinessDNAT 2.6121 0.0162 2.7132 0.0115 2.7425 0.0064 EmployeeDRLD 3.7111 0.0063 3.7727 0.0046 3.7735 0.0035 EmployeeDRLD2 3.8974 0.0064 3.9694 0.0047 3.9655 0.0036 EmployeeDNAT 3.2787 0.0063 3.3521 0.0046 3.3538 0.0035 Legend

1 – Red Light District (RLD)

- De Wallen (Eastern RLD) – RLD1

- Singelgebied (Western RLD) – RLD2

2 – Natural border main island (NAT)

3 – Outside the main island (outside NAT)

(24)

3 Empirical Method

The available data set opens up interesting avenues for data analysis. Of considerable interest to this study is the density of businesses and employees in the city centre of Amsterdam. By calculating the density it can be shown where hotspots are, whether there is a difference between the inside and outside of certain areas and how the density of businesses and employees changes across the RLD1, RDL2 or NAT border. Moreover, as there is data available for multiple years it is also able to see how this develops over time.

In the remaining part of this paragraph the methods that are used in this study are discussed. Firstly, I describe the basics of kernel density estimation that I use throughout by analysis and is necessary to give a good estimation of the density of businesses and employees. Secondly, I will present a two dimensional (2D) version of the kernel density estimation that can be used to plot density on 2D maps. Finally, I will show that with the use of the McCrary test it is possible to get insight in how the density of businesses and employees develops across a pre-set border.

3.1 Kernel density estimation

There are several methods available to compute the density. In essence, one can choose between a parametric and non-parametric approach, which approach is applicable depends on the nature of the data (Bagdonavicius et al., 2011). In the parametric procedure certain assumptions are made about the probability distribution of the variables that are assessed. In contrast, the non-parametric procedure has no fixed structure and depends entirely on all the data points to reach an estimate (Murphy, 2012). As in this study I do not assume a fixed distribution, it is chosen to address the data in a non-parametric way.

The first step is to determine a method with which to calculate the density. The most well-known and simple non-parametric method is using histograms. However, the use of histograms has generally been criticized, as they are not smooth, depend on the end points and width of the bins (Hwang et al., 1994; UESA, 2016). In essence, using different bins gives a different visualization of the same histogram and leads to different interpretations of the data (see first two plots in Figure 8).

(25)

(2) (1)

Figure 8: Different visualisations of same dataset

Source: Scikit, 2016

Adopting a kernel density estimator can alleviate this problem. Here, the dependency on the endpoints of the bins is removed by centring kernel estimator in a kernel function at each data point. Kernel estimators smooth out the contribution of each data point over the local neighbourhood of that data point. In turn, this effect depends on the type of kernel function and the bandwidth used. In essence, by using kernel density estimation it is possible to observe the estimated density at any point, see the second two plots in Figure 8 (Hwang et al., 1994; Scikit, 2016; UESA, 2016). Estimating the density by using a kernel estimator is executed with Equation 1.

! " =

1 %

&

' ()*

" − "

-ℎ

In this function, K resembles the kernel function and h its bandwidth. In this study, the Gaussian kernel will be used, as it is one of the smoother kernels (UESA, 2016). The Gaussian kernel is described in Equation 2.

1

20 1"2 −

1

2

3

4

In sum, kernel density estimation makes it possible to gain a more comprehensive insight in how density develops, by smoothing out the contribution of each data point.

(26)

(3)

3.2 Multivariate (2D) kernel density estimation

From the first paragraph it has become clear that with the use of kernel density estimation it is possible to create a smooth and better representation of the data. However, with the discussed version of kernel density estimation it is only possible to look at density based on one dimension (e.g. distance to the RLD) but not on multiple dimensions (e.g. latitude and longitude). As this method only presents a one-dimensional figure that is not particularly insightful. It would be more insightful to plot the density on a map, so it is directly visible where the concentration is the highest and how this changes from place to place. In order to do this it is necessary to adopt multivariate (2D) kernel density estimation.

Again, as in the case of 1 dimension, multivariate kernel density has an advantage over the use of histograms. The most striking difference here is that by making data inferences the contribution of each point is smoothed out into the space directly surrounding it. In this way, the smoothing of all points contributes to a better interpretable picture, see Figure 9 (Silverman, 1986).

Figure 9: Difference between 2D histogram (left) and 2D kernel density estimation (right)

Source: MKDE, 2016

In order to reach such a representation the initial kernel density estimator function needs to be slightly adjusted in order to account for multiple dimensions (Härdle et al., 2004). This leads to Equation 3, here K is a multivariate kernel function with d arguments.

!

₅

" =

1 %

1 ℎ

6

&

" – "

-ℎ

' ()*

(27)

The main contrast with 1D kernel density estimation is that the bandwidth matrix (h) also influences how the data is smoothened. As in this study, the density estimation will be plotted on a map all data points should be treated equally. Therefore, the S class kernel will be used, as this exerts the same amount of smoothing on every data point (Wand and Jones, 1993; Duong and Hazelton, 2003).

Although smoothing presents a nice picture it is not taking into account any borders. Without forcing discontinuity the 2D plot will look like the left plot in Figure 10. However, as the goal of this study is to compare the RLD with its direct surroundings this is not particularly insightful. In order to force this discontinuity separate density plots need to be overlaid. Here, it is crucial that every plot uses the same grid and bandwidth to avoid issues with different scales. When this is accounted for the 2D plot will look like the right plot in Figure 10.

Figure 10: Density plot on map without discontinuity (left), with discontinuity (right)

Computed in R

3.3 McCrary Test

As highlighted in Figure 10, it can be of particular interest to look at whether the density significantly changes around the RLD border. However, only so much can be implied from a plot. In order to indicate there is a real effect it is necessary to also conduct a study that underscores this with numerical evidence. An interesting avenue in research methodology that looks into this is regression discontinuity design (RDD), a research method that has become increasingly popular. The bottom-line of RDD is that one can compare observations that lay closely to either side of a pre-set cut-off point and estimate the average treatment effect (Imbens and Lemieux, 2008).

(28)

(4) McCrary (2008) has proposed a test that can be used to identify discontinuity of a running variable at a pre-set cut-off point. Here, the McCrary test takes into account that the treatment assignment is considered public knowledge. Thus, in other words, a subject is able to self-select to either side of the border of the running variable. Assuming that this is the case, one can expect that around the cut-off point the running variable will be discontinuous. If this is not the case, the treatment can be considered of no (or low) influence on the subject. The McCrary test is based on an estimator for the discontinuity at the cut-off point in the density function of the running variable. In essence, the McCrary test is a Wald test of the null hypothesis that the discontinuity equals zero. The result of the McCrary test is the log difference in heights (theta; 8) at the cut-off point, its standard deviation and a p value indicating at what significance level null hypothesis of no sorting can be rejected, see Equation 4 (McCrary, 2008). Figure 11 shows the example graphical output for a case of discontinuity and continuity.

9

_:

: 8 = 0

9

_*

: 8 ≠ 0

In this study the running variable will be the distance between a data point and the border of the RLD or NAT. The McCrary test is applicable here as a subject is assumed to be able to influence its own treatment assignment, thus whether he locates just inside or just outside the RLD. To this extent, if being inside or outside the RLD is of importance to the business owner one would expect that a discontinuity would be present at the border of the RLD. In essence, a discontinuity shows that one area is preferred over the other, and therefore can be seen as more attractive.

Figure 11: Example of McCrary Tests with discontinuity (left), with no discontinuity (right)

(29)

To gain a more comprehensive insight in whether the RLD (RLD1 + RLD2) attracts or repels business and employees and what the underlying cause of this could be, several McCrary tests should be performed. The use of multiple tests is that it enables to control for where the borders are set and to determine whether it is of influence that the RLD is inside the centre of Amsterdam. There are three distinct McCrary tests performed, for the sake of clarity their definitions are here shortly presented. Moreover, Figure 12 gives a visual representation of how these manipulations of how the different tests look like.

(1) The first McCrary examines how the density develops when the border of a RLD is crossed to either the inside or the outside the main island.

(2) The second McCrary examines how the density develops when the border of a RLD is crossed to the outside of the main island.

(3) The third McCrary test examines how the density develops when the border of the main island is crossed to the outside of the main island (excluding both RLDs).

Figure 12: Overview of executed McCrary tests (red vs yellow)

Legend (1) – Red Light District (RLD)

(2) – Natural border main island (NAT) (3) – Outside natural border (outside NAT)

Manipulations (red vs yellow) 1 RLD vs NAT + outside NAT [1 vs 2+3] 2 RLD vs OUTSIDE NAT [1 vs 3] 3 NAT vs OUTSIDE NAT [2 vs 3]

McCrary Test 1

McCrary Test 3

(30)

There are no widely accepted and clear-cut borders of where the RLDs start and end. Therefore, the first McCrary test that looks at the inside of the RLD to the outside of the RLD gives a biased representation. To overcome this problem, the second McCrary test is executed that only looks at the most undisputable border, the canals surrounding the RLD. The canals provide a better counterfactual as it functions as a natural border to the RLD. However, the estimated RLD effect of the second McCrary test is also influenced by the fact that it is located on the main island in the city centre. To control for the possibility that the observed effect is actually caused by the location of the RLD inside the city centre it is necessary to conduct a ‘placebo test’ to be sure the RLD effect actually exists and is not a ‘placebo effect’. To isolate the RLD effect from its location on the main island (NAT effect) the third McCrary test is needed to calculate the NAT effect. In turn, the least biased (isolated) RLD effect can be calculated by subtracting the result of the second McCrary test by the result of the third McCrary test.

(31)

4 Results

This section elaborates on the results of the empirical analyses and is separated into two parts. In the first part discusses the RLD effect on business. The second part gives an insight in the RLD effect on employees. Both parts follow the same structure. Firstly, I analyse the density maps of the centre of Amsterdam. Secondly, I take a closer look at the McCrary tests and make an effort to quantify the RLD effect.

4.1 RLD effect on Businesses

This subsection addresses to what extent the RLD influences the density of businesses inside the RLD. Firstly, in this subsection I interpret two-dimensional kernel density plots of the RLD and its outside area. Afterwards with the use of McCrary tests I quantify the effect of the RLD on business density.

4.1.1 Density Maps

With the use of two-dimensional kernel density estimation it is possible to derive the business density for three separate regions in Amsterdam (the red light districts, the natural border of the main island, and the rest of the city) and plot these densities on the same map. To identify whether this changes over time this is done fore three separate years (2000-2007-2014). The resulting plots form the basis of the analysis and are shown in Figure 13. In this section, with the use of Figure 13, I discuss some preliminary insights on the development of the RLD. In Figure 13, the top row shows the density plots of the entire city centre of Amsterdam. The bottom row in Figure 13 shows a zoomed in version that gives a more holistic view on the density spread inside the RLDs.

(32)

Figure 13: Density maps of businesses in Amsterdam. Top row shows plots off city entre, bottom row shows close up of RLDs

2000 2007 2014

(33)

Cross-sectional implications

In the plots of the year 2000, it is observable that on the main island (NAT) there are three centres where the density concentration is high. Here, the inside of De Wallen (RLD1) has the highest peak in business density. Moreover, this peak in business density seems to coincide with the concentration of prostitution windows. Especially between the inside of the RLD1 and the outside of NAT, there seems to be a strong contrast present, with a higher density inside than outside the RLD1. In contrast, when compared to the rest of NAT, the Singelgebied (RLD2) seems to have one of the lowest business densities. Here, it is even observable that the border of the RLD2 causes the discontinuity in the density peak of NAT. This gives some initial grounds that businesses rather locate outside the RLD2 area than inside the RLD2 area. Finally, the plot indicates that the business density develops fairly continuously when the border of NAT is crossed. In essence, the plots of 2000 imply that, the RLD1 is attractive to businesses, while the RLD2 does not seem to attract business activity.

The plots of the year 2007, resemble a similar picture to the plots of 2000. It is still observable that there are three main density centres inside NAT, of which the highest density can be observed in the RLD1. Moreover, the peak in business density is closely located to the place where most prostitution windows were closed. The RLD1 still seems to be an attractive region for businesses to settle in. The border of the RLD2 with NAT is still discontinuous. However, a centre has emerged inside the RLD2 that is located precisely in between two areas that contain prostitution windows. Moreover, it seems that the business density around the closed windows is slightly higher than its surroundings. In essence, the plots of 2007 imply that RLD1 is attractive for businesses and seems to have the highest density where windows have been closed. Furthermore, RLD2 has gained attractiveness at the exact spot where there are no prostitution windows.

In the plots of the year 2014, there are less peaks in density present, resulting in a flatter and more homogenous picture. The density peaks still seem to be at the same locations but have a larger centre. It is still especially clear that the inside of the RLD1 has the highest density. Interestingly, the plot shows that most of the closed windows are located where business density is the highest. In turn, the plot gives the impression that especially around the closed prostitution windows the business density is high. The density in RLD2 seems to develop continuously across its borders, and there seems to be no discrepancy in the density around the open or closed windows. It still seems to be the case that the density develops continuously across the border of NAT. In essence, RLD1 maintains to be the he most attractive hotspot for businesses in the centre of Amsterdam. For RLD2 this is less so the case, though it appears to be more attractive than before.

Changes over time: Inside the RLD

The density spread inside the RLD1 has stayed more or less the same between 2000 and 2014. The only thing that seems to have changed slightly over the years is that there is less of a peak in the density of

(34)

2014. So, in 2014 compared to 2007 and to a greater extent compared to 2000, there is a larger centre. It does seems to be observable that over time the density has risen at places where prostitution windows were closed, it may be so that the closing down of windows in the RLD1 has contributed to a larger density centre. In this sense, as more businesses tend to locate in the centre of the RLD1, the closing down of windows in RLD1 would have actually contributed to a greater attractiveness of the RLD towards businesses.

When looking at the density spread inside RLD2, a clearer pattern shift becomes apparent. In the plot of 2000, it appears that there is no real centre present in the district. However, as time progresses there seems to emerge a small centre in between the windows of RLD2 in the year 2007. However, this centre disappears again in 2014. Finally, what also grabs the attention is that the density around two windows at the Western end of RLD2 seems to change as the two windows are closed. Here, it is noticeable that as the windows are closed between 2000 and 2007, the density decreases. In contrast to what has been implied for RLD1, here it seems to be the case that closing down prostitution windows contributed to a lower attractiveness for businesses to locate inside the RLD.

Changes over time: Difference between inside and outside RLD / NAT

Figure 13 also shows that it depends where the border of a RLD is crossed. For instance, there is a bigger discrepancy between the business density inside RLD1 compared to the outside of NAT than compared to the inside of NAT. Interestingly, this observation holds for all observed years. Moreover, generally, the pattern of business density, thus where the hotspots are located, seems to stay the same as time progresses. However, it does seem to be the case that over time the density inside the RLD1 has increased more than the density inside or outside NAT. This observation implies that over time, the RLD has become more attractive in comparison to the areas close to it.

For RLD2 the difference between the inside and outside seems to progress differently. Here, in 2000 and even more so in 2007 there seems to be a significant discontinuity when the border from RLD2 towards the inside of NAT is crossed. However, the density seems to be continuous in 2014. Therefore, RLD2 attracted relatively more businesses than NAT. This observation implies that RLD2 has become more attractive to businesses over time as well and is more continuous with NAT.

Finally, looking at the border between the inside and the outside of NAT, it becomes apparent that at most places the density develops continuously across the border. With the notable exception at the border of the RLD1, here there seems to be a significant discontinuity between inside and outside.

In sum, the observations show that for every observed year RLD1 is a real hotspot for businesses. Moreover, the plots show that the highest density coincides with the location of the prostitution windows in RLD1. In essence, over time both RLDs appear to have attracted more businesses than the