House price dynamics in and between Amsterdam and Utrecht : the ripple effect

(1)

i

Master Thesis

House price dynamics in and between Amsterdam and Utrecht:

the ripple effect

Author: Floris Mathlener

Supervisor: Johan Conijn Student number: 10251650

MSc Business Economics: Finance & Real Estate Finance Track July 2017

(2)

ii

Statement of Originality

This document is written by Student Floris Mathlener 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)

iii

Abstract

This thesis examines the ripple effect within the Netherlands. The ripple effect exists when price shocks in a central area ‘ripple’ over to adjacent areas. The scope of this thesis is on Amsterdam and its

surroundings, as well as the city Utrecht. The sample period is between 1996-2016, using quarterly data. In this thesis, Vector Autoregressive (VAR) model analysis is performed to test whether lagged real price shocks in one district or city have an effect on current values of another city or district. Granger Causality (1969) tests are performed to elaborate on the results of the VAR models. The results show that

Amsterdam yields significant within-city ripple effects. The city centre seems to have a direct effect on the outer suburbs as well. The ripple effect seems less present in Utrecht. There are other districts that show equally or more important in the price formation of the city. The models including Amsterdam and Utrecht show that after 2001, Amsterdam shows significant ripple effects with regard to Utrecht. Before this time, the effect seems to be in the opposite direction. Finally, Amsterdam shows significant influence on its surrounding municipalities.

(4)

iv

1. Introduction

1.1 Motives and relevance

The Dutch housing market has gone through some strong fluctuations over past decade. Housing prices in the Netherlands suffered a substantial blow as a result of the recent economic downturn. The aftermath of this economic downturn made residential transactions a risky operation for both households and banks. After the financial crisis of 2008, it became apparent that real estate markets in some areas recover faster from the economic downturn than others. This thesis builds on this situation by researching the within-city housing dynamics.

In this thesis we further examine what are the driving factors as to if and why some cities or

areas ‘follow the leader’ faster than others. The Dutch housing market is researched by checking whether there exists some form of the ‘ripple effect’ in housing prices. This is, between and within two big cities in the Netherlands: Amsterdam and Utrecht. We check whether housing price shocks in one city or

neighbourhood have an effect on subsequent house price shocks in other cities or neighbourhoods. If this is the case, there is an interdependency of price fluctuations between the cities or neighbourhoods.

The existence of shock transmittance from one area to other areas might have important

consequences. Municipalities and governments are in charge of proposing regulation that induces long term sustainable growth. It is of vital importance to consider flows of house price shock transmittance when forming new urban planning policy. It is inevitable that these municipalities incorporate insights from across their own border, as the housing market is cross-regional. Exposing cross-border dynamics can help municipalities form a better strategy on how to deal with certain shocks and volatilities.

Investors might benefit from gaining insights on dynamics in the Dutch housing market. An

investor attempting to hold a diversified portfolio must be aware of certain links and cointegration between cities and neighbourhoods. Foreign investment is gaining increasing market share in the Netherlands (CBRE, 2017). It is often challenging for a foreign investor to account for all interlinkages when entering a market. This thesis might be helpful for these types of investors attempting to gain understanding of Dutch market dynamics. Investors will be better able to diversify between

neighbourhoods and cities. Banks and other lenders could also significantly benefit by gaining more insights on Dutch market dynamics. Banks face the decision of weighing risk against return with every mortgage it issues. Gaining knowledge of how a certain area respond to shocks in other areas is a vital factor in weighing the risk of a potential issuance of a mortgage in that region.

The ripple effect is widely researched in literature. The effect however, is not extensively studied for the Netherlands. This is, especially at the city level. Teye et al. (2016) research the Amsterdam housing market. The focus is on house price dynamics between the sub-districts in the city. The

researchers find that there is some kind of connected network of districts which affect one another. This thesis builds on this by examining whether these connected networks can be further specified into a ripple effect.

The Dutch market is especially fit for research on the city level. The country is relatively small.

When considering larger countries it is far more unlikely that people live in one city and work in another. Most people are bound to a city by their job. In 2011, more than 55% of the Dutch workforce worked and lived in a different municipality. To put this into perspective: almost 15% of the inhabitants of Almere commute to Amsterdam on a daily basis (CBS, 2011). The interlinkages of housing markets in Dutch cities are therefore expected to be larger than in other countries.

(6)

2

1.2 Study questions and hypotheses

Previous studies show that regional housing market shocks may have an effect on adjacent regions (Pollakowski & Ray, 1997). In this thesis, the focus is on whether price shocks in one area have a transitory or permanent effect on house prices in other areas. To research this, three perspectives are taken. First, within-city shocks are examined. The purpose of this thesis is to research whether price shocks to the centre of a city ripple over to the more outer parts of the city. The extent to which price shocks ripple over is quantified. Overall, the expectation is that price shocks ripple over from the city centre to outer districts within one or two quarters. Second, we test whether price shocks ripple over from Amsterdam to Utrecht, or the other way around. Third, house price diffusion from Amsterdam to directly surrounding cities is examined. The municipalities of Haarlemmermeer, Diemen and Ouderkerk aan de Amstel are included. The hypotheses per district and municipality are presented in part 4.

The expectations are in line with findings regarding the United Kingdom by Meen (1999), Cook

(2003), and Holly et al. (2011). It is expected that for the Dutch market, Amsterdam shows similar importance as London in the United Kingdom. It would therefore have a leading role in the price dynamics within the Netherlands. Research by Teye and Ahelegbey (2016), and van Dijk et al. (2009) suggests a leading role in the house price dynamics for the province Noord-Holland. It is expected that these effects also show on the city level, regarding Amsterdam and Utrecht. Lagged price shocks in Amsterdam are expected to have a significant effect on current price shocks in the surrounding cities.

Often, research on house price dynamics is focused on regions or country-wide data. Specific

characteristics of cities or neighbourhoods are often neglected. In this thesis, more emphasis is put on these characteristics by focusing on city-level dynamics. The extent to which the ripple effect is expected to be observed is linked to three categorising traits of the district in question. First, the geographic location is considered. It is expected that adjacent regions show similar price dynamics (Meen, 1999). Second, the regions are defined by certain price levels. The argumentation is that districts with similar price levels show similar price dynamics. Therefore, price shocks will ripple out more. Third, certain demographic characteristics are compared. Examples include household income, unemployment, and education level. Again, similar demographic characteristics are expected to have a direct or indirect effect on the way the districts show lag-dependency.

1.3 Methods and data

The NVM, which is a Dutch association for real estate brokers, provides an extensive database containing residential transactions between 1996 and 2016. This transaction data is the basis for constructing time indices per quarter per district. These time indices are the input for further analysis. The issue of ripple effects is formulated in terms of a vector autoregressive (VAR) model and Granger Causality testing (1969). The choice of this method is in line with research by Teye et al. (2016), Liao et al. (2015) and Chiang & Tsai (2016) who also implement VAR model analysis to research the ripple effect.

This thesis focuses on four main issues regarding the ripple effect. First, whether there actually

exists a ripple effect in house prices within two of the largest cities in the Netherlands. Some surrounding areas of Amsterdam are included to check whether this ripple effect is also present outside the city borders. The second part of this thesis focuses on answering the question whether there exists a some form of the ripple effect between the cities Amsterdam and Utrecht. The previous questions are examined using VAR analysis and Granger Causality testing.

(7)

3

1.4 Structure

The remainder of this thesis is structured as follows. Section 2 discusses the existent literature on the topic. The focus is on three topics: the economic background of the ripple effect, the existing research on the ripple effect worldwide and research on the Dutch market. Section 3 presents insights on the Dutch housing market. Section 4 presents the respective city markets, including the characteristics for the specific neighbourhood that are examined. The specific hypotheses that arise are also discussed. Section 5 discusses the used data, as well as the adjustments that are made to prepare the data. Section 6 discusses the used research methods. Section 7 contains the results of the analysis. Section 8 contains several robustness checks regarding the research. Section 9 contains the final conclusions, discussion and recommendations.

(8)

4

Literature review

2

The literature regarding house price dynamics and the ripple effect is extensive. Over the years, several methods have been introduced to research the topic. Geographically, most research concentrates on the United Kingdom. The London area is often examined as a leader region causing ripples on other regions. The Dutch market has not been largely covered in this type of research. The following literature review first discusses relevant literature regarding the economic background of the ripple effect. Second, findings per country or region regarding the ripple effect are discussed. Third, ripple effect research on the Dutch market is presented.

2.1 Economic background of the ripple effect

One of the first papers on house price dynamics is by Holmans (1990). Holmans raises several points of discussion that lead to further research on the ripple effect in the United Kingdom. Holmans observes a distinct spatial pattern in cycles of house prices onwards from 1960. Holmans observes that prices in the south-east and the north of England diverged in the 1980’s, and converged in the 1990’s. The south-east regions appear to lead the cycle. The magnitude of the movements in the south are greater than all others (1990).

Various studies have followed Holmans’ (1990) observations regarding house price dynamics.

One of the first papers written in the field of ripple effects and the implications for house price dynamics is by Drake (1995). Drake mentions that for a ripple effect to exist, one would observe the house price gap consequently widening in the upswings and narrowing back in the downswings. If this is the case this would become evident in the convergence tests (1995). Drake considers real estate in the United

Kingdom and finds no real evidence of certain equilibrating forces of convergence. Drake does however, observe casual empirical evidence for this ripple effect. It seems that house prices in the South East of England have a tendency to rise faster and earlier in economic upswings. In economic downturns this effect is observed in the opposite way (1995).

Meen (1999) sheds more light on the drivers behind this ripple effect. In his research on the

United Kingdom’s real estate market, Meen states that the ripple effect is the extent to which certain coefficients show certain distinct spatial patterns. This is, rather than purely random variation (1999). A model which includes coefficients that exhibit non-random spatial patterns is examined. Meen (1999) concludes that the ripple effect is deemed to be caused by changes within certain cities or regions, rather than changes between cities or regions. If the ripple effect would be in place, this would imply that transaction prices of houses between two areas would diverge in the short-term. Transaction prices would then converge in the long-term.

Meen (1999) constructs four possible explanations for the existence of the ripple effect. The first

explanation regards migrations. When house prices increase in one area due to increased demand, people will migrate to adjacent regions where housing is cheaper. This migration consequently increases prices in these areas as well. The second explanation concerns transfers of ownership. Repeated sales drive prices up, where previous sale prices get incorporated in current prices. When people move from one region to another, this drives prices up. The third explanation concerns spatial arbitrage. If markets are completely efficient, price differences for comparable dwellings are diminished due to arbitrage opportunities. Meen (1999) finds some evidence of a ‘positive feedback effect’. Information is spilled over into other regions, thereby affecting price dynamics. The last explanation concerns certain spatial patterns in house price growth. The basis is that price dynamics show a certain pattern due to similar patterns in the underlying regressors. If this is the case, house prices are expected to keep on behaving in the same way. The problem with this vision is that it is not clear why these shocks would have to keep emerging in the south.

(9)

5

Previous arguments by Meen (1999) are expanded by Hort (2000). Hort argues that the turnover

rate also has a significant effect on house price dynamics (2000). The idea is that sellers base the list price of their dwelling on the expectation they have of the market. These expectations are largely affected by the ‘ease of selling’. In upturns, the ease of selling usually increases fast in the most liquid areas. This argument might also be a factor in why the ripple effect would be more evident in Amsterdam than in Utrecht. The Amsterdam market is significantly more liquid, showing around 1.5 times more transactions over the period 1996-2016 (NVM, 2017). Also, it seems the Amsterdam housing market is more volatile. Large fluctuations are thus expected to have a faster and stronger effect on other districts.

Millington sheds more light on the dynamics of labour markets and housing markets (1994).

Millington claims that wages have an effect on housing prices. An unexpected positive shock on income will thus lead to incoming migrations, and have an upwards effect on local house prices. Theoretically, the inflow of labour and people will stop when the increased house prices offset the real increase in wages (1994). It’s unsure if these dynamics also hold when considering city-wide dynamics.

Miao et al. (2011) argue that it is likely that the ripple effect exists because of similar economic

bases. This is, as opposed to population migration and arbitrage in cross-regional properties. Certain wealth effects may have an effect on the amount of feedback between economic growth and housing prices. The writers argue that this could explain why the relationships between markets intensify during upturns in the residential market. Information spill-overs could also be an explanation. Home owners who observe a price increase in adjacent or similar regions will anticipate price increases in their own regions. This is in line with findings by Meen (1999). The second argument on information spill-overs is expected to have a larger effect on within-city ripple effects. This is, as economic bases within

neighbourhoods are less evident.

Capozza et al. (2004) identify several factors that drive real estate prices in metropolitan areas.

These factors include income, population and construction costs. The writers relate higher construction costs with lower levels of serial correlation and mean reversion. An increase in construction costs means that supply is less able to respond to shocks in demand. The last determinant Capozza et al. (2004) describe is the actual geographic distance.

Often, research on the ripple effect like Meen (1999), Miao et al. (2011) and Holly et al. (2011)

regards countries like the United Kingdom and the United States). In these countries absolute distances between cities are significantly higher than in the Netherlands. People working in Amsterdam might very well be willing to live in cities like Almere, Utrecht or Rotterdam. Shocks in housing prices in Amsterdam might lead to high migrations to not only surrounding villages, but also surrounding cities.

Buyers in the real estate market observe signals from the leader in the house price dynamics. On

the one side, they will value their own house higher, if they see that house prices in another city increase. Subsequently, people looking to buy will be driven out of the most popular regions, into the regions that they can actually afford. They will be looking for ‘the next best thing’ (Capozza et al. 2004). When assuming Amsterdam as the leader in the house price dynamics, the next best thing might, in the Netherlands, for instance be Utrecht or Rotterdam.

Yu (2015) focuses on city-level spill-over effects in real estate markets. Yu (2015) performs

analyses regarding 35 major cities in China. This paper is interesting as it regards a large group of

interacted metropolitan areas. The paper considers the dynamic properties of the model by studying the impact of three specific shocks on the market. The first shocks considered are income shocks. Yu finds that income shocks have limited effects on housing prices in other cities. This opposes findings by Capozza et al. (2004) and Millington (1994). Second, Yu finds that the interregional impact of housing price shocks is larger than the impact of income shocks within the city. This effect varies largely per city, but first-tier cities (Shanghai, Beijing) overall show the highest spill-over effects in house price shocks.

(10)

6

For the purpose of this thesis on the Dutch market, this could mean that Amsterdam would have a larger impact on Utrecht than on the surrounding smaller cities.

The last shocks Yu (2015) considers are interest rate shocks. All cities considered show a negative

response to adjustments in the interest rate. This makes sense as higher interest means higher mortgage service payments and thus lower prices. Yu finds that first-tier and large cities are affected more by interest rate shocks than second-tier cities (2015).

Holly et al. (2011) set up a Vector Error Correction Model (VECM) for the United Kingdom’s

market. Holly et al. (2011) elaborate on one of the main diffusing factors; commuting. The writers claim that relative prices between two regions are dependent on the price of commuting between the regions. This price of commuting can be measured by the mobility barriers in place. It reflects the infrastructure situation and the amount of facilities that are present. People that live in rural areas are more willing to travel to work as the relative supply of jobs is low. Holly et al. (2011) provide numbers on net outflow and net inflow of people travelling to work areas.

Interesting insights regarding city-specific determinants on housing price dynamics are provided

by Holmes et al. (2015). The paper on the Paris’ city housing market examines the extent to which districts converge over the long run. The researchers implement a probit model on the outcomes of unit root testing on all sub-district pairs. Holmes et al. (2015) conclude that the probability of stationarity in the district-specific differentials is positively affected by similar levels of unemployment, characteristics, and demographics.

2.2 The ripple effect worldwide

The ripple effect is researched in countless countries, regions and specifications. The most popular areas of focus are the United States and the United Kingdom. The methods can be categorised into two research currents: unit root modelling and VAR modelling. Establishing a ripple effect through unit root models builds on the fact that a ripple effect can be observed when two time series exhibit stationarity in their differenced levels. This stationarity is then explained as a form of mean-reversion when shocks occur, thus explaining the ripple effect. The other main method involves setting up some form of the VAR model. A VAR model is usually implemented to explain current values of a time series by past values of itself and past values of certain other variables. This method is more precise in testing for actual fortitude of impact of one area on another. Also, VAR modelling allows for multiple areas to be included in the model.

Chiang & Tsai (2016) implement this VAR model by researching the United States’ regional

housing markets. The researchers study certain relationships between cross-regional and regional housing markets. Granger causality tests are run to examine whether excess housing returns in one region can be explained by prior values of excess housing return of other regions. Los Angeles, New York and Miami, in their respective regions, are found to be the source of shocks in house prices.

Liao et al. (2015) implement a similar VAR model as Chiang & Tsai (2016) in their research on the

Singapore housing market. The researchers regard the effect of foreign liquidity on regional housing prices. The researchers find that foreign investments have big impacts on house price dynamics. Liao et al. add several endogenous and exogenous variables to their VAR model. Examples include private housing supply, land-sale intensity. The paper concludes that increased levels of foreign investment explain a substantial part of volatility in regional house prices in Singapore (2015).

Balcilar et al. (2013) provide further insights in ripple effect theory. In their paper on the South

African housing market, the researchers present the ripple effect as a unit root issue. If a ripple effect is in place, one will expect the ratios to exhibit some stationarity or mean reversion after a diffusion in

regional house prices. This method does not account for actual inter-urban relations, as the stationarity is related to the underlying trend level.

(11)

7

The researchers conclude that Cape Town and Durban are drivers in the South African real estate market. Similar conclusions could be expected for Amsterdam in the Netherlands.

Shi et al. (2009) pose the issue of ripple effects in a vector error correction model (VECM). The

model is applied to regional house prices in New Zealand. The researchers claim that the ripple effect in New Zealand is probably confined to regional boundaries, and will only spread nationally from one regional hub to another. This opposes findings for the United Kingdom (Meen, 1999), where house price shocks in the south-east ripple over to the rest of the country. The results are in agreement with research by Yu (2015), and Teye and Ahelegbey (2016) who find that there can be multiple starting points for a shock in house prices to ripple over to other regions. For the purpose of this thesis, the same might hold with regard to within city price shocks. Neighbourhoods like Oud-Zuid in Amsterdam are expected to show similar signalling effects towards the outer districts as the city centre. This would mean the ripple effect could also emerge from other neighbourhoods than the city centre.

Holly et al. (2011) conclude that London is the dominant region in the United Kingdom. This is in

line with prior research by Meen (1999) and Cook & Watson (2014). The writers find that regional house prices respond quite immediately to shocks in London house prices. This reaction is amplified by both interactions between the regions and certain internal dynamics. Shocks to London housing prices fade away after a certain amount of periods. On the contrary, shocks that spill over from London to other regions take much longer to evaporate. This prolongation of the shock effect is strengthened with geographical distance.

Research on the ripple effect is mostly performed on a regional or intra-city basis. There are

some researchers who concern the issue of within-city house price diffusion. First, Teng et al. (2017) focus on the Taipei metropolitan area in Taiwan. The researchers regard the house price diffusion with regard to fundamentals and bubble prices. By means of Granger Causality (1969) the writers find that housing bubbles force a certain extent of house price diffusion. Bubble prices are deemed to start in the central areas, rippling over to more suburban regions. The researchers claim that the house price bubble is larger in the suburban areas. This is, as prices in central areas are based more strongly on

fundamentals.

Prior literature provides with an interesting contradiction. As said, Teng et al. (2017) find that for Taipei, Taiwan, the city centre shows certain price diffusion patterns from the city centre towards the suburban areas. Oikarinen (2005) researches the ripple effect for the Finnish market, regarding Helsinki. The results show that suburban areas seem to diffuse price shocks towards the city centre. Oikarinen (2005) finds that diffusor areas in metropolitan areas are characterised by migrations and employment. The Helsinki metropolitan area covers a large surface. This might explain why non-central districts have such a strong effect in the price dynamics. This is, over solely a central location. This thesis on the Dutch market builds on this contradiction by examining the Dutch market.

Jones and Leishman (2006) examine the ripple effect on a somewhat broader basis. The writers examine whether Glasgow shows a significant effect on its neighbouring ‘local housing market areas’ (LHMA’s). The ripple effect is researched outside of the municipal boundaries. The writers find that Glasgow shows significant ripple effects with regard to its neighbouring areas. The extent to which this effect is present is dependent on the amount of household migrations. Similar results are expected to be found for the purpose of this thesis with regard to Amsterdam. Migrations from Amsterdam to areas like Ouderkerk aan de Amstel, Nieuw-Vennep and Hoofddorp are comparably high (CBS, 2017).

2.3 Ripple effects in the Dutch market

As said, the Dutch market is relatively underexposed with regard to the ripple effect. Research that considers the ripple effect in the Netherlands mostly regards regional house price dynamics. Teye and Ahelegbey (2016) consider the Dutch housing market by looking at provinces.

(12)

8

The focus is on whether house prices show temporal spatial dependence. Teye and Ahelegbey (2016) identify the provinces Drenthe, Noord-Holland and Limburg as the main diffusors of volatility in temporal house prices. Interesting insights are presented regarding the existence of a ‘hub’ which is the leader in price diffusion. The writers find that there is also a possibility that the dispersal of shocks in housing prices starts from a smaller region. Regional house prices are found to be temporally dependent in the Netherlands. This temporal interdependence however, seems to be varying throughout the study period. The researchers find that the interdependency was high between 1995 and 2003. After that, the

interdependency weakened until 2008, and increased after 2008 until 2016.

The Dutch market is researched on a regional level by Van Dijk et al. (2011) who design a panel time series model with ‘latent-classes’. This means the data is divided into two classes which are characterized by a pattern of conditional probabilities. Van Dijk et al. (2011) find that the Dutch market can be characterized by two classes. The first class yields mostly rural areas, close to larger cities. The second class contains the bigger cities, and some remote rural areas. The regions in the first class show high house price growth, and quite fast reactions to shifts in gross domestic product. Van Dijk et al. (2011) claim that a possible explanation for this is that regions in class one are in demand by commuters, working in large cities. This theory might also hold on the city level. Districts on the outskirts of town might benefit from welfare and house price increases in the city centre and the main business areas. This is, as these central areas provide with the most jobs. A price increase in the city centre can thus increase demand by commuters in the outskirts of town.

Zooming in, further analysis regarding the Amsterdam housing market is performed by Teye et

al. (2016). The researchers examine the interdependencies of housing prices in the sub-districts of Amsterdam. The city-wide impact is researched by implementing cointegration techniques and Granger causality. Teye et al. (2016) research the city-wide development as well as the impact of developments in house prices of Amsterdam on the national trend. This thesis builds on this by analysing city-specific effects and determinants of this ripple effect. Teye et al. (2016) find that house prices in the city centre grow stronger, and show more volatility than peripheral areas. The gap between housing markets in the more expensive and the cheaper districts became bigger between 1995 and 2014. Furthermore, the researchers prove that there is a long-term relationship between the Amsterdam housing market and the Dutch housing market as a whole. In this thesis, more attention is paid to actual house price shock ripples within Amsterdam and Utrecht.

(13)

9

The Dutch housing market

3 3.1 Urbanization and land use policy

The Dutch market shows some differentiating

characteristics compared to other countries (NVB, 2014). It is characterised by relatively large amount of smaller cities. These cities all have their own municipality, accompanied with their own urban- and space use planning policy. There is one main area which is a big determinant for country-wide developments: The Randstad (figure 1). Most major cities like Amsterdam, Utrecht and Rotterdam are located in this area.

The share of people living in a large metropolitan

area (500.000 inhabitants or more) is rather small. The Netherlands shows a ‘polycentric urban structure’. This means that most of the Dutch inhabitants reside in medium- and smaller-sized towns or villages. The rationale behind this is that the Dutch urban regions are built up from several surrounding core areas (PBL, 2016). This is quite different than for instance in the United

States. Predictions for the Netherlands show that an increase in inhabitants for largely urban

municipalities is expected. This increase is foreseen to affect the Randstad area most. Almost half of the inhabitants of the Netherlands live in the Randstad.

The Netherlands is a relatively small country. It is relatively easy to move, as absolute distances

are small. Following the ongoing urbanization, it is not uncommon for people to work in one (large) city and live in another. This is opposed to commuting dynamics in for instance the United Kingdom or the United Kingdom, where cities are further apart. The previous characteristic might have an impact on the way Dutch house prices show certain dynamics. If prices in one city rise above equilibrium levels, it is not unlikely that people will move between cities. The barriers to move are therefore expected to be lower. One would therefore expect that shocks in one Dutch city will have higher effects on neighbouring cities than in other countries like the United States and the United Kingdom. This is especially applicable to the cities in question; Amsterdam and Utrecht. The cities are highly connected, with travel times of about 20 to 30 minutes.

Van Fulpen (1988) claims the Dutch market can be characterised by a persistent shortage of supply. This shortage is due to several factors, including demographic trends and spatial constrictions of a physical and regulatory nature. This shortage might entail that people are more inclined to migrate from region to region, and city to city, as certain shocks unfold to the market.

The argument is supported by Neuteboom and Brounen (2010) in their paper on the accessibility

of the Dutch homeownership market. The paper analyses to what extent the market for homeownership is accessible for first-time buyers. The researchers show that first-time buyers have a relatively weak position. This inaccessibility for first-time buyers is especially applicable in recent times. The Amsterdam housing market is undergoing a strong boom, starting from 2013/2014 (ING, 2017). Neuteboom and Brounen (2010) find that the variation in market power per region is huge. Popular markets are flooded with richer, older buyers who have already built up some housing equity. Entrants to the market thus have a much lower probability of success in these regions. A consequence of this is that people might be stuck with consuming an amount of housing services that is insufficient to them. Households are thus tempted to migrate, or underconsume.

(14)

10

Broitman & Koomen (2015) identify two factors that affect the dynamics of the Dutch stock in

residential housing. The first one is economic performance. Broitman & Koomen (2015) find that there is a threat of uneven development following the current residential trends. The Netherlands is subject to an ongoing situation of urbanization. Housing demand in the core areas in the Randstad will have to be compensated by supply in the peri-urban areas. The zoning and land use plans for these regions are increasingly important. Housing shortages in popular areas like Amsterdam might lead to increased demand in the surrounding peri-urban areas. As the Dutch market yields low absolute distances, this might be visible in the directly surroundings as well as in nearby cities like Utrecht, Haarlem and Almere. Glaeser et al. (2006) elaborate by stating that the economy of a region is closely related to the amount of physical structures made by mankind existing in an area. It is therefore expected that economic welfare leads to higher levels or quality of residential stock, and possibly higher prices.

The Dutch housing market can be characterised by the amount of intervention policies. These

policies date back from several decades ago until now. The physically available space in the country is very limited. Governments thus implement policies to guard the landscape. These intervention policies can be categorised into two types: land use policies and housing policies. The Dutch government implemented spatial planning in a structured way in 1965. At this point the ‘Wet op de Ruimtelijke Ordening’ (Spatial Planning Act) was introduced. This law governs how urban planning is being handled in the Netherlands. The main implications for cities and municipalities are visible in the land use policy. Certain areas are to be maintained in their current use. Certain city parks or shops can thus not be converted to residential housing (Rijksoverheid, 2017). This has large implications for municipalities, as the physical capability to grow is small. Supply is inelastic, which has its effects on the house prices of the city in question. If there is an upward shock in housing demand, supply cannot increase by the increased demand. Prices are then expected to increase.

3.2 The Dutch housing market and its drivers

The Dutch residential transaction market in 2016 comprised a total worth of over 52.3 billion euros. This is around 8% of the total GDP (CBS, 2017). It is often said that the real estate market in the Netherlands works as a magnifying glass for the overall economy. In times of economic welfare, the growth is pushed upwards by real estate market growth. In economic downturns, the situation is worsened by stagnation in the housing market (FD, 2017). The percentage of owner-occupied housing is around 60 percent. Another 32 percent is socially let, and 8 percent is private rental. This division is quite remarkable from an European perspective. In Germany, the division is 40 percent owner occupied, 53 percent private rental and 7 percent social rental (Haffner et al. 2009). It seems the social sector is comparably large.

The NVB, which is an association for banks, highlights four distinguishing factors of the Dutch

market. First, interest payments on mortgages are (for now) 100% deductible from the income tax. Second, the rental market is dominated by social housing institutions that are not-for-profit. Third, the government imposes a regulation that involves high levels of rent control and protection of tenants. Last, Dutch municipalities have a strict zoning regime for the construction and development of new dwellings (NVB, 2014). The previous reasons might be determinants as to why the Dutch housing market shows different dynamics when it comes to ripple effects.

As said, characterising for the Dutch market is the heavy regulation by the government. Largely

influencing is the mortgage interest deductibility. The idea is that you can deduct your mortgage interest payments from your income tax. For some people this means that the discount on the interest payments is more than 50% (NVB, 2014). The consequence of this is that at least some of this benefit is capitalised into the house prices. Also, total mortgage debt is relatively high. In the years leading up to 2005, it was not uncommon for Dutch households to have a loan value of 110% relative to the house value. This consequently leads to severe risks.

(15)

11

In the subsequent years, this has come down to about 89% in 2014 (NVB, 2014). Right after the financial crisis of 2008 the situation worsened. People were losing their jobs, and the housing market collapsed. The problem was that certain loans were ‘under water’ leading to the situation where people can’t move because selling their house won’t pay off the mortgage they have on this house (NVB, 2014). This obviously led to large problems for banks and households.

The NVB (2014) argues that prices were driven by demand factors over supply factors. The household loan capacity plays a big role in this. As demand was high and banks were willing to supply more, prices went up. Banks developed certain mortgage products that optimised money supply to households. Loan to values increased as well. Supply factors did not play a big role in the dynamics of house prices. A main vulnerability of the Dutch market arises from this high household debt. A consequence is that the Dutch housing market is expected to behave more volatile than in countries where deducting the interest is not possible (DNB, 2016).

A further regulatory measure introduced by the government is the National Mortgage Guarantee

(NHG). This is a measure that ensures some form of insurance to banks when households default on their mortgage loan. This measure is introduced to increase the willingness to lend of banks. The system works in the following way. The borrower pays a fee of around 0.70 percent of the value of the dwelling in question. This fee is added to a fund that ensures that defaulting loans are paid off when needed. The lender usually gives the borrower a discount on the interest rate of between 0.3 percent and 0.6 percent. When the borrower defaults on his loan, the debt is repaid out of the fund. The borrower is still liable for the debt, but the bank passes over his responsibilities to the fund. The debt can only be cleared in case of decease, divorce, long-term unemployment or disability. The NHG can only be applied for when the borrower is buying their first house, and only up to an amount of €247.450 (NHG, 2017). At first the NHG seems like a solely positive measure for Dutch households. One issue is that the existence of a market where NHG is implemented is expected to yield higher prices. Banks are willing to lend more to households. People are thus able to pay more for a similar house. This is expected to drives prices up.

An interesting analysis on the Dutch market is performed by a commission in order from the

Dutch Parliament (2014). The commission analyses the developments in the Dutch housing market over the prior 20 years. The researchers conclude demand levels are far above the supply available. Between 1995 and 2008, house prices increased with 250 percent. Other countries experienced some growth as well, but simultaneously saw supply being increased significantly. The absolute supply of housing in the Netherlands did not respond adequately to demand developments. The elasticity of supply is low. Construction firms worry that prices and demand drop during the process of construction. This resulted in persistent shortages in the housing market. Young households especially suffer from this situation. Municipalities were more interested in building residential objects in the high-end of the market. This is as land profits are higher for high end construction. The availability of housing for young households in main urban areas following this was low (Dutch Parliament, 2014).

It is expected that the previous characteristics contribute to the ripple effect being in place. It is

observed that most of the regulations and drivers discussed have the effect of driving prices up, and increasing price volatility. The peaks and troughs are expected to be visible in the high-end of the market first, where prices are most volatile, and loan values high. The price shocks would then ripple over through to lower segments of the market.

Social sector 3.2.1

The Dutch government has been intervening in the housing market since the start of the 20th_century.

The grip of the government was strengthened in the post-war period, when the housing shortages were especially urgent. The government induced strong influence on the place, quality and amount of constructions. A system of social renting came into place.

(16)

12

The Netherlands knows quite a few housing associations that are non-for-profit and are aimed at

supplying housing for the lower-income group. These organisations are government instruments, but are mainly independently operating organisations. As they don’t have an aim for profit, they can construct and let dwellings at a relatively low price.

It is widely reported that the Dutch market is dysfunctional due to this social sector. In

Amsterdam, sixty percent of the dwellings are socially let (Municipality of Amsterdam, 2015). One might argue that this makes the accessibility of the social sector good, but this isn’t the case. The wait times for social rented housing are on average 11.5 years in Amsterdam. This ranges from 14 years in the popular areas to 8 years in the less popular areas (Municipality of Amsterdam, 2015).

The differences in prices between the social sector and the private sector are significantly large.

This leads to certain lock-in effects. For a large amount of households it is not possible to consume the desired amount of housing services through the market sector. This is, especially in popular areas like Amsterdam and Utrecht. Leaving the social sector would mean for them to move away from the city, or drastically decrease the house size they inhabit. The arising problem is that the free supply of socially let dwellings is low. Households tend to stay in their socially let house, long after they surpass the income bound that would make them eligible for the social sector (Haffner et al., 2009; Municipality of

Amsterdam, 2015). There are some implications for the owner-occupier market. As 60% of the dwellings is socially let, supply is low. Prices in the city are subsequently high. It also seems that the price volatility in larger cities like Utrecht and Amsterdam is high because of this shortage of supply.

(17)

13

District specification and hypotheses

4

In this part, the considered municipalities and districts are listed and explained. The characteristics are discussed. These characteristics are used to form certain hypotheses regarding the existence of the ripple effect.

4.1 The Amsterdam housing market

Amsterdam is the most popular city in the Netherlands. This goes to show in the real estate prices within the city. The average price per square meter in the capital is about two times the national average (CBS, 2017). A vital part of this excessive premium can be explained by a shortage of supply. After the financial crisis of 2008, construction was very limited. Some analysts expect that at the current rate of

construction, there will be a shortage of 270.000 dwellings in 2040. This shortage does not only come from an increase in inhabitants. There is also a rising number of single-person households (Municipality of Amsterdam, 2015). The following subsections will provide more background about the districts in

Amsterdam that are included in the empirical part of this thesis. An overview of the districts is provided in figure 2. The included districts per ripple model are presented in appendix 1. For the purpose of this thesis, six main hypotheses are set up regarding Amsterdam and Utrecht. Sections 4.1.1 until 4.3.3 present the sub-hypotheses. The first main hypothesis reads:

Hypothesis 1: Amsterdam shows a significant ripple effect, where shocks in house prices ripple out from the centre to the outer parts of the city.

(18)

14

Amsterdam Centrum 4.1.1

The city centre of Amsterdam is the oldest part of the city. The neighbourhood has an authentic appearance. The architecture is widely known to be a distinctive trait of the city centre of Amsterdam. For clarity, the city centre of Amsterdam is referred to as ‘Amsterdam Centrum’ from here on. The area resides around 86.500 people. This is around 4 times as much as the city centre of Utrecht. Roughly speaking, the area can be categorised into three neighbourhoods: ‘de Jordaan’, ‘Grachtengordel’, and ‘de Wallen’. The average price per m² over the sample period is around €3.870. This is the highest level in the whole city. Due to its good reputation, as well as high prices the area is expected to show significant house price impact on its surrounding districts.

Oud-Zuid 4.1.2

Oud Zuid is comparably popular as the city centre of Amsterdam. The area is known to be the expensive part of the city. The area resides about 92.000 people. The main neighbourhoods are ‘de Pijp’ and ‘Oud-Zuid’. De Pijp is a former working-class quarter. Due to its favourable location, and vibrant nightlife the neighbourhood has severely increased in popularity after the 90’s. The Oud-Zuid area is the most high-end part of Amsterdam. The area resides mostly wealthy households. The neighbourhood appearance is authentic, in line with the city centre of Amsterdam. This goes to show in the average prices per m². These are comparable with the city centre: €3.800. Strong ripple effects are therefore expected to spill over from the city centre. Oud-Zuid is one of the most prominent areas in Amsterdam. The district is therefore expected to show some form of a two-way ripple effect, where Oud-Zuid has a similar effect on the city centre as well.

Oud-West 4.1.3

District Amsterdam West is located on the western side of the city centre. For clarity in relation to Utrecht West, the district will be referred to as ‘Amsterdam Oud-West’ in this thesis. It is directly adjacent to the city centre. The area consists of four main neighbourhoods: Oud-West, De Baarsjes, Westerpark, and Bos en Lommer. The area is adjacent to Geuzenveld-Slotermeer on the west side, and Amsterdam Centrum on the east side. The location and popularity of the neighbourhoods are comparable. It must be said that Oud-West is the most popular area of the four, due to its favourable architecture and location next to the city centre. The average price over the whole district is around €3.319 per m². Due to its popularity and favourable locations, Oud-West is expected to be one of the main diffusors in the price dynamics within the city. This is, alongside Amsterdam Centrum and Oud-Zuid.

Watergraafsmeer/Oost 4.1.4

The city district Amsterdam Oost, which is the eastern part of Amsterdam, is divided into two

components. These two parts are: Watergraafsmeer/Oost and Zeeburg/IJburg. This is done to increase comparability of results. Watergraafsmeer/Oost is referred to in this thesis as ‘Amsterdam

Watergraafsmeer’. The main neighbourhoods in Watergraafsmeer are ‘Oosterparkbuurt’, ‘Indische Buurt’, and Watergraafsmeer. These neighbourhoods reside about 120.000 people. These are mostly older neighbourhoods, with authentic appearances. Watergraafsmeer has a reputation as a good

neighbourhood for families. Average prices are high: around €3.200 per m². The Watergraafsmeer area is expected to show significant lag dependency with respect to the city centre.

Zeeburg/IJburg 4.1.5

The second part of Amsterdam Oost that is considered includes Zeeburg and IJburg. For the purpose of this thesis, the combined district is referred to as ‘Zeeburg’. Both these neighbourhoods are distinctive in the way that they are comparatively new. Zeeburg and IJburg were mainly built from 1990 onwards. The area is mostly covered with medium- to high-end residential buildings.

(19)

15

The average dwelling size is high: over 100 m². The area is growing fast, and is now residing around 22.500 people (Municipality of Amsterdam, 2015). Average prices are about €3.100 per m² in this area, which is just above average. The Zeeburg area is expected to show significant lag dependency with the city centre. Due to differences in building period and inhabitant mix, Zeeburg is not expected to show lag dependency on Watergraafsmeer.

Zuideramstel 4.1.6

The Zuideramstel area comprises two neighbourhoods: ‘Rivierenbuurt’, and ‘Buitenveldert’. The area inhabits around 50.000 people. The Rivierenbuurt area is popular, as it is located on a favourable location. The architecture appears authentic, and there are ample retail facilities. Buitenveldert has an moderate reputation, and is known for its large residential flats. The ‘Vrije Universiteit’, which is a

university, is located in the area. It therefore for instance attracts students who want to live close to their university. Average prices are about €2.680. The Zuideramstel area is expected to show significant lag dependency with respect to Oud-Zuid.

Geuzenveld-Slotermeer 4.1.7

Geuzenveld is located in the far west of Amsterdam. For the remainder of this thesis, the district will be referred to as ‘Geuzenveld’. Historically, it is an unpopular area in Amsterdam, yielding high levels of crime and inhabitant dissatisfaction (Municipality of Amsterdam, 2015). The area inhabits around 45.000 people. The architecture is characterised by large residential flats. The area was mostly developed in the 1950’s. Average prices are about €2.100 per m², which is at the lower end in Amsterdam. Geuzenveld is located at the far west of Amsterdam. The area is adjacent to Amsterdam Oud-West. Geuzenveld is therefore expected to show lag-dependency with respect to Oud-West.

Slotervaart/Overtoomse Veld 4.1.8

Slotervaart and Overtoomse Veld are located in the west of Amsterdam. For the purpose of this thesis, the combined district is referred to as ‘Slotervaart’. The area has a somewhat better reputation than other parts of Amsterdam Nieuw-West. The number of inhabitants is around 38.000. The area is characterised by small single-family units for the working class, as well as large residential flats. The average price per m² over the sample period is about €2.300. Slotervaart is located in between Osdorp and Oud-Zuid. Osdorp yields similar architectural traits as Slotervaart and Overtoomse Veld. There might therefore be some spatial arbitrage (Meen, 1999). The Slotervaart area is expected to show lag

dependency with respect to Oud-Zuid and Osdorp.

Osdorp 4.1.9

Osdorp is located in the far west of Amsterdam. The number of inhabitants is about 67.000. Historically, the neighbourhood reputation is moderate. The municipality initiated a large redevelopment plan in 2001. Many thousands of dwellings are being renovated or demolished and rebuilt. Due to its location far from the centre, it is not surprising that prices are quite low: €2.100 per m². The area is expected to show lag dependency with Slotervaart as the price levels are comparable, and it is directly adjacent to this area.

Zuid-Oost 4.1.10

Amsterdam Zuid-Oost is also known as ‘De Bijlmer’. The area was developed in the 1960’s. The idea was to build a new city close to Amsterdam. When the construction was completed, it turned out the area was not as popular as previously expected. The neighbourhood attracted mainly lower-income

households. Amsterdam Zuid-Oost is nowadays still known for its high unemployment and crime rates. The architecture is characterised by large monotonous residential flats.

(20)

16

The average price is the lowest in the whole of Amsterdam: €1.830 per m². Districts like Watergraafsmeer and Diemen do not show similar characteristics, and are comparatively far away. Therefore, no spatial arbitrage of large extents of migration (Meen, 1999) is expected. The Zuid-Oost area is expected to show significant lag dependency with respect to the city centre. The city centre is seen as a determinant of price growth within the city. Price growth in the city centre signals prosperity and confidence in the real estate market to areas nearby.

Noord 4.1.11

Amsterdam Noord is distinctive in the way that it is cut off from the rest of Amsterdam by the river: 'Het IJ’. The central part of Amsterdam can be reached by ferry or bus. Interesting to note is that the metro line connecting Amsterdam Noord with the city centre is expected to be finished mid-2018. This might give the area a price boost, as well as increase the extent to which the ripple effect is observed. The number of inhabitants is around 93.000. The architecture in the district is quite diversified. The western part at the river is characterised by large logistics buildings. The rest of the area is mostly covered with small- to medium-sized single-family units, as well as large residential flats. The average price is about €2.250 per m². The area is expected to show lag dependency with respect to the centre of Amsterdam. This is, as Amsterdam Centrum is the only area with which it has a direct connection.

4.2 The Utrecht housing market

Historically, Utrecht is one of the most popular residential cities in the Netherlands. Some refer to it as ‘small Amsterdam’. It is characterised by an authentic city centre with canals and city squares. The Washington Post referred to Utrecht as ‘a hidden gem in Amsterdam’s shadow’ (2015). The city is one of the fastest growing cities in the Netherlands.

The municipality expects the city to grow with about 60.000 inhabitants in 2030. The city faces a

large shortage of housing supply in the central areas. The municipality is planning on housing 40.000 new inhabitants within the current city bounds. To achieve this, the available land has to be used more extensively. For the purposes of this thesis it is an interesting test-case. The city map including the districts is presented in figure 3. The included districts per ripple model are presented in appendix 1.

The ripple effect is expected to be present in a different way in Utrecht than in Amsterdam. The

city centre is significantly smaller than in Amsterdam. Price shock diffusion is therefore expected to originate in other districts as well. The following subsections provide more background about the districts in Utrecht that are included in the empirical part of this thesis. Also, the hypotheses regarding the ripple effect are presented. The second tested hypothesis is:

Hypothesis 2: Utrecht shows a significant ripple effect, where shocks in house prices ripple out from the centre to the outer parts of the city.

(21)

17 Figure 3: Utrecht city districts

A third hypothesis arises from theory. As said, Amsterdam is seen as a leader in house price dynamics in the country. It is therefore expected that Amsterdam shows significant effects on cities like Utrecht. This is in line with findings by for instance Meen (1999), and van Dijk et al. (2016). Hypothesis three therefore reads:

Hypothesis 3: Amsterdam shows a significant ripple effect with respect to Utrecht, where shocks in house prices in Amsterdam have a significant effect on future house price shocks in Utrecht.

The City Centre 4.2.1

The city centre is the smallest district in the city, residing around 18.000 people (Municipality of Utrecht, 2016). The city centre of Utrecht is known worldwide for its authentic appearance, brick wharfs and canals. For clarity, the city centre of Utrecht will be referred to as ‘Utrecht Binnenstad’ from here on. The average price per m² over the whole sample period is around €2.800. The Binnenstad area is expected to have a less dominant influence with regard to price shocks rippling out than the city centre of

Amsterdam. The area is considerably smaller in terms of inhabitants and surface. The architecture is similarly distinctive as in the city centre of Amsterdam. Price shocks due to economic prosperity or other macro-economic factors are expected to show simultaneously with other areas in Utrecht.

Utrecht West 4.2.2

As the name says, Utrecht West is located in the west of Utrecht. Utrecht West mainly consists of two neighbourhoods: Lombok and Oog in Al. The number of inhabitants is around 29.000 (Municipality of Utrecht, 2015). Oog in al is one of the most popular residential areas of Utrecht. Characterised by authentic 1930’s dwellings, the neighbourhood has grown in popularity over the past years.

(22)

18

Lombok is historically one of the neighbourhoods with large extents of low income households and people with a background of migration. Due to its preferable location and appearance, the

neighbourhood has grown in demand over the past years (Municipality of Utrecht, 2016).

The average price level per m² over the sample period is around €2.200. This is significantly lower

than the centre. Taking the previous into account, the ripple effect is expected to be present in Utrecht West. This is, with respect to the city centre. The neighbourhood is located directly adjacent to the city centre. The appearance is authentic, and popularity levels are high and increasing. The district is

therefore expected to show two-way effects with respect to the city centre. This means that price shock diffusion might originate both in Utrecht West or Utrecht Binnenstad in a similar fashion.

Utrecht Noordwest 4.2.3

Utrecht Noordwest is located in the north-west of Utrecht. Utrecht Noordwest also is defined by two main neighbourhoods: Zuilen and Ondiep. Ondiep is directly adjacent to the city centre. Ondiep is known as a ‘krachtwijk’, which is freely translated from Dutch as a ‘neighbourhood of force’. In 2007, the Dutch government introduced a list of the 40 neighbourhoods that yield the highest levels of criminality and inhabitant dissatisfaction. The municipality of Utrecht focuses their attention on improving the

appearance and inhabitant perception of these neighbourhoods. The eastern part of Zuilen is also put on the list of ‘krachtwijken’. This neighbourhood faces similar issues as Ondiep (Municipality of Utrecht, 2016).

The average price level per m² over the whole period is around €2.000. This is significantly lower

than the city centre. Due to its location, Utrecht Noordwest is therefore expected to show lag

dependency with respect to the city centre. The effect of shocks in Utrecht Noordoost are expected to be

higher, as it is also directly adjacent, and price levels are somewhat lower than the city centre.

Utrecht Overvecht 4.2.4

Utrecht Overvecht is also known as a ‘krachtwijk’. It is located in the north part of Utrecht The

neighbourhood was built in the 1960’s to accommodate growing demand for housing in the city. The area is built with mainly large multi-family residential flats. Most of the area is owned by the housing

associations, therefore decreasing the size of the owner-occupier market (Municipality of Utrecht, 2016).

The average price level per m² over the sample period is around €1.700. This is the lowest in the

city by distance. Overvecht is furthermore not adjacent to the city centre. The district of Overvecht is therefore not expected to show a ripple effect following shocks in the city centre. Shocks are expected to

be rippled over through areas like Utrecht Noordoost or Utrecht Noordwest.

Utrecht Noordoost 4.2.5

Utrecht Noordoost is located in the north-east of Utrecht. The main residential areas in Utrecht Noordoost are Tuinwijk, Tuindorp, Wittevrouwen and Voordorp. The district resides around 38.000 inhabitants. The residential stock is around 17.800.The district shows above average grading by inhabitants on topics like social cohesion and all-round neighbourhood perception (Municipality of Utrecht, 2016).

Wittevrouwen is one of the most popular residential areas in Utrecht. The housing in the area

was mainly built in and around 1900, inhabiting mostly wealthier households. The neighbourhood appearance is authentic. Tuindorp is also a popular neighbourhood for households, especially with young children. The area is covered with mainly single-family housing. Tuinwijk and Voordorp are also quite popular residential areas, showing somewhat lower prices. The overall price level is among the highest in Utrecht: €2.550 per m². The architecture in Wittevrouwen and Tuindorp is quite comparable with the city centre. The appearance of the area is authentic.

(23)

19

Due to its distinctive architecture and popularity, Utrecht Noordoost is expected to be a

determining factor in the rippling of price shocks over the city. The district is therefore expected to show high levels of price diffusion with the centre. This is, possibly in both directions.

Utrecht Oost 4.2.6

Utrecht Oost is located in the east of Utrecht. The housing stock is around 15.000, and the area inhabits close to 33.000 people. The neighbourhood perception is good, and crime rates are low Utrecht Oost consists mainly of the neighbourhoods Oudwijk and Rijnsweerd. Oudwijk is one of the most high-end residential areas of the city. The neighbourhood inhabits mainly wealthier households. There is a large extent of segregation within the neighbourhood (Municipality of Utrecht, 2016). The northern part is covered with larger residential villas and authentic dwellings. The southern part is a former working-class quarter.

As said, the price levels in the district are high. The average price per m² over the sample period

is €2.625. The neighbourhood is located directly adjacent to the city centre, the appearance is authentic, and popularity levels are high. The district is therefore expected to show two-way effects with respect to the city centre. This is, in a similar fashion as Utrecht Noordoost and Utrecht West.

Utrecht Zuid 4.2.7

Utrecht Zuid is one of the smaller districts in Utrecht. It is located in the south of Utrecht. The area resides around 27.000 inhabitants, living in 13.000 dwellings. The main neighbourhoods in the area are Hoograven and Lunetten. The criminality rates in this part of the city are high. The grades for

neighbourhood perception are average. Hoograven is also known as a ‘krachtwijk’. The municipality is trying to improve the neighbourhood by implementing several projects increasing the neighbourhood on themes like employment and education (Municipality of Utrecht, 2016).

Utrecht Lunetten is more popular. The district was built in the 1970’s. The municipality

experimented by letting the potential inhabitants have a say in how the neighbourhood should be built and should look like. The neighbourhood coherence and involvement was, and still is, high because of this. The area is mainly covered with single-family housing. The price levels in the area are among the lowest in Utrecht: €1.960 per m². The popularity of the neighbourhood is not too high, but the area is directly adjacent to the city centre. The district Utrecht Zuid is therefore not expected to be a strong determining factor within the city’s price dynamics. Price shocks are expected to ripple over from the city centre.

Utrecht Zuidwest 4.2.8

Utrecht Zuidwest is one of the larger districts in Utrecht. It is located in the south-west of Utrecht. The neighbourhood resides around 37.500 inhabitants. The area is known to have a large share of inhabitants obtaining social income aid (Municipality of Utrecht, 2016). The grades for district safety and youth hindrance are among the highest in the city. The main neighbourhoods in the area are Kanaleneiland and Rivierenwijk. Kanaleneiland is also categorised as a ‘krachtwijk’. The area is known for its large residential flats, as well as high levels of immigrants. The share of non-western immigrants in the area is over 50%. The area is known to be a quiet residential area. The build is categorised mainly by smaller single-family units (Municipality of Utrecht, 2016).

The price level in the district is average, showing prices per m² of about €2.100. The area is

directly adjacent to the centre. Due to its questionable reputation, and unpopular architecture Utrecht Zuidwest is not expected to be a determining factor within the city’s price dynamics. The district is expected to be subject to a ripple effect with respect to price shocks from the centre out.

(24)

20

Leidsche Rijn/Vleuten-De Meern 4.2.9

For the purpose of this thesis, the areas Leidsche Rijn and Vleuten-De Meern are combined into one district. Section 5 on data preparation elaborates on why this is done. The following part discusses Leidsche Rijn and Vleuten/De Meern separately, and formulates combined hypotheses. For the purpose of this thesis, the combined districts will be referred to as ‘Leidsche Rijn’.

Leidsche Rijn is known as the largest ‘Vinex’ area of the Netherlands. The Vinex areas where

mainly built between 1995 and 2005. The Vinex areas are put in place to accommodate the increasing amount of inhabitants in the Netherlands. Vleuten-De Meern is located in the far west of Utrecht (Municipality of Utrecht, 2016). Vleuten-De Meern used to be a separate municipality. Vleuten and De Meern are nowadays part of the Municipality of Utrecht. This is done in behalf of the build of Leidsche Rijn. The area was a vital link in connecting the Leidsche Rijn area to the city. Leidsche Rijn nowadays links Vleuten and De Meern to the city bounds of Utrecht (Municipality of Utrecht, 2016).

Price levels are average, showing prices per m² of around €2.150. It is interesting to note that

average house size is in fact higher in Leidsche Rijn than in the other parts of Utrecht. It is a rule of thumb that larger dwellings show lower prices per m². The area is not directly adjacent to the city centre. The Leidsche Rijn area is therefore expected to show lag-dependency with respect to Utrecht West over Utrecht Binnenstad.

4.3 Separate municipalities

As said, Amsterdam is the largest city in the Netherlands. It is therefore not unthinkable that the Amsterdam residential market has some effect on its surrounding areas. As with every large metropolitan, Amsterdam is surrounded by larger and smaller cities. This thesis on the ripple effect includes some of these cities to examine the effect of Amsterdam on its surroundings. The included municipalities are: Diemen, Ouderkerk aan de Amstel and Haarlemmermeer. A map of the area is provided in figure 4.

(25)

21

Diemen 4.3.1

The Diemen area is an interesting one, as it is quite attached to Amsterdam. The area is governed by its own municipality. It is closer to the city centre than the Zuid-Oost area. The area inhabits around 27.000 people (CBS, 2016). Diemen used to be a separate entity. Due to the ongoing growth of the city

Amsterdam, it has almost become a part of the city. The area still operates as an independent entity, yielding its own city centre. The hypothesis that is tested reads:

Hypothesis 4: Diemen shows significant lag-dependency with respect to the city of Amsterdam over its directly adjacent neighbourhoods.

The ripple effect is expected to be quite present in this area, as it can almost be seen as a part of Amsterdam. It is directly adjacent to the eastern part of Amsterdam.

Ouderkerk aan de Amstel 4.3.2

The second separate municipality that is included in the analysis is Ouder-Amstel. Ouder-Amstel is an area which comprises the town of Ouderkerk aan de Amstel. For the remainder of this thesis, the area is referred to as Ouderkerk aan de Amstel. This is also an interesting test case as it is a separated village, more so than Diemen, but is still close to Amsterdam. The distance from the city centre is around 5-6 kilometres. Ouderkerk aan de Amstel is known as a popular option for people wanting to leave the busy city. The average income in Ouderkerk aan de Amstel is comparably high, reaching around 29.300. This is 25% more than the country average. Ouderkerk aan de Amstel inhabits around 13.500 people (CBS, 2016). We test whether Ouderkerk aan de Amstel is significantly affected by price shocks in Amsterdam. The hypothesis that is tested reads:

Hypothesis 5: The city of Amsterdam shows significant ripple effects with respect to Ouderkerk aan de Amstel.

Haarlemmermeer 4.3.3

The last municipality that is included is Haarlemmermeer. The Haarlemmermeer used to be a lake before it was dried up in around 1850. The area covers around 185 square kilometres and is located on the west side of Amsterdam. The main residential areas are Hoofddorp and Nieuw-Vennep. Hoofddorp is located around 12 kilometres away from the Amsterdam city centre. This is around five kilometres less than Nieuw-Vennep. For the purpose of this thesis, the two cities are separately considered with respect to the ripple effect. The average income of the area is around €26.300. This is around 12% more than the national average (CBS, 2016).

In terms of size, Hoofddorp is the biggest. The number of inhabitants is around 76.000 (CBS,

2016), whereas Nieuw-Vennep inhabits around 31.500 people (CBS, 2016). The ripple effect is expected to show in the area. Price shocks in Amsterdam house prices are expected to ripple through Hoofddorp into Nieuw-Vennep. As Amsterdam is such an important hub in the area, direct effects are expected as well. The hypothesis that is tested thus reads:

Hypothesis 6: House price shocks are expected to ripple over from Amsterdam to the outer parts of Haarlemmermeer both directly and indirectly.

House price dynamics in and between Amsterdam and Utrecht : the ripple effect

Master Thesis