The effect of the expansion of Lelystad
Airport on house prices
Master Thesis
Faculty of Economics en Business
MSc Finance - Specialization Finance and Real Estate Finance
Author
Kristie Lek (10662022) Supervisor
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 1 Statement of Originality
This document is written by Kristie Lek who declares to take full responsibility for the contents of this document: I declare that the text and the work presented in this document are 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.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 2 Acknowledgements
Writing this thesis would not have been possible without the support of several people. I would therefore like to express my gratitude to them. I would like to thank my supervisor dr. F.P.W. Schilder, who was very helpful with his quick feedback and offered support and guidance in the entire process. I would also like to thank the Dutch Association of Realtors (NVM) for providing all necessary data and making an exception for this research to include slightly more municipalities in the dataset. Special thanks to my parents, who have always been there for me during the years of my studies.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 3 Abstract
This paper examines the effect of the expansion of Lelystad Airport and the changing flight routes regarding the expansion on house prices. The paper focusses on two different announcements of flight routes in the years 2007 and 2012. The effects are tested using data from the Dutch Association of Realtors from the year 2004 up until 2017. A difference-in-difference model is constructed and different variables of interest are created to be able to investigate both the overall effect and the effect of the different flight routes. The overall effect of the expansion of Lelystad Airport on house prices is -0.89 percent. This is equal to €2188 with an average transaction price of €246,148. When examining the effects of the different announcements, both announcements show negative and significant effects of -1.48 percent and -2.15 percent respectively. No conclusion can be drawn regarding the effect of the second announcement on houses below the flight route of the first announcement. Apart from that, the effect is also investigated for different altitudes and the results are +1.32 percent for altitudes below 1800, -5.21 percent for altitudes between 1800 and 2700 meters and -7.49 percent for altitudes of 2700 or more.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 4 Table of Contents
1. Introduction ... 5
2. Literature Review ... 8
2.1 House Price Models ... 8
2.2 Behavioral Economics ... 9 2.2.1 Behavioral Finance ... 10 2.2.2 Announcement Effects ... 11 2.3 Related Literature ... 13 2.4 Flight Routes... 15 2.5 Hypotheses ... 19 3. Methodology ... 21 3.1 Method ... 21 3.2 Variables ... 23 3.2.1 Dependent Variable ... 24 3.2.2 Variables of Interest ... 24 3.2.3 Control Variables ... 26 3.2.4 Fixed Effects ... 27
4. Data and Descriptive Statistics ... 28
5. Results ... 32
5.1 Effect of expanding Lelystad Airport ... 32
5.2 Effects of the different flight routes ... 35
5.3 Effects of different altitudes ... 36
6. Robustness Checks ... 39
7. Conclusion ... 41
8. Discussion and Limitations ... 44
8.1 Discussion and implications ... 44
8.2 Limitations ... 45
9. Reference list ... 46
10. Appendix ... 50
Appendix 1 – Zip Codes ... 50
Appendix 2 – Variables ... 53
Appendix 3 – Regression Results ... 55
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 5 1. Introduction
Schiphol Airport is one of the world’s busiest airports. According to Business Insider, it is the 11th busiest airport in the world and the 3rd busiest in Europe, with 68,515,425 passengers in
2017 (Business Insider, 2018). The increased popularity of flight vacations is making it extremely difficult for Schiphol to cope with the demand for flights. To be able to keep, and possibly strengthen, the economic function of Schiphol, it is decided to expand the smaller, regional airports. One of these regional airports is Lelystad Airport; a small airport close to Amsterdam. This airport used to facilitate business flights only, but because of the expansion Lelystad Airport is going to take over part of the (holiday) flights from Schiphol Airport. The expansion is going to be implemented in two phases. During the first phase, the number of aircraft movements will increase to 25.000 per year. In the final phase, the number of aircraft movements per year will increase to 45.000 (Lelystad Airport, 2018). There is a lot of critique on this decision. Inhabitants in the surrounding cities are afraid of noise, as the airport will be open from 6.00h till 23.00h with the possibility to extend this to midnight as explained by Omroep Flevoland (2015). Furthermore, the planes will be flying at a height of approximately 600 meters close to Almere, one of the bigger cities in the Netherlands. The planes have to fly at a low altitude (at approximately 1800 meters) in the Flevoland province because they would hinder air traffic from Schiphol Airport if they would fly higher. They are only allowed to climb higher after they have left the Flevoland province, but they have to remain low for the greater part of the flight routes over the Netherlands. Many people are worried that this will lead to too much noise in the provinces of Friesland, Groningen, Drenthe, Gelderland, Overijssel and Flevoland as these provinces will be part of the flight routes.
The housing market in the area surrounding the airport is booming. The expansion of the airport will even increase the demand for housing, as it will create a lot of new jobs according to the NVM (2017). Nevertheless, people are afraid that their houses will decrease in value due to the increased traffic noise. It would be interesting to investigate the significance of the effect of the increase in airplane noise on the house prices in the surrounding area. There is a lot of literature on traffic noise and its effect on residential properties. A lot of studies also investigate the effect of airport noise on house prices. There are even studies that examine the effect for Schiphol Airport. This effect will not be fully accurate for the area surrounding Lelystad Airport, as Schiphol Airport is much bigger, has a lot more flights and even has a railway station and highway next to it. It will be interesting to examine the effect of smaller airports on housing prices. In the case of Lelystad, it is interesting to investigate the effect of
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 6
the announcement of the flight routes; the behavioral aspect. As the airport is not yet operating at the higher level, it is not possible to investigate the effect of increased noise levels yet. Therefore, the effect of the announcement of the expansion will be investigated in this paper. The interesting part of Lelystad Airport is that two different flight plans have been announced at two different point in time and investigating the effect of this shift in flight routes on house prices of houses that are located underneath these flight routes has never been done. In this paper it will be done by testing if the first announcement had a negative effect on house prices underneath the first flight route and the second announcement on house prices underneath the second flight route. Besides, it needs to be investigated whether the negative effect of the first announcement has been undone with the announcement of the second flight plan.
Ever since the first plans to expand the airport were created, flight routes have been optimized to make sure as little residential cores would experience hinder from aircraft noise. The most detailed result is presented in the flight routes for the second announcement. Apart from investigating the shift in flight routes, it is also useful to see whether this route optimization process proves to be successful. This can be assessed by comparing the results of the effect of anticipated aircraft noise in the areas affected by the expansion of Lelystad Airport to results that are found in previous literature. All these assumptions result in the following research question:
What is the effect of the expansion of Lelystad Airport and the changing flight routes on house prices?
This research question will be tested with the use of data that was provided by the NVM from the year 2004 up until 2017. To be able to answer this question, three different hypotheses are formulated and tested using a difference-in-difference model with both time and zip code fixed effects, several housing characteristics as control variables and in which the standard errors are clustered at the zip code level. To investigate whether these hypotheses are true, different variables of interest have been created. These variables of interest are the interaction term of a dummy variable indicating whether the transaction is from after the announcement and a dummy variable indicating whether the house is located below a flight route. As this paper focusses on two different announcements and thus different flights routes over time, dummy variables are created for the different routes based on their zip codes. With the use of these variables of interest, the overall effect of the airport expansion as well as the shift in the flight routes could be investigated. Apart from that, the paper also investigates whether the effect is
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 7
different when the altitude of the plane changes. To be able to find an answer to this, the interaction term is again interacted with a dummy variable for the altitude of the plane.
The overall effect of the expansion of Lelystad Airport on house prices found in this paper is -0.89 percent, which is equal to a decrease of €2188 with an average transaction price of €246,148. This effect is in line with the negative effects found in previous literature. The effect is, however, much lower than in the article written by Jud and Winkler (2006). Houses affected by the expansion of Lelystad Airport are affected less severely than Greensboro/High Point/Winston-Salem metropolitan airport in North-Carolina and this suggests that the route optimization process for the flight routes to and from Lelystad Airport is successful as it has a great influence in the extent by which the houses are affected. When examining the effects of the two different announcements, both announcements show significant effects of -1.48 percent and -2.15 percent respectively. No conclusion can be drawn regarding the effect of the second announcement on houses below the flight route of the first announcement, because the coefficient of +1.32 is not significant. Apart from that, the effect is also investigated for different altitudes and the results are +1.32 percent for altitudes below 1800, -5.21 percent for altitudes between 1800 and 2700 meters and -7.49 percent for altitudes of 2700 or more. The results do not indicate the strongest effect for the first category as was expected, but for the last category. This contradicts with what is found in previous literature. Both Jud and Winkler (2006) and Dekker and van der Straaten (2009) find that the effect diminishes as houses are located further away from the airport. A possible explanation for this contradicting result is the increased job opportunities that houses in the first category benefit from together with the fact that the total effect of anticipated aircraft noise still had to be incorporated in house prices of houses in the third category.
The paper will continue with a literature review and this part results in the hypothesis development. The third section gives a description of the methodology and the data is provided and explained in the fourth section. After that, the results of the research are presented and robustness checks are shown. The conclusion is presented in the seventh section, followed by a discussion of the implications and limitations of this research.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 8 2. Literature Review
This section provides a literature review about forecasting house prices, behavioural economics, the effect of traffic noise on house prices and the different flight routes that have been announced. The first part of this section gives some background information about house price models. The second part explains the importance of behavioural economics. The next section reviews existing literature regarding the effect of traffic noise and airplane noise on house prices. Then an overview of the different flight routes that have been announced is given and the final part of this section contains the development of the hypotheses.
2.1 House Price Models
The efficient market hypothesis is an investment theory that states that capital markets are efficient. This theory states that all publicly available information is incorporated into share prices so that share prices reflect all relevant information. This also means that prices adjust quickly when new information becomes available (Fama, 1970). This does not hold for the real estate market. According to Lusht (2012), the real estate market is characterized by costly transactions, lack of transparency, a low number of buyers and sellers and a relatively slow reaction of supply. Within the real estate market, the housing market is regarded to be one of the most informationally inefficient. Pollakowski and Ray (1997) state that house prices do not immediately reflect changes in publicly available information. This makes the real estate market inefficient. Because of this inefficiency, returns are predictable and prices can be forecasted.
An accurate estimation of house prices is usually done by constructing house price models or indices. There are three different types of indices: appraisal-based indices, property share-based indices and transaction-based indices. Where appraisal-based indices are based on appraised values, property share-based indices use property share return for construction of the index. The main drawbacks of appraisal-based indices are appraisal lag bias (Geltner et al., 2001) and the fact that valuation methods might change over time (Eichholts, 1997). Property share-based indices are usually not very effective, as they are based on property share returns and the portfolios of property companies can change. Transaction-based indices are based on past transactions. The idea of transaction-based indices is to base the estimate of the price more directly on empirical evidence about the market value of a property (Eichholts, 1997). This is the reason why transaction-based indices are most widely used and this research will be transaction-based as well.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 9
There are two major approaches to construct a transaction-based index. The first approach is called the hedonic price method and the second approach is called the repeat-sales method. In a hedonic model, all transactions every period are used, and the qualitative differences between the properties that transact in different periods of time are controlled for in the regression. This is done by including control variables that describe the property characteristics that affect any given property’s value relative to other properties at any point in time (Geltner et al., 2001). The repeat-sales method is based on repeat-sales of the same structure and reflects the usage value of the location. In the repeat-sales approach, only properties that transact at least twice during the sample period are used and the regression is a model of changes in value across time within these properties (Geltner et al., 2001). The repeat-sales method only requires data regarding price, repeat-sales date and the location of the property, while the hedonic price method is based on all internal and external characteristics of the property. This means that the repeat-sales method is much less data intensive. However, much data is wasted as properties must have been sold twice. As aircraft noise is a characteristic of the location of a property, a hedonic model is used in this research.
2.2 Behavioral Economics
Mullainathan & Thaler (2000) describe behavioral economics as the combination of psychology and economics that investigates what happens in markets in which some of the agents display human limitations and complications. Behavioral economics studies the effects of psychological, social, cognitive and emotional factors on economic decision making of individuals and institutions and the consequences this has for market prices and returns. Because of the psychological influences, deviations from economic models are likely to occur. There are three different ways in which these deviations are possible. The first one is bounded rationality and means that rationality of individuals is limited by the information and cognitive abilities they have. The second one is bounded willpower and captures the fact that human decisions are not always in their own long-run interest. The last one is bounded self-interest and incorporates the fact that people are willing to sacrifice their own interests to help others. These three ways by which deviations from economic models occur are drivers of behavioral economics.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 10
2.2.1 Behavioral Finance
According to Camerer, Loewenstein and Rabin (2011) the explanatory power of economics is increased by providing it with more realistic physical foundations. By increasing the realism of the psychological foundations of economic analysis, the economics itself will improve as there are more theoretical insights and better predictions of phenomena. This in turn suggest a more sophisticated policy and that is what makes behavioral economics meaningful. Camerer et all. (2011) do, however, state that the neoclassical approach, which focusses on equilibrium, efficiency and utility maximization, should not be rejected. This approach provided the theoretical framework that can be applied to all types of economic behavior. A new approach to financial markets that is inspired by behavioral economics is behavioral finance. According to this approach, financial phenomena can be better understood when using models in which some agents are not fully rational (Thaler, 2005). In other words, some of the assumptions underlying individual rationality are relaxed; investors are no longer fully rational in their decision-making. An example of this might be an agent that bases a decision on beliefs that are incorrectly updated. With all these new insights into financial markets, the main drivers of market efficiency according to the Efficient Market Hypothesis, for example arbitrage, seem to be less strong than what was supposed. This led to the emergence of behavioral finance, which is an alternative view of financial markets that states that economic theory does not directly lead to the expectation that financial markets are efficient (Shleifer, 2000).
One of the first studies to test for context effects in economic behavior is the study done by Simonsohn and Loewenstein (2006). What they investigate is the extent to which home buyers are uncertain about their preferences, how they value goods and how easily they are influenced by arbitrary cues. They predict that people who move from more expensive cities are likely to rent more expensive apartments when comparing them to people who move from less expensive cities. Also, they presumed that as people got used to the new cities, they would adapt to the new price levels and readjust their housing expenditures. Apart from this, they expect that less tangible goods like diseases and clean air are just as unstable and susceptible to arbitrary cues. This study shows that the psychological concepts of anchoring, framing and contextual effects matter for housing market dynamics. Anchoring is using arbitrary baselines like list prices instead of valuations to formulate bids, framing is presenting market dynamics in a way that might lead to people paying different prices for the same object and contextual effects are about forming expectations in one market that influence prices paid in another market (Smith, 2011). Smith (2011) agrees with Simonsohn and Loewenstein, as she concludes
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 11
that housing markets may be rationally efficient but could also be influenced by an array vectors and driven by irrational behavioral psychologies. The main conclusions from these two studies is that economic fundamentals can fail to explain economic behaviors and that it might be useful to add psychological drivers to this analysis.
Several studies have investigated the different housing price bubbles and tried to understand the main drivers of house price formation. Shiller (2007) is one of them and found out that it does not seem possible to explain a house price boom in terms of fundamentals like construction costs or rents, but that psychological theory seems to be more reasonable. A boom occurs because of feedback mechanisms that encourage a view that states that housing is an important investment opportunity. This creates the incentive to buy a house even though the house is extremely overpriced. Shiller, among others, is convinced that this type of psychological theory fits the evidence better that normal economic fundamentals. The behavioral aspect of asset pricing is significant and should be considered when investigating house prices.
2.2.2 Announcement Effects
Announcement effects are part of behavioral economics, as responses to announcements are all about the way people respond to something that has not yet been implemented. On the one hand, people could be rational and make thought-through decisions based on announcements. On the other hand, people could as well be emotional and respond to announcements based on what they feel is going to be the effect. The reaction of market participants to different kind of announcements and timings has been investigated to be able to find an answer to the way people respond to something that is about to come. Multiple studies in financial literature find that late announcements result in lower abnormal returns and/or low stock return variability, while early announcements result in higher abnormal returns and/or high stock return variability. Researchers that have investigated these effects are Chambers and Penman (1984), Kross (1981) and Givoly and Palmon (1982). Kross and Schroeder (1984) find evidence that is in line with those previous studies, in combination with controlling for the timing effect. The timing effect persisted whether the earnings announcement:
1. contained either good or bad news, 2. contained fairly good or fairly bad news, 3. was made by either a large or a small firm, 4. was an annual or interim announcement.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 12
The main conclusion from all these studies is that people associate early earnings announcements with good news and that late earnings announcements are associated with bad news.
Hand, Holthausen and Leftwich (1992) examined daily excess bond returns that are the result of rating changes by rating agencies as Standard and Poor’s and Moody’s and excess bond returns that are the result of announcements regarding additions to Standard and Poor’s Credit Watch List. They observe that the announcements of additions to S&P’s Credit Watch List result in nonzero excess returns. Furthermore, announcements of downgrading and upgrading of ratings by rating agencies show effects on both bond and stock prices. The effect is more significant for downgrades and stronger for below investment grade bonds. According to Katz (1974) a 100 percent price adjustment to rating reclassifications takes place six to ten weeks after a reclassification. In this study, the price adjustment process of bonds to rating reclassifications is examined and announcements usually result in price adjustments some time after the announcement.
The price effect of announcements has been investigated a lot in finance literature. Jud and Winkler (2006) have applied this research to the real estate market and their paper even investigates the effect of airport expansions. They investigated the effect of the announcement of a new airport hub on house prices near the new airport. The airport they investigated is Greensboro/High Point/Winston-Salem metropolitan airport in North-Carolina. What they mention is that quite a lot of articles examine the effect of noise levels on property prices, but only few have measured the effect of announcements. Their main result shows that house prices in the 2.5-mile band from the Greensboro/High Point/Winston-Salem MSA declined with 9.2 percent after the announcement. The next 1.5-mile band shows house prices declines of 5.7 percent after the announcement. The announcement took place in April 1997 and their post-announcement period start from January 1999. They use a lag of approximately nine months between the announcement and the actual effect of the announcement.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 13 2.3 Related Literature
There is a lot of literature that focusses on the effect of traffic noise on the value of residential properties. Some of these papers examine the total effect of traffic noise, while others focus on either traffic or airport noise. The objective of a paper written by Wilhelmsson (2000) is to provide an analysis of the impact of traffic noise on the value of single-family houses. A hedonic price method is used under the assumption that negative externalities are capitalized into house prices or values. It is found that noise pollution has a statistically significant negative effect on the value of a house. Houses located near a road where noise is very loud would sell with a total discount of 30 percent. Another paper that found this relationship was written by Blanco and Flindell (2011). Their paper reveals the effects of similarly high road traffic noise on property values in very different residential areas. Their results show significant differences in the effect between the three different areas that are investigated. This indicates that it is interesting to investigate the effect in different areas.
For this research, papers focusing on airport noise are more important. One of the earlier studies that focused on the effect of aircraft noise on house prices was a paper written by Pennington, Topham and Ward (1990). The enormous growth in the volume of air traffic has brought an increasing concern about the level of aircraft noise. In their paper, they use house mortgage data and noise contour maps for Manchester International Airport. They find that properties in the parts of town most affected by noise from the Manchester International Airport have a market value that is on average 6 percent lower. Cohen and Coughlin (2008) find an even stronger relationship. They compare various spatial econometric models and estimation methods in a hedonic price framework to examine the impact of noise on 2003 housing prices near Atlanta airport. Houses located in the area in which noise disrupts normal activities sell for 20.8 percent less than houses that are in areas where noise does not disrupt normal activities. Espey and Lopez (2000) have investigated the same effect, but in monetary terms. They use a hedonic price method to estimate the relationship between residential property values and airport noise and the proximity to the airport in the Reno Sparks area in the United States. Their empirical results show that there is a statistically significant effect. There is a negative relationship between airport noise and residential property values. They find that a house in an area where there are noise levels of more than 65 dB sell for 2400 dollars less than equivalent homes in less noisy areas. From these papers, it can be concluded that aircraft noise has a strong and significant effect on house prices in affected areas.
This kind of research has also been applied to the Dutch market. Around 2004, large infrastructural projects like expansion of Schiphol Airport and construction of railways were
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 14
planned. This led to a social debate about the noise nuisance in the Netherlands. Theebe (2004) estimated the non-linear impact of traffic noise on property prices in the western part of the Netherlands that includes the provinces of North-Holland, South-Holland and Utrecht. The paper used a very large dataset with over 100,000 sales transactions. The data included individual property characteristics and noise levels for two million small areas. Spatial autocorrelation techniques are used and it was found that the impact of traffic noise averages around 5 percent with a maximum of 12 percent.
Apart from this, it would be interesting to know whether this effect changes when the amount of noise decreases. Dekkers and van der Straaten (2009) also examined the effect of noise in the area surrounding Schiphol Airport. The paper developed a spatially-explicit hedonic pricing model for house prices to quantify the social cost of aircraft noise disturbance in monetary terms. Background noise is accounted for, but the focus is on aircraft noise around Amsterdam airport. They find that higher noise levels lead to lower house prices and that air traffic has the largest price impact. They also find that a marginal benefit of 1459 euro per house for a 1 dB noise reduction, leading to total benefit of 574 million euros of 1 dB noise reduction.
From all these studies, it can be concluded that aircraft noise has a negative effect on house prices and that this effect is different for different levels of dB’s. The effect around Schiphol Airport has already been studied, but no one has ever focused on the smaller airports in the Netherlands that are in parts of the Netherlands that are not as populated as the area around Amsterdam. Investigating whether houses in those areas suffer as much from the airport noise as houses in the area surrounding Schiphol Airport and whether an increase in operating activity of those airports will lead to an even bigger reduction in house prices would be interesting. However, there is no data on increases in noise levels around Lelystad Airport yet, as the idea for expansion still has to be implemented. A study that investigates whether the announcement already influenced the house prices in the area is possible and would be of added value in the debate around Lelystad Airport.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 15 2.4 Flight Routes
On January 1st 1993, Schiphol Airport acquired Lelystad Airport. With the idea that originated
in 1991 to move flights from Schiphol to Lelystad, the first plans to expand Lelystad Airport were developed in 1994 (Atlas Regionale Luchthavens, 2007). These plans were mainly about turning Lelystad Airport into a business airport. The municipality of Lelystad was in favour of these plans and even created a masterplan ‘Versnelde Groei’ in 1996. In this masterplan, it is indicated that the expansion of the airport would have a positive impact on the development of the city of Lelystad. The expansion is allowed as long as it does not harm the liveability of the city (NACO, 2007). Noise disturbance is a key factor in accessing the liveability of a location. Schiphol traffic is flying at a height of approximately 7000 feet or 2100 meters above Lelystad and descend to about 4000 feet or 1200 meters when flying over urban areas before they start the landing (Gemeente Lelystad, 2007). This means that airplanes departing from Lelystad Airport have to take this into account. The planes cannot ascend directly but have to stay below a certain height. This causes a lot of noise and has therefore led to a discussion. Several route schemes have been developed to create the best flight routes that spare both nature and built-up areas. The very first routes of 2004 did not take this into consideration and would have caused a lot of noise pollution in urban areas, so these routes have been changed and were presented by the LVNL in December 2006 (Atlas Regionale Luchthavens, 2007). However, the Ministry of Infrastructure and Water could not provide any information on this and there are no publicly available maps of these first two flight schemes, so the effect of these routes on house prices cannot be investigated. Again, there was critique on the presented routes. 30 percent of the participated responses included proposals for routes. This indicates that the flight paths are of great importance to the involved. Based on these participated responses, the LVNL proposed nine new routes on November 16th 2007 (Adecs Airinfra, 2008). This resulted in the following flight path:
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 16
Figure 1: Old route
In the surroundings of the airport, there are multiple areas with protected nature called Natura 2000-areas. These include nature reserves like the Oostvaardersplassen, Lepelaarplassen, Markermeer & IJmeer, Ijsselmeer, Ketelmeer & Vossemeer, Veluewerandmeren, Eemmeer & Gooimeer, Arkemheen and the Veluwe (Lensink, & Smits, 2009). In ‘Effecten van de MER-alternatieven Lelystad Airport in relatie tot groene wet-en regelgeving’ it is investigated whether these nature reserves are preserved when expanding the airport. A cost-benefit analyses shows a positive outcome, even though there is disturbance in the reserves. After multiple changes to the flight routes this seemed to be the best alternative.
On the 5th of February 2009, the minister of Transport and Water gave Mr. J.G.M. Alders the order to investigate how Lelystad Airport could grow and what effects this would have on the economy, nature and people living near the airport. Mr. Alders initiated the Alderstafel to advice the parliament about the expansion of Schiphol Airport in coherence with Eindhoven Airport and Lelystad Airport. On March 30th 2009 the first Alderstafel Lelystad took place. The result of this advisory body was a widely supported advice in March 2012 regarding the expansion of Lelystad Airport. This advice was taken very seriously and the LVNL and CLSK came up with five new route alternatives that were to be investigated in the MER. Route B and route B+ are operationally feasible and they both meet all required conditions. Route B+ is the preferred route, as noise disturbance will be lowest and negative
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 17
effects are kept to the minimal (Adecs Airinfra & To70, 2014). This is achieved by making sure that as little residential areas and nature reserves are impacted by the flight routes. Below the B+ route is shown:
Figure 2: New route, Baan 23 – start and landing in southwestern direction
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 18
Figure 4 – Map of entire new flight route
This route optimization process has been set in motion to create the best flight routes that spare both nature and built-up areas. As explained before, the idea is to keep the costs of the expansion and the effects on the economy, nature and people living near the airport as minimal as possible. It is expected that the effect of anticipated airport noise on house prices will therefore be smaller than what is found in previous investigations that did not take this into account.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 19 2.5 Hypotheses
The research question will be tested by using data on house sales transactions in the area around Lelystad Airport. Considering the related literature from the previous sections, the expansion of Lelystad Airport will most likely influence house prices. The effect is expected to be negative. Noise disturbance is likely to decrease house prices, as people want to be compensated for living in an area where there is a lot of noise. This leads to the first hypothesis: H1: The announcement of the expansion of Lelystad Airport, and thus the effect of anticipated aircraft noise, will decrease house prices in the area surrounding the airport.
The effect of noise disturbance on house prices has already been investigated by multiple different researchers, as explained in the previous sections. Most of them do indeed find a negative and significant effect of aircraft noise on house prices. For example, Espey and Lopez (2000) find that there is a significant and negative relationship between airport noise and residential property values. However, this research focusses on an announcement effect and has been investigated by Jud and Winkled (2006). They find a negative and significant effect of 9.2 percent. Knowing this, the announcement of the expansion of Lelystad Airport is expected to already lead to a reduction in house prices.
In the case of Lelystad Airport, it is interesting to investigate whether the changes in the flight routes have influenced house prices of houses under these flight routes. The effect of the shift from the first route to the second route will be investigated. As is expected that the announcement of a flight route will decrease house prices, it is expected that the first route has resulted in a reduction of the house prices under this route. With the announcement of the second route, this effect is likely to be offset. When the routes change, different areas will be affected by the noise. People living under the first flight route will no longer have to deal with noise pollution, while people living under the second flight route will start to be affected by noise pollution. It is therefore expected that the second route will increase house prices of houses under the first route and decrease house prices on houses under the second route. Therefore, the second hypothesis is:
H2: The effect of anticipated aircraft noise on house prices will be negative for houses below the first flight route after the first announcement and for houses below the second flight route after the second announcement. The second announcement will offset the negative effect of the first announcement on houses below the first flight route.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 20
Furthermore, it is assumed that higher levels of expected noise will have a stronger negative effect on house prices. The higher the level of noise, the more disruption. When there is more disruption, it is a less attractive place to live and this will be incorporated into prices. This effect is also expected to be visible already after the announcement. Dekkers and van der Straaten (2009) have examined the effect of noise in the area surrounding Schiphol Airport and they find that higher noise levels lead to lower house prices. A 1 dB reduction in noise results in a benefit of 1459 euro per house. Noise levels reduce as planes fly higher, so the third hypothesis is:
H3: The anticipated effect of aircraft noise on house prices will diminish as the altitude of planes increases.
With these three hypotheses, the effect of the expansion of Lelystad Airport is going to be investigated and this will be a great addition to the existing literature regarding noise disruption and announcement effects.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 21 3. Methodology
This section contains the methodology that is used to answer the research question of this paper. Besides, the section gives information on the dependent variable and control variables in the regressions.
3.1 Method
To test the hypotheses, a difference-in-difference model is used as this will be the best model to test the effect of an announcement. The model is inspired by the model used by Carroll et al. (1996), with whichthey investigate the economic impact of a transient hazard on property values. They employ several housing characteristics and a time trend. Furthermore, Jud and Winkled (2006) examine the effect of a FedEx announcement regarding the expansion of the accommodation of an airport hub on the value of surrounding residential property and they use a similar kind of hedonic model. The following regression is a combination of the two models and will be run to test the first hypothesis:
(1) ln P𝑖𝑡 = 𝛼Airport𝑖 + 𝛿Post𝑡 + 𝛾Airport*Post𝑖𝑡 + 𝛽X𝑖 + 𝜈𝑛 + 𝜃𝑡 + 𝜖𝑖
For this difference-in-difference model, the treatment group will contain houses that are located under the flight route and standard errors will be clustered at the (6-digit) zip code level to correct for correlation between zip codes. Three different regressions will be run to see how the effect changes as the model increases in fit. The dependent variable is the natural logarithm of the transaction price. The ‘Airport’ variable is an indicator of whether a house is located under the flight route or not. This dummy variable equals 1 if the house is underneath the flight route. ‘Post’ is a dummy variable that equals 1 if the observation is from after the announcement and 0 otherwise. The variable of interest is ‘𝛾’, which measures the effect of the announcement of the airport expansion on the house prices under the flight route. When the sign of this variable is negative and the output is significant, the transaction price of houses sold after the announcement is lower for houses under the flight route than for houses that are located somewhere else. This is the relationship expected to be found. As house prices depend on a lot of different aspects, a second regression will be run in which quite a lot of control variables are added. These explanatory variables are included in the model under ‘X’. These control variables will be: surface area of the house, surface area of the parcel, type of house, period of construction, the state of the outside of the house, the state of the inside of the house, the state of the garden and whether the house is a monumental building or not. Besides, it is important to account for variables that cannot be measured or included in the regression as
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 22
control variables. This is why a third regression will be run that includes zip code fixed effects (𝜈) and time fixed effects (𝜃). In this model, the ‘Airport’ and ‘Post’ variables will be excluded from the regression to get rid of perfect multicollinearity between these variables and the fixed effects.
The first hypothesis can be tested by using the model above. For the second hypothesis, extra variables have to be added. It will not be possible to measure the effect of the shift in flight routes with the model above, as with this model it is only possible to measure a single announcement. In order to measure the shift, three different interaction variables are needed. The effect of the first announcement is measured with all variables that have a ’1’ added to the name. The variables with a ‘2’ in the name will measure the effect of the second announcement. Apart from this, the effect of the second announcement on the houses of the first flight route has to be investigated as well, as these houses will no longer be under a flight route. This effect is measured by the interaction term of ‘Airport1’ and ‘Post2’. Again two different models will be run, one with and one without the fixed effects, to see how the effects change as the model gets better predictive power. The model including fixed effects will be run without the ‘Airport’ and ‘Post’ variables.
(2) ln P𝑖𝑡 = 𝛼1Airport1𝑖 + 𝛿1Post1𝑡 + 𝛾1Airport1*Post1𝑖𝑡 + 𝛼2Airport2𝑖 + 𝛿2Post2𝑡 +
𝛾2Airport2*Post2𝑖𝑡 + 𝛾3Airport1*Post2𝑖𝑡 + 𝛽X𝑖 + 𝜈𝑛 + 𝜃𝑡 + 𝜖𝑖
To test the third hypothesis, it is needed to zoom in on the second announcement. A completely different model is needed. This model should include dummy variables for different flight heights. As can be seen in Figure 4, there are three main stages in the process of ascending and descending. A plane departing from Lelystad Airport has to stay below an altitude of 1800 meters when leaving the Flevoland province. This is mainly because the plane would hinder air traffic to and from Schiphol Airport. In the second stage, planes still have to remain low, but they can climb up until 2700 meters. Planes must stay at this altitude for the larger part of their journey over the Netherlands. The final phase is where the planes can really start climbing higher. This phase starts from a height of 2700 meters. Dummy variables for these three stages are included in a model. The dummy variable ‘1800’ equals 1 if the plane flies over the house at a height of 1800 meters or lower. The other two dummy variables will equal ‘0’ in this case. With the inclusion of the dummy variables it can be investigated if the effect of the announcement of the expansion of Lelystad Airport is different for different altitudes. Normally, the noise disruption decreases as planes fly higher. Therefore, it is expected that the
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 23
effect on house prices will decrease as altitude rises. As in the case of Lelystad Airport, it is still only an announcement, saying that the real disturbance has not yet been felt. It is interesting to investigate whether this is already incorporated into prices. This will be tested by including interaction terms for the three different stages of flight height and this results in the following model:
(3) ln P𝑖𝑡 = 𝛼Airport2𝑖 + 𝛿Post2𝑡 + 𝛾1Airport2*Post2𝑖𝑡*1800 + 𝛾2Airport2*Post2𝑖𝑡
*1800-2700 + 𝛾3Airport2*Post2𝑖𝑡*2700 + 𝛽X𝑖 + 𝜈𝑛 + 𝜃𝑡 + 𝜖𝑖
For this model the two different versions will be run as well; one with and one without the fixed effects. The fixed effects version will be run without the ‘Airport’ and ‘Post’ variables.
3.2 Variables
The aim of this research is to investigate the effect of anticipated aircraft noise on house prices. House price is the dependent variable. For this variable, the natural logarithm of the transaction price is taken and in the next section it is explained why this is decided. To examine the relationship between anticipated aircraft noise and house prices, it is important to control for unobserved heterogeneity. As explained in the literature review, a hedonic model price is based on all internal and external characteristics of the property. The qualitative differences between the properties that transact in different periods of time are controlled for by including property-specific control variables that describe the property characteristics that affect any given property’s value relative to other properties at any point in time (Geltner et al., 2001). By including these property-specific control variables, unobserved heterogeneity is controlled for. The next section will give more information about the dependent variable and the section after that explains which control variables are included in the regressions and why.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 24
3.2.1 Dependent Variable
The dependent variable in this hedonic model is the natural logarithm of the transaction price. The main reason for including transaction price as a logarithm is because it will result in reporting percentage price changes rather than absolute price changes (Stock and Watson, 2015). This makes interpreting the regression results easier. Apart from this, the relative instead of the absolute standard deviation is minimized and the residuals of logarithmic variables are usually closer to normality when compared to absolute variables. Outliers will in this case be less severe (Stock and Watson, 2015). This is clearly visible in the graphs shown below. The distribution of the ln version of the transaction price is closer to normality than the distribution of the transaction price itself.
Figure 5 – Transformation of transaction price
3.2.2 Variables of Interest
To be able to measure the effect of anticipated aircraft noise on house prices, an interaction term must be created. This is the interaction term between a dummy variable for time and a dummy variable for place. The dummy variable for time should indicate whether the transaction is from after the announcement and the dummy variable for place should indicate whether the transaction regards a house that is below a flight route.
Based on the correct postal codes, shown in Table 1 (will be explained in more detail in the ‘Data and Descriptive Statistics’ section), the ‘Airport’ variables are generated. The variable ‘Airport’ equals 1 for the four municipalities that are always close to the flight routes: Lelystad, Almere, Zeewolde and Dronten. The ‘Airport1’ variable is 1 for all municipalities under the old route and the ‘Airport2’ variable is 1 for all municipalities under the new route. The next dummy variable that must be created is the dummy variables that indicates whether the transaction is from before or after one of the announcements. Again, three different dummy
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 25
variables are created. The first dummy variable is ‘Post’ and indicates if the transaction has taken place after the first announcement. As the first announcement was in November 2007, this variable is equal to 1 when the transaction took place in February 2008 or later. There is always a delay is responsiveness between an announcement and the actual implementation. Jud and Winkler (2006) use a lag of approximately nine months between the announcement and the actual effect of the announcement. This period seems too long for this research, as the time between the two announcements is relatively short with just over four years. There has been a social debate regarding the expansion since the start of the first announcement. There already was a lot of critique on the very first flight route that was published in 2004; 30 percent of the participated responses included proposals for routes. This shows that people are aware of the effects of the expansion and it is expected that people started behaving accordingly soon after. With a longer lag, the effect of the first announcement will otherwise be incorporated in the effect of the second announcement. It is therefore more reasonable to use a smaller time lag than Jud and Winkler (2006) did. According to Katz (1974) a 100 percent price adjustment to rating reclassifications takes place six to ten weeks after a reclassification and the lag in this research is therefore shortened to three months. To check whether this assumption holds, robustness checks are done in which different lags between the announcement and the actual effect and the results are reported in Section 6. The variable that measures the effect of just the first announcement is ‘Post1’ and this variable is equal to 1 between February 2008 and May 2012. The ‘Post2’ variable is equal to 1 when the transaction took place after the second announcement, which was in March 2012. With a three-month lag, this means that the variable is 1 for transactions in and after June 2012. The variable of interest is the interaction term between the correct ‘Post’ variable and the correct ‘Airport’ variable, with which the effect of anticipated aircraft noise can be measured.
After generating the ‘Post’ and ‘Airport’ variables, the dummy variable for flight height is generated. Three different dummy variables are created; one for municipalities where planes fly over at an altitude of at most 1800 meters, one for municipalities where planes fly over at an altitude of between 1800 and 2700 meters and the last one for municipalities where planes fly over at an altitude of 2700 meters and higher (with a maximum of 5500 meters). These dummy variables are then again interacted with the interaction term of ‘Post2’ and ‘Airport2’ to measure the effect of altitude differences.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 26
3.2.3 Control Variables
Determinants of the transaction price of a house are both internal and external characteristics of the property. The most important characteristics are included in the regression as control variables. The control variables that are included in this study are surface area of the house, surface area of the parcel, type of house, period of construction, the state of the outside of the house, the state of the inside of the house, the state of the garden and whether the house is a monumental building or not. These variables will be discussed shortly. More information on these variables can be found in Appendix 2.
Surface Area of the House
The most important determinant of the price of a house is the size of the house. The bigger the house, the more money people are willing to pay for it. First, the variable ‘woon’ for the surface area of the house is winsorized to reduce the impact of outliers. After that, the variable is transformed to a logarithmic variable (ln_size), as this will make the distribution closer to normality.
Surface Area of the Parcel
Another important aspect of the price of a house is the area of the total parcel. This is the area on which the house is built and includes the area of the garden as well. A garden will likely increase the value of a house and the bigger the garden, the bigger this increase is expected to be. Therefore, the surface area of the parcel is included in the regression. The variable is transformed in the same way as the variable for the surface area of the house and included in the model as a logarithmic variable.
Type of House
The house type indicates what kind of house is looked at. Appendix 2 shows an overview of the different house types included in the dataset. In general, the value of an apartment differs from that of a family house. It is therefore important to control for house type in regression that use the transaction price of a house.
Period of Construction
Period of construction is a very important characteristic of a house. The older the house, the more likely the house suffers from physical deterioration and functional and external obsolescence. This is the reason why the year of construction should be included in the regression. As the dataset includes houses that have already been built around the year 1500, dummy variables for period of construction are generated. This results in total of 9 different periods of construction. Appendix 2 gives an overview of the division of the different periods of construction.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 27
State of the House and Garden
Just as the house may suffer from deterioration and obsolescence, the house may as well suffer from bad maintenance. Outdoor maintenance, indoor maintenance and maintenance of the garden are included in the regression to control for the different states of maintenance. These variables are dummy variables to indicate the current state. In Appendix 2, the division of these variables is shown. For outdoor maintenance (onbu) and indoor maintenance (onbi), a dummy variable is created that equals 1 if the level of maintenance is 7 or higher. For the state of the garden, the dummy variable equals 1 is the level of maintenance is 2 or higher.
Monumental Building
The final control variable is an indicator of whether a house is a monumental building or not. This variable is included in the regression as the price of a monumental house is expected to be higher than the price of a normal house.
3.2.4 Fixed Effects
Apart from the control variables, the model also incorporates fixed effects. Both zip code fixed effects and time fixed effects are added to the model. These fixed effects are included in the model to account for variables that cannot be measured or included in the regression as control variables. A fixed effects model treats unobserved differences between a certain observation as fixed parameters (Hedges, 1994). The zip code fixed effects are based on 6-digits. By including zip code fixed effects into the model, there is an automatic correction for the attractiveness of a location. A house on a street with some shops might be more attractive than a house on a street with a lot of traffic. Furthermore, including time fixed effects corrects for the economic situation of the period.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 28 4. Data and Descriptive Statistics
This section discusses the data used to find an answer to the main question of this research. In the first part, it is explained where the data is gathered from. The next part will give more insight in why this data is relevant for this research and how the data is put together. The latter part shows some descriptive statistics of the dataset.
To be able to measure an external effect like expected noise pollution on house prices, housing transactions and housing characteristics are needed. This kind of data is not usually publicly available and needs to be made available explicitly for researches. The data for this research is gathered from the Dutch Association of Realtors (NVM). This database can be used for data regarding house prices and housing characteristics. The data is made available by municipality and for this research 27 different municipalities have been made available for the period from 2004 up until 2017. This is quite a large dataset and it gives a total of 167.356 transactions before dropping outliers.
As stated in the previous paragraph, the dataset includes 27 municipalities. These municipalities are not necessarily located close to Lelystad Airport, because there is expected to be much noise pollution underneath the flight paths where planes have to stay low. As can be seen in Figure 4, the planes have not reached 1800 meters of height when they leave the Flevoland province. This is the area where noise levels reach the highest levels of dB’s, so all the municipalities in Flevoland have been added to the dataset. Furthermore, it is interesting to measure the effect along the route, but at different flight heights. Therefore, different municipalities in the 1800 to 2700 meters phase are included in the dataset as well. Besides, the dataset also includes municipalities where planes fly over at a height of more than 2700 meters. Apart from the municipalities suffering from the second flight route, municipalities that were originally underneath the flight route have been included as well. In this way, it is possible to measure whether the change in flight routes led to a shift in the effect of noise pollution on house prices. However, there is no information on flight heights for this first flight route, so it is not possible to make a distinction based on flight heights for the first route. Besides, a control group is included to be able to make the best comparison. When only looking at municipalities close to the flight routes, there will always be noise pollution. By including municipalities that have never been close to a flight route, it is possible to measure the real effect. Below is the list of municipalities included in the dataset. A list of zip codes that are included in the regressions can be found in Appendix 1.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 29
Table 1 – Municipalities included in the dataset
Route Height Municipality Code
Old - Ermelo 0233
Putten 0273
Nunspeet 0302
Nijkerk 0267
Leusden 0327
Old & New <1800 Lelystad 0995
Almere 0034 Dronten 0303 Zeewolde 0050 New <1800 Noordoostpolder 0171 Kampen 0166 Zwolle 0193 New 1800-2700 Raalte 0177 Heerde 0246 Deventer 0150 Apeldoorn 0200 Renkum 0274 New >2700 Haaksbergen 0158 Geldermalsen 0236 Zevenaar 0299 Brummen 0213 Zaltbommel 0297 Control - Hardenberg 0160 Staphorst 0180 Arnhem 0202 Barneveld 0203 Almelo 0141
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 30
The dataset contains a total of 167.356 observations without winsorizing variables and removing observations that miss input for a certain variable. After correcting for missing variables and outliers, 131,171 observations remain.
The descriptive statistics in Table 2 shows all variables that are included in the regressions and their mean, median, standard deviation, min and max. The indication of the numbers for the dummy variables can be found in Appendix 2. The independent variable is the natural logarithm of the transaction price. The mean transaction price of houses in the dataset is €246,148. The median transaction price is equal to €213,000. This number is lower, as some of the houses in the dataset are ranches that have very high transaction prices. The dataset covers transactions from the period 2004 until 2017. Most houses were constructed between 1991 and 2000, but both the mean and the median of the period of construction is around period (6), which is between 1971 and 1980. The mean surface area of houses in the dataset is almost 131 square meters. The median surface area is 123 square meters, with a minimum of 60 and a maximum of 400 square meters. This is after the variable has been winsorized to get rid of extreme outliers. This has also been applied to the variable ‘Surface Area Parcel’. The mean parcel area is approximately 360, while the median is only 197. The difference between these to number is driven by the small portion of houses in the dataset with extremely big parcels. The maximum parcel area is 4525 and the minimum 72. The median value for house type is a single-family home. This is also the house type of most of the observations in the dataset. The dataset does not contain any of the big cities in the Netherlands. Most of the municipalities used in this research are in the central part of the Netherlands where there is relatively much space. This is the reason why not that many observations are apartments and/or flats. All different types of apartments (21-27), see Appendix 2, are combined in one dummy variable. The average and median maintenance of the garden, the outside of the house and the inside of the house is normal/good. All these conclusions can be found in the table below.
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 31
Table 2 – Descriptive Statistics
Variable Obs Mean Median Std. Dev Min Max
Transaction price 131,171 246148.3 213000 119992.4 21406 1965000 Ln(p) 131,171 12.32553 12.26905 .4003733 9.971427 14.491 Year of Transaction 131,171 2010.464 2010 4.18991 2004 2017 Period of Construction 131,171 6.026363 6 2.18991 1 9 (1) 1500-1905 4,015 (2) 1906-1930 10,080 (3) 1931-1944 8,087 (4) 1945-1959 7,698 (5) 1960-1970 14,148 (6) 1971-1980 23,079 (7) 1981-1990 22,930 (8) 1991-2000 28,674 (9) > 2001 12,460
Surface Area House 131,171 130.8126 123 40.91921 60 400 Ln (House Size) 131,171 4.834437 4.812184 .2705829 4.094345 5.991465 Surface Area Parcel 131,171 360.5637 197 585.6418 72 4525 Ln (Parcel Size) 131,171 5.474786 5.283204 .7371655 4.276666 8.417373 Type of house 131,171 5.602931 5 2.066688 2 27 (2) Simple 3,943 (5) Single-family Home 103,966 (6) Canal House 93 (7) Mansion 9,454 (8) Residential Farm 1,726 (9) Bungalow 4,804 (10) Villa 5,750 (11&12) Estates 596 (21-27) Apartments 839 Garden Maintenance 131,171 3.621616 3 .8917838 1 5 Inside Maintenance 131,171 6.995731 7 1.031987 1 9 Outside Maintenance 131,171 7.027209 7 1.031987 1 9 Monumental 131,171 .0053061 0 .0726494 0 1
Master thesis Kristie Lek – The effect of the expansion of Lelystad Airport on house prices 32 5. Results
In this section the results of the impact of the expansion of Lelystad Airport on house prices in the affected areas are presented. The economic impact of the models previously discussed will be interpreted. The first section discusses the impact of anticipated aircraft noise on house prices in the Flevoland province. The second section contains the results of the effect of the different announcements and the third section the results for the different altitude heights.
5.1 Effect of expanding Lelystad Airport
Table 3 presents the estimated coefficients of the first regression model. Regression (1) in the first column is a simple model that only estimates the effect of the expansion of Lelystad Airport on house prices, without correcting for important aspects like house characteristics, house types, time and location. The coefficient of the interaction term for houses sold after the initial announcement located in the Flevoland province is -0.0504. This coefficient is significant at the 1 percent level. This result indicates that houses in the Flevoland province that were sold after February 2008 had transaction prices that were 5.04 percent lower than houses sold in different parts of the country or before the announcement. This is the effect that was expected. However, the model does not have a good fit, as it has a very low R-squared (8 percent). In the second column, the regression results of a regression with control variables for house type and house characteristics are presented. Again, the coefficient of the interaction term is highly significant, but the impact has decreased to -1.54 percent, suggesting that part of the effect is due to the quality of houses being sold. This model has a much better fit with an R-squared of 71.3 percent. The difference between these two models shows that correcting for house specific characteristics is very important in estimating house prices. The third column shows the regression results of the fixed-effects model. In this model, both zip code fixed effects and time fixed effects are included. This increases the overall fit of the model, as the R-squared is now 85.4 percent. The coefficient for the interaction term remains significant at the 1 percent level and the impact of anticipated aircraft noise on houses in the Flevoland province is -0.89 percent. This indicates that houses are sold for almost one percent less after the announcement of the expansion of the airport. With an average transaction price of €246,148, houses are sold at a discount of €2188. The effect is not extremely big, but it is significant. Houses are indeed sold at a discount because of anticipated aircraft noise. This result is in line with the first hypothesis.
