0
Universiteit van
Amsterdam
The influence of energy labels on house prices in
Dutch VINEX-districts
Jeroen Hesp 10538941
Supervisor: Drs. P.V. Trietsch, M.Phil.
BSc Economics and Business EC: 12 Finance & Organization track 16-02-2018
1
Statement of Originality
This document is written by Jeroen Hesp 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.
2
Abstract
In the global carbon emissions, the role of real estate is substantial. Energy efficiency in houses is important and rising numbers of properties receive better energy label ratings. In this paper the influence of energy labels on house prices is discussed. Past research mostly focused on large metropolitan areas. This paper will check if the energy label premium is also present in Dutch VINEX-districts. The empirical research shows that every upgrade from “C” labeled houses upwards, leads to a 1.2% higher square meter price of houses in VINEX-districts. Furthermore, when looking
3
Contents
1. Introduction ... 4 1.1 The problem ... 4 1.2 Research question ... 4 1.3 Literature ... 41.4 Factors that drive house prices ... 5
1.5 Data and methodology ... 5
1.6 Thesis structure ... 6
2. Literature ... 7
2.1 EPC and underlying energy efficient variables ... 7
2.2 Variables that drive house prices ... 7
2.3 Literature approving EPC premium ... 8
2.4 Literature nuancing the EPC premium ... 9
2.5 Data and research method used ... 10
3. Data and Methodology ... 12
3.1 Sample selection and data sources ... 12
3.2 Conceptual framework and hypotheses ... 14
3.3 Regression model and variables ... 15
4. Results ... 17
4.1 Descriptive statistics ... 17
4.2 Parametric tests ... 17
4.3 Results of regression analysis ... 18
4.4 Hypothesis testing ... 20
5. Conclusion ... 22
6. Limitations and future research ... 23
References ... 24
Appendix 1 - Examples of EPC labels ... 25
4
1. Introduction
Energy labels have proven to be successful in the white goods and car market, which increased the sale of products that are better for the environment (Jensen, Hansen and Kragh, 2016). The
Netherlands introduced energy performance certificates (EPC’s) in January 2008, one year before the official introduction in the European Union. For some houses, the energy costs are around half of the total monthly expenses of the house. If the price of energy becomes more volatile, this makes research on the effect of EPC’s on consumer benefit and the environment an interesting field of research.
1.1
The problem
The research by Brounen and Kok (2011) in the Netherlands has focused on large urban areas, specifically Amsterdam. However, the housing market of large urban areas like Amsterdam already show signs of overheating and the supply of houses is extremely divers, ranging from houses built in the 1600s to those built in recent years. Both these issues disturb the possible predictive effect of models of the effect of energy labels. Because of construction standards and a more uniform supply of houses, planned residential areas could show a more accurate sign of a possible EPC premium. Fesselmeyer (2017) also researched a more uniform area, namely high rise developments in
Singapore, finding a premium of 3 percent for green certification. Samples like this decrease the need to adjust for many energy efficiency, environmental and building characteristics (see table of chapter 2). In this paper a sample of 5,380 VINEX-district houses was used to research the possible EPC premium.
1.2
Research question
The aim of this research is to determine if a price premium exists in the transaction price of houses in planned residential areas. Hereby isolating the effect of the EPC label from other determinants of the transaction price like the location, period features and amenities. The research question formulated for this is:
What is the effect of upgrades in Energy Performance Certificates (EPC’s) on the transaction price of houses in planned residential areas (VINEX-districts) in the Netherlands?
The following sub questions will be discussed in the literature review; What are EPC’s and energy efficient housing features? Which factors drive house prices in general?
What articles support an EPC premium? What articles nuance an EPC premium?
Which data and research method is used in the literature?
1.3
Literature
The article by Brounen and Kok in 2011 is one of the first to empirically explain the implementation of EPC’s under a big scale certification program in Europe. Brounen and Kok find a premium of 3.6 percent in selling prices of “green” labeled houses compared to those were labeled “non-green”. “Green” energy label meaning an EPC rating of either A, B or C. “Non-green” energy labels are D or lower.
In the paper by Deng, Li and Quigley (2011) the impact of the introduction of the “Green Mark” programme in Singapore on transaction prices of houses was researched. A significant effect of 4 percent was found as a premium on housing prices.
Amecke (2012) mentions some indicators that measure positive influence of EPC’s on purchasing decisions. The research setup used contained a survey among 1,239 potential purchasers and tenants
5 in Germany. Amecke found that awareness, understanding, application and trustworthiness of the EPC have impact on the adoption and transaction prices. This paper, together with Murphy (2013), highlights a different view on EPC’s by doing a survey instead of log-level regression, as in most academic papers in this field of research.
Two Swedish papers that criticize the price influence of EPC’s are Hårsman, Daghbashyan and Chaudhary (2016) and Wahlström (2016). Hårsman et al. show the price influence of EPC’s is partly related to information that potential house buyers can observe during the sale of the property. Even though a small significant EPC premium is found, Hårsman et al. think that the observable energy efficient housing features have more impact on the premium than the information purely provided by the EPC label.
The results found by Wahlström (2016) do not find a price premium for energy efficient housing. Although, Wahlström finds that energy consumption has a positive impact on house prices. An explanation for these contradictory results might be that potential buyers base their expectation about future energy consumption on the energy efficient housing features, rather than the energy consumption of the previous property owner. Energy efficient housing features do give significant premiums on the transaction price.
1.4
Factors that drive house prices
In the paper by Jensen, Hansen and Kragh (2016) many factors that drive house prices are discussed. Property size, garden, view, location, terrace and energy efficient housing features all affect the transaction price of a house. In table 1 in chapter 2 more variables that drive house prices are mentioned. In particular EPC’s seem increasingly to be developing as an unique selling point. Jensen et al. (2016) found this was a consequence of the Danish introduction to oblige for public display of EPC’s in 2010.
Abelson (1978) researched 1,400 properties in Sydney from two different municipalities.
In general, the transaction price is influenced by the size and quality of the property, accessibility to work and recreation and environmental factors. The paper splits these components into 30
independent variables to predict the effect on transaction prices, all of which appeared significant at the 95 percent level. Two different municipalities in a metropolis like Sydney can quite diverge in energy efficient, building and environmental characteristics. As stated in paragraph 1.1, it would be interesting to check whether we can explain with fewer independent variables a possible EPC premium in VINEX-districts.
1.5
Data and methodology
The dataset contains 67,896 transactions from the Nederlandse Vereniging van Makelaars (NVM) and a list of all registered EPC’s (18,716) in the Netherlands from the Rijksdienst voor Ondernemend Nederland (RVO). 5,380 transactions were matched with an EPC label within 50 selected VINEX-districts. Instead of using a more heterogeneous dataset compared to Brounen and Kok (2011), a more homogeneous dataset is used, making it more efficient to determine the effect of EPC’s. This paper uses three different log-level time series regressions for the period 2008-2016, comparable to the method in Brounen and Kok (2011). The regression models explore two different EPC premiums. Firstly, the premium of different levels of EPC ratings. Secondly, the premium when the EPC score is higher than the average EPC score of the street on which a house is situated. This study contributes to the existing literature by adding new independent variables, while researching a more uniform sample of houses.
6
1.6
Thesis structure
Firstly, the existing literature on energy efficiency premiums is discussed. Secondly, the dataset and methodology will be described. Thirdly, the results of the time series regression are presented; while checking for various empirical terms as multicollinearity, heteroscedasticity and significance. Lastly, the conclusion and advice for further research are presented.
7
2. Literature
As EPCs are relatively new, a limited number of articles have been published on the subject. In this comprehensive literature review the main drivers of transaction prices for houses and the expected effects of EPC’s are presented.
2.1
EPC and underlying energy efficient variables
Due to the increased attention regarding global warming, there have been some increases in the energy efficiency of the built environment. In the beginning of 2013, Energy Performance of Buildings Directive (EPBD) was setup by the EU, as a main legislation to increase energy efficiency in buildings (Brounen and Kok, 2011). This directive introduced the energy performance certificate (EPC) for residential houses and utility buildings across Europe. The introduction of the EPC is to increase transparency of energy consumption across the real estate sector. The Dutch government has been at the forefront of this development by preceding the European directive by 7 years.
Brounen and Kok (2011) found that EPC’s create benefits for investors as well as tenants, since the energy savings from better energy efficient buildings result in less operating costs and higher house prices. As predictors for this, some papers include energy efficient housing features as variables in their empirical research. Central heating and insulation qualities are variables used by Brounen and Kok (2011). Hårsman et al. (2016) separates heating into electricity heating, district heating and sourced pump heating. Furthermore, Hårsman adds ventilation systems to his regression. Jensen et al. (2016) separates insulation quality into variables for the wall and roof type.
Hårsman, Daghbashyan and Chaudhary (2016) find that the premium on green-labeled houses is caused by the observable underlying energy efficient housing features rather than the EPC label itself. Additionally, Wahlström (2016) finds no price premium related to energy consumption, but does find a premium for energy efficient housing features. Wahlström suggests this is because potential house buyers observe the present energy efficient housing features, not the utility bill amount of the previous owner. These results indicate that the EPC label could serve as a proxy for the perceived benefits in energy performance of the house.
2.2
Variables that drive house prices
Abelson (1978) researched 1,400 properties in Sydney. The paper identified 30 independent variables to predict the effect on transaction prices. Significant examples of the independent variables are neighborhood live ability, noise from aircrafts or trains, view, distance to sea, accessibility to recreation and traffic jam likelihood. The biggest factors that explain house price variation are size, inflation and environmental factors. A 5.6 percent premium was found on houses near to a quiet road compared to a noisy road. A good view compared to a poor one was predicted to generate an 7 percent premium on an expensive house, while an 3.5 percent premium on average priced houses.
8
2.3
Literature approving EPC premium
Denmark was the first country to implement EPC’s, as early as 1997. Before July 2010, the impact of EPC’s on house transaction prices in Denmark was low. After July 2010, a significant effect was measured due to the obligation to publicly display EPC’s during house sales (Jensen, Hansen and Kragh, 2016). The research was based on 117,483 transactions with an EPC label. The transaction price of houses with an EPC of rating A had a positive trend between the year 2007 and 2012. While lower EPC ratings had a downward trend during the same period. Jensen et al. used four log-level regression models to isolate the effect of EPC’s. The premium for “green” houses (A, B or C) before July 2010 was 2.4 percent, while after July 2010 the premium was 10.1 percent. Due to the obligation to display EPC’s in house sales in 2010, people got more aware of the importance, seeing EPC’s as an investment for the future (Jensen, Hansen and Kragh, 2016).
The Netherlands introduced EPC’s in January 2008, as one of the first in Europe. Based on the data of 177,000 houses and apartments, Brounen and Kok (2011) estimated that “green” labeled houses sell for a premium of 2-4 percent compared to “non-green” labeled houses, ceteris paribus. In addition, using a differentiation between EPC ratings, it is shown that houses with the highest EPC rating sell for a 10 percent premium, which according to Brounen and Kok also exceeds the corresponding energy cost savings. This number is close to the 10.1 percent premium found in Jensen, Hansen and Kragh (2016) after the publicly display of EPC’s was introduced in July 2010.
Kok and Jennen (2012) researched the influence of EPC’s on the commercial property sector in the Netherlands. In this article Kok and Jennen found that the rent price of buildings with a “non-green” label is 6.5 percent lower than similar buildings with a “green” energy label. In this paper data from large real estate agents was used (CBRE, DTZ Zadelhoff (now Cushman and Wakefield) and Jones Lang LaSalle). In the log-level regression Kok and Jennen use rental price, size, distance to public transport, EPC label and renovation (dummy variable) as independent variables. The logarithm of rent price per square meter was used as dependent variable. The regression model was similar to the one used by Brounen and Kok (2011). The conclusion in this paper is in line with the article by
Eichholtz (2010), that also proves a premium for “green” labels in the US office market. This paper found a premium of about 3 percent. The premium in effective rent was estimated to be about 7 percent.
A third article about the Dutch residential housing market is by Murphy (2013). The research was based on a survey in two different samples, the first sample got the EPC during the sale transaction of the property and the second sample didn’t have any EPC. According to the survey results, 10 percent of the sample said that the EPC had an influence in the house purchasing process. Of this 10 percent, only 2.07 percent said that it increased the willingness to pay more for the house.
Nonetheless, among the EPC recipient sample group 36 percent of those that are looking
improvements in energy efficient housing features plan on improving their EPC.It’s interesting to know from a consumer perspective what the influence is of EPC ratings. Consequently, EPC’s seem to have little impact on the consumer psychology to pay more for a house. While some results in this paper are negative about EPC, consumers to see EPC labels as in investment for future lower utility bills.
Another paper, by Fuerst, McAllister, Nanda and Wyatt (2016), a big sample of 62,464 transactions is used. Fuerst et al. started with a cross sectional framework to investigate the impact of EPC ratings on transaction prices. Their second method was the repeat-sales methodology. The positive premiums are found to be 12.8 percent for A/B label. Additionally, relative to the label D, C has a premium of 3.5 percent, an E label has a discount of -3.6 percent and F a discount of -6.5 percent.
9 Compared to Brounen and Kok (2011) and Jensen et al. (2016) the premium at the highest energy labels is similar to the average 10 percent found in their papers.
Kahn and Kok (2013) describe the housing market in California, adjusting for Energy Star, LEED and GreenPoint (local standard) rated homes. LEED label is a broader standard than Energy Star, measuring sustainability performance beyond just the energy efficient housing features. The local GreenPoint label is similar to the LEED system, however the impact on sale price is bigger than for LEED labelled homes. This article showed that a house with a “green” label is more likely to be sold for a 2.1 percent higher price than a comparable house without the “green” certification.
Deng, Li and Quigley (2011) found that in the Singapore housing market there is a premium of between 4 and 6 percent if houses are labeled with a government green-mark. This was done with a two stage research setup. The first stage was a hedonic price model. This model was derived from the number of green and non-green transactions out of a sample of 697 houses. In the second stage, the fixed effects estimated for each of the 250 building projects were regressed on the location characteristics of the projects, together with control variables for a Green Mark rating. The two-stage estimation of Deng et al. (2011) showed a significant premium of 4 percent for building “green”. These studies have all found a positive effect of high EPC labels on the transaction price of houses in different areas. These effects have been measured in several ways, but the premium has been consistently positive, showing a 10% premium at most, to 2% at the least. Most studies indicate that the labels are proof of a possible added value of the energy performance of the residential houses.
2.4
Literature nuancing the EPC premium
Although most of the research found showed a positive relationship between a high EPC labelling and a premium in transaction prices, this effect is questioned by some studies.
Hårsman, Daghbashyan and Chaudhary (2016) researched if EPC’s of Swedish family single homes have a positive impact on transaction prices. With data from 2009 and 2010, a small but significant price impact was found. However, this paper also shows that the transaction price premium is partly related to the energy efficient housing features that buyers can observe during the sale. This
indicates that EPC labels by itself do not determine the energy efficiency stimulation in the housing market through a premium on property prices. They found that the observable energy efficient housing features have more impact on the house price, when compared to the EPC scores itself. Wahlström (2016) investigated the influence of energy consumption and energy efficient housing features on Swedish family single homes. His research contained 76,770 observations in the years 2009 and 2010. Furthermore, in the research almost 20 energy efficient housing features and neighborhood characteristics were used as variables in his regression. The results of this paper are contradictory; no significant relationship between energy usage and housing prices in family homes was found. However when looking for a premium in transaction price, there is a positive relationship with energy efficient housing features. Wahlström explains that this difference with the research of Brounen and Kok (2011) is due to the fact that potential house buyers are more likely to observe the energy efficient housing features, rather than look at the energy consumption of the previous owner. In this paper, the influence of EPC’s in not researched.
10 Another paper that doesn’t find a significant relationship between EPC’s and transaction prices of houses is Fuerst and McAllister (2011). This research is one of the first studies done in the UK. During this research, a small sample of 708 commercial properties and 24 BREEAM rates assets was used. There wasn’t found proof that the EPC label had any effect on market rental and transaction prices.
In a German paper of Amecke (2012) the effectiveness of EPC’s and specifically the impact on the private purchasing decisions for the residential house market was researched. In Germany, the energy usage of the housing market is about 40 percent of the total domestic energy usage. The EPBD increased the usage of EPC’s in the German market in 2013. The setup of the research is formed on an online survey; with a sample of 1,239 former customers of the company Immobilien Scout GmbH.
In the pool of respondents in the survey, the awareness and the average comprehension of energy labels was high. However, the use of energy labels is still restricted. Amecke concluded that EPC’s have a limited role in the purchase process of houses. The main reasons for this are as following. Firstly, the EPC’s are not trusted by purchasers. Secondly, the EPC’s don’t show the financial benefits clearly enough. Lastly, other factors count more than EPC’s. Such factors, for instance, are house size, location, terrace, overall price and garden.
The main nuances in the studies above for the EPC premium are the following: there are other factors that influence the transaction price of houses, that outweigh the effect of a high EPC label. Furthermore, the EPC literature does not take into account enough the process through which buyers select their house and assume that a label would be enough to determine the positive effects in terms of savings to pay a premium on a specific house.
The table on the next page summarizes the literature related to EPC influence on house prices. Abelson (1978) as a main source for the factors that predict house prices in general and is cited in many of these studies, is not included in the table.
0 Energy efficient characteristics; for instance: EPC rating, type of heating supply, roof type, type of wall and insulation quality.
Building qualities; for instance interior and exterior maintenance, construction period, terrace, garden, balcony, number of rooms, type of house and size.
Environmental characteristics; for instance noise disturbance from trains/cars, number of frost days per year, distance to public transport and number of shops in neighborhood. Household characteristics; often used survey researches; for instance: employment, education, credit risk and age of the residents.
1 Hårsman et al. state that the underlying energy efficient housing features cause the house price premium, not the EPC label 2 Discount of 6.5% was found if building was inefficient (D or lower), this gives a rounded premium of 7% when efficient 3 All EPC variables in this article are not significant, with EPC A being omitted.
4 In this article there are several dependent variables: ln (price/m2), house price appreciation/m2 and price per m2
Name authors Which variables Research method Dependent variable
Premium Country Sample size Property type Period 𝑹𝟐
Amecke (2012)
Energy efficient, building and demographic
Survey - - Germany 1239 Building sector 2010 -
Brounen & Kok (2011)
Energy efficient, building, environmental, political Log-level regression ln (price/m2) Yes, 3.6% and 10.2% Dutch 177318 & 31993 Residential houses 2007- 2009 0.53 Hårsman et al. (2016)
All OLS ln (price) Yes1 Sweden 77000 Single-family houses 2009-2010 0.52
Jensen et al. (2016)
Energy efficient Log-level regression ln (price/m2)
Yes, 2.4%, 10.1%
Denmark 72326 & 45157
General real estate market 2007-2012 0.60 Kok & Jennen
(2012)
Building and environmental
Log-level regression ln(rent price/m2)
Yes, 7% 2 Dutch 1100 Commercial property 2005-2010 0.71
Murphy (2013)
Building and household Survey - - Dutch 297 & 1027 Residential houses 2010 -
Wahlström (2016)
Energy efficient, building, environmental
OLS ln (price) No Sweden 76770 Single-family houses 2009-2010 0.73
Fesselmeyer (2017)
Energy efficient and building
Log-level regression ln (price/m2)
Yes, 3% Singapore 119826 High-rise development 2000-2016 ? Fuerst &
McAllister (2011)
Energy efficient and household
Log-level regression ln(rent price/m2)
No 3 UK 708 Commercial property 2011 0.57
Kahn & Kok (2013)
Energy efficient, building, environmental
Log-level regression ln (price) Yes, 2.1% US, California
4321 Residential houses 2007-2012 0.91 Eichholtz et al.
(2010)
Energy efficient and building Log-level regression ln (price/m2) Yes, 3% and 16% US 1813 Commercial buildings 2004-2007 0.43 Fuerst et al. (2016)
Energy efficient and building Log-level regression ln (price/m2) 4 Yes, 3.5% and 12.8%
Wales 62464 Residential houses 2003-2014 0.34
Deng et al. (2011)
Energy efficient and building
Log-level regression ln (price/m2)
Yes, 6.1% and 4.2%
Singapore 36512 Residential houses 2000-2010 0.44
10
2.5
Data and research method used
The research methods used in most papers is a regression with the square meter price as dependent variable and multiple hedonic characteristics & energy efficient features as independent variables. Energy efficient housing features ranging from roof and wall type to insulation, ventilation and heater types. In order to correct for economic growth, the area and time fixed effects are used. Some studies use yearly fixed effects, other papers prefer monthly fixed effects. An important aspect in the discussion chapter of Fuerst, McAllister, Nanda and Wyatt (2016) is that it is difficult to isolate the effect of the EPC in researching premiums on house prices. It is important to take into account other factors that have an impact on the price; such as transaction date, age, size, view, noise disturbance from cars or trains and location. Data for these variables are rarely available. They state that excluding these variables could lead to problems like omitted variable bias. Unobserved
variables like quality, condition or recent improvements could increase the premium for houses with higher energy labels, without contributing to energy efficient housing features. In the paper of Fuerst et al., this risk is reduced by removing some houses from the sample that are likely to have been recently improved.
Abelson (1978) did use several environmental variables in his regression and Wahlström (2016) divided all energy efficient housing features in different levels of heating, cooling and ventilation. Brounen and Kok (2011) is one of the most extensive researches on EPC premiums. The data used in their paper was provide by the RVO and the NVM who partnered in the research made more data available than generally possible.
Fesselmeyer (2017) researched the effect of green certification in high-rise buildings in Singapore. In Singapore, developers start selling apartments while starting the construction of high-rise projects. Most of the times they are not done until after the construction has been completed. Some
apartments are sold before the green mark certification program is awarded. Fesselmeyer found that green certification increased the transaction price per square meter by 3 percent. The premium is the biggest for apartments where the observable energy efficient housing features are less obviously “green” to potential buyers. Fesselmeyer used fewer variables than most papers. However, the sample used in this research is more uniform. Since most apartments in high-rise developments have identical construction elements and are built in the same period, it is easier to separate the green-mark certification premium from other variables influencing the price per square meter.
Not all recent literature was able to receive as much data as papers like Brounen and Kok (2011) or Hårsman, Daghbashyan and Chaudhary (2016). Fesselmeyer reduces the need for many
environmental and energy efficient control variables, by researching more uniform houses. In this paper, the problem mentioned in the paper by Fuerst et al. (2016) is also partly solved by researching more uniform areas of residential real estate. Houses within VINEX-districts are constructed at the same time, while having similar building styles. Therefore, the chance will be lower that omitted variable bias will occur. Since more environmental variables (noise disturbance, recreation, view, location) and building variables (size, type of house, terrace, garden) will be comparable in this sample.
Fesselmeyer (2017) used two variables in his regression identifying green certification. The first is ‘GM’, a dummy variable indicating if the apartment received a green-mark certification. The second is ‘post’, a dummy variable indicating whether the apartment has been sold before or after the green-mark certification. The research in this paper will also include two variables according to EPC ratings. The first is EPC_Score, indicating the rating of energy efficiency (where the letter scale is transformed into a numbered rank). It is assumed that EPC_Score has a positive influence on the
11 transaction price per square meter. The majority of literature finds a premium. The second is
EPC_Premium, a dummy variable indicating whether the EPC label of the house is higher than the average EPC label of the street. The reason this dummy variable is added, is to determine the relative advantage to neighboring houses in terms of transaction price because of a higher energy efficiency. Furthermore, the influence of the individual ratings A++ until B will be tested, by adding dummy variables.
12
3. Data and Methodology
In this chapter, the data collection and research methodology are discussed.
3.1
Sample selection and data sources
An EPC is designed to give potential house buyers an indication of their monthly energy costs and overall durability of the residence they are buying (Brounen and Kok, 2011). The RVO5 decides the ranking of the label, and validates the labels. This makes the ranking trustworthy. Energy labels stimulate house owners to make their house more energy efficient, since the owners get rewarded with a higher energy label, lowering the energy costs, and possibly increasing the transaction price when selling the property.
Several papers have done research on national scale or on specific markets like the housing markets in the main metropolitan areas of countries. These researches take quite a lot of control variables to control for the diversity of the housing supply. As the market for the main metropolitan area of country with a significant housing deficit could skew the results of the analysis, this research focusses on a specific subset of the Dutch housing supply.
The sample consists of transactions for houses that were built between 1995 and 2005 as part of the Vierde Nota Ruimtelijke Ordening (the fourth national plan on urban development) in the
Netherlands. This government note decides which areas are used for building planned residential areas called VINEX. These ‘VINEX-districts’ are of interest due to the following characteristics they have in common. Each of the areas in the sample is an expansion area for a city in the Netherlands. The areas were designed to be built against the city and not as a suburb. The areas typically are built with a more urban neighborhood, consisting of apartment buildings closer to the city center and a more suburban area consisting mainly of houses. The areas have a small number of amenities (some shops, a limited amount of schools (mostly primary education) and situated close to infrastructure (highway, rail and bus services to the city center or hubs of economic activity) (Ministerie van Volkshuisvesting, Ruimtelijke Ordening en Milieu, 1990).
The houses were built to the most recent building standards, specifying ceiling height, minimal floor size, automatic ventilation and central heating as a standard (Ministerie van Volkshuisvesting, Ruimtelijke Ordening en Milieu, 1990). Most houses were modularly built, meaning a concrete frame overlaid with brickwork on the façade. Most houses were also isolated to a specific standard.
5Due to a reshuffling of responsibilities the RVO has taken on the responsibilities of other governmental agencies, like those of the Ministry of Urban Planning & Housing.
13
Table 2
Data sources
Source Sample Size
NVM Transaction price Date Size Duration Address VINEX-district Neighborhood6 Provided number of transactions: 67,896 transactions RVO EPC label Date Available sample: 18,716 transactions Transactions matched with EPC label Final sample:
5,380 transactions
The transaction data from the NVM shows all the transaction prices of houses to about 50 years back in time. For this research, the NVM provided transaction information on address level for 50 VINEX-districts in the Netherlands. House characteristics are in this case; transaction price, selling duration and house size in square meters. Selling duration is the period of the sale until transaction had occurred. This variable is also used in the regression of Brounen and Kok (2011) under the name of average time on market in days. The location characteristics are VINEX district name, postal code, neighborhood code. Neighborhood code identifies the neighborhoods inside the VINEX-district, as there can be multiple neighborhoods inside one VINEX.
Although since January 2015 each transaction should include a formal EPC label (formalized in the “Besluit energieprestatie gebouwen”), the Dutch government has provided each home owner with a preliminary estimated label based on reference values like size, age and state of repair of the object. The preliminary labels were not available for analysis. Only the validated labels in the RVO database have been used for comparison. For the EPC rating of the houses, data was collected from the RVO. All registered EPC’s in the Netherlands are on this list. These were matched with addresses in the dataset.
As an EPC label has been made mandatory with house transactions since January 20157, of the 67.000 transactions only a fraction is complete. The sample with complete data consisted of 10.830 transactions. As label application and transaction do not necessarily match, the sample was trimmed down further to transactions of which the label was applied at most five years before of the sale of the house, to zero in on transactions of which it was sure the label could play a possible role in the buyer’s decision. This brings the sample size further down to 5.380 transactions for a time period between 2008 and 2016.The article by Kahn and Kok (2013) has a very large sample of control buildings, but only 4.321 EPC rated buildings in their sample. Abelson (1978) researched 1.400 properties. Eichholtz et al. (2010) have a sample of 1.813 houses. Kok & Jennen (2011) used a sample
6The neighborhood is identified by neighborhood code given by the Dutch National Bureau of Statistics (CBS).
7Although a formal EPC has become mandatory for home owners when selling their house as per January 2015
(besluit energieprestaties gebouwen), some homeowners have applied for a label for their house as early as 2005. These are included in the sample. Transactions of which a label was necessary, but not provided were excluded.
14 of 1.100. These are all qualitative academic papers on this subject, indicating that a sample of 5.380 EPC-rated transactions is enough to reach qualitative results.
The table below displays the distribution of the number of observations from the RVO between different energy labels. We can see that the 18716 observations are not evenly distributed. Only 11 observations are A++ and only 10 are A+.
Table 3
Distribution EPC labels in samples compared to the RVO database
EPC label RVO database Available sample Final sample
A++ 0.02% 0.06% 0.09% A+ 0.07% 0.05% 0.13% A 13.85% 55.12% 59.28% B 14.86% 37,57% 32.81% C 28.88% 6.90% 7.32% D 18.92% 0.20% 0.32% E 11.77% 0.04% 0.06% F 7.09% - - G 4.53% 0.06% -
This table states the frequencies of each type of validated label assigned to houses in the Netherlands and the research sample. The distribution for the RVO does not include the preliminary labels.
As can be seen in the table above, when comparing VINEX locations with the national distribution of EPC label, the distribution of VINEX locations is strongly skewed towards houses with a high energy performance (label “B” or up). The very low incidence of houses with a label lower than “C” indicates that several VINEX locations share postal code areas with houses that were built in the neighboring areas to cities. These are commonly old farms, built outside the city limits long before the districts were planned. The VINEX-districts have been built adjacent or around these houses.
3.2
Conceptual framework and hypotheses
In figure 1 the conceptual framework for this research is presented. The main effect that is measured is the effect of an investment in enhancing the energy performance of a house compared to other houses in the same street in a VINEX-district.
The conceptual framework is a variant of the conceptual models in Wahlström (2016) and Hårsman et al. (2016) in which a direct positive relationship is identified for the attaining a high EPC label and the transaction price of a house. Conform to the study by Wahlström (2016), this model identifies two types of controlling variables for the transaction prince. The control variable distinguished as either object-related: size and the duration (time to market) or location related: the price of other houses in the same street.
As no other attributes of the houses were available, size is the main object-related control variable for the transaction price per square meter.
15 As the selling price of houses does not grow proportionally to the growth in size, the transaction price per square meter is negatively skewed. The average square meter price of an apartment is higher than the square meter price of a house. This effect was also found in Fesselmeyer (2017). Figure 2 of the paper by Hårsman et al. (2016) shows five different factors influencing the transaction price of a house. Whereas this thesis researches a uniform area like VINEX, making the
“neighborhood attributes” and “local climate” factors less meaningful.
The average selling price of houses in the VINEX-district and the street are control variables for location. In the study by Wiley, Benefield and Johnson (2008), the average rent was included in the model as a control variable. In this research looking at transaction price, the averages of the VINEX-district and street are suitable control variables as well. In the paper by Wiley et al., the Energy Star and LEED ratings seem to have significant positive impact on the average rent variable.
Based on the literature review, specifically Brounen & Kok (2011), Jensen et al (2016) and Kahn & Kok (2013) and linked to the conceptual model the following two hypotheses can be stated:
Hypothesis 1 Attaining a high energy performance label will positively influence the transaction price of a house in a VINEX-district
Hypothesis 2 Attaining an energy performance label that is higher than other houses in the street will lead to a higher transaction price of a house in a VINEX-district
3.3
Regression model and variables
To measure the effect of the EPC premium on the transaction price a linear time series regression is used. Due to the fact that neighborhoods in several VINEX-districts in the Netherlands are compared, the data is seen as panel data and treated as such. This implies identifying groups in the regression analysis and taking into account that the transaction data is seen as time series.
The following regression models are used to determine the effect of an EPC premium on the transaction price per square meter. In the regression analysis, a model is used that is comparable to the one in Brounen and Kok (2011). In this model, the dependent variable is based on the natural logarithm of the transaction price of the house per square meter.
Regression model 1: 𝑃𝑟𝑖𝑐𝑒 = 𝛼 + 𝛽1AveragePV𝑡𝑖+ 𝛽2AveragePS𝑡𝑖 + 𝛽3𝑆𝑖𝑧𝑒𝑡𝑖+ 𝛽4Duration𝑡𝑖+ 𝛽5EPC_"A + +"𝑡𝑖+ 𝛽6EPC_"A + "𝑡𝑖+ 𝛽7EPC_A𝑡𝑖 + 𝛽8EPC_B𝑡𝑖 + 𝜀𝑡𝑖
Transaction Price Average price street Average price VINEX Size Duration EPC Premium
16 Price = The dependent variable indicating the natural logarithm of the price
per square meter. ln(transaction price/m2)
Size = Control variable on size of the house in square meters. ln(m2) AveragePS = Average selling price of houses in the same street
Duration = Time span the house is for sale in days
EPC_A++ = Dummy variable, valued “1” when the EPC label of a house is A++ EPC_A+ = Dummy variable, valued “1” when the EPC label of a house is A+ EPC_A = Dummy variable, valued “1” when the EPC label of a house is A EPC_B = Dummy variable, valued “1” when the EPC label of a house is B AveragePV = Average selling price of houses in the VINEX-district
In accordance with most other researches that tested the effect of EPC on the transaction price of houses, the transaction price per square meter was used to control for differences in house types (e.g. apartments, villas). To correct for extreme values, this measure was log mutated. The same applies to the size of the houses. Outliers in size skewed earlier research and therefore the variable is log mutated. The dummy variables EPC_A++ to EPC_B denote a significant investment in an energy efficient house.
Regression model 2: 𝑃𝑟𝑖𝑐𝑒 = 𝛼 + 𝛽1AveragePV𝑡𝑖+ 𝛽2AveragePS𝑡𝑖 + 𝛽3𝑆𝑖𝑧𝑒𝑡𝑖+ 𝛽4Duration𝑡𝑖+ 𝛽5EPC_Score𝑡𝑖+ 𝛽6EPC_Premium𝑖+ 𝜀𝑡𝑖
Price = The dependent variable indicating the natural logarithm of the price per square meter. ln(transaction price/m2)
Size = Control variable on the size of the house in square meters. ln(m2) AveragePS = Average selling price of houses in the same street
Duration = Time span the house is for sale in days
EPC_Score = Ranking variable of the EPC labels, scored “1” for label G and “8” for label A++ EPC_Premium = Dummy variable valued at “1” when the EPC score of a specific house is
higher than the average of the street
AveragePV = Average selling price of houses in the VINEX-district
Model 2 replaces the dummies for the separate labels with a ranking variable for EPC labels in the sample. EPC_Premium is a dummy variable that is 1 when the label of the house is higher than the average of other houses sold in the street.
Model 3 is a variant in which model 2 is tested for year fixed effects. In this model, a dummy for each year in the dataset is included to control for unobserved time-variant factors. Like Brounen & Kok (2011), Kahn & Kok (2013) and Jensen et al. (2016), the time fixed effects are used to control for macro-economic variables like changes in GDP, inflation and interest. Research in the last ten years has shown controlling for time fixed effects has become more effective than controlling for separate economic factors.
17
4. Results
In this chapter, the results of the empirical research are presented. The chapter first presents the descriptive statistics, then presents how the data fits the parametric assumptions. Next the findings of the empirical research are presented. Finally, the hypotheses are tested.
4.1
Descriptive statistics
In the table below the data has been split into the variables used in the analysis. The table shows the distribution of the data used.
Table 4
Descriptive statistics of variables in the regression model.
VARIABLES N mean sd min max skewness kurtosis
Price 5,380 7.627 0.200 5.948 8.505 0.0710 4.259 Size 5,380 4.784 0.285 3.466 6.907 -0.234 4.120 AveragePS 5,380 261,793 68,767 118,750 1,075,000 1.721 13.27 Duration 5,380 183.5 244.7 0 1,064 2.223 7.602 EPC_Score 5,380 5.517 0.656 2 8 -1.029 3.770 EPC_Premium 5,380 0.501 0.500 0 1 -0.00520 1.000 EPC_A++ 5,380 0.000929 0.0305 0 1 32.76 1,074 EPC_A+ 5,380 0.00130 0.0361 0 1 27.67 766.6 EPC_A 5,380 0.593 0.491 0 1 -0.378 1.143 EPC_B 5,380 0.328 0.470 0 1 0.732 1.536 AveragePV 5,380 267,185 44,433 195,389 373,882 0.138 2.535 Year 5,380 2014 1.670 2007 2015 -2.907 10.39 Number of neighborhoods 172 172 172 172 172 172 172
The dependent variable Price is measured as the natural logarithm of the selling price per square meter. Size is measured as the natural logarithm of the square meters of floor space in the house. Average PS and Average PV are the average selling price of houses in the street and VINEX-district respectively. These indicate a reference price for the transaction price of a dwelling. EPC_Score through EPC_B are different measures to indicate the availability of a certain EPC label for the property. EPC_Score is a recoded variable indicating “2” for an F-label, and “8” for an A++ label. EPC_Premium is a dummy that is “1” when the EPC label is higher than the average of the street. EPC_A++ through EPC_B are dummies with value “1” when that specific label is awarded to a house.
4.2
Parametric tests
To use a regression model to predict the effect of EPC on the transaction square meter price of houses, the data was tested for parametric assumptions. Linearity was tested by plotting scatterplots for the dependent variables and main predictive variables. As the data set is treated as set of time series for neighborhoods in VINEX areas, independence of cases is not met. This is corrected for by using a time series regression for the analysis. To make sure no multicollinearity would disturb the
18 analysis, correlation matrix of all variables was made and the model was tested for VIF. No disturbing multicollinearity was found.
1 2 3 4 5 6 7 8 9 10 11 1 PRICE 1.000 2 SIZE 0.006 1.000 3 AVERAGEPS 0.557* 0.431* 1.000 4 DURATION -0.017 0.022 0.022 1.000 5 EPC_SCORE 0.150* 0.120* 0.176* -0.009 1.000 6 EPC_PREMIUM -0.050* 0.057* -0.042* -0.032* 0.467* 1.000 7 EPC_A++ 0.026 0.017 0.003 0.008 0.115* 0.030 1.000 8 EPC_A+ -0.024 0.008 -0.015 -0.001 0.082 0.036* -0.001 1.000 9 EPC_A 0.124* 0.098* 0.160* -0.019 0.889* 0.447* -0.037* -0.044* 1.000 10 EPC_B -0.056* -0.050* -0.093* 0.026 -0.551* -0.308* -0.021 -0.025 -0.843* 1.000 11 AVERAGEPV 0.578* 0.097* 0.557* 0.012 0.189* -0.019 -0.004 -0.010 0.161* -0.077* 1.000
Normalcy has been tested with checking the kurtosis and skewness of the predictive variables. The variables that showed a strong disturbing kurtosis and skewness, have been winsorized, where outliers have been replaced by the values at the first and 99th percentile. Finally, the data was tested for equality of variance by employing a modified Wald test for heteroscedasticity. The test showed that heteroscedasticity is present in the data. As a consequence, the models have been run with robust standard errors.
4.3
Results of regression analysis
In table 6 below, the results for the regression analysis are given. Model 1 presents the effect of specific high quality EPC labels (label B and higher) on the transaction price per square meter. Model 2 replaces the replaces the specific EPC label with EPC score representing the ranking of energy efficiency. It also introduces the EPC_Premium, a dummy variable that is “1” when the EPC_Score of the house is higher than the average score of the street. Model 3 adds year fixed effects to the model to control for the development of the transaction price per square meter due to macro-economic factors, as is done in Brounen & Kok (2011), Kahn & Kok (2013) and Jensen et al. (2016). The year fixed effects were tested by adding year dummies (the complete table can be found in the Appendix).
Table 5: Correlation table
19
Table 6 Results of regression analysis
(1) (2) (3) VARIABLES Dependent: Price Dependent: Price Dependent: Price Size -0.109*** -0.108*** -0.115*** (0.0198) (0.0198) (0.0197) AveragePS 0.0000***8 0.0000*** 0.0000*** (0.0000) (0.0000) (0.0000) Duration -0.0001 -0.0001 -0.0001 (0.0000) (0.0000) (0.0000) EPC_Score 0.0122* 0.0083 (0.0068) (0.0075) EPC_Premium -0.0117* -0.0116* (0.00682) (0.0067) EPC_A++ 0.181* (0.107) EPC_A+ -0.0546 (0.0847) EPC_A 0.0199* (0.0110) EPC_B 0.0221** (0.0101) AveragePV 0.0000 0.0000 0.0000 (0.0000) (0.0000) (0.0000) Constant 7.780 7.737 7.831 (0.0000) (0.0000) (0.0000) Observations 5,380 5,380 5,380 R-squared 0.108 0.106 0.138 Number of neighborhoods 172 172 172
8Due to truncating the results for readability, these values seem non-significant, but due to the relative size difference
between the dependent and this independent variable the unstandardized coefficient is very small. The non-truncated values can be found in appendix 2
20
Neighborhood FE Y Y Y
Year FE N N Y
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
This table contains the results for the regression models (rounded at 4 decimals). Absolute value of t statistics in
parenthesis. Size is the log of the square meter floor surface of the house, Average PS and Average PV are the selling prices of houses in the street and VINEX-district. Duration is the number of days the house was for sale. EPC_A++, EPC_A+, EOC_A, and EPC_B are dummies that are “1” for the attained label. EPC_premium is a dummy that is “1” for houses that have a higher EPC label than the street it is situated on. EPC_score is the recoded value of the EPC label of the house.
The models have relative low predictive value, as the R2 varies from 0.108 for model 1 to 0.138 for model 3. The predictive value of the model stays consistent. Model 1 shows the main predictors for the transaction price (Size, AveragePS) have significant effect on the transaction price.
As expected, Size has a negative effect on the transaction price. The average price in the street has a positive effect. Model 1 furthermore gives evidence of the premium effect of attaining an EPC label indicating high energy efficiency. Attaining an A or B label leads to a 2% premium. Attaining an A++ label implies an 18% premium in transaction price. The data for an A+ label show a negative, non-significant effect. This could be due to a limited amount of observations. Supporting this explanation; people that decide to upgrade their current A labeled home choose in most cases to go for the A++ level, not the A+ level.
In line with Jensen et al. (2016) and Kahn and Kok (2013), neighborhood fixed effects have been used to adjust for the postcode within the VINEX-district. As stated in Kahn and Kok (2013) the
neighborhood fixed effects are used to control for the quality of the particular neighborhood and location specific factors such as the district being a part of a large metropolitan area near the capital of the Netherlands.
Model 2 indicates a consistent effect of Size and Average PS as significant predictors. The effect of EPC_Score on transaction price is positive (B=0.012, p<0.10). Notable in this model is the negative effect of attaining a better energy performance label than the houses in the street
(B= -0.012, p< 0.10).
When controlling for the year of sale in Model 3, the effect of the EPC on transaction price disappears. The year effect dummies show a development that is comparable to the price development of houses in the Netherlands in the period 2008-2015, showing a decline as the demand for house stalled to the global financial crisis, and deepened during the period 2012-2013, before recovering in 2015 (CBS, 2017).
4.4
Hypothesis testing
Based on the results of the regression analysis the stated hypotheses can be tested.
Hypothesis 1 – Attaining a high energy performance label will positively influence the transaction price of a house in a VINEX district
Is supported by the results of the regression analysis, as there is a positive relationship between attaining a high energy performance label and the transaction price. A generic EPC score showed a 1,8% premium (B=0,011, p< 0.10), testing for specific labels A++, A and B, showed a premium of almost 2% and upwards, culminating in an 18% premium for the A++ label.
Hypothesis 2 – Attaining an energy performance label that is higher than other houses in the street will lead to a higher transaction price of a house in a VINEX district
21 Is rejected due to a lack of evidence for a positive relationship between an EPC premium and the transaction price per square meter. This research does not support the positive effect of
outperforming the neighboring houses on energy efficiency as a way to obtain a premium on the transaction price of a house. Specifically, the research found negative effect on the transaction price of the EPC premium (B = -0.011, p< 0.10)
22
5. Conclusion
The aim of this paper was to research the existence of an EPC premium in specific sections of the Dutch housing markets. Isolating the possible effects of an EPC label without having to control for the diversity of a local supply of dwellings. The literature finds premiums ranging from 2.1% to 7% for “green” upgrades in general. While the “A” EPC labeled houses give premiums ranging from 10% to 12.8% according to Jensen et al (2016), Brounen and Kok (2011) and Fuerst et al. (2016).
The research question of this paper is; what is the effect of upgrades in Energy Performance
Certificates (EPC’s) on the transaction price of houses in planned residential areas (VINEX-districts) in the Netherlands?
With a sample of 5,380 transactions, using three regression models, the answer to this question is bilateral.
The first hypothesis is supported by the results of the regression analysis. There is a positive relationship between attaining a high energy performance label and the transaction price, with premiums ranging from 1.2% to 18%. The first premium found is predicted by the variable
EPC_Score. Every EPC label upgrade from C upward, leads to a 1.2% higher square meter price. This is close to the lowest EPC premiums found in the literature overview of table 1. The individual
premiums, for dummy variables A++ to B, range from 18% to 2.2%. The A++ EPC label leads to a 18% higher square meter price. An explanation for this could be that most people with A labeled homes in VINEX-districts, in the case of upgrading, choose for the highest possible label (A++), instead of A+. Several papers don’t include EPC labels higher than A. Therefore this premium is an addition to the existing literature on EPC labels.
The second hypothesis is not supported by the regression analysis. There is not a significant positive influence on the transaction price found, when the EPC rating is higher than the average EPC rating of the street. There is a positive confirmation on the research question when looking at the first hypothesis. However, when looking at the second hypothesis, we can conclude that in VINEX-districts relative improvements in energy efficiency compared to the neighboring houses are not worth the effort. Contradictory, homeowners have financial incentives to improve upon their EPC. The possible discount (of the higher EPC relatively to average of street) is diminished by the long term savings on energy costs.
This paper further contributes to the existing literature by researching a more uniform supply of houses. Accompanied with new independent variables included in the empirical research. According to Abelson (1978) the biggest factors that explain house price variation are size, inflation and environmental factors. In this research, the environmental factors are limited because uniform planned residential areas are used as a sample. Size is a good predictor of house prices, as can be seen in the results chapter. In all the academic papers about EPC, inflation is replaced by correction for yearly fixed effects. When correcting for yearly fixed effects, any positive effects of EPC are diminished. The yearly fixed effects only show a limited influence themselves for the years 2013 and 2014. These years are seen as the low point in the recent development of the Dutch housing market (CBS, 2017).
Hårsman, Daghbashyan and Chaudhary (2016) find that the premium on green-labeled houses is rather caused by the observable underlying energy efficient housing features rather than the EPC label itself. In this paper, a premium on EPC’s is found without adding much energy efficient housing features as variables. It seems that in VINEX-districts, because of the more uniform house supply, controlling for more energy efficient housing features is less necessary.
23
6. Limitations and future research
The R2 for the models is lower than the average R2 in table 1 of the literature chapter. Most of the previous papers include more predictor variables, thereby increasing the explanatory power of the models in those studies. Due to the fact that limited data was available from NVM and the RVO, choosing a uniform area like VINEX-districts decreases the need to control for many “neighborhood attributes” and “local climate” characteristics. Despite the fact that the premiums found are still in line with the literature, a higher R2 would be more desirable.
Another limitation is the distribution of data between energy labels. Only a small percentage of the 18,716 observations had an EPC of A+ or A++. This could be a reason that the relatively high 18% premium was found in the regression. Contra dictionary, also a very small percentage had an EPC of D, E, F or G. In ‘sample selection and data sources’ we discuss that this problem of unequal
distribution of EPC’s is also outside of VINEX-districts. The distribution of EPC’s in the Netherlands contains even fewer A++ and A+ EPC labels. However, the “non-green labels” indicating D or lower, are in higher quantities in the Netherlands as a whole. This could be related to the fact that VINEX-district houses are more energy efficient than the average house in the Netherlands.
From the RVO we could only receive the most recent list of EPC’s of all the registered houses in the Netherlands. The problem in our dataset is deciding what the original label of the particular VINEX-district was at the time of building, so the comparison with other transactions in the street is used as a proxy for an actual measurement of the energy performance of the houses in the street.
Furthermore, being able to distinguish between upgraded houses and those bearing the original label would be of worth in future research. Currently the data provided does not distinguish between the recent and original label. Knowing this information would help put an economic value on the improvements in energy efficiency in terms of future gains when selling the house.
Regardless of the limitations of this research it opens the possibility to compare the Dutch housing market with others in Europe, specifically recently built planned residential areas. Fesselmeyer (2017) researched a very uniform housing supply (high-rise developments) in Singapore. But we haven’t found any papers on planned residential areas.
Additionally, adding more energy efficient housing features as control variables would help determine how efficient the model in this study was. When researching planned residential areas, environmental, building and energy efficient characteristics are less needed as control variables. But the addition of heating, ventilation, isolation, HR efficient glass panes could certainly increase the likelihood of isolating the EPC premium.
Lastly, when researching a very specific type of neighborhood, it is harder to generalize this to the existing literature. Values of premiums could differ much from average premiums found in current papers on broader samples of houses. Making the study design more comparable would help validate the results found.
24
References
Abelson, P. W. (1978) Property prices and the Value of Amenities. Journal of Environmental Economics and Management, Vol. 6, 11-28.
Amecke, H. (2012) The impact of energy performance certificates: A survey of German home owners. Energy Policy, Vol 46, 4-14.
Brounen, D., & Kok, N. (2011) On the economics of energy labels in the housing market. Journal of Environmental Economics and Management, Vol. 62, Issue 2, 166-179. Centraal Bureau voor de Statistiek (2017) Huizenmarkt in Beeld, CBS, Voorburg.
Deng, Y., Li, Z., & Quigley, J. M. (2011) Economic Returns to Energy-Efficient Investments in
the Housing Market: Evidence from Singapore. Regional Science and Urban Economics, Vol 42,
506-515.
Eichholtz, P., Kok, N. & Quigley, J. (2010) Doing well by doing good: Green Office buildings. American Economic Review, Vol. 100, Issue 5, 2494-2511.
Fesselmeyer, E., (2017) The value of green certification in the Singapore Housing Market. Economic Letters, Vol. 163, 36-39.
Fuerst, F., McAllister, P. (2011) The impact of Energy performance Certificates on the rental and
capital values of commercial property assets. Energy Policy, Vol 39, 6608-6614.
Fuerst, F., McAllister, P., Nanda, A., & Wyatt, P. (2016) Energy performance ratings and house prices
in Wales: An empirical study. Energy Policy, Vol. 92, 20-33.
Hårsman, B., Daghbashyan, Z., & Chaudhary, P. (2016) On the quality and impact of residential
energy performance certificates. Energy & Buildings, Vol. 133, 711 - 723.
Jensen, O., Hansen, A., & Kragh, J. (2016) Market response to the public display of energy
performance rating at property sales. Energy Policy, Vol. 93, 229-235.
Kahn, M., Kok, N. (2013) The capitalization of green labels in the California housing market. Regional Science and Urban Economics, Vol 47, 25-34.
Kok, N., Jennen, M. (2012) The impact of energy labels and accessibility on office rents. Energy Policy, Vol 46, 489-497.
Ministerie Volkshuisvesting, Ruimtelijke Ordening & Milieu (1990) Vierde nota over de ruimtelijke
ordening: op weg naar 2015: extra (VINEX), Ministerie VROM, Den Haag.
http://publicaties.minienm.nl/documenten/vierde-nota-over-de-ruimtelijke-ordening-op-weg-naar-2015-extra-vinex
Murphy, L. (2013) The influence of the Energy Performance Certificate: The Dutch case. Energy Policy, Vol. 67, 664-672.
Wahlström, M., (2016) Doing good but not that well? A dilemma for energy conserving homeowners. Energy economics, Vol. 60, 197-205.
25 Wiley, J., Benefield, J., & Johnson, K. (2008) Green Design and the Market for Commercial
Office Space. The Journal of Real Estate Finance and Economics, Vol. 41, 228-243.
Appendix 1 - Examples of EPC labels
26
Appendix 2 - Expanded table of regression results
Table 7
27
(1) (2) (3)
VARIABLES Transaction Price Transaction Price Transaction Price
Size -0.109*** -0.108*** -0.115***
(0.0198) (0.0198) (0.0197)
AVERAGEPS 1.12e-06*** 1.11e-06*** 1.08e-06***
(9.00e-08) (9.01e-08) (8.28e-08)
Duration -1.17e-05 -1.16e-05 -1.14e-05
(8.20e-06) (8.20e-06) (8.22e-06)
EPC_Score 0.0122* 0.00825 (0.00683) (0.00750) EPC_Premium -0.0117* -0.0116* (0.00682) (0.00672) EPC_A++ 0.181* (0.107) EPC_A+ -0.0546 (0.0847) EPC_A 0.0199* (0.0110) EPC_B 0.0221** (0.0101)
AVERAGEPV 2.17e-07 2.18e-07 2.15e-07
(1.935e-06) (1.384e-06) (1.424e-06)
2008.Year 0.0159 (0.0519) 2009.Year -0.0517 (0.0547) 2010.Year -0.0544 (0.0529) 2011.Year -0.00285 (0.0521) 2012.Year -0.0690 (0.0517) 2013.Year -0.142***
28 (0.0533) 2014.Year -0.118** (0.0517) 2015.Year -0.0249 (0.0510) Constant 7.780 7.737 7.831
(5.170e-11) (3.698e-11) (3.804e-11)
Observations 5,380 5,380 5,380
R-squared 0.108 0.106 0.138
No. of neighborhoods 172 172 172
Neighborhood FE Y Y Y
Year FE N N Y
This table contains the results for the regression models. Above is the same table as presented in chapter 4 results, but the years are split into separate dummy variables. Absolute value of t statistics in parenthesis. Size is the log of the square meter floor surface of the house, Average PS and Average PV are the selling prices of houses in the street and VINEX-district. Duration is the number of days the house was for sale. EPC_A++, EPC_A+, EOC_A, and EPC_B are dummies that are “1” for the attained label. EPC_premium is a dummy that is “1” for houses that have a higher EPC label than the street it is situated on. EPC_score is the recoded value of the EPC label of the house.
