A hedonic price analysis of the value of industrial sites

Abstract

Hedonic price modelling is a widely used technique to explain the value of different types of individual property. Following the notion that areas within the city can suffer from devaluation, the question arises what factors influence the value of urban areas. In this paper, we use hedonic price analysis to answer this question for a specific type of urban area, the industrial site. We use the average property value per hectare as a representation of the value of an industrial site. A distinction is made between three types of explanatory variables: physical characteristics of the industrial site, regional economic characteristics and general economic trends. Although the overall explanatory value of our model appears to be modest compared to existing hedonic pricing studies of individual property, results show that most explanatory variables in our model have the expected coefficients and signs, indicating that this method can be applied in a meaningful way to gain insight into the valuation of urban areas.

Full text

This article was downloaded by: [Radboud Universiteit Nijmegen]
On: 04 November 2013, At: 01:14
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Journal of Property Research
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/rjpr20
A hedonic price analysis of the value of
industrial sites
Jasper Beekmansa, Pascal Beckersb, Erwin van der Krabbenac &
Karel Martensa
a Institute for Management Research, Radboud University
Nijmegen, Nijmegen, The Netherlands
b PBL Netherlands Environmental Assessment Agency, The Hague,
The Netherlands
c School of the Built Environment, University of Ulster, Belfast, UK
Published online: 31 Oct 2013.
To cite this article: Jasper Beekmans, Pascal Beckers, Erwin van der Krabben & Karel Martens ,
Journal of Property Research (2013): A hedonic price analysis of the value of industrial sites,
Journal of Property Research, DOI: 10.1080/09599916.2013.836556
To link to this article: http://dx.doi.org/10.1080/09599916.2013.836556
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &

Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

A hedonic price analysis of the value of industrial sites
Jasper Beekmans
a
*, Pascal Beckers
b
, Erwin van der Krabben
a,c
and Karel Martens
a
a
Institute for Management Research, Radboud University Nijmegen, Nijmegen,
The Netherlands;
b
PBL Netherlands Environmental Assessment Agency, The Hague,
The Netherlands;
c
School of the Built Environment, University of Ulster, Belfast, UK
(Received 3 January 2013; accepted 15 August 2013)
Hedonic price modelling is a widely used technique to explain the value of
different types of individual property. Following the notion that areas within the
city can suffer from devaluation, the question arises what factors influence the
value of urban areas. In this paper, we use hedonic price analysis to answer this
question for a specific type of urban area, the industrial site. We use the average
property value per hectare as a representation of the value of an industrial site.
A distinction is made between three types of explanatory variables: physical
characteristics of the industrial site, regional economic characteristics and
general economic trends. Although the overall explanatory value of our model
appears to be modest compared to existing hedonic pricing studies of individual
property, results show that most explanatory variables in our model have the
expected coefficients and signs, indicating that this method can be applied in a
meaningful way to gain insight into the valuation of urban areas.
Keywords: industrial property; brownfield sites; multiple regression analysis;
hedonic method; valuation
1. Introduction
The aim of this paper is twofold. First, Bryson’s(1997) notion that areas in the city
may suffer from devaluation poses the question what factors influence the value of
urban areas in the first place. This paper is inspired by hedonic price analysis of
individual (industrial) properties. Such studies, among others, were used to find
possible explanatory variables in an analysis that was carried out to research what
factors influence the value of a specific type of urban area: the industrial site.
The second aim of our study is to test whether a method that is inspired by
hedonic price modelling can be applied to urban areas instead of individual prop-
erty. The classic notion behind hedonic theory as defined by Rosen (1974) is that
‘goods are valued for their utility-bearing attributes of characteristics’(Rosen,
1974, p. 1). This implies that all individual characteristics of a good contribute to
the price of that good. However, these characteristics cannot be traded individually.
Although an urban area is not a good that is traded, and consequently does not
have a price as such, the rationale behind hedonic analysis that the overall price of
an object is determined by the implicit price of its characteristics, can be applied to
urban areas as well. Since urban areas are not a priced commodity per se, the value
*Corresponding author. Email: j.beekmans@fm.ru.nl
© 2013 Taylor & Francis
Journal of Property Research, 2013
http://dx.doi.org/10.1080/09599916.2013.836556
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

of an urban area is defined as the aggregate value of the property that is located in
that area. A similar approach is used by Reed (2012) in his study on factors that
influence the housing value of suburbs in Melbourne. Contrary to his study, we do
not use the median property price within an area, but an average value per hectare
(see Section 3.2).
This paper provides insight into the value of an urban area and can be useful
for policy-makers. Urban policies often aim at achieving the highest possible prop-
erty values in an area, when it comes to both maintenance of existing urban areas
as well as the development of new ones. This is especially true for, but not limited
to, industrial sites. In the Netherlands, this particular urban area has gotten signifi-
cant attention lately because of the rapid decline of these parts of the city (van der
Krabben & Buitelaar, 2011). It is often a goal of redevelopment policies to increase
the value of the properties on industrial sites as a means to stop decline
(Ploegmakers & Beckers, 2012). Studying what characteristics of the area influence
the value of the property within that area can thus be a useful tool for policy-
makers that are looking for characteristics of urban areas that can be influenced via
urban policies. This can help policy-makers to more easily make decisions on
implementing policies and the investments that come with it. Assessing value at the
level of an urban area instead of the much more common level of individual prop-
erty provides insights for policy-makers at a level at which policies actually aim to
have an effect.
2. Studies of property value and industrial sites
2.1. Hedonic price modelling
Hedonic price modelling is widely used as a technique for the assessment of prop-
erty value, estimating demand for specific attributes of housing and neighbourhoods
and analysing price indexes for different types of property (Páez, Long, & Farber,
2007). Hedonic price modelling can be applied to explain the value of heteroge-
neous goods (Adair et al., 2003; Dunse & Jones, 1998). This heterogeneity is
reflected in the different characteristics of an office building, house or industrial
property. A precondition that has to be met when assessing the implicit price of
each characteristic of a property is that the properties under research are of compa-
rable types. It is not useful to compare the implicit price of accessibility for an
office building with the implicit price of the same characteristic for housing. While
industrial sites are somewhat more heterogeneous than residential or office areas,
we argue that they are sufficiently homogeneous to apply hedonic price modelling.
This holds, as we make a distinction between different types of industrial sites
accommodating property ranging from heavy industrial to offices. A second consid-
eration that is taken into account here is the observation by Malpezzi (2003) that
practical issues, such as the functional form to use and the variables to include in
the analysis, often seem to be chosen ad hoc:
It is unfortunate, but the answer to ‘what does theory tell us about specification of
hedonic models?’is in brief: ‘not much’. Papers like Lancaster (1966) and Rosen
(1974) elegantly present models of housing characteristics without having much to
say about just what those characteristics are, or how exactly they are related to prices.
(Malpezzi, 2003, pp. 76–77)
2J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

Since a strong theoretical foundation for a functional form and the variables to
include in the analysis are lacking, studies from a different background are used;
hedonic pricing literature of individual (industrial) property, agglomeration econo-
mies literature, studies on industrial location and accessibility as well as policy
reports regarding industrial sites in the Netherlands.
2.2. Literature review: the location and valuation of industrial property
This section provides a brief overview of the literature on the factors that influence
the value and the location of industrial property in order to find possible explana-
tory variables that can be included in a model that aims at predicting the average
value of industrial sites. The underlying assumption is that some of the factors
explaining differences in individual property prices can also be used to help predict
differences in the average property value of an urban area. Location factors (for
instance, accessibility, agglomeration economies and distance to CBD) can also be
interpreted economically and studies on the underlying factors of industrial location
will therefore also be elaborated. Some of the studies elaborated here are studies or
policy reports of industrial sites in general. Although valuation is not the main
focus of these studies, they still might provide possible explanatory variables.
What characteristics are taken into account when hedonic price studies assess
the price of property depends on the type of property that is under research. In gen-
eral, property-specific characteristics are the most important. However, the charac-
teristics that are considered in housing studies are not necessarily relevant in
explaining the value of commercial property. Since our study will deal with indus-
trial sites, we will consider authors that have focused on industrial and commercial
property in more detail.
Hedonic studies of industrial property are mostly limited to studies undertaken
in the USA (Dunse & Jones, 2005a). Ambrose (1990), Black, Wolverton,
Warden, and Pittman (1997), Buttimer, Rutherford, and Witten (1997), Fehribach,
Rutherford, and Eakin (1993), Hoag (1980), Jackson (2002), Lockwood and
Rutherford (1996) and Ryan (2005) have all studied the prices of industrial prop-
erty, while Gunterman (1995) has studied the prices of industrial land. Physical
aspects of the property itself are found to be the most important explanatory
variables in the majority of these studies. Not surprisingly, size of the property is
found to be an important variable (Ambrose, 1990; Buttimer et al., 1997;
Fehribach et al., 1993; Lockwood & Rutherford, 1996). The same is true for the
age of a property, although some studies do not show significant results
(cf: Ambrose, 1990; Black et al., 1997; Buttimer et al., 1997; Dunse & Jones,
2005a; Dunse, Jones, Brown, & Fraser, 2004; Fehribach et al., 1993; Ryan,
2005; Sivitanidou & Sivitanides, 1995). A selection of other characteristics of
properties that are included in analyses are the number of dock-high doors
(Ambrose, 1990), the presence of sprinklers (Buttimer et al., 1997), the type of
tenant (Sivitanidou & Sivitanides, 1995) and the size of the office area within the
industrial property (Black et al., 1997). Unfortunately, our data-set does not
include these types of characteristics of individual property. The consequences of
this for our analysis will be elaborated shortly in Section 3.
Location characteristics are a second type of variables that are taken into
account in many studies of the price of property. Most of these can be regarded as
physical aspects of the (direct) environment of the property under research.
Journal of Property Research 3
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

The direct surroundings are important to owners of property as can be concluded
from studies on the location preferences of firms by, for instance, Pen (2002) and
STEC Groep (2005). From these studies, it can be concluded that firms prefer to be
housed in property that is located among property with a similar representativeness.
As a result, firms and their housing at a certain industrial site are often comparable
to a large extent. From the perspective of agglomeration externalities, the same con-
clusions can be drawn: locational behaviour of companies can be influenced by the
advantages agglomeration economies can offer. This can result in clustering and
specialisation (see e.g. Porter, 2000). De Vor and de Groot (2011) assume that
because of this the type of industrial site is a proxy for the appearance of an indus-
trial site since similar firms will tend to be located together. Sea ports can be con-
sidered an exception to this, because of the very distinct character of the firms
located there (PBL Netherlands Environmental Assessment Agency [PBL], 2008).
We will further elaborate on sea ports in Section 3.2.
Accessibility is another characteristic of the environment that, in a variety of
forms, has an influence on property values. Accessibility can be measured in
numerous ways. Lockwood and Rutherford (1996) and Sivitanidou and Sivitanides
(1995) show that the proximity of an airport has a positive effect on the price of
industrial property. Accessibility via road is also an important explanatory variable
for industrial property value (see e.g. Dunse et al., 2004; Sivitanidou & Sivitanides,
1995). Similar results for rental values of industrial property and regional accessi-
bility, in this case the proximity of a motorway junction, are reported in two studies
by Dunse and Jones (2005a,2005b) with the addition that proximity of motorway
access is only significant within 5 kilometres of the industrial property in their
data-set. Contrary to these findings, Ryan’s(2005) empirical work shows that a
location close to a freeway can be a disamenity for industrial property. Accessibility
by rail on industrial property prices is considered in many studies, although most
analyses show the influence is limited (see e.g. Ambrose, 1990; Black et al., 1997;
Lockwood & Rutherford, 1996; Ryan, 2005). A third form of accessibility has to
the best of our knowledge not been researched. This concerns water as a means of
transport. Ryan (2005) makes a brief suggestion that access to a port might be of
influence for industrial firms, but has not studied this. For industrial sites, the pres-
ence of water may be of importance since it creates an extra transport opportunity
for especially bulky or heavy goods.
In many studies, a factor that is commonly used to characterise the location of
property is the distance to a city centre or CBD. Based on Alonso’s(1964) seminal
work, this relation has been studied by many different authors. More sophisticated
approaches were developed by Di Pasquale and Wheaton (1996), Dunse et al.
(2004), and Dunse and Jones (2005b). These authors all find declining rental gradi-
ents from the CBD for industrial property. Lockwood and Rutherford (1996) and
Ryan (2005), on the other hand, do not find a significant relation between distance
from CBD and industrial property value. For industrial property, a location further
away from the CBD can mean better accessibility (because of less congestion
further away from the CBD) and as a result is not necessarily a negative location
characteristic. We will elaborate more on the location within the city and the conse-
quences for property values below.
Centrality and accessibility are two ways of looking at a location. A third
approach is to look at physical characteristics of a location that can influence the
value of a property. Dunse et al. (2004) provide an example of such a perspective.
4J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

From their study, they conclude that more prestigious location with better visibility
show higher property values. For industrial sites, this concerns mostly visibility
from a motorway (or other main road).
A location within an urban environment can have significant positive effects on
rental values. In their study of office rents, PBL (2009a)defined the urban environ-
ment based on the presence of amenities such as restaurants, theatres and shops.
For the industrial property in our database, unfortunately this information was
unavailable. Density was therefore chosen as a proxy for urbanisation rate of the
surroundings of the site. It is assumed that higher densities in the direct surround-
ings indicate a more urban environment. Note that this does not necessarily mean
that the industrial site is located close to the CBD.
Afinal physical characteristic that is elaborated here is not drawn from hedonic
pricing studies, but from policy reports on Dutch industrial sites. In practice, it is
observed that industrial sites can cause nuisance in the form of noise, risk, heavy
transport and pollution to neighbouring (residential) areas. Because of increasing
environmental regulation, the presence of housing can become a factor for the
industrial site itself since it affects the attractiveness of the location for certain firms
(CPB Netherlands Bureau for Economic Policy Analysis [CPB], 2001; Taskforce
Herontwikkeling Bedrijventerreinen, 2008). Measuring to what extent urban exter-
nalities influence property values remains a major issue that hedonics have tried
dealing with (Des Rosiers, Lagana, & Thériault, 2001). Nuisance is a factor that
can influence the attractiveness of property located at a certain location and for that
reason is taken into account when assessing the value of industrial sites.
In Hoag’s(1980) seminal work, financial and macroeconomic variables account
for the largest part of the explanatory power of the model used to explain the value
of industrial properties. Within macroeconomic variables, Hoag distinguishes
between regional and national variables. Other studies have followed this notion
and a variety of financial and economic variables are taken into account in almost
all hedonic price studies. This wide range includes proxies for economic circum-
stances, such as occupancy rates (Buttimer et al., 1997), manufacturing wage and
labour union strength (Black et al., 1997). Lockwoord and Rutherford (1996) have
set out to test whether both national and regional economic characteristics influence
industrial property value. Their results ‘do not support the inclusion of national
concomitants of value as hypothesised by Hoag’(Lockwood & Rutherford, 1996,
p. 269). Following this conclusion, in this study the emphasis will be on regional
economic characteristics.
Jackson (2002) differentiates between different (sub)counties in his study on
industrial property sales in southern California, indicating that agglomeration effects
are expected. Dunse and Jones (2005a)find similar differences between regional
property markets around Glasgow. Without specifying the exact spatial-economic
differences, de Vor and de Groot (2011) use a similar argumentation to predict
higher prices of residential property in the Randstad (the economic most important
region of the Netherlands) vs. a province outside the economic core region: ‘Since
the Randstad is the economic core region of the Netherlands, dwellings located in
this region are hypothesised to sell at a higher price than dwellings in
North-Brabant’(de Vor & de Groot, 2011, p. 615). Their results show that this is
indeed the case, providing an argument to include a region variable to control for
specific economical characteristics that are present at this level. Although it is
beyond the scope of this paper to discuss all regional effects in detail, one effect
Journal of Property Research 5
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

that we will highlight here is scarcity for industrial land. Sivitanidou and
Sivitanides (1995) have studied the relation between supply and rental levels of
industrial rents in Los Angeles and find that supply indeed influences rental levels.
According to Needham (1992), scarcity should be taken into account at a regional
level since the market for industrial land usually covers an area that is larger than a
municipality. Although it is not researched empirically, a relation is assumed
between the abundant provision of new industrial sites and negative effects such as
rapid decline of existing sites including high vacancy rates and declining property
values. The rationale behind this relation is that higher levels of scarcity for indus-
trial land make it more difficult for firms to move to another site since land is not
readily available. Instead, firms decide to invest in their existing premises (see
Louw, Needham, Olden, & Pen, 2004). The inclusion of a region variable will con-
trol for regional effects such as scarcity.
In addition to the influence of the economic characteristics of a region, there is
evidence that suggests that the size of the municipality is also a factor that should
be taken into account. In the case of industrial property in Dallas and Tarrant
county, a positive coefficient for the former is expected, ‘because Dallas is larger
and has larger manufacturing and distribution industries’(Fehribach et al., 1993,
p. 369). The size of the municipality in which the industrial site is situated should
thus be taken into account.
The distinction between regional and national economic factors was already
made above. Although we have concentrated on regional economic variables, there
are a number of studies in which year dummies are taken into account to control
for general economic trends. In her analysis of office and industrial rental values,
Ryan (2005) argues that the inclusion of a dummy variable for year of observation
adequately captures economic up- and downturns. Glascock, Jahanian, and Sirmans
(1990) and Wheaton and Torto (1994) use similar interpretations of year dummies
and find the expected significant results for these variables. From this, we conclude
that average industrial property values will also be affected by economic trends that
we will capture in a likewise fashion by including year dummies.
3. Empirical analysis: value of industrial sites
3.1. Introduction
In Section 2, various hedonic pricing studies and studies on industrial sites were
discussed. A large variety of characteristics that influence the value of individual
property or the functioning of industrial sites was presented. In this section, we
concentrate on the specification of the model that will be used to explain the value
of industrial sites.
We use ordinary least squares (OLS) regression on our data of the period 1997–
2008 to analyse the relationship between average property values and a number of
independent variables that are expected to influence this value.
1
Most studies that
use hedonic price modelling aim at explaining prices of property (rental, asking or
sales prices are the most commonly used variables). Collection of data on sales and
rentals is difficult since the amount of transactions of industrial property is rela-
tively low. Moreover, this data is scattered and not easily accessible. However,
appraisal data on property values, used for taxation purposes, are available. Based
on these values, the value per industrial site can be calculated.
6J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

The first step that was needed to be able to assign a value to individual sites
was to define industrial sites. The definition used here is based on the Dutch
national database on industrial sites, IBIS. It contains basic information, such as
surface area, land prices, available building land, geographical information on loca-
tion, etc., for every industrial site in the Netherlands. An industrial site is defined
as a site which, according to the land use plan, is suitable for the functions of trade
and industry and commercial and non-commercial services. This also includes a
limited number of (parts of) locations that are zoned and being used for offices
(IBIS, 2012). Dedicated office locations, typically located in or close to city centres
and railway stations, however, are not included in this definition. Due to privacy
reasons, industrial sites with fewer than five properties had to be excluded from our
analysis.
In Section 3.2, the dependent variable as well as the explanatory variables that
are based on the literature review above are presented. Table 1lists all variables as
well as the descriptive statistics and an operational definition for each variable. The
specification of the model is given in Section 3.3. Section 3.4 presents the results
of our empirical analysis.
3.2. Specification of variables
3.2.1. Dependent variable
For the construction of the dependent variable, data were obtained from CBS Neth-
erlands Statistics (CBS). Appraised values were derived from their database on
property taxation for the period 1997–2008. For every year, the total property value
of all property (except residential) for each industrial site was divided by its net
surface area in hectares,
2
which rendered the property value per hectare for every
year in the period under research. This value is referred to as the average industrial
property value. As was mentioned before, many hedonic pricing studies use rental
or selling prices as dependent variable. Appraisal values for taxation are believed
to be similar to such values, as the instruction for valuators states that the appraisal
value should represent transaction prices (Waarderingskamer, 2011).
3
It is therefore
assumed that appraisal value is a good indicator of the property values that are
commonly used in hedonic price studies, i.e. (listed) selling prices and rents. Der-
bes (2002) actually provides an argument to use appraisal value over transaction
prices:
(…) viable, profitable manufacturing plants seldom sell, since owners usually retain
them until they become unprofitable. When they do sell, the sale is usually a total
enterprise that includes machinery, equipment, patents and other intangibles. The
allocation of assets in such cases becomes extremely difficult. (Derbes, 2002, p. 40)
3.2.2. Explanatory variables
In Section 2, it was mentioned that in hedonic pricing literature the physical charac-
teristics of individual property normally account for a large part of the explanation
of the prices of individual property. Since the level of analysis here is the industrial
site, we focus on physical characteristics of the industrial site instead. The relation-
ships between various explanatory variables and the dependent variable will be
hypothesised here. All variables have reference categories as listed in Table 1.
Journal of Property Research 7
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

Table 1. Descriptive statistics of variables included in the model.
Variable name Operational description
Categories (when applicable)
Reference categories in bold Mean
Standard
deviation Source
Dependent variable
Average property
value
Average industrial property value (in €per
hectare) on site
–€
92,993
€179,430 CBS & IBIS
Explanatory variables
Physical characteristics of industrial sites
Type of industrial
site
Dummy variable for type (and proxy for
appearance) of industrial site
Mixed-use 69% –LISA & IBIS
Industry 17%
Transport 6%
Consumer services 5%
Financial and business services 2%
Miscellaneous 1%
Sea port Dummy variable for sea port –––IBIS & PBL
Accessibility road Travelling time (in minutes) to nearest motorway
exit
–6.5 5.4 NAVTEQ
Accessibility
public transport
Distance in metres to nearest bus stop –409 347 OV Reisinformatie
Accessibility water Amount of 10 x 10 metre grid cells of water
within 500 metre radius
–424 621 Top 10 Vector
(Kadaster)
Located along
motorway
Dummy variable. Equals 1 when motorway
intersects with industrial site
–––Nationaal
Wegenbestand &
IBIS
Distance from
centre
municipality
Distance in metres from centre of municipality in
which industrial site is located
–2664 1983 CBS
Distance from
CBD
Distance in metres from nearest CBD (22 largest
urban agglomerations)
–19,270 14,864 CBS
Age Dummy variable for the decade in which
industrial site was developed
1950s and before 24% Topographical
maps1960s 19%
(Continued)
8J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

Table 1. (Continued).
Variable name Operational description
Categories (when applicable)
Reference categories in bold Mean
Standard
deviation Source
1970s 18%
1980s 22%
1990s 14%
2000s 2%
Nuisance housing Amount of 10 x 10 metre grid cells with land use
‘housing’within 500 metre radius
–1487 1409 CBS
Nuisance open
space
Amount of 10 x 10 metre grid cells with land use
‘open space’within 500 metre radius
–2777 1967 CBS
Regional economic characteristics
Region Dummy variable for part of the country in which
the industrial site is located
Centre (Randstad) 30% –PBL
Intermediary zone 33%
Periphery 37%
Density Number of addresses within 1 kilometre radius –664 709 CBS
Urbanisation rate Dummy variable for type of municipality in
which industrial site is located
Urban agglomeration 19% –CBS
Suburban 17%
Other 64%
Economic trends
Year Dummy variable for year 1997–2008 ––
Journal of Property Research 9
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

With a few exceptions, all explanatory variables were calculated for every year
in the period under research. This is necessary for variables such as nuisance, type
of industrial site, density and accessibility since these are subject to change. The
variables region, urbanisation rate and age (defined as the period in which an indus-
trial site was developed, see below) are (much more) static and are assumed not to
have changed within the period under research. For the type of industrial site, for
example, yearly values were calculated by linking the geographical information in
IBIS to a database with addresses of individual firms which contains information
on the type of economic activity of the company. Based on the types of firms
located at an industrial site, a classification of types of industrial sites was made for
every year. Similar methods were applied to calculate yearly values for other vari-
ables.
Not all variables were available for every year in the period under research.
For the variables accessibility, water, centrality and located along motorway data
was available for the years 1997, 2002 and 2008. The values for missing years
were calculated assuming a moving average when changes had taken place between
two years, to allow for yearly analysis with all variables included.
3.2.3. Physical characteristics of the industrial site
Physical characteristics of the industrial site that are included in the analysis are
‘type of industrial site’,‘seaport, ‘accessibility’,‘nuisance’,‘centrality’,‘age’and
an interaction variable of ‘type of industrial site’and ‘accessibility’.
Six different types of industrial sites are distinguished: ‘industry’,‘mixed-use’,
‘transport’,‘consumer services’,‘business and financial services’and ‘miscella-
neous’.
4
Mixed-use industrial sites are the reference category for this variable. We
expect that ‘business and financial services’,‘consumer services’and ‘miscella-
neous’will have a higher average property value vis-à-vis ‘mixed-use’sites as
firms that are typically located on these types of industrial sites will have higher
demands regarding accessibility, representativeness and other value-adding charac-
teristics. The opposite is expected to be true for owners of property at ‘industry’
and ‘transport’sites. Therefore, the coefficients for these categories are expected to
show a negative sign.
Secondly, a dichotomous variable is added to the model to control for sea ports.
The identification of sea ports is based on the same existing database that was men-
tioned above (additional corrections were carried out by researchers from PBL
Netherlands Environmental Assessment Agency to identify sea ports more thor-
oughly). Sea ports do not belong to one of the categories of types of industrial sites
since the firms located at sea ports could be anything ranging from transport to
heavy industrial. Sea ports could thus be characterised as either transport, industry
or mixed use. Still, sea ports have certain distinct characteristics (such as a low
density) that influence the average property value that can be controlled for via the
inclusion of this variable.
The third explanatory variable in our model is ‘accessibility’. Three categories
are distinguished. Accessibility by road is measured in travelling time in minutes to
the nearest motorway exit.
5
The second category is ‘accessibility by public trans-
port’. This category is defined as distance in metres from the nearest bus stop. An
increasing distance from a motorway exit or bus stop is expected to have a negative
effect on the average industrial property value, which means that a negative sign is
10 J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

expected for both categories of this variable. A third category is presence of water.
The amount of open water (measured in 10 x 10 metre grid cells) within a 500
metre radius of the industrial site is determined using topographical maps. The
presence of water is expected to affect average property values positively since it is
seen as a proxy for an extra means of transport.
‘Located along a motorway’is the next variable included in the model. This
dichotomous variable takes value 1 if the industrial site intersects with a motorway.
Note that a location along a motorway does not automatically mean that the loca-
tion is very well accessible; the exit of the motorway does not necessarily have to
be nearby.
6
The location of property relative to a central place, such as a city centre, is studied
by different authors. For industrial property the relationship is not always straightfor-
ward. Although a central location can have advantages, such as agglomeration effects,
for industrial sites, being located further away from the centre of a municipality might
have advantages when it comes to, e.g., accessibility. This ambiguity seems less
apparent for offices or residential property. Ryan’s study seems to confirm this: ‘In
terms of industrial properties [as opposed to offices], the overall results demonstrate
that transport access may be weaker than localisation benefits’(Ryan, 2005, p. 763).
Dunse and Jones (2005a,2005b) found a negative correlation between industrial rent
and distance from nearest large town. Following these conclusions, we expect a nega-
tive sign for both distance to CBD and centre of municipality.
Topographical maps were also used to determine the age of an industrial site.
The present location of every industrial site was researched on historical topograph-
ical maps that were updated roughly every decade. Whenever the historical maps
showed the present location as developed, it gave us information on the decade in
which the industrial site was first developed.
7
The variable ‘age’is thus defined as
the decade in which the industrial sites were first developed. Six different age cate-
gories were used ranging from ‘1950s and before’to ‘2000s’, with evenly distrib-
uted age brackets in between.
The physical characteristic ‘nuisance’is included. Whether an industrial site is
hindered by functions surrounding it is determined by looking at the land uses sur-
rounding the industrial site. For every industrial site, the presence of the land uses
‘housing’and ‘open space’(again in 10 × 10 metres grid cells) within a 500 metre
radius of an industrial site was determined. Higher levels of housing surrounding
the industrial site are expected to have a negative relation with average property
value. Conversely, a positive sign is expected for the category ‘open space’.
The final physical characteristic that was included is an interaction variable of
‘type of industrial site’and ‘accessibility via road’. The importance of accessibility
can differ between types of firms. It is expected that the relationship between acces-
sibility and property value will be stronger for industrial sites dominated by logisti-
cal firms than for other uses since these firms appreciate accessibility better than
other types of firms. This variable was added to research the different influence that
accessibility may have on the value of properties located on the various types of
industrial sites.
3.2.4. Regional economic characteristics
Regional economic variables included in the model are ‘region’,‘urbanisation rate’
and ‘density’. As mentioned before, we distinguish between three different regions
Journal of Property Research 11
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

in the Netherlands based on economic significance: Randstad (the economic heart-
land of the Netherlands), an intermediary zone and the periphery (see Figure 1).
This division is based on quantile scores of the number of jobs in municipalities.
Figure 1. Distinction between three regions in the Netherlands (Source: LISA, 2007).
Visualisation by PBL Netherlands Environmental Assessment Agency. Municipal boundaries
from 2008 were used.
12 J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

For an earlier version and more information on the methods see van Oort (2004).
Considering the economic difference between these three parts of the Netherlands,
it is expected that the average industrial property values will be highest in the
central region, followed by the intermediary zone and the periphery, respectively.
Also, the ‘urbanisation rate’is included in the model. Three different categories
are distinguished: urban agglomeration, suburban and other. CBS provides the defi-
nitions that were used to characterise the municipalities. CBS has defined 22 urban
agglomerations throughout the Netherlands, based on morphological characteristics,
population and amount of jobs (for the exact definitions see CBS, 2005). The
municipalities that are located just outside the urban agglomeration are labelled as
suburban. The remaining municipalities are classified as other municipalities. Aver-
age industrial property values are believed to be highest in urban agglomerations,
followed by suburban and other municipalities respectively.
The final regional economic variable that is included in the model is ‘density’.
Again, this variable is defined by CBS and measures the number of addresses
within a 1 kilometre radius of the industrial site. An average density of addresses
was calculated for every industrial site, using the method mentioned before.
3.2.5. Economic trends
A year dummy was included in the model to control for general economic trends.
Economic trends for industrial sites in the Netherlands have been described by
Louw et al. (2004). Following their study, we expect average industrial property
values to increase annually compared to the reference year 1997. Furthermore,
based on the data of sales of industrial property (Vastgoedmonitor, 2012), we
expect a bust in the coefficients for the years 2003–2006.
3.3 Model specification
The above leads to the following functional form of our model (see Table 1for a
description of variables):
lnPVit ¼^
aþX
N
n¼1
^
baPCit þX
N
b¼1
^
cbRECrt þETt
where
lnPVit Natural logarithm of the average industrial property value (in €per hectare)
on site (i) in year (t);
PN
n¼1
^
baPCit Physical characteristics of site (i) in year (t) (type of industrial site, sea
port, accessibility road, accessibility public transport, accessibility water,
located along motorway, distance from centre municipality, distance from
CBD, age, nuisance housing, nuisance open space, type of industrial site x
accessibility via road);
PN
b¼1
^
cbRECrt Regional economic characteristics in region (r) in year (t) (region,
urbanisation rate, density);
ETtEconomic trend in year (t).
The dependent variable was transformed using a natural logarithm. Although
the literature is ambiguous on which method should be applied as this depends on
the specific situation (Malpezzi, 2003), for this study a semi-logarithmic model was
Journal of Property Research 13
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

chosen. The main reason for this is that it allows easy interpretation of the coeffi-
cients of the explanatory variables. These can be interpreted as the change in value
in terms of percentages when the explanatory variable increases with one unit (for
more on the advantages of a semi-logarithmic model see: PBL, 2009a). Multicollin-
earity is a common problem when applying hedonic price models. To test for this,
we follow Belsley, Kuh, and Welsch (2004) and calculate the variance inflation fac-
tors of all explanatory variables included in the model. The results of this analysis
show that our findings will not suffer from multicollinearity.
3.4. Results of the analysis
In this section, the results of our OLS model will be presented. The individual
coefficients of the variables are listed in Table 2, along with the change in average
property value in terms of percentages. Note that although year dummies were
included in the model to account for economic trends, these were excluded from
presentation to increase the readability of Table 2. The coefficients of the year
dummies are discussed below.
Our model includes more than 27,000 cases for the period under research. The
variables included in our model explain 37% of the variation of the average indus-
trial property value on the industrial sites in our data-set. The overall explanatory
value of the model is .370 (adjusted R
2
) which is considered reasonable taking into
consideration property-specific characteristics are not included (see Nitsch, 2006 for
a similar argumentation in a study on the influence of location factors on office
rents). A first closer look at our explanatory variables shows that almost all vari-
ables are significant at the 99% level. At this level, there is sufficient evidence that
the variables ‘type of industrial site’‘water’,‘age’,‘nuisance’,‘region’and ‘den-
sity’are significant in explaining the average property value of industrial sites. The
variables ‘accessibility’,‘sea port’and ‘located along motorway’are significant at
the 95% level. Two of the interaction variables show to be significant at the 95%
level. In terms of the expected signs, only nuisance related to the presence of hous-
ing does not show the expected sign. The variables that do not significantly influ-
ence the average property value are ‘urbanisation rate’and both categories of the
centrality variable. Below, some of the most notable results are discussed in more
detail.
The coefficients of the year dummies show the expected signs and patterns.
Coefficients roughly show the expected signs, although compared to the reference
year 1997 the first three years show a negative sign. The years in which periodic
appraisals (2001, 2005 and 2007) were done can be recognised in shocks in the
coefficients. Although the coefficients do not show the expected bust in the years
2003–2006, they appear to reflect inflation and corrections in value due to
extensions.
If we take a closer look at the individual explanatory variables, the first notable
results are the large differences in average value in terms of percentages between
the categories of the variable ‘type of industrial site’. Miscellaneous sites have the
highest property values, with a difference of 209% compared to mixed use sites.
Sites that are dominated by consumer services and financial and business services
show a difference of 144% and 121%, respectively. Also, transport sites and sites
dominated by industrial firms show higher average property values than mixed-use
sites, with more than 67% and 12%, respectively, although a negative sign was
14 J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

Table 2. Results of OLS regression.
Dependent variable: Natural logarithm of average industrial property value (per hectare) on
industrial site
B t % value
Constant 9.311 96.73 –
Physical characteristics of industrial sites
Type of industrial site (0/1)
Mixed industrial-business (ref) ––
Industry .120
**
2.340 12.7%
Transport .514
*
7.620 67.2%
Consumer services .890
*
9.220 144%
Financial and business services .794
*
8.140 121%
Miscellaneous 1.130
*
7.440 209%
Sea port (0/1) −.356
**
−2.130 −30.0%
Accessibility
By road (in minutes) −.016
*
−2.890 −1.6%
By public transport (in km) −.107
*
−2.870 −10.2%
Water (ha within 500 m radius) .026
*
8.260 2.6%
Located along motorway (0/1) .107
**
2.210 11.3%
Centrality
Distance from centre municipality (in km) .088 .094 –
Distance from nearest CBD (in km) −.001 −.850 –
Age (0/1)
1950s and before −.671
*
−11.690 −48.9%
1960s −.488
*
−8.680 −38.1%
1970s −.426
*
−7.630 −34.7%
1980s −.212
*
−4.180 −19.9%
1990s (ref) –––
2000s .322
*
3.520 38.0%
Nuisance
Presence of housing (ha within 500 m radius) .034
*
19.300 3.5%
Presence of open space (ha within 500 m radius) .029
*
22.350 2.9%
Density (addresses per ha) .015
*
3.910 1.5%
Type of industrial site x accessibility by road
Mixed industrial-business –accessibility by road (ref) –– –
Industry –accessibility by road .010 1.580 –
Transport –accessibility by road −.004 −.550 –
Consumer services –accessibility by road −.028
*
−2.920 −2.8%
Financial and business services –accessibility by road −.026
**
−2.110 −2.6%
Miscellaneous –accessibility by road −.018 −1.040 –
Regional economic characteristics
Region (0/1)
Centre (Randstad) .255
*
5.570 29.0%
Intermediary zone (ref) –––
Periphery −.164
*
−3.960 −15.1%
Urbanisation rate (0/1)
Urban agglomeration .088 1.450 –
Suburban −.056 −1.170 –
Other (ref) –––
Notes: Standard errors are corrected to account for multiple observations of industrial sites over time.
N= 27,141. Adj. R
2
= .370.
*
Significant at the 99% level.
**
Significant at the 95% level.
Journal of Property Research 15
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

expected for these two categories. The small number (only 2%) of miscellaneous
sites in the data-set might be an explanation for the relatively high value this cate-
gory shows. The method of appraisal (see note 3) that is most commonly used for
properties typically located at this type of industrial site might also have added to
the high value for this particular category. Mixed-use industrial sites are plentiful in
the Netherlands (this type makes up almost 70% of the sites in the dataset) and this
type represents by far the lowest average property values. Specialisation and clus-
tering of similar economic activities on industrial sites appears to be related to
higher average property values, indicating agglomeration externalities. Higher aver-
age property values are frequently mentioned as a goal in redevelopment plans for
declining urban areas. This means that for policy-makers it can be interesting to
regulate the types of firms to be located at a certain site when developing new sites
or redeveloping existing ones.
The coefficients and corresponding changes in value of the accessibility vari-
ables are somewhat contrary to the conclusions by Ryan (2005). Accessibility by
public transport and road both show the expected negative signs, indicating that
access indeed significantly influences the average value of industrial sites positively,
whereas Ryan’s results suggested a negative relationship between accessibility via
road and property value. However, our results are similar to the conclusions on the
influence of accessibility via road by e.g., Sivitanidou and Sivitanides (1995), Dun-
se et al. (2004) and Dunse and Jones (2005a). Accessibility via water has a posi-
tive, although small, influence on average property value. This appears to indicate
that firms value this extra form of accessibility.
As was expected, visibility from a motorway has a positive influence on the
average value of an industrial site. When new industrial sites are planned, a loca-
tion that is visible from a motorway can be favourable from the perspective of
achieving the highest property values possible (for that reason, industrial firms usu-
ally pay higher prices for purchasing the land). A downside to this is that locations
that are visible from the motorway are often greenfield locations. For the Nether-
lands, developing new industrial sites on greenfield locations adds to the loss of
open space and is believed to have a harmful effect on the landscape (PBL,
2009b).
Both categories of centrality, distance from CBD and distance from centre of
municipality, do not show significant results. A possible explanation is the overlap
between these variables and density. Also, the linkages between location and acces-
sibility, as elaborated by Ryan (2005), are interesting to research more thoroughly.
The coefficients of the distinguished age classes show the expected signs and
pattern. This is in line with hedonic pricing studies on individual property, where
age is an important explanatory variable in many studies (see, e.g., Buttimer et al.,
1997; Dunse et al., 2004; Jackson, 2002). From the results it can be concluded that
the earlier the industrial site is developed, the lower the average property values
are. The changes in value are not completely similar between decades, indicating
that industrial sites from certain decades have lower average property values. An
important explanation for lower average property values on older sites is the
decrease in value because of decline.
8
Industrial sites developed in the 1970s show
relatively high property values compared to sites developed in the 1960s with only
4% difference in value between these two decades. The difference between sites
from the 1970s and 1980s however is almost 15%. The same goes for industrial
sites developed in the 1980s with almost 20% difference in average property value
16 J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

compared to the reference category 1990s. This indicates that especially industrial
sites from the 1970s and 1980s represent relatively low average property values.
We do not have a proper explanation for this, other than that we can speculate that
the initial quality of the buildings that was built in this time was significantly lower
than in earlier and later periods.
The signs of the variables that were included in the model to research the influ-
ence of nuisance are ambiguous. Both land use ‘open space’and ‘housing’have a
positive effect on the dependent variable. However, the expectation was that the
presence of housing and the nuisance for residents could turn into a negative loca-
tion aspect for firms, reflected in lower average property values. More thorough
research of the relationship between the average property value of an industrial site
and surrounding land uses is needed to get more understanding of how these affect
one another.
The region in which an industrial site is situated has a significant effect on the
average property value. In line with the expectations, values are highest in the
Randstad, followed by the intermediary and peripheral zones, respectively. Also,
density shows the expected sign; a higher number of addresses in the vicinity of
the industrial site has a positive effect on the average property value of the site.
Finally, for the urbanisation rate we did not find a significant relation with the
dependent variable. The regression results for this variable indicate that average
property values for industrial sites do not differ between urban agglomerations, sub-
urban municipalities and other municipalities. The differences between regions
appear not to be present at this spatial level, indicating that a location in an urban
agglomeration, vis-à-vis suburban and other municipalities, does not have the same
advantages as a location in the Randstad, vis-à-vis the intermediary and peripheral
zone.
The interaction variables of type of industrial site and accessibility by road were
included in the model to research how the value of different types of industrial sites
benefits from accessibility. The coefficients of consumer and financial and business
services are the only two interaction terms that show a significant difference from
the reference category ‘mixed industrial-business'. Both categories show a negative
sign, indicating that for these two types of industrial sites the effect of reduced
accessibility on average property value is more severe than for mixed industrial-
business. The effects of accessibility on property value for industry, transport and
miscellaneous site are comparable to those for mixed industrial-business, although
it was expected that accessibility would be more important for transport sites as
compared to the reference category.
Summarising the most important conclusions, we conclude that the highest aver-
age property values of industrial sites in the Netherlands can be found in the
Randstad. Sites dominated by public services, hospitals and educational facilities
(i.e. miscellaneous sites) show the highest average values, followed by sites that
are characterised as consumer services sites and financial and business services
sites. A location along a motorway leads to an average 11% higher property value
as compared to sites that are not visible from the motorway. Industrial sites that
were developed recently show the highest average property values and the coeffi-
cients indicate that the value of industrial sites decreases with age. The influence of
other functions close to industrial sites is ambiguous, with both the presence of
housing and open space showing a positive influence on the property value of an
industrial site. The analysis shows mixed results when it comes to differences in
Journal of Property Research 17
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

the contribution of accessibility via road between different types of industrial sites
with most of the interaction variables showing insignificant coefficients. Firms that
attract customers such as consumer and financial and business services do appear
to appreciate accessibility.
4. Conclusions and further research
4.1. Conclusions
Bryson (1997) argues that urban areas represent a certain value which may change
over time. The first goal of this paper was to analyse which variables influence this
value of urban areas. To identify relevant variables, we used a method inspired by
classic hedonic pricing studies. Moreover, we wanted to test whether this method
that is conventionally used in studies of individual property can also be applied in
a meaningful way at the level of urban areas, notably industrial sites in the
Netherlands.
In our study, we considered the average property value of industrial sites to be
an adequate representation of the value of this type of urban area. We used OLS
regression to identify the relevancy of three categories of explanatory variables in
the period 1997–2008: physical characteristics of the industrial site, regional eco-
nomic characteristics and general economic trends. The selection of variables to be
included in our model was based on the hedonic price literature of (industrial)
property.
By and large, the results of the analysis appear to be in line with our hypothe-
ses that were based on existing hedonic pricing literature. Both regional economic
characteristics and physical characteristics of the urban area itself have the hypothe-
sised significant influence on the value of the industrial sites under research. From
this, we conclude that the findings of this paper generate meaningful insight into
the factors that influence the average property value of industrial sites. Our most
notable findings in line with prior literature are with regard to the variables ‘age’,
‘accessibility by road’and ‘located along motorway’. Furthermore, the expected
regional economic differences in average property value are reflected in the coeffi-
cients of the three distinguished regions of the Netherlands. Apart from the afore-
mentioned confirmatory findings, we also obtained outcomes that differ from what
we would have expected based on prior literature. The variables ‘centrality’(both
distance to CBD and centre municipality) and ‘urbanisation rate’do not signifi-
cantly influence property values. The results of both of these variables could be
studied more extensively in relation to the accessibility of a location. Although not
completely in line with our expectations, the results of the analysis for the variable
‘type of industrial site’add new insights to the existing literature: the influence that
the composition of firms located on an industrial site can have on the average prop-
erty value has not been researched much in urban and property literature. Literature
on agglomeration economies could help to further interpret these findings.
‘Mixed-use’sites show the lowest average property values and at the same time
make up a substantial part (69%) of all industrial sites in the Netherlands. Speciali-
sation in terms of the composition of firms located on an industrial site appears to
have a positive influence on the average property value with specialised types of
industrial sites, such as ‘transport’and ‘consumer services’, showing significantly
higher average property values.
18 J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

These insights are useful for policy-makers involved in urban regeneration and
are applicable to urban areas that face decline. For now, we have chosen to use
industrial sites in the Netherlands as a case study, since industrial sites in the
Netherlands have received considerable attention because of rapid decline (van der
Krabben & Buitelaar, 2011). Keeping property values at a high level is often an
explicit goal of urban policies such as regeneration. The outcomes of this study can
be useful for policy-makers since it provides insight into the factors that affect the
value of an urban area as a whole, contrary to a more narrow focus on factors that
influence individual property value. This broader perspective can be of particular
interest since it provides information on property value at a spatially relevant level,
i.e. the level at which the outcomes of processes such as urban decline and gentrifi-
cation are notable and could be influenced. For instance, via the aforementioned
specialisation of types of firms to be located at newly developed or regenerated
industrial sites.
4.2. Directions for further research
This study shows important relations between locational characteristics and regional
economic characteristics on the one hand, and average property value of designated
urban areas on the other hand. For this study, industrial sites in the Netherlands
have been used as a case study to test the proposed method and expected out-
comes. Further research could concentrate on other urban areas that face processes
such as decline and gentrification. More insight into what causes the differences in
value between urban areas could result in more successful regeneration policies for
example, or could even help to prevent urban areas from declining.
Future work will be aimed at the construction of an index, based on property
values, that allows us to gain insight into the process of decline of urban areas.
While constructing this indicator there are a few factors that will have to be taken
into account. The idea that neighbourhoods go through life cycles was introduced
by Hoover and Vernon (1959) and recently applied theoretically to industrial sites
in the Netherlands by Louw et al. (2004). If we assume that during the phases of
the life cycle the average property value of an industrial site increases with growth,
stabilises over time and eventually falls, this would mean that we must consider in
more detail the changes in average property values. In this regard, it seems safe to
assume that the average property value of an industrial site will be at its lowest
point at the moment that the site has reached the end of its life cycle. However,
one could argue that decline is most severe right after the phase of stabilisation,
when the average property value (according to the life cycle) will fall sharply.
These and other challenges will also be topics of future research.
Acknowledgements
Our own calculations on property values are based on taxation data (WOZ) by
CBS Netherlands Statistics (CBS), provided by Dutch municipalities. These data
were linked to individual industrial sites using the national employment database
(LISA). This paper is part of an ongoing research programme and received
financial assistance from the Netherlands Institute for Cities and Innovation Studies
(NICIS). We are grateful for comments by Friso de Vor, Edwin Buitelaar and Jan
Schuur on earlier versions of this paper.
Journal of Property Research 19
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

Notes
1. To counter the problem of the locational coefficients that are not fixed over time, a dis-
tinction was made between two time periods. The analysis shows that the differences in
coefficients for the locational characteristics between these periods are small, so we can
assume that they are indeed fixed over time.
2. IBIS lists different measures for the surface of an industrial site. The one used here to
calculate the average property value of a site is the surface that is actually in use by
firms. This thus excludes public space, roads and land that was serviced, but was not
yet sold to end users.
3. The valuation techniques for all property that is appraised for taxation under the so-
called Real Property Act (Wet Onroerende Zaken or WOZ) which was first introduced in
1997 will be elaborated here shortly. Since 2007 property is appraised annually, from
1997 till 2004 this was done once every four years and between 2004 and 2006 the
appraisal was done bi-annually. During these periods, the appraised value was only cor-
rected for inflation and not for redevelopment or extensions. For industrial and business
property, the method used for appraisal is based on known transactions of similar prop-
erties within the municipality. Some types of industrial property with few transactions
(for example, energy plants, hospitals, waste processing plants, schools, etc.) are
appraised by calculating the replacement costs of the property. Although this method
will result in property values which are not affected by obsolescence (nor deterioration),
this will not influence our results since they only make up a small number in our data-
base. There are two reasons for this. First, outliers were defined and a considerable
number of these types of properties have (very) high replacement costs. Second addi-
tional research shows that only a very small percentage of the properties are schools,
hospitals and the like. Results of this analysis are available upon request.
4. Although most types are straightforward, the categories ‘mixed-use’and ‘miscellaneous’
need some extra explanation. ‘Mixed-use’refers to a mix of different types of
businesses, rather than a mix of different land uses, such as residential, shopping, etc.
‘Miscellaneous’refers to a final category of industrial sites that, based on the types of
firms located there, mostly houses schools, hospitals and government services.
5. This is confirmed by a weak correlation (−.36) between the explanatory variables
‘accessibility by road’and ‘located along motorway’.
6. The travelling time is calculated as the travelling time from every individual property
divided by the total amount of properties. This is called the average travelling time (of
the industrial site). The same goes for all variables that are defined in terms of distance,
unless mentioned otherwise.
7. Note that on some sites developments may continue after this date. Also, redevelopment
may have taken place making new developments possible. Still, the decade in which
the first development took place is a good indicator for the age of the industrial site as
a whole.
8. Although the relation between decline and decrease in value can be debated extensively,
here we assume that age in general is the most important driver of decline.
Notes on contributors
Jasper Beekmans is a PhD student at the Institute for Management Research, Department of
Geography, Planning and the Environment at the Radboud University Nijmegen, the
Netherlands. After obtaining his MSc degree in Spatial Planning at the Radboud University
Nijmegen, he joined the department in October 2007. He is involved in a NICIS-funded
research programme concerning (re)development of industrial sites. His research is focused
on explaining the differences in decline between industrial sites.
Pascal Beckers is an international business economist and human geographer, with special
expertise in Economics, Migration and Integration, and Urban Studies. He has both aca-
demic as well as professional experience in these fields. His expertise in Economics, Migra-
tion and Integration and Urban Studies stems from his PhD work on migrant integration in
the Netherlands, his current function at the PBL Netherlands Environmental Assessment
20 J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

Agency, and his current function as assistant professor in Geography and Planning at the
Radboud University Nijmegen, the Netherlands.
Erwin van der Krabben holds a position as professor of Planning and Property Development
in the Institute for Management Research, Department of Geography, Planning and the Envi-
ronment at the Radboud University Nijmegen and a position as professor of real estate in
the School of the Built Environment, University of Ulster, Belfast. He has published in a
variety of academic journals with respect to land policy, area development and the regula-
tion of land and property markets. He has acquired substantial funds for scientific research
and supervises a number of PhD projects in these academic fields.
Karel Martens is an associate professor in Transport Planning at the Institute for Manage-
ment Research, Department of Geography, Planning and the Environment, Radboud Univer-
sity Nijmegen, the Netherlands. He has nearly 20 years of experience as an academic and
practitioner in the fields of transportation planning and urban planning in the Netherlands,
Israel, Belgium and, recently, the USA. His topics of expertise include parking modelling
and policy, the nexus between transport and land use, multi-modal travel and participatory
governance. He supervises a number of PhD projects in these academic fields.
References
Adair, A., Berry, J., McGreal, S., Hutchison, N., Watkins, C., & Gibb, K. (2003).
Urban regeneration and property investment performance. Journal of Property
Research, 20, 371–386.
Alonso, W. A. (1964). Location and land use: Toward a general theory of land rent.
Cambridge, MA: MIT Press.
Ambrose, W. (1990). An analysis of the factors affecting light industrial property valuation.
The Journal of Real Estate Research, 5, 355–369.
Belsley, D. A., Kuh, E., & Welsch, R. E. (2004). Regression diagnostics: Identifying influen-
tial data and sources of collinearity. New York: John Wiley & Sons.
Black, R., Wolverton, M., Warden, J., & Pittman, R. (1997). Manufacturing versus
distribution: Implicit pricing of real property characteristics by submarket. Journal of
Real Estate Finance and Economics, 15, 271–285.
Bryson, J. M. (1997). Obsolescence and the process of creative reconstruction. Urban
Studies, 24, 1439–1458.
Buttimer, R., Rutherford, R., & Witten, R. (1997). Industrial warehouse rent determinants in
the Dallas/Fort Worth area. Journal of Real Estate Research, 13,47–55.
CBS Netherlands Statistics. (2005). Grootstedelijke agglomeraties en stadsgewesten afge-
bakend [Delineation of urban agglomerations and city regions]. Voorburg: Author.
CPB Netherlands Bureau for Economic Policy Analysis. (2001). Veroudering van bedrijven-
terreinen: een structuur voor herstructurering [Decline of industrial sites. A framework
for redevelopment]. The Hague: Author.
de Vor, F., & de Groot, H. (2011). The impact of industrial sites on residential property val-
ues: A hedonic pricing analysis from the Netherlands. Regional Studies, 45, 609–623.
Derbes, M. J. (2002). Non-comparable industrial sales. The Appraisal Journal, 70,39–45.
Des Rosiers, F., Lagana, A., & Thériault, M. (2001). Size and proximity effects of primary
schools on surrounding house values. Journal of Property Research, 18, 149–168.
Di Pasquale, D., & Wheaton, W. (1996). Urban economics and real estate markets.
New York, NY: Prentice Hall.
Dunse, N., & Jones, C. (1998). A hedonic price model of office rents. Journal of Property
Valuation & Investment, 16, 297–312.
Dunse, N., & Jones, C. (2005a). Rental depreciation, obsolescence and location: The case of
industrial properties. Journal of Property Research, 22, 205–223.
Dunse, N., & Jones, C. (2005b). UK roads policy, accessiblity and industrial property rents.
In D. Adams, C. Watkins, & M. White (Eds.), Planning, public policy & property
markets (pp. 128-147). Oxford: Blackwell.
Dunse, N., Jones, C., Brown, J., & Fraser, W. D. (2004). The spatial pattern of industrial
rents and the role of distance. Journal of Property Investment & Finance, 23, 329–341.
Journal of Property Research 21
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

Fehribach, F., Rutherford, R., & Eakin, M. (1993). An analysis of the determinants of
industrial property valuation. The Journal of Real Estate Research, 8, 365–371.
Glascock, J. L., Jahanian, S., & Sirmans, C. F. (1990). An analysis of office market rents:
Some empirical evidence. American Real Estate and Urban Economics Association
Journal, 18, 105–119.
Gunterman, K. L. (1995). Sanitary landfills, stigma and industrial land values. The Journal
of Real Estate Research, 10, 531–542.
Hoag, J. (1980). Towards indices of real estate value and return. The Journal of Finance,
35, 569–580.
Hoover, E. M., & Vernon, R. (1959). Anatomy of a metropolis: The changing distribution of
people and jobs within the New York metropolitan region. Cambridge, MA: Harvard
University Press.
IBIS. (2012). Toelichting begrippen IBIS [Clarification of concepts used in IBIS].
The Hague: Ministry of Infrastructure and the Environment.
Jackson, T. (2002). Environmental contamination and industrial real estate prices. Journal of
Real Estate Research, 13, 179–199.
Lancaster, K. (1966). A new approach to consumer theory. Journal of Political Economy,
74, 132–157.
Lockwood, L., & Rutherford, R. (1996). Determinants of industrial property value. Real
Estate Economics, 24, 257–272.
Louw, E., Needham, B., Olden, H., & Pen, C. J. (2004). Planning van bedrijventerreinen
[Planning of industrial sites]. The Hague: Sdu Uitgevers.
Malpezzi, S. (2003). Hedonic pricing models: A selective and applied review. In T. O.
Sullivan & K. Gibbs (Eds.), Housing economics and public policy: Essays in honor of
Duncan Maclennan (pp. 68-89). Oxford: Blackwell.
Needham, B. (1992). A theory of land prices when land is supplied publicly: The case of
the Netherlands. Urban Studies, 29, 669–686.
Nitsch, H. (2006). Pricing location: A case study of the Munich office market. Journal of
Property Research, 23,93–107.
Páez, A., Long, F., & Farber, S. (2007). Moving window approaches for hedonic price
estimation: An empirical comparison of modelling techniques. Urban Studies, 45,
1565–1581.
PBL Netherlands Environmental Assessment Agency. (2008). Werkgelegenheidsgroei op
bedrijventerreinen [Employment growth on industrial sites]. The Hague: Author.
PBL Netherlands Environmental Assessment Agency. (2009a). De waarde van de kan-
tooromgeving [The value of the surrounding environments for offices]. The Hague:
Author.
PBL Netherlands Environmental Assessment Agency. (2009b). De toekomst van bedrijven-
terreinen: van uitbreiding naar herstructurering [The future of industrial sites: From
extension to redevelopment]. The Hague: Author.
Pen, C. J. (2002). Wat beweegt bedrijven? Besluitvormingsprocessen bij verplaatste
bedrijven [What makes companies move? Decision making processes of relocated firms].
Utrecht: Royal Dutch Geographical Society (KNAG)/Faculty of Spatial Sciences Univer-
sity of Groningen.
Ploegmakers, H., & Beckers, P. (2012). Evaluating regeneration policies for rundown
industrial sites in the Netherlands. The Hague: PBL Netherlands Environmental
Assessment Agency.
Porter, M. E. (2000). Locations, clusters and company strategy. In G. L. Clark, M. P. Feld-
man, & M. S. Gertler (Eds.), The Oxford handbook of economic geography
(pp. 253–275). Oxford: Oxford University Press.
Reed, R. (2012, June). The increasing importance of social sustainability –the contribution
of social area analysis in housing market analysis. Paper presented at the European Real
Estate Society 19th Annual Conference, Edinburgh.
Rosen, S. (1974). Hedonic prices and implicit markets: Product differentitation in pure
competition. The Journal of Political Economy, 82,34–56.
Ryan, S. (2005). The value of access to highways and light rail transit: Evidence for
industrial and office firms. Urban Studies, 42, 751–764.
22 J. Beekmans et al.
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013

Sivitanidou, R., & Sivitanides, P. (1995). Industrial rent differentials: The case of Los
Angeles. Environment and Planning A, 27, 1133–1146.
Stec Groep. (2005). Bedrijfsruimtegebruikers in Beeld: Resultaten Database Bedrijfsruim-
tegebruikers Nederland 2005 [Focus on occupants of commercial property: Results from
database on occupants of commercial property The Netherlands 2005]. Nijmegen:
Author.
Taskforce Herontwikkeling Bedrijventerreinen. (2008). Kansen voor kwaliteit: een ontwikk-
elingsstrategie voor bedrijventerreinen [Opportunities for quality: A development strat-
egy for industrial sites]. The Hague: Ministry of Housing, Spatial Planning and the
Environment/Ministry of Economic Affairs.
Van der Krabben, E., & Buitelaar, E. (2011). Industrial land and property markets: Market
processes, market institutions and market outcomes: The Dutch case. European Planning
Studies, 19, 2127–2146.
Van Oort, F. (2004). Urban growth and innovation: Spatially bounded externalities in the
Netherlands. Aldershot: Ashgate.
Vastgoedmonitor. (2012). Gemiddelde verkoopprijs bedrijfsruimte 2000–2007 Nederland
[Average sales prices of commercial property 2000–2007 The Netherlands]. Retrieved
from http://www.vastgoedmonitor.nl
Waarderingskamer. (2011). WOZ-Uitvoering: (modelmatige) waardebepaling niet-wonin-
gen [Appraisal of taxation values: (Modelwise) appraisal of non-residential prop-
erty]. Retrieved from http://www.waarderingskamer.nl/kiosk/Product.aspx?id=93&view
=demands&demands=2745
Wheaton, W., & Torto, R. (1994). Office rent indices and their behaviour over time. Journal
of Urban Economics, 35, 121–139.
Journal of Property Research 23
Downloaded by [Radboud Universiteit Nijmegen] at 01:14 04 November 2013