Land Use Policy 47 (2015) 6677

Contents lists available at ScienceDirect

Land Use Policy

journal homepage: www.elsevier.com/locate/landusepol

Negotiation issues in forming publicprivate partnerships for

browneld redevelopment: Applying a game theoretical experiment
B. Glumac a, , Q. Han a , W. Schaefer a , Erwin van der Krabben b,c
a
b
c

Eindhoven University of Technology, Department of the Built Environment, Den Dolech 2, P.O. Box 513, 5600 MB Eindhoven, The Netherlands
Radboud University, Geography, Spatial Planning and Environment Department, P.O. Box 9108, 6500 HK Nijmegen, The Netherlands
University of Ulster, School of the Built Environment, United Kingdom

a r t i c l e

i n f o

Article history:
Received 24 February 2014
Received in revised form 8 March 2015
Accepted 27 March 2015
Keywords:
Browneld redevelopment (BR)
Publicprivate partnership (PPP)
Game theory
Fuzzy Delphi method (FDM)

a b s t r a c t
The redevelopment of a browneld can provide a range of societal, environmental but also economic
benets for a number of entities. In the Netherlands (and elsewhere), publicprivate partnerships are
common practice for such projects, because of two main reasons. First, limitations to public funding have
led governments to invite the private sector into various long-term arrangements for capital-intensive
projects. Second, a comprehensive approach for the whole browneld area may be more efcient and
protable, compared to piecemeal development via interventions by individual owners. This article investigates, with respect to browneld redevelopment, the interaction behavior of two key parties in forming
partnerships: the municipality and a private developer. It is assumed that, apart from their mutual interest to redevelop the browneld area, they will have different interests as well. In order to indicate their
specic interest and the negotiation outcome regarding the forming of a public private partnership, this
paper makes use of an experimental game theory approach. Three specic negotiation issues were analyzed in our research: a building claim, future land use and reparcelling of the land. In addition, this paper
suggests an eight-step procedure to conduct a game theoretical experiment. A survey was conducted in
order to gather the required data for the experiment. The data have been used to estimate the payoffs
variations between the two key parties in the mentioned negotiation games. Finally, by comparing sub
game perfect Nash equilibrium generated game outcomes and direct expected outcomes of respondents,
this paper experimentally proves that the game theoretical analysis provides a valid representation of
a real world browneld redevelopment negotiation within the Dutch institutional-economic context.
The outcome of the experiment conrms the Dutch tradition of public private partnerships in urban
development practice, with public and private bodies willing to share nancial risks and returns in these
projects.
2015 Elsevier Ltd. All rights reserved.

Introduction
Several denitions for a browneld can be found in the literature
(CABERNET, 2002; Yount, 2003). This paper uses the following: A
browneld site is any land or premises which has previously been
used or developed and is not currently fully in use, although it may
be partially occupied or utilized. It may also be vacant, derelict
or contaminated. Therefore, a browneld site is not available for
immediate use without intervention (Alker et al., 2000). Numerous
authors (Carroll and Eger Iii, 2006; Chen et al., 2009; De Sousa, 2002;
Ganser and Williams, 2007; Lange and McNeil, 2004a,b; Wang et al.,

Corresponding author. Tel.: +31 402472373.

E-mail addresses: b.glumac@tue.nl (B. Glumac), q.han@tue.nl (Q. Han),
w.f.schaefer@tue.nl (W. Schaefer), e.vanderkrabben@fm.ru.nl (E. van der Krabben).
http://dx.doi.org/10.1016/j.landusepol.2015.03.018
0264-8377/ 2015 Elsevier Ltd. All rights reserved.

2011) have argued that the redevelopment of a browneld can

provide a range of economic, social, and environmental benets.
Leaving brownelds unmanaged brings a potential loss of economic
opportunities to the community in which they are located.
In most cases, a browneld redevelopment (BR) seeks a form
of partnership. A public private partnership (PPP) is a concept frequently used in development practice (Koppenjan and Enserink,
2009) although a uniform denition is still lacking (Weihe, 2005).
PPPs are particularly useful when circumstances are not favorable for a piecemeal development via interventions by individual
owners (Grimsey and Lewis, 2002). In such cases a comprehensive
integrated approach, with private owners/developers collaborating with the responsible public authorities, may be more efcient
and protable. Another important reason for the establishment
of a PPP can be limitations to public funding available, making a public sector-led redevelopment impossible. This has led

B. Glumac et al. / Land Use Policy 47 (2015) 6677

local governments to invite the private sector into various longterm arrangements for capital-intensive real estate development
projects.
Forming a PPP can be problematic as a consequence of differences in goals amongst potential partners. The existing literature
addresses the general diversication of goals and interests by providing various typologies of potential parties (Coiacetto, 2001;
Hieminga, 2006). The preferences of the potential parties involved
in BR may vary substantially. It is also possible that the selfinterest of an individual decision-maker can be heavily inuenced
by the other parties that are present in a certain decision moment.
Providing more insight into these interactions may be of help
when forming new publicprivate coalitions for BR. This paper
introduces concepts of game theory in order to improve the understanding of the interactions among two key decision-makers in
forming PPP for a BR project: the municipality and a private developer.
Game theory has been applied in many elds of research, but
only few applications can be found in urban development practice.
Most game-theoretical applications with respect to urban development focus on negotiations, applying game-theoretical concepts
with regard to the interaction of players. Samsura et al. (2010,
2013) and Samsura and van der Krabben (2011, 2012) have used
game theory to model negotiation processes with respect to value
capturing in land and property development. In addition, pricing strategies with respect to land use have been modeled with
help of game-theoretical concepts (Forester, 1987; Ma et al., 2007;
Martnez and Henrquez, 2007; Mu and Ma, 2007; Sibdari and Pyke,
2010; Wu et al., 2014; Zellner et al., 2009). Modeling this kind of
negotiations has proved to be able to generate practical advice.
For example, Pfrang and Witting (2008) have demonstrated how
lease contract negotiations can be smoothened and how a social
environment between the tenant and the landlord can be cultivated.
Regarding the application of game theory with respect to
decision-making processes for BR, analyzing negotiations may be
useful to decide how to allocate cost and benets in browneld
redevelopment negotiations (Liang et al., 2008; Wang et al., 2007,
2011), to compare the costs and benets of BR and greeneld
development, in order to support BR with effective policies (Liang
et al., 2008), and to evaluate the potential conict in engaging public opinion in redevelopment processes (Tam and Thomas, 2011;
Tam et al., 2009). Most applications, however, refer to improving decision-making processes in establishing various partnerships
(Blokhuis et al., 2012; Sounderpandian et al., 2005; Walker et al.,
2008; Wang et al., 2008; Youse et al., 2007, 2010). Ultimately,
the latter research helps to develop decision-support tools, claries interests, identies tradeoffs, recognizes party satisfaction,
and generates optimal solutions, preparing a decision maker to
optimally benet from the negotiation (e.g. Youse et al., 2010)
Although game theory can help to negotiate favorable conditions
related to different partnerships types, little attention so far has
been put on isolated negotiable issues in forming a PPP for a BR
project. This paper elaborates on three specic issues in these
negotiations (building claim, future land use and reparcelling the
land; see Section Dening the institutional-economic context of
the game) and aims to contribute to the further development of
game-theoretical approaches to urban development practice by
suggesting a formal procedure for applying a game-theoretical
experiment.
Classical game theory has been largely criticized due to the
notion of a homo economicus, a completely rational decisionmaker (e.g. Camerer, 2003; Raiffa, 2002). Therefore, instead of using
a classical game-theoretic approach, this paper provides the ndings based on experimental game theory results. Rather than only
modeling the outcome of the negotiations, the games have been

67

experimentally tested. Usually, an experiment consists of several

phases: description of the game environment, the assumptions
underlying the game, and estimation of players preferences. This
experiment introduces an eight-step procedure. First, the game is
set in a proper institutional-economic environment. This phase has
been divided into ve separate steps: (1) selecting a game class
cooperative vs. non-cooperative and conict vs. common interest;
(2) selecting a game form strategic vs. strategic; (3) selecting a
game solution concept; (4) describing the institutional-economic
context of the game, here it is important to mention that the study
was conducted in the Netherlands; (5) designing the game conditions within the game set environment. Further (6), two different
types of games have been assumed (ultimatum and bargaining
game) for specic negotiation issues in forming PPP for BR. In addition, both games are experimentally validated by a survey among
BR experts. The players in both games are a public party and a private party. In the remaining of the paper, these two parties are
referred to as a municipality (M) and a developer (D). To estimate
the respondents preferences (7), a standard phase in game theoretical experiments, the fuzzy Delphi method (FDM) with similarity
aggregation method (SAM) has been applied. (8) Finally, the experiment ends with the analysis of the outcomes.
To collect the data for validating the results an on-line survey
tool is used (Berg Enqute System 2007).
The experiment explores whether the self-prediction of the
respondents about the game outcome corresponds to the gametheoretical predictions. This provides insight in the suitability of the
application of game theory in predicting real-world actor behavior
concerning BR projects. In addition, based on the outcomes of the
analyses, interventions can be designed and through them, various policies may be considered. The eventual new policies would
aim at supporting the cooperation between relevant parties, thus
reducing the number of conicts and stimulating the actual implementation of BR projects.
This paper rst explains the basic elements of a game tree and
argues for the implementation of game theoretical experiments in
forming PPP for BR projects (Section Game theory applications in
forming PPP for a BR project). Further, Section Designing games
over negotiation issues in forming PPP for a BR project explains the
eight-step procedure of conducting a game theoretical experiment
for an urban development project. Section Data collection and
respondents characteristics reports on the data collection technique and the background of respondents that are used for the
game theoretical experiment. Section Game experiment results
summarizes the empirical results of validated game trees and estimated game outcomes. Finally, Section Conclusions concludes on
the importance of using rigorous procedural steps of conducting
a game theoretical experiments and the contribution that such
experiments can provide to represent real world browneld redevelopment negotiation.

Game theory applications in forming PPP for a BR project

Often, in urban development the outcome of a decision-making
process does not only depend on individual choice but is also inuenced by choices made by other decision-makers. Game theory
is a suitable theory to test the behavior of interactive decisionmaking situations (e.g. Neumann et al., 1944). Even more, game
theory assumes that decision-making is always interdependent;
game theory mainly aims to provide a better understanding of situations in which decision-makers interact (Colman, 1995; Osborne,
2004; Rasmusen, 2007; Shoham and Leyton-Brown, 2009; Stengel,
2008).
A decision-maker a player in game-theoretical terminology
has to think ahead and is assumed to devise a strategy based

68

B. Glumac et al. / Land Use Policy 47 (2015) 6677

Fig. 1. An example of game tree

on the expected countermoves of the other player(s). Therefore,

the main objective of game theory is to determine what strategies the players, rationally, ought to choose in order to pursue
their own interests and what the outcomes will be if they do
so. Game theory is well-suited to describe and analyze situations in which players have both conicting and supplementary
interests and interdependency in behavior in urban development
and related decision-making situations in which two or more
players are involved (e.g. Liang et al., 2008; Mayer et al., 2005;
Mu and Ma, 2007; Samsura et al., 2010; Wang et al., 2007).
Other research has already demonstrated that game theory can
provide insight in the negotiation strategies over specic issues
relevant for BR (Arentze and Timmermans, 2003; Forester, 1987;
Sounderpandian et al., 2005; Wang et al., 2011; Youse et al.,
2010).
The basic arguments for applying game theory in PPP negotiations for a BR can be summarized as follow. First, game theory
is based on the premise of relational interdependency between
players. This responds to the multi-actor environment in forming PPPs for BR projects. Secondly, game theory assumes that
players make rational decisions based on their utility function
representing their needs and interests. Having in mind that two
key players in BR (municipality and developer) are a group of
individuals responding in a certain institutional-economic context, it has to be veried if their decisions are indeed rational.
Finally, game theory analysis reveals the possible course of actions
based on certain payoffs, structure and present players in any
game. Controlled changes of these terms might inuence different
negotiation outcomes, thus making game-theoretical applications a useful policy or strategy tool in forming a PPP for a BR
project.

Game elements: players, strategies, payoffs, equilibriums

Basic assumptions in game theory include that decision-makers
pursue well-dened, exogenous objectives (they are rational), they
have an innite good memory (perfect recall), and they take into
account their knowledge or expectations of other decision-makers
behavior (they reason strategically). Game theoretical models are
highly abstract representations of real-life situations, which allow
them to be used to study a wide range of phenomena. In order to
predict interaction outcomes a game consists of at least three basic
elements: players, strategies, and payoffs.
The players are the decision-makers; a player n is assumed to
be a solitary actor who makes decisions as a single decision body.
The strategy sn is a plan of all possible actions An = {an }, dening
what player n might do in any given situation during the game,
aiming for utility maximization. All players make their own choices
by selecting a strategy, but the result for each player depends on
the choice of the other player. The resulting set of strategies for
each of the N players in the game is denoted as a strategy prole
SN = {sn ,. . .,sN }. If the game has only two players, a strategy prole
is therefore a pair of strategies with one strategy per player.
Finally, the payoff for player n is denoted as Un (s1 ,. . .,sn ).
This can be dened as a number associated with each possible outcome resulting from a complete set of strategic selections
by all the players in a game. Generally, higher payoff numbers
are attached to outcomes that are better in the players rating
system.
The conjunction of chosen strategies and related payoffs is
dened as the outcome of the game. A clear distinction has to be
made between the concepts of outcome and payoff; an outcome is
the decision, if any, arrived at by the players collectively, while the
denite payoff of an outcome for a player is the value of that outcome for the player. Because players will have different valuation
systems over the set of possible outcomes, and hence have different
preferences over the outcomes, this is where conicts can arise.
In order to predict the outcome of a game, game theoretical analysts focus on possible strategy proles and on selecting one or
more strategy proles as a reection of the most rational behavior
by the players. A strategy prole that consists of the best strategy
for each of the n players in the game is dened as an equilibrium
s* = (s* 1 ,. . .,s* n ). Players choose equilibrium strategies in trying to
maximize their individual payoffs. In order to nd equilibriums,
the players most preferred strategies should be dened. Solution
concepts are suitable for dening such preferred strategies; a solution concept F: {S1 ,. . .,Sn , U1 ,. . .,Un } s* is a rule that denes an
equilibrium based on the possible strategy proles and the payoff
functions.

A game tree and its elements

Experimental game theory

In general, games can be represented in strategic or extensive form. The main difference between them underlines how the
players act. They can either act simultaneously (strategic form)
or sequentially (extensive form). It is important to underline that
games in extensive form are represented with a game tree. There is
no convention, however, how to design a game tree. The following
example (Fig. 1) draws the tree downwards starting with the root
on top. The nodes represent the state of the game while two successor nodes represent the move of the player. Lines called edges
connect them. The line that connects two successors nodes are a
possible players move or an action (X, Y, a, b, c, d). If a decisionmaker (1, 2) performs an action, this node is called decision node.
Alternatively, an action can be determined by nature: in that case
the node is called chance node. Then the move or action is random,
predened with the probability. The nodes without a successor are
named terminal nodes. At such a node, every player gets a payoff
that is a real number.

Generally speaking, the emergence of game experiments can

be identied by a need for empirical information about the principles of strategic behavior and the ability of experiments to provide
such information (Crawford, 2002). The predictions of classical
game theory are very sensitive to the structure of the game. Relying only on the existing (data) input of a certain research context
often leads to an unobserved or uncontrolled structure of the
game. Although experiments often share some of these problems,
the control and observation given by applying modern experimental techniques provides a notable advantage in identifying
the relationships between strategic behavior and the environment (Crawford, 2002). Still, theory and experiment may have
strongly complementary roles. While theory provides a framework
to gather empirical data and interpret the respondents behavior,
experiments indicate which parts of the theory are most useful
in predicting and identifying behavioral parameters that have not
been reliably determined by theory (Crawford, 2002).

B. Glumac et al. / Land Use Policy 47 (2015) 6677

Designing games over negotiation issues in forming PPP for

a BR project
Besides the argued applicability, detailed insight is required
when an analyst wants to implement a game-theoretical experiment in a context of BR. Therefore, this section provides special
attention to the procedure for conducting a game theoretical experiment. The purpose of setting these dened methodological steps
when applying a game theoretical framework for an urban development problem should be regarded as a constraint to pre-inuence
the game in order to reduce the biased assumptions of an analyst.
Particularly in this study, the construction of a game experiment
consists of the mentioned eight steps. This section provides a
detailed insight for each of these steps.
First, the game class and form needs to be identied. Here the
major tasks for an analyst are identifying one out of the four possible game class combinations and identifying if it is more suitable
that players act simultaneously or sequentially. Further, there are
many solution concepts that might be used as a reference. However, only one proper game solution concept has to be adopted
that ts best to the problem itself. Further necessary step would
be to describe the institutional-economic context of the game and
designing the statistically valid game conditions within the context. Without dening and presenting them to the respondents,
experimental results would not be meaningful. This article also
suggests that the assumed game structures should be experimentally validated as well. Estimating the respondents preferences and
predictions of the game outcomes can be performed in numerous
ways by applying various methods. In this article FDM with SAM has
been used, it enabled experimentally valid results although having
a relatively small sample of the respondents. A nal step described
in this section is interpreting the respondents feedback with SPNE
and direct game outcome predictions that provided on one hand the
proof that in this experiment player act rationally and on the other
that the experimental game theory can be successfully applied in
BR negotiation.
Identifying a game class: non-cooperative conict game
Games can be classied as cooperative and non-cooperative
games. Both are used to study players interaction (Shoham and
Leyton-Brown, 2009). Cooperative game theory deals with situations in which groups of players already agreed to cooperate. These
players aim for coordinating their actions, eventually resulting in
joint prots. Because these joint prots often exceed the sum of the
individual prots, cooperative game theory deals with the interaction of players within binding agreements such as PPP. In this case,
an interaction could address the division of the PPPs expenses and
prots. However, often some of the interactions are not part of a
binding agreement. This might be the case when public and private players negotiate about the division of risks or development
potential before creating a PPP. Such negotiations are characterized
as non-cooperative games, which model how actors strategically
behave toward each other when the cost of bargaining and coordination is too high.
Moreover, game theory distinguishes between conict and common interest games (e.g. Bowles, 2004). In a common interest game,
the interactions have a pattern of a trafc jam that is an overall poor
outcome. Therefore, it is benecial for everyone to avoid it. Contrarily, in a conict game, the interests of several decision-makers are
opposed or only partly coincide. Each decision-maker will usually
choose an option contributing to his own interest, which has not to
be the interest of others. For example, negotiating about the players
inuence on a future land-use means more for one and less for the
other. As an aside, these individual decisions could result in worse
outcomes for all players compared to a coordinated decision.

69

Given the explanation of four possible classes of games, this

paper selects a non-cooperative, conict game as a credible mimic
of the negotiation issues in forming PPP for BR project between
a municipality (player M) and developer (player D). In particular, three negotiation issues have been observed: building claim,
future land use and future reparcelling of the land. Two games
that are well known in the game theoretic literature are selected
to represent these negotiation issues, respectively, the ultimatum
and the bargaining game, where the building claim negotiation
issue is represented by a ultimatum game and the future land
use and reparcelling of the land negotiation issues by a bargaining
game.
Identifying a game form: extensive form
As mentioned in Section A game tree and its elements, games
can be represented in strategic or extensive form. The main difference between them underlines how the players act. They can either
act simultaneously (strategic form) or sequentially (extensive
form). Given the time perspective of the negotiation, in strategic
form representation the players act in a single time sequence while
in the extensive form players are being aware of all moves of players
in the previous time sequences. The negotiation between previously mentioned player M and D ts better with the concept of the
sequential games, because of two reasons. First, both players are
aggregated entities. This implies a multitude of opinions between
both players thus resulting in a structured decision-making process
before confronting each other. Secondly, urban redevelopment processes usually consist of several phases (e.g. Hieminga, 2006; Miles
et al., 2007; Peiser and Frej, 2003). This makes urban redevelopment almost by denition a time consuming multi-phase process.
It is therefore unlikely that the negotiation will be resolved in a
single time sequence negotiation.
In our study we assume perfect information for M and D when
negotiating over forming a PPP. Perfect information implies that
a player knows the status of the game and therefore the complete history of the game up to then. Negotiating about building
claims, future land use and reparcelling of the land can be considered common practice for both M and D (Samsura et al., 2010; Van
der Krabben and Jacobs, 2013). On the other hand, the players may
lack information with respect to for instance market conditions and
land and real estate prices. Nevertheless, this application of game
theory assumes perfect information for players.
Finding a solution concept: sub game perfect Nash equilibrium
(SPNE)
A solution concept is a formal rule for predicting how the game
will be played. The central concept of non-cooperative game theory
is the so-called Nash equilibrium that is named after John Nash who
introduced this solution concept in the 1950s for games in strategic
form.
In extensive games, a backward induction process always
produces Nash equilibrium, also called Sub game Perfect Nash
Equilibrium (SPNE), since it represents strategy prole for both
players. The optimal play of any player should maximize his payoff. This payoff can be decided irrespectively of the others players
actions when observing the players last action. That is the reason why the backward induction process always starts with the
decision node closest to the leaves. A player naturally chooses the
action at this node giving him the maximum payoff. An action is
chosen in this way for every decision node when all subsequent
actions have been decided. In the example given above (Fig. 1),
player 2 chooses between the payoffs 3 and 0 on the left side node
and between payoffs 2 and 0 on the right side node. Therefore,
player 2 chooses actions a and c (indicated by the two arrows).

70

B. Glumac et al. / Land Use Policy 47 (2015) 6677

Going backward in time, player 1 chooses action Y (also indicated by

an arrow) assuming the previously described behavior of player 2.
The action selected by the backward induction is not always unique,
since it is possible that more than one action provides the maximal
payoff. This procedure denes every action in all decision nodes
in a complete game tree and describes a strategy for each player.
The result of backward induction is therefore a strategy prole. In
extensive form games, it is then also possible to state that SPNE and
strategy prole obtained by backward inductions are synonyms.
Dening the institutional-economic context of the game
To set up the game, we rst dene the institutional-economical
context. For this purpose, we refer to the common land and property development models applied in the Netherlands (Samsura
et al., 2010): (1) public land development, (2) building claim model,
(3) PPP model and (4) private land development model. For an
extensive analysis of the characteristics of these models, we refer
to Van der Krabben & Jacobs (2013). Each development model is
characterized by the initial situation on the land market related to
the ownership structure, the dened parties that acquire the land,
service it and reparcel the land into building plots, and the parties
that acquire the building plots.
This study addresses only the building claim development
model as a specic type of PPP. In this PPP development model,
in the initial situation the original owner or a private developer
owns the land. Subsequently, the land is acquired by a joint venture
company (JVC) a type of PPP that is formed by the municipality
and the developer that already owned (part of the) land on the BR
location. The land is serviced and reparcelled by the same JVC. The
private developer participating in the JVC has usually claimed the
right of rst buyer of the building plots, against a xed price (building claim; in return for his willingness to take part in the JVC) of the
newly formed building plots (see Section Sample size and response
rate for further explanation).
The games presented in this paper address certain negotiation
issues that nally inuence the decision of both players: to form the
JVC or not. The municipality invites a developer to form a JVC for a
single BR project. In order to simplify the game, it is assumed that
the municipality also owns part of the land on the BR site. When
formed, the JVC will service the land and reparcel it into ready-tobuild-on building plots. The building plots can be seen as the nal
product that is produced by the JVC.
The players involved in the game base their decision to form
the JVC or not on several general conditions. First, the respondents
(actual players) should be familiar with the proper browneld definition (Alker et al., 2000). Second, the decision problem is limited
only to the initiative phase of a BR (to produce building plots or
not). Third, the size of a browneld is in the range of one to ten
hectares. Further, future land use is assumed to be a mix land use
(affecting the price of the building plots that can be sold). Finally, it
is assumed that the decisions that will be made are country-specic.
Therefore, this study focuses only on the Netherlands.
Setting a game condition state
The descriptive part of the game is composed from a limited
set of the relevant attributes that represent changing BR negotiation settings. An attribute is the characteristic of a product (e.g. BR
project) that consists of various levels (Louviere et al., 2000). This
rigorous delineation is made for the purpose of the statistically valid
game experiment named game condition. In order to describe the
exogenous conditions of the BR negotiation upon building claim,
future land-use, and reparcelling of the land, four attributes have
been used:

Table 1
Condition state: attributes allocation and the treatment combinations.
Treatment combination

1
2
3
4
5
6
7
8
9

Attributes allocation
L

AS

0
0
0
1
1
1
2
2
2

0
1
2
0
1
2
0
1
2

0
1
2
1
2
0
2
0
1

0
2
1
1
0
2
2
1
0

Location: refers to the proximity and accessibility of a site. Proximity is a distance to the key city locations (e.g. CBD); and
accessibility refers to how good the access is to the site by car
and public transport.
Embeddedness: the extent to which the redevelopment area can
be integrated into the existing urban fabric.
Administrative support: refers to the transparency and perceptions of continuity in governance, politics and the administration.
Synergy with surrounding users: the extent to which the surrounding area inhabitants/users support the redevelopment.
These attributes emerged as signicant in related studies
(Adams et al., 2001; Glumac et al., 2011; Lange et al., 2013; Page
and Berger, 2006; Peiser, 2007; Syms, 1999; Thomas, 2003; Wang
et al., 2013). Each attribute that is used in the game condition consists of three levels (e.g. location excellent, moderate, and poor).
A level is an ordinal value of an attribute that enables changing BR
negotiation settings.
For an experimental design, this study generated orthogonal
fraction factorial design (Hahn and Shapiro, 1966) to secure the statistical validity of this experiment. Given these four attributes all
having three levels, the full factorial design would suggest 34 = 81
different decision moments. Instead, an orthogonal fractional factorial design reduced the number of treatments. The following design
(Table 1.) shows the attributes allocation over the nine treatment
combinations (Hahn and Shapiro, 1966). Predened orthogonal
design assures that the attributes levels are unconfounded with
each other thus providing a statistical validity. As an aside, each
respondent responds to only one treatment combination or one
negotiation setting when rating and selecting the game outcomes
(Fig. 5). For example, the treatment combination in the rst row (1)
consists of the lowest levels for each of the four attributes (0, 0, 0,
0), see Table 1.
Game assumption
As mentioned above, two games are assumed to be a valid representation of three different negotiation issues in forming PPP for a
BR project. Therefore, this subsection provides argumentation for a
choice to select these two non-cooperative conict games and their
structure. They both investigate negotiation or more specic how
bargaining occurs between two rational actors. First, it is important to underline for both games that they are non-cooperative,
because players estimate the costs of bargaining and cooperation
prior to the binding agreement. This leads to a players decision
to establish a PPP or not. Second, both games are conict games
because the players larger inuence on a future land-use, for example, means more for one and less for the other player. As an aside,
if we would assume that the partnership had already been agreed
on, than we would investigate the interaction under binding agreement and therefore classify these games as cooperative, conict

B. Glumac et al. / Land Use Policy 47 (2015) 6677

Fig. 2. Ultimatum game: building claim

games such as cake-cutting or fair division games (e.g. Shoham and

Leyton-Brown, 2009).
The ultimatum game: building claim
We have assumed that the ultimatum game can represent the
negotiation on one issue: the building claim negotiation. In the
Dutch context of land development, the building claim refers to
a situation in which a private developer that owns a piece of land
on a (re)development location agrees with the municipality to sell
its plot of land to the municipality (or the JVC) under the condition that the municipality will service and reparcel the land into
building plots and offers the private developer the rst right to buy
building plots (the building claim; equal in size to the plot of land it
originally owned) against a price that has been agreed upon at the
time the original plot of land was sold (Samsura et al., 2010). This
development model has been common practice in land and property development in the Netherlands since the 1990s. It is based
on two assumptions: (1) acquiring (previously built-on) land and
servicing it and selling it again as building land is not without risks
and therefore, private developers require a building claim as a compensation for sharing this risk with the municipality; (2) building
houses (or other properties) on the building plots is expected to
be protable and therefore, in general, the demand for building
plots from the private sector exceeds the supply. The latter makes
it attractive for private developers to have the rst right to buy.
This building claim model is assumed to be favorable both to the
municipality and the private developer, because it allows for a more
efcient and protable comprehensive development strategy for
the location. To select or design a game, the following postulate is
used. Two options of a negotiation issue need to be accommodated
in the structure of the game. This postulate is addressed by setting
each option as an action in the game. Therefore, the rst negotiation issue regarding the building claim has two options: to offer
a building claim (BC) or not (NBC) (Fig. 2). As an alternative to the
basic ultimatum game it is also possible to assume an altering or
counteroffer ultimatum game. In this case, that altering or a counteroffer would appear at the decision node of player D where he
would ask for BC after player M offers NBC. Such an assumption,
however, would be redundant due to the nature of this negotiation
issue. The negotiation issue is regarded as dummy ordinal variable
where NBC has less value. Having in mind that this negotiation issue
is described by the ordinal dummy variable, altering or counteroffer extension to the game structure would only be appropriate in
the case of interval or ratio variable where quantitative divisions
are possible. Also, it would not be realistic to present solely this
counteroffer as a new negotiation round since a building claim is
one out of four urban development models (Samsura et al., 2010).
This implies that player D would also include three other development models at this decision node. Further, this would lead to
the construction of the game where players choose the optimal
development model (Samsura et al., 2010) instead of deciding to
create a PPP or not. As an aside, the building claim game is set in
the previously described game condition.

71

Player description, information, and strategy. The ultimatum game

is regarded as a 2 2 game. That stands for a game where there are
only two players each having only two strategies. This ts with the
proposed building claim ultimatum game since there are only two
proposals available: BC and NBC. The two players are the municipality (M) and developer (D) that would potentially form a JVC. The
player M is the initiator of the game, which is consistent with the
Dutch context of public land development (Samsura et al., 2010).
Information available to the players is dened as follows: (1)
perfect each player knows his position in the game tree and all
players know the previous moves of the other players; (2) certain
all players know the payoff of playing a particular strategy given the
strategies of other players; (3) asymmetric players have different
pay-offs; (4) incomplete a player does not know the other players
pay-offs.
Fig. 2 illustrates the game. At the rst decision node, player M
offers to player D a deal in which a building claim is either available (BC) or not available (NBC). For both possible actions of player
M, player D can accept (a) or reject (r) the deal on the succeeding
decision node in the game. The game stops when the end nodes are
reached.
This procedure practically explains the complete plan of possible actions (strategy by denition) of the players M and D. Their
actions differ and a branch represents each action. Player M can
choose from two possible actions: BC, NBC. These two actions dene
the plan of possible actions Am = {BC, NBC}. Similar, the actions of
the player D are: a, r as a reaction on the BC and a, r as a reaction on
NBC. They dene the Ad = {aBC , rBC , aNBC , rNBC }. Note that, because of
the readability of Fig. 2 the actions aBC and aNBC are marked as a.
The payoffs are estimated empirically with the fuzzy Delphi
method (FDM). Rather than providing a game solution, this subsection explains the game design. In Section Estimated results
with fuzzy Delphi method we further analyze the outcome of the
game.
The bargaining game: future land use and reparcelling of the land
This game addresses two other negotiation issues between M
and D: respectively inuence on the future land use and on the
reparcelling of the land. Both issues are relevant to the negotiation process, because they may inuence the protability of the
BR project. The future land use allowed on the location directly
affects the price per m2 of the building plots, while the way the
location is reparcelled in building plots may inuence the total size
of building land that can be sold (e.g. Zellner et al., 2009). The decision regarding future land use and the reparcelling of the land is a
public decision (as part of a decision on the land use plan), but the
private developer may be able to inuence it, because for the actual
implementation of the land use plan the municipality depends on
the willingness of the private developer to buy the building plots
and redevelop the site. The inuence over the future land-use and
the reparcelling of the land has been expressed in the ordinal scale
(high, medium and low inuence of developer in future development). The outcome of the negotiation indicates if JVC would be
established or not. For instance, if M would allow D only minor
inuence, D might reject to join JVC. As an additional relevancy, this
negotiation can be linked to the division of the future JVC shares.
For instance, if M would allow D high inuence, D might accept a
smaller share in the returns from the JVC.
Similar to the ultimatum game, this game can be seen as a bargaining game that has one additional proposal in the table. Also
similar to the previous game, the postulate is that the ordinal values of an attribute (or variable) need to be accommodated in the
structure of the game. Therefore, three inuence values are identied for this issue: high (H), medium (M), and low (L). The level
of the players (e.g. developer) inuence (H, M, L) expresses the
potential to adjust the land use ratio within the mixed-use zoning.

72

B. Glumac et al. / Land Use Policy 47 (2015) 6677

Fig. 3. Bargaining game: future land use and parcellation

The (H) inuence means that a player can carry out any land use
regulated by a mixed-use zoning plan and completely determines
the size and the shape of any parcel in the land that will be redeveloped. Medium inuence (M) grants a developer less and low
inuence only minimal possibilities to adjust the land use ratio.
Fig. 3 describes the bargaining game. Contrary to the previously
presented ultimatum game, in the bargaining game it is possible to
assume the existence of the counteroffers that might cause costly
delays. This is appropriate since this negotiation can occur in any of
the four mentioned urban development models (Section Dening
the institutional-economic context of the game).
Player description, information, and strategy. The players in the bargaining game are again municipality (M) and developer (D). Their
information is set to be the same as in the previous game.
At the rst decision node, player M offers to player D one of
the deals H, M, or L linked to different inuence on the future land
use and reparcelling of the land. For each of the possible actions
of player M, player D can react differently on each of the decision
nodes. The structure of every sub-tree has been designed in a way
that the highest level of inuence can be reached no matter what
was the initial offer from the player M (H, M, L). For example, when
player M offers H, then player D can only accept (a) or reject (r) the
deal on the succeeding decision node in the game. This is because
the highest level of inuence is already offered. However, if M offers
M, then D can ask for the highest inuence (h) or either accept (a)
or reject (r) the offer. If D ask for (h) there is one more succeeding
decision node where M can accept (A) or reject (R) that offer. This
is similar for the branch when the player M offers L at the initial
node. In any case, the game stops when the end nodes have been
reached.
Similar to the previous game, this procedure practically explains
the complete plan of possible actions related to the strategies of
players M and D. Their actions differ and a branch represents each
action. Player M may choose from nine possible actions in this
game: H, M, L, A, R, A, R, A, R. These actions dene the plan of all
actions Am. Player D may also choose from nine actions: a, r as a
reaction on the H. Then h, a, r, as a reaction on M, and h, m, a, r, as a
reaction on L. Together they dene the plan of all possible actions
for the player D, Ad.
Fig. 3 also presents the payoffs. In Section Estimated results with
fuzzy Delphi method we further analyze the outcomes of this game.
Game tree validation
As a second part of the game experiment, this study introduces
the validation of the game trees (Fig. 4.), with help of a semistructured questionnaire with multiple-choice answers.

For both games, every decision node is textually described. For

example, the rst question in the experiment (Fig. 4.) corresponds
to the initial decision node in the building claim game (Fig. 2.). In
this example, the description is: A municipality (as initiator) negotiates with a developer over the building claim. What are the possible
negotiation options? Given this description, every respondent states
the possible actions (branches) at that decision node by lling in the
multiple-choice answers. In each question, a respondent can rst
verify if all assumed actions at the specic decision node are correct or not. Therefore, the ideal validation of the game tree structure
would be that all respondents selected all of the multiple answers
meaning that they exist in real market negotiation situations. Secondly, a respondent checks if there is a missing action(s) at certain
decision node. If an action is missing, a respondent chooses the
answer other. Additionally, a respondent can provide the description of a missing action in the text line below every question. In this
case, the ideal validation would be that none of the answers were
selected meaning that there are no missing actions. The results of
the game validation are presented in Section Validated game tree
structure.
The preferences and predictions of the game outcomes
The nal part of this experiment concerns the estimation of the
respondents preferences and predictions over the game outcomes
end nodes in both games (Figs. 2 and 3).
First, the respondents are asked to give their estimation on a
game outcome in a specic decision moment. This moment is set as
a game condition within the given institutional-economical context
(Section Identifying a game class: non-cooperative conict game)
and it is described with specic set of attributes and levels (Table 1).
Second, each game outcome is textually described. The text description captures all lines of action leading to the game outcome. For
example, Fig. 5 describes outcome 1 as: A developer rejects the offer
and stops the negotiation when building claim is available. This corresponds to the following line of actions: player M plays BC, and then
player D plays r (Fig. 2). Same logic is used to describe every game
outcome in a game tree. Finally, every respondent provides its preference toward a certain game outcome and its prediction that this
outcome will occur. This is repeated for every outcome in a game
tree. A preference is regarded here as a payoff and by knowing the
payoff it is possible to apply backward induction and estimate SPNE
game outcome. By comparing the SPNE solution with the estimated
most probable outcome, it is possible to investigate whether the
players are rational and if market behavior in a BR project can be
explained with the application of the game theory.
Therefore, there were two estimations per outcome in the survey (Fig. 5). Further on, each estimate requires three input values
(min, max, and optimal) from a respondent in order to meet the
requirements of FDM.
Data collection and respondents characteristics
To collect the data two approaches have been employed. First, a
semi-structured questionnaire with multiple-choice answers has
been used to validate a game tree. Second is the fuzzy Delphi
method (FDM) and it is used to estimate preferences and predictions of the game outcomes. FDM derived from combining the
traditional Delphi method and fuzzy set theory. Various researchers
contributed to the origin of this approach (Hsu and Chen, 1996;
Ishikawa et al., 1993; Murray et al., 1985; Noorderhaven, 1995).
FDM mainly aims to improve the characteristics of the traditional
Delphi questionnaire that has the tendency that both the questions and the answers are indistinct. Additionally, there is a notable
problem to solve the fuzziness in expert consensus in group decision making. These two key issues resulted in the proposed (Hsu

B. Glumac et al. / Land Use Policy 47 (2015) 6677

73

Fig. 4. Game validation experiment

and Chen, 1996) similarity aggregation method (SAM). Although,

Hsu and Chen (1996) used a trapezoidal fuzzy number to estimate
the group consensus, this study uses the triangular fuzzy number because it is the least demanding for the respondents (e.g. Klir
and Yuan, 1995). Results are discussed in Section Estimated results
with fuzzy Delphi method.
The data in this experiment is collected with the survey tool
Berg Enqute System 2007; the period the survey took place was
April to September 2011. This section reports briey on the sample
size, response rate, and the distribution of respondents characteristics.

The experiment consists of two data collection parts in which the

response rate is different. For the validation of the game trees,
86 respondents participated in the survey out of 563 that visited
initially the on-line survey; response rate: 15.28%. For estimating
the preferences and most expected game outcomes, the number of
respondents dropped during the survey. Precisely, 43 respondents
completed this part of the experiment thus making the corresponding response rate of 7.64%. The distribution of the respondents
characteristic is similar in the two parts of the experiment.

Sample size and response rate

The following table gives an insight of the respondent types

(Table 2). The division of respondents between player M and player
D shows a ratio of 3:7. Preferably, this should be improved. However, since the main concern is to have the minimum requirement
of the number of experts per group (1215 respondents per group
for fuzzy Delphi method), the unfavorable ratio has been overlooked. The overall experience in years and the BR experience of
the respondents are presented in Table 3. This does not completely
meet with the descriptive purpose, but it is nevertheless used in the
FDM-SAM calculation. Overall, it is possible to validate the game
trees in this experiment because the response rate is regarded as

The experiment relies on FDM to collect and asses the

respondents observations. The minimum requirements of this
experiment are determined by the rule of thumb that each group
of respondents should have 1015 people each, as recommended
in classical Delphi literature (Delbecq et al., 1975).
The groups of respondents that were investigated are: (1) independent developers; (2) contractors; (3) asset developers; (4)
development agencies; (5) municipalities. Type (1), (2), and (3)
respondents refer to player D; type (4) and (5) refer to player M.

Types of respondents, years of experience and BR experience

Fig. 5. Game rating and game choice experiment

74

B. Glumac et al. / Land Use Policy 47 (2015) 6677

Table 2
Type of respondents.

Table 5
Building claim: SPNE.

Type of respondent

Frequency

(1) Independent developers

(2) Contractors
(3) Asset developers
(4) Development agencies
(5) Municipalities
Total

21
26
13
12
14
86

Percent
24.4
30.2
15.1
14.0
16.3
100.0

End branch

Player

W = (a, b, c)

BC a

M
D
M
D
M
D
M
D

4.74
5.37
1.78
1.74
2.53
4.58
2.61
2.78

BC r
NBC a
NBC r

6.29
6.51
4.17
4.26
6.59
6.11
4.50
5.34

7.85
8.22
6.34
6.93
7.98
7.91
6.53
7.56

SPNE

6.30
6.70
4.10
4.31
5.70
6.20
4.55
5.23

Table 3
Characteristics: years and BR experience.
Characteristic

Levels

Frequency

Percent

Years of experience

<10 years
10 years

51
35

59.3
40.7

BR experience

Yes
No

35
51

40.7
59.3

acceptable and the characteristics of the respondents tend to be an

objective representation of reality.
Game experiment results
Four parts of the experiment have already been previously
explained above. In this section, the results are described for the
validation of the game tree and the estimation of the respondents
preferences and predictions of the game outcomes.
Validated game tree structure
A valid game tree structure guarantees that the game is correctly
modeled. For that purpose, multiple answers were used to check if
the actions (branches) are correctly assumed (Fig. 4.) as mentioned
previously.
In Table 4, the rst column lists the assumed and eventually
new branches that form the structure of the game tree. The list of
branches is formed as the game tree is drawn, downwards starting
with the root on top. For example, the rst listed branch is BC. This
action belongs to the initial decision node when the player M plays
BC. Further, the branch BC a denotes an action from the decision
node when player D plays a as a reaction on player Ms previous
move BC. In addition, all OTHER branches refer to the eventual
new branch starting from the certain decision node. The second
Table 4
Building claim: Game tree structure validity.
Branch

Valid

Frequency

Percent

BC

No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes

13
73
76
10
84
2
17
69
74
12
73
13
40
46
55
31
69
17

15.1
84.9
88.4a
11.6
97.7
2.3
19.8
80.2
86.0a
14.0
84.9
15.1
46.5
53.5
64.0a
36.0
80.2
19.8

NBC
OTHER
BC a
BC r
BC OTHER
NBC a
NBC r
NBC OTHER

a
Note: The branches that have the high percentage of the respondents replying
that the branch is not valid; the explanation follows in the text below.

column in the table reports the status if an assumed branch is valid,

expressed by no and yes. Finally, the frequency and related percent
column shows the number and percentage of respondents that
validated a certain branch by using the multiple answer buttons
in the survey tool. In total, 86 respondents participated in this
part of the experiment. Both the frequencies and percentages are
estimated by SPSS .
Although at a rst glance this is a doubtful feedback, the main
reasons for such a report are problems with the quality of the
data. Some respondents did not correctly observe instructions.
They gave mainly mutually exclusive responses, instead of multiple answers as designed in the questionnaire. Taking for example
the branch NBC at the initial decision node, 88.4% of the respondents did not check this branch in multiple answers when using
the survey tool. The assumption is that the respondents stated their
preferences instead of stating if this branch could be present at
the negotiation table at all. In addition, some of the respondents
claimed that there are other possible branches in certain decision nodes, like 19.8% of respondents for the branch NBC OTHER.
In this case, they mainly suggested completely new negotiation
issues.
Similar results but also the problems occur in the bargaining
game. Tables are available in Appendix 1.
Estimated results with fuzzy Delphi method
The nal part of the experiment concerns data collection and
the estimations of the respondents preferences about the negotiation outcome over a single negotiation issue. In addition, this
sub-section provides the game analysis using SPNE.
As explained, experimentally estimated game data provide the
empirical evidence about the principles of strategic behavior on the
one hand, and the information necessary for any game analysis on
the other (Crawford, 2002). For that purpose, a questionnaire based
on FDM was employed. As described previously, each of the end
nodes reects the possible outcomes of a game. In Tables 5 and 6,
two different estimations are presented and both calculated with
FDM in the Matlab program. The rst column describes all end
nodes or branches of the ultimatum game. The next column indicates if estimation relates to player M or player D. The following
three columns (W) represent the fuzzy number estimations of an
outcome, while column (S) provides the corresponding defuzzied
Table 6
Building claim: most probable outcome.
End branch

Player

W = (a, b, c)

BC a

M
D
M
D
M
D
M
D

2.41
2.24
1.77
1.70
2.27
2.41
2.56
2.61

BC r
NBC a
NBC r

4.23
4.47
3.79
3.33
4.41
3.97
4.25
3.97

6.56
6.79
5.10
5.78
6.22
5.92
6.43
5.78

Outcome

4.40
4.50
3.55
3.60
4.30
4.10
4.40
4.12

X
X

B. Glumac et al. / Land Use Policy 47 (2015) 6677

number. All calculations are based on Hsu and Chen (1996). Still a
distinction must be made between these two tables.
Table 5 represents the estimation of players M and D general
preferences over the building claim game outcomes. In technical
terms, the general preference covers all nine treatment combinations discussed in the setting up of the game-theoretical
experiment. In addition, the preferences (S) are regarded as indication of the payoffs. This can also be traced in the game tree (Fig. 2).
After generating the payoffs, the game can be solved by backward
induction referring to SPNE within the perfect information games,
explained previously. In the far right column, the indicated SPNE is
marked with X. On the other hand, Table 6 reports on the respondents aggregated opinion about the most probable outcome. The
X in this table refers to the highest score (S) for the most probable
outcome. The score X in this table is assigned separately for player
M and player D. This implies that the most probable outcome could
be reported differently for each player. For the bargaining game,
the same procedure is conducted. The results of this game are in
Appendix 2.
When comparing the last columns of Tables 5 and 6 and
Appendix 2, there is an evident match between the gametheoretical solution provided by SPNE based on the preferred
outcomes and the most probable outcome estimated by the respondents. Several conclusions can be drawn based on this perfect t.
In the studied context, the rst conclusion is that the players
act rationally. This perfect t indicates that players can perceive
their own moves and strategies, but also the other players moves
and strategies. Therefore, the most common critic of game theory about ill-assumed rationality of a decision maker has been
avoided in the studied context. This rational behavior probably can
be traced to the fact that the players are represented by the aggregated opinion. Besides the implemented fuzzy Delphi method to
collect aggregated opinions, this study introduces the fractional factorial experimental design (Table 1) to provide a statistically valid
experimental set-up. Such design equips respondents (player) with
a proper information background about a decision problem thus
avoiding biased reports on both preferred and expected game outcomes. Finally, this methodological approach evidently led to the
empirical results that discover rational behavior of the players in
the studied context and the proof of players rational behavior is
the precondition to solve a SPNE with the backward induction.
Second, the perfect t between estimated SPNE and probable
outcome implies that the bargaining games are suitable to interpret
the behavior of the real-life negotiations in BR projects. When the
ultimatum game is representing the real world problem, it is often
seen that the bargaining pie has even or near-even splits, like in this
experiment as well. This implies that the game outcome estimated
for example with SPNE is based on very small differences in the
payoffs value. However, this is not an imperfection of game theory
but rather a common mistake of an analyst to dene all attributes
inuencing each players payoff. Commonly, an analyst misses to
include hidden attributes such as fairness that would indicate a
willingness of a player to reject one-sided offers. Another common
critic is the relevance of the estimated SPNE in the case when the
payoffs increase in value but the ratio over all game outcomes
remains the same. This is only a problem when the payoffs are
represented by the attributes levels that have certain value and not
represented by the players utility that captures all attributes and
their levels. By applying an experimental game theoretical framework, both obstacles have been overcome. Together the described
institutional-economic context of the game and designed game
conditions represent the most important attributes and their levels
that are crucial for the game. In addition, any hidden attributes are
also reected in the payoff because the respondents themselves
reported the payoffs by fuzzy Delphi method (Fig. 5). The same
method enables to represent payoffs as utility based on the players

75

preference of previously identied attributes and their levels. The

perfect t shows that the players behave exactly as we would
expect them to, meaning that the mentioned common mistakes
has been avoided and that experimental game theory can be
successfully used in elaborated context.
Besides the included set of attributes, their levels, and a way
of quantifying the payoffs, the projected SPNE might also be inuenced by a negotiation culture specic for a certain region or a
country. Therefore, the game outcome in which the players enter
a PPP with building claim agreement can be regarded as typically
Dutch: public and private bodies in urban development are commonly prepared to share nancial risks and returns. This outcome is
a well-established Dutch urban development model (Samsura et al.,
2010), therefore the reported SPNE can be regarded as a part of a
repeated game. In this type of games, players play the game multiple times. In addition, this repeated game is a nite game with
undened time span. It is nite because the institutional-economic
environment is changeable, and it is not dened in time because
these processes are long and unpredictable. Usually, these games
lead to a socially optimum strategy. This can be interpreted as a
negotiation culture and one essential part of nitely repeated
games with undened time span is punishing players who deviate
from this strategy. The punishment may reect in lower payoffs
for both players. Therefore, the estimated SPNE should again be
regarded as typical for the Dutch context. An interesting addition
to this study would be a comparison with SPNE of a similar game
but within a different institutional-economic context.
Within this institutional-economic context, it is also possible to
conclude on both players preferences by only observing the game
payoffs. First, it can be noticed that player D has higher payoffs
for each of the games outcomes compared to player M (Table 5
and Fig. 2). This can be interpreted as that player D is more motivated to join a PPP. Such an outcome should contribute to higher
preferences to develop a site because this is the key activity of a
developer. Further, if the deal would be rejected (BC r; NBC r) then
player D reports higher payoff when rejecting a deal without building claim agreement (BC r < NBC r). Therefore, it is clear that player
D perceived the potential consequence of accepting a less favorable deal. On the other hand, the higher payoff of player M for the
same end branch is counterintuitive at rst. However, such a payoff
indicates that actually having a building claim agreement is more
favorable for this player too. This can be mainly explained that it
is in municipal interest to have also a rmly attached developer
that will provide secure development of building plots. Alternatively, we might say that player M also prefers to share the BR risks
that usually exceed greeneld development risks (De Sousa, 2002).
These preference ndings could be used when negotiating other
issues not covered by this study, such as nancial obligations of
both players.
In addition, these ndings provide a base to check the possibility of reaching a different game outcome. In the game theoretical
framework, this can be achieved with interventions such as: (1)
changing the information of the involved players, (2) changing the
pay-offs, (3) changing the playing rules (Jost and Weitzel, 2008).
Conclusions
Decision processes in urban (re)development projects have
become more complex. Therefore, it is useful to nd theories that
can support the governance of such processes. Of special interest in this paper are urban (re)development processes that do not
depend on the individual choice by one stakeholder, but depend
on multiple stakeholder decisions. A game-theoretical framework
has been introduced in this paper to provide a better understanding of decision-makers interactive behavior and expected decision
outcomes, along with the recommendations concerning the

76

B. Glumac et al. / Land Use Policy 47 (2015) 6677

application of intervention strategies in conict situations. Game

theory also provides a relevant framework to gather empirical data
in order to interpret the respondents interactive behavior. Therefore, an experiment has been carried out to indicate which parts
of the basic theory are most useful in predicting and identifying
behavioral parameters that theory does not reliably determine. For
that purpose, an illustration of urban (re)development process is
provided by elaborating the three specic negotiation issues in the
initiative phase of a BR project.
An experimental game theory procedure has been introduced
and it consists of eight consecutive steps: (1) selection of a game
class; (2) selection of a game form; (3) selection of a game solution
concept; (4) description of the institutional-economic environment
(5) design of the game conditions within the game set environment; (6) validation of the assumed game structure; (7) estimation
of the respondents preferences and predictions of the game outcomes; (8) interpretation respondents feedback with SPNE and
direct game outcome predictions. The suggested procedure guarantees that the decision moment is properly isolated and that the
experiment is statistically valid.
In general, the benets of using an experiment game theoretical framework to planning and development practice can be found
in applying a prescriptive decision-making approach, instead of
relying only on the normative nature of game theory. The idea of
the perspective approach is to provide usable outcomes such as
novel perspectives through decision aids. Therefore, this approach
is not so much concerned with a theoretical contribution but rather
with the pragmatic value provided to the end user. In the described
institutional-economic, an analyst would be able to investigate the
inuence of different external effects (Table 1) that would inuence the outcome of negotiations. However, to be able to perform
such analysis additional data is required since it is not possible to
provide an empirical evidence of the use of a certain method and
to show its performance on the same data set.
Furthermore, this paper provides an empirical example of the
negotiations taking place in a BR project with respect to the choice
for the building claim model, the future land use and reparcelling
of the land. In the studied context, it is rst shown that the assumption of rational players is correct. The evident match between the
game-theoretical solution provided by SPNE and the most probable outcome estimated directly by the respondents suggests that
the players do act rationally. This test is required to assure that the
backward induction can be used to estimated SPNE. In addition,
the perfect match implies that negotiations in BR can be successfully described by the bargaining games because players behave
exactly as we would expect them to. However, the estimated SPNE
should be regarded as typical for the Dutch context. The game is a
well-established Dutch urban development model (Samsura et al.,
2010), therefore the reported SPNE can be regarded partly as the
outcome of a repeated game with nite and undened time span.
As such, our analysis provides empirical evidence of the Dutch tradition of public and private bodies entering into public private
partnerships in urban (re)development, willing to share nancial
risks and returns, even in very complex situations. Anyhow, for
each game, all negotiable attributes are described with the according ordinal values. As the ndings reveal, the increase in attributes
values leads to the higher-level acceptance of a deal. This reects
potentials for interventions in order to reach a certain outcome. As
an alternative, these ndings can be used to set a borderline for an
attractive offer. Therefore, this article has demonstrated how different negotiation issues in BR can be quantied and what are the
strategies under dened institutional-economic environment for
the two major players in urban development process, a municipality and developer.
Finally, the reader should realize that game theory always
presents an abstraction. Not all engagements of real-life

interaction processes in BR projects have been covered, and

deliberately so. The aim is to represent the interaction structure
as a tool to understand the behavior of the involved parties, not to
completely mimic the real-world to every detail. As mentioned,
the negotiation attributes have been represented as separate
single games. This implies that the proposed games provide a rm
base (structure and procedure) for the similar attributes (with the
similar institutional-economic context and type of variable) that
need to be studied in the negotiation context.

Appendix A. Supplementary data

Supplementary data associated with this article can be
found, in the online version, at http://dx.doi.org/10.1016/j.
landusepol.2015.03.018.

References
Adams, D., Disberry, A., Hutchison, N., Munjoma, T., 2001. Ownership constraints to
browneld redevelopment. Environ. Plan. A 33, 453477.
Alker, S., Joy, V., Roberts, P., Smith, N., 2000. The denition of browneld. J. Environ.
Plan. Manage. 43, 4969.
Arentze, T., Timmermans, H., 2003. A multiagent model of negotiation processes
between multiple actors in urban developments: a framework for and results of
numerical experiments. Environ. Plan. B 30, 391.
Blokhuis, E., Snijders, C., Han, Q., Schaefer, W., 2012. Conicts and cooperation in
browneld redevelopment projects: application of conjoint analysis and game
theory to model strategic decision making. J. Urban Plan. Dev. 138, 195205.
Bowles, S., 2004. Microeconomics: Behavior, Institutions, and Evolution. Princeton
Univ. Press, Princeton, NJ.
CABERNET, 2002. Browneld Denition. Concerted Action on Browneld and Economic Regeneration Network.
Camerer, C.F., 2003. Behavioral Game Theory: Experiments in Strategic Interaction.
Carroll, D.A., Eger Iii, R.J., 2006. Brownelds, crime, and tax increment nancing. Am.
Rev. Public Admin. 36, 455477.
Chen, Y., Hipel, K.W., Kilgour, D.M., Zhu, Y., 2009. A strategic classication support
system for browneld redevelopment. Environ. Model. Softw. 24, 647654.
Coiacetto, E., 2001. Diversity in real estate developer behaviour: a case for research.
Urban Policy Res. 19, 4359.
Colman, A., 1995. Game Theory and Its Applications in the Social and Biological
Sciences. Routledge, London.
Crawford, V., 2002. Introduction to experimental game theory. J. Econ. Theory 104,
115.
De Sousa, C.A., 2002. Measuring the public costs and benets of browneld versus
greeneld development in the Greater Toronto area. Environ. Plan. B 29,
251280.
Delbecq, A.L., Van de Ven, A.H., Gustafson, D.H., 1975. Group Techniques for Program
Planning: A Guide to Nominal Group and Delphi Processes. Scott. Foresman,
Glenview, IL.
Forester, J., 1987. Planning in the face of conict: negotiation and mediation strategies in local land use regulation. J. Am. Plan. Assoc. 53, 303314.
Ganser, R., Williams, K., 2007. Browneld development: are we using the right targets? Evidence from England and Germany. Eur. Plan. Stud. 15, 603622.
Glumac, B., Han, Q., Smeets, J., Schaefer, W.F., 2011. Browneld redevelopment features: applying fuzzy Delphi. J. Eur. Real Estate Res. 4, 145159.
Grimsey, D., Lewis, M.K., 2002. Evaluating the risks of public private partnerships
for infrastructure projects. Int. J. Project Manage. 20, 107118.
Hahn, G., Shapiro, S., 1966. A Catalog and Computer Program for the Design and Analysis of Orthogonal Symmetric and Asymmetric Fractional Factorial Experiments.
General Electric, Research and Development Center, New York, pp. 73.
Hieminga, G., 2006. Projectontwikkeling Marketperspectief & Integrale Gebiedsgerichte Herstructurering. ING Bank, Economisch Bureau ING, Amsterdam, pp.
116.
Hsu, H.-M., Chen, C.-T., 1996. Aggregation of fuzzy opinions under group decision
making. Fuzzy Sets Syst. 79, 279285.
Ishikawa, A., Amagasa, M., Shiga, T., Tomizawa, G., Tatsuta, R., Mieno, H., 1993. The
maxmin Delphi method and fuzzy Delphi method via fuzzy integration. Fuzzy
Sets Syst. 55, 241253.
Jost, P.-J., Weitzel, U., 2008. Strategic Conict Management: A Game-Theoretical
Introduction. Elgar, Cheltenham, UK.
Klir, G.J., Yuan, B., 1995. Fuzzy Sets and Fuzzy Logic: Theory and Applications. Prentice Hall PTR, USA.
Koppenjan, J.F.M., Enserink, B., 2009. Publicprivate partnerships in urban infrastructures: reconciling private sector participation and sustainability. Public
Admin. Rev. 69, 284296.
Lange, D., McNeil, S., 2004a. Browneld development: tools for stewardship. J. Urban
Plan. Dev. 130, 109116.
Lange, D., McNeil, S., 2004b. Clean it and they will come? Dening successful browneld development. J. Urban Plan. Dev. 130, 101108.

B. Glumac et al. / Land Use Policy 47 (2015) 6677

Lange, D., Wang, D., Zhuang, Z., Fontana, W., 2013. Browneld development selection
using multi-attribute decision-making. J. Urban Plan. Dev. 140, 04013009.
Liang, Y.H., Guo, P., Hu, J.F., 2008. Analysis of browneld redevelopment by evolutionary game. In: IEEE International Conference on Industrial Engineering and
Engineering Management, Singapore, pp. 481485.
Louviere, J.J., Hensher, D., Swait, J.D., 2000. Stated Choice Methods: Analysis and
Applications. Cambridge University Press, Cambridge.
Ma, L., Arentze, T., Borgers, A., Timmermans, H., 2007. Modelling land-use decisions under conditions of uncertainty. Comput. Environ. Urban Syst. 31,
461476.
Martnez, F.J., Henrquez, R., 2007. A random bidding and supply land use equilibrium model. Transport. Res. B 41, 632651.
Mayer, I.S., van Bueren, E.M., Bots, P.W.G., van der Voort, H., Seijdel, R., 2005. Collaborative decisionmaking for sustainable urban renewal projects: a simulation
gaming approach. Environ. Plan. B 32, 403423.
Miles, M.E., Berens, G., Weiss, M.A., 2007. Real Estate Development: Principles and
Process. Urban Land Institute.
Mu, L., Ma, J., 2007. Game theory analysis of price decision in real estate industry.
Int. J. Nonlin. Sci. 3, 155160.
Murray, T.J., Pipino, L.L., van Gigch, J.P., 1985. A pilot study of fuzzy set modication
of Delphi. Hum. Syst. Manage. 5, 7680.
Neumann, J., Morgenstern, V.O., 1944. Theory of Games and Economic Behavior.
Wiley, New York.
Noorderhaven, N., 1995. Strategic Decision Making. Addison-Wesley, UK.
Osborne, M.J., 2004. An Introduction to Game Theory. Oxford University Press, New
York.
Page, G.W., Berger, R.S., 2006. Characteristics and land use of contaminated browneld properties in voluntary cleanup agreement programs. Land Use Policy 23,
551559.
Peiser, R.B., 2007. Regenerating Older Suburbs. Urban Land Institute, Washington,
USA.
Peiser, R.B., Frej, A.B., 2003. Professional Real Estate Development. Urban Land Institute, Washington.
Pfrang, D.C., Witting, S., 2008. Negotiating ofce lease contracts: from a
game-theoretical towards a behavioural view. J. Eur. Real Estate Res. 1,
88105.
Raiffa, H., 2002. Negotiation Analysis: The Science and Art of Collaborative Decision
Making. Harvard University Press, Cambridge.
Rasmusen, E., 2007. Games and Information: An Introduction to Game Theory. Blackwell Publishing, Malden, UK.
Samsura, A., van der Krabben, E., 2011. Funding transport infrastructure development through value capturing: a game theoretical analysis. In: Nunen, J.A.E.E.,
Huijbregts, P., Rietveld, P. (Eds.), Transitions Towards Sustainable Mobility.
Springer, Berlin, Heidelberg, pp. 5980.
Samsura, A.A., van der Krabben, E., 2012. Negotiating land and property development: a game theoretical approach to value capturing. J. Eur. Real Estate Res. 5,
4865.
Samsura, D.A.A., van Deemen, A.M.A., van der Krabben, E., van der Heijden, R.E.C.M.,
2013. Bargaining for value capturing: a game-theoretical analysis and experimental approach. Environ. Plan. B 40, 234253.
Samsura, D.A.A., van der Krabben, E., van Deemen, A.M.A., 2010. A game theory
approach to the analysis of land and property development processes. Land Use
Policy 27, 564578.

77

Shoham, Y., Leyton-Brown, K., 2009. Multiagent Systems: Algorithmic, GameTheoretic and Logical Foundations. Cambridge University Press.
Sibdari, S., Pyke, D.F., 2010. A competitive dynamic pricing model when demand is
interdependent over time. Eur. J. Oper. Res. 207, 330338.
Sounderpandian, J., Frank, N., Chalasani, S., 2005. A support system for mediating brownelds redevelopment negotiations. Ind. Manage. Data Syst. 105,
237254.
Stengel, B.V., 2008. Game Theory Basics. Department of Mathematics, London School
of Economics, London, UK.
Syms, P., 1999. Redeveloping browneld land: the decision making process. J. Prop.
Invest. Finance 17, 481500.
Tam, C.M., Thomas, K.L.T., 2011. Conict analysis study for public engagement
programme in infrastructure planning. Built Environ. Project Asset Manage. 1,
4560.
Tam, C.M., Zeng, S.X., Tong, T.K.L., 2009. Conict analysis in public engagement
program of urban planning in Hong Kong. J. Urban Plan. Dev. 135, 5155.
Thomas, M.R., 2003. Browneld redevelopment: information issues and the affected
public. Environ. Pract. 5, 6268.
Van der Krabben, E., Jacobs, H.M., 2013. Public land development as a strategic
tool for redevelopment: reection on the Dutch experience. Land Use Policy
30, 774783.
Walker, S., Hipel, K.W., Inohara, T., 2008. Attitudes and coaliions within browneld
redevelopment projects. In: IEEE International Conference on Systems, Man and
Cybernetics, Singapore, pp. 29012906.
Wang, L., Fang, L., Hipel, K., 2011. Negotiation over costs and benets in browneld
redevelopment. Group Decis. Negot. 20, 509524.
Wang, L., Fang, L., Hipel, K.W., 2007. A game-theoretic approach to browneld redevelopment: negotiation on cost and benet allocation. In: IEEE International
Conference on Systems, Man, and Cybernetics, Montreal, QC, pp. 18001805.
Wang, Q., Hipel, K.W., Kilgour, D.M., 2008. Conict analysis in browneld redevelopment: the erase program in Hamilton, Ontario. In: IEEE International Conference
on Systems, Man and Cybernetics, Singapore, pp. 29132918.
Wang, W.-M., Lee, A.H.I., Peng, L.-P., Wu, Z.-L., 2013. An integrated decision making model for district revitalization and regeneration project selection. Decis.
Support Syst. 54, 10921103.
Weihe, G., 2005. Publicprivate partnerships: addressing a nebulous concept. In:
Kaspersen, L.B. (Ed.), 10th International Research Symposium on Public Management (IRSPM X). International Center for Business and Politics Copenhagen
Business School, Copenhagen.
Wu, Y., Zhang, X., Skitmore, M., Song, Y., Hui, E.C.M., 2014. Industrial land price and
its impact on urban growth: a Chinese Case Study. Land Use Policy 36, 199209.
Yount, K.R., 2003. What are brownelds? Finding a conceptual denition. Environ.
Pract. 5, 2533.
Youse, S., Hipel, K., Hegazy, T., 2010. Considering attitudes in strategic negotiation over browneld disputes. J. Legal Aff. Dispute Resol. Eng. Construct. 2,
240247.
Youse, S., Hipel, K.W., Hegazy, T., Witmer, J.A., Gray, P., 2007. Negotiation characteristics in browneld redevelopment projects. In: 2007 IEEE International
Conference on Systems, Man and Cybernetics, Institute of Electrical and Electronics Engineers (IEEE), New York City, NY, USA, pp. 18661871.
Zellner, M.L., Page, S.E., Rand, W., Brown, D.G., Robinson, D.T., Nassauer, J., Low, B.,
2009. The emergence of zoning policy games in exurban jurisdictions: informing
collective action theory. Land Use Policy 26, 356367.