Diffusion Dynamics of Energy-Efficient Innovations in the
Residential Building Environment
Stefan N. Groesser, MBA (Dipl.-Kfm. Techn.)!
Ph.D. Candidate, University of St. Gallen, Institute of Management, Switzerland
Research Associate, University of St. Gallen, Institute for Management, Switzerland
Research Associate, University of Beme, Center for General Ecology, Switzerland
stefan.groesser@ unisg.ch
www.ikaoe.unibe.ch/personen/groesser
Silvia Ulli-Beer, Dr.
Research Associate, University of Beme, Switzerland
Interfacultary Center for General Ecology, Switzerland
silvia.ulli-beer@ ikaoe.unibe.ch
Mohammad T. Mojtahedzadeh, Ph.D.
Attune Group Inc., Albany, U.S.A.
Mohammad@attunegroup.com
I Abstract
The overall target of the Swiss Energy Policy is to reduce CO -emissions and thereby to
achieve the vision of an energy-efficient ‘2000 Watt per Capita Society’. A major instrument
to contribute to the vision in the residential building sector is standards about energy-
efficient building designs. However, new residual houses according to the energy-efficiency
standard are not as often constructed as they ought to be. This short-term decision has a
long-term consequence. In the paper, a preliminary System Dynamics model about the system
‘residential building environment’ will be presented with the purpose to explain the
development of key variables of the building environment system. The analysis of possible
policies put forward that a sound understanding of the decision process of potential building
owners is necessary to create a beneficial system design which foster the diffusion of energy-
efficient building designs. Furthermore, the paper should enable discussions about the model
and methodological issues of the model creation process.
Keywords: Sustainable Development, Energy-Efficiency, Building Environment,
Diffusion, Innovation, Kyoto Protocol
1 Stefan Groesser is the contact person for further information about the model, feedback and criticisms or other
issues related to the paper.
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
1. Introduction?
The overall target of the Swiss Energy Policy is to reduce CO -emissions and thereby to
achieve the vision of an energy-efficient ‘2000 Watt per Capita Society’. A major instrument
to accomplish the vision is the Minergie-Standard, a standard about energy-efficient building
designs. However, new residual houses according to the energy-efficiency standard are not as
often constructed as they ought to be in order to achieve the 2000 Watt society vision
(Koschenz and Pfeiffer 2005). The research project ‘Diffusion dynamics of energy-efficient
Buildings (DeeB)’aims at carving out reasons for the comparably low acceptance rate of
energy-efficiency standards. For this purpose, a System Dynamics model about the system
‘building environment’ will be created with the purpose to explain the development of key
variables of the building environment system, such as number of energy-efficient and
traditional houses and sqm of living area in residual housing. The purpose of the paper is to
show the results of a first modeling workshop and the status of the system dynamics modeling
part of the research project. Ulli-Beer et al. provide a more comprehensive summary of the
status of the ‘DeeB’ research project (Ulli-Beer et al. 2006). Furthermore, the paper should
enable discussions about the model and methodological issues of the model creation process.
2. Context of the Problem
In the next chapters, relevant positions of the Swiss Environment Policy will be shown in
order to elaborate the position of the Minergie measure. Then, the research project will be
described for which the System Dynamics model will be built.
2.1 Environment and Energy Politics of Switzerland
The Swiss Constitution is the most basic institution anchoring environmental issues in
Switzerland and the Swiss Society. The Swiss Government is bound to the mandate stated in
the Constitution. Especially article 73 (sustainable development), article 74 (protection of the
environment), and article 89 (energy politics) are relevant with regards to environment and
energy politics. In addition, Switzerland signed the Kyoto Protocol in 1997 and committed
itself to reduce emissions of climate affecting green houses gases. On the basis of the Kyoto
Protocol, the Swiss Parliament passed a CO2-Law implementing the emission reduction
objective (BUWAL 1999). According to the CO2-Law, the emission of COz has to be reduced
by 10 percent below the reference value of 1990 (Swiss Federal Assembly 1999) until the
year 2010. The subfield of energy politics comprises concretizations of the high-level
objectives of CO2-emission reduction. Its content can be divided into the Energy Policy
Vision of the 2000 Watt Society and the programs elaborated to compass the vision. In the
following, the vision and the programs will be discussed separately.
The vision of a post-industrial 2000 Watt per capital Society? was developed by the council of
the Swiss Federal Institute of Technology in the year 1998 (ETH Council 1998). Its main goal
is to reduce the primary energy consumption to 2000 Watt per capita by means of
technological innovations and to maintain or increase the actual standard of living
simultaneously (Jochem et al. 2004; Spreng and Semadeni 2001). What the measure ‘2000
Watt’ signifies will be explained by a comparative example. Today’s global energy
consumption is on average 2000 Watt per capita. The energy consumption figures, which
? The abbreviation ’ee’ stands for energy-efficiency or energy-efficient and is used throughout the paper.
> The 2000 Watt measure is the calculated average energy consumption for a period of one year per person. In
physics, the unit ‘Watt’ signifies power and not energy consumption or work. If, in addition to absorbed power,
time is considered, the energy consumption is obtained. I. e., 2000 Watt*Y ear/(Y ear*Person) would be the
correct unit of measure for the energy consumption target. This is equivalent to 48 kWh/day, which is a more
common notation for energy consumption.
Page 2 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
depend highly on the technological level of the observed country, vary from 500 Watt in
Ethiopia to more then 10.000 Watt in the United States. Unfortunately, a further increase in
per capita energy consumption is expected (Novatlantis 2005; Spreng and Semadeni 2001).
The per capita energy consumption in Switzerland is approximately 5100 Watt. In the year
1960, the consumption per Swiss inhabitant was around 2000 Watt which is the present
objective of the vision. However, this does not indicate that the comfort and living standard
has to be reduced to the situation in 1960. It rather implies a change in the consumption
behavior towards a modem life style based on innovative technical solutions, new
management concepts and societal innovations which improve the efficiency of energy
utilization and which in turn reduce the total energy consumption (Spreng and Semadeni
2001). However, not only should there be a reduction of consumed energy; the consumption
must also transform from fossil to non-fossil fuel resources. The actual energy consumption
can be divided into two parts: energy obtained by usage of fossil and non-fossil fuel
resources. In the year 2000, the average Swiss energy demand of 5100 Watt was generated by
2/3 fossil and 1/3 non-fossil fuel resources. The total consumption per capita should be
reduced by the factor of 1.6 until the year 2050 whereas the fossil fuel consumption should be
halved (cf. Figure 1).
5000
Fy
a
§
#4000 {2100 —
i Non-Fossil|
3000
i 0) Fossil
8
1700
i 2000 250 I
q 3000
6 1000 2000 oy)
tae 1500
500
0
Year2000 Year2000CH Year2030 -- Year2050 ~—-Year2150
World TargetCH — TaetCH Target CH
Figure 1: Development of the Swiss Gross Energy Consumption until the Y ear 2150 (Koschenz
and Pfeiffer 2005). The consumption should be reduced from 5100 Watt to 2000 Watt what
corresponds to a reduction of 60%.
In the long-term, the gross energy consumption, both fossil and renewable fuel, should be
reduced from 5100 Watt to 2000 Watt with the fossil fuel part being 500 Watt, which is
equivalent to the long-term objective producing maximal one ton of CO per capita and year
(Novatlantis 2005). Noteworthy is that approximately one-half of the 5100 Watt is used for
the construction and maintenance of buildings (BFE 2005; Koschenz and Pfeiffer 2005;
Spreng and Semadeni 2001). Hence, to reduce the energy consumption of the building
environment is crucial to achieve the objectives of the 2000 Watt Society. Furthermore, the
vision’s long time horizon is mainly caused by the long renewal and refurbishment cycles in
the building environment.
After the vision, the second part of the Swiss Energy Politics, the programs, will be discussed.
The Swiss Energy Policy is transformed in 10-years lasting measure programs, which are
based on the cooperation with the private sector and follows the subsidiary principle of
federal interventions to achieve the 2000 Watt vision (Nauser 2005). The program lasting
from 1991 until 2000 was named ‘Energie2000’, an investment program with the objective to
increase the acceptance of energy-efficient technologies. Its budget added up to 560 Mio.
Page 3 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
CHF (BFE 2005). The follow-up program of ’Energie2000’ is ’EnergieSchweiz’. Since the
situation of CO2 emissions has worsened due to strong increases in energy consumption, the
government focuses its financial resources on improvements of innovative technologies and
measures in four areas: transportation sector, industrial processes, energy-efficient building
technologies, and renewable energies. Despite the current development in energy
consumption, the government retains the '2000 Watt Society’ as the strategic long-term
objective (BFE 2001). The soft target of the program is the development of energy awareness
in the Swiss Population. It is one prerequisite for voluntary actions, improvements in
education, creation of new technologies and efficient energy utilization (BFE 2001). The
measurable objective of the program is to reduce consumption of fossil fuels by 10 percent in
the upcoming 10 years. Thereby, the Swiss Government tries to achieve the Swiss CO2-Law
reduction target of 10 percent until the year 2010 with 1990 as base value. ’EnergieSchweiz’
is therefore the main program for the Swiss Energy and Climate Policy. Its total budget is not
pre-determined for the 10-year period. The budget allocation is negotiated on a yearly basis.
For 2001, it added up to 75 Mio. CHF. In the years 2004 and 2005, it was reduced to 49 Mio.
and 45 Mio. CHF, respectively (Nauser 2005).
The measures of the program ‘EnergieSchweiz’ can be distinguished in three parts: (1)
support of voluntary measures, (2) promotion measures for positive energy-related behavior,
and (3) regulations regarding target values and specifications (BFE 2001; Nauser 2005).* The
introduction of the Minergie-Standard is a measure which is located both in the first and
second category. In other words, the acceptance and application of the Minergie-Standard
occurs nowadays on a voluntary basis and can be funded by cantonal promotion programs.
Other current measures of the ‘EnergieSchweiz’ program are the climate levy for gasoline
and diesel as well as the CO -levy for combustibles. Both measures are parts of the third
category (BUWAL 2005; Swiss Federal Assembly 1999).
2.2 Project about Diffusion Dynamics of Energy-Efficient Buildings
As described above, the diffusion of energy-efficient building technologies is a focal point for
the Swiss Government. The Project ‘Diffusion of energy-efficient Buildings (DeeB)’ is
funded by the Swiss Federal Research Fund as part of the Research Program SNP 54.° The
project will be described in the following.
The project aims at analyzing and accelerating managerial and organizational adaptation
processes that foster the diffusion of pioneering energy efficient technologies in the building
sector. Psychological, managerial, and economic theories as well as the results of empirical
investigations about antecedents of behavior choices will be synthesized into a simulation
model for a middle-sized Swiss city. The model will shed light on dynamic interactions
between behavioral factors and contextual factors, thus explaining the diffusion of energy
efficient buildings in a community (Kaufmann-Hayoz, Bruppacher, and Ulli-Beer 2005). The
project team is hosted at the Interdisciplinary Center for General Ecology at the University of
Beme, Switzerland.
3. Review of Existing System Dynamic Models for the
Building Environment
A System Dynamics oriented literature review puts forward that less than a dozen articles are
publicly available which cover the issue of house construction or urban development. V ennix
provides a model which addresses how a housing association can maintain low cost housing
* Neuser provides an comprehensive overview of the specific measures in the ’EnergieSchweiz’ program
(Nauser 2005).
> Cf. also http://www.nfp54.ch/e.cfm.
Page 4 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
and at the same time guarantee its financial continuity (Vennix 1996). Eskinasi’s and
Rouwette’s model concentrates on the effect of a building development programs on the
social housing vs. commercial housing development (Eskinasi and Rouwette 2004). The
purpose of their model is in line with Forrester’s Urban Dynamics model (Forrester 1969).
Hennekam and Sander extend Forrester’s work and explicitly represent 40 urban sectors and
their interconnections via spatial interactions (Hennekam and Sanders 2002). Hong-Minh et
al. concentrate particularly on the determinants creating the demand for new private houses
(Hong-Minh, Childerhouse, and Naim 2000). And finally, Lee and Choi explore the effects
between recent Green Belt policies and the urban subsystem of Seoul (Lee and Choi 2004).
To conclude, the literature review shows that several studies about the building environment
have been conducted, either on the side of demand generation, or from the perspective of
single actors, or to assess effects of a specific policy. However, a comprehensive
representation of the system ‘building environment’ does not exist, which would enrich the
understanding of all involved actors, such as home owners, architects, building firms, and
official authorities.
4. A Preliminary Simulation Model
The preliminary model, which will be described and used for policy analysis, was created
during a group model building session held at the end of a one week introductory class in
System Dynamics at the University of Beme/Switzerland. The group consisted of 14
participants. The session lasted for five hours and the modeling team consisted of one
facilitator, one modeler and one reflector. In the following, we will describe the resulting
model in several steps according to the standard modeling process (Sterman 2000).
1. Problem Articulation
on (Boundary Selection) .
5. Policy
Formulation = [=
& Evaluation }
ame
3. Formulation
Figure 2: The Iterative Model Creation Process is a systematic way to articulate, formulate, test
and evaluate the created model and leads to a sound dynamic hypothesis about the problem
statement (Sterman 2000).
4.1 Problem Articulation and Boundary Selection
The first step in a model creation process is to formulate a problem statement and to define
the model purpose. Since this System Dynamics model will be used in a scientific research
project, we pose several research questions that have to be answered to achieve the goals of
the project.
Problem Statement
Energy-efficient innovations for the residential building environment can significantly
contribute to reduce the energy-consumption of the residential building stock (Groesser 2006;
Koschenz and Pfeiffer 2005). However, the system ‘Residential Building Environment’
Page 5 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
consists of several independent actors resulting in distributed decision making and is
therefore highly complex and unintuitive in its behavior. In addition, the system exhibits a
strong tendency toward policy resistance. The problem exists because the system reveals a
distributed decision structure due to the supply chain organization leading to sustained
suboptimal decision making. This phenomenon is well recognized and documented in the
literature (Duggan 2004; Fiddaman and Peterson 2000; Milling 1999; Sterman 1989, 1992;
Sterman and Fiddaman 1993). In other words, the system will most probably not cure or
optimize itself because of the individual decision possibilities of the agents.
Model Purpose
The model will shed light on dynamic interactions between behavioral factors and contextual
factors, and thus, first, explaining the diffusion of energy efficient innovations in the
residential building environments in a community. The second objective is to recommend a
system design that fosters reinforcing processes towards a sustained diffusion of energy
efficient innovations in the residential building environment within a community.
Research Questions for the Simulation Model
1. What is the physical structure of the residential building environment?
2. What are essential system variables that determine the rate of growth of the energy
efficient and traditional building stock for residential buildings? (Main Research
Question)
3. What are main forces that steer management processes’ of relevant actors in the value
creation chain of energy efficient buildings?
4. What are external conditions’ and behavioral factors® that inhibit or facilitate the
diffusion of energy efficient buildings?
5. How do these factors interact with each other?
6. What are possible intervention policies to foster the innovation adoption?
The elaborated preliminary model, presented in this paper, will not be extensive enough to
answer the posed questions, but it provides first ideas about the structure and sectors of the
system ‘building environment’.
Reference Modes as Dynamic Problem Definition
The reference mode is a graphical representation of the answers to the questions ‘What is the
historical behavior of the key concepts and variables? What might their behavior be in the
future?’ Furthermore, the reference mode defines the problem in a dynamical manner. It is the
red thread needed to guide the model conceptualization and simulation development.
However, criticism about the selected reference modes is appropriate. The question has to be
raised, if the selected reference modes are adequate definitions of the problem being studied.
Possible reference mode variables for the diffusion of energy-efficient innovations in the
residential building environment are:
e number of energy-efficient and traditional residential houses,
m? of energy-efficient and traditional residential area,
Energy demand per capita and m? of residential area,
Energy consumption for heating per capita and m2,
Energy consumption for warm water per capita and m?.
® Such as strategic planning process, supply capacity and hiring decisions.
7 Such as technical and cost factors, technological innovations, public initiatives, and market conditions.
8 Such as planning, decision making, and routines.
Page 6 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
The variables chosen for the reference mode are first ideas and need further improvement and
validation. Especially the following criteria will be used to determine if the selected variables
have the potential for key reference variables:
e Validation of the chosen variables by real data must be possible,
e Variables must represent the problem development, i.e., the root causes of the
problem must be included,
e Development of the variable must not be biased by any other development/trend
occurring at the same time horizon, or the trends must be detachable.
Figure 3 (left side) shows a possible development of the number of houses according to
energy-efficient and traditional building standards. The number of traditional houses (blue
graph, #1) increases in the time from 1970 until 2012. Energy-efficient building designs do
not exist, or are not known widely or not accepted. However, the number of traditional houses
does not further increase after knowledge about energy-efficient building designs has
disseminated through the society and new buildings are constructed according to the energy-
efficient building standards. The number of energy-efficient houses (red graph, #2) increases
with an average slope of 40 houses per year after the 2024 because the required information
and knowledge is available that foster the dissemination of the products.
Reference Mode: # of Houses Reference Mode: sqm of Residential Building Area
4.000 12M
3.000 900.000
2.000 600.000
1.000 300.000
0 ols
1970" 1988 2006 2024 2042 2060 2078 2096 2114 2132 2150 197071968 2006 2024 2042 2060 2078 2096 2114 2132 2150
years years
Houses: Traditional ons Houses Residential Area: Traditional: ————_j—_—J— FJ Houses
Houses: Enery- Efficient ——2———2—_—— 2D Houses Residential Area: Energy-Efficient —< i? Houses
Figure 3: Expected Reference Modes for the Simulation. The reference mode diagrams for the variables ‘# of
Houses’ and ‘sqm of residential building area’ are supported by statistical data until the year 2000. Thereafter,
the development is sketched according to the best information and considerations about the diffusion process.
Figure 3 (right side) depicts the reference mode ‘sqm of residential building area’ for the
categories ‘traditional’ and ‘energy-efficient’. The development of traditional residential area
is according to the development of number of traditional houses. Until the year 2010, the sum
of traditional living area (blue graph, #1) increases slightly and falls thereafter with a low
slope due to building demolition. Energy-efficient living space (red graph, #2), on the other
hand, increases strongly starting around 2004. The growth rate is further increased by the
trend to more residential area per capita. The development is damped and stopped when the
demand for residential area is satisfied. Considering the peak level at around the year 2010, it
is improbable that the tuming point will occur already in the year 2010. It could appear at a
later point in time. Further analysis has to determine when this peak will most probably
happen.
Both figures show typical graphs of a logistic function about the diffusion process indicating
that the system structure has to be related to Bass’s Diffusion Model (Bass 1969; Sterman
2000). Even though the graphs appear reasonable, further validation with statistical and non-
statistical data is required to determine if the chosen variables are good choices as reference
modes. However, there will always be a caveat regarding the reference mode because the
Page 7 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
model purpose is to model the diffusion process of an innovation that is about to occur. Two
possibilities exist to cope with the problem: First, use of data about diffusion processes in
another country, e.g., use the data from a country in which the energy-efficient building
innovations have already occurred, or second, use data from a similar diffusion process within
Switzerland that matches the diffusion process in the building environment as close as
possible.
Time Horizon
The determination of time horizon and reference mode is intertwined and subject to an
iterative process. The issue of time horizon selection puts forward the questions about:
e How far back in the past lie the roots of the problem?, and
e Whatis the reactivity of the system, and thus, when can intended effects be seen?
e How far into the future should we consider to avoid unintended effects of your policy
intervention?
We employ the following criteria for the selection of the time horizon:
e Availability of data for key reference variables,
e Roots of the problem lie in the historical time period,
e Time horizon is corresponding to the time frame indicated in the problem formulation
(short- vs. mid- vs. long-term) (see also the time horizon of the 2000 Watt Society in
(Koschenz and Pfeiffer 2005),
e Consideration of the characteristics of the problem context. In the building
environment, it is important to consider the long lasting building stock structure
leading, e.g., to long-term renovation or demolition cycles.
For the simulation model, we chose the time horizon from 1970 until 2150. Our decision was
guided by the aforementioned criteria.
4.2 Dynamic Hypotheses
The modeling step ‘dynamics hypotheses’ comprises the development of a theory about the
dynamics characterizing the problem in terms of underlying feedback and stock and flow
structure of the system (Sterman 2000). The dynamic hypotheses and the system structure
will be developed according to the following steps: model boundary chart, definition of
model variables and discussion of assumptions, subsystem diagram, formulation of causal
loop diagrams and finally, representation of the system structure as stock and flow diagrams.
Model Boundary C hart
The model boundary chart in Figure 4 shows the endogenously modeled variables in the
simulation model (yellow area), the variables being used as exogenous inputs (green area)
and variables that are not considered for the formulation of the simulation model (framed by
the blue oval). The description of relevant variables is provided below.
Definitions of Model Variables and Discussion of Assumptions
House Building Population
The house building population consists of entities that all want to build houses. In the
simulation, an exogenous driven value is used to represent the goal quantity of people who
want to build and own houses. It is rather unnatural and methodologically artificial to use a
fixed exogenous value for a dynamic value such as population. Apart from the fact that the
model is a first sketch and has to be elaborated in certain dimensions, a basic rationale is
behind this choice of modeling. According to the available statistical data from the year 1980
until 2004, the house building population has an arithmetic average of 14803 [persons] with a
Page 8 of 27
Stefan N. Groesser et al. 24th International System Dynamics Conference, Nijmegen
standard deviation of 330 [persons]. This indicates very weak population dynamics and
results in a relatively stable population size. The development of the male and female
population matches each other almost perfectly with an average difference of approximately
580 [persons]. Noteworthy is that the time series data (1980 - 2004) shows an increasing
trend for foreign inhabitants and a declining trend for the native inhabitants what could have
significant influences on social trends of family size and inhabited residential area.
Figure 4: Model Boundary Chart. It shows the variables included and excluded in the simulation.
model. Emphasis is laid especially on the endogenous and exogenous variables that describe and
explain the dynamic development of the problem being studied.
Potential Home Owners
The variable ‘potential home owners’ represents the amount of people who can possibly and
who want to build a home. It is assumed that the term home refers to single family houses. In
the model, we discuss only the development of home ownership of single-family houses and
do not concentrate on ownership of multi-family houses or on leasing of either single or multi
family houses. The potential home owners are differentiated in potential traditional and ee
home owners. The first want to build a house according to the traditional energy standard. Or
more precisely, they do not want to or cannot build an energy-efficient building. The second
group is the potential ee-home owners.
Decision for ee-Fraction
The decision for ee-fraction is a highly-aggregated variable embodying factors important to
the potential home owner's decision about an energy-efficient or a traditional home design. It
is a key variable because it strongly determines the path the potential home owner will take
due to, e.g., irreversible investment costs in building material and social capital. Possible
hidden factors could be classified in several clusters: psychological, context, social,
technological and political factors (Groesser 2006). In order to accelerate adaptation of
energy-efficient technologies, it is essential to understand the hidden factors indicating a need
for further disaggregation.
Page 9 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
Current Home Owners
People who currently inhabit and own a single-family house are named ‘current home
owners’. The definition comprises only private home owners inhabit their home. Institutional
investors are not included in the definition. In the simulation model, two groups of home
owners are differentiated: traditional and energy-efficient home owners. But, what is the
distinctive feature between traditional and energy-efficient houses? A first possibility is when
the building design complies with existing energy-efficiency standards, e.g., Minergie® or
Minergie-P®. Problematic is the fact that not all houses which comply with the
aforementioned standards actually have the label, because an additional bureaucratic process
is required to obtain the label. Consequently, not all houses complying with the standards
have been registered and labeled as energy-efficient houses. This problem of ‘complying with
the standard, but not officially labeled as complying’ becomes more important because the
dark figure grows daily (Groesser 2005). A second possibility would be to define a threshold
in energy-consumption, measured in MJ/(m**a). According to this definition, a house would
be energy-efficient when its average energy-consumption figure would be lower than this
threshold. Certainly, the threshold ought to evolve dynamically over time corresponding to
enhancements in energy-efficient technologies.
Transition to Energy-Efficient Home Owners
In case a home owner has built his house the traditional way, one possibility to change the
characteristic of the house is to refurbish it with energy-saving components. Transition to a
energy-efficient home owner represent the result of the refurbishment, i.e., in case houses are
upgraded and comply to the energy-efficiency standard, the home owners then belong to the
class of energy-efficient home owners. This formulation shows only one side of the
medallion. The second, more intuitive side is that the decision process of the traditional home
owner to obtain an ee-house is modeled which, in tum, causes the transition of the house type
from traditional to energy-efficient. For this formulation alternative, more decision variables
would be required. Further research has to determine if a disaggregation is required since
refurbishment is not part of the current research project, however, important in the near
future.
Physical Structure of House Environment in Construction
The physical structure of house environment in construction captures the physical creation
process of houses once it has started. The preceding process of house planning and multi-
party negotiations are not considered. In case the house is completed, it changes its
affiliations and no longer belongs to this category. A gain, a differentiation between traditional
and energy-efficient houses is drawn. While houses are in construction, an inhabitation is not
possible. The aggregation level of houses has been chosen because more valid data is
available from statistics. Other possible aggregation levels would have been: number of
apartments or m’ of residential area. The first measure can also be derived from statistics and
will probably incorporate in a future version of the model. The m?-measure would be derived
from the number of apartments and buildings with an average value of residential area for
each apartment size. Thus, by this estimation impreciseness would be added and model
accuracy would be reduced. However, if an estimation of the development of the energy-
consumption per residential area (MJ/m?*a) is intended, this would be a reasonable
calculation method.
Completed Houses
Completed houses are physical building structures either according to energy-efficient or
traditional standards. In the model, only completed houses can be inhabited by home owners.
The affiliation of the houses to certain classes is determined by the decision at the beginning
Page 10 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
and can only be changed by severe refurbishment. Small refurbishment projects cannot
change energy consumption of the house and, therefore, are not considered.
Active Architects
Architects in Switzerland are distinguished between ones that offer only traditional building
layouts and ones that offer building layouts in compliance with traditional and energy-
efficient standards. The difference between both types of architects being competencies and
knowledge about how the requirements are met in order to create layouts that comply with
energy efficient standards.
Elaboration of Single Hypotheses
In the System Dynamics model, some hypotheses are integrated which we want to make more
explicit and subject to further research and validation by means of literature, expert judgment
and empirical data.
H1: The sum of persons about energy-efficient innovations has a positive influence on the
decision for ee-fraction of potential home owners
Hypothesis H1 is well supported by scientific literature about diffusion of innovations.
Sterman formulates a diffusion model in System Dynamics syntax (Sterman 2000) based on
Bass’s theory of innovation diffusion (Bass 1969) and elaborates several versions. Hypothesis
H1 can be supported by the effect known as ‘word of mouth (WOM) effect’ in the first
diffusion phase. However, once the market demand is going to be satisfied, the word of
mouth effect is compensated by the effect of market saturation. H1 is, therefore, only valid if
the market is not saturated. Stated differently, H1 is too coarse and needs further detailing.
H2: The probability to decide for an energy-efficient house has a positive influence on the
house conversion rate from tradition houses to energy-efficient houses
The probability to decide for an energy-efficient house is operationalized by the variable
‘decision for ee-fraction’ and is formulated to aggregately represent the decision process of a
future home owner and is valid for the decision to build a new home. Hypothesis H2
implicitly assumes that the factors influencing the decisions of future home owners are also
valid for current owners of traditional houses with respect to the refurbishment of their
traditional house to an energy-efficient house. H2 proposes an indirect causal relationship
between the variable ‘decision for ee-fraction’ and ‘house conversion rate’. Bond et al.
support the reasoning with the theory about social interdependency in decision making among
probands (Bond and Smith 1996).
H3: The higher the demand for traditional houses, the lower will be the intention of architects
for further education
Hypothesis H3 implicitly assumes an interaction between the market forces of supply and
demand. The higher the demand for traditional houses, the higher will be the expected profit
of architects who offer this kind of service and therefore further learning as means to acquire
knowledge about energy-efficient building designs does not seem to be necessary to stay in
business. The current representation of the causal effect needs further disaggregation in order
to obtain reliably testable and relationships that can be validated. But the basic rationale
seems plausible.
Page 11 of 27
Stefan N. Groesser et al. 24th International System Dynamics Conference, Nijmegen
H4: The higher is the variable ‘decision for ee-fraction’, the higher will be the graduation
rate of architects with an education in ee
Hypothesis H4 proposes a relationship which influences the capacity of the supply side of the
market. If an indicator of the demand side, the variable ‘decision for ee-fraction’, increases in
value then architects in education will be more willing to change their course of study in
order to graduate with latest knowledge about energy-efficient building technologies to meet
the higher demand that is assumed to exist then. H4 has to be validated in expert interviews,
because we expect that literature will not provide further information about the issue.
Preliminary Sector Diagram
By means of the subsystem diagram, the variables of the simulation model are divided in
several subgroups and are graphically represented in an aggregated and coarse manner. Three
sectors could be identified for the discussed simulation model: Building owner sector,
physical building sector, and architect sector.
The building owner sector comprises the different types and states of ownership. The other
sectors are connected via the decisions of the home owners for traditional or energy-efficient
home designs and via the quantities of different home owners. The development of the
building owners quantities correspond to the development of the house structure quantities
indicating a tight relationship between both sectors. The sector ‘physical building traces the
changes of the physical building stocks for both traditional and energy-efficient houses. The
connections to other sectors is created by, first, the number of architects resulting in the
offered architect capacity, and second, by decision of future home owners about the types of
buildings. The last sector concerns the architects. Their decisions about first education (first
graduation) and second education (further learning) is represented. Links are created by the
potential home owners’ decisions. Figure 6 shows the sectors, the main variables for each
sector and the connections between them.
\\__ Amount of Completed Houses (trad/
Figure 5: Sector Diagram consists of Architect Subsystem, Building Owner Subsystem, and
Physical Building Subsystem. The major interconnections represent information dissemination
processes especially between the building owner and architect as well as between the building
owner and the physical building sector.
Page 12 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
A further elaborated model could also include sectors of the industry which comprises supply
of materials, labor, education demand, and financial comparison between different building
designs. Moreover, sectors about technology and the support system could be included, i.e.,
actors which have a supportive influence on the diffusion of energy-efficient technologies in
the building environment.
Feedback Loops
A feedback loop consists of variables connected by arrows denoting the causal influences
among the variables (Sterman 2000). In the following, only the most relevant feedback loops
incorporated in the simulation will be explained, because the explanation of approximately
200 feedback loops of this relatively small model is not feasible. Several feedback loops can
be identified which drive or balance the system. Loops L1 until L3 are positive feedback
loops fostering the diffusion process, whereas loops L4 until L8 are negative feedback loops
which lead to balancing or dampening tendencies of the system.
L1: Word of Mouth Effect by Potential ee Home Owners (Figure 6)
The word of mouth effect is a phenomenon that fosters diffusion processes. People with
knowledge about a new technology pass it to people who do not have this knowledge (Bass
1969; Sterman 2000). In the building environment, the total amount of persons convinced
about energy-efficiency has a positive influence on the decision of potential home owners to
build an energy-efficient house. As a result, more potential home owners become potential
energy-efficient home owners. Consequently, the sum of persons convinced to energy-
efficiency will further increase leading to a reinforcing process which is controlled by the
balancing mechanism L4 ‘Everybody has a house’.
L2: Market Pull for ee Architects (Figure 6)
The feedback loop ‘market pull for ee architects’ represents the effect the decision for energy-
efficient houses has on the supply of architects with special knowledge about energy-
efficiency. In more detail: An indicator of the market demand for energy-efficient houses,
represented by the potential ee home owners and indicated by the decision for ee-fraction, has
an effect on the amount of architects graduating with special knowledge about energy-
efficient designs. Production management literature coins the aforementioned mechanism
effect market pull effect (De Toni, Caputo, and Vinelli 1988; Kimura and Terada 1981). The
more of these architects are available, the higher is the supply and capacity to build energy-
efficient houses. Given the ceteris paribus condition, the number of potential energy-efficient
home owners will increase leading to a higher amount of convinced people and finally will
result in an increase of the decision fraction of ee.
L3: Word of Mouth Effect by Current ee Home Owners (Figure 6)
The first reinforcing loop is created by the word of mouth effect of potential ee home owners.
Also people, who already live in an ee-house, tell their acquaintances about their experience
and thus contributing to the word of mouth effect, which will lead to new potential ee home
owners and in tum to a higher number of current home owners of ee-buildings. Since the
causal chain is similar to the aforementioned feedback loop L1, it will not be explained in
detail. But why do we distinguish between the word of mouth of potential ee home owners on
the one hand and current ee home owners on the other hand? We expect the feedback loops to
have different strengths, e.g., that the word of mouth effect created by current ee home
owners is more powerful because the people currently living in energy-efficient houses are
either convinced by experience that their house choice was right or act in a way to be
consistent with their choice made and to avoid dissonant behavior (Stroebe, Jonas, and
Hewstone 2001). However, this stronger effect of current ee-home owners might contribute to
Page 13 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
the total word of mouth effect with a delay until the home owners actually inhabit the
building.
L4; Everybody Has a House (Figure 6)
The negative loop ‘Everybody has a house’ captures the fact that the finite house building
population will be satisfied, amongst others, by the word of mouth effects and limit the
diffusion process once the whole population is either owner of an energy-efficient home or in
the phase of building such a home. Starting with many people without own houses, more
potential home owners exist and resulting in a higher number of new potential owners of ee-
houses. In turn, they will increase the amount of ee-home owners and therefore the total
number of enrolled and current home owners. Consequently, the number of people without
houses will be reduced. The more positive feedback cycles foster the spread of information
and experience about energy-efficient building technologies, the faster will the number of
people without houses be reduced, and the more rapid will L4 stabilize the system at it upper
value.
L5: EE House Building Slows Down the Diffusion Process (Figure 7)
L5 describes the effect that ee-house construction requires architects with knowledge about
energy-efficiency leading to a smaller future construction rate of energy-efficient houses
because the architects supply is reduced by occupations. This loop could slow down the
diffusion process when not enough architects with knowledge about energy-efficiency are
available. L5 indicates that the number of architects with special knowledge is a bottle-neck
variable which needs further attention. However, since the occupation of the architects is
rather short (short time delay), this feedback loop is considered to have only insignificant
influences.
new potential ee
home owners
jy hongowners +)
L1: WOM effect by
potential ee home +
owners |? potential ee
potential home architects graduating home owners
owners with ee Dd tt)
+ + tt)
L3: WOM effect by current ee home
L2: Market Pull current ee home owners
home owners
for Architects owners
sedision for ee , sum of conte
new potential rachon fH ee persons
L4: Everybody
has a house total enrolled and
people without _; current home owners
houses —_ house building
population
Figure 6: Feedback Loop Diagram about potential and actual home owner dynamics. The
reinforcing loops, which are driven by word of mouth and market pull effects, lead to a strong
positive development of the number of current ee-home owners. These three loops are
controlled by the powerful balancing loops representing the limitation to this growth process
by the satisfaction of the house building population.
Page 14 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
L6: EE Architect Capacity Clear the ee-D emand (Figure 7)
Loop L6 depicts the reduction of the demand for energy-efficient houses by the transition of
architects from traditional to the energy-efficient branch, e.g., by further education. The
balancing process comes into play when the ee-architects exploit the demand for ee-houses.
The rationale is that more architects with knowledge about energy-efficiency will increase the
number of built energy-efficient houses. The more of such houses exist, the lower will be the
demand for energy-efficient houses relative to the demand for traditional houses, given a
fixed number of the total building population. Thus, architects having a traditional education
will be more resistant to change their competence portfolio and will rather remain in the
architect branch they are currently in. In other words, the ee-architects construct houses and
fulfill the demands of the population willing to build according to ee-standards. This leads to
a perceived market saturation effect of the potential ee-house owners. Hence, traditional
architects delay their decision to enlarge their product portfolio by energy-efficient designs,
because they want to serve the potential traditional home owners. This mechanism works
against the WOM-effect, but is comparably weak.
new ee house
building rate
fl
aval CC 15: eehous@ building © houses in
architects slows down the construction
“ diffusion
architects with ee t=) ee houses
knowledge = completion rate
We L6: ee architect
capacity clear the ee
demand
+ <house building
population>
iid, PRG es Gre
transition rate P
7 owners ratio _
Figure 7: Feedback Loops L5 and L6 are both balancing loops. They represent both the effect
the house construction has on, first, the number of available ee-architects, and second, the
motivation for traditional architects to become ee-architects.
L7: Traditional Architect Capacity Clear the Traditional Demand (Figure 8)
Loop L7 is similar to loop L6. It represents the fact that the reduction of the demand for
traditional houses will increase the number of architects who want to expand their business
and want to offer energy-efficient building designs as well. Just like L6, the strength of the
feedback loops is weak compared to L1 to L3.
L8: Traditional House Building Exhausts the ee-Architects Capacity (Figure 8)
The loop L8 is similar to the loop L5. It describes the effect that traditional house
construction requires architects with knowledge about traditional building design leading to a
smaller future construction rate of traditional houses because the required architect supply is
reduced. This loop hinders the diffusion of traditional buildings and thus fosters indirectly the
diffusion of energy-efficient buildings. Again like L5, the effect of this loop is considered
weak compared to L1 to L3.
Page 15 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
NX i ow]
potential ee home owners
ee to potential trad home
owners ratio
(—}
L7: traarchitect
capacity clear the trad trad houses
F demand completion rate
architects with trad 4
knowledge
¥ =
e trad houses in
L8: trad house buildin
available trad exhausts the capacity of construction
architects ee-architects
new trad house
building rate
Figure 8: Feedback Loops L7 and L8 are both balancing loops. They represent the effect the
house construction has on the number of available ee-architects, and the motivation for
traditional architects become ee- architects.
Stock and Flow Diagrams
The formulation of the simulation model as stock and flow diagrams will be provided in the
appendix. We assume the detailed diagrams are self evident and refrain from describing them.
Table 1 shows the values of the parameter used in the model.
No. Parameter Name Value Unit
1 |House Building Population 13287-14582 (var.) (1970-2004) Person
2 |Decision Time for House Building Population 25 Year
3_ [Initial Traditional Houses in Construction 20 House
4 [Initial Potential House Owners 100 Person
5_ [Initial Traditional Houses 2099 House
6 [Initial ee-Houses 0 House
7_|Inital Potential ee-House Owners 0 Person
8 Avg. Inhabitants per House 2 Dmnl
9 Initial ee- Houses in Construction 0 House
10 [House Completion Time 2 Year
11 [Demolishing Time Traditional Houses 60 Year
12 [Demolishing Time ee-Houses 80 Year
13 [Conversion Decision Time 1 Year
14 [Architects Learning Time i Year
15 Architects Work Life Time 35 Year
16 [Initial Traditional Architects 30 Person
17 [Initial ee-Architects 0 Person
18 [Graduating Architects 5 Person/Y ear
19 [Normal Graduation Ratio 0.1 Dmnl
20 [Architect Capacity 2.75 Person/Y ear
21 [Normal Architects Switching Ratio 0.25 Dmnl
22 [Start of EE-Movement 1998 Year
23 [Decision Time ‘Building Type' 3 Y ear
Table 1: Parameter Values of the Simulation Model. The parameter values with the grey
background color are values that are validated by statistical data. The other values have to be
validated in further expert interviews or by other means.
Page 16 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
4.3 Formulation of a Simulation Model
In the phase ‘formulation of a simulation model’, the specification of structure and decision
rules, estimation of parameters, behavioral relationships and initial conditions, as well as the
test for consistency with the purpose and boundary has to be addressed (Sterman 2000). The
stock and flow diagrams have been created with this step already in mind. Therefore, this sub-
step is already completed. Since real values were only partially available for the specific
region of Switzerland, the missing data was reasonably estimated. The purpose of the
simulation model is consistent with the stated purpose. The most crucial values in the
simulation model are up to now the number of traditional and energy-efficient houses, and
also the number of home owners in both classes. The other variables are supportive. The
stated model purpose is to shed light on dynamic interactions between behavioral factors and
contextual factors, and thus explaining the diffusion of energy efficient innovations in the
residential building environment in a community. The model boundary of the simulation
model is considered to be too narrow in the current state. As we have pointed out in Chapter
4.2, several other sectors have relevant influences on the diffusion of energy-efficient
buildings and need to be included in an elaborated version of the model.
4.4 Model Validation
The validation of System Dynamics models has two important aspects: structure and behavior
validation (Barlas 1996). Structure validation is about warranting that the model’s internal
structure is a sufficiently accurate description of the real system, with respect to the model
purpose and issue of interest. Structure validation can be differentiated in several direct and
indirect structure tests respectively (Barlas 1996; Sterman 2000). Behavior validation means
that the output behavior of the model reproduces closely enough the dynamic behavior of the
real system under study.
In the validation process, different structure validity tests were carried out (e.g., dimensional
consistency tests, direct extreme-condition tests, stress tests, behavior sensitivity tests) (Coyle
1996; Roberts et al. 1981; Sterman 2000). The model was improved continuously in the
course of validation. The validation, however, can be improved significantly once more
specific data about the reference mode is available. As pointed out previously, it is possible
that the data selection and data management could be prohibitively high. Further research
about available data and its selection and gathering is required.
4.5 Standard Runs, Policy Formulation and Evaluations
Standard Simulation Run
Figure 9 shows the development of some key variables of the system. Traditional houses
(blue line, #1) show an increasing tendency that peaks at the year 2008 followed by a
decreasing trend. The market powers, especially the word of mouth effect (L1, L3) and
supply of ee-architects (L2), foster energy-efficient instead of traditional building designs, but
these effects are only active when a large enough pioneer group exists. Time is required for
the group to evolve, thus, explaining the time delay before the traditional houses number
decreases. The amount of energy-efficient houses (red graph, #2) increases around the year
1997 due to the formal establishment of an energy-efficiency association. The growth rate
peaks at the inflection point around 2045 and thereafter reveals a diminishing growth because
of the market saturation process. L4 is active and weakens the L1 to L3. The growth rate of
the energy-efficient houses will be equal to the demolition rate of tradition houses, after the
house building population has been satisfied. This leads to a slow transition from the
traditional building stock to energy-efficient houses.
Page 17 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
A differing development can be seen for the number of traditional architects (grey line, #4)
and energy-efficient architects (black line, #5). The changes in the traditional architect
quantity lags behind the changes in the house quantities because the traditional architects are
only willing to change or graduate with an ee-education when the demand for traditional
houses shrinks. In addition, the retirement of traditional architects reduces the number of
traditional architects. In case of the ee-architects, the variable concurs with the number of ee-
houses because ee-houses can only be built when ee-architects are available. The higher
market demand for ee-houses, driven by the WOM-effects (L1 and L3), pulls the graduating
architects towards an ee-education (L2). Hence, the number of ee-architects grows steeply. A
saturation effect occurs around the 2090 when the number of retiring architects equals the
graduation architects resulting in a dynamic equilibrium.
The green graph (#3) shows the development of the total enrolled and current home owners
indicating a goal seeking behavior towards the goal ‘total building population’. This graph
relativises the considerably fast spread of energy-efficient houses because the demand for
houses is nearly satisfied at the end of the simulation. A further spread of energy-efficient
houses can only occur when traditional houses are demolished which have a life time of 50 to
100 years. However, at the year 2150 only a few traditional houses still exist indicating that
the residential building system changed from a traditional- oriented building stock towards an
energy-efficient one. However, the assumption that technology has not advanced during the
time duration of 180 years is unrealistic and has to be relaxed in a further version of the
model.
Simulation Run "Standard"
6.000 House
20.000 Person
200 Person
3.000 House
10.000 Person
100 Person
0 House
0 Person
0 Person
1970 1997 2024 2051 2078 2105, 2132
Time (Year)
4 4 i
‘Traditional Houses House
Energy-Efficient Houses House
Total Enrolled and Current Home Owners ———_— Person
Architects (only traditional) Person
Architects (energy-efficient) ly si Person
Figure 9: Development of #of Houses, and # of Architects, and Total Enrolled and Current
Home Owners. The system behavior reveals a long-term diffusion process due to long cycle
time of the building stock.
Figure 10 shows the development of the decision for ee-fraction (blue line, #1). To remember,
this variable represents the main driver for the decision to build an energy-efficient house or
not. The more the value tends toward the value 1.0 [dmnl], the greater is the amount of
potential building owners who will decide to build an ee-house. In the current model, the
combined word of mouth effects of ee-convinced persons and current ee-home owners is the
only factor that influences the decision process. Thus, the growth of the sum of the convinced
persons directly affects the decision about energy-efficiency and subsequently leads to a
stronger word of mouth effect. The goal seeking growth behavior of the decision for ee-
fraction is solely caused by the increasing ee-population (red line, #2).
Page 18 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
Simulation Run "Standard"
1 dm
20.000 Person
0,5 dmnl
10.000 Person
0 dm
0 Person
1970 1997 2024 2051 2078 = 21052132
Time (Year)
Decision foree-Fraction <= = = = a: dmnl
Sum of ee- Convinced Persons — es Person
Figure 10: Development of Decision for ee-Fraction. The variable represents the decision of
potential home owners for energy-efficient buildings. The sum of ee-convinced persons is the
antecedents variable for the development of the ee-decision variable.
The variable ‘decision for ee-fraction’ seems to be a highly important leverage point of the
system. Thus, additional factors that influence potential home owners’ decision about energy-
efficiency have to be included, because they do not only decide due to the word of mouth
effects, but also due to financial restrictions, objective feasibility and so forth (Groesser
2006). Figure 11 shows the mathematical formulation of the decision for ee-function.
f, = f((EF(x,,X,),X5...,X,) = EF(pot_ee+cur_ee)* N_Dec_ee;
with: f, = decision foree- fraction
EF(x,,X») = effect of convinced persons on decision for ee fraction
pot_ee = potential ee home owners
cur_ee= current ee home owners
N_Dec_ee=normal decision fraction for ee
Figure 11: Formulation of the decision function of potential home owners for an energy-
efficient home. Currently, the sum of ee-convinced persons is the argument for the decision
function.
Sensitivity Runs as Parameter Policy Runs
The objective of policy experiments is to improve or optimize the system behavior according
to some set of objective variables or objective functions. Coyle distinguishes policy into its
structure and its parameters within the structure (Coyle 1996). Not changing the
aforementioned, we argue that in an early stage of the modeling the objective of a policy
experiment is not to optimize the system behavior, it is rather to understand the system better
and detect structural inaccuracies as well as get a more intuitive grasp about the system
behavior under different structural parameter sets. Table 2 provides an overview about the
parameter policy runs.
For all three policy runs, sensitivity analyses have been used to explore the effects of
parameter changes on the system behavior. Monte Carlo simulations have been employed to
conduct the sensitivity analyses.’ Monte Carlo simulations are stochastic techniques, i.e., they
® Basic parameters: n=500 runs; noise seed 1234, intervention parameter is distributed according to a random
uniform distribution with values within the interval indicated in table 2.
Page 19 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
are based on the use of random numbers and probability statistics to investigate the behavior
of the modeled system with different parameters values.
No. Policy Name Randomized Variable Lower Bound | Upper Bound
pi] Arhiects with ee Knowledge Flood the formal graduation ratio 0 [dmnl] 0.9 [dmnl]
P2] Supply of Further Education for Architects architect switching ratio 0 [dmnl] 0.9 [dmnl]}
Influence the Decision of Potential Home ic
P3 Owners towards Energy-Efficient Homes normal decision for ee 0 [dmnl] 0.25 [dmnl]
Table 2: Parameter Policy Runs. The three different policy runs give an indication about the model behavior.
The randomized uniformly distributed variables take values in reasonably bounds.
The interpretation of the sensitivity graphs will be provided upfront. The simulation results
are displayed in the graphs as confidence bounds. These are computed at each point in time
by ordering and sampling all the simulation runs. Thus, for example, for a confidence bound
at 50, one-quarter of the runs will have a value larger than the top of the confidence bound
and one-quarter will have a value lower than the bottom. The color code (yellow, green, blue,
and grey) corresponds to the confidence different bounds (50%, 75%, 95%, and 100%). In the
following, three policy runs are analyzed and their results will be compared with each other.
P1: Architects with EE-Knowledge Flood the Market (Variable: normal graduation ratio)
The first policy run analyzes the effects changes in the education support system, e.g.,
financial support for universities for the supply of education courses about energy-efficiency
in the building environment, has on the graduation rate of architects. The first variable
‘Architects with traditional knowledge’ (Figure 12, left side) shows relatively sensitive
behavior to changes of the parameter values from 0.0 [dmnl] to 0.9 [dmnl]. After the
initiation of the energy-efficiency topic, the normal graduation ratio has significant influence
on the system. The lower the normal graduation rate, indicating that people do not graduate
with knowledge about energy-efficiency, the more architects are available to construct houses
according to the traditional standard. Because the word of mouth effect is rather low and
people are not convinced to build ee-houses a demand for traditional houses exist. Thus, the
feedback loops L1 and L2 are reduced in their strength resulting in fewer ee-houses.
When the parameter has values p > 0.1 [dmnl], the number of architects with traditional
knowledge decreases rapidly due to a large decision for ee-fraction that is driven by L1 and
L3. In addition, the decision for ee-fraction leads to a faster transfer of the traditional
architects to the ee-cohort. The result on the ee-houses is a relatively strong inclination
between 2015 <t < 2050. The behavior of the ee-house development is considered robust
because three-quarters of the random draws create an envelope with a maximal width of 900
entities.
icy_flooding the market
policy_flooding the market. Lace =y foods z
50% | 75% 95% (100% 50% 75% (N95 % J 100% Ja
“Architects (with trad knowledge)" ee houses
200 6.000
150 | 4.500
100 3.000
Soro 2015 2060 2105 2150 Yon 2015 2060 72105 72150
‘Time (¥ ear) Time (Y ear)
50 1.500
Figure 12: Analysis of the Policy Run ‘Architects with ee-Knowledge Flood the Market’. For p < 0.1 [dmnl], the
system shows highly sensitive behavior. The behavior is for p > 0.1 [dmnl] much more robust.
Page 20 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
P2: Supply of Further Education for Architects (Variable: architect switching ratio)
The second policy run analyzes, again, the effect of changes in the education support system.
This time, however, not the first education, in other words, the education in the university, but
the further education of already graduated architects is financially supported and enabled by
the federal government. The effects are shown in Figure 13 and indicate almost no changes
when varying the randomized parameter inputs. Where does this robustness of the system
come from?
Figure 13 (left side) shows that the variable ‘architects with traditional knowledge’ is non-
sensitive to variations in the normal switching ratio until t > 2070. This is because the ratio of
potential traditional home owners to energy-efficient home owners is quite small leading to
transition effect of nearly 0. In other words, the architects stick to their business and try to
satisfy the demand for traditional houses. Then, L7 becomes active and dampens graduation
of traditional architects because the market demand is perceived to saturate. Hence,
graduating architects decide for an ee-education. The remaining traditional architects serve
the still existing demand for traditional houses. At the same time, the demand for ee-houses
increases because of the word of mouth effect. Hence, the ratio between potential energy-
efficient and traditional home owners increases. Consequently, the first architects leave the
traditional architecture market and move towards energy-efficient building designs by further
education. This development starts at t = 2070. Depending on the value of the normal
switching ratio, the architects switch more rapidly (for p > 0.1 [dmnl]) or more slowly (for p
< 0.1 [dmnl]). In either case, the effect of switching architects does not influence the
development of the ee-building stock because the limiting variable in the development of the
ee-building stock is not the number of ee-architects; it is the number of potential ee-home
owners.
arcitects switching axcilets switching
50% 75% 05% 100% 50% 75% 95% 100%
"Architects (with trad knowledge)" ec houses
200 | 6.000
150 | 4.500
100 | 3.000
1.500
Bo) wis 2060 2105 2150 Som 2015 7060 2105 2150
Tine (¥ eat) Tine (¥ ea)
Figure 13: Analysis of the Policy Run ‘Supply Further Education of Architects’. The
switching of architects occurs when the future market is considered to be the energy-efficient
building market. This happens at t > 2070. However, the switching of architects does not have
an influence on the ee-building stock because the bottleneck variable is the potential ee-home
owners.
P3: Influence the Decision of Potential Home Owners towards Energy-Efficient Homes
(Variable: normal decision fraction for ee)
The last policy run analyzes the effect of changes in the building owner decision system. By
these changes, the effects of further decision factors on the system behavior are explored.
Figure 14 (left side) shows that the number of traditional architects is highly sensitive to the
normal decision fraction because it drives the decision for ee-fraction which leads to a higher
value number of potential ee-home owners. The more home owners are interested in ee, the
stronger is the motivation for traditional architects to become ee-architects. If the normal
decision for ee-fraction is smaller than the standard value, the architects will switch more
slowly due to the inverse mechanism just described. Consequently, the number of traditional
architects indicates a strong dependency on the normal decision for ee-fraction.
Page 21 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
The variable ‘ee-houses’ is also influenced by randomized changes in the variable ‘normal
decision for ee-fraction’, because the bottleneck variable ‘potential ee-home owners’ is
indirectly influenced by it. Hence, higher values than the standard value lead to a steeper
growth of the potential ee-home owner stock that, in turn, will result in more ee-houses. For
parameters smaller than the standard value, the opposite is true. Both cases are depicted by
Figure 14 (right side).
nomtal decision fraction forge. noma decision faction foree
50% 75% NNI95% 100% 50% 75% 95% 100% I
"architects (with trad knowledge)" ee houses
200 6.000
150 4.500
100 3.000
50 1.500
9 0 —
1970 2015 2060 2105 72150 1970 2015 2060 2105 2150
‘Tame (¥ ear) Time (Year)
Figure 14: Analysis of the Policy Run ‘Influence of the Decision of Potential Home Owners’.
Traditional architects and ee-houses exhibit sensitive behavior to changes in the normal
decision fraction for ee because both variables depend on the variable ‘potential ee-home
owners’ which is steered by decision for ee-fraction.
Comparison of the Policy Runs and First Insights Derived from the Model
A comparison of the simulation runs indicates beneficial and less beneficial intervention
points for the federal government with the objective to increase the diffusion and adoption of
the innovation. In order to evaluate the different policy runs, e.g., policy A is more effective
than policy B, an objective has do be defined. In our case, the model purpose is to explain
innovation diffusion and to derive a system design that fosters this diffusion process. Hence,
the objective is that the ee-building stock reaches a value as high as possible and in a time
period as short as possible. The variable ‘ee-houses’ is a highly and reasonable representative
of the system development.
Given this objective of the system, the important policy graphs are the graphs on the right
side in the Figures 12, 13, and 14 (‘ee-houses’). The graphs on the left represent an
anteceding variable for the ee-houses and are not considered for the evaluation of the policies.
The first policy P1 (‘Architects with ee-knowledge flood the market’) influences the
parameter ‘normal graduation ratio’ as control level. The system behavior is improved when
value p > 0.1 [dmnl] because this leads to a higher graduation rate of ee-architects by which
the market demand can be more satisfied. On the other hand, values of p < 0.1 [dmnl] reduce
the number of ee-houses substantially. The maximal width of the envelop Ep; of policy P1 is
about 2800 [houses]. P1 manipulates the control variable ‘normal graduation ratio’ to which
the system is highly sensitive. The results of the policy exhibit the highest degree of
uncertainty if compared with the other policy runs.
The second policy ‘Supply of Further Education for Architects’ (P2) influences the variable
‘architect switching ratio’ within reasonable bounds. As Figure 13 (right graph) shows, the
behavior is not influenced by the uniformly randomized changes in the control lever. The
development of ee-houses is highly robust. Policy P2 is therefore dispensable and should not
be considered as improvement strategy.
Policy P3 (‘Influence the Decision of Potential Home Owners towards Energy-Efficient
Homes’) is the most promising improvement strategy because it increases the number ee-
houses even higher than P1 and avoids strong downside movements, simultaneously. The
maximal width of the envelope Ep3 of policy P3 is about 2000 [houses] which is about 30%
less than P1. However, it is not clear which measure will have guaranteed influences on the
Page 22 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
variable ‘normal decision for ee-fraction’. This indicates a requirement to elaborate the
decision function (Figure 11) and include additional mechanisms of decision making.
4.6 Feedback Loop Management
Positive feedback loops are motors for growth or collapse of a system. They push towards
exponential growth or decay and thrive towards disequilibrium. Feedback loop management
intends to regulate them by actively managing different loops in favor of the system specified
by a system objective. In the existing model, three positive feedback loops are controlled by
five balancing loops. Hence, the controlling potential rather outweighs the reinforcing forces
in number. However, not all feedback loops have the same influence on the system’s behavior
in strength and time. Table 3 shows a first feedback loop strength analysis. L3 is the strongest
reinforcing feedback loop of the system. Moreover, loops L1, L2, and L3 are active when
people have ee-houses or are convinced to construct ee-houses, and the population without
own houses is sufficiently large. Later in the simulation, the population without own houses
decreases (cf. Figure 9, consider that the population without own houses is the total
population less the total enrolled and current home owners [line #3]). Thereby, the strongest
balancing loop L4 is activated and weakens the three mentioned positive loops. From a
management perspective, the diffusion process should occur as fast as possible. The limit to
growth is determined by natural resources, i.e., by humans without own houses, and cannot be
changed. If all humans in the considered region have fulfilled their demand for an own house,
the diffusion process will stop. To foster the diffusion process, measures have to be
introduced that enable the loops L1 to L3 to work, i.e., people have to be convinced to build
ee-houses. One possibility is to introduce a timely determined advertisement program and
subsidies to enable the reinforcement of the system.
No. Feedback Loop Polarity Strength
Word of Mouth Effect by Potential ee Home
Li Owners Reinforcing 2
L2 Market Pull for ee Architects Reinforcing 2
Word of Mouth Effect by Current ee Home
L3 Owners Reinforcing 3
L4 Everybody Has a House Balancing 3
L5 | EE House Building Slows Down the Diffusion] Balancing 1
L6] EE Architect Capacity Clear the EE Demand Balancing 1
Traditional Architect Capacity Clear the Trad
L7 Demand Balancing 1
Traditional Architect Capacity Hinders the
L8 Diffusion of Trad Houses Balancing 1
Table 3: Feedback Loop Management. L1, L2, and L3 are active leading to a diffusion of the
innovation. L4 successfully controls them in the last quarter of the simulation time frame. Loops
L5 to L8 are rather weak. Management activities should be concentrated on the first three positive
loops.
5. Conclusions and Further Research
The overall target of the Swiss Energy Policy is to reduce CO -emissions and thereby to
achieve the vision of an energy-efficient ‘2000 Watt per Capita Society’. Major instruments
to contribute to the vision in the residential building sector are the standards about energy-
efficient building designs. However, new residential houses according to energy-efficiency
standards are not as often constructed as they ought to be. In order to investigate the problem,
a preliminary System Dynamics model was created. With Coyle, we share the view that the
main aim of System Dynamics is to develop system policies which improve the dynamic
Page 23 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
behavior of a system (Coyle 1996). However, the stage of the model is considered to be
preliminary and therefore this goal is not achievable with the existing model. Our intention
was rather to gain first insights about the diffusion of innovations in the residential building
environment. Key insights drawn from the model formulation and analysis are: (1) The most
important variable to enable a fast transition to an energy-efficient building stock is the
potential ee-home owners. The demand side is the bottleneck for the diffusion process. The
supply side anticipates the future market volume and demand. Throughout all simulation runs
the demand side was the limiting factor for a successful diffusion. (2) The decision for ee-
fraction is a high leverage point in the system as shown by policy P3. Changes in this variable
can lead to significant improvements of the system behavior. Up to now, the function
constituting the decision for ee-fraction is rather a black-box. This indicates a further research
demand to comprehend the diffusion mechanisms better. (3) Most important mechanisms
active to foster the diffusion of innovation process are the word of mouth effect, market pull
effect and market saturation effect.
Further Research
As mentioned previously, the model is a preliminary version and needs more elaboration.
Important questions that have to be addressed in the further research process are:
e Transition process of the legal requirements about energy-efficient building standards,
e.g., SIA-Standard to Minergie- Standard,
e More precise definition of the limit between energy-efficient and non-energy-efficient
buildings,
e Disaggregating the variable ‘decision for ee-fraction’, e.g., by utilization of additional
scientific theories (e.g., theory of planned behavior (Ajzen 1991), or diffusion theory
(Rogers 1995)),
e Inclusion of additional variables to improve model accuracy guided by psychological,
managerial, economical theories as well as current legislation (SFOE 2002), empirical
investigation about antecedents variables (Kaufmann-Hayoz, Bruppacher, and Ulli-
Beer 2005),
e Further elaboration of the physical building structure, e.g., single-, double-, and
multiple-family buildings, connection between buildings and owners, and
consideration of the lessee-owners relation, vacant and occupied buildings or
apartments,
e Additional parameter estimation and data analysis are required to improve the fit
between model behavior and actual data,
e Validation and elaboration of the model with system experts and by means of real
data,
e Investigation of the population dynamics, especially the increasing trend of foreign
inhabitants and the effects of the trend on other trends and variables such as the
inhabitants per house development or the inhabited residential area. Also the social
trend to smaller families and the higher motivation to home ownership instead of
home leasing has to be researched about in more details,
e Answer to the questions under which conditions a market pull effect occurs, e.g., if
the market pull is determined by the system structure, and if the market pull effect can
be intentionally altered in a market push effect,
Improvement of the objective function for a ‘good’ or intended system behavior,
e Elaboration and inclusion of the house building support system, e.g., supplier and
producer of building material.
The inclusion of further theory is an important step to (1) increase model accuracy and
validity and (2) to foster the success of policy initiatives, because initiatives, such as the
Page 24 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
Swiss energy program by the Swiss Federal Office of Energy, aiming at the diffusion of
energy efficient buildings depend not only on technological innovations and fair market
conditions (external opportunities) but also on adequate knowledge, decision rules,
management, and networking (behavioral variables) of the various actors involved in the
value creation chain of energy-efficient buildings (e.g., investors, architects, building crafts,
producers, political authorities). The completed model will allow analyzing and
demonstrating effects of policies on the variables of interest, such as energy consumption of
the building stock, or decisions of future home owners about energy-efficient building
designs depending on external legal requirements.
II. Bibliography
Ajzen, I. 1991. The Theory of Planned Behavior. Organizational Behavior and Human
Decision Processes 50 (2):179-211.
Barlas, Y. 1996. Formal Aspects of Model Validity and Validation. System Dynamics Review
12 (3):183-210.
Bass, Frank. 1969. A New Product Growth Model for Model Consumer Durables.
Management Science 15 (5):215-227.
BFE. 2001. EnergieSchweiz: Das Nachfolgeprogramm von Energie2000. Beme: Swiss
Federal Office of Energy.
—-——. 2005. Nachbefragung und Erfolgskontrolle Investitionsprogramm Energie 2000:
Schlussbericht. Berne: Swiss Federal Office of Energy.
Bond, R., and Peter B. Smith. 1996. Culture and Conformity: A Meta-Analysis of the Asch
Line Judgment task. Psychological Bulletin 119 (2):111-137.
BUWAL. 1999. Schweiz beschliesst Gesetz zur Umsetzung des Kyoto-Protokolls. In
Medienmitteilung des Informationsdiensts des Bundesamts fir Umwelt, Wald und
Landwirtschaft. Bern: Bundesamt fiir Umwelt Wald und Landwirtschaft.
———. 2005. Klimarappen: Zielvereinbarung unterzeichnet. In Medienmitteilung des
Informationsdiensts des Bundesamts fir Umwelt, Wald und Landwirtschaft. Bern:
Bundesamt fiir Umwelt Wald und Landwirtschaft.
Coyle, R. Geoffrey. 1996. System Dynamics Modelling: A Practical Approach. London:
Chapman and Hall.
De Toni, A., M. Caputo, and A. Vinelli. 1988. Production Management Techniques: Push-
Pull Classification and A pplication Conditions. International J ournal of Operations
and Production Management 25 (3):457-472.
Duggan, Jim. 2004. Policy Diffusion in the Beer Game. Paper read at 22nd Intemational
Conference of the System Dynamics Society, July 25-29, at Oxford, England.
Eskinasi, Martijn, and E.A.J.A. Rouwette. 2004. Simulating the Urban Transformation
Process in the Haaglanden Region, the Netherlands. Paper read at International
System Dynamics Conference, at Oxford.
ETH Council. Strategie Nachhaltigkeit im ETH-Bereich: 2000 Watt-Gesellschaft - Modell
Schweiz. Wirtschaftsplattform, ETH Zurich 1998 [cited 01.10.2005. Available from
http://www.novatlantis.ch/projects/2000W /brochure/resources/pdf/ge_brochure.pdf.
Feng, XiWen, Fujun Lu, and Xinhua Wang. 2001. Strategic Study and Dynamics Decision
for Development Program of City Housing. Paper read at The 19th International
Conference of the System Dynamics Society, July 23-27, at Atlanta, Georgia.
Fiddaman, Thomas S., and David W. Peterson. 2000. Beating The Production Distribution
Game. Paper read at 18th Intemational Conference of the System Dynamics Society,
August 6-10, at Bergen, Norway.
Forrester, Jay Wright. 1969. Urban Dynamics. Cambridge MA: Productivity Press. Original
edition, MIT Press.
Page 25 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
Groesser, Stefan N. 2005. Interview with an Expert from the Minergie Association, Beme,
Switzerland. Beme: Exhibition 'Hausbau 2005'.
———. 2006. Environmental Politics and the Minergie®-Concept: Position and Potential of
the Standards, and Possible Explanations for the Low Acceptance Rate. St. Gallen and
Beme: University of St. Gallen, Institute of Management & University of Beme,
Interdisciplinary Center for General Ecology, Unpublished Manuscript.
Hennekam, Hans, and Frank M. Sanders. 2002. Making Complex Network Analysis in
System Dynamics. Paper read at Proceedings of the 20th International Conference of
the System Dynamics Society, July 28 - August 1, at Palermo, Italy.
Hong-Minh, Séverine, Paul Childerhouse, and Mohamed Naim. 2000. Developing A System
Dynamics Model Of The UK Private House Building. Paper read at 18th International
Conference of the System Dynamics Society, August 6-10, at Bergen, Norway.
Jochem, E., D. Favrat, K. Hungerbihler, Rudolf von Rohr, D. Spreng, A. Wokaun, and M.
Zimmermann. 2004. Steps towards a Sustainable Development - A White paper on
R&D of Energy-Efficient Technologies. Zurich: CEPE/ETH and Novalantis.
Kaufmann-Hayoz, Ruth, Susanne Bruppacher, and Silvia Ulli-Beer. 2005. Project Proposal:
Diffusion Dynamics of Energy Efficient Buildings (DeeB). Simulation of the
Dynamic Interactions between Relevant A ctors' Managerial Leaming, Technological
Innovations, and Public Policy. Beme: University of Beme, Interdisciplinary Center
of General Ecology.
Kimura, O., and H. Terada. 1981. Design and Analysis of Pull System: A Method of Mulit-
Stage Production Control. J ournal of International Production Research 19 (3):241-
253.
Koschenz, Markus, and Andreas Pfeiffer. 2005. Potenzial Wohngebaude: Energie- und
Gebaudetechnik fiir die 2000-Watt-G esellschaft, Schriftenreihe Nachhaltigkeit.
Ziich: Faktor-V erlag.
Lee, Man-Hyung, and Nam-Hee Choi. 2004. Green Belt Policy Change and Uninvited
Aftereffect in Seoul. Paper read at 22nd Intemational Conference of the System
Dynamics Society, July 25-29, at Oxford, England.
Milling, Peter M. 1999. Simulating Different Decision Rules and Market Demands: New
Insights into the Beer Game. Paper read at 17th International Conference of the
System Dynamics Society and 5th Australian & New Zealand Systems Conference, at
Wellington, New Zealand.
Moon, Tae Hoon. 2002. Environmental Carrying Capacity of the Seoul Metropolitan A rea:
Estimation, Implication and Limitations. Paper read at Proceedings of the 20th
International Conference of the System Dynamics Society, July 28 - August 1, at
Palermo, Italy.
Nauser, Markus. 2005. Vierter Bericht der Schweiz zuhanden der UNO-Klimakonvention.
Beme: BUWAL, Sektion Okonomie und Klima.
Novatlantis. 2005. Leichter leben: Ein neues V erstaéndnis fir unsere Ressourcen als Schliissel
zu einer nachhaltigen Entwicklung - die 2000-Watt-Gesellschaft. Berne: Swiss
Federal Office of Energy.
Roberts, Nancy H., David F. Andersen, Ralph M. Deal, Michael S. Grant, and William A.
Shaffer. 1981. Introduction to Computer Simulation. Waltham, MA: Pegasus
Communications.
Rogers, Everette M. 1995. The Diffusion of Innovations. Vol. 4th Edition. New Y ork: Free
Press.
SFOE. 2002. Wirkungen der kantonalen Energievorschriften im G ebaudebereich im Jahr
2002. Berne: Federal Office for Energy, EnergieSchweiz.
Spreng, D., and M. Semadeni. 2001. Energie, Umwelt und die 2000 Watt
Gesellschaft.Grundlage zu einem Beitrag an den Schlussbericht
Page 26 of 27
Stefan N. Groesser et al. 24th Intemational System Dynamics Conference, Nijmegen
Schwerpunktsprogramm Umwelt (SPPU) des Schweizerischen National Fonds (SNF).
Working Paper No. 11, CEPE, ETH-Zurich, Zurich.
Sterman, John D. 1989. Misperceptions of Feedback in Dynamic Decision Making.
Organizational Behavior and Human Decision Processes 43 (3):301-335.
———.1992. The Beer Distribution Game: An Annotated Bibliography Covering its History
and Use in Education and Research: MIT System Dynamics Group.
———. 2000. Business Dynamics: Systems Thinking and Modeling for a Complex World.
Boston: McGraw-Hill.
Sterman, John D., and Thomas S. Fiddaman. 1993. The Beer Distribution Game Flight
Simulator. Cambridge, MA: MIT Sloan School of Management.
Stroebe, Wolfgang, Klaus Jonas, and Miles Hewstone. 2001. Introduction to Social
Psychology. Berlin, Heidelberg, New Y ork: Blackwell Publishers Ltd.
Swiss Federal Assembly. 1999. Bundesgesetz ttber die Reduktion der CO2-Emissionen
(CO2-Gesetz). Berne: Bundesversammlung der Schweizerischen Eidgenossenschaft.
Ulli-Beer, Silvia, Susanne Bruppacher, Stefan N. Groesser, Stefanie Geisshiisler, Matthias
Miller, Mohammad Mojtahedzadeh, Markus Schwaninger, Fran Ackermann, David
Andersen, George Richardson, Roland Stulz, and Ruth Kaufmann-Hayoz. 2006. An
Action Science A pproach for Understanding and Accelerating the Diffusion Process
of Energy-Efficient Buildings: Challenges and Preliminary Guidelines of the Research
Task. Paper read at 24th International Conference of the System Dynamics Society, at
Nijmegen, The Netherlands.
Vennix, Jac A. M. 1996. Group Model Building: Facilitating Team Learning Using System
Dynamics. Chichester: Wiley.
Voyer, John J. 2004. Affordable Housing and Urban Sprawl Policy Choices in Y ork County,
Maine: A System Dynamics Approach. Paper read at 22nd International Conference
of the System Dynamics Society, July 25-29, at Oxford, England.
Page 27 of 27
