32nd International Conference of the System Dynamics Society
Delft, Netherlands July 20 - 24, 2014
Market Penetration of Alternative Powertrain C oncepts in
Heavy Commercial Vehicles: A System Dynamics A pproach
Claudio Seitz*”*, Orestis T erzidis*
* Karlsruhe Institute of Technology
Institute for Entrepreneurship, Technology Management and Innovation (EnTechnon)
Fritz-Enler-Str. 1-3, 76133 Karlsruhe, Germany
» Robert Bosch GmbH, Stuttgart, Germany
i Corresponding author: claudio.seitz@ partner.kit.edu, +49 (0)711 811 41743
Diffusion of alternative powertrain concepts in heavy commercial vehicles will start in the upcoming
years after electrification and natural gas engines have already been introduced for passenger cars.
Numerous quantitative forecasting and technology diffusion models exist for passenger cars but cannot
be transferred unchanged to heavy commercial vehicles. A system dynamics model for the diffusion of
alternative powertrain concepts in heavy commercial vehicles is developed by adaptation of existing
simulation models for passenger cars. The structural validity is assured by changing the structure and
parameterization based on stakeholder interviews, secondary studies, and theoretical foundations. The
results reveal the significance of a satisfying refueling infrastructure for alternative fuel trucks and the
transitional market potential of hybrid electric trucks. The discussion of the system dynamics model
emphasizes the analysis of customer demand as an essential field for future research of alternative
powertrain diffusion in heavy commercial vehicles.
Keywords: technology diffusion, system dynamics, market penetration, altemative powertrain concepts,
heavy commercial vehicles, organizational adoption of innovation
1 Introduction
The diffusion processes of alternative powertrain technologies are widely studied based on the example of
passenger cars (PC) and, partly, light-duty-vehicles. The d market ion of low carbon
technologies within automotive vehicles provides a solid basis for forecasting future greenhouse gas
(GHG) emission levels of individual passenger transportation. In contrast, the road freight and public
passenger transportation is not widely understood, despite it is a main originator of GHG emissions.
Heavy commercial vehicles (HCV) are not part of any — to the best of our knowledge — quantitative
forecasting method or technology diffusion model. This might be caused by the dominating Diesel
powertrain technology or the different industry structure. Nevertheless, a transition towards alternative
powertrain concepts can be expected in the upcoming years due to political will and customer demand: an
increasing cost pressure forces transport ies to minimize their fuel 1 and the European
Commission will propose a strategy targeting fuel consumption and CO» emissions from heavy and
medium duty commercial vehicles to push low carbon transportation (European Commission, 2011,
2012).
Quantitative forecasting methods for the transportation system are mainly based on system dynamics
(SD), agent-based modeling or diffusion and times series models (Shafiei et al., 2013; Al-Alawi &
Bradley, 2013). System dynamics is particularly suitable to study the fundamental market dynamics and
understand the interdependencies of influencing factors. Thus, this method is assumed to be appropriate
for levering the basic und ding of i diffusion processes on the HCV market.
Since empirical data for the heavy commercial vehicle market are equally rare, a transferred model from
the PC market could provide a basis for future research to predict market penetration of low carbon
technology on this market as well. Nonetheless, this transferred model should be discussed considering
some fundamental differences caused by the business-to-business (B2B) HCV market structure and
specific customer requirements.
From a mathematical perspective, such system dynamics’ models are simulation models of nonlinear,
coupled differential equations (Sterman, 2000). Therefore, one major challenge in implementing SD
models lies in the formulation of the matt ical equation ing the empirically observable
reality: this represents the heuristics of the problem at hand. Although Sterman, for instance, claims there
are no valid models, verification tests help developing confident and reliable SD models (Sterman, 2000).
These tests can be clustered into direct structure tests, structure-oriented behavior tests, and behavior
pattern tests. Direct structure tests draw comparisons between the market model, represented by the
system of differential equations, and the reality. Structure-oriented behavior tests involve simulation of
the whole model as well as decoupled sub-models of it and yield to evaluate the model generated
dynamics. Behavior pattern tests use graphical and visual measures to compare typical behavior features
(Barlas, 1996).
Against this background, the objective of this article is to discuss the application of existing PC system
dynamics models aiming to forecast the market penetration of powertrain technologies to the HCV
market. This is primarily under consideration of the structural model validity to provide a basis for future
research. The SD model is exemplarily developed for the German HCV market. A thorough discussion of
the HCV market itself is not intended. It’s rather the question whether existing simulation models of the
PC market are transferrable to the HCV market, providing an appropriate approach for future research.
Thus, the remainder of this article is organized as follows. The second section provides the theoretical
foundation of the HCV market simulation model. In section 3 the market model framework is developed
by the adaptation of existing SD models for the PC market. Section 4 discusses the simulation model in
terms of validity and the implication of the results. A conclusion for stakeholders and future fields of
research for HCV market technology diffusion models is drawn finally in section 5.
2 Theoretical foundations for the simulation of powertrain technology
diffusion in HCV markets
2.1 Simulation models for powertrain concepts on the passenger car market
There are multiple research streams about automotive market modeling by means of system dynamics.
These cover different aspects regarding the technologies to be analyzed (fuel cell vehicles, electric and
hybrid electric vehicles, natural gas vehicles), the stakeholders considered (customer, filling stations,
energy supply system, manufacturer, government, dealer), the country specific market and the aimed
scope (market penetration of technologies, total market sales, overall fuel demand or GHG emissions,
policy deployment, manufacturer’s actions) (Shafiei et al., 2013). Targeting this paper’s aim, we
primarily focus in the following on SD models dealing with diffusion processes of alternative powertrain
concepts.
Besides some strongly explorative SD studies by two master theses from MIT, Struben was among the
first developing a SD model for examining the market diffusion of alternative fuel vehicles. The original
work about the diffusion of fuel cell vehicles in California was gradually expanded and applied to other
2
powertrain concepts. These models put special hasis on the spatial di: ion of the fuel demand
and the corresponding refueling infrastructure. Additionally, the buying process is discussed in detail,
using a multinomial logit choice model (MNL) and the construct of “familiarity” by social factors as
media attention, marketing effectiveness, and word-of-mouth (Struben, 2004, 2006; Struben & Sterman,
2008).
For the European PC market Janssen identified the most influencing stakeholders on the diffusion of
natural gas vehicles in the Swiss car fleet. “Customers sector”, “filling station sector”, “car import, retail,
and service sector” are modeled as endogenous stakeholders, whereas the government, natural gas
industry, and “non-Swiss car industry sector” are as keholders. In his analysis he
described the effectiveness of policy actions on the diffusion of natural gas and fuel cell vehicles. The
system dynamics model implementation, calibration, and validation is predominately achieved by
dynamic test runs of model modules, which have been primarily derived by empirical observations
(Janssen et al., 2006; Janssen, 2005).
Bosshardt expanded the work of Janssen by integrating competition between altemative powertrain
concepts and analyzing multiple European markets. Thereby he regarded aspects of cost, availability of
car models, refueling infrastructure coverage, powertrain attractiveness, and social norm. The model is
used for analyzing the influence of different strategies enhancing the diffusion of alternative powertrain
concepts. Validation of the SD model is achieved by graphic representation and subsystem tests, extreme
condition tests, sensitivity analysis, and a mathematical analysis of the “social norm loop” (Bosshardt,
2009; Bosshardt et al., 2007).
Keles et al. analyzed the market penetration of fuel cell vehicles in Germany based on the action of
different stakeholders, they | focused on the interactions of CORSUMES, automotive manufacturers, filling
station owners and p r y, they impl a “fuel cell vehicle demand and supply
module”, a “filling station module”, an “attractiveness module” and the resulting governmental “balance
of payment”. Thereby, they distinguished between the available urban and highway refueling
infrastructure, latter being dependent of the urban ones. A profound discussion of the model validity is
not provided (Keles et al., 2008).
Mainly based on the work of Struben and Janssen, Weikl simulated the future market shares of alternative
powertrain concepts for the German market. He put special emphasis on the emotional powertrain related
buying criteria in the decision process of new car buyers. Additionally, he aimed to model the
manufacturers’ actions to improve the characteristics and availability of alternative powered PC models in
detail. The validity of the model is discussed by structure and behavior pattem tests (Weikl, 2010).
Recently, Keith extended the work of Struben by incorporating effects of supply constraints of alternative
powertrain concepts. Thereby, the scope of platform models availability that is negatively influencing the
market share was introduced similar to the work of Weikl. However, the utility reduction perceived by
customers and that is caused by missing models of an emerging powertrain concept in all market
is idered by an logistic form. Additionally, he discussed the future role of hybrid
vehicles as a transitional technology and the spatial diffusion of alternative vehicles (Keith, 2012).
To summarize, existing SD models for the diffusion processes of automotive powertrain concepts
essentially comprise four main feedback loops: Infrastructure, familiarity (or social norm), technology
attractiveness, and vehicle model availability. Additionally, there are factors, which comprise
governmental actions, international energy prices, and societal trends. This principal setup of common
standard structures of alternative automotive powertrain technology diffusion should form the basis for
the HCV market as well. Nevertheless, adjustments are required to furthermore assure the structural
validity of the SD model.
2.2 Organizational adoption of emerging technologies in HCVs
Bearing in mind the different structure of the HCV market as a B2B industry, some general aspects
should be taken into account when transferring the main feedback loops from the PC to the HCV market.
On the one hand, organizational buying processes are different to individual ones. On the other hand,
diffusion of emerging technologies takes place in a setting of organizational rather than individual
adoption of innovations. Organizational adoption is principally influenced by the environmental context,
the organizational context, and the perceived technological characteristics (Rogers, 2003; Frambach &
Schillewaert, 2002; Tomatzky & Fleischer, 1990). This is predominantly congruent to individual adoption
processes and fits to existing SD models for the PC market: Infrastructure density, technology attractive-
ness, and vehicle availability affect both the environmental context and the perceived technological
characteristics. Additionally, Frambach & Schilleweart stresses the relevance of the social network as
well as the observability of an innovation (Frambach & Schillewaert, 2002). Thus, aspects of familiarity
should play a role in this context as well. At the same time, organizational buying is primarily dominated
by the process orientation of companies. Bansch mentioned four generic criteria to differentiate
organizational from consumers’ buying behavior: higher specifity of demand, higher number of persons
involved, stronger tendency towards rationality, and a longer purchase decision process (Bansch, 2002).
Summarizing differences of the HCV and PC market regarding buying and adoption processes, customers
on the HCV market are assumed to act generally more rational in aspects of costs and suitability to the
transport task. Nevertheless, familiarity affects the buying process by the awareness of the people
involved within the process and the observability of new products or technologies. At the same time the
organizational processes differ from company to company. Same applies to the organizations’ size,
structures, and slack (Tomatzky & Fleischer, 1990). Therefore, there are no uniform preferences and
adoption rates within the HCV market expected, equal to the PC market.
3 Development of the market model framework
The adaptation of SD models from the PC to the HCV market is set in the context of the German market.
Thereby we exemplarily focus on a conventional powertrain (e.g. Diesel), an alternative fuel powertrain
(e.g. Liquefied Natural Gas), requiring a new independent refueling infrastructure, and a hybrid electric
powertrain (HEV), not requiring additional filling stations. Thus, a HEV is not changing the user behavior,
but as an ing, innovative technology it has ly agreed characteristics. As commercial
vehicles are used for very diverse applications (Law et al., 2011), this initial study solely focuses on
HCVs used for long-haulage applications, since they represent the highest share of new registered HCVs
(Hill et al., 2011).
Table 1 — Overview of interviews conducted
Targeted Contact with Interviews Otherform Content
stakeholder _ organizational of input
group representative of
Customers Freight forwarders 7 = Buying decision, decision process, preferences
HCV industry OEM S Yes 2
Automotive supplier 3 Yes Market dynamics, technological input data
HCV dealer 5 7 Market structure, buying decision, preferences
Consultancy : Yes Market dynamics, technological evaluation
Filling stations Gasindustry 1 Yes Parameterization, general strategy, prices
LNG station provider 2 . Parameterization, general strategy, prices
Govemment Govemment . Yes Regulation, taxes & incentives
3.1 Stakeholder assessment
Secondary market studies, the theoretical framework for organizational adoption, and a qualitative
primary study yielding to gather missing data for model parameterization are used to develop the market
model. Thereby we have interviewed experts from the automotive industry, LNG providers, LNG filling
station manufacturers, and long-haulage companies (cf. Table 1). As research design, semi-structured
guided interviews and a selective content analytical transcription has been chosen (Luna-Reyes &
Andersen, 2003).
There are several stakeholders, which are influencing the diffusion of alternative powertrain concepts on
automobile markets. Albeit the stakeholders on HCV market are generally the same as on the PC market,
their behavior patterns and objectives differ. Thus, we a analysis and evaluated the
corresponding influencing factors for the diffusion process by a cross-impact study.
Customers are defined as freight forwarders using HCVs for long-haulage transportation. Although there
is a huge variety of such companies — ranging from owner drivers to multinational companies — we cluster
them in solely two different customer groups based on Roger’s Innovation Diffusion Theory: innovative
and conservative companies (Rogers, 2003). Within these customer groups the firms are assumed having
uniform as well as constant preferences and requirements. The investment decision upon a new HCV, and
a powertrain concept respectively, is reached by an organizational process of the buying center according
to the perceived technologies’ characteristics, the technology availability, and awareness. The rationality
in decision making on B2B markets has been stressed. Furthermore, surveys revealed the most relevant
criteria when buying a new commercial vehicle. Reliability, total cost of ownership and usefulness are
among the highest ranked factors (Dressler et al., 2012; Kelp & Stolz, 2011; Diez & Krauss, 2006; Frost
& Sullivan, 2010). Recently, image consideration also gains in importance, since B2B markets are
generally driven by derived demand and therefore are still d dent on end-
Hence, the societal trend towards green, innovative, and sustainable transport solutions is conveyed
through the entire supply chain.
HCV manufacturers (OEM) develop and offer commercial vehicles on the market. Due to simplification,
the supplier industry is incorporated within this stakeholder group. Based on expected customer demand,
governmental policies, and market trends, R&D expenditures are allocated for technology characteristics’
improvement and the expansion of a powertrain concept’s availability. Referring to Tomatzky and
Fleischer, the availability is a decisive factor within the adoption decision process (Tomatzky & Fleischer,
1990). We define it as the variety of power classes which are available in different HCV models for the
specific powertrain concept. Consequently, a low availability reduces the probability of a powertrain
concept to be chosen by customers, because it could be ineligible to the customers’ transport task.
Refueling infrastructure incorporates managers of public filling stations. They decide upon their expected
profitability whether they invest in alternative filling stations or not. Hence, the decision upon built-up or
removal of filling stations is mainly influenced by a powertrain concepts’ stock in the market. However,
many freight forwarders are using their on-site filling stations. In addition, freight forwarders are widely
concluding contracts with a few of public filling stations spread over their general cruising radius. In
doing so, bulk consumer prices for fuel are achieved and costs are reduced. In summary, considerably
fewer filling stations per square kilometer are required compared to PC market. A spatial disaggregation
of the refueling infrastructure, as shown in Struben (Struben, 2006), is not used in the model. Instead a
homogenous distribution is assumed and the infrastructure is separated in urban and highway stations. For
long-haulage trucks primarily highway stations are necessary for an unrestrained usage.
The Government sets market regulations, fuel and vehicle standards, taxes, and incentives. Despite the
fact governmental decision making depends on market fleet GHG emissions or OEM lobbying attempts,
modeling this as endogenous would led us out of scope of this article. Thus, governmental regulation is
5
assumed to be exogenous. The influence of governmental ion is i 1 by policies
as subsidies, taxes, and infrastructure development (Bosshardt, 2009; Zhang et al., 2011).
Besides the influencing factors of the stakeholder groups, there are further impacts. Perceived technology
characteristics defines the attributes of a HCV powertrain concept, perceived by a customer buying a new
vehicle. The development of technology characteristics doesn’t underlie solely the OEMs’ decisions but
also general market rules, e.g. experience curves and economies of scale. Decisive factors for the
organizational adoption decision are the technology availability (Tomatzky & Fleischer, 1990), total cost
of ownership, purchasing price sensitivity and basically on an intra-organizational perspective the
perceived usefulness, perceived ease of use (Frambach & Schillewaert, 2002), and image to achieve
social desirability (Venkatesh & Davis, 2000; Johnston & Lewin, 1996; Frambach & Schillewaert, 2002).
External factors are exogenous effects out of influence by the model behavior. In addition to the
governmental regulation, primarily European fuel prices, the fuel consumption and the maximum
improvement of technologies are regarded. Minor relevant factors could influence the HCV market
dynamics, but are not assumed having a major impact. Thus, these factors are not used explicitly or are
fully disregarded. Among them are interdependencies with the PC market, HCV drivers preferences or 2™
life car market (implicitly used), and biofuels, dealer I , a hic change or urt i
among others (disregarded).
3.2 Model generation
In this section we discuss the adaptation of appropriate parts of PC SD models in order to establish a SD
model for the HCV market. The aim of this model is to simulate the market shares of alternative
powertrain concepts in new registered HCVs to analyze the future market penetration. Based on the four
main feedback loops of powertrain diffusion patterns on automobile markets and the relevant stakeholders,
the market model is developed. To assure the structural validity of the transferred model, we highlight the
required changes to structure and parameterization of the original model parts, based on empirical and
theoretical foundations. The dynamic hypothesis of the SD model resulting from the four main feedback
loops and the system analysis is stated by the generalized causal loop diagram shown in Figure 1.
Organizational buying decision module
The market share of a powertrain concept is calculated based on a buying decision model using a utility
choice model, similar for instance by Struben (Struben, 2006) or Weikl (Weikl, 2010). Regarding the
organizational buying, we don’t focus on the intra-group processes of the decision making unit. For
simplification we do rather assume all people involved in the organizational buying process decide upon
one combined decision rule. Therefore, the market share of the conventional as well as the alternative
powertrain concepts is a function of an organization’s familiarity with it, the perceived technological
attractiveness, and the vehicle availability.
Furthermore, due to the rationality on B2B markets, we assume organizations judge product options fully
independent of the product option currently in use. Additionally, there is no observable error of
organizations, when evaluating a product option. Hence, the market share of a technology is directly
derived by the probability an organization is choosing a certain powertrain concept. Consequently, the
market share calculation is not based on the bass diffusion model; instead we are using a utility choice
model. Therefore, we use the Bradley Terry Luce decision rule (Green & Krieger, 1988). If an error part
in the utility evaluation of the organizational buying center is assumed, a multinomial logit model could
be used (Struben, 2006; Keith, 2012).
Due to missing empirical studies on European freight forwarders’ preferences when investing in a new
truck, utility functions are approximated based on the Prospect Theory. Thereby, losses in attribute
characteristics are judged higher then gains by the same extent (Kahneman & Tversky, 1979).
Figure 1 - Generalized Causal Loop Diagram
Filling Station Filling Station
Construction _- Coverage
——
+
RAay Required Filling
Refueling Stations
Infrastructure Loop
fo. as Innovative
yee Image
+ +
OEM R3 Market Ant) Market
Interest Vehicle Model Share Position B2c
Availability Loop & 4, +
+ +,
+
¥ Technology
Image
+ *
R&D |
investment
Economical cae
Attractiveness Ea
AL
Transport Task
7 Suitability
Economies of
Scale
R22
Peradived ialCoct
Technological ue ees
Attractiveness Loop P
Fuel Prices
Peters et al. discussed the application of the Prospect Theory for automotive markets (Peters et al., 2008).
Each utility function is parameterized with a reference utility of 1, based on current costs and preferences
of freight forwarders in Germany (cf. figure 2).
The decomposed attribute utilities A of the perceived technological attractiveness (Figure 3) are aggre-
gated to the total utility U using a polynomial non-compensatory decision rule with the attribute
coefficient B. To sum up, we define the probability p at time t an organization i is choosing a product
option j under familiarity F and availability O by:
U, (t)
p, (t) =| = |«0,(t) x F, () (1)
iy 25 (t) is i)
x k
with U,(t)=[]Ay(t (2)
Figure 2 - Prospect Theory: utility function
Utility
0 Attribute value
Perceived technological attractiveness loop
Based on customer requirements, the perceived technological attractiveness is determined by the
economical attractiveness, the suitability to the transport task and the technology image (Figure 3). The
modeling of this feedback loop is partly similar to models for PC market. The structure is slightly adopted
in order to account for HCV customer specific requirements; in contrast the parameterization is
particularly different. Technology data is mainly derived from secondary studies (Law et al., 2011; Hill et
al., 2011) and expert interviews. Referring to Weikl, we implement the technological attractiveness
endogenously (Weikl, 2010). The customer preferences represent the major difference between the PC
and HCV market. OEMs as the supply side of the market are assumed to act on both markets identical.
This means technological and economical progress of the powertrain concepts in the HCV industry
follows the same functional rules as in the PC industry.
Initially, alternative powertrain. concepts have a reduced attractiveness due to lower transport task
ility, caused by additi ll space and load, the immature technology, and missing filling
and service infrastructure. The filling station coverage is given by the infrastructure development loop.
Figure 3 — Decomposition of the perceived technological attractiveness
Economical Transport Task Technology
Attractiveness Suitability Image
TCO oe EaseofUse Usefulness Teen) |anovatves
+ Fuel costs + Component + Filling + Load *C02- + Vehicle stock
«Taxes costs stationdensity ° Space emissions
+Maintenance *OEM + Range + Noise
+ Acquisition mark-up + Service + Torque
costs * Market + Competitive
maturity position
The remaining factors of the transport task suitability are improving with increasing market share, due to
experiences and higher R&D expenditures. This functional relationship is implemented by a S-shaped
increase of the different transport suitability factors. At the beginning, experiences and expenditures
remain limited caused by low market share S. After a certain market penetration the technological
improvement rises until the transport task TS reaches asy ically the technol i
maximum. Generalized, the mathematical formulation of the improvement rate i at time t reads as follows:
i(TS(t), S(t)) = (1-TS(t))x S(t) (3)
The structure of the economical attractiveness loop is transferred widely unchanged from PC market. It is
given by the acquisition costs of a powertrain concept and the expected total cost of ownership (TCO).
According to the learning curves concept, economies of scale reduce the fixed cost (Sterman, 2000; Weikl,
2010). The variable costs are implemented as exogenous input by constant governmental regulation and
the exogenous improvement of fuel consumption.
The relevance of image consideration is increasing within the HCV industry. On the one hand, “Green
Logistics” gets a crucial decision factor in the transportation sector (McKinnon & Piecyk, 2009; Dressler
et al., 2012; Klink et al., 2010). On the other hand, Rogers mentioned image as one core construct in his
Innovation Diffusion Theory, similarly to Johnston and Lewin for organizational buying in general
(Rogers, 2003; Johnston & Lewin, 1996). Thus, the technology image is defined by an innovative image,
determined by the market share, and a green image, determined by the CO; emissions. Furthermore, the
perceived technology image is also dependent, whether a company is innovative or conservative.
Familiarity loop
Familiarity does not solely play a role for individuals but for organizations as well. Drivers of HCVs
share their experiences of trucks among themselves and convey knowledge about new powertrain
concepts into the organization. Managers of transportation companies broaden their know-how by visiting
exhibitions or consulting of HCV salesmen. Thus, word-of-mouth and advertising are a factor for the
adoption of alternative powertrain concepts by forming the consideration set and levering the
observability (Frambach & Schillewaert, 2002).
Struben defined familiarity as the “cognitive and emotional process through which drivers gain enough
information about, und ding of, and to a platform for it to enter their
consideration set” (Struben, 2006). For the diffusion of alternative powertrain concepts we transfer
Struben’s model from PC market widely unchanged to the HCV market to consider the effect of
marketing and promotion efforts as well as word-of-mouth. We extend the model by implementing the
two different B2B customer groups: innovative and conservative companies. Innovative companies are
less risk-averse and are actively seeking for new technologies improving their business performance.
Consequently, their familiarity rises faster than within conservative ones. Compared to passenger cars,
social expenditures between members of transportation firms are more rare then between private persons,
which is regarded by an adapted parameterization.
Refueling infrastructure loop
Based on an evaluation of different modeling approaches for refueling infrastructure development, we
implemented this loop according to the model of Janssen (Janssen, 2005). A spatial disaggregation of the
infrastructure as by Struben is not considered (Struben, 2006). Routes of HCVs are principally more
predetermined than those of PCs. A spatial disaggregation of the refueling infrastructure would require
very detailed information about transport flows. The simplified approach of Weikl (Weikl, 2010) doesn’t
seem to be appropriate, since the filling station (FS) density is assumed to have major importance for the
diffusion of alternative fuel HCVs. The approach of Janssen (Janssen, 2005) and Keles & Wietschel
(Keles et al., 2008) are similar, however, Janssen’s work is preferred due to the simplified replicability.
Nevertheless, the idea of Keles & Wietschel about separating the refueling infrastructure into urban and
highway filling stations is implemented. Companies in transport sector use - in contrast to private
passenger cars — company-owned on-site filling stations and contracts with public filling stations. Thus, it
is assumed fewer public filling stations are required in order to achieve a satisfying coverage.
Additionally, we solely consider highway filling stations, since those are predominantly important for
long-haulage HCVs. Nonetheless, there exists a central uncertainty about the number of filling stations
required for a satisfying coverage for the HCV customers. By developing different scenarios, this issue is
regarded in the upcoming section.
The adoption parameters for the Bass Diffusion model of the filling station managers are assumed to be
similar to PC market (Janssen, 2005). However, the parameters for profitability are adjusted according to
secondary studies and expert interviews.
Vehicle model availability loop
The interest of an OEM for a powertrain concept j depends on the societal and economical pressures on
the OEM to innovate. The economical pressure is determined by the actual market share S and the
awareness of technologies to customer, perceived by the OEM Market Research R. The societal pressure
is explained by the average change of ecological awareness E and the political will for GHG emission
reduction. Based on the assumption PC OEM are acting according to comparable decision miles as HCV
OEM, the OEM interest O is explained similar to Weikl, incorporating the HCV market specificities
(Weikl, 2010).
0', (t) =(1-0, (t))x(S, (t) +E, (t) +R, (t) +x 5°" (t) 4)
Finally, the vehicle model availability is directly derived by the interest of the OEM, delayed by the
development time for new powertrain concepts and model generations of the OEM (Weikl, 2010;
Sterman, 2000).
4 Discussion of the simulation model
The SD model is implemented based on the market model framework. By using an exemplary Base Case,
the structural validity of this model is discussed, the model generated dynamics are analyzed and the most
sensitive are highli d. The ization of the Base Case corresponds to the referenced
models, if not otherwise stated. The HCV market specific parameters are provided in Table 2.
Table 2 - Parameterization Base C ase
Loop Variable Value Source/Comment
Required FS density [Dmnl] 10% “Assumption based on expert interviews
Interval proactive FS managers [year] [SOP-2025] Assumption based on expert interviews
: Standard gross margin on LNG [€] 0,05, Calculation based on expert interviews
FS investment cost [Te] 500 Expert interviews
E Numberof highway fuel stations [FS] 350 prea oa gare ee aja
Govemmental FS program [FS/year] 2 Expert interviews, project LNG Blue comidor 2015 — 2020
5 D HEV LNG
Gi Startof Production (SOP) [year] = 2016 2018 Expert interviews
2 OFM maricup [Dmal] 50% 50% 50% Profit margin of OEM on unit costs
Engine noise [0,10] 8 9 10 _ Expertjudgments, noise as a potential restriction
‘Add. load [eg] + 4350 4300 TIAX (2011) & expert judgements
® Add installation space [D mal] 10 85 9 _ Expertjudgments & TIAX (2011)
3 Power! torque (Dmnl] 9 95 8 _ LNG:lowertomue, HEV: Engine forpeak-power
E Tank volume [ltr] / [kg] 720 720 200 Expert interviews
E& —Avg.fuelconsumption 2020 {lt /[kg] 31 ltr 291tr 26kg AEA (2011) & expert interviews, HEV: 6% fuel efficiency
Unit costs 2020 [Te] 17 245 30 —_Expertinterviews, no additional resale value for LNG & HEV
Leaming rate [Dmnl] -1% 12% -10% — Costreduction in percent by doubling of production volume
Fuel price 2035 {ef /[e/ka] 2,34 2,34 1,67 interviews and 2025 — 2030 to 2035
Period of usage [year] 4 Expert interviews
di Interest rate 10% Avg. capital costs, expert interviews
Effective contact rate user / non-user 10% / 5% Own assumption: Less B2B contacts than private contacts
8] _Riskaverso conservative companies 10% - 0% Decreasing with increasing market share Rogers (2003)
O eeemeren eee 20% Hee ol seer get es ee ae
10
4.1 Structural validity
Regarding the structural validity of the simulation model it can he stated so far: based on validated SD
simulation models for the automotive market aiming to forecast the market penetration of altemative
powertrain concepts in passenger cars a market model for the same purpose for HCV has been derived.
By evaluating existing models, those model parts fitting best to the problem at hand were identified.
Nonetheless, the differences of both markets have been highlighted by using empirical and theoretical
findings. These differences are considered by structural changes and an adjusted parameterization. Thus,
the model should represent — to certain extent as a heuristic — the real system. Additionally, a dimensional
consistency test has been performed successfully. To sum up it can be stated, the model passes the direct
structure test (Barlas, 1996).
Besides the direct structure tests, there are structure-oriented behavior tests for evaluating the structural
validity. Thereby we use extreme-condition and behavior sensitivity tests (Barlas, 1996). Moreover, we
analyze the dynamic behavior of crucial feedback loops. With the given structure and parameterization
we simulated a Base Case of the market share development for the three powertrain concepts:
conventional Diesel engine powered truck [D], Diesel hybrid electric truck [HEV], and a monovalent
liquefied natural gas truck [LNG] (Figure 4).
Figure 4 - Base Case simulation
%
100 +
80 5
60 4 [LNG]
40 5
[HEV]
20 4 — 0)
0-7 T T T 1 Year
2010 2015 2020 2025 2030 2035
The simulation of the Base Case shows that the diffusion of the HEV is starting right after the start of
production (SOP) in year 2016. Due to lower TCO and a green innovative image, at first Innovators and
later on Mainstream buyers start adopting this new technology. With i ing market ion the
transport task suitability improves and the customers get more and more familiar with this technology.
The successful diffusion process gets interrupted by a strong and fast market penetration of the LNG.
After the SOP in 2018 it takes roughly 10 years until LNG reaches a noticeable market share. Despite
marketing efforts and a governmental LNG station program, the filling station density remains lower than
needed. With an increasing economical attractiveness and familiarity, suddenly a tipping point is reached
and a self-sustaining process of infrastructure built-up, transport task suitability improvement and
increasing familiarity starts. As a result, the diffusion of the HEV fails to certain extent and LNG would
get the most widely used technology. C ly, the c ional Diesel engine would get a niche
application towards 2035.
Based on this Base Case simulation, in the following we highlight those parameters to which the model is
highly sensitive and discuss whether this behavior could fit to the real system. Additionally, we evaluate
the impact of uncertain parameterization. Therefore, we increase all parameters by 1% and compare the
average relative effect on the market share of LNG for the years 2020, 2025, 2030, and 2035. Figure 5
shows the most sensitive parameters of each feedback loop.
11
Figure 5 - Most sensitive parameters on LNG market share 2035 by module [% ]
Economical Transport OEM Familiarity Infra-
Attractiveness Task structure
Suitability
0/0) amle)20] a
-0,9
TCO
Fud consumption
Endcustomerprice
The exponential effect of the learning rate in decreasing costs of alternative powertrain concepts has a
major impact on the market share development. The relevant buying criteria for freight forwarders have a
significant influence as well, according to the sensitivity of costs, fuel consumption, and transport task
suitability. Likewise, the sensitivity of the infrastructure loop highlights the particular importance of the
filling station density. The marketing effectiveness has a significant impact on the familiarity loop,
whereas the effective contact rate between users and non-users has a minor influence. On the one hand,
this fits to the parameterization derived by theory and studies but contradicts partially some expert
interviews. Thus, a future refinement of the familiarity loop by using empirical insights is advisable.
Varying a parameter by 1% provides a static sensitivity analysis, since it solely evaluates the effect in an
incrementally changed market environment. Dynamic or disruptive changes can’t — and shouldn’t — be
expected. Therefore, we take up the results of the HCV market system analysis and evaluate whether
effects cause an intended or unintended disruptive model change. Consequently, we discuss the following
hypotheses:
H1: Using a MNL buying decision model will not cause a disruptive model change.
The different methods to convey a customer’s utility into the resulting market share are heuristics of the
reality. Nevertheless, the decision rules shouldn’t cause a disruptive model change. The comparison
between the Base Case with the Bradley-Terry-Luce utility choice model and a MNL-model reveals
similar behavior and trends of the market share development. However, using MNL the market
penetration of LNG starts immediately after the SOP caused by the compensatory choice model.
Additionally, the unobservable part of the utility causes a lower total sensitivity of the model. To sum up,
we suggest evaluating this issue by a thorough customer study, e.g. conjoint analysis, to determine the
choice model empirically. theless, the general i of the model remains.
12
Figure 6 - Base Case simulation with Bradley-Terry-Luce (left) vs. MNL-model (right)
% %
100 + 100 +
07 5 ‘hy 804
radley-Terry-
60 Luce 60 4 MNL-model
40 4 104 [LNG]
[HEV]
—
204 Nn. a4 (D]
0 T 7 T 1 0 T T T T 1 Year
2010 «2015 = 2020S 2025. 2030S 20385. 2010 2015 2020» 2025S 2030)» 2035
H2: A high required station density will prevent LNG from market diffusion
In order to test this hypothesis we conduct a Monte Carlo simulation of the Base Case with varying the
parameter required FS density using a normal distribution with 1p = 10% and o = 5%. The individual
traces show, this parameters prevent the diffusion of the LNG significantly. The plot of the LNG market
share depending on required FS density shows the functional relationship on the LNG market share in
2035. Therefore, the hypothesis is corroborated. However, it shows the outstanding influence of this
parameter and highlights the importance of a valid and empirically derived value.
Figure 7 - Monte Carlo Simulation of the LNG market share (left) and results plot for the year 2035
depending on the parameter required FS density (right)
80 80
zg — BaseCase: 10% FS density 2035
& — 15% FS density 60
g — 20% FS density
: 40 &
8
: 20 \
= = 0
2010 2015 2020 2025 2030 2035 #5 10 15 20 25
Year FS density [%]
H3: If LNG fuel costs equal the costs for Diesel, the LNG powertrain concept will not penetrate the
market
Referring to Figure 8 we corroborate this hypothesis. The major benefit of the LNG powertrain concept of
substantially lower fuel costs doesn’t exist anymore. Due to higher acquisition costs and lower transport
task suitability, the diffusion of this powertrain concept would fail.
H4: A focused governmental filling station program with 50 operating filling stations towards 2020 will
push the diffusion of the LNG powertrain concept
Towards 2035, the increased governmental station program solely has minor influence on the LNG
market share. In contrast, the diffusion starts right after the SOP, since the transport task suitability is not
reduced by missing refueling infrastructure. Consequently, we corroborate this hypothesis. However, the
lack of familiarity, availability, and disadvantageous attributes of the transport task suitability limit the
market penetration in early years. Concluding the results of Figure 8 and Figure 7, a market saturation of
the LNG powertrain concept of roughly 65% towards 2035 is recognizable in absence of an
infrastructural constraint.
13
Figure 8 - Scenario simulation for H4 and H5
80
60 — BaseCase
— H3:LNG equals Diesel price
40
20
0-7 7 T T T 7 1 Year
2014-20169 2018 = 2020S 2022) 2024 »=— 2026 )9=— 2028 )3=— 2030» 2032» 2034 =. 2036
In summary, the sensitivity analysis and the disruptive parameterization changes reveal reasonable
behavior patterns, which fit to the real system. Thus, the model passes the structure-oriented behavior test.
Therefore, the transfer of existing SD models from the PC to the HCV market has been successful in
terms of the structural validity. As a transient behavior model, Barlas suggests the “use of graphical and
visual measures of typical behavior features” for behavior pattern tests (Barlas, 1996). Regarding the
behavioral validity, the model shows correct patterns in terms of trends and general behavior in the Base
Case parameterization. However, we have outlined factors, which are on the one hand highly sensitive
and on the other hand highly uncertain due to missing empirical insights. Hence, we propose to judge the
behavioral validity finally after more empirical data has been gathered on the HCV market.
4.2 Discussion of the model results
Despite the fact that the model hasn’t conclusively passed the behavioral pattern test, we are confident in
the general behavior. For this reason, we discuss the revealed interdependencies and outcome of the
diffusion patterns of alternative powertrain concepts on the HCV market.
The Base Case simulation shows that hybridization of long-haulage trucks is a promising measure to
reduce GHG emissions of the transportation system initially. Customers quickly adopt this new
technology, due to TCO advantages, similar handling, and a green as well as innovative image. However,
the HEV turns out as a transition technology towards a low carbon transportation system. In our case the
LNG powertrain concept, but potentially other alternative fuel trucks as well, has significantly higher cost
saving potential. Y et, these technologies require a new independent infrastructure and are fundamentally
driven by its development. The network effect of the infrastructure development loop for long-haulage
HCVs seems to be much more essential than on the PC market. On the one hand, a missing infrastructure
hinders the usage of such HCVs; on the other hand, high annual mileage and fuel consumption offer a
large as well as stable turn-over potential for LNG stations. With the given parameterization, for 2035 we
prospect an annual LNG consumption in Germany of about 3 bn kg. Adducing a standard gross margin on
LNG of 5 €Cent/kg, the profit potential amounts to over 150 m€ annually.
The investigation of hypothesis 4 shows, an initial satisfying German LNG infrastructure network of 50
stations could be achieved by investment costs of roughly 25 m€ (cf. Table 2). Therefore, we assume the
largest lever for the diffusion of altemative LNG powertrain concepts by the refueling infrastructure
development. In contrast, a subsidy for LNG trucks would not significantly influence the diffusion before
the refueling infrastructure development has started. Nevertheless, increasing LNG taxes would hinder
the LNG diffusion as hypothesis 3 reveals. In the case of HEV, a legislation to compensate for the
additional load and a technology subsidy provides the most promising lever for the future market share.
By using the prospected market share of the Base Case simulation, we estimated the sales potential of
alternative powertrain concepts in long-haulage HCVs in Germany. We assume a long-haulage truck
14
share of 40 % (Hill et al., 2011) and sales forecasts for Germany based on a large automotive industry
market information provider. Figure 9 shows esti for the ing years and highlights the
enormous market potential of alternative powertrain concepts. This could justify the manufacturer’s
in the devel of these technologies. Furthermore, the average GHG emissions of the
new registered long-haulage truck fleet would reduce from roughly 900 g/km per truck by 20% to 725
g/km in 2035. However, these numbers should solely provide a tendency towards 2035, due to some
uncertain parameters.
Figure 9 - Sales potential of alternative powertrain concepts for long-haulage trucks in Germany
[in thousands units per year]
unc)
Grey)
0 | il
2020 2025 2030 2035
To sum up, the model forecasts a substantial transition of the long-haulage truck market. Albeit we
adduced LNG and HEV as exemplary powertrain concepts, the shift towards low carbon transportation
could also be driven by other concepts. Synthetic fuels or fuel cells, for instance, could equally substitute
the conventional Diesel powertrain. However, without a satisfying filling station density this transition
will not take place. Particularly in early years, conservative transportation companies are skeptical
towards new powertrain concepts. Marketing for the superior economical attractiveness and the deletion
of transport task suitability restrictions by the OEMs are effective measures against the skepticism.
5 Conclusion and Outlook
The aim of this paper is to discuss the application of existing PC system dynamics models for analyzing
the future market penetration of alternative powertrain concepts on the HCV market. Thereby we focused
primarily on the structural model validity, in order to get a deeper understanding of the transition towards
low carbon transportation and outline future fields of research. Summarizing, the model turns out to be
structural valid as well as capable to study the fundamental market dynamics and highlight the sensitive
factors of the diffusion process. Y et, there are several limitations due to missing empirical data and the
comprising simplification caused by focusing solely on the German long-haulage HCV market.
Implications for practitioners (Stakeholders)
Governmental stakeholders have an outstanding influence on the adoption of alternative fuel trucks. By
an external intervention for an immediate infrastructure built-up, the tipping point of the network effect
could be reached soon. Especially compared to PC market, the effect of external infrastructure built-up is
higher, as lower station density is sufficient for a satisfying coverage. Moreover, supervising the refueling
infrastructure built-up could reduce the required FS density by an optimized location planning.
A large turn-over potential for LNG providers and filling stations is expectable, if a successful market
penetration of LNG trucks would take place in Europe. Therefore, we suggest an analysis of the effect of
this scenario on the energy supply system in terms of LNG availability, distribution, and prices.
15
Implications for researchers
The adaptation of SD models from existing markets to related markets facilitates the modeling process
significantly: a basic understanding of the problem is gained much faster, collection of required data can
be conducted more purposefully, and central hypotheses can be stated early in the modeling process.
Nonetheless, a profound discussion of the structural validity by the change of structure and
parameterization is required.
Moreover, future research for a more reliable and holistic understanding of the market penetration of
alternative powertrain concepts in HCVs is needed. Qualitative and quantitative studies of the demand
side of the HCV market are rare. Analyzing the organizational adoption behavior in more detail,
determining empirically derived utility functions, and evaluating the required filling station density could
contribute extensively to the und ding of diffusion of low carbon technologies in the
freight transportation system and the behavioral validity of the SD model.
Additionally, we propose analyzing the diffusion of other powertrain concepts (e.g. fuel cell) and the
influence of vehicle measures (e.g. aerodynamics, waste heat recovery) on those. Finally, a stepwise
adoption of alternative powertrain concepts from urban, to regional, and international applications could
be possible. Thus, other HCV use cases, such as regional distribution or urban applications (garbage, city
bus, lower distribution) should be taken into account as well.
References
Al-Alawi, B. M., & Bradley, T. H. (2013). Review of hybrid, plug-in hybrid, and electric vehicle market
deli Studies, R ble and Sustainable Energy Reviews, 21(0), 190-203.
Bansch, A. (2002). Kauferverhalten. (9., durchges. und erg. Aufl). Miinchen: Oldenbourg.
Barlas, Y . (1996). Formal aspects of model validity and validation in system dynamics. System Dynamics
Review, 12(3), 183-210.
Bosshardt, M. M. (2009). Fleet dynamics: Identifying the main micro processes of technological change
within the European passenger car fleet. Diss., Eidgendssische Technische Hochschule ETH Ziirich,
Nr. 17984, 2099. Ziirich: ETH.
Bosshardt, M. M., Ulli-Beer, S., Gassmann, F., & Wokaun, A. (2007). Developing a diffusion model of
competing altemative drive-train technologies (cadt-model). In International Conference (Ed.),
Proceedings of the 25th International Conference of the System Dynamics Society and 50th
Anniversary Celebration: July 29-August 2, 2007, Boston. New Y ork N.Y: System Dynamics Society.
Diez, W., & Krauss, H. D. (2006). Die europaische Nutzfahrzeugindustrie im Zeichen der
Industrialisierung.
Dressler, N., Gundermann, S., Shen, J., Nagashima, S., Keese, S., Winterhoff, M., Nilsson, P. M., &
Pietras, F. (2012). Truck Transportation 2030: Impacting the commercial vehicle industry.
European Commission (2011). White Paper: Roadmap to a Single European Transport Area - Towards a
competitive and resource efficient transport system. Brussels.
European Commission (2012). EU transport in figures. EU transport in figures. Luxembourg:
Publications Office of the European Union.
Frambach, R. T., & Schillewaert, N. (2002). O: ional innovation adoption: a multi-level framework
of determinants and opportunities for future research. Marketing Theory in the Next Millennium,
55(2), 163-176.
Frost & Sullivan (2010). Strategic Analysis of North American and European Hybrid Truck, Bus and Van
Market. London.
Green, P. E., & Krieger, A. M. (1988). Choice rules and sensitivity analysis in conjoint simulators.
Journal of the Academy of Marketing Science, 16(1), 114-127.
Hill, N., Finnegan, S., Norris, J., Brannigan, C., Wynn, D., Baker, H., & Skinner, I. (2011). Reduction
and Testing of Greenhouse Gas (GHG) Emissions from Heavy Duty Vehicles: Lot 1: Strategy. Final
Report to the European Commission - DG Climate Action.
16
Janssen, A. (2005). Modeling the Market Penetration of Passenger Cars with New Drive-train
Technologies. DISS. ETH, 15855. Zitrich.
Janssen, A., Lienin, S. F., Gassmann, F., & Wokaun, A. (2006). Model aided policy development for the
market penetration of natural gas vehicles in Switzerland. Transportation Research Part A: Policy
and Practice, 40(4), 316-333.
Johnston, W.J., & Lewin, J. E. (1996). Organizational buying behavior: Toward an integrative
framework. Journal of Business Research, 35(1), 1—
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk. Econometrica:
Journal of the Econometric Society, 263-291.
Keith, D. (2012). Essays on the Dynamics of Alternative Fuel Vehicle Adoption: Insights from the
Market for Hybrid-Electric Vehicles in the United States. Dissertation. Massachusetts Institut of
Technology.
Keles, D., Wietschel, M., Rentz, O0., & Mdst, D. (2008). Market penetration of fuel cell vehicles —
Analysis based on agent behaviour. International Journal of Hydrogen Energy, 33(16), 4444-4455.
Kelp, R., & Stolz, L. (2011). European Truck Customer 2010: Customer expectations in the commercial
vehicle industry.
Klink, G., Krubasik, S., & Graf, A. (2010). Nutzfahrzeuge werden grin. Diisseldorf.
Law, K., Jackson, M. D., & Chain, M. (2011). European Union Greenhouse Gas Reduction Potential for
Heavy-Duty Vehicles. Cupertino.
Luna-Reyes, L. F., & Andersen, D. L. (2003). Collecting and analyzing qualitative data for system
dynamics: methods and models. System Dynamics Review, 19(4), 271-296.
McKinnon, A. C., & Piecyk, M. I. (2009). Measurement of CO. emissions from road freight transport: A
review of UK experience. Energy Policy, 37(10), 3733-3742.
Peters, A., Mueller, M. G., de Haan, P., & Scholz, R. W. (2008). Feebates promoting energy-efficient cars:
Design options to address more and possible ing effects. Energy Policy, 36(4),
1355-1365.
Rogers, E. M. (2003). Diffusion of innovations. (5th ed.). New Y ork, NY : Free Press.
Shafiei, E., Stefansson, H., Asgeirsson, E. I., Davidsdottir, B., & Raberto, M. (2013). Integrated A gent-
based and System Dynamics Modelling for Simulation of Sustainable Mobility. Transport Reviews,
33(1), 44-70.
Sterman, J. D. (2000). Business dynamics: Systems thinking and modeling for a complex world. Boston:
Irwin/McGraw- Hill.
Struben, J. (2004). Technology transitions: identifying challenges for hydrogen vehicles: DRAFT. In M.
Kennedy, G. W. Winch, R. S. Langer, J. I. Rowe, & J. M. Yanni (Eds.), Proceedings of the 22nd
International Conference of System Dynamics. Oxford.
Struben, J. (2006). Identifying Challenges for Sustained Adoption of Altemative Fuel Vehicles and
Infrastructure. SSRN Electronic J ournal.
Struben, J., & Sterman, J. D. (2008). Transition challenges for alternative fuel vehicle and transportation
systems. Environment and Planning B: Planning and Design, 35(6), 1070- 1097.
Tomatzky, L. G., & Fleischer, M. (1990). The processes of . (4th ed.). Issues in
organization and management series. Lexington, Mass: Lexington Books.
Venkatesh, V., & Davis, F. D. (2000). A Theoretical Extension of the Technology Acceptance Model:
Four Longitudinal Field Studies. Management Science, 46(2), 186-204.
Weikl, R. (2010). Simulati zur Absch der Mark icklung unterschiedlicl
Antriebsvarianten am deutschen Fahrzeugmarkt. Di i ihe / GUC, G t fiir
Unternel Mt und C: ling, 42. Chemnitz, Chemnitz: GUC Verl. der Ges. fir
Unternel i und C li
Zhang, T., Gensler, S., & Garcia, R. (2011). A Study of the Diffusion of Alternative Fuel Vehicles: An
Agent-Based Modeling A pproach*. Journal of Product Innovation Management, 28(2), 152-168.
17
