Identifying uncertainties in the development of Oil
Sands in Nigeria: An exploratory SD Model
S. Koul*!, 0. A. Falebita!, J-F. K. Akinbami’, J. B. Akarakiri®
Saroj Koul*!
1yindal Global Business School (JGBS), OP Jindal Global University Delhi, India
E-mail: skoul@jgu.edu.in
Oluwabunmi A. Falebita’*
1yindal Global Business School (JGBS), OP Jindal Global University Delhi, India
E-mail: 14jgbs-phd-oafalebita@ jgu.edu.in
John-Felix. K, Akinbami”
> Centre for Energy Research and Development (CERD),
Obafemi Awolowo University Ile-Ife, Nigeria
E-mail: akinbamijfk@ gmail.com
Joshua B. Akarakiri?
3 A frican Institute for Science Policy and Innovation (AISPI),
Obafemi Awolowo University Ile-Ife, Nigeria
E-mail: jakarakiri@ yahoo.co.uk
*!Corresponding A uthor
Identifying uncertainties in the development of Oil
Sands in Nigeria: An exploratory SD Model
Abstract:
The rapid increase in attention and interests in unconventional resources have been attributed
to their vast occurrences across the globe; the dwindling reserves of conventional resources,
and improvements in technology making their extraction more profitable. The occurrence of
oil sands in Nigeria has been known for close to a century with estimated reserves of about 43
billion barrels of recoverable crude oil. This resource however is still not commercially
developed despite several attempts made by the government. It is therefore presumed that there
are myriad of uncertainties surrounding this development as is the case with fossil fuel
resources especially unconventional like oil sands. The uncertainties may be due to certain
factors including technological, environmental and political.
In this exploratory study, firstly these uncertainties associated with the development of oil
sands in Nigeria are identified and categorized. Secondly, the system dynamics approach is
employed to explore cause-and-effect relationships among the uncertainties in relation to oil
sands development.
This study is made possible through the funding received from Organization for Women in
Sciences for the Developing World (OWSD) and Swedish International Development
Cooperation Agency (SIDA).
Key Words: Uncertainties, Oil (Tar) Sands, Nigeria, System dynamics, causal loop diagram,
exploratory study
1. Introduction
Energy is the underpinning of all aspects of our “Global Economy” (Hall et al., 1986; Bassi et
al., 2010); it is also central to crucial world issues such as climate change, food security,
international trade and economics, national security and geopolitics (Ate and Nwoke, 1998;
Smil, 2003 and Odell, 2004). Without adequate actions aimed at maintaining energy
availability, the well-being of our increasingly urbanized, industrialized and growing world
population faces the prospects of a number of severe concems such as; reduced standard of
living, declining access to food (Pimentel, 2008) and clean water supplies (Gleick et al., 2006),
and the contraction of global trade and GDP (IPCC, 2007a,b,c). A nation that can competently
tackle its energy development concems, and equally manage essential global energy resources
is bound to play a prominent role in international markets (Odell, 2004; IEA, 2010).
Nigeria as a nation can be referred to as “energy rich” because of the presence of several
mineral resources widely distributed in various states of the nation; this is aside her crude oil
which is responsible for about 95 percent of her foreign earnings (Sambo, 2009; Mbasuen &
Darton, 2012b). However, the country still faces difficulties in tackling her energy demand
issues, coupled with the fear of dwindling crude oil reserves (Oketola, 2014). Besides, it has
2
been put forward that unconventional hydrocarbon resources especially Oil Sands (OS) have
the potentials to generate more royalty and tax revenues for the Nigerian government, as well
as cost reduction and foreign exchange reservation on a long-term basis (Adedimula, 2000;
Ayodele, 2011; and Falebita, 2014).
The desire to increase foreign earnings, spread out energy sources as well as diversify the
economy through the development of other energy sources, spurred the attention on the
Nigerian OS, especially with its various applications in several industries. In addition, other
potential benefits include increased foreign earnings for the nation, reduction of importation of
bitumen, asphalt and other derivatives of OS; in addition to increase in oil reserves of the
nation to combat the fear and challenge of dwindling conventional oil reserves.
Yet the energy system is a very dynamic system and its developments such as that of the
OS are bedevilled by myriad of uncertainties such as technological, resource, political, and
environmental. This study is therefore aimed at developing capabilities for identifying the
uncertainties associated with the development of OS in Nigeria with the aid of system
dynamics methodology. The paper is consequently structured as follows: first we attempt to
introduce OS, its occurrence and associated uncertainties; and then the definition of
uncertainty; followed by techniques for identifying uncertainties and the SD approach to
uncertainties. Lastly we present the findings and conclusions.
2. Oil Sands: Availability and Usefulness, Economy, Uncertainty in its Development
Oil sands is a naturally occurring mixture of sand, clay or other minerals, water and
bitumen, which is heavy and extremely viscous oil that requires treatment before it can be
extracted and put to use by refineries for the production of usable fuels such as gasoline and
diesel. Bitumen is soluble organic matter derived from degradation of oil either as seeps that
come to surface or within shallow subsurface reservoirs (EMD Report, 2013) Bitumen is so
viscous that at room temperature it behaves like cold molasses. A variety of treatment methods
are currently available to OS producers and new methods are put into practice as more research
is completed and new technology is developed (EMD Report, 2013; Alberta Energy, 2014a).
Over the last twenty years, the increase in non-conventional oils has contributed largely to the
renewal and increase seen in global reserves (World Energy Council, 2013).
OS resources are found in various nations across the globe including Venezuela, US,
Russia, former Soviet Union, Cuba, Indonesia, Brazil, Trinidad and Tobago, Jordan,
Madagascar, Colombia, Albania, Romania, Spain, Portugal, Nigeria and Argentina (Isaacs,
2011; Alberta Energy, 2014b); with the major deposits in Canada and Venezuela.
Canada: The Athabasca deposit in Alberta is the largest commercial deposit, most
developed and makes use of the most advanced technological processes for production.
Alberta’s oil reserves play a vital role in the Canadian and global economy, through providing
steady, dependable energy to the world. The OS of Alberta have been described as "Canada's
greatest buried energy treasure" (Alberta Energy, 2014a). In terms of proven reserves, Alberta
OS contains proven crude oil reserves of about 173 billion barrels (bbl) making it the third-
largest proven crude oil reserve in the world (Table 1) subsequent to Saudi Arabia and
3
Venezuela; 97 percent (168 bbl) of those reserves recoverable with today’s technology are in
the OS (World Energy Council, 2013).
Total investment in Alberta OS projects should be greater than $514 billion; generating
revenues more than $2,484 billion (2013 Canadian dollars). Each dollar put in the OS
creates about $8 worth of economic activity; where 25 percent of that economic value is
produced outside A lberta that is in Canada, US and around the world (Alberta Energy, 2014b;
CERI, 2014). Projected revenues for federal and provincial government on inflation-
adjustment basis; from OS-related investment is $79.4 billion between 2012 and 2035
(Conference Board of Canada Report, 2012).
In addition to all these benefits that are being derived from Canadian oil sands, others are
(CERI, 2014);
1. Employment and job creation of about 121, 500 Albertans are employed in Alberta’s
mining, and oil and gas industries;
2. Huge federal and provincial revenues through taxes amounting to about $574 billion and
$302 billion (2013 Canadian dollars) respectively, and through
3. Carbon capture Storage, (CCS is a technology that can be used in a number of industries to
reduce CO? emissions) carbon emissions have been reduced to between 26 and 50 percent
per barrel since 1990 (Alberta Energy, 2014b).
Venezuela: The oil resource base of this country is very enormous, with remarkable increase
from proven reserves estimates of around 99.4 bbl in 2009 to about 211bbl in 2011; and then to
an approximate value of 298.4 bbl as of 2014. The recent estimates dwarfs that of Canada;
making it presently the largest in the world and the increase has been traced to massive
reserves inclusion from Orinoco’s OS deposits. Hence, she currently represents the principal
contributor (Table 1) to OPEC oil (Oil and Gas Journal, 2013; World Energy Council, 2013;
and OPEC, 2014). As at 2009 Venezuela produced over half a million barrels of oil per day
from four OS development projects: Petroanzoategui, Petromonagas, Petrocedeno and
Petropiar (Energy Information Administration, 2009a).
At least 10 percent of the annual investment in these deposits goes into social programmes
such as the provision of free health care, discounted food for poor neighbourhoods, job
creation programmes, education, and indigenous land-tilling and discounted oil prices for
exports to neighbouring Caribbean countries (Alvarez and Hanson, 2009).
The exploitation of OS in Canada and Venezuela OS (168 bbl and 220 bbl of reserves
respectively; details in Table 1) have contributed very extensively to available reserves in
these two countries by an increase of a factor of four since the 1990s. Venezuela leads the
world in terms of oil reserves, followed by Saudi Arabia and Canada (World Energy Council,
2013 and OPEC, 2014). This however is not to say that these developments do not have their
downsides such as environmental impacts on water, land, and air among others; contributing to
existing uncertainties as well as creating new ones.
Table 1: Contributory Oil reserves in billion barrels (bbl)
(Source: Modified after Oil and Gas Journal, 2013, and OPEC, 2014).
Countries Proven Crude Oil Percentage (%) Notes
reserves (bbl)
Venezuela 298.4 24.7 220 bbl is from OS
SaudiArabia | 265.8 22.0
“Canada 173.0 NotApplicable 168 bbl is from OS
IR Iran 157.8 13.1
Traq 144.2 12.0
Kuwait 101.5 84
UAE 97.8 8.1
Libya 48.4 40
Nigeria 37.1 3.1 Presently no contribution from OS
Qatar 25.2 21
Algeria 12.2 1.0
Angola 9.0 0.7
Ecuador 88 0.7
*Canada is a Non-OPEC Country
Nigeria
Nigeria presently has the second largest proved oil reserves in A frica with about 37.1 bbl (from
her conventional crude) after Libya (World Energy Council, 2013 and OPEC, 2014; see Table
1 for details). According to existing sources the country has about 43 billion barrels of
recoverable crude oil within the coastal region of Ondo State (Geological Consultancy Unit
(GCU), 1980), and probably twice this amount for the entire OS belt of 120 by 4-6 kilometres
which is yet to be developed (GCU), 1980). The OS belt (Ekweozor, 1990; and Enu, 1990)
extends from parts of Edo State through Ondo, Ogun and Lagos States and its outcroppings
have been known to be present in Westem Nigeria for about a century (Adegoke, 1974,
(Ministry of Solid Minerals Development (MSMD), 2006; (Ministry of Mines and Steel
Development (MMSD), 2010).
The distinctiveness of Nigerian OS such as; medium to good sorting of the sand grains, low
clay content, higher bitumen content and low heavy metal content give it significant advantage
over the Canadian OS (Coker, 1990). This deposit also possesses the following potential for
easy development; amenability to gravity assistance, potential for steam assistance,
amenability to open cast mining similar to and in some cases better than the Canadian OS; in
addition to its technological, economic and environmental benefits to the nation are some of
the inherent advantages of the development of this resource.
Based on studies carried out on Nigerian OS, its characteristics such as; water-wet nature
of the sand grains, textural parameters, oil saturation, general chemical properties and facies
and age relationship compare favourably with Athabasca OS (Oluwole et al., 1985 and
Ekweozor and Nwachukwu, 1989).
Tar sands, oil sands or bituminous sands are all names for a combination of clay, sand,
water, and bitumen which is viscous extra-heavy crude oil. It describes sandstones or friable
sands (quartz) impregnated with bitumen (a hydrocarbon soluble in carbon disulfide). Bitumen
is just one of the many products derivable from OS (Adegoke et al., 1974; Oblad et al., 1987;
Meyer, 1995 and Speight, 1997). However, despite all these advantages and potential benefits
derivable from the development of this resource, it is yet to undergo commercial exploration
and exploitation. The question is why? It is safe to say that there are uncertainties associated
with this development; responsible for its present state. What are they? Where are they
located? What types of uncertainties are they? What are their effects on this resource
development? These are some of the questions to be explored in this study.
3. Uncertainty Definition, Identification and Methodology
A number of authors have attempted to define the concept called uncertainty; some of the
definitions are: Uncertainty is the deficiency of knowledge (Funtowiez and Ravetz, 1990);
Walker et al., 2003 defined uncertainty as “a departure from the unattainable state of complete
determinism resulting from either a deficiency of knowledge or natural variability in a
system”; uncertainty is a situation of inadequate information revealed as inexactness,
unreliability or border with ignorance; Brugnach et al., 2007 introduced another viewpoint of
“ambiguity” in the definition of uncertainty with ambiguity defined as “...concurrent existence
of numerous equally valid frames of knowledge” (Dewulf et al., 2005). Uncertainty is
becoming a progressively more important factor in the decision making and policy analysis
because of the swiftly changing, complex, and unpredictable nature of our world (Walker,
2003).
Walker identified the following types of uncertainty encountered in decision making and
policy analysis: uncertainty about model form; uncertainty about the values of a model's
parameters; uncertainty about underlying probability distributions and structural uncertainty
“this refers to uncertainty in future structural elements of the world, unknowable at the time of
analysis” (Walker, 2000). He then classified uncertainty (Walker et al., 2003) into different
levels described as follows;
1) Location of uncertainty: where the uncertainty manifests itself in the model complex; in the
context, in the model itself (‘model technical’ or ‘model structure’ uncertainties), in the
input, in parameters or in the output. Here the marker “output” is introduced by Janssen et
al., 2010.
Level of uncertainty: where the uncertainty manifests itself along the continuous spectrum,
between deterministic knowledge and total ignorance. Here the markers are ‘statistical’,
‘scenario’ and ‘recognized ignorance’; while qualitative level of uncertainty (which cannot
be quantified, but can be described) was added by Janssen et al., 2010.
Nature of uncertainty: whether the uncertainty is due to imperfection of knowledge
(epistemic) or inherent variability of the phenomena being described; ambiguity is
introduced here by Janssen et al., 2010.
Context uncertainty: The uncertainty in the model context concerns choices made in the
step from natural system to conceptual model. Questions about the model boundaries and
&
2
fas
6
choice of input and output variables are associated with this type of uncertainty.
Assumptions or scenario’s are usually used to address these uncertainties.
Other categorizations of uncertainties include:
a) Model structure uncertainty can be described as ‘arising from a lack of sufficient
understanding of the system of reference, including the behaviour of the system and the
interrelationships among its elements’ (Walker et al., 2003). According to, Van Asselt
and Rotmans (2002), this is the most difficult type of uncertainty to address.
Model technical uncertainty concerns ‘uncertainties related to the computer
implementation of the model’ (Walker et al., 2003); it consists of both software and
hard-ware errors.
Input uncertainty is both uncertainty about ‘driving external forces that produce
changes within the system’ and uncertainty about ‘the system data that ‘drive’ the
model and quantify significant features of the reference system’; it is considered to be
stochastic in nature.
Parameter uncertainty is uncertainty related to the a-priori chosen parameters,
described by Walker et al. (2003).
Aggregated uncertainty results from all uncertainties above (Janssen et al., 2010).
=
e.
=
2
Janssen et al., 2010, based on the additions to the work of Walker et al., (2003); then came
up with an analysis framework for uncertainties. Van Asselt, (2000) developed a taxonomy for
the sources of uncertainty; here she identified 2 classes variability (ontological dimension of
uncertainty) and lack of knowledge (epistemological dimension of uncertainty). Each class is
further sub-divided as follows:
Variability: a) Inherent randomness of nature, b) Value diversity, c) Human behaviour, and
d) Societal and technological randomness;
Lack of knowledge: a) Unreliability from inexactness and b) Structural uncertainty
resulting from irreducible ignorance.
Identifying key uncertainties is one of the steps in the thinking phase of the adaptive
policymaking process as defined in the work of (Walker, Cave, and Rahman, 2001; Walker,
2000; Walker et al., 2003). Further, Walker et al., (2003) provides a theoretical basis for the
treatment of uncertainty systematically in model-based decision support activities, this
involves the use of uncertainty matrix. The uncertainty matrix enables the categorization of
uncertainties based on the three categories of; location (also referred to as ‘source’ by Eker and
van Daalen (2012a, and b), level, and nature. Other Methodologies / approaches for identifying
and analysing various types of uncertainty are summarized in Table 2.
Table 2: Different approaches for various types of Uncertainty
Uncertainty Types Methodologies/Analytic approach
Structural uncertainty 1. Ignoring uncertainties
2. Conduct ‘what if” policy analysis
3. Muddling through an adaptive approach (Walker, 2000;
Walker et al., 2003).
Quantitative Analytical approach (Walker and Haasnoot,
2011)
Level 1 Deterministic (optimization, sensitivity)
Level 2 Probabilistic (sensitivity, expected value, confidence intervals)
Level 3 Scenario analysis
Level 4 Exploratory (scenario) analysis, adaptive pathways
Approaches by J anssen et al., 2010
Model technical uncertainty Multiple simul model impl
Input uncertainty Monte Carlo analysis
Parameter uncertainty Calibration OR Selected fixed value
Aggregated uncertainty Compositional framework
4. Techniques that help deal with uncertainties:
A number of techniques have been employed in dealing with and exploring uncertainties over
the years; arising from the ease of use of cheap, powerful computing capability, as well as
combination of computational speed, graphical display, and data handling ability of modem-
day computer (Walker, 2000). The techniques that have been identified in this study are
presented in Table 3. These techniques include SD, EM and EMA, MCDA, MAMCDA,
ABM, NUSAP, PRIM, combination techniques of SD with one or more of these other
techniques (see Table 3).
Inherent advantages of SD in dealing with uncertainty that have been identified include:
ability to produce qualitative interpretations, conclusions and recommendation; through
addition of rich information; use of multi-model approaches; use of sensitivity analysis and
uncertainty analysis (Pruyt, 2014 and Walker et al., 2014); use of new advanced techniques
and tools; potential for dealing with deep uncertainty; basic foundation for most of other
techniques; and ability to work successfully with the other techniques briefly described as
follows:
EM and EMA represent a quantitative approach to uncertainty analysis, useful for
exploring deep (multi-faceted and multi-dimensional) uncertainties. These techniques allow
the generation of myriad scenarios, for analysing dynamic behaviours and testing robust
policies. The resulting policies are adaptive and capable of accommodating unexpected turn of
events.
MCDA and MAMCDA are techniques that involve the modelling of multi-dimensional
issues or uncertainties, and thereafter identifying most appropriate solutions rather than
optimal ones through the ranking of a countable set of policy alternatives on a multiple criteria
basis. ABM as a technique utilizes algorithms to quantify and describe uncertainties.
NUSAP as a technique provides an analysis and diagnosis of uncertainty in science-for-
policy, through the assessment of qualitative and quantitative uncertainties based on the five
attributes of the acronym. These are Numeral, Unit, Spread, Assessment, and Pedigree, useful
8
Table 3: Techniques for dealing with uncertainties
Techniques
Description
Examples of use
System Dynamics (SD)
1. making quantitative simulation models
2. generating plausible behaviours over time
3. results in qi i
Ford et al., 1989; Ford 1989; Ford &
Bull, 1989, Ford, 1990 Lowry et al.,
2012; Walker et al., 2014
Exploratory Modelling (EM)/
Exploratory Modelling Analysis
(EM/EMA)
EM/EMA is a quantitative uncertainty analysis approach
1. it systematically explores deep uncertainty
2. tests the robustness of policies
3. generates and explores plethora of scenarios
4. simulates and analyses dynamic behaviours
Bankes, 1993; Pruyt, 2007; Agusdinata,
2008; Pruyt, et al 2011a);
Multiple Criteria Decision Analysis
(MCDA)
MCDA is an iterative technique for
1. modelling and analysing multi-dimensional issues
2. finding most appropriate solutions instead of optimal ones
i i ing, modelling and exploring
Roy and McCord 1996; Figuera, Greco
and Ehrgott 2005; Pruyt, 2007
Multi-A tribute Multiple Criteria
Decision Analysis (MAMCDA)
MAMCDA is used for describing, choosing or ranking countable sets of alternative policies on multiple
criteria
Pruyt, 2007
ABM describes elements in a model with the use of
Hamarat et al., 2013
‘Agent Based Modelling (ABM)
Compositional Fuzzy-rule based
combines the evaluation of model structure, input and parameter for their extensive
Janssen et al., 2010
Numeral Unit Spread Assessment
NUSAP is a heuristic for good practice in science for policy, which
Funtowicz and Ravetz (1990); Van der
Pedigree (NUSAP) 1. enhances reflection on the various dimensions of uncertainty Sluijs et al., 2003
2. makes uncertainties explicit
3. qualifies uncertainties using these five qualifiers: Numeral, Unit, Spread, Assessment and Pedigree
4. diagnoses uncertainties on the basis of spread and strength
Patient Rule Method (PRIM) | PRIM is used to find areas in the input space for an observed behaviour Moorlag et al 2014
SD +1 or more other Tech
E ry System Dynamics ESDMA is a mainly a quantitative multi-method that involves; Auping, 2011; Eker and van Daalen
Modelling and Analysis (ESDMA) = 1. exploration to broaden the horizon 2012 (a, and b); Pruyt, et al, 2011;
SD +EMA 2. deep robustness aimed at testing, design of adaptive policies, and resilient systems. Pruyt and Kwakk el al., 2012 a& b
3. particularly appropriate for systematically exploring and analysing plethora of plausible dynamic
behaviours over time
4. testing robustness of policies over all these scenarios
5. it leads to qualitati and
SD +MAMCDA Pruyt, 2007
SD +EM Pruyt, 2007
SD +ABM + PRIM
Mooniag, et al., 2014
CALIBRATING SYSTEM DYNAMICS MODELS OF
TECHNOLOGY DIFFUSION WITH STRUCTURAL BREAKS:
THE CASE OF ANDROID HANDSETS
Abhinay Puvvala*
Tata Research Design and Development Centre, Systems Research Laboratory
54, Hadapsar Industrial Estate, Pune — 411013, India
Tel: +91 9923137399
Email: abhinay.puvvala@tcs.com
Amitava Dutta
School of Management, George Mason University
4400 University Drive, Fairfax, Virginia 22030-4444, USA
Tel: +1-703-993-1779
Email: adutta@gmu.edu
Rahul Roy”
Indian Institute of Management Calcutta
Joka, Diamond Harbour Road, Kolkata — 700104, India
Tel: +91-33-2467-8300
Email: rahul@iimceal.ac.in
Abstract
Diffusion of new technologies has been a major application domain for system dynamics (SD) models.
A common assumption when calibrating such SD models is that the key parameters driving diffusion,
such as contagion strength, are constant over the duration of analysis. However, particularly in the
context of new technology difffusion, these parameters may change over time, sometimes
dramatically. This can result in so-called called structural breaks in the diffusion pattern.
Calibrating SD models in the presence of structural breaks presents some challenges. We discuss
these issues in the context of Android handsets, using quarterly sales data for the period 2009-2012,
and referring to specific events in its evolution.
Keywords: Diffusion models, dynamic parameters, calibration, structural break, Android.
1 Introduction
Calibration is a critical part of the System Dynamics (SD) model building process, as it helps
build confidence in the validity and usefulness of the model (Forrester & Senge, 1980),
(Barlas & Carpenter, 1990), (Sterman, 2000), (Oliva, 2003). It entails tuning the model
structure as well as parameters, so that the model-generated behavior is “right for the right
reason” (Oliva, 2003 p3). While calibration of model structure involves a process of
refutations, negation and qualitative simulation (Barlas, 1996), (Qudrat-Ullah et al., 2009),
calibration of parameters requires use of statistical techniques that results in simulated
® This work was carried out as a Doctoral Student of Indian Institute of Management Calcutta
» Author for correspondence
32" System Dynamics Conference,Amsterdam 2014 1
for providing insight on two independent properties related to uncertainty in numbers; that is
spread and strength, which are then combined in a Diagnostic Diagram mapping.
PRIM as a method that deals with uncertainties by identifying areas in a selected input
space, responsible for observed patterns and behaviour.
5. Methodology:
In this study, both primary and secondary data were collected, to obtain adequate
information on the uncertainties associated with OS development in Nigeria. Secondary data
were derived from literature, while primary data were obtained from survey of stakeholders.
These stakeholders considered of paramount importance to OS development in Nigeria,
comprised the following categories: researchers conducting studies on Nigerian OS, policy
makers in government agencies involved in policy formulation process for OS development
in Nigeria, and residents of host communities to OS deposit.
SD methodology developed by Forrester is useful for analyzing and understanding the
behavior of complex and dynamic systems. It is based on the use of informal maps, causal
relations, and feedback loops, with the aid of computer simulations to generate plausible
scenarios of the system under study. It entails the use of causal loop diagrams (CLD’s);
which are diagrams that help to depict interactions—causes and effects in existence among
variables of interest within a system—as-well-as create and indicate the direction of flow.
In uncertainty classification, Walker et al., (2003) defined what is known as location of
uncertainty in their model-based analysis of uncertainty. However, in conceptual uncertainty
analysis, Meijer (2008) introduced the term ‘source of uncertainty’ to replace location, and
identified six sources of uncertainties. These comprise the following; technological, resource,
competitive, supplier, consumer and political, Eker and van Daalen (2012b) further
introduced one more source known as societal.
In this study, we also introduced two other sources of uncertainty known as
environmental and infrastructural. These eight sources are used to categorize the myriad
uncertainties identified as associated with OS of Nigeria, and are presented in the next
section.
6. Findings: Use of SD for identifying the cause-and-effects relationships among the
uncertainties
In this section, myriad uncertainties that have been identified in association with the
development of OS in Nigeria are categorized using Walker et al (2003) framework, and
presented in the following table (Table 4) following which cause-and-effect interactions
among the uncertainties were investigated using SD methodology.
Table 4: Uncertainties associated with OS in Nigeria and their associated factors
Sources of Uncertainties Associated F actors/Variables
Technological Availability of required technology, technological ad number and type
of wells required, time needed for extraction, importation of technology
Environmental Carbon emissions, tailings generation, water requirements, land use, reclamation
and ion of land, other envi ll effects
Political Policy and legal framework, tax regimes, political will, nature and interest of
leaders, political regime/politics, land approval requirements and time,
government support
Social/societal Host community perspective and support or otherwise, livelihood of residents,
public opinion, security issues, environmental protection policy
Infrastructural Road network, drilling equipments, pipelines and other social amenities; housing
facilities, medical conveniences
Consumer/Demand Existence of conventional crude and other substitutes, availability of market,
consumer preference, price of products
Supply Availability of drilling and other equipments, quality of equipments delivered,
delivery time of equipments,
Resource (OS reserves and | Estimates of reserves, value of recoverable reserves, geological characteristics of
Human) reserves, nature of occurrence of reserves, availability of skilled and un-skilled
man-power required (human resources), quality of human resources available,
time needed for employment of required man power
Financial Investment requirements, estimated period to before break-even, return on
investments, profitability and viability.
Causal loops diagrams (CLD): the CLDs’ are diagrams that provide insight, enhance the
exploration of plausible behaviours, and help to depict the interactions; causes and effects in
existence among variables of interest within a system; they create and indicate the direction
of flow. These diagrams drawn with Vensim PLE software, showing the interactions and
effects of the identified uncertainties are as shown (Figures 1{a & b}, 2, 3 and 4).
Figure 1a and 1b is an illustration of the CLD of the uncertainties identified and their
effects on OS development; portraying probable interrelationships and effects of these
uncertainties on development of OS in Nigeria and vice versa (where a + sign implies a
positive effect, and a - sign a negative effect). The development of OS is expected to result in
a decrease in uncertainties associated with the development of this resource; however this
may not be the case especially for environmental and social as the development could lead to
creation of new uncertainties in this group that were initially not envisaged. Furthermore,
major causal links and feedback loops among these uncertainties are shown in figure 1b. For
instance, four balancing loops (B) and five reinforcing loops (R) are identified and shown, to
capture overall causal relations among uncertainties, and between uncertainties and Nigerian
OS development. Reinforcing feedback loops indicates that all the factors eventually increase
in the same direction; that is any change in a variable in this loop, results in a change in the
variable in the same direction again. While balancing feedback loops implies that the factors
increase in such a way that they counter one another; this implies that the self-change in any
variable in the loops occurs in an opposite direction. Knowledge of these causal factors is
useful in future analysis of the uncertainties for the identification of effective treatments and
generation of robust policies for the resource development. Figure 2 depicts the relationship
between FIR uncertainties and the contributory variables associated with them.
11
Social) Societal Uncertainty puppy Uncertainty
Technological Uncertamty =< Resource Uncertainty
Sama
Consumer! Demand Uncertainty
Environmental i ee
Va 3 Political Uncertainty
1 Wo
Infrastructural Uncertainty &
Financial Uncertainty
Figure 1a: CLD of Uncertainties in OS Development
8)
Infrastructural Uncertainty
‘
a Sécietal Uncertainty
B
Technological Unc Ts @
“* Uncertainty
Environmental Uncertainty
R
a > olitical Uncertainty
Supply Uncertainty
ey
Financial vai cee atl Consumer Demand Uncertainty
Figure 1b: CLD of Uncertainties in OS Development
12
Pipelines
Investment Road Network Drilling equipments
requirements
+ Social Amenities
Estimated
break-even period at
Finabcial Uncertainty
Retumon 4
Investment Estimate of
Geological recoverable reserves
Profitability and Characteristics Quality of human
Viability resource Time needed for
‘echnology employment
availability
Reserves Estimate 9 ncehainty
+ Human Resource
“ +\\ Availability
Nrasirectura
Nature of Occurrence
of Reserves Extraction
Capacity
Figure 2: CLD of Relationship b Fi ial, Infrastructural and Resource (FIR)
Uncertainties in OS Development.
Number of wells Time needed for
required extraction Other environmental
a effects
Importation of _
technolo Technological Uncertainty Carbon emissions —
N oo
Technological
Availability of advancement a sy
required technology ailings generation___-“
Policy and legal
framework a use
Resource - Environmental ition and
development policy potitical Uncertai protection policy restoration of land
Be
i et
Nature of leaders
Government
Political regime
4 PoliticaYwill power 5
Interest of leaders support
Land approval Coruption
requirements
Figure 3: CLD of Relationship b Technological, Envir 1, and Political (TEP)
Uncertainties in OS Development.
13
Availability of drilling
and other equipments eae or
Public opinion Livelihood of residents rt +
in resource area Supply Unvertainty
an : .
58 Delivery time of
Social Societal Utcertaini$ecutity issues equipments
+
; Consumer
Host C ommunity Environmental preference Availability of
perspective protection et
WW
Consumer/ Demand Uncertainty
Ph Price of products
Existence of Existence of other
conventional crude substitutes
Figure 4: CLD of Relationship between Social/Societal, Consumer/Demand, and Supply (SCS)
Uncertainties in OS Development.
The depiction (Figure 3) of the interactions and effects on TEP uncertainties by their
related variables is as shown; while Figure 4 represents the relationship between SCS
uncertainties and their variables/factors. These are anticipated relationships which may
change as more information become available during the course of the investigation and
analysis.
7. Conclusions:
This exploratory study has contributed significantly to the subject matter of associated
uncertainties in the projected development of Nigerian OS; having firstly investigated various
classes, types and taxonomy of uncertainties as well as methodologies and techniques for
approaching and handling these uncertainties. Also, the advantages of SD technique for
identifying uncertainties as well as uncertainties in the development of OS in Nigeria were
identified with their cause and effects depicted.
These uncertainties including; technological, social/societal, supply, environmental,
infrastructural, resource, consumer/demand, political and financial were identified and their
contributory variables were identified for their place in the development of OS in Nigeria.
The cause and effects (CLDs’) are anticipated relationships identified from the survey
and investigation, conducted both in literature and sampling of stakeholder groups. The work
being exploratory and work-in-progress, we present the CLDs’ useful for identifying
important feedback structures and interrelationships among the uncertainties. These are to be
developed further into modules of the major uncertainties and their likely effects on OS
development. Following which an SD model will be developed to incorporate the modules. In
order to identify adaptive policy options that could enhance the development of OS in the
near future.
14
ACKNOWLEDGEMENTS
Authors appreciate the funding received from Organization for Women in Sciences for the
Developing World (OWSD) and Swedish International Development Cooperation A gency
(SIDA). Weare also grateful to our reviewers for their valuable suggestions.
REFERENCES
Adegoke, O.S. (1974): Preliminary proposal for the exploration and Utilization of Bitumen
of Western State of Nigeria. Unpublished Report, Geological Consultancy Unit,
University of Ife.
Adegoke, 0.S., Ako, B.D., Enu, E.L., Petters, S.W., Adegoke, A.C.W., Odebode, M.O. and
Emofurieta, W.O. (1974): Geotechnical investigation of the Ondo State bituminous
sands. Report, Geological Consultancy Unit, University of Ife, Ile-Ife. Pp 257.
Alberta Energy (2014a): What is Oil Sands? http://www.energy.alberta.ca/OilSands/793.asp,
accessed 3/3/2015.
Alberta Energy (2014b): Facts and Figures, http://www.energy.alberta.ca/OilSands/791.asp,
accessed 6/3/2015.
Ate, B.E. and Nwoke, C.N. (1998): Towards a Viable Energy Policy for Nigeria. Seminar
Organized by the Nigerian Institute of International Affairs, Lagos, Jan 5-7.
Auping, W.L. (2011): The uncertain future of copper - An Exploratory System Dynamics
Model and Analysis of the global copper system in the next 40 years. Delft University
of Technology, Delft.
Auping, W.L., DeJong, L., Pruyt, E., and Kwakkel, J.H. (2014): The Geopolitical Impact of
Shale Gas: The Modelling Approach. Int’! System Dynamics Conference, Delft, the
Netherlands, 2014.
Bankes, S. (1993): Exploratory Modelling for Policy Analysis, Operations Research, 41(3),
435-449,
Bassi. A. M., Powers, R., and Schoenberg W. (2010): An integrated approach to energy
prospects for North America and the rest of the world. Energy Economics, 32, 30-42.
Brugnach, M., Dewulf, A.,Pahl-Wostl, C., Tallieu, T., (2007): Towards a relational concept
of uncertainty: incorporating the human dimension. In: CAIWA Conference, Basel.
Canadian Energy Research Institute (CERI) (2011): Canadian Oil Sands Supply Costs and
Development Projects (2010-2044). Study no 122, May 2011. www.ceri.ca accessed
10/3/2015
CERI, (2014): Canadian Economic Impacts of New and Existing Oil Sands Development in
Alberta (2014-2038) http://www.ceri.ca/images/stories/CDN_ Economic Impacts _of_
New_and_ Existing Oil Sands Development in Alberta - November 2014 - Final
-pdf, accessed 15/3/2015.
15
Conference Board of Canada (2012): Fuel for Thought: The Economic Benefits of Oil Sands
Investment for Canada’s Regions. http://www.albertacanada.com/files/albertacanada
/AIS_FuelforThought.pdf , accessed 8/3/2015
Dewulf, A., Craps, M., Bouwen, R., Taillieu, T.,Pahl-Wostl, C., (2005): Integrated
management of natural resources: dealing with ambiguous issues, multiple actors and
diverging frames. Water Sci. Technol. 52, 115-124.
Eker, S., and van Daalen, C.E., 2012(a): Uncertainties in the Development of Unconventional
Gas Production in Europe. 9” International Conference on the European Energy
Market.EEM 12, Florence, Italy.
Eker, S. and van Daalen, C. E., 2012(b): Investigating the Effects of Uncertainties Associated
with the Unconventional Gas Development in the Netherlands. 3" Intemational
Engineering Systems Symposium CESUN 2012, Delft University of Technology, 18-20
June 2012. Delft, the Netherlands.
Ekweozor, C.M. (1990): Geochemistry of Oil Sands of South-western Nigeria. In Occurrence
Utilization and Economics of Nigerian Bitumen B.D. Ako and E.I. Enu (Eds.). Proc. of
the workshop on Tar-Sands/ Ogun State University, Ago-Iwoye. Ogun State, Nigeria.
Pp. 50-62.
Enu, E.I. and Adegoke, O.S. (1984): Potential Industrial mineral resources associated with
the Nigerian bitumen, 27" International Geological Congress. Vol. VII, pp. 345.
Enu, E.I. (1990): Nature and Occurrence of Bitumen in Nigeria. In Occurrence Utilization
and Economics of Nigerian Bitumen B.D. Ako and E.I. Enu (Eds.). Proc. Of the
workshop on Tar-Sands/ Ogun State University, Ago-Iwoye. Ogun State, Nigeria. Pp.
11-15.
Falebita O. A. (2014b): The Nigerian Tar Sands: A Techno-Economic Analysis. ISBN 978-3-
659-61698-3, Lambert Academic Publishing, Saarbriicken, Germany.
Geological Consultancy Unit (GCU) Report (1980): Geotechnical Investigations of the Ondo
State Bituminous Sands. Vol. 1, the Caxton Press Limited Ibadan.
Hall, C.A.S., Cleveland, C. J., Kaufmann, R., (1986): Energy and Resource Quality the
Ecology of the Economic Process. John Wiley and Sons (Wiley Interscience New
Y ork).
Intemational Energy Agency (2010): World Energy Outlook, OECD/IEA, Paris
http://www.responsiblenergy.org/tarsands.asp, accessed 26/03/2011
Janssen, J.A.E.B., Krol, M.S., Schielen, R.M.J., Hoekstra, A.Y., de Kok, J.-L., (2010):
“Assessment of uncertainties in expert knowledge, illustrated in fuzzy rule-based
models”. Ecological Modelling 221 (2010) 1245-1251.
Meyer, R. (1995): Bitumen, In Encyclopedia of Energy Technology and the Environment,
Vol.1, A. Bisio and S. Boots (Eds.), John Wiley and Sons, New Y ork, N. Y.
Ministry of Solid Minerals Development (2006): Technical Overview; Nigeria's Bitumen
Belt and Development Potential, Pp 27.
16
Ministry of Mines and Steel Development (2010): Bitumen and Bitumen Exploration
Opportunities in Nigeria. Pp 14.
Oblad, A. G., Bunger, J. W., Hanson, F. V. (1987): Tar Sand Research and Development,
Ann. Rev, Energy 12. Pp 283-336.
Odell, P. R. (2004): Why Carbon Fuels will dominate the 21% Century's Global Energy
Economy, Multi-Science Publishing Company Ltd., Brentwood, Essex, UK.
Ojeme, S. (2011): FG moves to develop bitumen deposits. The Punch Newspaper, Nigeria,
Thursday, August 4, 2011. Pp 32.
OPEC, (2014): OPEC Share of World Crude Oil Reserves, in Annual Statistical Bulletin.
http://www.opec.org/opec_web/en/publications/202.htm, accessed 14/3/2015.
Pimentel, D. (Ed.), 2008: Biofuels, Solar and Wind as Renewable Energy systems: Benefits
and Risks. Springer.
Pruyt, E., Kwakkel, J., Yucel, G., and Hamarat, C. (2011a): Energy Transitions towards
Sustainability I: A Staged Exploration of Complexity and Deep Uncertainty. Int’]
System Dynamics Conference 2011a, Washington DC, USA.
Pruyt, E. (2014): System Dynamics and Uncertainty. Int’] System Dynamics Conference
2014. Delft, the Netherlands.
Smil, V. (2003): Energy at the Crossroads: Global Perspectives and Uncertainties, The MIT
Press. 0262194929.
Speight, J.G. (1997): Bitumen, in Kirk-Othmer Encyclopedia of Chemical Technology.
van Asselt, M. (2000): Perspectives on uncertainty and risk: the PRIMA approach to
decision support, Kluwer Academic: Dordrecht. 1.2.3.4.5.
Walker, W. E (2000): Uncertainty: the Challenge for Policy Analysis in the 21‘ Century,
Lecture presented on November 29, 2000 at Delft University of Technology on the
occasion of the inauguration of the author as Professor of Policy Analysis and Decision
Support Systems in the Faculty of Technology, Policy, and Management.
Walker, W.E., Jonathan Cave, J. and Rahman, S.A., (2001): Adaptive Policies, Policy
Analysis, and Policymaking, European Journal of Operational Research, 128(2).
Walker, W.E., Harremoes, P., Rotmans, J., Van Der Sluijs, J.P., Van Asselt, M.B.A., Janssen,
P., and Krayer Von Krauss, M.P., (2003): “Defining Uncertainty - A Conceptual Basis
for Uncertainty Management in Model-Based Decision Support’, Integrated
Assessment, 4(1), pp. 5-17.
Walker, L.T.N., Malczynski, L.A., Kobos, P.H., and Barter, G., (2014): The Shale Gas
Phenomenon: Utilizing the Power of System Dynamics to Quantify Uncertainty. Int’l
System Dynamics Conference 2014, Delft, the Netherlands.
World Energy Council (2013): World Energy Resources: Oil. http://www.worldenergy.org/
wp-content/uploads/2013/10/WER_2013 2 Oil.pdf, accessed 14/3/2015.
17
Puvvala et al. /Calibration in presence of structural breaks
behavior mimicking observed behavior closely (Forester & Senge, 1980), (Barlas, 1989),
(Sterman, 2000), Oliva (2003).
A common assumption during calibration is that the parameters being calibrated are constant
over the period of analysis. However, this assumption can be violated, particularly for new
technology diffusion, due to evolution of technology capabilities, changing consumer
perceptions and changing regulations. Examples of statistical models of technology adoption
with non-stationary model parameters can be found in studies such as Meade and Islam
(2006). In the case of SD models of technology diffusion, the presence of non-stationary
model parameters raises two issues with respect to calibration. First, the timing of the
parameter changes, or break points, needs to be determined. Once these break points are
identified, model parameters need to be calibrated to accommodate these breaks.
We examine this problem by specifically modeling the diffusion of Android based mobile
handsets, a technology that is about six years old and is still evolving. We first develop an SD
diffusion model based on a contagion mechanism. We accommodate changes in the strength
of the contagion parameter that determines how rapidly the new technology diffuses. The
effect of Apps appears as complementary goods that contribute to network effects. The
development of a mobile operating system (MOS) such as Android is marked by a discrete
events and a steady stream of incremental enhancements. We accommodate this by allowing
breaks to occur in the contagion parameters at specific points in time.
In the next section we trace the history of Android growth and briefly review relevant
literature on diffusion of new technology. We then develop a contagion model of Android
diffusion using SD. This model is then calibrated in a way that accommodates the presence of
structural breaks in the diffusion pattern.
2 Android Mobile Devices
Android is a Linux-based operating system for mobile telephones and tablets developed by
the Open Handset Alliance in partnership with Google and other companies (Burgelman et
al., 2009). The source code is available under free and open source software licenses.
Devices running on Android adhere to the Compatibility Definition Document (CDD). The
hardware manufacturer has complete freedom to utilize and customize, and Google does not
charge any royalty from the hardware manufacturers for OS distribution rights. The first
handset device running on the maiden version of Android OS was launched on 22nd October
2008. Table 1 shows a history of Android handsets for the period 2008-2011 (German, 2011).
Table 1: Changes in Handset Model, Handset Feature and User Satisf:
Month- | Number | Min of User Min MSRP Max of Talk Min of Max of Screen
Year _| of Models Rating time Minute | Weight Size (Inch)
Oct-08 1 3.50 330 300.00 5.60 32
Oct-09 6 2.50 179 385.00 5.70 3.7
Mar-10 12 2.00 100 350.00 4.70 3.1
Nov-10 54 1.00 30 540.00 3.60 3.8
Jul-11 95 2.00 129 624.00 3.88 43
It is evident that handset price dropped dramatically and phone features improved
significantly. Interestingly, Table 1 shows that User Rating has gone down in the same
period. Google search patterns also point to growing usability issues until December 2011
2
Puvvala et al. /Calibration in presence of structural breaks
(Google Trends, 2013). In addition to handset attributes, another factor that has contributed to
the diffusion of Android handsets is the availability of complementary goods in the form of
‘apps’ that users can buy from the ‘Play Store’. Developers of Android apps enjoy a low
barrier-to-entry, but the huge diversity in the combination of different input mechanisms,
processor types and screen sizes, has caused difficulty in developing and testing of apps.
Coexistence of different Android versions has complicated the situation further.
Google made several changes in the Android play store over time. Some changes were
targeted towards handset users, while others, like the release of higher version software
development kit (SDK) and Native Development Kit (NDK) were targeted towards
developers. In July 2012, Google also made a major overhaul of developer policy (Google
Play, 2012) with the aim of improving user experience.
The preceding discussion shows that the evolution of Android platform has been marked by
both incremental as well a discrete changes.
3 Modelling Growth of Android Mobile Devices
Diffusion of a new technology typically follows a contagion process that results in an S-
shaped growth pattern (Rogers, 1976). Two functions that have been commonly used to
approximate this pattern are the Logistic and Gompertz functions (Bass, 1969), (Mahajan and
Muller, 1979). Using time series data on Android handset sales (Gartner, 2013) from 2009-
2012, we were able to fit the Logistic curve with a Mean Absolute Percentage Error (MAPE)
of 13.86% and the Gompertz curve with an MAPE of 10.12%.
For purposes of this paper, the significant point is that these diffusion curves assume that the
process underlying the diffusion is stable. Moreover, they lack the ability to explain the
mechanics of growth. As discussed earlier, however, technology diffusion processes may not
be stable and its contagion parameters may indeed change over time. In the following section
we present an SD model of Android diffusion that accommodates the possibility of changes
in the contagion parameters during the calibration process.
3.1 A Causal Model of Android Diffusion
Figure | presents a causal model of Android diffusion. At the core of diffusion is the basic
contagion mechanism (loop L1) and that of market saturation (L2). Loop L3, which connects
Installed Base Android with Contag Strength Android represents a Network Externality
effect. The externality effect can be both positive (a growing user group increases the value
of adopting an Android hand set) and negative (growth in user base creates more
opportunities for dissatisfied users through negative word of mouth and congestion). Loop L4
represents the response of handset manufacturers to growing Adoption Rate Android. The
impact can also be both positive (growing sales brings in new entrants, lowering price and
inducing more people to adopt) and negative (quality of low priced handsets fail to satisfy
market and drives away potential adopters).
The model of Figure | is converted into a stock-flow simulation model by using appropriate
policy equations that drive the growth. Two important equations that the simulation model
would have are one for Adoption Rate Android and the other for Contag Strength Android.
Since diffusion is based on a contagion process, Adoption Rate Android can be modeled
based on standard Infection Rate equation (Sterman 2000) and be written as:
Puvvala et al. /Calibration in presence of structural breaks
PA;
MOS ()
AR iq = CS. iq * IB. a
android android ‘android PAwos + [Buaarata-+ Bower
<Time>
Installed Base of
Other MOS
fo)
Potential —_L2: Market uu:
eee, LsiMarket adopters Saturation Adoption Rate Word-of-Mouth Installed Base
re Android
of MOS ndroi aa
s Mo 4
L4: Supplier
Reponse 3: Network
Market Loop Externality
Growth Ss Loop
Fraction
Contag Strength
Figure 1: Causal Model used for Calibration of Android Diffusion
Where, ARandroia 18 Adoption Rate Android, CSandroia is Contag Strength Android, 1Bandroia
is Installed Base Android, PAmos is Potential Adopter of MOS and [Botner is Installed Base of
Other MOS. The sum in the denominator is a measure of the total addressable market of all
MOS at any point in time. Contag Strength Android by this model depends on ARandroia and
TBandroid and hence can be expressed by Equation 2.
CSandroid = f UBandroia) * 9 (ARandroia) (2)
The Market Growth Loop (L5) was operationalized in the form of equation (3).
Market Expansion = Market Growth Fraction * PAyos (3)
3.2. Calibration of SD diffusion model with Structural breaks
The SD simulation model built developed from Figure 1 had to be calibrated for structure and
parameter values. The calibration process has been aided greatly by software features
commonly termed as Automatic Calibration (AC). The AC methods typically minimize the
sum of errors (between observed and simulated values) of chosen model parameters across all
observed data points. We followed the three-stage heuristic suggested by Oliva (2003).
Recognizing that the diffusion of Android handsets is marked by a combination of
incremental changes and discrete events, the first step in calibration was to identify the
presence of structural breaks in model parameter values. We used the Bai-Perron test with
Bayesian Information Criterion (BIC) to do so. Table 3 gives the BIC values for different
number of breaks for our Android handset sales data. The lowest BIC value is obtained with
4 breakpoints (at 14, 20, 26 and 34 months).
Table 3: BIC Values for Different Number of Breaks in Sales Growth
No.ofBreaks| 0 | 1 | 2 | 3 [ 4 | 5
BIC Values _ | 29.0428 | 28.6121 | 28.4617 | 28.0805 | 28.022 | 28.2671
Puvvala et al. /Calibration in presence of structural breaks
The timing of these breaks was incorporated into the parameter estimation process for
Contagion Strength using the functional form given in equation (4).
Contagion Strength android = Kandroia * e&ar4roia.1* Bandroid + Kandroid,2*ARandroid) (4)
Kandroido is a constant. Equations (5) and (6) were written to express Kandroid,1, Kandroia in Way
that information on structural breaks could be incorporated.
Kandroid! = Kandroia,11 * Pulse(0,14) + Kandroid,12 * Pulse(14,6) + Kanaroid,13 * Pulse(20,6)
+ Kanaroia,i4 * Pulse(26,8) + Kandroid,is * Pulse(34,6) (5)
Kandroid,2= Kandroia21 * Pulse(0,14) + Kandroia22 * Pulse(14,6) + Kandroia23 * Pulse(20,6) +
Kandroid24 * Pulse(26,8) + Kandroid.2s * Pulse(34,6) (6)
We calibrated the model using the optimization feature of Vensim Professional
(http://vensim.com/optimization/#vensim82 17s-optimizer-provides-fast-calibration-of-
models-and-discovery-of-optimal-solutions). Table 4 presents fit statistics for the model-
generated behavior with calibrated parameters.
Table 4: Fit Statistics for Model Generated Values
LR Square | MAPE
Android Sales | 0.9991 | 8.92%
The calibration also yielded polarity of Kanaroia,1, Kandroia2, Which determine the polarity of
loops L3 and L4 and offer insight into how Installed Base and Android Sales impacted
Android diffusion.
4 Conclusion
In this paper, we have focused on calibration issues in SD models of Android diffusion. This
technology evolution is marked by discrete events and incremental changes. Hence there is
the possibility of changes in model parameters over the period of analysis. We followed a
two-step calibration process. The first step consisted of identifying the breakpoints in a
rigorous way using appropriate statistical techniques, keeping in mind the dangers of over-
fitting inherent in identifying an excessive number of breakpoints. The second phase
consisted of calibrating the causal model where parameter values were allowed to change
across the breakpoints. Minimization of the gap between observed sales and simulated
adoption rate was used as the objective function. The calibration process could also identify
changing causal influences in the model structure itself, albeit in a limited way in that
changing link polarities may emerge from the analysis. In summary, our work demonstrates
the need to be sensitive to the presence of structural breaks in the calibration of SD models.
Technology diffusion is one domain in which this is likely to occur, but there can be other
domains where this situation may arise as well.
References
Bai J., Perron P. 1998. Estimating and Testing Linear Models with Multiple Structural
Changes. Econometrica 66(1): 47—78.
Puvvala et al. /Calibration in presence of structural breaks
Barlas, Y. 1989. Multiple tests for validation of system dynamics type of simulation models.
European journal of operational research, 42(1): 59-87.
Barlas Y., Carpenter S. 1990. Philosophical roots of model validation: two paradigms, System
Dynamics Review 6(2): 148-166.
Barlas, Y. 1996. Formal aspects of model validity and validation in system dynamics. System
dynamics review 12(3): 183-210.
Bass, FM. 1969. A new product growth model for consumer durables. Management Science
15(5): 215-227.
Burgelman RA., Silverman A., Wittig C., Hoyt DW. 2009. Google’s Android: Will It Shake
up the Wireless Industry in 2009 and Beyond. Stanford Graduate School of Business Case.
Product No. SM176-PDF-ENG
Burnham, KP., Anderson DR 2004. Multimodel inference: understanding AIC and BIC in
Model Selection. Sociological Methods and Research 33(2): 261-304
Forrester JW., Senge PM. 1978. Tests for building confidence in system dynamics models.
System Dynamics Group. Sloan School of Management. Massachusetts Institute of
Technology.
Gartner 2013. News Articles on Smartphone Sales, http://www.gartner.com/it. Web accessed
on April 15, 2013
German K. 2011. A brief history of Android phones, Cnet Reviews, 02 August 2011, Web.
15 April 2013, http://reviews.cnet.com/8301-19736_7-20016542-25 1/a-brief-history-of-
android-phones/
Google Play Developer Program Policy 2012. http://play.google.com/about/developer-
content-policy.html. Web accessed on March 13 2014
Google Trends 2013. Web Search interest: android issues. Worldwide, 2004 — present,
google.com/trend. Web. Accessed on April 14, 2013
IDC 2012. Worldwide Mobile Phone Growth Expected to Drop to 1.4% in 2012 Despite
Continued Growth Of Smartphones, According to IDC”. IDC Press Release,4th Dec 2012,
Web. Accessed on May 25, 2013.
IDC 2013. “Smartphone sales to touch | billion-unit mark in 2014: Credit Suisse”. Web.
Accessed on May 25, 2013.
Mahajan, V., Muller E. 1979. Innovation Diffusion and New Product Growth Models.
Journal of marketing 43(4): 55-68
Meade N., Islam T. 2006. “Modeling and forecasting the diffusion of innovation — A 25-year
review”. International Journal of Forecasting 22(3): 519-545.
Oliva R. 2003. Model calibration as a testing strategy for system dynamics models. European
Journal of Operational Research 151(3): 552-568.
Qudrat-Ullah H., Seong BS. 2010. How to do structural validity of a system dynamics type
simulation model: the case of an energy policy model. Energy Policy 38(5): 2216-2224.
Rogers EM. 1976. New Product Adoption and diffusion. Journal of Consumer Research 2(4):
290-301.
Sterman JD. 2000. Business Dynamics: Systems Thinking and Modeling for a Complex
World. Irwin/McGraw-Hill, Boston.
